Sei sulla pagina 1di 428

EDITOR-IN-CHIEF

PETER W. HAWKES
Laboratoire d Optique Electronique
du Centre National
de la Recherche Scientifique
Toulouse, France

ASSOCIATE EDITOR-IMAGE PICK-UP AND DISPLAY

BENJAMIN KAZAN
Xerox Corporation
Palo Alto Research Center
Palo Alto, Calijornia
Advances in
Electronics and
Electron Physics

EDITEDBY
PETER W. HAWKES
Laboratoire d Optique Electronique
du Centre National
de la Recherche Scientifique
Toulouse, France

VOLUME 68

1986

ACADEMIC PRESS, INC.


Harcourt Brace Jovanovich, Publishers
Orlando San Diego New York Austin
Boston London Sydney Tokyo Toronto
C O P Y R I G H T 0 1986 BY ACADEMIC PRESS. INC
A L L RIGHTS RESERVED
NO PART O F T H I S PUBLICATION MAY BE R E P R O D U C E D O R
T R A N S M I T T E D IN ANY FORM O R BY ANY M E A N S , ELECTRONIC
OR MECHANICAL, INCLUDING PHOTOCOPY. RECORDING. OR
ANY INFORMATION STORAGE A N D RETRIEVAL S Y S T E M , W I T H O U T
PERMISSION IN WRITING FROM T H E PUBLISHER

ACADEMIC PRESS, INC


Orlando. Florida 32XX7

United Kingdom Edition publiyhed b\


ACADEMIC PRESS INC (LONDON) LTD
23-2K o v d i Rodd London N W I 7 D X

L I B R A ROYF C O N G R b S S C A T A L O G CARD NUMBtR 49-7504

ISBN 0-12-014668-1 (alk. paper)

P K I \ I I L) IN 1116 I N I T F C FT4TES 01 4 M t K I C A

8 6 x7 xx 89 9 8 7 6 5 4 1 2 1
PREFACE

In Volume 57 of these Advances, an article by H. W. Redlien and R. J. Kelly


was devoted to “The microwave landing system: A new international standard,”
and I am assured that this has been heavily used by the aviation community.
Since then, precision distance measuring equipment (DME/P) has come into
widespread use as an integral and critical part of the MLS, a development in
which R. J. Kelly has been deeply involved. I am therefore delighted to be able
to publish his detailed account with D. R. Cusick of this new addition to the
instrumentation that enables modem aircraft to land and take off so safely. In
view of the importance of the subject and the authors’ concern to make their
account at once complete, self-contained, and yet readable, this article dominates
the volume and I have no doubt that members of the aviation world will be
grateful for so clear and full a document.
In 1935 A. Einstein, B. Podolsky, and N. Rosen published a paper in Physical
Review entitled “Can the quantum-mechanical description of physical reality be
considered complete?” This analysis of the orthodox interpretation of the quan-
tum theory revealed a paradox, now universally known as the EPR paradox. This
in turn has generated a voluminous literature, among which a paper by J. S. Bell
is something of a landmark, adding Bell’s inequalities to the scientific vocabu-
lary. The fiftieth anniversary of the EPR paper seemed a good occasion to survey
this work, and I am extremely happy that W. De Baere, himself an active
contributor to the debate, has agreed to put all these complex ideas into perspec-
tive. The point at issue is a difficult one, but is vitally important for the under-
standing and future development of quantum mechanics.
The final article is devoted to two common elements of systems of charged-
particle optics that have been neglected in comparison with round lenses, prisms,
and multipoles. These elements, mirrors and cathode lenses, have been regarded
as difficult to analyze owing to the fact that some of the usual mathematical
approximations that are fully justified in lenses and prisms can no longer be
employed. Considerable progress with this problem has been made in the group
led by V. M. Kel’man, co-author of the standard Russian book on electron
optics, in the Nuclear Physics Institute in Alma-Ata. This work is available in
translation since most of it appeared in the Zhurnal Tekhnicheskoi Fiziki (Soviet
Physics: Technical Physics), but it is not widely known and I hope that this
connected account in a widely accessible series will help to remedy the situation.
The authors have taken considerable trouble to make their article self-contained,
and thus reference lists of key formulas, for aberration coefficients in particular,
are included.
vii
...
Vlll PREFACE

As usual, I conclude by thanking all the authors and by listing forthcoming


articles.

Monte Car10 Methods and Microlithography K . Murata


Invariance in Pattern Recognition H. Wechsler

PETERW. HAWKES
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS, VOL. 68

Distance Measuring Equipment and


Its Evolving Role in Aviation
ROBERT J. KELLY AND DANNY R. CUSICK
Bendix Communications Division
Allied Bendix Aerospace Corporation
Towson, Maryland 21204

I. INTRODUCTION

For almost 40 years, distance measuring equipment (DME) has been a


basic element in world air navigation. Increasing needs of automation in
aviation-and the increasing levels of sophistication that resulted-fueled its
evolution, significantly improving its capabilities, while the soundness of its
principles guaranteed its wide use. As a result of this continuing evolution, a
new version of DME is now part of the new international standard for landing
systems, the microwave landing system (MLS).
Because of its persistent presence, DME has played a dominant role in the
development of air navigation, and its principles are common to other parts
of the air traffic control system (ATC). This article discusses these basic prin-
ciples by placing them in perspective with other types of air navigation sys-
tems as part of a broader, integrated discussion of navigation requirements.
Specifically, the discussions relate to the needs of the air traffic control sys-
tem and to the dynamic flight characteristics of the aircraft operating within
that system. To expose the diverse requirements that the DME element must
satisfy in the modern and evolving National Airspace System, substantial
background information is provided on the elements of the ATC system
and on the implementation of automatic flight control systems.
Two main types of DME have been standardized by the International
Civil Aviation Organization (ICAO). The DME that was standardized ;n
1952 and that has been in general use during the past 40 years is referred
to as the DME/N. The new precision DME, to be used with MLS, is known
as DME/P. The ICAO standardized it in 1985.

Copyright @ 1986 by Academic Press, Inc.


All rights of reproduction in any form reserved.
2 ROBERT J. KELLY A N D D A N N Y R. CUSICK

The DME principle is simple. Distance information is computed in the


aircraft by measuring the round-trip time between interrogations from an
airborne transmitter and replies to those interrogations from a gound beacon.
Because the DME has proven itself to be a robust technique, its use and
implementation will continue into the next century.
This chapter views both DME/N and the new DME/P from a broad
systems engineering perspective. Like many other navigation aids DME is
characterized in terms of a signal format, accuracy (error budgets), coverage
(power budgets), radio frequency assignment(channel plan), the measurement
technique used to extract the DME navigation parameter (range), and the
output guidance information rate. A DME must also satisfy certain levels of
signal integrity and availability.
The signal format is the blueprint of the system; it ensures that the DME
ground and airborne elements will be compatible. Range accuracy must
facilitate safe arrival of aircraft at their destinations. The ground facility must
provide a usable signal that covers the volume of airspace where navigation
information is required. Well-conceived radio frequency (rf ) channel plans
assure that ground facilities can be geographically located so that their
individual service volumes unambiguously “combine” to provide navigation
guidance throughout the navigable airspace. The techniques used to make
navigation parameter measurements must be simple so that ground and
airborne equipment designs are reliable and economical while achieving the
necessary accuracy and coverage. Finally, the guidance output information
rate must be consistent with the aircraft flight control system’s (AFCS)
dynamic response so that it will not limit the aircraft’s maneuvers.
The DME system discussions in this article will follow the functional
theme outlined herein. Emphasis will be directed toward systems con-
siderations rather than toward hardware/software implementations or flight
test data analysis. Section I1 provides the background and historical
perspective necessary to view DME in the broader context of air navigation
and its specific role in the National Airspace System (NAS). Section I11 is
devoted to DME/N, and Section IV addresses the most accurate version of
DME, the DME/P.
The navigation applications discussed in this article are restricted to the air
traffic control system as it is defined by the International Civil Aviation
Organization and as it is implementedin the United States.Therefore,only the
domestic enroute, terminal, and approach/landing phases of flight are
addressed. Enroute and approach/landing scenarios are defined using only
the conventional takeoff and land (CTOL) aircraft type. However, DME is not
restricted to these applications or to CTOL aircraft. Both helicopter and
remote area operations, for example, come under the broad envelope of DME
applications.
DISTANCE MEASURING EQUIPMENT IN AVIATION 3
11. GENERAL
BACKGROUND TO DISTANCE
MEASURING
EQUIPMENT
APPLICATIONS

A . Basic Operational Scenario and Introduction to the


Navigation Services

A commercial jet aircraft leaving Chicago’s O’Hare International Airport


for New York’s LaGuardia Airport will fly along a national airway system
divided into enroute and terminal areas. During the enroute portion of the
flight, the aircraft flies airways defined by a network of straight-line segments
or radials. These radials are generated by very-high-frequency omni range
(VOR) and DME navigation aids that provide direction and distance
information to known ground locations. Using these radials, the pilot can
navigate to the destination terminal area.
Arriving in the LaGuardia terminal area, the aircraft enters a region of
converging tracks and high-density traffic. Position accuracy becomes
increasingly more important as the aircraft approaches its runway destination,
particularly when visibility is reduced. Here, the enroute guidance supplied by
the VOR/DME radials is replaced with approach and landing guidance such
as that generated by the MLS. The three-dimensional guidance provided by
MLS is of sufficient .accuracy to guide safely the aircraft to a landing, even
under zero-visibility conditions.
Accurate range data are essential during both phases of flight. The
guidance aid which generates this ranging information is DME. It provides
range data accurate to 0.2 nautical miles (nmi) along the enroute airways.
Within the terminal area, its range error decreases to 100 feet or less for the
landing guidance applications.
DME has evolved from the radar beacon concept to satisfy the navigation
needs of both the civilian and the military aviation community. This evolution
is reviewed in Subsection B. The general principles of aircraft position estima-
tion, as applied to the radio navigation aids used in the NAS, are summa-
rized in Subsection C. Both bearing and range measuring systems are treated.
In order to place the DME in its proper operational context, other
navigation aids that operate in conjunction with it must be discussed. In
Subsection D the role of VOR, ILS, MLS, and the ATC surveillance system
(secondary radar) in the overall ATC system are briefly discussed. Also
included is the Tactical Air Navigation System (TACAN) upon which the
military currently depends for its enroute navigation. In this system, DME
plays a central role in a signal format that is arranged to contain both range
and bearing information.
Both military (TACAN) and civil (VOR/DME) enroute navigation
4 ROBERT J. KELLY AND DANNY R. CUSICK

systems are combined on the ground in the configuration called VORTAC.


For the United States NAS and many of the European ICAO member states,
the principal ATC enroute airways are based on range and bearing
information from VOR/DME or VORTAC facilities.
For the precision phases of flight conducted in the terminal area-
approach, landing, missed-approach, and departure- the principal aids are
ILS and MLS. A high-precision DME is the range-providing element of the
microwave landing system. When used with the instrument landing system
(ILS), another version of DME/N provides the range information normally
generated by the ILS marker beacons (which "mark" a few specific points
along the straight-line approach path).
DME, VOR, ILS, MLS, and TACAN are navigation aids (navaids) that
provide the primary navigation parameters of range and bearing. These
navaids can be used directly for navigation, or they can be integrated into the
aircraft by an area navigation system known as RNAV. Navigation derived
from RNAV is becoming increasingly important and has led to the develop-
ment of RNAV routes in addition to the VORTAC routes. Subsection D
briefly describes RNAV. Table I summarizes the navigation services. With the
exception of MLS, they have been in use for the past three or four decades.
It is essential to understand the interaction of navigation signals with the
Aircraft Flight Control System (AFCS) in order to comprehend a navigation
system. To assist in this understanding, background information on aircraft
guidance and control is given in Subsection E. Accuracy performance
comparisons between VOR/DME, DME/DME, and TACAN, using RNAV
to calculate the position fixes, are given in Subsection F.

TABLE I
PRESENT NAVIGATIONAL
SERVICES

Service Frequency band (MHz)

VHF omnidirectional range (VOR) 112-118


Distance measurement, DME/N 960-121 5
Air traffic control surveillance system"
Interrogator 1030
Transponder 1090
Instrument landing system
Localizer 108-112
Glide slope 328-335
Marker 15
MLS
Azimuth and elevation 5031-5091
DME/P (Same as DME/N)

Includes mode S and the traffic alert collision avoidance


system (TCAS).
DISTANCE MEASURING EQUIPMENT IN AVIATION 5
B. History of the DME Principle

This section introduces the principal elements of DME, the airborne


interrogator, and the ground transponder, by tracing the early history of their
development. Both the DME and the ATC Secondary Radar were derived
from the radar beacon concept developed during World War 11. All of the
techniques described herein were operational in the early 1940s (Roberts,
1947).
Conventional radar operates by sending out high-energy pulses of radio
waves. The radar receiver detects the echo of the transmitted energy that is
reflected back from a target. The elapsed time between the transmission of the
pulse and the reception of the echo is a measure of the range (or distance) to
the reflecting target. Further, by use of a suitable antenna design, the radio
energy is concentrated into a narrow beam so that the echoes are received
only when the radar is ‘‘looking’’ at targets residing within the antenna’s
beamwidth. By proper coordination of the motion of the antenna and the
sweep of an intensity-modulated cathode-ray tube (CRT) display device, a
plan view (like a map) of the reflecting targets in the region of the radar set is
traced out. This map-type display, which is illustrated at the top of Fig. 1, is
called the plan-position indicator (PPI). Thus the target’s position is identified
by a “bright spot” on the PPI. The radial trace through the spot is the target’s
bearing relative to the radar’s reference. The distance from the center of the
PPI to the bright spot is proportional to the target’s range.

I . The Beacon Concept


Conventional radar pulse transmissions are reflected by targets of different
sizes regardless of their individual importance to the radar user. Sometimes
echoes are too weak to be observed, as are those from a small airplane at a
great distance. Under other circumstances, strong echoes from buildings or
natural features such as mountains may mask weaker echoes from aircraft.
Furthermore, the exact location of a geographical point on the ground may be
of importance to a radar-equipped aircraft even though the terrain gives no
distinguishing echo. In all such cases, radar beacons find their use as
cooperative aids in pinpointing specific targets.
A radar beacon is a device that, upon reception of the original radar pulse,
responds with its own transmitter to radiate a strong signal of one or more
pulses which is easily distinguished from the weaker radar echoes from
surrounding objects. The beacon, therefore, is a device which amplifies the
echo. The beacon transmitter need not be very powerful to be able to give a
reply much stronger than the target echo itself. Most importantly, the beacon,
when placed upon the ground, can provide an accurate reference point which
can be used to determine an aircraft position relative to the reference point.
6 ROBERT J. KELLY AND DANNY R. CUSICK
NORTH

INTERROGATOR

e = BEARING
R = RANGE
DIRECTIONAL
ANTENNA

RECEIVER TRANSMITTER
BEACON

FIG. 1. The beacon concept.

In a typical beacon application (Fig. l), the scanning narrow beam


emanates from the aircraft. The position of the beacon on the ground (its range
and bearing relative to the aircraft) can be determined by viewing the airborne
PPI display. The ground beacon’s antenna is designed to be omnidirectional,
that is, to receive signals equally well from all azimuth directions. The beacon
DISTANCE MEASURING EQUIPMENT IN AVIATION 7
bearing is derived from the narrow scanning beam while the “round-trip”
delay of the radar transmission and its reply from the beacon is proportional
to the beacon’s range to the aircraft, resulting in a unique measurement of
position. The measurement of range and bearing constitute one of the most
fundamental procedures in determining a position “fix” by a radio navigation
aid. It is the basis, for example, by which guidance is obtained to fly the
VORTAC routes.
The process by which a radar set transmits a signal suitable for triggering
the beacon is known as interrogation; the corresponding beacon transmission
is termed the reply. Radar beacons which reply to interrogations are called
transponders and the radar set used to interrogate a beacon is called an
interrogator.
The provision of a special character to the pulsed signals of either the
interrogator or the transponder is called encoding. A device that sets up such a
coded signal is called an encoder; a device that deciphers the code at the other
end of the link is called a decoder.
Although the transponder reply is like an echo, it differs from one in several
significant respects. The strength of the response is independent of the
intensity of the interrogating signal, provided only that the interrogating
signal exceeds a predefined minimum threshold intensity at the beacon
receiver. Also, the response frequency is different from that of the inter-
rogation frequency. The response signal pulse may also differ from the
interrogation signal in form. It may even consist of more than one pulse, the
duration and spacing of the pulses being an arbitrary design choice. Thus
echoes from a desired beacon on the ground can be distinguished from others
on the ground not only in terms of pulse amplitude (nearer beacons have
larger amplitude returns) but also in terms of radio frequency and pulse
spacing (code). All of the important components necessary to form a beacon
system are illustrated in Fig. 1. Roberts (1947) is a complete treatise on the
systems-engineeringconsiderations of radar beacon designs.
The complete downlink/uplink cycle starts with the generation of a trigger
pulse in the airborne radar interrogator. This trigger initiates the transmission
of an interrogating signal and, after a predetermined delay (called the zero mile
delay),starts the cathode-ray tube sweep on the indicator. On the ground, the
beacon transponder receiver detects the interrogating signal and generates a
video signal that is passed to the decoder. The decoder examines the video
signal to see if it conforms to an acceptable code. If not, it is rejected. If
accepted, a trigger pulse is passed to the blanking gate that prevents the
receiver from responding to any further interrogations for a time sufficient to
permit the complete coded reply to be transmitted. This time, called the reply
dead time, varies generally from about 50 to 150 ys. Because the beacons
require a reply dead time, there is a limited number of interrogations (called
8 ROBERT J. KELLY AND DANNY R. CUSICK

the traffic load) that can be accommodated by the transponder (see Sec-
tion 111,C).
The transponder decoded output, a single pulse, is delayed for a time which
when added to the reply code length equals the “zero mile delay” used in the
interrogator. The delayed signal is then encoded with a reply code and
transmitted equally in all directions on a different rf frequency. At the airborne
radar antenna the reply is detected by the interrogator receiver, decoded, and
displayed to the operator on the PPI indicator.
If the interrogator is “on top of” the transponder, i.e., if the range were zero
miles-there would be no propagation delay through space, and the
transponder reply signal’s arrival time would equal the zero time delay. The
PPI trace would then show a “spot” at the center of its display. This system
was developed almost 50 years ago and forms the basis of the secondary
surveillance radar (SSR) system and the DME.
If the airborne element is the transponder with nondirectional antenna
and the interrogator with its directional antenna is placed on the ground, the
configuration is the basis of the SSR system. Range and bearing and, thus, a
position fix on the aircraft are made available to the air traffic controller on the
ground. On the other hand, if the beacon is located on the ground and the PPI
scope in the aircraft is replaced by an automatic range tracking circuit, the
range to the fixed beacon is displayed to the pilot. This is the DME concept
(see Fig. 2). Interestingly enough, the principal components-interrogator
transmitter and receiver, encoder, decoder, transponder dead time, echo
suppression, transponder receiver, and transmitter -remain the principal
elements of today’s DME system. Only elements of performance such as
accuracy have been improved over the years in response to more demanding
applications.
Since range is the only parameter of interest in DME, bearing information
is sacrificed in exchange for use of simple omnidirectional ground and
airborne antennas. In addition, a higher data rate is possible since the ground
and airborne equipments are always “looking” at each other. An aircraft
“range-only’’ position fix can be obtained by measuring the range from two
separated ground transponders. See Subsections C and D,5 on scanning
DME.

2. History of the Beacon Concept


The radar beacon was invented in 1939 by a group at the Bawdsey
Research Station of the Air Ministry in the United Kingdom (Williams et al.,
1973). It was developed in response to a military need, and its invention was
not immediately made public for reasons of security. During the war, radar
beacons were used by Germany, Japan, and the Allies. Because these countries
DISTANCE MEASURING EQUIPMENT IN AVIATION 9

TRIGGER

AIRBORNE
INTERROGATOR
4
ENCODER DECODER

4 t
TRANSMITTER RECEIVER

\ a

AIRCRAFT SKIN

El GROUND
BEACON

FIG.2. DME operation.

veiled the invention in secrecy, there were no pre-World War I1 nonmilitary


uses of beacons.
The initial purpose of the beacon was for Identification of Friend or Foe
(IFF), where appropriate transponder responses to coded interrogations
indicated that a particular military vehicle was either a “friend” (correct code
response) or “foe” (incorrect response).
It was soon discovered that beacons could be used for offensive purposes
by helping to locate target areas. That is, they were used not only for
identification but for navigation as well. One such system which employed
lightweight ground beacons was the Rebecca-Eureka system. It included an
airborne interrogator called “Rebecca” and a ground beacon called “Eureka.”
Originally designed by the Telecommunications Research Establishment
(TRE) for British use, the Rebecca-Eureka system was later adopted by the
United States Army and used in many operations. The Rebecca-Eureka
system operated at a radio frequency of 200 MHz. Table I1 shows its channel
plan, which is the forerunner of the DME channeling concept (Roberts, 1947).
10 ROBERT J. KELLY AND DANNY R. CUSICK

TABLE I1
CHANNEL
REBECCA-EUREKA PLAN

Interrogation Response
channels channels
(MW (MW Characteristics

A 209 209 Only cross-channel used;


B 214 214 e.g., interrogation on A,
C219 219 response on B, C, D, or E
D 224 224
E 229 229

3. Development of the D M E Concept


As remarked in Dodington (1980), “it must have occurred to many people
that the addition of a relatively simple automatic tracking circuit would
enable an interrogating aircraft to constantly read its distance with respect to a
ground beacon.” However, it was apparently not until 1944 that such a device
was actually built (by the Canadian Research Council), operating in the
Rebecca-Eureka band of 200 MHz. In November 1945, the Combined
Research Group at the U.S. Naval Research Laboratory built and demon-
strated such a system operating at lo00 MHz. It was largely based on work
that the Hazeltine Corporation had been doing on the Mark V IFF, at 950-
1150 MHz. This system, however, did not use precise frequency-control
techniques. A crystal-controlled DME system was developed at ITT Nutley
Laboratory in 1946 (Dodington, 1949) and had a large impact on the final
DME channel plan selected by ICAO. Whether the DME had crystal control
or not was not simply a hardware question. It was most importantly a signal
format question involving the efficiency of radio frequency spectrum use. The
use of precise frequency control permitted more DME channels to be
accommodated within a specified frequency bandwidth.
The development of DME from 1946 to 1959 was very complex, involv-
ing the U.S. military, Congress, industry, and national pride and will receive
only a cursory review. The reader is referred to several excellent references
(Dodington, 1980; Sandretto, 1956;Rochester, 1976)for details on this subject.
As described in Sandretto (1956),
the air-navigation and traffic-control system in the United States grew as the product of
necessity and without being planned. It began about 1919 with the installation of 4-course
radio ranges for furnishing guidance to aircraft along defined routes. Later, other
navigational aids were added, such as markers (both low and high frequency), to indicate
points along the airways. Still later, low-approach equipment was added. When a
controlling agency was established on the ground, it was equipped with slips of paper
to designate aircraft, and utilized limited knowledge of airspeed, heading, and winds to
DISTANCE MEASURING EQUIPMENT IN AVIATION 11
predict their future positions. Reliance was placed on communicationsfrom the aircraft to
the ground controller to learn of the aircraft’s actual positions. Communications were
furnished largely through the high-frequency equipment owned and operated by the
airlines.Teleprinter communicationswere later added between the various control centers.
In reviewing this system, it is clear that it was slow and inefficient. Because
of this situation, several companies (ITT, Hazeltine, RCA, and Sperry)
proposed schemes to alleviate this problem through implementation of a
comprehensive navigation and traffic control system. An ITT system called
NAVAR was based upon the following concepts developed in 1945 (Busignies
et al., 1946):
(1) means whereby the pilot can know his position in three dimensionsso
he can navigate to his destination;
(2) means whereby a ground authority, capable of regulating the flow of
air traffic, can know the position of all aircraft;
(3) means whereby the ground authority can forecast the future positions
of all aircraft;
(4) means whereby the central authority can issue approvals to pilots of
aircraft to proceed or to hold.
When the present ATC system is discussed later, it will be interesting to
note how closely it approached ITT’s original planning concepts.
The first navigation elements of NAVAR were the DME and an associated
bearing device. They were demonstrated for three weeks in October, 1946, at
Indianapolis where the U.S. played host to the Provisional International Civil
Aviation Organization (PICAO).
A major result of the Indianapolis demonstrations was that ICAO later
decided that the future short-range navigation system would be p-8, i.e.,
range-bearing [as opposed to GEE, which was the British VHF hyperbolic
navigation system (Colin and Dippy, 1966)l. Further, it was decided that the
DME part would be located at 1000 MHz, not 200 MHz. It was argued that
the p-8 system was the natural choice since man is born with the knowledge of
left and right and learns about distance when he learns to walk (Sandretto,
1956).
Meanwhile, subcommittees for both the Radio Technical Commission for
Aeronautics (RTCA) and the ICAO began working on standards for a
coordinated navigation system, including DME.
During 1946, RTCA Special Committee SC-21 developed a DME
channeling plan at L band (the 1000-MHz region) that ignored crystal control
(RTCA, 1946).Its 13interrogation channels and 13 reply channels were spaced
at a relatively wide separation of 9.5 MHz based on using free-running
oscillators whose stability was to be k 2 MHz. Later in the summer of 1947,
RTCA organized committee SC-31 on Air Traffic Control. This committee’s
12 ROBERT J. KELLY AND DANNY R. CUSICK

report (RTCA, 1948a) stated that


The Navigation Equipment is a transmitter-receiver having multiple channels. This
equipment, with associated ground equipment:
(1) Provides distance and bearing information for navigation. These data, when used
in conjunction with a computer, will allow the pilot to fly any desired course.
(2) Provides precise slope, localizer, and distance information for instrument
approaches.
(3) Provides information for airport surface navigation to enable the pilot to taxi his
aircraft.
(4) Provides air-ground aural communication of a reliable and status-free type.
( 5 ) Provides a situation display in pictorial form which enables the pilot to monitor
traffic conditions in his vicinity or receive other pertinent data such as holding areas,
airlane locations, and weather maps from the ground.
(6) Provides suitable output to allow the aircraft to be automatically flown, either
enroute or during final approach and landing.
In retrospect this appears to be the first attempt to define the ATC system
as it exists today. The SC-31 committee’s navigation system is further
described to have “the airborne navigation equipment and its directly
associated ground elements wholly contained in the 960-1215 MHz band. The
airborne navigational equipment for the ultimate system will have accuracy of
k0.6“ in bearing and & 0.2 nautical mile or 1% of the distance, whichever is
greater.” As will be discussed in Subsection F, none of the bearing systems
finally placed in operation achieved this & 0.6” accuracy. However, DME/N
did achieve a performance of 0.2 nautical mile and-with the recent
development of DME/P-has surpassed that goal by almost twentyfold.
Although the equipment proposed by the SC-31 committee for the
ultimate or target system was very dissimilar to the NAVAR system, it
constituted a coordinated system with essentially the same elements that had
been described by ITT. Committee SC-31 had brought together the thinking
of the Air Force and the Navy as well as the Civil aviation industry. By
participating in the committee’sactivities, each group learned of the work that
had been done in the other groups.
As proposed by SC-31, only the DME and the Secondary Surveillance
Radar remained in the 960-1215 MHz band. The bearing information (VOR)
was assigned to the VHF band while the aircraft landing systems ended up at
VHF for ILS and C-band (in the 5000-MHz region) for MLS (see Table I).
That is, through the consensus of committee participation, the NAVAR and
other proposals were discarded. In early December, 1947, another RTCA
Committee (RTCA, 1948b),SC-40, was convened for the purpose of studying
two channeling techniques under development for DME. One of these was
the multiple radio frequency channel, crystal-control technique, and the other
the pulse-multiplex system which was similar to that proposed by SC-21. The
committee made its report in December of 1948 with a recommendation that
DISTANCE MEASURING EQUIPMENT IN AVIATION 13
both crystal-control and pulse-multiplex techniques be included in the
channel plan.
The envisioned DME system would interrogate on ten crystal-controlled
channels at the low end of the band and reply on ten crystal-controlled
channels at the high end of the band. Channels were spaced 2.5 MHz apart and
used 10 pulse codes in multiples of 7 ps, starting with 14 ps. Thus, there were
10oO combinations of frequency pairs and codes, of which it was proposed to
pair 100 with the VOR and the ILS localizer (see Table VI).

4. DME Versus TACAN


The U.S. Civil Aviation Agency (CAA)-the forerunner of the Federal
Aviation Administration-adopted the SC-40 plan as its official position and
convinced ICAO to adopt it. The first ICAO Channel Plan, given in Tables V
and VI, is essentially the same as that proposed by SC-40. However, as noted in
Dodington (1980), there was little enthusiasm for this system outside the CAA,
both within and outside the US. The main reason was that the US. military
was developing an entirely different system in the same frequency band. When
the SC-40 system was finally abandoned in 1956, there were only 340 airborne
sets in existence.
The trouble started shortly after World War 11, when the CAA began to
install the recently developed VOR and DME units that were expected to form
the nucleus of the emerging “common” system. The Navy found VOR/DME
to be unsatisfactory for aircraft carrier operations because of the large size and
complex siting requirements of the relatively low-frequency VOR antenna.
They contracted with the ITT’s Federal Telecommunications Laboratories in
February, 1947, to design an alternative system that could be used effectively
at sea. By 1951,the Navy project had not only acquired the name TACAN but
also the support of the Air Force, which had originally endorsed VOR/DME.
The Air Force was converted to TACAN by its promise of greater flexibility
and accuracy and its integration of the bearing and range elements into a
single unit and, accordingly, a single frequency band. Thus, scarcely before the
SC-40 “common system” document had circulated, civil and military agencies
were following separate courses of action. Rochester (1976) provides a
complete history of the VOR/DME/TACAN controversy. It was resolved in
August, 1956,by discarding the SC-40 system and replacing it with a TACAN-
compatible DME. Part of the solution involved a fixed pairing of 100TACAN
channels one-for-one with 100 VOR/ILS channels. This proposal by
Dodington (1980) thus allowed a single-channel selector for VOR/ILS/DME
and removed the objections of the airlines which were concerned that blunders
would result from VOR bearing information received from one station and
DME from another.
14 ROBERT J. KELLY AND DANNY R. CUSICK

The new enroute navigation system, christened VORTAC, retained the


VOR bearing component but replaced civil DME with the distance compo-
nent of TACAN. Each VOR station, instead of being collocated with a DME,
was now collocated with a TACAN beacon (which, of course, also provides the
DME service).
Figure 3 shows the general VORTAC arrangement. At the ground station,
the VOR central antenna is housed in a plastic cone that supports the TACAN
antenna. Leads to the TACAN antenna pass through the middle of the VOR
antenna, along its line of minimum radiation, and thus do not disturb the
VOR antenna pattern.

i' 1
I I -
I
TACAN

L - _M l L l-T l
- - - -----
I
I
I

I
I
I
CHANNEL
SELECTOR

CHANNEL
7 1 SELECTOR

CHANNEL
SELECTOR

' VOR
TACAN
ANTENNA
I
VOR
ANTENNAS NORTH

NORTH

BEARING

70 m.-
BEARING
TO BEACON

NOTE: W E AND VOR CHANNELS ARE PAIRED

FIG.3. The VORTAC system.


DISTANCE MEASURING EQUIPMENT IN AVIATION 15
Civil aircraft read distance from the TACAN beacon and bearing from
VOR transmitter. Military aircraft read both distance and bearing from the
TACAN beacon. Both types of aircraft therefore fit into the same ATC
network.
Although the CAA and DOD had thus resolved their differences, the
system defined by ICAO in 1952 was still the SC-40 format. Most of ICAO’s
member countries did not implement it and, in fact, some took the opportunity
to implement a hyperbolic system such as DECCA, the British low-frequency
navigation system. After more years of debate, ICAO finally adopted the
TACAN-compatible DME in 1959, and the VORTAC system has remained
intact since that time. The evolution of the international DME standards will
be detailed in Section 1II.A.

C. Principles of Air Navigation Systems

In this section basic concepts of navigation, position fixing, and dead


reckoning are defined and applied to air navigation.

I . Dead Reckoning, Fixing, and Navigation Systems


The fundamental task of a navigator is the selection of a route to be
followed by his aircraft. To do this the navigator must know his aircraft’s
present position and destination before he can choose a route and, from that, a
course and heading. All position-determination schemes can be classified as
dead reckoning, position fixing, or a combination of both.
A position, once known, can be carried forward indefinitely by keeping
continuous account of the velocity of the aircraft. This process is called dead
reckoning. Because velocity v is a vector, it has a direction and magnitude
(speed)u. From a given starting point p o and a velocity vo, one can determine a
new position p after time Ato by stepping off a distance Ad = vot, in the
direction of the velocity vo. At position p l , the process is repeated to find the
next position p z using v1 and Atl, and so on. Unfortunately, direction and
speed cannot always be precisely measured, particularly if direction is derived
from aircraft heading only and effects of winds are approximated. Thus, on the
average, the errors of dead reckoning increase in proportion to the length of
time it is continued. Ordinarily, such an elementary dead reckoning process is
seriously in error after an aircraft has traversed a few hundred miles without
any indication of position from some other source. The Doppler radar
navigator is an example of a good and more complex dead reckoning system
(Kayton and Fried, 1969). It can guide an aircraft across the Atlantic Ocean
and have a position error at its destination of less than 10 nautical miles. In
16 ROBERT J. KELLY A N D DANNY R. CUSICK

recent years, the Doppler radar navigator has been replaced by the inertial
navigation system (INS) for transatlantic flight navigation.
A position fix, in contrast to dead reckoning, is a determination of position
without reference to any former position. The simplest fix stems from the
observation of a recognizable landmark. In a sense, any fix may be thus
described; for example, the unique position of stars is the recognizable
landmark for a celestial fix. Similarly, some radio aids to navigation provide
position information at only one point, or relatively few points, on the surface
of the earth. In this case the landmarks are the physical locations of the
ground stations.
To obtain a position fix using radio navaids such as DME or VOR requires
measurements from at least two different ground facilities;for example, the use
of two differentDME stations (DME/DME) requires measurementsfrom two
different ground stations or, for DME/VOR, two different measurements
(range and bearing) from the same location. A single measurement yields only
a locus of position points; two measurements are necessary to identify the
point corresponding to an aircraft’s position with respect to the surface of the
earth. Since a fix is based upon essentially “seeing” landmarks whereas dead
reckoning is more associated with a computational process, it is “natural”
that most electronic aids to navigation are based on the determination of
fixes.
Dead reckoning can be combined with position fixing to realize a more
accurate navigation system. These configurations are referred to as multisen-
sor navigation systems in the literature (Fried, 1974; Zimmerman, 1969). As
will be shown in Subsection E, dead reckoning and position fixing comple-
ment each other; each provides an independent means of checking the
accuracy of the other. Where position fixing is intermittent, with relatively
long intervals between fixes, dead reckoning is appropriately considered the
primary navigation method. In these systems, position fixing constitutes a
method of updating the dead reckoning calculations such as in the “aiding”of
an INS by a VOR and DME position fix, as shown in Fig. 4. Notice how the
position fix corrected the dead reckoning measurement. Britting (1971) gives
an in-depth discussion of INS.
This correcting combination is extremely important in modern air
transports which may have the navigation guidance autocoupled to the flight
control system (Subsection C,3).A major concern is controlling the aircraft so
that it maintains its desired course in the presence of wind shears and wind
gusts. Small tracking errors are achieved by having the aircraft tightly coupled
to the guidance system. Forcing the aircraft to follow the guidance signal
closely requires low-noise navigation sensors. The INS is such a low-noise
device, but, as mentioned above, it has a tendency to drift from its initial setting
with time (about 1 nmi/h). These drift errors are corrected periodically by
DISTANCE MEASURING EQUIPMENT IN AVIATION 17
NORTH

t NORTH

$ = HEADING \
FINAL APPROACH

--
GT-TRACK ANGLE WAY POINT

-
R RANGE
V AIRCRAFT GROUND VELOCITY

FLIGHT PATH CONSTRUCTION


DESTINATIONS

- .WAY POINTS

position-fixing radio navaids. With position-fix aiding, the INS becomes the
desired accurate windproof navigation system.
If fixes are available continuously or at very short intervals, position fixing
becomes the primary navigation method. Rapid position fixing constitutes the
most common primary navigation system currently in use over the continental
US. (see Subsection D,5 on RNAV). Its popularity arises from the low
acquisition and maintenance cost of the airborne element. In some cases, dead
reckoning measurements are necessary to augment the rapid position-fix
information. Such a case arises when excessive noise from the VOR/DME
sensors causes the aircraft flight control system to have undesirable pitch and
18 ROBERT J. KELLY AND DANNY R. CUSICK

roll activity. The aircraft may also follow the sensor noise because it is tightly
coupled to the guidance signal (high signal gain). Unfortunately, reducing the
noise by uncoupling from the guidance signal (low signal gain) can degrade
the aircraft’s transient response so that it cannot closely follow the course in
the presence of wind turbulence.’ On the other hand, a stable aircraft tran-
sient response requires accurate rate information (velocity) to generate the
necessary dynamic damping so that the aircraft will not oscillate about the
desired course. Since low-noise rate information cannot always be derived
from the position-fix information, accurate rate information may be derived
from dead reckoning measurements. An accelerometer output, for example,
can be integrated once to provide the velocity. In this case dead reckoning
measurementsfrom inertial sensors aid the VOR/DME position-fix guidance.
This augmentation is called inertial aiding. In the context of the above
discussion, both velocity data as well as position data are necessary to control
an aircraft accurately along its desired flight path.
Complementing position-fix data with dead reckoning information and
vise versa is a notion which will be encountered again in detail in Section V. It
is one of the most important solutions to the problem of obtaining accurate
navigation, particularly during the approach/landing applications.
As stated in Pierce (1948), navigation does not consist merely of the
determination of position or the establishment of a compass heading to be
followed. Navigation requires the exercise of judgment; it is a choice of one out
of many courses of action that may lead to the required result based on all
available data concerning position, destination, weather, natural and artificial
hazards, and many other factors. Therefore, there can be no electronic
navigational system-only electronic aids to navigation. An aircraft may be
automatically guided along a predetermined course by the use of equipment
that performs the dead reckoning function, or may be made to follow a line of
position known to pass through a desired objective. Neither of these
achievements constitutes navigation by equipment. The automatic devices
simply extend the control exercised by the navigator in time and space.
A navigation system comprises: (1) Sensors or navigation aids that
generate primary guidance parameters; (2) a computer that combines the
guidance parameters to obtain the aircraft position; (3) flight path commands
(the path to be followed) that, when contrasted with the aircraft’s present
position, generate the steering commands that define needed corrections in
flight path; and (4) a flight control system, which causes the aircraft to correct
its flight path while remaining stable. Figure 4 functionally describes the
principal sensor and position calculation elements for a two-dimensional(2D)
lateral guidance navigation system. Three-dimensionalflights would include

Alternatively, filtering the sensor output data is not always a viable solution because the
added phase delays may reduce the aircraft flight control system’s stability margin.
DISTANCE MEASURING EQUIPMENT IN AVIATION 19
the aircraft’s altitude which would be used to derive the vertical guidance
component. A navigation system provides output information in a variety of
forms appropriate to the needs of the aircraft. If the information is to be
directly used by the pilot, it involves some type of display; other outputs can be
steering signals sent directly to the autopilot. In some applications such as
manual flights along the VORTAC radials, the sensor outputs-range and
bearing-are sent directly to the displays without additional data trans-
formations and computations.
Subsection C,3 will describe how the outputs of the navigation system are
used to “close the loop” with the aircraft to form the total aircraft flight
control/guidance system. Only position-fix navigation systems will be treated.
Bearing “only” navaids, such as the airborne direction finder (ADF) using
nondirectional beacons (NDB),and not discussed in this article (see Sandretto,
1959).

2. Principles of Navigation Parameter Measurement:


Range and Bearing
Electronic navigation aids are defined as those systems deriving naviga-
tionally useful parameters-for example, range and bearing- through the
application of electronic engineering principles and electronic technology.’
Following the partition developed in Sandretto (1959), these systems are of
two general classes: the path delay and the self-contained. Path delay systems
are characterized by employment of at least one radio transmitter and at least
one radio receiver. The transmitter emits energy that travels to the receiver.
The navigational parameter is derived by measurement of the delay incurred
in the transmission. All path delay systems are based on the assumption that
radio wave propagation is rectilinear and that its velocity of propagation is
constant. Therefore, parameters such as phase delay or pulse arrival time delay
between a reference and a distant object can be measured and made
proportional to range or bearing. The self-contained systems consist of devices
that sense certain natural phenomena and, with the aid of computers, derive
navigational parameters. The INS is such a system.
Path delay systems are characterized by the various types of position lines
which they produce. There are only two fundamental types: the single-path
and the multiple-path system. Figure 5 notes the two main variations of each
of these types. The single-path system measures absolute transmission time
and produces circular “lines of position”- that is, a single-path navigation
system determines aircraft range. DME is such a system.

The terms navigation aid (navaid)and navaid sensor are used synonymouslyin this article.
In the system context, a radio navaid is a sensor which includes both the ground and airborne
elements and thus includes cooperative techniques such as DME. Traditionally,MLS, ILS, and
VOR are characterized by a signal in space (ground element) and a sensor (airborne element).
20 ROBERT J. KELLY AND DANNY R. CUSICK
NAVAIDS USED I N ATC SYSTEM

I
SELF-CONTAINED SYSTEMS
MEASUREDELAYOVER OBSERVES NATURAL
TRANSMISSION PATH PHENOMENA & COMPUTES
POSITION FIX

I
FlFl RECKONING

SATELLITE SINGLE PATH


(CIRCULAR LOP)

I
MULTIPLE PATHS

+ ’
PATH DELAYS
COMPARED
.CELESTIAL
FIXES
(NOT USED I N mzri:R
.INS
RNAV

FI
SYSTEM) NAVIGATOR

I 4 .AIRDATA

TRUE HYPERBOLIC
ONE-WAY
FOR EQUAL DELAY
(HYPERBOLIC LOP) (RADIAL LOP)

VOR
RADAR BEARING
DME ILS

-
MLS ANGLE
LOP LINES OF POSITION .TACAN BEARING

‘UNDER DEVELOPMEN7

FIG.5. ATC navaid family tree.

The multiple-path system measures difference in, or compares, transmission


times. Since the locus of all points having a constant difference from two other
points is a hyperbola, these systems should produce lines of positions which
are hyperbolas. Many of these mutliple-path systems are so instrumented that
they can only determine when the differencein transmission time is zero (that
is, when the times are equal). In this case, the line of position is a hyperbola of
zero curvature, or a striaght line. That is, the locus of all points that are
equidistant from two other points is a straight line (the bisector). These
simplified multiple-path systems are known as radial systems; those with
nonzero time delays are called true hyperbolic systems. Radial systems
measure aircraft bearing. Examples are VOR and the TACAN bearing signal.
True hyperbolic systems include DECCA, LORAN, and OMEGA (Kayton
and Fried, 1969). In summary, multiple-path systems must use transmissions
which differ in either time or frequency.
a. Single-Path Systems. The simplest system in concept is the single-path,
one-way system; it employs a transmitter at one point and a receiver at
DISTANCE MEASURING EQUIPMENT IN AVIATION 21
another. Transmission occurs over a single path between the two points. The
delay in transmission over this path is the important element in securing the
navigational parameter. The parameter which is delayed, for example, could
be the phase of a subcarrier whose frequency is maintained very accurately.
The receiver similarly employs an oscillator matched in phase to the
transmitter subcarrier. The transmitter-receiver phase difference, therefore, is
a measure of the time and, therefore, the distance taken by the wave to traverse
the transmitter-to-receiver distance. Such systems are called coherent systems.
They require very stable references such as atomic clocks and are not used in
the navigation systems discussed in this article.
A second type of single-path system employs a round-trip path as shown in
Fig. 6. Energy from the transmitter illuminates a “reflector.” A reflector may be
passive, such as an airplane hangar, or active, such as a radio beacon. The
reflected energy is returned to the originating transmitter location where its
delay (time of arrival) is compared with the initial reference time. Con-
ventional radar applies the passive reflector concept, while DME and the
ATC secondary radar are based on the beacon idea.
The two-way, single-path system overcomes the costly reference problem
associated with the one-way, single-path system because only one reference
oscillator with greatly reduced frequency stability characteristics is necessary
to achieve a coherent system. Thus, the stringent frequency stability require-
ments for coherent signal processing need be maintained only over the
equivalent of several round-trip delay periods. Coherent signal processing is
desirable when high accuracy performance is necessary. It permits both the
amplitude and the phase information in the reflected echo to be used in
determining the target range and velocity. However, in systems where the
accuracy requirements are modest, such as the DME application, noncoher-
ent signal processing is sufficient. Frequency stability requirements can then
be reduced even further. Noncoherent systems measure only the time delay of
the return signal’s pulse envelope, as illustrated in Fig. 6. In these systems the
frequency spacing of the rf channel usually dictates the frequency stability
requirements.
As mentioned in Subsection B, the conventional airborne radar is not an
effective navigation aid because it cannot easily identify ground reference
points. Position-fix navigation using beacons can easily identify the ground
reference landmark via the channel plan using a unique radio frequency
and/or pulse code.
Because radar-to-beacon and beacon-to-radar transmissions are each one
way, the signal power varies as the inverse square of the range, rather than as
the inverse fourth power, as do ordinary radar echoes. This means that the
range of a radar beacon may be doubled by increasing the radar transmitter
power or the receiver sensitivity fourfold, whereas a 16-fold increase would be
+
22 ROBERT J. KELLY AND DANNY R. CUSICK
d = DISTANCE BETWEEN RADAR AND REFLECTOR
c = VELOCITY OF LIGHT
At = 9 ;PULSE LEADING EDGE DELAY MEASUREMENT
TRANSMITTED ECHO OR
PULSE ENVELOPE

W E IS NON COHERENT RANGE SYSTEM (PULSE ENVELOPE)

COHERENT RADAR MEASURES BOTH AMPLITUDE AND PHASE


TO OBTAIN TARGET RANGE AND RANGE RATE

MEASUREMENT
PHYSICAL
REFLECTOR
RECEIVER

4 - - - - d d

MEASUREMENT

RECEIVER
- REFLECTOR
(BEACON)
TRANSMITTER

FIG.6. Two-way single path systems

required to double the range of ordinary echoes. Consequently, beacon


systems are relatively modest in size and require less transmitter power than
conventional radar.
b. D M E : A Single-Path Two-way Navaid. For the above reasons, the DME
is based upon the beacon principles. Ground transponders are used to mark
and identify known positions on the ground so that the aircraft can determine
its position relative to them. Such a system is cost-effective because the user
requires only the purchase of the airborne unit, while the ground equipment is
DISTANCE MEASURING EQUIPMENT IN AVIATION 23
purchased and maintained by local, state, or federal governments. The system
design permits low-cost airborne equipment having small weight and size for
two reasons. One, the codes and different radio frequencies permit omni-
directional antennas, and two, the transmitter power requirements have to
overcome only the inverse square of the range propagation loss as noted
above. System costs are further reduced because the two-way, single-path
system does not require a coherent reference. That is, the operational accuracy
(0.2 nautical miles for DME/N and 0.017 nmi for the DME/P) can be achieved
with simple noncoherent time-of-arrival “leading edge” measurements as
indicated in Fig. 6. In addition, since the system does not have to be coherent,
the uplink radio frequency can be different than the downlink without
degrading the system, thus permitting a flexible channel plan while contribut-
ing to the low cost of the airborne equipment. Finally, the beacon concept
permits the use of an omnidirectional antenna on the ground, allowing 360”
azimuth coverage in the horizontal plane with low transmitter power at a
sufficient data rate. Aircraft can then obtain range guidance at all points
necessary for enroute, terminal, and landing applications.
c. Radial Navaids. Radial systems used in the ATC are radio navaids
providing bearing information. They generate a zero delay time by having
coincident paths (same ground antenna). This means that the two path signals
must be coded differently: A reference signal is required to establish zero time
delay while a second signal provides the variable delay. Figure 7 illustrates the
principle of the radial system used in VOR and TACAN. The rotating
cardioid pattern generates a sinusoidal signal, which is a delayed version of the
reference signal. In order to separate the reference from the signal, it must vary
in time or have a different subcarrier modulation f ~ r m a t . ~
Radial systems used in ATC system are VOR, TACAN (bearing), ILS, and
MLS (angle).The VOR and TACAN derive bearing by measuring phase delay
between the reference and the signal, which are diverse in subcarrier
modulation formats. In MLS (angle) this bearing information is proportional
to a time delay. In ILS the signals are diverse in frequency, and bearing is
derived from the amplitude differences of the two signals.
d. Radio Navaid Position-Fix Conjigurations. Radio navaids whose locus of
points define a circle are called p systems, while those whose locus of points
define radials are called 8 systems. The four systems used in the ATC system
are summarized in Fig. 8 in terms of the p/8 navigation parameters. Not
In terms of performance,it is noteworth that two-way systems have their reference signal in
the aircraft. The reference clock (delay) is “turned on” with the interrogation transmissionand is
“turned off” with the reply signal. The elapsed time on the clock is proportional to the range.
Unlike radial systems, the reference element in two-way range systems is not corrupted by
propogation effects and receiver noise.
24 ROBERT J. KELLY AND DANNY R. CUSICK
NORTH

GROUND STATION
RADIATES BOTH
REFERENCE AND
ROTATING CARD10

CARDIOID ANTENNA
PATTERN ROTATES
AT FREQUENCY f

INFORMATION
SIGNAL (CARDIOID)

FIG.7. Principle of radial navaid for VOR and TACAN.

shown in Fig. 8 is the VOR/DME configuration, which is not widely


implemented in the enroute application.
The ATC enroute and terminal navaids can determine position fixes by the
intersections of radials and circles (see Fig. 9). In the 2 D navigation
application, the slant range from the ground station (DME or VOR) is
projected onto a plane tangent to the earth. This plane is illustrated later in
Fig. 10. Thus the lines of position shown in Fig. 9 are assumed to also lie in the
tangent plane. In ILS, the position fixes are determined by the intersection of a
sphere (DME), a vertical plane along the runway center line (localizer),and a
lateral plane perpendicular to the localizer plane (glide path). For MLS, the
point in space is the intersection of two cones (azimuth and elevation scanning
DISTANCE MEASURING EQUIPMENT IN AVIATION 25
-
AIR
DMEIP
DMEIP
IIA)

GROUND

NOTE:
NOT SHOWN IS A VORlDME CONFIGURATION

FIG.8. Air-derived p/B navaids in the ATC system.

beams) and a sphere (DME/P). Wax (1981) demonstrates how several different
position-fix techniques can be integrated into an algorithm to estimate an
aircraft’s position.
All the navigation aids described above are air-derived systems; that is, the
relevant navigation parameter is calculated in the aircraft. The Secondary
Surveillance Radar is an example of a ground-derived system; that is, the
relevant information is derived on the ground and then transmitted to the
aircraft via a communications link.
The global positioning system (GPS) is a new position-fix system which
provides three-dimensional position and velocity information to users
anywhere in the world. Position fixes are based upon the measurement of
transit time of rf signals from four of a total constellation of 18 satellites
(Milliken and Zoller, 1978). The exact role that GPS will play in the future
NAS has not yet been defined. One of the tasks of RTCA Special Committee
159 is to define that role (RTCA, 198%). As stated by the Department
of Defense/Department of Transportation (DOD/DOT) (1984), the DOD
intends to phase out its Air Force TACAN systems beginning in 1987. They
will be replaced with GPS.

3. Additional Navigation System Considerations


The purpose of this section is to complete the general discussion of
navigation principles by describing how a navigation system is “connected” to
the aircraft to form a closed-loop control system. The central notion represents
26 ROBERT J. KELLY AND DANNY R. CUSICK

VO R IDME
OR TACAN

VOR/VOR

FIG.9. Position fixes determined by p / B intersection.

an aircraft in flight as a time-varying velocity vector. It has a direction in which


it is going and a speed at which it is going there. The time integral of this
velocity vector is its flight path. The navigation system’s role is to generate
position and velocity corrections so that the aircraft’s velocity vector can be
altered to follow the desired flight path in an accurate and timely manner.
To continue the conceptual development of navigation principles further
requires a coordinate system to reference the desired flight path and the
navigation system’s output signal. Since the aircraft’s guidance system is useful
only if the aircraft closely tracks the desired flight path, accuracy performance
definitions are necessary to compare the AFCS’s tracking performance. This
subsection briefly describes coordinate reference frames and accuracy con-
siderations as they apply to a closed-loop navigation system.

a. Coordinate System. An aircraft in flight has a position and velocity at


some altitude above the earth at some instant in time. In order to specify these
aircraft states, a reference coordinate system needs to be defined. The subject
of navigation coordinate frames is a vast and difficult topic and will not be
pursued in any depth here. For example, seven reference systems, useful to
navigation are defined in Chapter Two of Kayton and Fried (1969). One of
these systems, geodetic spherical coordinates, has wide application in the
mechanization of dead reckoning and radio navaid systems. It gives the
aircraft lateral position in terms of longitude I and latitude CD on an ellipsoidal
(earth), where I and @ are measured in degrees. The coordinate system is used
DISTANCE MEASURING EQUIPMENT IN AVIATION 27
in current RNAV equipment with the waypoints also being defined in A and (D.
The vertical dimension is given by altitude h above the reference ellipsoid. See
Appendix D of RTCA (1982).
Navigation system designers assume that the earth‘s surface can be
approximated by an ellipsoid of rotation around the earth’s spin axis. In this
representation it is postulated that, when the earth cooled from its molten
mass, its surface assumed contours of equipotential gravity. The resulting
gravitation field g is very nearly perpendicular to the earth’s surface. It is called
the “apparent” gravity and is the vector difference between the Newtonian
field of gravitational attraction and the centrifugal forces due to the earth‘s
rotation. The reference ellipsoid is chosen to minimize the mean-square
deviation between the normal to the ellipsoid and g. In order to keep the
narrative simple and to focus on the essential ideas, the normal to the ellipsoid
will be assumed to lie along g (the direction that a “plumb b o b falls). Use of
the geodetic coordinates which are given in degrees allows the lateral position
(say A. and (Do) of the aircraft above the earth to have the same coordinates
(Ao, (Do) on the earth’s surface. The aircraft’s altitude is its height above the
reference ellipsoid as measured along the ellipsoid’s normal (i.e., approxi-
mately along g).
Enroute navaids provide guidance so that the pilot can fly between two
waypoints specified in terms of A and (D. For example, assume that the
waypoints were at the same height. Steering commands would be generated by
the navigation system to guide the aircraft along the geodetic bearing and
distance. The aircraft’s trajectory would trace an arc following the earth’s
curvature at a constant attitude, as shown in Fig. 10(a).
For short-range navigation (50 to 100 nmi for enroute and 20 nmi for
approach and landing), the lateral position can be referenced to Cartesian
coordinates instead of to the geodetic coordinates. As a result both the
calculations and the geometric picture of the flight path are greatly simplified.
This is accomplished by using a tangent plane to approximate locally the
ellipsoid, as shown in Fig. lqa).
It can be shown that, for short-range navigation, the length of the aircraft’s
ground track on the earth’s surface will not deviate significantly from the
length of the flight path’s projection onto a plane tangent to the earth’s surface
near the derived waypoints. At a distance of 50 nmi from the tangent point the
difference in length is about 16 ft. Measuring the aircraft’s height as the
perpendicular distance to the tangent plane would, however, deviate signifi-
cantly in an operational sense from that height measured along the normal to
the earth’s surface. At 50 nmi the altitude difference is about 1300 ft, which is
unacceptable when compared with the lo00 ft aircraft separation distance
defined for the ATC route structure.
Thus, for short-range enroute and terminal area navigation, lateral
guidance can be defined with respect to the tangent plane. However, the
28 ROBERT J. KELLY A N D DANNY R. CUSICK
A/C FLYING
CONSTANT ALTITUDE
\
_c--

TANGENT PLANE

9
(a)

AIRCRAFT CM

i
9

FIG.10. Earth reference coordinate system.


DISTANCE MEASURING EQUIPMENT IN AVIATION 29
vertical guidance must be measured along the normal to the earth’s surface
using, for example, a barometric altimeter. For enroute navigation it is the
aircraft’s altitude directly above the earth‘s surface which is operationally
important, not the aircraft’s height above the tangent plane reference point. In
the final approach landing application the operationally significant coor-
dinate system is referenced to the tangent plane which contains the runway.
Thus the aircraft’s position can be given in xyz Cartesian coordinates. The
decision height at the terminus of the final approach is measured with respect
to the glide path intercept point (GPIP) on the runway (see Subsection D,4).
The elevation angle glide path is chosen so that the aircraft will be safely above
the hazardous obstacles along the approach path. This article, then,
will define a very simple Cartesian coordinate system that permits description
of the most essential concepts of navigation, guidance, and control. The
Cartesian reference frame is defined at some point on the earth’s surface (e.g.,
airport runway). As shown in Fig. 10(b) this fixed Cartesian reference frame is
called an earth-fixed reference. A tangent plane is defined through this point.
Two of the Cartesian coordinates lie in the tangent plane, one of which lies
parallel to the meridian (North). A third coordinate lies along the earth’s
gravity vector and is normal to the tangent plane. It defines the aircraft altitude
above the tangent plane. However, as long as the aircraft’s altitude is
determined consistent with the enroute, terminal area, and approach/landing,
then the earth-fixed reference will accurately describe the aircraft flight path.
In other words, the tangent plane altitude calculation must be corrected for the
earth’s curvature during enroute flight missions. The earth-fixed reference,just
like the geodetic reference, permits the navigation problem to be separated
into aircraft lateral positions and altitudes.
Making a second assumption that the earth is fixed in space allows an
inertial frame of reference to be defined on the earth’s surface; that is, unknown
acceleration effects are neglected (McRuer et al., 1973). This inertial frame,
which is accurate for relatively short-term navigation, guidance, and control
analysis purposes, permits a description of aircraft motion in terms of inertial
sensor^.^ It does have practical limitations for very long-term navigation
(worldwide).
With respect to this earth-fixed coordinate system, the aircraft’s center-of-
mass (c.m.) can now be specified by a position vector with components
(Xa,x,Za).In addition, three orthogonal axes exist within the aircraft. They
are the pitch, roll, and heading axes of the airframe. The origin of this body-
fixed triad “sits” on the tip of the aircraft position vector, which is coinci-
dent with the aircraft’s c.m.
Aircraft equipped with INS can establish a local vertical using Schuler tuning (Schneider,
1959), thereby forming the required inertial reference. Accurate position data can then be obtained
from the platform stable accelerometers.
30 ROBERT J. KELLY AND DANNY R. CUSICK

The attitude of the aircraft is defined by the Euler angles (a, /?,and y )
through which the earth-fixed axis must be rotated to be coincident with the
airframe body-fix axis.
Understanding aircraft navigation techniques cannot be confined to just
the motion of the aircraft’s c.m. Attitude must also be considered, because
the flight path traced out by the motion of the aircraft c.m. is controlled by
changes in the aircraft’s heading, roll, and pitch. The navaid guidance signal
is also referenced to the earth-fixed system so that the steering commands will
be in the correct form to modify the aircraft’s velocity vector when necessary.
Subsection E will show that the AFCS for conventional fixed-wing aircraft
has two orthogonal channels: a lateral channel and a longitudinal (vertical)
channel. Since the earth-fixed reference frame is an orthogonal system, as
described above, the lateral navigation problem can be separated from the
vertical navigation problem as a consequence of this simplification. This is an
important distinction because later it will be shown that the lateral enroute
navigation control laws are similar in form to the lateral (azimuth) ap-
proach/landing control laws. With this simplifying division of the navigation
problem, the narrative and the examples given in the balance of this article will
be selected with respect to the lateral navigation application, wherein the
aircraft’s altitude is assumed to be constant. When the aircraft’s altitude and

TRUE N O R T H
4

= HEADING
e= BEARING
SIDE SLIP ANGLE = 0

FIG.11. Definition of true track angle during coordinated flight.


DISTANCE MEASURING EQUIPMENT IN AVIATION 31
time are included as steering commands, the navigation problem becomes a
3D or a 4D problem, respectively.
The following terms-bearing, heading, ground track velocity, and track
angle-are important for the understanding of lateral navigation discussions.
As shown in Fig. 11,the bearing 8 is the angular position of the aircraft c.m. to
its intended destination as measured clockwise from a reference such as true
North. Track angle tjT is the true angular direction of the aircraft’s ground
velocity. The heading tj is the angular direction that the longitudinal axis of
the aircraft is pointing. Air speed is the speed of the aircraft in the direction of
its heading through the air mass. From these definitions, three distinctions in
the aircraft flight path can be given. If range to a destination can also be
determined, then a desired flight path (course) can be defined as a connected
time series of intended bearing and range points. The aircraft’s actual flight
path would be a connected time series of true track angles and true ranges,
while the indicated flight path is a time series of range and track angle
measurements.
As part of a recurring theme in this article, track angle or ground velocity is
the critical parameter for enroute navigation. When accurate navigation is
necessary, it can be determined from a series of position measurements or it
can be derived from accelerometers on board the a i r ~ r a f t Guidance
.~ and
control are simply the corrections of the velocity vector in such a manner
that the aircraft follows its intended flight path. Clearly, if air speed and
heading are known, then, as indicated in Fig. 11, a simple but less accurate
method is available; one uses the wind vector and the air speed vector to
obtain the ground speed velocity.

b. The Aircraft Flight Control/Guidance System. The task of a navigation


system is not only to derive the aircraft’s position, but also to generate com-
mands to change its velocity vector. In order to accomplish these changes,
its output signals are fed into the aircraft flight control system, thereby forming
a “closed loop” composed of the aircraft and the navigation system. This
“closed loop” is called the aircraft flight control/guidance system. If the
navigation system correction signals are sent to a display, then the loop is
’ The central concept of INS consists of a platform suspended by a gimbal structure that
allows three degrees of rotational freedom (Maybeck, 1979).The outermost gimbal is attached to
the body of the aircraft so that the aircraft can undergo any change in angular orientation while
maintaining the platform fixed with respect to, for example, the earth-fixed coordinate frame.
Gyros on the platform maintain a desired platform orientation regardless of the orientation of the
outermost gimbal. Thus, the platform remains aligned with respect to a known reference
coordinate system. Accelerometers on this stabilized platform can then provide aircraft
acceleration with respect to that known set of reference coordinates. After local gravity is
subtracted, the resulting signals can be integrated to yield aircraft velocity and acceleration. An
alternative to the gimbaled implementation is the strap-down inertial system (Huddle, 1983).
-
32 ROBERT J. KELLY AND DANNY R. CUSICK
(a) MANUAL-COUPLED FLIGHT
- 1
AIRCRAFT
AIRCRAFT POSIT1ON
A
COURSE
DEVIATIONS
NAVIGATION
SYSTEM
.
~

NAVAID SENSOR
ENROUTE
DME
VOR
DESIRED FLIGHT PATH
APPROACH &
LANDING
I LS
M LS

(b) AUTOCOUPLED FLIGHT

AIRCRAFT
POSITION

STEERING

DESIRED FLIGHT PATH


FIG.12. Principal aircraft flight control/guidancesystems.

closed through the eyes and hands of the pilot, as shown in Fig. 12(a). When
the pilot closes the loop, the configuration is called a manual-coupled flight
control system (FCS). Similarly, if the guidance signal is sent to an autopilot,
then the configuration is called an auto-coupled FCS, as shown in Fig. 12(b).
Guidance then involves the instrumentation and control of the six aircraft
coordinates. For the 2D problem, the guidance system provides corrections to
the aircraft’s velocity vector by changing its bank angle and heading, such that
the aircraft’s c.m. follows the desired 2D flight path. The aircraft’s c.m. velocity
is called the ground track velocity. When altitude corrections are included,
the aircraft velocity is a 3D vector composed of the ground track velocity
plus its vertical speed.
The navigation system outputs for the autocoupled 2D enroute appli-
cation are lateral steering commands. For manual-coupled flight, the visual
display will present bearing and range guidance. To maintain the intended
DISTANCE MEASURING EQUIPMENT IN AVIATION 33
flight path the guidance corrections are displayed as (1) range along the
straight-line segments between selected waypoints, and (2) bearing or cross-
course deviations (left/right) normal to the straight-line segments. The
guidance correction signals for autocoupled flight are steering commands
expressed in terms of cross-course deviations. The cross-course corrections
are coupled to the lateral channel of the autopilot. Some aircraft are
instrumented to adjust their power (thrust) to maintain a range schedule along
the flight path.
For the straight-in final approach and landing application, the aircraft
velocity vector lies in and points down an inclined plane called the glide path.
The navigation system, or more accurately, the landing guidance system,
shown in Fig. 12, is now an ILS or MLS. Both use DME and both are
described in Subsection D,4.
For manual flight on final approach, landing guidance is displayed as
left/right indications on the course deviation indicator (CDI),which are cross-
course deviations about the extended runway center line. Vertical deviations
about the inclined plane are displayed on up/down indications on the CDI.
The CDI is also used in enroute applications. Lateral steering corrections in
autocoupled flights are sent to the lateral channel of the autopilot, and vertical
corrections are sent to the autopilot’s longitudinal or vertical channel.
Fixed-wing, conventional take-off and land aircraft will be considered in
the closed-loop aircraft/flight control system discussions and analyses. The
other major aircraft types, short takeoff and land (STOL) and vertical takeoff
and land (VTOL), will not be considered. However, the ATC navaids
summarized in Subsection D (especially MLS and its DME/P) can and do
provide navigation guidance to both the STOL and VTOL aircraft types.
c. Navaid Accuracy Considerations. In navigation, the accuracy of an esti-
mated or measured position of an aircraft at a given time is the degree of
conformance of that position with the true position of the aircraft at that time.
Because of its statistical basis, a complete discussion of accuracy must include
the following summary of the statistical assumption made and the measuring
techniques used.
A navigation system’s performance must be evaluated in the context of the
total system, aircraft plus navigation system. Several measures of a navigation
system’s performance require that the aircraft flight path remain inside its
designated air lanes, and that the pilotage error of the aircraft about the
indicated navigation signal as well as its attitude be maintained within
prescribed limits. Keeping the aircraft within its air lane is directly related to
the navigation system’s bias accuracy; while maintaining a small tracking
error and prescribed attitude are direct functions of the closed-loop systems’
bandwidth, autopilot gain, and the navigation system’s noise error. Under-
standing the performance of the closed-loop guidance system requires an
34 ROBERT J. KELLY A N D DANNY R. CUSICK

understanding of the navigation system’s error mechanisms. The navigation


system, in turn, is a configuration of navaid sensors, a computer, and output
signals that drive an autopilot or a cockpit display, as shown in Fig. 4.
Determining the performance of the navigation system, therefore, also entails
determining the performance of the navaid. This section discusses the
necessary considerations.
The navaid sensor measures some physical phenomena such as time delay
of the electromagnetic radiation. As indicated in Fig. 4, the sensor measure-
ments are then transformed to a coordinate system where they are combined
by the navigation system to estimate the aircraft’s position and velocity vector.
The aircraft’s position, as determined by the navigation system, will differ from
its true position due to errors in the sensor measurements, assumptions in the
earth coordinate system reference model, errors in computation, and, finally,
departures in the mathematical model implemented and the real world. This
difference is called the RNAV system error, as shown in Fig. 13. The figure
also indicates an additional error, namely the flight technical error, which
combines with the previously addressed Navaid system error to produce the
total system error or tracking error. The flight technical error (or autopilot
error for autocoupled flight) is the difference between the indicated flight path
and the desired flight path. The “flight technical error” (FTE) refers to the
accuracy with which the pilot controls the aircraft as measured by his success
in causing the indicated aircraft position to match the indicated command or
desired position on the display. FTE error will vary widely, depending on such
factors as pilot experience, pilot workload, fatigue, and motivation. Blunder
errors are gross errors in human judgment or attentiveness that cause the pilot
to stray significantly from his navigation flight plan, and are not included
in the navigation system error budget. Blunder tendency is, however, an
important system design consideration; their effects are monitored by the ATC

AIC TRUE A/C TRUE


/FLIGHT PATH / POSITION

OESl RE D
ERROR ERROR COURSE

RNAV ROUTE
WIDTH
WHERE RNAV
SAYS A/C IS FLIGHT TECHNICAL ERROR
RNAV
FLIGHT PATH
DESIRE0 A/C
POSIT ION

FIG. 13. RNAV position error.


DISTANCE MEASURING EQUIPMENT IN AVIATION 35
controller. Roscoe (1973) discusses flight test results which attempted to
measure flight technical errors and blunder distributions. To complete the
picture, the total system error (tracking error) consists of the navigation
system error and the flight technical error; it is the difference between the
aircraft’s actual position and the desired flight path.
An example of a navigation system error budget is given in Table V and
illustrated in Fig. 40 in Subsection F. A navigation system can be the
sophisticated configuration of sensors, as shown in Fig. 4, or it can be simply
the navaid measurements themselves, as in VOR/DME, where bearing and
range are directly displayed to the pilot.
As discussed previously, the intended aircraft’s position (c.m.)as well as its
measured position are determined relative to the earth-fixed coordinate
system. The position errors can be defined in at least two differentways: They
can be referenced directly to the earth-fixed coordinates or they can be
referenced to the aircraft’s intended flight path, which, in turn, is referenced to
the earth-fixed system. To keep the discussion focused on the important
concepts, only the two-dimensional plane used in lateral navigation will be
treated. (The vertical component, altitude, is simply a one-dimensional
measurement, orthogonal to the lateral plane.)
In the first method, the error components are traditionally defined in terms
of a location error (bias) and a dispersion error (variance). When the error
components are defined along the coordinates of a Cartesian system, the
position error dispersion (variance) is given in terms of the circular error
probability (CEP). The CEP is the radius of a circle such that the probability
is 0.5 that an indicated position will lie within the circle (Burt, et al., 1965).
The center of the circle is chosen at the center of mass of the probability
distribution, which is the desired destination point of the aircraft. For a biased
system, the center of the circle will be displaced from the aircraft’s desired
destination.
The second method measures 2D navigation system performance in terms
of its error about the intended flight path. It is the method used in this article to
evaluate the accuracy performance of navigation systems. As shown in Fig. 13,
the error normal to the indicated flight path is called the cross-track (XTRK)
error, while the error along the indicated flight path is called the along-track
(ATRK) error. These error components are thus total system errors or
tracking errors.
Referencing the aircraft’s true position against the desired flight path is the
natural method of specifyingnavigation errors. The cross-track error must be
consistent with the width of the enroute air lanes or RNAV routes, while the
along-track error must not degrade the aircraft longitudinal separation
criteria from other aircraft. This cross-track/along-track theme is also present
in the straight-in approach and landing application where, for example, the
36 ROBERT J. KELLY AND DANNY R. CUSICK

cross-track errors are the lateral azimuth errors. In this case, the allowed
errors must be consistent with the runway width, while the along-track errors
must be consistent with the runway touchdown zone. In the simple
VOR/DME navigation application, the VOR error is the cross-track error,
while the DME error is the along-track error.
The aircraft path deviations-cross-track (lateral), along-track, and
vertical-are assumed to be orthogonal random variables. Since most of the
operational accuracy questions in this article address only one of these
components, the statistical problem is one dimensional. For example, the 2D
navigation application usually addresses only the cross-track corrections.
Therefore, the 95% confidence limits are one-dimensional statistics using the
variance and probability distribution of the cross-track deviation measure-
ment. On the other hand, when analyzing a 3D navigation flight path,
additional considerations arise because the error surface can be envisioned as
a tube in space. In this case, the 95% probability error contour corresponding
to the tube’s cross section is determined using a bivariate distribution of the
lateral and vertical flight path deviations. [See Table 2.1 in Mertikal et al.
(198511.
To summarize the tracking error or total system accuracy is a combination
of the autopilot accuracy and the navigation system accuracy. The accuracy
performance of navigation systems is specified in terms of 95% probability.
This means that for a single trip along a defined flight path (e.g., between two
waypoints) 95% of all position measurements must lie within the specified
XTRK error allowance. This implies that out of 100 flights, 95 of them must
satisfy the specified XTRK error allowance. In other words, the measurement
defines a confidence interval and its confidence limits are compared to the
accuracy standard. It is a confidence interval because the desired flight path is
an unknown constant parameter which the navaid is estimating. The
computed 95% confidence interval is a random variable which covers the
desired flight path 95 out of a hundred times. The confidence interval is used to
evaluate one-dimensional guidance signals (e.g., XTRK deviation).
Discussion will now center on the navaid error components which are a
subset of the navigation system components. The navaid sensor measurement
differs from the true navigation parameter due to variations to the pheno-
mena (e.g., time delay errors induced by multipath) and due to errors in
measurement and computations (instrumentation errors).Navaid errors fall in
three separate categories: outliers, systematic, and random. Outliers are the
results of miscalculations, electrical power transients, etc., which can be
detected by comparing a new measurement with respect to the three or four
standard deviation values of past position estimates (see Section IV,D).
Systematic errors obey some deterministic relation, such as a circuit temper-
ature dependency, or are the result of faulty calibration procedures. They
DISTANCE MEASURING EQUIPMENT IN AVIATION 31

usually can be removed when the deterministic relation is known or the


calibration procedure is corrected. In other words, if an ensemble of
navigation instruments were known to produce nonzero average errors,
simple recalibration or computation could nullify these effects. No generality
is lost, therefore, in assuming zero mean ensemble errors. Random errors are
unpredictable in magnitude and they are best described by statistical methods.
It is these errors which are addressed by the navaid accuracy performance
standards. Other accuracy types, such as predictable, repeatable, and relative
accuracy are defined in DOD/DOT (1984). Predictable and repeatable
accuracies are included in the error components described next and in
Subsection F. Relative accuracy does not apply to the applications addressed
in this article.
In summary,it is assumed that both the outliers and systematicerrors have
been removed and that only the random errors remain. This means that
although any particular navaid can have a bias AxB,the average bias of the
ensemble of navaids is zero, i.e., E[AxB] = 0, where E is the expectation
operator. Further, each navaid has an equipment variance (noise)component
a; that is assumed to be identical for each member of the class of navaids.
Therefore, as with the navigation system as a whole, the navaid errors may be
specified in terms of bias and noise (variance). The assumptions above are
important, and they are discussed further in the next section.
Once equipment errors have been accounted for, the only remaining major
error component is site-dependent errors such as multipath. Unlike equip-
ment noise errors, which vary with time but are essentially independent of
aircraft motion or position, site-dependent errors remain unchanging for long
periods of time, while exhibiting extreme variations as the aircraft changes
position. Thus site-dependent errors are said to be spatially distributed. A
time-varying error is generated when the aircraft flies through the spatially
distributed interference fields. It is implicitly assumed that the variance
component associated with site-dependent errors and with aircraft motion is
detected by all radio navaids having simple signal processing algorithms. The
degree to which this variance component is reduced depends upon the level of
output data filtering applied.

d. Error Budget Considerations. The purpose of an error budget is to allow


facility planning by local, state, and federal governments, including the ob-
stacle clearance needed to define enroute airways, terminal approaches, and
runway clearance areas. They are also used to prepare equipment procure-
ment specifications.Error budgets must be simple if they are to be useful. Also
they must be simple because there usually is no database to substantiate
anything more than simple statistical procedures. It is for this reason that the
root-sum-square (RSS) calculation procedure is used extensively throughout
38 ROBERT J. KELLY AND DANNY R. CUSICK
the navaid industry to estimate system performance. The purpose of this
section is to identify the assumptions upon which RSS calculations are valid.
Even though these assumptions may only be partially justified, the RSS
procedure is applied anyway because there is no other alternative short of
directly adding the maximum excursion of the error components. Direct
addition of the error components is not statistically or economically
defensible. The true navaid system accuracy lies somewhere between these two
calculation procedures.
The problem is how to use ensemble statistics which were obtained by
averaging over all terminal areas, equipment classes, and environmental
conditions to predict the performance of a given aircraft on a single mission to
a given runway. An estimate of the total system error is the mean square error
(MSE) or second moment. If uncorrelated ensemble statistics can be used and
the ensemble average of each error component is zero, then the value of each
component in the error budget (including the bias errors) can be given by their
respective variance. Under these conditions, the MSE can be easily calculated
by simply summing the variances and extracting the positive square root. The
MSE is thus the variance for the total system. In the literature, this calculation
is known as the root-sum-square.
The difficulty with the RSS approach is that the random variables which
represent the slowly changing drift errors may be correlated during flight
missions to the same airport. To understand the problem, consider the error
due to temperature effects. The equipment design engineer can determine the
maximum temperature drift errors from the specified operating temperature
range. Civil navigation equipments are usually designed to operate over
temperature ranges between 70 and - 50 "C. These excursions usually have
daily and seasonal cycles whose average values would be nearly zero when the
equipments are calibrated for the average temperature of a geographic area.
This means that, although the drift errors are a nonstationary process, they
will have a near zero average over a long time interval. This point can be used
to advantage because it implies that repeated flights by a given aircraft to the
same runway (for example) will at worst only be partially correlated. In any
event, the ensemble average over all conditions and missions will be nearly
zero. Although the value of the ensemble variance is less than the peak drift
error, it is common practice to use the peak drift value for the drift error
variance component. Thus the error budgets for some geographic loca-
tions may be pessimistic when the ensemble variance of the temperature
drift errors are added with the other error component variances. There is
generally no statistical correlation computation problem with the noiselike
errors.
Physically the bias error component and the noise errors are associated
with separate error source mechanisms. They can be viewed in the frequency
DISTANCE MEASURING EQUIPMENT IN AVIATION 39
domain as separate spectral components and thus as separate random
variables. The low-frequency component when grouped together are called the
bias error, while the sum of the high-frequency components is called the noise
error. Both spectral groups, bias and noise, are defined as random variables
having zero means and variances equal to otiasand 0.;
In principle, each flight mission can be considered as a statistical
experiment where each biaslike error component AxB is selected from separate
batches, each having a zero ensemble mean E [ A x B ]and variance otias.Before
each flight mission, one error component is drawn from each batch and is
directly added with samples from the noise component (a;) batches. The result
should, with high probability, be within the total specified system error
allowance.
In the next section the discussions above will be further refined to include
quantitative definitions of the time average, ensemble averages, and the RSS
calculations.
e. Statistical Dejnitions and Root-Sum-Square Applications. Navaid output
data can be viewed as a combination of a static or quasi-time-invariant
component (bias) and a dynamic or fluctuating component (variance). The
dynamic component or noise term can be further partitioned into spectral
components, corresponding to errors, which (1) aircraft can follow and are
called scalloping course errors or “bends”; (2) can affect the aircraft’s attitude;
or (3) have no effect on the AFCS. (The degree of data filtering performed
determines whether the noise component of a navaid disturbs the aircraft
position or its attitude or both.) DME/P and MLS angle guidance systems
define the noise error components in terms of these spectral components. ILS
defines the course errors in terms of bias and noise with no distinction made
between the spectral groups comprising the noise. Most navaids, such as VDR,
DME/N, and TACAN, relax the definition error further and make no
distinction between bias and noise. For these navaids the accuracy is described
by a single value equivalent to the mean square error. The MSE describes in a
rudimentary way the general intensity of random data. Before defining MSE,
two types of averaging processes, time averages and ensemble averages, will be
defined.
The accuracy performance of a class of navaids is determined by first
calculating the time average on one member of the class. To do this M time
sample measurements x ( t i )are taken over a short period of 10 to 40 seconds,
where i = 1, 2,. . .,M. From these measurements, M error terms A x ( t i )are
calculated. Using the M error terms, the time-average sample error AX^, the
time sample variance S i , and the time sample mean square error
40 ROBERT J. KELLY AND DANNY R. CUSICK

are determined. Figure 14(a) illustrates these concepts for a series of range
measurements from a s i n g l a M E interrogator/transponder combination.
The positive square E is called the RMS error (Bendat and Piersol,
1971) and is = ,/- when the measurements are
uncorrelated.
To obtain the ensemble average of the short interval time averages, time
measurements are repeated on (N - 1)additional units. The ensemble average
is then determined from the N sets of measurements. Let

where j is the jth navaid and the bar indicates the ensemble average
calculations. The sample variance of the bias measurements is
I N

In this discussion, both N and M are sufficiently large that the sampling errors
are negligible- and thus represent the population statistics for a given type of
navaid, i.e., AxB N E[AxBj], S;,, 2: dbiasr
2
S i s' o i , and MSE = E[Axs].
Further, it is assumed that all the bias measurements are uncorrelated, i.e.,
E[AXB~,AXB~] = E[AXBj12. AS assumed above, E[AxBj] = 0 and the noise
component of each navaid in the ensemble is equal, that is, a i j = dj?,k = o i .
Therefore, with these assumptions and as shown in Fig. 14(b), the ensemble
mean-square error is simply MSE = nj?,+ otias.Its positive square root is
defined as the root sum square, RSS = ,/=. The distinction in this article
between RMS and RSS is that when E[AxBj] = 0, the RSS is only a sum of
variances.
With the above-defined measurement procedures, it is now possible to
illustrate how both a navaid error budget and a navigation system error
budget are constructed using the RSS procedures. Let A X AR, and A F be
different error components of a navigation system such as RNAV. The XTRK
component of the sensor error could be represented by AYwhile AR and A F
could represent the RNAV computation error and the XTRK flight technical
error, respectively. Assuming a linear model, the system error AS is given by
AS = A Y + AR + AF. Assume further that these error components are zero-
mean, uncorrelated random variables; then the ensemble mean square is

BIAS NOISE

The total system error is simply the RSS = which itself is the standard
deviation for the total system as indicated in Fig. 14(b).
DISTANCE MEASURING EQUIPMENT IN AVIATION 41
X AXIS
b
PROBABILITY DME MEASUREMENTS
DENSITY FUNCTION

ARIANCE (NOISE)

DME
GROUND STATION @
2

r
Y AXIS
MEASUREMENTS ON A SINGLE
INTERROGATOR /TRANSPON DER
COMBINATION

(a) TIME AVERAGE MEASUREMENT

E [ A X j ] = O

MSE = E [ A X f]= UN2+ UBfAs

j = 1, 2 ' . ' ' . N NAVAID UNITS


j = 1.2 ' ' ' ' ' N NAVAID UNITS
RSS =

DME ENSEMBLE BIAS ERRORS DME ENSEMBLE MEAN SOUARE ERRORS

(b) ENSEMBLE AVERAGE MEASUREMENT


FIG.14. Range error definitions.
42 ROBERT J. KELLY AND DANNY R. CUSICK

For the DME/N, VOR, and TACAN area navigation systems, the total
system error is the single number, RSS = (a: + ... + a;)”’, where a:,
a:, . ..,a: are the individual variances for each of the P error components. In
ILS system error analysis, the RSS of the bias variance components are
calculated separately from the RSS of the noise variance components, as
indicated in Table 111. For DME/P and the MLS angle guidance, the error
components are defined in Sections E,7 and IV,C,1 and in Table 111.
Combining error components on an RSS basis is almost universally
adopted in the literature (RTCA, 1984, 1985a; DOD/DOT, 1984). They
implicitly assume that the error components satisfy a linear model and that
they are uncorrelated, zero-mean, random variables. The linear model as-
sumption further implies that there is no coupling between the navigation
system error (NSE) and the flight technical error (FTE). In fact, there is a small
nonlinearity between these terms because of the NSE changes with aircraft
position, which, in turn, is a function of the FTE. This nonlinearity is
negligible, and the NSE and FTE can be combined on an RSS basis if the pilot
has no a priori knowledge of the NSE (e.g., severe VOR course bends). The
validity of the uncorrelated error component assumption depends upon
several factors. Clearly the error components are uncorrelated over the
ensemble of all flight routes. For repeated flights over the same route the site-
dependent errors may be correlated while the equipment bias errors depend
upon the nature of the drift mechanism. Bias errors are assumed to be constant
over the duration of a single mission, however. In any event for a given facility
installation the site and equipment are selected and configured such that the
system accuracy for a single mission is achieved. In addition, the FAA
periodically has flight inspections to check a facility’s performance. Since the
exceptional sites are treated on an individual basis to ensure flight safety, and
since the statistical distributions are not known anyway, the uncorrelated
assumption is retained because of the simplicity of the RSS calculation. Simple
addition of the error components would unrealistically limit the approach
path and skew the facility implementation economics.
Error budgets are defined in terms of 95% confidence limits, rather than
20 limits, because the two specifications are essentially equivalent only for
Gaussian random processes. The probability that a given event will occur is
the important quantity not the 20 value. The probability of exceeding
specified limits is the only meaningful measure by which air-lane route widths,
decision windows, touchdown foot prints, and obstacle clearance surfaces can
be operationally defined.
Although each error component can be given in terms of a sample time
variance, S i , the calculation of the 95% confidence interval requires
knowledge of the underlying probability distribution function which is, in
general, not known. In practice this lack of knowledge is circumvented by
DISTANCE MEASURING EQUIPMENT IN AVIATION 43
estimating the confidence intervals using the 2.5 and 97.5 percentile limits as
determined by the measured data. The overall system error is then determined
by the RSS of the 95 percentile value of each error component. The confidence
limits of the combined measured error samples is then compared to the
appropriate accuracy standard. Figure 83 illustrates the measurement metho-
dology for the DME/P. In particular, it shows how the 95% confidence limits
are determined from the system flight test results and the individual navaid
component test results. Caution should be exercised when invoking the
central limit theorem to approximate the total navigation system error
distribution. [See Fig. 2 in Hsu (1979).] Based upon the above definition, it
should be emphasized that the term “bias with 95% probability” means that,
over the ensemble of navaids, 95% of the equipment bias measurements lie
between the 2.5 and 97.5 percentile limits. The specified bias error value is the
upper and lower percentile limits, as noted in Fig. 14. Figures 43 and 44 in
Subsection F show how errors are combined to yield position-fix RSS
accuracy estimates. All error components defined in Sections I11 and IV of this
article are 95% confidence level specifications.
One operational basis for using the single RSS value is that guidance
obtained from enroute navigation sensors for manual flight is usually filtered
extensively by the flight control system. Errors varying at a very high
frequency are readily filtered out in the aircraft equipment, leaving only low-
frequency error components. Since this error component can be tracked by the
aircraft, it can, when combined with the bias component, displace the aircraft
from its intended flight path. Thus, the single RSS value is a useful measure of
how well the aircraft will remain in its designated air lane.
This rationale does not apply to autocoupled approaches. Biaslike errors
again must be controlled so that the aircraft will remain in its air lane. The
variance of the noise errors cannot always be reduced by heavy filtering if
certain AFCS stability margins are to be maintained. Therefore, the unfiltered
portion of the noise may induce aircraft pitch and roll motions (control
activity) that may be unacceptable to the pilot (see Subsection E).
As noted earlier, the single RSS specification is not adequate for the
approach and landing application. Two numbers are defined for the ILS
accuracy standard, the mean course error, crbias, and the “beam bends” noise,
crN. The noise term is specified separately to limit the control activity and to
limit, for example, lateral velocity errors away from runway centerline. (Note
that an ILS localizer or MLS azimuth bias error always directs the aircraft
toward the runway centerline, whereas a low-frequency noise error component
may direct the aircraft away from runway centerline.)
With the advent of the MLS, accuracy standards made a significant
departure by expanding the traditional notions of bias and noise to include the
spectral content of the navigation signal. With these new definitions it is
44 ROBERT J. KELLY AND DANNY R. CUSICK

intended that the MLS angle and DME/P accuracy standards will more
accurately represent the qualities needed to achieve successful landings.
The lateral error components for the enroute, terminal, and landing phases
of flight are summarized in Table 111, Subsection D,2. The table indicates the
increasing level of sophistication in the error component definitions. Clearly,
the magnitude, nature, spectral content, and distribution of errors as a
function of time, terrain, aircraft type, aircraft maneuvers, and other factors
must be considered. The evaluation of errors is a complex process, and the
comparison of systems based upon a single error number could sometimes be
misleading.
As stated earlier, the purpose of an error budget is to estimate the accuracy
performance of a proposed navigation system hardware (and software)
implementation. The idea is to select the 95% confidence interval for each
component of the navigation system, such that the RSS of the total
combination is within the 95% probability error limits specified for a given
application. If the assumed conditions are correct, then there is high
confidence that 95% of the aircraft’s position measurements will be within the
allowed error limits during a specified segment of the aircraft’s Bight mission.

4 . Summary
The basic notions of a navigation system were defined with particular
emphasis on the DME navaid. DME, a position-fix navaid, was contrasted
with dead reckoning techniques. DME can be used in a navigation system in
two ways. If it is the primary method of navigation, then its data may require
inertial aiding so that its guidance corrections do not affect the AFCS dynamic
response. However, if dead reckoning is the primary navigation aid, then
DME may assist it by periodically correcting its drift errors. It was shown that
the navigation parameters follow a common theme whether the application is
enroute navigation or aircraft approach and landing guidance. The desired
flight path and indicated flight path coordinates can be defined for both
applications with DME playing a central role. Error components for the
approach and landing application are smaller and are more refined than in the
enroute navigation application.

D. The Air TrafJic Control System and Its Navigational Aids

This section describes the principal elements of the Air Traffic Control
System, thereby helping to illuminate the operational role of DME in the
context of the enroute, approach, and landing phases of flight.
DISTANCE MEASURING EQUIPMENT IN AVIATION 45
1. Goals of the National Airspace System
The National Airspace System (NAS) is a large and complex network of
airports, airways, and air traffic control facilities that exists to support the
commercial,private, and military use of aircraft in the United States [Office of
Technology Assessment (OTA), 19823.
The NAS is designed and operated to accomplish three goals with respect
to civil aviation:
(1) safety of flight;
(2) expeditious movement of aircraft; and
(3) efficient operation.
These goals are related hierarchically with safety of flight as the primary
concern. The use of airport facilities, the design and operation of the ATC
system, the flight rules and procedures employed, and the conduct of
operations are all guided by the principle that safety is the first consideration.
In the U.S., the FAA has defined a single nationwide airway system shared
by all users of the NAS based on strategically located elements of VOR, DME,
and TACAN navaids (Fig. 15). These facilities are geographically placed so as
to provide for continuous navigation information along the defined airways.
The airways serve as standard routes to and from airports throughout the U.S.
DME provides the means to determine the position of the aircraft along a
given route and to aid in verifying arrival at intersecting airways.
The airways are divided into two flight levels. Below 18,000 feet mean sea
level (MSL), the airways that run from station to station make up the system of
VOR, or “VICTOR,” airways. These airways can be successfully navigated
utilizing VOR bearing information only. The use of DME is not required but

FIG.15. Depiction of how air routes are constructed from VOR (and TACAN) defined
radials.
46 ROBERT J. KELLY AND DANNY R. CUSICK

FIG.16. Excerpt from low-altitude enroute chart depicting the VICTOR airway system in
Southern Virginia.

certainly simplifies the process of position determination along an airway.


VICTOR airways are depicted on low-altitude navigation charts (Fig. 16),
marked with a letter V and a number, e.g., V23. VOR stations in these airways
are located an average of 70 miles apart.
The high-altitude routes, or jet routes, are defined for navigation at
altitudes of 18,000 to 45,000 ft and are utilized by high-performance turbine
powered aircraft. These routes are identified on high-altitude navigation
charts by the letter J followed by a number, e.g., J105. Unlike the VICTOR
airways, DME, or its equivalent, is required for navigation along the jet routes.
In addition to the VICTOR and JET airways the FAA (Federal Aviation
Administration, 1983d) recently authorized random route using area naviga-
tion (RNAV).
VOR, DME, and TACAN derived position information is also utilized for
terminal area navigation (Fig. 17).
DISTANCE MEASURING EQUIPMENT IN AVIATION 47
VORlDME or TACAN 1 RWY 14

FIG. 17. A terminal instrument approach procedure utilizing a DME defined arc with
position fixes along the arc defined by VOR radial intersections. Missed approach procedure (to
BOAST intersection)is defined by a DME arc.

The FAA provides planning and advisory services to guide the aviator in
making use of the NAS under either of two basic sets of rules-visual flight
rules (VFR) and instrument flight rules (1FR)-which govern the movement
of all aircraft in the United States. Similar visual and instrument flight rules
are in force in foreign countries that are members of ICAO. In many cases,
ICAO rules are patterned on the US.model. In general, a pilot choosing to fly
VFR may navigate by any means available to him; visible landmarks, dead
reckoning, electronic aids such as the VORTAC system, or self-contained
systems on board the aircraft. If he intends to fly at altitudes below 18,000 ft, he
need not file a flight plan or follow prescribed VOR airways, although many
pilots do both for reasons of convenience and safety. The basic responsibility
for avoiding other aircraft rests with the pilot, who must rely on visual
observation and alertness-the “see-and-avoid” principle.
48 ROBERT J. KELLY A N D DANNY R. CUSICK

In conditions of poor visibility or at altitudes above 18,000 ft, pilots must


fly under IFR. Many also choose to fly IFR in good visibility because they feel
it affords a higher level of safety and access to a wider range of ATC services.
Under IFR, the pilot navigates the aircraft by referring to cockpit instruments
(e.g., VOR and DME for enroute flights) and by following instructions from air
traffic controllers. The pilot is still responsible for seeing and avoiding VFR
traffic when visibility permits, but the ATC system will provide separation
assurance from other IFR aircraft and, to the extent practical, alert the IFR
pilot to threatening VFR aircraft.
The distinction between VFR and IFR is basic to ATC and to the safe and
efficient use of airspace since it not only defines the services provided to
airmen, but also structures the airspace according to pilot qualifications and
the equipment the aircraft must carry. Much of the airspace below 18,000 ft
is controlled, but both VFR and IFR flights are permitted. The altitudes be-
tween 18,000 and 60,000ft are designated as positive control airspace; flights
at these levels must have an approved IFR flight plan and be under control
of an ATC facility. Moreover, the aircraft must be equipped with VOR and
DME in order to fly the VORTAC routes or other approved navigation aids
if using RNAV. Airspace above 60,000 ft is rarely used by any but military
aircraft.
The airspace around and above the busiest airports is designated as a
terminal control area (TCA), in which only transponder-equipped aircraft
with specific clearances may operate regardless of whether using VFR or IFR.
All airports with towers have controlled airspace to regulate traffic movement.
At small airports without towers, all aircraft operate by the see-and-avoid
principle, except under instrument meteorological conditions. Figure 18 is a
schematic representation of the U.S. airspace structure.
Aircraft flying under IFR, on the other hand, are required to have a radio
and avionics equipment allowing them to communicate with all ATC facilities
that will handle the flight from origin to destination. Figure 19 illustrates a
hypothetical VORTAC and RNAV route from takeoff to touchdown.

2. The ATC System


The third major part of the National Airspace System offers three basic
forms of service: navigation aids (including landing), flight planning and in-
flight advisory information, and air traffic control.
The essential feature of air traffic control service to airspace users is
separation from other aircraft. The need for this service derives from the simple
fact that under IFR conditions the pilot may not be able to see other aircraft in
the surrounding airspace and will therefore need assistance to maintain safe
separation. Figure 20 represents the four functional elements of the ICAO
z=
:ONTINENTAL
CONTROL
AREA
DISTANCE MEASURING EQUIPMENT IN AVIATION

t
60,OOOFT.
(FL600)

45,000 FT.
(FL450)

18,000 FT. MSL


A

PoS'TIVE
4

TRANSPONDER
WITH ALTITUDE
ENCODING
REQUIRED
49

14.500 FT MSL

FIG.18. Airspace structure. AGL,above ground level; MSL, mean sea level; FL,flight level
(FAA).

standardized air traffic control system: radar, VHF/UHF communications,


enroute navigational aids, and landing aids. The ATC navigation service is, of
course, a subject of interest in this article because it intimately involves DME.
In the U.S. the Department of Transportation (DOT), through the Federal
Aviation Administration (FAA), operates these four systems plus a primary
radar system to fulfill its statutory responsibilities for airspace management.
Controllers use these systems to provide navigation services, separation
between aircraft, and ground-proximity warnings. Pilots require these systems
for navigation and to receive ATC services. Specifications for these systems
50 ROBERT J. KELLY AND DANNY R. CUSICK
ZNM

SECONDARY AREA

4NM
I

t
IAWP

VORIDME

INWP FAWP RUNWAY


THRESHOLD
IAWP = I N I T I A L APF’ROACH WAVPOINT
INWP = INTERMEDIATE APPROACH WAVPOlNl
FAWP = F I X E D APPROACH WAYPOINT
MAWP = MISSED APPROACH WAVPOINT

0
_ _ ---_
I ,

DEPARTURE
AIRPORT

.C..I.....^.
AZIDMEIP AREA

AREA I
AZIDMEIP

FIG. 19. Hypothetical VORTAC and RNAV f l i g h t paths and route structure.

such as frequency and power output are determined by both national and
international standards. These “service volume” standards are absolutely
necessary. Unacceptable interference in the air-ground or ground-air links of
these systems could deny or distort information essential to the controller
or pilot in providing or receiving safe ATC service. Sections III,C,2 and 3
describe the procedure for ensuring interference-free DME service volume.
The systems presently in use for ATC are:
(1) A means for the controller to detect and identify aircraft using the
primary or secondary radar;
DISTANCE MEASURING EQUIPMENT IN AVIATION 51

OTHER AIRCRAI

USED BY CONTROLLER

FIG.20. The ATC system.

(2) A common enroute navigation system (VORTAC);


(3) Precision approach and landing guidance (MLS/ILS);
(4) Nonprecision approach guidance (e.g., VOR/DME);
(5) Facilities for direct controller-to-pilot communications.
Note how closely this ATC system satisfies the four requirements stated by
ITT in 1945.See the Institute of Electrical and Electronics Engineers (1973)for
an in-depth review of the ATC system.
a . Enroute and Terminal Area IFR Route Structures. In addition to air-
craft separation, the pilot under IFR conditions requires navaid guidance
to avoid obstacles as he descends from the enroute structure to the terminal
area and executes a precision or nonprecision approach.
The following discussion illustrates how the navaid accuracy performance
requirements are consistent with the airway route structure. The narrative
will be in the context of area navigation (RNAV) because the VOR/DME
radials are really special cases of the more general RNAV route. RNAV
and its present status in the NAS is reviewed in Subsection D,6. It will be as-
sumed that the primary navigation guidance is derived from position-fix
radio navaids such as VOR/DME, TACAN, or DME/DME.
In addition to the navaid guidance, flight procedures and obstacle
clearance surfaces must be defined so that aircraft can safely fly along the
52 ROBERT J. KELLY A N D D A N N Y R. CUSICK

enroute airways and maneuver in the terminal area. Enroute procedures are
given in FAA (1975b, 1984)while terminal area procedures are documented in
FAA (1976) and ICAO (1982).
Enroute airways and terminal area approach paths are defined by a series
of waypoints connected by straight-line segments. The enroute airways are
usually specified by VORJDME radials with waypoints defined directly over
the ground stations or at the intersection between the radials.
A waypoint may be identified in several ways, i.e., by name, number, or
location. Waypoint location is necessary in the computation of navigation
information and to minimize pilot workload during time-critical phases of
flight. At a minimum, enough waypoints are provided to define the current and
next two legs in the enroute phase of flight, and to define an approach and
missed approach.
In some mechanizations, waypoints are entered into the equipment in
terms of their latitude and longitude. In other equipment a waypoint’s
location may be specified in terms of a bearing and distance from any place
whose position is itself known to the pilot.
Recalling earlier discussions in Subsection C, lateral guidance using path
deviation correction signals is the principal means of enroute and terminal
area navigation. Vertical guidance is typically not derived from path deviation
signals; aircraft separation and obstacle clearance is achieved by measuring
height with a barometric altimeter. Although the most inexpensive and
therefore most implemented form of lateral guidance is essentially bearing
information using NDB or VOR, the trend is toward bearing and range
position-fix navigation. From the position fixes, deviation signals about the
desired course called steering commands are input to the aircraft flight control
system. In particular, for the RNAV application, the lateral navigation
information displayed to the pilot commonly takes the form of bearing and
distance from present aircraft position to the defined waypoint and deviation
from a desired track proceeding to or from the waypoint. As noted in
Subsection C,2, when the displayed guidance permits lateral position-fix
navigation the process is called 2D navigation. In summary, the most utilized
form of area navigation is 2D lateral guidance. The role assigned to vertical
information as derived from the barometric altimeter is minimum descent
altitude alerts and aircraft descent rate. When vertical deviation signals are
derived from the altitude data, then RNAV guidance is called 3D navigation.
To avoid collisions with other aircraft, precipitous ground terrain, and
airport obstacles, the enroute airways and terminal approaches are separated
both laterally and vertically. Lateral separation is defined by primary and
secondary protection areas, as shown in Fig. 19. The full obstacle clearance
is applied to the primary area, while for the secondary area the obstacle
clearance is reduced linearly as the distance from the prescribed course is
DISTANCE MEASURING EQUIPMENT IN AVIATION 53
increased. The widths of the air route are based upon the accuracy of the
navaid guidance, the flight technical error, and the RNAV equipment errors.
Blunder errors are monitored by enroute and airport surveillance radars and
are not included in the flight technical error component. The minimum
authorized flight altitude is determined by the obstacle clearance, which is
the vertical distance above the highest obstacle within the prescribed area.
For precision approaches this distance is based upon a collision risk
probability of lo-’ over the entire flight mission. The minimum flight alti-
tude may equal the obstacle clearance or it may be raised above it under
certain conditions. If there are no obstacles then the obstacle clearance is the
height above the ground. A displacement area is defined about each waypoint.
It is a rectangular area formed around the plotted position of the waypoint.
The dimensions of this area are derived from the total navaid system along-
track and across-track error values, including the flight technical errors (see
Subsection F). In other words, the waypoint displacement area defines the
accuracy with which an aircraft can reach a waypoint and is directly related to
the airway route width or terminal area approach path width. In the final
analysis, it is the displacement area of the approach path which defines the
minimum descent altitude.
For the enroute airways, the lateral guidance protected area is k4 nmi
about the route center line and the secondary area extends laterally 2 nmi on
each side of the primary area, as shown in Fig. 19. The required vertical
separation is lo00 ft below 29,000 ft altitude and 2000 ft above 29,000 ft
altitude. The minimum required altitude (MRA) is lo00 ft and is equal to the
obstacle clearance. In general, the minimum enroute altitude (MEA) is equal
to the MRA. However, as noted earlier, the MEA may be raised above the
MRA when, for example, the air route is over mountainous terrain. For
+
the mountains in the eastern United States MEA = MRA 1000 ft; over
the Rocky Mountains, MEA = MRA + 2000 ft.
Terminal area approach paths are also protected by lateral and vertical
separation distances. When in the terminal area, a series of position fixes
define the approach path. They are initial approach fix (IAF), intermediate
approach fix (INAF), final approach (FAF), missed-approach fix (MAF), and
the runway fix (RWYF). The path between each pair of fixes is called a
segment; e.g., the intermediate segment is between the INAF and the FAF.
About each segment there is defined a protected primary area and a secondary
area. Terminal protected areas are 2 nmi about the route center line and the
secondary obstacle clearance areas extend laterally 1 nmi beyond the primary
area.
FAA Handbook 8260.3A, called the TERPS (FAA, 1976), prescribes
criteria for the design of instrument approach procedures. Although it does
not at the present time contain special criteria for the design of RNAV
54 ROBERT J. KELLY A N D DANNY R. CUSICK

procedures, most of the TERPS criteria are applicable to RNAV procedures


with minor modifications. Guidelines for implementing RNAV within the
NAS are given in AC 90-45A (FAA, 1975), which is oriented toward
VOR/DME. Future revisions to AC 90-45A will reflect the more general
configurations of multisensor RNAV as defined in RTCA (1984).In applying
TERPS to a RNAV procedure, the term “fix”isequivalent to “waypoint”;thus
“final approach fix (FAF)” becomes “final approach waypoint (FAWP).”
Therefore, an RNAV instrument approach procedure may have four separate
segments. They are the initial, the intermediate, the final, and the missed-
approach segments. The approach segments begin and end at waypoints or
along-track distance (ATD) fixes, which are identified to coincide with the
associated segment as shown in Fig. 19.
The RNAV instrument approach procedure commences at the Initial
Approach Waypoint (IAWP).In the initial approach, the aircraft has departed
the enroute phase of flight and is maneuvering to enter the intermediate
segment.
The purpose of the intermediate segment is to blend the initial approach
into the final approach and provide an area in which aircraft configuration,
speed, and positioning adjustments are made for entry into the final approach
segment. The intermediate segment begins at the Intermediate Waypoint
(INWP) and ends at the FAWP. For a standard procedure, the INWP is about
8 nmi from the runway with a minimum segment length of 3 nmi. During the
final approach segment, the aircraft aligns itself with the runway and begins
the final descent for landing.
The final approach segment beings at the final approach waypoint or
along-track distance fix and ends at the missed-approach point, normally the
runway threshold waypoint. When it is not the runway threshold, it is an ATD
fix based on a distance to the runway waypoint. The optimum length of the
final approach segment is 5 miles. The maximum length is 10 miles. The final
approach primary area narrows to the width of the FAWP displacement area
at the runway threshold. A missed-approach procedure is established for each
instrument approach procedure. The missed approach must be initiated no
later than the runway threshold waypoint.
The obstacle clearance for the initial approach segment is 1000ft and thus
matches the MEA of the enroute airway. Its secondary area obstacle clearance
is 500 ft at its inner edge, tapering to zero at the outer edge. The optimum
descent gradient to the intermediate fix is 250 ft/mi. Obstacle clearance for the
intermediate segment is 500 ft. Because the intermediate segment is used to
prepare the aircraft speed and configuration for entry into the final approach
segment, the gradient should be as flat as possible. The optimum descent
gradient in this area for straight-in courses should not exceed 150 ft/mi. The
DISTANCE MEASURING EQUIPMENT IN AVIATION 55
obstacle clearance for the final approach segment begins at the final approach
waypoint and ends at the runway or missed-approach waypoint, whichever is
encountered last. Associated with each final approach segment is a minimum
descent altitude (MDA). It is the lowest altitude to which descent shall be
authorized in procedures not using a glide slope. Aircraft are not authorized to
descend below the MDA until the runway environment is in sight, and the
aircraft is in a position to descend for a normal landing. As mentioned earlier,
sometimes the MDA must be raised above the obstacle clearance. Conditions
which necessitate raising the MDA are precipitous terrain, remote altimeter
setting sources, and excessive length of final approach. The required obstacle
clearance for a straight-in course is 250 ft. The decision height (DH) minimum
is also 250 ft when the visibility is 1 nmi.
The above final approach discussions are directed toward nonprecision
approaches, as contrasted with precision approaches. Precision approaches
require a descent path from which vertical deviation correction signals can be
derived, as described in Subsection D,4. The MDA and DH are determined by
the size of the final approach displacement area and the obstacle clearance.
Although typical nonprecision approach heights are 400 ft and above, the
lowest authorized minimum is 250 ft. Precision approaches routinely achieve
decision height at 250 ft and below because the final approach displacement
area is smaller. Precision navaids and cockpit displays achieve smaller
displacement areas because (1) positive vertical guidance is required (i.e., a
glide path as well as azimuth guidance); (2) the navaid guidance errors are
smaller; and (3) the display sensitivities are increased to reduce the flight
technical errors. For example, the full-scale sensitivity of the CDI is L 1.25 nmi
at 20 nmi and is +350 ft at runway threshold. The full scale sensitivity is
always designed to be within the obstacle clearance surface.
With smaller displacement areas the obstacle clearance for precision
approaches can be satisfied more easily because the projected footprint of its
displacement area is smaller. Consequently, with precision approaches, lower
decision height minima are more readily achievable to airports or runways
with smaller obstacle-free surfaces.
The principal distinction between nonprecision and precision approaches
as used by the FAA is that a combined azimuth and elevation guidance (glide
slope) is required for precision approaches (FAA, 1976). The term nonpreci-
sion approach refers to facilities without a glide slope, and does not imply an
unacceptable quality of course guidance. It is the reduction of the final
approach displacement area which is important to achieving lower minima.
Positive vertical guidance is not necessary down to 250 ft because adequate
obstacle clearance is provided. At that altitude the pilot has visible cues and is
in the region of “see and avoid.” He can correct the attitude and position of the
56 ROBERT J. KELLY AND DANNY R. CUSICK

aircraft to accomplish a manual flare and have a safe landing. Nonetheless,


the pilot is more comfortable executing nonprecision approaches down to
higher minima such as 400 ft. It is clear that precision approaches, with their
accurate descent path, provide additional margins of integrity and pilot
confidence. Table 111 summarizes the lateral route width accuracy require-
ments for the radio navaids servicing the enroute, terminal area, and the
approach/landing application.

TABLE I11
LATERAL
ACCURACIESIN NAS"

XTRK system Navaid XTRK error


error (95% prob.) (95% prob.)
Route Route
type width Includes FTE No FTE

Enrouteb +4 nmi Random + 4.0 nmi Random k 3.8 nmi


Random JJV + 4 nmi JIV f3.0 nrni J/V & 2.8 nmi
Terminalb + 2 nmi k 2 nmi f1.7 nmi
RNAVb Runway (k0.7 nmi) Runway kO.5 nmi
Non displacement displacement
precision area area
approach
ILS Cat I k455 ft & 455 ft Bias 45 ft
Noise 46 ft
ILS' f400 ft + 75 ft bias 25 ft k27.5 ft
Cat I1 Obstacle (runway width) Noise + 12.7 ft
approach clearance
ILSd k 7 5 ft k 27.5 ft Bias +loft
Cat 111 (Runway Noise + 11.5 ft
landing width)
MLS' +400 ft +75 ft + PFE 21.7 ft
Cat I1 95% Prob. +PFN 12.5 ft
approach (runway width) +BIAS 10.8 ft
kCMN 11.3 ft
MLS" +75 ft +27.5 ft + PFE 20.0 ft
Cat 111 (Runway About runway +PFN 11.5 ft
1anding width) centerline +BIAS 1O.Oft
*CMN 10.4 ft

a J = Jet routes, V = VOR routes, random = RNAV; FTE = flight technical


error; runway length 10,000 ft
RTCA (1984).
ICAO (1972b); FAA (1970).
ICAO (1972b); F A A (1971, 1973).
ICAO (1981a).
DISTANCE MEASURING EQUIPMENT IN AVIATION 57
Adding a third dimension of vertical guidance to the two-dimensional
RNAV systems can achieve significant operational advantages. Briefly, a 3D
RNAV capability permits altitude change by following vertical routes (tubes)
of known dimensions; thus vertical guidance is available for stabilized descent
in instrument approach procedures using computed glide path information. In
some cases, the computed glide path can make it possible to safely eliminate
obstacles from consideration. To provide vertical guidance during ascent or
descent, the 3D RNAV equipment compares the indicated altitude with the
desired altitude and presents the computer correction instrumentally.
In prescribing obstacle clearance for 3D RNAV, it is useful to think of the
vertical route as being the center of a tube of airspace. The lateral dimension
of the tube is the width of the RNAV route as described in Table 111. The
vertical dimension of the tube is sufficient to contain the combined 3D RNAV
vertical errors. The longitudinal dimension of the tube is limited only by its
operational use. For example: The distance required to climb 10,000 ft at a
climb angle of 2", or the descent from the final approach waypoint to the
missed approach waypoint at a descent angle of 3". For obstacle clearance, it is
necessary to consider only that portion of the tube which is at and below the
designed vertical flight path. An aircraft is protected from obstacles when no
obstacles penetrate the tube from below. The along-track error also has
significance in vertically guided flight. When an aircraft is ahead or behind its
assumed position, it will be either above or below its intended path.
The possibilities for a hazardous incident are analyzed using the total
navigation system error and the distance to an obstacle. For the approach/
landing application, the hazard probabilities are small and the obstacles
are known fixed objects, e.g., towers, hangars, etc. Although the probability
of exceeding the route width of an enroute airway is only 95%, the lateral
obstacle is in effect another aircraft in a parallel route; thus the joint prob-
ability of a collision is very small, especially when ATC aircraft separations
are maintained using ground radar control. The longitudinal separation
using the ground radar (ARSR) is 3 to 5 nmi. Without radar it is 5 nmi with
DME and the faster aircraft is in front of the slower aircraft (Kayton and
Fried, 1969).
The principal ideas associated with terminal approaches are worth
restating. They are: (1) The collision risk probability establishes the obstacle
clearance; (2) the total navigation system error determines the approach
displacement area; and (3) the height of the displacement area for a given
approach path is determined by the obstacle clearance. Thus the displacement
area is lowered until it equals the obstacle clearance. For a nonprecision
approach the displacement area is larger than the precision approach
displacement area. Therefore, its MDA or DH will generally be higher than
that required for a precision approach.
58 ROBERT J. KELLY A N D D A N N Y R. CUSICK

6. Surveillance Radar. The primary radar is intended to detect all aircraft


by processing echoes of ground facility transmissions-no airborne equip-
ment is required. There are two types. The first, air route surveillance radar
(ARSR), operates in the 1300-1350 MHz band and normally has a range of
200 nmi. The airport surveillance radar (ASR),in the 2700-2900 MHz band,
has a range of 60 nmi. ARSR systems have peak power outputs of 1.0 to 5.0
MWatts, and ASR systems from 400 kW to 1.0 MW. The power output for a
particular site is determined by site conditions, i.e., terrain, elevation, and the
proximity of adjacent sites.
Bearing and range of the aircraft are displayed to the controller on a plan
position indicator (PPI)called a plan view display (PVD).The primary radar
permits the controller to see any aircraft within the coverage area, regardless
of whether or not the pilot is using the ATC system. The ATC primary
surveillance radar is a self-contained system whose purpose is to maintain
aircraft separation by dicect communication with the pilot when the aircraft is
carrying a radio. Since the ARSR and ASR do not involve equipment on board
the aircraft, international standardization is not required.
The secondary radar, as noted in Subsection B, or air traffic control radar
beacon system (ATCRBS), is a system wherein the airborne element is the
transponder and the ground element is the interrogator. The ATCRBS power
output is 200-2000 W and has a nominal 250 nm range when used in enroute
applications. The secondary radar ground-air link, or interrogator, operates
at 1030 MHz and activates the 1090 MHz aircraft air-ground link, or
transponder. The transponder reply among the 4096 available codes generally
includes an identification code and an altitude code. Transponder replies are
received by the ground interrogator, and heading and ground speed are
calculated by tracking devices. Heading and ground speed are displayed on
the PVD along with the aircraft identification number and altitude. A rotating
directional ground-based interrogator transmits approximately 400 inter-
rogations per second.
The cooperative ATCRBS is one of the keys to advanced ATC auto-
mation. The ATC system is devloping additional automated features to
economically meet the high demand for services in the 1990s. In the U.S. there
are about 300 ATC radars in use, over 100 ARSRs and about 200 ASRs. All
ATC primary radars are equipped with the ATCRBS secondary radar system.
Unlike the primary radar system, the secondary radar is an ICAO standard; its
“Standards and Recommended Practices”(SARPs) are given in ICAO, (1972b,
para. 3.8).
3. Navigation Aids
Aircraft users require short-range navigation aids to probide enroute
guidance over the defined airways under IFR conditions. The VORTAC
DISTANCE MEASURING EQUIPMENT IN AVIATION 59
routes were established to provide a simple, common system that is easy to use,
thereby reducing pilot workload. DME/N and VOR provide the guidance
with RNAV augmenting the VORTAC route structure by providing random
routes. Table IV places the VORTAC system in perspective by comparing the
number of units in use with all the air and marine navigation aids in use
throughout the world. The table was excerpted from Dodington (1984).
The accuracy and coverage of navigational aids are determined by the
enroute and terminal area operational requirements. Systems deriving bearing
information (VOR and TACAN) must offset the effects of multipath and
siting,which are the major sources of error (Subsection F). Signal coverage is a
design problem only because adjacent channel effects must be considered (see
Section III,C,3). In other words, it is easy to obtain coverage with a single
facility, but it takes special care to eliminate spillover into other channels so
that coast-to-coast service can be provided using a limited number of channel
frequencies.
The next several subsections describe the VOR, TACAN, ILS, and MLS
angle systems. The DME/N and DME/P are treated in Sections 111and IV of

TABLE IV
SYSTEMSUSEDTHROUGHOUT
MAINRADIONAVIGATION THE WORLD'

Users
Number of
System Frequency stations Air Marine

Omega 10-13 kHz 8 10,600 7000


Loran-C 90-1 10 kHz 33 chains 2000 60,000
Decca 70-130 kHz 50 chains 1000 30,000
ADFb 200-1600 kHz 5000 200,000 500,000
ILSb 75,108-1 12, 1600 70,000
329-335 MHz
VORb 108-118 MHz 2000 200,000 -
Transit 150,400 MHz 5 satellites - 38,000
Tacan 960-1215 MHz 2000 17,000 -
DMEb 960-1215 MHz 1000 80,000
SSRb 1030, 1090 MHz 100,000 -
MLSb 5031-5091 MHz Implementaticm about to' start
Estimated number of U.S. aircraft: 224,000 General aviation (Quinn;
1983)
2500 Air carrier
20,800 Military aircraft

Dodington, 1984.
a

Signal format standardized by the International Civil Aviation


Organization.
60 ROBERT J. KELLY AND DANNY R. CUSICK

this article. These navaids are viewed from a system level viewpoint. Each
system is characterized by its signal format, accuracy, coverage, channel plan,
and data rate. Two other system level characteristics are needed before the
navaids used in the ATC system are complete. They are integrity and
availability.
Integrity is a measure of the truthfulness of the signal in space. It is ensured
by a ground monitor system, which shuts the facility down if certain critical
signal characteristics exceed predetermined limits. Integrity is quantified by
the reliability (mean time between outages, MTBO) of the ground equipment
and its monitor.
A navigation system must be available to the pilot, especially under IFR
conditions. Although a facility which has been shut down has good integrity, it
is not available. The system availability A is given by A = MTBO/(MTBO +
MTTR), where MTTR is the mean time to repair. Integrity and availability
are critical to the successful operation of approach/landing systems.
Radio navaid accuracy performance comparisons are made in Hogle et al.
(1983). Integrity and reliability comparison of civil radio navaids are made in
Braff et al. (1983).

a. VOR. The very-high-frequencyomnidirectional range is the ICAO inter-


national standard for providing short-range enroute bearing information.
Paragraph 3.3. of ICAO (1972b) is the ICAO SARPs for VOR. As noted in
Subsection C , it is a radial determining navigation aid, and along with the
automatic direction finder (ADF) (see Table IV) is the most popular system
currently in use. Since it is a radial system, it must have a reference signal and a
signal which carries the bearing information. The reference signal is frequency
modulated (FM) at 30 Hz on a 9960 Hz subcarrier. A cardioid antenna pattern
rotates 30 times per second (Fig. 21), generating 30 Hz amplitude modulation
(AM)in the receiver, and thus provides the bearing information. The airborne
receiver then reads bearing as a function of the phase difference between the
FM reference signal and the AM modulated signal. The system is graphically
described in Fig. 2 1.
The VOR estimates the aircraft bearing in the same manner as the
TACAN; it employs a phase detector. Early implementations of phase
detectors were electromechanical devices using resolver phase shifters which
were rotated by a servo motor until a peak or null was obtained (Hurley et al.,
1951). The bearing display was connected directly to the phase shifter shaft.
Today’s systems are all-electronic digital implementations.
The VOR ground transmitter radiates continuous-wave signals on one of
20 channels between 108 and 112 MHz (interleaved with ILS localizer
frequencies) and 60 channels between 112 and 118 MHz. Transmitter power
is from 50 to 300 W.
DISTANCE MEASURING EQUIPMENT IN AVIATION 61
CHANNEL A

f, = 9960 Hr
AIRBORNE RECEIVER

PHASE
DETECTOR e^

CHANNEL B

CHANNEL B

CHANNEL A

PAlTERN
ROTATES

FIG.21. Principles of VOR.

Because of the frequency used, it is difficult to get an antenna with good


vertical directivity. Consequently, significant radiation strikes the ground in
the vicinity of the transmitter which, upon reflection, interferes with the
desired radiating signal causing a bearing error. This error, which is sometimes
2" to 3" in magnitude, appears as course variations along the radials of the
VOR station. Several approaches to fixing this problem have been tried,
including elevating the antenna to get it farther away from reflecting objects
and using a large counterpoise (ground system) under the antenna. Both of
these methods have been somewhat successful, but the development of
Doppler VOR (DVOR) and Precision VOR (PVOR) offered better solutions.
Severe constraints, however, were placed upon these solutions. They were not
to affect the service of the already existing airborne equipment and, if possible,
should improve the system's overall performance.
Doppler VOR accomplished this by simply reversing the roles of the
reference signal and the information signal. Because their phase relations
remain the same, they allow a standard airborne receiver to operate without
62 ROBERT J. KELLY AND DANNY R. CUSICK
modification. With this modification, the effective aperture of the bearing
information signal increased more than 10:1 (antenna diameter of 44 ft); thus,
the effects of site error could theoretically have a tenfold reduction.
Experiments have confirmed this; site errors of 2.8" were reduced to 0.4".
Steiner (1960)describes DVOR.
The second solution used the same idea as TACAN; that is, it modified the
ground antenna to produce multiple lobes, thereby creating a two-scale
system. The coarse information was 30 Hz and the multilobe antenna (11
lobes) generated a 330 Hz harmonic for the fine information. It could
potentially obtain and 11:l reduction not only in site errors but also in
airborne receiver instrumentation errors (Winick, 1964; Flint and Hollm,
1965).The primary reasons the FAA selected the DVOR solution and did not
pursue the PVOR further were that the conventional VOR receivers could not
decode the multilobe harmonic and DVOR sufficiently reduced site errors to
less than 0.5". Since a DVOR installation is a rather extensive and expensive
process, it is reserved for the most difficult sites, where the simpler techniques
do not suffice.
Given the deficiencies of VOR, many airlines are making use of the
scanning DME as their primary enroute aid to form a p-p navigation system
based on RNAV principles. VOR/DME is then used as an alternative if
acceptable DME/DME geometric configurations cannot be found (see
RNAV, Subsection D,6).
b. TACAN. TACAN range and bearing information defines a natural p-8
system. Unlike VOR/DME, which uses two different rf transmissions,
TACAN provides range and bearing information using a single rf emission.
TACAN is not an ICAO standard.
Range information is derived within the TACAN system in the same way
as for DME: A transponder responds to an interrogation. It uses the same
frequencies, pulse characteristics (coding, form), and interrogator pulse rates
as DME/N.
Bearing information is determined at the interrogator in a manner that is
functionally identical to the VOR phase detector. That is, it generates both a
reference signal and bearing information signal and is, therefore, a radial
system. Because DME is a pulse system, the TACAN bearing information is
equivalent to a sampled data VOR system. Moreover, TACAN bearing
information is a two-scale system similar to precision VOR. The coarse
bearing is 15 Hz and the fine bearing is the ninth harmonic or 135 Hz.
Bearing information is obtained with the following additions to the DME
signal format:
(1) A circular antenna array replaces the simple omnidirectional DME
antenna. It generates a cardioid pattern in space and sequentially rotates it at
15 Hz. At a fixed point in the coverage sector, the DME pulses appear to be
DISTANCE MEASURING EQUIPMENT IN AVIATION 63
amplitude modulated at 15 Hz. A second pattern composed of 9 lobes
spatiallymodulates the cardioid pattern, which generates a 135 Hz sinusoid or
ninth harmonic at a fixed receiver in space.
(2) The 15 Hz reference signal is generated by a North reference burst
which is composed of 24 pulses, the spacing between pulses being alternately
12 and 18 ps. When decoded in the airborne equipment they become 12 pulses
spaced 30 ps apart. This pulse train is referenced once per revolution of the
cardioid antenna pattern.
(3) The 135 Hz reference signal is generated 8 times per revolution of the
cardioid antenna pattern. It consists of 12 pulses spaced 12 ps apart.
(4) The transponder operates at a constant duty cycle of 2700 pps. When
interrogations are lacking, pulse pairs with random spacing are generated
(called squitter pulses) so that the output rate is kept constant at 2700 pps. This
constant duty cycle permits the sampled 15 and 135 Hz sinusoids to be
reconstructed by zero-order hold (ZOH) circuits in the aircraft with complete
fidelity. It also permits AGC to be generated in the aircraft such that the 15 and
135 Hz AM modulated signals can be detected without distortion.
In the airplane the 15 Hz reference burst provides a time reference mark
from which the phase of the 15 Hz information signal can be determined.
Similarly, the 135 Hz reference burst provides a time reference mark from
which the phase of the 135 Hz information signal can also be determined. After
reconstruction from the samples, the information signal is sent to the 15 and
135 Hz phase detectors where the operations are functionally identical to that
used in the VOR receiver. The 135 Hz phase detector accurately estimates the
aircraft's bearing; it is, however, ambiguous over each 40" sector (there are
nine of them). The 15 Hz phase detector resolves the ambiguity by determining
in which of the 9 sectors the actual bearing angle lies (see Fig. 22).
Because TACAN utilizes a multilobe spatial antenna pattern, it can
employ a two-speed phase detection system which is less susceptible to siting
effects and equipment error than is conventional VOR. TACANs accuracy is
comparable to the PVOR. The ninth harmonic reduces the bearing error by
about 4 of those for the 15 Hz signals. TACAN's accuracy is discussed further
in Subsection F.

4 . ApproachfLanding Aids
A guidance system for approach and landing is simply a precise, low-
altitude navigation aid with the additional accuracy and reliability needed for
landing aircraft under reduced visibility. The standard system now in use is
the Instrument Landing System; its SARPs are defined in paragraph 3.3 of
ICAO (1972b). The Microwave Landing System SARPs are defined in para-
graph 3.11 of ICAO, (1981a);it is scheduled to replace ILS in 1998. ILS and
MLS permit precision approaches at or below a 250 ft weather minimum
64 ROBERT J. KELLY AND DANNY R. CUSICK
HORIZONTAL ANTENNA PATTERN

MAGNETIC NORTH
15 RESOLUTlONSlSECOND

WEST
270'

SOUTH
180'

---
-
HORIZONTAL PATTERN COMPONENT THAT GENERATES 15 Hz SIGNAL
COMPOSITE PATTERN THAT GENERATES 15 Hz AND 135 HZ SIGNALS

TIME WAVEFORM AND REFERENCE BURSTS

15 Hz MODULATION

-b 1/15 SECOND I.
.
.ANTENNA PATTERN ORIENTATION SHOWN AT THE TIME OF
NORTH REFERENCE BURST
LET THE NORTH REFERENCE BURST DEFINE TIME 0.
THE RECEIVED SIGNALS ARE THEN

.
15Hz : -SIN r m o n t - 8 )
1 % ~ -SIN~ : rz701~t-se)
WHERE e IS THE BEARING FROM THE STATION.
AUXILIARY REFERENCE BURSTS ARE TRANSMITTED EACH
TIME A 135 Hz PATTERN PEAK POINTS EAST

FIG.22. Principles of the TACAN bearing measurement.


DISTANCE MEASURING EQUIPMENT IN AVIATION 65
using a barometric altimeter. If in addition a middle marker or DME mea-
surement is included, the minimum can be reduced by 50 ft to 200 ft. Non-
precision approaches do not require positive vertical steering commands;
minimum descents down to 250 ft are permitted depending upon the obstacle
clearance.
Before describing ILS and MLS, this section will discuss the landing
operation itself, using MLS as an example. The landing process consists of
curved or segmented flight paths for noise abatement and the transition to the
final centerline approach (Fig. 23). The decision heights where the pilot must
be able to “see to l a n d are 200 and 100 ft ceilings-Categories I and 11,
respectively. The flare maneuver, touchdown, and rollout complete the
landing. MLS must support each of these maneuvers under IFR conditions.
The elevation element, typicallly located 861 ft from threshold and offset from
the centerline by 250 ft, provides elevation guidance to the decision height of
Category I and I1 operations and to threshold in Category I11 (zero ceiling).
The approach-azimuth element supports lateral guidance to the decision
height in Categories I and I1 operations and to touchdown and rollout in
Category 111. The flare maneuver is typically performed manually for

FIG.23. The landing operation with MLS.


66 ROBERT J. KELLY AND D A N N Y R. CUSICK

Categories I and I1 by visual reference, but, for Category 111, positive azimuth
and vertical control are required to touchdown.
The Category I1 “window” is centered at the 100 ft decision height. Its
height is f 12 ft around the indicated glide path and its width is the lateral
dimensions of the runway (typically f 7 5 ft). If the aircraft is not in the
window, the pilot will execute a missed approach. If the aircraft passes through
the Category I1 window within specified variations of pitch, roll, speed, and
lateral drift velocity, then the probability is high that a successful landing can
be achieved when the pilot commences the flare maneuver. The decision height
must be determined within f 5 ft. On the other hand, no Category I window is
necessary because the pilot can maneuver his aircraft to the correct position
and attitude before he executes his flare maneuver (FAA, 1970).
MLS facilitates the autoland maneuver with its increased accuracy,
which is extended to lower elevations further down the landing path than
provided by ILS. Flare to touchdown requires the height above the runway
and, in MLS, it may be provided by a combination of the approach elevation
angle and DME prior to runway threshold and then by transitioning to the
radar altimeter in the vicinity of runway threshold. Terminal area IFR
procedures for MLS are currently under development by the FAA.

a. ZLS. The Instrument Landing System is a radial system providing


guidance for approach and landing by two radio beams transmitted from
equipment located near the runway. One transmitter antenna, known as the
localizer, forms a single course path aligned with the runway center line. The
other transmitter, the glide slope, provides vertical guidance along a fixed
approach angle of about 3”.These two beams define a sloping approach path
with which the pilot aligns the aircraft, starting at a point 4-7 mi from the
runway.
Instrument landing systems operate on frequencies between 108.1 and
111.9 MHz for localizers, and between 328.3 and 335.4 MHz for glide slopes.
Both localizer and glide slope are paired on a one-to-one basis. The service
volumes for an ILS are bounded by wedges of an 18 or 25 nmi radius from the
localizer antenna.
Lateral guidance (Fig. 24) is provided by the localizer located at the far end
of the runway. Two identical antenna patterns, the left-hand one modulated
by 90 Hz and the right-hand one by 150Hz are provided. The vertical needle of
the airborne display is driven right by the 90 Hz signal and left by 150 Hz
signal, and centers when aircraft is on course.
Vertical guidance (Fig. 24) is generated by the glide slope antenna located
at the side of the approach end of the runway. A 150 Hz amplitude-modulated
signal is provided below course and 90 Hz modulation is provided above
DISTANCE MEASURING EQUIPMENT IN AVIATION 67
LOCAL1ZER

A\ ANTENNA

- RUNWAY
FIG.24. Principles of ILS.

course. The horizontal needle of airborne display is driven up when the


amplitude of the 150 Hz signal exceeds that of the 90 Hz signal and is driven
down when the reverse occurs. The needle is horizontal when the aircraft is on
course. (The “cross-pointer” is a course deviation indicator combining a
vertical localizer needle and a horizontal glide slope needle.)
Along-course progress is provided by fan markers that project narrow
beams in the vertical direction and operate at 75 MHz. The middle marker is
placed at the 200 ft decision height; the outer marker about 5 miles out. The
outer marker corresponds to the final approach fix. The outer marker can be
viewed as the final approach fix. Since DME may be used as a replacement
for the markers, 20 DME channels are provided for hard pairing with ILS
localizer frequencies. ILS’s limitations are detailed in Redlien and Kelly,
(198 1).
b. MLS. Like ILS, the angle portion of MLS is an “air-derived’’ radial
system where signals are radiated from ground antennas in a standard format
and then processed in an airborne MLS receiver. The angle information is
derived by measuring the time difference between the successive passes of
highly directive narrow fan-shaped beams, as shown in Fig. 25. Range
information, required to obtain the full operational benefits of the MLS angle
data, is derived from precision DME, which Section IV describes in detail.
MLS operation is detailed in Redlien and Kelly, (1981).
68 ROBERT J. KELLY A N D DANNY R. CUSICK

m
.-dI TIME-ANGLE 4-

AIRCRAFT RECEIVER
FIG.25. MLS angle measurement technique.

MLS is a highly modular system and it may be implemented in simple


configurations. The ground systems may contain approach azimuth, the
“bearing” facility, approach elevation, back azimuth, and DME/P, as shown
in Fig. 26.
The signal format is time multiplexed; that is, it provides information in
sequence on a single-carrier frequency for all the functions(azimuth, elevation,
basic, and auxiliary data). The format includes a time slot for 360” azimuth
guidance with provisions for growth of additional functions. The angle
guidance and data channel plans provide 200 C-band channels between 5031
and 5091 MHz.
Narrow fan-shaped beams are generated by the ground equipment and
scanned electronically to fill the coverage volume. In azimuth, the fan beam
scans horizontally and has a vertical pattern that is shaped to minimize
illumination of the airport surface. In elevation, the arrays are designed to
minimize unwanted radiation towards the airport surface, thereby providing
accurate guidance to very low angles. Unlike ILS, the MLS conforms to a
single accuracy standard equivalent to that required to perform fully
automatic landings.
A ground-to-air data capability is provided throughout the angle guidance
coverage volume by stationary sector coverage beams that are also designed to
DISTANCE MEASURING EQUIPMENT IN AVIATION 69
77ms SEQUENCE

BASIC DATA NO. 2

FIG.26. MLS signal format.

have sharp lower-sidecutoff characteristics. This capability is used to transmit


the identity of each angle function and to relay information (basic and
auxiliary data) needed to support simple and advanced MLS operations.
The wide angle capability of MLS is graphically displayed in Fig. 27.
Because proportional guidance is available over a +40" azimuth sector,
RNAV curved and segmented approaches down to the 200 ft decision height
can be executed. Shown in the figure are the airborne elements which comprise
the MLS RNAV equipment configuration. RTCA Special Committee 151 is
currently defining the minimum operational performance expected of the
airborne equipment intending to perform MLS/RNAV approaches (RTCA,
1985a). For approaches below the 200 ft decision height, the MLS data will
bypass the RNAV box and be utilized directly by the aircraft flight control
70 ROBERT J. KELLY AND DANNY R. CUSICK

MLSRCVR I
A2
DMEP
FLIGHT
DIRECTOR

I RNAV

CONTROL
PANEL

W
I

FIG.27. MLS avionics for segmented approaches

system. RTCA (1985a) recommends that the RNAV error contribution be


limited to 35 ft (95% probability) for MLS area navigation applications.
c. Nonprecision Approach. Aircraft equipped with the appropriate enroute
navaids are permitted to fly final approaches down to MDAs which are
consistent with the obstacle clearance.The nonprecision approach procedures
additionally require that the final approach displacement area be 20.7 nmi,
as indicated in Table 111. Over 30% of the runways with nonprecision
approaches use on-airport VOR (DOD/DOT, 1984). See the discussion in
subsection D.5, which addresses the role that the scanning DME may play in
nonprecision approaches.

5. Communications
Radio communications between controllers and pilots take place on VHF
frequencies between 118.0 and 136.0 MHz, and on UHF frequencies between
225.0 and 400.0MHz. ATC communications are necessary to ensure the safe,
orderly, and expeditious flow of air traffic.Chapter 4 of ICAO (1972b)defines
the ICAO standard for communication systems.
DISTANCE MEASURING EQUIPMENT IN AVIATION 71
6. Special Radio Navaid Techniques
This section discusses two techniques which enhance the operational
performance of DME and other navaids in the ATC system. They are RNAV
and the “scanning DME.” RNAV enhances enroute navigation and nonpreci-
sion approaches by providing point-to-point navigation along line segments
other than the VORTAC radials. Scanning DME is a single interrogator that
obtains the range to several different ground transponders by using frequency
hopping or scanning frequency techniques. From these data a very accurate
position fix is determined.
a. RNAV. Since the introduction of VORTAC, air navigation has been
improved by providing a means for unlimited point-to-point navigation. This
improved method of navigation, which utilizes a computer to determine
courses that need not lie along VORTAC radials, is called Area Navigation. It
permits aircraft operations on any desired course within the coverage of
various stations providing referenced navigational signals, i.e., DME or VOR,
or within the limits of self-contained airborne systems, such as the INS. Sensor
inputs to the RNAV may be p-p, 8-8, or p-8.
As stated in Quinn (1983), there are three general operational concepts in
area navigation: (1) two dimensional (2D), involving only horizontal move-
ments; (2)three dimensional (3D), in which vertical guidance is combined with
horizontal; and (3) four dimensional (4D), in which time of departure and
arrival are combined with horizontal and vertical movements for complete
aircraft navigation. At present, 2D area navigation is the most widely used
version. Many pilot-related operational issues regarding 3D and 4D approach
procedures have been and still are under study (Jensen, 1976; RTCA, 1985a).
The conventional implementation of2D RNAV divides the flight course into a
series of waypoints connected by straight-line segments (see Fig. 19). The
waypoints are stored in the RNAV computer, and are identified by suitable
coordinates such that DME/DME or VOR/DME measurements can gen-
erate the steering commands to guide the aircraft along the straight-line
segments.In the more simple implementations, the position fixes derived from
VOR and/or DME measurements are sometimes aided by inertial sensors so
that a waypoint turn can be anticipated and thus minimize the error incurred
in capturing the next flight path segment (Tyler et al., 1975).
In the early 1970s, the FAA established a high-altitude RNAV airway
structure. As stated in RTCA (1984) these airways were generally aligned to
avoid all special-use airspace and coincide with regional traffic flows. They
consequentlyoffered little in mileage savings over VOR airway structures. The
RNAV structure did not take into consideration the center-to-center traffic
flow that had evolved. Since preferential procedures are usually tied in to the
VOR system, the high-altitude RNAV airway structure was not used to any
72 ROBERT J. KELLY AND DANNY R. CUSICK

great degree after a short period of interest in the early 1970s. In 1978, the
RNAV airway structure was significantly reduced, and in early 1981 all
published high-altitude RNAV routes were revoked.
By the end of the 1970s, it became apparent that the potential economy
and utility of RNAV could best be realized by random area navigation routes
rather than by the previous grid system. Without a fixed route structure to use
in normal flight planning, pilots and air traffic controllers developed an
informal system in which pilots requested “RNAV direct destination” routing
from controllers after they were airborne and beyond the airport terminal
area. Such direct clearances are granted when possible. Figure 19 illustrates
such a requested direct destination route. These procedures reflect the original
intent of the RNAV concept wherein the “R’in RNAV signifies “random.”
An operational evaluation by the FAA showed that there were no adverse
effects in using latitude and longitude coordinates for domestic routes having
direct random route clearance (FAA, 1981). Advisory circular 90-82 (FAA,
1983d) was issued, authorizing random routes for RNAV equipped aircraft
having 1and (D coordinates and flying above 39,000 ft.
The evident benefits in the use of RNAV are a reduction in flight time
obtained by following the shortest or best route between origin and destina-
tion, and reduced fuel consumption. Other benefits include more efficient use
of airspace, reduced pilot workload, and weather avoidance.
A status review of RNAV is given in Quinn (1983),where an informal 1981
FAA study estimated that the largest number of RNAV avionics units was
VOR/DME-based (over 33,000 units). About 20,000 of these units accepted 4
or 5 waypoints. Costs range from $2,000 to $20,000. The balance of the units
included Omega/VLF (6,400 units) with costs ranging from $20,000 to
$80,000; LORAN-C (700 units) with costs ranging from $2,000 to $20,000;and
expensive INS units ($100,000 and above). It is estimated that today about
20% of the civil fleet has RNAV. The RTCA minimum standards on airborne
RNAV equipment using VOR/DME imputs are documented in RTCA (1982).
As shown in Fig. 4,RNAV outputs can be displayed to the pilot for manual
coupled flight or they can be coupled directly as steering commands to the
autopilot. Autocoupledflightsusing RNAV steeringcommands result in small
tracking errors (Fig. 13).This means that the optimal maneuvers necessary to
fly the desired course can be chosen using, in some cases, a flight management
system (DeJonge, 1985), which minimizes fuel consumption. Moreover, since
the pilot has confidence in his RNAV coupled AFCS, he can reduce his work-
load. He is free to fix his attention on other cockpit concerns.
For manual coupled flights, the RNAV inputs can be as simple as a
position-fix VOR/DME or DME/DME, with little or no inertial aiding
(Bryson and Bobick, 1972). Autocoupled systems use VOR/DME or
DME/DME with inertial aiding for dynamic damping to reduce control
DISTANCE MEASURING EQUIPMENT IN AVIATION 73
activity (ROLL) and tracking errors. In the most sophisticated RNAV
implementations, INS is used as the primary navigation input with radio
aiding by VOR/DME or DME/DME [Karatsinides and Bush, 1984; Bobick
and Bryson, 1973; Zimmerman, 1969). Minimum operational peformance
standards for these multi-sensor RNAV systems have recently been published
in RTCA (1984).
With over 750 VORTAC stations currently in service in the continental
U.S., VOR/DME signal coverage is not a problem in flying defined air routes
above the minimum enroute altitude. It is, however, incomplete for aircraft
that regularly operate at low altitudes (above 2000 ft) in offshore and
mountainous areas in the eastern part of this country. For direct requested
RNAV routes, signal coverage is a factor which must be considered. According
to Goemaat and Firje (1979),RNAV coverage throughout the 48 states above
10,OOO feet using VOR/DME is essentially complete. Coverage criteria were
based upon the FAA Standard Service Volume (SSV) (see Section 111,C).
RNAV coverage using DME/DME was more complete than VOR/DME
because a line-of-sight criterion could be employed instead of the more
restrictive SSV. For example, 86%of the continental U.S. (CONUS) had dual
DME coverage at 3000 ft above ground level.
Phasing out the current airway structure and converting to a more flexible
system of area navigation is a process that will require many years to complete.
At present, the FAA is upgrading VORTAC stations to solid-state equipment
at a cost of roughly $210 million (fiscal year 1980 dollars). At the same time,
ihe question of adopting a new navigation technology to conform to
new international standards is scheduled for consideration by the ICAO.
The issue is not so much selection of a single new navigation system to
replace VORTAC as it is a question of adopting procedures for worldwide
navigation using RNAV, which will be compatible with several possible
technologies.
Nevertheless, the national airspace plan has been updated to ensure the
continuation of VORTAC service until 1995. Present indications are that
these systems will remain prime elements in a mix of future navigation system
plans which extend to beyond the year 2000. Subsection F compares the
accuracy performance of DME/DME, VOR/DME, and TACAN, using an
RNAV computer, and discusses additional RNAV system considerations.

b. Scanning DME. A position fix can be determined from the range


measurements obtained from two DME ground stations. The intersection of
the circular locus of points representing each DME range measurement
defines the airborne receiver position. This position fix technique is referred to
as DME/DME. In areas that have suitable DME/DME coverage the
potential accuracy of dual DME offers a significant improvement over
74 ROBERT J. KELLY AND DANNY R. CUSICK

VOR/DME. The degree of accuracy improvement is dependent upon the


geometry between the interrogator and the selected DME ground stations. In
principle these measurements can be made using two or more interrogators,
requiring provisions for channel selection and synchronization of the range
measurements from the individual interrogators. In one implementation
which complements an INS (Karatsinides and Bush, 1984), two DME
measurements are processed every 5 s. The “scanning” DME, however,
performs all these tasks in a single interrogator.
The “scanning” DME innovation provides the potential for rapid position
fixes (position update rate) consistent with the AFCS control laws used in
many jet transports. The idea is to use rf frequency-hop techniques to scan the
spectrum of usable ground transponders in a given geographic area. The
scanning DME, as defined in Aeronautical Radio, Inc. (ARINC) (1982), has
two frequency scanning modes, a directed frequency foreground mode and a
background mode.
The tuning source may designate from one to five stations for geometry
optimization, with one of these stations available for cockpit display. These
stations are designated as foreground stations and are placed in a loop which
cycles through the full list at least once every 5 s. All five selected stations are
identified by separate digital words containing both the rf channel frequency
and the distance information.
A second loop, called the background loop, provides the acquisition and
output of data obtained during the background scan mode. When in the
background mode the interrogator scans through the entire DME channel
spectrum, excluding those channels already designated in the foreground loop.
For any station where squitter pulses are received during the background
scan, the interrogator computes the distance to that station. Both the distance
measurement and the channel frequency are given to the aircraft navigation
systems utilizing the DME data. The ARINC characteristic states that the
initial scan of the background channels should be completed in a maximum of
three minutes. This full spectrum scan time is based upon 20 channels,
occupied by a ground transponder with a 70%reply efficiency and a minimum
reply signal strength of - 87 dB m.
In effect, the scanning DME has six time slots available every second. Five
are utilized for obtaining the range to the designated foreground stations while
the sixth, or “free scan” time slot, obtains range data on channels in the
background loop whenever stations are present. As required by ARINC
(1982), the data output rate (on the ARINC 429 bus) shall not fall below six
outputs per second and the measurement age of the data output should not
exceed 0.2 s. This means that each time slot (or data output) per frequency can
occur at least once per second. Since the interrogation PRF is limited to 30
pulses per second, this implies that each time slot has 3 to 4 replies available for
DISTANCE MEASURING EQUIPMENT IN AVIATION 75
filtering before being sent to the 429 data bus, assuming a 70% reply efficiency.
As the aircraft proceeds along its desired route, foreground stations are
discarded and replaced by background stations having better position fix
geometries. Algorithms have been developed that select foreground channels
to be used in determining a position fix.
The advantage of the ARINC 709 characteristic for scanning DME is its
short dwell time per channel (0.067 s.), which permits rapid position fixes.
Because of the short dwell time, the range to 5 or more ground transponders
can, in principle, be used by an RNAV computer to determine the aircraft’s
position. Using special signal processing algorithms (e.g., regression analysis
(Latham, 1974) demonstrated achievable 100 ft CEP. However, since RTCA
(1984) requires only a cross-track error of 3.8 nmi for enroute RNAV systems,
there is no motivation to reduce the errors to less than that obtained with a
two-station DME fix. In general, the noise is reduced only about 50% for a
five-station fix. Most of the error reduction is achieved by removing the bias
contribution from each ground station.
Because of the above considerations some airframe manufacturers use the
five foreground stations as follows: position fix (2),ILS (l), display (2). They
then use the background stations to ensure that the facilities chosen for their
planned navigation flight route are operating correctly. As the geometry of the
foreground stations used to obtain the position fix deteriorates (geometric
dilution of precision, GDOP), these foreground stations are replaced by the
appropriate background channels.
Station selection algorithms are under development that pick not only
stations within signal coverage but also provide acceptable position-fix
geometries whenever they exist (Ruhnow and Goemaat, 1982).This means the
stations should be offset to the left and right of the flight path. Subsection F
discusses the DME/DME GDOP considerations.
Goemaat and Firje (1979) analyzed the number of major airports within
CONUS in which an aircraft can make a landing approach down to 400 ft.
(nonprecision)using scanning DME. They concluded that 22% of the airports
examined provided DME/DME approach capability on at least one major
runway (5000ft in length or more). This coverage will increase with the
implementation of DME/P as the FAA begins its deployment of the planned
1200 MLSs.
In some systems a flight management computer system (FMCS) contains a
navigation database for all navaids (DME and VOR) in the flight area. The
selection process used in the Boeing 737-300 FMCS is detailed in Karatsinides
and Bush (1984). Failure modes for navigation systems which use INS to aid
DME/DME position-fix navigation are still under investigation. For
example, the Boeing 737-300 FMCS, in the absence of inertial inputs, switches
into a second mode and works on DME information only, with some
76 ROBERT J. KELLY AND DANNY R. CUSICK

degradation of the navigation performance. When DME fails temporarily, the


system continues in a crude inertial dead-reckoning mode.

7 . Summary
In this section, today’s ATC system was described, both from the
operational and navaid equipment viewpoint. The narrative highlighted the
roles of the conventional DME/N and of the new DME/P. For enroute
navigation the DME/N is one of the central elements to the VORTAC system
of air lane radials. Today it participates as one of the principal inputs to
RNAV. Tomorrow, it is expected that the DME/N will have an even bigger
role in the enroute air structure and in nonprecision approaches where the
scanning DME, because of its high accuracy, can replace the VOR bearing
information as a principal input to the RNAV.
With respect to the approach/landing applications, DME/Ns role is
expanding, where it will supplement the ILS middle markers and will provide
three-dimensional guidance at those locations where middle markers cannot
be installed because the necessary real estate is not available. Today and in the
future, DME/P plays a critical role in MLS. This role will become even more
visible and necessary as all the MLS operational features are exercised (1) by
increasing demand for more certified Category I1 runways, (2) by coordinate
conversionsto achieve curved approaches and (3)by providing MLS service to
more than one runway using a single ground facility.

E. Principles of Air Navigation Guidance and Control

The purpose of this section is to present the principles upon which the
guidance and control of aircraft are based, with particular emphasis on the
DME as it is used in the enroute (VOR/DME) and the approach/landing
phases of flight (MLS). For the most part, the material will be qualitative with
emphasis on the four or five key concepts associated with aircraft control
systems. It is not intended that the material represent, even in summary form,
the current thinking of today’s AFCS engineers.
These concepts are necessary to understand the requirements imposed
upon the output data characteristics of a radio navaid. An aircraft follows the
desired flight path when the AFCS control law, which requires position and
velocity information as inputs, is satisfied. The ideal or natural solution is to
directly measure the position and velocity deviations of the aircraft from its
intended path. Radio navaid and referenced accelerometers make these
measurements directly; other inertial sources are indirect measurements.
Radio navaids are desirable sources for this information because they are less
expensive and more easily maintained than referenced accelerometers.
DISTANCE MEASURING EQUIPMENT IN AVIATION 77
There are, however, constraints imposed upon the radio navaid when
velocity data as well as position displacement information are derived directly
from the navaid guidance signal. The two applications of interest in this
article, enroute navigation and approach/landing guidance, have widely
different requirements in terms of output signal noise and information
bandwidth, the most restrictive being the approach/landing application down
to 100 ft decision height and lower. At these weather minima imperfections in
the approach/landing guidance signal usually require that the velocity or the
rate information be complemented by inertial sensors.
This article develops the background necessary to explain the com-
plementation process, and extends it to include the concepts of path following
error (PFE) and control motion noise (CMN) which form the basis of the
DME/P and MLS angle accuracy specifications.

I . Feedback-Controlled Aircraft- General Principles


Aircraft use enroute navaids to fly between terminal areas and, once in the
terminal area, landing guidance aids are used to complete their flight. These
navaids can be coupled to the AFCS manually (through the pilot) or
automatically via an autopilot. In either case, the guidance loop is closed and a
feedback control system is created. Feedback control systems tend to maintain
a prescribed relationship between the output and the reference input by
comparing them and using the difference as a means of control. Such systems
play a dominant role in flight control because they suppress wind disturbance
and reduce the phase delay of the corrective signals and the effects of
nonlinearities. For example, in the presence of disturbances such as wind
gusts, feedback control tends to reduce the difference between the reference
input (the desired flight path) and the system output of the actual aircraft
position.
When the pilot is operating in clear air, information about the attitude of
his aircraft, its position, and velocity with respect to objects on the ground and
other aircraft are fed back to him through his eyes. Under these VFR
conditions, a navaid is not necessary because the pilot navigates using visual
“landmarks.” The feedback system is then a special case of that shown in
Fig. 12, where the block marked nauaid is replaced with a straight line and
the display is eliminated.
When visual information is unavailable, the pilot can use his cockpit
instruments to obtain pitch, roll, heading, air speed, rate of turn, altitude, and
rate of descent and, when combined with considerable skill, may reach his
destination. In general, however, he does not have the information to obtain
the guidance needed to follow a prescribed path over the ground, much less for
approach/landing maneuvers in mixed air traffic.
78 ROBERT J. KELLY AND DANNY R. CUSICK

It is the purpose of the navigation system to help the pilot follow his
desired flight path by providing electronic “landmarks” under IFR conditions.
This desired flight path can be stored, for example, in computer memory. In
some cases, only several numbers need be preprogrammed; this is possible
when the flight path is simply a straight-line segment connecting two known
enroute waypoints or is a line segment connecting the ILS outer marker with
the decision height during a landing approach.
Information from the navaid “tells” the pilot where the aircraft is. The
difference between the desired flight path and the actual aircraft position is the
correction (or error) signal. The pilot flies along his desired flight path by
maneuvering his aircraft toward an indicated null position on his display.
This null seeking process is mechanized using a negative feedback control
system, as shown in Fig. 28. In accordance with control system terminology,
the aircraft is the controlled element, the navaid generates a measurement
relative to a known reference, and the controller is a human pilot or autopilot
which couples the correction signal to the aircraft. The correction signal is
maintained near its null position by changing the direction of the aircraft’s
velocity vector by controlling the attitude of the aircraft. Inducing a roll angle
command causes the aircraft to turn left or right from its flight path. Similarly,
a change in its pitch angle causes the aircraft to fly above or below its intended
flight path. An early paper which discusses the general problem of automatic
flight control of aircraft with radio navaids is by Mosely and Watts (1960).
As stated in Graham and Lothrop (1955), “the human pilot himself
represents an equivalent time lag of 0.2-0.5 second, therefore, he is seldom
able to operate at ‘high‘ gain without instability (overcontrolling). For this
reason the pilot is generally unable to achieve the accuracy and speed of
response of an automatic system. He represents, however, a remarkable

STEERING -
‘(t?;A;pdN’
.
CONTROLLER AIRCRAFT
(AUTOPILOT) (CONTROLLED
ELEMENT)

POSITION
a
ESTIMATE

-
A
r(t)
AIRFRAME
ATTITUDE c
SENSORS

t OBSE&VATlON
RADIO
NAVAID
(SENSOR)
at)
DISTANCE MEASURING EQUIPMENT IN AVIATION 79
variable-gain amplifier and nonlinear smoothing and predicting filter. He can
stabilize and control a wide variety of dynamics. Memory and rational
guesswork often play an effective part in deciding his control actions.”
An autopilot can be used to overcome the limitations of the human pilot.
Autopilot feedback control provides speed of response and accuracy of con-
trol. These particular advantages are enhanced by high gain. Unfortunately,
the combination of fast, high-gain response will adversely affect dynamic
stability. If time lags are present in the control system, high gain can cause the
system to hunt or oscillate. It also usually increases the susceptibility of the
system to spurious signals and noise. For these reasons, an autopilot, unlike
the human pilot, usually gives satisfactory performance in only a narrow range
of system parameters.
As noted in Graham and Lothrop (1955), design of an automatic
feedback control system for an aircraft begins with a thorough understanding
of the inherent mechanics of aircraft response to its controls. The conventional
airplane has at least four primary flying controls-ailerons, rudder, elevator,
and throttle-which produce forces and moments on the airframe.6Actuation
of these controls gives rise to motions in the six output variables: roll, pitch,
heading, forward velocity, sideslip, and angle of attack. These motions (in
particular the aircraft attitude change) redirect the velocity vector (whose time
integral is the flight path), as shown in Fig. 29. Qualities of an aircraft which
tends to make it resist changes in the direction or magnitude of its velocity
vector are referred to as stability, while the ease with which the vector may be
altered to follow a given course are referred to as the qualities of control.
Stability makes a steady, unaccelerated flight path possible, and maneuvers
are made with control.
2. Aircraft Guidance and Control Using Navigation Systems
As shown in Fig. 29, the aircraft’s center of mass (c.m.) is positioned on the
tip of a position vector r(t). Its time derivative dr(t)/dt is the aircraft velocity v.
Its components are measured relative to the same earth reference coordinate
system as the position vector. The aircraft’s attitude (pitch, roll, and heading)is
measured relative to a translated replica of the ground coordinate system
which rides on the tip of the moving position vector. That is, the rotation of
the aircraft’sairframe coordinate system (Fig. 29) is defined by the Euler angles
relative to the translated earth-fixed coordinates. Similarly, the guidance
signal after a suitable coordinate transformation in the airborne navaid is
defined with respect to the earth-fixed coordinate frame. Since the aircraft’s
position as measured by the navaid and the pilot’s desired position are in the

In the cockpit, rotation of the pilot’s wheel moves the ailerons.Pushing the wheel’s column
back and forth raises and lowers the elevators.
80 ROBERT J. KELLY A N D DANNY R. CUSICK
A/C ATTITUDE COORDINATE SYSTEM
PITCH = ep
ROLL = 0
HEADING = 0

AIC
POSITION REFERENCE LATERAL
VECTOR SYSTEM CH-ANNE L COMMAND
DEVIATION SIGNAL

FIG.29. Aircraft attitude, position, and velocity coordinate definitions.

same coordinate system, then their difference generates the correct deviation
signal.
The AFCS for a CTOL aircraft is a two-axis system having lateral (roll)
and longitudinal channels. The longitudinal channel is further partitioned into
a vertical mode (pitch) and a speed control (thrust) mode. The input signals to
these channels depend upon the deviation signal’s relation to the aircraft’s
velocity vector. For example, assume as shown in Fig. 29 that the aircraft’s
velocity vector v(t,) is in the horizontal plane (tangent plane). Then the
component of the deviation signal which is assumed to be also in the
horizontal plane and perpendicular to v(t2) will enter the lateral channel of
the autopilot.
It is important to note that the aircraft, for the practical reasons given
below, does not maintain its commanded course using angular correction
signals. Each navaid sensor must eventually have its output signal trans-
formed to the rectilinear coordinates of the earth-fixed reference.
Although an AFCS can be designed to receive guidance signals in any
coordinate system, rectilinear coordinates are preferred because the flight
mission is always defined in terms of linear measure (feet, meters). In the
aircraft landing application, the aircraft must ultimately fly through a decision
window having linear dimensions so that it will land in the touchdown zone
located on the runway. This is why the MLS angle accuracy standards (ICAO,
1981a) are given in linear measure and not in angular measure (degrees).
DISTANCE MEASURING EQUIPMENT IN AVIATION 81
Similarly, in the enroute and terminal area application, the flight paths (air
lanes) are defined and separated in terms of linear measure such that the
aircraft cannot “bump” into each other. In both cases, it is the linear deviations
from the intended course in feet or nautical miles at each point in space, which
are important for aircraft navigation. Guidance data can be derived from any
navigation sensors as long as they are transformed into the appropriate
rectilinear coordinate system of the autopilot channels.
Although not immediately obvious, even straight-in MLS (or ILS)
approaches, as illustrated in Fig. 30, require this transformation. In the
literature the process is usually called course softening or desensitization.
Lateral conversion is obtained by simply multiplying the azimuth guidance
angles by the range to the antenna site, using information provided by the
DME/P element of the MLS ground equipment.
In Fig. 31, the aircraft and the navigation aids are placed in their automatic
feedback control system configuration. The block diagram functionally
represents one of the two autopilot channels. The intended flight path is the
input signal command with the output being the actual aircraft path. The
outer loop, also called the guidance loop, contains the aircraft and flight
control systems, the ground reference navigation aid (MLS, ILS, DME,
TACAN), the airborne sensor, and coordinate transformer. A display device
which allows the pilot to monitor or use the measured deviation of the aircraft
from the selected flight path completes the system. Unlike today’s ILS, which
supports straight-in approaches only, the MLS guidance loop must be
prepared to accept any of several types of input signals as itemized in Fig. 31.
In addition to the guidance loop, there is a series of inner loops involving the
feedback of airframe motion quantities such as attitude (roll and pitch) and
attitude rate which serve to control the aircraft. These are called control loops
in Fig. 31.
As with any closed-loop control system, the AFCS guidance loop has
specific gain, bandwidth, and stability properties based on the components of
the loop. Separate guidance loops exist for each of the aircraft channels

I-:tf(
APPROACHING AIRCRAFT

5
WITH ITS COORDINATE
SYSTEM

R = DMEIP

AY R A 6 AZ

FIG.30. Simple straight-in approach coordinate system.


82 ROBERT J. KELLY AND DANNY R. CUSICK
CURVED APPROACH PATH
WAYPOINTS FOR
SEGMENTED APPROACH
PATH
mSELECTEDGLIDE PATH
AND RADIAL A I R C R A F T. -[NEAR
._...-..-. I
FROM

-
Y - DEVIATION
COMMAND
INPUT
SIGNAL 4- FLIGHT POSITION (FEET)
GAIN +CONTROL AIRCRAFT
(COMMANDED @OEVIA!IOtFROM SYSTEM
POSITION) -f 3

COORDINATE
TRANSFORM

t AIRBORNE
ANTENNA
- RADIO
NAVAID
L

mentioned above. For CTOL aircraft the guidance loop bandwidth is less than
0.5 rad/s for the lateral autopilot channel and less than 1.5 rad/s for the ver-
tical (pitch) mode of the longitudinal channel.

3. Qualitative Description of the Aircraft Control Problem


It is important in the following discussions to understand the distinctions
between the navaid output observations, the navigation system steering
commands, control law, and control action. These distinctions will be clarified
below using as an example the system shown in Fig. 28. Illustrated in the figure
is a functional description of an aircraft/guidance system which uses inertially
aided position-fix measurements from a radio navaid. The navigation system
is composed of the radio navaid sensors, airframe attitude sensors, aircraft
position estimator, and the steering command summer. The problem is to
control the aircraft such that its flight path r(t) follows the commanded path
c(t).How closely r(t) follows c(t)is determined by how timely and accurately
the aircraft velocity vector d r / d t can be changed.
Automatic flight control systems are easily comprehended using the
notation and bookkeeping of the state-variable formalism (Schmidt, 1966).In
DISTANCE MEASURING EQUIPMENT IN AVIATION 83
this formalism, drldt is related to r and the airframe attitude states through a
system of first-order differential equations. Integrating dr/dt then yields the
aircraft flight path.
Controlling the aircraft trajectory proceeds according to the following
sequence:
(1) Radio navaid sensors (e.g., VOR/DME) measure the aircraft position
state observable z(t), which contains the aircraft true position states r ( t ) plus
random disturbances.
(2) Estimating techniques on board the aircraft attempt to extract r ( t )
from z(t). The resulting estimate r ( t ) can be obtained by suppressing the
random disturbances using a simple low-pass filter or a sophisticated
technique such as a Kalman filter.
( 3 ) The estimate r(t) is compared with c ( t )to determine if control action
is required. If they are equal, no added control is necessary. That is, the
steering command remains unchanged, which, in turn, does not change the
aircraft control surfaces (e.g., ailerons or elevators).
(4) A “control law” is used to calculate the control action necessary to
make the actual flight path r ( t ) at time t + At equal the command c(t + At).
The “control law” must also use the velocity vector dr/dt to ensure that r(t) is
stable and does not oscillate around c(t). The velocity input dr/dt can be
obtained from r ( t ) directly, or it can be derived from the airframe attitude
sensors.
( 5 ) Application of control action in accordance with the “control law” is
achieved by changing, for example, the aileron position to bank the aircraft SO
that its c.m. position r ( t ) will move to follow the command c(t).
Using the above sequence to fix the ideas contained in Fig. 28 permits the
terms guidance and control to be succinctly defined. Guidance (McRuer et al.,
1973) is the action of defining the velocity relative to some reference system to
be followed by the aircraft. Control is the development and application of
appropriate forces and moments that (1) establish some equilibrium state of
the aircraft and (2) restore a disturbed aircraft to its equilibrium state. Control
thus implicitly denotes that r ( t ) will be a stable flight path which is achieved
using both position displacement feedback, r(t), and rate feedback, dr/dt.
When performing an analysis, it is the deviations about the intended flight
path c ( t )which are important. In this case, a perturbation model can be used
which, in many applications, may reduce a nonlinear AFCS to a system of
linear first-order differential equations. The following discussions assume that
the controller is an autopilot.
The key issues associated with a radio navaid and its interaction with
AFCS can be clearly presented using the lateral guidance loop as a discussion
84 ROBERT J. KELLY AND DANNY R. CUSICK

example. In the following the terms A+, A$, and Ay are deviations from the
desired attitude or course. Further the notation A j will represent the velocity
of the deviation signal at Ay.
The detailed mechanism for turning an aircraft onto a desired flight path
proceeds as follows. A lateral position correction signal Ayc (ft) is transformed
into a roll command A4,(deg) which causes the aircraft ailerons to deflect a
given number of degrees. Relative to position changes in the aircraft c.m., this
action occurs rapidly (< 1 s). The moment generated by the air flow across the
deflected ailerons causes a change A 4 in the aircraft attitude about its roll axis.
After about 1 s A 4 = A$c. Slowly the aircraft velocity vector begins to rotate
such that the aircraft’s c.m. departs laterally from its previous flight path until
Ay = Ayc. During this time the aircraft heading $ is also changing. A wide-
body jet will reach this new equilibrium state in about 10seconds. This process
continues until the aircraft is on the desired flight path as represented by a new
velocity vector with track angle $T.
The feedback loop which tracks attitude changes A$ are called control
loops; those which track the c.m. changes Ay are called guidance loops. Thus
the time constants of the control loops are faster than the time constants for
the guidance loops. The accuracy with which the null-seeking loops track the
attitude and position commands is dependent on loop gain, which, in turn,
must be chosen to ensure that they remain stable over the range of input
parameter changes. An unstable control loop would lead to uncontrolled
attitude variations, which, unfortunately, would not be related to the steering
input command signals. Under such conditions, the aircraft cannot be
controlled to stay on the flight path.
In the following discussions it will be assumed that the control loops are
stable. Attention will now focus on the requirements for a stabilized guidance
loop. A small tracking error, sufficient guidance loop stability margin, and
good pilot acceptance factors are the critical performance indices necessary to
achieve a successfulautocoupled guidance. Each of these points will be treated
separately.
Clearly, an autocoupled aircraft must accurately follow the desired track
angle in the presence of steady winds, gusts, and wind shears. To correct the
differencebetween the desired course (input) and the aircraft track (output),
the guidance system must obtain the position and velocity of the aircraft and
determining its heading. Assume a pilot wants to turn the aircraft to follow a
new course which intersects his present course to the right. His navigation
system will provide a displacementcorrection signal Ay to the autopilot which
in turn generates a roll command to bank the aircraft.
If only the position displacement signal were used, it is apparent that a
stable approach could not be achieved, because the aircraft would continu-
ously turn as it approaches the new course and overshoot it. That is, the air-
DISTANCE MEASURING EQUIPMENT IN AVIATION 85
craft’s c.m. would be accelerated toward the new course in proportion to the
distance off course as if it were constrained to the course by a spring. Without
damping, the c.m. would oscillate continuously about the new course.
To prevent this unstable maneuver, a term related to the rate of closing
toward the new course must be introduced. That is, to use the above analogy,
the spring requires damping. Such a “rate” term A j is available either by
differentiatingdirectly the position displacement A y with respect to time or by
using the relative angle between the aircraft track angle and the new course
bearing. In either case, a signal proportional A j can be derived to provide
dynamic damping.
An equally important situation arises when the aircraft is subject to steady
winds and wind turbulence. Because an aircraft has a response time inversely
proportional to its guidance loop bandwidth, presence of wind should be
detected as soon as possible. Inserting a rate signal A j in addition to Ay into
the autopilot anticipates the wind disturbance so that the aircraft’s attitude
can be adjusted to counter the effects of wind force and moments which may
push it off course. These examples qualitatively demonstrate why both
position and velocity information are necessary for stable flight path control.
To achieve the tracking accuracy necessary to guide the aircraft through
the Category I1 window on a final approach places additional demands on the
feedback signals. The aircraft must be tightly coupled to the guidance signal,
which, in engineering terms, means that its guidance loop gain must be large.
High guidance loop gain, however, is synonymous with high guidance loop
bandwidth. A wide guidance loop bandwidth is necessary so that the high-
frequency components of the wind gusts can be counteracted by the AFCS.
This is desirable to decrease the influence of wind, wind shears, and gusts
which tend to cause the aircraft to depart from its intended approach and
landing path. On the other hand, tight coupling is undesirable if the guidance
signal is noisy or has guidance anomalies caused by multipath. In this case, the
tightly coupled aircraft will tend to follow the anomalies and depart from its
intended approach and landing path. Regardless of whether the departures
from the intended path are because of winds, shears, and gusts or due to
anomalies in the received signals, they are tracking errors. With this
background the discussion will move on to examine the performance of the
radio navaid.
Unlike wind disturbances, which can be attenuated by increasing the loop
gain, guidance signal noise will not be attenuated because its source is not in
the forward portion of the loop. Increasing the loop gain (or bandwidth)
amplifies the effects of the noise. The situation is further compounded when
the rate feedback signal is derived from the guidance signal, since the
differentiating process accentuates the high-frequency components of the
noise.
86 ROBERT J. KELLY AND DANNY R. CUSICK

Reducing the guidance loop gain is not always a solution because the
effects of wind gusts and shears may increase. Removing the higher-frequency
components from the wind disturbance means that they are not available as
correction signals. The consequence is larger tracking errors. Narrow loop
bandwidths (low gain) will also degrade the aircraft's transient response,
perhaps to a point which is unacceptable to the pilot.
Heavy filtering of the navaid output signal does not alleviate the above
problems. The data filtering process introduces phase delays on the guidance
signal that may erode the AFCS stability margin, which is only about 60" to
90" to begin with. That is, an additional phase delay of 60" to 90" will
destabilize the system.
There is yet a third problem. To maintain a sufficient phase margin, the
navaid data filter bandwidth must be much wider than the guidance filter
bandwidth. Guidance signal noise may disturb the roll and pitch control
loops and also the control surfaces (ailerons and elevators). Since the wheel is
coupled to the ailerons and the column is coupled to the elevators, unwanted
control activity may also induce unwanted wheel and column motion. During
an autocoupled approach, excessive control activity can cause the pilot to lose
confidence in the AFCS. Nevertheless, although these noise components cause
airframe attitude fluctuation and control surface activity, their spectral
content lies outside the guidance loop bandwidth. The induced motions are
too rapid to influence the direction of the aircraft's c.m. and therefore do not
afect the aircraf's Jlight path.
How are the above noisy navaid issues resolved? For some autocoupled
enroute applications using VOR/DME and for some ILS approaches, the
solutions include (1) doing nothing and accepting the unwanted control
activity, (2) decreasing the loop position gain and accepting larger tracking
errors, (3) using inertial aiding for dynamic damping, (4)reducing the control
activity by reducing the autopilot gain and the navaid filter bandwidth.
Clearly the solutions involve trade-offs between tracking error, stability
margin, windproofing, control activity, and costs. A more sophisticated
solution is the complementary filter.' Subsection E,6 describes the comple-
mentary filter approach used for more demanding ILS applications.
MLS system designers took the approach that the radio navaid designs
must begin with a low-noise output signal. The MLS angle and the DME/P
standards, which are the most recent navaid standards approved by ICAO,
' A complementary filter blends the good high-frequency characteristics of the inertial
sensors with the good low-frequency characteristics of the radio navaids. The overall combination
will then have the desired wideband frequency response with low sensor noise and minimum data
phase lag. A complementary filter permits the radio navaid output data to be heavily filtered if
necessary.
DISTANCE MEASURING EQUIPMENT IN AVIATION 87
attempt to correct the deficiencies of the previous systems by purposely
defining low-noise accuracy standards even in the presence of multipath.
Despite these more stringent requirements, the MLS signal-in-space may
require some degree of inertial aiding or complementary filtering to provide
the high dynamic damping and the position accuracy necessary for achieving
successful Category I1 approaches, Category I11 landings, and automatic
landings under VFR conditions.

4. Radio Navaid Data Rate


In all AFCSs the data sampled guidance signal must eventually be
reconstructed back into an analog signal since, in the final analysis, the aircraft
moments and wind forces are analog by nature. For sampled data systems
such as MLS, DME, and TACAN, the position samples must be provided at
an adequate data rate so that the signal can be faithfully reproduced.
Undersampling can generate an additional tracking error component and can
induce an additional phase delay after the signal is reconstructed. For
example, guidance deviations caused by wind gusts must be accurately
reconstructed so that the control loops can apply the appropriate compensat-
ing corrections in a timely manner. Reconstruction circuits (e.g., a zero-order
hold, ZOH) introduce phase delay in addition to the data smoothing filter.
In order to provide a low-noise guidance signal with a bandwidth that
satisfies the requirements developed in Subsection E,3, the navaid data rate
must be high enough to (1) minimize the signal reconstruction errors and the
reconstruction circuit phase delays, and (2) allow data smoothing while still
providing enough independent samples to reconstruct the signal. It can be
shown that if thebutput data filter satisfies a six-degree phase delay criteria,
then it will also satisfy the data reconstruction requirements for independent
samples (Ragazzini and Franklin, 1958).Therefore, only the six-degree phase
delay filter criterion need be satisfied, as discussed in Subsection E,7.
The second requirement is that the data rates must be high enough to
permit filtering of the raw data samples if the CMN is excessive. For a given
noise reduction factor g, the second data requirement can be derived from
Eq. (40)in Section IV,D,l. It is data rate = f , = 29' (noise BW), where (noise
BW) = (a/2)wfi, for a single-pole filter. For example, if the raw CMN is
0.14, then to achieve 0.05" CMN at the input to the AFCS requires that
g 2 = 8. Assuming mfi, = 10 rad/s, then the data rate must be at least 40 Hz.
Figure 37 illustrates the concepts described above. The data rate relation
applies when the noise source has a broad power spectral density such as the
MLS azimuth function. If the radio navaid sensor cannot provide an ade-
quate number of independent samples, some form of inertial ordering will
85 ROBERT J. KELLY AND DANNY R. CUSICK

be necessary. Example implementations are given later in Fig. 38.’ The need
for small tracking errors and wide-bandwidth navaid filters that minimize
phase delay points toward data rates which are about 20 times the loop
guidance bandwidth.

5 . Functional Description of AFCS


The conclusions in Subsections E,3 and E,4 placed qualitative accuracy,
bandwidth, and data rate constraints on the output data of radio navaid
sensors. With these constraints an aircraft guidance system can be designed
that is accurate, responsive, and stable. In this section, the key issues will be
revisited using the mathematical formalism of control theory. This elementary
review will serve to fix clearly the ideas treated earlier and to indicate how the
aircraft and its navigation system can be analyzed quantitatively.
For the moment, the AFCS functional description will be general, applying
either to the lateral or to the vertical channels. Figure 32 is the basic control
system, where K is the autopilot gain, H,(s) and H2(s) represent the transfer
functions of the autopilot, aircraft, and the aircraft dynamics. F(s) is the
transfer function of the sensor, s is the Laplace transform variable, and the
input/output functions, c(t) and r(t), are scalar quantities.
In an ideal control system the controlled output r ( t ) should followthe
commanded input c(t) with the smallest mean-square tracking error e(t)’,
where e(t) = c ( t ) - r(t). This is the optimum meaning of accuracy, fast
response, and stability. The intended purpose of the AFCS is to keep the
aircraft as close as possible to its intended course c(t), even in the presence of
wind gusts d(t) and with an imperfect radio navaid having error n(t).
Let R(s) and C(s) be the Laplace transforms of r ( t ) and c(t), respectively.
For the time being, let the sensor error source n(t) = 0 and the disturbance
d ( t ) = 0; then, from Fig. 32,

Equation (1) is the closed-loop transfer function of the aircraft’s guidance


loop. The 3 dB bandwidth of R(s)/C(s)is the loop guidance bandwidth w 3dB;
the forward transfer function is KH,(s)H,(s), and the open-loop transfer
function is KF(s)H,(s)H,(s). For radio navaids, F(s)is equivalent to a low-pass
filter. As stated in Subsection E,3, the 3 dB bandwidth of the navaid sensor
transfer function F(s) should be much larger than the loop guidance
bandwidth w~~~ for two reasons. One, the phase delays of the corrective
signal’s components should not erode the phase stability margin of R(s)/C(s).
* Accuracy and data rate limitations are among other considerations that may preclude GPS
from being used as a primary navaid in approach/landing applications.
DISTANCE MEASURING EQUIPMENT IN AVIATION 89
WIND DISTURBANCE
d(t)

GAIN K

AIRCRAFT

FIG.32. Generalized block diagram of attitude controlled A/C and navaid.

Two, the spectral component of c ( t ) and d ( t ) should be faithfully reproduced


so that e(t)2 is maintained within design limits. When F(s) cannot satisfy these
requirements, the radio navaid may require some form of inertial aiding.
An important design consideration is the effect of sensor disturbances
upon the tracking error. The AFCS should respond to the desired signal c(t),
but not respond to extraneous noise sources. Let N(s)be the Laplace transform
of n ( t ) the error term of the navaid sensor.
In Fig. 32, let c(t)=O and d ( t ) = O ; then the closed-loop transfer function
R(s)/N(s)for the noise source is identical to Eq. (1). Again, the navaid sensor
determines the guidance system performance. Unfortunately, the character-
istic that F(s) should possess for small tracking errors, when combined with an
imperfect navaid, is in conflict with those required for the command signal c(t).
For the command signal, the bandwidth of F(s) should be greater than co3dB,
while, for the navaid error source n(t), the bandwidth of F(s) should be small.
In other words, the spectral content of c ( t ) should be unaltered while that of
n(t) should be suppressed when the variance of n(t) is excessive.
An important property of a closed-loop AFCS transfer function is its
capability to suppress wind disturbances. Again the essential design con-
sideration is high loop gain. In Fig. 32 let D(s)be the Laplace transform of d ( t )
and let c(t) = n ( t ) = 0. Using Fig. 34 the closed loop transform for a wind
disturbance is
R(s)
- - - H,(s)
(2)
D(s) 1 + KF(s)H,(s)H*(s)
The effects of negative feedback on d ( t )can be compared with c ( t )and n ( t ) by
inserting the wind disturbances on the input side of H(s). The transfer
functions for all three sources are now equal to Eq. (1) if D(s) is transformed
90 ROBERT J. KELLY A N D D A N N Y R. CUSICK

into D(s)/KH,(s).Wind disturbances are therefore reduced by large autopilot


gain.
In summary, the loop bandwidth, w 3 d B , must be sufficiently large to
reproduce c ( t ) faithfully and attenuate d ( t ) , but not to the extent that n(t)
introduces excessive tracking errors. After choosing w3 dB, the navaid noise
can be filtered to a level consistent with the desired control activity and
windproofing using a complementary filter. The formalism above implicitly
assumes the control loops needed for dynamic damping. The tracking error
e ( t ) thus contains three components: the effects of the band-limited signal c(t),
the wind disturbance d(t), and the radio sensor error n(t).
a. Lateral Channel Control Law. The control law for the lateral channel will
now be developed. The lateral channel is chosen because essentially the same
control law is used in the enroute, terminal, approach, and landing phases
of flight. As stated in Subsection E,3, it is necessary that the bank angle com-
mand be proportional to the lateral deviation from the desired flight path.
Moreover, in order for the flight path to be stable, the bank command must
also be proportional to the rate of change of the path deviation. Consequently,
the bank angle control law assumes the form
A 4 c = K , A y +- K + A j (3)
Not included in Eq. (3) is integrated feedback, which removes any dc offset
errors. Control law gain coefficients are chosen to ensure a given stability
margin while maintaining an operationally acceptable tracking accuracy. It is
an involved analysis using root locus techniques (see Bleeg, 1972,Appendix B).
Although the derivation of these coefficients requires extensive analysis, once
they are determined, 03dB can be derived using simple relations such as those
given in Fig. 33(b). This is fortuitous because the guidance loop bandwidth
(03 dB) is critical to the navaid filter design. The control loop's performance,
on the other hand, will not succumb to a simple analyses; a full-blown
computer simulation is necessary.
At this point, the design engineer must make several decisions. From what
sensor should the lateral path deviation A y and its velocity A j be measured?
Ideally, one would want to obtain A y from the ground reference navaid, and
using that position measurement, derive the velocity data needed for dynamic
damping. Relative to inertial aiding this is the most economical approach.
However, as noted in Subsection E,3 there are several important con-
siderations depending upon which phase of flight is being addressed.
As expected, the autocoupled final approach/landing phase has the most
stringent requirements. Anomalies in A y and A j must be limited such that
unwanted roll commands A$c as expressed by Eq. (3) are limited to less than 2"
(95% probability) (paragraph 2.1.4, Attachment C, ICAO, 1972b; Kelly, 1977)
and Fig. 38. It is the opinion of many fight control engineers that the 2" roll
DISTANCE MEASURING EQUIPMENT IN AVIATION 91
requirement for autocoupled flight also applies to the terminal area and
enroute applications. Control law considerations for MLS are discussed in
Section II,E,6.
For the enroute application, a wide variety of control law implementa-
tions are available to the user, depending upon whether he flies the VORTAC
radials or uses RNAV routes, and whether he files manual or has his RNAV
outputs coupled directly to the autopilot. When flying manual, the displayed
VOR and DME raw data may be adequate because the pilot is more forgiving
when he is in control of the aircraft. However, when the aircraft is in the
autocoupled mode, autopilot tracking errors are viewed more critically by the
pilot because he is not in direct control of the aircraft. Thus, DME and VOR
position data for the autocoupled mode may require velocity inertial aiding.
Anomalous attitude or control surface activity under no wind conditions is
disconcerting to the pilot and may result in negative pilot acceptance factors.
On the other hand, the pilot expects control activity when he encounters air
turbulance. Small tracking errors in the presence of wind and good pilot
acceptance factors are desirable because pilot work loads can be reduced and
fuel consumption can be minimized. Only when the pilot can use the
autocoupled mode with confidence will these goals be achieved. Although
some ILS landing operations require the position data to be complemented
with inertial aiding (see Fig. 38c), the major concern is the accuracy of the
velocity information. Only the velocity information considerations will be
treated below.
In order to address explicitly which variable can be used to provide the
velocity feedback components, the physics and geometry of how an aircraft
negotiates a lateral turn must be analyzed. Figure 33 illustrates such an
analysis, where a small-signal perturbation model has been developed, i.e.,
c ( t ) - r ( t ) = Ay when the desired path has changed from c ( t ) = 0 to c(t) = Ay.
The necessary variables can be determined using Eqs. (1)-(5) in Fig. 33(b).
They are derived by successive integrations beginning with the equation of
motion for a coordinated turn [Eq. (I)]. Equations (1)-(5) illustrate how the
velocity vector is changed so as to achieve a desired flight path. For ease of
exposition the equations of motion in Fig. 33, assume that the direction of the
aircraft’s c.m., ICIT, equals $; i.e., the effects of wind are negligible. Also assumed
is the complete decoupling between the roll and directional axes. Since o3dB is
small compared with the natural frequency of the bank angle transfer function
6/OCN 1. The actuator dynamics have also been neglected.
The variables available for dynamic damping are:

(1) A j as derived from Ay [Eq. (4)]


(2) bL as integrated from 4, where 6 is measured by a vertical gyro
CEq. (113
92 ROBERT J. KELLY AND DANNY R. CUSICK
a

i
AIRBORNE
RECEIVER
e.g VORIDME
P. 0 GROUND REFERENCE
NAVAIDS
e.9. VORIDME
-
@ ‘BANK ANGLE (DEGREES)
$ = AIC HEADING IDEGREESI
”,
p w =WIND GUST IDEGREESl

W = WIND SPEED IFEETISECOND)


V, = AIC AIRSPEED (FEETISECOND)
9 = GRAVITY I32 FEETISECOND’)
$ = LAPLACE INTEGRATION OPERATOR

S = LAPLACE DIFFERENTIATION OPERATOR

NOTE: A y IS I N THE HORIZONTAL PLANE NORMAL TO c(tl

FIG.33. (a) Lateral channel perturbation model.

(3) $ (heading), as measured by the directional gyro or radio compass


rEq. (3)1
(4) $T (track angle), as determined from v and measured by referenced
accelerometers. Under wind conditions $T # $ at point A in Fig. 33. $T =
tan-’(j/i), where v = Jm.From v the feedback correction term A$T
is calculated, where is the difference between the commanded track angle
and the true track angle $T. (See Fig. 11.)
Assuming that A j cannot be derived accurately from the radio navaid, A$T
is the ideal input for dynamic damping, since it is based upon the true tracking
angle of the aircraft. However, the determination of At)T requires an expensive
INS platform or its equivalent in strap-down inertial sensors. Less accurate,
but perhaps adequate, tracking performance can be obtained from the heading
$ and the bank angle 4 to obtain the damping signals.
In the 1950s and 1960s, heading feedback was the solution used in many
autopilots for approach and landing operations. However, when the aircraft is
subjected to a low-frequency crosswind, the tendency of the airplane is to
DISTANCE MEASURING EQUIPMENT IN AVIATION 93

ASSUME va = 7

(1) -
mv2
= mg Tan A@
Ra
BUT V = A $ X R, THEREFORE

t
mg
FIG.33. (b) Physics of a lateral turn.

"weathercock" in the wrong direction. This produces a heading change which


is not related to +T (Bleeg et al., 1972). In general

$*=*,+*=*,+- :sd 4dt

where $, is the initial wind heading transient, 4 is a function of time, and the
(4)

wind disturbance is injected as indicated in Fig. 33.


Thus, using $ as a damping term, is not an ideal solution in the presence of
steady wind and low-frequency wind shears. Before the advent of INS, the
means to obtain dynamic damping and windproofing was to measure the bank
angle directly and to perform a long-term constant integration, as indicated by
94 ROBERT J. KELLY AND DANNY R. CUSICK

Eq. (4). A solution was to incorporate the operation called lagged-roll, as


indicated in Fig. 33. However, pure integration was not desirable because the
airplane may be mistrimmed; i.e., it may have a steady roll angle which could
prevail throughout the approach.’ Another approach to obtaining dynamic
damping is to combine heading with a rate signal derived from the imperfect
navaid sensor in a complementary filter. The heading signal has good high-
frequency response and correspondingly poor low-frequency wind response
while the navaid rate signal has good low-frequency characteristics. Such an
example is given in Brown (1983) for the ILS application.
Because the exposition which follows will be simpler and clearer, $ will be
used to obtain A j with the full understanding of the limitation mentioned
above. The use of K , A$ = K+A j will allow an estimate to be made in the
tracking error as generated by the sensor noise errors. Equation (3) now
becomes
A4= = K , A y + (5)
The block diagram in Fig. 33 is obtained by starting with Eq. (5) and using
Eqs. (2)-(5) of Fig. 33. The blocks with dotted inputs and outputs are the
alternate feedback terms A j and A&. Wind effects which push the aircraft c.m.
off course can be included by a summing junction in Fig. 33. For a steady
crosswind, the input to the summing junction is a step function. The resulting
path deviation A y would be a ramp function. A wind shear would be
represented by a ramp function at the summing junction input.
The composite transfer function using only heading feedback is given in
Fig. 34. The analysis assumes that the sensor output data filter bandwidth
is large compared with the guidance loop bandwidth (look ahead to
Fig. 38). As shown in the figure, the transfer function is a low-pass filter of
second order, which is given in standard form in terms of the damping
factor t and the natural frequency wo to emphasize the important perfor-
mance factors. Also given is the expression relating the loop guidance band-
width w3dB to w o . By choosing K , and K,, the loop guidance bandwidth
wJdBand t can be adjusted to suit the enroute, terminal, approach, and
landing phases of flight. For the approach/landing phase 5 z 0.5-0.7 and
w3dB2: 0.1-0.2 rad/s. Note that if K , = 0 (no damping, t = 0) then the guid-
ance loop is conditionally stable; i.e., it will become unstable with any input.
This condition dramatizes the need for velocity information in the aircraft
guidance loop. As shown in Fig. 34, increasing or decreasing K y changes w 3dB
as mentioned several times in Subsection E,3. If rate feedback were incor-
porated as shown in Fig. 33(a), then wo = 4- and 5 = 3(Kj/K,)wo.
Using I) in the control loop to maintain a constant aircraft heading will cause the aircraft
c.m. to drift off course in the presence of wind. To avoid drift errors in the approach/landing
application the aircraft “decrabs”just before touchdown.
DISTANCE MEASURING EQUIPMENT IN AVIATION 95

NATURALFREQUENCY
OF THE GUIOANCE LOOP

STANOAROIZEO TRANSFER FUNCTION

LOOP GUIDANCE BANOWIOTH w3db =wo Jc 1-2 t 2, + 2-4 t 2+ 4


KII, IS IN '$)/ OJ/
NOTE:
KylSIN O@/FT THE COEFFICIENTS VARY SLOWLY
WITH TIME AS A FUNCTION OF THE
W o IS I N RAOIAN / SEC AFCS GAIN SCHEDULING
FIG.34. Transfer function of lateral channel perturbation model.

Using certain simplifying assumptions, the desired tracking error can be


easily calculated. The mean-square tracking error e ( t ) = c ( t ) - r ( t ) is defined
by
-
e(t)2 = lim -
S'
e 2 ( t ) d t=
T - ~ -~T T
where E(s) is the Laplace transform of e(t).
If the input command signal c(t), the wind distance d(t), and the radio
navaid errors n ( t ) are zero mean uncorrelated random variables, then
3 2 = C(t)++ d(t)++ n(t)+ (7)
96 ROBERT J. KELLY AND DANNY R. CUSICK

where each component on the right is the tracking error contribution due to
c(t),d ( t ) , and n(t), respectively. Data sampled reconstruction effects have not
been included. (See Subsection E,4.) Note that inertial aiding component
errors are assumed to be negligible. Assume in the following that the
bandwidth of c(t)is well inside w 3dB, then c(t)gwill be small. Assume also that
the autopilot gain is sufficient to suppress the effects of d(t), then under these
assumptions, e(t)’ z n(t)+.
Since the guidance loop transfer function H ( s ) has been derived in Fig. 34
using heading as the damping term, the tracking error due to sensor error
perturbations can now be calculated. Using Eq. (6)

where E [ s ] = H ( s ) N ( s )and N ( s ) is given by (9) below. Let the sensor noise be


represented by a Gauss-Markov process (Brown, 1983), then its power
spectral density is

where the sensor filter bandwidth 0, >> W j d B and 0,’is the noise power at the
output of the sensor. Then

N(s) = m / ( S + w,)
where, for example, in the autocoupled approach/landing application
w , = 10 rad/s.
Performing the integration in Eq. (8) using the integrals evaluated in James
et al. (1947), the tracking error is
-
e(t)2= o ~ K , u / w , K , ( 5 7 . 3 ) (10)
Thus, although the tracking error due to a sensor with noise errors is
proportional to autopilot gain K,, it is also inversely proportional to K,. For a
given noise power spectral density S,(w) and aircraft speed u, the tracking error
can be reduced by decreasing the ratio K,/K,. Since there are two degrees of
freedom, K , is selected to achieve the desired dynamic damping, while K, is
adjusted to obtain the desired tracking error. The above analysis assumed that
the noise error of the heading feedback was negligible. If the navaid error n ( t )
has a bias component, it would be added to Eq. (10) on a root-sum-square
basis as indicated in Subsection C,3.
Equation (10) has a simple physical meaning when w, >> w J d B . The noise
power spectral density given in Eq. (9) becomes S , N 2a,’/w,, a constant.
Using the relations for K , and K, in Fig. 34 and letting = 1/& then
DISTANCE MEASURING EQUIPMENT IN AVIATION 97
wo = w3dB and Eq. (10) becomes
= Snw3 dB/(2 a>
The tracking error is simply that portion of the noise power residing inside the
guidance loop bandwidth which is intuitively satisfying. A similar analysis
could be performed to determine the effects of wind gusts on tracking
accuracy. Equation (9) can also be used to simulate wind gusts.
The above discussion was meant only to introduce the basic uses of AFCS
design and to show why it is necessary to have radio navaids which can
provide wide-band guidance (about 10 x w3dB)with a low-noise output. A
more complete discussion of the lateral channel as it applies to the approach
and landing maneuver is given in McRuer and Johnson (1972). In the next
section it is shown how the complementary filter permits the designer to
achieve system stability without regard to the time constant introduced by
heavily filtering the noisy radio navaid output.
6. The Notion of the Complementary Filter
In the preceding sections, the problem of using a radio navaid sensor with
a noisy output continually figured highly in all the discussions. The problem
was how to filter the noise without introducing excessive data delays. It was
suggested in the earlier narrative that the solution involved the use of inertially
derived position data which would augment the filtered radio navaid data.
Also discussed were ways of providing dynamic damping using inertial
sensors such as vertical gyros to replace rate signals derived from the radio
navaid position fixes.
The problem now is how to augment the position data fixes with dead
reckoning position data obtained from an inertial sensor. The idea is that the
radio navaid can obtain good low-frequency data after filtering while the
inertial sensors (accelerometers) have good high-frequency characteristics
(i.e., small high-frequency noise components). Drift errors are typically the
source of the poor low-frequency performance exhibited by inertial sensors.
Radio navaids have goods-frequency response because ground field monitors
ensure that the bias errors are maintained within acceptable limits.
The technique of combining the two sensor outputs to yield a wide-band
filter without phase delays, signal distortion, or excessive noise error is called
the complementary filter. The concept assumes each source has the same
signal, but the measurement noise from each source is different. That is, both
the radio navaid and the inertial sensor are measuring the same navigation
parameter, only their measurement noises are different. The development
below follows Brown (1973).
Let s(t) be the position of the aircraft in flight, and let nl(t) be the inertial
sensor noise and n2(t)be the radio navaid noise. Configure the filter as shown
98 ROBERT J. KELLY AND DANNY R. CUSICK

in Fig. 35(a). The output is the estimate $(to = z(t). The Laplace transform of
the output is
Z ( S )= S(S) + N,(s)[l - G(s)] + N ~ ( s ) G ( s ) (1 1)
Signal term Noise terms

From Eq. (11)the signal term is not affected by the choice of G(s) while the
noise terms are modified by the filter. The point here is that the noise can be
reduced without affecting the signal in any way. If, as in the navigation
problem, the two sensors have complementary spectral characteristics, then
G(s) can simply be a low-pass filter and the configuration in Fig. 35(a) will
attenuate both noise sources. Finding the optimum cutoff point of the low-
pass filter G(s) is an optimization problem which can be solved using the
Wiener filter theory. The filter configuration in Fig. 35(a) can be rearranged
into the feed-forward configuration shown in Fig. 35(b). An equivalent
feedback configuration is discussed in Gelb and Sutherland (1971).
The feed-forward implementation of Fig. 5(b) provides some insight into
how the complementary filter works. The goal is to eliminate the signal in the
lower leg by substitution thereby producing a new input signal nl(t) - n2(t).A
filter design is chosen to obtain the best estimate of nl(t) in the presence of a
distrubing signal n2(t).The estimate n,(t) is then subtracted from the sensor
+
output s(t) nl(t), yielding an estimate of s(t). Intuitively, the filter mechanism
is clarified if n,(t) is predominantly high-frequency noise and n2(t) is
predominantly low-frequency noise. In this case, 1 - G(s) is a high-pass filter
which passes nl(t) while rejecting n2(t). The Kalman filter gain optimization
for the case where aircraft range is determined from a radar and the double

FIG.35. (a) Basic complementary filter. (b) Feed-forward complementary filter


configuration.
DISTANCE MEASURING EQUIPMENT IN AVIATION 99
integration of a strap-down accelerometer is noted in Maybeck (1979). The
Kalman filter overcomes the limitations of the Wiener filter which is only
optimum for stationary random processes.
The configuration in Fig. 35(d) forms the basis of the inertially aided
navigation system, which is the cornerstone of many modern flight manage-
ment systems currently in use on new jet transports. Figure 36 (Brown, 1983;
Bobick and Bryson, 1970; Gelb and Sutherland, 1971; Karatsinides and Bush,
1984) is the state vector representation of Fig. 35(b), where multiple sensors
are used as inputs. It employs a linear Kalman filter which assumes that the
difference between the inertial and the aiding sensor errors is small and
therefore justifies the use of a linear filter. Thus the aided INS is a dead
reckoning system which uses redundant navigation information from other
sensors to compensate for its own error sources. The comparison of the
complementary filter with the Kalman filter is given in Brown (1973) and
Higgins (1975). It should be noted that the complementary filter can also
be used to obtain the dynamic damping feedback by complementing radio
navaid rate data with an inertial source such as the track angle error The
output of the complementary filter is multiplied by the appropriate gain
constant and inserted into the control law defined by Eq. (3). For example, a
velocity complementary filter output would be A j c .
The state vector configuration of the complementary filter is known in the
literature as a “multisensor navigation system” (Fried, 1974; Zimmerman,
1969). The concept is very general. Although the example in Fig. 36 used the
INS as the primary sensor, depending upon the application, radio navaid
can also serve as the primary sensor.

7. Concepts of PFE and CMN


In this section the concepts of path following error (PFE) and control
motion noise (CMN) are introduced together with the rationale for defining
the guidance errors in terms of these concepts. They are generalizations of the
traditional measures of error, namely, bias and noise. These concepts,

TRUE VALUES + INERTIAL ERRORS


BEST ESTIMATE
INS OF POSITION &
I VELOCITY

I\
AIDING
INERTIAL
SENSORS KALMAN
FILTER ERRORS

TRUEVALUES
tMEAS ERROR

FIG.36. State vector complementaryfilter configuration.


100 ROBERT J. KELLY AND DANNY R. CUSICK

originally defined in Kelly (1974), are the accuracy terminology used in


specifyingthe MLS angle and the DME/P signal in space. These concepts are
very general and apply to all navaids as well as MLS.
The MLS angular error or DME/P range error is the difference between
the airborne receiver processed sampled data output and the true position
angle at the sampling time. The MLS guidance signal is distorted by
ground/airborne equipment imperfections and by propagation-induced per-
turbations such as multipath and diffraction. To fully delineate the signal-in-
space quality, these perturbations are viewed in the frequency domain, that is,
as an error spectrum." This representation allows the interaction between
MLS errors and the AFCS to be understood in terms of the parameters which
define a successful landing. Associated with the signal error spectrum is the
idea of the AFCS frequency response. The frequency response of the aircraft
dynamics can be divided into three major spectral regions-a low-, middle-,
and high-frequency region.
Guidance error fluctuations in the low-frequency region fall within the
aircraft's guidance loop bandwidth and, since they can be tracked, cause
physical displacement of the aircraft. The effect of these lateral or vertical
displacements is measured in terms of unsuccessful landings. The middle
frequencies result in attitude changes (roll and pitch) and in control surface
activity whose motions are too rapid for the aircraft to follow. They also
include induced wheel (lateral channel) and column motions (longitudinal
channel). Control surface motions (ailerons and elevator) can lead to
component wear-out and fatigue, whereas the column and wheel motion
affect pilot acceptance factors. Figure 37 summarizes the azimuth control
activity criteria used by the airframe industry.
One measure of these pilot acceptance factors is the Cooper Rating
(Hazeltine, 1972). Although subjective and not universally accepted, it is the
only test currently in wide use for measuring these effects.The high-frequency
regime, which begins around 10 rad/s, does not affect the aircraft guidance
system or pilot and passenger comfort measures. Figure 37, which was taken
from Neal (1975),illustrates the lateral channel spectral response of a typical
transport aircraft. Note that the roll and aileron responses cannot be
approximated by a simple filter as can the guidance loop response (Fig. 34).
Based upon the above considerations, ICAO defined PFE and CMN
in the MLS angle and DME/P SARPs (ICAO, 1981a, 1985). The SARPs
definitions for DME/P are:
PFE-that portion of the guidance signal error which could cause aircraft
l o The error spectrum is defined as the Fourier transform of the flight test error trace. The
flight test error trace is the difference between the MLS measurement and the actual position of
the aircraft.
DISTANCE MEASURING EQUIPMENT IN AVIATION 101

NAVAIO OUTPUT OATA ZERO OROER


ANALOGUE
SENSOR OlGlTAL FILTER

BW-fJZ BW 'WF IL OELAV - :WJdbT,


OUTPUT

w3dB = GUIDANCE
LOOP BANOWIOTH

' W3db wFIL fJz fr

AZIMUTH CONTROL ACTIVITY CRITERIA

COOPER RATING: COOPER 11, PLEASANT TO FLY


LATERAL ACCELERATION: 0.04 0
ROLL ATTITUDE VARIATION: 2 1' +5 2'
CONTROL WHEEL VARIATION: i2' +5 So
AILERON DEFLECTION: 2 20
CONDITION: 15.OOO FT RUNWAY, AT THRESHOLD

0.01 1 0.10 0.8 1.0 2.0 10


SIGNAL E ~ O R
SPECTRUM
FREOUENCY IRADISECI __*

FIG.37. (a) Output data filter and data reconstruction unit. (b) Typical transport aircraft
power spectral response.
102 ROBERT J. KELLY AND DANNY R. CUSICK

displacement from the desired course and/or glide path. PFE frequency
regions are 0-1.5 rad/s for elevation and 0-0.5 rad/s for azimuth. For the
angle SARPS the PFE is partitioned into 2 additional components, path-
following noise (PFN) and bias. PFN is the PFE component with the bias
removed.
CMN-that portion of the guidance signal error which could affect
aircraft attitude angle and cause control surface, wheel, and column
motion during coupled flight, but which does not cause aircraft displacement
from the desired course and/or glide path. The frequency regions for CMN
are 0.5-10 rad/s for elevation and 0.3-10 rad/s for azimuth.
Clearly, an excellent guidance signal would result if all frequency
components above the path-following error regime were suppressed by a
smoothing filter placed at the output of the navaid receiver. However, this
cannot be accomplished without introducing a large phase lag in the higher
path-following frequency components with the resulting adverse effects on
AFCS stability discussed in Subsections E,3 and E,4. The MLS community
chose a 10-rad/s, single-pole filter that induces less than a 6" phase lag for a
1-rad/s path-following signal component. In other words, the navaid filter
should be 2 10 times the guidance loop bandwidth. The penality is, of course,
that the higher nonguidance (i.e., control motion) frequency components
become inputs to the AFCS:As shown in Fig. 37, the MLS and DME/P
SARPs permit all MLS angle receivers and DME/P interrogators to band
limit their output data to 2 10 rad/s using a single-pole filter or its equivalent.
Based on the 0.2 rad/s guidance loop bandwidth shown in Fig. 37, the MLS
data can be filtered (down to 2 rad/s) while still maintaining the I6" phase lag.
A measurement methodology is also defined to determine if the MLS angle
and DME/P equipment satisfy the PFE and CMN error specifications(Kelly,
1977; Redlien and Kelly 1981).
Using the definitions of PFE and CMN above, the question of obtaining
rate information from the radio navaid can be revisited. The control law in
Eq. (3) requires knowledge of the bank-angle transfer function, the pseudo-
derivative transfer function, and most importantly the power spectral den-
sity of the MLS azimuth signal before the unwanted control activity can be
estimated (see Fig. 37). An example of an MLS control law proposed for
autocoupled approaches in the terminal area (Feather, 1985) is
A4c = 0.026 Ay + 0.49 A j
Complementary filters using simple inertial sensors (i.e., non-INS) were
used to ensure that the anomalies in Ay and A j did not cause A4c to exceed the
2" limit.
MLS must be operational for several years before a statistical represen-
tation of the MLS signal can be defined. Only then will the required degree of
inertial aiding for each aircraft type be accurately known. Based upon flight
DISTANCE MEASURING EQUIPMENT IN AVIATION 103
test data collected during the MLS development program, there is reason to
believe that the MLS error signal at many airports may approach an
essentially flat power spectral density function. Multipath bursts, when they
exist, may be of short duration. If this is the case, then at the very least
complementary rate filter designs, when required, will be easier to implement
in MLS-equipped aircraft.
MLS control laws, as well as the tracking error and control activity
criteria, are currently under development not only for the autocoupled mode,
but also for flight director and manual-coupled flight. These development
programs are directed at approach and landing operations in the terminal
area. Critical issues are the following questions: What are the operational
benefits of a curved approach? Given that the autopilot error is less than 30 ft
at the Category I1 window in the presence of steady wind and wind turbulance,
is an autopilot error larger than 100 to 150 ft acceptable in the terminal area?
Similar control law design considerations are relevant to the enroute
application for the "cruise mode." One control law currently in use for
aircraft speeds of 500 knots at 30,000 ft is A4c = (0.0025 deg/ft)Ay +
(1.6deg/deg)A+,. Using 5 = l/fi and the equation in Fig. 34, the guidance
bandwidth for this control law is 0.04 rad/s. In this case, the position deviation
Ay can be derived from VOR/DME or DME/DME. The dynamic damping is
derived from the track angle deviation A+T as determined from an INS. The
roll anomaly for the above cruise mode control law will be well within 2".
In summary, navaid sensors (not only MLS) should satisfy the following
three criteria or inertial aiding may be required:
(1) PFE, when combined with either the flight technical error or the
autopilot error, should permit the aircraft tracking accuracy to be achieved.
(2) Filter bandwidth should not introduce more than 6"lag or more than
1 dB of gain at the guidance loop bandwidth.
(3) The CMN level must not include more than 2" rms unwanted
roll activity for autocoupled flight.
The MLS guidance signals satisfy the first two criteria. Only the third criterion
must be analyzed when developing MLS control laws for the terminal area
and the approach/landing phases of flight. If criterion 3 cannot be satisfied,
then a control law implementation having one of the forms illustrated in
Fig. 38 may be necessary.

8. Interaction of DMEIP Signal with AFCS


In the ICAO SARPs, the DME/P accuracy components are described
in terms of PFE and CMN. The purpose of this section is to provide the tech-
nical basis for specifying the range measurement in these terms rather than
the more conventional ones of bias and noise.
104 ROBERT J. KELLY AND DANNY R. CUSICK

(a) CONVENTIONAL

COMPLEMENTARY 2 K%zc
drp
9

(b) INERTIALLY DERIVED RATE COMPLEMENTARY FILTER

RADIO , AY
-
b
NAVAID
COMPLEMENTARV Ayc
FILTER
INERTIAL AV INERTIAL.
DATA
h i

(GI COMPLEMENTARY FILTER

FIG.38. Alternate control law implementations.

Depending upon the aircraft’sorientation with respect to the MLS ground


facility, azimuth, elevation, and DME signals can be inputs to any of the two
autopilot channels when properly converted to linear path deviations. The
precision DME accuracy standard must ensure that the DME/P has
approximately the same operational accuracy as the angle MLS in specific
portions of the approach path. For example, two-dimensional (constant
altitude) curved and segmented approaches may require that the DME/P
DISTANCE MEASURING EQUIPMENT IN AVIATION 105
accuracy be equivalent to azimuth guidance cross-course deviations. On the
other hand, for most straight-in approaches, range data will not be required in
either the lateral or vertical channels of the AFCS.
As mentioned earlier, CMN has a negative influenceupon pilot acceptance
factors. Resolution of this difficultyrequires the CMN to be at least limited
by the system design to a maximum of 0.1" everywhere in the MLS cover-
age sector while utilizing an airborne output data filter bandwidth which
limits the phase delay to 5" or 6" (see Section IV,D).
It is against this angular error requirement that the DME/P CMN
accuracy standard (when it is applicable) is derived. Since the range
information is already linear, it is specified to be equivalent to the angular
error of 0.1" at all points within the coverage volume. Future tests are re-
quired to determine whether 0.1" is an adequate CMN upper limit.
The situation is different for the PFE component. At the MLS entrance
gate, the PFE can be large because the aircraft separation standard is large
( rfi 2 nmi); however, as the aircraft proceeds toward runway threshold, the
guidance error must tighten up so that the aircraft can comfortably fly through
the 100-ft decision height window. Thus, unlike CMN, the PFE can degrade
with increasing distance from runway threshold by a 10:1 factor. Figure 39
graphically illustrates the differences between PFE, intended course, and
autopilot error.
It should be emphasized again that for DME/P, CMN is only a
consideration when the application requires vertical or lateral guidance using
the DME/P, e.g., curved approaches. It is not critical for example in
determining the 100-ft decision height.

MLS
INDICATED /
POSITION

'<
/
/
INTENDED COURSE
PATH
/ FOLLOWING ERROR

, AIRCRAFT ACTUAL
POSITION

AUTOPILOT ERROR
FIG.39. Definition of autopilot error and path following error.
106 ROBERT J. KELLY AND DANNY R. CUSICK

9. Summary
This section developed the background necessary to understand why an
aircraft radio navaid for use in the NAS should provide low-noise, high-
information-rate output data. Aircraft guidance and control in the enroute
navigation and the approach/landing application requires accurate position
and velocity data with small phase lag at several samples per second. If the
radio navaid is unable to achieve these position and velocity data require-
ments, then the radio navaid position fix data will require some form of inertial
aiding. The aiding may be as simple as gyro attitude measurements to obtain
dynamic damping or it may be as sophisticated as a complementary filter.
Also introduced in this section were the concepts of PFE and CMN. They
are new terms which form the basis of the MLS angle and the DME/P
accuracy standards.

F. DME, TACAN, and VOR Enroute Navigation


Accuracy Performance Comparisons

The accuracy performance of three position-fix systems, DME/DME,


VORTAC, and TACAN, are compared in this section. The results will show
that the DME/DME position-fix system is by far the most accurate method of
providing enroute guidance in the ATC system,particularly when operating in
a strong multipath environment.

I . General Error Analysis


An RNAV system using position-fix sensors has the following error
components:
(1) ground element instrumentation error,
(2) airborne element instrumentation error,
(3) site effects (multipath),
(4) RNAV instrumentation errors,
(5) flight technical error.
Two analyses are presented. First, position-fix accuracy obtainable along a
VORTAC radial is examined. Navigation along such radials represents the
standard mode of flight utilized in the past and will continue to be utilized for
some time to come. Second, since the use of RNAV is expected to grow rapidly
in the future, a general RNAV route will be used as a means to compare the
position-fix performance of four different configurations of sensors:
VOR/DME, DVOR/DME, TACAN, and DME/DME. In the discussions
DISTANCE MEASURING EQUIPMENT IN AVIATION 107
below it is assumed that the slant range (DME) has been converted into
ground range using the aircraft height measurement.
The ground instrumentation errors are defined by the FAA for VOR and
DME (FAA, 1982).The airborne sensor instrumentation errors are defined by
equipment manufacturers, who may elect to comply with the FAA’s TSO
(FAA, 1966) or the ARINC characteristics (ARINC, 1982), depending upon
the customer’s operational requirements. Potentially the largest radio navaid
error sources are multipath reflectionsin the vicinity of the ground station site.
Bearing determination systems are more sensitive to enroute or terminal area
site effects than a range determining system. This point will be justified and
explored further under Subsection F,2.
The RNAV instrumentation errors include computation and resolution
errors as well as the course selection error (CSE). The final error component is
the flight technical error and, as noted in Subsection C, it is the difference
between the intended flight path and the indicated flight path as displayed in a
manual-coupled AFCS. For autocoupled flight, the flight technical error is
called the autopilot error and for the purposes here is considered to be
negligible. As discussed in Subsection C,3 the error components are the mean-
square-error (MSE) values. Figure 40 functionally illustrates these errors.
Table V lists those error components as they apply to the four VORTAC
sensors: DME, TACAN, VOR, and DVOR. The analysis was based upon the
most elementary RNAV flight path, a VORTAC radial.
In general, VOR/DME RNAV performance accuracy is determined at a
waypoint or at any point along the RNAV route by using the concept of the

P
-L
t
I
I
- ALONG
TRACK

AIR 0.1 nm.

I
L -------------- 4
RNAV SYSTEM I N A IR C R A F l

FIG. 40. Two-dimensional RNAV error budget.


108 ROBERT J. KELLY AND DANNY R. CUSICK

TABLE V
ERRORBUDGETSFOR VORTAC NAVAIDS'
NOMINAL

Error component TACAN


(1) (95% prob.) bearing VOR DME/N DVOR

Airborne 0.11" 1" 0.1 nmi 1"


Ground 0.1 1" 1" 0.08 1"
Instrumentation only errors 0.16" 1.4" 0.13 nmi 1.4"
[RSS of (1) & (2)]
Linear instrumentation error at 0.14 nmi 1.75 nrni 0.13 nmi 1.75 nrni
50 nmi Cross Cross Along
Convert (3) track track track
Site error 1.91" 5.4" 0.04 nmi 0.53"
1 = - 10 dB for DME and TACAN
2 = - 20 dB for VOR and DVOR; p = multipath-to-direct signal ratio
Subtotal 1.92" 5.58' 0.14 nrni 1.5"
RSS (3), (5)
Subtotal linear 1.67 nrni 4.85" nmi 0.14 nmi 1.28 nmi
Convert (6) at 50 nmi
CSE 1.6" 1.6" 0 1.6"
Flight technical error 2.0" 2.0" 0 2.0
Total system error 3.2" 6.1" 0.14 nmi 2.97"
RSS (31, (51, (71, (8)
RNAV computation 0.5 nrni 0.5 nmi 0.5 nmi 0.5 nmi
Total linear system error at 50 nmi 2.7 nmi 5.3 nmi 0.52 nmi 2.6 nmi
Convert (9) Cross Cross Along Cross
track track track track

a Assuming air carrier quality airborne elements and a manual coupled VORTAC flight path
radial.

tangent point. It is the point from which a line perpendicular to the RNAV
route centerline passes through a specified VOR/DME location. The tangent
point distance is the distance from the VOR/DME to the tangent point, as
shown in Fig. 41. The along-track distance (ATD) is the straight-line distance
from the tangent point to the waypoint. The waypoint displacement area is
calculated using Fig. 41 and Eqs. (1) and (2) in Fig. 42. In Fig. 42, Do is the
tangent point distance.
The flight geometry used for the Table V comparison is an aircraft
positioned 50 nmi along a VORTAC radial. It is the same course range used by
the FAA (Del Balzo, 1970) to define the VOR and RNAV course widths. The
bearing angular errors and the range errors can be transformed into the cross-
track and along-track errors using Eqs. (1) and (2) in Fig. 42 with a = 0. Thus
ALONG TRACK DISTANCE (ATD)-O

a \

N
4

D = d and the tangent point distance Do = 0. For a VORTAC radial, the


along-track error is simply the DME range error, while the cross-track error is
the bearing error component of the VOR, DVOR or the TACAN.
The error component values listed in the table were determined as follows.
The RNAV computation errors, the CSE, and the flight technical errors are
the values used in RTCA (1982).The equipment errors are the nominal values
used in the air carrier class of equipment. Only the multipath site errors which
can comprise the largest error component require further discussion.
Specular multipath can be characterized by three parameters, the multi-
path amplitude to direct signal ratio p, the separation angle 4 between the
multipath and the desired signal, and their relative rf phase. Only p and the
separation angle will be considered in estimating the multipath error
component; the relative rf phase will be considered to be zero.
In order to proceed, it is necessary to determine the site multipath level p
for the different systems. Its value depends upon the geometry, dimensions,
and material of the reflectors in the vicinity of the ground station antenna.
It is assumed that only the rf frequency dependence of the lateral dimension
of the first Fresnel zone is significant in determining the relative difference
between the DME and VOR reflection coefficients (Kelly, 1976; Redlien
and Kelly, 1981). Assume for a given site that the VOR reflection coefficient
pVoR= -20 dB. Further assume that the reflection coefficient of the given
VOR reflector increases for DME signals to pDME = - 10 dB. The multipath
geometry is given in Fig. 43.The error analysis below derives the multipath
error for the DME, TACAN, VOR, and DVOR using - 10 and -20 dB for
pDME and pVOR, respectively. The results were used in Table V.
110 ROBERT J. KELLY A N D DANNY R. CUSICK

6,j= VOR SYSTEM ERROR

-
6~ = DME SYSTEM ERROR
d AIRCRAFT DISTANCE FROM TANGENT POINT
Do =TANGENT POINT DISTANCE

[&I = 0 6 2 = VOR ERROR VARIANCE

-
E

= DME ERROR VARIANCE


E 16~21 U R ~

(1) ~ X T K Z =I ~ R SIN%
Z + GCJ 0 2 C O S Z ~] 1’2

NOTE hRAND 60 ARE INDEPENDENT RANDOM VARIABLES


(2) “ A T K ~ = [ ~ CRO~S ~ Q+ 020g2 S I ~ 2 a ] 1 / 2
(3) 0 = [ Doz + d2 ] ”‘
(4) cos a = d

TOTAL SYSTEM ERRORS ARE


(51 u X T K Z (TOTAL) = o X T K Z + uRNAVZ + oFTEZ
(6) UATKz(TOTAL) = UATKZ+ (JRNAVZ

\
--
\
\
\

\
\
\ COURSE
\
\

VORTAC
STATION

FIG.42. Geometry of DMEjVOR error equation computation.

In summary, Table V states that the DME along-track errors for the
VORTAC geometry are negligible when compared with the bearing errors of
either the DVOR, VOR, or the TACAN. The TACAN accuracy performance
is comparable to the DVOR under the conditions assumed for the multipath.
Note that, because of the site error, the VOR would not satisfy the 4 nmi
enroute accuracy requirement noted in Subsection D. In this case, a DVOR
would be required for that site. As shown in the table its total system error is
only 2.6 nmi and is well within the 4 nmi error limit.
DISTANCE MEASURING EQUIPMENT IN AVIATION 111

ER = REFLECTED SIGNAL, ED= DIRECT SIGNAL


EC = COMBINED SIGNAL
p=ERIED
@ = SEPARATION ANGLE

NORTH
HANGER, HILL, OR OTHER REFLECTOR

20 DEGREE

FIG.43. Multipath geometry and error mechanism.


112 ROBERT J. KELLY A N D DANNY R. CUSICK

The RNAV will now use DME/DME, VOR/DME, or TACAN as input


sensors to estimate an arbitrary course instead of the VORTAC radial
discussed above. Also, the outputs of the RNAV are assumed to be coupled to
an autopilot and the flight path is preprogrammed.
With these assumptions, the model shown in Figure 40 is modified with
CSE E 0 and the flight technical error is negligible. The along-track and cross-
track errors can be calculated for any RNAV course AB using the equations
and geometry of Fig. 42 for a DME/VOR configuration or Fig. 44 for the
DME/DME configuration.
The DME/DME position-fix configuration is given in Fig. 44 assumes a
scanning DME interrogator. Since the scanning interrogator is common to
both the DME measurements, the errors of the DME/DME are correlated.
Most importantly, it is clear from Eq. (1)in Fig. 44 why the station selection
algorithms described in Subsection D,4 choose only those sites where 8 is more
than 30" but less than 150". In this range, the cross-track error degradation
will only be double the optimum value at 90".
The DME/DME position error equations used in Fig. 44 are derived in
Braff (1966). The geometry and equations in Figs. 42 and 44 will be used to
compare the RNAV performance when DME/DME, VOR/DME, TACAN,
and DVOR/DME are used as sensor inputs. Let the course line A B = 70 nmi,
b = 50 nmi, D, = 20 nmi, and D, = 15 nmi as defined in Fig. 44 for the
DME/DME calculation. Assume the same course line AB for the VORTAC
calculations in Fig. 42 and assume the VORTAC station location is identical
to the DMEl ground station. Figure 45 graphically displays the computer
simulation results for an aircraft traversing the RNAV course AB. The
simulation used the equations in Figs. 42 and 44 and the error components
were those used in Table V. As shown in the figure, the DME/DME was the
most accurate, and, as expected, the DVOR and the TACAN had equivalent
performance. Because of the assumed site error defined in Fig. 43, the
VOR/DME configuration was unacceptable because it exceeded the k4 nmi
enroute RNAV course width.
Figure 46 gives the position of the aircraft relative to the ground
transponder coordinates in terms of DME/DME range measurements. Note
that the ambiguity in they coordinate has to be resolved from either one more
DME station measurement or from some other reference. The equations are
derived in Porter et al. (1967).

2. Approximate VORTAC Multipath Error Relations


In this section, the approximate multipath error relations are derived for
the DME, TACAN, VOR, and DVOR navaids. The method of analysis will
treat both the DME range element and the VOR/TACAN bearing element
DISTANCE MEASURING EQUIPMENT IN AVIATION 113
= DME ERROR FOR STATION 1 A N D SCANNING DME INTERROGATOR
hR1
= DME ERROR FOR STATION 2 A N D SCANNING DME INTERROGATOR
,,6
d = AIRCRAFT DISTANCE ALONG RNAV COURSE AB

E [6R,2] = OR12 = DME 1 ERROR VARIANCE

E [ 6R22] = UR22 = DME 2 ERROR VARIANCE

R1 = DME 1 MEASUREMENT
R2 = DME 2 MEASUREMENT

p,, =
[&R1 &R2 1= CORRELATION BETWEEN ,6
, + ,,6
OR1 OR2

STATION 2
STATION 1

FIG.44. Geometry of DME/DME error equations computations.


114 ROBERT J. KELLY AND DANNY R. CUSICK
1

l5O0
0m0 7 14 21 28 35 42 49 56 63 70
DISTANCE ALONG COURSE AB (nm)

---- ALONG COURSE


CROSS COURSE

FIG.45. Position fix accuracy performance of DME, VOR, and TACAN.

using the same mathematical formulation. Let the desired DME pulse eD(t) be
represented by a half-sine wave of period z. Since the DME pulse width
between the half-voltage points of the leading and trailing pulse edges is 3.5 ps,
then z = 10.5 ps. The DME pulse is therefore
eD(t) = EDsin(271ft) for 0 I t I 2/2, where T = l/f
(12)
eD(t)= 0 for 212 < t I z
Let the multipath disturbance 6n be a delayed version of the direct signal with
separation angle 4. The disturbance is given by
6n = ERsin(2nfi+ 4) (13)
ED and E R are the peak amplitudes of the direct signal and the multipath
signal, respectively.
For DME/N the leading edge of the pulse is detected at the half-voltage
point, which corresponds to t = 2/6. The leading edge time of arrival error 6t
for small disturbances 6n is

where the derivative is evaluated at the half-voltage point to = ~ / 6 The


.
DISTANCE MEASURING EQUIPMENT IN AVIATION 115
POSITION F I X COORDINATES ( x , y I USING MEASURED
R 1 A N D R 2 A N D G R O U N D STATION COORDINATES

2 2
x = R1 -R2
4x,

--
-1 X1
b
1

+ x AXIS
FIG.46. Geometryfor determiningposition fix of DME/DME equippedaircraft relative to
ground transponder coordinates.

result is

where p = E,/E,. Using Eq. (15) the maximum disturbance 6tM occurs at
4 = x/6:

The distance error 6RDM,in feet is


1 1
6~,,, =- x 1096t, =-
2 2 4 p T log ft
The estimated peak distance error in the presence of -10 dB multipath
116 ROBERT J. KELLY AND DANNY R. CUSICK

(p = 0.32) using Eq. (17) is 103 ft. The exact relation is given in Sec-
tion IV,B,4.
With suitable modifications,Eq. (13) can be used to derive the phase error
for the VOR system. The phase error is evaluated at the zero crossing, namely,
to = 0. Using Eq. (13) the zero crossing error 6t, is
6to = (p sin 9)/2nf (18)
To convert to bearing error in degrees, multiply Eq. (18) by 57.3 (27cf).
The bearing error in electrical degrees is
88VOR = ( 5 7 . 3 ) ~sin 9 (19)
where 360 electrical degrees equals 360 spatial degrees.
For the geometry of Fig. 43 and - 20 dB multipath, the VOR bearing error
6OvO, = 5.4". The exact equations are given in Flint and Hollm, 1965.
The reason bearing measurement systems have less accuracy than range
measurement systems can be explained as follows: Bearing errors are pro-
portional to the effective beam width of the ground station radiation pat-
tern. The VOR rotating cardioid antenna pattern has a 3 dB beam width of
about 133". DME errors in general do not increase with range R; however, the
bearing error 68 when expressed in a linear measure 6Y does increase with
range by definition (6Y = R68). In order to offset this linear increase, the
bearing ground station effectivebeam width must be drastically reduced so
that 68 is small to begin with.
This cannot be accomplished in any practical way. However, as stated in
Subsection D, the multipath performance of VOR can be improved using
techniques which in effect reduce the antenna beam width. One approach
which culminated in the DVOR effectivelyincreased its antenna aperture by a
factor of 10 (Steiner, 1960). Thus, its effective beam width was reduced about
10:1, with an equivalent reduction in multipath error. For the multipath
configuration shown in Fig. 43, the DVOR would have about a 0.53" error,
thus yielding a 10: 1 improvement over the VOR implementation.
To understand the error mechanism of TACAN, it is necessary to revisit
the two-scale concept. If the TACAN were implemented using only the 15 Hz
cardioid, then the error would be identical to the VOR. However, the error is
reduced by a factor of 9 when the nine-lobe antenna pattern which generates
the 135 Hz (ninth-harmonic) component is considered. The key point is that
both the 15 and 135 Hz reflected signals are delayed by the same amount as
they traverse path Q + R, as shown in Fig. 43. The processing of the received
data consists of determining the phase of the first harmonic (15 Hz
component) in the received pattern relative to the first-harmonic time
reference markers and the phase angle of the ninth harmonic (135 Hz
component) in the received pattern relative to 9th harmonic reference markers.
DISTANCE MEASURING EQUIPMENT IN AVIATION 117
The function of the first harmonic is to determine which 40" sector the
measured phase of the ninth harmonic falls. Thus, the phase measurement on
the 15 Hz channel component constitutes a coarse measurement while the
135 Hz channel component constitutes the fine measurement.
To determine the multipath improvement which can be expected, Eq. (19)
will be modified to reflect the fact that 360 electrical degrees equals 40 spatial
degrees. Thus, the TACAN bearing ~ T A , is given by
~ 68vo,/9
6 6 ,= ~ N ~ 4
6 . 4 sin (20)
For the configuration described in Fig. 43, the TACAN bearing error is
66,,, = 1.92".This value was used in Table IV. The exact equation for the
TACAN bearing error is given in Sandretto (1958).Introducing the multilobe
radiation pattern can be viewed as reducing the effective beam width of the
TACAN source from 132"to about 15".
In summary, the site error is reduced by 6 of that for the VOR for equal
reflection coefficients. An important consequence of the TACAN multilobe
antenna pattern is that both the site and instrumentation errors are reduced.
As stated in Subsection D, the DVOR reduces the site errors, but not the
instrumentation errors. As a consequence, the instrumentation errors for
TACAN in Table V are reduced by a factor of 9 relative to the VOR errors.
TACAN ground facilities have to be sited very carefully because a multipath
level of, for example, - 6 dB would induce an error of 30" and thus could cause
a large ambiguity by jumping to the wrong 40"sector. This is one of the known
drawbacks of systems like TACAN which use a two-scale measurement
system.
The DVOR and TACAN bearing errors would have equivalent DME
linear errors at distances of about 3100 and 11,000 ft, respectively. On the
other hand, the new MLS 1" azimuth antenna has an equivalent DME linear
error at 12.9 nmi. The MLS ICAO standard requires 20-ft accuracy at a
distance of 15,000 ft. Interestingly enough, the same formulation can be used
to derive the MLS angle error equation (Kelly, 1976; Frisbie, 1976).
Finally, both Eqs. (14) and (18)can be used to approximate the rms receiver
noise performance or diffuse multipath if 6n represents the standard deviation
of the random error source. In this case p = l/(s/6n), the inverse of the voltage
signal-to-noiseratio or signal-to-interference voltage ratio.
In all of the formulation above the reference signal was not included in the
analysis. Such an analysis is not required because for VOR the reference signal
is embedded in an FM format which is not sensitive to amplitude disturbances.
In the case of TACAN, the reference signal is derived from a series of reference
pulses whose maximum timing errors, as calculated using Eq. (16), is less than
103 ft for the - 10 dB multipath level. The corresponding electrical phase
error would be negligible. Note that, depending upon the reflector distance
118 ROBERT J. KELLY AND DANNY R. CUSICK

used in Fig. 43, the DME multipath induced error may be zero if the pulse time
is more than 1 ps. In this case, there is no error because the multipath occurs
after the pulse leading edge measurement point.

3. Additional Accuracy Considerations


Additional comments will now be made with respect to the DME/DME
position-fix system. The error described in Fig. 44 for the DME/DME can be
reduced if more DME ground facilities are tracked. Test results in Latham
(1974) indicated that the use of regression analysis on the data taken with at
least 8 to 9 ground stations in a flight run resulted in a DME/DME position
error of 100 ft CEP. Latham (1974) collected data from 17 different DME/N
ground stations in southern New England. The measured bias error from these
stations had an ensemble standard deviation of 400 ft. The ensemble average
was 200ft and, because the ensemble average of the ground station was
assumed to be zero (Subsection C), this average error was attributed to the
airborne interrogator and the distribution geometry of the ground stations.
The standard deviation of the range noise from each station was measured to
be about 200ft. The RSS error for the 17 ground stations investigated,
including the airborne bias, was
+
RSS = -J(400)2+ (200)2 (200)2 = 490 ft
which reinforces the earlier statement that for many ATC navaids it is the bias
error which predominates. If the distributions in the above study are assumed
to be normally distributed, then the 95% probability of the DME measured
system error is about 2 x 490 = 980 ft. This is consistent with the new
DME/N error standard of 0.2 nmi (1215 ft). A very general treatment of the
accuracies achievable from multiple range measurements is given in Grogin-
sky (1959) and Wax (1981).

111. THEDME/N SYSTEM

Section I1 of this article has presented the origins of the DME concept
together with the evolving role of the DME within the ATC system. In
addition, the characteristics of the information provided by any navigation
sensor (accuracy-noise and bias, data rate, and bandwidth) for use by
manual and automatic flight control systems were detailed and justified.
Section 11,therefore, serves as the foundation for the discussions that follow by
not only justifying the importance of DME within the ATC system but by
justifying the need for the design constraints imposed on the DME as well. In
this section the focus is on details of the DME/N system, which provides the
DISTANCE MEASURING EQUIPMENT IN AVIATION 119
reader with a knowledge of how the system is put together and why it is
uniquely qualified for its role in the ATC system.
Today’s DME system represents a refinement of the beacon concept
introduced in Section II,B. It is an internationally defined method of
determining aircraft slant range to known ground locations by measuring the
elapsed round-trip time between the transmission of an interrogation by an
airborne interrogator and the receipt of a reply to that interrogation from a
ground-based transponder. The characteristics of the DME system are
defined by the International Civil Aviations Organization in the form of
Standards and Recommended Practices (ICAO, 1972b). The scope, structure,
and evolution of the DME SARPs is the subject of Subsection A. In
Subsection B, an overview of the DME/N system and its major components is
presented. The emphasis is both on how the system is put together and on the
functional elements that comprise the ground and airborne equipment.
In Subsection C, the emphasis shifts to the details of the DME/N signal-
in-space. Specifically addressed are the quantitative aspects of accuracy,
coverage, the channel plan, and ground system capacity. The goal of the
section is not only to detail the basic system characteristics but to stress the
interrelationships between each.

A . DME Standards and Recommended Practices

In December, 1982, the ninth meeting of the All Weather Operation’s


Panel (AWOP) of the International Civil Aviation Organization completed
technical standards for the precision distance measuring equipment, DME/P.
Also approved were the revision and updating of the conventional DME
(DME/N) technical standards (ICAO, 1972b). After minor review by AWOP-
10 in September, 1984, the technical standards were adopted by the ICAO
Air Navigation Council on December 6, 1984, and became applicable as
amendments to Standards and Recommended Practices of the ICAO Annex
10 paragraph 3.5 on November 21,1985 (ICAO, 1985). From the first draft of
the DME SARPs put forth in 1950 to the 1985 version, this document has
evolved in its level of specification reflecting increased operational need (e.g.,
more channels, improved accuracy) as well as the available technological
capability. This 1985 revision is the latest improvement to a system whose role
in the ATC system has been expanding for the last 40 years.
The present document defines three types of DME ground facilities:
DME/N (narrow), DME/W (wide), and DME/P (precision). Each of these
systems operate in the L band of the frequency spectrum between 960 and
1215 MHz. DME/N is the primary ranging system used in aviation today. In
conjunction with VOR or TACAN bearing or with other DMEs, it is used to
120 ROBERT J. KELLY AND DANNY R. CUSICK

obtain position fixes for enroute and terminal area navigation. DME/W,
operationally equivalent to DME/N, has a less restrictivespectrum constraint
on the transmitted waveform. This rarely used system, to be phased out by
January, 1987, was intended to serve as an inexpensive means of providing
DME service in isolated regions where frequency protection of the service
volume is not a concern. DME/P serves as the MLS range element and
provides precision range information (accuracy on the order of 100 ft) having
characteristics suitable for approach and landing operations. The DME/P
standards represent the newest addition to the ICAO SARPS.
The purpose of the SARPs is threefold. First, it defines the required
minimum performance of the DME system in terms of accuracy, coverage,
and capacity. These specifications define a system having accuracy and
coverage capable of satisfying enroute and terminal area navigation needs
with the capability of a single facility to serve 100 aircraft simultaneously.
Second, the SARPs define the system characteristics necessary to ensure
compatibility among ground and airborne equipments of different manu-
facture. These characteristics include:
(1) Signal format: Characteristics of the signal format are specified that
include the transmitted pulse shape, the zero mile delay, and the calibration
technique.
(2) Ground facility signal power-in-space/receiver sensitivity: The
SARPs specify the minimum power densities to be provided by the ground
facility at the extremes of its defined coverage volume. The size and shape of
the coverage volume are determined by the governing authority based on
operational need. Ground facility receiver performance is based on received
power density at the antenna. Interrogator radiated power requirements and
receiver performance are required to be consistent with the ground system
specifications. Note that the SARPs avoid specifying implementation details
such as the transmitter power, cable losses, antenna characteristics, and
receiver performance, leaving the details of how the requirements are satisfied
to the equipment designer.
(3) Signal spectrum, stability, and rejection: The SARPs specify radiated
signal spectrum characteristics, carrier frequency stability, and off-channel
rejection characteristics.These characteristics provide a basis for the develop-
ment ground facility channel assignment criteria.
(4) Channel plan: The channel plan defines interrogator/reply frequency
pairs and the associated pulse-pair spacings or codes. In addition, the plan
explicitly provides for the “hard pairing” of DME channels with other
navigation aids, namely VOR, ILS, and MLS.
Finally, the SARPs serve as the basis for the development of ground
equipment procurement specifications, frequency planning, and field deploy-
DISTANCE MEASURING EQUIPMENT IN AVIATION 121
ment by ICAO member countries. It serves as the basis of airborne equipment
specifications as well. It must be emphasized that the SARPs are not
equipment specifications. Rather, they are an internationally agreed docu-
ment that defines the system characteristics to a level that ensures com-
patibility and performance among units of differentmanufacture and national
origin.
Traditionally, the emphasis of the SARPs has been on standards
pertaining to system performance and the ground equipment with airborne
specificationsplaying a lesser role. It is the ground transponder that can better
bear the burden of increased size, weight, power consumption, and cost
incurred in adhering to strict performance standards. Specifications pertaining
to the interrogator are limited to those absolutely necessary to ensure
performance and compatibility, thereby reflecting the primary goal of light-
weight, small-size, low-power, and low-cost airborne equipment. Detailed
airborne characteristics have been and will most likely remain in the domain
of the various user groups such as ARINC, RTCA, and EUROCAE.
Some countries incorporate the ICAO SARPs DME specification into
their national aviation standards in a way that provides equal or more
stringent performance than the SARPs. In the US., the DME/N character-
istics have been included in the VOR/DME National Standard (FAA, 1982),
while the corresponding ground equipment specification, for the FAA's
second-generation VORTAC, is found in FAA (1978). For DME/P, the
national standard is defined in FAA (1983b) while its equipment specification
is detailed in FAA (1983~).
In the US., airborne equipment specifications have been defined by three
different organizations, each addressing different, but perhaps overlapping,
user communities. The FAA requirements for DME/N performance are in the
form of a Technical Service Order (TSO) (FAA, 1966). The DME/P TSO has
not been developed at this time. Although compliance with the TSO is not
mandatory for DME equipment utilized in US. airspace, a user can more
easily obtain the required aircraft certification for operations in conditions
where a DME is required (e.g., above 18,000 ft under IFR conditions) if TSO'd
equipment is utilized.
The Radio Technical Commission for Aeronautics publishes Minimum
Operational Performance Specifications (MOPS) pertaining to both DME/N
and DME/P interrogators (RTCA, 1978, 1985b). Usually, the MOPS forms
the basis of the FAA's TSO requirements. RTCA (1985b) reflects the latest
improvements to the SARPs that apply to DME/N and represents the first
detailed specification of the new DME/P interrogator.
Airborne equipment for use by the airlines typically conforms to specifica-
tion that are more stringent than those defined by the RTCA MOPS. Such
specifications are developed and published by Aeronautical Radio, Inc.
122 ROBERT J. KELLY A N D DANNY R. CUSICK

(ARINC) and are called ARINC Characteristics. ARINC (1982)defines airline


requirements for DME/N, while the DME/P requirements remain to be
defined.
In Europe, the European Organization for Civil Aviation Electronics
(EUROCAE) establishes both ground and airborne equipment specifications
(EUROCAE, 1980). As in the U.S., EUROCAE is currently developing
DME/P equipment specifications.
The present ICAO SARPs begin with a system description which defines
not only the function of the DME system but its relationship to other systems,
particularly VOR, ILS, and MLS. Next, system characteristics concerning
coverage, accuracy, channel plan, capacity, and facility identification are
presented. Detailed characteristics of the ground element (transponder) and
the airborne element (interrogator) are presented in seperate sections.
Guidance material is provided at the end of the SARPs “white pages.” These
“green pages” are not specifications but rather recommendations concerning
system details such as channel assignment criteria and example ground and
airborne equipment power budgets.

1 . Evolution of the DME Standards


The first official notice by ICAO concerning DME was a recommendation
by its Special Radio Technical Division in late 1946 for the development of
distance measuring equipment (ICAO, 1947). In January, 1949, ICAO
recommended in Section 2 of “Aid to Final and Approach and Landing” that,
along with requirements for inner, middle, and outer markers, a U H F distance
measuring equipment be added to the ILS at the earliest possible date. At the
same time, it required that a DME become an integral part of the ILS not later
than January 1, 1954 (ICAO, 1949). In the same year, ICAO recommended
that the DME become a basic component of the VOR at the earliest practical
date.
In May, 1950, the first draft specification for DME was published in
Attachment B of Annex 10 of the SARPs for Aeronautical Telecommunica-
tions. As mentioned previously in Section II,B of this article, the channel plan
was essentially that proposed by RTCA SC-40, having 10 frequencies and 10
codes for a total of 100 channels (Table VI). The ground transponder
transmitter frequency stability was limited to & 200 kHz, implying that the
oscillators need not be crystal controlled. The radiated spectrum constraint, so
important in DME today, was to be determined at a later date.
Table VII summarizes the principal SARPs requirements at various
critical times during its evolution and includes the upgraded DME/N and new
DME/P parameters added with the 1985 revision. Figure 47 shows the
DISTANCE MEASURING EQUIPMENT IN AVIATION 123
TABLE VI
THEDME CHANNEL PLANAS PROPOSEDI N THE
FIRSTDRAFTDME SPECIFICATTONINCLUDEDI N THE
SARPs OF MAY,1950'

Pulse Spacings
Mode Interrogation (ps) Reply (PSI

A 14 17
B 21 70
C 28 63
D 35 56
E 42 49
F 49 42
G 56 35
H 63 28
I 70 21
J 77 14
Interrogation/Reply Frequency Pairs
Interrogation Band Reply Band
(960-986 MHz) (1188.5-1215 MHz)

Channel Frequency Channel Frequency

1 963.5 00 1188.5
2 966.0 10 1191.0
3 968.5 20 1193.5
4 971.0 30 1196.0
5 973.5 40 1198.5
6 976.0 50 1201.0
I 978.5 60 1203.5
8 981.0 70 1206.0
9 983.5 80 1208.5
10 986.0 90 1211.0

'The plan utilized 10 interrogation/reply fre-


quency pairs and 10 pulse codes for a total of 100
channels.

evolution of the pulse-shape specification over the same time period.


Compliance with the upgraded DME/N standards, identified in the SARPs by
a dagger, is required after January 1, 1989. Attention should be given to the
continual tightening of the specifications with each revision, as well as the
increased requirements imposed on the airborne equipment with the latest
revision. These changes reflect both the growing need for enhanced perfor-
mance from the DME as well as the availability of improved technology.
TABLE VII

Specification 1950 1952 1960 1972 1985


(Draft) DME/N DME/P

System level
Coverage 200 nrni enroute REC 200 nrni REC 200 nmi REC 200 nrni 200 nmi; same as 22 nmi and at least that
(enroute) 360" up (enroute) 360" up (enroute) 360" up equip. associated of MLS azimuth
to 60,000 ft to 75,000 ft to 75,000 ft with VOR, ILS coverage
System accuracy None REC (0.5 nrni REC (0.5 nrni REC (0.5 nrni k0.2 nmi; 95% prob Standard 1-100 ft
or 3%) or 3%) or 3%) PFE, 60 ft CMN
Standard 11-40 ft
PFE, 40 ft CMN
Channel plan 10 freq 10 freq X - 126 X-Y 252 x,y, w, x, Y,w,x-200
F
10 codes 10 codes (352 channels) channels
*
h)
Interrogation PRF 30 pp/s 30 PP/S 30 PP/S FA Mode: 40 pp/s;
IA mode: 16 pp/s
Aircraft capacity None 50 100 100 100 or peak traffic 100 or peak traffic
Transponder
Antenna 6 dB gain 360"; 4 dB gain 360"; 4 dB gain 360"; 4 dB gain None None
+fdipole +fdipole + f dipole ++dipole
gain 6" vert. beam 6" vert. beam 6" vert. beam
Transmitter
freq. stability
k 200 kHz 200 kHz *0.002% * 0.002% +0.002% Same as DME/N

Pulse rise time None <0.3 ps 2.5 k 0.5 ps NOM 2.5 ps < 3 ps < 1.6 p, 200-300 ns
< 3 ps partial rise time
Pulse width None f0.2 2.5 + 0.2 ps 3.5 f 0.5 ps 3.5 t 0.5 /.is 3.5 0.5 ps Same as DME/N
Pulse fall time None <0.5 ps 2.5 f 0.5 ps NOM 2.5 ps 2.5 ps NOM Same as DME/N
< 3.5 ps < 3.5 ps
Pulse spacing f 0.5 p s f 0.5 p s f0.25 p s f0.5 ps f 0.5 p s Same as DME/N
(REC f 0.10)
Effective radiated > 2 kW >8 kW > 8 kW -83 dB W/m2 -89 dB W/m2 - 89 dB W/m2 7 nmi
power (under all coverage limit
conditions) -75 dB W/m2
<7 m i
- 70 dB W/mz MLS
app. ref. datum
-79 dB W/mZ 8 ft
above runway
Transmission rate None None None REC REC Same as DME/N
(REC 2700 pps) (REC 2700 pps) 2700 pps 2700 pps
Receiver freq. k 100 kHz f 100 kHz f 60 kHz +60 kHz f 0.002% Same as DME/N
stability
Transponder -82 dBm at -90 dBm at -95 dB m at -95 dB m at - 103 dB W/mz IA Mode:
N
sensitivity receiver receiver receiver receiver -86 dB W/m2
FA Mode:
- 75 dB W/m2

Dynamic range None None None None -22 dB W/m2 Same as DME/N
Squitter rate 300 None > 700 pps >700 pps = 700 pps < 1200 pps
Frequency 50 dB at 50'dB at 80 dB; 80 dB; 80 dB; Same as DME/N
rejection 2 MHz away 10 MHz away >900 kHz >900 kHz >900 kHz
Recovery time None 8 p s after 8 p s after 8 p s after 8 p s after Same as DME/N
60 dB pulse 60 dB pulse 60 dB pulse 60 dB pulse
Decoder rejection None None None None -75 dB at f 2 p s Same as DME/N
Decoder accept. <0.5 p s None None None f 1 p s (1 dB degrad.) Same as DME/N
Time delay None f 500 ft k 500 ft f 500 ft f 500 ft Standard 1-33 ft
accuracy PFE, 26 ft CMN
Standard 11-16 ft
PFE, 16 f: CMN
TABLE VII (continued)

Specification 1950 1952 I960 1972 1985


(Draft) DME,” DMEjP
Time delay None 115 ps 50 ps 50 p s X channel Depends on channel Same as DME/N
56 ps Y channel code
Time delay None 2nd pulse at 50% 2nd pulse at 50% 2nd pulse at 50% pt 1st pulse at 50% pt 1st pulse at virtual
measurement origin
Reply efficiency 67% 67% 70% 70% 70% FA Mode: 80%
IA Mode: 70%
Dead time 133 ps 10 ps 60 ps NOM 60 ps NOM 60 p s NOM Same as DME/N
FA Mode has priority
Spectrum None - 18 dB at 1 MHz -60 dB at 800 kHz 200 MW at 800 kHz 200 MW at 800 kHz Same as DME/N
away away in 500 kHz away in 500 kHz away in 500 kHz
- 3 0 dB at 10 MHz BW BW BW
away -65 dB at 2 MHz 2 MW at 2 MHz 2 MW at 2 MHz
away in 500 kHz away in 500 kHz away in 500 kHz
BW BW BW
Monitor None Time delay Time delay Time delay Time delay Time delay (PFE)
XT power (REC) XT power (REC) XT power (REC) XT power (REC) XT power
RX sensitivity (REC) RX sensitivity (REC) RX sensitivity (REC) RX sensitivity (REC) RX sensitivity
Pulse spacing (REC) Pulse spacing (REC) Pulse spacing (REC) Pulse spacing (REC) Pulse spacing
Frequency (REC)
Interrogator
Transmitter f400 kHz & 400 kHz +400 kHz 100 kHz +
- 100 kHz Same as DME/N
stability
Pulse rise time None <0.3 ps 2.5 f 0.5 ps 2.5 ps NOM < 3 ps < 1.6 p s
( < 3 PS) 200-300 ns partial
rise time
Pulse width None k0.2 ps 2.5 5 0.2 ps 3.5 f 0.5 ps 3.5 f 0.5 ps 3.5 0.5 ps Same as DME/N
Pulse fall time None <0.5 ps 2.5 f 0.5 ps 2.5 p s NOM 2.5 ps NOM Same as DME/N
(< 3.5 ps) ( < 3.5 ps)
Pulse spacing k1 p s +1 ps
- T0.5p s k0.5 p s T0.5p s Same as DME/N
(REC 0.25 p s )
Spectrum None - 14 dB at 1 MHz 90% in 0.5 MHz 90% in 0.5 MHz 90% in 0.5 MHz Same as DME/N
away BW BW BW
- 30 dB at 10 MHz
away
Jitter interrogation Yes Yes Yes Yes Yes Yes
Effective radiated None None None None None None
power
Time delay None 2nd pulse at 50% pt 2nd pulse at 50% pt 2nd pulse at 50% pt 1st pulse at 50% pt 1st pulse at virtual
measurement origin
Time delay 115ps 115 p s 50 p s 50 p Depends on channel Same as DME/N
code
Time delay None None None None 0.17 nmi (REC) IA Mode: 100 ft
5
L

accuracy PFE, 50 ft CMN


FA Mode: 50 ft PFE,
33 ft CMN,
Standard I
23 ft PFE, 23 ft CMN,
Standard I1
Receiver None None None None - 89 dB W/m2 Consistent with uplink
sensitivity specifications
Dynamic range None None None None - 18 dB W/m2 Same as DME/N
Receiver None None None None -42 dB at 900 kHz Same as DME/N
selectivity away (REC)
Decoder rejection None None None None -42 dB at T 2 p s Same as DME/N
Decoder accept. <1ps None None None None None

Abbreviations: PROB, probability; REC, recommended specifications; BW, bandwidth; NOM, nominal; XT, transmitter; RX receiver.
128 ROBERT J. KELLY A N D DANNY R. CUSICK

w 1.OA w 1.OA
5 0.9A 5 0.9A
f k
:
d
L:0.5A $ 0.5A
m m
a
!i 5
9 0.1A 9 0.1A \
_I
TIME
<2.5m

1952 PULSE SHAPE I S ESSENTIALLY 1960 PULSE SHAPE IS ESSENTIALLY


RECTANGULAR WITH FAST RISE GAUSSIAN WITH AN ABSOLUTE
TIME AND WEAK RELATIVE ADJACENT CHANNEL RADIATED
ADJACENT CHANN€c RAQCATED POWER LIMITATION.
POWER LIMITATION.

1
VIRTUAL ORIGIN PARTIAL
T*

RISE TIME
200-300nr
1985 DME/N PULSE SHAPE AND SPECTRUM REOUIREMENT REMAIN UNCHANGED.
PARTIAL RISE TIME AND LINEARITY OF LEADING EDGE ARE SPECIFIED FOR
THE D M E P APPLICATION. USE OF A WIDE SPECTRUM PULSE NECESSITATES A
REDUCTION IN RADIATED POWER TO MEET THE ABSOLUTE ADJACENT
CHANNEL LIMITATION.

FIG.47. Evolution of the D M E transponder transmitted pulse shape.

B. A Functional Description of the DMEIN System

Today’s DME system is based on the beacon concept introduced in


Section II,B and is composed of the following components:
(1) Interrogator encoder and transmitter
(2) Transponder receiver and decoder
(3) Transponder encoder and transmitter
(4) Interrogator receiver and decoder
( 5 ) Predetermined zero mile delay
(6) Common pulse arrival time measurement technique
(7) Interrogator range tracking and display unit.
DISTANCE MEASURING EQUIPMENT IN AVIATION 129
Recall that the use of the ground based beacon has the following obvious
advantages:
(1) In traditional radar, as the range (R) increases, the power of the echo
decreases by a factor of l/R4. Since the ground beacon serves as a repeater, the
power of the transponder reply only decreases by a factor of 1/R2.
Consequently, the radiated power requirements in DME are substantially less
than those necessary in traditional radar ranging applications.
(2) Unambiguous range information is provided to a known location on
the ground which can have physical significance such as an airport, the stop
end of a runway, or a waypoint in the VICTOR airway or JET route system.
(3) The interrogation and response frequencies are different. Hence, a
beacon receiver will not be confused by its own transmissions.
(4) The use of an omnidirectional antenna at the ground beacon provides
distance information simultaneously to a multiplicity of aircraft.
All of the above aspects will become evident in the discussions that follow.
In this section, a qualitative description of DME/N operation, components,
and system considerations is provided. Detailed discussions are deferred to
Subsection C. In Subsection B,1, the operation of the DME/N system along
with a description of the major functional components that comprise the
ground and airborne equipment is presented. In Subsection B,2, those aspects
of the DME that make it a viable navigation aid are described with the major
goal of illustrating the interrelationship of the major system considerations. In
Subsection B,3, the problem of automatically acquiring and tracking the
desired beacon replies is introduced. Finally, the general trends apparent in
DME designs today are reviewed in Subsection B,4.
I . Operation and Components of the DME/N
The operation of the DME/N system can be concisely summarized as
follows. The airborne unit (interrogator) interrogates a single ground facility
that transmits a reply following a known calibrated fixed delay (the zero mile
delay). The airborne unit then computes the slant range to that facility by
measuring the elapsed time to the receipt of the transponder reply. The
measured range is then provided to the pilot and other aircraft systems as
required.
Each interrogation consists of a pulse pair where the spacing of the pulses
(or code) together with the rf frequency define the operating channel thereby
allowing the interrogations to be addressed to a specific ground facility.
Similarly, the transponder reply consists of a pulse pair having a code and rf
frequency corresponding to the channel in use thereby allowing the airborne
unit to distinguish desired ground facility transmissions from those of other
facilities within line of sight of the interrogator.
130 ROBERT J. KELLY AND DANNY R. CUSICK

The DME interrogator and transponder are similar in that both must
identify valid DME signals using the three discriminates of pulse width,
frequency, and code. In DME, the rf frequency and pulse-pair spacing or code
of the DME signal (interrogation and reply) designate the operating channel.
The width of the DME pulse is used to discriminate a valid pulse from those
due to noise which are of short duration.
The interrogator and transponder differ in function. An interrogator must
accurately measure the time delay between the transmission of the inter-
rogation and receipt of the reply. It does so by starting a clock coincident with
the transmission of each interrogation and stopping the clock upon reception
of the reply. On the other hand, the transponder must accurately generate a
fixed time delay between receipt of the interrogation and the transmission of a
reply. This is done by starting the time-delay clock upon receipt of an
interrogation, and after a fixed time transmitting a reply such that the round-
trip propagation delay is accurately extended by a fixed amount.
A ground transponder will “hear” many interrogations; those from aircraft
utilizing the transponder and from other aircraft within line of sight that are
utilizing other DME beacons. To minimize the load on the beacon and to
ensure unambiguous range information to the user, the transponder must only
respond to those interrogations intended for it-those having rf frequency
and pulse code corresponding to the assigned channel.
Although a reply to an interrogation will only be transmitted by the
desired ground facility, the airborne unit will “hear” the desired ground station
and others that are within its line of sight. Consequently, the interrogator must
sort out the reply to its own interrogation from the thousands of pulse pairs
that it may actually receive. It does so by identifying those beacon trans-
missions that are consistently synchronous with the interrogations. Those
pulse pairs which do not have the proper rf frequency and pulse codes are
eliminated from consideration as valid replies by the interrogator signal
processor.
In addition to reply pulse pairs, the transponder radiates squitter and a
facility identification signal. Both consist of pulse pairs that are identical to
those transmitted in response to interrogations. Squitter consists of random
transmissions of on-channel pulse pairs radiated at a rate of 700 to 3600 pulse
pairs per second. The intertransmission times are randomized to prevent
synchronization with replies to interrogations. Squitter is utilized by many
interrogators to determine appropriate automatic gain control (AGC) settings
and, as noted in Section II,D, squitter serves as the source of the TACAN
bearing information which is conveyed by modulating the amplitude of the
beacon transmissions.
When using any navigation facility, it is important that the pilot know with
certainty the source of the displayed navigation information. DME, VOR,
DISTANCE MEASURING EQUIPMENT IN AVIATION 131
ILS, and even MLS ground facilities identify themselves by transmitting a
3- or 4-character identifier in Morse code that corresponds to the facility
identifier shown on air navigation charts. For example, receipt of the Morse
code identifier “ D C A uniquely identifies the VORTAC facility located at
Washington National Airport in Washington, D.C.
The DME identification signal consists of on-channel pulse pairs sent at a
periodic rate of 1350 pulse pairs per second. Appropriate circuitry in the
airborne equipment converts the periodic train of decodable pulse pairs into
an audible tone. The periodic sequence is keyed on and off in order to form the
dots and dashes of the Morse code identification signal. During the key-down
period of an I D transmission, squitter transmissions and replies are inhibited,
thereby providing a clear identification signal to the pilot.
Both airborne and ground DME receivers must handle signal levels from
the weakest levels which occur when the aircraft is at the limits of the ground
facility coverage (either above, below, or beyond) to the strongest which can
occur when the aircraft is on the ground near a DME facility. Up to an 80 dB
variation in signal strength can be expected with levels from - 90 dB m on the
low end to - 10 dBm on the high end.
The dynamic range required of the interrogator receiver is operationally
quite different from that of the transponder. Since the interrogator is only
concerned with replies from a single source (i.e., the desired ground facility), a
receiver having a dynamic range on the order of 10 dB will suffice. Once the
signal level being received from the desired facility is known,” changes of the
level can be tracked by automatic gain control circuitry. As noted above,
squitter from the desired transponder often serves as the basis of AGC settings.
Such is not the case for the transponder. Since the transponder does not
know the amplitude of the next pulse to be received, conventional AGC
implementations cannot be utilized, and a wide dynamic range receiver is
required. Consequently, logarithmic amplifiers or log/linear hybrid con-
figurations are typically used in the receiver IF to provide the required
dynamic range.
The mechanisms which affect the accuracy of the range information are the
same for both the interrogator and transponder. The major error mechanisms
are simple, controllable, and readily identified. The accuracy and stability of
the clocks, counters, and/or shift registers used to measure the round-trip time
in the interrogator and to generate the reply delay in the transponder directly
affect the accuracy of the range measurement. As the clock count-to-elapsed
time relationship changes (i.e., oscillator drift due to temperature, age, etc.)
from that assumed when performing the range computation, a bias error will
l 1 The DME system is designed such that the squitter from the desired ground station is the
most persistent signal. The AGC voltage is derived on the assumption that the most persistent
signal is the desired signal.
132 ROBERT J. KELLY AND DANNY R. CUSICK

result. Since digital timing techniques are typically used, quantization noise
error is inevitable. These quantization, errors are reduced by utilizing high-
frequency clocks. Finally, range measurement errors may also exhibit a
dependence on pulse amplitude and rise time. Although such errors are
insignificant in the DME/N application, they are of great concern in the
design of DME/P equipment (see Section IV,D).
In order to prevent the radiation of erroneous navigation information, the
ground facility is actively monitored in real time by what is in essence a high-
quality interrogator. The monitor system interrogates the transponder and
measures those parameters that affect system accuracy and coverage: mean
reply delay, radiated power, receiver sensitivity, frequency stability, and
transmitted pulse code. Because mean reply delay can affect safety of flight, it is
a primary parameter. Primary alerts shut down a ground facility or switch in
redundant equipment if available. Shutdown occurs when the reply delay
error exceeds predefined limits: nominally 500 or 250ft when the ground
facility is associated with a landing aid such as ILS. Radiated power, receiver
sensitivity, frequency stability, and transmitted pulse code are not considered
hazardous and are therefore designated secondary parameters. Secondary
alerts cause an indication to be provided to a control point (such as a Flight
Service Station) when variations exceeding allowable limits are noted by the
monitor.
The remaining components of the DME system are the ground and
airborne antennas. Ground facilities typically provide horizontal coverage
over a full 360". The vertical lobe of the ground antenna must include the
minimum and maximum altitudes of the facility's service volume-for
example, from the horizon to an altitude of 45,000 ft at 130 nmi for the
standard high-altitude service volume defined by the FAA. Pattern shaping to
improve ground system coverage near the horizon results in gains of 9 to
11 dB in the vertical plane with an upper coverage limit on the order of 60".
This upper limit results in a "cone of silence" in the airspace above the station.
An aircraft requires range information independent of its direction of
flight with respect to the ground facility and throughout roll and pitch
variations of up to f30". Typically, the airborne antenna is a 4-wave
monopole providing nearly uniform gain laterally around the aircraft and
maximum gain of 5 db i in the aircraft's vertical plane. In the power budgets
presented later, a gain of only 2 dB i is assumed. This value represents a lower
bound on the antenna gain that can be expected over nominal flight attitude
and aircraft-to-ground facility geometries and therefore determines the
minimum received signal level. For example, consider the common occurrence
of an aircraft in level flight at the limits of the service volume. In this case,
the ground facility lies along a line at a shallow angle with respect to the
longitudinal axis of the aircraft (approximately 3" below the nose when the
DISTANCE MEASURING EQUIPMENT IN AVIATION 133
aircraft is at an altitude of 30,000 ft and 100 nmi from the facility) yielding a
gain on the order of 2 dBi. Typical vertical patterns for both the ground
facility and the aircraft are shown in Fig. 48 (Office of Telecommunications,
1978).

2. An Overview of DME System Considerations


As a viable navigation aid, DME/N satisfies the following system
considerations:
(1) It provides coverage and accuracy that satisfy operational require-
ments for enroute and terminal area navigation. In its most accurate
configuration,it provides position fixes suitable for nonprecision approaches.
(2) It has the capacity to service 100 aircraft desiring distance service
from the same ground transponder.
(3) It provides unambiguous distance information to known ground
locations.
(4) It will service the user demands beyond the year 2000, providing over
200 channels that can be assigned in a way that minimizes mutual interference
and favors simple inexpensive hardware implementations in the aircraft.
Service volumes are defined with range/altitude coverage characteristics
that reflect the intended use of the facility, i.e., high-altitude enroute, low-
altitude enroute, and terminal. The coverage of a single facility is limited to
line-of-sight operations, a characteristic of the L-band portion of the radio
spectrum utilized by DME.' The minimum available coverage is determined
by the radiated power, receiver sensitivity,and antenna characteristics of both
the ground and airborne units.
Enroute facilities are geographically placed so as to have overlapping
service volumes, providing for continuous navigation along defined airways
throughout virtually all of the United States. Terminal facilitiesare associated
with airports and are intended to satisfy the need for short-range navigation in
the terminal area. Some terminal facilities are associated with landing aides
such as ILS and serve as replacements for marker beacons.
Today the system accuracy of DME/N derived range information is on the
order of 0.2 nmi (95%probability). By defining a common calibration method,
system accuracy is preserved among equipments of different manufacture.
Specifically, the 4 amplitude point on the pulse leading edge serves as the
reference point for declaring pulse arrival. All time measurements, both for
range and pulse code, are referenced to the 5 amplitude point of received and
transmitted pulses and are thus independent of pulse rise time and amplitude.
Line of sight is defined as the height above an obstacle equal to the first Fresnel zone
radius. (See Appendix C of BrafT, 1966.)
134 ROBERT J. KELLY A N D DANNY R. CUSICK
ZENITH

AIRCRAFT BOTTOM MOUNTED h 14 MONOPOLE VERTICAL


ANTENNA PATTERN (MAXIMUM GAIN 5.0dBi)

90 80 70 60 50 40

30

20

10

DO

-10

-20

-30

-90 -80 -70 -60 -50 -40


FREE SPACE VERTICAL ANTENNA PATTERN MODEL FOR
TACAN GROUND FACILITY (MAINBEAM GAIN 7.4dBi)

FIG.48. Typical vertical plane antenna patterns for DME ground and airborne elements.
DISTANCE MEASURING EQUIPMENT IN AVIATION 135
The ground beacon has a finite capacity to process interrogations, a
consequence of the dead time that results from the time required to identify a
valid interrogation and transmit the reply pulse pair following the appropriate
fixed delay. Valid interrogations arriving during this dead time are ignored
and result in missed range measurements at the interrogator. This effect
defines an important system performance parameter known as the reply
efficiency. It is simply the ratio of the number of interrogations from a single
interrogator to the number of replies generated by the ground beacon in
response to the interrogations. The reply efficiency decreases as the total
number of interrogations increases, thus limiting the number of aircraft which
can be accomodated by a single ground beacon.
The DME/N system is designed to provide a 70% reply efficiency or better
in the presence of 100 aircraft. In order to serve all 100 aircraft and satisfy the
minimum reply efficiency requirements, both the beacon dead time and
interrogation rate must be limited. Dead times are at least 60 ps while the
interrogation rate for a single interrogator is limited to 30/s for track and
15O/s for search. Computations of the peak load capacity typically assume
that 95% of the interrogators utilizing the beacon are in track while the
remaining 5% are in search.
The reply efficiency is of prime importance during acquisition when the
interrogator is attempting to sort out replies to its own interrogations from all
other beacon transmissions. Once the desired signal has been acquired, the
minimum range information update rate determines the acceptable beacon
reply efficiency, a value which is typically less than that required for successful
acquisition.
Three features of the DME/N system ensure the availability of unam-
biguous distance information to a known ground location. First, ground
beacons are assigned distinct channels characterized by an airborne in-
terrogation frequency/pulse code and a corresponding beacon reply
frequency/pulse code. Second, recall that range information is derived by
identifying those beacon transmissions (replies) which are consistently syn-
chronous with the interrogator transmissions. In order to prevent syn-
chronization with other interrogators utilizing the beacon which can lead to
false range indications, the time between consecutive interrogations is
randomized (jittered). Finally, the Morse code identifier transmitted by the
beacon allows the pilot to verify that the desired station is indeed the one in
use.
252 DME/N channels are defined in the L-band portion of the radio
spectrum at frequencies from 960-1215 MHz. Each channel consists of an up-
and downlink frequency pair (63 MHz apart) with adjacent channels spaced
on 1 MHz intervals. Pulse codes are utilized to allow multiple channels on a
single frequency.
136 ROBERT J. KELLY AND DANNY R. CUSICK

Since the field deployment of DME is necessarily composed of many


overlapping service volumes, due consideration must be given to minimizing
the potential for mutual interference. This is accomplished through a
combination of equipment performance specifications and the appropriate
geographicseparation of the ground facilities.Together these techniquesfavor
simple, inexpensiveairborne implementations while accomodating a sufficient
number of facilities to satisfy air navigation needs.
Today’s DME system is designed to place the burden of signal integrity in
the hands of the ground transponder. In the airborne interrogator, simple
modulators generating square pulses are permitted while rejection of signals
and adjacent frequencies is accomplished with simple filters. To allow for this
airborne simplicity,strict controls must be placed on the uplink spectrum, and
more sophisticated adjacent frequency signal rejection techniques (such as the
Ferris discriminator) must be utilized. The four key aspects of interference
protection are summarized below.

a. Ferris Discriminator. The frequency rejection characteristics of the


ground transponder must ensure that off-frequency interrogations will not
result in undesired replies and unnecessarily degrade the ground system reply
efficiency. Since essentially square pulses can be used by the airborne
interrogators, simple IF filters do not provide adequate attenuation of
adjacent channel signals. Hence, an alternate means of rejecting an off-
frequency pulse is necessary.
Adequate frequency rejection is provided in the ground transponder
receiver by comparing the output levels of a narrow and a wide band IF filter.
The narrow-band filter, with a typical bandwidth of 200 kHz, will pass nearly
90% of the energy contained in an on-frequency pulse while an off-frequency
pulse is attenuated. The output of the wide-band filter (typically 800 kHz or
larger) is needed to make the range measurement since it preserves the high-
frequency information contained in the pulse leading edge. A received pulse is
declared on-frequency when the output level of the narrow-band filter exceeds
a predetermined wide-band filter output level. In DME engineering circles,
this circuit configuration is called the Ferris discriminator, named after its
originator, Hal Ferris (Dodington, 1949).
Rather than simply attenuating the off-frequency pulse as is done in the
airborne equipment, the Ferris Discriminator “identifies” the off-frequency
pulse, allowing it to be excluded from further processing by the ground
transponder receiver. It does so by making use of the fact that 90% of the
energy in a DME pulse is confined to within 0.5 MHz of the carrier frequency,
a consequence of the fact that a relatively wide (3.5 ps) pulse is used. This
technique readily provides the equivalent of 80 dB of frequency rejection, a
level greater than that which can be provided by a simple 800 kHz I F filter.
DISTANCE MEASURING EQUIPMENT IN AVIATION 137
b. Uplink Pulse Spectrum. To accomodate the use of a simple filter in the
interrogator, the spectrum of the uplink pulse is strictly controlled. Specifi-
cally, the effective radiated power (ERP) in the first adjacent channel is limited
to 200 mW and to 2 mW in the second adjacent channel. This is in contrast to
the simple relative adjacent frequency signal level requirement imposed on the
downlink waveform. The implication is that any ERP may be utilized as long
as the power in the adjacent bands is maintained below the specified absolute
levels. This limitation requires careful shaping of the uplink pulse so that
adequate signal power can be provided to achieve the desired coverage while
simultaneously providing a rise time consistent with system accuracy
requirements.

c. Decoder Rejection Characteristics. Pulse codes are utilized to provide for


multiple channels on a single interrogator/reply frequency pair. Thus,
although the spectral resource is limited, more facilitiescan be accomodated in
a given geographic area. To obtain a true increase in the number of available
channels, off-code signals must be rejected by both the ground and airborne
equipment is effectively as off-frequency signals. This is achieved by imposing
strict decoder tolerances on both the airborne and ground equipment. These
tolerancesrequire that pulse pairs having spacingsdeviating more than & 2 ps
be rejected.This ensures that off-code interrogations do not result in undesired
replies and that off-code replies received by an interrogator are excluded from
consideration as valid replies.

d. Channel Assignment Procedures. Channel assignment procedures are


utilized to guarantee that the uplink desired-to-undesired( D / U )signal ratio at
all points within a defined service volume falls within the defined signal
rejection capabilities of the airborne equipment. This is accomplished by
appropriate geographic separation of facilities. Co-frequency/co-code as-
signments are made such that an 8 dB D / U or greater is provided. Adjacent
frequency assignments (both same and different codes) are made such that
when the airborne frequency rejection characteristic and ground system
radiated spectrum are assumed, a post filter D/U of + 8 dB or better is again
obtained. On-frequency/off-code assignments are on equal footing with first
adjacent frequency assignments-a result of the defined decoder character-
istics. This method of frequency protection ensures correct interrogator AGC
operation since the desired signal will always be stronger than any decodable
undesired signal.
It is clear that much effort and analysis has been devoted to defining
equipment characteristics and channel assignment criteria to minimize
interference between DME equipments. In addition, interference pro-
tection from secondary radar systems operating in the 1030 and 1090 MHz
138 ROBERT J. KELLY A N D DANNY R. CUSICK

portions of the L band is afforded by utilizing the immediately adjacent DME


channels only when absolutely necessary. However, such protection cannot be
afforded from other systems utilizing the same frequency band, especially
those operating on an essentially uncontrolled basis.
The Joint Tactical Information Distribution System (JTIDS) is one such
system which is permitted only limited operations within the L band (960-
1215 MHz). It is being developed in the US. and NATO countries to provide
the Allied Forces the level of command, control, and communications
required to ensure effective and affordable military operations in the 1980s and
1990s. JTIDS is a secure,jam-resistant, high-capacity, digital communications
system. It is a spread spectrum system transmitting information in pulsed form
using both pseudorandom coding of each transmitted pulse and frequency
hopping across the entire L band. It will be utilized on airborne, shipboard,
and ground platforms.
If JTIDS operations were allowed to occur “anywhere/anytime,” the level
of JTIDS signals received in both ground and airborne equipment are clearly
uncontrolled and can, for example, be as high as -36 dBm when civilian
aircraft is in close proximity to a JTIDS equipped aircraft. Such high signal
levels can degrade the performance of DME interrogators, with the de-
gradation manifesting itself as a change in the acquisition sensitivity. This is
often due to the JTIDS signals “capturing” the AGC of the interrogator and
thereby desensitizing the receiver.
With the possibility that JTIDS can interfere with not only DME ground
and airborne equipment but the secondary radar systems as well, the FAA,
NTIA, and DOD are participating in an electromagnetic compatibility (EMC)
study to determine the effects of JTIDS on all L-band ATC and navigation
systems, particularly DME/N and DME/P. Once the effects are determined,
necessary restrictions, such as limits on the transmitted pulse rate and power
as well as geographic separation requirements, will be imposed to ensure the
continued performance of all the L-band systems. An earlier JTIDS-EMC
study is documented in Office of Telecommunications (1978). The results of
the most recent effort are expected to be published in 1987.

3. Interrogator Range Tracking Considerations


As noted earlier, the DME interrogator determines the range to the ground
transponder by measuring the time between transmission of its interrogation
and the receipt of the reply to that interrogation. However, the reply to the
interrogation is simply one of many ground facility transmissions that will be
“heard” by the interrogator following an interrogation. Hence, before the
range to the facility can be determined, the interrogator must identify which, if
any, of the received pulse pairs is the reply to its own interrogation.
DISTANCE MEASURING EQUIPMENT IN AVIATION 139
When the interrogator is first tuned to a new channel, it must locate the
“time slot” in which the desired replies are actually occurring. This process
(mathematically known as correlation) is referred to as SEARCH and in modern
interrogators takes some 2 s when a valid reply signal is present. Once the
reply is identified, the interrogator can determine the slant range to the station
and the slant range rate and therefore track the desired replies. In other words,
the interrogator now knows where to find the desired replies (to within a few
p s ) on subsequent interrogations. This process is referred to as TRACK.
The search or acquisition process identifies the correct “time slot” by
finding the interval that contains more replies per interrogation than would
result from the reception of the randomly transmitted squitter. Historically,
search was done by a range gate, or time window, that was slowly (6 nmi/s)
moved out from zero range to a maximum range of say 200 miles. If, at any one
interval during this process, the range gate received a statistically significant
number of replies in excess of that which would be expected from random
ground station squitter, then the range gate would be stopped and tracking
initiated on the basis that synchronous replies were occurring in that time (or
range) interval. With this technique, acquisition of the synchronous range
could take 20 to 30 s. When tracking, the gate would be made to follow the
reply, moving in a direction dependent upon the position of the reply within
the gate, i.e., either early or late.
A popular circuit used to mechanize the function described above was the
phantastron (Kelly, 1963), whose output generated a delay which moved the
early/late range windows. The delay was generated by a voltage picked off of a
potentiometer driven by a servomotor whose error signal was derived from the
early/late range window. Many DME interrogators designed in the late 1950s
and early 1960s used an improved two-scale system which were variations of
the Meacham Range unit developed by Bell Telephone Laboratories (Chance
et al., 1949).
In modern interrogators, range correlation is accomplished by dividing the
covered range into cells. Associated with each cell is a counter where the
counter value represents the confidence that a synchronous reply is falling in
that cell. Following each interrogation, those cells in which valid replies occur
have their associated counters incremented. Those cells where no reply occurs
have the associated counters decremented. The counter corresponding to the
synchronous range will attain a higher value than all the others. This
technique can be implemented by utilizing a random access memory (RAM),
where each RAM location corresponds to a range cell. A specific im-
plementation is described in Cusick (1982).
Figure 49 graphically represents the search process. Assuming a 200 nmi
maximum range, the maximum elapsed time between an interrogation and a
reply is approximately 2500 ps. Hence, the interrogator need only “listen” for
140 ROBERT J. KELLY AND DANNY R. CUSICK
SQUITTE R

F
OWN REPLIES
INTERROGATIONS I SYNCHRONOUS
TO INTERROGATIONS

1
-SEC

5th 11-1 LESS THAN


TRANSPONDER
DEAD TIME

II :- RANDOM OCCURRENCE OF THREE


I I CONSECUTIVE SOUITTER
I

I
REPLIES TO OTHER AIRCRAFT
I
I
d t h

2 5 O O / l S z 200 nmi *
NOTE:
EACH MARK REPRESENTS A PULSE PAIR.
RANGE MEASUREMENTS ARE MADE ON
THE FIRST PULSE.

FIG.49. DME search process.

its reply for a period of 2500 p s following the transmission of an interrogation.


Since the ground transponder is transmitting an average of 2700 pulse
pairs/second, each interrogation sweep will indicate an average of six to seven
randomly spaced pulse pairs. If each sweep is compared, there will be very little
correlation between the times at which the pulse pairs are received from sweep
to sweep except for the desired replies, which will be highly correlated. For the
example shown, the valid reply would have a count of five after six sweeps as
compared to a maximum count of three in the other cells. Therefore,
depending upon the search algorithm utilized, the desired reply could be
identified after four or five sweeps.
The design of acquisition and track algorithms is an interesting and
important aspect of DME interrogator design. Although details of the many
possible implementations and the analysis of their performance goes beyond
the scope of this article, some general comments can be made.
The algorithm must minimize the acquisition time when a valid reply
signal is present. The algorithm must with certainty acquire and track the
DISTANCE MEASURING EQUIPMENT IN AVIATION 141
correct transponder reply. It must never acquire and continue to track the
wrong range, as this could lead to a hazardous situation. In doing so,
the following aspects of the DME environment must be accomodated:
(1) Interrogation rate limits: Clearly, high interrogation rates result in
short acquisition times since the interrogator can collect more reply samples
for correlation in a shorter period of time. However, search interrogation rates
are limited to a maximum of 15O/s in order to minimize the loading of the
ground facility.
(2) Reduced reply efficiency: As has been discussed, the interrogation
load on a DME beacon reduces the reply efficiency as viewed from the
interrogator. The reply efficiency can be as low as 50% under certain loading
conditions and equipment configurations (FAA, 1982). The implication for the
acquisition/track algorithm designer is that the valid reply is present only on a
probabilistic basis, making it more difficult to separate from the randomly
occurring squitter.
(3) Aircraft motion: As the aircraft moves, the desired replies will move
from cell to cell. During acquisition, high slant range velocities may cause
desired replies to appear as uncorrelated replies on subsequent interrogations.
Hence, provision must be made for retaining accumulated range cell
confidence as replies traverse the range cell boundaries.

4 . Hardware Implementation Considerations


A review of various hardware DME realizations during the past four
decades will not be undertaken in this article. However, four design and
component trends are apparent. First, since the 1970s transmitters and
receivers have evolved from vacuum tube designs to all-solid-state designs.
Second, in the middle 1960s, analog circuits were replaced by digital circuits,
especially in the video, timing, and logic circuits. All electro-mechanical
ranging circuits (phantastrons, range potentiometers, and servo loops) have
been replaced by digital configurations using counters, clocks, and shift
register delay lines. Third, with the advent of the microprocessor, all control
functions and data processing functions are performed by computer software
programs. Finally, the most recent innovations have incorporated custom LSI
designs.

C. The DMEIN “Signal in Space”: Accuracy, Coverage,


Channel Plan, and Capacity

In the previous section, a qualitative description of the DME/N system


was presented, illustrating both how the system works and why it is put
142 ROBERT J. KELLY AND DANNY R. CUSICK

together as it is. In this section, the emphasis shifts to the details of the
DME/N signal in space.
Four topics are fundamental to the definition of the DME/N “signal in
space”: accuracy, coverage, channel plan, and capacity. As is evident from
earlier discussions, these topics are interrelated in that the selection of a system
parameter related to one cannot be done without due consideration of
the effect on the complete system. The purpose of this section‘is to detail
these fundamental aspects of DME while stressing the inherent complex
interrelationships.
Subsection C,1 reviews current DME/N accuracy requirements and how
they are achieved. Subsection C,2 discusses various aspects of DME/N system
coverage. Subsection C,3 defines the details of the DME channel plan in terms
of its structure, frequency-assignment considerations, and the “hard pairing”
of the DME channels with the VHF navaid frequencies. Finally, Subsection
C,4 presents details concerning the computation of ground beacon capacity.

1. Accuracy
The accuracy requirements imposed on the DME/N system reflect its
intended use. They satisfy basic enroute and terminal area navigation needs
when utilized in conjunction with bearing information derived from VOR and
TACAN ground stations.
For the purpose of facility planning and equipment procurement specifica-
tions, DME/N system errors must be partitioned into instrumentation errors
and site-dependent errors (i.e., those resulting from multipath and random
pulse interference or garble). Detailed error budgets, such as those to be
presented for DME/P in Section IV of this article, are traditionally not
defined for DME/N. For the DME/N, only the total system error and the
equipment error components are specified. Thus, site-dependent errors
represent the margin after the equipment error components have been
removed from the total system error. As stated in Section II,c, the error
components are assumed to be zero mean random variables and hence can be
combined on a root-sum-square basis. Further, no distinction is made
between noise and bias errors; only a single number is specified. The error
limits are interpreted on a 95% probability basis.
Table VIII illustrates a partitioning of the DME/N system range errors
among the instrumentation and site-dependent components. The values
shown are based on SARPs ground equipment and system error specifications
and SARPs, ARINC, and RTCA airborne specifications (ICAO, 1985;
ARINC, 1982; RTCA, 1985b). The SARPs ground equipment error limits
are more stringent when the facility is associated with a landing aid such
as ILS.
DISTANCE MEASURING EQUIPMENT IN AVIATION 143
TABLE VIII
DME/N SYSTEMERROR
BUDGET (95% PROBABILITY)

Specification (nm)

SARPs RTCA ARINC

Instrumentation
Ground (SARPs) 0.08 - -

Airborne 0.17 0.17 0.10


RSS total 0.19 0.19 0.13
System error (SARPs) 0.20 0.20 0.20
Margin for site-dependent errors 0.07 0.07 0.15
(multipath,garble)

The ground system instrumentation error is 0.04 nmi when the


DME is associated with a landing aid (ILS). Total instrumentation
error and margin for site-dependent effects are modified accordingly.
Ground and airborne instrumentation error components are
those defined in the new DME/N SARPs (ICAO, 1985) and the new
RTCA MOPS(RTCA, 1985b).

Instrumentation errors are specified for received signal power levels that
are at minimum sensitivity and thus include receiver noise effects. Other
factors affecting instrumentation error were noted in Subsection B,1. Sufficient
margin remains for the site-dependent components since:
(1) Multipath errors will typically not exceed 0.04 nmi-a result of the
fact that multipath-to-direct signal ratios can be expected to be less than
- 6 dB since the aircraft is above most of the severe multipath.
(2) Garble-induced errors, which result from extraneous pulses that are
coincident with the ranging pulse (first pulse of an interrogation or reply pair),
are negligible. This is a result of the frequency rejection characteristics of the
ground and airborne equipment as well as the channel assignment method-
ology that ensures an 8 dB post IF filter D / U in the airborne equipment. That
is, extraneous pulses will always be below the amplitude point of the desired
pulse and hence can contribute little to the range error.
It must be emphasized that the site-dependent components of the DME/N
range errors (i.e., multipath and garble) are identical to those that must be
considered when assessing the performance of the I A mode of DME/P. A
more detailed analysis of these effects is presented in Section IV of this article.
Two system level definitions are key in ensuring that the system error
budgets are achieved: the shape of the envelope of the transmitted rf pulse and
the calibration technique. These are detailed in Figs. 50 and 51.
144 ROBERT J. KELLY AND DANNY R. CUSICK

A
0.9A

0.5A

0.1A
0.05A
I
a b c d e f g h

TIME-

Pulse envelope: Measured w i t h a wideband linear detector.


Note: The characteristics of the DME pulse (shape
and spectrum) are defined w i t h respect t o the radiated
signal-in-space and d o n o t include the effects of
receiver f iItering.
Leading edge @ 0.5A: Reference point for all ranging and delay measurements.

Pulse Amplitude (A): Peak amplitude of pulse envelope

Pulse duration (h-a): Pulse duration utilized in determination of average


transmitted power.

Pulse rise time (d-b): Rise time as measured between the 10% and 90%
amplitude points on the pulse leading edge.
(d-b) 2 3ps

Pulse width (f-c): Time interval between the 50% amplitude points on the
leading and trailing edges.
(f-C) = 3.5 * 0 . 5 ~ ~

Pulse decay time (g-e): Decay time as measured between the 90% and 10%
amplitude points on the pulse trailing edge.
(g-e) 5 3.5~s
FIG.50. DME/N transmitted pulse specifications.

2. Coverage
A single nationwide airway system shared by all users of the National
Airspace has been defined with the use of VOR, DME, and TACAN radio
navigation aids. These airways, described earlier in Section II,D, serve as
standard routes to and from airports throughout the U.S. The navaid facilities
(VOR, DME, TACAN, or combinations thereof) are geographically placed so
DISTANCE MEASURING EQUIPMENT IN AVIATION 145
TRANSMITTED INTERROGATION TRANSMITTED REPLY PAIR
INTERROGATION PAIR RECEIVED REPLY RECEIVED AT
PULSE PAIR A T TRANSPONDER PULSE PAIR INTERROGATOR
INTERROGATION CODE REPLY CODE
1 Z U S (XI 12 us (XI
30us I Y I

t
b-
DOWN LINK
PROPOGATION DELAY
(6.18 pslnmil
&$::

TOTAL ROUND TRIP


TRANSPONDER
DELAY
IZERO RANGE
REFERENCE)
up LINK DELAY
PROPOGATION

(6.18 pslnmil

DELAY MEASURED BY INTERROGATOR -4


Total Round Trip Delay-Transponder Delay
Range(nmi) = 12.36psl nmi

The ‘12 amplitude point (i.e., 6dB below the pulse peak) on the pulse leading
edge is the reference pont for all time measurements: range, transponder
delay, and code. Hence, range measurements are independent of pulse shape.

Range measurements are based on first pulse timing

Transmitted RF interrogation pulse pair is processed by interrogator receiver


to begin range timing. Bias errors due to delays in the processing path are
minimized since the received reply pulse pair is processed by the same
receiver path.

The transponder begins transmission of the reply pair such that the desired
delay results between the ‘/z amplitude point on the leading edge of the first
interrogation pulse received and the same point on the first pulse of the reply
pair.
FIG.51. DME/N calibration and range coding.

as to provide for continuous navigation information along the defined


airways.
The volume of airspace served by a single DME ground facility is
influenced by a large number of factors: antenna characteristics, propogation
path variation, ground cover, natural and manmade obstructions, equipment
characteristics such as transmitter power and receiver sensitivity of both the
ground and airborne equipment, and the interrogation load on the ground
beacon. Since the L-band portion of the radio frequency spectrum is utilized,
the maximum coverage is limited to line of sight. The factors which determine
the coverage of a single DME facility are discussed below.
a. Standard Service Volumes. The volume of airspace to be served by a single
DME ground facility is required by the SARPs to be consistent with the
coverage of the facility with which it is paired, i.e., VOR, TACAN, ILS, or
146 ROBERT J. KELLY AND DANNY R. CUSICK

MLS azimuth. In the United States, the FAA has chosen to define “standard
service volumes” to satisfy this requirement. The standard service volume
(SSV) is characterized by a defined volume of airspace and the signal
characteristics (e.g., signal level and frequency protection) that are guaranteed
within that volume. Each service volume definition reflects operational need
and user experience with various types of ground facilities. Three service
volumes applicable to DME/N are defined: terminal, low altitude, and high
altitude.
Each service volume is pictorially depicted in Figs. 52 and 53. Collectively
they connect the airspace up to 60,000 ft (enroute jet routes) down to the 200 ft
decision height (final approach). The shape of the lower edge reflects the effects
of signal attenuation at low elevation angles, the combined result of the free-
space antenna pattern and signal reflections from the ground. In each case,
360”azimuth coverage is provided, with vertical coverage limited to 60” above
the horizon (not shown), thereby defining a “cone of silence” in the area above
the station.
The SSVs reflect the airspace about a given facility that is subjected to
flight inspection prior to commissioning the facility for operational use (FAA,
1980,1982;ICAO, 1972a).Local terrain and obstruction conditions that result
in inadequate system performance are restricted in use, with public notifica-
tion via an appropriate notice to airmen (NOTAM). Where terrain condi-
tions are such that operational performance is achievable outside the SSV,
and an operational need for expanded service is required, an extended service
volume (ESV) may be defined. ESVs typically consist of air routes defined
along a VOR or TACAN radial that extends over water or flat terrain-
areas where adequate signal power is assured at long ranges and low angles
relative to the ground facility due to the lack of signal obstructions. The
same signal standards and flight check certification procedures applicable
to SSVs are required prior to commissioning of the ESV.

b. Power Budgets. Power budgets play a role similar to that of error


budgets. They identify both ground and airborne equipment losses, pro-
pogation losses, and antenna characteristics to ensure that the up- and
downlink signal in space is adequate for the accurate determination of range.
In essence, they allow the determination of the coverage provided by a single
DME ground facility.
The coverage problem consists of two basic parts: the downlink and the
uplink. First, to ensure that the ground facility will respond with high
probability to an interrogation, the airborne interrogator must provide a
signal level at the ground facility that exceeds the beacon receiver sensitivity.
The beacon receiver sensitivity is defined as the signal level that results in 70%
reply efficiency from an unloaded beacon. Second, the ground facility must
DISTANCE MEASURING EQUIPMENT IN AVIATION 147
STANDARD LOW STANDARD HIGH
ALTITUDE SERVICE VOLUME ALTITUDE SERVICE VOLUME

100 nrni
40 nmi

c18,OOO ft.

45,000 h . 4

-1,000 h.

DEFINITION OF THE LOWER EDGE OF THE STANDARD


LOW AND HIGH ALTITUDE SERVICE VOLUMES

W
CI
400

0
0 5 10 15 20 25 30 35 40
DISTANCE TO THE STATION IN nrni
FIG.52. Low- and high-altitude Standard Service Volumes (SSVs) as defined in the U.S.
VOR/DME/TACAN National Standard (AC-00-31A). All elevations are with respect to the
station elevation (AGL).
148 ROBERT J. KELLY AND DANNY R. CUSICK

STANDARD TERMINAL
SERVICE VOLUME

25 nmi

-12,mft.

DEFINITION OF THE LOWER EDGE OF THE STANDARD


TERMINAL SERVICE VOLUME

1000

0
0 5 10 15 20 25
DISTANCE TO THE STATION IN nmi
FIG.53. Terminal Standard Service Volume as defined in the U.S. VOR/DME/TACAN
National Standard (AC-00-31A). All elevations are with respect to the station elevation (AGL.)

provide a signal level at the airborne interrogator that exceeds the airborne
receiver sensitivity. The airborne receiver sensitivity is the minimum signal
level at which the interrogator can acquire and track the desired beacon replies
when the beacon reply efficiency is 70%.
Clearly, trade-offs exist in the selection of radiated power and receiver
sensitivity on both the air-to-ground (AIG)and ground-to-air (GIA) links. An
increase in the radiated power on one end of the link allows a decrease in the
sensitivity of the receiver on the other end of the link and vice versa. Since
weight, size, power consumption, and cost are inherent constraints in avionics
DISTANCE MEASURING EQUIPMENT IN AVIATION 149
design, the burden of high transmitter power and sensitive receivers has
traditionally fallen on the ground equipment.
The DME SARPs as well as the US. National Standard define the
minimum uplink power density available within the standard service volume
as well as the minimum power density required at the ground facility to obtain
70% reply efficiency. With these details defined, the interrogator designer is
free to select transmitter power and receiver sensitivities to obtain the desired
performance (i.e., range coverage). The power density specifications and
typical airborne transmitter/receiver parameters are noted in Table IX.

TABLE IX
SUMMARY OF GROUNDSYSTEMRADIATEDPOWERAND SENSITIVITY
RECEIVERSENSITIVITYREQUIREMENTS,AND TYPICAL
REQUIREMENTS, AIRBORNE
TRANSMITTERPOWERLEVELS

Ground-to-air “available signal power” requirements‘


SARPs -89 dB W/m2 All points in defined coverage volume
US. National Std -86 dB W/m2 Below 18,000 ft
-91.5 dB W/m2 Above 18,000 ft
Beacon receiver sensitivities (70% reply efficiency)
SARPs - 103 dB W/m2 All facilities-input to antenna
US. National Std - 101.5 dB W/m2 DME facilities-input to antenna
- 104.5 dB W/m2 TACAN facilities-input to antenna
Airborne receiver sensitivity requirements
SARPs - 89 dB W/m2 Input to antenna
RTCA (1978) -80dBm Input to receiver terminals
ARINC -90dBm Input to receiver terminals
RTCA (1985b) -83 dBm Input to receiver terminals
Typical transmitter power levels
Ground beacon (US.) Interrogator

Terminal 100 W US. National Std 2 8 0 W EIRP Below 18,000 ft


Low altitude 1000 W 2 800 W EIRP Above 18,000 ft
High altitude 5000 W RTCA 250 W Below 18,000 ft
2250 W Above 18,000 ft
ARINC 21000w
Useful relations
Free-space propogation loss (dB)
= - 37.8 - 2010g(f&) - 20 lOg(R,,J
Conversion of power density (dB W/m2) to isotropic power (dB m)
P(dBm) = 68.5 - 2010g(fM,) + P(dB W/m2)

All levels are provided on a 95% time-of-availability basis.


150 ROBERT J. KELLY A N D D A N N Y R. CUSICK

Signal level specifications in terms of power density (dB W/m’) ensure


independence from implementation details. The ground system designer must
select an appropriate combination of transmitter cabling and antenna
characteristics to provide the required power density in space. No as-
sumptions regarding the airborne installation (i.e., net antenna-cable gain)
utilizing the ground facility are required. The airborne equipment designer
must select antenna, cable, and receiver characteristics that provide the desired
performance in the presence of the minimum available signal power density.
Examples of power budgets that reconcile the ground and airborne equipment
losses are noted in Table X.
c. The Signal Level in Space. The previous discussions have defined the
standard service volumes and the minimum signal levels to be provided within
those service volumes. For a particular equipment configuration, the pre-
diction of the resulting signal levels within the service volume is a complex and
nonexact task. The discussion below is neither rigorous nor complete, but is
intended as an introduction to the problem of predicting facility coverage.
For operations in free space, away from all reflecting surfaces and in a
nonchanging medium (atmosphere), the computation of the available signal in
space can be accomplished when the transmitter power, antenna gains, cable
losses, and distance between the transmitter and receiver are known. This
problem is described in Fig. 54.
The gain of an antenna G is defined as the ratio of the power density at a
point in space caused by a specific antenna to the power density at the same
point caused by an ideal lossless isotropic antenna where each antenna has the
same input power. The antenna gain will vary as a function of its horizontal
and vertical angles from antenna boresight.
The isotropic antenna is one which is perfectly omnidirectional. The
effective capture area A is that area (square meters) which when multiplied by
the power density at the antenna location results in the power produced at the
antenna output port. For an isotropic antenna, the capture area A is A2/47c,
where A is the wavelength of the transmitted signal. Note that the physical
cross section of an antenna can be much smaller than its capture area. For
example, the capture area of a +-wavelength dipole is 1.64 (A2/4x),while the
physical cross section may be 100 times less.
The relationship of antenna gain to the capture area A (Silver, 1949) is
A = (A2/4n)G (21)
Hence, as it should be, the gain of an isotropic radiator is unity. For the DME
application the antenna radiation pattern, G is assumed to be omnidirectional
in the horizontal plane and is assumed to vary only as 0 in the vertical plane.
DISTANCE MEASURING EQUIPMENT IN AVIATION 151
TABLE X
EXAMPLE
GROUND
AND AIRBORNE
DME/N POWERBUDGETS'

Airborne receiver
SARPs Low alt and terminal High altitude

Power available at antenna


airborne -82.0 dB m -79.0 dB m - 84.5 dB m

Antenna gain + 2.0 dB + 2.0 dB + 2.0 dB


Aircraft cable loss - 3.0 dB - 3.0 dB - 3.0 dB

Signal available at receiver


terminals -83 dBm - 80.0 dB m - 85.5 dB m

Receiver noise video noise


figure 12 dB; IF bandwidth
0.8 MHz -106dBm -106dBm -106dBm
Signal-to-noise ratio (video) 23 dB 26 dB 20.5 dB
Beacon receiver
US National Standard
SARPs Low alt and terminal High altitude

Power available at ground


antenna -96 dBm -94.5 dBm -97.5 dB m
Antenna gain + 8.2 dB + 8.2 dB +4.5 dB
Cable loss - 1.5 dB - 1.5 dB - 1.5 dB

Signal available at beacon


receiver terminals - 89.3 dB m - 87.8 dB -94.5 dB
Receiver noise video noise figure
6 dB; I F bandwidth 0.8 MHz -112dBm -112 dBm -112dBm
Signal-to-noise ratio (video) 22.7 dB 24.2 dB 17.5 dB

Notes: (1) Power density (dB W/m2) = power (dB m) - 7; (2) SIN (video)= SIN (IF) +
10 log (IF noise bandwidth/video noise bandwidth); (3) These budgets are intended as examples
only. Other combinations of antenna/cable coupling are possible and should reflect installed
conditions.

That is, G, = GT(8), where 8 is the angle that a given ray makes with the
horizon for the ground antenna. Similarly, let 6, be the angle that a given ray
makes with the aircraft pitch axis for the airborne antenna; i.e., G, = G,(8,).
Let PT equal the power radiated by an isotropic radiator. The power
radiated in any direction is PT/471 watts per steradian. Assume that the
isotropic radiator is replaced by an antenna with gain GT(6) over isotropic.
The power radiated in that direction is P&(8)/4n watts/per steradian.
152 ROBERT J. KELLY A N D DANNY R. CUSICK

GT GR

P i = TRANSMITTER POWER (WATTS)


LT = TRANSMITTER CABLES LOSS
GT = TRANSMIT ANTENNA GAIN (WATTSMTATT)
R = TRANSMITTER TO RECEIVER DISTANCE (METERS)
GR = RECEIVER ANTENNA GAIN (WATTSMTATT)
LR = RECEIVER CABLE LOSS
PR = RECEIVED POWER (WATTS)

FIG.54. Definition of parameters for the analysis of a transmit-receive link

Placing a receiver antenna with effective capture area A at distance R from the
transmitting antenna subtends a solid angle of AIR2 steradians. Thus the
power at the receiver PR is
PR = P,GT(B)A/4ZR2 (22)
The effective receiving antenna capture area is
A = (A2/47r)G,(8,) m2 (23)
Let LT and L R represent the cable losses between the antenna and the
transmitter or receiver, respectively, as shown in Fig. 54. Hence, the received
power at the input to the receiver is simply the product of the power density at
the antenna, the capture area of the receiving antenna, and the cable losses:

The term (A/47rR)’ is the space or propagation loss and is proportional to the
DISTANCE MEASURING EQUIPMENT I N AVIATION 153
inverse square of the range. At L band, the propagation loss is conveniently
expressed as
free-space propagation loss (dB)
= -37.8 - 20 log(&Hz) - 20 log(R,,i) (25)
If beacons and interrogators operated in free space, the above discussion
would provide the means to predict the useful range of operations that
could be achieved with an interrogator/beacon pair whose operational
parameters-antenna characteristics, transmitter power, and receiver sensi-
tivities-were known. But they do not. Operations are conducted near a
spherical Earth from which radiated signals are reflected and scattered and in
an atmosphere that has pronounced and not always predictable effects. These
factors must be considered in predicting the useful range of the DME system.
The signal at the interrogator receiver is the sum of the direct path signal
and reflected signals from the ground (assuming that neither signal path is
blocked by obstructions and that long delay reflections are absent). This
results in a received signal which can deviate significantly from the direct path
signal (Fig. 55). The reason is simply that the two signals are related in phase
and add vectorially. When the direct and reflected signals are in phase, the
received signal will be larger than the free-space value. When they are 180" out
of phase, the received signal is the difference between the direct and reflected
path signal levels and hence the received signal level is reduced below the free-
space value.

FIG.55. Geometry for signal strength calculation in the presence of ground multipath.
154 ROBERT J. KELLY A N D DANNY R. CUSICK

Ground reflection effects can be approximated using Fig. 55 as follows: Let


the direct signal be

where o is the rf carrier frequency and cpD is the path delay. Let the reflected
signal be

where pantis the magnitude of the reflected ray at the transmitter antenna, pRis
the ground reflection coefficient, I$ is the reflection coefficient phase, and cpR is
the reflected signal path delay. The above assumed that the antenna gain
GT(6D) = 1 for ZD and CT(6R) = pan,for i f R . RD and RR are the path lengths for
the direct and reflected signals, respectively.
A pattern propagation factor I FI can be defined which permits the effects of
ground reflections to be incorporated into the range equation. IF1 is the ratio
of the resultant signal at the point in question to the value it would have been
under free-space conditions.
Using Fig. 55

where

and
AR = RR - RD N 2hi h,/R
Note that h , and h, have been measured from the tangent plane shown in
Fig. 55. For a flat earth, h , = hT and h, = hR.
The range Eq. (24) becomes
h = PTGT(6)GR(e)A)IF12(~/4~R)2 (27)
DISTANCE MEASURING EQUIPMENT IN AVIATION 155
where GR(OA)is equal for both the direct and reflected signals. Equation (27)
applies both to the DME up- and downlinks.
The phase change on reflection 4 and the reflection coefficient pRdepend
on, in addition to the factors already noted, the electrical properties and the
degree of roughness of the reflecting surface. The roughness of the surface
determines the resemblance of the surface to a mirror. A surface which appears
smooth to long-wavelength radiation may be rough for short-wavelength
radiation.
+
For the case of a single reflection, IF1 varies between (1 p ) and (1 - p).
When pRis nearly unity, that is, the reflecting surface is mirrorlike (little or no
attenuation occurs), and pantN 1, IF1 will vary between 0 and 2 times the free-
space value. On the other hand, for p << 1, IF1 will approximate the free-space
value.
When AR is small, pR 2: 1 and 4 = 180°, then Eq. (26) becomes (F12 =
( 4 ~ h , h , / Rand
) ~ Eq. (27) becomes

This condition can occur when the aircraft is close to the ground during the
landing phase of flight. Since the received power decreases as l/R4, a careful
antenna design is warranted as discussed in Section IV,C,4 for the DME/P
siting considerations.
It is clear that if the system parameters of transmitter power, antenna
characteristics, and the reflection parameters of 4 and pR are known, the
effects of reflections can be accounted for and the received signal level
computed for a specified receiver/transmitter geometry. Although the system
parameters are known, the parameters which determine the characteristics of
the reflected signal are subject to wide variation since the ground and its
covering vegetations change with season, weather, and wind.
In order to make prediction of the signal level in space under a wide range
of conditions, a statistical treatment of the reflecting characteristics and
atmospheric effects is required. Effects of the reflecting surface and at-
mospheric losses must be treated as random variables whose density functions
must be convolved to determine the distribution of the power density in space.
This complexity requires the use of computer models to perform the
computations accurately.
A computer-based propagation model, developed by the Institute for
Telecommunication Sciences, was developed to perform just this function (see
Johnson and Gierhart 1978a,b, 1980, 1983). The model is applicable for
ground/air telecommunication links operating at frequencies from 0.1 to
20 GHz for aircraft altitudes below 300,000 ft and is utilized by the FAA to
156 ROBERT J. KELLY AND DANNY R. CUSICK
100

I-
w
Y

LL 80
0
:
260
3
0
I-
z
-40
-I

-a
(1

x
3 20
t
5a
0
0 20 40 60 80 100 120 140 160 180 200 220
DISTANCE FROM THE DESIRED STATION I N nmi

FIG.56. Signal strength in space for a SO00 W TACAN beacon primarily utilized for high-
altitude enroute navigation. This graph does not show the upper coverage limit, i.e., the signal
strength in the “cone of confusion” above the station.

DISTANCE FROM THE DESIRE0 STATION I N nmi

FIG.57. Signal strength in space for a lo00 W DME beacon primarily utilized for low-
altitude navigation. This graph does not show the upper coverage limit, i.e., the signal strength in
the “cone of confusion” above the station.

define standard facility performance for a given set of transmitter power


levels and antenna characteristics. Since the DME system is a safety service,
signals must be highly reliable, and thus conservative assumptions are
made as inputs to the propogation model. Figures 56, 57, and 58 illustrate
typical results of the propagation model for actual DME ground facilities.
DISTANCE MEASURING EQUIPMENT IN AVIATION 157

0 20 40 60 80 100 120 140 160 180 200 220


DISTANCE FROM THE DESIRED STATION IN nmi

FIG.58. Signal strength in space for a 100 W DME beacon primarily utilized for terminal
area navigation. This graph does not show the upper coverage limit, i.e., the signal strength in the
“cone of confusion” above the station.

Additional such “Power Available” curves are provided in FAA (1980) and
Johnson and Gierhart (1980).
Note, for example, in Fig. 56 that the -79 dBW/m2 contour equals the
free-space loss at 140 nmi. At 70 nmi the free-space loss should have been only
-73.5 dBW/m2 but it remains at -79 dBW/mZ, thus reflecting the above
ideas. That is, propagation losses are a random phenomenon and can only be
described statistically.
The signal strength curves shown in Figs. 56,57, and 58 are in terms of 95%
time of availability. This means that the instantaneous signal strength will be
greater than or equal to that value 95% of the time. Similarly, the D/U ratios
that comprise the channel assignment rules are also defined on a 95%
availability basis.
The curves in Figs. 56, 57, and 58 contain many propagation effects not
included in the Fig. 55 analysis. One of these effects, radio beam bending
caused by atmospheric refraction, requires comment. The atmosphere above
the earth has an index of refraction which decreases with increasing altitude.
This decreasing refractive index causes radio beams near the earth’s surface to
bend downward, as indicated in Fig. 59a. Therefore, the distance to the radio
horizon is greater than the geometric horizon distance. In order to perform
intuitively simple calculations, it is desirable to transform these curved radio
beams into straight-line transmission paths. The transformation is accom-
plished by subtracting the radius of curvature of the refracted beam from the
earth‘s curvature (Johnson and Gierhart, 1978a). The result is an effective
158 ROBERT J. KELLY AND DANNY R. CUSICK

earth radius K a = (4/3)a, where a is the actual earth radius. As noted in Fig.
59b, the radio beams now plot as straight lines. The geometry used in Fig. 55
was based upon an equivalent 4/3 radius earth. As long as the receiver is within
the radio horizon, the propagation losses are simply the straight-line losses.
Beyond the radio horizon, the radio beam is in the diffraction region;
therefore, an additional diffraction loss factor must be included in the power
budgets. The diffraction loss factors are 1.6 dB/nm and 2.8 dB/nm at 1000 and
5000 MHz, respectively.These considerations are important when performing
cochannel protection ratio calculations in Subsection III,C,3,b.
d . Uplink/Downlink Design Considerations. DME is a two-link system in
that both the up- and downlinks must operate successfully for range
information to be obtained by the interrogator. Specifically, the signal at the
ground beacon must have sufficient strength to result in a reply efficiency that
is adequate for acquisition to occur with the available signal power at the
interrogator. The signal level at the interrogator that results in acquisition of
the desired signal, determined by considering the performance of both the up-
and downlinks, is referred to as the beacon-coupled acquisition sensitivity
(Office of Telecommunications, 1978). It may be different from a bench
measurement of interrogator sensitivity, since the interrogator transmitter
power and ground facility receiver performance must be taken into account.
The analysis method described below illustrates how the appropriate values of
interrogator transmitter power and receiver sensitivity can be chosen to
ensure that an interrogator will acquire uplink replies to its downlink
transmissions.
Using Eq. (24), the ratio of the power at the beacon receiver terminals PRb
to the interrogator transmitter power PTiis

R‘ b
- = GTGRLTLRJFJ2
pTi

The ratio of the power at the interrogator terminals PRito the beacon
transmitter power PTbis

Since the interrogating signal and the reply signal traverse the same paths, the
right-hand sides of Eqs. (28) and (29) are equal, and the ratios can be equated,
yielding
DISTANCE MEASURING EQUIPMENT IN AVIATION 159

INTERFERENCE REGION

REFRACTION

RAY CURVATURE OVER SPHERICAL EARTH OF RADIUS a

INTERFERENCE REGION

WITH
REFRACTION

RAYS WHEN USING EOUlVALENT4/3 RADIUS EARTH.

RADIATION PATHS WITHIN~RADIOHORIZON HAVE STRAIGHT LINE RAYS


RADIATION PATHS BEYOND RADIO HORIZON IDIFFRACTION REGION1
HAVE l.BdB/nm ADDITIONAL LOSS P L-BANDA N 0 2.8dBInm B C-BAND.
RADIO HORIZON DISTANCE dlnml = d1.51h (FEET)

FIG. 59. Definitions of radio horizon and geometric horizon.

The power at the interrogator receiver terminals therefore differs from that
received at the beacon receiver terminals by PTb(dBm)- PTi(dBm).
Equation (30) is applied as follows. The ground beacon transmitter power
and receiver sensitivities are specified by the FAA (1982)and listed in Table IX.
For the low-altitude service volume PRb= -94 dBm and PTb= 60 dBm.
Furthermore, as shown in Table IX, the FAA provides a signal in space of
- 79 dBm, (- 86 dBW/m2) with 95% availability for the low-altitude service
volume. Using Table X, the power at the interrogator receiver input is PRi=
- 80 dBm. The manufacturer must design a receiver with at least a - 80 dBm
to detect and measure the signal at the limits of the low-altitude service
volume. Inserting PRb= - 94 dBm, PTb= 60 dBm, and PRi= - 80 dBm into
160 ROBERT J. KELLY AND DANNY R. CUSICK

Eq. (30) yields PTi= 46 dBm = 40 W [the MOPS (RTCA, 1978) requires at
least 50 W]. Thus at least a 40 W transmitter is necessary to ensure that the
downlink is balanced with the uplink.
A calculation to determine the required transmitter power for the high-
altitude service volume proceeds as follows. From Table IX the uplink signal
in space is - 91.5 dBW/m2 or - 84.5 dBm. Therefore, the receiver design must
successfully detect PRi= -85.5 dBm (ARINC requires -90 dBm). Using
Table IX, PTb= 67 dBm and PRb= -96 dBm for the TACAN beacon. The
required interrogator transmitter power is 446 W.

3. Channel Plan
The purpose of a channel plan is to allocate the available spectral resource
in an efficient manner while satisfying the operational requirements of the
system. In the case of TACAN/DME, the spectral resource is the 960-1215
MHz portion of the L band, and the operational requirement is to provide
unambiguous high-quality range data along all points of defined airways. To
be complete, the channel plan concept must include the following:
i. Channel Dejnition. Defines the frequency separation that accomodates
the pulse shapes and signal processing bandwidths necessary to achieve
system accuracies.
ii. Channel Assignment. Provides a required number of ground facilities in
an manner that precludes or at least minimizes mutual interference.
iii. Channel Pairing. Simplifies selection of co-located stations, i.e., VOR-
DME, VORTAC, ILS-DME, and MLS-DME, thereby minimizing pilot
workload and blunders.
Accommodating the above involves a multitude of system considerations.
For example, channel frequency spacing impacts equipment transmit and
receive frequency stability requirements-the closer the channel spacing in
frequency, the more stable the equipment must be to prevent interference due
to drift. In contrast, accuracy considerations dictate the receiver bandwidth
requirements which in part determine susceptibility to off-frequency extra-
neous signals. In essence, this defines the channel assignment criteria. Channel
assignment criteria are then made manageable by strict spectral limitations
imposed on the transmitted waveforms.
a. TACANIDME Channel DeJnition. The TACANIDME system operates
in the 960-1215 MHz (L-Band) portion of the radio spectrum. This band,
allocated for aeronautical radionavigation, is reserved on a worldwide basis
for the use and development of airborne electronic aids to air navigation and
any directly associated ground-based facilities [National Telecommunication
DISTANCE MEASURING EQUIPMENT IN AVIATION 161
and Information Administration (NTIA), 19843. TACAN/DME interroga-
tion and reply frequencies are defined on 1 MHz increments between 962
and 1213 MHz, leaving 2 MHz guard bands at both ends of the band. Fre-
quencies immediately below the 960-1215 MHz band (942-960 MHz) are
allocated for use in mobile and fixed communications as well as broadcast
functions. Frequencies immediately above the 960-1215 MHz band (1215-
1240 MHz) are allocated for radiolocation and radionavigation-satellite
(space-to-earth) functions (e.g., the NAVSTAR global positioning system
operates in this band).
The basic TACAN/DME channel plan defines 252 distinct operating
channels numerically designated 1-126. A channel is defined by an
interrogation/reply frequency pair and a pulse code, either X or Y. The
interrogation frequency is the transmission frequency of the airborne
equipment, while the reply frequency is the transmission frequency of the
ground equipment. These two frequencies are 63 MHz apart, with 63 MHz
serving as the IF frequency in typical TACAN/DME equipment. The use of a
63 MHz I F takes advantage of the interrogation/reply frequency separation
in that the transmit frequency can be used as the local oscillator in the receive
channel, thereby reducing the number of crystals required in early DME
equipment and simplifying the frequency synthesizers utilized in modern
DME equipment. The code of the channel refers to the spacing of the pulse
pairs as utilized by the interrogator and the ground beacon. This “encoding”
of interrogation and reply transmissions allows for multiple channels on a
single frequency.
The channel plan utilizes 126 interrogation frequencies on 1 MHz
increments between 1025 and 1150 MHz inclusive. Channel numbering is
sequential with increasing interrogation frequency. Associated with each
interrogation frequency are two channels formed by the X and Y codes. Each
channel is uniquely distinguished by beacon reply frequency and the pulse-
pair code used for interrogation and reply. X channels 1-63 utilize reply
frequencies between 962 and 1024 MHz, while the corresponding Y channels
utilize reply frequencies between 1088 and 1150 MHz. X channels 64-126
utilize reply frequencies between 1151 and 12 13 MHz, while the corresponding
Y channels utilize reply frequencies between 1025 and 1087 MHz. Note that,
unlike the X channels, the Y channels reuse frequencies within the inter-
rogation frequency block. This basic channel plan structure is summarized in
Fig. 60. Although the DME/P channel plan is not detailed until Section
IV,A,4, the interrogation and reply frequencies for the newly defined W and Z
coded channels, used exclusively for DME/P, are shown for future reference.
This channel pairing has been adopted by ICAO for facilities supporting
operations in the international air navigation service and also serves as the
basis for all frequency planning and assignments in the U.S. national airspace
162 ROBERT J. KELLY AND DANNY R. CUSICK

I
-
960 MHz 1213 M H r

96; M H r 11215 MHz


1025 MHr 1088 M H r 1151 MHz

- '.
0'
H0

",,1 - 6
'.
1

\
'.
I

0'
'0
0
64-
0

/-
'
0

,yO
1-

-.
7-562
"'
1

-
30-119Zc
, 0

KW:
CH = C H A N N E L NUMBER (1-1261
INTERROGATION FREQUENCIES
I = I N T E R R O G A T I O N FREOUENCY IMH7.
1-1 REPLY FREQUENCIES R = REPLY FREQUENCY (MHzl

X CHANNELS Y CHANNELS

I = 1025 t (CH-11 I = 1025 + (CH-1 I


R=I-63 1 < CH G 63 R=It63
R=It63 64 < C H < 126 R=I-63
12 us INTERROGATION CODE 36 us
12 us REPLY CODE 30 vs
50 us TRANSPONDER REPLY D E L A Y 56 vs

NOTE: W A N D Z CODED CHANNELS ARE USED EXCLUSIVELY FOR DME/P A N D ARE


CODE M U L T I P L E X E D ON THE X A N D Y FREQUENCIES RESPECTIVELY

FIG.60. TACAN/DME channel plan with interrogation reply frequency pairing, codes,
and transponder reply delays.
DISTANCE MEASURING EQUIPMENT IN AVIATION 163
system (NAS).In the International channel plan, channels 1-16X are reserved
for national use. Additionally, channels 1-16Y, 60-69X, 60-79Y, and 124-
126Y are to be utilized on a secondary basis to afford frequency protection to
secondary surveillance radar systems such as ATCRBS and MODE S, as
well as the newly defined TCAS. These systems utilize 1030 MHz as the
ground radar interrogation frequency (airborne interrogation frequency in
TCAS) and 1090 MHz as the airborne transponder reply frequency. In the
United States, TACAN/DME channels 17-59 and 70-126 are designated
for use in the national airspace system, while channels 1-16 and 60-69 are
reserved for tactical military operations.

b. TACANIDME Channel Assignment. Several basic principles govern


TACAN/DME ground facility channel assignment. These principles, as
described earlier in Subsection B,2, make optimum use of the frequency and
code rejection characteristics of the ground and airborne equipment as well as
the radiated pulse spectrum to maximize the number of facilities that can be
accommodated in a geographic area.
Ground facility channel assignments are made in a way that provides a
minimum uplink free-space desired-to-undesired ( D / U ) signal ratio at all
points within a desired facility’s service volume. When making a frequency
assignment, each facility must be considered a desired facility with all others
treated as undesired facilities. Only when the D / U requirements are satisfied
for all facilities can an assignment be made.
The specific free-space D / U criteria is chosen such that an effective (post-
processing) D / U of 8 dB is achieved in the interrogator when its frequency and
code rejection capabilities are considered. In this case, the undesired signal
level is the peak power from another ground facility. Thus, the interrogator
designer is guaranteed that, within the defined operational service volume, the
desired signal will always have a minimum 8 dB advantage over all undesired
beacon transmissions. The designer is free to utilize this discriminate as
desired.
Frequency assignment rules are based on both the spectrum of the ground
system transmitted waveform and the airborne interrogator receiver fre-
quency rejection characteristic as summarized in Table XI and Fig. 61.
Ground system radiated spectrum specifications limit the radiated power in
adjacent frequency bands to an absolute level of 200 mW on the first adja-
cent channel and 2 mW on the second adjacent channel. This implies that the
transponder design may utilize any ERP level so long as (1) the pulse shape
specifications are satisfied, (2) the minimum power densities are provided at
all points within the coverage volume, and (3) the absolute spectrum require-
ments are satisfied. Thus, for a facility with a 43 dBW ERP, the first adjacent
channel will be 50 dB down with respect to the on channel signal. The
164 ROBERT J. KELLY AND DANNY R. CUSICK

-50dB for 43dBW ERP


RADIATED SIGNAL -47dB for40dBW ERP
-3718 for 3OdBW ERP

11 f0
ADJACENT ON CHANNEL
CHANNEL
FREllUENCY

1st ADJACENT FREOUENCY

0 Assumes 50 d B Rejection in Interrogator

2nd ADJACENT FREQUENCY

oAtsumer 7 0 dB Rejection in Interrogator

FIG.61. Ground system frequency assignment criteria and related interrogator frequency
rejection assumptions.
DISTANCE MEASURING EQUIPMENT IN AVIATION 165
TABLE XI
SPECTRUM
CONSTRAINTS
ON rn DME TRANSMITTED
WAVEFORM

Adjacent frequency band Power with respect


relative to transmit Maximum ground system to peak on-channel
frequency radiated power" power

0.55-1.05 MHz 200 mW -23 dB


1.75-2.25 MHz 2 mW -38 dB

,I Radiated power = total radiated energy in band/pulse duration; pulse


duration is the time between the 5% points on the leading and trailing edges.

interrogator frequency rejection characteristics, whether achieved by IF fil-


tering or by the use of a Ferris Discriminator, are such that the undesired
main lobe ground transmission is attenuated at least 50 dB. RTCA
(1978) specifies that all interrogators satisfy this requirement.
For undesired ERP levels (P,,) >43 dBW, the frequency assignment
process is governed by the main lobe of the radiated signal. This results from
the fact that even after the 50 dB rejection by the interrogator receiver, the
main lobe of the undesired signal will be larger than the undesired sidelobe
signal in the interrogator receiver. This dictates the need for a free-space
DIU > -42 dB to achieve the required post processing 8 dB DIU protec-
tion. The required DIU is achieved by physically separating the two beacons
until D/U > -42 dB. For undesired ERP levels <43 dBW, it is the side-
lobes of the radiated signal which dominate in the interrogator receiver. In
this case, a free-space DIU = 1 - P,, is required to ensure the desired 8 dB
post-processing DIU. Note that the two ground facilities are physically sepa-
rated until the 200 mW sidelobe is 8 dB below the desired signal. Reducing
P,, does not mean that the facilities can be brought closer together. Defining
DIU = 1 - P,, is just a means to make the D/U measurement. For example,
suppose the channel assignment conditions are satisfied at a separation dis-
tance of 50 nmi with P,, = 40 dBW, then reducing the power to 26 dBW, for
example, does not reduce the separation distance.
For second adjacent frequency assignments, the same 8 dB post-processing
DIU criterion applies. This results in an allowable free-space DIU of - 62 dB
for undesired ERP values >63 dBW and a preprocessing DIU = - 19 - P,,
(P,, in dBW) for P,, <63 dBW. The additional 20 dB in each of the various
criteria reflects the 20 dB decrease in the adjacent channel power.
Until the adoption of DMEJP, there was only one code (X or Y)
associated with each uplink frequency. There was no need to consider co-
frequencyloff-code assignments except for air-to-air interference. The as-
signments assumed that the interrogator decoder rejection was 8 dB with a
166 ROBERT J. KELLY AND DANNY R. CUSICK

+ 6 p s aperture. However, in order to provide 200 DME/P channels for


hard pairing with the 200 MLS angle guidance channels, additional pulse
codes (W and Z ) were introduced. The W channels are formed by code
multiplexing with X channel frequencies, while the Z channels are formed
by code multiplexing with the Y channel frequencies. Channel plan details
specific to DME/P are discussed in Section IV,A,4. The W and Z codes can
be used on a restricted basis after Jan. 1, 1989, and for general use after
Jan. 1, 1995 [All Weather Operational Panel (AWOP), 19821.
With the addition of multiplexed codes, channel assignment criteria and
interrogator pulse code rejection characteristics must be revised accordingly.
Ground facility code assignment criteria for co-frequency assignments require
a -42 dB free-space DIU based on 50 dB interrogator decoder rejection of
received signals with pulse spacings more than 2 p s from the nominal code
spacing. For the off-code second adjacent channel, the required interrogator
decoder rejection is 70 dB, where the additional 20 dB arises from the fact that
the second adjacent power cannot exceed 2 mW.
With these design principles, the uplink assigned codes are on an equal
footing with respect to the uplink frequency assignments. That is, co-
frequencyloff-code channels are rejected as effectively as co-code/off-
frequency assignments. The new SARPs airborne decoder tolerances for both
the DME/N and DME/P decoder apertures have been tightened to & 2 ps.
Because Ferris Discriminators are necessary, DME/P interrogators reject off-
frequency channels at least 50 dB. Therefore, in DME/P, the off-frequency
channels have the same rejection characteristics as the off-code channels. Note
that the term “rejection,” as used here, refers to the ability of the interrogator
signal processing to ignore off-code and off-frequency signals during acquisi-
tion and in determining the appropriate AGC levels. It does not refer to the
ability of the signal processing to handle garble-induced range measurement
errors and reply efficiency degradation when desired and undesired signals are
coincident in the interrogator receiver.
The channel assignment process explicitly considers only the assignment
of beacon reply frequencies. Air-to-ground frequency protection is provided
by appropriate specifications imposed on the ground equipment in terms of
off-code and off-frequency signal rejection. These specifications require that
ground transponders be equipped with both:

(1) A Ferris Discriminator which will inhibit the generation of replies for
all signals 900 kHz and more removed from the desired channel nominal
frequency; and
(2) A decoder which will inhibit the generation of replies for all signals
less than 75 dB above the minimum receiver sensitivity level with the received
pulse pair 2 p s removed from the nominal channel code spacing.
DISTANCE MEASURING EQUIPMENT IN AVIATION 167
The above results are summarized in Table XII, Column A. The protection
criteria defined in Column A are applicable until Jan. 1, 1989. After Jan. 1,
1989, the protection criteria listed in Column B will become effective. The
criteria change after 1989 includes not only added emphasis on off-code
rejection and improved interrogator performance; it also changes the
protection criteria for off-frequency assignments. After 1989 the post process-
ing D/U 2 8 dB for off-frequency assignment will be abandoned. The co-
channel criterion of 8 dB remains in effect. As shown in Column B, the first
adjacent frequency co-code criterion is no longer (1 - P,) dB but has been
relaxed to -42 dB. This means that a new facility can be physically moved
toward an undesired facility (with P, = 25 dBW, for example) until the free-
space D/U = -42 dB. Then the 200 mW adjacent spectal sidelobe will not be
8 dB below the desired signal level at the output of the IF Filter. In fact, the
adjacent sidelobe of the interfering signal will be 10 dB above the desired
signal. Therefore, after 1989, Ferris Discriminators or their equivalent will be
required in DME/N interrogators as well as in DME/P interrogators.
The frequency assignment process for all navaids within the National
Airspace System-TACAN/DME, VOR, ILS, MLS, and DME/P-is
clearly a complex task where the assignment criteria are interrelated due to the
collocation of various facilities and the “hard pairing” of various channel
plans. Given the enormous number of navaids, the use of ITS propagation
models for determining the desired and undesired signal levels within the

TABLE XI1
PROTECTION
RATIO D/U (dB)

Type of assignment A (before 1989)” B (after 1989)

Cofrequency
Same pulse code 8 8
Different pulse code 0 - 42

First adjacent frequency


Same pulse code -(P” - 1)* - 42
Different pulse code -w” + 7) - 75
Second adjacent frequency
Same pulse code -(Pu - 19) - 75
Different pulse code -(Pu + 27) - 75

a The DjU ratios in Column A protect those DME/N inter-

rogators operating on X or Y channels. Column A applies to


decoder rejection of 8 dB and a 6 ps aperture.
* Puis the peak effective radiated power of the undesired signal
in dB W. The frequency protection requirement is dependent upon
the antenna patterns of the desired and undesired facility and the
ERP of the undesired facility.
168 ROBERT J. KELLY A N D DANNY R. CUSICK

defined service volumes, the need to maximize the number of facilities which
can be accommodated within the NAS, and the fact that frequency as-
signments for new facilitiesmust satisfy the specified D / U relationships with all
existing navaids, the complexity of the channel assignmentprocess dictates the
need for computerized assignment procedures. Such procedures have been
developed. These and other details concerning navaid assignment are found in
Hensler (1984) and FAA (1980).
c. TACANIDME Channel Pairing with Other Navaids. In practice,
TACAN/DME stations may be collocated with VOR facilitiesto provide both
range and azimuth information to civil aircraft. Similarly, TACAN/DME
stations may be collocated with ILS installations to provide terminal area
range, azimuth, and glide slope information to properly equipped aircraft. To
facilitate the collocation of these equipments, the VOR/ILS (VHF) channels
are “hard paired” with the TACAN/DME (UHF) channels. This “hard
pairing” between the VHF and DME facilities, summarized in Table XIII, is
used to advantage to reduce pilot work load. Typical installations provide for
automatic DME channel selection consistent with the manual selection of the
VHF facility by the pilot.
VOR/ILS localizer facilities use frequencies in the 108-118 MHz band
with channels spaced on 50 kHz increments. Forty ILS localizer (azimuth)
channels are defined between 108.10 and 111.95 MHz on each odd 100 kHz
pair (i.e., 108.10, 108.15, 108.30, 108.35, . .. , 111.90, 111.95). The remaining
frequencies in the VHF band are assigned for VOR. ILS glide slope facilities
TABLE XI11
PAIRING
TACAN/DME AND VHF NAVAIDCHANNEL SCHEME
~ ~~ ~

TACAN/DME channel VHF Navaid


200 NAS channels 200 VHF frequencies

17-59 X & Y 108.0-1 17.95 MHz


70-126 X & Y 50-kHz spacing
18-56 X & Y even ILS localizer: 108.10-111.95 MHz
(40channels) odd 100 kHz and next higher 50 kHz
(ie., 18X with 108.10, 18Y with 108.15,...)
17-55 X & Y odd VOR: 108.0-111.05 MHz
(40 channels) even 100 kHz and next higher 50 kHz
(i.e., 17X with 108.00, 17Y with 108.05,.. .)
57-59 x&Y VOR: 112.0-1 17.95 MHz
70-126 X & Y SO-kHz spacing
(120 channels) (i.e., 57X with 112.0, 57Y with 112.05,...)

Note: ILS localizer frequencies are non-sequentially paired with ILS


glide slope frequencies (329.15-335.0 MHz, 150 kHz spacing).
DISTANCE MEASURING EQUIPMENT IN AVIATION 169
use the 328.6-335.4 MHz band with assignments beginning on 329.15 to
335.0 MHz on 0.15 MHz increments. Localizer and glide slope frequencies are
also hard paired.
Even-numbered X and Y TACAN/DME channels 18-56 are hard paired
with the 40 ILS localizer frequencies. The pairing is in sequence with increasing
channel number and frequency: 18X with 108.10 MHz, 18Y with 108.15 MHz,
etc. Odd-numbered X and Y TACAN/DME channels 17-57 are hard paired
with the VOR frequencies between 108.0 and 112.05 MHz. The remaining
TACAN/DME channels in the NAS (58,59,70-126) are hard paired with the
remaining VOR frequencies 112.1-117.95 MHz in 50 kHz increments.
Complete listings of the VHF/UHF/MLS frequency channeling and pairing
can be found in NTIA (1984).

4 . System Capacity
The ground beacon has a finite capacity to process interrogations, a
consequence of the dead time that results from the time required to identify a
valid interrogation and transmit the reply pulse pair following the appropriate
fixed delay. Valid interrogations arriving during this dead time are ignored
and result in missed range measurements at the interrogator. This important
system performance parameter, introduced in Subsection B and referred to as
reply efficiency, is simply the ratio of the number of interrogations from a
single interrogator to the number of replies generated by the ground beacon.
The reply efficiencydecreases as the number of interrogations increases and
thereby affects the number of aircraft which can be accomodated by a single
ground beacon. As noted earlier, ground facilities must serve a minimum of
100 aircraft while maintaining a reply efficiency of at least 70%.
Figure 62 illustrates the relationship of beacon dead time to the received
interrogation pulse pair and the transmitted reply. In the example shown, the
dead time is 60 ps, for both the X and Y pulse codes, and is easily implemented
by inhibiting the decoder output. Thus, the receipt of a valid second pulse
during the dead time period will not result in a reply. Note that the actual dead
time may be longer than 60 p s if receiver blanking is utilized during ground
facility transmissions.
At sites where long delay multipath is present, such as in mountainous
terrain, additional dead time is introduced by the use of a retriggerable
blanking gate. This gate, which may be up to 250 p s in duration, is intended to
inhibit replies to interrogation echoes on the basis of an amplitude test. The
direct path interrogation triggers the blanking gate and sets a threshold level
to which subsequent interrogations are compared. Interrogations arriving
during the blanking time having amplitude less than the direct path signal are
ignored while those having higher amplitudes result in replies. The assumption
170 ROBERT J. KELLY AND DANNY R. CUSICK

1st 2nd X CHANNEL


PULSE PULSE

I 50pr REPLY DELAY REPLY


12us
MARGIN
10us

t
1st
t
2nd
PULSE PULSE

Y CHANNEL
1r t 2nd
PULSE PULSE

1 1
REPLY
56pr REPLY DELAY
30 us

t
I I I

t
1st 2nd
PULSE PULSE

REPLIES TO INTERROGATIONS WITH DECODES DURING THE


DEAD TIME PERIOD ARE INHIBITED

B 1011s MARGIN ALLOWS ADEQUATE TIME FOR THE COMPLETE


TRANSMISSION OF THE SECOND REPLY PULSE

Fic;. 62. Dead time definitions for [)ME N X and Ycodes.

is simply that interrogations that arrive shortly after a previous interrogation


and are of lesser amplitude are probably due to echoes. Replies to the echoes
would appear as an interrogation load and further degrade the facility
capacity.
The reply efficiency (RE) is a function of the interrogation load (A) and the
effective dead time (7)and is given by

That is, the reply efficiency is simply the “free time “-to-” total time” ratio for
a given interrogation load. It is a special case of the more general equation
[Eq. (36)] applicable to DME/P ground equipment shown in Section IV,B,S.
DISTANCE MEASURING EQUIPMENT IN AVIATION 171
As an example, consider 100 aircraft, each with 2 interrogators having an
average interrogation rate of 30 Hz. When a 60 p s dead time is assumed a 74%
reply efficiency results. This exceeds the minimum acceptable value of 70%
and implies that 100 aircraft can be readily served by a single ground station.
At sites where a 250 p s retriggerable blanking gate is utilized, a 40% reply
efficiency results under the same traffic loading conditions. The traffic load
must be reduced in order to keep the reply efficiency above 70%. Hence, such
sites will have less capacity than sites that do not utilize the retriggerable
blanking gate.

IV. SYSTEMS
CONSIDERATIONS
FOR THE NEWDME/P
INTERNATIONAL
STANDARD

The Precision Distance Measuring Equipment (DME/P) standard is


based upon the same principle as DME/N, which provides distance in-
formation in the aircraft by measuring total round-trip time between
interrogations from an airborne transmitter and replies from a ground
transponder. The difference is that, unlike DME/N, DME/P provides
accuracy and data rates that will support approach and landing operations in
conjunction with the MLS angle data. The DME/P functions have been
successfully integrated with those of DME/N and represent the newest
addition to the ICAO DME specification. Kelly (1984) provides a concise
summary of the DME/P concept adapted by ICAO. The material in this part
is an extension of that paper.
Subsection A serves to introduce the basic DME/P concepts by reviewing
the history of the DME/P standards development, reviewing the operational
requirements for precision range information, and describing the operation of
the DME/P system and its associated channel plan. Subsection B provides
detailed discussions of the DME/P technical characteristics and an analysis of
the major sources of performance degradation: noise, multipath, and garble.
Subsection C combines this information to provide system level error and
power budgets. Implementation considerations for both the ground and
airborne equipment are discussed in Subsection D, while Subsection E
summarizes the results of DME/P field tests.

A . An Overview of the D M E / P System

The purpose of this section is to introduce the basic DME/P system


concepts, beginning in Subsection A,1 with the history of the development of
the DME/P concept and the international standard.
172 ROBERT J. KELLY AND DANNY R. CUSICK

The first priority in approaching the task of defining a precision DME was
to establish the DME/P operational requirements (OR). The OR is simply the
definition of the user requirements that the DME/P must satisfy so that
measured range data in conjunction with the MLS angle data can lead to a
successful aircraft approach and landing. In addition, the DME/P OR had to
ensure interoperability with existing DME/N system. Subsection A,2 briefly
summarizes the OR, which contain an interpretation of the operational need
in terms of the range accuracy values which must be achieved in various
regions of the MLS coverage volume.
DME/P is a two-pulse/two-mode system. As in DME/N, pulse pairs are
utilized for interrogation and reply transmissions where the code or pulse
spacing identifies the operating channel. Two modes are provided: the initial
approach, or IA mode, which functions identically to the DME/N and the
final approach, or FA mode, which provides the accuracy required for the final
approach and landing operations. System operation and the rationale for its
selection is detailed in Subsection A,3.
The DME/P channel plan, described in Subsection A,4, is simply a subset
of the DME/N channel plan. Although the same interrogation/reply fre-
quency pairing is utilized, additional pulse codes have been defined to
accommodate the “hard pairing” with the MLS angle channels. The plan is
configured to handle estimated aircraft traffic densities well beyond the year
2000 by minimizing the potential for DME self-interference or garble.

1. Historical Background
The ICAO Seventh Air Navigation Conference, April, 1972, established
the All Weather Operations Panel (AWOP) as the designated body to develop
SARPs for a new nonvisual precision approach and landing guidance system
(MLS) (Redlien and Kelly, 1981). While several states included proposals for
precision DME as part of their submissions on MLS, AWOP determined that
its initial obligation was the preparation of SARPs for the angle and data
functions of the MLS. This decision recognized that the MLS could be
implemented, if necessary, with conventional distance measuring equipment
as specified in ICAO (1974),to satisfy early operational uses. The realization of
the full benefits of MLS, which required standardization of a precision DME
(DME/P) capability, was addressed accordingly as a second priority.
The eighth meeting of AWOP (AWOP-8) in March, 1980 (AWOP, 1980),
resulted in adoption of SARPs (ICAO, 1981a) by the panel for the angle
guidance portion of the MLS, which were accepted by the Communications
Divisional Meeting in Montreal, April, 1981 (ICAO, 1981b).
In conjunction with the final activity in developing the angle SARPs, the
panel began its consideration, in a very limited way, of DME/P SARPS in
DISTANCE MEASURING EQUIPMENT IN AVIATION 173
September, 1979, at its Working Group M (WG-M) meeting in Seattle.I3 At
this meeting, WG-M agred that the DME/P specificationscould be integrated
into the existing text of Annex 10. At the March, 1980, AWOP-8 meeting, it
was decided to review and update the DME/N standards where necessary.
It was during this meeting that the first “updated” DME/P concept was
offered in the form of a technical paper (Kelly and LaBerge, 1980). A complete
analysis and design was provided for range guidance for CTOL, STOL, and
VTOL operations. The paper presented, for the first time, a proposed accuracy
standard (including error budgets, power budgets, and the concept of system
efficiency) and the signal processing functions necessary to achieve the
proposed standard. The concept it described was a synthesis of ideas
developed and field tested by investigators under contract with the Federal
Aviation Administration in the early 1970s.Chronological details are given in
Kelly and LaBerge (1980).
Since the AWOP-8 meeting, DME/P became the principal activity of
AWOP. The chronology of events was as follows:
(1) Rio de Janeiro, Brazil, September, 1980, Working Group M-4 (WG-
M4) Meeting-Operational requirements for DME/P were developed and
the Traffic Loading/Channelization, Multipath, and System Concepts
Subgroups could, after a thorough examination of suggested techniques,
arrive at a recommendation to the WG-M concerning the best DME/P
concept.
(2) Amsterdam, The Netherlands, January, 1981, System Concepts
Subgroup Meeting-Power budget and error budgets were normalized to
permit the assessment of the various DME/P technique suggestions.
(3) London, England, February, 1981, Traffic Loading/Channelization
Subgroup Meeting-A standard traffic loading model was defined and its use
to predict uplink and downlink loading effects was agreed upon.
(4) Montreal, Canada, April, 1981, Informal WG-M meeting-WG-M
was briefed on the progress of the three subgroups.
( 5 ) Amsterdam, The Netherlands, August, 1981, Traffic Loading/
Channelization Subgroup Meeting-The channel plan was developed.
Ralph Vallone of the U.S. was the originator of the new DME Channel Plan.
(6) Munich, Federal Republic of Germany, September, 1981, Systems
Concepts Subgroup Meeting-DME/P interoperability with DME/N was
addressed. The original concept described in Kelly and LaBerge (1980)carried
with it an error component that would be incurred, if a DME/N interrogator
were to obtain distance service with a DME/P transponder. The error arose in
l 3 Membership in WG-M included Australia, Brazil, Canada, France, International Air
Transport Association (IATA), International Federation of Airline Pilots’ Associations
(IFALPA), Italy, Japan, The Netherlands, U.S.S.R., U.K., US., and W. Germany.
174 ROBERT J. KELLY AND DANNY R. CUSICK

this concept because it was thought that the ground transponder could not be
made to differentiate reliably between DME/N and DME/P interrogations,
and means to attempt such a distinction were intentionally omitted. Since the
pulse time of arrival decision threshold for the two applications must be
different, a bias error will be generated in mixed operations because, without
prior information, the ground transponder zero-range delay can only be
calibrated with respect to one or the other threshold.
Three new system concepts were informally submitted to the Concepts
Subgroup for evaluation, each with a unique method for resolving the
interoperability error. All concepts contained the notion of a low-threshold
precision mode and a - 6 dB threshold nonprecision mode. They were:
(a) A two-pulse system concept (France/United Kingdom) much like
the system proposed in Kelly and LaBerge (1980), except a pulse rise-time
recognition circuit in the transponder was proposed to differentiate between
DME/N and DME/P interrogators. The integrity of the pulse recognition
circuit became an issue.
(b) A three-pulse system concept (Federal Republic of Germany/United
States) using a conventional pulse pair followed 2 0 0 ~ slater by a single
precision mode pulse; DME/N and nonprecision DME/P range measure-
ments were to be made on the conventional DME pulse pair; precision range
measurements would be performed on the third pulse. This concept was
viewed as a large departure from the present DME system.
(c) A two-pulse/two-mode concept (France), which was the technique
unanimously chosen by the subgroup (Carel, 1981);Josef Hetyei of France was
the originator of this concept.
(d) Italy and Japan proposed alternative pulse-shaping implementa-
tions which are applicable to the two-pulse/two-mode technique.

(7) Paris, France, September, 1981, AWOP WG-MS-The Working


Group accepted the recommendation of the System Concepts Subgroup
and initiated the development of DME/P SARPs predicated on the two-
pulse/two-mode technique.
(8) Moscow, USSR, March, 1982, AWOP WG-M6-A first draft of the
DME/P Transponder SARPs was developed.
(9) Rome, Italy, September, 1982, AWOP WG-M7-The final draft of
the DME/P SARPs was developed by the group, including revisions to the
DME/N SARPs and the DME Guidance Material.
(10) Montreal, Canada, November, 1982, AWOP-9-DME/P Draft
SARPs approved by full AWOP-9 panel meeting.
With an approved SARPs in the wings, work began in earnest on the
formulation of airborne and ground equipment specifications. In the United
DISTANCE MEASURING EQUIPMENT IN AVIATION 175
States, a ground facility equipment specification (FAA, 1983C)for a complete
MLS system including DME/P was released in conjunction with a request
for proposal in April, 1983. This led to a contract award by the FAA in
January, 1984, to the Hazeltine Corporation and represents the first pro-
duction procurement of MLS equipment (208 systems) by a government
body. E-Systems of Salt Lake City, Utah, would build the DME/P ground
facility.
In the avionics arena, the Radio Technical Commission for Aeronautics
convened Special Committee 149, whose primary charter was to update the
existing DME Minimum Operational Performance Standards (RTCA, 1978)
to reflect important changes in the SARPs and to add the DME/P to the
avionics standard. Only after final completion of the DME/P MOPScould the
avionics industry truly begin the development, design, and marketing of
DME/P airborne equipment.

2. DME/P Operational Requirements


The DME/P is an integral part of the microwave landing system for
aircraft approach, landing, and missed approach operations. Thus, the
DME/P system must be capable of providing high-accuracy range in-
formation in a potentially severe multipath environment such as that
encountered during landing operations. The accuracy required (100 ft-30 m)
for such operations is at least an order of magnitude better than that provided
by the present conventional DME system (DME/N).
The operational requirements defined by ICAO state that the DME/P
must provide high-quality guidance to a range of 22 nmi from the MLS
ground facility and be fully compatible and interoperable with conventional
DME/N. This implies that
(1) the ICAO-defined (ICAO, 1974) radiated pulse spectral character-
istics, applicable to DME/N, must be satisfied;
(2) a conventional interrogator may be used with a DME/P transponder
to obtain at least the accuracy expected with a conventional transponder; and
(3) a DME/P interrogator may be used with a conventional transponder
to obtain at least the accuracy expected with a conventional transponder-
interrogator combination.
In addition, the OR states that, when operationally required, the DME/P
distance information shall permit, in combination with the appropriate MLS
angle and data information, the functions in Table XIV to be performed. The
two-pulse/two-mode technique selected by ICAO as the basis for DME/P
implementation satisfies these requirements. This technique is discussed in the
next section.
TABLE XIV
ACCURACY
NECFSARY TO SATISFY
DME/P OPERATIONAL (95% PROBABILITY)
REQUIREMENT

Distance from
Function ref. datum (nmi) PFE CMN

Segmented approach"
Extended runway centerline 20 f250m 68 m
At 40" azimuth +375 m 68 m
Segmented approachb
Extended runway centerline 5 k85 m 34 m
At 40" azimuth k127 M 34 m
Marker replacement
Outer marker 5 f800m NA
Middle marker 0.57 f400 m NA
30 m decision height' determination (100 ft)
3" Glide path (CTOL) 0.3 +30 m
- NA
6" Glide path (STOL) +l5m
- NA
Flare initiation over uneven terraind
3" Glide path (CTOL) 0 f30m 18 m
6" Glide path (STOL) +12m 12 m
Autopilot gain scheduling' 20 to 0 f250m NA
Flare maneuver-CTOLI with MLS flare 0 f30 m 12 m
elev.-STOL f12m 12 m
CTOL high-speede rollout/turnoffs Runway
region +12m 30 m
Departure climb and missed approach 0105 +100m
- 68 m
VTOL approachh 0.5 to 0 f12m 12 m
Coordinate translations' - f12to f30m 12m

a DME/P accuracy approximately corresponds to the azimuth function PFE at a distance of

20 nmi from the MLS reference datum both along the extended runway centerline and at an
azimuth angle of 40".Also the DME/N error at the limits of MLS coverage is consistent with the
DME/N 0.2-nmi system accuracy. The CMN is the linear equivalent of the 0.1" azimuth angle
CMN.
PFE corresponds to azimuth angular error; CMN is approximately the linear equivalent of
the 0.1" azimuth angle CMN.
Equivalent to radar altimeter performance for 3" glide path. The 30-m PFE corresponds to a
1.5-m vertical error for a 3" elevation angle.
Flare initiation begins in the vicinity of the MLS approach reference datum; MLS elevation
and DME provide vertical guidance for automatic landing when the terrain in front of the runway
threshold is uneven.
Sensitivity modification or autopilot gain scheduling requirements are not strongly
dependent on accuracy.
It is intended that this specification apply when vertical guidance and sink rate for
automatic landing are derived from the MLS flare elevation and the DME/P.
9 The rollout accuracy requirement reflects system growth potential. The requirement is to

optimize rollout deceleration and turnoff so as to decrease runway utilization time.


It is intended to assure the pilot that the aircraft is over the landing pad before descending.
'Figures in the table are typical of a VTOL application, where MLS coordinates are
translated to center of landing pad.
DISTANCE MEASURING EQUIPMENT IN AVIATION 177
Based on the accuracy requirements in Table XIV, two accuracy standards
were defined in the SARPs to accomodate a variety of applications in an
economical manner. Standard 1 is intended to satisfy the accuracy require-
ments of CTOL applications where “flare height” is obtained by radar
altimeter. Standard 2 is intended to satisfy the accuracy requirements of
STOL and VTOL applications as well as operations utilizing the MLS flare
function. For both standards, accuracy degradations with distance and
azimuth angle are specified. These accuracies, defined on a 95% probability
basis, are summarized in Table XV. No degradation is permitted with
elevation angle but a 1.5:l degradation is permitted in azimuth. Between the
lateral limits at f40”,the accuracy is permitted to degrade linearly with
increasing range.
3. The Two-PulselTwo-Mode System
The two-pulse/two-mode technique is the method selected by ICAO as the
basis of DME/P system implementation. It provides high accuracy in a final
approach mode (FA) and full interoperability with DME/N in an initial
approach mode (IA).
In the final phases of the approach and landing operation, the major
source of accuracy degradation is due to multipath rather than interrogator
and transponder instrumentation errors. Multipath perturbations, which
always occur later in time than the desired signal, are minimized by
thresholding the received pulse at a point which has not been significantly
corrupted by multipath. Therefore, the necessary accuracy for DME/P is
achieved by thresholding low (about 17 dB below the pulse peak) on the
leading edge of a fast rise-time pulse. This is in contrast to the 50% level
thresholding and slow rise-time pulses used in DME/N.
By achieving enhanced accuracy in this manner, questions are raised
concerning interoperability with DME/N. First, the use of a low threshold
level impacts transmitter power in the sense that more power is required to
obtain an adequate threshold-to-noise level throughout the coverage
v01ume.l~At the same time, a fast rise-time pulse is desired, which increases
the spectral width of the radiated pulse. These two requirements are in conflict
with the need to satisfy the adjacent channel spectrum requirements noted
earlier. Hence, both the pulse rise time and the maximum range at which the
low-threshold technique can be used must be restricted. Second, conventional
interrogators are calibrated assuming that transponder reply delay timing is
l4 A low threshold-to-noiseratio can cause pulse arrival-timemeasurement anomalies which
induce noisy range error measurements. At threshold-to-noise levels of 12 dB or more, these
effects are insignificant when, in addition, a pulse-width discriminatoris employed. When these
conditions are satisfied,the magnitudeof the receiver noise errors is simply inversely proportional
to the signal-to-noisevoltage ratio.
178 ROBERT J. KELLY AND DANNY R. CUSICK

TABLE XV
DME/P ERRORS
ALLOWABLE (95% PROBABILITY)'

Location Std. PFE CMN

20 to 5 nmi +250 m 68 m
from MLS approach reducing linearly to reducing linearly to
reference datum +_85m 34 m
5 nmi to MLS -f85m 18 m
approach reference reducing linearly to
datum -+30 m
k85 m
reducing linearly to
i12m
k100rn 34 m

At MLS approach +30m 18 m


reference datum
and through +12m
- 12 m
runway coverage
Throughout back
azimuth coverage 68 m
volume

68 m

" At distances from 5 nmi to the MLS approach reference datum and
throughout the back azimuth coverage the IA mode may be used when the F A mode
is not operating.

based on the 50% threshold level. However, transponder reply delay initiation
based on low-level thresholding of a conventional DME pulse will be different
from that based on the 50% points. Thus, if a conventional interrogator were
to interrogate a DME/P transponder which used only the low-level threshold
technique, an interoperability error (actually a bias) could occur.
The two-pulse/two-mode technique solves the interoperability problem
by providing two modes of operation, a wideband final approach mode
and a narrowband initial approach mode. The F A mode utilizes the low-
thresholding technique in both the interrogator and transponder for ranges to
7 nmi from the transponder. By limiting the FA mode coverage range to 7 nmi,
the adjacent channel radiated power constraints on the transponder can be
satisfied while an adequate threshold-to-noise ratio (for low-threshold pulse
detection) is maintained throughout this region. Beyond 7 nmi, the IA mode is
used, and its operation is identical to that of DME/N. In addition, a transition
between the two modes is provided in the 7-8 nmi region. Therefore, in the 1A
mode, no DME/N interoperability bias error occurs because both the
DISTANCE MEASURING EQUIPMENT IN AVIATION 179
interrogator and the transponder threshold on the 50% pulse levels. Further,
use of the 50% threshold allows an adequate threshold-to-noise ratio to be
maintained to the coverage limit of 22 nmi.
The interrogator selects the mode, F A or IA, according to the measured
range from the ground facility. The interrogation mode is characterized by the
pulse code (i.e., pulse-pair spacing). The transponder reply delay timing is then
based on the proper threshold point for the type of interrogation received.
Within a channel assignment, the transponder reply code remains the same
regardless of the type of interrogation. Note that both the DME/P and the
revised DME/N SARPs require range measurements on the first pulse,
thereby avoiding the effects of first pulse multipath echoes which can disturb
range measurements predicated on second pulse timing.
The operation of the DME/P system is summarized with a specific
example (an X-channel ground facility) in Fig. 63. An aircraft located 8 to
22 nmi from the transponder interrogates with the IA mode interrogation
code of 12 ps at a rate of 40 Hz for search and 16 Hz for track. In the 7-8 nmi
region, an “approaching” interrogator transmits in both IA and F A modes to
effect a smooth transition to the FA mode by 7 nmi. In the event that no FA
mode replies are received, the interrogator will remain in the IA mode with an
appropriate warning to the pilot and aircraft systems utilizing the range data.

FIG.63. DMEjP two-pulse/two-mode system operation using X-channel codes.


180 ROBERT J. KELLY AND DANNY R. CUSICK

Within 7 nmi of a DME/P transponder, the aircraft obtains precision range


information with FA mode interrogations at 40 Hz. These interrogations
are identified by the X-channel code of 18 ps. Receipt of this code at the
transponder causes the incoming waveform to be thresholded low.
The FA mode replies use the same code as the IA mode replies: 12 ps for the
X-channel example. In the FA mode, both the interrogator and the
transponder process received signals in a 3.5-4.5 MHz bandwidth. The wide
bandwidth is required, for reasons to be noted in Subsection B, to ensure that
the pulse leading edge is essentially linear in the threshold region.
As was discussed in Section II,E, aircraft control requires on the order of
five data samples each second. Clearly, with an FA mode interrogation rate of
40 Hz, more than enough data are available. Thus, raw range measurements
can be filtered to reduce the effects of noise, multipath, instrumentation-
induced measurement errors, and to compensate for mising data.
DME/P IA and F A mode functions can be included in an interrogator
package also having the conventional DME/N capability. On the other hand,
some users may elect to implement only the IA mode such that the
combination would yield, in effect, an enhanced DME/N having increased
accuracy and all of the additional DME/P channels.

4. DMEfP Channel Plan


The channelization scheme for DMEfP makes use of the same
frequency/code concept used in the TACAN/DME channel plan detailed in
Section III,C,3 and summarized in Fig. 60. The major difference is the use of
additional pulse codes necessitated by (1) the need to uniquely identify IA and
F A mode interrogations and (2) the need to accommodate “hard pairing” of
DME/P with the 200 MLS angle channels.
The DME/P-related additions to the existing channel plan are such that
they provide an optimum pairing between MLS C-band angle and L-band
range channels. They do not interfere with present DME services while
minimizing the potential for DME garble interference. A major concern in the
development of the channel plan was the traffic loading effects of candidate
channel plans on pulse garble interference, which is directly proportional to
the number of pulse-code-multiplexed channels that are defined on each
frequency. The selected channel plan minimizes these interference effects by
spreading the number of channels over a large number of frequencies without
incurring any meaningful degradation of existing or planned DME/N enroute
services. As a result, the channel plan can accomodate channel assignments in
extremely dense environments such as those postulated for the Northeast and
Southwest United States. Furthermore, the selected channel assignment plan
was assessed against a worst-case traffic loading environment.This assessment
DISTANCE MEASURING EQUIPMENT IN AVIATION 181
assumed that the system efficiency would never be less than SO%, given an F A
mode transponder reply efficiency of 80%. Another consideration was to pair
the uplink codes with the uplink frequencies so that all codes are spaced at
least 12 ps apart. Existing DME/N interrogators with f 6 ps decoder aper-
tures can therefore successfully operate without interference from the new
channels.
Associated with each DME/P channel are two interrogation codes and a
single beacon reply code. The two interrogation codes allow the transponder
to discriminate between IA and FA mode interrogations as described in the
previous section. Four codes are defined: W, X, Y, and Z. Interrogation/reply
frequencies for DME/P X and Y channels utilize the same pairings as the
TACAN/DME channel plan. Additionally, the IA mode interrogation and
reply codes are identical to those used in the TACAN/DME channel plan,
thereby ensuring interoperability of TACAN/DME equipment with
DME/P equipment. Interrogation/reply frequencies associated with DME/P
Z channels utilize the same frequency pairs as the Y channels, while the W
channel frequency pairs correspond to the X channel structure. The channel
plan structure is summarized in Table XVI.
All of the DME/P channels are “hard paired” with the 200 MLS angle
channels,as summarized in Table XVII. The angle channels are in the C-band
portion of the radio spectrum between 5031 and 5091 MHz in 300 kHz
increments and are numerically designated 500-699. Note that the even
channels 18-56 X and Y correspond to the TACAN/DME channels that are

TABLE XVI
DME/P CHANNELPLAN

Pulse pair
Number spacing (ps)
Channel Pulse of Operating Reply
designation code channels mode Inter Reply delay (ps)

18-56 even WE 20 IA 24 24 50
FA 30 24 56
18-56 even X b 20 IA 12 12 50
FA 18 12 56
17-56/80-119 Yb 80 IA 36 30 56
FA 42 30 62
17-56/80-119 Z‘ 80 IA 21 15 56
FA 27 15 62

‘W channels are pulse-code multiplexed on the X frequencies.


DME/N X and Y channels are fully compatible with DME/P X and Y IA mode
channels.
Z channels are pulse-code multiplexed on the Y frequencies.
182 ROBERT J. KELLY AND DANNY R. CUSICK

TABLE XVII
DMEjP MLS CHANNELPAIRING

DME/P Channels MLS Channels


200 channels 200 channels

18-56 X and W even 500-699


17-56 Y and Z 5031.0-5090.7 MHz, 300 kHz spacing
80-119 Y and 2
18-56 X and W even Channels 500-539
(40 channels) 5031.0-5042.7 MHz
(i.e., 18X with 500, 18W with 501,. ..)
17-56 Y and Z Channels 540-619
(80 channels) 5043.6-5066.7 MHz
(i.e., 17Y with 540, 1 7 2 with 541 ,...)
80-119 Y and Z Channels 620-699
(80 channels) 5067.0-5090.7 MHz
(k, 80Y with 620, 802 with 621 ,...)

“hard paired” with the 40 ILS channels. This correspondence is intended to


accommodate a single DME/P at joint ILS-MLS sites while maintaining the
“hard pairing” of DME with ILS and DME/P with MLS.

B. Pulse Shape, Time-of-Arrival Determination, and


the Major Sources of Peformance Degradation

This section details the technical characteristics of the DME/P (pulse


shape and time-of-arrival estimation) that are the keys to obtaining the desired
ranging accuracy in the presence of noise, multipath, and garble.
A new pulse waveform, discussed in Subsection B,1, was defined to
accommodate the accuracy specification while at the same time satisfy the
existing DME/N adjacent channel interference specification. Subsection B,2
describes the method of pulse time-of-arrival determination. It takes advan-
tage of the new pulse shape to achieve accuracy in the presence of severe
multipath.
Three sources of error in DME/P which must be accommodated in order
to achieve the desired performance are receiver noise, multipath, and garble.
Effects of receiver noise are primarily determined by the received signal level
(i.e., signal-to-noise ratio) and the receiver signal processing. These effects are
detailed in Subsection B,3. Multipath, the single most important threat to
DME/P accuracy performance, is treated Subsection B,4. It is shown that the
multipath effects are controlled by selecting the critical parameters associated
DISTANCE MEASURING EQUIPMENT IN AVIATION 183
with pulse rise time and time-of-arrival technique so that they match the
runway geometry and obstacle clearance zones. Garble, the subject of Sub-
section B,5, degrades two major performance parameters: system efficiency
and accuracy. Unlike DME/N, garble becomes a critical issue in DME/P due
to the use of wide bandwidth receivers, and the multitude of DME equipments
operating simultaneously in the environment. System efficiency is a new
DME/P concept which influences many aspects of the system design,
specifically the interrogator data filter. Accuracy degradation occurs due to
coincidence with extraneous pulses. Methods of determining the level of
accuracy degradation which occurs for a given garble environment are
described.

1. DME/P Pulse Shape


Three factors govern the selection of the DME/P pulse:
(1) The waveform must be compatible with DME/N ground and
airborne equipment.
(2) The pulse must have a fast rise time leading edge in order to provide
timing measurements of the required accuracy in the presence of noise,
multipath, and garble.
(3) The DME/P waveform must satisfy the same spectrum constraints
imposed on the DME/N waveform as noted in Section 111 of this article.
ICAO (1985)provides a template for the DME/P waveform where the key
parameters are the 10-90% rise time, 3 amplitude width, partial rise time,
and partial rise time linearity. The same specifications apply to both the
interrogator and transponder equipments. Although it is noted below that the
IA and F A mode requirements differ, only a single waveform satisfying the FA
mode specifications need be generated, thereby simplifying modulator design.
The IA mode pulse is specified to have a 10-90% rise time 1600 ns or less
and a nominal width as measured at the $-amplitude points of the leading and
trailingedges of 3.5 ps. [RTCA (1985b) specifies that the rise time be I1200 ns
for the reasons noted in Subsection D,2. The 1200ns maximum rise time
specification will be used in this article.] These requirements also satisfy the
DME/N waveform specifications, thereby maintaining interoperability be-
tween the conventional and precision DME systems.
Three additional requirements, detailed in Fig. 64, are imposed on the FA
mode waveform. First, the 5-30% leading edge rise time, referred to as the
partial rise time, is specified to be 200-300 ns. This segment of the pulse
leading edge is where time-of-arrival measurements are made when operating
in the FA mode. The specification is key to achieving high accuracy in the
presence of multipath as discussed in Subsection B,4.
184 ROBERT J. KELLY A N D DANNY R. CUSICK

28--

L I M I T I +20%OF THE
24-- MEASURED PARTIAL
RISE TIME SLOPE)
22--

-
-ae 2 0 - -
Y

RISE TIME SLOPE1

------ BOUNDARY FOR PULSE


TURN-ON TRANSIENTS 8 -- RISE TIME SLOPE
-me- MEASURED PULSE

o r - d --------
1%----
r :

VIRTUAL ORIGIN
L I M I T 1OOOnr

I 1 I I I I
0 100 200 300 400 500
PARTIAL RISE TIME Ins)

FIG.64. The DME/P pulse: FA mode requirements and the cos/coszwaveform.


DISTANCE MEASURING EQUIPMENT IN AVIATION 185
Second, a linearity requirement is imposed on the partial rise time segment
of the pulse leading edge and requires that the instantaneous slope at all points
in this segment may not deviate by more than 20% from the slope of a straight
line passing through the 5% and 30% points.15 This ensures that the virtual
origin of the pulse can be accurately determined (assuming the use of the
delay-attenuate-compare (DAC) time-of-arrival estimator, as described in
Subsection B,2) and that calibration of the DME/P airborne and ground
equipment is independent of DAC parameters and pulse rise times.
Finally, the instantaneous magnitude of any pulse turn-on transient that
occurs prior to the virtual origin of the transmitted pulse must not exceed 1%
of the peak pulse amplitude. In addition, any such transient must not occur
more than 1OOOns prior to the virtual origin. The reason for such a
specification is as follows. First, a transient signal prior to the pulse can
prevent a pulse from being detected by the DAC circuit. Limiting the
amplitude of the turn-on transient eliminates this possibility. Second, such
transients can be declared pulse arrivals by the DAC. Limiting the duration of
the transient to a maximum of lo00 ns allows the resulting false pulse to be
discarded by a pulse-width discriminator.
Wide-band signal processing is necessary to preserve the shape of the fast
rise time DME/P pulse and therefore achieve the up- and downlink accuracy.
Thus, unlike DME/N, Ferris Discriminators or their equivalent are required
in both the interrogator and the ground transponder to provide the means of
identifying and rejecting off-frequency DME pulses.
As in DME/N, the downlink spectrum requirements are relative and are
easily satisfied.The most severe restriction is on the uplink, where the radiated
power is limited to absolute levels in the adjacent channels. As noted in
Section 111,B,2,any ERP can be utilized by a ground facility provided that the
radiated power is limited to 200 mW in the first adjacent frequency channel
and 2 mW in the second. Thus, a fast rise time pulse (as required in the
DME/P application) having greater spectral width than the conventional
pulse requires that the ERP must be reduced such that the stated limita-
tions on adjacent frequency radiated power are satisfied. On the other hand,
the radiated signal power must result in a threshold-to-noise ratio throughout
the coverage volume such that the effects of receiver noise do not adversely im-
pact ranging accuracy. These conflicting goals, fast rise time versus coverage,
are a major reason for the inclusion of two modes of operation (IA and FA)
into the DME/P signal format.
A pulse shape which satisfies the requirements described above is the
cos/coszpulse. Figure 64 depicts this pulse for partial rise times at the extremes

l 5 The intersection of the line connecting the 5% and 30% points of the pulse leading edge
with the time axis is known as the virtual origin.
186 ROBERT J. KELLY AND DANNY R. CUSICK

of the allowable limits. The pulse consists of a leading edge which is a portion
of a cycle from a sinusoidal function, a trailing edge which is a portion of a
cycle from a sinusoidal function which has been squared, and stretched at the
peak to provide the desired $-amplitude width.
The characteristics of this pulse, for the range of allowable partial rise
times, are summarized in Table XVIII. In addition, the maximum uplink ERP
which can be used based on the adjacent channel specifications is defined.
Note that the ERP limitations for the first and second adjacent channels are
nearly equivalent. In this sense, the cos/cos2 pulse is nearly optimum.

2. Pulse Time-of-Arrival Estimation Technique


To measure range in the interrogator or to begin the reply delay timing in
the transponder, the DME pulse must be detected and its time of arrival
precisely determined. Envelope detection is used in DME; the phase
information is discarded.
Methods for estimating the pulse time of arrival (TOA) must satisfy the
accuracy specifications appropriate to the aircraft class subject to power

TABLE XVIII
SUMMARY
OF cos/cos2 ~ L S ECHARACTERISTICS

5-30% partial rise time (ns) 200 250 300


Maximum deviation from linearity 5% 5% 5%
10-90% rise time (ns) 800 1000 1200
90-10% fall time (ns) 3000 2919 2677
f amplitude width (ns) 3500 3500 3500
5%-5% pulse duration-t(ns) 5800 5800 5800
Equivalent energy rectangular
pulse duration (terr) 2650 2615 2615
Relative first adjacent channel
energy (W,lW, dB) -28.4 -31.9 - 34.5
Relative second adjacent channel
energy W21WdB) -48.6 -50.1 -51.9
Max ERP" allowed by first adjacent
channel limit (dB m) 54.9 58.4 60.9
Max ERP allowed by second adjacent
channel limit (dB m) 55.0 56.6 58.4

a Computation of maximum ERP: (W/w)(f/ferr)p&, where =


energy in ith adjacent channel, i = 1, 2; W = total energy in pulse;
p:,,, = maximum average power permitted in ith adjacent channel;
t = 5%-5% pulse duration; and ref( = duration of equivalent energy
rectangular pulse.
DISTANCE MEASURING EQUIPMENT IN AVIATION 187
budget considerations. Trade-offs are then necessary for satisfactory perfor-
mance taking the following into account: (1) noise, (2) lateral specular
multipath, (3) signal amplitude and rise-time variations, (4) threshold-to-noise
anomalies, and ( 5 ) circuit parameter variations.
In general, emphasis on noise performance is counterproductive to good
multipath resistance. Matched filter designs are not necessary because, as will
be shown in Subsection B,3, the system noise performance when wide-
bandwidth signal processing is utilized is well within the system error budgets
given in Subsection C,2. Therefore, wide-bandwidth signal processing is
preferred so as to obtain the necessary multipath discrimination. This means,
given the constraints described earlier, the TOA measurement must be made
on the pulse leading edge. Since the operating environment produces high
levels of ground multipath interference, the leading edge threshold technique
must not be sensitive to signal-strength-dependent errors.
A technique which satisfies these conditions for the FA mode is the delay-
attenuate-compare circuit. This circuit compares a delayed version of the
pulse to an attenuated version of the same pulse. Pulse arrival is declared when
the delayed signal exceeds the attenuated signal. A delay of 100 ns and an
attenuation of 5 to 6 dB will result in a nominal threshold level 15 to 18 dB
below the pulse peak when a 1000 ns rise time cos/cos2 pulse is utilized. These
values represent a compromise between good multipath performance and the
avoidance of noise anomalies. In any case, the delay and compare parameters
must always be chosen such that the threshold crossing occurs during the
partial rise time; that portion of the received pulse between the 5 and 30%
amplitude levels having guaranteed rise time and linearity. The importance of
these factors will become evident below.
The key characteristic of the DAC is that for a linear leading edge pulse,
the threshold time relative to the start of the pulse is independent of the pulse
rise time and amplitude. It is for this reason that the DME/P precision pulse
shape specification requires that the pulse leading edge be linear in the
threshold region. Thus rise-time variations within the SARPs specification
and amplitude variations that result from signal power and propagation
effects do not induce TOA estimation errors when a DAC processor is used.
General design relations for DAC are given in Fig. 65, as well as an example
design.
Just as in DME/N, pulse detection in the IA mode is based on the 50%
( - 6 dB) pulse amplitude level using a peak-amplitude-find (PAF) circuit
(sometimes called a +-amplitude find circuit). In this technique the pulse is
delayed (e.g., 2000 ns) until the peak is found and a threshold, 6 dB down from
the peak, is generated (Fig. 66). Since multipath is generally not a critical factor
within the IA mode service volume, the higher threshold level provides a
188 ROBERT J. KELLY AND DANNY R. CUSICK

I DECODE TIME ITD) FOR


ARBITRARY PULSE
I TD:p(T~-7)'Ap(T~)

= 0 : p (1- 7 <Ap It1


DAC OUTPUT (LOGIC SIGNAL)
= 1;p (t- 7 ) > A p (t)

DEFINITION OF LINEAR
LEADING EDGE PULSE P It) = { :iR<yo<t<TR

1
DECODE TIME FOR LINEAR
LEADING EDGE PULSE 1 'D=&

RELATIVE THRESHOLD OF ATTENUATED T


LINEAR LEADING EDGE PULSE (1-A) TR

I RELATIVE THRESHOLD OF DELAYED


LINEAR LEADING EDGE PULSE
AT

0 50 100 150 200 250 300 350 400 450 500


TIME (nr)

PULSE LEADING EDGE SHOWN I S LINEAR WITH A SLOPE CORRESPONDING TO


THE 200 ns 0.05-0.30 RISE TIME SPECIFICATION. THIS CORRESPONDS TO A N
800 nr TOTAL RISE TIME TR .
7 = DAC DELAY TIME, 100 ns
TD = DAC DECODE TIME, 200 ns = 7 /1-A

RELATIVE THRESHOLD LEVEL OF ATTENUATED PULSE = TII1-A) TR = 0 . 2 5


RELATIVE THRESHOLD OF DELAYED PULSE = A T / I l - A ) TR = 0.125

FIG. 65. Delay-attenuate-compare design relations


DISTANCE MEASURING EQUIPMENT IN AVIATION 189

DELAY +
MAX
ATTENUATE
-
4
COMPARE
4 t
J

FIG.66. The +-amplitude find circuit.

satisfactory means to obtain greater range coverage than the low-level


threshold technique required in the F A mode. Most importantly, the
interoperability error is eliminated because the &amplitude level is the same
calibration point used in DME/N.
a. Transponder Calibration. Calibration of the IA mode of the DME/P
transponder is identical to the procedure used for DME/N described in
Section III,C,l. Although the situation is more complex, the calibration of the
FA mode of the DME/P transponder using the DAC time-of-arrival circuit is
amazingly straightforward for waveforms which are linear in the partial rise
time region. To appreciate this, note that the SARPs define the transponder
delay using the notion of the “virtual origin.” The virtual origin is the point at
which the straight line through the 30% and 5% amplitude points on the pulse
leading edge intersects the zero amplitude axis. Except for a constant offset
given by the DAC detection time equation, the DAC declares the arrival of the
pulse virtual origin. Thus, the delay between the interrogation pulse and the
reply pulse for a DAC detection circuit is the time interval between the virtual
origins of the two pulses. Remarkably, the DAC parameters in the inter-
rogator, transponder, and the external test sets which calibrate them can all
have different values, and the round-trip delay will be measured correctly. The
measured delay will be correct even if uplink and downlink pulse rise times are
different (Fig. 67). All this is a consequence of the linear partial rise time
assumption and is the motivation for the partial rise time linearity speci-
fication in the SARPs. For any combination of DAC parameters in the range
of 100 ns f 10% delay and 6 L 1 dB attenuation, the nonlinearity error for a
+20% slope variation is less than 10 ft (3 m). It should be emphasized that
other F A mode TOA measurement techniques are not excluded as long as
their equipment calibration procedures satisfy the virtual origin concept.
190 ROBERT J. KELLY AND DANNY R. CUSICK

DAC DECISION TIME


DAC DECISION TIME ROUND TRIP DELAY = T l i T + T D ~ ; n y + T R i p L Y

DAC DECISION TIME ROUND TRIP DELAY = ITlNT - TDI + TDC) I TDELAy + ITREpLy - To2 t TD,l

DAC DECISION TIME ROUND TRIP DELAY = TINT + TDELAY + TREpLy

DAC DECISION TIME ROUND TRIP DELAY = VIRTUAL ORIGIN ROUND TRIP DELAY

FIG.67. Calibration of the DMEIP.

3. Receiver Noise
Receiver noise, which perturbs the leading edge of the received DME
pulse, is a source of pulse time-of-arrival errors that translates into range
errors at the interrogator. These errors are considered part of the ground and
airborne instrumentation errors in that the minimum received signal levels
must be accommodated within the equipment instrumentation error budgets.
The purpose of this section is to define the effects of receiver noise on DME/P
range measurement accuracy.
The thermal noise component in the receiver IF is given by
IFnoise= kTB (32)
where kT =4 x mW/Hz ( T = 290 K) and B = IF noise bandwidth
(W.
The signal-to-noise ratio in the receiver video (i.e., ratio of the peak signal
to the rms value of the video noise) is then
S/N(video) = S/N(IF)
IF noise bandwidth
+ lolog video noise bandwidth
- (receiver noise figure) (33)

The receiver noise figure is typically assumed to be 6 dB for DME/P ground


DISTANCE MEASURING EQUIPMENT IN AVIATION 191
equipment and 9 dB for the less expensive airborne receiver (see Subsection
C,3 on power budgets).
The FA mode of the DME/P utilizes the delay-attenuate-compare
circuit to determine the time of arrival of a received pulse. A pulse width
discriminator is utilized to discard pulses in the receiver video that are due to
noise and have insufficient width to be considered valid DME pulses.
Discarding pulses of less than 2 ps will result in satisfactory discrimination.
Figure 68 summarizes the performance of the DAC circuit over a wide
range of signal-to-noise ratios for two I F bandwidths (3.5 and 4.5 MHz) and
two different DAC attenuation parameters (5 and 6 dB). The time-of-arrival
errors increase rapidly as the SNR drops below 15 dB. Such behavior is
expected since 15-17 dB is approximately the level at which the received pulse
is thresholded by the attenuated signal. Similarly, the probability of pulse
detection decreases rapidly as the SNR drops below 10 dB. Note that the reply
efficiency would drop off as the square of the detection probability, since both
pulses of the pulse pair must be received by the ground (or airborne) unit for a
valid range measurement to be made. The following fact is evident from the
plots: The signal levels required to reduce the noise-induced time-of-arrival
errors to an acceptable level exceed those required to obtain an adequate reply
efficiency. This is in contrast to DME/N, where the accuracy requirements are
readily satisfied for signal levels that provide 90% or greater reply efficiency.
Figure 69 shows the PFE and CMN components of the noise-induced
time-of-arrival errors for SNR values that are typical of those encountered in
operation. Contrary to intuition, the performance improves when the I F
bandwidth is increased from 3.5 to 4.5 MHz. This is attributed to the fact that
the wider bandwidth allows for more rapid signal transitions. Hence, pulse
edges, including noise pulses, are better defined and the desired pulse leading
edge becomes easier to distinguish from the noise pulses particularly when
a noise pulse occurs just before the valid pulse leading edge. Note that the
1 MHz increase in the IF bandwidth results in only a 1 dB decrease in video
SNR.
For both DME/N and the IA mode of DME/P, pulse time-of-arrival
determination is accomplished with the $amplitude find circuit. Figure 70
summarizes the performance of the +-amplitude find for an I F bandwidth of
800 kHz and a 2 p s pulse width discriminator. As in the F A mode case, the
time-of-arrival errors increase rapidly as the SNR drops. Similarly, the
pulse detectability decreases with decreasing SNR. For SNRs greater than
30 dB, the range errors are well accommodated by the instrumentation error
budgets, as will be shown in Subsection C,2. Note that at the point where
the reply efficiency begins to drop, the range errors are on the order of
100 ft and hence are negligible when compared to DME/N range accuracy
requirements.
192 ROBERT J. KELLY AND DANNY R. CUSICK

840

5 700

I"
c

0600
e
f m
-: -
LL

E"
a
w 100

0
0 10 20 30 40 50 60
PEAK VIDEO-TO-RYSNOISE RATIO (dB)

- - KEY IF BANDWIDTH
3.5 mnz
DELAY
1Wnr
ATTENUATION
5 dB
-.---
----
....... ......
3.5 mnz
4.5 mnz
100 nr
100 nr
6 dB
5 d0
~

4 . 5 ~ ~ 100 M 6 dB

FIG.68. Noise performance of delay-attenuate-compare time-of-arrival detection circuit.


A 2 ps pulse-width discriminatorwas utilized.
CONTROL MOTION NOISE PATH FOLLOWING ERROR
(FEET-=% PROBABILITY) (FEET-%% PROBABILITY)
o s l o g g g

L
\o
w
1wl- 500
-z 90 - - 450
6
E - Gi 400
-
80
$0 70 350
w
v) 6 0 - 8300

2 50- P250
P
B 40- 200
> U
f 30- z 1 5 0
=8 20-
g,,
gn 10- 50
0 - 0

-40 +
t
t
i
m
a
$30
E \

-‘\.
‘FWN
--\-
--\
----------__
FIG.70. Noise performance of the f-amplitude find time-of-arrival detection circuit for an
IF bandwidth at 800 kHz and a 2 p pulse-width discriminator.

4. Multipath
The single most important DME/P approach and landing performance
index is the system’s resistance to pulse leading edge distortion by multipath.
Without a multipath solution, the DME/P would not be viable. Interestingly
enough, using the solution of low thresholding on the pulse leading edge, the
multipath error component can be constrained to a value comparable to the
equipment instrumentation errors. Multipath was not a concern with the
DME/N because the accuracy allowance is 10 to 50 times that allowed for the
DME/P. Moreover, for the enroute application, the multipath has a long
delay time (which affects principally the second pulse) rather than the short
DISTANCE MEASURING EQUIPMENT IN AVIATION 195
delay multipath (less than 1 p s ) encountered in the approach and landing
environment.

a. Specular Multipath ( F A Mode). Specular multipath is produced by


scatterers which are sufficiently large (e.g., a substantial fraction of a Fresnel
zone) to yield a reflection level within about 10 dB of the direct signal. Example
sources include sizable buildings, wide-body aircraft, and vehicles near the
ground facility.
Specular multipath is characterized by four parameters: the relative
strength of the reflected to the direct signal (MID ratio), the relative delay of
the reflected to the direct signal, the relative rf phase of the reflected and direct
signals, and the rate of change of relative phase (scalloping frequency). Given
these parameters and the pulse time-of-arrival estimation technique, the
specular multipath effects in the vicinity of the runway threshold can be
calculated. Figures 71 and 72 estimate the multipath performance when a
DAC is utilized to determine pulse time of arrival. The results are consistent
with the accuracy specification (Subsection C,2) for the FA mode.
The ground and airborne pulse arrival time measurement circuits are
designed to prevent large specular multipath errors along the extended
runway center line. This is accomplished in a manner similar to the angle
components of the MLS (Redlien and Kelly, 1981). That is, the pulse arrival
time decision point is chosen to be less than the shortest multipath delay time
associated with large obstacles around the runway. The multipath delay times
are determined by the runway obstacle clearance surfaces. Obstacles such as
hangars, which are located so that they conform to the obstacle clearance
surfaces, typically generate large multipath reflections, having delays in excess
of 300 ns. Figure 72 is a plot of constant-delay ellipses about the runway. Such
plots can be used to estimate the multipath delay that results from an obstacle
at a given position relative to the aircraft, runway, and ground facility.
Computer simulations and experimental measurements have shown that
representative signal processors (such as the DAC with a 3.5-4.5 MHz
bandwidth) will not be susceptible to multipath with delays in excess of 300 ns,
as noted in Fig. 71. Multipath environment studies and field measurements
(Evans, 1982; Evans and Swett, 1981) suggest that high-level multipath with
delay less than 300 ns will occur at less than 20% of the runways. In addition, if
the multipath delays are less than 300 ns, most of these reflections will be less
than -3 dB. Under this -3 dB multipath assumption, the errors will be
bounded at about 33 ft (10 m) for DAC parameters of 100 ns delay and 5 to
6 dB attenuation values. This error component was included in the system
error budget (Subsection C,2) proposed for the 100 ft (30 m) system accuracy
standard.
- . - - - -
.
100.0 - 1 - 1 - 1 - ~ . . - . -

- 80.0
60.0 -
'
NO IFFILTER PULSE SHAPE, COS/COs
RISE TIME: 800 NS

.
v)
2 DAC DELAY: 100 NS
DAC AlTENUATION: -6dB
MID: -6 dB

g -60.0 -
-80.0 .
-1-0 ' . ' ' I * I 1 I - ' 1 * ' -

MULTIPATH ERRORS WITH N O IF FILTERING

DELAY PEAK ERROR

TIME DELAY FOR


PEAK NEG. ERROR

TIME DELAY FOR


PEAK POS ERROR

TIME DELAY FOR NO ERROR T


0
(DAC DECODE TIME) TD- I-A
p = MULTIPATH-TO-DIRECTAMPLITUDE RATIO
-
A = DAC ATTENUATION
7 DAC DELAY
TD = DAC DECODE TIME

NOTE:
THESE EQUATIONS DO NOT APPLY WHEN IF FILTERING IS USED.
THE ERROR CURVES ARE MODIFIED AS SHOWN ABOVE.

FIG.71. Multipath errors as a function of delay and phase for DAC processing of cos/cosz
pulse.

The spatial frequency content of the multipath interference field con-


tributes to the CMN component of system error. As the aircraft moves
through the interference field, the relative phase of the direct and reflected
signals varies at a rate (scalloping frequency) dependent upon the geometry
and the aircraft velocity. Hence, the range errors are time varying and
therefore contribute to the CMN component of the system error. The high-
frequency components (> 10 rad/second) can be reduced by the interrogator
data filter (see Subsection D), while nothing can be done to attenuate those
DISTANCE MEASURING EQUIPMENT IN AVIATION 197
SCALLOPING FREQUENCIES
LESS THAN 10 RAD/SEC
FOR REFLECTORS WITHIN
THESE LIMITS
2500
2ooo
Y
f
=r 1000
zc 500
o w
gf 0
z'c -500
Y
0 -1000
2 -1500
5 -2000
-2500
-1250 0 1250 2500 3750 5OOO 6250 7600 8750 loo00 1126
DISTANCE ALONG CENTERLINE I N FEET
NOTE:
SCALLOPING FREQUENCY LIMITS ASSUME 200 FT/SEC
AIRCRAFT VELOCITY ALONG RUNWAY CENTERLINE

FIG.72. Contours of constant multipath delay for aircraft-to-transponderdistance of


10,oOo ft.

components of the error signal falling within the data filter bandwidth. Fig-
ure 72 illustrates, for the given geometry, the region where the multipath
error components will be filtered. Kelly and LaBerge (1980) discusses addi-
tional multipath control techniques which are applicable to those few sites
where multipath exceeds - 3 dB in the region where the scalloping frequency
is less than 10 rad/second.
b. Specular Multipath (ZA Mode). Multipath performance for the IA mode
may be different from that encountered in the FA mode because the TOA
measurement threshold is much later on the pulse and thus the system is more
sensitive to long-delay multipath (Fig. 73). The critical range is 7 nmi because
the error allowance is most stringent at that range. Fortunately, however, the
approach geometry is such that the aircraft is above most of the severe
multipath. Studies indicate that at most runways along the runway center line
extended (< 20"), the MID is less than - 6 dB. More severe conditions may
exist on the wide radials (30"-40"). For these rare cases, special multipath
control measures are available which will minimize the multipath interference
at wide angles (Kelly and LaBerge, 1980).
c. Difuse Multipath (FA and IA Modes). Diffuse multipath is produced by a
"collection" of small objects (i.e., small relative to the Fresnel zone radius in at
least one dimension) which individually yield very low level reflections (e.g.,
198 ROBERT J. KELLY AND DANNY R. CUSICK

-160.0
@=18o,MID=-6dB
-240.0

-320.0
-400.0 - " l ' l . ' . l ' l . ' * l * l .
0.0 300.0 600.0 900.0 1200.0 1500.0 1800.0 2100.0 2400.0 2700.0 3OW.O

-
DELAY (nr)
6 RELATIVE PHASE OF REFLECTED SIGNAL
MID = RATIO OF REFLECTED SIGNAL LEVEL TO
DIRECT SIGNAL LEVEL

FIG.73. IA mode multipath performance for a 1200 ns rise time cos/cosz pulse.

20 dB below the direct signal level), but which may be significant as a group
due to the large number and disposition of such scatterers. Examples include
boulders, posts, approach lights, ocean waves, trees, and small buildings or
aircraft. As viewed by the receiver, diffuse multipath in flight (test) error traces
appears noiselike and will be reduced by the interrogator output filter.
Analytical and experimental work (Evans, 1982; Evans and Swett, 1981)
indicates that the diffuse multipath component will be satisfactorily bounded
by the error budget allocations. Flight test traces have shown diffuse multipath
error components to be less than 10 ft (3 m) (Edwards, 1981).

5 . Garble-System Eficiency and Accuracy Eflects


The DME/P system accuracy requirements must be satisfied in the
presence of pulses other than just the desired interrogation and reply pulse
pairs. The interference caused by the extraneous pulse environment is referred
to as garble and, in contrast to DME/N, is an important system consideration
in DME/P. There are two sources of garble interference of interest, DME and
non-DME. DME garble is simply due to the effects of other DME equipment
operating at or near the same frequency in the surrounding environment.
Non-DME components are, for example, those caused by other L-band
systems such as the ATC Radar Beacon System and the Joint Tactical
Information Distribution System. The effects of these non-DME garble
sources are beyond the scope of the present discussion.
The effect of the DME extraneous pulse environment is twofold, involving
both accuracy and system efficiency. Accuracy degradation occurs when an
DISTANCE MEASURING EQUIPMENT IN AVIATION 199
undesired pulse is coincident with a desired pulse, resulting in a distorted pulse
leading edge. Such coincidence can occur on both the up- and downlinks. The
occurrence of garble-induced range errors is a probabilistic event and
primarily adds to the CMN component of the system error. Since the DME/P
receiver processing parameters (i.e., bandwidth, time-of-arrival determination,
pulse-width discrimination, etc.) are selected to satisfy other considerations
such as multipath performance, the garble-induced error component must be
controlled by the interrogator data filter and channel assignment procedures.
The total garble error for the up- and downlinks combined must be less than
20 ft (6 m) for both the CMN and PFE components.
System efficiency is defined as the percentage ratio of valid replies
processed by the interrogator in response to its own interrogations. It is an
important parameter because system accuracy must be achieved when all of
the expected reply samples are not present. It is degraded when coincidence
with an undesired pulse prevents detection of the desired pulse or when an
interrogation falls in the transponder’s dead time and cannot result in a reply.
The interrogator’s output filter must therefore be designed such that the effects
of the missing data are minimized.
a. Downlink DME Garble. Downlink garble occurs when valid inter-
rogations at the ground transponder are coincident with interrogations from
other aircraft and transmissions from other ground facilities,resulting in either
a loss of signal or a pulse time-of-arrival measurement error. The garble can be
on-channel interference from aircraft in the same service volume or it can be
interference from aircraft whose interrogations are within the bandwidth of
the serving transponder. The garble effect is then a function of the number of
interrogations, and the corresponding distribution of interrogation freq-
uencies and signal amplitudes received at the transponder.
b. Uplink DME Garble. Uplink garble occurs when valid replies at the
interrogator are coincident with transmissions from other transponders and
interrogators and results in either a range measurement error or a missed reply
at the interrogator. The garble can be interference from any transponder
whose frequency is within the bandwidth of the interrogator, including those
on the same frequency, but with different pulse coding. This undesired ground-
to-air loading is a function of the number of transponders in the vicinity of the
interrogator and the corresponding distribution of reply frequencies and
signal amplitudes received at the interrogator.
c. Transponder Reply EfJiciency. As in DME/N (Section III,C,4), the
DME/P transponder has only a finite capacity to process received inter-
rogations. Recall that this results from the fact that each interrogation that
results in a reply creates a period of time during which arriving interrogations
200 ROBERT J. KELLY AND DANNY R. CUSICK

will not result in replies. This inhibit time is the result of dead time gates, reply
transmission gates, receiver blanking gates, and, indirectly, the priority
determination technique for replies, squitter, and facility identification pulse
pairs.

d. Interrogator Processor EfJiciency. The interrogator signal processor


efficiency is the ratio of the number of replies processed by the interrogator to
the number of interrogations in the absence of garble and transponder dead
time effects. This efficiencydepends on the receiver noise level and, as shown in
Subsection B,3, will exceed 95% for SNR values that are adequate to satisfy the
system accuracy requirements.
The interference rate due to garble (i.e., the rate at which range
measurements are in error or missing as a result of garble effects) is dependent
upon the channel assignment plan, traffic loading, and the ground trans-
ponder and interrogator receiver bandwidths. Garble effects are not a critical
factor in the DME/N system because of the use of narrow-bandwidth
receivers, which limit the garble susceptibility to co-frequency and first
adjacent frequency extraneous pulses. DME/N system efficiency is therefore
primarily dependent on the transponder reply efficiency, rather than on the
up- and downlink garble effects, and the accuracy degradation due to garble is
negligible when compared with the DME/N accuracy requirements.
In DME/P, the garble problem becomes significant because of tighter
accuracy standards imposed on the IA mode and the use of wide-bandwidth
receivers and low-level thresholding in the FA mode. As one might expect, the
F A mode reflects a greater susceptibility to garble in terms of accuracy and
system efficiency than the IA mode. These factors have been accommodated in
the DME/P system definition and, under the channel assignment constraints
detailed in Section III,C, do not require special consideration by the using
authority.
Worst-case traffic loading for an interrogator in the FA mode occurs at the
extremes of coverage, that is, at the 7-8 nmi transition region at a 2000 ft
altitude and the runway rollout region, where the signals received at the
ground equipment are at the minimum operational values. In the IA mode, the
worst environment will occur at the 22 nmi coverage limit, where not only are
the down- and uplink signal levels at low values, but the interrogator is at the
extremes of the frequency protected service volume.
Garble environments can be estimated by considering the geographic
separation of DME ground facilities, the number and location of aircraft
using each facility, the location of the victim ground and airborne equipment,
transmitter powers, and antenna patterns. Although the interference environ-
ment is composed of downlink and uplink components, only one of the links
dominates in terms of effect. For example, if the desired interrogator is at a low
DISTANCE MEASURING EQUIPMENT IN AVIATION 20 1
altitude (final approach at 2000 ft), then the downlink is critical because not
only is the ground transponder exposed to all of the aircraft present in a large
volume about the transponder, but the received signal level will be at or near
the transponder sensitivity value. The uplink effect is minimized since, at a low
altitude, the number of transponders within the line of sight of the victim
interrogator is reduced.
Table XIX defines one of the DME extraneous pulse environments used by
the ICAO to develop the DME/P channel plan and the 50% system efficiency
specification. These pulse interference rates and amplitude levels represent the
worst-case downlink environment for the FA mode at 7 nmi. This analysis is
summarized in Koshar and Smithmeyer (1982), where, in addition, garble
environments at other critical points in the service volume such as runway
threshold and at 22 nmi are provided.
e. Estimating System EfJiciency. System efficiency is the combined effect of
downlink garble, ground transponder dead time, uplink garble, and inter-
rogator signal processor efficiency. Since each of these efficiency components
are statistically independent, they can be computed individually and then
combined to yield the overall system efficiency. The effect of a single
component is defined as the percentage ratio of valid replies processed by the
interrogator in response to its own interrogations, assuming all other
components are not present. The system efficiency is then the product of the
individual components which must exceed the 50% SARPs requirement.
Figure 74 defines a procedure for calculating the system efficiency,making
use of the data shown in Fig. 75. The “detection probability” data (PDET) were

TABLE XIX
AIR-TO-GROUND INTERROGATIONLOADING
AT THE EL MONTE, CALIFORNIA,
DME/P
FOR A DESIREDINTERROGATOR
TRANSPONDER LQCATED7 nmi FROM THE BEACON
AND AT 2000 ft
ALTITUDE
~~ ~ ~ ~~

Number of interrogations per second”

Relative DIu DJU DJU DJU DJU DJU DJU


frequency I+30 I+20 I+14 I + 9 I + 6 I0 2-6

Cofrequency 2848 2112 1264 672 368 112 80


k1 MHz 6264 2432 1560 840
+ 2 MHz 4736 2016 704 432 208 112
k3 MHz 6208 2356 960 540 360 120
Total individual pulses 40,112 17,960 8276 4968 1872 688 160
~

a Number of interrogations greater than or equal to an undesired-to-desiredsignal power


ratio (DJU of 30,20, 14 dB, etc.) are represented in each column.
202 ROBERT J. KELLY A N D DANNY R. CUSICK

ik
p,, = ith pulse (of pulse pair) down-link garble effiimncy from kth adjacent
dunnel; k = 0 is the "on" channel frequency.

prePly = ($a8 equations 35 and 36).

p:p = uplink garble efficiency

Nota:

where fk = ka adjacent frequency garble rate

7 = Intaractionwindow about desird pulw (lour)


FIG.74. Computation of system efficiency.

derivea by computer simulation of the effects of different interfering DME


pulses. Range gate widths and decoder apertures were assumed to be 2 1.5 p.
Figure 76 illustrates the time interval z during which interactions with a
desired pulse can occur.
As an example, assume that the downlink garble rate on the first adjacent
frequency is fi= 10 kHz and that the D/U = 10 dB. From Fig. 75, PDET =
0.65. Therefore,
P& = 1 - (1 - e-f")(l - PDET)
= 1 - (0.1)(0.35)
= 0.96 (34)
This is but one of the many possible components of the system efficiency.
To determine the system efficiency, one need only repeat the computation for
all other components and combine them as shown in Fig. 74.
The computation of reply efficiency for the DME/P transponder is more
complex than that already noted for DME/N. This is due to the fact that a
single transponder is providing service to both IA and F A mode users. The IA
and FA mode reply efficiencies can be estimated using
DISTANCE MEASURING EQUIPMENT IN AVIATION 203

FIG.75. Graph depicting the detection probability determined by computer simulation of a


garble source of given frequency and D / U given that the undesired pulse having random rf phase
falls at a random position in a 10 ps window about the desired pulse. Assumes a 4 MHz IF
bandwidth, DAC parameters of 100 ns delay and 5.5 dB attenuation, and 2 ps pulse-width
discriminator.

where R,A is the FA channel reply efficiency, W is the number of FA channel


interrogations/second, R,, is the IA channel reply efficiency, N is the number
of IA channel interrogations/second, T,, is the FA inhibit interval due to
earlier FA interrogation, TWNis the IA inhibit interval due to earlier FA
interrogation, TNwis the FA inhibit interval due to earlier IA interrogation,
and T ” is the IA inhibit interval due to earlier IA interrogation.
T,, and T N N are simply measures of the effect (or dead time) an
interrogation has on its own channel. TwNand TNwindicate the effect (or
dead time) of an interrogation on the opposite channel. For the DME/N
transponder, TNw= TWN= 0, and Eq. (36) reduces to the classical reply
efficiency relation given in Section III,C,C
Typical values of T, TwN,TNw, and TNNare 80,86,50, and 76 p s for the
Y channel, assuming a 60 p s transponder dead time and a 10 &transmitter
pulse blanking time.
f. Estimating Garble Errors. The range error component due to garble is
a complex function of the extraneous pulse environment and the receiver
processing parameters, particularly the interrogator data processing al-
204 ROBERT J. KELLY AND DANNY R. CUSICK

OUTLIER WINDOW

DESIRED PULSE

RANGE GATE
OR DECODER APERTURE

GARBLE PULSE

TIME -+
APPLIES TO ANY PULSE
OF A TWO PULSE GROUP

GARBLE
INTERVAL
INTERACTION
7= tour
4
FIG.76. Garble interaction interval. A garble pulse in the interval will cause an error or
missed pulse. A garble pulse outside the interval has no effect.

gorithm. As already noted, the garble environment is characterized by the


pulse arrival rate and the undesired signal level of the extraneous pulse on each
frequency about the on-channel frequency.
The occurrence of a garble event on a given interrogation depends on the
probability that an undesired pulse is positioned close enough to the desired
pulse to result in an error or missed reply. Assuming Poisson arrivals at rate f
and an interaction interval z, the probability of a garble event appearing on a
single interrogation is 1 - e p f r .
The nature of the garble event, that is, whether the result is a range
measurement error or missed reply, depends on the relative undesired pulse
amplitude, phase, frequency, and position with respect to the desired pulse.
The signal processing parameters, that is, the IF bandwidth, Ferris Discrimi-
nator, pulse width discriminator, and track gate, all influence the outcome of a
garble event. Figure 77 illustrates an ungarbled DME pulse. Figure 78 is an
example of where an on-frequency garble pulse caused a missed pulse
detection (top figure) while the bottom figure resulted in a loo0 ns error. The
effects of adjacent frequency garble are depicted in Fig. 79.
The overall effect of garble events on the range accuracy is highly
dependent on the interrogator data filter algorithm (Subsection D,2). The
garble-induced errors occur at random times in the sequence of range
measurements and typically will have large amplitudes relative to other
DISTANCE MEASURING EQUIPMENT IN AVIATION 205

lo
0
T
_-

.10 _-
DAC DETECT TIME: 8993.0 ns DECLARED PULSE
-20-

-m-.

a, : ; : ; : ; : ; :: : I : : : :

DELAV.ATTENUATE-COMPARE I
I
PULSE WIDTH DISCRIMINATOR
TRACK GATE
0 m uylo 6Ooo 8wo lmao 12M)o 1uylo 1 m

TIME Ins1

FIG.77. Processingof ungarbled DME pulse. Pulse time of arrival indicated by rising edge
of “ D A C inside track gate that is associated with valid pulse-width discriminatoroutput.

sources of measurement error. Hence, garble errors appear as impulse noise.


Linear filters are good for removing the effects of random noise errors (such as
those due to receiver noise) from the range information. However, the memory
of the filter causes the effect of a single impulse error to be seen over many
subsequent samples. Hence, nonlinear filtering techniques, such as slew rate
limiting, noted in Subsection D, are useful in minimizing the effects of garble
errors. Additionally, the data filter must have the ability to “coast” through
missed data samples due to garble and transponder reply efficiency effects
while satisfying the accuracy requirements.
There are two approaches to predicting the accuracy degradation due to
garble. One can simulate a given multisource garble environment and record
the results. However, the simulation must be completely rerun if any of the
garble conditions or signal processing parameters is changed. A more rigorous
and useful approach is to generate (by simulation) data defining the error
statistics as a function of D / U and relative frequency. The functional relation
between the error and the garble rate is then determined using procedures
similar to those outlined in Figs. 74 and 75. In this way, the statistics of the raw
error can be estimated for a general garble environment.
As is evident, predictions of the magnitude of the garble error component
that are generally applicable are difficult if not impossible to make due to the
complexity of the phenomenon. In general, garble effects must be experiment-
ally determined via bench measurements or studied with the aid of a time
domain computer simulation.
Figure 80 shows the error statistics (mean, standard deviation) when the
garble pulse has a D / U level ranging from - 30 to + 30 dB and rf frequency of
206
10-

0 --
-lo--
Y

3
t -20-L NO DECODE
k DECLARED PULSE ARRIVAL
5 .m--

-40,: I : ! : ; : I ; I :

DELAY-ATTENUATE-COMPARE -I
WLSE WIDTH DISCRIMINATOR
TRACK GATE
0 2000 4000 6000 8000 ioom 12000 14000 16ooo

TIME In4

UNDESIRED PULSE: D I U I S 3 dB
ONFREOUENCV
SAME RF PHASE
3000 ns I N FRONT OF DESIRED PULSE

1°T

0 --

-lo--
DECLARED PULSE
-20- -1000 ns ERROR

-30.-

DELAY.ATTENUATE-COMPARE

PULSE WIDTH DISCRIMINATOR


TRACK GATE
-I
-
0 2000 4000 MMO 8000 loo00 12000 14Wa 1-
TIME Ins1

UNDESIRED PULSE. DIU I S 3 dB


ONFREOUENCY
SAME RF PHASE
1000 ni IN FRONT OF DESIRED PULSE

FIG.78. Processing of garbled (on-frequency)DME pulse.

0 to 4 MHz relative to the desired signal. Pulse detectibility data for the same
conditions were shown in Fig. 75. The results at a particular DIU and
frequency are simply statistics of multiple samples taken for various relative rf
phases and garble pulse positions in a 10 ps window about the desired pulse. A
garble pulse falling outside the 10 ps window was found to have no effect on
the measurement of range with the desired pulse.
DISTANCE MEASURING EQUIPMENT IN AVIATION 207

,; ;+;:/!:
0.-
-
rn
;
n
Y
-10-
3
t -30
.20-- +266.5 ns ERROR

3
~

OF TRACK GATE
-30-

-40 : ; : I : ; : ; : I ;

DELAY-ATTENUATECOMPARE
I
PULSE WIDTH DISCRIMINATOR ”
TRACK GATE
0 aooo 4&
4000 wbo
woo 8000 lwoo lZOw
1ZOw 14wo 16Mo

TIME hi1
UNDESIRED PULSE: D/U IS 3 dB
ZMHZ OFF FREOUENCY
FREOUENCY

3wo ns IN FRONT OF DESIRED PULSE

DELAY-ATTENUATE.COMPARE - “ “ I

PULSE WIDTH DISCRIMINATOR


TRACK GATE
0 2000 4wo M)o 8ow 10000 lzwo ~4ooo 1wlm
TIME Ins1

UNOESIREDPULSE: D I U l S 3 d B
2MHZOFFFREQUENCY

1000 n i IN FRONT OF DESIRED PULSE

FIG.79. Processing of garbled (off-frequency)DME pulse.

The data presented in Fig. 80 are applicable when a garble event is present
on each interrogation. In an actual environment garble events occur at
random during the sequence of interrogations. These random events are taken
into account by using the pulse detectibility data in Fig. 75. The statistics of
the raw (unfiltered) garble-induced error component (mean and standard
deviation) resulting from N sources (source i is characterized by specific
DIU, rf frequency, and arrival rate) can be determined from the data given
208 ROBERT J. KELLY AND DANNY R. CUSICK
400

350

300
250

200

150

100

50

-50

-100
-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30
DESIRED-TO-UNDESIRED SIGNAL RATIO (dB)

-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30


DESIRED-TO-UNDESIRED SIGNAL RATIO (dB)

- -
-
KEY GARBLE FREQUENCY

0 mH,
1 mH,
---------
a
-
.
-
.
-

2 mH,
....-.,..........-
---_----.---
3 mH,
4 mH,

FIG.80. Graphs depicting the statistical effect determined by computer simulation of a


garble source of given frequency and D/Ugiven that the undesired pulse having random rf phase
falls at a random position in a 10 p s window about the desired pulse. Assumes a 4 MHz IF
bandwidth, DAC parameters of 100 ns delay and 5.5 dB attenuation, and a 2 p s pulse-width
discriminator.
DISTANCE MEASURING EQUIPMENT IN AVIATION 209
in Figs. 75 and 80 using the following equations:
N

with the constraint that


N
2pi=1 (39)
i= 1

The symbols utilized are defined as follows: N is the number of garble sources
of a given D / U , rf frequency, and arrival rate f;ui, the mean error component
due to source i as determined from Fig. 80; a:, the error variance due to source
i as determined from Fig. 80; and pi,the percentage of interrogations in which
an error due to garble source i is present. It is the product of two terms:
(1 - e-fr), the probability that a garble pulse falls in the interval r; PDET,
the probability that the desired pulse is detected given the presence of a garble
pulse in the interval T as obtained from Fig. 75. Finally, f is the pulse arrival
rate; while z is the interaction interval; 10 p s for the data shown in Figs. 75
and 80.
Although the filtered error component cannot be accurately determined
using the above procedures, the magnitude of the raw error resulting from the
garble is determined. The utility of this computation is in comparing the
relative effects of different garble environments on D M E / P performance.
Figure 81 shows the predicted raw error standard deviation for a pulse rate of
16,200 pulses/s. Figure 82 shows the filtered garble error statistics for the same
case that results when an a-8 filter with a 40 ns slew rate limit is used as the
interrogator data filter (see Subsection D,2). These results were derived by
direct simulation of the garble environment.
The garble error calculation procedures outlined above are summarized
by the following examples. Assuming an on-frequency garble error source with
D / U = 10 dB and a rate of 16,200 pulses/s, what is the standard deviation of
the raw range error? From Figs. 75 and 80, p1 = 0 ns, a1 = 160 ns, and
PDET = 0.67. The percentage of interrogations in which a garble source in-
duces an error is calculated using Eq. (39). That is, p1 = (1 - e-0.16)(0.67)=
0.1. The raw noise error is then determined from Eq. (38): a, =
Jm
=
= 51 ns.
In Fig. 81, the complete simulation resulted in a 55 ns noise error.
Therefore, the generalized procedure described here permits a close approxi-
mation to those estimates obtained from a complete simulation of the garble
environment.
210 ROBERT J. KELLY A N D D A N N Y R. CUSICK

-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30


DESIRED-TO-UNDESIRED SIGNAL RATIO IdB)

-
KEY ~
GARBLE FREQUENCY
0 mH,
-.-.-- 1 mH,
--------- 2 mH,
-.............-...
--- --- --- ---
3 mH,
4 mH,

FIG.81. Standard deviation of range error that results for unfiltered measurements made in
the presence of a 16,200 pulse/second garble source of specified D/U and frequency.

The filtered error for this example can be estimated from Fig. 82. The
filtered standard deviation is only about 55.5 ns. The efficiency of the DME/P
signal processing, particularly the data filter shown later in Fig. 91, is clearly
evident by this example, where the garble noise error component was reduced
almost by a factor of 10.

6. Summary
This section has shown how the performance required of the DME/P is
achieved in the presence of noise, multipath, and garble by taking advantage
of a fast rise time pulse, the delay-attenuate-compare circuit for pulse time-
of-arrival determination, and a wide IF bandwidth that preserves the shape of
the received DME. Summarized below are the major points presented in this
section.
The fast rise time pulse (1000 ns), called cos/cosz, satisfies not only the need
for DME/P to be fully interoperable with DME/N but the strict spectral
limitations imposed on the DME ground stations as well. The delay-
attenuate-compare detector simply takes maximum advantage of the fast rise
time pulse to enhance the performance of the DME/P in the presence of
multipath.
DISTANCE MEASURING EQUIPMENT IN AVIATION 211
15.00
13.50

12.00

10.50
c
9 9.00
20 7.50
O
K 6.00
c
0
z 4.50
a
k 3.00

1.50
0.00
-20 -16 -12 -8 -4 0 4 8 12 16 20
DESIRED-TO-UNDESIRED SIGNAL RATIO (dB)

-
KEY GARBLE FREQUENCY
0 mH,
-.-.-. 1 mH,
--------- 2 mH,
-................-
--- --- --- ---
3 mH,
4 mH,

FIG.82. Standard deviation of range error that results for filtered range measurements
made in the presence of a 16,200 pulse/second garble source of specified D/U and frequency.
Interrogator data filter is an a-p filter using outlier limit detection and slew-rate limiting.
a = 0.125, fi = 0.0078, outlier limit = 300 ns, slew-rate limit = 40 ns. Data derived by direct
computer simulation of the garble environment.

Accuracy and system efficiency degradation due to receiver noise was


shown for both the IA and FA modes. As will be shown in Subsection C , the
noise-induced errors are readily accomodated within the IA and FA
instrumentation error budgets when the minimum required signal levels are
available. For the important FA mode, the effects of receiver noise were shown
for a variety of DAC parameters and I F bandwidths. Of particular interest is
the fact that when a pulse-width discriminator is used, increasing the I F
bandwidth from 3.5 to 4.5 MHz will reduce the noise-induced range errors.
Additionally, it is clear that accuracy, and not reply efficiency, is the limiting
consideration.
The pulse shape, pulse detection technique, and the wide I F bandwidth are
the keys to multipath immunity of the DME/P FA mode. Unlike DME/N,
where strong multipath can result in errors of 200 or more feet, FA mode
errors are negligible for reflected pulse delays in excess of 350 ns while the
maximum error is less than 33 ft for short delay multipath. In the IA mode, the
aircraft is only exposed to low-level (6 dB or more below the desired)
212 ROBERT J. KELLY AND DANNY R. CUSICK

multipath, and the resulting errors are readily accomodated within the
accuracy requirements.
Finally, due to its wide bandwidth and tighter accuracy requirements,
DME/P exhibits a greater susceptibility to garble than DME/N. Although
difficult to analyze, the effects of garble have been shown to be both
predictable and manageable. Prediction of the garble effects requires a
knowledge of the rate, rf frequency, and levels of the pulses that comprise the
garble environment, the particular signal processing in use, and a database of
error and pulse detectibility data that can only be obtained by time domain
computer simulation or accurate laboratory measurements. The effects of
errors and missing data are minimized by appropriate data processing
techniques like those to be discussed in Subsection D.

C . D M E / P Signal in Space: Accuracy and Coverage

The purpose of this section is to bring together the detailed performance


discussions of the previous sections and thereby define DME/P accuracy and
coverage requirements. First, Subsection C,1 defines the DME/P error
components, path following error, and control motion noise, which describe
the effect of the error on aircraft flight control system performance.
Additionally, methods of measuring these components are described. Subsec-
tion C,2 presents the system error budgets for Standard 1 and Standard 2
DME/P indicating that the system accuracies can be achieved for specified
values of site-related errors and instrumentation errors.
Subsection C,3 presents example DME/P power budgets where the values
shown reflect the parameters required to obtain the desired signal coverage.
Subsection C,4 treats one of the elements necessary to achieve the desired
vertical coverage, the siting of the DME antenna.

I . DMEIP Error Measurement Methodology


The DME/P system accuracies, like the MLS angle system accuracies, are
specified in terms of path following error and control motion noise. These
parameters describe the interaction of the DME/P guidance signal with the
aircraft in terms directly related to aircraft position errors and flight control
system design.
The PFE is that part of the error signal having frequency components less
than 1.5rad/s. These low-frequency components can displace the aircraft from
its intended course over the ground. The portion of the error signal having
frequency components between 0.5 and 10 rad/s is the CMN. CMN
components fall above the aircraft guidance loop bandwidth and thus are too
rapid for the aircraft to follow. However, the CMN components of the error
DISTANCE MEASURING EQUIPMENT IN AVIATION 213
signal may induce aircraft control activity such as aileron and cockpit wheel
motion as well as aircraft roll variations. This unwanted control activity affects
the pilot’s confidence in both the guidance system and his flight control system
during autocoupled approaches.
Performance evaluation of the DME/P system in terms of PFE and CMN
is summarized in Fig. 83. The range error signal is generated by forming the
difference between the range output of the interrogator and the reference
range. The PFE and CMN components are determined by processing the
error signal with the appropriate digital filters. Digital implementation of a
PFE and a CMN bandpass filter are shown in Figs. 84 and 85.
For the purposes of determining compliance with the accuracy standard,
the PFE and CMN components are evaluated over any 10 s measurement
interval of the F A mode flight error record taken within the DME/P coverage
limits. The measurement interval for the IA mode is 40 s. The 95% probability
requirement is interpreted to be satisfied if the PFE and CMN components do
not exceed the specified error limits for a total period that is more than 5% of
the evaluation interval. Determination of the 95th percentile of the error data
is illustrated in Fig. 86. This same technique can also be utilized to determine
the transponder and interrogator instrumentation error components where
the overall system range error is replaced with transponder delay measure-
ments or interrogator range measurements.

2. DMEIP System Error Budgets


As noted earlier in Section III,C,l, error budgets serve to aid facility
planning and equipment procurement specification by identifying, quantify-
ing, and combining the instrumentation and site-dependent errors that
determine overall system performance. These budgets are critical to the
specification of the DME/P due to the tight accuracy constraints. An
allowance in the budget must be made for each of the error components
discussed earlier-multipath, garble, and noise-as well as the ground and
airborne equipment instrumentation errors. Receiver noise effects are in-
corporated by satisfying the instrumentation error requirements at the
minimum received signal power levels.
Table XX is the error budget recommended for guidance by ICAO. The
itemized components are to be satisfied on a 95% probability basis.
Allowances for the FA mode Standards 1 and 2 as well as the IA mode are
shown.
To be useful, an error budget must have a simple procedure for combining
the individual components. If the error sources are additive, independent
random variables with zero mean, they can be combined on an RSS basis.
Although these conditions are not always fulfilled for DME/P, experience
214 ROBERT J. KELLY A N D D A N N Y R. CUSICK

DME/P INTEROGATOR

-@
RAW DISTANCE PATH FOLLOWING PATH
ESTIMATES OUTPUT ERROR F I L T E R O O

-
ERROR

CONTROL
POSITION MOTION NOISE
REFERANCE FILTER W

GUIDANCE CORNER FREQUENCIES


RECEIVER OUTPUT FILTER -W2
StW 2

wg* , w2- PATH


FOLLOWING
ur2
I

DME IP 1.5 0.5 '10 FILTER S2 + ZSW, +W:


~~~ ~-
*Wo=O64Wn CONTROL MOTION NOISE FILTER

[*]
* * THIS CORNER FREQUENCY M A Y BE LESS
THAN 10 RADlSEC FOR SOME APPLICATIONS.

PATH FOLLOWING FILTER (LOW PASS)

CONTROL MOTION NOISE FILTER

-----

I
I I I I
I I \ RADlANSlSEC
w, Wo \ 3 4
FILTER CONFIGURATION AND CORNER FREQUENCIES

OUTPUT OF PFE
OR CMN FILTER

T-SLIDING WINDOW

MLS-DMEIP measurement methodology. E: error specification: T: region


t o be evaluated: T1, T 2 T3 _...
: time intervals that error exceeds specification
For ground equipment t o be acceptable in t h i s region, the following inequality
h w l d be true: (T1 + T2 + T j + -.) / T 5 0.05.

FIG.83. MLS-DME/P measurement methodology. E : error specification T region to be


evaluated. T,, T2, T3,...: time intervals that error exceeds specification. For equipment to be
acceptable in this region, the following inequality should be true: (TI T2 + T3 + ...)/ T I0.05.+
DISTANCE MEASURING EQUIPMENT IN AVIATION 215
DEFINITIONS
H(S) = ANALOG FILTER TRANSFER FUNCTION
RATIO OF THE LAPLACE TRANSFORM OF THE OUTPUT SIGNAL
TO THE LAPLACE TRANSFORM OF THE INPUT SIGNAL
H(2) = D I G I T A L FILTER TRANSFER FUNCTION
RATIO OF THE 2-TRANSFORM OF THE OUTPUT DATA TO THE
2-TRANSFORM OF THE INPUT DATA
'n = INPUT DATA SAMPLE AT TIME n
Yn = OUTPUT DATA SAMPLE A T TIME n
ANALOG PFE FILTER
5.49
H(S) =
S2 + 4.698 + 5.49

DIGITAL PFE FILTER-FA MODE (40 Hz DATA RATE)

HE) = 7.9762 X lop4

Yn = 7.9762 X ( X n + 2XnP1 + Xn-2 ) + 1.887Yn-1 - 0.8902Yn-2

D I G I T A L PFE FILTER-1A MODE (16Hz DATA RATE)

H(Z0 = 4.5942 X

Yn = 4.5942 X 10-3 (X, + 2XnP1 + X,-2 1 + 1.7288YnP1 - O.7472Ynp2

FREQUENCY RESPONSE

0
-4
-8
-*
-
U
LU
-12

-16
n
3
-20
z
W -24
a
= -28
-32
-36
-40
0 2 4 6 8 10 12 14 16 18 20
FREQUENCY (RADIANS/SEC)

FIG.84. Analog and digital implementations of PFE filter.


216 ROBERT J. KELLY AND DANNY R. CUSICK

DEFINITIONS
H(S) = ANALOG FILTER TRANSFER FUNCTION
RATIO OF THE LAPLACE TRANSFORM OF THE OUTPUT SIGNAL
TO THE LAPLACE TRANSFORM OF THE INPUT SIGNAL
H(2) = DIGITAL FILTER TRANSFER FUNCTION
RATIO OF THE 2-TRANSFORM.OF THE OUTPUT DATA TO THE
2-TRANSFORM OF THE INPUT DATA
Xn = INPUT DAJA SAMPLE AT TIME n
Yn = OUTPUT DATA SAMPLE AT TIME n
ANALOG BANDPASS CMN FILTER
H(S) = 1. 10
S+O.5 S+10

DIGITAL BANDPASS CMN FILTER-FA MODE (40 Hz DATA RATE)

H(2) = 0.11094

=
I (22- 1 )
(2’ - 1.76432 + 0.7671)
0.1 1094Xn - 0.1 1094Xn + 1.7643YnP1 - 0.7671 YnP2
Yn

DIGITAL BANDPASS CMN FILTER-1A MODE (16 Hz DATA RATE)

H(2) = 0.2404

Yn = 0.2404Xn
I 2:0;::!
(2’ + 0.4958)
- 0.2404Xn-2 + 1.4808Ynp1
I - O.4959Ynp2

FREQUENCY RESPONSE

-3

-12

-15
0 3 6 9 12 15 18 21 24 27 30
FREQUENCY(RADIANS/SEC)
FIG.85. Analog and digital implementationsof bandpass CMN filter.
DISTANCE MEASURING EQUIPMENT IN AVIATION 217

(a)

ao 240.0 480.0 720.0 960.0 12w.o 1440.0 1680.0 1920.0 2160.0 2400.0
SAMPLE NUMBER

1m.o
90.0
INPUT TO SLIDING WINDOW

:
Gi 80.0 PROCESSOR ABSOLUTE
VALUE OF CMN
7013
4 400 POINT/lO SE
5 60.0 SLIDING W I N D 0

(b) 50.0
OUTPUTOF SLIDING WINDOW
PROCESSOR: 95th PERCENTILE
OF DATA IN WINDOW
5, 40.0

s 30.0
5 zao
10.0

0.0
ao 240.0 480.0 720.0 960.0 12w.o 1440.0 1680.0 1920.0 2160.0 2400.0
SAMPLE NUMBER

1w.o . , , , , I , , ' " ' . , ' I .


1 95th PERCENTILE
00.0
&j
z
-
.-
J

z
Z

Y
70.0

60.0
.-
+
Y
50.0 -
5= 40.0 --
(C) c .

all.lrLL I.VI.IDL"

FIG. 86. Determining the CMN component of DMEjP range error. (a) CMN determined
by filtering interrogator range error with CMN filter. (b) Processing of CMN with 400 point (10 s)
sliding window to determine 95th percentile of CMN data in window relative to 0 ns error.
(c) Output of sliding window processor.
218 ROBERT J. KELLY AND DANNY R. CUSICK
TABLE XX
DME/P ACCURACY SPECIFICATTONS”
(METERS-95% PROBABILITY)

Standard 1 Standard 2 IA mode

Error source Error component PFE CMN PFE CMN PFE CMN

Instrumentation Transponder 10 8 5 5 15 10
Interrogator 15 10 7 I 30 15
RSS Subtotal 18 13 8.6 8.6 33 18
Site related Downlink specular multipath 10 8 3 3 31 20
Uplink specular multipath 10 8 3 3 31 20
Nonspecific (diffuse)multipath 3 3 3 3 3 3
Garble (DME) 6 6 6 6 6 6
Garble (non-DME) 1 1 1 1 1 1
RSS Subtotal 15.6 13.2 8.0 8.0 53 29
System Combined instrumentation and
site-related error components
not allowed to exceed system
accuracy standard 30 18 12 12 85 34

’The figuresfor “nonspecular multipath and for “garble” are the RSS totals of the uplink and downlink
components. The pulse waveform is assumed to have a cos/cosz shape and a 1200-nsrise time. The error used
for specular multipath is the peak value obtained with an MID = - 6 dB for IA mode and MID = - 3 dB for
FA mode. Interrogator instrumentation error includes CMN component due to transponder reply efficiency.

indicates that reasonable estimates can be obtained in this manner. In practice,


the combination of site-related components with the instrumentation errors
cannot exceed the overall system errors noted earlier in Subsection A,2. Thus,
if the specified error components are not individually exceeded, it can be
expected that the overall system performance will be achieved.

12.3 NM

--------_____ HORIZONTAL
0.85’- 1-

FIG. 87. DME/P vertical coverage.


DISTANCE MEASURING EQUIPMENT IN AVIATION 219
3. System Coverage and Power Budgets
A DME/P ground facility is required by the SARPs to provide range
information throughout a coverage volume (in azimuth) that is at least as large
as that provided by the MLS azimuth facility with which is paired. Typically,
360" DME azimuthal coverage to 20 nmi from the MLS reference datum will
be provided. Vertical coverage is summarized in Fig. 87. Note that coverage is
to be provided down to 8 ft above the runway surface, ensuring the availability
of range information following touchdown.
As is the case with DME/N, the signal in space to be provided by the
transponder is specified in terms of power density (dB W/m2),thereby keeping
the specification independent of transmitter, antenna, and cabling con-
siderations. Specifications are similar for the transponder sensitivity require-
ments. Table XXI summarizes the relevant signal-in-space, sensitivity, and
transmitter requirements for DME/P ground and airborne equipment.
TABLE XXI
SUMMARY
OF GROUND
AND AIRBORNE
RADIATEDPOWERAND RECEIVER
SENSITIVITI.
REQUIREMENTS

Transponder requirements (SARPs)"


Uplink peak ERP at - 89 dB W/m2 At any point within the defined coverage
interrogator antenna volume beyond 7 nmi of transponder
antenna
- 75 dB W/m2 At any point within the defined coverage
volume within 7 nmi of transponder
antenna
- I 0 dB W/m2 At MLS approach reference datum
-79 dB W/m2 At 8 ft above the runway surface at the
MLS datum, or the farthest point on the
runway centerline within line of sight of
the antenna
Downlink sensitivity -75 dB W/mZ FA mode
at transponder antennab -86 dB W/m2 IA mode
Interrogator requirements (RTCA)'
FA 1A

Transmitter power 250 W 50 W


Receiver sensitivity at input terminal -73 dBm -73 dBm At 7 nrni range
- 83 dB m At 22 nmi range

SARPs requires interrogator performance to be consistent with ground system


specifications.
' FAA (1983~)requires - 89 and - 79 dB W/m2 for the IA and FA modes, respectively.
RTCA interrogator requirements assume a 2-dBi antenna gain and a 3-dB cable loss in
determining receiver input power.
220 ROBERT J. KELLY A N D D A N N Y R. CUSICK

TABLE XXII
CTOL AIR-TO-GROUND
POWERBUDGET
(DME/P)

22 7 Reference
Power budget items nmi nmi datum Rollout

Peak interrogator transmitter power (dB m). 54 54 54 54


Aircraft antenna gain (dBi) 0 0 0 0
Aircraft cable loss (dB) -4 -4 -4 -4
Peak effective radiated power (dBm) 50 50 50 50
Ground and lateral multipath loss (dB) -5 -3 -4 - 17
Path loss (dB) - 125 -115 - 107 - 103
Polarization and rain loss (dB) -1 -1 0 0
Received signal at transponder antenna (dB m) -81 -69 -61 - 70
Ground antenna gain (dBi) 8 8 8 8
Pattern loss (dB) -4 -2 -5 -5
Cable loss (dB) -3 -3 -3 -3
Received signal at transponder (dB m) - 80 - 66 - 61 - 70
Receiver noise video (dB m)
(NF = 6 dB)
IF bandwidth is 3.5 MHz - 106 - I06 - 106
IF bandwidth 0.8 is MHz - 112
Signal-to-noise ratio (video) (dB)* 32 40 45 36

Power density (dB W/m2) = power (dB m)- 8.


* S/N(video) = S/N(IF) + 10log(IF noise bandwidth/video noise bandwidth).
As noted earlier in Section 111,C,2,power budgets play a role similar to that
of error budgets. They are a means to identify, quantify, and combine the
ground and airborne equipment losses, propagation losses, and antenna
characteristics to ensure that the up- and downlink signal in space is adequate
to satisfy all relevant accuracy and system efficiency requirements.
Tables XXII and XXIII are the DME/P CTOL power budgets as defined
in AWOP (1982).The permitted ground system ERP is based on the use of an
800 ns cos/cos2 pulse. The calculations assume that the aircraft antenna is not
shielded by the aircraft structure. In addition, free-space propagation loss is
assumed. This is an accurate method of loss computation due to the short
ranges (< 22 nmi) involved. However, the effect of ground multipath that
results in signal fading is an important consideration that will be addressed in
the next section.
It is important to note that the receiver sensitivity (in the case of both
ground and airborne equipment) is the minimum signal which satisfies both
the system efficiency and accuracy requirements. However, in contrast to
DME/N, the sensitivity of a given link is limited by the CMN component
arising from the receiver noise rather than the system efficiency. This is clear
from the discussions of Subsection B,3.
DISTANCE MEASURING EQUIPMENT IN AVIATION 22 1
TABLE XXIII
CTOL GROUND-TO-AIR
POWERBUDGET
(DME/P)

22 7 Reference
nmi nmi datum Rollout

Peak effective radiated power (dB m)” 55 55 55 55


Ground and lateral multipath loss (dB) -5 -3 -4 - 17
Antenna pattern loss (dB) -4 -2 -5 -5
Path loss (dB) - 125 -115 - 107 - 103
Monitor loss (dB) -1 -1 -1 -1
Polarization plus rain loss (dB) -1 -1 0 0
Received signal at aircraft (dB m) -81 -6 1 - 62 -71
Aircraft antenna gain (dBi) 0 0 0 0
Aircraft cable loss (dB) -4 -4 -4 -4
Received signal at interrogator (dB m) - 85 -71 - 66 - 75
Receiver noise video (dB m)
(NF = 9 dB)
IF bandwidth is 3.5 MHz - 103 - 103 - 103
IF bandwidth is 0.8 MHz - 109
Signal-to-noise ratio (video) (dB)b 24 32 37 28
~ ~ ~ ~ ~

a Power density (dB W/mZ) = power (dBm) - 8.


S/N(video) = S/N(IF) + 10log(IF noise bandwidth/video noise bandwidth).

It is also important to note that the IA mode is not the limiting


consideration; rather it is the FA mode rollout maneuver for which
consideration must be given. This is primarily because the*-amplitude find (or
PAF) circuit is utilized for pulse detection in the IA mode as opposed to the
- 20 dB thresholding DAC utilized in the FA mode. Additionally, a low SNR
results from the 22 dB signal attenuation due to ground multipath and
antenna pattern behavior at the low coverage angle.

4 . DMEIP Antenna Considerations


The DME/P antenna should be located laterally as close to the runway
centerline and near the MLS azimuth function antenna site as possible,
consistent with obstacle clearance criteria. The relation of the distance L in
back of the runway stop end where a DME antenna of height H must be
sited is L = 61 + 50H,where L and H are in meters.
Locating the DME/P antenna at the azimuth site simplifies the coordinate
conversion computation when MLS is used to determine three-dimensional
position information as in MLS area navigation operations. Additionally, a
near centerline location keeps the range information meaningful when
unconverted information is utilized to determine decision height and rollout
distance.
222 ROBERT J. KELLY AND DANNY R. CUSICK

TABLE XXIV
SUMMARYOF SIGNALATTENUATION
DUE TO GROUND
MULTIPATH
AS A FUNCTION
OF PHASE-CENTER
HEIGHT

Phase- Signal attenuation"


center (including ground multipath) Antenna-
height to-threshold
(m) 15.2 m 2.4 m (rollout) distance (m)

4.6 - 8 dB -22 dB 3048


9.1 -3.0 dB - 19 dB 4513

Relative to free-space peak signal.

Siting of the DMEIP antenna in the vertical dimension is dependent on


runway length, runway profile, and local terrain. These siting factors are
particularly important when it is necessary to assure that adequate signal
power density is available in the runway threshold region or near the runway
surface. Table XXIV gives the expected attenuation for a 7 ft (2.1 m) vertical
aperture onmi antenna. This subject and the antenna design are covered in
more detail in Kelly and LaBerge (1980).

D . DMEIP Signal Processing and Transponderllnterrogator


Implementation considerations

The previous sections have identified all of the components required to


successfully implement DME/P ground and airborne equipment as well as
how these components affect system performance. This section is intended to
unify these concepts by first summarizing the signal processing functions that
are key to the DME/P and then illustrating the ground and airborne
equipment configurations at the functional level. It must be emphasized that
many of the signal components required for DME/P are not unlike
those utilized in DMEIN, as described in Section I1 of this article. Hence,
the discussions below will detail only those aspects which are unique to
DMEIP.

I . Summary of DMEIP Signal Processing functions


Several signal processing techniques are critical to the successful im-
plementation of DME/P and are common to both the ground and airborne
equipment. As in DMEIN, these functions are receiver bandwidth and
dynamic range, signal identification and validation, and pulse detection.
DISTANCE MEASURING EQUIPMENT IN AVIATION 223
Although the signal processing circuits described in this section may appear to
represent one hardware concept among many, it must be emphasized that
these concepts are fundamental and necessary in one form or another to
achieve the required accuracy in most approach and landing situations. Fig-
ure 88 summarizes all of the functional components by illustrating signal flow
through the DME/P interrogator. Figure 89 summarizes the operational
environment, with an emphasis on the pulse interference environment, which
necessitates the signal processing concepts addressed below.

a. Receiver Considerations- Bandwidth and Dynamic Range. In order to


achieve the desired accuracy in the FA mode, the receiver must faithfully
reproduce the sharp linear leading edge of the received pulse in the vicinity of
the threshold point. To do so requires an IF bandwidth on the order of 4 MHz
in contrast to the 400-800 kHz bandwidths typical of DME/N equipment. As
noted in Subsection B,4, the use of wide-bandwidth processing and the delay-
attenuate-compare time-of-arrival detector ensures excellent multipath im-
munity. On the other hand, the exposure to noise and garble effects is
increased.
In the IA mode, an 800 kHz I F bandwidth is all that is required due to
reduced accuracy requirements and a less stressful multipath environment.
The narrow bandwidth, combined with the +-amplitude thresholding of the
received pulse, results in better immunity to garble and receiver noise eirects.
The dynamic range requirements of the DME/P airborne and ground
receivers are no different from those defined for DME/N in Section III,B.

b. Signal ldentijcation and Validation. As in DME/N, the DME/P signal


processing must identify and validate received DME signals before they are
utilized to generate replies or for range measurement at the interrogator. The
same characteristics of the DME signal, pulse width, rf frequency, and code,
are used to perform this function.
Recall that the purpose of the pulse-width discriminator is to discriminate
a valid pulse from those due to noise which are of short duration. This function
takes on a more important role in the implementation of the DME/P F A
mode. In the presence of noise, the delay-attenuate-compare circuit will
indicate many pulse arrivals. If a noise-induced pulse occurs near enough to a
desired pulse (e.g., within the interrogator track gate or on the leading edge of
the pulse itself), it may be misconstrued as a valid pulse and errors may result.
This is, in part, due to the use of a wide-bandwidthIF and a low threshold level
relative to the peak of the pulse. However, because of the wide receiver
bandwidth, the average duration of the noise bursts are much shorter than the
nominal 3.5 p s DME pulse width. Therefore, a pulse-width discriminator is
Low Pan Equivalent DME Signal Combined Desired Signal
Estimated and Interference with Interferance
Time of
Arrival

, IcL Zero Error Reference

m
T Limiter

Data

+ PAF

I A Mode

-=a-
Valid Code Decoder Filter
12r2us IFBW-
800kHz
Valid on.channel signal
I A OutputbFA output

6of
Estimate
db
Arrival
Time A transmitter interrogation sample turns the range counter "ON,"
I The synchronized reply turns the counter "OFF."
The number of counts is proportional to the range.

FIG.88. Interrogator signal processing functions.


DISTANCE MEASURING EQUIPMENT IN AVIATION 225

15

F A Mode Wideband
4.5 MHz Filter

FIG.89. DME/P interference environment.

employed to reject the noise pulses and identify the received DME pulse. A
pulse-width criterion of 2 p s or more is sufficient to discriminate between valid
DME pulses and noise-generated pulses in both the IA and FA modes.
As in DME/N, only pulses that are on the assigned channel frequency must
be considered for further processing. In the DME/P, the channel frequency
separation of 1 MHz is less than the FA mode receiver bandwidth. Thus,
pulses on at least the first and second adjacent frequencies will be indistin-
guishable in the receiver video from those on the desired frequency.
Consequently, unlike DME/N, some variant of the Ferris Discriminator (see
Section III,B) must be used not only in the ground transponder but in the
airborne interrogator as well.
A received signal which satisfies the pulse width and frequency tests must
also have the proper pulse spacing for the operating code. Requirements for a
pulse pair decoder are the same as those for DME/N equipments. Note that
the DME/P ground transponder decoder is more complex than that utilized in
conventional DME equipment. It must distinguish not only between channel
codes (e.g., X and 2 )but also between IA and FA mode interrogation codes
as well, ensuring the use of the proper processing when generating the reply.
c. Pulse Time-of-Arrival Determination. As in DME/N, the arrival time of
the DME pulse must be determined in both the airborne and ground
equipment. As noted in Subsection B,2, pulse time of arrival in the FA mode is
226 ROBERT J. KELLY AND DANNY R. CUSICK

determined using the delay-attenuate-compare circuit, with nominal pulse


thresholding 18 to 20 dB below the pulse peak for DAC parameters of 100 ns
delay and attenuations of 5 to 6 dB. As in DME/N, the IA mode utilizes a
*-amplitude find circuit to determine pulse time of arrival.
2. Transponder/InterrogatorImplementation Considerations
As is evident, the functional components of the DME/P transponder and
interrogator are much like those required in DME/N and discussed earlier in
Section II1,B. The major differences are the inclusion of the FA mode channel
and the need for strict attention to the effects of each on system accuracy.
a. Transponder. Functionally, the DME/P transponder, illustrated in
Fig. 90, can be viewed as two separate transponders, one which responds to
F A mode interrogations and one which responds to IA mode interroga-
tions. However, in order to reduce cost, hardware implementations will most
certainly integrate functions common to the IA and F A modes.
For the X channel example shown, the IA mode transponder responds to
the receipt of the 12 p s code. A 12 p s coded reply is transmitted 50 ps after the
receipt of the interrogation. The reply delay timing is adjusted to 50 ps
between the +-amplitude point on the leading edge of the first interrogation
pulse to the same point on the leading edge of the first reply pulse. The FA
mode transponder responds to interrogations having a code of 18ps. A reply is
then generated 56 ps after the arrival of the first pulse with a reply code of 12
ps. The reply delay timing is adjusted to 56 p s between the virtual origin of the
first interrogation pulse and the virtual origin of the reply pulse, as discussed in
Subsection B,2.
As in DME/N, the DME/P transponder must transmit squitter and the
Morse code identification signal. Squitter is provided by the DME/P ground
system to ensure compatibility with DME/N interrogators, some of which

TRANSMITTER PRIORITY LOGIC

a
DELAY
5011%
\

DECODER X-CHANNEL
RECEIVER I A MODE 1211s

5611s DELAY

FIG.90. DME/P transponder.


DISTANCE MEASURING EQUIPMENT IN AVIATION 227
utilize the squitter signal amplitude for performing automatic receiver gain
control.
The priority logic determines the type of transmission to be made when
conflicts occur between the transmission of squitter pairs, ID pairs, and
IA/FA mode replies. FA mode replies, due to their critical nature, will always
take priority over all other ground system transmissions, while IA mode
replies and squitter transmissions are suppressed during the identificationkey
down time.
An integral part of the transponder function is the need to ensure the PFE
instrumentation accuracy of the FA and 1A mode reply delays in an
environment of parameter variations with time and temperature. Tbis is
accomplished in real time via automatic delay stabilization (ADS), a technique
also utilized in modern DME/N ground transponders. The ADS function is
performed by simply measuring the reply delay in both the FA and IA modes
just as an interrogator measures range. Since the ADS is at zero range, the
ADS measurement corresponds to the transponder delay associated with the
assigned channel. If this estimate differs from the desired value, the tran-
sponder delay is adjusted to compensate accordingly.
The ADS corrects all errors that vary slowly with time for a fixed signal
level. It cannot, however, correct rapidly time varying errors which contribute
to the transponder CMN component, such as receiver noise and clock
quantization effects, nor can it correct for components which induce delays
that are amplitude or rise-time dependent such as those which arise due to the
log IF receiver.
The key to understanding the transponder instrumentation error is the
realization that the mean reply delay is the important parameter rather than
the reply delay which occurs on a single interrogation. The reply delay on a
single interrogation may deviate from the desired value, for reasons which will
be discussed below. Such deviations are of two types: those which exhibit a
slowly changing magnitude over a period of many interrogations and those
which change on each interrogation. Those which change slowly over time
contribute to the PFE component of the instrumentation error. Those which
change with each interrogation contribute primarily to the CMN component
of the error. Transponder instrumentation errors arise primarily from the
following sources.
ADS. The transponder reply delay is controlled by the ADS. The ADS
measures the reply delay, estimates the mean reply delay, and corrects the
reply delay accordingly. Errors in this correction result directly in errors in the
transponder mean reply delay, ultimately translating to range errors as viewed
from an interrogator. Hence, the accuracy of the transponder can be no better
than the accuracy of the ADS error estimates and the resolution of reply delay
corrections.
228 ROBERT J. KELLY A N D DANNY R. CUSICK

Reply Delay Clock Rate. The accuracy of the transponder delay is


determined in part by the accuracy of the clock used to time this delay.
Assuming that a digital clock is used, the reply delay will have a small random
component which changes with each interrogation due to the quantization
effectsof the clock, the magnitude of which is directly proportional to the rate.
That is, the higher the clock rate used for reply delay timing, the smaller the
quantization error.
Log Amplijiers. Log amplifiers are recommended for use in each of the I F
channels in order to provide the necessary dynamic range receiver. However,
log amps typically exhibit an imperfect log-linear amplitude transfer curve in
the form of amplitude ripples about the desired response curve. These ripples
distort the pulse leading edge such that pulses of different amplitudes will
exhibit different detection times. In contrast to DME/N, the magnitude of
these deviations is of concern in the DME/P application. The error can be
reduced either by appropriate design methods, which may be costly, or by
utilizing received pulse amplitude information and compensating the reply
delay according to a stored calibration table.
IF Filter. When narrow-bandwidth filters are utilized, such as in the IA
mode channel of the transponder, the delay variation with pulse rise time can
be significant. Specifically, rise-time variations across the population of
interrogators must be considered in the design if the received pulse rise time
can be expected to be different from that used for system calibration. For
example, a five-pole 800 kHz bandwidth Butterworth filter will have an 11 ns
delay variation for input pulses having rise times between 800 and 1200 ns, the
range of rise-time values permitted by interrogator specifications. Bessel filter
implementations will have negligible delay variations. Since a wide bandwidth
is utilized in the FA mode, delay variation with pulse rise time (within the
range allowed by the specifications)is not a concern.
Finally, other factors, such as circuit parameter drifts in the rf, IF, and
digital circuitry, as well as the reply delay clock oscillator frequency,
potentially affect the transponder accuracy. However, such slowly varying
drifts are compensated by the ADS.
b. Transponder Monitor. Since the DME/P transponder is utilized during
critical approach and landing maneuvers where the quality of the data directly
impacts flight safety, the importance of a reliable, fail-safe monitoring system
cannot be over emphasized.
The structure of the DME/P monitor parallels that of the DME/N
monitor described in Section II1,B. However, it must monitor both the IA and
F A modes of the ground system. The primary parameters reflect the status of
the system accuracy and coverage and are the mean reply delay (PFE), the
DISTANCE MEASURING EQUIPMENT IN AVIATION 229
radiated power, the reply efficiency, and the receiver decoder aperture. As in
DME/N, a primary parameter that exceeds its predetermined limit causes the
transponder to be removed from service, or redundant circuitry (if available)is
utilized to keep the system in an operational mode. Such transfers or
shutdowns must occur in less than 1 s to minimize the time that erroneous
guidance is provided to aircraft utilizing the DME/P ground facility. The
facility may remain on the air in the IA mode in the event that the failure is
isolated to FA mode circuitry.
Secondary parameters are those for which a maintenance action only is
required and are pulse-pair output rate, pulse partial rise time, and CMN.
Since these parameters are not critical to flight safety, only an indication to the
ground facility need be provided in the event that a parameter exceeds its
limits.
c. Interrogator. Aside from the inclusion of two operating modes and
improved accuracy, the DME/P interrogator functions in the same way as its
DME/N counterpart. It must (1)identify and validate received pulse pairs; (2)
derive range measurements from replies which are synchronous with the
interrogations; and (3) process the measured range data for output to displays
and the aircraft flight control system.
Identification and validation of received pulse pairs is identical to the
techniques utilized in the ground transponder. Identification of valid pulse
pairs that are synchronous with interrogations is identical to the process used
in DME/N and described in Section 111,B,3.The raw range measurements are
filtered by an appropriate data filter that reduces the effects of the measure-
ment errors due to instrumentation, noise, multipath, and garble and
accommodates missing data conditions for system efficiencies down to 50%.
This filtering of the range data is typically not done in DME/N airborne
systems and is therefore unique to DME/P applications.
The error mechanisms in the DME/P interrogator are identical to those
noted earlier for the transponder, with the exception of the automatic delay
stabilization function. The ADS is replaced by a similar concept where timing
is begun in the interrogator by the interrogation pulse, which is processed
using the same circuitry used to process the replies. In this way, slowly varying
circuit parameters common to both paths are compensated. However,
amplitude and rise-time-dependentdelay variation cannot be so compensated
and must be handled separately.
d. Signal Acquisition. For the IA and F A mode, the interrogator must
acquire and validate the reply signal within 2 s. After loss of the tracked signal,
the interrogator must provide a suitable warning after 1s for the FA mode and
10 s for the IA mode. Signal acquisition is defined to be based on the IA mode.
230 ROBERT J. KELLY AND DANNY R. CUSICK

For DME/P, a more sophisticated range correlation and acquisition


scheme is required than that described for DME/N. Not only must the
interrogator acquire and track in the IA mode, it must automatically begin the
transition to FA mode processing 8 nmi from the transponder. To accomplish
this transition, the interrogator must acquire confidence in the FA mode while
simultaneously tracking in the IA mode. Once FA mode confidence is
acquired, FA and IA mode data are blended to prevent discontinuities in the
range output due to differing bias errors in the two modes. That is, an
increasing proportion of FA mode data is used to generate the output range
information. Once the interrogator moves within 7 nmi of the transponder, the
IA mode interrogations are stopped and only the FA mode is used to ge:ierate
range information. In the absence of FA mode data, tracking may continue in
the IA mode provided that suitable indication of this condition is provided to
the pilot and aircraft systems utilizing the DME/P range data.
e. Interrogator Data Processing Algorithm. The SARPs accuracy specifica-
tions, as well as the error budgets discussed in Subsection C,2, assume that
the higher-frequencynoise contributions are limited by an interrogator output
data filter. Depending upon the user’s application, additional filtering for noise
reduction can be used provided the dynamic response of the aircraft’s flight
control system is not adversely affected.
For a linear filter, the noise reduction factor is given by
sample rate
= (2 x (output filter noise bandwidth) >”’ (40)

and is applicable when the noise spectrum is flat from dc to the sampling
frequency. The 2a range noise level is thus reduced, by filtering, to l/g of its
initial value. Large noise reduction factors correspond to narrow-bandwidth
filters. For a single-pole filter, the noise bandwidth is n/2 x (3 dB bandwidth).
For example, a 10 rad/s single-pole filter utilizing a 40 Hz sample rate has a
noise reduction factor of 2.8 and a phase delay of 4” at 1.5 rad/second.
In addition to data smoothing with adequate transient responses, the data
processing algorithm should perform two other functions that will enhance
the quality of the range data: prediction of missing data samples and limiting
of spurious measurement values (e.g., those due to garble) which would
otherwise deviate significantly from their predicted values.
When the system efficiency is less than loo%, replies will not be received
for all interrogations. To “coast” through periods of missing data, the data
filter algorithm replaces the missing samples with predictions based on past
position and velocity data. A filter with velocity memory is the a-j? filter (Kelly
and LaBerge, 1980; Benedict and Bordner, 1962).An a-j? filter which satisfies
CTOL FA mode applications has a = 0.125 and j? = 0.0078. The noise
DISTANCE MEASURING EQUIPMENT IN AVIATION 23 1
reduction factor for this filter is 3.3 and it will satisfy the “not-to-exceed
aircraft flight control phase delay requirement of 6” (i.e., 4.6” at 1.5 rad/s).
Assuming that the output of the data filter is close to the true range,
significant deviations (greater than 300 ns) from the predicted range are then
physically unrealistic in terms of aircraft response and can therefore be
removed. This process is called “outlier rejection.” If the rejected sample is
replaced by a predetermined value, the entire algorithm is called a “slew-rate
limiter.” The outlier window width and the slew-rate limit value are
determined by the error budget associated with the user application and the
data filter noise reduction factor. Acceptable values are a 300 ns outlier limit
with a 40 ns slew-rate limit. These parameters, together with the above
specified a-8 filter coefficients, permit the DME/P to accurately track all
CTOL maneuvers.
Figure 91 summarizes the a-B filter algorithm which incorporates outlier
rejection and slew-rate limiting.
f. Interrogation Rate Jitter. It is common practice in DME to “jitter” the
time between interrogations to preclude the occurrence of synchronous
interference between interrogating aircraft. This jittering process has the
additional effect of reducing the aliased spectrum peaks of the range error
frequency response such that the sampled frequency components outside the
filter pass band are uncorrelated. This means that sinusoidal errors, such as
those that can be generated by multipath having nonzero scalloping
frequency,appear noiselike to the filter and are therefore attenuated according
to the filter noise reduction factor given in Eq. (40). A jitter value of f50% of
the intersample time is adequate to remove the aliased peaks, as illustrated
in Fig. 92.

E. System Field Tests

The feasibility of the DME/P concept was confirmed by a comprehensive


FAA field test program. Figure 93 is an example flight error trace recorded in
June, 1980, as part of this test program. It illustrates many of the accuracy
concepts discussed earlier. The ground transponder was a modified Cardion
conventional 100 W terminal area DME. The airborne interrogator was a
modified version of a Bendix enroute DME. This equipment constituted the
range element of the FAA’s basic wide MLS which was contracted to Bendix.
The ground station was installed on Runway 22 at Wallops Station Flight
Center, Chincoteague, Virginia, December, 1979. A laser ground tracker was
used within 5 nmi of the runway threshold, and a beacon tracker was used at
ranges greater than 5 nmi. The airborne antenna used on the FAA’s Convair
CV-880 was a standard L-band omni type.
232

i f (data sample available) then


-
Fk = uk - Yk
else if (missing data) then 4
Fk = 0 YK
end i f
A
i'k

&-
I
--- - -- _+_a % ? _ __
OUTLIER
REJECTION
-- AND
-- SLEW R A T E
- --_ -- LIMITER

41

yk

Uk ~ RAN GEMEASUR EMENT


Ek = DIFFERENCE BETWEEN MEASURED A N D PREDICTED RANGE
AUGMENT ED BY OUT LlER iSLEW R AT EiMlSSlNG D A T A ALGORITHM
Vk = VELOC IT Y ESTIMATE
Xk = POSITION ESTIMATE
Yk = PREDICTED POSITION
a,C = F ILT ER COEFFICIENTS
T - INTER-SAMPLE TIME
la1 = OUT LIER REJECTION L I M I T
Ibl ~ SLEW R AT E L I M I T

DESIGN R ELAT ION S FOR OPT IMAL F ILT ER

G I VEN a (3- az
12-ai

3dB BAN D WIDT H IRADIANSISECI ~ JTa


F3dB ~

NOISE BAND WIDT H (RADIANSISECI


FNB - 3 FgdB

PERFORMANCE MEASURES
Variance ~n Parition Output v a m n c e I" VeloClty 0"tP"t
=
'x Variance in ~ a Position
w ~nput 'i = variance in Raw Position InDut

Dxz ~ z0
m ( Unit Ramp)
Lpos,tmn Ramp
-

RBS~OI~IC] Di2 = zo
m IVeloCity o f Unit Rampl-
~ V e l o c ~ tRamp
y Respanse~

(2-a) (i-aI2 ~-i a2iz-al+zP i i - a i


ap re-p-zai T~ a0 i4-0-2ar

FIG.91. Block diagram of a-B filter with outlier rejection, slew-rate limiting, and missing
data capability. The optimal filter provides the maximum noise reduction while providing the best
transient tracking capability to a unit ramp input.
DISTANCE MEASURING EQUIPMENT IN AVIATION 233

-.-.-. .O

--------- .1
.2
-
-................
--- _----.
.3
.4
----
-I-

.5
FIG.92. Frequency response of an a-fi filter for uniformly distributed interrogation jitter
that is Ox,lo%, 20%, 30%, 40%, and 50% of the nominal intersample time of 25 ms. The aliased
peaks are reduced as the jitter percentage is increased. 40 Hz sample rate, a = 0.125, = 0.0078.

A detailed description of these field tests, which verified the validity of the
precision mode concept, is found in Edwards (1981).

F. Conclusion

This article has described the internationally standardized radio aid to


navigation known as “Distance Measuring Equipment” and its evolving role
in support of air navigation in the world today. Based on the beacon concept
developed in the 1940s, DME was first standardized as an international system
by ICAO in the early 1950s.In conjunction with VOR and TACAN, DME has
since become one of the primary navigation aids in the U.S. The most recent
revision of the DME standard provides for two forms of the DME system. The
DME/N serves the needs of the enroute navigation, while the newly adopted
DME/P standard provides the high data rates and accuracy necessary for
approach/landing operations in conjunction with MLS.
VOR and TACAN, together with DME, have defined the basic air route
structure in the U.S. with aircraft flying station to station along inter-
234 ROBERT J. KELLY AND DANNY R. CUSICK
3 DEG CENTERLINE APPROACH RUN 1
JUNE 5,1980
OUTPUT DATA
(FROM FILTER)
150
N 2 2 2 FT (MEASURED)
;
I-
100
U
a- 50
ga o
.
'
-50
w -100
-150
STANDARD 1

-*O0 t
STANDARD 1
PFE ERROR LIMITS

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
N M FROM AZ PHASE CENTER
FIG.93. Example flight error trace from the Wallops Island precision DME field tests.

connecting radials. To put the DME system in its proper perspective, this
article has described the enroute navaids, VOR and TACAN, and the landing
aids, ILS and MLS, showing the role of the DME as an integral component of
these systems.
With the availability of economical airborne computing facilities,
aircraft need no longer fly along VOR radials. Instead, with an RNAV
computer, arbitrary flight paths are possible. In an RNAV environment, the
navigation sensor data, be it DME, VOR, or TACAN, are used to compute the
aircraft's position. The superior position fix performance of a multi-DME-
based RNAV clearly emerges from the alternative position fixing techniques
based on VOR and TACAN.
The basic concepts of aircraft navigation, guidance, and control were
introduced by defining how the information from the navigation sensors is
utilized to execute these functions.In doing so, the characteristics of the sensor
DISTANCE MEASURING EQUIPMENT IN AVIATION 235
output data that affect the performance of an aircraft in both manual and
autocoupled flight were defined.
This article has described the conventional DME system-DME/N-
detailing its primary components and its evolution. The most important
aspects of the DME/N signal in space, accuracy, coverage, the channel plan,
and service capacity, were discussed. The standardization of DME/N has
been achieved through the definition of a handful of system parameters,
particularly the pulse detection and calibration methods. It has been shown
that, through a combination of equipment characteristics, namely frequency
rejection, decoder rejection, and channel assignment rules, an adequate
number of facilities can be accommodated without mutual interference to
serve the enroute, terminal, and approach/landing navigation needs of an
expanding and dynamic aviation industry.
Finally, this article addressed the newest member of the DME family, the
DME/P. DME/P, which provides slant range accuracy on the order of 100 ft
at runway threshold, is to be an integral part of the MLS. Only with DME/P
can the full capabilities of MLS (i.e., full 3D guidance within the airport
environment) become reality. The newly defined ICAO standard represents
the successful integration of the precision DME function with that of
DME/N. Its signal format provides for full interoperability between DME/N
and the initial approach mode of DME/P, eliminating the need for dual
equipment on the ground and in the air. In addition, the DME/P waveform
satisfies the strict spectral requirements imposed on ground facilities and
therefore minimizes frequency compatibility issues.
The discussions have emphasized those factors which make DME/P a
viable system, even in an environment characterized by strong multipath, 50%
missing data, and an ever-increasing number of undesired pulses. Careful
selection of the pulse shape (cos/cos2),the IF bandwidth (4 MHz), and the
pulse detection technique (delay-attenuate-compare), ensures immunity to
multipath in nearly all foreseeable runway environments. Use of even simple
data filter algorithms, incorporating such features as outlier rejection and
slew-rate limiting, will minimize the effects of missing data and accuracy
degradations due to the presence of undesired pulses. Finally, the prevalence
of high-speed digital technology has rendered obsolete the analog and
mechanical timing/tracking techniques of the past, ensuring the availability of
economicalground and airborne equipment exhibitinga 10-foldimprovement
in instrumentation error. When combined, these factors yield a precision
DME system that is compatible with its DME/N counterpart and that
provides the accuracy required to support complex operations in the terminal
area, including automatic landings. In summary, the DME system is one
whose capability and utility will well serve the aviation community into the
next century.
236 ROBERT J. KELLY AND DANNY R. CUSICK

LISTOF ACRONYMS

AiC Aircraft
AFCS Aircraft flight control system
AGC Automatic gain control
ARINC Aeronautical Radio, Inc.
ASR Airport surveillance radar
ATA Air Transport Association
ATARS Automatic Traffic Advisory and Resolution Service
ATC Air traffic control
ATCRBS Air traffic control radar beacon system
AWOP All Weather Operations Panel
CMN Control motion noise
Diu Desired-to-undesired signal ratio
DAC Delay-attenuate-compare
DME Distance measuring equipment
DOD Department of Defense
DOT Department of Transportation
EUROCAE European Organization for Civil Aviation Electronics
FA mode Final approach mode
FAA Federal Aviation Administration
FAR Federal Air Regulation
GPS Global positioning system
IA mode Initial approach mode
ICAO International Civil Aviation Organization
IF Intermediate frequency
IFR Instrument flight rules
ILS Instrument landing system
INS Inertial navigation system
Mi0 Multipath-to-direct signal ratio
MLS Microwave landing system
Mode S A digital data link system (formerly DABS)
NAS National airspace system
NOTAMs Notices to Airmen
PAF Peak amplitude find
PFE Path following error
rf Radio frequency
RNAV Area navigation
RSS Root-sum-square
RTCA Radio Technical Commission for Aeronautics
SARPs Standards and recommended practices
SNR Signal-to-noise ratio
SSR Secondary surveillance radar
ssv Standard service volume
DISTANCE MEASURING EQUIPMENT IN AVIATION 237
TACAN Tactical air navigation system
TCAS Traffic alert and collision avoidance system
VFR Visual flight rules
VOR Very high frequency omnirange transmitters
VORTAC A TACAN colocated with a VOR station

ACKNOWLEDGMENTS

The authors wish to dedicate this chapter to Jonathan and Anne Kelly and Woodrow and
Christopher Cusick. We wish to thank C. LaBerge, W. Reed, and A. Sinsky of the Allied
Corporation-Bendix Communications Division for their helpful technical comments and
G. Jensen and K. Jensen for their much-needed editorial comments. Finally, we would like
to thank M. Morgan and M. ODaniell for their efforts preparing the manuscript and Dick
Neubauer for the excellent graphics.

REFERENCES

Aeronautical Radio, Inc. (ARINC) (1982). “Mark V Airborne Distance Measurement


Equipment -Characteristic 709- 5.” ARINC, Annapolis, Maryland.
All Weather Operations Panel (AWOP) (1980).“Report of the All Weather Operations Panel-
8th Meeting,” Doc. 9310, AWOP/8. ICAO, Montreal, Canada.
A11 Weather Operations Panel (AWOP) (1982). “Report of the All Weather Operations Panel-
9th Meeting,” Doc. 9400, AWOP/9. JCAO, Montreal, Canada.
All Weather Operations Panel (AWOP) (1984). “Report of the All Weather Operations Panel-
10th Meeting,” Doc. 9449, AWOP/lO. ICAO, Montreal, Canada.
Bendat, J., and Piersol, A. (1971). “RANDOM Data; Analysis and Measurement Procedures.”
Wiley (Interscience), New York.
Benedict, T., and Bordner, G. (1962). Synthesis of and optimal set of radar track-while-scan
smoothing equations. IRE Trans. Autorn. Control AC-7,No. 4.
Bennett, J., and Hung, J. (1971). Applications of statistical techniques to landmark navigation. J .
Inst. Nauig. 17, No. 4.
Bleeg, R., et al. (1972).“Inertially Augmented Automatic Landing System: Autopilot Performance
with Imperfect ILS Beam,” FAA Final Rep. No. FAA-RD-72-22. Federal Aviation Admin.,
Washington, D.C.
Bobick, J., and Bryson, J. (1970). Updating inertial navigation systems with VOR/DME
information. AIAA. 11, No. 10.
Braff, R. (1966). “VORTAC Error Analysis for Helicopter Navigation: New York City Area,”
FAA Rep. No. RD-66-46. Federal Aviation Admin., Washington, D.C.
Braff, R. (1972).The effect of geometry on area navigation system errors. J. Znst. Nauig. 19, No. 1.
Braff, R., Shively, A,, and Zeltser, M. (1983). Radio navigation system integrity and reliability.
Proc. ZEEE 71, No. 10.
Britting, K. R. (1971).“Inertial Navigation System Analysis.” Wiley (Intercience), New York.
238 ROBERT J. KELLY AND DANNY R. CUSICK
Brown, R. (1973). Integrated navigation systems and Kalman filtering: A perspective. J . Inst.
Nauig. 19, No. 4.
Brown, R. (1983). “Introduction to Random Signal Analysis and Kalman Filtering.” Wiley,
New York.
Bryson, A., and Bobick, J. (1972). Improved navigation by combining VOR/DME information
and air data. J. Aircr. 9, No. 6.
Burt, A,, et al. (1965). “Mathematical Considerations Pertaining to the Accuracy of Position
Location and Navigation Systems,” AD-629-609. Stanford Res. Inst., Menlo Park,
California.
Busignies, H., Adams, P., and Colin, R. I. (1946). Aerial navigation and trafiic control with
navaglobe., NAVAR, Navaglide, and Navascreen. Electr. Commun. 23(2).
Carel, 0.(1981).“Proposal for a DME/P Employing a Two Pulse Format,” ICAO All Weather
Operations Panel-System Concepts Subgroup, Working Pap. No. M-10. ICAO, Munich,
West Germany.
Chance, B., Hughes, V., MacNichol, E., Sayne, P., and Williams, F. (1949). “Electronic Time
Measurements,” MIT-Radiat. Lab. Ser., Vol. 20. McGraw-Hill, New York.
Colin, R. I., and Dippy, R. J. (1966). The hyperbolic navigation system. IEEE Trans. Aerosp.
Electron. Syst. AESZ, 475-487.
Cusick, D. R. (1982). “A Microprocessor-Based Precision DME Interrogator for the Microwave
Landing System,” Proc. Micro-Delcon ’82, University of Delaware. IEEE Computer Society
Press.
DeJonge, M. (1985).“Flight Management Systems: Their Role is Steadily Increasing,” ICAO Bull.
ICAO, Montreal, Canada.
DelBalzo, J. (1970). Inertial systems and area navigation in the U.S. domestic airspace. J. Inst.
Nauig. 17, No. 4.
Department of Defense/Department of Transportation (1984). “Federal Radionavigation Plan,”
DOT-TSC-RSPA-84-8. DOD/DOT, Washington, D.C.
Dodington, S. H. (1949). Crystal control lo00 MHz for aerial navigation, Electr. Commun. 26(4).
Dodington, S . H. (1980). Development of la00-MHz distance measuringequipment. IEEE Trans.
Aerosp. Electron. Syst. AES16(4).
Dodington, S. H. (1984). Civil navigation aid in ITT. Electr. Commun. 58, No. 3.
Edwards, J. (1981). “DME/P Flight Test Data,” ICAO All Weather Operations Working Group
M-4, Background Inf. Pap. 2, ICAO, Paris.
European Organization for Civil Aviation Electronics (EUROCAE) (1980). “Minimum Perfor-
mance Specification for Ground Distance Measuring Equipment,” ED-31. EUROCAE,
Paris.
Evans, J. E. (1982). “L-Band DME Multipath Environment in the MLS Approach and Landing
Region,” Proj. Rep. ATC-116. MIT Lincoln Laboratories, Cambridge, Massachusetts.
Evans, J. E., and Swett, P. H. (1981).“Results of L-Band Multipath Measurements at Operational
U.S. Airports in Support of the MLS DME-P,” Rep. ATC-109. MIT Lincoln Laboratories,
Cambridge, Massachusetts.
Farrell, J. (1976). “Integrated Aircraft Navigation.” Academic Press, New York.
Feather, J. B. (1985). “Guidance Studies for Curved, Descending approaches using MLS,”
Douglas Aircraft Company, Rep. No. Cl-E87-TN-386, Long Beach, California.
Federal Aviation Administration (FAA). (1985). “Airman’s Information Manual. Part 1.
Basic Flight Manual and ATC Procedures (issued quarterly). U.S. Govt. Printing Office,
Washington, D.C.
Federal Aviation Administration (FAA) (1966). “Technical Standard Order: Airborne Distance
Measuring Equipment, TSO-C66A, Part 37 of Federal Aviation Regulations.” FAA,
Washington, D.C.
DISTANCE MEASURING EQUIPMENT IN AVIATION 239
Federal Aviation Administration (FAA) (1970). “Criteria for Approving Cat. I and Cat.
I1 Landing Minimum for FAR 121 Operations,” FAA Advis. Circ. 120-29. FAA,
Washington, D.C.
Federal Aviation Administration (FAA) (1971).“Automatic Landing System,” FAA Advis. Circ.
20-57A. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1973). “Criteria for Approval of Cat. I11 A Landing
Weather Minima,” FAA Advis. Circ. 120-28A. FAA, Washington, D.C.
Federal Aviation Administration (FAA)(1975a).“Approval of Area Navigation Systems for use in
the US. National Aerospace System,” AC-90-45A. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1975b).“Flight Procedures and Airspace,” FAA Handb.
8260.19. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1976).“United States Standard for Terminal Instrument
Procedures (TERPS),” FAA Handb. 8260.3B, 3rd ed. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1978). “VORTAC Equipment Replacement and Facility
Modernization,” FAA-B2678A. FAA, Washington, D.C.
Federal Aviation Administration (FAA)(1980). “Frequency Management Engineering Principles:
Geographical Separation for NAVAID Frequency Assignments,” Order 6050.5B. FAA,
Washington, D.C.
Federal Aviation Administration (FAA)(1981).“Operation Free Flight,” FAA Rep. No. FAA-AT-
81-1. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1982). “US. National Standard for the Very High
Frequency Omnidirectional Radio Range (VOR)/Distance Measuring Equipment/Tactical
Air Navigation (TACAN) Systems,” AC 00-31A. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1983a). “United States Standard Flight Inspection
Manual,” 3rd ed. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1983b). “Microwave Landing System Interoperability
and Performance Requirements,” FAA-STD-022b. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1983~). “Microwave Landing System Ground
Equipment/Precision Distance Measuring Equipment (DME/P),” FAA-E-2721/3b. FAA,
Washington, D.C.
Federal Aviation Administration (FAA)(1983d).“Random Area Navigation Routes,” FAA Advis.
Circ. AC 90-82. FAA, Washington, D.C.
Federal Aviation Administration (FAA) (1984). “Procedures for Handling Airspace Matters,”
FAA Handb. 7400.2C. FAA, Washington, D.C.
Flint, R. B., and Hollm, E. (1965). VOR evolutionary system improvements in the US. IEEE
Trans. Aerosp. Nauig. Electron. ANEIZ(1).
Fried, W. (1974). New developments in radar and radio sensors for aircraft navigation. IEEE
Trans. Aerosp. Nauig. Electron. AESIO( 1).
Frisbie, F. (1976). “MLS Performance in Rising Terrain,” London BIP No. 12 AWOP Working
Group A, ICAO.
Gelb, A,, and Sutherland, J. R. (1971). Software advances in aided inertial navigation systems. J .
Inst. Nauig. 17, No. 4.
Goemaat, M., and Firje, E. (1979). Analysis of VOR/DME signal coverage over the continental
United States. J. Inst. Nauig. 26, No. 2.
Graham, D., and Lothrop, R. C. (1955). Automatic feedback control and all-weather flying.
Aeronaut. Eng. Rev., October.
Groginsky, H. (1959).Position estimation using only multiple simultaneous range measurements.
IRE Trans. Aerosp. Nauig. Electron. ANE-6(3).
Gum, S., and Walters, B. (1982).“An Invitation to Fly: Basics for the Private Pilot.” Wadsworth
Publishing, Belmont, California.
240 ROBERT J. KELLY AND DANNY R. CUSICK

Hazeltine Corporation (1972). Sperry Rand Report on “MLS Data Format: Analog Computer
Studies,” Rep. 10926, DOT-FA72WA-3804. Hazaltine Corporation, Greenlawn, New York.
Hensler, T. C. (1984). “Upgraded MLS Channel-Assignment Model.” Electromagnetic Compat-
ibility Analysis Center/IIT Res. Inst., Annapolis, Maryland.
Higgins, W. (1975). A comparison of complementary and Kalman filtering. IEEE Trans. Aerosp.
Electron. Syst. AES-11, No. 3.
Hogle, L., Markin, K., and Bradley, J. (1983). Navigation systems performance versus civil
aviation requirements. Proc. IEEE 71, No. 10.
Honold, P. (1971).“Secondary Radar.” Heydon & Son, New York.
Hopkins, J. (1981). Integrated satellite navigation and strapdown attitude and heading reference
for civil air carriers. J. Inst. Nauig. 28, No. 3. 1981.
Hsu, D. (1979). Long-tailed distributions for position errors in navigation. Appl Stat. 28, No. 1.
Huddle, R. (1983).“Inertial Navigation System Error Model Considerations in Kalman Filtering
Applications.” Control and Dynamic Systems, Vol. 20. Academic Press, New York.
Hurley, H. C., Anderson, S. R., and Keary, H. F. (1951). The CAA VHF omnirange. Proc. I R E
39(12).
Institute of Electrical and Electronics Engineers (IEEE) (1973). Special issue on aeronautical
communications. I E E E Trans. Commun. COM-21, No. 5.
International Civil Aviations Organization (ICAO) (1947). “Final Report-First Session,”
Special Radio Technical Division of the Provisional ICAO. ICAO, Montreal, Canada.
International Civil Aviations Organization (ICAO) (1949). “Standards and Recommended
Practices for Telecommunications and Radio Aids to Air Navigation.” ICAO, Montreal,
Canada.
International Civil Aviations Organization (ICAO) (1972a). “Manual on Testing of Radio
Navigation Aids,” Doc. 8071,3rd ed., Vol. 1. ICAO, Montreal, Canada.
International Civil Aviations Organization (ICAO) (1972b). “International Standards and
Recommended Practices- Aeronautical Telecommunications Annex 10,” 3rd ed., Vol. 1.
ICAO, Montreal, Canada.
International Civil Aviations Organization (ICAO) (1981a). Microwave landing system (MLS)
characteristics. In “International Standards and Recommended Practices- Aeronautical
Telecommunications Annex 10,” 3rd ed., Vol. 1, Sect. 3.11, Chapter 3. ICAO, Montreal,
Canada.
International Civil Aviations Organization (ICAO) (1981b).“Report of the ICAO Communica-
tions Divisional Meeting,” Doc. 9341. ICAO, Montreal, Canada.
International Civil Aviations Organization (ICAO) (1982). “Procedures for Air Navigation
Services, Aircraft Operations,” Doc. 8168-0PS/611, 2nd ed., Vol. 2. ICAO, Montreal,
Canada.
International Civil Aviations Organization (ICAO) (1985).UHF distance measuring equipment.
I n “International Standards and Recommended Practices- Aeronautical Telecommunica-
tions Annex 10,” 4th ed., Vol. 1, Sect. 3.5, Chapter 3. ICAO, Montreal, Canada.
James, H., et al. (1947). “Theory of Servomechanisms,” MIT-Radiation Lab. Ser., Vol. 25.
McCraw-Hill, New York.
Jensen, R. (1976). Effects of 3-D RNAV approach procedural complexity. J. Inst. Nauig.22, No. 2.
Johnson, M. E., and Gierhart, G. D. (1978a). “The IF-77 Electromagnetic Wave Propagation
Model,” DOT/FAA/ES-83. US. Department of Commerce, Institute for Telecommunica-
tion Sciences, Boulder, Colorado.
Johnson, M. E., and Gierhart, G. D. (1978b). “Applications Guide: Propagation and Interference
Analysis Computer Programs (0.1 to 20 GHz),” FAA-RD-77-60. U.S. Department of
Commerce, Institute for Telecommunication Sciences, Boulder, Colorado.
Johnson, M. E., and Gierhart, G. D. (1980).“An Atlas of Basic Transmission Loss for 0.125 to 15.5
DISTANCE MEASURING EQUIPMENT IN AVIATION 24 1
GHz,” FA-RD-80-1. US. Department of Commerce, Institute for Telecommunication
Sciences, Boulder, Colorado.
Johnson, M. E., and Gierhart, G. D. (1983). “Computer Programs for Air/Ground Propagation
and Interference Analysis(0.1 to 20 GHz),”FAA-RD-73-103.U.S. Department of Commerce,
Institute for Telecommunication Sciences, Boulder, Colorado.
Karatsinides, S., and Bush, R. (1984). “Navigation Processing of the Flight Management
Computer System for the Boeing 737-300,” Proc. 40th Annu. Meet. Institute of Navigation,
Cambridge, Massachusetts.
Kayton, M., and Fried, W. (1969). “Avionics Navigation Systems.” Wiley (Interscience),
New York.
Kelly, R. J. (1963).“A New Light-Weight DME,” Proc. 10th Annu. East Coast Conf. IEEE Aerosp.
Navig. Electron.
Kelly, R. J. (1974). “System Design and Flight Test Results of the Bendix/Bell CAT II/III
Elevation Approach Guidance Function,” Proc. AIAA Mech. Control Flight Conf., AIAA
Pap. 74-909. Am. Inst. Aerosp. Admin., Anaheim, California.
Kelly, R. J. (1976) Time reference MLS multipath control techniques. J. Inst. Naoig. 23,No. 1.
Kelly, R. J. (1977). Guidance accuracy considerations for the microwave landing system. J . Inst.
Naoig. 24, No. 3.
Kelly, R. J. (1984). System considerations for the new DME/P international standard. IEEE
Trans. Aerosp. Electron. Syst. AE-520,No. 1, 1-24.
Kelly, R. J., and LaBerge, E.F.C. (1980). Guidance accuracy considerations for the microwave
landing system precision DME. J. Inst. Naoig. 27,No. 1.
Kelly, R., Redlien, H., and Shagena, J. (1978). Landing aircraft under poor conditions. IEEE
Spectrum 15(9).
Koshar, A,, and Smithmeyer, J. (1982). “MLS Channel Plans and Traffic Loading,”
DOT/FAA/RD-81-113. Federal Aviation Admin., Washington, D.C.
Kuo, B. (1962). “Automatic Control Systems.” Prentice-Hall, Englewood Cliffs, New Jersey.
Latham, R. (1974). Aircraft positioning with multiple DME. J. Inst. Nauig. 21, No.2.
Latham, R., and Townes, R. (1975).DME errors. J. Inst. Nauig. 22,No. 4.
McRuer, D., and Johnson, W. (1972).“Development of Approach Control System Requirements
with Application to a Jet Transport,” NASA CR-2023. Natl. Aeronaut. Space Admin.,
Washington, D.C.
McRuer, D., et al. (1973). “Aircraft Dynamics and Automatic Control.” Princeton Univ. Press,
Princeton, New Jersey.
Maybeck, P. (1979). “Stochastic Models, Estimation, and Control,” Vol. 1. Academic Press,
New York.
Mercer, J. (1954). A quantitative study of instrument approach. J . R. Aeronaut. Soc. 58, No. 518.
Mertikal, J., et al. (1985).Treatment of navigational accuracies: Proposals for the future. J. Insr.
Nauig. 32, No. 1.
Milliken, R., and Zoller, C. (1978). Principles of operation of NAVSTAR and system
characteristics. J. Inst. Nauig. 25,No. 2.
Moseley, F., and Watts, C. (1960). Historic paper on automatic radio flight control. IRE Trans.
Aeronaut. Nauig. Electron. ANE-’I(1).
National Telecommunication and Information Administration (NTIA) (1984). “Table of
Frequency Allocations and Other Extracts From: Manual of Regulations and Pro-
cedures for Federal Radio Frequency Management.” U.S. Department of Commerce,
Washington, D.C.
Neal, G. (1975) MLS-navigation, guidance, and control. J. Inst. Naoig. 22,No. 4.
Office of Technology Assessment (OTA) (1982).“Airport and Air Traffic Control System,” DTA-
ST1-175. OTA, Washington, D.C.
242 ROBERT J. KELLY AND DANNY R. CUSICK

Office of Telecommunications (1978). “EMC Analysis of JTIDS in the 960-1215 MHz Band,”
Vols. 1-7. US. Department of Commerce/Office of Telecommunications, Washington, D.C.
Pierce, J. A. (1948). Electronic aids to navigation. Adu. Electron. 1,425.
Porter, W. A., ODay, J., Scott, R., and Volq R. (1967). “Modern Navigation Systems,”
Engineering Summer Course. University of Michigan, Ann Arbor.
Quinn, G. (1983). Status of area navigation. J. Inst. Nauig. 30.
Radio Technical Commission for Aeronautics (RTCA) (1946). “Report of SC-21 Meetings.”
RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1948a).“Air Traffic Control,” RTCA DO-
12, Spec. Comm. 31. RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1948b). “Final Report of SC-40,” Doc.
121/48/CO 24. RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1963). “Standard Performance Criteria for
Autopilot/Coupler Equipment,” Doc. DO-1 18 (SE-79). RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1978). “Minimum Performance
Standards Airborne DME Equipment Operating Within the Radio Frequency Range of
960-1215 MHz,” Doc. DO-151A. RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1982). “Minimum Operational Perfor-
mance Standards for Airborne Area Navigation Equipment Using VOR/DME Reference
Facility Inputs,” RTCA DO-180. RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1984). “Minimum Operational Perfor-
mance Standards for Airborne Area Navigation Equipment using Multi-Sensor Inputs,”
RTCA DO-187, RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1985a). “Minimum Operational Perfor-
mance Standards for Airborne MLS Area Navigation Equipment,” RTCA Pap. 532-85/SC-
151-79,5th Draft. RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1985b). “Minimum Performance
Standards Airborne DME Equipment Operationwithin the Radio-Frequency Range of
960-1215 MHz,” Doc. DO-189. RTCA, Washington, D.C.
Radio Technical Commission for Aeronautics (RTCA) (1985~).“Minimum Aviation System
Performance for the Global Position System,” Spec. Comm. SC-159. RTCA, Washington,
D.C.
Ragazzini, J., and Fanklin, G. (1958).“Sampled Data Control Systems.” McGraw-Hill, New York.
Redlien, H., and Kelly, R. (1981). Microwave landing systems: The new international standard.
Adv. Electron. Electron Phys. 57,311.
Roberts, A. (1947). “Radar Beacons,” MIT-Radiat. Lab. Ser., Vol. 3. McGraw-Hill, New York.
Rochester, S. I. (1976). “Take-off at Mid-Century: FAA Policy 1953-1961.” Federal Aviation
Admin., Washington, D.C.
Roscoe, S. (1973). Pilotage error and residual attention: The evaluation of a performance control
system in airborne area navigation. J. Inst. Nauig. 20, No. 3.
Ruhnow, W., and Goemaat, J. (1982).VOR/DME automated station selection algorithm. J . Inst.
Nauig. 29, No. 4.
Sandretto, P. C. (1956). Development of TACAN at federal telecommunications laboratories.
Electr. Commun. 33(1).
Sandretto, P. C. (1958). “Electronic Aviation Engineering.” International Telephone and
Telegraph, New York.
Sandretto, P. C. (1959). Principles of electronic navigation systems. IRE Trans. Aeronaut. Nauig.
Electron. ANEX(4).
Schmidt, S . (1966).Application of state space methods to navigation problems. Adv. Control Syst.
3.
DISTANCE MEASURING EQUIPMENT IN AVIATION 243
Schneider, A. (1959). Vector principles of inertial navigation. IRE Trans. Aerosp. Nauig. Electron.
ANE-6(3).
Silver, S . (1949). “Microwave Antenna Theory and Design,” MIT-Radiat. Lab. Ser., Vol. 12.
McGraw-Hill, New York.
Steiner, F. (1960).Wide-base doppler VHF direction finder. IRE Trans. Aerosp. Nauig. Electron.
ANE-1.
Tanner, W. (1977). Inertially aided scanning DME for area navigation. J . Inst. Naoig. 24, No. 3.
Tyler, J., Brandewei, D., Heine, W., and Adams, R. (1975). Area navigation systems: Present
performance and future requirements. J . Inst. Nauig. 22, No. 2.
Wax, M. (1981). Position location from sensors with position uncertainty. IEEE Trans. Aerosp.
Electron. Syst. AES19, No. 5.
Williams, C., Bowden, B., and Harris, K. (1973). Pioneer award for contribution in field of
secondary radar. IEEE Trans. Aerosp. Electron. Syst. AES-9, No. 5.
Winick, A. B., and Hollm, E. (1964). “The Precision VOR System,” IEEE International
Convention, New York.
Zimmerman, W. (1969). Optimum integration of aircraft navigation systems. IEEE Trans.
Aerospace Electron Syst. AFS-5(5).
ADVANCES IN ELECTRONICS A N D ELECTRON PHYSICS, VOL. 68

Einstein-Podolsky-Rosen Paradox and Bell’s Inequalities

W. DE BAERE
Seminarie voor Wiskundige Natuurkunde
Rijksuniversiteit Gent
B-9000 Gent, Belgium

1. INTRODUCTION

Quantum mechanics (QM) is a statistical theory that makes predictions


about the outcome of measurements on microphysical systems. The peculiar-
ity of this physical theory is, according to the standard, or Copenhagen,
intepretation of QM, that its statistical predictions are claimed to be inherent
and cannot be explained further by completing our knowledge by some
supplementary physical parameters.
This inherent statistical feature of QM is so strange as compared with all
known (classical) theories before the advent of QM, that resistance to accept
such a counterintuitive world view has always been present in the scientific
community, from the early days of QM. On the contrary, classical theories
give us a rather comfortable feeling just because our most intuitive notions of
physical systems (e.g., having definite properties of their own, being localized
and separable from the rest of the world) are somehow inherent in this classical
kind of theory.
Moreover, the dogmatic claim of the proponents of the Copenhagen
School, that at the atomic level physical events are no further analyzable in
more detail than is allowed by QM (e.g., the impossibility of a space-time
picture or analysis of individual events) has resulted in a critical attitude by a
large number of scientists,among whom are de Broglie, Schrodinger, Einstein,
and Bohm. In the words of Stapp (1982123, the origin of the uneasy feeling of
those people with the present state and orthodox interpretation of quantum
theory (QT) may be understood as follows: “Some writers claim to be
comfortable with the idea that there is in Nature, at its most basic level, an
irreducible element of chance. I, however, find unthinkable the idea that
between two possibilities there can be a choice having no basis whatsoever.
Chance is an idea useful for dealing with a world partly known to us. But it has
no rational place among the ultimate constituents of Nature.”
245
Copynght @ 1986 by Academic Press, Inc.
All rights of reproduction in any form reserved.
246 W. DE BAERE

The most important critical quantitative analysis of QM appeared in 1935


when Einstein, Podolsky, and Rosen (EPR) (Einstein et al., 1935) made a
thorough analysis of the orthodox interpretation of QT, based on a
Gedankenexperiment with correlated systems. Starting from a realistic
conception of nature, according to which physical systems may have
independent intrinsic properties (whether observed or not), EPA came to
conclusions which contradicted QM (the EPR paradox), from which they
inferred that QM was not a complete theory. This conclusion was at the origin
of the famous Bohr-Einstein debate over the interpretation of QT, the end of
which is now, 50 years later, apparently not yet in sight, in spite of claims to the
contrary.
An important step forward in the clarification of the EPR problem came
from a development by Bell (1964), who used the ideas of EPR to set up a
simple inequality, the Bell inequality (BI), which has to be satisfied if EPR are
correct. There are several reasons for which this BI is interesting. First of all, it
is violated by QT in appropriate circumstances. Furthermore, the original
EPR issue about the completeness(a rather vague concept which is not easy to
define)of QT has shifted to the question whether or not QT is compatible with
locality and separability. Finally Clauser et al. (1969) showed that this issue of
basic importance may be decided on the empirical level. Hence, as a result of
these developments initiated by Bell, the original EPR Gedankenexperiment
has become since then within the reach of experimental realization. The result
of most of the existing correlation experiments agree very well with QT and
violate the BI, from which it is concluded that on the quantum level
nonlocality is now a basic property. However, it appears that, due to various
criticisms, this conclusion is not yet final.
It is the purpose of this article to present a critical review of the problems
related to the EPR argumentation and to the Bell inequalities. As a
complement to this review we advise the reader to consult other books or
reviews on these subjects, e.g., Hooker (1972),Belinfante(1973),Scheibe(1973),
Jammer (1974),Clauser and Shimony (1978), Selleri and Tarozzi (1981),Selleri
(1983b),and DEspagnat (1983, 1984).
In the following we will use the terms “EPR paradox,” “EPR problem,”
and “EPR argumentation” interchangeably,without bothering about which is
preferable.

11. THEEINSTEIN-PODOLSKY-ROS~
ARGUMENTATION

A . Original Version

EPR made their analysis on a pair of correlated physical systems for which
they showed that, according to some set of particular definitions and
EINSTEIN-PODOLSKY -ROSEN PARADOX 241
hypotheses, a physical system can have both position and momentum, in
contradiction to the QM postulates, according to which noncommuting
observables cannot be determined or known simultaneously.
The main definitions in EPR are those of “complete theory” and of
“element of physical reality.” As a necessary requirement for a complete
theory they adopt the following definition:
(Dl) Complete theory: “Every element of the physical reality must have a
counterpart in the physical theory.” As to the concept of “element of physical
reality” which is contained in (Dl), EPR use the following definition, which
they find reasonable enough for their purpose.
(D2) Element of physical reality: “If, without in any way disturbing a
system, we can predict with certainty (i.e., with probability equal to unity) the
value of a physical quantity, then there exists an element of physical reality
corresponding to this physical quantity.”
To illustrate the meaning of these definitions, consider a quantum-
mechanical system S in one dimension x, described by a wave function $(x). If
$(x) is an eigenfunction of an operator A with eigenvalue a

444 = M x ) (1)
then measurement of A will give with certainty the value a. Moreover, from the
preparation of the state of the ensemble of systems S, this result is known even
without measuring A, i.e., without disturbing S. According to (D2)there exists
in this case an element of physical reality corresponding to the physical
quantity A. Using (Dl) this element of physical reality must have a
counterpart in the underlying theory, QT. This must then be the case for every
other physical observable quantity C with operator C for which Eq. (1) is valid

C W ) = c$(x) (2)
i.e., [A, C] = 0. Hence, according to (Dl) the only allowed elements of
physical reality within orthodox QT are those corresponding to a complete set
of commuting observables if the ensemble of systems S is described by an
eigenfunction of one of them.
If, on the contrary, the operator C corresponding to the physical quantity
C does not commute with A: [ A , C ] # 0, then one has

W(X) # clCl(4 (3)


i.e., the result of measuring C on this state is not known without disturbing S.
Certainty about the value of C, among the different possible values, will in this
case only be obtained by actually measuring C, i.e., by disturbing S. Hence, as
long as C is not measured actually, with C there does not correspond an
248 W. DE BAERE

element of physical reality of S , according to (D2). Up to now all this is


consistent with orthodox QM.
However, at this point EPR remark that, in order to criticize the
completeness of QT, it is sufficient to find at least one particular situation for
which all the above is not valid. This means that, if it can be shown that A and
C are elements of physical reality of S, even if [ A , C ] # 0, then the alleged
completeness of QT would no longer be tenable. Now, to show that this
actually happens in some particular cases is the intention of the EPR
reasoning. In particular by their argumentation they try to show that
Heisenberg’s uncertainty principle may be violated.
To this end EPR consider two correlated systems S , and S, which are
supposed to interact only in a finite time interval (0, T ) .Suppose that for t > T
the wave function $(x1,x2)of the combined system ( S , , S , ) is known
completely.
At the basis of the EPR argumentation, which leads to their contradictory
conclusions, are furthermore the following three hypotheses:

(H 1) Einstein locality (separability), or the validity of the relativity


principle: in any reference frame physical influences between nonseparated
physical systems S , and S2(ie., between interacting systems) should propagate
with a maximal velocity c, the velocity of light. If S , and S2 are separated, then
no mutual influence at all should exist. In both cases, the idea of action at a
distance or influences propagating with infinite velocity is considered as an
absurdity and empirically incorrect. As a consequence, a measurement on one
of two separated systems, say S , , cannot influence at all the outcome of a
measurement made on the second system S , , if the two measurements are
separated in space. Einstein, in his proper and direct way of expressing his
intuitive insights in such fundamental matters, expressed his view on locality
as follows (Einstein, 1949):“But on one supposition we should, in my opinion,
absolutely hold fast: the real factual situation of the system S, is independent
of what is done with the system S , , which is spatially separated from the
former.”
(H2) Universal validity of Q M : In particular it is supposed that the process
of state vector reduction applies when a certain measurement is made, or when
information about a system is gained.
(H3) Counterfactual dejiniteness (counterfactuality)(Stapp, 1971) is valid:
This is essentially the assumption that, if on a certain system S , an observable
A is measured with the result a, it does make sense or one is allowed to
speculate about what would have been the result if, instead of A , another
(eventually noncompatible) observable B were measured on precisely the same
system S, in exactly the same individual state [which is not contained in the
quantum formalism, but may be contained in some future more fundamental
EINSTEIN-PODOLSKY-ROSEN PARADOX 249
hidden variable theory (HVT)]. Hence the above hypothesis amounts to
assuming that physically meaningful conclusions may result from combining
actual measurement results with hypothetical ones.
Suppose then that at t > T we measure the physical quantity A on S , , with
eigenvalues a , , a , , . . . , and corresponding eigenfunctions u , ( x , ) , u,(x,), . . . :
= akuk(xl) (4)
According to standard QM rules we have to write II/(x,,x2) in terms of the
wave function U k ( x 1 )
x2)
$(xl, = uk(x1)$k(x2) (5)
k

with $ k ( x z ) the expansion coefficients. If the result of the measurement is ak,


then the prescription of the reduction of the wave function says that I&,,x 2 )
reduces to one single term, namely uk(xl)II/k(xz).In QM this must be
interpreted to mean that, after measurement of A on S , , S , must be described
quantum mechanically by uk(x1) and S, by II/k(x2). Again, up to this point
everything is quantum mechanically correct. I think that at this point it is
important to note that, in order to understand the origin of the EPR paradox,
the foregoing conclusion is correct only if the measurement of A on S , has
actually been carried out. If not, the wave function $(xl, x,) does not reduce to
one single term; in particular S, is not described further by $ k ( x Z ) .
Now, EPR would propose to measure, instead of A, another quan-
tity B with eigenvalues b , , b,, . . ., and corresponding eigenfunctions
u 1(x 1 u 2 (x 11, . * *
19

BUk(X1) = bkUk(X1) (6)


Instead of Eq. (5) one should now write
$(xl,xZ) =~uk(x1)(Pk(x2) (7)
k

This means that if B is measured actually (say at t = t o ) with the result bk,then
$ ( x l , x 2 )reduces again to one single term, namely uk(xI)(Pk(xz),and after this
measurement any further predictions for measurements on S , should be
derived from its wave function (Pk(X2).
In the foregoing we have stressed the condition that the measurements
were actually carried out. It is important at this point to note that the EPR
reasoning violates what is allowed by standard QT: Indeed, if A and B are
incompatible, then either their values cannot be measured or known
simultaneously, or A and B are measured simultaneously but on different
systems, prepared analogously by different sources. Now it is precisely this
violation of QM which EPR allow themselves in their argumentation, namely,
as they state it: “If, instead of this, we had chosen another quantity, say B,
250 W.DE BAERE
having the eigenvalues bl, b,, ...,and eigenfunctions ul(xl), uz(xl), u3(x1),.. .,
we should have obtained,. . ..”
When discussing the Bell inequalities in Section 111, we will see, however,
that this counterfactual reasoning, “If, instead of this,. .,we had chosen . ..we
should have obtained ...” lies also at the origin of the disagreement between
QM predictions and the BIs. Furthermore we will see (Sections I11 and IV)
that the BIs are violated empirically in favor of QM. We will also make
plausible there that it is the breakdown of counterfactuality (H3), rather than
of locality (Hl), that is at the origin of the disagreement between BI and QT.
Hence, at this point it already seems reasonable to suspect that it is incorrect
and unjustified to speculate about the outcome of an already-carried-out
experiment if some parameters of the total experimental arrangement were
changed. In fact, on closer examination, it appears that it is reasonable to
assume that, once an experiment is carried out, a subsequent experiment can
never be done again under exactly the same conditions as those which were
responsible for the specific outcome in the actual experiment.
Therefore, it is to be expected that the result of any argumentation, like the
EPR, which violates at some stage some QM principle, will lead to conclusions
which are contradicted by QM (see below).
Before continuing the original EPR reasoning to its end, let us remark that,
even before choosing the alternative to measuring B, there already exists a
conflict with the initial assumption, namely that for t > T the wave function is
represented by the pure state, Eq. (5). Let us consider therefore the
consequences of the hypothesis of locality (H1)and of the universal validity of
QT (H2) and show that these hypotheses lead to contradictory statements.
Indeed, for simplicity suppose for a moment that the sum on the right-hand
side of Eq. (5) contained only a finite number of terms, and let us introduce
time

Then the measurement of A = uk at c = t o > T reduces J/(x1,xZrt) to


uk(x1, tO)+k(XZ, to). Now, because of the absence of any mutual influence
between S1 and S, from t = T on, it may be concluded that S2 is also de-
scribed by +&, t) for all t satisfying T < t c t o . This means that a statisti-
cal ensemble of couples (Sl,S,) for t > T may be decomposed into a series
of subensembles described by ul(xl,t)+l(xZ,t),uZ(xl,t)+z(xZ,c),...; i.e., the
original ensemble should be described by a mixture of states uk(x1, t)+k(xz, t),
which is in contradiction to the original assumption of the ensemble being
described by the pure state, Eq. (5). There is also an observable difference
between the two possible descriptions (Furry, 1936a; Selleri, 1980; Selleri and
Tarozzi, 1981). In the literature, the last description is by means of so-called
EINSTEIN-PODOLSKY -ROSEN PARADOX 251
“state vectors of the first type” (corresponding to proper mixtures), while the
first way is by means of “state vectors of the second type” [i.e., Eq. ( 5 ) ]
(corresponding to improper mixtures) (DEspagnat, 1965). QM observables
which have different values in the two schemes are sometimes called “sensitive
observables” (Capasso et al., 1973; Baracca et al., 1976; Bergia and Cannata,
1982; Cufaro Petroni, 1977a,b; Fortunato et al., 1977). For later reference we
shall call the above argumentation, which uses only hypotheses (Hl) and (H2),
the modified or simplified EPR argumentation.
From the assumed legality of measuring B instead of A on S,, EPR
conclude: “We see therefore that, as a consequence of two different measure-
ments performed upon the first system, the second system may be left in states
with two different wave functions. On the other hand, since at the time of
measurement the two systems no longer interact, no real change can take place
in the second system in consequence of anything that may be done to the first
system. This is, of course, merely a statement of what is meant by the absence
of an interaction between the two systems. Thus, it is possible to assign two
diflerent wave functions (in our example, $k and rp,) to the same reality (the
second system after the interaction with the first).”This conclusion contradicts
the QM formalism and constitutes what is known as the EPR paradox. Note
again that here the spirit of QM is violated because it is supposed that the
wave function represents the system itself.
Inasmuch at the time of EPR much attention had been paid to whether or
not physical systems (viewed as “particles”) followed individually a well-
determined path (a world line) in space-time, in the remaining part of their
paper EPR are concerned with showing by an explicit example of the above
conclusion that one and the same physical reality may be described by
eigenfunctionsof the two noncommuting operators P and Q: [P, Q ] # 0. To
this end EPR consider as the wave function describing the couple (S,, S,)

vwX19 xz) = 1 m

-m
expCi(x1 - X Z + XO)P/51 dP (9)

with xo being a constant. For observable A they choose the momentum


PI = h/i a / a x , of S , with eigenvalues p and eigenfunctions
u,(x 1) = exp(ipx 1 lh) (10)
Because of the continuity of p , Eq. (5) takes the form

$(XI,XZ) = 1
m

-m
up(x1)$p(x2)dp (1 1)

where the wave function for S,


$Jx2) = exPC - i(x, - xo)p/hl (12)
252 W.DE BAERE
is an eigenfunction of P2 = h/i a/ax, with eigenvalue -p. Applying the above
general results of the EPR analysis we have to conclude the following. Sup-
pose the measurement of Pl on s, gives the result p. Then $(x1,x2) in
Eq. (11) reduces to ~ , ( x , ) ~ ~ ( x ,This
) . allows us to conclude that P2 = - p
without disturbing in any way S, (S, and S, are separated by assumption).
Hence P, = --p is an element of reality for S,.
For the observable B EPR take the coordinate Q , of S,: B = Q1,with
eigenvalues x and eigenfunctions ux(xl) = 6(x, - x). $(x,,x,) may now be
written

rm
with

(Px(x2)= expCi(x - x2 + XO)P/hl dP

= hd(x - x, + xg) (14)


which is an eigenfunction of Q , with eigenvalue x2 = x + xo. Applying again
the EPR criteria we have: If the measurement of Q1 on S, has the result
Q1 = xl, then $(xl, x,) reduces to u,(xl)~,(x2).This allows EPR to conclude
+
that x2 = x xo without in any way disturbing (again by assumption) S,.
+
Hence, x2 = x xo is also an element of reality for S,. Now [P2,Q2] =
h/i # 0, and from this particular example EPR conclude that the same physi-
cal reality, S,, may be represented by two wave functions which may be
eigenfunctions of two noncompatible observables P2 and Q 2 , in contradic-
tion with what is allowed by orthodox QT. Hence they conclude that QT is
not complete, because in its formalism either position or momentum is con-
tained, whereas according to EPR both should be contained in it as elements
of physical reality. But remember the crucial role played by the hypothesis
of counterfactuality (H3) in the EPR argumentation. The particular situa-
tion above considered by EPR (measurement of P, or Q1 on one member of
two coupled systems) is up to now only a kind of Gedankenexperiment.
However, recently Bartell (1980b) has shown how the EPR proposal can be
transformed into a real experiment. It has been shown that inequalities of
the Bell type (Section 111) can be constructed, on the basis of which the va-
lidity of the EPR hypotheses (Hl) and (H3) may be verified. However, we
consider it as not justified, as Bartell argues, that violation of the proposed
inequalities should be considered as evidence for the existence of nonlocal
influences. Indeed, as we will show in Section V, locality or the relativity
principle may be saved by assuming a breakdown of counterfactuality (H3)
as a more plausible origin of an eventual violation.
EINSTEIN-PODOLSKY -ROSEN PARADOX 253
B. E P R Paradox in the Bohm Version

Instead of measuring two noncompatible observables A = P and B = Q


which have a continuous spectrum of eigenfunctions, Bohm (1952a)proposed
to consider two systems S , , S , having the property of possessing spin 3.
Instead of the original EPR Gedankenexperiment, this proposal has after-
wards turned out to be realizable experimentally.
In this case $ ( x , , x , ) has to be replaced by x ( x l , x z ) $ ( a l , a z ) , with
$(u1,a2)being the spin part of the wave function and the Pauli matrices,
a,,a, representing the spins of S, ,S, . Suppose the couple (S, ,S,) is prepared
somehow in the singlet spin state: $(u1,a2)=I+s) = 100). For simplicity it
may be supposed that we have a spin-0 system S(0)which decays according to
+
S(0) + Sl(# S,(i). If this process ends at T, then for t > T, S1 and S2 may
be supposed to be separated or at least no longer influencingeach other. From
elementary QM it follows that, if angular momentum is conserved in the decay
process, 100) may be written

I$J = 100) = (1/Jz)(lu:Y)Iub3 -b


.I">":).!I (15)
in terms of basis vectors II&) for S , and IU:~) for S2 in which S , , S 2 have
spin projections & 1 (in appropriate units) along an arbitrary unit vector a.
Let us now illustrate the EPR argumentation on this conceptually simple
example of correlated systems with only dichotomic variables. Take as a
-
physical observable A of s,:A = a, z = q Z= an operator for the measure-
ment of the spin projection m,,(z, t) along the unit vector z . The eigenvalues
are f 1, and the respective eigenvectors, lu$!), Iui!!).
An alternative option is to take as an observable B for S,: B = a, y = -
olY= an operator for the measurement of the spin projection ms,(y, t) along
the unit vector y. The eigenvalues are again k l , and the respective eigen-
functions, It&)>, IuY2).
-
EPR option ( I ) : Measure A = a, z, with the result m,,(z, t ) by sending S ,
through a Stern-Gerlach device SG, ( z ) with its analyzer direction (i.e., its
magnetic field) along z . According to Eq. (15) we should write for this case

loo) = ( l / J z ) ( ~ u : ~ ) ~ u : ? )- ~ u ~ ! ! ) ~ u : ~ ) ) (16)
From Eq. (16) it is seen that m,,(z, t) = + 1 and m,,(z, t) = - 1 will be found
with equal probability = 3. Suppose that at t = to the result obtained is
ms.(z, t o ) = + 1. According to QT, 100) reduces to l u ~ ! ) l u : ? ) .It follows then
that for t 2 to measurement of msz(z,t 2 to) will give - 1 with certainty
without disturbing S, . According to the EPR criterion, this means that with
msz(z, t 2 t o ) = - 1 there corresponds an element of physical reality and this
must have a counterpart in a complete theory.
254 W. DE BAERE

EPR option (2): Applying the counterfactuality hypothesis (H3), one can
argue that one could equally well have measured the quantity B = a, * y
with the result ms,(y, to), by replacing SG,(z) by SG,( y). Instead of Eq. (16)
we must now start from
loo) = (l/fi)(luy2)lup) - luy?)lu$3) (17)
Again msl(y, to) = + 1 and ms,(y, t o ) = - 1 will be found with equal prob-
ability 3. Suppose again that at t = to we would have found msl(y, to) = - 1.
This measurement may now be considered as a state preparation of S2 for
t 2 to: In fact 100) is reduced by the measurement process to the single term
1ur?)1u1‘+‘). As a result of this process, the result ms,(y, t 2 t o ) = + 1 is then
predicted with certainty, again without disturbing S2. Hence with this value
for ms,(y, t 2 to) there must also correspond an element of physical reality
and this must again have a counterpart in a complete theory. Now this is
certainly not the case in orthodox QT because of the noncompatibility of oZty
and 02*: [02,, 02,] # 0. Hence the EPR conclusion that QT is incomplete.
As already mentioned, this incompletness conclusion may be avoided by
insisting that for this to be true, ms2(z,t) and ms,(y, t) should be available at
the same time. This way out of the problem is anticipated by EPR; indeed they
offer it themselves at the end of their analysis: “One could object to this
conclusion on the grounds that our criterion of reality is not sufficiently
restrictive. Indeed, one would not arrive at our conclusion if one insisted that
two or more physical quantities can be regarded as simultaneous elements of
reality only when they can be simultaneously measured or predicted. On this
point of view, since either one or the other, but not both simultaneously, of
the quantities P and Q can be predicted, they are not simultaneouslyreal. This
makes the reality of Q and P depend upon the process of measurement carried
out on the first system, which does not disturb the second system in any way.
No reasonable definition of reality could be expected to permit this.” How-
ever, their remark in the last sentence of this quotation can be circumvented
rather easily by extending their definition of an element of reality; indeed each
measured value should be considered as representing an element of reality not
only of the object system but of the combined system, object plus measuring
apparatus.
However, the simplified EPR argumentation in the Bohm version is again
much more difficult to reject. To see this, let us apply the EPR reasoning on
each of the couples ( S , , S 2 ) (Selleri and Tarozzi, 1981):
(1) Measure mS,(a,tO)= + 1 at an instant t = to when S,,S2 are sep-
arated (located) so far that it may be supposed that they no longer influence
each other (principle of separabilty).
(2) From the standard QM formalism (assumed to be universally valid
EINSTEIN-PODOLSKY -ROSEN PARADOX 255
by EPR), it follows then that at this moment the state 100) reduces to
lu:!+))lu:?), which describes henceforth (Sl,S,) for t 2 t o . In particular, as a
result of measuring ms,(a, to), the state describing S, is now luf!) for t 2 t o .
(3) Because of the supposed separability and absence of any influence,
this must also have been the case for t < to (at least for those times where
separability holds); the state vector of S , must have been lu:?) and
mS2(a,t < t o ) = - 1.
(4) From msI(a, t) + ms2(a, t) = 0 (conservation of the component of the
total spin along an arbitrary direction) it follows further that mS1(a,t <
to>= -1 and that the state vector of S , is It&!) for t < t o (notice that
this is already in conflict with the usual QM formalism, according to which
+
it is only after ms2(a, to) = 1 is measured by SGl(a), thus for t 2 to that the
state vector of S , is 1.)~':
(5) Making the same reasoning for each couple ( S , , S,), one arrives at
the contradictory conclusion that, the original ensemble was not described
by the pure state, Eq. (15), but by a mixture of states Iu::')lu:?) and
I u ! , ! ! ) l u : ~ ) with equal probability (see also Table I).

TABLE I
SG,(a) IN PLACE,
SG2(a)NOT IN PLACE:
EPR PARAWX

Time SI s2

t < to State vector of (Sl,S2):100)


msl(a,t < t o ) = ?
m,,(a, t < t o ) = -ms,(a, t < to)=?
f = to ms,(a,to)is measured: S, moves through No measurement on S,, but
SGI(a)(physical, local interaction) m,,(a, t o ) = -m,,(a, t o ) predicted
t L to Reduction of (00): if State vector becomes Iu:?!) by nonlocal
ms,(a,to)= + 1:100> reduces to influence
Iu~!)tu~?!)for (s,,s,)
Inferences:

1
m,,(a,t = to) = - I
(separability,locality)
m,,(a, t < to) = + 1 c- m,,(a,t < to) = - 1

I
state vector = 1)~':;
(conservation of
angular momentum) I
state vector = Iu;!!)

(Sl,S2jdescribed by mixture of
Iub:' )I ub? ) and 1 ub? )I ub:' )
= observably different from (00)
EPR Paradox
256 W. DE BAERE

In order to show the contradiction in a quantitative way, let us introduce


the triplet state
I$,) = 110) = (1/&)(lu:Y)lu:9 + lu:?)lu:Y)) (18)
Then one may write
lu:Y)lu:3 = (l/JZ)(lW + 110)) ( 194
lu:?)iu:Y) = (l/Jz)(lOo) - 110)) (19b)
such that the original ensemble of couples ( S , , S , ) should now be a
superposition of singlet and triplet spin states, instead of a pure singlet state.
In fact there are observable differences between the two kinds of descriptions
(Selleri and Tarozzi, 1981). In the pure state one has
( o o ) J ~ ~ o =o )0 (20)

Obviously there is something that has to be wrong in the above reasoning.


Noting that no contradiction arises if one does not go outside the strict QM
framework [Bohr's resolution of the EPR paradox (Bohr, 1935a,b)]; Selleri
and Tarozzi (1981) arrive at the conclusion that it is step (3) in the EPR
reasoning, namely the locality hypothesis (Hl), which is incompatible with
QM, and which is responsible for the above contradiction.

C . Critical Analysis of the Original EPR Argumentation

When the EPR paper appeared in 1935, it was already clear that the
quantum formalism was very powerful and successful in the treatment of
physical problems at the atomic level. As a consequence, it had become the
daily working tool of a large majority of physicists who did not bother about
fundamentals. Yet the original EPR argument was considered a serious
challenge of the completeness of QT, especially by those interested in the
soundness of its basis.
In this section we will review the main reactions immediately after the EPR
argumentation was published. We will see that most of the early reactions
emphasized the difference between the definition of state of a physical system
in EPR, which is that of classical mechanics (CM), and that adopted in QM.
In CM the state is supposed to be a representation of the system itself, while
in QM the state represents, at least in the Copenhagen interpretation (Stapp,
EINSTEIN-PODOLSKY-ROSEN PARADOX 257
1972),merely our knowledge about the system. Also attention is drawn to the
fact that measurement of noncompatible observables requires mutually
exclusive experimental arrangements. This may be seen as an indirect criticism
of EPRs implicit hypothesis of counterfactuality (H3).
Kemble (1935) was one of the first to defend the orthodox viewpoint
against EPR. He argues that, if it were true, as EPR contend, that “it is
possible to assign two different wave functions ... to the same reality ...,”
the QM description would be erroneous.
According to Kemble’s view, the EPR problem may be avoided by stick-
ing to “. . . the interpretation of quantum mechanics as a statistical mechanics
of assemblages of like systems.” Hence in the above EPR contention, one
of the wave functions describes the future statistics of one subassemblage,
while the other describes the future of another, different, subassemblage. And
which wave function is to describe the subassemblage depends on the choice
of what will be measured on one of two correlated systems in the original
assemblage. Kemble concludes that there is no reason to doubt the complete-
ness of QM on the grounds advanced by EPR. Once it is accepted that the
wave function does not describe the intrinsic state of a system itself, but rep-
resents only the regularities between future observations on a system, the
original EPR problem on the completeness of QM disappears.
According to Wolfe (1936), the origin of the EPR troubles lies in an
unjustified extrapolation of ideas on “physical reality,” which are certainly
satisfactory and reliable in classical physics, to the domain of QM. EPRs
own specific example of a system having both position and momentum
illustrates this point. In CM both may indeed be determined and known
simultaneously, and states in CM accordingly contain this information.
However, in QM only one of the two may be known, and this is again reflected
in the quantum state. According to Wolfe, the quantum state only represents
our knowledge about a system. Hence Wolfe concludes that “Viewed in this
light, the case discussed by Einstein, Podolsky, and Rosen simply dissolves.
Two systems have interacted and then separated. Nothing that we do to the
first system after this affects the “state” of the second. But ..ieasurements on
the first system affect our knowledge of the second and therefore affect the
wave function which describes that knowledge. Different measurements on the
first system give us different information about the second and therefore
different wave functions and different predictions as to the results of
measurement on the second system.”
The reply of Bohr (1935a,b) goes along similar lines. After explaining in
detail his general viewpoint, called “complementarity,” Bohr shows that the
QM formalism is mathematically coherent and that the limitations on the
measurement of canonically conjugate observables are already contained in
the formalism. According to Bohr, the origin of the EPR contradiction is
258 W. DE BAERE

“. .. an essential inadequacy of the customary viewpoint of natural philoso-


phy for a rational account of physical phenomena of the type with which
we are concerned in quantum mechanics.”
Bohr continues to explain in detail that the experimental arrangements
and procedures for measuring noncompatible observables, in particular x
and p , are essentially different and mutually exclusive. Choosing one arrange-
ment means the renunciation of knowledge of the variable measured by the
other. Otherwise stated: Measurement of one variable means an uncontrol-
lable reaction on the system which definitely prohibits the measurement of
the other variable. Hence Bohr concludes that only if two or more observ-
ables can be known or measured simultaneously,can they have simultaneous
“reality” status. This was also the point made by Ruark (1935). According to
Bohr then, it is therefore clear that the original EPR argumentation is invalid
for noncommuting observables A and B. Also EPR were apparently aware
of the above criticism because of their final remarks towards the end of their
paper (Section 11,B): “... one would not arrive at our conclusions if one
insisted that two or more physical quantities can be regarded as simulta-
neous elements of reality only when they can be simultaneously measured
or predicted.” They rejected this solution to their problem on the ground
that “This makes the reality of P and Q depend upon the process of mea-
surement carried out on the first system, which does not disturb the second
system in any way. No reasonable definition of reality could be expected to
permit this.” However, the last viewpoint is not correct because it must ad-
mitted that any measurement result is somehow characteristic for elements
of reality of both the instrument and of the physical system, and not only
of the system itself. Hence, according to Stapp (1971), it appears that EPR
violates “. . . Bohr’s dictum that the whole experimental arrangement must
be taken into account: The microscopic observed system must be viewed in
the context of the actual macroscopic situation to which it refers.” For a re-
cent discussion of the debate between EPR and Bohr, see de Muynck (1985).
An interesting discussion of the EPR criticism against QM has been given
by Furry (1936a,b). According to Furry (1936a) the essence of Bohr’s reply to
EPR is that “one must be careful not to suppose that a system is an
independent seat of “real” attributes simply because it has ceased to interact
dynamically with other systems.”However, one should be warned that such an
extreme positivistic standpoint may also be criticized. Indeed, it may well be
assumed that both the apparatus and the system itself have their own “real”
and independent (in the sense of not being influenced instantaneously by the
rest of the world) attributes of which any measurement result is but a repre-
sentation. It will be shown in Section V,B that such a viewpoint is perfectly
consistent with the Bell inequalities and the hypothesis of Einstein locality
(H l), provided the hypothesis of counterfactuality (H3) is abandoned.
EINSTEIN- PODOLSKY-ROSEN PARADOX 259

Furry then studies in a quantitative way the observable differences


between the above assumption of EPR that the QM wave function describes
the “real” attributes of the system itself and the orthodox QM standpoint that
applies the usual QM formalism and denies that physical systems possess
independent real properties.
Consider then two correlated systems S1,S 2 which are described by a QM
state vector l$(xl,x2)). According to von Neumann (1955; see also Baracca
et al., (1974), l $ ( x l , x2)) may in general be written

Moreover in Eq. (22) all &’ are different, and also all P k . The state vector
Eq. (15) or (16),corresponding to the Bohm version of the EPR paradox with
correlated spin-) systems in the singlet state, may be considered as an
illustration of Eq. (22), for which moreover the expansion on the right-hand
side is not unique. From the one-to-one correspondence between & and p k in
Eq. (22) it follows that if L = & is measured on S1,then R = Pk is predicted on
S2 and vice versa.
Furry then gives a definite form to the above-mentioned opposite
viewpoints of systems possessing real attributes and of standard QM in the
following way:
Method A assumes that, as a result of the interaction which causes the
correlation between S , and S 2 , S, made a transition to one of the states
I q A k ( x , ) and
) S2 made a transition to the corresponding state 15Pk(x2)).Also it
is assumed that the probability for transition to JCP,,(X,)> is wk. Note that the
assumptions of Method A are precisely the conclusions of the modified EPR
argumentation (Sections II,A and I1,B) according to which a mixture of states
1 q A k ( x l )1)t p k ( x 2 ) )should describe an ensemble of correlated systems (S,, S2),
instead of a pure state as in Eq. (8) or (22),which was the assumption from the
outset.
Method B, on the contrary, uses only standard QM calculations based on
the pure state, Eq. (22).
Furry considers then four types of questions which may be answered by
using either Method A or Method B. The particular type of question for which
the two methods give different answers is the following.Suppose observable M
is measured on S1with eigenvaluesp and eigenstates1 &(x1)) and observable S
is measured on S , with eigenvalues c and eigenstates 147,(x2)). The question
260 W. DE BAERE

is: If M = p on S , is measured, what is the probability that S = (r will be


measured on Sz?
According to Method A the probability is

and according to Method B

IT ~<lGp(xl)l~~k(xl)>(rl,(xz)l 5pk(XZ)>
(25)
1k
wkl <%k(x 1 )I$p(x I
1)>

It is seen that the difference between Eqs. (24) and (25) consists in the
appearance of interference terms in Eq. (25) which are not present in Eq. (24).
The origin of these terms is usually ascribed to the influence of the measuring
apparatus on the internal conditions within the system, whose future behavior
has therefore been influenced. We will see in Section III,F in a direct way that
this must happen with each system that passes some apparatus. Moreover
there does not exist a way by which Method A can be made consistent with
Method B. Hence, Furry claims to have shown mathematically the inconsis-
tency between standard QM and the EPR viewpoint that physical systems
may have objective properties which are independent of observation.
However, we will show below that this is not entirely correct. It is only correct
if Furry’s assumptions of his Method A are accepted.
The link with the original EPR argumentation is obtained by consideringa
particular example for which the expansion of Eq. (22) is not unique. Furry
then applies his Method A to each of the expansions, which amounts to
assuming that in each case the system S1 or Sz made a transition to a definite
state for which a certain observable has some well-defined value. As in the case
of EPR, contradiction with QM occurs when both observables are noncom-
patible. To illustrate these points clearly, Furry analyzes a thought experiment
of the EPR type and shows that the final conclusions are in conflict with
Heisenberg’s uncertainty relations.
Furry (1936b) concludes that “. . . there can be no doubt that quantum
mechanics requires us to regard the realistic attitude as in principle
inadequate.”
In defense of the realistic attitude of EPR we should remember, however,
what has been said at the start of the discussion of Furry’s contribution. Also it
may be questioned whether Furry’s Method A is identical to that which
Einstein had precisely in mind. Indeed, the eigenstates l q A k ( x l ) ) of the
observable L, which Furry ascribes to subensembles of systems S 1 , are QM
EINSTEIN-PODOLSKY -ROSEN PARADOX 26 1
states. This means that, for the subensemble described by l q A k ( x l ) (selected
) by
retaining only those S1 for which measurement of L is A&, each subsequent
measurement of L will again give the result 1,.However, if another observable
M is subsequently measured instead of L, then only the probabilities
~ ( $ , , ( x , ) ~ q A k ( x l for
) ) ~ obtaining
z the value p are known. Now the idea of
Einstein was probably that, once the state of S, is known completely, it should
be possible to predict with certainty all possible measurement results at one
instant. Perhaps EPR’s only weak point was to assume that all these possible
measurement results also correspond to actual results; i.e., they assumed
implicitly the validity of counterfactuality (H3). We will see in Section V that,
on account of the empirical violation of the BIs either Einstein locality (Hl) or
counterfactuality(H3) has to be abandoned. In the latter case this means that,
among all possible measurement results, there is only one that has physical
sense, namely the one that is actually obtained in a real measurement.
It is seen that the difficulty with considerations on states of individual
quantum systems is that there does not exist any formal scheme within which
such problems can be discussed. Tentatively this may be done along the
following lines.
Suppose we represent the individual state of each quantum system S at
each instant t by a set of (field) functions { d ( x , t ) } ( i = number of field
components).In the spirit of Bohr, we shall consider the individual state of the
apparatus, say a Stern-Gerlach device SG(a) with analyzer direction a, as
equally important for the determination of the measurement result, e.g., the
spin projection ms(a, t) along a at time t. This means that SG(a) too should be
represented by a similar set {tl,&a)(x, t)}. Measurement of the spin projection
ms(a, t) of S along a should then formally be represented by the mapping

In this formal scheme the “elements of reality” corresponding to the


particular value rn,(a, t) should therefore be represented by { a i ( x ,t ) } for S and
t ) } for SG(a); i.e., rn,(a, t ) should not be considered as an attribute
by {cL$~(,,)(x,
of S alone, but of S and SG(a) together. Hence, within such a scheme, a
reconciliation between the realistic standpoint of EPR and the orthodox or
Copenhagen standpoint should be possible (see also de Muynck, 1985).

D . First Attempt at an Experimental Discrimination


between EPR and Standard QM

We have seen above that the simplified EPR argumentation leads to the
conclusion that an ensemble of correlated systems S,, Sz should be described
by a mixture of product states, instead of a pure state as required by QM. The
262 W. DE BAERE

quantitative analysis of Furry (1936a)(Section I1,C)has shown very clearly the


difference between the predictions based on both kinds of description.
Subsequently, Bohm and Aharonov (1957) were the first to suggest that,
eventually, the original situation of correlated systems, described in standard
Q M by means of a pure state, could evolve spontaneously to a situation which
should be described by means of a mixture of product states of the individual
systems when the separation becomes sufficiently large. In the special case of
correlated photons it is assumed that their polarizations are opposite, and, in
order to retain the QM rotational symmetry, that the polarization directions
are uniformly distributed in an ensemble of coupled systems.
In order to discriminate between the two alternatives Bohm and Aharonov
discuss an experiment (Wu and Shaknov, 1950)in which correlated photons y1
and y2 originate from the annihilation of an electron and a positron, each of
which is supposed to be at rest. Under these conditions the state of the couple
(yl, y 2 ) has total spin and parity J p = 0-(Kasday, 1971)and according to QT
one can write either
= ( l / f i ) ( I R l ) l R Z ) - IL1)ILz)) (27)
or
l~ylyz) = (1/JZXlX,>lY2) - IYl)lX2)) (28)
for the pure state vector of the coupled photons. In Eq. (27) the notation is
such that IR1)(R2)(ILl)lL2)) represents photons moving in opposite direc-
tions (e.g., along z) which are right (left) circularly polarized. Similarly, in
Eq. (28), IXl)l Y,) represents two photons which are linearly polarized in,
respectively, the x and the y direction. Hence, according to standard QM,
measurement of one kind of polarization on y1 immediately gives information
on the corresponding polarization of y 2 , because of the correlation in Eq. (27)
or (28). Now, because such ideal polarization measurements on individual
annihilation photons do not yet exist, this kind of one-to-one correlation
between individual photons is not verifiable. Therefore, to distinguish between
the two alternatives, one has to proceed in an indirect way, namely by means
of the process of Compton scattering, which also depends on the photon
polarization.
Suppose then that y1 and y 2 , with momenta pyl and py2= -py,, are
Compton scattered at points A and B with final momenta PI,
and ptz (Fig. 1).
In Bohm and Aharonov (1957)two different situations are considered, namely
for the scattering planes (pyl,pi,) and (pyz,pi2)being perpendicular or parallel.
In each case coincidence counts of photons are registrated for which the
scattering angle 8 is the same for both photons. The ratio of the number of
coincidence counts in both cases is called R. It is this experimentally
determined ratio R, for an ideal angle of 8 = 82", which is compared with the
EINSTEIN-PODOLSKY -ROSEN PARADOX 263

FIG.1. Compton scattering of correlated annihilation photons.

predictions following from three assumptions:


(Al) Standard QT is universally correct, i.e., for large as well as for small
separations between y,l y2 ;
(A2) Standard QT applies only for small separations, e.g., when the wave
packets overlap. For those separations for which this is not the case, it is
assumed that the photons are already in QM states which are circularly but
oppositely polarized about their direction of motion;
(A3) The same as (A2), but now it is assumed that each photon y1 is
already in a state of linear polarization in some arbitrary direction, and the
other, y2, is in a state of perpendicular polarization. All directions are equally
probable.
The theoretical predictions as calculated by Bohm and Aharonov (1957)
for the different assumptions, taking into account the characteristics of the
experimentalconfiguration (e.g., finite experimental solid angle), are as follows

RA1 = 2.00, R,, = 1.00, RA3 = 1.5 (29)


These values are to be compared with the experimental result R = 2.04 f.0.08
of Wu and Shaknov (1950). From this it is evident that in this case standard
QM gives the correct result. More recent, similar experiments on annihilation
photons (Bertolini et al., 1955; Langhoff, 1960; Section IV,A) give further
evidence for the correctness of the QM description of correlated and widely
separated photons; hence the above hypothesis of Bohm and Aharonov Lie.,
assumption (A2) or (A3)] is not correct.
The relevance of the experiment of Wu and Shaknov with respect to the
EPR paradox has been criticized by Peres and Singer (1960). They argued that
photons, being massless, have their spin always along their direction of
propagation, and that the spin components orthogonal to this direction do not
have physical meaning because these quantities are not gauge invariant.
However, Bohm and Aharonov (1960) subsequently showed that these
arguments were invalid because they were based on an incorrect interpretation
of photon polarization in QT.
264 W. DE BAERE

E. Attempts to Resolve the EPR Paradox

Margenau (1936) apparently was the first who tried to resolve the EPR
paradoxical conclusions by emphasizing that the assumed validity of the QM
process of state-vector reduction may be the origin of the troubles. Following
EPR,Margenau remarks that, if this process is correct, then “. . . the state of
system 1, which, by hypothesis, is isolated from system 2, depends on the type
of measurement performed on system 2. This, if true, is a most awkward
physical situation, aside from any monstrous philosophical consequences it
may have.” However, note again that from this statement it is seen that
Margenau too accepts implicitly the validity of counterfactuality (H3).
Margenau then argues with respect to the EPR problem that “. . .if it be denied
that in general a measurement produces an eigenstate, their conclusion fails,
and the dilemma disappears.” However, we have already shown above that the
EPR contradictions are most striking when (H3) is not used at all. In this
respect, we will discuss below (Section II,F) some interesting recent proposals
for a direct experimental verification of the correctness of the process of state-
vector reduction in the case of correlated systems, or of the validity of locality
or the relativity principle.
According to Breitenberger (1965) there can be no question of a real
paradox, neither in the original EPR situation nor in the later Bohm version,
because the contradictory conclusions result from statements which are
unverifiable (i.e., counterfactual) and therefore are devoid of physical sense. It
is also noted that the EPR reasoning is based implicitly on the assumption of
the knowledge of a precise value of a conserved physical quantity, the
importance of which was first emphasized by Bohr (1935b).
Moldauer (1974)re-examines the EPR argumentation from the standpoint
that physical theories, in order to be verifiable, should deal with reproducible
phenomena. It is argued that under this requirement the EPR conclusion of
incompleteness of QT does not apply, although it is admitted that non-
reproducible events, such as the decay of a radioactive atom, are objective
observations which endow physical reality to such events. EPR claim that
these events too should be described in a complete theory. According to
Moldauer such inherently nonreproducible events (e.g., no ensemble of
radioactive atoms can be prepared which decay at precisely the same instant)
ought not to be the object of theoretical investigation.
Zweifel (1974) uses Wigner’s theory of measurement to resolve the EPR
paradox. This theory is based on the idea of the existence of some kind of
interaction between the system under investigation and the mind of the
observer. It is argued that S , and S , are interacting with each other, because
both interact with the mind of some observer via the Wigner potential. The
observed correlation between S , and S, is then viewed as a direct consequence
EINSTEIN- PODOLSKY -ROSEN PARADOX 265
of their interaction with the mind of the respective observers. It is argued that
both observers must exchange information in order to be sure that their
measurements on S , and S2 concern two systems which have a common origin,
and this exchange should be responsible for the correlation.
The physics of the EPR paradox has also recently been re-examined by
Kellett (1977). It is argued that the EPR argumentation against the
completeness of Q T is unsatisfactory on two grounds:
(1) The EPR argument is invalid as it stands because it is based on a
Gedankenexperiment which is physically unrealizable [however, see Bartell,
1980b, Section II,A and Section 11,F);
(2) The basic EPR assumptions are equivalent to assuming that the QM
description of physical systems is independent of any observation or
measurement. It is argued that the present experimental evidence for the
violation of the BI (Section IV) rules out these basic assumptions, hence make
invalid the EPR argumentation itself. Kellett admits, however, that the
concept of electron “in itself‘’ is not necessarily meaningless and that
individual electrons may well exist.
The difficulty is that up to now no theory for such individual systems exists
and that questions pertaining to such systems are not relevant to the state
vector II/ of QM. The completeness claim of Q M is considered as being the
claim that all information that is necessary to any future observation is already
contained in I). Hence, I) is a summary of all possible outcomes of future
measurements. Nevertheless, in spite of admitting the existence of individual
systems, according to Kellett “. . . it is clearly the EPR definition of physical
reality that is at fault.”
An approach to the EPR paradox that sounds rather strange consists in
assuming that influences are not only propagating into the future, but equally
well into the past (retroactivity). Adherents of such a view are, e.g., Costa de
Beauregard (1976, 1977, 1979), Rietdijk (1978, 1981, 1985), and Sutherland
(1983). However, it seems very difficult to grasp the physical significance, if
any, of the idea that events or processes that already happened in the past
could have been influenced by events that happen just now. The only
significance the present author may recognize in such a picture is within a
completely deterministic evolution of the world: What happens in the future
depends on what happened in the past, and in this way both the future and the
past are related. Yet most of us will prefer the picture that all events which will
happen later on are influenced by events that happened earlier, and not the
other way around. Of course, in such a scheme, knowledge of the present
would also provide knowledge of the past, however without making it
necessary that influences are propagated backwards in time toward the past.
Also Sutherland (1985) has recently shown that such a model is implausible,
266 W. DE BAERE

because a new paradox arises from the requirement that a nonlocal influence
of a measurement on S,, exerted on the measurement on S,, should be
independent of whether S, is inside or outside the forward light cone of S, .
Another approach to the EPR problem is that of Destouches (1980), in
terms of De Broglie's theory of the double solution. It is shown that within this
framework a more general and more satisfactory theory than QT may be set
up, satisfying the following set of conditions: (1) the results of QM are
recovered, (2) the Bell inequality is violated, (3) the EPR paradox disappears
and (4) there do not exist retroactive influences in the sense of Costa de
Beauregard (1976,1977,1979), of Rietdijk (1978,1981,1985), or of Sutherland
(1983).
Other attempts to resolve the EPR paradox are in terms of the density
matrix formalism of Jauch (1968) and of Cantrell and Scully (1978). In both
cases it is claimed that within this formalism a satisfactory answer to the EPR
problem may be given. However, Whitaker and Singh (1982) remark that in
this way one is implicitly using the ensemble interpretation of QT, inside
which no paradoxical conclusions appear (see also Ballentine, 1970,1972).It is
stated that the EPR paradox requires a resolution only within the Copenha-
gen interpretation (CI) (Stapp, 1972) of QM, because this maintains that it
gives a complete (probabilistic) description of individual systems. The EPR
paradox has been discussed in different QM interpretations in Whitaker and
Singh (1982).
Muckenheim (1982) presents the following solution to the Bohm version
of the EPR problem. It is based on the allowance of negative probability
distributions. It is assumed, in the EPR spirit, that each of the correlated
spin4 systems S,, S, has a definite spin direction on its own, e.g., s and -s.
Hence for each system it is accepted that spin components along different
directions have simultaneous reality, even if these correspond to noncom-
patible observables, although it is admitted that these cannot be observed
simultaneously. Call w,(a,s) the probability for obtaining the result s* = ki
along a, with w,(a,s) + w-(a,s) = 1. It is assumed further that w+(a,s)-
- -
w_(a,s) = (const)a s, such that w,(a,s) = 3 k a s. With s2 = 2, it follows
-
from a s = ( a / 2 ) c o s 8 that the probabilities w+ may become negative. This
model allows Miickenheim to reproduce all QM results. For a single system it

s.
is found that

(s(a,s)) = p(s)[w+(a,s)s+ + w-(a,s)s-]dR =0 (30)

-
with p(s) = (47c-',a s = &/2)cos0, dC2 = sinOd8dz and s+ = -l while
+,
for correlated systems the result for the correlation function is
P(a,b) = (s(a,s)s(b,s)) = -$a -b (304
EINSTEIN-PODOLSKY -ROSEN PARADOX 267
in accordance with standard QM. However, no explanation is given of the
physical significance of the negative probabilities.
In an interesting paper de Muynck (1985) investigates the relation between
the EPR paradox and the problem of nonlocality in QM. In this paper it is
discussed, along the lines of Fine (1982a,b) (Section V,C) whether the EPR
pr6blem and the violation of the BIs by recent experiments has anything to do
with a fundamental nonlocality or inseparability at the quantum level. In this
respect it is interesting that de Muynck makes a clear distinction between
unobserved, objective EPR reality and observed reality that is described by
QT. Because of the absence of any experimentally verifiable consequence on
the basis of this EPR reality, the EPR analysis itself is incomplete and on a
metaphysical level. Remember also that similar points were stressed when
pointing to the necessity of introducing counterfactuality as well in the EPR
reasoning (Section II,A,B) as in the Bell reasoning (Section 111,C).
In fact it cannot be denied that on the quantum level the result of what
would have happened if, instead of an actual experiment, one had carried out
another one, with other macroscopic apparatus parameters, is completely
unverifiable. de Muynck repeatedly stresses not to forget the central lesson of
QM (and of Bohr) that it is impossible to gain knowledge without taking into
account the measuring arrangement. However, by remarking that the
existence of an objective reality is not excluded at all by QM, de Muynck tries
to show that “. .. the positions of Bohr and Einstein with respect to reality are
less irreconcilable than is often taken for granted.” Hence, in this article no
extreme standpoints are taken and it is a virtue that a reconciliation between
EPR and Bohr is strived for. As de Muynck remarks: “. . . Bohr’s completeness
claim,. . . restricts the possibility of defining properties of a system . . .,”and:
“Einstein, indeed, may be right in pointing at the possibility that we might
obtain knowledge about a system exceeding the quantum mechanical
knowledge.” Hence, “. .. the Einstein-Bohr controversy is not a matter of
principle. It is just about the domain of application of quantum mechanics.”
Other proposals for a resolution of the EPR paradox and of the problem
of locality in relation with the Bell inequalities will be discussed in Section
II1,G. A generalization of the EPR paradox by Selleri (1982) and the resulting
new Bell-type inequalities will be mentioned in Section III,D,6.

F. Recent Proposals for Testing the Validity


of Einstein Locality

In Section II,D we have seen that the agreement of the experimental results
of Wu and Shaknov (1950) with Q M (and also with the results of more recent
similar experiments) was used by Bohm and Aharonov (1957) to rule out the
268 W. DE BAERE

hypothesis that a pure state vector evolves spontaneously into a mixture of


product states (known as the Bohm-Aharonov hypothesis).Such an evolution
is indeed expected if Einstein locality (Hl) is supposed to be valid. Hence the
conclusion that (Hl) must be wrong.
In Section IV,A we will see that further evidence supporting this
conclusion comes from atomic cascade experiments suggested by the Bell
inequality, such as the very convincing recent experiment of Aspect et a!.
(1982). However, a drawback of this kind of experiment is that the validity of
the hypothesis of counterfactuality (H3) has to be assumed in order to justify
the combination of results obtained in at least two different experiments. If
one neverthelesscontinues to stick to nonlocality,then it appears that the EPR
paradox is avoided by introducing another paradox, namely the violation of
the relativity principle.
Now, eliminating paradoxes by introducing new ones is certainly not a
recommended method to solve physical problems. Moreover Fine (1982a,b)
and others have recently shown (Section V,B,l) that locality is irrelevant for
the derivation of the BI. Therefore, there is some ground to suspect that the
resolution of the EPR paradox, which consists in the rejection of the existence
of separated entities and a finite speed of propagation of influences,is not the
correct one. Bearing in mind that the process of state-vector reduction plays a
crucial role in the EPR paradox, we will discuss below two recent proposals of
Gedankenexperiments for verifying (non)locality.
Although it was not the purpose of EPR to criticize the QM formalism, it
will appear that the EPR reasoning may be used to criticize the CI of Q M
(knowledge of the result of an experiment is sufficient to determine the
subsequent state vector)as it is applied to the case of two correlated systems S ,
and S,. Remember that Margenau’s resolution of EPR (Margenau, 1936)
consisted precisely in the rejection of state-vector reduction. In fact in the
following we shall assume unlimited validity of the relativity principle (i.e., no
action at a distance) and ask what may be wrong in the use of the CI of the
quantum formalism to two correlated systems S , and S , . We will try to make
plausible that it is indeed the reduction of the state vector for the system
( S , , S,) which may be subjected to criticism.

I . Poppers’s New E P R Experimenl


As a first example in which state-vector reduction in the case of correlated
systems may be seriously in trouble, we mention Popper’s recent EPR
experiment.
Recently Popper (1985) proposed a new version of the classic EPR
Gedankenexperiment. As in the original EPR paper, he is mainly concerned
with showing that physical systems may have both position and momentum at
EINSTEIN-PODOLSKY -ROSEN PARADOX 269

BL
I
*L
TY I
*R

FIG.2. Setup for Popper's new EPR experiment.

the same time, and hence follow a (classical) trajectory. However, we will use
this proposal to criticize state-vector reduction in the case when only one
measuring device is in place.
Popper's new proposal is essentially as follows (Fig. 2): 0 is the origin of
correlated systems ( S , , S,) which move in opposite directions: S, moves to the
right towards a screen A, which has a slit of extension Ay, while S, moves to
the left. Because of their common origin, S, and S, may be represented by a

s
state vector which reflects their correlation

I$S1S$O)) = 14Y(1))l -d,YfZ) = - Y 9


x (d,ycl'; -d,y(,) = - ~ ' " ~ $ ~ , s , ( t o ) )dy'" (3 1)
It follows that if at t = t o S , is observed at (d,y(')), then at this moment the
state of Eq. (31) reduces to Id,y('))l -d,y'" = -y(')). This means that at
t = to the position of S , will be (- d , - y'") if this were actually measured. In
particular, those systems S , passing through the slit Ay will be diffracted (to be
observed at 4)as a result of the Heisenberg relation Apy x h/Ay; with the
knowledge of the position of S, with an accuracy Ay corresponds an
uncertainty Apy of its corresponding momentum component. Now, because of
the correlation, the same must happen with S,; indeed, the resulting state-
vector reduction provides us with the knowledge of the position of s, with an
accuracy of the same order Ay. Hence there must be, according to standard
QM, an uncertainty Apy of the momentum, which means that S, will be
diffracted too (to be observed at screen B,), even when the screen A, with a slit
Ay is not in place.
If this were actually the case, this would be a direct confirmation of the
validity of state-vector reduction in the case of correlated and widely
separated physical systems, and as a consequence also a strong argument for
the existence of action at a distance. Thus the physical interaction of S, with
the slit in A, would be responsible for both the diffraction of S , and, by a
nonlocal influence, for a diffraction of S, at the imagined screen A,. On
account of separability and locality, however, diffraction of S, is not expected
270 W. DE BAERE

to occur (at least as long as A, is not in place);to all observations of diffraction


on 4 should correspond observations on B, in a small region around the x
axis.
In a quantitative analysis of Popper's proposal, taking into account the
initial (inevitable) uncertainties at the source, Bedford and Selleri (1985) have
shown that the above experiment is possible in principle under certain
conditions. With annihilation photons, for example, the uncertainty at the
source is such that an observed effect on S , cannot be related to the observed
diffraction of S , . In the case of photon emissions in opposite directions in the
center of mass, however, with sufficiently large mass of the emitting source, a
relation between both diffractions should be possible. Indeed, these authors
have shown that the angular deviation between the momenta of S, and S2 in
the laboratory system is inversely proportional to the mass of the source,
hence in principle may be made arbitrarily small.

2. Double-Slit Experiment with Correlated Systems


As a second example of a situation in which troubles with state-vector
reduction in the CI of QM may be expected, we discuss a double-slit
experiment connected with an EPR-type situation (Fig. 3) (De Baere, 1985).
0 is the source of two correlated systems S,, S,. S, moves towards a screen
A , which contains two slits s1 and s2 . S, moves toward a screen A , where its
position can be observed. To the right of A, we have a third screen 4 where S,
may be observed, in the case S , has passed A,. Suppose further that all
relevant parameters (distances OAR,OA,, slsz, momentum of S,) are such
that on 4 an interference pattern is observed in the case that 0 is the source of
systems S, only. The question we are interested in then is the following: If S, is
observed at the screen A, at about the moment that S , reaches A,, will the
subsequent interference pattern at 4 subsist or will it disappear (as demanded

"i

FIG.3. Double-slit arrangement with correlated systems.


EINSTEIN-PODOLSKY -ROSEN PARADOX 27 1
by the CI)? Again, the argumentation of Bedford and Selleri may be used here
to conclude that, with each impact of S , on A,, there must correspond a
unique position of impact of S , on A,.
To answer the above question, let us look at the QM description of this
situation. The correlated couple ( S , , S,) is again described by the state vector
of Eq. (31). Suppose we observe the impact of S, at A, at t = t o . At this
moment the state vector I$s,s,(to)) reduces to Id, y(l) = -y(”)1 -d, y‘”). If it
happens that y(,) = y:) = -s then we may say that observation of S , on A,
has indirectly determined the slit s, through which S, will move. More
precisely, if a counter were placed behind the slit sl,S1 would have been
observed by it. In any case (counter behind s, in place or not), by the
observation of S, on A, the CI assures that at t = to the state vector for S, is
Id,y(’) = s). Having determined in this way the slit s, through which S , will
pass, standard QM predicts that the corresponding part of the incident S,
beam will give rise to a diffraction pattern at &. The same can be said for
observations of S, at A, for which y(’) = yji’ = +s. For those cases, state-
vector reduction steers the corresponding s, systems in the state vector
Id, y“’ = - s) (which corresponds in the usual terminology to those systems S ,
which “pass” slit s,), which again will give rise to a diffraction pattern at 4 .
For all other observations of S , at A,, the original state vector will reduce to a
state Id, y ( ’ ) # fs), which will correspond to absorption (or at least an impact)
of the corresponding systems S , on the left side of screen A,. In any case, these
systems will not pass either through slit s, or s,, and hence cannot contribute
to the final pattern observed on &.All this can be visualized by imagining that
photographic plates are mounted on the appropriate sides of screens A,, A,,
and &.
Therefore, one may expect that, under the validity of the QM process of
state-vector reduction (as used by EPR and in the CI), the pattern observed at
4 will be a superposition of two diffraction patterns, each corresponding to
observations of systems S, at positions s; and s; . On the contrary, under the
conditions of Einstein locality and separability, a measurement on S, will not
disturb or influence the conditions within S, at the moment it interacts with
the screen A,. It follows that the interference pattern at 4 will subsist in this
case, independently of what is done with S,. If this turns out to be the case
experimentally, then this would be evidence for our conjecture that in the case
of correlated systems S , , S,, observation on S , does not steer S, in some QM
state, unless an appropriate measurement apparatus is present. At the same
time this would constitute a resolution of the EPR paradox (see also Table 11).
However, we want to remark here that for single systems the so-called no-
result Gedanken experiment of Renninger (1960) may be used to gain
information about the system (and its wave function) without subjectingit to a
measurement apparatus.
272 W. DE BAERE

TABLE I1
SG,(a) I N PLACE,SG,(a) IN PLACE:EPR PARADOX

Time S, S,

t < to State vector of (Sl,S2):100)


m,,(a, t < t o ) = ?
m,,(a, t < to) = -m,,(a, t < t o ) = ?
t = to m,,(a, t o ) is measured: S, moves through ms,(a, t o ) is measured: S, moves through
SG,(a) (physical, local interaction) SG,(a) (physical, local interaction)
t 2 to Reduction of 100): if
m,,(a, to) = + 1: 100): reduces to 1)~’: m,,(a, t o ) = - 1: 100) reduces to Iub!!)
for S , for S ,
-+ State vector for ( S , , s , ) : lu!,:’)Iu~?!)

Inferences
ms,(a,t 2 to) = 1 + ms,(a, t 2 t o ) = - 1
No inference about S, for t < to No inference about S, for t < t o
3 No EPR Paradox

Considering the double-slit experiment with a source of single quantum


systems, then according to Feynman et al. (1965) this is “. . . a phenomenon
which is impossible, absolutely impossible, to explain in any classical way, and
which has in it the heart of quantum mechanics. In reality, it contains the only
mystery. We cannot make the mystery go away by explaining how it works.”
Perhaps the proposal made in this section with a source of correlated quantum
systems may help to get some insight in this mystery by allowing for the
possibility, via coincidence measurements on A, and 4 , to explain or to
follow how the pattern on 4 is built up.
Let us note in this respect that recently Wootters and Zurek (1979) studied
in a quantitative way, in the context of information theory, Bohr’s comple-
mentarity concept and Einstein’s version of the double-slit experiment (with a
source of single photons), in which both the path and the interference pattern
of the photons are attempted to be observed, it was found that the more
information one has about the path, the less one has about the interference
pattern. Subsequently Bartell (1980a) made a proposal of simpler realizable
systems than that of Wootters and Zurek for observing intermediate wave-
particle behavior. It is based on Wheeler’s reformulation of the Einstein
version of the double-slit experiment (Wheeler, 1978).

G . Conclusion

We have seen that the EPR argumentation [using (Hl), (H2) and (H3)],
which resulted in the claim of incompleteness of QM, has been criticized
EINSTEIN-PODOLSKY -ROSEN PARADOX ’ 273
rather convincingly, mainly on the grounds of the impossibility in principle of
knowing simultaneously noncompatible observables. We have made plausible
that this amounts to the invalidity of the hypothesis of counterfactuality (H3),
i.e., that one is not allowed to combine in an argumentation actual
measurement results and hypothetical ones (supposed to be carried out on the
same systems in exactly the same individual states). A disagreement on the
observational level exists between QM and what we called the simplified
argumentation [using only (Hl) and (H2)]. The discussion above has shown
that the problem of QM nonlocality and nonseparability deserves further
critical analysis, both on the theoretical and the experimental level. According
to a new approach to the EPR paradox (Section II,F), it may be interesting to
question the validity of the CI of the QM formalism, especially in its
application to correlated but widely separated physical systems.

111. THEBELLINEQUALITIES

A . Introduction

The status of the EPR paradox, after the early criticism, remained
unchanged for about 25 years. Several reasons may be given for this. First of
all, EPR were not able to elevate their objections to QM from the purely
epistemological level to the empirical level. This was done for them by Bell
(1964). Moreover, QM predicted with great success and accuracy all results in
atomic and molecular physics, and for every proposed new and realizable
experiment a definite and unambiguous answer could be given. Hence, there
was no practical need at all to supplement this successful formalism by means
of hypothetical or hidden variables (HVs). Also, attempts to derive QM from
classical ideas (Fenyes, 1952; Weizel, 1953a,b) were in general not very
convincing because they did not lead to clear, verifiable differences with QM.
As a result of all this, opinion had settled that Bohr’s (1935a,b) reply to EPR
was convincing and final.
Another important reason for this was the existence of the von Neumann
theorem. Indeed, already in 1932 von Neumann (1955)had proved mathemat-
ically, starting from a number of plausible assumptions, that no deterministic
theory for the individual system, based on the idea of dispersion-freevariables,
is able to reproduce all statistical predictions of QT.
However, since Bohm (1952b,c) proposed a concrete counterexample of
von Neumann’s theorem, and Bell (1966) and Bohm and Bub (1966b) showed
that one of the underlying assumptions, namely the linearity hypothesis, was
not necessarily valid in all HV theories, the interest in the fundamentals of QM
274 W. DE BAERE

was again awakened. Also, at about the same time, Bell (1964)made his famous
analysis of the EPR paradox which consisted in translating the EPR ideas in
precise mathematical form and deriving a simple inequality, the Bell
inequality. The importance of this BI stems from the fact that it made very
clear the conflict between QM and the EPR point of view, and that Clauser et
al. (1969)showed that by generalizing the BI the whole issue could be brought
to the experimental level.

B. Von Neumann’s Theorem

The essentials of von Neumann’s proof of the impossibility of dispersion-


free states are as follows (von Neumann, 1955; Albertson, 1961; Bell, 1966;
Bohm and Bub, 1966a; Capasso et al., 1970; Jammer, 1974, pp. 265-278). The
starting points are a number of hypotheses:
(vN1) To each observable R there corresponds in a one-to-one way a
Hermitian operator R in Hilbert space. It is assumed that R is hypermaximal,
i.e., that its eigenvalue problem is solvable;
(vN2) With the observable f ( R )there corresponds the operator f ( R ) ;
(vN3) If to the observables R, S , . . . (not necessarily simultaneously
measurable) there correspond Hermitian operators R, S, . . ., then to the
observable R + S + ... there corresponds the operator R + S + ” ’ ;
(vN4) Hypothesis of linearity: If R, S , . . . are arbitrary observables and a,
b, .. . real numbers, then the following relation holds between expectation
values
(uR + bS + ...) = u ( R ) + b ( S ) + ... (32)
with
(R) = CWn($nIRI$n) (33)
n

if the ensemble is described by the pure states I$,) with probabilities w,.
The dispersion A R of an observable R in an ensemble may be defined by
(AR)’ = ( ( R - ( R ) ) ’ ) = ( R 2 ) - (R)’ (34)
A dispersion-free ensemble is then defined by the condition AR = 0 for all R,
i.e.
< R 2 ) = <R>’, VR. (35)
A further definition introduced by von Neumann was that of the homog-
eneous or pure ensemble. Any partition of such an ensemble results in a series
of subensembles, which, by definition, have the same statistical properties as
the original one.
EINSTEIN-PODOLSKY -ROSEN PARADOX 275
On the basis of the above postulates and definitions, von Neumann was
able to prove his theorem on the impossibility of reconstructing QM from any
kind of theory which starts from dispersion-free ensembles, usually identified
with a HVT. In the words of von Neumann “we need not go any further into
the mechanism of the ‘hidden parameters,’ since we now know that the
established results of quantum mechanics can never be re-derived with their
help” (von Neumann, 1955, p. 324).
The proof proceeds by showing that the assumption of dispersion-free
ensembles leads to inconsistencies. In the particular case of spin 3,this proof
runs as follows. Consider a dispersion-free ensemble of spin-3 systems
corresponding to a definite HV 1.For such an ensemble, each observable has a
well-defined value when measured. Consider in particular the observables
R = ox, S = oy,and T = R + S = ox + ay,with a,, oy,and a, the Pauli
matrices. Then the results of measuring R, S, and T on each element of the
ensemble may be represented by r(A), s(A), t(A). Hence
(R) = r(4, ( S ) = s(4, ( T ) = t(4 (36)
and, according to (vN4), one should have
(R +S) =(T)=(R) +(S)
or t(2) = r ( 4 + s(1) (37)
Now, because of the fact that eigenvalues of R, S, and T are, respectively, 1,
f 1, and Eq. (37) cannot be true, from which follows von Neumann’s
impossibility theorem.
Challenged by these assertions, Bohm constructed in 1952 (Bohm,
1952a,b) a concrete counterexample of von Neumann’s theorem. In Bohm’s
model, which is a reinterpretation of QM, the notion of quantum potential is
introduced, and position and momentum are treated as hidden variables.
Subsequently, Bell (1966) and Bohm and Bub (1966a) criticized the general
validity of von Neumann’s hypotheses, in particular the hypothesis of
linearity (vN4). Bohm and Bub (1966a) came to the conclusion that only a
limited class of HV theories was excluded by von Neumann’s theorem [in fact,
only those satisfying the von Neumann assumptions (vNl)-(vN4)]. Moreover,
Bell (1966)explicitly constructed a physically reasonable HV model for spin-4
systems, which did not satisfy (vN4). Bell showed (Selleri, 1983b, pp. 49-52)
that, although (vN4) is correct for standard QM states, it is unjustified to
require it to be valid for all conceivable HV models.
Other impossibility proofs for HV theories were set up by Gleason (1957),
by Jauch and Piron (1963), and by Kochen and Specker (1967). All these are
apparently more general than von Neumann’s proof and are not based upon
the criticized hypothesis (vN4).The proof of Jauch and Piron has in turn beer,
276 W. DE BAERE

refuted by Bohm and Bub (1966b). The general problem with proofs of this
kind is not their mathematical correctness but the physical relevance of their
basic assumptions. For this reason it may be stated that no such proof will ever
be able to rule out a priori all possible HV models. For more details we refer to
the already mentioned papers, Clauser and Shimony (1978)and Jammer (1974,
pp. 296-302).

C . The Original Bell Inequality

1 . Derivation and Discussion


In 1964 Bell re-examined the Bohm version of the EPR argumentation.
Bell investigated the properties of any local deterministic HV theory which
satisfies the requirements of EPRs realistic conception of reality. The main
result is the famous Bell inequality, which has to be satisfied by any such
theory. The fact that the BI is violated by QM constitutes Bell's theorem: No
local deterministic HVT can reproduce all results of QT.
To prove this, a couple of correlated spin-+systemsS,($), &(+) in the singlet
state is considered, and it is assumed that the internal state of each system may
be represented, because of their correlation, by some common set of time-
dependent parameters (HVs),collectively denoted by A(t). Suppose that s, and
S2 move in opposite directions toward Stern-Gerlach devices SG,(a) and
SG2(b)which measure the spin component m,, along unit vector a and of mS2
along unit vector b.
Bell then remarks that, under the validity of Einstein locality (separability)
(H l), the value of m,, is independent of band the value of ms2is independent of
a. Furthermore, these values are assumed to depend on the HVs A(t),
representing the individual state of S, and S 2 . This suggests the existence of
functions msl(a,A(t)), ms2@,A(t)) which completely determine the spin pro-
jections (whether measured or not). Hence the basic assumption is that at any
time t , S , and S, have correlated properties which are definite and locally
determined (in the above sense),in agreement with the EPR idea of elements of
physical reality. To simplify notation we shall write
A(a,A) = m,,(a, A) = k 1, B(b,A) = m,,@,A) = k1 (38)
and adopt the convention that the spin projections are k 1 instead of i-3 (in
units of h).
The central quantity in Bell's reasoning is the correlation function

A(a, A)B(b,A ) p ( l )dA (39)


EINSTEIN-PODOLSKY-ROSEN PARADOX 277
in which A represents the whole HV space, and p(A) a normalized distribution
function of the HVs il

(40)

Now, the QM prediction for the correlation function P(a, b) is (Peres, 1978)
. -
P(a, b) = (OOJa, aa, b100) = -a -b (41)
. .
In Eq. (41), 100) is the singlet state vector of Eq. (15) and a, a,@, b are
the operators for the spin projections along a and b. In particular, one has
P(b,b) = - 1. Hence, if the HV expression in Eq. (39) reproduces this result,
then one should have that B(b,A) = - A @ , A) and Eq. (39) becomes

P(a,b) = -
I A(u,il)A(b,A)p(A)dA

Bell then goes on to reason along the lines of EPR and considers the
(42)

hypothetical situation that instead of analyzer direction b one had chosen


another direction c. Now, it is seen immediately that at this point,just as in the
EPR argumentation, the validity of counterfactuality (H3) is needed to allow
such reasoning. As remarked while discussing the EPR argumentation, QM
forbids, however, in principle such a reasoning because the spin projections
B(b,A) and B(c,A) are incompatible quantities, and hence cannot be known
simultaneously because of the practical impossibility of measuring them
together in one single measurement. We believe that this is a central lesson
from the successful QT, that one necessarily has to incorporate in any future
theory.
Simultaneous measurements of incompatible observables was studied
long ago, e.g., by She and Heffner (1966), Park and Margenau (1968), and
recently by de Muynck et al. (1979) and by Busch (1985).In de Muynck et al. it
has been shown that a simultaneous measurement of such observables causes
an unavoidable mutual disturbance of both results; i.e., the value of B(b,A) will
depend on whether it has been measured alone or together with B(c,A) (after
having given an operational meaning to such combined measurements).
Otherwise stated, the joint probability distributions will not reproduce the
marginal distributions. We will see in Section V,B that this is precisely a point
in the derivation of the BI which is severely criticized.
Yet, to make physical sense of the above counterfactual hypot!iesis
(Lochak, 1976; De Baere, 1984a,b), it may be assumed that in similar
correlation experiments (which may be carried out either simultaneously at
different places, or subsequently at the same place) the HVs are distributed
according to the same HV distribution p(A). Under this assumption one may
278 W. DE BAERE

write then, according to Eq. (42)

h
P(a, c ) = - A(a, A)A(c,A)&)

Subtracting Eq. (43) from Eq. (42) one obtains


dA (43)

P(a, b) - P(a, c) = - lA A)A(c,4 l P ( 4


CA(a,4A(b,4- 4,

= --JA &a, W ( b ,4L-1 - A@, AMc, AllP(A)dJ. (44)

Because A(a,A) = & 1, B(b,A) = f 1, it follows that

or
IP(a, b) - P(a, c)l I
I [1 - A(b, A)A(c,A)]p(A)dA (45)

IP(a,b) - P(a,c)l I 1 + P(b,c) (46)


This is the original form of Bell's inequality, which expresses in very simple
mathematical terms the consequences of Einstein locality (H 1) and counter-
factuality (H3). Equation (46) is the inequality that each local, deterministic
HVT in which counterfactuality is valid has to satisfy.
It is easily seen that Eq. (46)is violated by the QM prediction in Eq. (41).To
this end it suffices to take the following configuration of spin analyzer
directions (Fig. 4)
O,, = O,, = O,, f 2 = 60" (47)
for which
P(a,b) = P(b,c) = -P(a,c) = -3 (48)
Inserting Eq. (48) into Eq. (46), one gets the contradiction 1 5 3 . From this
result follows Bell's theorem: No local deterministic HVT is able to reproduce

FIG.4. Configurationof orientations for which the BI is violated by QM.


EINSTEIN-PODOLSKY -ROSEN PARADOX 219
all predictions of QT. We will see in Section IV that most correlation
experiments verify QM, and hence violate the BI, from which it is concluded
almost generally that the locality hypothesis (H 1) is invalid. In other words, it
is believed that any acceptable HVT which does not violate QT a priori should
be nonlocal. However, we have seen that Bell implicitly assumes the validity of
counterfactuality (H3) as self-evident. From the above conclusion of nonlo-
cality, it may be argued then that perhaps the importance of counterfactual-
ity (H3) is being overlooked. It may be stated, therefore, that locality may
be saved if counterfactuality is abandoned.

2. Wigner‘s Version of Bell’s Theorem


Wigner (1970) considers two correlated spin-$ systems S,($), S,($) in the
singlet state and assumes that the HVs A determine the spin components in any
number of directions. Wigner’s argument uses only three directions a, b, and c.
For a given 1, and under the condition of locality, the HVT gives the
predictions A(a,A) = a, = & 1, A(b,A) = b, = k 1, and A(c, A) = c, = k 1 for
the spin components of S , and predictions B(a,A) = a, = & 1, B(b,A) = b, =
L- 1, and B(c,A) = c, = & 1 for S,.
Let us denote then by p(a,, bl,cl;az,b,,c,) the probability for having the
configuration (al,b,, c,; a,, b,, c,) of spin components for S , and S, . In the
singlet state one must have that a, = -a,, b, = -b,, c, = -cl. Again QM
forbids from first principles that the above characterization be verified
experimentally. Indeed, Heisenberg’s uncertainty principle forbids the simul-
taneous measurement, hence knowledge, of spin components such as a,, b,, c,
for S, or of a2,b2,czfor S,.
Refusing to accept this basic truth will again necessarily lead to contradic-
tory conclusions, as we will see. Also, and for the same reason, only one couple
of values such as, e.g., a,, a, or b,,c, will represent values that correspond to
actual results, while all the other combinations such as b , , a , , etc. are then
hypothetical. Hence, as in all other derivations of Bell-type inequalities, the
validity of counterfactuality is already built in implicitly.
To see that the above HV scheme contradicts QM, let us derive
probabilities from p(a,, b,, c,; a,, b,, c,) which are experimentally verifiable
and, hence, for which QM makes specific predictions. Consider then the
+ +
probability p ( a , + ,b 2 + )for the results a, = 1, b, = 1
p ( a , + , b , + )= ~ ~ ( ~ , + , ~ , ~ ~ =
~ -l C J~ ~ z - ~ ~ z + ~ ~
c1
and thus one should have from Eq. (50)

which is not satisfied if c bisects oab.


Wigner’s argumentation has been criticized by Bub (1973), who shows that
the assumption of a probability distribution for the values of incompatible
observables leads to inconsistencies in the case of single spin-4 systems.
Freedman and Wigner (1973) reply that Bub’s criticism is based on a
misinterpretation of Bell’s locality postulate.

D. Generalized Bell Inequalitiesfor Dichotomic Variables

It has been remarked that a weak point in the derivation of the original BI
is the requirement that the condition P(a,a) = - 1 has to be satisfied exactly.
Now, no experimental setup is perfect; so that this condition will never be
realized in practice. The solution to this problem led to the first derivation of
the generalized Bell inequality (GBI) by Clauser et al. (1969).
It has been argued, furthermore, that neither the restriction to a
deterministic scheme nor the use of the HV concept is necessary for the
derivation of generalized forms of the BI. It appears that even the concept of
locality may be defined in different ways with respect to Bell’s inequalities
(Eberhard, 1978; Rastall, 1981; Bastide, 1984; Stapp, 1985). As a result, a large
number of generalized Bell inequalities (GBI) has been constructed since the
appearance of the original BI, the most well known of which we will review
below.
EINSTEIN-PODOLSKY -ROSEN PARADOX 28 1
I . Generalization by Clauser et al.
Clauser et al. (1969) were the first to derive a GBI and to show how it may
be transformed so as to allow direct experimental verification. Consider again
the correlation function P(a, b) defined by Eq. (39). With the notation and
conventions of Section II1,C one may write the following inequality for
IP(a, b) - P(a, b’)l

IP(a,b) - P(a, b’)J I


I IA(u, l)B(b,A) - A(a, l)B(b‘,l ) l p ( l )d l

= I I A k , WW,4lC1 - B(b,W ( b ’ ,m(4dl

= C1 - B(b, WW,
4 l p ( 4 dl

= 1- B(b,l)B(b’,l ) p ( A )d l (53)

If the criticized condition A(b,A) = - B ( b , l ) were valid, Eq. (53) would


lead immediately to the original BI, Eq. (46). However, suppose that we now
have
P(u’,b) = 1 - 6, 0I 6 I 1 (54)
for some directions a‘ and b, and such that 6 differs from zero for a’ = -b.
Writing A = A+ u A - , with

A+ = (A1 A(a’,A) = &B(b,A)) (55)


it may be shown that

With this result the right-hand side of Eq. (53) may be written

B(b,WV’,M 4 d l

= P(a’, b’) - 2
J1,- A@’, /Z)B(b’,&(A) dl

IA(a’,A)B(b’,I)(p(A)dA

= P(u‘,b’) - 6 = P(u’,b‘) + P(u’,b) - 1 (57)


‘i=
282 W. DE BAERE
+

-a‘
450 2

FIG.5. Analyzer directions for which the GBI is violated by QM.

From Eqs. (53)and (57) it follows that


IP(u, b) - P(u, b’)l + P ( d , b) + P(u’,b’) I2 (58)
This is the GBI, also called the Clauser-Horne-Shimony-Holt or CHSH
inequality, which again is violated by QM; e.g., for %,b = %,,./3 = Oarb= %arb, =
45” (Fig. 5) one has the contradiction 2& I2.
The GBI of Eq. (58) may be transformed to an inequality which contains
only directly available experimental quantities for correlated optical photons
yl, yz, originating from an atomic cascade. In this case the Stern-Gerlach
apparatuses SG,(a), SGz(b) are replaced by linear polarization filters Pl(a),
Pz(b). Normally A(a) = + 1 would correspond to the detection of y , which
passed P,(a), and A(a) = - 1 would denote nondetection. However, because
of problems connected with small photoelectric efficiencies for optical pho-
tons, another convention must be introduced: A(a) = - 1 means emergence
from P,(a) and A(a) = - 1 nonemergence. With this convention, P(a, b) is an
emergence correlation function, and in order to relate it to experimental data
one has to introduce the supplementary assumption that the probability for
coincidence detection of y,, y, is independent of a, b.
If, furthermore, the following experimental counting rates are introduced:
R(a,b) = rate of coincidence detection of y1,y2 with Pl(a),P,(b) in place;
R, = R,(a) = rate of detection of y , with P,(b) removed, assumed independent
of a; R, = R,(b) = rate of detection of yz with P,(a) removed, assumed
independent of b; then Eq. (58) may be transformed to
IR(u,b) - R(u,b’)l + R(u’,b) + R(u’,b’) - R , - R , I 0 (59)
For R(a,b) = R(a - b) and relative polarizer orientations
la - bl = la - b‘1/3 = la’ - bl = la’ - b’l = ~p (60)
EINSTEIN-PODOLSKY -ROSEN PARADOX 283
Eq. (59) simplifies to
3R(q) - R ( 3 q ) - R , - R , I 0 (61)
Defining
Ncp) = C3R(cp)- R(3cp) - R , - R,l/Ro (62)
with Ro being the rate of coincidence detection of yl, y z with Pl(a),
P,(b)
removed, Freedman and Clauser (1972)arrived at a lower bound - 1 for A(cp).
Hence, instead of Eq. (61) one may write finally
- 1 I [3R(cp) - R(3cp) - R1 - R , ] / R o I 0 (63)
Now, it may be shown that if the QM prediction for the correlation
function is of the form r + scosncp, then there is maximal violation of the
upper limit of Eq. (63) for ncp = n/4 and of the lower limit for ncp = 3n/4. In
the case of correlated cascade photons one has n = 2, and these angles
become, respectively, cp = n/8 and 3n/8. Inserting these two values for cp into
Eq. (63),and noting that R(9n/8) = R(n/8),one has
- 1 5 [3R(n/8)- R(3n/8) - R1 - R,]/Ro I 0 (64)
and
-1I [ 3 R ( 3 ~ / 8-) R(n/8) - R1 - R , ] / R o I 0 (65)
By subtracting Eq. (65)from Eq. (64),both inequalities may be combined to
one single inequality
(R(22.5")/R0- R(67.5")/R0(- I 0 a (66)
which no longer depends on R , , R , . Actually it is Eq. (66) which has been
tested in most cascade-photon correlation experiments.

2. The Proof by Bell


In his own derivation of the GBI, Bell (1972) starts from the idea that the
respective Stern-Gerlach apparati SG, (a)and SG,(b) carry their own hidden
information, which may be represented formally by HVs A,, & with domains
A,, I \ b and normalized distributions pu(Aa),p b ( & ) satisfying

JAa pa(Au) d l a = 1, la E Aa (67)

and
n
284 W. DE BAERE

The measurement results are now A(u,A,&) = f 1,O and B(b,A, &) =
+
- 1,0, the value 0 assigned to A or B if S1 or S2 is not detected. For the
correlation function P(a,b) we now have

with

44 4=
1. N u , A,&)pa(Aa) (704

and
B(b,A) =
1., B(b,A, i b ) p b ( & ) d& (70b)

I&&i)I I 1, lB(b,i)l I 1 (71)


If for SGl(a),SG2(b)alternative settings a’ and b’ are considered, then one
may write

P(a, b) - P(a, b’) =


I [A(a,i)B(b,A) - A(a,A)B(b’,i ) l p ( l )dA

= b A(a,A)B(b,A)[ 1 f A@’,i)B(b’,i ) ] p ( i )d i

A(a,i)B(b’,i)[l & A(a’,i)B(b,A ) ] p ( A ) d l (72)


-

or, because of Eq. (71)

IP(a,b) - P(a, b’)l I


I [l k A(a’, A)B(b‘,i)]p(i) d A

+ jAf c1 A(a’,i)B(b,41p(A)dA

< 2 f [P(a’,b’) + P(a’,b)] (73)


which finally may be written
+
IP(a,b) - P(~,b’)l IP(a’,b’) + P(~’,b)lI 2 (74)
This form of the GBI is similar to that first derived by Clauser et al. (1969),
EINSTEIN- PODOLSKY-ROSEN PARADOX 285
Eq. (58). It is seen that, as a result of the conditions in Eq. (71), the GBI of
Eq. (74) has to be satisfied by any stochastic HVT for which locality and coun-
terfactuality are assumed to be valid.

3. The Proof by Stapp


In his version of Bell’s theorem, Stapp (1971,1977) considers an ensemble
of N correlated spin4 systems S,(i), S2($) moving toward Stern-Gerlach
devices SG,(a), SG,(b). We will review this version in some detail because it is
claimed and believed (DEspagnat, 1984) that the concept of HV is not needed.
However, we will see that this is not correct, because the HVs I may be
considered as a formal representation of some elements in Stapp’s
argumentation.
It is assumed that each device has two alternative settings: a, a’ for SG,,
b, b’ for SG,. Call Aj(a,b) the prediction of a hypothetical, more complete
theory for the measurement of the spin component of the jth system S , along
a, and likewise for Bj(a,b). If Aj(a,b) and Bj(a,b) are actually measured, then
Stapp supposes that the theory predicts definite values for the numbers
Aj(a,b’), Bj(a,b’),Aj(a’,b), Bj(a’,b), Aj(a‘,b‘),and Bj(a’,b‘) which correspond to
alternative settings (a, b’), (a‘,b) and (a’, b’) of the Stern-Gerlach devices.
Moreover, it is assumed that these numbers would have been the measurement
results if, instead of the couple of actual settings (a,b), one of the three
alternative possibilities were chosen. This assumption is the essence of what
Stapp calls the hypothesis of counterfactual definiteness, and he further
assumes that QT also makes correct predictions for these hypothetical
measurement results.
Now, if Aj(a,b) and Bj(a,b) are measured, then according to QT Bj(a,b’)
can never be measured simultaneously with Bj(a,b) (unless b and b’ are
parallel). Hence, if the values Bj(a,b’),j = 1,. . .,N are to be used to define a
correlation function whose value is determined in a real experiment, then the
hypothesis of counterfactuality is equivalent to the following assumption.
Suppose that in one actual experiment with settings (a,b) we have obtained N
couples of results Aj(a,b),Bj(a,b), j = 1 , . .. ,N and from another actual
measurement with settings (a,b’) we have obtained the series of results
Aj(a,b’),Bj(a,b’). Counterfactuality then assumes that between the [ ( N / 2 ) ! I 2
permutations which make the series Aj(a,b‘), j = 1 , . .. ,N coincide with the
series Aj(a,b), j = 1,. ..,N there is at least one permutation such that all
internal conditions (which are supposed to be responsible for each actual
result) within S, and S, coincide exactly in both series.
It is readily seen that these “internal conditions” may be identified with the
former concept of HV 1. Hence, although HVs do not appear explicitly in
Stapp’s reasoning, they are implicitly contained in it. Note that only when the
286 W. DE BAERE

above condition is satisfied does it have sense physically to consider what


would have been the result if instead of an actual apparatus setting, one had
chosen another possible one. Suppose then further that such a reasoning is
allowed. Consider then the series of results

Aj(u,b), Bj(u,b), j = 1,. ..,N (754


Aj(u,b’), Bj(u,b’), j = 1,. . ., N (75W
Aj(U’,b), Bj(U’,b) j = 1,. .. ,N (754
A ~ ( u b’),
’, b’),
B~(u’, j = 1,. .. ,N (754
If the first series are actual results, then the other three are hypothetical, but,
on account of the assumed validity of counterfactuality, may be assumed to be
equivalent to actual results in some subsequent experiment. Under this
condition, then, Stapp introduces locality by the following requirements:

Aj(u,b) = Aj(u,b’) = Aj(u), j = 1,. . .,N (764


’ , = A ~ ( u b’)
A ~ ( u b) ’ , = A~(u’), j = 1,. . . ,N (76b)
Bj(u,b) = Bj(u’,b) = Bj(b), j = 1,. . . ,N (764
b’) = Bj(b’),
b’) = B~(u’,
B~(u, j = 1,. . .,N (764
The relevant correlation functions are then defined by (for N sufficiently
large)
1 N
P(u, b) = - 1
N j = 1 Aj(u)Bj(b) (774

1
c N
P(u’,b) = - Aj(U’)Bj(b)
N j=1

P(u, b’) = -
1
Nj=l
1N Aj(a)Bj(b’)
P(a’,b’) = -
1
N j=1
1N
Aj(U’)Bj(b’) (774

Stapp shows then that Eqs. (77a-d) are not compatible with the Q M
prediction of Eq. (41) for all conceivable settings a, a’, b, b’. Indeed, for O,, = O”,
eaZb= 135”, Oob, = 0”,and Oarb,= 45” one arrives at the contradiction f i I
1. Stapp concludes that the locality conditions of Eqs. (76a-d) have to be
false.
Recently Stapp (1985) has been defending his counterfactuality hypothesis
further. According to Stapp there are two different concepts of locality, both
EINSTEIN-PODOLSKY-ROSEN PARADOX 287
due to Einstein. The first is the one that is used in Einstein’s relativity theory,
according to which no physical signal can travel faster than light. The second is
the idea that events separated in space may in no way disturb each other, and
this should be incompatible with QM. However, both concepts must
necessarily have something to do with the propagation of influences, and it is
not clear why both propagation speeds should be different, at least not in a
unified world view. In Section V we will see, on the contrary, that many
arguments point to the validity of locality in both the above senses.

4 . The Proof by Selleri


Selleri (1972) has given a very simple proof of the GBI by using the result
[Eq. (53)] of Clauser et al. (1969)

IP(a, b) - P(a, b’)(= 1 -


I B(b,A)B(b’,A)p(A)d I

Noting that on both sides of Eq. (78) the sign may be changed and that the
(78)

right-hand side does not depend on a, one may also write

IP(a’,b)
I
+ P ( d ,b’)l = 1 + B(b,A)B(b’,A)p(A)d A

Adding Eqs. (78) and (79), one immediately gets the GBI of Eq. (74).
(79)

5 . The Proof b y Clauser and Horne


An alternative to the proof by Bell (1972) of the GBI which is valid for
deterministic, as well as for stochastic HV theories, is given by Clauser and
Horne (1974).
Assume analyzer orientations a and b, and suppose measurements are
made on N couples (Sl,S2).Call N , ( a ) and N,(b) the number of counts at
detectors D,and D, behind the respective analyzers. Call N,,(a, b) the number
of coincidence counts. If N is large enough, then one may write for the
probabilities of these counts

p12(a,b) = N,,(a, b ) / N (804


The internal conditions within the correlated systems S , , S , are again
represented formally by means of HVs I with a normalized distribution p ( l ) .
Clauser and Horne then assume that knowledge of these HVs and of the
288 W. DE BAERE

analyzer directions determines only the probabilities pl(a, A), pz(b,A), and
p12(a,b, A) for counts to be registrated at Dl, at D,,and at both D1 and D,.
Now, here locality is introduced by assuming that pl(a, A), pz(b,A) do not
depend on b and a and that
Plz(a9 4= P l h 4P,(b, 4 (81)
Equation (8 1) is known as the Clauser-Horne factorability condition. Also

Pl(4 =
I Pl(44 P ( 4d l (82a)

P2(W =
I P& A M 4d l (82b)

Pl&,b) =
I Plz(a,b,A)P(A)dA

To prove the GBI, Clauser and Horne consider two alternative orien-
(824

tations a, a’ for the first analyzer and two alternatives b, b’ for the second one.
From the inequality
- X Y Ix y - xy’ + x ’ y + x’y’ - X’Y - yx I 0 (83)
which is valid for real numbers x , x’,y , y‘, X , Y satisfying 0 I x , x’ 5 X ,
0 5 y , y’ IY, it is concluded that
- 1 5 p,,(a,b,A) - Pl,(4b‘,A) + Pl2(af,b,4 + P1z(a’,b’,4
-P1(U’,4 - pz(b,A) 5 0 (84)
Multiplying Eq. (84) by p ( l ) and integrating over A, one gets
- 1 5 P l 2 ( 4 b) - P l 2 ( 4 b’) + P1&’, 4 + P l Z b ’ , b’) - P 1 ( 4 -P 2 W I0

In the above inequality the probabilities may be replaced by the observed


counting rates &(a‘), R2(b),R(a,b), etc. defined in Clauser et al. (1969)(Section
III,D,l), to obtain
R(u,b) - R(u,b’) + R(u’,b) + R(u’,b’)
I1 (86)
&(a’) + R2@)
which is essentially the same as Eq. (59).
As already remarked when discussing the derivation by Clauser et al., to
compare Eq. (86) with atomic cascade experiments necessitates, as long as no
EINSTEIN-PODOLSKY-ROSEN PARADOX 289
perfect polarizers exist, the introduction of a supplementary assumption. This
is the so-called “no-enhancement’’ assumption, according to which the prob-
ability for photon detection with a polarizer in place, e.g., pl(u,A), is not
larger than the corresponding probability pl(A) with the polarizer removed
0Ip1(a,A) Ip1(A) s 1 (874
0 I pz(b, A) Ip#) s1 (87b)
for all I, a, and b. To save locality, the validity of Eqs. (87a,b) in the existing
cascade experiments has recently been seriously criticized (Section V,C).
The general validity of the factorability condition [Eq. (Sl)] may be
criticized (Selleri and Tarozzi, 1980; Selleri, 1982) by making the following
remarks. In the above scheme, the correlation function P(a, b) may be written
as

p(a?b) =
h
-Pk-
PAa+ A)PZ@+? 4- PI@+

9
9

A)PZ(b+ 9 4+ P l k - 9
9 4Pz(b-

A)P,(b-
7

2
4
M A )&
f

with pl(a*, A) the probabilities for A(a,A) = f 1. Now, suppose that A is just a
short-hand notation for the set of HVs A’,A“, . . .: A = (A’,A”,. . .). Then one may

s
write

P(a, b) = ql(a,A‘, I“,. . .)q,(b, 2,il”,. . .)

x p ( X , A”,. . .) dA’ d l ” . . . (89)


and, integrating Eq. (89) formally over A’, the new expression should also
satisfy the factorability condition, i.e.,

and one should have that


s
P(a, b) = rl(a,I”,.. .)rz(b,A”, . . .)p(A“, ...) dI“ dA”’ . . . (90)

rl(u, A”,. . .)r,(b,A“,. . .)

= jql(..r,A!!,. ..)qz(b,Z,i”, .. .)p(A’, A”,. .. ) d X (91)


290 W. DE BAERE

Dividing Eq. (91) by its a derivative, one obtains


r

with
r\(a,I’’, . . .) = dr,(a,I”,. . .)/da
(93)
q\(a, I’,A“, . . .) = dql(u, I’,I”,.. .)/da
It is seen that the left-hand side of Eq. (92) is independent of b, while its right-
hand side depends on b. This cannot be true in general; hence the same applies
to the factorability condition [Eq. (81)] on which Eqs. (88) and (90) are based.

6 . The Proof by Eberhard and Peres


Like Stapp, Eberhard (1977) and Peres (1978) claim to give a proof of the
GBI which starts only from the principle of locality and does not need to
introduce the HV concept nor determinism. The validity of counterfactuality
is considered as self-evident or at least a “rather natural way to thinking.” To
give an outline of the simple proof, we shall use the same notation as in Section
III,D,3 and start from Eqs. (77) for the correlation functions and from
Eqs. (76), which express the requirement of locality.
Eberhard’s derivation is based on the remark that for the j t h event one has
Xj = Aj(a)Bj(b)+ Aj(~’)Bj(b)
+ Aj(a)Bj(b’)- Aj(~’)Bj(b’)
I 2 (94)
because Aj(a’)Bj(b’)is the product of the other three terms, and Aj = f 1 and
Bj = 5 1. It follows from Eq. (94) that
1
-C Xj = P(u, b) + P ( d , b) + P(a, b’) - P ( d , b’) I2 (95)
N j
which is the generalized Bell inequality.
From the violation of Eq. (95) by QT, Eberhard concludes that Eqs. (76)
are not correct, hence requiring nonlocality, because he unconditionally
assumes counterfactuality to be correct.
Eberhard also gives an alternative formulation of Bell’s theorem: It
amounts to saying that, for certain relative orientations of four analyzer
directions a, a’, b and b, it is impossible to find four sets of actual spin
measurement results {Aj(a,b),Bj(a,b ) } , {Aj(u’,b), Bj(a’,b)}, {Aj(a,b’),Bj(a,b ’ ) } ,
and { Aj(a’,b’), Bj(a’,b‘)}for which Eq. (76) and hence the GBI Eq. (95)are valid.
EINSTEIN-PODOLSKY- ROSEN PARADOX 29 1
However, according to us, in this reformulation of Bell‘s theorem the
invalidity of Eq. (95) does not imply the breakdown of locality but rather the
breakdown of counterfactuality.
A proof of the GBI along similar lines has been given by Peres (1978).
Remembering that the derivation of the GBI is the result of combining one
series of actual data with three series of hypothetical ones, Peres distinguishes
two possible attitudes with respect to the violation of the GBI Eq. (95) by QM:
“One is to say that it is illegitimate to speculate about unperformed
experiments. In brief, ‘Thou shalt not think.’ Physics is then free from many
epistemological difficulties. For instance, it is not possible to formulate the
EPR paradox.” At this point Peres refers to the EPR statement “If ... we had
chosen another quantity . .. we should have obtained .. .,” which is according
to him “The key point in the EPR argument.” The second attitude is:
“Alternatively, for those who cannot refrain from thinking, we can abandon
the assumption that the results of measurements by A are independent of what
is being done by B.” Here A and B stand for SG,(a) and SG,(b) in our notation.

7 . Other Bell-Type Inequalities


We have seen that the hypothesis of locality (Hl) and of counterfactuality
(H3) leads in various ways either to the original BI [Eq. (46)] or to the GBI
[Eq. (58)l.Although the theoretical and experimental disagreement between
these inequalities and QT (Section IV) is already sufficient to discard one of
these underlying hypotheses, various people have extended the BI further by
constructing what we shall call Bell-type inequalities. In most of these
generalizations Eq. (46) or (58) is somehow contained.
Pearle (1970) considers n possible analyzer orientations a,, a,,. . . ,an for
one measuring device and n orientations b,, b,, . ..,bn for a second one. In
terms of a HV A with a normalized distribution p(A), and dichotomic
measurement results A(a,,A), B ( b j ,A) = 5 1, a correlation function P(a,, b j ) is
defined in the usual way as

P ( a i ,b j ) = A(a,,A)B(bj,A)p(A)dA (96)

Starting from the identity


292 W. DE BAERE

one easily derives the inequality


P(a,,b,) + P(a,,b,) + ... + P(a,,b,) I2n - 2 + P(a,,b,) (98a)
or
n
1 CP(ai,bi) + P(ai+ ,, bill I 2n - 2 + P(a,, b,)
i= 1
(98b)

For equal angles between adjacent vectors in the sequence a,, b,, a,,. . .,b,
Eq. (98b) simplifies to
(2n - I)P(a,b ) I(2n - 2) + P(a,, b,) (984
in which a, b represent any two adjacent vectors.
Applying the approach of Wigner (1970),DEspagnat (1975) considers an
ensemble, for which it is somehow possible to predict the number
n(rl,r2,. . ., r m ) of systems having dichotomic values r , , r , , . . .,rm for ob-
servables R,, R,, . . .,R,. A correlation function P(ri,r j ) is defined as the mean
value of the product rirj. D’Espagnat is then able to derive the general
inequality

On the basis of locality and counterfactuality Herbert and Karush (1978)


derived the following set of Bell-type inequalities
-(m - 1) ImP(8) - P(m8) - (m - 1) I0
-n InP(8) - P(n8) - n + P(0) I0 (100)
with m representing any odd integer and n any even integer.
Another method to generate Bell-type inequalities (Selleri, 1978)is to start
from the inequality
CaA(a.1) + /3B(b,A)+ yA(c,A)]’ 2 1 (101)
with a,& y = k 1, A(a,A) = f 1, etc., and B(a,A) = -A(u, A). It may be shown
that Eq. (101) generates inequalities such as
P(a,b) + P(a,c) + P(b,c) I 1 (102a)
P(u, b ) - P(u,c) - P(b,c) I1 (102b)
-P(a,b) + P(a,c) - P(b,c) I 1 ( 102c)
-P(a,b) - P(a,c) + P(b,c) I1 (102d)
Garuccio (1978) generalizes this method by considering the inequality
[aA(a, A) + /?A@, A) + yA(c, A)]’ 2 min( k a k /3 k y)’ (103)
EINSTEIN-PODOLSKY -ROSEN PARADOX 293
in which a, fl, and y are positive, but otherwise arbitrary, and a 2 B 2 y > 0.
With these conditions
min(_+a_+ a _+ Y )=
~ (-c( + + y)‘ ( 104)
and the following Bell-type inequality may be deduced
P(u, b ) + xP(u,C ) + yP(b,C) I 1 + x - y (105)
with a = y//? and y = y/a.
Other equivalent inequalities may be derived, such that instead of Eqs.
(102a-d) one may write
P(u,b) + xP(u,C ) + yP(b, c) I 1 + x - y (106a)
+ x -y
P(u,b ) - xP(u,c)- y f ( b , c ) I 1 (106b)
-P(u, b ) + xP(u,c)- y f ( b , c ) I 1 + x - y (106c)
- P(u,b ) - xP(u,C ) + yP(b,C ) I 1 + x - y (106d)
All these inequalities are again violated by QM for certain values of u, b, c.
From Eqs. (106a-d) follows an inequality with four correlation functions:
From Eqs. (106a) and (106b) we have
P(u, b ) + IxP(u,c)+ yP(b,c))I 1 + x - y (107a)
and from Eqs. (106c) and (106d), replacing x by x’, y by y’, and c by c’
+
- P ( u , ~ ) Ix’P(a,c’)- y’P(b,c’)l I1 + X ’ - y’ (107b)
Adding Eqs. (107a) and (107b), one obtains
IxP(a,C ) + yP(b,c)[ + Ix’P(u,c’) - y ’ f (b,c’)l
I 2 + x + x’ - y - y’ (108)
in which 1 2 x 2 y > 0, 1 2 x‘ 2 y’ > 0.
From numerical calculations for variable parameters x, y, x’, y’ it follows
that maximal violation of Eq. (108) occurs for x = y = x’ = y’ = 1; i.e., the
original GBI is the strongest inequality with respect to the QM singlet state.
A systematic derivation of all Bell-type inequalities for dichotomic
measurement results has been given by Garuccio and Selleri (1980). They use
the probabilistic formulation of Clauser and Horne (1974), in which the
correlation function is given by Eq. (88):

%(a, 4q2(b,4 P ( 4d i (109)


294 W. DE BAERE

with

and - 1 Iq,(u,A.) < 1, - 1 < q , ( b , l ) I 1.


Suppose we have analyzer directions a,, a,,. . .,a,,, for S1 and analyzer
directions b, ,b,, . ..,b, for S,. The intention of Garuccio and Selleri is then to
derive inequalities, for all possible linear combinations of correlation
functions, of the type
1cijP(ai,b j ) I M
i,j
(111)

in which cij are real coefficients. The result of their study is that appropriate
upper bounds M are given by

in which ti = & 1, yli = f 1.


Another method to construct all generalizations of the BI has been
proposed by Froissart (1981).The method appears to be very general because
it covers not only an arbitrary number of orientations of each measuring
device, but also an arbitrary number of correlated systems. By introducing a
“quality factor” for two-systems inequalities, Froissart shows that the original
Bell inequality has the best quality factor. This is consistent with the
conclusion of Garuccio (1978).
Other Bell-type inequalities have been derived by Selleri (1982) as a result
of a generalization of the current versions of the EPR paradox. The original
EPR definition (D2) of an “element of physical reality” has been criticized by
Selleri because of its insistence on predictability “with certainty.” Selleri
argues that, on account of the existence of the Wigner-Araki-Yanase
theorem (Wigner, 1952; Araki and Yanase, 1960; Yanase, 1961), it is in
principle necessary to have imperfect measurements, if Q T is to be correct.
Therefore, it is not justified to base an argumentation on the possibility in
principle to predict some physical observable with certainty.
Selleri proposes some new, generalized criteria for elements of reality and
separability, which are based on the possibility of predicting probabilities pi
for measurement results ui of an observable U on objects belonging to a
subensemble E’ of an ensemble E, without disturbing the objects. In this case
Selleri speaks of a physical property l l ( U ; p , , p , , . . .) of the subensemble E’.
Furthermore, it is said that, if another subensemble E“ of E is considered, E‘
and E” are separated, if the physical property l l ( U ; p , , p , , . . .) of E‘ does not
depend on what property ll( V; q l , q 2 , .. .) is measured on E“.
EINSTEIN-PODOLSKY -ROSEN PARADOX 295
On the basis of these generalized definitions, Selleri then derives the
following Bell-type inequalities
IP(a,b) - P(a,b’)l + IP(a’,b) + P(a’,b’)l + 81P(a,a)l I 10 (113)
P(a, b) = - (P(a,a)la.b ( 1 14)
Combining Eqs. (1 13) and (1 14), and choosing all analyzer orientations such
that maximal violation of the resulting inequality occurs, one arrives at the
inequality
2 4 + 8 I10/IP(a,a)l (115)
which is violated if IP(a,a)l > 0.923.

E. Generalization of the Bell Inequality to Arbitrary Spin

Until now we have considered only generalizations of Bell’s inequality


which were concerned with two-state systems or with two-dimensional
subspaces of many-state systems for which only dichotomic measurement
results are possible. In recent years, however, it has been shown that the
conflict between QM and all the above-mentioned Bell-type inequalities is not
restricted to spin-$ systems, but may be extended to systems with arbitrary
spin s.
The first consideration of correlated systems on which nondichotomic
quantities A(a, A), A(a’,A), B(b, A), and B(b’,A) may be measured came from
Baracca et al. (1976). It was shown that, if values M, and M, exist such that
IA(441IMI, t’@’,4l 4 M ,
IB(b,A)I I M , , lB(b’,A)l IM , (116)
then the following generalization of the GBI [Eq. (74)] exists:
IP(a, b ) - P(a, b’)l + /P(U’,b’) + P ( d , b)l I 2M2 (117)
with
M 2 = M1M2
However, inequality (1 17) is apparently not a very strong one for the following
reason. If A(a,A), B(b, A), . .. are the results of measuring the observables
- -
j a, j b,. . . for the spin projections of systems S,, S , along a, b,. . ., then
Baracca et al. (1976) show that in the singlet state one has
P(a,b)=(001j-aj-b100) = - * j ( j + 1)a.b (1 19)
and that this QM result agrees with Eq. (1 17) for all possible orientations a, a’,
b, and b’ for j > 5.
296 W. DE BAERE

Yet it appears that observables exist, without having a direct physical


meaning such as spin projections, for which the QM predictions violate
Eq. (117).
Interesting spin-s Bell-type inequalities which are more successful with
respect to their violation by nondichotomic observables that have a clear
physical meaning were constructed by Mermin (1980).
The basic inequality from which Mermin starts is
slms,(a) + ms,(b)I 2 -ms,(a)ms,(c) - ms,(b)ms,(c) (119a)
in which mS,(a),m,,(b), and m,,(c) denote spin components of S , along
arbitrary directions a, b, and c, s the spin value, and
S1)* als, ms,(4> = ms,(a)ls, ms,(a)> ( 120)
.
etc., with Sf1) a the operator corresponding to the spin projection of S , along
a, and Is, ms,(a)) the corresponding state vector. Assuming perfect correlation
between S1, S2,for which mS,(a)= - ms2(a),and taking expectation values of
both sides of Eq. (119), one has
s(OO11(S(’). a - S(’) b)llOO)-
-
2 (OO1S(’) as(’) clOO) + (001S(’)- b S 2 )- C l W (121)
in which (Edmonds, 1957)
1 +S

1
l o o ) = ~ m m s , = - s
Is,ms,(a))ls,ms2(a) = -ms,(a)>(- l)s-msJe) ( 122)

Mermin shows then that for the above expectation values the following
results hold true:
-
(001S(” * as(’) b100) = -$(s + l)a - b (1 23a)
and

with direction y perpendicular to the plane (a, b), and the spin components m
and m‘ both taken along a. If it is assumed that a, b, and c are coplanar and a
and b make an angle 8 + n/2 with c (Fig. 6) then Mermin’s GBI for spin s takes
on the form

This inequality has to be satisfied for all 8 if locality is supposed to be


compatible with QM. However, it may be shown that Eq. (124) is violated by 8
EINSTEIN-PODOLSKY -ROSEN PARADOX 297
*
C

FIG.6. Relative orientiations a, b, and c for which Mermin’s spin-s version of the GBI
applies.

values in the range


0 < sin0 < sine, = 112s (125)
If x = sin 8, then explicit expressions of Eq. (124) and corresponding
values 8, for low spin values are as follows:
spin 3:x 2 2 x, 0, = 90”

spin 1: 2x2 - x 4 2 x, e, = 38.170


spin 3: 5x2 - 6x4 + 3x6 2 $x, 0, = 24.08”

spin 2: 20x2 - 42x4 + 48x6 - 20x8 2 5x, 0, = 17.58”


In the classical limit s + co, the above contradiction disappears and
ms,(a), etc., become simply components of a classical angular momentum
along a. It may be shown further that in this limit the inequality (119) reduces
to s sin 0 2 $ 9 sin 8 and hence is always satisfied.
Further work on spin-s Bell-type inequalities has been pursued by Bergia
and Cannata (1982), Mermin and Schwarz (1982), and Garg and Mermin
(1982, 1985).
It may be observed that all possible BIs and GBIs can be classified
according to how they are derived in a deterministic local HVT [in which the
results A(a,A),B(b,A) are fixed] or a probabilistic one [in which only
probability functions pl(a,, A ) , p 2 ( b i , A) exist]. The relation between both
kinds of approaches has been discussed by Garuccio and Selleri (1978) and by
Garuccio and Rapisarda (1981).

F. Direct Proof of QM Nonlocality

Because of the crucial importance of the issue of locality, it is interesting to


note that recently Rietdijk and Selleri (1985) claimed to have proven “that the
locality hypothesis is in any case untenable if the predictions of quantum
mechanics are all correct.” We shall give below a physical outline of their
argumentation.
unpolarized beam A I.$>' No transmission
7 2

To illustrate how the process of state-vector reduction, applied to an


ensemble of correlated systems S,, S,, strongly suggests the existence of
nonlocal influences, we shall consider the following four steps:
Step I. (Fig. 7) An unpolarized beam of spin-$ systems S,($) moves in the
+ z direction towards a polarizer Pl(+x) with its spin analyzer in the + x
direction. Suppose that only systems polarized along this direction are
transmitted; i.e., the spin state of the transmitted beam is described by the state
vector luili ). A second polarizer P,( - x) will subsequently stop all transmitted
systems.
Step 2. (Fig. 8) Insert a third polarizer P3(a)between Pl(x)and P,(-x).
Now we have
luilj) = cos(8/2)lu!y) - sin(8/2)lubl)) ( 126a)
luil?) = sin(8/2)lu;y) + COS(~/~)~U:~!) (126b)
which means that a fraction c0s2(8/2) of the beam transmitted by Pl(x)will
pass P3(a).From
Iu!?) = cos(8/2)lui11) + sin(8/2)lui1!) (127a)
Iub'?) = -sin(8/2)1ui1i) + COS(~/~)~U~~!) (127b)
it follows that P,( - x) will transmit a further part sin2(8/2); i.e., P,(a) is
responsible for the fact that from the beam transmitted through Pl(x)a part
cos2(8/2)sin2(8/2) will now pass Pl(-x), instead of zero in Step 1. Hence, it
may be concluded that the passage of S, through some apparatus (e.g., a

FIG.8. Unpolarized beam moving toward three polarizers.


EINSTEIN-PODOLSKY-ROSEN PARADOX 299

?”

FIG.9. Two correlated systems moving toward two polarizers with opposite analyzer
directions.

polarizer) changes by a local interaction the internal physical conditions


within S , .
Step 3. (Fig. 9) Consider now a source which produces a beam of spin-*
systems S,($) and S,($) in the singlet state IOO), such that S , moves toward a
polarizer P,(a), and S2 toward a polarizer in the opposite direction, P2( - a)
(Fig. 3). With each S , that passes P,(a), there corresponds an S2 that passes
Pz( -a). From the state vector of Eq. (15), it follows that the beams that
are transmitted through P,(a) and P2(- a) are described by the state vector
Iu;J)luh’l). From Step 2 it follows that the internal conditions within each S2
that has passed P2(- a) are changed by a local interaction between S , and
PZ(- 4.
Step 4. (Fig. 10) The same situation as in Step 3, but now P2(- a) has been
removed. Quantum mechanically we have the same description as in Step
3: The statistical properties of the couples ( S , , S,) for which S , passed P,(a)
must again be calculated by means of the reduced state vector Iub?)lub”l).
Now we have that S , has changed locally by passing through P,(a), and
apparently this process has steered the correlated system S2 such as to become
a member of an ensemble described by a well-defined state vector, namely,
Iuh?). The difference with Step 3 is that there a local physical interaction was
responsible for this steering, while here this is lacking. Rietdijk and Selleri
argue that the only way to avoid the conclusion that these systems S, were
already described by lub2!) before S , passed through P,(a) is to assume that

FIG. 10. Two correlated systems; only one polarizer is present.


300 W. DE BAERE

precisely at this moment the internal conditions within S, are changed


nonlocally (i.e., by action at a distance), so as to become described by luL2!). In
this way state-vector reduction in the case of correlated systems strongly
suggests a nonlocal behavior at the quantum level.

G . Recent Developments

An understanding of the origin of the BIs or of the nonlocality of QT in


terms of Einstein’s own unified field theory has been attempted by Sachs
(1980) and by Bohm and Hiley (1981). It is well known that Einstein’s attempts
to construct a fundamental theory of matter are based on an extension of his
general relativity and on a representation of matter and radiation by a
continuous field, obeying deterministic nonlinear field equations. It was hoped
that the probabilistic and alleged nondeterministic predictions of QT could be
derived somehow from such a deterministic and classical scheme.
It is argued by Sachs that in his own approach Einstein finally rejected the
concept of “Einstein locality,” in the sense of the existence of physical systems
which could be supposed to be described independently of anything else in the
world, in particular of measuring devices. Hence, according to Einstein, the
idea of separability of physical systems should not be considered as a
fundamental one (Einstein, 1949). Sachs then goes on to discuss the conflict
between QT and the BI within his own version of the above deterministic
scheme, in which a maximum velocity c (the velocity of light) for the
propagation of physical influences is built in, and which reproduces QT in the
linear limit.
A similar attempt at understanding the Q M nonlocality along the same
lines comes from Bohm and Hiley (1981). After recalling that Einstein rejected
the “. . . fundamental and irreducible feature of the quantum theory, i.e.,
nonlocality,” and describing Einstein’s views on locality, these authors
propose a way how nonlocality could follow from Einstein’s approach.
Remembering that in Wheeler and Feynman’s absorber theory of radiation
(Wheeler and Feynman, 1949) a photon is emitted when there is somewhere
matter to absorb it, Bohm and Hiley suggest a similar mechanism in the case
of correlated photons. If the analyzer directions are a and b, then it may well
happen that these circumstances help to determine the way the photons leave
the source. In this way they are led to assume that the distribution of HVs will
no longer be represented by p(A), but instead by p(a, b, A). And this would lead
to an HV framework in which the usual BIs can no longer be derived.
Cramer (1980) constructs a QM version of the Wheeler-Feynman
absorber theory (Wheeler and Feynman, 1949) as an explanation for the
nonlocality which follows from the theoretical and experimental violation of
the BI. It is shown that within this approach there can be a nonlocal
EINSTEIN-PODOLSKY -ROSEN PARADOX 30 1
communication between measurement devices separated in space, so as to
produce a violation of the BIs.
In recent years an interesting connection has been revealed between the
existence of a joint probability distribution for the values of some observables
and the validity of the BI. As in Clauser and Horne (1974) (Section III,D,5) let
us introduce the experimental probability distributions plz(a, b), p , ,(a, b’),
plz(a’,b), plz(a’, 0 p l ( 4 pda‘), p,(b), and p,(b’) for spin projections or
polarizations of systems separated in space S,, S, along directions a, a’ and
b,b. Then Fine (1982a,b)has proved that in any theory (with or without HVs,
deterministic or stochastic, local or nonlocal) whose framework provides a
joint probability distribution p(a,a‘, b, b’) for the values a *, a;, b +, b+ ( = f1)
of the respective dichotomic observables, such that the above experimental
distributions are obtained as marginals of p(a,a’, b, b’), the GBI [Eq. (85)]
should hold:
-1 Plz(a,4 - Pl2@,6’) + Plz(a’,b) + P,&’,b’) -P l ( 4 -PAN I 0
(128)
It may be shown that by appropriate interchanging of the parameters in
Eq. (128) seven similar equivalent inequalities exist. Writing for the previously
defined correlation functions P(a, b)

P(a,b) = P I Z ( Q + , ~ + - +) ~ i z ( a + , b + )(129)
) Plz(a+,b-) - P I Z ( ~ - , ~ +

and similarly for the other correlation functions P(a, b‘), P(a‘, b), and P(a’,b’),
one may arrive (de Muynck, 1986)at the GBI [Eq. (58)l in the form given by
Clauser et al. (1969) and by Bell (1972).
These results are very interesting because it appears that for the derivation
of the BIs the locality hypothesis does not play any role. In this respect it may
be worthwhile to mention that Edwards (1975) and Edwards and Ballentine
(1976) have constructed nonlocal HV models which satisfy the BI. And in
contrast with this Scalera (1983) and Caser (1984b)(see below) have developed
local models which violate the BI. These examples seem to support the thesis
of the irrelevance of the locality condition with respect to the BIs.
As a first illustration of such models with anomalous properties, we note
that Scalera (1983) constructed a kind of local classical helix model for two
correlated photon beams moving in opposite directions. Each beam is
assumed to be represented by a continuous helix-shaped ribbon. If a photon
passes some polarizer with its analyzer direction along a, it means that the
associated ribbon passes undistorted, but somehow the directions f a are
marked on it. If a photomultiplier is to receive and to count such a photon, it is
assumed that the energy contained within two marked directions is integrated.
A correlation between the two ribbons produced at the source originates by
302 W. DE BAERE

their being marked in the same direction. Scalera then shows that the BIs may
be violated by this model.
Another local model that violates the BIs has been constructed by Caser
(1984b).The peculiarity of this model is that it starts from the assumption that
the state of a measuring apparatus depends on some or all previous
measurements made by means of it. It is argued that QM is compatible with
such a memory effect. By showing quantitatively that the model satisfies all
physical requirements of locality, however without being identical to Bell’s
definition, it appears that the BIs can be violated.
That the violation of the BIs is not a peculiar property of correlated
quantum systems has been shown by Aerts (1982a, 1984).If one considers two
macroscopic systems which are correlated because of the fact that they are not
completely separated (e.g., there are some connecting tubes or wires), then
results of measurements made on them will also be correlated, and the GBI
may be shown to be violated. Hence, according to Aerts, the problem is to
separate physical systems completely, whether quantum mechanical or
classical. If it were possible really to produce such separate systems, then the
GBI would be verified and QM should turn out to be wrong in describing such
systems. Hence, Aerts (1982b) concludes that the present QM is unable to
describe separated systems. For this reason, a more general scheme than QM
is proposed that encompasses both CM and QM (Aerts, 1983).
To explain the observed violation of the GBI, hence the alleged
nonlocality, one may attempt to set up concrete physical models. This has
been done by Caser (1982), who remembers that, according to Mach’s
principle, the inertial properties of matter are determined by the distribution
of far-away matter in the form of stars, etc. In the same way, Caser argues, it
may be that, although the magnetic field H of a Stern-Gerlach analyzing
device is different from zero only in the region of the device, the vector
potential A (H = V x A) is nonvanishing in a much larger region, which may
include the source of correlated spin4 systems. Now from the existence of the
Aharonov-Bohm effect (Aharonov and Bohm, 1959), it follows that the
behavior of a quantum system depends on the local values of A. Hence, it may
be, according to Caser, that the values of A at the region of the source
influence the distribution of HVs in such a way that one has to write p(a,b, A)
instead of p(A) (a,b being the orientations of the fields H in both Stern-
Gerlach devices). It follows that Bell’s theorem is no longer valid under these
conditions, and it is shown explicitly how such a model can reproduce the
results of QM.
In a subsequent work Caser (1984a) shows that in an anisotropic space, in
which two measuring devices act in a different way on physical systems, a local
HVT may reproduce the QM results for spin correlation experiments.
Moreover it is shown that, unlike the argumentation of Clauser et al. (1969),
EINSTEIN-PODOLSKY -ROSEN PARADOX 303
no supplementary assumption about the detection probabilities is needed
when the HVs lie in the plane of the analyzer directions.
A further illustration of the disagreement between Q M predictions and the
BI when the relevant observables do not all commute pairwise, has been given
by Home and Sengupta (1984). They show that the violation of the BIs is not
only characteristic for systems having correlated components separated in
space. It is argued that instead one may equally well consider an ensemble of
single systems and measure the compatible pairs (A(a),B(b)), @(a), B ( b ) ) ,
(A(a’),B(b)),and (A(a’),B(b’))on the individual members of the ensemble, of
course under the condition that only one pair is measured at a time.
Adopting a specific labeling scheme which transforms nondichotomic
results to dichotomic ones, they are able to derive Bell-type inequalities of the
form
- P(a, C) + P(b, C ) +
P(b, a) I 1 (130)
or
IP(a,b) + P(a, b’) + P ( d , b) - P ( d ,b’)l < 2 (131)
with the usual definition [Eq. (39) or (77)] of the correlation functions.
They consider then an ensemble of atoms (e.g., alkali atoms), which have a
single valence electron in the 2p1,, state. In such a state the total angular
momentum is J = +,the orbital angular momentum is L = 1, and the spin of
the valence electron is s = +. If m,, rnL, and m, denote the components of the
respective angular momenta along some arbitrary direction a, then one may
write for m, = ++
1.1 = +,m, = $) = ( ~ / J I )=(l,m,
~ ~= 1)ls
I L= + , m , - +>
-IL = l , m L = O)ls = +,m, = ++)) (132)
Now Home and Sengupta take as observables and their operators
A(a):L a, . A(b):L b, - A(c): L c,-
B(a):a a, . B(b):a b,- -
B(c):a c (133)
and calculate P(a, c), P(b,c), and P(b,a)
P(a,c) = (+,+1L . a a - c l + , + )
= $c0s2(e,/2) - $sin2(8,/2) - 3 (134a)
P(b,c) = (+,+lL*ba-c1+,3)
= +os2(el - e,/2) - $sin2(6,/2) - 5 (134b)
- -
P(b,a) = (f,+IL b a a[+,+)
= $cos2 o1 - 71 (1 34c)
304 W. DE BAERE

In Eqs. (133) and (134), a, b and c are coplanar unit vectors and 8, = Oab,
8, = Oat. The left-hand side of Eq. (130) becomes
-5 + 4 2 +cos2(ez/2)
~ 0 ~ - 0 ~ + $cosz(e, - 02/2) (135)
which for, e.g., 0, = 7c/6 and 0, = 2n/3 becomes 4, such that the BI [Eq. (130)]
is violated.
Franson (1985) remarks that the recent experiment of Aspect et al. (1982)
(Section IV,A) does not exclude local theories in which a measurement result
may be known only some time after the measurement event has occurred.
Under these circumstancesit should be possible that information between two
measurement devices could be exchanged with subluminal velocities.
In a recent paper Summers and Werner (1985) show that the violation of
the BIs is also a general property of a free relativistic quantum field theory. Use
is made of the Reeh-Schlieder theorem (Reeh and Schlieder, 1961; Streater
and Wightman, 1964), according to which any local detector has a nonzero
vacuum rate. It is shown that the vacuum fluctuations in free-field theories are
such that the BIs are maximally violated. It is concluded that all physical and
philosophical consequences with respect to the violation of the BI apply also
in free quantum field theories.
Some proposals for a resolution of the (non)localityproblem have already
been discussed in Section II,E,F. Other recent investigationson that issue with
respect to the BI will be reviewed in Section V, where the significance of the BIs
and their violation will be considered from a critical point of view.
A number of papers which may be of help in the study and clarification of
the problems raised by EPR and the BIs are those of Berthelot (1980), Bub
(1969), Corleo et al. (1975), Flato et al. (1975), Gutkowski and Masotto (1974),
Gutkowski et al. (1979), Liddy (1983,1984), Rastall (1983,1985), and Stapp
( 1979).

Iv. EXPERIMENTAL
VERIFICATION OF BELL’SINEQUALITIES

In this section we will give a brief survey of the results of existing


experiments which were intended to verify the GBI [Eq. (86) or (66)] of
Clauser et al. (1969) (Section III,D,l) and at the end we will mention
some new proposals. For more details on the experiments themselves and for a
thorough discussion of the experimentalproblems connected with verification
of the GBI [Eq. (86) or (66)], we refer to the review papers of Paty (1977),
Pipkin (1978), Clauser and Shimony (1978), or to the original papers
themselves.
EINSTEIN-PODOLSKY -ROSEN PARADOX 305
In their paper Clauser et al. (1969) also discussed the inefficiency of
the then-available data on polarization correlation of photons (Wu and
Shaknov, 1950; Kocher and Commins, 1967) with respect to an experimental
discrimination between QM and HVT via their GBI [Eq. (58) or (63)]. They
proposed a concrete new experiment which would allow the verification of
Eq. (63).
The idea of Clauser et al. was to optimize the experiment of Kocher and
Commins (1967),i.e., to choose optimal relative polarizer orientations of 22.5"
and 67.5O, to increase polarizer efficiencies, and also to make observations
with one polarizer and then the other removed. Under these conditions one
will get sufficiently reliable data to be used in the GBI or CHSH inequality (66)
lR(22.5")- R(67.5")]/RoI I (136)
They actually made two proposals: One was for the J = 0 -+ J = 1 +
J = 0 electric-dipole cascade, while the other was for the J = 1 J = 1 --+ --+

J = 0 cascade. In terms of the relative angle cp between the two polarizer


orientations and of the half-angle 8 of the cone in which the photons are
detected, the QM predictions for the polarization correlations are for both
cases
R(cp)IRo = &f +E m f i +4
+ $(&;I - E:)(Efi - Efi)F1(8)COS243 (137)
R i / R o = ( ~ f l+ ~ 1 ) / 2 , i = 1,2 (138)
In Eqs. (137) and (138) and E\ (i = 1,2) are the efficiencies of the ith
polarizer for light polarized parallel and perpendicular to the polarizer axis
and, further, for the cascade J = 0 + J = 1 J =0 -+

Fl(8) = 2G:(8)[G2(8) +
iG3(8)]-1 (139)
G,(8) = a(+
- cos 8 + sin2 8 - 3cos38) ( 140a)
G2(@ = 8 - +(sin28 + 2) cos 8 ( 140b)
G3(8)= 4 - cos 8 - ~ C O 0S ~ ( 140c)
For the second cascade J = 1 --+ J = 1 -+ J = 0, the relations in Eq. (140)
remain valid, but in Eq. (137) -Fl(8) should be replaced by
F2(8) = 2G:(8)[2G2(8)G3(8) iG:(8)]-' + (141)
Assuming that E: and E: are vanishingly small and takings1 = ~ f =i it may
be seen that Eq. (137) may be violated in both cascades for experimental
parameters satisfying the relation
J Z e ( 8 ) + 1 > 2 / ~ ,, , j = 1,2 (142)
306 W. DE BAERE

A , Cascade Photon Experiments

1 . Experiment of Kocher and Commins (1967)


This was the first correlation experiment with cascade photons. The source
was the 6 IS, excited state of Ca. Via a two-step cascade J = 0 + J =
l+J=O
(6 IS,)
4
(4 lP1) 4,= 4227
(4 lS0)
4 (143)
towards the ground state, two photons y1 and y2 are emitted. Because the total
angular momentum of y1 and y2 is zero, the QM transition probability is
proportional to (el .e2)', with e, the polarization vector of photon y i . The
results were in agreement with this prediction: For perpendicular polarizer
settings no coincidences above background were measured, while there was a
high coincidence rate for parallel polarizer settings. However, the experiment
was not appropriate for a verification of the GBI [Eq. (136)] because the
efficiencyof the polarization was not high enough and only the angles cp = 0"
and 90" were considered.

2. Experiment of Freedman and Clauser (1972)


In this experiment correlated photons y1 and y2 were obtained in the
following J = 0 + J = 1 + J = 0 cascade in calcium
. 5513 A I,,, = 4227 A
(4p2 IS,)
=
* (4P4S Pl) ,(4s IS,)
The polarizer efficiencies had the following values
~ f =i 0.97 f 0.01, E: = 0.038 f 0.04
(144)
~ f =i 0.96 f 0.01, E: = 0.037 0.004
The half-angle 6, within which the photons were detected, was 8 = 30".
In different runs of the experiment the results R(22.5")/R0= 0.400 f 0.007
and R(67.5")/R0 = 0.100 f 0.003 were determined. Denoting by dexp,and, ,a
the experimentally determined and theoretically predicted values for the
left-hand side of Eq. (136), then it is found that aexpt= 0.300 k 0.008, which
violates the BI of Eq. (136), but is in agreement with the Q M prediction
+
, ,a = 0.301 0.007.
This was the first clear evidence for the validity of Bell's theorem which
excludes all local HV theories.

3. Experiment of Holt and Pipkin (1973)


Here the source was the spin-0 isotope Ig8Hg of which the J = 1+
J = 1 + J = 0 cascade was used to produce photons y1 and y 2
EINSTEIN - PODOLSKY -ROSEN PARADOX 307

The polarizer efficiencies were as follows:


E;, = 0.910 0.001, E: <
(145)
E ; = 0.880 k 0.001, E! <
and the half-angle 0 = 13". In this case one has dexpt= 0.216 k 0.013,
in agreement with the GBI 6 I0.25, but contradicting the Q M prediction
6,M = 0.266.
At the time this experiment was carried out, this result was exciting because
now HVT was favored over QT. However, as reported by Clauser and
Shimony (1978), the origin of the disagreement may lie in an improper
mounting of some components, such as lenses, in the experimental arrange-
ment of Holt and Pipkin. Some indication for this came from a repetition of
this experiment by Clauser (1976a,b).

4 . Experiment of Clauser (1976a,b)


Stimulated by the previous surprising results, Clauser repeated the
experiment of Holt and Pipkin with only minor changes, such as the use of a
spin-0 isotope 'O'Hg instead of 19'Hg. The experimental parameters were
&;I= 0.965, E: = 0.01 1

~ f =i 0.972, E: = 0.008 (146)


e = 18.60
The result was hexpt= 0.2885 k 0.0093, violating the GBI 6 I 0 . 2 5 , but in
agreement with the Q M prediction: 6,, = 0.2841. It is seen that these results
do not support the conclusion of Holt and Pipkin.

5 . Experiment of Fry and Thompson (1976)


The experiment consisted in the study of the J = 1 +J = 1+J =0
cascade in the spin-0 isotope '"Hg
(7 3s,) 4, = 4358 A, (6 3p1)' 7 1 = 2537 4 (6 lS0)
The experimental parameters were as follows:
&i
= 0.98 f 0.01, E: = 0.02 k 0.005
&; = 0.97 f 0.01, E: + 0.02 f 0.005 (147)
0 = (19.9 f 0 . 3 ) O
308 W.DE BAERE
The result Bexp1 = 0.296 k 0.014 was again in agreement with BQM = 0.294
+
- 0.007, but violated the GBI: 6 I 0.25.

6. Experiment of Aspect et al. (1981)


Use is made of the following J = 0 + J = 1 -P J = 0 cascade in calcium
(4p2 Is,) 17, = 5513 5
(4S4P Pl )
I,, = 4227 A
44s2 IS,)
There were several improvements with respect to the results of the previous
experiment of Fry and Thompson (1976). First of all the excitation rate was
more than ten times greater. Hence the statistical accuracy was much better
than ever before. Correlation measurements for the full 360” range of relative
polarizer orientations were in agreement with the Q M predictions,for various
distances (up to 6.5 m) between the source and the polarizers. In addition this
means that, up to these distances, the Aharonov-Bohm hypothesis (Section
II,D) of a pure state evolving spontaneously towards a mixed state is not
correct.
The experimental parameters and results are
&i
= 0.971 0.05, E! = 0.029 & 0.005
6; = 0.968 & 0.005, E: = 0.028 & 0.005
( 148)
F(9) = 0.984
Bexpt = 0.3072 Ifr 0.0043, B Q M = 0.3058 & 0.002
In order to avoid the Clauser-Horne assumption of rotational invariance
(Section III,D,4), the more general GBI [Eq. (86)]
- 1 I S = [R(a,b) - R(a, b’) + R(a’,b)
+ R(a’,b’) - &(a’) - R,(b)]/R, I 0 (149)
was tested for relative orientations
$,b = eazb = eaPb* = gab!f 3 = 22.5” (150)
The result Sexpl= 0.126 & 0.014 was in good agreement with the QM value
SQM = 0.1 18 k 0.005.

7 . Experiment of Aspect et al. (1982)


In all previous experiments the polarizer directions were fixed during each
run of a correlation measurement. In these cases it could be criticized that
perhaps somehow these static directions influenced the way the source emits
the correlated photons, such that Bell’s locality conditions [Eq. (76)] would no
longer be valid: This would then be a plausible explanation for the observed
EINSTEIN-PODOLSKY -ROSEN PARADOX 309
experimental violation of the GBI. Hence, Bell (1964) argued that, in order to
prevent such a possible influence, it would be of crucial importance to carry
out “. . . experiments of the type proposed by Bohm and Aharonov, in which
the settings are changed during the flight of the particles.”
The experiment of Aspect et ul. (1982)was a refinement of that of Aspect et
al. (1981), in which it was attempted to realize these requirements. However,
they were fulfilled only partially, because the switching was by means of an
acousto-optical device which operated periodically each 10 ns. Yet the two
switchesoperated independently of each other, so that it is plausible to assume
that their global effect on the measurements on both sides was uncorrelated.
Also, as the distance between source and polarizers was up to 6.5 m, the
condition of measurements separated in space was very well satisfied.
The version of the GBI that was verified in this experiment was

+ R(co’,
R(a’,b’)
-
R(a’,co)
-
R(co,b )
00’) R(co’,co) R(co, 00)
I 0 (151)

which is similar to Eq. (149). In Eq. (151) the symbol co corresponds to


coincidence measurements in which the respective polarizer is removed, e.g.,
R(oo’, 00) means that the polarizers with orientations a’ and b are removed.
Again the relative orientations of Eq. (150) were chosen, and an average
value S,,,, = 0.101 t 0.020 was determined from two runs. This corresponds
with the predicted value S,, = 0.112 and violates S I0 by five standard
deviations.
The results of this experiment are at this moment considered as the
strongest evidence against all local realistic HV theories.

8. Experiment of Perrie et al. (1985)


The most recent correlation experiment with cascade photons is that of
Perry et al. (1985),who used a source of metastable atomic deuterium D in the
state 2s. In this case, as a result of a return to the ground state lS, the two
correlated photons are emitted simultaneously (in contrast with the former
cascade experimentswhere emission is consecutive due to the finite lifetime of
an intermediate state). This implies that, if the sites of observation are
equidistant from the source, the condition of events separated in space is
fulfilled.
For the polarizer transmission efficiencies it is found that
= C; = 0.908 k 0.013, E: = E: = 0.0299 k 0.0020

and with these values the QM prediction (137) for R((p)/R,is well verified.
310 W. DE BAERE

Furthermore, one has that dexp,= 0.268 & 0.010, in agreement with the
prediction d,, = 0.272 f 0.008, but violating the form of Eq. (136) of the BI
by two standard deviations.
It is seen that the above-mentioned experiments by Aspect et al. (1981,
1982) provide up to now the clearest evidence against local HV theories.
However, the main reason for carrying out the present experiment was to
eliminate a recent criticism (Section V,C) of the current interpretation of the
results of the former two-step cascade experiments. In essence this criticism
amounts to the idea that absorption and subsequent re-emission of photons in
the source could be appreciable and significant, such that the conclusion of
complete agreement with QM and nonlocality would not be justified. Now it
appears that with a source of D(2S) this problem is negligible, and from the
above results, confirming QM, it would follow that the only remaining
loophole for concluding nonlocality would be the “no-enhancement”
hypothesis (Section III,D,5).

B. Positronium Annihilation Experiments

Another source of correlated photons is the annihilation of an electron


and a positron at rest. From the experiment of Wu and Shaknov (1950), it
followed that the spin parity J p for the ground state of positronium is J p = OK.
If spin parity is conserved during the annihilation, then according to QM the
correlated photons should be described by the state vector
I$> = (1/Jz)(lX>lI 0,- I V l I X ) 2 ) (152)
if both photons move opposite in the z direction.
In Eq. (152) I X ) , represents a state of photon y,, which has the property of
being linearly polarized in the x direction, and similarly for the other states. If
it were possible to measure, by means of ideal polarization analyzers, the
polarization of y1 along a (lz)and of y2 along b (lz),then it can be shown that
QM predicts for the correlation function P(a, b) the result
P(a, b ) = P(rp) = - cos 2rp (153)
with rp the angle between a and b.
Now, as the above dichotomic measurement results are of the type
considered by Bell, the GBI [Eq. (74)] should also be valid for the above
situation, provided locality holds
IP(a, b ) - P(a, b’)l + IP(a’,b’) + P(a’,b)l I2 (154)
Comparing with the QM prediction [Eq. (41)] for the spin-9 case, it is seen
immediately that Eq. (153) violates Eq. (154), too.
EINSTEIN-PODOLSKY -ROSEN PARADOX 31 1
However, the problem with annihilation photons is that, although the
efficiency for detection is very good, the efficiency of linear polarizers is by far
insufficient. Therefore, information about the correlation between the planes
of polarization of the respective photons can only be obtained indirectly, e.g.,
via measuring the polarization-dependent joint distribution for their
Compton scattering. From this reliance on QM and from the fact that
supplementary assumptions are needed (see below), in order to allow for a
comparison between QM and local HV theories via the GBI, this type of
correlation experiment is considered less reliable than cascade-photon
experiments.

I . Experiment of Wu and Shaknov (1950)


The first experiment with photon annihilation was that of Wu and
Shaknov, whose results were already used by Bohm and Aharonov (1957) to
disprove the assumption that each of the two correlated photons was already
in a well-defined quantum state, before any measurement is carried out
(Section 11,D).
In this experiment coincidence rates between the Compton-scattered
photons are registrated for two different arrangements: The first rate, r L , is
for perpendicular scattering planes (Fig. 1, Section II,D) while the second rate,

,
r l l ,is for parallel scattering planes. QM predicts a value of 2 for the ratio r =
r l / r , which agrees very well with the experimental result r = 2.04 k 0.08.
However, the GBI [Eq. (154)] cannot be verified directly because it is not
possible, from this kind of experiment, to obtain a clear answer about the
precise polarization of each photon of the pair. The information one gets is
only statistical.

2. Experiment of Faraci et al. (1974)


A more recent experiment with photon annihilation was that of Faraci et
al. Referring again to Fig. 1 (Section II,D), let us call 0, the angle between pyl
and pyI,0, the angle between pi2 and py2,and cp the angle between the planes
(Pt,,P,,) and (Pi2,PY2).
From the method of partial-polarization analysis (Snyder et al., 1948),it
follows that for the coincidence rate W ( 4 ,B,, cp) we have

+ 2sin2 0, sinZ0,[(2k - l)sin2 cp + 1 - k]} (155)

in which K O is the wave number of the annihilation photons, K i the wave


number of yi scattered in direction pii with scattering angle 0,, a, = K J K ,
+ K , / K i , k = 1 for the pure Q M state [Eq. (152)], and k = i f o r a mixture of
312 W. DE BAERE

products of photon states, such as


l ~ ~ l l ~ ~ 2 ~ l ~ ~ 2 l ~ ~ l ~

As a measure for W(O1, 8,, cp), the number of coincidence counting rates
N ( 8 , , 02,cp) of Compton-scattered photons was observed and the ratios
N(8,8, 90°)/N(0,8, Oo) and N(60",60°,rp)/N(60",60°,Oo) determined for
various values of 8 and cp. The results were within the limit permitted by the
BIs, and disagreed with the QM predictions.
Furthermore, the ratio N(60",60°,90")/N(60°,60",Oo)was determined as a
function of the difference in the flight paths of the two photons. It was found
that there was a disagreement with QM and that the data allowed for a
description in terms of a mixture of product states.

3. Experiment of Kasday et al. (1970, 1975)


In order to use two annihilation photons for verifying the GBI [Eq. (154)],
Kasday et al. make the following two supplementary assumptions. The first is
that it is possible in principle to construct ideal analyzers for linear
+
polarization. If results 1 and - 1 are assigned if a photon is polarized
parallel or perpendicular to the polarizer axis, then the QM prediction for the
correlation function P(a,b) for annihilation photons is given by Eq. (153).
Now, such ideal polarizers do not exist, so that recourse has to be made to
Compton scattering of the respective photons. The second assumption in
snnihilation experiments is then that the results that would have been
obtained with ideal polarizers and those obtained via Compton scattering are
correctly related by QM. This means that one may use the Klein-Nishina
formula for Compton scattering, which depends on the plane of polarization
of the incoming photon.
The quantity measured by Kasday et al. was the coincidence rate W ( q )=
W(8, = 90°,8, = 90°,cp), for which the QM prediction is
W V ) = 1 - M,M,P(cp) (156)
In Eq. (156) MI and M 2 are instrumental factors and P(cp) is the correlation
function [Eq. (153)] for ideal polarizers. The results of this experiment were in
agreement with QM and the GBI was violated.

4 . Experiments of Wilson et al. (1976)and of Bruno et al. (1977)


The experiment of Faraci et al. was repeated with different experimental
equipment by Wilson et al. (1976) and by Bruno et al. (1977). The results of
both experiments agreed very well with the predictions of QM, and no
influence of the difference in the flight paths of the two photons was observed.
EINSTEIN-PODOLSKY -ROSEN PARADOX 313
C. Proton-Proton Scattering Experiment

Following a suggestion by Fox (1971), Lamehi-Rahti and Mittig (1976)


studied low-energy S-wave proton-proton scattering at 13.2 and 13.7 MeV.
The idea was to investigate the spin correlations between the protons in the
final state of the interaction. In the ideal case it should again be possible to
determine the spin components of both correlated protons along directions a
and b.Now, as in the case of annihilation photons, such ideal analyzers do not
exist, and the correlation function P(a,b) as defined by Eq. (39) cannot be
determined directly. Therefore, other processes with the correlated protons
must be used, which depend on their polarization state. The extra process
which is used here is the interaction of the protons with carbon foils C , and C,
(Fig. 11).
Detectors A, A' are placed behind C1,and detectors B, B' behind C,. It is
assumed that plane (B,B') coincides with the reaction plane (pl, p,) and that 6
denotes the angle between the planes ( A ,A') and (p,, p,). In the experiment, a
correlation function P(@, defined as

is determined. In Eq. (157), N A B is the number of coincidence counts in


detectors A and B, etc.
As in the annihilation case, the link with the GBI [Eq. (154)] can again be
established only at the cost of introducing similar supplementary as-
sumptions. It is found that with these assumptions good agreement is obtained
with QM, and that the GBI is violated. However, a drawback of this
experiment is that there is no separation in space between the measurements.

FIG. 11. Experimental setup in proton-proton scattering.


314 W. DE BAERE

TABLE 111
LISTOF CORRELATION
EXPERIMENTS

Type of experiment Source Authors Results agree with:

Atomic cascade photons Ca (1) Kocher and Commins (1967) QM HVT?


Ca (2) Freedman and Clauser (1972) QM
'"Hg (3) Holt and Pipkin (1973) HVT
'"Hg (4) Clauser (1976a,b) QM
'''Hg (5) Fry and Thompson (1976) QM
Ca (6) Aspect et al. (1981) QM
Ca (7) Aspect et al. (1982) QM
D (8) Perrie et al. (1985) QM
Annihilation photons (1) Wu and Shaknov (1950) QM
(2) Faraci ef al. (1974) HVT
(3) Kasday et al. (1975) QM
(4) Wilson et al. (1976) QM
(5) Bruno et a2. (1977) QM
Proton-proton scattering (1) Lamehi-Rahti and Mittig (1976) QM

In fact, it is one of the extra assumptions that this does not influence the
mutual dependence of the measurements.
The conclusions of the above correlation experiments are summarized
in Table 111.

D . Recent Proposals for New Experiments

As a complement to the current EPR experiments, a number of new


experimental tests for Einstein locality or for the BIs have been proposed
recently.
Roos (1980) considers the interaction e+e- -+ K ° K o and shows that if
locality holds then some decays K,K, have to be observed, which is excluded in
the QM scheme if charge conjugation C and parity P are conserved.
Selleri (1983a) and Six (1984) also discussed the observable difference
between the Q M description of the .Ip = 1-- state of the K ° K o system and
the description based on a mixture of product states [which follows from the
simplified version of the EPR argumentation (Section ll,A,B)]. Selleri showed
that, because of the absence of interference terms in the latter description
(Furry, 1936a; Section II,C), certain correlated observations on the K ° K o
system are predicted to be 12% lower than predicted by QM.
Tornqvist (1981) investigated the reaction e+e- -+ A1\ -+ n-pn'p and
considered the decay distribution W(p,+, pz-) of n- and TL+ in the final state. It
appeared again that there is an observable difference between the predictions
EINSTEIN-PODOLSKY -ROSEN PARADOX 315
of both the above kinds of description: According to QM
- + ap,+ ‘Pn-
W(P,+,P,-) 1 (158)
whereas were Einstein locality valid, one should have, instead of Eq. (1 58)
WP,+,P,-) - + @/3)P,+ ‘Pn-
1 (1 59)
In Eqs. (158) and (159) a is some parameter depending on Q M S- and P-wave
amplitudes. Up to now none of these proposals has been tested in real
experiments.

v. BELL’SINEQUALITIES
INTERPRETATION OF
AND OF RESULTS
EXPERIMENTAL

A . The Current Interpretation

As a result of the almost general agreement with QM of those experiments


which were set up to discriminate between Q M and all local HV theories
(Section IV), the prevailing opinion has settled that it has now been decided, on
rather safe empirical grounds, that nonlocality is a basic feature of Nature at
the quantum level.
The origin of this state of affairs is probably that in almost all derivations
of the BIs or similar inequalities (Section III,C,D), the importance of the
locality hypothesis has not only been overestimated, but in many of them it is
considered as the only condition. Therefore, when reviewing these derivations,
we have called attention to the equal importance of the hypothesis of
counterfactuality, which is made either implicitly or explicitly and then
considered as self-evident. We will see below that the neglect of this last
hypothesis is at the origin of the main criticism of the nonlocality interpre-
tation of the BIs.
The large attention that is paid to the locality hypothesis is further reflected
in various statements. According to Stapp (1975) “Bell’s theorem is the most
profound discovery of science,” and Vigier (1979) is of the opinion that, with
respect to the standard predictions of QM, “Its experimental issue might
prove as important for the future evolution of physics as the negative result of
the Michelson-Morley experiment.” Vigier continues that “. . . the theoretical
discussion started by Bell has shown that a positive experimental test of
the truth of quantum-mechanical measurement predictions in the case of
correlated particles.. . has far-reaching consequences which.. . imply a des-
truction of the Einsteinian concept of material causality in the evolution of
Nature, since they would establish the physical reality of nonlocal interactions
316 W. DE BAERE

between spacelike separated instruments of measurements.. . .” In the view of


Bohm and Hiley (1981), “It seems clear from both the theoretical analysis and
from the experiments that some kind of nonlocality is a fact ...,” and on
account of the results of Aspect et al. (1981, 1982), Rohrlich (1983) concluded
that the “local hidden variables theory is dead.”
A very long list of similar statements may certainly be found in the
literature. Remember also that recently Rietdijk and Selleri (1985) claimed to
have proven nonlocality (Section 111,F). Furthermore, various people, con-
sidering nonlocality as a fact, have been constructing rather detailed models
for its explanation (Vigier, 1979; Halbwachs et al., 1982; Section 111,G).
Whether or not the alleged nonlocality property may be used for superluminal
signaling is still a matter of debate (Hall et al., 1977; Herbert, 1982; Ghirardi
and Weber, 1983).

B . Criticism of the Current Interpretation and


Alternative Interpretations

I . Joint Probability Distributions and Bell’s Inequalities


Although for many people nonlocality is now an established fact, there are
others who on no account can accept such a conclusion, because it violates
the relativity principle. However, in Sections III,C,D we have seen that the
counterfactuality hypothesis is as important as the locality hypothesis for
the derivation of all the BIs. Einstein locality can therefore be retained by
abandoning counterfactuality.
The first criticism along these lines came from de la Peiia et al. (1972). These
authors remark that, on account of Bell’s theorem, an eventual nonlocal HVT
reproducing all results of QM requires the instantaneous propagation of
signals and can therefore not be Lorentz invariant. The alternative, the
acceptance of the ultimate and irreducible random character of physical
events as expressed by the quantum formalism, is according to de la Peiia et al.
equally unacceptable. Therefore, they re-examine the BI and show under what
conditions the BIs are to hold.
Their criticism directs against the simple derivation of the BI by noting the
following: Because all terms in the derivation of the BI [e.g., in Eq. (44) or in
Eq. (53)] use the same HV A for the quantities A(a, A), A(a’,A), B(b,A), B(b’,A),
they must necessarily refer to four measurements carried out on the same pair
( S , , S2). Now, if A(a, A) and A(a’,A) are not compatible, then these cannot be
known or measured simultaneously, at best measured by subsequent measure-
ments. Because of the large empirical success of QT (which incorporates in its
formalism this restriction), this must simply be admitted as true.
EINSTEIN-PODOLSKY -ROSEN PARADOX 317
Now, de la Peiia et al. remark that under these conditions (of subsequent
measurements) it may happen that the original density p ( l ) , effective for
A(a, A), may be changed by the measurement of A(a, A). In a recent paper
Rietdijk and Selleri (1985)(Section III,F) have shown that A is indeed changed
by the measuring process. According to de la Peiia et al., when such a local
change happens, this does not require in any way a nonlocality in the theory.
The result of their analysis is a modified Bell inequality

IP(a, b) - c(a’,b) + Pb(a, b’) + &(a‘, b’)l 2 ( 160)


in which P,(a’,b), P,(a, b’), and &,(a’, b’) may be called conditional correlation
functions. For example, C(a’,b) uses the results A@’) and B(b) after A(a) has
been determined previously. The same applies to Pb(u,b’) and to P,,(a’, b’).
With this new interpretation of the BI, it is seen that the QM predictions
for the various correlation functions, for the singlet spin state, are as follows
(de la Peiia et al., 1972; De Baere, 1984b)
P(u, b) = -COS (161a)
c(a’, b) = -cos 6,b cos eao# (161b)
Pb(a, b’) = - cos 60, cos e b b , (161c)
eb(a’,b’) = - cos 6,b cos eaa,cos e b b , (161d)
Inserting these predictions into Eq. (160), one obtains the inequality
lcos 6,b[cos oaa!(1 + cos ebb?) + cos 6bbf - 1115 2 (162)
which is satisfied for all possible orientations. de la Peiia et al. conclude: (1)the
original BIs [Eq. (46) or (58)] are valid only if the measuring process does not
affect the distribution of the HVs A (which is not the case; see above) and all
measurement operators commute [which is not the case in actual experiments
(Section IV)]; (2) the modified BI [Eq. (160)] applies under the assumption
that the measurement process affects the distribution p(A) and it is not
necessary that all the operators commute.
In further work Brody and de la Peiia Auerbach (1979) show that the BI
already holds under the condition of “joint measurability.”According to these
authors Peres’ conclusion that “Unperformed Experiments Have No Results”
(Peres, 1978) is equivalent to well-known properties of the state vector for
correlated systems, and that in correlation experimentswe “. ..merely observe
the persistence of a correlation when there is no interaction available to
destroy it.” It is shown very clearly that, just as in later work by Fine (1982a,b),
the only condition necessary to derive the BIs isjoint measurability of the four
quantities A(a), A(a’),B(b), and B(b‘), the first two being measured on S , , the
318 W. DE BAERE

last two on S,. Hence, what experiments with respect to the BI tell us is
whether or not these four quantities can be measured simultaneously.
According to Brody and de la Peiia Auerbach, then, “We see that four
independent series of data allow us to determine whether the system under
study possessesjoint measurability or not. This point, we believe, is central to
an understanding of Bell’s inequality and yet has been almost universally
ignored.” Furthermore, “In other words, the Bell inequality requires a
condition of joint measurability of spin projections along two different
directions on the same particle; it is this fact, rather than any nonlocality of
hidden variables, which helps us to explain why the inequality is inconsistent
with quantum theory.. .,” and “We conclude that the Bell inequality tells us
nothing about hidden variables, not even about their nonlocality.. ..” For an
interesting discussion, from this standpoint, of the most important derivations
of the BIs we refer to two recent papers by Brody (1985a,b).The present author
is also of the opinion that the above points are basically correct and that they
amount to the nonvalidity of counterfactuality, as suggested at various places
in this paper.
Similar criticisms against the BI and Bell’s theorem were raised by Bub
(1973), Lochak (1976), and Lehr (1985).
Bub shows that the Wigner procedure of taking a probability measure
over the six-dimensional subspace of phase space [which leads to the Wigner
inconsistency [Eq. (52)] in the case of two correlated s p in 3 systems], leads to
the same inconsistencies in the case of a single spin-4 system. The point is again
that it is assumed that there exists a distribution for the values of incompatible
observables.
Lochak‘s criticism is against the use of one single distribution p(A) for the
HVs 1 in all the correlation functions figuring in the BI. It is shown that this
amounts to the implicit hypothesis that the values of two noncompatible
observables may be determined via one single measuring instrument or via
two mutually compatible instruments. This, however, is not allowed in the
Q M formalism and hence it is not a surprise that Q M predictions violate the
BIs. Also, Bell’s theorem that, on account of this violation, all local theories are
impossible is necessarily incorrect, just because it is based on the above
incorrect implicit assumption. We have seen (Section 111) that this implicit
hypothesis (equivalent to the counterfactuality hypothesis) is used in any
proof of the BIs.
An interesting clarification of the significance of the BIs came from Fine
(1982a,b) and from de Muynck (1986) (Section 111,D). Fine’s results are
valuable because they show that the problem at issue, locality, is irrelevant for
the BIs. This has been illustrated in Section III,D by various nonlocal HV
models which satisfy the BIs and by local models which violate the BIs.
Moreover, it appears that the only requirement for the validity of the BIs is the
EINSTEIN-PODOLSKY-ROSEN PARADOX 319
existence of a joint probability distribution (JPD) for the simultaneously
measured values of the relevant observables [i.e., A(a),A(a’),B(b),and B(b’)],
and such that the observed distributions, figuring in the BIs [i.e., p12(a,b),etc.],
are recovered as marginals. Some points in the argumentation of Fine have
been criticized by Eberhard (1982) and by Stapp (1982a)and have been replied
to by Fine (1982~).
As stressed by various authors (de la Peiia et al., 1972;Lochak, 1976;Brody
and de la Peiia Auerbach, 1979;Fine, 1982a,b;de Muynck, 1983;Rastall, 1983;
de Muynck and Abu-Zeid, 1984; Kraus, 1985a,b; De Baere, 1984a,b; Brody,
1985a,b),if this possibility of simultaneous measurement is not fulfilled the BIs
may be violated. According to de Muynck and Abu-Zeid (1984) “If this
interpretation is correct, then the EPR-type experiments do not constitute a
test of the Bell inequalities because only two of the four observables are
measured jointly in these experiments. In order to relate such experiments to
the Bell inequalities we would have to construct a J P D of the four observables
that are involved, from the probability distributions of the pairs of ob-
servables measured in the EPR experiments. Even if such a construction is
possible mathematically the physical significance of merging the outcome of
difSerent experiments into one single probability statement is completely
obscure in a theory dealing only with measurement results. In orthodox
quantum mechanics the joint measurement of incompatible observables is
held to be impossible. J P D s of incompatible observables are judged to be
meaningless. For this reason the Bell inequalities cannot be derived in this
theory, as necessary properties to be obeyed by all EPR experiments.”
In recent work (De Baere, 1984a,b,c)the above points have been illustrated
by concrete examples. As a first example one may consider the following decay
chain of a spin-zero object S(0)
S(0) + Si(1) + SL(1) -P S,(+) + S,(+) + S,(+) + S,(+) (163)
For this situation one may choose as compatible observables the spin
projections of S , , S,, S, ,S4 along, respectively, the unit vectors a, b, a’,b’, with
results denoted by A,(a),A,@), A,(a’),A4(b’) if one prefers locality to be valid,
or by A,(a;a’,b, b’), A,@; a, a’, b’), A3(a’,a, b, b’), A,(&; a, a’, b) in a nonlocal
scheme. Using the Eberhard-Peres procedure (Eberhard, 1977; Peres, 1978;
Section III,D,6), one arrives in both schemes at the GBI [Eq. (74)]. Now, in
order to calculate the correlation functions between spin-component measure-
ments on s,,. . ., S, one has to consider the following state vector for the
situation of Eq. (163)
I$> = loo> = (l/~)Clu:)lu:)Iu3)Iu4)
-(1/$)(lu: > b 2 ) + lu’>lu: ))(1/$)
x (lu:)lu:) + Iu3)Iu4+))+ lu’)lu2)lu:>lu4+>1 (164)
320 W. DE BAERE

The results for the relevant correlation functions are


P(U, b) = (0()1U1 * aU2 * bJOO) = 3 C O S (165a)
~ ( u , b ’ ) =<OOlul. au,. b’lOO) = -+cos&,, (165b)
P(a’,b) = <001U3 *a’U2 *b(OO)= -$COSe,.b (16%)
P(u’,b) = ( ~ ~ u , ~ a ’ u , ~ b=+cost&,.
~00) (165d)
Inserting (165a-d) into the GBI [Eq. (74)] gives
13cos 6,b + jcos &,* + 13cOs oath! - jcos 8,Pbl <_ 2 (166)
which is satisfied by all possible directions of the spin analyzers.
Another example (De Baere, 1984c) is the decay of a spin-4 system into
three distinguishable spin-4 systems
sc+, + Sl(9 + S2C9 + S,(+) (167)
such that all orbital angular momenta are zero. In this case one may consider
the simultaneous measurements of spin components A,(u), A,@), A,(c) of
S,, S,, S,. One of the relevant Bell-type inequalities is
IP(u, b) + P(u,c ) + P(b,c) - 11 I 2 (168)
and for the QM predictions of the respective correlation functions one has
P(U, b) = - ( F i - 3F:) COS 6,b (169a)
P(u,C) = - 2Fl(FO/JT +- F1/3)cos O,, (169b)
P(b,C ) = 2F,(Fo/JT - F1/3)Cos 8bc ( 169c)
In Eq. (169) Fo and Fl are the so-called fractional parentage coefficients (Rose,
1957),which appear in the QM description of the process in Eq. (167). It may
be shown that the correlation functions of Eq. (169) never violate the BI
[Eq. (168)l. The example of the QM correlation functions [Eq. (161)] always
satisfying the conditional BI [Eq. (160)] is a further illustration of the points
discussed above.
Now, we have seen that no one of the existing correlation experiments
fulfills the condition of joint measurability, because the correlation functions
figuring in the BIs [Eqs. (46) and (58)] are determined in separate experiments.
Recently it has been argued (de Muynck and Abu-Zeid, 1984; De Baere,
1984a,b,c;Kraus, 1985a,b; Todorov, 1985a,b,c)that an interesting alternative
interpretation may be given to the violation of these BIs. The argumentation is
based on the idea that HVs 1are representations of the state of individual
systems, or, equivalently, for their internal physical conditions. It has been
EINSTEIN-PODOLSKY - ROSEN PARADOX 321
shown that, for the original BIs to be valid, it is necessary that in different
(similar) experiments these physical conditions are somehow reproduced, or,
in technical terms, that the respective HV distributions p(A) remain the same.
Under this condition, then, it appears that one is allowed to suppose that spin
components A(a,A), A(a’,A) are known for one single individual state A. It is
seen that this is Stapp’s counterfactuality hypothesis, which basically contra-
dicts QM. Now, the fact that the BIs for such situations (i.e., separate
experiments) are violated may well point to a new basic property of Nature
which we have called the nonreproducibility of physical situations at the HV
or quantum level. It follows also that Einstein locality may be maintained. The
concrete examples discussed above may be considered as supporting such a
viewpoint, because precisely there the reproducibility hypothesis does not
have to be made.
With respect to this picture, let us note that Feingold and Peres (1980)
studied the probability that Bell’s locality conditions [Eq. (76)] are violated.
They concluded that in at least ,/? - 1 = 41% of the cases the result of a spin-
component measurement on one of two correlated systems must depend on
what is measured on the other system. In our picture this result would cor-
respond to a measure of the nonreproducibility of physical situations.

2. First- and Second-Kind States and Bell’s Inequalities


Another interpretation of the BIs (Capasso et al., 1973; Baracca et al.,
1976) is that these allow for a discrimination between a description of
ensemblesof correlated systemsin terms of pure states (also called states of the
second kind, describing an ensemble then called an improper mixture) or in
terms of product states (also called states of the first kind, describing ordi-
nary or proper mixtures). Observables which are able to discriminate be-
tween the two kinds of description are called sensitive observables (Capasso
et al., 1973; Baracca et al., 1976; Cufaro Petroni, 1977a,b; Fortunato et al.,
1977; Section 11,A).

3 . Angelidis’ Criticism on the Clauser-Horne Factorability Condition


In Section III,D,S we discussed Selleri’s arguments (Selleri and Tarozzi, 1980;
Selleri, 1982) against the general validity of the Clauser-Horne factorability
condition [Eq. (82)]. Another criticism of this condition came from Angelidis
(1983), who criticizes the general validity of the Clauser-Horne version
[Eq. (63)] of the GBI, (A(cp)l I 1, for all values of cp and for all local HV
theories.
Remember that such theories were defined by Clauser and Horne
(1974) (Section 111,D,5) as those for which there exist probabilities
322 W. DE BAERE

pz(b,A), plz(a,b, A) and a unique normalized HV distribution ~ ( 2 )If.


pl(a, i),
the factorability condition of Eq. (82) holds, if p ( l ) does not depend on a, b, and
if p l ( 4 = pl,pz(b) = p2,p12(a,b)= p 1 2 ( b - bl), then IA(cp)I 5 1 has to be
valid for any cp.
The point Angelidis makes is that the above Clauser-Horne universality
claim is incompatible with what he calls conservation of angular momentum,
expressed by
P l ( 4 - PZ@) =0
or

for any a and b. From this condition Angelidis concludes that

Pl(a94 - p,(b,A) = 0
hence

P h 4 = Pl(b,4 = Pl(4 (1 72a)


and similarly

P2(% 4= Pz@, 4= P 2 ( 4 (172b)


But then, on account of the Clauser-Horne factorability condition [Eq. (Sl)]
one should have that
PlZ(4 b, 4= P1(4P2(4 (173)
is independent of a, b, which is physically impossible and contradicted by
observation.
As expected with such controversial matters, Angelidis’argumentation has
been criticized vigorously recently by a number of authors (DEspagnat, 1984;
Garg and Legett, 1984; Barut and Meystre, 1984; Horne and Shimony, 1984;
Cushing, 1985). Most of them point to a logical error in Angelidis’ argumenta-
tion: Namely, in order to infer Eq. (171) from Eq. (170), it is necessary that
Eq. (170) be valid for arbitrary square-integrable functions p(A). However,
as Horne and Shimony (1984) note, in Eq. (170) p(A) is not arbitrary at all.
Moreover, Garg and Legett remark that in the most recent correlation
experiments by Aspect et al. (1981,1982) it is not IA(cp)I I 1 which is used, but
instead the inequality (151) which is free from Angelidis’ criticism. It follows
that, according to the above authors, nonlocality is unavoidable, In his replies
Angelidis (1984a,b) claimed that the above critics have misunderstood his
argumentation.
EINSTEIN-PODOLSKY-ROSEN PARADOX 323
4 . Nonmeasurable Sets and Bell’s Inequalities
Another attempt to save locality and to resolve the apparent problem
of nonlocality came from Pitowski (1982a, 1983), who constructed a local
deterministic model that reproduces the QM statistical predictions for
correlated spin-3 systems.
The model introduces spin functions s(x) for any unit vector x with the
property s(-x) = -s(x). It is supposed that the precise form of these spin
functions is not known. Introducing some particular assumptions about the
way the set {xls(x) = *} is distributed over the sphere 1x1 = 1, Pitowski proves
mathematically that the observed coincidence frequencies, compatible with
QM, are reproduced. The idea behind it is that there exist distributions of spin
values that are nonmeasurable in the mathematical sense. The above
approach is based on an extension of the concept of probability to
nonmeasurable sets.
This approach has been criticized subsequently by Mermin (1982) and
Macdonald (1982) on the grounds that from Pitowski’s model Bell-type
inequalities may be derived which are violated by the QM predictions. In his
reply Pitowski (1982b) stresses that these inequalities are based also on the
counterfactuality hypothesis (besides the hypothesis of locality) and hence do
not invalidate his proposed model.
This model is interesting because it shows clearly that it is not impossible to
build local deterministic models which reproduce the results of QM for EPR-
like situations with correlated systems. However, the physical significance of
sets of points which are nonmeasurable in a mathematical sense is not very
clear.
Based on these ideas Gudder (1984) has shown further that in a generalized
probability theory the concepts of reality, locality, and HVs can be made
compatible with QM for systems with spin.

C . Criticism of Supplementary Assumptions

We have seen (Sections III,D,5 and IV,A,B,C) that, in order to be allowed


to use the available experimental data for the verification of QM or of the BIs,
it has always been necessary to introduce some plausible, but not directly
verifiable, supplementary assumptions. The number of such assumptions is
least in cascade photon experiments and for this reason their results are
considered as the most reliable. Generally, these additional assumptions have
to do with the relatively low efficienciesfor optical photon detection and with
the so-called “no-enhancement’’condition, expressed by Eq. (87).
Because these hypotheses are by no means trivial, and on account of the
lack of any indication for their correctness, the current interpretation of the
324 W. DE BAERE

correlation experiments with cascade photons has been seriously criticized


during the past years (Marshall et al., 1983a,b; Marshall, 1983, 1984; Santos,
1984; Selleri, 1984, 1985; Garuccio and Selleri, 1984; Marshall and Santos,
1985). Local HV models were proposed which either violated at least one of
the additional hypotheses (mainly the condition of no-enhancement) or even
did not need them.
The first who showed the crucial importance of the no-enhancement
assumption were Clauser and Horne (1974). They constructed a local HV
model that reproduced all QM results for correlation experiments, based
on the (according to them) physically implausible idea of enhancement.
Furthermore, Marshall et al. (1983a) proposed “. .. a very plausible model.. .
with enhancement features, which, though not giving identical predictions to
the quantum-mechanical model, fits the experimental data very closely.” In
particular this is illustrated for the available data of Aspect et al. (1981).
Marshall et al. (1983b) and Selleri (1984) also called attention to the necessity
of absence of collective effects in cascade-photon experiments in order to
supply reliable data. Such an effect may be, e.g., the resonant rescattering of yz
in the atomic cascade due to absorption by other atoms in the ground state
and subsequent re-emission. These authors calculated that in concrete
experiments such resonant scattering should be very important. As a result of
these developments, it is concluded that the evidence for nonlocality, coming
from the existing cascade experiments (mainly from those of Aspect et al.,
1981, 1982), cannot yet be considered as conclusive. It is insisted that further
experimental work be done.
Results of such work has been reported very recently by Aspect and
Grangier (1985), who show, on theoretical and experimental grounds, that the
above resonant scattering has been overestimated by Marshall et al. (1983b)
and Selleri (1984). It appears then that in the experiments by Aspect et al.
(1981, 1982) such collective effects are negligable. Moreover, the most recent
data by Aspect and Grangier (1985) again clearly violated the BIs, even
without subtraction of accidental coincidences. Furthermore, it is worth
noting that in the recent correlation experiment by Perrie et al. (1985) (Sec-
tion IV,A) the process of absorption and re-emission of photons does not
play a significant role, yet the results confirm QM.
Another possible origin of the apparent nonlocality has been given by
Pascazio (1985). It is remarked that a further supplementary assumption,
which is always made implicitly, is that the coincidencemeasurements on both
photons occur locally. Now, Pascazio argues that this is not the case because
the electronic equipment on both sides of the arrangement may be considered
as a source of nonlocality. The argumentation is accompanied and supported
by a simple model.
EINSTEIN-PODOLSKY -ROSEN PARADOX 325
VI. CONCLUSIONS

In this work we have given a survey of past and recent work on the
problems related with the (original and simplified) EPR argumentation and
the Bell inequalities. These problems have been considered mainly as being the
completeness of QM and QM nonlocality. We have seen that both topics
(EPR, BI) are related because they start from common hypotheses, namely
locality (H 1) (explicitly) and counterfactuality (H3) (implicitly).
From a compilation of existing correlation experiments (Section IV),
testing QM and the BIs, during the last decade, it may now be concluded
almost with certainty (however, see Section V,C, for recent criticism on the
validity of supplementary assumptions and of the data) that these BIs are
violated and QM confirmed. If this is correct, then the crucial problem is to
know how this evidence has to be interpreted, or to bring about its real
significance.
According to the current view (Section V,A), opinion has apparently settled
that locality is responsible for the violation of the BI and is incompatible with
the QM framework. Hence QM, and any future theory with it, should be
basically nonlocal. This conclusion originates certainly from the dispropor-
tion importance attributed to the hypotheses (Hl) and (H3).
However, there exists a relation between the hypothesis of counterfactu-
ality (H3) and the assumption of the existence of a joint probability
distribution (JPD) of observables which are, in the QM framework, incom-
patible (Section V,B,l). The important point now is that if such a JPD is
assumed, the BIs are satisfied automatically and the converse is also true. It
follows that from this point of view the whole issue of locality is irrelevant with
respect to the BIs and that Bell’s theorem does not apply. The lesson from the
empirical violation of the BI is then clear: (1) Any future (HV?) theory,
superseding QT, should not contain in its formalism a prescription for
constructing JPDs for incompatible observables; (2) (H3) is invalid. It follows
immediately that the original EPR argumentation, which leads to the
incompleteness statement of QM, is also invalid. However, this does not
necessarily mean that Einstein’s concept of elements of physical reality is
wrong, nor his idea of an independent reality with its own intrinsic properties.
We have seen that a small extension of the original EPR ideas may realize a
reconciliation between EPR and Bohr (Section 11,F).
At first sight it would seem then that the BI should be deprived of irs
original importance (the issue being the fundamental property of locality).
However, that is not the case. This may be seen by trying to give a physical
interpretation of the violation of (H3). In each verification of the BIs (Sec-
tion IV) one combines individual measurement results coming from at least
326 W. DE BAERE

two different experiments (corresponding to different experimental param-


eters). We have shown (Section III,C,D) that, in order to be allowed to do so,
(H3) should be valid. Now, this is apparently not true, and we have argued that
a physical reason for this may be that on the quantum level the conditions
(tentatively represented by HVs A) which determine the choice between several
possibilities (e.g., spin up, spin down) in a real experiment, do not always
reproduce themselves in the future, and hence are not interchangeable. For-
mally, one could say that one should write p(A.,t) instead of p(A). And this
“nonreproducibility” would be as interesting and astonishing as nonlocality.
In this way one may be led to the wholeness idea of Bohm (1980), in which
the world does not contain separated systems and is eventually evolving from
a starting point to an end point, in such a way that the propagation of in-
fluences are in agreement with the relatively principle, and hence are not
instantaneous.
As to the locality problem, this should be decided in single experiments
such that the predictions depend in a clear way on whether locality is valid or
not. Indeed, this cannot be done when combining results from diferent
experiments, because then it is apparently the existence of a JPD or (H3) which
is at issue. It is argued in Section II,F that the simplified EPR paradox [based
on the validity of locality (Hl) and of QM (H2)] could be avoided by assuming
a breakdown of the CI of the QM state-vector reduction (also suggested by
Margenau, 1936) in the case of correlated systems when only one measuring
device is present. Some candidates of experiments for verifying this are
proposed there.
We shall conclude by stating that the locality problem has not yet been
solved completely, and that further theoretical and experimental work should
be done. Perhaps it may be rewarding not to have too much confidence in the
alleged universal validity of QM, and try to invent situations where QM may
eventually get into trouble. At any rate, the claim of the proponents of the
orthodox viewpoint, that there is no basis at all for a choice between different
possibilities, seems to be too conservative and a stumbling block for future
research. This is also reminiscent of universality claims of classical physics
towards the end of the last century. Seen in this historical perspective, it is not
prudent to advance and defend extreme standpoints (Section V,A). This holds
especially with respect to the issue of locality, because other secure pillars of
modern physical theory, Einstein’s relativity theories, are founded on it. Let us
end finally then with the following careful statement of Dirac (1976): “. . . I
think it might turn out that ultimately Einstein will prove to be right, because
the present form of quantum mechanics should not be considered as the final
form. There are great difficulties.. . in connection with the present quantum
mechanics. It is the best that one can do till now. But, one should not suppose
EINSTEIN-PODOLSKY -ROSEN PARADOX 327
that it will survive indefinitely into the future. And I think that it is quite likely
that at some future time we may get an improved quantum mechanics in which
there will be a return to determinism and which will, therefore, justify the
Einstein point of view.”

ACKNOWLEDGMENTS

The author thanks Prof. T. A. Brody, Prof. F. Selleri, Dr. W. M. de Muynck, and Dr. J. Kriiger
for clarifications and/or discussions about some topics treated in the article, and the N.F.W.O.
Belgium for financial support.

REFERENCES

Aerts, D. (1982a).Example of a macroscopic classical situation that violates Bell inequalities. Lett.
Nuouo Cimento SOC.Ital. Fis. [2] 34, 107.
Aerts, D. (1982b). Description of many separated physical entities without the paradoxes
encountered in quantum mechanics. Found. Phys. 12, 1131.
Aerts, D. (1983). Classical theories and nonclassical theories as special cases of a more general
theory. J. Math. Phys. 24,2441.
Aerts, D. (1984).The missing elements of reality in the description of quantum mechanics of the
E.P.R. paradox situation. Helu. Phys. Acta 57,421.
Aharonov, Y., and Bohm, D. (1959). Significance of electromagnetic potentials in the quantum
theory. Phys. Rev. 115,485.
Albertson, J. (1961). Von Neumann’s hidden-variable proof. Am. J. Phys. 29,478.
Angelidis, T. D. (1983). Bell’s theorem: Does the Clauser-Horne inequality hold for all local
theories? Phys. Rev. Lett. 51, 1819.
Angelidis, T. D. (1984a).Angelidis responds. Phys. Rev. Lett. 53, 1022.
Angelidis, T. D. (1984b). Angelidis responds. Phys. Rev. Lett. 53, 1297.
Araki, H., and Yanase, M. M. (1960).Measurement of quantum mechanical operators. Phys. Reo.
120,622.
Aspect, A., and Grangier, P. (1985). About resonant scattering and other hypothetical effects in the
Orsay atomic-cascade experiment tests of Bell inequalities: A discussion and some new
experimental data. Lett. Nuouo Cimento Soc. ItaL Fis. [2] 43,345.
Aspect, A,, Grangier, P., and Roger, G. (1981). Experimental tests of realistic local theories via
Bell’s theorem. Phys. Rev. Lett. 47,460.
Aspect, A., Dalibard, J., and Roger, G .(1982).Experimental tests of Bell’s inequalities using time-
varying analysers. Phys. Rev. Lett. 49, 1804.
Ballentine, L. E. (1970).The statistical interpretation of quantum mechanics. Rev. Mod. Phys. 42,
358.
Ballentine, L. E. (1972). Einstein’s interpretation of quantum mechanics. Am. J. Phys. 40, 1763.
Baracca, A., Bergia, S., Bigoni, R., and Cecchini, A. (1974).Statistics of observations for “proper”
and “improper” mixtures in quantum mechanics. Riv. Nuovo Cimento SOC.Ital. Fis.[114,169.
Baracca, A., Bergia, S., Livi, R., and Restignoli, M. (1976).Reinterpretation and extension of Bell’s
inequality for multivalued observables. Int. J. Theor. Phys. 15,473.
328 W. DE BAERE

Bartell, L. S. (1980a). Complementarity in the double-slit experiments: On simple realizable


systems for observing intermediate particle-wave behavior. Phys. Rev. D 21, 1698.
Bartell, L. S. (1980b). Local realism and the Einstein-Podolsky-Rosen paradox. On concrete new
tests. Phys. Rev. D 22, 1352.
Barut, A. O., and Meystre, P. (1984). Rotational invariance, locality, and Einstein-Podolsky-
Rosen experiments. Phys. Reu. Lett. 53, 1021.
Bastide, C. (1984). The different non-localities in the Einstein-Podolsky-Rosen paradox. Phys.
Lett. A 103A, 305.
Bedford, D., and Selleri, F. (1985).On Popper’s new EPR-experiment. Lett. Nuouo Cimento SOC.
Ital. Fis. [2] 42,325.
Belinfante, F. J. (1973). “A survey of Hidden Variable Theories.” Pergamon, Oxford.
Bell, J. S. (1964).On the Einstein Podolsky Rosen paradox. Physics(Long Island City, N . Y.)1, 195.
Bell, J. S. (1966). On the problem of hidden variables in quantum mechanics. Rev. Mod. Phys. 38,
447.
Bell, J. S. (1972). Introduction to the hidden variable questions. In “Foundations of Quantum
Mechanics” (B. DEspagnat, ed.), Course 49, p. 171. Academic Press, New York.
Bergia, S., and Cannata, F. (1982). Higher-order tensors and tests of quantum mechanics. Found
Phys. 12, 843.
Berthelot, A. (1980). Comments on determinism, locality, Bell’s theorem and quantum mechanics.
Nuovo Cimento SOC.Ital. Fis. B 1111 57B, 193.
Bertolini, G., Bettoni, M., and Lazzarini, E. (1955). Angular correlation of scattered annihilation
radiation. Nuovo Cimento 2,661.
Bohm, D. (1952a).“Quantum Theory.” Prentice-Hall, Englewood Cliffs, New Jersey.
Bohm, D. (1952b). A suggested interpretation of the quantum theory in terms of “hidden”
variables. I. Phys. Rev. 85, 166.
Bohm, D. (1952~).A suggested interpretation of the quantum theory in terms of “hidden”
variables. 11. Phys. Rev. 85, 180.
Bohm, D. (1980). “Wholeness and the Implicate Order.” Routledge & Kegan Paul, London.
Bohm, D., and Aharonov, Y.(1957).Discussion of experimental proof for the paradox of Einstein,
Rosen, and Podolsky. Phys. Rev. 108,1070.
Bohm, D., and Aharonov, Y. (1960). Further discussion of possible experimental tests for the
paradox of Einstein, Podolsky and Rosen. Nuovo Cimento 17,964.
Bohm, D., and Bub, J. (1966a). A proposed solution of the measurement problem in quantum
mechanics by a hidden variable theory. Reu. Mod. Phys. 38,453.
Bohm, D., and Bub, J. (1966b).A refutation of the proof by Jauch and Piron that hidden variables
can be excluded in quantum mechanics. Rev. Mod. Phys. 38,470.
Bohm, D., and Hiley, B. J. (1981). Nonlocality in quantum theory understood in terms of
Einstein’s nonlinear field approach. Found. Phys. 11,529.
Bohr, N. (1935a). Quantum mechanics and physical reality. Nature (London) 136,65.
Bohr, N. (1935b). Can quantum-mechanical description of physical reality be considered
complete? Phys. Rev. 48,696.
Breitenberger, E. (1965). On the so-called paradox of Einstein, Podolsky and Rosen. Nuouo
Cimento [lo] 38,356.
Brody, T. A. (1985a). The Bell inequality. I. Joint measurability. Preprint. Found. Phys. (to be
published).
Brody, T. A. (1985b). The Bell inequality. 11. Locality. Preprint. Found. Phys. (to be published).
Brody, T. A,, and de la Peiia Auerbach, L. (1979). Real and imagined nonlocalities in quantum
mechanics. Nuouo Cimento SOC.Ital. Fis. B [l 13 54B,455.
Bruno, M., DAgostini, M., and Maroni, C. (1977). Measurement of linear polarization of
positron annihilation photons. Nuovo Cimento SOC. Ital. Fis. B [I 11 40B, 143.
EINSTEIN-PODOLSKY -ROSEN PARADOX 329
Bub, J. (1969). What is a hidden variable theory of quantum phenomena? Int. J. Theor. Phys. 2,
101.
Bub, J . (1973). On the possibility of a phase-space reconstruction of quantum statistics: A
refutation of the Bell-Wigner locality argument. Found. Phys. 3,29.
Busch, P. (1985). Indeterminacy relations and simultaneous measurements in quantum theory.
lnt. J . Theor. Phys. 24,63.
Cantrell, C. D., and Scully, M. 0.(1978).The EPR paradox revisited. Phys. Rep. 43,499.
Capasso, V., Fortunato, D., and Selleri, F. (1970). Von Neumann's theorem and hidden-variable
models. Riv. Nuouo Cimento SOC.Ital. Fis. [l] 11, 149.
Capasso, V., Fortunato, D., and Selleri, F. (1973). Sensitive observables of quantum mechanics.
Int. J. Theor. Phys. 7 , 319.
Caser, S . (1982).Objective local theories and the symmetry between analysers. Phys. Lett. A 92A,
13.
Caser, S. (1984a). On a possible role of the vector potential in the Einstein-Podolsky-Rosen
paradox. Phys. Lett. A 102A, 152.
Caser, S. (1984b). On a local model that violates Bell's theorem. LPTHE ORSAY&4/51, Nov.
Clauser, J. F. (1976a). Experimental investigation of a polarization correlation anomaly. Phys.
Rev. Lett. 36, 1223.
Clauser, J. F. (1976b). Measurement of the circular-polarization in photons from an atomic
cascade. Nuouo Cimento SOC.Ital. Fis. B [111 33B, 740.
Clauser, J.F., and Horne, M. A. (1974). Experimental consequences of objective local theories.
Phys. Rev. D 10,526.
Clauser, J. F., and Shimony, A. (1978). Bell's theorem: Experimental tests and implications. Rep.
Prog. Phys. 41, 1881.
Clauser, J. F., Horne, M. A., Shimony, A., and Holt, R. A. (1969).Proposed experiment to test local
hidden-variable theories. Phys. Reu. Lett. 23,880.
Corleo, G., Gutkowski, D., Masotto, G., and Valdes, M. V. (1975). Are Bell's type inequalities
sufficient conditions for local hidden-variable theories? Nuouo Cimento SOC.Ital. Fis. B 25B,
413.
Costa de Beauregard, 0. (1976). Time symmetry and interpretation of quantum mechanics.
Found. Phys. 6,539.
Costa de Beauregard, 0. (1977).Time symmetry and the Einstein paradox. Nuouo Cimento SOC.
Ital. Fis. B [ l l ] 42B, 41.
Costa de Beauregard, 0.(1979).Time symmetry and the Einstein paradox. 11. Nuouo Cimento SOC.
Ital. Fis. B [111 51B, 267.
Cramer, J. G. (1980). Generalized absorber theory and the Einstein-Podolsky-Rosen paradox.
Phys. Rev. D 22,362.
Cufaro Petroni, N. (1977a).On the observable differences between proper and improper mixtures.
I. Nuouo Cimento SOC.Ital. Fis. B 1111 40B, 235.
Cufaro Petroni, N. (1977b).On the observabledifferences between proper and improper mixtures.
11. Nuouo Cimento SOC.Ital. Fis. B [ll] 40B, 381.
Cushing, J. T. (1985). Comments on Angelidis's universality claim. Phys. Rev. Lett. 54,2059.
De Baere, W. (1984a). On the significance of Bell's inequality for hidden-variable theories. Lett.
Nuouo Cimento SOC.Ital. Fis. [2] 39,234.
De Baere, W. (1984b). On conditional Bell inequalities and quantum mechanics. Lett. Nuouo
Cimento SOC.Ital. Fis. [2] 40,488.
De Baere, W. (1984~).Evidence for nonreproducibility of physical situations at the hidden
variable level as the origin of the violation of Bell's inequality for the singlet state.
Unpublished, University of Ghent.
De Baere, W. (1985). EPR paradox considered as a criticism of the state vector reduction in the
330 W. DE BAERE
quantum mechanical description of correlated separated systems. In “Proceedings of the
International Conference on Microphysical Reality and Quantum Formalism.” Reidel,
Dordrecht, Netherlands. (to be published).
de la Peiia, L., Cetto, A. M., and Brody, T. A. (1972). On hidden-variable theories and Bell’s
inequality. Lett. Nuouo Cimento SOC.Ital. Fis. [2] 5, 177.
de Muynck, W. M. (1983). Objectivity, nonlocality and the Bell inequalities. Phys. Lett. A 94A,
13.
de Muynck, W. M. (1985). On the relation between the Einstein-Podolsky-Rosen paradox and
the problem of nonlocality in quantum mechanics. Preprint, Eindhoven University of
Technology.
de Muynck, W. M., and Abu-Zeid, 0. (1984). On an alternative interpretation of the Bell
inequalities. Phys. Lett A IOOA, 485.
de Muynck, W. M., Janssen, P. A. E. M., and Santman, A. (1979).Simultaneous measurement and
joint probability distributions in quantum mechanics. Found. Phys. 9,71.
de Muynck, W. M. (1986). The Bell inequalities and their irrelevance to the problem of locality in
quantum mechanics. Phys. Lett. A 114A, 65.
DEspagnat B. (1965). “Conceptions de la Physique Contemporaine.” Hermann, Paris.
DEspagnat, B. (1975). Use of inequalities for the experimental test of a general conception of the
foundation of microphysics. Phys. Reo. D 11, 1424.
D’Espagnat, B. (1983). “In Search of Reality.” Springer-Verlag, Berlin and New York.
DEspagnat, B. (1984). Nonseparability and the tentative descriptions of reality. Phys Rep. 110,
201.
Destouches, J. L. (1980). Usingde Broglie’sphysical waves to solve the Einstein-Podolsky-Rosen
paradox. Nuouo Cimento SOC.Ital. Fis. B [113 58B, 73.
Dirac, P. A. M. (1976). In “Directions in Physics” (H. Hora and J. R. Shepansky, eds.).
Sidney.
Eberhard, P. H. (1977). Bell’s theorem without hidden variables. Nuouo Cimento SOC.Ital. Fis. B
[11] 38B, 75.
Eberhard, P. H. (1978). Bell’s theorem and the different concepts of locality. Nuouo Cimento SOC.
Ifal. Fis. B Ell] 46B, 392.
Eberhard, P. H. (1982). Constraints of determinism and of Bell’s inequalities are not equivalent.
Phys. Rev. Lett. 49, 1474.
Edmonds, A. R. (1957). “Angular Momentum in Quantum Mechanics.” Princeton Univ. Press,
Princeton, New Jersey.
Edwards, J. (1975). Sufficient conditions for objective local theories. Nuouo Cimento SOC.Ital. Fis.
B [11] 29B, 100.
Edwards, J., and Ballentine, L. (1976). Sufficient conditions for objective local theories. 11. Nuovo
Cimento SOC.Ital. Fis. B [11] MB, 91.
Einstein, A. (1949). Autobiographical notes. In “Albert Einstein: Philosopher-Scientist’’ (P. A.
Schilpp. ed.), p. 85. Tudor Publishing Company, New York.
Einstein, A,, Podolsky, B., and Rosen, N. (1935).Can quantum-mechanical description of physical
reality be considered complete? Phys. Rev. 41, 777.
Faraci, G., Gutkowski, D., Notarrigo, S., and Pennisi, A. R. (1974). An experimental test of the
EPR paradox. Lett. Nuouo Cimento SOC.Ital. Fis. [2] 9,607.
Feingold, S. J., and Peres, A. (1980). How often are Bell’s inequality premises violated? J . Phys.
Math Gen. A 13, 3187.
Fenyes, I. (1952). Eine warscheinlichkeitstheoretische Begrundung und Interpretation der
Quantenmechanik. Z. Phys. 132, 81.
Feynman, R. P., Leighton, R. B., and Sands, M. (1965).“The Feynman Lectures on Physics,” Vol.
111. Addison-Wesley, Reading, Massachusetts.
EINSTEIN-PODOLSKY-ROSEN PARADOX 331
Fine, A. (1982a). Hidden variables, joint probability, and the Bell inequalities. Phys. Rev. Lett. 48,
291.
Fine, A. (1982b). Joint distributions, quantum correlations and commuting observables. J . Math.
Phys. 23, 1306.
Fine, A. (1982~).Comments on the significance of Bell’s theorem. Phys. Reo. Lett. 49, 1536.
Flato, M., Piron, C., Grea, J., Sternheimer, D., and Vigier, J. P. (1975). Are Bell’s inequalities
concerning hidden variables really conclusive? Helv. Phys. Acta 48, 219.
Fortunato, D., Garuccio, A,, and Selleri, F. (1977). Observable consequences from second-type
state vectors of quantum mechanics. Int. J . Theor. Phys. 16, 1.
Fox, R. (1971). Low-energy proton-proton scattering as a test of local hidden-variable theory.
Lett. Nuovo Cimento Soc. ftal. Fis. [2] 2, 565.
Franson, J. D. (1985). Bell’s theorem and delayed determinism. Phys. Rev. D 31,2529.
Freedman, S. J., and Clauser, J. F. (1972). Experimental test of local hidden-variable theories.
Phys. Rev. Lett. 28, 938.
Freedman, S. J., and Wigner, E. (1973). On Bub’s misunderstanding of Bell’s locality argument.
Found. Phys. 3,457.
Froissart, M. (1981). Constructive generalization of Bell’s inequalities. Nuooo Cimento Soc. Ital.
Fis. B 64B,241.
Fry, E. D., and Thompson, R. C. (1976). Experimental test of local hidden-variable theories. Phys.
Rev. Lett. 31,465.
Furry, W. H. (1936a). Note on the quantum-mechanical theory of measurement. Phys. Rev. 49,
393.
Furry, W. H. (1936b). Remarks on measurements in quantum theory. Phys. Rev. 49,476.
Garg, A,, and Legett, A. J. (1984). Comment on “Bell’s theorem: Does the Clauser-Home
inequality hold for all local theories?” Phys. Reo. Lett. 53, 1297.
Garg, A,, and Mermin, N. D. (1982). Bell inequalities with a range of violation that does not
diminish as the spin becomes arbitrary large. Phys. Rev. Lett. 49,901.
Garg, A., and Mermin, N. D. (1985). Farka’s lemma and the nature of reality: Statistical
implications of quantum correlations. Found. Phys. 14, 1.
Garuccio, A. (1978). Generalized inequalities following from Einstein locality. Lett. Nuooo
Cimento Soc. Ital. Fis. [2] 23, 559.
Garuccio, A,, and Rapisarda, V. A. (1981). A comparison between deterministic and probabilistic
local hidden-variable theories. Lett. Nuovo Cimento Soc. Ital. Fis. [2] 30,443.
Garuccio, A,, and Selleri, F. (1978). On the equivalence of deterministic and probabilistic local
theories. Lett. Nuooo Cimento Soc. Ital. Fis. [2] 23, 555.
Garuccio, A., and Selleri, F. (1980). Systematic derivation of all the inequalities of Einstein
locality. Found. Phys. 10,209.
Garuccio, A., and Selleri, F. (1984). Enhanced photon detection in EPR type experiments. Phys.
L e f t . A 103, 99.
Ghirardi, G. C., and Weber, T. (1983). Quantum mechanics and faster-than-light communica-
tions: Methodological considerations. Found. Phys. 12, 1 1 71.
Gleason, A. M. (1957).Measures on the closed subspaces of a Hilbert space. J . Math. Mech. 6,885.
Gudder, S. P. (1984). Reality, locality, and probability. Found Phys. 14,997.
Gutkowski, D., and Masotto, G. (1974). An inequality stronger than Bell’s inequality. Nuovo
Cimento Soc. Ital. Fis. B [ l l ] 22B, 121.
Gutkowski, D., Masotto, G., and Valdes, M. V. (1979). On the sufficiency of Bell’s conditions.
Nuooo Cimento Soc. Ital. Fis. B [ l l ] SOB, 323.
Halbwachs, F. Piperno, F., and Vigier, J. P. (1982). Relativistic Hamiltonian description of the
classical photon behavior: A basis to interpret aspect’s experiments. Lett. Nuooo Cimento Soc.
Ital. Fis. [Z] 33, 311.
332 W. DE BAERE

Hall, J., Kim, C., McElroy, B., and Shimony, A. (1977).,Wave-packet reduction as a medium of
communication. Found. Phys. 7,759.
Herbert, N. (1982). FLASH-A superluminal communicator based upon a new kind of quantum
measurement. Found. Phys. 12,1171.
Herbert, N., Karush, J. (1978). Generalization of Bell’s theorem. Found. Phys. 8,313.
Holt, R. A., and Pipkin, F. M. (1973). Harvard University preprint.
Home, D., and Sengupta, S. (1984). Bell’s inequality and noncontextual dispersion-free states.
Phys. Lett. A 102A, 159.
Hooker, C. A. (1972). The nature of quantum mechanical reality: Einstein versus Bohr. In
“Paradigms and Paradoxes”(R. G. Colodny, ed.), p. 67. Univ. of Pittsburgh Press, Pittsburgh,
Pennsylvania.
Home, M. A., and Shimony, A. (1984). Comment on “Bell’s theorem: Does the Clauser-Horne
inequality hold for all local theories?” Phys. Rev. Lett. 53, 1296.
Jammer, M. (1974). “The Philosophy of Quantum Mechanics.” Wiley, New York.
Jauch, J. M. (1968). “Foundations of Quantum Mechanics.” Addison-Wesley, Reading,
Massachusetts.
Jauch, J. M., and Piron, C. (1963).Can hidden variables be excluded in quantum mechanics? Helv.
Phys. Acta 36,827.
Kasday, L. R. (1971). Experimental test of quantum predictions for widely separated photons. In
“Proceedings of the International School of Physics ‘Enrico Fermi”’ (B. d‘Espagna1, ed.),
p. 195. Academic Press, New York.
Kasday, L. R., Ullman, J. D., and Wu, C. S. (1970). Bull. Am. Phys. SOC.[2] 15,586.
Kasday, L. R., Ullman, J. D., and Wu, C. S. (1975). Angular correlation of Compton-scattering
annihilation photons and hidden variables. Nuovo Cimento SOC.Ital. Fis. B [ll] 25B, 633.
Kellett, B. H. (1977). The physics of the Einstein-Podolsky-Rosen paradox. Found. Phys. 7,735.
Kemble, E. C. (1935). The correlation of wave functions with the states of physical systems. Phys.
Rev. 47,973.
Kochen, S., and Specker, E. P. (1967). The problem of hidden variables in quantum mechanics. J .
Math. Mech. 17,59.
Kocher, C. A,, and Commins, E. D. (1967). Polarization correlation of photons emitted in an
atomic cascade. Phys. Rev. Lett. 18, 575.
Kraus, K. (1985a). Einstein-Podolsky-Rosen experiments and macroscopic locality. In “Open
Questions in Quantum Physics” (G. Tarozzi and A. van der Merwe, eds.), p. 75. Reidel,
Dordrecht, Netherlands.
Kraus, K. (1985b). “Quantum Theory, Causality, and EPR Experiments,” Preprint. Center for
Particle Theory, University of Texas, Austin.
Lamehi-Rahti, M., and Mittig, W. (1976). Quantum mechanics and hidden-variables: A test of
Bell’s inequality by the measurement of the spin correlation in low-energy proton-proton
scattering. Phys. Rev. D 14,2543.
Langhoff, H. (1960). Die Linearpolarisation der Vernichtungs strahlung von Positronen. Z. Phys.
160,186.
Liddy, D. E. (1983). On locality, correlation and hidden variables. J . Phys. A: Math. Gen. 16,
2703.
Liddy, D. E. (1984). An objective, local, hidden-variables theory of the Clauser-Horne
experiment. J . Phys. A: Math. Gen. 17,847.
Lochak, G. (1976). Has Bell’s inequality a general meaning for hidden-variable theories? Found.
Phys. 6, 173.
Macdonald, A. I. (1982). Comment on “Resolution of the EPR and Bell Paradoxes.” Phys. Rev.
Lett. 49, 1215.
EINSTEIN-PODOLSKY -ROSEN PARADOX 333
Margenau, H. (1936). Quantum-mechanical description. Phys. Rev. 49,240.
Lehr, W. J. (1985). On the hypothesis that quantum mechanics has a nonlocal structure. Nuovo
Cimento Soc. Ital. Fis. B WB, 185.
Marshall, T. W. (1983).The distance separating quantum theory from reality. Phys. Lett. A 99A,
163.
Marshall, T. W. (1984).Testing for reality with atomic cascades. Phys. Lett. A 100A, 225.
Marshall, T. W., and Santos, E. (1985). Local realist model for the coincidence rates in atomic-
cascade experiments. Phys. Lett. A 107A, 164.
Marshall, T. W., Santos, E., and Selleri, F. (1983a). Collective effects in the atomic-cascade
experimental tests of Bell inequalities. Lett. Nuovo Cimento SOC.Ital. Fis. [2] 38, 417.
Marshall, T. W., Santos, E., and Selleri, F. (1983b).Local Realism has not been refuted by atomic
cascade experiments. Phys. Lett. A 98A, 5.
Mermin, N. D. (1980). Quantum mechanics vs local realism near the classical limit: A Bell
inequality for spin s. Phys. Rev. D 22,356.
Mermin, N. D. (1982).Comment on “Resolution of the EPR and Bell Paradoxes.” Phys. Rev. Lett.
49, 1214.
Mermin, N. D., and Schwarz, G. M. (1982).Joint distributions and local realism in the higher-spin
Einstein-Podolsky-Rosen experiment. Found. Phys. 12,101.
Moldauer, P. 0. (1974). Reexamination of the arguments of Einstein, Podolsky, and Rosen.
Found. Phys. 4, 195.
Miickenheim, W. (1982). A resolution of the Einstein-Podolsky-Rosen paradox. Lett Nuouo
Cimento Soc. Ital. Fis. [2] 35,300.
Park, J. L., and Margenau, H. (1968). Simultaneous measurability in quantum theory. Int. J.
Theor. Phys. 1,211.
Pascazio, S . (1985). Experimental tests of Bell inequalities. Are all local models really excluded?
Phys. Lett. A 111A, 339.
Paty, M. (1977). The recent attempts to verify quantum mechanics. In “Quantum Mechanics, a
Half Century Later” (J. Leite Lopes and M. Paty, eds.), p. 261. Reidel, Dordrecht,
Netherlands.
Pearle, P. M. (1970). Hidden-variable example based upon data rejection. Phys. Rev. D 2,
1418.
Peres, A. (1978). Unperformed experiments have no results. Am. J . Phys. 46,745.
Peres, A. and Singer, P. (1960). On possible experimental tests for the paradox of Einstein,
Podolsky and Rosen. Nuouo Cimento 15,907.
Perrie, W., Duncan, A. J., Beyer, H. J., and Kleinpoppen, H. (1985). Polarization correlation of
the two photons emitted by metastable atomic deuterium: A test of Bell’s inequality. Phys.
Rev. Lett. 54, 1790.
Pipkin, F. M. (1978).Atomic physics tests of the basic concepts in quantum mechanics. Adu. At.
Mol. Phys. 14,281.
Pitowski, 1. (1982a).Resolution of the Einstein-Podolsky-Rosen and Bell paradoxes. Phys. Rev.
Lett. 48, 1299.
Pitowski, I. (1982b).Pitowski responds. Phys. Rev. Lett. 49, 1216.
Pitowski, I. (1983).Deterministic model of spin and statistics. Phys. Rev. D 27, 2316.
Popper, K. (1985).Realism in quantum mechanics and a new version of the EPR experiment. In
“Open Questions in Quantum Physics” (G. Tarozzi and A. van der Menve, eds.), p. 3. Reidel,
Dordrecht, Netherlands.
Rastall, P. (1981).Quantum mechanics, locality and completeness. Nuovo Cimento SOC. Ital. Fis. B
1113 62B, 175.
Rastall, P. (1983).The Bell inequalities. Found. Phys. 13, 555.
334 W. DE BAERE

Rastall, P. (1985). Locality, Bell’s theorem, and quantum mechanics. Found. Phys. 15,963.
Reeh, H., and Schlieder, S. (1961). Bemerkungen zur Unitaritatsaquivalenz von Lorentz-
invarianten Feldern. Nuouo Cimento 22, 1051.
Renninger, M. (1960). Messungen ohne Storung des Messobjects. 2. Phys. 158,417.
Rietdijk, C. W. (1978). Proof of a retroactive influence. Found. Phys. 8,615.
Rietdijk, C. W. (1981). Another proof that the future can influence the present. Found. Phys. 11,
783.
Rietdijk, C. W. (1985). On Nonlocal Influences. In “Open Questions in Quantum Physics”
(G. Tarozzi and A. van der Merwe, eds.), p. 129. Reidel, Dordrecht, Netherlands.
Rietdijk, C. W., and Selleri, F. (1985). Proof of a quantum mechanical nonlocal influence. Found.
Phys. 15, 303.
Rohrlich, F. (1983). Facing quantum mechanics reality. Science 221, 1251.
Roos, M. (1980). Tests of Einstein locality. HU-TFT-80-5, Helsinki preprint.
Rose, M. E. (1957). “Elementary Theory of Angular Momentum.” Wiley, New York.
Ruark, A. E. (1935). Is the quantum-mechanical description of physical reality complete? Phys.
Rev. Lett. 48,466.
Sachs, M. (1980). Bell’s inequalities from the field concept in general relativity. Nuouo Cimento Soc.
Ital. Fis. A 58A, 1.
Santos, E. (1984). Microscopic and macroscopic Bell inequalities. Phys. Lett. A 101A, 379.
Scalera, G. C. (1983). On a local hidden-variable model with unusual properties. Lett. Nuouo
Cimento SOC.Ital. Fis. [2] 38, 16.
Scheibe, E. (1973). “The Logic Analysis of Quantum Mechanics.” Pergamon, Oxford.
Selleri, F. (1972).A stronger form of Bell’s inequality. Lett. Nuouo Cimento SOC.Ital. Fis. [ 2 ] 3,
581.
Selleri, F. (1978).On the consequences of Einstein locality. Found. Phys. 8, 103.
Selleri, F. (1980). Quantum correlations and separability. Lett. Nuouo Cimento SOC.Ital. Fis. [2]
21, 1.
Selleri, F. (1982). Generalized EPR-paradox. Found. Phys. 12,645.
Selleri, F. (1983a). Einstein locality and the K ° K o system. Lett. Nuouo Cimento SOC.Ital. Fis. [2]
36,521.
Selleri, F. (1983b). “Die Debatte um die Quantentheorie.” Friedrich Vieweg & Sohn,
Braunschweig/Wiesbaden.
Selleri,F. (1984). Photon scattering with Doppler-shift in the atomic-cascade experimental tests of
Bell inequalities. Lett. Nuouo Cimento SOC.Ital. Fis. [2] 39,252.
Selleri, F. (1985). Local realistic photon models and EPR-type experiments. Phys. Lett. A 108A,
197.
Selleri, F., and Tarozzi, G. (1980). Is Clauser and Horne’s factorability a necessary requirement for
a probabilistic local theory? Lett. Nuouo Cimento SOC.Ital. Fis. [2] 29, 533.
Selleri, F., and Tarozzi, G. (1981). Quantum mechanics reality and separability. Riu. Nuouo
Cimento SOC.Ital. Fis. [l] 4, 1.
She, C. Y., and Heffner, H. (1966). Simultaneous measurement of noncommutable observables.
Phys. Rev. 152, 1103.
Six, J. (1984). Test of the non separability of the K ° K o system. In “The Wave-Particle Dualism’’
(S. Diner, D. Fargue, G. Locak, and F. Selleri, eds.), p. 391. Reidel, Dordrecht, Netherlands.
Snyder, H. S., Pasternak, S., and Hornbostel, J. (1948). Angular correlation of scattered
annihilation radiation. Phys. Rev. 73,440.
Stapp, H. P. (1971). S-matrix interpretation of quantum theory. Phys. Rev. D 3, 1303.
Stapp, H. P. (1972). The Copenhagen interpretation. Am. J . Phys. 40, 1098.
Stapp, H. P. (1975). Bell’s theorem and world process. Nuouo Cimento SOC.Ital. Fis. B [I I] 29B,
270.
EINSTEIN-PODOLSKY PROSEN PARADOX 335
Stapp, H. P. (1977).Are superluminal connections necessary? Nuooo Cimento SOC.Ital. Fis. B [I 11
4OB, 191.
Stapp, H. P. (1979). Whiteheadian approach to quantum theory and the generalized Bell’s
theorem. Found. Phys. 9, 1.
Stapp, H. P. (1982a).Bell’s theorem as a nonlocality property of quantum theory. Phys. Rev. Lett.
49, 1470.
Stapp, H. P. (1982b). Mind, matter, and quantum mechanics. Found. Phys. 12,363.
Stapp, H. P. (1985).Einstein locality, EPR locality, and the significancefor science of the non-local
character of quantum theory. LBL-20094. In “Proceedings of the International Conference
on Microphysical Reality and Quantum Formalism.” Reidel, Dordrecht, Netherlands (to be
published).
Streater, R. F., and Wightman, A. S. (1964). “PCT, Spin and Statistics, and All That.” Benjamin,
New York.
Summers, S. J., and Werner, R. (1985).The vacuum violates Bell’s inequalities. Phys. Lett. A llOA,
257.
Sutherland, R. I. (1983). Bell’s theorem and backwards-in-time causality. Int. J. Theor. Phys. 22,
377.
Sutherland, R. I. (1985).A corrolary to Bell’s theorem. Nuouo Cimento SOC.Ital. Fis. B [ l l ] 88B,
144.
Todorov, N. S. (1985a).On Bell’s argument. Ann. Fond. L. de Broglie 10,49.
Todorov, N. S. (1985b). The Einstein-Podolsky-Rosen and Bell arguments revisited. I. On the
validity of certain critiques against the current interpretation of Bell’s argument. Ann. Fond.
L . de Broglie 10,207.
Todorov, N. S. (198%). The Einstein-Podolsky-Rosen and Bell arguments revisited. 11.
Restrictive character of the entire system of Bell’s axioms and unfalcifiability of the EPR
standpoint. Ann. Fond. L. de Broglie 10, No. 4.
Tornqvist, N. A. (1981). Suggestion for Einstein-Podolsky-Rosen experiments using reactions
like efe- + AA + n-pn-p. Found. Phys. 11, 171.
Vigier, J. P. (1979). Superluminal propagation of the quantum potential in the causal
interpretation of quantum mechanics. Lett. Nuouo Cimento SOC. Ital. Fis. [2] 24,258.
von Neumann, J. (1955). “Mathematical Foundations of Quantum Mechanics.” Princeton Univ.
Press, Princeton, New Jersey.
Weizel, W. (1953a).Ableitung der Quantentheorie aus einem klassischen, kausal determinierten
Modell. Z. Phys. 134,264.
Weizel, W. (1953b). Ableitung der Quantentheorie aus einem klassischen, kausal determinierten
Modell. 11. Z. Phys. 135, 270.
Wheeler, J. A. (1978). The “past” and the “delayed-choice” double-slit experiment. In
“Mathematical Foundations of Quantum Theory” (A.R. Marlow, ed.), p. 9. Academic
Press, New York.
Wheeler, J. A,, and Feynman, R. P. (1949). Classical electrodynamics in terms of direct
interparticle action. Reo. Mod. Phys. 21,425.
Whitaker, M. A. B., and Singh, I. (1982).Interpretations of quantum mechanics and some claimed
resolutions of the EPR paradox. J. Phys. A : Math. Gen. 15,2377.
Wigner, E. P. (1952).Die Messung quantenmechanischer Operatoren. Z. Phys. 133, 101.
Wigner, E. P. (1970).On hidden variables and quantum mechanical probabilities. Am. J . Phys. 38,
1005.
Wilson, A. R., Lowe, J., and Butt, D. K. (1976).Measurement of the relative planes of polarization
of annihilation quanta as a function of separation distance. J. Phys. G 2,613.
Wolfe, H. C. (1936). Quantum mechanics and physical reality. Phys. Reu. 49, 274.
Wootters, W. K., and Zurek, W. H. (1979). Complementarity in the double-slit experiment:
336 W.DE BAERE

Quantum nonseparabilityand a quantitative statement of Bohr’s principle. Phys. Rev. D 19,


473.
Wu, C. S., and Shaknov, I. (1950). The angular correlation of scattered annihilation radiation.
Phys. Rev. 71, 136.
Yanase, M. M. (1961). Optimal measuring apparatus.Phys. Rev. 123,666.
Zweifel, P. F. (1974). Measurement in quantum mechanics and the EPR paradox. Int. J . Theor.
Phys. 10,67.
ADVANCES IN ELECTRONICS AND ELECTRON PHYSICS,

Theory of Electron Mirrors and Cathode Lenses


E. M. YAKUSHEV AND L. M. SEKUNOVA
Institute of Nuclear Physics
Academy of Sciences of Kazakh SSR
SU-480 082 Alma-Ata, USSR

I. INTRODUCTION

Electron mirrors and emission systems are used in various electron-


optical devices: electron mirror microscopes (Oman, 1969; Luk’yanov
et al., 1973), cathode ray tubes (Zworykin and Morton, 1954; Zhigarev
and Shamaeva, 1982), image converters (Artamonov et al., 1968; Butslov and
Stepanov, 1978),brightness amplifiers (Kasper and Wilska, 1967),and so on.
In practice it is of a great interest to make use of a mirror for correction
of spherical and chromatic aberrations of electron lenses (Zworykin et al.,
1945; Septier, 1966; Slowko, 1975). Application of a mirror in a multipassage
mass spectrometer (Fortin and Baril, 1972; Noda, 1963) is worthy of noting.
A mirror is applied for monochromatization of electron beams, e.g., in an
ion microanalyzer (Leleyter and Slodzian, 1971), in an ion detector of the
mass spectrometer (Vasil’ev et al., 1974), and in an emission microscope
(Vorob’ev, 1959; Gruner et al., 1971).
Recently a new field of ion mirrors application comes to light, namely,
nonmagnetic time-of-flight mass spectrometry (Mamyrin et al., 1973;
Wechsung et al., 1978; Daumenov et al., 1981; Gohl et al., 1983; Berger,
1983). A peculiarity of this field is the use of a substantially nonstationary
flux of charged particles-an extremely narrow (in the axial direction) ion
packet, which, while traveling from the source to the detector, splits up, due
to the differences in velocities of ions having different masses, into several
fragments. Herewith the mirror is to perform spatial focusing of these frag-
ments in the radial and, primarily, in the axial directions. It should be noted
that one has also to deal with the nonstationary flux problem in ordinary
emission and scanning microscopes and image converters if one investigates
the processes the duration of which becomes comparable with the time of
particle flight in the system considered.

337
Copynght @ 1986 by Academic Press, Inc
All nghts of reproduction In any form reserved
338 E. M. YAKUSHEV A N D L. M. SEKUNOVA

When stating the general theory of electron mirrors and emission sys-
tems, we shall take into account the exclusive variety of physical situations
where these elements are used. We shall also consider some concrete applica-
tions of the theory.

11. EQUATIONS
OF MOTION

It is characteristic for both the electron mirror and the cathode lens that
in the space occupied by the electric magnetic field and the particle flux there
is, as a rule, a rather narrow specific region where the particles' velocity is
close to zero, and the electric field is quite different from zero. In electron
mirrors particles are reversed in this region. In cathode lenses one of the
equipotential surfaces of this region is the cathode surface emitting particles.
Herewith, if in mirrors the field is analytic everywhere including the neigh-
borhood of the specific region, in cathode lenses the field has a discontinuity
on the cathode surface. However, for theoretical study of electron mirrors and
cathode lenses this difference is not substantial, if the field on the internal
side of the cathode surface is assumed to be the analytical continuation of
the external field. It provides a possibility of studying properties of the
appropriate cathode lens as well when investigating particle motion in the
analytical field of the mirror and choosing one of its equipotentials as an
emitting surface.
In a cylindrical coordinate system, r, I), z , where the z axis coincides with
the symmetry axis of the field, the electric field is described by a scalar po-
tential cp = cp(r,z), and the magnetic one by a vector potential A having
components A , = A , = 0, A , = A,@, z). The equations of motion in this field
for a particle having charge e and mass m are as follows:

2e
i 2 + + r2$2
i 2 = --('
m
+ E) (2)

mr2$ + erA, = -eC, (3)


where C , is a constant, and E is the total energy of a charged particle expressed
in terms of a potential. The potential cp is normalized in such a way that it
equals zero at the place in space where the velocity of a chosen particle equals
zero. Then the energy of any other particle in this place, due to the spread of
particle velocities, may differ by E.
ELECTRON MIRRORS A N D CATHODE LENSES 339
The quantity A , is taken to be equal to zero on the symmetry axis of the
system.
Using the complex variable
r = reio (4)
where

and t is the time, we can rewrite the set of equations (1)-(3) as follows:

-_
1
i 2+ cp = lrr*
1 .. -E + -(2C,
k2 + A,r)TA , (7)
k2 k 4
where
k = J q
C, = (i/k2)(ri* - r * i )

and an asterisk (*) denotes the complex conjugate.


Equations (6) and (7), together with Eqs. (5) and (9) describe the motion of
a charged particle in a stationary axially symmetric field. Here Eq. (7)
represents energy conservation for a particle, whereas Eq. (9) is the con-
servation law for the azimuthal component of its generalized momentum.
With the methodology of the static electron-optic system investigation
developed by Scherzer (1937), Glaser (1952), Sturrock (1955), the law of energy
conservation for a particle is used in order to eliminate time of flight from the
equations of motion and to introduce the coordinate of the main optical axis z
as an independent variable. This transformation implies certain assumptions.
First of all, the particle flux is assumed to be stationary, and the time-of-flight
characteristics of the electron-optical system are eliminated from further
consideration. The other assum tions are the results of the fact that the
structures like d G and*J (the prime denotes differentiation with
respect to z ) appear in the equations. Because of the presence of these
structures, when linearizing the equations, one has to adopt both the
requirement of smallness r, and the following conditions of the paraxial
approximation
E/(P << 1, Ir'l << 1 (10)
In doing so, mirror and emission systems are not considered, because the
340 E. M. YAKUSHEV AND L. M. SEKUNOVA

inequalities (10) can not be satisfied for them in the neighborhoods of the
reversal points of electron trajectories or in the near-cathode region.
When constructing the theory of electron-optical mirrors Recknagel
(1937) used Eqs. (6) and (7) without eliminating time of flight. His method
assumes solely small deviations of electrons from the axis, allowing whatever
large angles of electron trajectory slopes in the neighborhood of a reversal
point, where particle energies are close to zero. Generalizing the concept of
paraxial approximation in such a way, he obtained the equations giving the
first approximation, then found the expression for the correction term and
showed that it was third order with respect to the first approximation.
The method by Recknagel was then developed further by Zworykin et al.
(1945) and Kas'yankov (1966). They derived the final formulas which had
complicated forms and required rather serious computational difficulties to be
overcome. The insufficiency of the theory sets one seeking for new ways of
mirror study. Bernard (1952), in particular, proposed using v = & as an
independent variable. Hahn (1971) used the azimuthal angle characterizing
the particle output from the Larmor coordinate system as such a variable.
Slowko and Mulak (1973) used the model representation according to which
particle reflection takes place directly in the zero equipotential of the field
obeying the law of incident and reflected angle equality, and the entire field of
the mirror represents an ordinary electron lens. However, one could not
manage to obtain the finished theory of electron mirrors as complete and
accessible for practical applications as the theory of electron lenses. Up to now
necessary practical calculations have been carried out, mainly, without using
aberration representations-either via direct integration of the equations of
motion (e.g., Regenstreif, 1947), or using the transformation proposed by
Recknagel (1937) allowing one to overcome computational difficulties
(Lafferty, 1947; Berger and B a d , 1978).
In other papers (Kel'man et al., 1971, 1972, 1973a,b, 1974a,b; Sekunova
and Yakushev, 1975a,b; Sekunova, 1977; Daumenov et al., 1981; Bimurzaev
et al., 1983a,b) a new method of theoretical study of electron mirrors allows
one to overcome analytically the above-mentioned mathematical difficulties
and to obtain final expressions for aberration coefficients up to third order
in the forms of dependences on the principal optical axis coordinates.

A . Trajectory Equations

Consider two synchronously moving particles with equal specific charges,


e/m, with their initial conditions somewhat different (Fig. 1). One of these
particles, 1, moving along the axis z with E = 0, will be called the guiding
particle. Its position at the given time instant t will be denoted by (.Then from
ELECTRON MIRRORS AND CATHODE LENSES 341

FIG. 1. Synchronous motion of (1) guiding and (2) arbitrary particles in the field of an
electron mirror.

Eq. (7) the precise equality follows:


i=a k r n (11)
where +(z) = cp(0,z)is the potential distribution along the mirror axis, is +(c)
c,
the same function of the argument and a is the sign factor characterizing the
direction of the particle's travel along the z axis. The position of the other
particle, 2, at the same time instant twill be characterized by the radius vector,
p = q + r, drawn to it from the guiding particle. Then the equality takes place:

z=c+q (12)
Any particle 2 can be associated with a great number of particles, each
moving along the axis with E = 0. As a guiding particle choose the particle 1
from the set for which the identified equality holds:
fJ = i/lil (13)
where i is the axial component of the velocity for particle 2. This condition
implies that synchronization of motion of these particles is produced over the
time instant of their reflection.
Using Eqs. (11) and (12), we introduce a new independent variable and a c
dynamical variable q into Eqs. (6) and (7), eliminating from them t and z. It
implies introducing the movable coordinate system, r, $, u. associated with the
guiding particle, the time being measured by the distance covered by the
guiding particle along the z axis. On performing these transformations, we
obtain
342 E. M. YAKUSHEV AND L. M. SEKUNOVA

Herein

where B ( z ) is the distribution function of the magnetic field induction along


the principal optical axis. B ( i ) is the same function of the argument i

The concept of a guiding particle implies that y equals zero for it. It means
that there is a particular solution of Eq. (15) y = y(i), and its magnitude is
small for small E and r. We shall use just this solution. As no limitation is used
when deriving Eqs. (14) and (15), they are valid for any values of r and E. We
shall seek the solution of these equations for the region close to the principal
axis of the mirror and for small values of the energy spread. Equations (14) and
(15) are put down in such a form that admits a solution by the method of
successive approximations under specified conditions. The fact that these
equations are derived without limitations for the trajectory slope does not
mean that the limitation is removed altogether. As r is required to be small,
large slopes may be allowed only at small sections of the trajectories, in
particular, in neighborhood of the reversal points. We shall assume that,
excluding these sections, everywhere, and also beyond the field, the trajectory
slope with respect to the principal optical axis is the value of the order of
smallness such as r.
To define geometrical aberrations of third order and chromatic aberra-
tions of order rc, it is necessary to preserve in the trajectory equations terms of
order not higher than third in r and of first order in rc. We shall stick to this
requirement further on. Using the known representation of the field potentials
near the symmetry axis of the system
ELECTRON MIRRORS AND CATHODE LENSES 343
and the expansion

we reduce Eqs. (14) and (15) to the form

with

Here

The primes mean differentiation with respect to the independent variable i.


The right-hand side of Eq. (23) is of third order in r and of first order in rE, as
the constant C , is the value of second order in r; q is of second order in r and of
first order in E. The estimation of the order of C 2 follows from Eq. (18) if one
takes into account that lr’l is small far from the reversal points. The fact that lr’l
increases without limit at reversal points does not contradict the smallness of
C , , as q5(i)= 0 in this point.
To substantiate the order of r], we consider Eq. (24) in more detail. It is
necessary to choose such a particular solution of this equation for the value of
r] which equals zero for the guiding particle. So, due to the linearity the
equation under consideration, the order of r] will be the same as the order of its
right-hand side. This part may be reduced to the following:

s(3) =
- 4arbrt‘ - aQ,,r,’ - a Q;r2 d i - 3kBC2 + E (29)

because in the approximation considered


(4r’r*’ + a Q 2 r 2 ) ’= $Q;r2
Here the subscript a refers to values associated with the point which is far from
the reversal region. From Eq. (29) it follows that the right-hand side of Eq. (24)
remains everywhere the value of second order in r and of first one in E.
344 E. M. YAKUSHEV AND L. M. SEKUNOVA

Thus, the smallness of the quantities S'3) and d3)necessary for application
of the method of successive approximations is guaranteed either by the
smallness of 4 which is valid in the reversal region, or by the smallness of lr'l
rather far from this region. This is the difference between Eqs. (23)and (24) and
the corresponding equations for electron lenses in which the smallness of the
right-hand sides is provided solely by the smallness of the trajectory slope, and
so they are not suited for analyses of mirror systems.
In order to solve the sets of equations (23)and (24),we shall find at first the
quantity U determining r in the first approximation, setting S(3)= 0 in
Eq. (23). Substituting U into the right-hand side of Eq. (24), we determine q.
Then, substituting U and into the right-hand side of Eq. (23), we obtain a
linear nonhomogeneous differential equation of third order for the correction
term x
z=r-U (31)
With S'3' = 0 we have
4,'' + $$'U' + $Q2U = 0 (32)
The equation has a singularity at [ = zk determined by the equality + ( Z k ) = 0.
The point is a first-order pole for the coefficients at U' and U, because it is
assumed that the electric field differs from zero in the vicinity of electron
trajectory reversal points. In accordance with the theory of appropriate
equations (Smirnov, 1974), choose two linearly independent real partial
solutions of Eq. (32), one of which, p ( 0 , is the analytic function, and the other,
g ( l ) , can be represented as
S(l)=4 ( t ) r n (33)
where q(5) is the analytic function. On substituting the function g in Eq. (33)
into Eq. (32), instead of U we obtain
4q" + &$'q' + $(34"+ $ k 2 B 2 ) q= 0 (34)
Let us specify the initial conditions at the point 5 = zk

Pk = qk = (35)
Then from Eqs. (32) and (34) it follows

The subscript k denotes the quantities referring to 4(Zk) = 0.


ELECTRON MIRRORS AND CATHODE LENSES 345
Note that, using Eqs. (32) and (34), we can find the derivativesfrom p and q
of any order at = z,; they may be needed for numerical integration of these
equations.
The particular solutions, p and g , are related as follows:
&(Pd - P’d = 4LP (38)
We shall use these solutions as the fundamental set of solutions of Eq. (32).
Then the general solution of this equation may be represented in the form
U = UP + bg (39)
where a and b are arbitrary, complex, in general, constants of the first order.
On substituting this solution into Eq. (26), instead of r, and taking into
account Eqs. (18) and (38), we obtain
d3)= - u u * ( ~ P +
’ ~2Q2p2) - (a*b + ab*)(4p‘g’+ 2 Q 2 ~ g )
+ ia(a*b - ab*)$kB+; - bb*(c$g’2+ 2Q2g2)+ E (40)
We shall seek the solution of Eq. (24) in the form
q = eqq + aa*vl + (a*b + ab*)q2+ bb*q3 + ia(a*b - ab*)q4 (41)
Then Eq. (24) decomposes into a set of equations the solutions of which are

1
v3 = 24?l2- (45)

The quantities qrp,ql - q4, and their derivatives are finite everywhere.
According to Eq. (41), it provides the second order for the functions v and 1’.
On substituting into the right-hand side of Eq. (23), determined by Eq. (25),
instead of r and v, their approximate values, in accordance with Eqs. (39) and
(41) and using the relations in Eqs. (18), (31), (32), and (38), we obtain a linear
nonhomogeneous differential equation for defining x, whose particular
346 E. M. YAKUSHEV AND L. M. SEKUNOVA

solution will be sought in the form


x = a2a*x1+ a2b*x2 + aa*bx3 + a*b2x4 + abb*xs
+ bZb*x6+ &(ax7+ bxs) (47)
In doing so, the differential equation for x decomposes into a set of equations
cfq; + &YxJ + $Q2xj = Si3), j = 1,2,. . . , 8
where

Solving Eq. (48) by the method of variation of parameters, we find

Equations (31), (39), and (47) together with Eq. (57) determine r as a
function of [. Now in order to find r and $ as functions of [ based on Eqs. (4)
and (9,it is necessary to eliminate time t from Eq. (5).Using Eqs. (1 l), (20),and
(22),we write, neglecting the excess terms
o=*-o--Ic
where

When deriving Eq. (58), the integration constant is set equal to zero; however,
this fact does not restrict generality of consideration, because the constants a
and b are arbitrary. It follows from Eqs. (4), (31), and (58)
= rei(+e-K) =
u+x (61)
ELECTRON MIRRORS AND CATHODE LENSES 347
hence
re'(^-') = U + x + ilcU
Transform Eq. (60), substituting U and q into it in accordance with Eqs. (39)
and (41)
IC = a a * q + (ab* + a*b)K, + bb*K3 + E I C ~+ i(a*b - ab*)K5 (63)
Herein

Using Eqs. (39), (41), (47), and (63), we write in a parametric form the
equations of a charged particle trajectory in the field of a mirror

These equations determine particle trajectories in an electron mirror through-


out their paths including reversal points. They are valid up to terms of third
order of smallness in r and of first order in re.
348 E. M. YAKUSHEV AND L. M. SEKUNOVA

+
In order to obtain the explicit dependences of r and on z, it is necessary to
c
define = c(z)from Eq. (70) and to substitute it into Eq. (69). The solution of
Eq. (70) with respect to ( will be defined by the method of successive
approximations. Up to terms of second order in r and of first order in E
i= - rl(4 (71)
On substituting this equality into Eq. (69) and expanding the entering
functions of the argument z - q(z) in series in q, which is necessary for an
analysis of aberrations, we find

= rein = aP(z) + b d z ) + x(z) - c a m + bg’(z)lrl(z)

+ i[ap(z) + bg(z)] K ( Z ) -
( 0
-
4
kB(z) rl(z))
m
+
where the angle R = - 8 characterizes a particle going out of the plane
+= 8(z) rotating about the z axis, and

The functions p(z), g(z),~ ( z ~) ,( z and


) , K ( Z ) entering Eq. (72) were defined earlier
as the functions with argument [.
Thus, Eq. (72) describes charged particle trajectories in the rotating
coordinate system r, R, z as the explicit dependence of rand R on z. Equation
(72) is valid everywhere, excluding the neighborhood of trajectory reversal
points because the values of the derivatives u’(c), x’((), ~’(c), and 8’(c) increase
without limit there, and, due to this, the expansion becomes invalid. However,
if an image is formed far from the reversal point, the limitation mentioned is
unimportant.
To determine a concrete trajectory, it is necessary to express the arbitrary
constants a and b via the initial conditions. When the initial conditions are
specified far from the reversal point, the arbitrary constants can be found by
Eq. (72). When the initial conditions are specified near the reversal point (e.g.,
when calculating the trajectories in an electron emission/mirror microscope),
only Eqs. (69) and (70), also valid in this region, are suitable for determining the
arbitrary constants. From Eq. (72) it follows that in the first approximation the
particle trajectory is described by the dependence u = U(z),where U ( z )is the
solution of the equation
+
~ ~ ( z ) U ” ( Z+’(z)U’(z)
) +
+QZ(z)U(z)= 0 (74)
This equation fully coincides with the paraxial equation for axially symmetric
electron lenses; however, in contrast with lenses, it cannot describe the entire
ELECTRON MIRRORS AND CATHODE LENSES 349
trajectory in a mirror. Near the reversal point z = zk, both Eqs. (74) and (72)
become invalid. This circumstance is insignificant when observing an image
situated, just as the object itself, far from the reversal point.

B. Time of Flight

In the derived trajectory equations, Eqs. (69), (70), or (72), the explicit
dependence of particle coordinates on time of flight is eliminated. Let us show
that the information on time-of-flight characteristics of electron-optical
systems is contained in q. Indeed, due to the synchronous motion of an
arbitrary particle and the appropriate guiding particle, their time of flight
from the instant of reflection, corresponding to 5 = zk, to any other instant,
corresponding to 5,will be equal, according to Eq. (1l), to

During this time any synchronously moving particle of the set reaches its
plane, z = const. These planes, as seen from Eq. (12),do not coincide, and their
parallel displacements relative to each other are determined by the de-
pendence of q on the initial conditions for the entire particle set. On the
contrary, all the particles of the set intersect some specified plane, z = const, at
various instants of time. The spread of time instants A t of the intersection of
the chosen plane by these particles characterizes time-of-flight aberrations. To
define the spread, we rewrite Eq. (75) taking account of Eq. (12) in the form

Here

where

represents the time of flight for the guiding particle, and the quantity

means the total time-of-flight aberration. Equation (79) is rigorous. If one


restricts oneself by the terms of order not higher than second in r and of first
350 E. M. YAKUSHEV A N D L. M. SEKUNOVA

order in E, then for the region far from the reversal points of electron
trajectories we obtain a simple and physically clear equality

At = --=w --
B
[&qs + aa*ql + (a*b + ab*)q2
kv% kv%
+ bb*q3 + i(a*b - ab*)q4] (80)
determining time-of-flight aberrations in any plane z = const.

111. ELECTRON-OPTICAL
PROPERTIES OF AXIALLY
SYMMETRIC
ELECTRONMIRRORS

The previous section presented the method of calculation of axially


symmetric mirrors, allowing one to overcome mathematical difficulties
associated with a breakdown of classical paraxiality conditions [Eq. (lo)] for
charged particle trajectories in the neighborhood of reversal points. It gives
one a possibility to carry out investigations of electron mirrors and to find the
expressions for aberration coefficients in a form like that for electron lenses,
i.e., depending on the field distribution over the mirror axis. Below, the
electron-optical properties of a mirror in a paraxial approximation are
investigated, aberration classification is produced, and the aberration coeffi-
cients are calculated. The object plane, z = z,, and the image plane, z = zb, are
assumed to be located out of the mirror field; hence, there is a possibility to
study the electron-optical properties of a mirror using Eq. (72). The z
coordinate enters the equation as an independent variable. All the functions
considered in this section are functions of this coordinate; the prime denotes
differentiation with respect to it.
It is the peculiarity of electron mirrors that there are two branches of a
trajectory- the direct one, corresponding to the particles incident on the
mirror, and the reversed one, corresponding to reflected particles. It is evident
that for the values a, b, 0 given, Eq. (72) describes one of the two branches of a
trajectory-the direct or the reversed. For every branch there is its own sign of
c and, generally speaking, its own meanings of the arbitrary constants a and b.
We refer B, a, b to the incident particle before its reflection. After reflection
these meanings change and turn out to be 5,Z, 6. Then the following equalities
hold:
u = u(z, a, a, b), = x(z, a,a, 6)
- (81)
yI = q(z, 5,a,b), 2 = K ( Z , 5,a, b )
ELECTRON MIRRORS AND CATHODE LENSES 351
Herein and after, the bars refer to the quantities or the functions related to the
reversed branches of trajectories.
Let us establish the connection between the quantities characterizing the
direct and the reversed branches of a trajectory. According to the identity in
Eq. (13), 5 = - 6.We shall find the relation between the constants ii,band the
constants a, b using continuity of a particle trajectory and of its velocity
in the neighborhood of a reversal point. A s ~ ( 2 ,7 ) 0, i ( z k )= 0 in this point, it
follows from Eq. (61) that u(z,) = U(z,), u(z,) = U(z,). Hereof, using Eq. (39),
the relations in Eqs. (1 1) and (33), and taking into account for differentiability
of the functions p , q at the reversal point, we obtain: up, = Zp,, bak&q, =
bCk+;q,. Thus the constants before and after reversal are connected by the
relations
-
5=--o, Z=a, b=-b (82)
These relations, together with Eqs. (81), give a possibility, based on Eq. (72), to
write the equation for the reversed branch of a trajectory for specified initial
conditions corresponding to the direct branch of this trajectory.

A . Determination of the Constants a and b

We shall determine the constants a and bin terms of the coordinate u, and
the trajectory slope ub in the object plane. The rectilinear section of the
trajectory of a particle going out of the object plane z = z, may be represented
in the form

Here c1 and jl are complex variables of first order determined by the initial
conditions in the object plane

The subscript a denotes the quantities referring to the object plane. The initial
value of SZ, is set to be equal to zero (u, = r,); however, this fact does not
restrict the generality of consideration because the system is axially sym-
metric. R is a constant value and is equal to
352 E. M. YAKUSHEV AND L. M. SEKUNOVA

On the other hand, based on Eq. (72) for the same section of the trajectory,
we can write

u = (a + x p - ubq, + ioic)p(z)+ (b + xg - ubq, + ibic)g(z) (87)


In this expression the quantities xp, xs,q p ,qg are connected with the functions
~ ( zand
) q(z) by the relations

As follows from Eqs. (47) and (57)

The quantities qp,qg may be found by proceeding from the conditions for
+ +
q and q’ in the object plane: qa = paqp gaqg; qb = p:qp g:qg. Thus we
obtain

It is worth noting that the values x p ,xe, qp,qg are the constants independent of
the position of the object plane z = z,.
Now, on comparing Eqs. (83) and (87), we may obtain

(95)

In doing so it is enough to substitute into x p , xg, qa, and ic, instead of the
constants a and b, their approximate values alp: and jllg:, respectively.
Thus, Eqs. (94) and (95) with account of Eqs. (84) and (85) determine the
constants a, b in terms of the initial conditions for the object plane with
accuracy up to third order.
ELECTRON MIRRORS AND CATHODE LENSES 353
B. Mirror Properties in the First Approximation

Using the developments made in the previous section, the trajectory


equation can be written as

u = rein = ap + bg + x - u'q + iu
( --
K
:$q)
for the direct branch, and

for the reversed one.


At first, let us consider mirror properties in the first approximation. For
this purpose, in Eqs. (96) and (97) we restrict ourselves by terms of first order in
r and set E = 0. Then, on taking into account Eqs. (73), (82), (84), (85), (94), and
(95), the equation of the trajectory in this approximation takes the form

Herein, with the double sign (f), the upper refers to the direct branch of a
trajectory, while the lower refers to the reversed. This notation will also be used
further on for trajectories in a mirror.
Let us determine the cardinal elements of an electron mirror. For this aim
consider several specific mirror trajectories. It is clear from Eq. (98), provided
that
ub = raPb/Pa (99)
in the meridional plane R =0 of a rotating coordinate system there is a
family of trajectories

r = rbp(z)/pb (100)
the direct and the reversed branches of which coincide (Fig. 2). Depending on
the function p(z), these trajectories may or may not intersect the z axis. By
analogy with light optics, we shall assume that such trajectories determine the
curvature center of an electron mirror located in the point zc of intersection of
the mirror axis and the continuations of the out-of-field rectilinear sections of
the trajectories. Out of field

P(4 = (z - Z dPb (101)


354 E. M. YAKUSHEV AND L. M. SEKUNOVA

Iz=zk
FIG.2. Trajectories determining the center of curvature of an electron mirror (Sekunova,
1977): (a) divergent mirror with repulsive potential & = -q$, (b) convergent mirror with
repulsive potential do = -0.185 4d.
LM is an equivalent optical mirror.

So the curvature center of the mirror will be at the point z = zc belonging to its
principal optical axis.
When satisfying the condition
4 2 = (ra/sa)sh ( 102)
Eq. (98) describes the trajectories lying in the meridional plane R = 0 and
symmetric with respect to the principal optical axis (Fig. 3)
r = +rbg(z)/gb ( 103)
Intersection of trajectory out-of-field section continuations with the z axis
determines the point z = zB playing the role of the electron mirror vertex. The
value of zB may be found if the function g(z) is known, because out of field
g(z) = (z - zB)gh (104)
The constant R introduced above [see Eq. (86)] and equal to
R = palpb - 9.19: = ZB - zc (105)
has a sense of the curvature radius of an electron mirror. With oR 0 the =-
mirror is convergent, while with OR < 0 it is divergent. Note that, according to
ELECTRON MIRRORS AND CATHODE LENSES 355

,r
@a I I
I

FIG.3. Trajectories determining the electron mirror vertex. Notation as for Fig. 2.

Eq. (38), there is a relation between the electric field intensity in the reversal
point of a trajectory and the curvature radius of an electron mirror
R = 4;/2PhgbJ, (106)
The trajectories initially parallel to the principal optical axis (Fig. 4) in the
plane !2 = 0 of the rotating coordinate system may be represented in the form

It follows from Eqs. (101),(104),(106), and (107) that both for the electron and
the light-optical mirror the following relations hold: The focal distance is
f = R/2 ( 108)
the position of the focal plane z = zF is determined by the equality
ZF = +(ZB + zc) (109)
while the principal plane passes through its vertex.
It follows from Eq. (98) that the positions of mutually conjugated planes
z = z , and z = zb are determined by the relation
Pagb + Pbga = (110)
356 E. M. YAKUSHEV A N D L. M. SEKUNOVA

I
+f--i
z=zFI !z=& IZ=tb

FIG.4. Trajectories determining the electron mirror focus. Notation as for Fig. 2.

On incorporating Eqs. (101), (104), and (109, this relation is reduced to the
form

which is also characteristic of light-optical mirrors. Magnification of an


electron mirror equals
M =f/(za - zF) (112)
It is necessary to take into account that, due to the presence of a magnetic
field, the image plane is turned with respect to the object plane by the angle

y = *b - *a = --

C . Aberrations

In order to find the mirror aberrations, we substitute Eqs. (94) and (95) for
the constants a and b into Eq. (97) and write the latter for a rectilinear section
ELECTRON MIRRORS AND CATHODE LENSES 357
of a trajectory out of field. Subdividing the equation obtained according to the
orders of the values, we represent it as
ii=U+Au (1 14)
where is the paraxial approximation for a rectilinear section of a trajectory
determined on the basis of Eq. (98), and Au is a correction term of third order
characterizing the image distortion (aberrations) in an electron mirror. On
taking into account Eqs. (84),(85), (88), and (98),the expression for Au takes the
form

- U'(z)V(z) + A(z) - iu(z)o(z) (115)


where
= (lip - X,)P(Z) + (lig + xg)s(z) (1 16)
o=lC-E (117)
Using Eqs. (47), (49)-(56), (82),(88), (90), (91), (94), and (95), we represent A as
follows:
A(z) = aza*Alg + a(afl* + a*fl)A,p + aflp*A3g
+ aa*pA4p + fl(ap* + a*p)ASg + flzfl*Atjp
+ ia(ap* - a*fl)A,g + ip(afl* - a*P)A8p
+ aEA9S + BEAlOP (118)
Constant quantities from A1 to Al0 are defined by the expressions
358 E. M. YAKUSHEV A N D L. M. SEKUNOVA

and do not depend on initial conditions, because the plane z = za is located


beyond the mirror field, where the integrands of these expressions equal zero.
Involving Eqs. (63)-(68) and (82) represents the function o as the sum
0 = a a * o l + flp*02 + i(afl* - a*fl)o3 + E 0 4 ( 129)
where

ok 1
0 4 =- __ B’qp dz (133)
JZk

are constants independent of initial conditions. Taking into account that


beyond the field
qba = lI24a> via =
1
--pa I2
> via = -3pbgi> (134)
1 r2
?;a = 3 vko =0
as well as involving Eqs. (39), (41), (63), (82), (94), (95), and (1 18), we rewrite
Eq. ( 1 15) as the expansion
ELECTRON MIRRORS AND CATHODE LENSES 3 59

+ m{[A9gb + k(2qpa 4a ~b-".>I" gb - iu4:}

There is a number of correlations between the quantities entering the last


expression. On comparing Eqs. (120) and (123) it is clear that
€ 7 3 2 = sbA5 (136)
As shown below, the following relations hold

[A3g: + k(2+ ;)] - [A4& + f($ + F)] =1


(137)

A7gb---2 ~ 4 -a - 0 1 (138)
R Phsb

The left-hand side of Eq. (137), involving Eqs. (121) and (122), may be
represented as follows:
360 E. M. YAKUSHEV AND L. M. SEKUNOVA

On differentiating the equations


241; - 4‘11 = - 4 ~ ”- ‘4Q 2 p’ (141)
241; - 4 ’ 1 3 = -4d2 - $Qzg’ (142)
we obtain
241; + 4‘ql - 4”11 = -+Q;p’ (143)
241; + 6’1;- 4”13 = -$Q;g’ (144)
taking into account that
’ $QzP’)’ = $Qh2
( 4 ~ ’+ (145)
(4s’’ + bQzgZ)’= $Qh2 (146)
because p(z) and g(z) satisfy Eq. (74). Multiplying Eq. (143)by q 3 ,and Eq. (144)
by ql ,subtracting the first from the second, and dividing the obtained equation
by 2&, we find
C f i ( ~ i & - 1;13)I’ = (Q;/~&)(P’v~ - g21i) (147)
Using the resulting equation, we carry out the integration in Eq. (140) directly
proving Eq. (137), if we take into account Eqs. (106) and (134).
To verify Eq. (138), substitute A7 and q4 in the forms in Eqs. (125) and (46)
into its left-hand side. The twofold integral entering the resulting expression is
carried out by parts, taking into account Eqs. (145) and (106)

( 148)
This expression is rewritten realizing that p satisfies Eq. (74)

On carrying out the latter integration by parts, we find, after eliminating the
uncertainty occurring at the point Zk
ELECTRON MIRRORS AND CATHODE LENSES 36 1
The right-hand side of Eq. (138), determined by Eq. (130), may be reduced to
the same form, if q1 is substituted into Eq. (138) in the form of Eq. (43), and
double integration is carried out by parts. Equation (139) is verified in the same
way.
Taking into account Eqs. (136)-( 139), we reduce the expansions of
Eq. (135) to the form

where

1
5
1 =- 2’pCQ4~3 + 4Q;p2p’ + 32p”(W2 + aQ2p2)]dz (152)

1 9
JZ = -CQ4p2g
84kPaga I z k f i
+ 4 Q W g + 32~”(4p’g’+ aQ2pg)l dz (153)

ok
+
J , = 16pbz~ ~ ~ & ( B ” p4B’p’ + 8Bp”)dz
J
ok
6 - (157)
362 E. M. YAKUSHEV A N D L. M. SEKUNOVA

The expansions in Eqs. (135) and (151) differ by the circumstance that there is
no double integrationin the latter. We have eliminated them by the same way
used for verifying Eqs. (137)-( 139).
Using Eqs. (84), (85), (101), (104), (105), and (109), we rewrite the expansion
in Eq. (151) as follows
Equation (162) determines the aberration of an electron mirror in an
arbitrary plane, located beyond the field. If this plane is the plane of the
Gaussian image, z = z b , then using Eqs. (108)-(112), we may simplify the
expressions for abberrations

Equation (163) determines the image aberration in an electron mirror


when a limiting diaphragm is absent.
If an incident ray is determined by the coordinates of an object point and
by the coordinates rD and R, of the point of its intersection with the plane of
the diaphragm z = z D ,which also located beyond the field, then the following
relation is valid
uD= D e i R ~= r, + uh(zD - z,) = r, + ubZDa (175)
364 E. M. YAKUSHEV AND L. M. SEKUNOVA

Using it, we write expression for the image aberrations in a mirror with the
limiting diaphragm

where

With E = 0 Eq. (176) gives geometrical aberrations of a mirror in a form


quite analogous to the case of geometrical aberrations of electron axially
symmetric lenses. This allows one to use for electron mirrors the classification
of aberrations adopted for electron lenses. According to this classification, G D
is the coefficient of spherical aberration, FD is the coefficient of isotropic
coma, while f , is the coefficient of anisotropic coma, C, is the coefficient
characterizing astigmatism, c, is anisotropic astigmatism, D D is the image
curvature coefficient, and ED and e, are the coefficients of isotropic and
anisotropic distorsions, respectively.
The coefficients at c/4a give one the possibility to determine the mirror
chromatic aberrations. K I D represents the coefficient of chromatic aperture
aberration, the coefficient K 2 , characterizes the change of the image scale,
and the coefficient k, denotes the change of the image turning angle. The
coefficients f,, c,, e,, and k , differ from zero only in the presence of a
magnetic field.
ELECTRON MIRRORS AND CATHODE LENSES 365
D . Time-o
f-Fligh t Focusing

Investigation of time-of-flight focusing of nonstationary fluxes of charged


particles is one of the basic problems when calculating and designing a large
number of electron-optical and ion-optical devices, in particular, time-of-
flight mass spectrometers. Now we study this problem theoretically, in the
general case, i.e., without imposing any restriction, besides their rotational
symmetry.
The time of flight of a particle from the initial plane z = z, to an arbitrary
plane z = const after reflection, according to Eqs. (11) and (12), equals

When deriving this equality, a change of the sign of has been taken into
account in the particle reversal point at i= zk, and the notation q, = ~(2,) has
been adopted. The planes z = z, and z = const are assumed to be located
beyond the mirror field. In this case Eq. (188) takes a simpler form
t = to + (d/k&)(?, + 'I) (189)

is the time of flight of a particle along the z axis with E = 0 between the planes
considered, z = z, and z = const, before and after reflection, 4a denotes the
potential of the field-free space. The second term of the sum in Eq. (189)
determines time-of-flight aberration. We shall find the structure of this
aberration, taking account the quantities of first order in E and of second order
in I , substituting q in the form of Eq. (41)into Eq. (189) and using Eqs. (81) and
(82).
t - to =( +
~ / 4 a ) ~ t e (a/k&){aa*Cqt(za) + V~(Z)I
+ (ab* + a*b)Cqz(z,) - qz(z)l + ia(a*b - ab*)C~4(z,)+ G'4(Z)I
+ bb*C%(Z,) + %(Z)1) (191)
where
Dra(Zu,Z) = Dre = ( D / ~ ) & C V J Z J + V,(Z)I (192)
is the time-of-flight energy dispersion of the first order.
Let us show that the particle time-of-flight focusing in energy may be
achieved with a mirror and deduce the equation of focusing. As follows from
366 E. M. YAKUSHEV AND L. M. SEKUNOVA

Eq. (42), the function q,(z) in a field-free space may be represented as being
linearly dependent on z
110 = (1/24cz)(z
- zT) (193)

Note that the quantity zT does not depend on the presence of the magnetic
field in a mirror or on the choice of the point z, beyond the field and is
determined only by the electric potential distribution along the z axis.
Taking into account Eq. (193), Eq. (192) takes the form

The first-order condition of time-of-flight focusing will be understood as the


independence of the time of flight of particles on their spread in initial energy,
i.e.,

Then it follows that


2, + ZB = 22, (197)
This expression means that time-of-flight focusing is achieved if the plane
z = zar from which particles are emitted, and the plane z = zB, where they
are registered, are located symmetrically with respect to the plane z = z T ,
determined, according to Eq. (194), by the electric field distribution over the
principal optical axis of a mirror. The plane z = zT will be called the principal
plane of time-of-flight focusing. In mirrors, time-of-flight focusing of particles
with a small spread in energy is realized based on the fact that, when reflecting,
the higher the particle energy, the longer the path it covers in the reflecting
field.
Assuming that the principal plane z = zT is located before a mirror, in the
field-free space we rewrite Eq. (190) as follows:

where
ELECTRON MIRRORS AND CATHODE LENSES 367
is the time interval between the instants of intersection of the plane z = zT by a
particle before and after its reflection. This interval will be called the time-
focusing interval. Its dependence on the particle mass determines the amount
of a mirror time-of-flight mass dispersion
dTo 1
D,, = m- =-T
am 2 O

It is a characteristic feature of the principal plane z = zT of time-of-flight


focusing that the time-focusing interval To does not depend, to a first
approximation, on the initial energy of the particles. It means that particles
with equal specific charges, elm, and with a small spread in energy traversing
the principal plane simultaneously before reflection, will also traverse the
plane simultaneously after reflection. The time-focusing interval Todetermines
the effectivedrift distance L

It is worth noting that both L and zT depend only on the mirror electric field.
If the position of the principal plane of time focusing is known, the
position of two mutually conjugated, in the sense of time-of-flight focusing,
planes, z = z, and z = zB is determined by Eq. (197). The latter will be called
the equation of time images, an object/image understood as the structure of
substantially nonstationary flux (of a short current pulse) in the plane of the
object/image. The mirror time magnification, M,, determined as a ratio of the
current pulse duration (At)Bin the image plane, z = z B ,to the initial duration
of the same pulse, (At),, in the object plane, z = z,, is always constant, due to
the stationary nature of the electric and magnetic fields forming a mirror, and
equals

It means that to a first approximation a mirror adequately transmits the time


structure of the nonstationary charged particle flux from the object plane to
the image plane. As it is seen from Eqs. (197)and (198),the object and the image
are shifted, relative each other, in time by To,independent of the values of z,
and zg.
Now we shall find the position of the mutually conjugated planes, z = z1
and z = z 2 , for which both space and time-of-flight focusing are realized.
Setting into the equations of an image, Eqs. (111) and (197), z, = z, = zl,
zb = zB = z2, and solving these equations with respect to z l and z 2 , we obtain

21,2 = zT kJ(zT - zB)(zT - zC) (203)


368 E. M. YAKUSHEV A N D L. M. SEKUNOVA

When this equality is satisfied, the correct point-by-point space and time
image of an object is formed in the same plane, i.e., space-time-of-flight
focusing of the nonstationary flux of charged particles is realized. For an
electrostatic mirror having a specific distribution of the field along its axis,
only a pair of the mutually conjugated planes satisfying the conditions of
space-time-of-flight focusing, Eq. (203),may be found. When a magnetic field
is present, one may find a set of such pairs due to the fact that the quantities zB
and zc depend now both on electric and magnetic fields.
Note that, as for the possibility of realizing space-time-of-flight focusing
of a charged particle nonstationary flux, a mirror is a unique element in the
class of the axially symmetric fields considered.
Now let us consider the second-order time-of-flight aberrations, tracing
the motion of two particles with initial energies E = 0 which simultaneously
traverse the plane z = z , located before a mirror. Let one of them move along
the z axis and, after reflection in a mirror, reach the plane z = const in a time
interval t o . Then the time of flight up to this plane for the other particle remote
from the z axis, due to the presence of aberrations determined by Eq. (19l), will
differ somewhat from to
t = to + (At)!” (204)
where
(At)!2)= ( ~ ~ / k f i ) { ~ ~ *-p&z, + +
b ~ [ Z~) c R I , ]

+ $(ab* + a*b)pbgb(z - z,) + 2ia(a*b - ab*)q,


+ bb*gbZ[zB- +(z, + + R I , ] }
Z) (205)
and

When deriving these expressions, it was taken into account that the initial and
final planes, z = z , and z = const, are located beyond the field of a mirror. In
this case, according to Eqs. (101) and (104), the functions p ( z ) and g ( z ) are
approximated by the straight lines p = pb(z - zc) and g = gb(z - zB), and
Eqs. (43), (44), and (45) for the quantities q l ,q,, q3 are simplified; the quan-
tity q4, determined by Eq. (46), becomes constant.
Represent the structure of geometrical time-of-flight aberrations as the
dependences on a radial deviation of particles ra and on a slope of their
trajectories r: and r,& given in the initial plane z = z,. For this purpose, we
use the expressions in Eqs. (84),(85), (94),and (95) determining the parameters
ELECTRON MIRRORS AND CATHODE LENSES 369
a, b and the relation uh = rk+ ir,$k. Then, combining in a proper way the
quantities of second order, we find
+ r:$i2)
(At):2) = T,(ri2 + T2r,rh + T,r:$i + T,r; (208)
where

T - -b,
ff

-k R f i
- zJ21, + (za - z,)~Z,

The aberration associated with the coefficient TI describes the deformation of


the time structure of the charged particle's nonstationary flux emitting from
the point located in the z axis and in the plane z = z,. The aberration of this
sort may be called the time-of-flight spherical aberration. The coefficient T4
determines the time deformation of a structure of the flux entering the mirror
field parallel to the z axis. An aberration of this sort depends only on the
particle coordinates in an object plane, and so it may be called the time-
of-flight distortion. The coefficients T2,T3 determine mixed time-of-flight
aberrations, and in pure electrostatic fields T3 = 0.
If the first-order time-of-flight focusing in an energy E is available as well as
the initial spread in energy which is relatively great, it is of interest to calculate
time-of-flight aberrations of second order and above. In order to calculate
these aberrations, it is necessary to consider the particle motion along the z
axis and to define, with a required accuracy, the quantity q = q ( ( ) as the
dependence on the initial energy E. Then for the space free of field the total
time-of-flight chromatic aberration may be written as follows

It is possible to calculate chromatic aberrations with any specified


accuracy, as the equations of motion of axial particles can be integrated
strictly at arbitrary values of E. For axial particles (at r = 0) Eq. (15), with the
370 E. M. YAKUSHEV A N D L. M. SEKUNOVA

use of Eqs. (12) and (17), reduces to an extremely simple form

1:'s
allowing solution by the method of separation of variables
dz
=j : + V k d m

where the quantity q k satisfies the equality


4(zk + qk) + = (216)
Equation (215) combined with Eq. (216) determines the required dependence
9 = V(Z,&).
In order to make Eq. (215) more convenient for carrying out the analysis of
time-of-flight aberrations, it is necessary to fulfill certain transformations in it.
We rewrite it using Eq. (216) as follows:

The expression under the radical on the right-hand side of the equation will be
rewritten in the form
4(z+ q k ) - 4 ( z k + q k ) = 4(z)[1+ pL(z,q k ) ] (2 18)
where p is a small quantity. Expanding the functions 4,4(z + v k ) , + ( z k + q k ) in
a power series in q k , we obtain

It is seen from this expression that the function p is continuous and


differentiable everywhere including the singularity z = zk. The value of p = pk
at the singularity will be found after eliminating the uncertainty of the 0/0 type

In the space free of field the quantity p becomes constant, pa = &Ida.Recall


that the object and its image are located beyond the field of a mirror. This
provides a possibility of solving Eq. (217) with respect to 9. On carrying out
this operation and involving Eqs. (218), we find
ELECTRON MIRRORS AND CATHODE LENSES 371
This equation, combined with Eqs. (213), (216), and (219), determines the
overall time-of-flight aberration (At), with any specified accuracy with respect
to the spread in energy in the nonstationary flux of particles. So, restricting
ourselves by the values of second order, on the basis of Eq. (221), one may
obtain

+ Wi2’
(At), = (&/4Pt, (222)

where D,, is the energy dispersion determined by Eq. (195), and


(At)L2’ = (e2/&)TV (223)
is the time-of-flight chromatic aberration of second order, and

(224)
with
PI = (4‘ - 4;M p2 = (4“ - 4;1/4 (225)
The second-order total time-of-flight aberration of an electron mirror is
composed of geometrical aberrations [Eq. (208)] of four kinds and of
chromatic aberration [Eq. (2231 of one kind
(At)‘,’ = (At)!,) + (At)!” (226)
In the case for which space-time-of-flight focusing is realized, the coefficients
of geometrical and chromatic time-of-flight aberrations in the image plane
take the form
372 E. M. YAKUSHEV A N D L. M. SEKUNOVA

IV. MIRROR
ELECTRON
MICROSCOPE

An electron mirror possesses a remarkable feature-it creates an image of


its reflecting electrode. For the first time this effect was noticed by Hottenroth
(1936, 1937). Investigating experimentally electron-optical properties of
electrostatic mirrors, he noticed that, with a solid reflecting electrode, tiny
roughness of the electrode surface is observed in a mirror image. The mirror
ability to create a magnified image of the surface of its electrode reflector
allows one to use it as an original electron microscope like a shadow light-
optical microscope.
The scheme of a mirror electron microscope (MEM) is shown in Fig. 5. The
main part of the microscope is the mirror M (in the simplest case: aflat sample,
a cathode and a diaphragm with an aperture, and an anode). An electron beam
emerges from the source S and, retarded in the field of the mirror objective,
reflects at a very short distance from the sample surface K. When near a sample
a slow electron beam is modulated by microfields produced by micronon-
homogeneities of the sample surface, which may be of electric, magnetic, or
geometrical character. After changing the direction of its motion, the beam
again traverses the objective field, which now acts upon it as accelerating one.

FIG.5 . Scheme of the simplest mirror microscope: (a) of the direct construction (the
incident and the reflected beam axes coincide), (b) with separation of the incident and reflected
beams by virtue of the magnetic field B.
ELECTRON MIRRORS AND CATHODE LENSES 373
On the screen P the reflected beam creates a brightness picture which
represents the magnified image of the nonhomogeneities of the sample
surface. It is the advantage of the device that in it a sample does not undergo
bombardment by an electron beam. As a result, the sample is not destroyed, its
subcharging is reduced to minimum, there is no sample heating, and no
secondary emission of its surface or other undesirable phenomena. MEM is
extremely sensitive to the micrononhomogeneitiesof the surface under study,
because in the region occupied by fields associated with these micronon-
homogeneities which are very close to the sample surface, the velocities of
particles belonging to the beam incident on the sample are extremely small,
and even insignificant disturbances of the field in the near-cathode region lead
to noticeable redistribution of the velocities in its components. In the
neighborhood of the reversal point the character of the particle motion is such
that its trajectory is parallel to the surface under consideration, and it still
raises the device sensitivity because of an increase in the time of interaction
between the beam and microfields. If the potential difference between the
sample and the cathode of the source of illuminating electrons (bias voltage)
varies, it is possible to study the surface microfields of the sample at various
distances from its surface. Owing to the merits mentioned above, the
application field of MEMs is rather wide (Luk’yanov et al., 1973).
An image in a mirror electron microscope operating in a shadow regime is
formed in a different way than in a microscope of a transmission type. In
a mirror microscope an image represents a brightness picture formed by a
reflected beam in which the density of electron current is redistributed as
a result of its interaction with micrononhomogeneitiesof the cathode surface.
Conformity between the object micrononhomogeneities and the received
image is of a rather complex character. The contrast theory explaining this
conformity is constructed based on the calculation of the current density of
the reflected beam of electrons moving in the field of a mirror objective
disturbed by the nonhomogeneities of the cathode surface in any cross sec-
tion. As the region of action of micrononhomogeneitiesin the direction per-
pendicular to the cathode surface is small compared with the objective field
extent, one may agree with Wiskott (1956) that the field disturbance caused by
them, is concentrated in a narrow near-cathode region, and that the entire
additional pulse is imparted to the electron at its reversal point. The problem
of contrast calculation then involves two tasks. The first is to define the value
of an additional pulse which is imparted to the electron in the disturbed field (a
microfield) near the cathode. Here, in the absence of micrononhomogeneities
on the cathode surface, the field may be considered as homogeneous. The field
disturbance is determined by a concrete character of micrononhomogeneities
of various types on the cathode surface. Calculations are based on deter-
mination of electron motion in such a field. The study by Wiskott (1956)
3 74 E. M. YAKUSHEV A N D L. M. SEKUNOVA

underlying the theory of a mirror microscope was the pioneering work. Later
much effort was devoted to this problem. The most comprehensive papers to
be mentioned are those by Barnett and Nixon (1967-1968) and Sedov (1970).
The second task is to calculate the electron trajectory in the basic field of a
mirror objective (in a macrofield). In doing so it is necessary to take into
account that in the reversal point of an electron, due to a disturbance, the
initial conditions for its motion in the reversed direction are changed. When
interpreting the image contrast, the calculation of trajectories in a macrofield
is no less important than the calculation of disturbances in a microfield. It is
known (Sedov et al., 1971) that disappearance of the contrast and even its
inversion are possible for certain geometries of a mirror objective and for
appropriate parameters of illuminating electron beams. However, the
complete studies of trajectories in the macrofield of a mirror objective are not
enough. Therefore, the image deformation associated with geometrical
aberrations of electron mirrors have not been studied, and theoretical
investigations in a paraxial approximation have been carried out only in
application to the simplest designs of the MEM objectives (e.g., Luk’yanov
et aZ., 1968; Artamonov and Komolov, 1970; Schwartze, 1965-1966).
Here the geometrical optics of MEMs operating in the shadow regime is
considered taking into account the mirror’s geometrical and chromatic
aberrations of third order in r and of first order in rE without imposing any
restriction on its design but the rotational symmetry of the field forming a
mirror. In doing so it is taken into account that, equally with electrostatic
fields, magnetic fields can take part in forming a shadow image, and the
cathode surface (a mirror reflecting electrode), in general, has a form of the
surface of an axially symmetric body.
Electron-optical properties of a mirror, including its aberrations, were
dealt with, as a whole, in Section 111. However, a mirror operating in the MEM
regime has a peculiarity that it creates on a screen an image of one of its
electrodes-a cathode, i.e., the object located in the neighborhood of the
reversal points of electron trajectories. Therefore, the formulas obtained in
Section 111, to deduce which it is assumed that both the object and the image
are located beyond the mirror field, cannot be applied for calculation of mirror
microscope aberrations. However, such calculations may be made using the
equations deduced in Section I1 that determine the trajectories of charged
particles in the field of an electron mirror taking into account the terms
proportional to the initial energy spread of the particles and the terms of third
order with respect to a particle’s deviation from the axis of field symmetry. An
electron mirror is considered to be the entire optics of a mirror microscope,
except for the system of separation (when it is available) of the direct and the
reversed beams, as well as the emission system specifying the illuminating
beam of electrons. Then the source of illuminating electrons is understood as
ELECTRON MIRRORS AND CATHODE LENSES 375
the electron beam crossover formed by the emission system. The position of an
electron source and its size are assumed to be the position of the crossover and
its size that are formed by the emission system in the absence of a mirror
field. Such definition of a source gives one the possibility of separating
investigations of the mirror electron-optical properties and those of the
emission system. We shall assume that there is a field-free region between the
mirror and the illuminating system, and it is there that the microscope screen
is located.
Introduce the cylindrical coordinate system r, $, z , whose origin is set into
the point of intersection of the z axis and the cathode surface, while the
positive direction of the z axis is chosen along the direction from the cathode
to the screen. The potential on the cathode surface is assumed to be constant
and equal to zero. Under these conditions the equalities hold

o=-l, z,=o

which are to be introduced when using the formulas deduced in Section 11. The
energy of illuminating electrons will be characterized by the quantity cp + 5,
5 now being associated not only with the initial energy spread E of the
illuminating particle beam, but also with the bias voltage 6 between the
cathode reflector and the cathode of the emission system specifying the flux of
illuminating electrons
(=&+S (233)
To investigate the mirror electron-optical properties in the MEM regime
we shall use both Eq. (72), determining the explicit dependences of I and $ on z
in a region far from the cathode surface, and the parametric dependences of the
trajectory [Eqs. (69) and (70)], which are also valid in the near-cathode region.
The constants a and b entering these equations are to be expressed in terms of
the coordinates of the point on the cathode surface and the coordinates of the
point of an electron source. In doing so, it is necessary to keep in mind the
specificity of the MEM operating in the shadow regime, the essence of which is
that when illuminating monoenergetic electrons by a homocentric beam, the
information at a certain point on the cathode surface is transmitted by a
unique trajectory (a reflecting trajectory) which passes by this point at the
smallest distance (in the limit, it is a tangent of the cathode surface at this
point). Every reflecting trajectory is associated with the appropriate point on
the cathode surface by definite conditions: First, it intersects the normal to the
cathode surface drawn through this point, and, second, in the place of their
intersection the reflecting trajectory and the normal are mutually per-
pendicular (Fig. 6). If the source of illuminating electrons has a finite size and
emits electrons with an energy spread, then every point belonging to the
376 E. M. YAKUSHEV AND L.M. SEKUNOVA

FIG.6. Scheme determination of the connection between a point on the cathode and the
reflecting trajectory. ANB, cross section of the cathode surface by a meridional plane; A'N'B',
projection of the reflecting trajectory section on this plane; NN', normal to the cathode surface.

cathode surface is associated not with one reflecting trajectory, but with a set
of trajectories (a reflecting beam of electrons), each of which satisfies the
specified conditions.

A . Determination of Arbitrary Constants

In order to establish the connection between the constants a and b referring


to the reflecting trajectory and the coordinates of the point on the cathode
corresponding to this trajectory, it is necessary to consider the trajectories
traversing within close proximity of the cathode surface. Then, on account of
aberrations, one should keep the terms up to and including third order in r. Let
the cathode surface be an equipotential; then its equation near the axis may be
approximated by the paraboloid of revolution
= (4i/44;)uku? (234)
where U f = rkei$*;rk,$k are the coordinates of the point on the cathode sur-
face. The equation of the normal to this surface will be

NOWwe shall find the point of intersection of this normal and the trajectory.
Using Eq. (62),we write
re'* = [u(()+ x ( ( ) + i ~ ( ( ) ~ ( ~ ) ] e ' @ ' ~ ' (236)
Setting Eqs. (235) and (236) equal, and taking into account that z = [ + '1, we
ELECTRON MIRRORS AND CATHODE LENSES 377
obtain

c
The value = incorresponding to the intersection point will be found from the
condition of perpendicularity of the normal and the reflecting trajectory,
which, in the required approximation, may be written as follows:
(4;:/44;)[uku*’(c) + u:v’(c)15=5, = (238)
Then, using the equality U = ap(c) + b g ( c ) and g(c) = q ( c ) m , as well as
the initial conditions of Eq. (35) for p ( c ) and q(c), we obtain
= $4;r)(Ukb* + Uzb) (239)
From this equality it is clear that enis the fourth-order value. Equations (237)
and (239) give

When deducing these expressions, due to the smallness of en, the following
equations are adopted:
p(in) = 1, g(cn) = a?
4’(cn) = 4;? q ( c n ) = q(O) = q k ,
ei8(‘n)= 1 + Wn), O(5,) = - kBk&$5/24;, 42 = x(i,) =0
Equation (240)connects the constants a, b of the reflecting trajectory and the
coordinates of the corresponding point on the cathode surface. The relation
between these constants and the coordinates rs, Rs,zs of the source point may
be determined on the base of Eq. (72).Extrapolating the rectilinear section of
the trajectory direct branch from a certain plane z = z, located beyond the
mirror field to the source plane z = zs, we obtain
us = IseQ - aps + bgs + x s - ( a ~ h+ bgb)qs + i(aps + bgs1k.a (241)
where
ps = (zs - zdpb, g s = (zs - Zs)gb, v ~ s= ~ ( 2 , )+ q’(za)(zs - za)
Ka = K(Za), x s = x(za) + ~ ’ ( z , ) ( z-s z,)
the points z = zB,z = zc determining, as before, the positions of the vertex and
the center of curvature of an electron mirror on the z axis. On solving the set of
equations, Eqs. (240) and (241) with respect to a, b with accuracy up to values
378 E. M. YAKUSHEV A N D L. M. SEKUNOVA

of third order, we obtain

- x s - 4aPS + BSS)~,l (243)


Here a, p are values of first order in r
a = uk (244)
B = (l/gs)(us - Psuk) (245)
As can be seen from the expressions in Eqs. (242) and (243), to a first
approximation one may set a = a, b = /?.This approximation is enough to
calculate the values I],x,K to an order not higher than first by Eqs. (41), (47),
and (63). Thus, Eqs. (69) and (70) [or (72)] combined with Eqs. (242) and (243),
completely determine the charged particle trajectories in the field of a mirror
objective under specified conditions of the object point and of the point of an
electron source.

B. Electron-Optical Properties of MEM


in the First Approximation

Now we shall investigate the properties of a mirror microscope in the first


approximation. In it, the equation of a reflecting trajectory in the region far
from the cathode surface, according to Eqs. (72) and (242)-(245), will be
T @(@I =
ukP(z) f [ g ( z ) / g s l ( u s - Psuk) (246)
Recall that the functions p(z),g(z) describe the trajectories lying in the
meridional plane of the rotating coordinate system and determining the
cardinal elements. In the field-free region these functions are approximated by
straight lines and, according to Eqs. (101), (104), (105), (108), and (109), are as
follows:
dZ)= ( z - zF + f)pb (247)
g ( z ) = ( z - zF - f)gb (248)
ELECTRON MIRRORS AND CATHODE LENSES 379
where f is the focal distance, zF is the position on the z axis of the focal plane
z = zF.Just as in light optics, the position of the mirror curvature center, z c ,
and the position of its principal plane, z B , are given by the relations
zc = ZF - f (249)
z g = ZF +f (250)
In the screen plane, z = z b , which is the plane of a shadow image, Eq. (246),
written for the reversed branch of the trajectory, gives

where b - r be i[#b+@(Zb)l P b = ( z b - ZF + f ) P b , g b = ( z b - Z F - f ) g b . Equation


(251) means that it is the correct image of the cathode surface that is formed in
the screen plane for the illuminating electron point source. Indeed, if such a
source of electrons is found on the mirror axis (r, = 0),then Eq. (251) becomes

Hence, it follows that there is a linear conformity between the points of the
cathode and the microscope screen. The microscope magnification is constant
in all directions and equals
= rb/rk = P b + (Ps/gs)gb (253)
From the expression in Eq. (208) it also follows that, in the presence of a
magnetic field, the image plane is rotated with respect to the object plane
around the z axis by the angle

(254)

The infinity sign in the upper limit of the integral means that the latter is
calculated up to a certain arbitrary value z = z , belonging to the region free of
field. Now we consider the case when the point source of illuminating
electrons is off the z axis (us # 0). In this case, from Eq. (251) it follows that the
central point of the cathode (rk = 0,z = 0)corresponds to the point (q,,u b o ) of
the plane of the microscope screen
ObO = -(gb/gs)% (255)
The above conclusions on the adequacy of a shadow image, its rotation with
respect to the object plane, and the microscope magnification also remain
valid when the point source of illuminating electrons is displaced. However,
380 E. M. YAKUSHEV AND L. M. SEKUNOVA

establishing the point-to-point conformity between the object and the image, it
is necessary to take into account the displacement of the image central point
from the z axis caused by the displacement of the point source and determined
by Eq. (255).
Equation (253)gives one the possibility to determine the magnification of
the mirror microscope in any position of the planes of the screen, z = zb, and
the source, z = z,. It is common for these planes to coincide (z, = zb); the
+
magnification of a microscope is M = 2(z, - zF f ) p b . If the source is at
infinity (the illuminating beam is parallel), then M = 2(zb - zF)ph.In the case
when the source is in the center of curvature of a mirror, M = (zb - z F - f)&.
When the positions of a point source and the principal plane of a mirror are
brought close together (z, + zB), the microscope field of view degenerates to
a point, while its magnification tends to infinity. In this case an electron probe
is formed which illuminates the only point of the cathode surface uk = u,/p,.

C . Image Forming

Now we consider the reflecting trajectories in the presence of small


disturbances near the surface of the mirror cathode. These disturbances may
be caused both by (electric or magnetic) microfields on the cathode surface and
by micrononhomogeneities of the surface relief. However, whatever the
character of disturbances, it is common that after their action redistribution of
electron velocities in its components takes place, and, as a result, the
intersection of the proper reflecting trajectory and the device screen is
changed. If one disregards the details of the disturbing process, the
mathematical result may be characterized by some changes of the constants u
and P in Eq. (246) for the reversed branch of the reflecting trajectory, writing
the equation for this branch as follows:

Taking into account that the disturbance is concentrated in the neighbor-


hood of the reversal point of the reflecting trajectory, we find from the con-
dition of continuity of a trajectory in this point that Au = 0. In order to
express the value Ab in terms of the particle velocity change caused by the
disturbance, we find the dependence of /3 on the particle velocity in the reversal
point of the reflecting trajectory. Taking into account values of third order,
this dependence is determined by Eqs. (69) and (70), in which it is necessary
to set [ = T k m ) . If we restrict ourselves by values of first order, this
dependence may also be derived from Eq. (246), if we set i = T k m) On.
differentiating Eq. (246) with respect to time in the neighborhood of the
ELECTRON MIRRORS AND CATHODE LENSES 38 1
reversal point, we find

Analogously, from the expression for the reversed branch of the reflecting
trajectory [Eq. (256)], we find the value fl + Afl and, comparing it with
Eq. (257), obtain

Afl = -L e i * k ( A i k + irkA$k)
k4L
where Aik and i k A$k are, respectively, the radial and azimuthal electron
velocity changes undergone in the perturbation process. Then, caused by this
disturbance, the displacement Aub of the point the reflecting trajectory
traverses the microscope screen is determined by the equality
2 .
At+, = y e r S (Aik
k + irkA$k)(Zb - zF - f)gb (259)
k4k
Note that the electron velocity, as a result of a disturbance, changes only its
direction and, therefore
(ik + At,)’ + r i ( $ k + A&)’ + (Aik)’ = i,’ + li$i (260)
where A i k is the electton velocity component normal to the cathode surface
received by an electron in the perturbation process. If the disturbance is so
+ +
small that the quantity [(Aik)’ ri(All/k)2 (Aik)2] is negligible compared
+
with (if ri$:), then, instead of the expression in Eq. (260), we may write the
equality
ik Aik + rf$k =0 (261)
implying the perpendicularity condition of the velocity change .vector Av =
Aik + irkAIjk and the velocity component vector vll = i, + irk$k parallel to
the cathode surface in the neighborhood of the reversal point. With this
condition Eq. (259) takes the form

where

y = arctan( - i k / r k $ k ) (264)
The quantities ik and rk$k entering the last equation may be found with the
382 E. M. YAKUSHEV AND L. M. SEKUNOVA

help of the trajectory equation [Eq. (69)]. Differentiating it with respect to


i
time and taking into account that = f k& as well as Eqs. (242)-(245), we
obtain in the neighborhood of the reversal point

rssin($s - $k - 0,) - 7kBk


gsrk
2d)k
)] (265)

Then Eq. (264) takes the form

Thus, Eq. (262) combined with Eqs. (263) and (266) gives one the possibility to
determine the deviation of the disturbed trajectory relative to the undisturbed
one both in the magnitude and in the direction in the screen plane.
Note that it is a very common case for the source of illuminating electrons
to be on the axis (rs = 0), and the magnetic field to be absent (Bk = 0) in the
near-cathode region, for the velocity of the reflecting beam in the place of
reversal to have only a radial component (rk& = o), and for the velocity
change to come about in the azimuthal direction ( A i , = 0). It is just this case
for which the value Au is calculated for some concrete forms of nonhomo-
geneities of the sample surface in a number of papers (e.g., Sedov et al.,
1968a,b; Sedov, 1970). It is evident that these calculations are also suitable
for the case when the source of illuminating electrons is far from the axis;
however, now Av denotes the change of the velocity component parallel to
the cathode surface and not that of the radial velocity. In any case, if the
velocity change Av, caused by nonhomogeneities of various types, is known,
then Eq. (262) combined with the expression for magnification [Eq. (253)],
gives one the possibility to calculate the contrast of a shadow image in any
mirror. Here, besides electrostatic fields, a magnetic field may also take part
in forming an image; in the general case, the cathode surface may be curved,
and the source of illuminating electrons may be located outside the symmetry
axis of the system.
As follows from Eq. (262), the deviation of the disturbed trajectory from
the undisturbed one, Arb, in the plane of the microscope screen equals

Arb = K Av/kJ& (267)


where
ELECTRON MIRRORS AND CATHODE LENSES 383
The last quantity has a sense of the microscope sensivity to relatively small
changes of the electron velocity caused by the disturbing nonhomogeneities.
This increase of sensivity is always desirable. However, one should keep in
mind that such an increase involves deterioration of the microscope
resolution, because in this case identification of the micro-objects located
close to each other becomes difficult. In reality, if two identical objects are at a
distance Ark < Arb/M, where M is the microscope magnification, then their
shadow images are superimposed. When the resolution is high, the case seems
to be optimal when
Arb/M 2: 2 (269)
where I is the limiting resolution of the microscope.

D . Impact Parameter and the Circle of Secondary Emission

The effect of disturbance is determined, to a great extent, by the distance h


(see Fig. 6) between the cathode surface and the reflecting trajectory in the
neighborhood of the analyzed point. Taking into account third-order values,
this distance may be calculated on the base of Eqs. (12) and (234).In doing so
one should, in general, substitute the value 5 = [, into Eq. (12) corresponding
to the intersection of the reflecting trajectory and the normal to the cathode
surface. However, according to Eq. (239),5, is of fourth order; therefore, in the
adequate approximation one may write

where

While deducing the last relations, besides Eqs. (12) and (234), Eqs. (41)-
(46),(232),and (242)-(245), as well as the conditions in Eqs. (35),(36),and (37),
imposed on the functions p , q near the cathode surface, are taken into account.
For a point source of monoenergetic electrons ( E = 0,d # 0) Eq. (270)
determines the surface which represents the set of the reversal points of
reflecting trajectories. In the adopted approximation this surface is a
paraboloid of revolution with the axis parallel to the z axis and intersecting the
cathode surface at the point uk = u, = r,ei*". As seen from Eq. (270), the
quantity h, in the general case, is not a constant for all points of the cathode
surface. The only exception is when there is no magnetic field near the surface
of the cathode (Bk = 0), and the source of illuminating electrons is located in
384 E. M. YAKUSHEV A N D L. M. SEKUNOVA

the plane z = z c , passing through the center of curvature of a mirror. If the


+
negative bias 6 is given to the cathode, so that ( = E 6 < 0, then electrons
never reach cathode surface.The minimal distance at which they can approach
the surface equals ISl/4;. With positive bias (6 2 0) some part of the electrons
reach the cathode, forming the circle of absorption. The coordinates of the
circle center coincide with r, and II/. and are determined by Eq. (271),while its
radius, as follows from Eq. (270),equals

= 4gs(44;’p: s++k2Bzg:
E YI2

Note that the center position of the circle of absorption does not depend on
the bias voltage 6.

E. Aberrations of a Shadow Image

When deducing the formulas describing the first-order electron-optical


properties of a mirror microscope operating in the shadow regime, the source
of illuminatingelectrons was supposed to be pointlike, and the influenceof the
terms of third order in r and of first order in rE on forming a shadow image was
assumed to be small. The account of these effects determines aberrations of a
shadow image.
We shall consider, first, aberrations associated with a finite-sizedsource of
illuminating electrons. Suppose that in the plane z = z, the source represents a
circular spot of a radius p , , and its every point may be considered as a self-
dependent point source of illumination. If the quantity us is given a sense of
the coordinates of the spot center, the coordinates of any spot point upmay be
written in the form up = us + peinp,and 0 5 0,5 2n, p 5 p , . In order to take
into account the influence of finite-sized source on the microscope resolution,
one should substitute up instead of us into Eq. (246). Then for the reversed
branch of a trajectory we obtain

U + -us
9 uk
- -(pg, + p,g) = -
9
-peiRp (273)
gs gs ss

From the equality obtained it is seen that after reversal the cross section of the
reflecting beam by the plane z = const is a circle. Thus, single-valued
conformity between the points of an object and its shadow image is broken in
this case. In the screen plane the radius of the spot of diffusion equal to

represents the aberration determining the smearing of the shadow image


caused by the finite-sized source of illuminating electrons. The aberration
ELECTRON MIRRORS AND CATHODE LENSES 385
limits the microscope’s resolving power up to the value

For a specified aperture angle of the electron reflecting beam yp = psgb/gs,the


dependence of microscope resolution on the positions of the source and the
screen is determined by the equality

Now we turn to a consideration of geometrical aberrations of third order


in rand of chromatic aberrations of the order of rE. Using Eqs. (72),and (242)-
(245),we represent the coordinates of the point of intersection of the reflecting
trajectory reversed branch and the plane of a microscope screen, z = zb,
located beyond a field, in the form
rbeiRb= u b + AM (277)
where 0, gives the first-order coordinates of the point determined by Eq. (251),
and AMis the total aberration in the plane of a shadow image

Bk
- i-a(aP* + - %Pb + Bsb)a - x s - 4aPS + /?s,)..l
8 9s
- - bgb)wb + x b + i(aPb - Pgb)lcb
Using Eqs. (41), (47), and (63), in which one should set a = a,b = /?,and
Eqs. (82), (232), and (233), one may reduce Eq. (278) to the form
AM= A u ( ~ +
’ AM< (279)
where the quantity
AM(^) = a2a*L, + .(a/?* + a*B)L2 + aPP*L, + C~CC*PL,
+ /?(aP* + a*B)L, + PZP*L, + ia(aP* - a*P)L,
+ $(a/?* - a*P)L, (280)
determines the total geometrical third-order aberration, and
AM< = SaL, + 5PL1, (281)
the aberration associated with the initial energy spread of illuminating
electrons E and the bias voltage 6. The coefficients from L , to L,, entering the
386 E. M. YAKUSHEV AND L. M. SEKUNOVA

above two relations equal

Here M is the microscope magnification, and


= (l/gs)(Pbgs - Psgb)
The quantities from J1to J19are integrals which do not depend on the position
of the illuminating electron source, z = z,, and of the microscope screen plane,
z = Zb
ELECTRON MIRRORS AND CATHODE LENSES

J4 -- - 7 j "1 E ( Q 4 g 2 - 8Q;v3)dz
164k 0 fi
388 E. M. YAKUSHEV AND L. M. SEKUNOVA

Due to the fact that the integrands equal zero beyond the mirror field, the
integration region is extended to infinity. There are a number of intercon-
nections between the integrals entering the expression for geometrical
aberrations
4; (312)
-iJ4 - J 7 ) = qZs?;s - q i s q 3 s = q 2 b q ; b - ?;by/3b
4 a
4;
-(JS - J 6 ) = qlsq;s - q i s q 3 s = q l b q ; b - Y ] ; b r / 3 b (313)
4 a

The proof of the given interconnections may be made in the same way as done
in Section I11 for substantiating Eqs. (1 37),(1 38), and (1 39). In a similar way we
eliminate the double integration in Eqs. (293)-(311) associated with the
presence of the quantities qq, ql, q,, q 3 ,q4 in the integrands. On carrying out
these transformations and substituting the values u, /3 into Eqs. (280)and (281),
according to Eqs. (244)and (245), we find, taking into account Eqs. (312)-(3 16)
Ad3) = (A, + ia1)u,2u: + ( A , + ia,)uku:us + ( A 3 + ia3)u,2ur
+ (A4 + ia,)u,usu,* + (AS + ia5)u:u: + + ia,)u:u,* (A6 (317)
AUg = (A7+ ia7)ukt + (A8 + ia8)ust; (3 18)
where

- 3Mf1,
P2 P3
-N f l , (319)
gs gs
ELECTRON MIRRORS AND CATHODE LENSES 389

(333)
390 E. M. YAKUSHEV AND L. M. SEKUNOVA

and

I, =7
164k
jmf i
0
%s2(Q4p + 4 Q Y ) + 8p”(44d2 + Q2g2)]dz (338)

+ Q2g2)]dz
+ 8q”(4$~g’~ (340)

(343)

(344)
ELECTRON MIRRORS AND CATHODE LENSES 39 1
Thus, in the general case, geometrical aberrations of a mirror microscope
operating in the shadow regime are determined by twelve aberration
coefficients, while chromatic aberrations are determined by four coefficients;
their magnitudes depend both on the position of the illuminating electron
source plane and on the position of the microscope screen plane. To calculate
all these coefficients, together with a number of quantities characterizing a
mirror in the first approximation-zF, &,pb,gb, as well as the quantities
determined by the positions of the source and the screen

Ps = ( z s - zF + f)ph? 9s = ( z s - zF - f)gb,
Pb = (zb - zF + f)Pb, gb = (zb - zF - f)gb
it is necessary to calculate the quantities from I , to I,,, depending only on the
distribution of the electric and magnetic fields along the mirror axis. If the
magnetic field is absent, the coefficients from a, to a8 turn out to be zero, and
the geometrical/chromatic aberrations are characterized by six/two aberra-
tion coefficients, whereas I , = 1 8 = I , = I,, = I , , = 0. As can be seen from
Eqs. (317) and (318), most of the aberrations are caused by the fact that the
point source of illuminating electrons is far from the mirror symmetry axis.
If the source is on the axis (us = 0), then
A d 3 )= ( A , + ia,)u;u: (349)
AUc = (A7 ia7)ukt (350)
In order to investigate in more detail the image deformations associated
with geometrical aberrations, we introduce, instead of the rotating cylindrical
coordinate system, I , R, z, the rotating Cartesian coordinate system, x , y, z, the
x axis being drawn through the source point. In this system us = x s , uk =
x k + iy,, and A d 3 ) = A x + i Ay. Then from Eq. (3 17) it follows that

Ax = A I X k ( X ? + Y,") - a l y k ( x i + Y:) + A Z ( x i + Y i ) &


+ A3(X; - y i ) x s - 2a3xkykXs + A4XkXf - a4ykx:

+ A5XkXf + a5ykxf + A,X: (351)


AY = A l y k ( x f + y.?) + a l X k ( X i + Y : ) + + y,')xs
+ 2A,xkykx, + a,(x; y,')xs + A4ykx: +
- a4XkXf

- A 5 y k x l + agxkxf + a 5 x i (352)
Let a test net be drawn on the cathode surface, along the lines of its
intersection with the planes parallel to x z and yz, and its projection on the
plane z = 0 represent a square net with constant-size steps along the x and y
axes. Using Eqs. (351), (352), (251), and (253), we consider the image of the net
392 E. M. YAKUSHEV AND L. M. SEKUNOVA

in the presence of this or that type of aberrations. If all the aberration


coefficients are zero, the mirror gives an adequate, magnified image of the net

The image of the test net will be a square net with an M-fold increased step.
Now we assume that all aberration coefficients but A, equal zero. In this case
the quantities Ax and Ay determine deformation characterized by the
aberration coefficient A,. The test net image in the presence of this
deformation is shown in Fig. 7(a). Such a type of aberration represents
isotropic distortion. Anisotropic distortion [Fig. 7(b)], occurring only in the
presence of a magnetic field, is characterized by the coefficient a,.
Deformations of these types are also inherent to electron lenses. All other
types of geometrical aberrations characteristic of a shadow image have no
analogue among aberrations of electron lenses. In Fig. 8(a) the image of a test
net is given for the case when only one coefficient, A 2 , is not zero. The
displacement of the net node image in the rotating coordinate system takes
place in the direction of the point-source shift from the symmetry axis of a

FIG.7. (a) Isotropic ( A , # 0) and (b) anisotropic (al # 0) distortion of the image
ELECTRON MIRRORS AND CATHODE LENSES 393

FIG.8. Image deformationscaused by aberrationsdetermined by (a) the coefficient A 2 and


(b) the coefficient a2.

mirror. In the perpendicular direction a similar node shift occurs in the


presence of a magnetic field when all coefficients, besides u 2 , equal zero
[(Fig. 8(b)]. Deformation of the test net image associated with aberrations
characterized by the coefficients A , and a, is represented in Fig. 9. The
presence of the coefficients A , and a, does not lead to an image deformation.
Aberration associated with A, results in a uniform and isotropic change of the
image scale, and aberration caused by a, gives a change of the image rotating
angle. Aberrations caused by A, and u5 lead to the stretching of the image
along a certain direction and to the straining across the image (Fig. 10).If all
aberration coefficientsbesides A , and u6 equal zero, a certain shift of the entire
image by Ax = A,X: and Ay = a6x$ occurs without changing the image scale
or deformation. Note that, although the presence of geometrical aberration
does not affect the resolution of a mirror microscope, taking them into
account is necessary for a strict interpretation of a shadow image.
Now we shall consider chromatic aberration of an electron mirror
microscope associated with the bias voltage 6 and the spread in the initial
394 E. M. YAKUSHEV AND L. M. SEKUNOVA

FIG.9. The image deformations caused by aberrations characterized by (a) the coefficient
A, and (b) the coefficient a,.

energy E of illuminating electrons. The quantity 5 entering Eq. (318) equals


+
t = E 6. We shall consider the aberrations associated with 6 and E sepa-
rately. First we set E = 0. Then the aberrations associated with the bias volt-
age in the rotating Cartesian system take the form
Ax, = A7Xk6 - a7yk6 + A8xS6 (355)
Ay, = A,yk6 + a7x,6 + a,x,6 (356)
Hence, the coefficient A , characterizes the change of the image scale; a 7 ,the
angle of rotation; and A 8 and as, the displacement of the entire image along
the x and y axes, respectively. Since the bias voltage is a constant, the
aberrations associated with it do not cause image diffusion. If 6 = 0 while
E # 0, Eq. (318) determines the diffusion of image points associated with the
initial energy spread of the illuminating electrons. The value of the diffusion in
ELECTRON MIRRORS AND CATHODE LENSES 395

FIG.10. (a) The image scale change at A, # 0; (b) image rotation angle change at a4 # 0.

the rotating Cartesian coordinate system is determined by the equalities


Ax&= (A7xk - a7yk + A8Xs)E (357)
A Y ~= ( A 7 y k + a7xk + a8Xs)E (358)
These equalities mean that, in the plane of a microscope screen in the presence
of the initial energy spread of the electrons, an object point is depicted as a
stroke whose length depends both on the position of the source point and that
of the point on the cathode. The length divided by magnification determines
the limiting resolution of a microscope associated with the initial energy
spread

+ A ~ X S+)(A7yk + + a g x ~ ) ~ (359)
&
= -4(A7x, - ~ a7xk
M
396 E. M. YAKUSHEV A N D L. M. SEKUNOVA

which may be achieved with a sufficiently small size of the illuminating


electron source, if the value A, determined by Eq. (275) can be neglected. From
Eq. (359) it follows that, in the general case, on the cathode surface there is only
one point with coordinates

the shadow image of which is free of chromatic aberration. If the source is


located on the mirror axis ( x , = 0), then this point is the central point of the
cathode surface, xk = yk = 0. In order to make a shadow image achromatic
throughout the entire viewing field of a microscope, a simultaneous satisfac-
tion of three conditions
A, = 0, U, = 0, X, = 0 (362)
is enough. For any distribution of fields along the mirror axis the quantity A ,
may be set equal to zero by a special choice of the microscope screen position,
z = z b . This position will be found on the basis of Eq. (325), setting its right-
hand side equal to zero
k2Bt
~- Pb z, - zc
I12 -7 - 113

(S
zb - z B
-- _- 84k2 S a Zs - ZB
(363)
f +Ill + -
zs - zF + j& zs - zC

Pa z, - zB g; (2, - ZB)2'13

From Eq. (333) it is clear that the second condition, a, = 0, may be satisfied
only by an appropriate choice of the field distribution along the mirror axis. In
particular u, = 0 if B = 0. The third condition, x, = 0, is satisfied when the
source of illuminating electrons is located on the mirror axis. Thus, all three
conditions [Eq. (362)] may be satisfied simultaneously. However, realization
of these conditions does not mean that the limitations for the initial energy
spread of illuminating electrons is eliminated completely. There is a limitation
caused not by the requirements of resolution, but by the necessity to obtain
the image contrast which, evidently, dissappears if e / 4 ; >> 1, where 1 is a
characteristic size of analyzed nonuniformities (of topography or microfields)
in the direction perpendicular to the cathode surface. The influence of the
initial energy spread on the image contrast is considered, in particular, in the
work by Sedov and Nazarov (1972); however, it should be noted that the
effects of image diffusion and contrast forming in the screen plane are
propagated, as seen from Eqs. (262) and (318), in opposite directions. For
ELECTRON MIRRORS AND CATHODE LENSES 397
instance, when considering in an electrostatic mirror the meridional motion of
a reflecting beam originating from a source located on the axis, these effects
occur in mutually perpendicular directions.
Note that when determining the electron-optical properties of a mirror
operating in the regime analogous to a light shadow microscope, we do not
impose the requirement of large magnification which is characteristic of
microscopes. This allows one to apply the above-stated theory for all electron-
optical devices using an electron mirror in the indicated regime -for example,
in a mirror image converter (Artamonov et al., 1968)and a brightness amplifier
with a mirror system (Kasper and Wilska, 1967).

V. CATHODE
LENSES

Cathode lenses are widely applied in electron devices of various types:


electron guns, image converters, emission electron microscopes, and so on. In
this connection, it is of great interest to investigate these lenses theoretically,
including their aberrations. Study in this field, started with well-known papers
by Recknagel(l941,1943) and Artzimovich (1944), have a long history, and, in
total, there is a rather large number of appropriate works. Recently
electrostatic cathode lenses were studied by Ignat’ev and Kulikov (1978),
Dodin and Nesvizhski (198 l), Nesvizhski (1984),and Takaoka and Ura (1984).
The works by Monastyrski and Kulikov (1976, 1978), Kulikov et al. (1978),
Monastyrski (1978a,b), and Ximen Ji-ye et al. (1983) are devoted to the
investigation of cathode lenses with combined electric and magnetic fields.
However, the involved assumption on a finite (although small) value of the
longitudinal component of the electron initial velocity contradicts the specific
nature of a cathode lens in which, in principle, the electrons having small, even
zero energies take part in image formation.
In a cathode lens the particle initial energy E and the distance r from the
symmetry axis are assumed to be small. Although these quantities are different
in nature, the initial energy, if it is not zero, may be considered (Artzimovich,
1944)as values of second order in r. When stating the theory of cathode lenses,
we shall follow this statement, and the values proportional to E are assumed to
be of second order in r. Then the value orders will be determined by their
powers of r; e.g., re is a third-order value.

A . Determination of Arbitrary Constants

The origin of the cylindrical coordinate system (r, rl/, z ) used below will be
placed in the central point of the cathode surface. The z axis will coincide with
398 E. M. YAKUSHEV A N D L. M. SEKUNOVA

the principal optical axis of the lens, while its direction is from the cathode to
the anode. In doing so, in all the equations derived in Section I1 we set
zk=o, a-1 (364)
As long as the potential of the cathode surface is set equal to zero, the
equation ofthis surface is
cp(r,z) = 0 (365)
Here in the general case the cathode may be not flat. The radius of curvature of
the central region of the cathode surface is related to the field distribution in
the near-cathode region by the well-known relation
Rk = -24;/4: (366)
It is assumed that Rk > 0 for the convex cathode and Rk < 0 for the concave.
The charged-particle trajectory in the region far from the cathode surface is
described by Eq. (72), which is valid with an accuracy up to values of third
order. However, the constants a, b entering this equation are to be determined
proceeding from the conditions on the cathode surface. In this case it is
necessary to use the set of equations, Eqs. (69) and (70), describing the
trajectory of this particle also depending on the variable ( in the near-cathode
region, including its point of emission. First we define the value [ = [ k
corresponding to the particle emission point. At this point
if = k2&Cos2 (367)
where & is the angle between the particle initial velocity and the z axis. On the
other hand, on the grounds of Eqs. (1 1) and (12), we can write

(3681

From the last two equations we obtain

whence, neglecting the values of the order of smallness higher than the third,
we obtain
[k = (E/4;)CoS2
$k (370)
While deriving this, it is taken into consideration that, according to
Eq. (24), r] is a second-order value.
When determining the arbitrary constants a, b, it is necessary to take into
account that the direction of the normal n to the cathode surface at the point of
ELECTRON MIRRORS AND CATHODE LENSES 399
particle emission is the natural direction specifying symmetry of the emitted
particle angular distribution. So, when studying cathode lenses with curved
emitting surfaces, it is reasonable to specify the initial conditions for particles
with respect to the normal by two angles: Y k , O k (see Fig. 11). Y k is the angle
between the normal o and the particle initial velocity v k . This angle can have
any magnitude over its entire range: - 4 2 I y k I 4 2 . a k is the angle between
the two planes, one of which contains the point of the particle emission and the
z axis while another involves the vectors v k and n. It also can have all the
magnitudes of its range: 0 I 0, I 211. The connection between the angles
$ k , y k , a k has the form

cos 8, = cos y k - ( r k / R k ) cos w k sin y k (371)


When deriving the last equation, it has been taken into account that in the
approximation considered the angle between the normal n and the z axis
equals r k / R k .
At the point of particle emission Eq. (69) gives
rke-iN5k) =
ap(ck) + b ( c k ) + x ( c k ) + iK(ck)[ap(ck) + bg(ck)l (372)
Herein, it is assumed that the plane $ = 0 contains the point of emission. It
does not restrict the generality of the subsequent conclusions because of the
system’s axial symmetry.
On differentiating Eq. (69) at this point with respect to time and using
Eq. (1l), we obtain
[ik + irk& -ikrk6’(ck)m]e-ie(ck)

= k{ap’(ck) +k’(ck) -k X’(ck) + i[ap’(Ck) + bg’(ck)IIC(ck)


iCap(ck) + bg(ck)lK’(ik)) (373)
It is necessary to solve Eqs. (372) and (373) with respect to the constants
a, b. The solution is to be carried out with an accuracy up to values of third

FIG. 11. Scheme for determination of the initial conditions of motion in a cathode lens
400 E. M. YAKUSHEV AND L. M. SEKUNOVA

order. For convenience of calculation we shall transform these equations into


the form

(374)

In doing so Eq. (38) is taken into account. The quantities + k , r k $ k entering


Eq. (375) can be expressed in terms of the particle’s initial energy E and
angles of emission Y k ,

i k $k &sin
=k Yk sin o k (377)
The quantities e([,‘),
g ( c k ) in the approximation considered can be written,
using Eqs. (370) and (371), as follows:

Substituting Eqs. (370), and (376)-(379) into Eqs. (374) and (375) and
producing the expansion in small values, we find
ELECTRON MIRRORS AND CATHODE LENSES 40 1

2J.5 . Bk
a=rk, ,!3- elmksin y k - irk-
4; 24;
As the constants a , b are now known, further, when investigating
aberrations sufficiently far from the cathode surface, one can use a more simple
equation [Eq. (72)] without involving the set [Eqs. (69) and (70)] which is valid
everywhere including the near-cathode region. A study of trajectories in the
region far from the cathode surface, where 4 ( z ) >> E, is of the greatest practical
interest, as this region contains the place of smallest cross section of a beam in
electron guns and the image plane in image converters and emission electron
microscopes.

B. Near-Axial Approximation

Cathode lenses can be used for solving problems of two kinds: first, to
produce an image of the cathode surface and, second, to focus the electron flux
emitted by the cathode surface into a spot of a small diameter-a crossover. In
both cases the trajectory sections far from the cathode surface are of interest.
Let us consider, first, near-axial trajectories, preserving in Eq. (72) the
terms of first order in I and substituting into Eq. (72),instead of the constants
a, b, their values a, p defined in the first approximation according to Eqs. (382).
Then the trajectory equation becomes

rein = U = ap(z) + Pg(z) = rk


402 E. M. YAKUSHEV AND L. M. SEKUNOVA

We shall consider the cathode surface as the object plane. According to the
definition of the function g(z) [Eq. (33)], g(0) = 0, because 4(0)= 0. Then, in
accordance with Eq. (383), the image plane, z = zb, will be determined by the
equation
dZb) =0 (384)
In this plane, as seen from Eq. (383), the cathode lens field forms a similar
electron-optical image of the cathode surface. The cathode lens magnification
is constant in all directions and equals
=dZb) (385)
It is necessary to keep in mind that the magnetic field of the cathode lens, in
accordance with Eq. (73), leads to the rotation of the image about the z axis at
an angle y b equal to

All the electrons emitted by the cathode surface can take part in the forma-
tion of the electron image of this surface in the approximation considered.
The change of the emitted particle’s initial energy E from zero to its maximum
value E,(O IE IE,,,), as well as the change of the initial angles of the particle
emission over the entire range ( I r k 1 In/2, lWkl I274 causes neither defor-
mation nor diffusion of a point-to-point image.
Now we shall consider the peculiarities of the electron image formation
in the presence of a limiting diaphragm. Introducing the latter becomes ne-
cessary in those cases when a cathode lens is used to receive the image of
(geometrical,electrical, or magnetic) micrononhomogeneitieson the cathode
surface, which deform the energy or the angular distribution of the particles
emitted by every point of the cathode surface. In these cases the electron flux
diaphragm is used for obtaining an image contrast. Visualization of the
cathode surface potential relief is an example of a problem of this kind. In
this case the particle’s initial energy E is considered as a function of co-
ordinates rk, t,bk of the point on the cathode surface. When using a diaphragm,
the portion of the electron flux emitted by every point of the cathode surface
is removed by an operating diaphragm, and in the image plane z = zb of the
cathode lens a brightness picture corresponding to the function E = & ( r k , $k)
is formed.
Let the diaphragm be the equipotential surface p(r, z ) = 4 ( z D ) in
, which a
coaxial (with the axis z ) round aperture of small diameter d is cut. The plane
z = z D tangent to this surface will be called the diaphragm plane. The equa-
tion of the trajectory traversing the two points-of the object and of the
ELECTRON MIRRORS AND CATHODE LENSES 403
diaphragm-has the form

where pD = P(Z,), gD = g(z,) # 0, and uD = rDeinDare the coordinates of the


intersection point of the trajectory and the diaphragm (rD 5 4 2 ) .
When using Eq. (387), one should keep in mind that now not all electrons
emitted by the cathode surface take part in the formation of its electron image,
because only electrons whose initial conditions, according to Eq. (383), satisfy
the nonequality

can pass through the limiting diaphragm. In the general case, Eq. (388)
determines the phase volume of the beam passing through the diaphragm.
If the field of view of the emission system is determined by the electron
emitted normal to the cathode surface (&sin y k = O), then, according to
Eq. (388), for this field of view

(389)

From this inequality it follows, in particular, that the field of view reaches its
maximum value if there is no magnetic field (Bk = 0) in the near-cathode
region, while the diaphragm plane z = zD contains the ray crossing point
satisfying the equation
P(zD) = (390)
In this case, as follows from Eq. (388), the variation range E and yk of the beam
passing through the diaphragm does not depend on the emitting point
coordinates and is determined by the inequality

The problem the formation of the electron flux emitted by the cathode
surface as a small spot (a crossover) is, as usual, solved using electric fields. As
for an electrostatic cathode lens, it is of interest to determine the beam cross
section in the neighborhood of the crossing point (z = zD)satisfying Eq. (390).
Here the area of the beam cross section does not depend on the cathode size
and can be made significantly smaller than the cathode surface area. Thus the
crossing point, z = zD, determines the crossover position on the principal
optical axis of an electrostatic cathode lens. In the presence of the magnetic
404 E. M. YAKUSHEV A N D L. M. SEKUNOVA

field the formation of a similar crossover may occur, if the magnetic field near
the cathode surface equals zero (Bk= 0).

C . Aberrations of Spatial Focusing

In order to analyze aberrations of a cathode lens in the region far from the
cathode surface, it is necessary to represent Eq. (72) as a power series depicting
an explicit dependence of the trajectory on the initial conditions, rk, E, y k , a k ,
characterizing the particles leaving the cathode. In accordance with this, we
substitute into Eq. (72)the functions ~ ( z )~ ,( z )and
, JC(Z), according to Eqs. (41),
(47), and (63), as well as the constants a, b, according to Eqs. (380) and (381),
and then combine therein the terms of third order in powers of a and fl. Then,
neglecting terms of orders higher than third and taking into account Eqs. (382)
and (383), we find

( + i7g(z)
Bk
u = rein = U - p ( z )
24k

+ a 3 [ L l ( z )+ ill(z)] + C C ~ ~ ~ * +[ Li12(z)]
~(Z)
+ a 2 f l [ L 3 ( z )+ i13(z)] + af12[L4(z)+ i14(z)]
+ a f l P * [ L 5 ( z ) + i15(z)1 + f12fl*[IL6(z) f ii6(z)i

+ ae[L,(z) + il,(z)l + P E [ L f j ( Z ) + iZ,(z)] (392)


This equation describes the trajectories of particles in the region far from the
cathode surface under the specified initial conditions on the cathode. When
deriving Eq. (392), the following notations are used

(393)

k2B: k4B:)y
+-8&4; ~

644i2

L4(4 = -(%+ JJP + (J, --


ELECTRON MIRRORS AND CATHODE LENSES 405

(397)

+ 4Q;p‘) + 8p“(44p” + Q z p 2 ) ]dz (410)


These integrals are obtained as a result of transforming Eqs. (42)-(46), (57),
(64), and (65)-(68). Double integrals entering Eqs. (57), and (64)-(68) are
reduced to single integrals via integration by parts just as was done in Section
I1 for the proof of Eqs. (137)-(139).
ELECTRON MIRRORS AND CATHODE LENSES 407
It can be seen from Eq. (392) that aberrations of a cathode lens in an
arbitrary plane z = const located sufficiently far from the cathode can be
represented as the sum of the terms
Au = u - lJ = A#(') + Au(3) (422)
where A d 2 ) is the second-order aberration, while A d 3 ) is the overall third-
order aberration.
The second-order aberration in an arbitrary plane is described by the
expression

(
A d 2 )= - p(z) + i7g(z)
kBk
2+k ) i k
sin 2yk
-+Ok (423)

In the image plane of a cathode lens, z = z b ,this aberration takes, according to


Eqs. (384) and (385), the familiar form
&
A@ = - M-eiWk sin 2yk (424)
4L
Diffusionof the point image caused by this aberration should be defined using
the energy and the angular spread of the particles emitted by the cathode. It is
worth noting that the magnitude of this diffusion referred to the emission
system magnification does not depend on the magnetic field. We shall find the
radius of the diffusion circle associated with the second-order aberration,
setting E = E,, = n/4 in Eq. (424)

Ar(2)= M&,,,/q5; (425)


The overall third-order aberration of a cathode lens in the image plane is
determined by the expression
Au(3) =
a3 [L1(zb) + i l l ( z b ) l + a 2 P * [ L 2 ( z b ) + ilZ(zb)]
+ a2fl[L3(zb) + iz3(zb)l + a P 2 [ L 4 ( z b ) + i14(zb)l
+ aflfl*[L5(zb) + i15(zb)l + P Z P * [ L 6 ( z b ) + i26(zb)l
+ a & [ L 7 ( Z b ) + i17(zb)l + P E [ L 8 ( Z b ) + i18(zb)l (426)
According to Eq. (382), we substitute the values a, j? into this expression.
Then, after reducing the like terms, we obtain
+ (5- if,)rk&e2i"ksin2yk
Aui3)= Gr&&eiWksin3yk
+ [2(~, + if,)]rkesin' + (c,+ O,)r,"&eiok sin y k
Yk

+ (c,+ icr)r~&e-imksinyk+ (E, + ie,)rf


+ ( K +~ikrm)&&ei"'*sin
~ ykcos2y k + ( K ~+, ik,)&rk (427)
408 E. M. YAKUSHEV A N D L. M. SEKUNOVA

where

Here p;.gb,Bb,4b are the values of the functions p',g',B,#~ at z = zb.


Equation (427) describes various kinds of deformation of a cathode surface
image, when using a nondiaphragmed beam.
ELECTRON MIRRORS AND CATHODE LENSES 409
Introduce a limiting diaphragm, placing it into the plane z = zD.Using the
first-order equation [Eq. (387)], we express the quantity b in terms of the
coordinates of the cathode point and the diaphragm uD = rDeiRD

b = (uD - rkPD)/gD (440)

Substitute this value of /3 and ct = rk into Eq. (426). Then, combining the terms
in the appropriate way, we obtain
Auk3' = (CD + ibD,)uk~g + (FD - ifD)rkuk
+ [2FD + FDm + i(2fD + f D m ) l r k U D U g
+ (CD + DD + idD,)riUD + (CD + icD)rZug + (ED + ieD)r;
+ (KID + ikDm)&UD+ ( K ~ +D ikD)&rk (441)

where

(443)

(446)

(447)
410 E. M. YAKUSHEV AND L. M. SEKUNOVA

(453)
(454)
(455)

The coefficients G D , F D , C,, OD, E D , f D D , c D , e D determine the geometrical aber-


rations: spherical aberrations, coma, astigmatism, the field curvature, distor-
tion, anisotropic coma, anisotropic astigmatism, and anisotropic distortion,
respectively, also characteristic of ordinary electron lenses (Glaser, 1952).The
coefficient K I D determines aperture chromatic aberration, while K Z D and
kD are responsible for the changes in magnification and the angle of the
image turning, respectively, associated with nonzero energies of the particles
leaving the cathode. Chromatic aberrations of this kind are inherent in
ordinary lenses as well. The other four kinds of geometrical aberrations
(FD,, bDm,fDm,dD,) and one kind of chromatic aberration (kD,) are not to be
found in ordinary lenses. Moreover, in combined cathode lenses they are also
absent if there is no magnetic field (Bk = 0) on the cathode surface.
As shown in Section V,B, if the magnetic field in the near-cathode region
Bk = 0, and the diaphragm plane, z = zD, contains the crossing point where
p D = 0, the value of the field of view of the emission system does not depend
on the size of the aperture in the limiting diaphragm and can be made
significantly greater. Satisfying the same conditions (Bk = 0,pD = 0)leads to a
great simplification of the expressions in Eqs. (442)-(457) for the aberration
ELECTRON MIRRORS AND CATHODE LENSES 41 1
coefficients

However, the dependence of aberration coefficients on Bk,pDcan be used in


order to decrease separate kinds of aberration. In this case, one should have in
mind that such a decrease of aberrations can result in some limitation of the
emission system’s field of view.

D . Time-of -Flight Aberrations of Cathode Lenses

When designing the electron devices intended for investigation of rapid


processes whose duration is comparable with the time of particle flight, a
problem arises associated with the time structure of the analyzed charged-
particle flux. Therefore, investigation of the particle’s time of flight in emission
systems becomes of primary importance owing to the wide application of
these systems in various kinds of electron devices.
The particle’s time of flight from the cathode surface to an arbitrary plane
z = const, according to Eqs. (11) and (12), is determined by the expression

where [k is the value of corresponding to the instant the particle goes out of
the cathode surface. This value is determined strictly-by Eq. (369), or
approximately-by Eq. (370).
Represent Eq. (459) as the sum of two integrals
412 E. M. YAKUSHEV A N D L. M. SEKUNOVA

Calculation of the first with an accuracy up to values of second order is given


in Section II,B. We shall calculate the second integral using the smallness of [ k .
In doing so values of third order in r will be neglected. In this approximation
we can write

When deducing this relation, Eqs. (366), (369), and (371) are taken into
account, and the values of order higher than second are neglected. Using
Eqs. (76)-(80), (460), and (461) and taking into account Eqs. (380)-(382), we
represent the particle's time of flight in the region far from the cathode surface,
where 4 ( z ) >> E as follows

t =T +D , m COS Y k + At(') (462)


where

is the time of flight of an axial particle with zero initial energy E = 0 from the
cathode to some plane z = const

is the first-order dispersion of the particle's time of flight along the component
uI of their initial velocity normal to the cathode surface
At(2) = T1(E/4)sin2yk-t T2lkmCOSc%ksinyk
+ T3rkJ&7sinw,sinyk+ T4r; + T5E/4 (465)
is the second-order time-of-flight aberration of cathode lenses in the plane
z = const. The coefficients of these aberrations are

(467)
ELECTRON MIRRORS AND CATHODE LENSES 413

T,(z) = -- m (470)
k Is
The quantities qs, ql,q 2 ,v 3 ,q4 entering the last expression are determined by
Eq. (42)-(46).
Equation (465) describes the time-of-flight aberrations of a non-
diaphragmed flux of particles. In order to obtain the proper expression for the
aberrations of a diaphragmed flux, one should carry out the transformations
used earlier when obtaining Eq. (465);however, one should now substitute
into Eq. (80)the value j3 determined by Eq. (440)instead of b. On making these
transformations we obtain
At‘2’ = T1Du D u g + 3 T 2 D r k ( U g + + (i/2)T3Drk(Ug - uD) + T4Drk2 + T 5 D E / 6
(471)
where

(473)

(474)

As can be seen from Eq. (464),the first-order time-of-flight dispersion of a


cathode lens is determined by the intensity of the electric field near the cathode
surface. It depends neither on curvature of the cathode surface, nor on the
electric field distribution or the presence of a magnetic field in the cathode lens.
Let us consider the second-order time-of-flight aberrations. As seen from
Eq. (47l), these aberrations are subdivided into a geometrical one, associated
with the coefficients Tin, TZD, T3,,,T4D,and a chromatic one, determined by
the coefficient T S D . The geometrical aberration associated with the coefficient
Ti, determines the spread in time of flight up to the plane z = const of the
monoenergetic particles emitted by the central point of the cathode surface
(rk = O,uD # 0). The aberration of this kind may be called the time-of-flight
spherical aberration occurring owing to the existence of the particle’s initial
414 E. M. YAKUSHEV AND L. M. SEKUNOVA

velocity component parallel to the cathode surface. The aberration associated


with the coefficient T 4 D determines the spread in time of flight, provided
Ti, = T,, = T3D= TsD = 0. The aberration of this kind depends solely on
the coordinates of the point of particle emission, and so it may be called the
time-of-flight distortion. The aberration associated with the coefficients T,,
and T3, will be called the mixed geometrical time-of-flight aberrations. It is
worth noting that in electrostatic cathode lenses T3, = 0.
The analysis of the time-of-flight aberrations of a cathode lens have been
carried out for the region located sufficiently far from the cathode surface. In
some cases it is interesting to study the peculiarities of the time-of-flight
aberration formation in a near-cathode region, where the expansion in Eq. (80)
is invalid. In these cases one should use the strict relation of Eq. (459) or its
expansion for a near-cathode region.

VI. CONCLUSION

Let us emphasize several aspects of the theory developed which may be


used for solving the modern problems of physical electronics that are not
associated directly with axially symmetric electron mirrors.
One of these aspects is a special choice of the basic set of particular
solutions of the near-axial trajectory equation. This choice is based on the
character of the equation singularity, and not on the initial conditions of
particle motion. It allows one to consider in the same way the qualitatively
different systems-electron mirrors, mirror microscopes, and emission sys-
tems. As the presence of a singular point (the particle reversal point) is the
peculiarity of electron mirrors which is not associated with the type of field
symmetry, the methodology developed for theoretical study of electron
mirrors with rotational symmetry may be applied, with appropriate modifi-
cation, to the systems which do not possess rotational symmetry.
The concept of arbitrary and guiding particles is the other aspect. This
idea, expressed by Eqs. (11)-( 18), may be extended, without serious difficulty,
into arbitrary static electron-optical systems with curved axes and relativistic
particles. Thus, the perspective for development of a general theory of
charged-particle nonstationary flux focusing and flux control by virtue of
static electromagneticfields is open. This problem refers to a poorly developed
field of physical electronics which is located at the junction of static and
dynamical electron optics. At the same time, in the techniques and practice of
physical experiments the use of nonstationary fluxes is widened, in particular,
owing to the implementation of lasers into electronics. The studies carried out
here deal with this fundamental problem only partly.
ELECTRON MIRRORS AND CATHODE LENSES 41 5
ACKNOWLEDGMENTS

We are greatly indebted to Dr. P. W. Hawkes for stimulating our writing this article.
We also wish to express our gratitude to Prof. V. M. Kel’man for his support and constant
cooperation.

REFERENCES

Artamonov, 0.M., and Komolov, S. A. (1970). Radiotekh. Elektron. 15 (l),220.


Artamonov, 0. M., Bolotov, B. B., Gerasimova, N. B., and Yudovina, G. A. (1968). Elektron.
Tekhn. 4 (S), 267.
Artzimovich, L. A. (1944). Izv. Akad. Nauk SSSR,Ser. Fiz. 8,313.
Barnett, M. E., and Nixon, W. C. (1967-1968). Optik 26,310.
Berger, C . (1983). J. Appl. Phys. 54,3699.
Berger, C., and Baril, M. (1978). Rev. Phys. Appl. 13,271.
Bernard, M. (1952). C. R . Hebd. Seances Acad. Sci. 234,606.
Bimurzaev, S . B., Daumenov,T., Sekunova, L. M., and Yakushev, E. M. (1983a). Zh. Tekh. Fiz. 53,
524.
Bimurzaev, S. B., Daumenov, T., Sekunova, L. M., and Yakushev, E. M. (1983b). Zh. Tekh. Fiz. 53,
1151.
Bruche, E., and Scherzer, 0.(1934). “Geometrishe Electronenoptik.” Springer-Verlag, Berlin and
New York (Russian transl.: Lenizdat, Leningrad, 1943).
Butslov, M. M., and Stepanov, B. M. (1978). “Image Converters and their Applications in
Research Investigations.” Nauka, Moscow.
Daumenov, T., Sapargaliev,A. A., Sekunova, L. M., and Yakushev, E. M. (1981). Zh. Tekh. Fir. 51,
1137.
Dodin, A. L., and Nesvizhski, M. B. (1981). Zh. Tekh. Fiz. 51,897.
Fortin, M., and Bad, M. (1972). Rev. Sci. Instrum. 43, 1140.
Glaser, W. (1952). “Grundlagen der Electronenoptik.” Springer-Verlag, Vienna (Russian transl.:
GIITL, Moskva 1957).
Gohl, W., Kutscher, R.,Laue, H. J., and Wollnik, H. (1983). Int. J . Mass Spectrom. Ion Phys. 48,
41 1.
Gruner, H., Gaukler, K. H., and Mollenstedt, G. (1971). Optik 33,255.
Hahn, E. (1965). Jenaer Jahrb (Jena),p. 107.
Hahn, E. (1971). Wiss. Z . Tech. Univ., Dresden 20 (2). 361.
Hottenroth, G. (1936). Z. Phys. 103,460.
Hottenroth, G. (1937). Ann. Phys. (Leipzig) [S] 30,689.
Ignat’ev, A. N., and Kulikov, Ju. V. (1978). Radiotekh. Elektron. 23,2470.
Kachiarz, J., Jaworski, M., and Zarembinski, S. (1982). Optik 61, 157.
Kasper, E., and Wilska, A. P. (1967). Optik 26,247.
Kas’yankov, P. P. (1961). Opt. Spektrosk. (Leningrad) 11,765.
Kas’yankov, P. P. (1966). Tr. Opt. Inst. im. Vavilova 33, N162.
Kel’man, V. M., Fedulina, L. V., and Yakushev, E. M. (1971). Zh. Tekh. Fiz. 41,1489.
Kel’man, V. M., Sekunova, L. M., and Yakushev, E. M. (1972). Zh. Tekh. Fiz. 42,2279.
Kel’man, V. M., Sekunova, L. M., and Yakushev, E. M. (1973a). Zh. Tekh. Fiz. 43,1799.
Kel’man, V. M., Sekunova, L. M., and Yakushev, E. M. (1973b). Zh. Tekh. Fiz.43, 1807.
Kel’man, V. M., Sapargaliev, A. A., and Yakushev, E. M. (1974a). Zh. Tekh. Fiz. 44,928.
416 E. M. YAKUSHEV AND L. M. SEKUNOVA

Kel’man, V. M., Sapargaliev, A. A. and Yakushev, E. M. (1974b). Zh. Tekh. Fiz. 44,938.
Kulikov, Yu. V., Monastyrski, M. A., and Feigin, Kh. I. (1978). Radiotekhn. Elektron. 23, 167.
Lafferty, J. M. (1947). Proc. IRE 35,778.
Leleyter, M., and Slodzian, G. (1971). Rev. Phys. Appl. 6,65.
Luk’yanov, A. E., Spivak, G. V., Sedov, N. N., and Petrov, V. I. (1968). Izv. Akad. Nauk SSSR, Ser.
Fiz. 32,987.
Luk’yanov, A. E., Spivak, G. V., and Gvozdover R. S. (1973). Usp. Fiz. Nauk 110 N(4), 623.
Mamyrin, B. A., Karataev, V. I., Shmikk, D. V., and Zagulin, V. A. (1973). Zh. Exp. Teor. Fiz. 64,
82.
Monastyrski, M. A. (1978a). Zh. Tekh. Fiz. 48, 1117.
Monastyrski, M. A. (1978b). Zh. Tekh. Fiz. 48,2228.
Monastyrski, M. A., and Kulikov, Yu. V. (1976). Radiotekh. Elektron. 21,2251.
Monastyrski, M. A., and Kulikov, Yu. V. (1978). Radiotekh. Elektron. 23,644.
Nesvizhski, M. V. (1984). Radiotekh. Elektron. 29,972.
Noda, T. (1963). J . Appl. Phys. 2, 128.
Oman, R. M. (1969). Adv. Electron. Electron Phys. 26,217.
Recknagel, A. (1936). Z . Tech. Phys. (Miinchen) 17,643.
Recknagel, A. (1937). Z . Phys. 104,381.
Recknagel, A. (1941). Z . Phys. 117,689.
Recknagel, A. (1943). Z . Phys. 120,331.
Regenstreif, E. (1947). Ann. Radioelectr. 2, 348.
Scherzer, 0.(1937). Beitrage zur Electronenoptik, 33.
Schwartze, W. (1965-1966). Optik 23,614.
Sedov, N. N. (1970). Izv. Akad. Nauk SSSR, Ser. Fiz. 34,1529.
Sedov, N. N., and Nazarov, M. V. (1972). Radiotekh. Elektron. 17,2391.
Sedov, N. N., Luk‘yanov, A. E., and Spivak, G. V. (1968a).Izv. Akad. Nauk SSSR, Ser. Fiz. 32,996.
Sedov, N. N., Spivak, G. V., Petrov, V. I., Luk’yanov, A. E., and Rau, E. I. (1968b). Izv. Akad. Nauk
SSSR, Ser. Fiz. 32,1005.
Sedov, N. N., Labutin, N. I., and Wolk, Ch, P. (1971). Prib. Tekh. Eksp. No. 5 , p. 211.
Sekunova, L. M. (1977). Zh. Tekh. Fiz. 47,2030.
Sekunova, L. M., and Yakushev, E. M. (197Sa). Zh. Tekh. Fiz. 45,723.
Sekunova, L. M., and Yakushev, E. M. (197Sb).Zh. Tekh. Fiz. 45,732.
Septier, A. (1966). Adv. Opt. Electron Microsc. 1,204.
Slowko, W. (1975). Opt. Appl. 5 (2), 27.
Slowko, W., and Mulak, A. (1973). Opt. Appl. 9 (2), 37.
Smirnov, V. I. (1974). “Kurs vysshei matematiki,” Vol. 2. Nauka, Moskva.
Sturrock, P. (1955). “Static and Dynamic Electron Optics.” Cambridge Univ. Press, London and
New York (Russian transl.: Inostrannaya Literatura, Moskva 1958).
Takaoka, A., and Ura, K. (1984). Optik 68, 107.
Vasil’ev, M. A,, Tchenakin, S. P., and Tcherepin, V. T. (1974). Prib. Tekh. Eksp. No. 4, p. 224.
Vorob’ev, Yu. V. (1959). Izv. Akad. Nauk SSSR, Ser. Fiz. 23,694.
Wechsung, R., Hillenkamp, F., Kaufmann R., Nitsche R., and Vogt, H. (1978).Mikroskopie 34,47.
Wiskott, D. (1956). Optik 13,463.
Ximen Ji-ye, Zhou Li-wei, and Ai Ke-cong (1983). Optik 66, 19.
Zhigarev, A. A., and Shamaeva, G. G. (1982).“Electro-ray and Photo-electron Devices.” Visshaya
Shkola, Moskva.
Zworykin, V. K., and Morton, G. A. (1954). “Television.” Wiley, New York (Russian transl.:
Inostrannaya Literatura, Moskva, 1956).
Zworykin, V. K., Morton, G. A., Ramberg, E. G., Millier, J., and Vance, A. W. (1945). “Electron
Optics and Electron Microscope.” Wiley, New York.
Index

A ATD, see Along-track distance


ADF, see Airborne direction finder ATRK, see Along-track error
ADF, see Automatic direction finder Attitude of the aircraft, 30
ADS, see Automatic delay stabilization Autocoupled guidance, 84, 90
Aeronautical Radio, Inc., 74, 75, 107, 122 Automatic delay stabilization, 227
ARINC Characteristics, 122 Automatic direction finder, 59, 60
AFCS, see Aircraft flight control system Autopilot, 79
AGC, 63 autopilot feedback control, 79
Aharonov-Bohm effect, 302 Autopilot error
hypothesis, 308 definition, diagram, 105
Airborne direction finder, 19 Availability of a system, 60
Aircraft flight control system, 2 , 4, 30, 39 AWOP, see All Weather Operation’s Panel
frequency response, 100
guidance loop, 81 B
stability margin, 86 Bawdsey Research Station, U.K., 8
Air navigation guidance and control, 76-106 Beacon concept, 5-9
Air navigation systems, principles of, 15-45 diagram, 6
2D, 17-18 interrogation, 7
Airspace structure transponder, 7
diagram of, 49 Beacon performance curve, 158
Air traffic control radar beacon system, 58, Bell’s inequalities, 245-335
198 direct proof of QM nonlocality, 297-300
Air traffic control system, 1, 48-58 generalized BI for dichotomic variables,
and its navigational aids, 45-76 280-295
Secondary Radar, 5 original Bell inequality, 276-280
special radio navaid techniques, 71-76 Wigner’s version of Bell’s theorem, 279-
surveillance system, 3 280
All Weather Operations Panel, 119, 172 recent developments, 300-304
Along-track distance, 54, 108 Bell’s inequalities, experimental verification
Along-track error, 35-36 of, 304-315
Angelidi’s criticism on Clauser-Home fac- cascade photon experiments, 306-3 10
torability condition, 321-322 Clauser experiment, 207
Annihilation photons, 263, 31 1 Freedman-Clauser experiment, 306
Approach-azimuth element, 65 FryThompson experiment, 307-308
Approachilanding aids, 63-70 Holt-Pipkin experiment, 306-307
Appropriate equations, theory of, 344 Kocher-Commins experiment, 306
Area Navigation, 71 Perry experiment, 309-310
ARINC, see Aeronautical Radio, Inc. positronium annihilation
ARSR, 57, 58 experiments, 3 10-312
Aspect experiments (1981), (1982), 308-309 proton-proton scattering, 3 13-3 14
Astigmatism, 364, 410 recent proposals for new experiments, 3 14-
anisotropic, 364, 410 315
ATC. see Air traffic control BI, interpretation of, 315-325
ATCRBS, see Air traffic control radar beacon criticism of current interpretation, 3 16-321
system current, 315-316
417
418 INDEX

Bell’s locality conditions, 286, 308 Closed-loop transfer function, 88


Bell-type inequalities, 291-295, 320 closed-loop tranform for wind disturbance,
for dichotomic measurement results, 293 89
BI, see Bell inequality CMN, see Control motion noise
Boeing 737-300 FMCS, 75 Collision risk probability, 53
Bohm- Aharonov hypothesis, 263 Coma, anisotropic, coefficient of, 364, 410
Bohm version of EPR paradox, 253-256 Coma, isotropic, coefficient of, 364, 410
EPR option ( I ) , 253 “Complementarity,” 257
EPR option (2), 254-256 Complementary filter, 86, 97-99
Bohr-Einstein debate, 246 Completeness, 246, 257, 265, 325
Bohr’s complementarity concept, 272 Complete theory
Bohr’s resolution of the EPR paradox, 256 definition of, 247
Brightness amplifiers, 337 Compton scattering, 262, 311-312
Bruno experiment, 312 of correlated annihilation photons, 263
Klein-Nishina formula for, 3 12
“Control law,” 83
C
lateral channel, 90-97
CAA, see Civil Aviation Agency Control motion noise, 77, 87, 199
Cascade photon experiments, 306-310 definition of, 102
cascade in calcium, 306 and PFE, 99-103
Cathode lenses, 397-416 CONUS, 75
aberrations of spatial focusing, 404-41 1 Conventional takeoff and land aircraft type, 2,
electrostatic, 397, 414 80
near-axial approximation, 401 -404 Cooper rating, 100
brightness picture, 402 Copenhagen interpretation, 245, 256
time-of-flight aberrations of, 41 1-414 Correction signal, 78
equation for, 412 Correlated photons
geometrical, 414 state vector for, 310
spherical, 413 Correlation experiments, 246
Cathode ray tubes, 337 list of, 314
intensity-modulated, 5 Correlation function, 276-277, 292-293, 303,
CDI, see Course deviation indicator 313
CEP, see Circular error probability Counterfactuality, 248-3 18, passim
Channel pairing Course deviation indicator, 33
in TACANIDME, 168 Cross-track error, 35-36
Channel plan, 160-169 CRT, see Cathode ray tube
CHSH, see Clauser-Home-Shimony-Holt CTOL, see Conventional takeoff and land
CI, see Copenhagen interpretation
Circular error probability, 35
Civil Aviation Agency, 13 D
Clauser experiment, 307 DAC, see Delay-attenuate-compare
Clauser-Home assumption of rotational invar- Dead reckoning, 15-19
iance, 308 Dead time, 169-170
Clauser-Home factorability condition, 288, De Broglie’s theory of the double solution,
321-322 266
general validity of, 289 DECCA, British navigation system, 15, 59
Clauser-Home proof of GBI, 287-290 Decision height minimum, 55, 57
Clauser-Home-Shimony-Holt inequality, Delay-attenuate-compare time-of-arrival es-
281-283 timator, 185, 189, 195
Clauser-Horne universality claim, 322 design relations, 188
Closed-loop guidance system, 33 Density matrix formalism, 266
INDEX 419

DH, see Decision height receiver noise, 190- 194


Dichotomic variables, 253, 280-295 signal in space, 212-222
Dispersion-free ensemble, 274-275 signal processing and transponder-inter-
Dispersion-free variables, 273 rogator implementation, 222-23 1
Distance measuring equipment, in aviation, 1 system field tests, 231-233
243 transponder, diagram of, 226
channel plan, 123 two-pulseitwo-mode system, 177- 180
operation diagram, 9 DMEIVOR
range data, 3 geometry of error equation computation, 110
range error, 3 DODIDOT, 37, 42
scanning DME, 73-76 Doppler radar navigator, 15
standards, 120-127 Doppler VOR, 61-62, 117
evolution of, 122-127 Double-slit experiment with correlated sys-
versus TACAN, 13- 15 tems, 270-272
Distortions, coefficients of, 364, 410 Einstein’s version, 272
isotropic and anisotropic, 364, 410 setup for, 270
DME, see Distance measuring equipment with source of single quantum systems, 272
DMEiDME Downlink/uplink cycle, 7
enroute navigation accuracy, 106-1 18 Drift distance, effective, 367
geometry of error equation computations, DIU
1 I3 desired-to-undesired signal ratio, 137, 163
DMEiN system, 1, 12, 118-171, 200 specific free-space DIU criteria, 163
accuracy, 142-144 DVOR, see Doppler VOR
system error budget, 143
calibration and range coding, 145
coverage, 144- 160 E
ground transponders, 227 Eberhard-Peres procedure, 3 19
operation and components, 129- 133 Eberhard-Peres proof of GBI, 290-291
power budgets, table of, 151 Einstein-Bohr controversy, 267
“signal in space,” 141-171 Einstein locality, 248, 258, 271, 276, 298,
system capacity, 169-171 300, 321
DME navigation parameter, 2 consequences of, 278
DMEiP international standard, I , 12, 165, recent proposals for testing validity of, 267-
171-235, see also Precision Distance 272
Measuring Equipment double-slit experiment with correlated sys-
accuracy specifications, table of, 218 tems, 270-272
antenna considerations, 22 1-222 Popper’s new EPR experiment, 268-270
channel plan, 180-182 Einstein-Podolsky-Rosen paradox, 245-335
decoder, 166 attempts to resolve, 264-267
DME used with MLS, I critical analysis of original EPR argumenta-
error measurement, 212-213 tion, 256-261
garble-system efficiency and accuracy ef- Feynman and, 272
fects, 198-212 first attempt at an experimental discrimina-
interaction with AFCS, 103-105 tion between EPR and standard QM,
interference environment, 225 261-263
multipath, 194-198 hypotheses
operational requirements, 175- 177 counterfactuality, 248
accuracy necessary, table, 176 Einstein locality, 248
allowable errors, table, 178 universal validity of QM, 248
pulse shape, 183- 186 modified EPR argumentation, 251
pulse time-of-arrival estimation, 186- 189 original version, 246-252
420 INDEX

Einstein’s relativity theories, 326 Federal Telecommunications Laboratories, 13


Einstein unified field theory, 300 Feedback-controlled aircraft, 77-79
Electron lenses, 337 feedback loop, 84
Electron mirror microscopes, 337, see also Ferris discriminator, 136, 165, 166, 167, 185,
Mirror electron microscope 204
aberrations, 356-364 Final approach fix, 53, 54
time-of-flight spherical aberration, 369 Final approach waypoint, 54
conservation law for azimuthal component First- and second-kind states and Bell’s in-
of its generalized momentum, 339 equalities, 321
electron-optical properties of axially sym- Fixing position, 15-19, see also Position fix
metric, 350-371 Flight control system
constants a and b auto-coupled, 32
mirror properties in first approximation, manual-coupled, 32
353-356 Flight technical error, 34, 42
energy conservation law for a particle, 339 Fractional parentage coefficients, 320
equations of motion, 338-350 Freedman-Clauser experiment, 306
time of flight, 349-350 Fry-Thompson experiment, 307-308
trajectory equations, 340-349 R E , see Flight technical error
mirror time-of-flight mass dispersion, 367
time-of-flight energy dispersion of the first
order, 365 G
time-of-flight focusing, 365-37 1
Garble, 198-212
principal plane of, 366
downlink DME garble, 199
space-time-of-flight focusing, 37 1
garble source, 208
time-focusing interval, 367
interaction interval of, 204
trajectories determining focus, 356
uplink DME garble, 199
Electron mirrors theory of, 337-416
Gaussian image, plane of, 363
Element of physical reality, 247, 261, 294
Gauss-Markov process, 96
definition of, 247
GBI, see Generalized Bell inequalities
Encoding and decoding, 7
GDOP, see Geometric dilution of precision
Ensemble mean-square error, 41
Generalization of Bell inequality to arbitrary
EPR, see Einstein-Podolsky-Rosen
spin, 295-297
EPR argumentation, see EPR paradox
Generalized Bell inequalities, 280-295
EPR-Bohr debate, 258 Clauser generalization, 281-283
EPR problem, see EPR paradox proof by Bell, 283-285
Error analysis, 106- 112
proof by Clauser and Home, 287-290
Error budget, 37-39
proof by Eberhard and Peres, 290-291
bias error, 38 proof by Selleri, 287
noise error, 38
proof by Stapp, 285-287
from temperature effects, 38
Geometric dilution of precision, 75
time sample mean square error, 39
Glide path, 33
EUROCAE, see European Organization for
Glide path intercept point, 29
Civil Aviation Electronics
Global positioning system, 25
European Organization for Civil Aviation Elec-
GRIP, see Glide path intercept point
tronics, 122
GPS, see Global positioning system
Expectation values, 296
Guidance loop bandwidth, 90
Guidance output information, 2
F
FAF, see Final approach fix
Faraci experiment, 31 1-312 H
FAMP, see Final approach waypoint Hazeltine Corporation, 10
FCS, see Flight control system Mark V IFF, 10
INDEX 42 1

Heisenberg’s uncertainty relations, 260, 269, Joint probability distributions and Bell’s in-
279 equalities, 316-321
Hermitian operator, 274 Joint Tactical Information Distribution System,
Hidden variable theory, 249, 273 138, 198
hidden parameters, 275 JPD, see Joint probability distribution
Hilbert space, 274 JTIDS, see Joint Tactical Information Distribu-
Holt-Pipkin experiment, 306-307 tion System
HVT, see Hidden variable theory
K
I Kalman filter, 83, 98-99
gain optimization, 98
IAF, see Initial approach fix
Kasday experiment, 3 12
IAWP, see Initial approach waypoint
Klein-Nishina formula, 3 12
ICAO, see International Civil Aviation
Kocher-Commins experiment, 306
Organization
ICAO Channel Plan, 13
Identification of Friend or Foe, 9 L
IFF, see Identification of Friend or Foe Lagged-roll, 94
IFR, see Instrument flight rules Lateral channel control law, 90-97
ILS, see Instrument landing system
perturbation model, 92
Image converters, 337
transfer function of, 95
Image curvature, coefficient of, 364
Lateral guidance, 52, 66-67
Impossibility proofs, 275
Lateral turn, physics of, 93
INAF, see Intermediate approach fix
“Leading edge” measurements, 23
Inertial aiding, 18 Leading edge time of arrival error, 114
Inertial navigation system, 16 Linear filter
Information theory, 272
noise reduction factor, equation for, 230
Initial approach fix, 53
Linearity hypothesis, 273, 275
Initial approach waypoint, 54
Locality, 248, 250, 326, see also Einstein
INS, see Inertial navigation system locality
Instrument flight rules, 47-48, 58
LORAN, 20, 59, 72
route structures, 51-57 Lorentz invariant, 3 16
Instrument landing system, 4, 39, 66-67
system error analysis, 41
Integrity, 60 M
MTBO, 60 Mach’s principle, 302
Intermediate approach fix, 53 Mark V IFF, 10
Intermediate waypoint, 54 MDA, see Minimum descent altitude
International Civil Aviation Organization, I , MEA, see Minimum enroute altitude
48 Mean square error, 38
Seventh Air Navigation Conference, April, Mean-square tracking error, 95
1972, 172 Mean time between outages, 60
Interrogator, 58 MEM, see Mirror electron microscope
Interrogator performance curve, 158 Mermin’s GBI for spin, 296-297
Interrogator range tracking, 138- 141 Microwave landing system, I , 3, 44
INWP, see Intermediate waypoint air-derived radial system, 67
ITT Nutley Laboratory, 10 angle accuracy standards, 80
control laws, 103
guidance loop, 81
J guidance signals, 103
Jitter MLSIRNAV, 69
interrogation rate jitter, 231 signal format, 69
422 INDEX

straight-in approaches, 81 “No-enhancement” hypothesis, 3 10, 323-324


Minimum descent altitude, 53, 5 5 , 57 Nondirectional beacons, 19
Minimum enroute altitude, 53 Nonlocality in QM, direct proof of, 297-300
Minimum Operational Performance Specifica- Nonmagnetic time-of-flight mass spectrometry,
tions, 121 337
Minimum required altitude, 53 Nonmeasurable sets and Bell’s inequalities,
Mirror electron microscope, 372-397 323
aberrations of a shadow image, 384-397 Nonreproducibility of physical situations, 32 1
electron lens aberrations, 392 No-result Gedunken experiment, 271
geometrical aberrations, 39 1 NOTAM, see Notice to airmen
test net image, 392-393 Notice to airmen, 146
arbitrary constants, 376-378 NSE, see Navigation system error
diagram of, 372
electron beam crossover, 375
electron-optical properties, in the first ap- 0
proximation, 378-380 Office of Technology Assessment, 45
image forming, 380-383 OMEGA, 20, 59, 72
impact parameter and circle of secondary OTA, see Office of Technology Assessment
emission, 383-384 Outlier rejection, 231-232
microscope magnification, equation for, 379
MLS, see Microwave landing system
MOPS, see Minimum Operational Performance
P
Specifications Path following error, 99-103, 199
MRA, see Minimum required altitude and CMN, 99-103
MSE, see Mean square error definition of, 100- 102
MTBO, see Mean time between outages definition, diagram, 105
Multipassage mass spectrometer, 337 Perry experiment, 309-310
Multipath geometry and error mechanism, 11 1 PFE, see Path following error
“Multisensor navigation system,’’ 99 Phantastron, 139
Photon annihilation, 31 1
Photon polarization, 262-263
N
PICAO, see Provisional International Civil
NAS, see National Airspace System Aviation Organization
National Airspace System, I , 45 Plan-position indicator, 5, 58
Navaids, see also Navigation aids Plan view display, 58
accuracy consideration, 33-37 Polanzer efficiencies, 306-307
bias accuracy, 33 Popper’s new EPR experiment, 268-270
error budget, 41 setup for, 269
error components, 36 Position fix, 15-19, 26
guidance signal, 30 Position fix accuracy, I14
NAVAR, I 1 Position-fix aircraft guidance system, 78
Navigation accuracy performance of DME, Positive control airspace, 48
TACAN, and VOR, 106-1 18 Positive vertical guidance, 55
Navigation aids, 4, see also Navaids Positronium annihilation experiments, 3 10-3 12
Navigational services, table of, 4 Bruno experiment, 312
Navigation parameter measurement, 19-25 Kasday experiment, 3 I2
DME, 19 Wilson experiment, 312
INS, 19 Wu-Shaknov experiment, 3 I1
range and bearing, 19-25 Power budgets, 146
Navigation system error budget, 35, 42 PPI, see Plan-position indicator
NDB , see Nondirectional beacons Precision Distance Measuring Equipment,
Negative probability distributions, 266 171-235
INDEX 423

PRF, 74 RTCA, 1984, 42, 54, 69, 75


Protection ratio DIU, 167 RTCA Special Committee 159, 25
Proton-proton scattering experiment, 3 13-314 Runway fix, 53
experimental setup, 313 RWYF, see Runway f i x
Provisional International Civil Aviation Orga-
nization, 11
S
PVD, see Plan view display
SARPs, see “Standards and Recommended
Practices”
Q SEARCH, 139
QM rotational symmetry, 262 Secondary surveillance radar, 8, 12, 25
“Quality factor,” 294 “See-and-avoid’’ principle, 47
Quantum potential, 275 Segment, 53
Selleri proof of GBI. 287
Separability, 271, 276, 300
R SG, see Stem-Gerlach
Radar, 5 Short takeoff and land
Radial navaids, 23-24 Slew-rate limiting, 231-232
Radial systems, 20 Specular multipath, 109
TACAN, 20 Spherical aberration, coefficient of, 364, 410
VOR, 20 Spin correlation experiments, 302
Radio navaid data rate, 87 Spin projections, 277
Radio navaid position-fix configurations, 23- Squitter, 130
25 signal amplitude, 227
Radio navigation systems SSR, see Secondary surveillance radar
table of, 59 SSV, see Standard service volumes
Radio Technical Commission for Aeronautics, “Standards and Recommended Practices,” 58,
11, 12 120
RAM, see Random access memory Microwave Landing System, 63
Random access memory, 139 for nonvisual precision approach, 172
Rebecca-Eureka system, 9 table of, 122, 124-127
table of. 10 for VOR, 60
Reeh-Schlieder theorem, 304 Standard service volumes, 145, 147, 148
Relativity, 248, 316, see also Einstein’s theory Stapp proof of GBI, 285-287
of relativity State of a physical system in EPR
Reply efficiency, 169 definition of, 256
equation for, 170 State vectors of the first type, 251
Resonant scattering, 324 State vectors of the second type, 251
Retroactivity, 265 State-vector reduction, 264, 269, 270, 271,
equipment, 27 298
RNAV, 4, 27, 34, 50-52, 107 Stem-Gerlach device, 253, 261, 276, 285,
flight paths, 50 302
instrumentation errors, 107 STOL, see Short takeoff and land
position error, 34 Successive approximations, method of, 348
routes, 35 System efficiency
system error, 34 definition of, 199
3D navigation, 52, 57 estimating, 201-203
Rocky Mountains, 53
Root-sum-square calculation, 37-44
T
RSS, see Root-sum-square
RTCA, see Radio Technical Commission for TACAN, see Tactical Air Navigation System
Aeronautics Tactical Air Navigation System, 3, 13- 15, 5 1
424 INDEX

area navigation systems, 41 von Neumann theorem, 273, 273-276


enroute navigation accuracy, 106-1 18 VOR, see Very-high-frequency omni range
error definition for, 109 VORIDME, 36, 54, 72, 108
TACAN bearing measurement, diagram, 64 error definition for, 109
Tangent point, 108 navigation application, 36
tangent point distance, 108 VORIDME position-fix guidance, 18, 51
TCA, see Terminal control area VORIDMEITACAN controversy, 13
Telecommunications Research Establishment, VORIILS
9 localizer facilities, 168
Terminal control area, 48 VOR/ILS/DME, 13
TERPS, 53, 54 VORTAC,4, 14, 19, 44-50, 59,73, 112-118
Time images, equation of, 367 flight paths, 50
Time-of-flight aberrations, 349 multipath error relations, 1 12- 1 18
Total system error, 35 system diagram, 14
TRACK, 139 TACAN plus VOR, 4
Tracking error, see also Total system error VORTAC radials, 19, 44
Transponder, 58, 63
TRE, see Telecommunications Research
W
Establishment
True hyperbolic systems, 20 Wallops Station Flight Center, 23 I , 234
DECCA, 20 Wheeler-Feynman absorber theory of radia-
LORAN, 20 tion, 300
OMEGA, 20 Wholeness idea of Bohm, 326
Two-dimensional lateral guidance navigation Wiener filter theory, 98, 99
system, 18 Wigner-Araki-Yanase theorem, 294
Wigner potential, 264
Wigner’s theory of measurement, 264
U
Wigner’s version of Bell’s theorem, 279-280
Universal validity of QT, 248, 250 Wilson experiment, 312
U.S. Naval Research Laboratory, 10 Wind turbulence, 18, 85, 89
Working Group M meeting in Seattle, 173
V Wu-Shaknov experiment, 31 1

Vertical guidance, 52, 66-67


Very-high-frequency omni range, 3, 60-62 X
Doppler VOR (DVOR), 61-62 XTRK, see Cross-track error
enroute navigation accuracy, 106- I 18
Precision VOR (PVOR), 61-63
Z
principles of, diagram, 61
“scalloped, ” 43 “Zero mile delay,” 8
VFR, see Visual flight rules Zero-order hold, 63, 87
VICTOR airways, 45-46, 129 ZOH, see Zero-order hold
Visual flight rules, 47
von Neumdnn’s impossibility theoiem, 275

Potrebbero piacerti anche