154 IEEE TRANSACTIONS ON GEOSCIENCE ELECTRONICS, VOL. GE-17, NO.

4, OCTOBER 1979

Complementary Code and Digital Filtering for

Detection of Weak VHF Radar Signals from
the Mesosphere
GERHARD SCHMIDT, RUDIGER RUSTER, AND PETER CZECHOWSKY

Abstract-Measurements of the structure and the dynamics of the radar, the antenna gain or the pulse length. If the pulse length
middle atmosphere with a fine height resolution have been carried outis increased, and the pulses are phase coded, the radar can be
using the SOUSY-VHF-Radar (SOUSY = SOUnding SYstem), operated operated with the maximum average power without deteriorat-
by the Max-Planck-Institute for Aeronomy in the Harz mountains in
Germany. Since the echoes from the middle atmosphere are coherent ing the height resolution.
within a time scale of the order of a second, the received signals are Only a few VHF radars have been used to make measure-
over-sampled and added coherently for that time period. This integra-ments in the mesosphere. The Jicamarca radar was used with
tion, which is equivalent to a digital combfiltering, improves the signal-
a peak power of 1-2 MW and a height resolution of 5 km [3].
to-noise ratio. It is carried out by a hardware adder. In order to detect
This radar, however, was designed primarily for observations
echoes from the mesosphere, an increase in the effective radiated power
is necessary. By applying a 32-element complementary code to long in the ionosphere. A few months ago MiXler et al. [8] reported
transmitted pulses, the average power is increased and radar returns the first results obtained with the Urbana radar, using a peak
from the height range between about 65 and 90 km have been recorded power of 4 MW, a duty cycle of 0.8 percent, and a height
with a height resolution of 300 m. The special hardware and software resolution of 3 km. In order to get an insight in mesospheric
used is described. Some observational results from mesospheric mea-
surements are presented. dynamics with scale sizes of the order of 100 m, the height
resolution has to be increased by at least a factor of 10. The
application of sophisticated phase coding techniques offers a
I. INTRODUCTION means for measuring these processes with the high sensitivity
necessary for these investigations. In 1978, measurements
OPPLER radars operating in the lower VHF-band (30-60 with the SOUSY (SOUnding SYstem) radar applying comple-
MHz) are potential tools for observing the structure and mentary code and combfiltering techniques were carried out.
dynamics of the lower and middle atmosphere up to 100-km The peak pulse power was 600 kW, the maximum duty cycle
altitude. Using the Jicamarca radar facility near Lima, Peru, was 4 percent, and the height resolution was 300 m. For the
Flock and Balsley [1] observed VHF radar returns from the first time, therefore, the fine structure of mesospheric dynam-
D region of the equatorial ionosphere. Crane [2] detected ics could be investigated. By applying pulse scanning tech-
echoes from stratospheric heights using the Millstone L-band niques, special investigations in the troposphere have been
radar. Woodman and Guillen [3] were the first to observe carried out with a height resolution of 30 m using the SOUSY
radar returns from the stratosphere and mesosphere and to radar [17].
measure winds and turbulence in these height regions with the The first aim of the present paper is to describe the SOUSY-
Jicamarca radar. Further investigations in the mesosphere VHF-Radar facility. In particular, the functions of the radar
using VHF radars have been carried out by various authors controller and the hardware adder will be discussed in detail.
[4] -[11]. Secondly, the technique of using a complementary code for
The principle of the VHF radar technique is based on the the detection of weak radar signals is described, as well as the
scattering of radio waves from atmospheric dielectric constant signal processing. Finally, some results from high resolution
fluctuations with scale sizes of the order of half the radar measurements of mesospheric structures are presented.
wavelength. These fluctuations are tracers of large scale back-
ground motions [3]. Recently Rottger and Liu [12] have II. THE SOUSY RADAR FACILITY
shown that, in addition to turbulence scattering, partial reflec- The SOUSY-VHF-Radar operates at a frequency of 53.5 MHz
tion from thin layers can contribute significantly to the re- in a valley in the Harz mountains, Germany, 90 km from Han-
ceived echo power. Fluctuations of the dielectric constant in nover. Fig. 1 presents a block diagram of the radar facility
the troposphere and stratosphere are caused mainly by varia- with the main system parameters. The radar controller, which
tions of the atmospheric temperature and humidity, whereas is programmed by a 16-bit computer, is the central unit. It
those in the mesosphere result from electron density variations. holds 1024 program steps in core and controls, via 8 channels,
Mesospheric radar echoes are very weak. The power losses the whole radar system: in particular the master oscillator, the
can be compensated for by increasing the peak power of the transmitter, the transmit-receive-switch, the receiver, the analog
to digital converter, and the hardware adder.
Manuscript received May 3, 1979; revised August 3, 1979.
The authors are with the Max-Planck-Institut fur Aeronomie, D-3411 The transmitter operates as a linear amplifier with a band-
Katlenburg-Lindau 3, Germany. width of 2.5 MHz. The peak pulse power is 600 kW and the

0018-9413/79/1000-0154$00.75 © 1979 IEEE

SCHMIDT et al.: DETECTION OF RADAR SIGNALS FROM THE MESOSPHERE 155

___-------c ADDER CONTROL

Fig. 2. Block diagram of the radar controller.

A. The Radar Controller

The central unit of the radar facility is the radar controller,
which synchronously controls the whole system. The radar
controller is programmed by the computer using simple
MT| | DISC TTY IPTRPTP [IAY Fortran IV statements. After the program has been loaded
Fig. 1. Block diagram of the SOUSY-VHF-Radar facility. (MT-Mag- and the computer has started the radar controller, it runs auto-
netic Tape, DSC-Moving Head Disc, TTY-Teletype, PTR, PTP-Paper matically, stopping at the program end. In case of errors or
tape read/punch, DISPLAY-Storage Oscilloscope, for definition of
abbreviations see Table I). failures occurring during the radar operation, the radar con-
troller is shut off caused either by a safety circuit from the
transmitter and the transmit-receive-switch or by a power fail-
duty cycle 4 percent. The pulse length can be varied from 0.8 ure circuit or by a parity check system in the radar controller.
to 100 ,us. The maximum height resolution is somewhat larger The response time of a shutdown is only 50 ns.
than 120 m because the receiver bandwidth is 1 MHz. Pulse Fig. 2 shows the block diagram of the radar controller. Its
leng-th, pulse coding, and pulse repetition frequency are easily program, generated by the computer, is stored in two different
adjustable to the requirements of the different observational memories and two latches. The core size of each memory is
programs by software instructions. 1024 storage locations of 8-bit length. An additional bit is
The transmitter, the receiver, and the antenna are connected used for parity check. The first memory (program memory)
to a high speed transmit-receive-switch with a recovery time of contains the various commands and the second (rate memory)
less than 5 us, corresponding to a minimum radar range of the respective time intervals for which the commands have to
about 750 m. The transmitting antenna consists of a system be carried out. These execution times are counted in clock
of 196 four-element Yagis with a total gain of 31 dB and a periods, which can be chosen to be 0.5, 1, 2, or 4 ,s each.
half-power beamwidth of 5°. To reduce interference with Since a storage location in the rate memory consists of 8 bits,
commercial TV stations and to reduce the influence of ground the maximum length to be stored is 255 clock periods. If,
clutter from the surrounding mountains, the suppression of however, a program step longer than 255 clock periods is re-
the first side lobes of the antenna pattern is about 20 dB and quired, this command is divided into several sequential pro-
the antenna response far off-axis (which corresponds to low gram steps. The two latches contain the number of repetitions
elevation angles) is approximately 40 dB. At present the an- of the radar cycle (cycle latch) and the maximum address for
tenna beam is directed to the zenith. By means of a system of the program in the rate and program memory (maximum ad-
4-bit phase shifters, which are computer controlled, the ex- dress latch). Three cascaded counters with comparators for
tended version of this array will be steerable continuously in reset are used as address counter, rate counter, and cycle
any direction within a cone of 300 vertex angle centered on counter.
the vertical. The high-sensitivity receiver has a dynamic range The radar controller output consists of 8 bits. The different
of 70 dB and a video bandwidth of 1 MHz. The complex sig- signals of the respective eight parallel channels are listed in
nals at the receiver output are digitized by an analog-to-digital Table I. The prepulse (PRP) signal activates the transmitter
converter. The maximum sample frequency is 2 megawords/s, and the radio frequency control (RFC) signal causes the ampli-
with a word length of 10 bits, corresponding to a dynamic tude modulator to generate the HF-pulse to be transmitted.
range of 57 dB. These digitized complex signals can be coher- The phase flip (FLP) signal controls a biphase modulator in the
ently added by a hardware adder with a maximum core for frequency oscillator as well as the add-subtract (A/S) operation
1024 complex values. The adder also is used for coding tech- in the adder (see Section II-B). This signal is used to generate
niques (Section III). Phase coding schemes are applied, in radar cycles with alternating phases (O°/ 1800) in order to elim-
particular for investigations at mesospheric heights, in order to inate offset bias of the quadrature demodulator and of the
carry out measurements with the maximum duty cycle and the analog to digital converter. In addition, the FLP signal is used
maximum height resolution. The computer takes the data to generate phase coded transmitter signals used for the detec-
from the adder to store it on magnetic tape or disc. The adder tion of weak radar returns (see Section III). The transmit-
and the radar controller are described in detail in the following receive-switch (TRX) signal multiplexes the antenna between
sections. transmitter and receiver, if the radar is used in the monostatic
156 IEEE TRANSACTIONS ON GEOSCIENCE ELECTRONICS, VOL. GE-17, NO. 4, OCTOBER 1979

BIT PRP- 1-I ADDER

2 RFC - - --
x (t)
3 FLP - L

0 4 TRX

tI
<J RAA

6 J RG L
2 7 SP
cr-
0 CS , __l WIIl _ DELAY LINE
0m I- RADAR CYCLE -
Fig. 4. Equivalent circuit for the coherent integration of M input sig-
RATE MEMORY - - nals x yielding the output signaly (see (2)).
Fig. 3. Timing diagram of the 8-bit output of the radar controller.
be chosen such that it is much smaller than the characteristic
TABLE I time of the physical processes of interest. Earlier measurements
RADAR CONTROLLER OUTPUT (8 BIT) made by Woodman and Guillen [3] show that this character-
Bit Signal Function istic time is of the order of a second. Since, however, short
period statistical variations are superimposed on the signals, T0
1 PRP Transmitter Prepulse has to be of the order of milliseconds. If sampling is desired at
2 RFC Transmitter Radio Frequency Control about 1000 heights, the required data rate from the analog-to-
3 FLP + A/S Transmitter Phase Flip and Adder: Add/Subtract
4 TRX Transmit-Receive-Switch Control digital converter would be of the order of several megawords
5 RAA Reset Adder Address per second, which is beyond the acquisition rate of a minicom-
6 RG Receiver Gate Control puter. Due to the coherence of the radar returns within the
7 SP Spare
8 CS Conversion Start for Analog Digital Converter time scale of seconds, the received signals are oversampled and
added coherently for that period of time. This integration
procedure reduces the amount of data to be stored in the com-
TABLE II puter for data processing. This summation is equivalent to
Rate Memory Prog. Memory combfiltering the data.
Address (rates in clock periods) (octal format*) Fig. 4 shows the principle of coherently integrating M signals
resulting from M radar pulses with the repetition period ro. In
0 1 0 order to analyze this integration process, the transfer function
1 8 54
2 2 55 of the adder, which acts as a combfilter, is determined. A
3 3 57 monochromatic input signal with the amplitude X(w) and the
4 1 53 frequency X is given by
5 2 57
6 1 53
7 6 57 x(t) = X(co) e (1)
8 4 54
9 5 4 The output signal resulting from averaging M input signals is
10 4 204
11 1 4 1 M-1
12 2 0 y(t)=M- Z x(t+kTo)
Mk=O
The max. address latch contains a "1 2".
*bit 1: LSB M-i
= X(@o)~
()
=

M exp (jwk7ro) exp (jwt)

k=o
mode. The reset adder address (RAA) signal controls the hard-
ware adder processing of the phase coded signals and will be = Y(W) - exp (ijt). (2)
discussed in Section III. The receiver gate (RG) signal protects
the receiver against saturation during the transmission phase, The transfer function H(o), defimed by
because the transmit-receive-switch has an isolation of only 69 Y(co) = H(w) * X(@) (3)
dB. Bit 7 is spare bit. The conversion start (CS) signal clocks
the analog-to-digital converter which, in turn, controls the becomes with (2)
adder to take the data. sin (Mo(roI2))
Fig. 3 presents an example of a timing diagram of the eight H(co)=ep(I(M-
H(Cw) = exp (}(M -l)6 2
2 )MoMsin(o(,ro/2))' (4)
parallel signals at the radar controller output. The program
producing this sequence is given in Table II, which lists sepa- The magnitude of H(c) is presented in Fig. 5 for different
rately the contents of the rate and the program memory. It values of M. The filter function has maxima (combs) at fre-
should be pointed out that due to the high efficiency in pro- quencies
gramming the radar controller, the program presented in Table v
II only needs 13 storage locations. fv = , r=O,--,M- 1. (5)
'o
B. The Adder Only signals with frequencies which are multiples of the recip-
In measuring dynamical processes in the lower and middle rocal radar cycle 1/To are not attenuated by the filter. The
atmosphere with a radar, the pulse repetition period ro has to half-bandwidth of one tooth of the combfilter determines the
SCHMIDT et al.: DETECTION OF RADAR SIGNALS FROM THE MESOSPHERE 157

M=1

¶
[HI I 1--
1
to
2
to
3
'o

1 2 3 1 2 3
to To to to to to
Fig. 5. Magnitude of the transfer function of a combfilter for different
values of M. MULTIPLEXER
Fig. 6. Block diagram of the hardware adder.
frequency wm at the 3-dB point for a given value ofM:
sin (MCm ((To/2)) -0.7. (6) During the summation of M successive repetitions of the total
M sin (wm)r (T-o /2)) signal the periodic component will add coherently, the random
component incoherently. Thus the signal-to-noise power ratio
If M > 10, equation (6) yields RM becomes:
1.4 M-1 2
M (To/2)(7
(k=O {M - x(t + kro)}2
Since the Doppler frequency X = 2ir * f and the velocity v of
the reflecting or scattering volume are related to the angular RM =- M-i J2 M. 2
frequency Q of the transmitter by k=0

92 2 X=f2
V=-C (8) M2x2(t + kro)
Mu~~MR 1.
_
(13)
Wm. in turn, determines the velocity cuttoff vm, where the Thus the integration of M successive signals results in an im-
sensitivity is down by 3 dB, so that provement by a factor of M in the signal-to-noise power ratio,
1.4c
*M
X
(9) or by a factor of VAT in the respective signal-to-noise ampli-
vU tude ratio.
Fig. 6 shows the block diagram of the digital hardware adder
with c being the speed of light, and X the radar wavelength. used in the SOUSY radar system. The two quadrature com-
Equation (9) shows that the velocity vm is inversely propor- ponents of the signal (e.g., real and imaginary part) are pro-
tional to the number M of additions. cessed in parallel, allowing for a maximum of 1024 heights.
This combfiltering process not only reduces the amount of They are coherently added in a 20-bit adder and stored in the
data but at the same time also improves the signal-to-noise respective memory. If the first of the M radar cycles is pro-
ratio. A noisy signal can be expressed by the sum of a periodic
signal x(t) and random noise n (t). The periodic signal cessed, the multiplexer stores a zero in the adder instead of the
content of the memory. The sign of the operation (addition
x(t)=x(t+kro), k0, ,M- 1 (10) or subtraction) is set by the radar controller. An alternating
adds coherently, whereas the random element n (t) is suppressed sign change together with the corresponding phase control
in the averaging due to its incoherence. signal (FLP) to the transmitter is used in eliminating the in-
Two reasonable assumptions about the noise make the calcu- strumental bias, as well as in processing phase coded signals
lation tractable. (see Section III). The reset adder address (RAA) command
a) The mean value of the noise over a sufficiently long period from the radar controller enables different sequences of data
is zero, i.e., to be stored in the adder array of 1024 words. Using comple-
mentary coding techniques only 512 heights can be stored in
I M-i the adder, since data for each height are stored in two words
lim M , n(t + kTo) = °. (I 1)
M-+co M k=O of the memory related to the two codes used.
b) The variance of the noise an is constant over the data III. PULSE CODING TECHNIQUE
collection interval. The effective radiated power Peff of a radar is given by
Under these assumptions the signal-to-noise power ratio R Peff = G - P (14)
for the kth sample is
where G is the gain of the antenna system including losses
{x(t + kro)}2 (12) caused by the transmission lines and P the average power out-
an put of the transmitter. The latter is the product of the peak
158 IEEE TRANSACTIONS ON GEOSCIENCE ELECTRONICS, VOL. GE-17, NO. 4, OCTOBER 1979
pulse power Pmax and the duty cycle d,:
P =dc 'Pmax- (1 5)
The duty cycle, in turn, depends on the pulse length A tp,
which determines the range resolution Ar, and on the pulse
repetition period T0, which determines the unambiguous range
r. The physical quantities to be measured, such as the layer
thickness, determine independently of the system the required
range resolution Ar at range r, and these in turn predetermine
the duty cycle dC. Thus, Peff finally becomes
Peff =G -
Ar
'max = G
A tp
~Pmax- (16)
r T
From (16) it follows that the detection of signals from greater
ranges by radar systems requires an increase in peak power, if
the effective radiated power as well as the range resolution is
to stay unchanged. Since, however, the transmitting tube peak
power is limited, an increase of the average power can only be
achieved by means of a wider transmitted pulse, resulting in a
decreased range resolution Ar.
The technique of pulse coding offers a means for using the
maximum average power capability of a pulse radar so that
there is neither an increase in peak power nor a degradation of
range resolution. This is accomplished by transmitting a wide
pulse in which the carrier is modulated and then, by proper
signal processing methods, causing a time compression of the
received signal to a much narrower pulse of high effective peak
power. A wide variety of coding schemes exists for different
applications. In the measurements, reported in Section V, a
complementary code has been used.
A. Complementary Code
A complementary code consists of a pair of two equally long
finite sequences of binary elements [13], [14]. A sequence of
binary elements is denoted by
(pi), i=l, ,N (17)
with pie{+l, -l1}.
The binary elements are assumed in this paper to be either
"+1" or "- 1". A complementary code has the property, that
the sum of the autocorrelation functions of the two sequences
(p1) and (pI2)) is equal to zero for all delays, except for the
delay zero, where it becomes 2N, i.e.:
N-T
RI() (1)
i=l
2-
N-T
RI, (r) = iPy
2 I
E
i=1
Fig. 7 shows an example of two such code sequences, their
autocorrelation functions RI (T) and RII (r) for N = 32, as well
as their sum. The main advantage of the complementary code
is evident. The sum of the autocorrelation functions does not
show any sidelobes within the repetition time of the coded sig-
Code l
RI (o1) Code 11.,
Autocorreoation
function for code
.64
I32
RI (E) Rn (0)
Sum of auto-
REl (T) +32 correlation functions
Autocorrelation
function for code 11
.s 32- 32T 32T
W~% -
rvlV75 " 1=0 t
Fig. 7. Autocorrelation functions of a 32-element complementary code.
nals. The dynamic range of the signal, therefore, is much
higher than, for example, that of the Barker code [15].
The use of the complementary code, of course, implies that
the radar targets do not change within the time period r0, the
time needed to sample the signal with both codes. This as-
sumption is fulfilled in the measurements, reported in Section
V, since the observed coherence times of mesospheric echoes
are of the order of a second, and the pulse repetition period is
about 1 ms.
It should be pointed out that the complement of a sequence
is not equal to the sequence of the complements of each ele-
ment. In the case of binary elements like {+1, - 1}, the latter
sequence corresponds to the negative of the original sequence,
i.e.:
(pi)* / (Pi) (-Pi)
= (19)
where the star denotes the complement. Equation (19) im-
plies that (p*) and (p*)* form a pair of complementary codes,
if the pair (pi) and (pi)* is one.
Using a 32-element complementary code, measurements in
the mesosphere have been carried out. The two parts of the
complementary code are generated by a Fortran program in
the computer and stored in the radar controller. The clockrate
for the code is 2 ,us so that the transmitted signals have a length
of 64 Ms. The height resolution, therefore, is 300 m. In order
to eliminate the instrumental bias, sequences of 4 codes each
are transmitted at -0 = 1.6 ms intervals: code I positive, code
II positive, code I negative, and code II negative. The number
M of coherent integrations (equations (2) and (9)) therefore,
is restricted to be a multiple of four.
Using the nomenclature introduced above, the sequence of
codes is
f(p.l'), (p(2,), (-p.l)), (-p.2,)j (20)
I I I
which is, in terms of (p$l)) only, equivalent to
{(0()), (p(l)) (0()*), (0()*)*I'
with i= 1, ,N. (21)
When received the adder converts the signals from negative
codes to those from positive codes by subtraction. Therefore,
signals from code I and code II only need to be stored sepa-
rately in the adder memory. Since the memory size of the
adder is 1024 complex words, data from at most 512 heights
SCHMIDT et aL: DETECTION OF RADAR SIGNALS FROM THE MESOSPHERE 159

can be stored in the adder. The received signals y(l) from B which transforms s' into s also transforms the respective
codeIandy f2romcodeII,( 1,**--,L)sampledatLheights covariance matrices R. The resulting similarity transformation
are crosscorrelated with the transmitted coded signals pI1) and is
pi2), respectively, yielding the decoded signal R(s)=B-l -R(sl)-B. (26)
1N
=
(Pil)yg1 +p;(2)y(2)_,)' Since the covariance matrices are known, B can be deter-
mined. A detailed description of this procedure is given by
l=1, * * ,L -N+ 1. (22) Ruster and Woodman [16]. The temporal and spatial varia-
tion of different physical quantities is calculated from the
It should be noted that the number of heights L is reduced to compressed, decoded, and calibrated data by means of spectral
L - N + 1 when using an N-element code. Due to the temporal analysis. The signals s(h1, tj) observed at height h, (I= 1,
coherence, M signals YI, k can be averaged in order to improve *--,L) and times t (j=O,- , J - 1) is fully characterized
the signal-to-noise amplitude ratio by a factor of VA: (see by its time Fourier transform, the signal spectrum
Section II-B):
F(h,, Wk) =5f{s(hl, tj)}, with t, = jAt = jMrO
' 2 M/2, ,
>,=m2 Yl, k, with Il=lI,-...L -N +l. (23)
At ' (2 ) 2
(27)
Because the decoding and averaging operations are linear, The minimum frequency resolution, therefore, is
their application can be reversed, such that the received coded
signals are first averaged and than decoded. This procedure, 1
fmin = JA t (28)
first proposed by Woodman (in a private communication,
1976), reduces the computing time by several orders of magni- corresponding to a velocity resolution of
tude:
-(1) 2 M/2 Vmi =J (29)
k=1 (X being the radar wavelength).
2 M/2
The maximum unambiguous frequency is given by
-(2) (2)
k=1 f1
fmax 2At (30)
)'g 1E (p(l)y(MA p2)y(2)
+p corresponding to a maximum unambiguous velocity of

with I=1,* *, L -N+l. (24) Vmax 4At' (31)

The averaging process is performed on-line, whereas the co- According to equation (9), the last two equations are only
herently integrated signals are decoded off-line. valid if the signals have not been combfiltered. If M, the num-
IV. DATA PROCESSING ber of coherent integrations, is greater than 10, the maximum
velocity vma. is reduced to
The data, which have been compressed and decoded, are pro-
cessed off-line in several sequential steps. The constant parts x
of the signal arising from stationary radar targets such as moun- Vm +4.5A t (32)
(2
tains can be very strong, so that the weaker temporal fluctua-
tions containing the physical information are concealed. The V. RESULTS OF MESOSPHERIC MEASUREMENTS
best possible estimate of this so-called ground clutter is given Several measurements have been carried out in 1978 in con-
by the time average of the signal taken over the whole observa- nection with a special program for investigating the seasonal
tional period. After subtracting these constant parts, the two variation of mesospheric structures [9]. Fig. 8 presents a
components of the resulting signal s' (i.e., the real and imagi- characteristic three-dimensional plot of the echo power ob-
nary parts) should be statistically independent and their average served in the height range between 64 and 94 km during a
power should be the same for a sufficiently long time interval. period of four hours. A general feature of the autumn season
If, however, the quadrature demodulator contained phase is the occurrence of thin layers with a maximum extension of
errors and the two channels of the signal have been differently about four kilometers which are present continuously for up
amplified s' has to be calibrated. For that purpose, a transfor- to 7 h. In several cases more than three layers could be ob-
mation matrix B has to be determined, such that served at the same time at different altitudes, within a limited
s = B * s' (25) height interval of 12 km ranging from 70 to 82 km. Between
11.25 and 13.00 UT, the echo intensity of the lowest layer
where s is the calibrated signal, the components of which are shows a fluctuation of more than 10 dB with a period of about
statistically independent and have equal intensities. The matrix 30 min. It is interesting to note that this layer seems to move
160 IEEE TRANSACTIONS ON GEOSCIENCE ELECTRONICS, VOL. GE-17, NO. 4, OCTOBER 1979

POWER PROFILES 2 OCT 1978 16. JUNE 1978

-15.20 .7-56 V -
.

80- 0 5

f-b'-
*i l **

P/dB
16]

13
70 -
10 j
o- A
' 11.25
90 so 70
z/km
65 -
Fig. 8. Power P of the mesospheric echoes received with the SOUSY- 11.26 11.28 11.30 11.32
VHF-Radar as a function of height z and time t. 0 dB corresponds to t/UT
noise leveL
Fig. 10. Contour plot of the received echo power of mesosopheric
structures.
2 OCT 1978
16. JUNE 1978 11.26 UT

74.1km

4el0- 73.8km
-
3

5
- -------------------------- E3
8
1

E 73.5km
N 0- _ ^ 5
_

73.
L3 N

0*- s -- 4

.1 72.9km
4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
0
13.00 14.00
--------------------
--s t 3
t/UT 4]3i 72.6km I'
Fig. 9. Time series of the vertical velocity of mesospheric structures, O 1~ - -
5r
E3
__
-
_ - _ - 4F
the power of which is shown in Fig. 8. If the echo power was below g
given level, no velocities have been calculated. 0 lO 200 320
tIs

down with an apparent velocity of 1 m/s. A second type of Fig. 11. Time series of the echo power PR and spectral width for the a

thin structure presented in Fig. 10.

signal, which can be descnbed as blobs, was received from alti-
tudes above 82 km. These echoes occur sporadically and can
be characterized as thin short-lived structures.
The vertical velocities of the structures shown in Fig. 8 have
been determined for the same height range and time period.
The time interval between two successive time averaged sam-
ples was At= 150 ms, resulting in a maximum unambiguous
Doppler shift of 3.3 Hz, which corresponds to a radial velocity
of 9.4 m/s. This value, in turn, is reduced to 8.4 m/s by comb-
filtering according to (32).
A 32-sample frequency analysis, i.e., 4.8 s, yields a frequency
resolution of 0.2 Hz. The resulting velocity traces for 40 height
steps during a period of about 3.5 h from 11.25 UT to 15.10
UT is presented in Fig. 9. The gaps indicate that the signal-to-
noise ratio was smaller than the level estimated to be the lower
limit for defining significant velocities. Fig. 9 thus shows the
presence of two layers where the signal-to-noise ratio was
above the specified level. The bottom of the layer in the
upper part of this figure displays a vertical oscillation with an
amplitude of more than one kilometer. Three kilometers
below that level, a second layer is present, which appears to
move down from 72.0 km to 69.0 km. Within these layers
short term velocity variations with a period of the order of
minutes and an amplitude of about 3 m/s are observed. In
several cases they are correlated over more than six height
steps, corresponding to 1.8 km.
Fig. 10 presents a contour plot of the echo power observed
in the height range between 65 and 82 km for a time period of
about 6 min. According to the nomenclature reported by
Czechowsky et al. [9], the structure shows the characteristic
features of a sheet with a thickness of some hundred meters
and very steep gradients at the boundaries. The time variation
of the relative echo power PR is plotted in Fig. 11 (solid lines),
for six successive heights, each separated by 300 m. Only at
heights of 73.5, 73.2, and 72.9 km does the echo intensity
significantly exceed the noise level. The echo power starts
first to increase at the altitude of 73.5 km and, with a time
SCHMIDT et aL.: DETECTION OF RADAR SIGNALS FROM THE MESOSPHERE 161

delay of 2 min, appears at the next lower height. The resulting stratosphere using the Millstone Hill L-band radar," in Proc. 14th
apparent velocity of 2.5 m/s agrees quite well with that deduced Radar Meteor. Conf. Amer. Meteor. Soc. (Boston, MA), pp. 101-
106, 1970.
from the Doppler shift of the signals, indicating that the ob- [3] R. F. Woodman and A. Guillen, "Radar observations of winds
served sheet is really moving downwards with a velocity of and turbulence in the stratosphere and mesosphere," J. Atmos.
about 2-3 m/s. The dashed curves represent the width of the [4] Sci., vol. 31, pp. 493-505, 1974.
P. K. Rastogi and R. F. Woodman, "Mesospheric studies using the
spectrum. It is interesting to note that there is a strong anti- Jicamarca incoherent scatter radar," J. Atmos. Terr. Phys., vol.
correlation between the power PR and the spectral width a 36, pp. 1 217-1231, 1974.
Rastogi and S. A. Bowhill, "Scattering of radio waves from
[10]. Consequently the observations presented in Figs. 10 [5] P.theK.mesosophere-I. Theory and observations," J. Atmos. Terr.
and 11 show that in these cases strong echoes occur in regions Phys., vol. 38, pp. 399-411, 1976.
with the least turbulence, which seems to indicate that partial [6] -, "Scattering of radio waves from the mesosphere-Il. Evi-
reflection from thin sheets also contributes to the echo power dence for intermittent mesospheric turbulence," J. Atmos. Terr.
Phys., vol. 38, pp. 449-462, 1976.
[12]. [7] R. M. Harper and R. F. Woodman, "Preliminary multiheight
radar observations of waves and winds in the mesosphere over
VI. CONCLUSIONS Jicamarca," J. Atmos. Terr. Phys., vol. 39, pp. 959-963, 1977.
The SOUSY-VHF-Radar facility has been described. It has [8] K. L. Miller, S. A. Bowhill, K. P. Gibbs, and I. D. Countryman,
"First measurements of mesospheric vertical velocities by VHF
been used for investigations of weak radar returns from the radar at temperate latitudes," Geophys. Res. Letters, vol. 5, pp.
mesosphere. In order to obtain the maximum height resolu- 939-942, 1978.
tion while operating at the maximum average power a pulse [9] P. Czechowsky, R. Ruster and G. Schmidt, "Variations of meso-
spheric structures at different seasons," Geophys. Res. Letters,
compression technique, particular
in a 32-element complemen- vol. 6, pp. 459-462, 1979.
tary code, has been applied. Due to the coherence of the re- [10] R. Ruster, P. Czechowsky and G. Schmidt, "Dynamical processes
ceived signals, the signal-to-noise ratio was improved by digitally in the mesosphere," to be submitted to Geophys. Res. Letters,
1979.
combfiltering the data. Some initial results have been presented [11] J. Rottger, P. K. Rastogi, and R. F. Woodman, "High resolution
to show both the variety of structures in the mesosphere and VHF radar observations of turbulence structures in the meso-
their dynamics. sphere," submitted to Geophys. Res. Letters, 1979.
[12] J. Rottger and C. H. Liu, "Partial reflection and scattering of
ACKNOWLEDGMENT VHF radar signals from the clear atmosphere," Geophys. Res.
Letters, vol. 5, pp. 357-360, 1978.
The authors acknowledge fruitful discussions with their col- [13] M. J. E. Golay, "Complementary series," I.R.E. Trans. Inform.
leagues from the SOUSY project group of the Max-Planck- 7heory, pp. 82-87, 1961.
[14] P. Leppanen, "On complementary code pairs and matched filters
Institut fur Aeronomie. They would also like to thank H. for them," Internal Rep., Inst. Telecommunication, Univ. Oulu,
Becker for constructing the digital electronics and K. Meyer Oulu, Finland, 1976.
for operating the radar system. [15] R. H. Barker, "Group synchronizing of binary systems," in Com-
munications Theory, W. Jackson, Ed. New York: Academic
REFERENCES Press, 1953, pp. 273-287.
[16] R. Ruster and R. F. Woodman, "Digital filtering, calibration and
[1] W. L. Flock and B. B. Balsley, "VHF radar returns from the correlation analysis of radar-echoes from the tropo- and strato-
D-region of the equatorial ionosphere," J. Geophys. Res., vol. 72, sphere," Kleinheubacher Ber., vol. 21, pp. 239-246, 1978.
pp. 5537-5541, 1967. [17] J. R6ttger and G. Schmidt, "High resolution VHF radar sounding
2] R. K. Crane, "Measurements of clear air turbulence in the lower of the troposphere and stratosphere," this issue, pp. 182-189.