T VT NonIsoCorrSpec Revised

Page1of27
AParametricModelfortheDistributionoftheAngleofArrivalandthe
AssociatedCorrelationFunctionandPowerSpectrumattheMobileStation
AliAbdi
1
,JanetA.Barger
2
,andMostafaKaveh
3
1
Dept.ofElec.andComp.Eng.,NewJerseyInstituteofTechnology
2
Dept.ofElec.Eng.,UniversityofTexasatTyler
3
Dept.ofElec.andComp.Eng.,UniversityofMinnesota
Abstract__OneofthemainassumptionsintheClarkesclassicchannelmodelisisotropicscattering,
i.e. uniform distribution for the angle of arrival of multipath components at the mobile station.
However,inmanymobileradiochannelsweencounternon-isotropicscattering,whichstronglyaffects
the correlation function and power spectrum of the complex envelope at the mobile receiver. In this
contribution,weproposetheuseoftheversatilevonMisesangulardistribution,whichincludesand/or
closely approximates important distributions like uniform, impulse, cardioid, Gaussian, and wrapped
Gaussian, for modeling the non-uniform angle of arrivals at the mobile. Based on this distribution,
associatedcorrelationfunctionandpowerspectrumofthecomplexenvelopeatthemobilereceiverare
derived. The utility of the new results is demonstrated by comparison with the correlation function
estimatesofmeasureddata.
Pointofcontact:AliAbdi
Dept.ofElectricalandComputerEngineering
NewJerseyInstituteofTechnology
323KingBlvd.
Newark,NJ07102,USA
Phone:(973)5965621
Fax:(973)5965680
Email:abdi@adm.njit.edu
The work of the first and the third authors have been supported in part by the National Science Foundation, under the
WirelessInitiativeProgram,Grant#9979443.
AParametricModelfortheDistributionoftheAngleofArrivalandtheAssociatedCorrelationFunctionandPowerSpectrumAliAbdietal.
Page2of27
I.INTRODUCTION
Thereceivedsignalcorrelationfunctionsandpowerspectraatthemobilestation(MS)dependon
theprobabilitydensityfunction(PDF)oftheangleofarrival(AOA)ofascatteredwave.Clarkestwo-
dimensionalisotropicscatteringmodel(auniformAOAPDFover ) , [ )givesrisetothezero-order
Bessel function for the autocorrelation function and the U-shaped power spectrum for the complex
envelope of multipath components at the MS [1, p. 40-43]. However, it has been argued [2]-[6], and
experimentally demonstrated [7]-[16] that the scattering encountered in many environments is non-
isotropic, resulting in a non-uniform PDF for AOA at the MS. As has been discussed in [16], the
assumption of a uniform PDF for the AOA introduces small errors on the first order statistics of the
received signal, but a significant error on the second order statistics, like correlation functions and
levelcrossingrates.
In[17]and[18],twogeometricallybasedangleofarrivalPDFsarederived;whilethreedifferent
models are assumed for such a PDF by other authors: quadratic PDF [19], Laplace PDF [10], and
cosine PDF [20]. In addition to showing good fit to measurements, another important factor in
selecting a candidate for the angle of arrival PDF is the mathematical convenience it provides to the
determinationofclosed-formsolutionsforthecorrelationfunctions,powerspectra,spectralmoments,
etc..ThepreviouslymentionedPDFsdonotsatisfythelatterrequirement.Achievingthisgoalisthe
mainpurposeofthispaper.
Theremainderofthepaperisorganizedasfollows.InSectionIIwepresentthetwoparametervon
Mises PDF as flexible model for the PDF of the angle of arrivals. The utility of this model is
demonstrated by the derivation of closed-form expressions for the correlation function and power
spectrum of the complex envelope at the MS in Section III. This is followed by a study of the
differences in the characteristics of the correlation function and power spectrum for cases which
deviate from the isotropic scattering model introduced by Clarke. Finally, measured data is used to
confirm the validity of the proposed model in predicting the scattering characteristics encountered in
actualmobilecommunicationsituations.
Page3of27
II.AFLEXIBLEPDFFORTHEAOA
A PDF was introduced by R. von Mises in 1918 to study the deviations of measured atomic
weightsfromintegralvalues[35].ThisPDFplaysaprominentroleinstatisticalmodelingandanalysis
of angular variables [21, pp. 57-68]. Let the random variable represent the AOA of a multipath
component(scatterandspecular),receivedattheMS.ThevonMisesPDFforthescattercomponentof
, ) (
scat
p ,isgivenby:
) , [ ,
) ( 2
)] cos( exp[
) (
0

=
I
p
p scat
, (1)
where (.)
0
I isthezero-ordermodifiedBesselfunction, ) , [
p
accountsforthemeandirectionof
AOAofscattercomponents,and 0 controlsthewidthofAOAofscattercomponents.Figs.1and
2 show ) (
scat
p in linear and also in polar coordinates, respectively, for different values of and
0 =
p
(non-zerovaluesof
p
simplyshiftFig.1circularlyorrotateFig.2).
As can be observed in Fig. 1, for 0 = we obtain =
2 1 ) (
scat
p (isotropic scattering), while
= yields ) ( ) (
p
scat
p =
(extremely non-isotropic scattering), where (.) is the Dirac delta

function.Forsmall ,thisfunctionapproximatesthecardioidPDF[21,p.60],whichisrathersimilar
to the cosine PDF [20]; while for large it resembles a Gaussian PDF with mean
p
and standard
deviation 1 [21, p. 60]. In general, the von Mises PDF can approximate the wrapped Gaussian
PDFquitewell[21,p.66].ItisinterestingtonotethatthevonMisesPDFappearsinanumberofother
communication contexts. For example, this PDF is referred to as the Tikhonov PDF in partially
coherent communication [22, p. 406], has been used in phase-locked loop related problems [23], and
hasbeenshowntorepresentthePDFofthephaseofasinewaveinGaussiannoiseforlargesignalto
noiseratios[24].
A useful geometrical interpretation for can be obtained by the fact that it is the slope of the
tangent to the curve ) (
scat
p , in polar coordinates, at the point 2 = . This result can be obtained
using the relation ) cos sin ( ) sin cos ( + + [25, p. 176], where ) ( = represents a
curveinpolarcoordinatesandprimedenotesdifferentiationwithrespectto .AlookatFig.2verifies
thisresult.Forlarge ,say 3 , p
scat
( ) takesaunidirectionalshape,ascanbeobservedinFig.2.
Page4of27
InordertoobtainameasureforthespreadofAOA'sinthepolarcoordinatesforlarge ,weconsider
the inflexion points of p
scat
( ) which are approximately equal to 1 , assuming is large [21,

p.60]. So, for a unidirectional scattering scenario (large ), the width of the AOA of scatter
componentsatMSisroughlyequalto 2 .For 10 , 3 = ,weobtain
o
66 and
o
36 ,respectively.
III.CORRELATIONANDSPECTRUMOFTHECOMPLEXENVELOPEOFMULTIPATHCOMPONENTS
Let the received complex envelope at the MS due to scatter components be represented by
) ( ) ( ) ( t r j t r t r
scat
q
scat
i
scat
+ = ,with 1 = j ,where ) (t r
scat
i
and ) (t r
scat
q
arethein-phaseandquadrature
scatter components. Then, the normalized autocorrelation function of ) (t r
scat
, ) (
~

scat
r r
, is given by
) (
~
) (
~
) (
~
+ =
scat
r r
scat
r r
scat
r r
q i i i
j , where
~
( )
~
( )
r r
scat
r r
scat
i i q q
= and
~
( )
~
( )
r r
scat
r r
scat
i q q i
= are normalized
autocorrelationandcrosscorrelationfunctionsof ) (t r
scat
i
and ) (t r
scat
q
.Notethatthetermnormalized
indicates 1 ) 0 (
~
=
scat
r r
.Itcanbeeasilyshownthat )] cos 2 [exp( ) (
~
=
m
scat
r r
f j E [1,p.40],whereE
representsmathematicalexpectationand
m
f isthemaximumDopplerfrequency.Using ) (
scat
p in(1)
forcarryingouttheexpectationwithrespectto yields[26,p.357]:
) (
cos 4 4
) (
~
0
2 2 2 2
0
|
.
|
\
|
+
=
I
f j f I
m p m
scat
r r
. (2)
For 0 = ,(2)resultsinthecorrelationfunctionsfortheClarkestwo-dimensionalisotropicscattering
model, i.e. ) 2 ( ) 2 ( ) (
~
0 0
= =
m m
scat
r r
f J f j I
i i
and 0 ) (
~
=
scat
r r
q i
, where (.)
0
J is the zero-order Bessel
function. Note that for both analytical and numerical calculations, formula (2) and its real and
imaginarypartsaremuchsimplerthan,andthusmorepreferabletothecomplicatedresultsin[19].
The power spectrum of ) (t r
scat
can be obtained based on
2 2 2
)] ( ) ( ) ( ) ( [ ) ( f f p G p G f S
m
scat scat scat
r r
+ =

,
m
f f [1, p.42], where and f are
relatedaccordingto ) ( cos
1
m
f f
= ,
2
isthetotalpowerfromallscattercomponentsattheinputof
the receiver antenna, and (.) G is the receiver antenna gain pattern. For a vertical monopole antenna,
2 3 ) ( = G . So the normalized power spectrum of ) (t r
scat
can be defined as
2
3 ) ( 2 ) (
~
= f S f S
scat
r r
scat
r r
, i.e.
2 2
)] ( ) ( [ ) (
~
f f p p f S
m
scat scat scat
r r
+ =

,
m
f f . The term
normalizedcomesfrom 1 ) (
~
=
}
m
m
f
f
scat
r r
df f S ,whichcanbeeasily verified.Thisisinagreementwith
Page5of27
1 ) 0 (
~
=
scat
r r
,since ) (
~

scat
r r
and ) (
~
f S
scat
r r
areFouriertransformpairs.Byusingthe ) (
scat
p in(1)and
aftersomemanipulationsweobtain:
m
m p m p
m
scat
r r
f f
I
f f f f
f f
f S

= ,
) (
) ) ( 1 sin cosh( ) cos exp(
1
) (
~
0
2
2 2
, (3)
where cosh(.) isthehyperboliccosine.For 0 = ,(3)reducestothepowerspectrumfortheClarkes
two-dimensionalisotropicscatteringmodel,i.e.
2 2
1 ) (
~
f f f S
m
scat
r r
= ,
m
f f .
Forthespecularcomponent,thePDFofAOAcanbewrittenas:
) , [ ), ( ) (
0
=
spec
p , (4)
where ) , [
0
istheAOAofthespecularcomponent. Letthereceived complexenvelopeatthe
MS due to the specular component be represented by ) (t r
spec
. Then, the normalized autocorrelation
function of ) (t r
spec
can be obtained based on )] cos 2 [exp( ) (
~
=
m
spec
r r
f j E and using the given
) (
spec
p :
) cos 2 exp( ) (
~
0
=
m
spec
r r
f j . (5)
The normalized power spectrum of ) (t r
spec
, ) (
~
f S
spec
r r
, can be easily obtained by taking the Fourier
transformof(5):
m m
spec
r r
f f f f f S = ), cos ( ) (
~
0
. (6)
Notethat 1 ) (
~
) 0 (
~
= =
}
m
m
f
f
spec
r r
spec
r r
df f S .
Wheninadditiontothescattercomponents,aspecularcomponentisalsopresent,thePDFofthe
AOAofmultipathcomponents(scatterandspecular)attheMScanbeexpressedas[19]:
) , [ ,
1
) ( ) (
) (
+
+
=

K
Kp p
p
spec scat
, (7)
whereKistheRicefactor,defined astheratioofthepowerinthespecular componenttothatinthe
scatter components. ) (
scat
p and ) (
spec
p are also given in (1) and (4), respectively. For ) (t r , the
received complex envelope at MS due to multipath components (scatter and specular), we also have
Page6of27
1 )] ( ) ( [ ) ( + + = K t Kr t r t r
spec scat
, where ) (t r
scat
and ) (t r
spec
are independent. Now, based on (7), the
normalizedautocorrelationfunction )] cos 2 [exp( ) (
~
=
m r r
f j E of ) (t r becomes:
1
) (
~
) (
~
) (
~
+
+
=
K
K
spec
r r
scat
r r
r r
, (8)
where ) (
~

scat
r r
and ) (
~

spec
r r
aregivenby(2)and(5),respectively.Bydefining ) (t r
i
and ) (t r
q
asthe
in-phase and quadrature multipath components of ) (t r , i.e. ) ( ) ( ) ( t jr t r t r
q i
+ = , we obtain
) (
~
) (
~
) (
~
+ =
q i i i
r r r r rr
j , where
~
( )
~
( )
r r r r
i i q q
= and
~
( )
~
( )
r r r r
i q q i
= are normalized
autocorrelation and crosscorrelation functions of r t
i
( ) and r t
q
( ) . The normalized power spectrum of
) (t r issimplytheFouriertransformof ) (
~

r r
in(8),i.e.:
1
) (
~
) (
~
) (
~
+
+
=
K
f S K f S
f S
spec
r r
scat
r r
r r
, (9)
with ) (
~
f S
scat
r r
and ) (
~
f S
spec
r r
givenby(3)and(6),respectively.Formultipathcomponents(scatterand
specular)wehave 1 ) (
~
) 0 (
~
= =
}
m
m
f
f
rr rr
df f S .
AnimportantadvantageofthePDFforAOAin(1)istheclosed-fromresultsobtainedfor ) (
~

scat
r r

and ) (
~
f S
scat
r r
, presented in (2) and (3), respectively. This can significantly simplify the derivation of
correlation or spectral parameters, required for various system calculations. For example, assume the
multipathcomponentsaresuchthat ) (t r
scat
isazero-meancomplexGaussianprocessand 0 ) ( = t r
spec
,
hence ) ( ) ( t r t r
scat
= (this corresponds to the Rayleigh multipath fading channel, i.e. Rayleigh
distribution for the envelope
2 2
)] ( [ )] ( [ ) ( t r t r t r
scat
q
scat
i
scat
+ = and uniform distribution for the phase
] ) ( ) ( [ tan ) (
1
t r t r t r
scat
i
scat
q
scat
= ). If we want to estimate the speed of the mobile [27], which is
proportionalto
m
f ,wemayusethezerocrossing rateof ) (t r
scat
i
,givenby =
scat scat
i
b t r ZCR
2
~
)} ( {
[1,p.60],where
scat
n
b
~
isthenormalizedspectralmomentsof ) (t r
scat
,definedby:
}
=
m
m
f
f
scat
r r
n n scat
n
df f S f b ) (
~
) 2 (
~
.
Notethatduetonormalization, 1
~
0
=
scat
b .BasedontheFourierrelationbetween ) (
~

scat
r r
and ) (
~
f S
scat
r r
,
we have
0
) (
~ ~
=
=
n scat
r r
n n scat
n
d d j b . Hence, according to (2), we derive a result more general than
thatgivenin[27]:
Page7of27
) ( 2
sin ) ( 2 cos )] ( ) ( [
) 2 (
~
0
2
1
2
2 0 2
2

+ +
=
I
I I I
f b
p p
m
scat
.
So,forthezerocrossingrateof ) (t r
scat
i
wearriveat:
m
p p
scat
i
f
I
I I I
t r ZCR 2
) (
sin ) ( 2 cos )] ( ) ( [
)} ( {
0
2
1
2
2 0

+ +
= .
Theaboveclosed-formformulaisausefultoolforstudyingtheeffectofnonisotropicscatteringonthe
performanceof )} ( { t r ZCR
scat
i
forestimatingthespeedofthemobile.For 0 =
p
and 0 = ,Clarkes
isotropic scattering, the above result reduces to
m
f 2 , in agreement with [1, p. 60]; while for
2 =
p
and 7 = we obtain
m
f 2 51 . 0 , half of the isotropic case. Therefore, we may conclude
that )} ( { t r ZCR
scat
i
is not a robust estimator against nonisotropic scattering and may change
significantly as
p
and vary. Higher order spectral moments like
scat
b
3
~
and
scat
b
4
~
, needed for
developing more efficient and robust estimators for the speed [27], can be simply obtained by taking
higherorderderivativesof ) (
~

scat
r r
in(2).
IV.ACASESTUDY
Here we consider three different urban scenarios studied in [19] and derive the associated
autocorrelation functions and power spectra of the complex envelope of multipath components at the
MS.LettheMSmovefromlefttoright.Forthefirstscenario,S1,scattercomponentsarriveattheMS
headon,i.e.
p
= 0 .Forthesecondscenario,S2,scattercomponentsarrivefromboththe
p
= 0 and

p
= directionswithequalprobabilities.Forthethirdscenario,S3,scattercomponentsarriveatthe
MSfromadirectionperpendiculartoitsdirectionofmotion,e.g.
p
= 2 .ForS1andS3, ) (
scat
p is
equal to ) ( 2 ) cos exp(
0
I and ) ( 2 ) sin exp(
0
I , respectively; while for S2 we have
) ( 2 ) cos cosh( ) ( 5 . 0 ) ( 5 . 0
0
0
= +
=
I p p
p p
scat scat
. These three PDFs are plotted in Fig. 3 for
3 = ,togetherwithClarkesclassicisotropiccase 0 = .
Thenormalizedautocorrelationfunctionsofthecomplexenvelopeduetothescattercomponents
fortheabovethreescenarioscanbederivedaccordingto(2):
Page8of27
) (
4 4
) (
~
0
2 2 2 2
0
1 ,
|
.
|
\
|
+
=
I
f j f I
m m
S scat
r r
, (10)
) (
4 4 Re
) (
~
0
2 2 2 2
0
2 ,
|
.
|
\
|
+
=
I
f j f I
m m
S scat
r r
, (11)
) (
4
) (
~
0
2 2 2 2
0
3 ,
|
.
|
\
|

=
I
f I
m
S scat
r r
, (12)
where Re[. ] gives the real part of its complex argument. Notice that for S1, the imaginary part of
) (
~
1 ,

S scat
r r
, i.e. ) (
~
1 ,

S scat
r r
q i
, is nonzero, contrary to Clarkes classic isotropic scenario. By substituting
(10), (11), and (12) into (8), we can readily obtain
~
( )
r r
S1
,
~
( )
r r
S2
, and
~
( )
r r
S3
. Figs. 4 and 5 show
~
( )
r r
S
i i
1
,
~
( )
r r
S
i i
2
, and
~
( )
r r
S
i i
3
, the real parts of
~
( )
r r
S1
,
~
( )
r r
S2
, and
~
( )
r r
S3
, for ( , ) ( , ) K = 0 10 and
( , ) 1 3 , respectively (for the case of nonzero K, we considered
p
=
0
). Comparing with Clarkes
model ( , ) ( , ) K = 0 0 shown in Figs. 4 and 5, we observe the noticeable effect of non-isotropic
scatteringindifferentscenariosonthecorrelationpropertiesattheMS.Fornonisotropicscattering,it
is possible to have large correlations when Clarkes isotropic scattering model predicts small
correlations.
Thenormalizedpowerspectraofthecomplexenvelopeduetothescattercomponentsforthethree
scenariosS1,S2,andS3canbeobtainedbasedon(3):
m
m
m
S scat
r r
f f
I
f f
f f
f S

= ,
) (
) exp( 1
) (
~
0
2 2
1 ,
, (13)
m
m
m
S scat
r r
f f
I
f f
f f
f S

= ,
) (
) cosh( 1
) (
~
0
2 2
2 ,
, (14)
m
m
m
S scat
r r
f f
I
f f
f f
f S

= ,
) (
) ) ( 1 cosh(
1
) (
~
0
2
2 2
3 ,
, (15)
NoticethatforS1,thepowerspectrum ) (
~
1 ,
f S
S scat
r r
isnotsymmetricwithrespectto 0 = f ,contraryto
Clarkesscenario.Bysubstituting(13),(14),and(15)into(9),wereadilyobtain ) (
~
1
f S
S
r r
, ) (
~
2
f S
S
r r
,and
Page9of27
) (
~
3
f S
S
r r
. Figs. 6 and 7 show ) (
~
1 ,
f S f
S scat
r r m
, ) (
~
2 ,
f S f
S scat
r r m
, and ) (
~
3 ,
f S f
S scat
r r m
for 10 = and 3,
respectively.ComparingwithClarkesclassicisotropiccase, 0 = ,superimposedontheplotsinFigs.
6 and 7, we clearly see the significant effects of different non-isotropic scattering scenarios on the
powerspectrumattheMS.
V.EXPERIMENTALRESULTS
InthissectionweshowtheutilityoftheproposedAOAdistribution,andtheassociatedparametric
correlation function in describing the correlation properties of a range of a variety of measured data.
Firstwe giveabriefsummary ofthedataused (thedynamicstatisticalcharacteristicsofthisdataset
arediscussedin[28]).Inthesubsequentsubsections,wetalkaboutenvelopecorrelationandenvelope
extraction,twonon-isotropicscatteringcorrelationmodelforfittingtodata,estimationoftheenvelope
correlationofdataandparameterestimationforthetwomodels,andfinally,themeritsandlimitations
ofthetwomodelsinfittingtomeasureddata.
A.TestLocationsandDataCollection
Two sites were used for the collection of data: a suburban housing development in Greenville,
Texas,andanurbanareaamongthebuildingsofthecampusoftheUniversityofTexasatArlington,
Texas. These test locations were chosen to provide a random signal environment that would be
representative of actual operational situations. By choosing environments with irregular structures,
data reflects the general angular spread of the signal and not yield results found only in very regular
environments.
Inthesuburbanarea,parkedvehicles,vegetation,andinconsistencyinthesetbackofthehousing
structures produced very irregular reflective conditions. Measurements for records #0011 through
#0015weretakeninthesuburbanneighborhoodcharacterizedbyhomesofawood/brickconstruction
withanominal7mofseparationbetweenthestructures.Thetransmitantennawasataheightof2m
above a hillside location to produce a low grazing angle over the homes, simulating a cellular radio
signalata greaterdistance.Themeasurementsweremadewiththereceiveantennainatotalshadow
Page10of27
environment. Records #0016 through #0018 represent data taken near a large intersection in a
suburbanareawiththetransmitantennaheightraisedtoabout5m.Fortheintersectionmeasurements
there was a significant direct ray component of the signal. During these sample periods there was
significanttrafficactivitynearthetestvehicle.
Theurbanenvironmentischaracterizedbybuildingsofonetosixstoriesinheight,heavytraffic,
largeopenareas,andhighlyinconsistentsetbackfromthestreet.Records#0019through#0022were
madeintheurbanlocationwiththetransmitantennaataheightofabout21m,producingasignalata
moderatelylowangleofradiationoverthebuildingssurroundingthepath.Amapoftheurbanlocation
is shown in Fig. 8. Buildings shown are of modern brick construction. Building heights are indicated
by a number inside the outlines shown in Fig. 8. The urban location contained significant vehicular
traffic, and the measurements were made over paths that had no direct ray component, but
demonstratedthedeepshadowenvironment.
The UHF transmitter generated a 910.25 MHz carrier at a nominal power of 0.2 W (the
wavelengthwas0.33m).Signalsamplesweretakenatdistancesof0.16kmto0.25km.Thecollected
dataconsistoftwelvesetsofnarrowbandinphaseandquadraturecomponentstakenattwelvedifferent
locations. Each data set represents the signal over a traveled distance of 47 m, or 7 s of time. These
signals were digitized with a sampling rate of 35156.25 Hz, to produce one sample per 0.2 mm of
distancetraveled(28sintime)bythemovingreceiverplatform(atafixedspeedof6.7m/s).Hence,
the number of samples for each record is approximately 250,000. The antenna at the mobile receiver
wasa1/4waveverticalstub,0.6cmindiameter,mountedontopofa1.9cmdiameteraluminumtube.
Thetubeserved asa rathersmall groundplane. Forthe fixedtransmitter,thesame antenna wasused
except a 1/4 wave diameter ground plane (flat) was attached to the base of the antenna stub. Both
antennashadomnidirectionalpatterns.
B.EnvelopeCorrelationandEnvelopeExtraction
Since for many practical applications, the envelope of multipath components and its correlation
propertiesareofconcern,weconsidertheenvelopeofmultipathcomponent ) ( ) ( t r t z = .When ) (t r is
Page11of27
a complex Gaussian process with non-zero mean, the autocovariance function of ) (
2
t z ,
)]}] ( [ ) ( )]}{ ( [ ) ( [{ ) (
2 2 2 2
2 2
+ + = t z E t z t z E t z E
z z
, can be expressed in terms of the
autocorrelation function of the complex envelope of multipath components ) (t r , having both scatter
and specular components [29]. The normalized version of ) (
2 2

z z
, ) (
~
2 2

z z
, assuming no specular
component(whichmakes ) (t r azero-meancomplexGaussianprocess),isgivenby[1,p.55]:
2
) (
~
) (
~
2 2
=
scat
rr
scat
z z
, (16)
where ) (
~
2 2

scat
z z
is the normalized autocovariance function of ) (
2
t z due to scatter components. Again
normalizedindicates 1 ) 0 (
~
2 2
=
scat
z z
.
In order to extract ) (t z , the multipath (or fast fading) component of the envelope, which we are
interested in, the shadow (or slow fading) component of the envelope must be removed. Let
) ( ) ( ) ( t jR t R t R
q i
+ = denote the received complex envelope at the MS in the presence of both
multipath and shadow components, with ) (t R
i
and ) (t R
q
as the inphase and the quadrature
components. The shadow (or slow fading) component of the envelope, ) (t R , is represented by its
mean, ] ) ( [ t R E , which is a function of time because ) (t R is nonstationary. The multipath (or fast
fading) component of ) (t R can be represented as ] ) ( [ ) ( ) ( t R E t R t z = [1, p. 91]. A method for
estimating ] ) ( [ t R E from ) (t R usingalocalslidingwindowisdescribedin[30].
C.TwoNon-IsotropicScatteringPropagationModel
Weconsider ) (
~
2 2

scat
z z
forfittingtothedata,accordingtotwopropagationmodels:thesimplenon-
isotropic model and the composite non-isotropic/isotropic model. In the simple non-isotropic model,
we assume that the MS receives the signals mainly from one direction. The AOA PDF and the
normalized autocorrelation function for the simple non-isotropic model are given in (1) and (2),
respectively.Forthecompositenon-isotropic/isotropicmodel,theMSreceivessignalsuniformlyfrom
alldirections,inadditiontoacomponentcenteredaroundaspecificdirection.TheAOAPDFandthe
normalizedautocorrelationfunctionforthecompositenon-isotropic/isotropicmodelcanbewrittenas:
Page12of27
) , [ ,
2
1
) 1 (
) ( 2
)] cos( exp[
) (
0

+

=
I
p
p scat
, (17)
) 2 ( ) 1 (
) (
cos 4 4
) (
~
0
0
2 2 2 2
0
+
|
.
|
\
|
+
=
m
m p m
scat
r r
f J
I
f j f I
, (18)
where 1 0 indicates the amount of directional reception. The composite non-isotropic/isotropic
modelreducestothesimplenon-isotropicmodelfor 1 = ,andsimplifiestoClarkesmodelfor 0 = .
Forthesimplenon-isotropicmodelandbasedon(2)and(16)wehave:
) (
cos 4 4
) (
~
2
0
2
2 2 2 2
0
2 2
|
.
|
\
|
+
=
I
f j f I
m p m
scat
z z
, (19)
whileforthecompositenon-isotropic/isotropicmodelandaccordingto(18)and(16)wederive:
) (
) ( ) 2 ( ) 1 ( cos 4 4
) (
~
2
0
2
0 0
2 2 2 2
0
2 2
+ |
.
|
\
|
+
=
I
I f J f j f I
m m p m
scat
z z
. (20)
D.EstimationoftheEnvelopeCorrelationandtheParametersoftheTwoModels
Theestimateforthenormalizedautocovariancefunctionoftheenvelope-squaredduetomultipath
components, ) (
~
2 2

z z
,wasobtainedusingthebiasedmethod[31,p.743],scaledsuchthat 1 ) 0 (
~
2 2
=
z z

(the bias is very small because of the large number of samples 250000). Recall that (16) and
consequently(19)and(20)arevalidwhen ) (t r isazero-meancomplexGaussianprocess(nospecular
componenthence 0 ) ( = t r
spec
and ) ( ) ( t r t r
scat
= ).ThismeansthatatleasttheunivariatePDFsofthe
envelope, ) ( ) ( t r t z = , and the phase, ) (t r , have to be Rayleigh and uniform, respectively,
independent of each other (note that a univariate Rayleigh PDF for the envelope by itself does not
implythat ) (t r isazero-meancomplexGaussianprocess[32]).Accordingtothedataanalysisresults
partlyreportedin[33],records#0011,#0017,#0019,#0021,and#0022meettheaboverequirements.
So, for these records, equation (16), and subsequently (19) and (20) hold exactly, while for other
records,thethreeequationsmaybeconsideredasapproximations.However,ourcollecteddatathatis
Page13of27
analyzed and discussed in the sequel, suggest that irrespective of the apparent underlying probability
distribution for the complex envelope ) (t r , the parametric models given in (19) and (20) give very
goodfitstotheautocovariancefunctionoftheenvelope-squared.
In (19) and (20), the maximum Doppler frequency is 20 =
m
f Hz. The unknown parameters
) , (
p
in (19) for the simple non-isotropic model and ) , , (
p
in (20) for the composite non-
isotropic/isotropic model were estimated by the nonlinear least squares method (which provides
consistent estimates [34, p. 207]):
Simple
) , (
MSE min arg )
, (
p
p

= where
=
n
scat
z z z z
n
1
2 1
Simple
)] (
~
) (
~
[ MSE
2 2 2 2

with ) (
~
2 2

scat
z z
given in (19), while
Composite
) , , (
MSE min arg )
, (

=
p
p
where
=
n
scat
z z z z
n
1
2 1
Composite
)] (
~
) (
~
[ MSE
2 2 2 2

with ) (
~
2 2

scat
z z

given in (20) (note that MSE stands for mean-squared-error). In the estimation procedure we had
350 = n and 1 ..., 1, s, 10 14 . 0
3
1
= =

+
n

. This indicates that the estimation was carried out
overthecoherencetime [1,p.72],whichranges from 0 = through
m
f 1 = .Theestimatedvalues
)
, (
p
and )
, (
p
for bothmodels are listed in Table I. The mean-squared-error of the Clarkes
model, defined by
=
n
m
z z
f J n
1
2 2
0
1
Clarke
)] 2 ( ) (
~
[ MSE
2 2

, is also given in Table I, together
with the minimum values of
Simple
MSE and
Composite
MSE for all twelve data sets. In Fig. 9 and for
severaldatarecords, ) (
~
2 2

scat
z z
sin(19)and(20)areplottedusingtheestimatedparameters )
, (
p
and
)
, (
p
,togetherwith ) (
~
2 2

z z
andtheassociatedClarkesresult ) 2 (
2
0

m
f J .Thecorrelationplots
ofrecords#0012,#0020,#0021,andalso#0015arenotshowninFig.9,asthefirstthreearesimilarto
thecorrelationplotof#0011,whereastheforthoneresemblesthatof#0014.
E.Discussion
Asexpected, ) MSE min( ) MSE min( MSE
Composite Simple Clarke
> forallthetwelvesetsofdata.Note
thesignificanterrorofClarkesisotropicscatteringmodelinpredictingtheamountofcorrelationfor
most of the cases. For example, for records #0011 and #0014, the maximum error in predicting the
correlations is about 0.5, which is not negligible at all. Regarding the characteristics of the two
proposed models, let us first focus on the simple non-isotropic model. In this case, only for few
situations the maximum deviation of the model predictions from the empirical correlations over the
Page14of27
range 1 0 < <
m
f is about 0.2, andfor the rest it is less than 0.1. In general, for 1 > , the empirical
correlations approach zero slower than the predictions of Clarkes model. Notice the very slow
convergence of the empirical correlations to zero for record #0018, 3 . 3 = , followed well by the
simplenon-isotropicmodel.Onlyforonecase(record #0019), issmallandequalto0.6,yielding
a result close to that given by Clarkes model. One disadvantage of the simple non-isotropic model,
whichgaveusthemotivationtodevelopthecompositenon-isotropic/isotropicmodel,isitsinabilityto
follow the oscillations observed in the empirical curves of records #0013 - #0017 and #0022. On the
otherhand,thecompositenon-isotropic/isotropicmodelisabletofollowallkindsofvariationsofthe
empirical curves. Specifically, note the improvements obtained in data fitting for records #0013 -
#0017, #0019, and #0022. Of course those improvements are achieved at the expense of adding the
parameter tothesimplenon-isotropicmodel.
Notethatsincethesimplenon-isotropicmodelshouldaccountfortheincomingwavesattheMS
from all directions just by a single von Mises PDF, estimated values are all small for this model
( 3 . 3 ), which correspond to large AOA spreads (
o
63 2 ). On the other hand, since the
composite non-isotropic/isotropic model is equipped with two PDF for modeling the directions of
incoming waves at the MS, von Mises PDF can take care of the directional reception much better,
whichinturnyieldsmoderateandlargevaluesfor ( 88 12 and 700 = ).Thiscorrespondsto
moderate and small values for the AOA spread of the directional incoming waves at the MS
(
o o
33 2 12 and
o
3 . 4 2 = ).
Regardingthephysicalinterpretationoftheestimatedvalues
p
forbothmodelsisTableI,letus
assumethata
o
20 toleranceinestimatingtheapproximateorientationstothefixedtransmitter,listed
inTableI,isacceptable(ofcoursethetolerancerange
o
20 shoulddependon ,butherewefixitto
make a simple conclusion). In Table I, a perpendicular direction corresponds to either
o
90 =
p
or
o
270 =
p
,andaparalleldirectionisassociatedwitheither
o
0 =
p
or
o
180 =
p
.Nowitiseasyto
verify that within the tolerance limit
o
20 , the simple non-isotropic model correctly estimates
p

only for records #0013, #0014, #0018, and #0021, while the composite non-isotropic/isotropic model
successfully estimates
p
for #0011 - #0013, #0016, #0017, #0020 and #0022. This confirms the
Page15of27
superiorityofthecompositenon-isotropic/isotropicmodelincharacterizingthepropagationofwaves
inwirelesschannels.
VI.CONCLUSION
In this contribution we have proposed the application of the von Mises PDF for modeling the
distributionoftheanglesofarrivalofscattercomponentsreceivedatthemobilestation.Thisflexible
parametricfamilyofangularPDFsincludesavarietyofnon-isotropicscatteringscenarios,including
the isotropic one as a special case. This PDF for angles of arrival results in closed-form and easy-to-
use formulas for the correlation function and power spectrum of the complex envelope of scatter
components at the mobile station. These formulas include Clarkes Bessel function-based correlation
and U-shaped power spectrum as special cases. As a consequence, those correlation or spectral
parameters which are important for various system calculations can be easily derived. For example,
spectralmomentsofthereceivedsignalatthemobilestation,usefulforestimatingthespeedofmobile,
canbeexpressedinsimpleclosedforms.Themathematicaltractabilityofthecorrelationfunctionand
power spectrum derived in this paper make them useful for other applications such as efficient
simulation of fading channels [36], Karhunen-Loeve expansion of the multipath fading process [37],
derivation of closed-form expressions for the error probabilities in channels with significant Doppler
spread, hence having fast variations [38] [39, pp. 523-524] [40, pp. 365-367] (in place of numerical-
onlycalculations[41]),etc..Theutilityoftheproposedmodelandnew resultshavebeenverified by
fitting the parametric model of the correlation function to the estimates of this function based on
measureddata.
REFERENCES
[1] G.L.Stuber,PrinciplesofMobileCommunication.Boston,MA:Kluwer,1996.
[2] J. S. Sadowsky and V. Kafedziski, On the correlation and scattering functions of the WSSUS
channelformobilecommunication,IEEETrans.Vehic.Technol.,vol.47,pp.270-282,1998.
Page16of27
[3] M. Patzold, Y. Li, and F. Laue, A study of a land mobile satellite channel model with
asymmetrical Doppler power spectrum and lognormally distributed line-of-sight component,
IEEETrans.Vehic.Technol.,vol.47,pp.297-310,1998.
[4] M.Patzold,U.Killat,Y.Li,andF.Laue,Modeling,analysis,andsimulationofnonfrequency-
selective mobile radio channels with asymmetrical Doppler power spectral density shapes,
[5] F.P.Fontan,M.A.V. Castro,J.Kunisch,J.Pamp,E.Zollinger,S.Buonomo,P.Baptista,and
B. Arbesser, A versatile framework for a narrow-and wide-band statistical propagation model
fortheLMSchannel,IEEETrans.Broadcasting,vol.43,pp.431-458,1997.
[6] W. R. Braun and U. Dersch, A physical mobile radio channel model, IEEE Trans. Vehic.
Technol.,vol.40,pp.472-482,1991.
[7] A.Kuchar,E.A.Aparicio,J.P.Rossi,andE.Bonek,Azimuth,elevation,anddelayofsignals
at mobile station site, in Wireless Personal Communications: Emerging Technologies for
Enhanced Communications. W. H. Tranter, T. S. Rappaport, B. D. Woerner, and J. H. Reed,
Eds.,Boston,MA:Kluwer,1999,pp.99-110.
[8] J. P. Rossi, J. P. Barbot, and A. J. Levy, Theory and measurement of the angle of arrival and
time delay of UHF radio waves using a ring array, IEEE Trans. Antennas Propagat., vol. 45,
pp.876-884,1997.
[9] J. Fuhl, J. P. Rossi, and E. Bonek, High-resolution 3-D direction-of-arrival determination for
urbanmobileradio,IEEETrans.AntennasPropagat.,vol.45,pp.672-682,1997.
[10] Q.Spencer,M.Rice,B.Jeffs,andM.Jensen,Astatisticalmodelforangleofarrivalinindoor
multipath propagation, in Proc. IEEE Vehic. Technol. Conf., Phoenix, AZ, 1997, pp. 1415-
1419.
[11] P.C.FanninandA.Molina,Analysisofmobileradiochannelsoundingmeasurementsininner
cityDublinat1.808Ghz,IEEProc.Commun.,vol.143,pp.311-316,1996.
[12] J. G. Wang, A. S. Mohan, and T. A. Aubrey, Angles-of-arrival of multipath signals in indoor
environments,inProc.IEEEVehic.Technol.Conf.,Atlanta,GA,1996,pp.155-159.
Page17of27
[13] S.Guerin,Indoorwidebandandnarrowbandpropagationmeasurementsaround60.5GHzinan
empty and furnished room, in Proc. IEEE Vehic. Technol. Conf., Atlanta, GA, 1996, pp. 160-
164.
[14] T. Lo and J. Litva, Angles of arrival of indoor multipath, Electron. Lett., vol. 28, pp. 1687-
1689,1992.
[15] D. O. Reudink, Large-scale variations of the average signal, in Microwave Mobile
Communications.W.C.Jakes,Jr.,Ed.,NewYork:Wiley,1974,pp.79-131.
[16] W.C.Y.Lee,Findingtheapproximateangularprobabilitydensityfunctionofwavearrivalby
usingadirectionalantenna,IEEETrans.AntennasPropagat.,vol.21,pp.328-334,1973.
[17] P.Petrus,J.H.Reed,andT.S.Rappaport,Effectsofdirectionalantennasatthebasestationon
theDopplerspectrum,IEEECommun.Lett.,vol.1,pp.40-42,1997.
[18] J.C.LibertiandT.S.Rappaport,Ageometricallybasedmodelforline-of-sightmultipathradio
channels,inProc.IEEEVehic.Technol.Conf.,Atlanta,GA,1996,pp.844-848.
[19] K. Anim-Appiah, Complex envelope correlations for non-isotropic scattering, Electron. Lett.,
vol.34,pp.918-919,1998.
[20] M.D.AustinandG.L.Stuber,Velocityadaptivehandoffalgorithmsformicrocellularsystems,
[21] K.V.Mardia,StatisticsofDirectionalData.London:Academic,1972.
[22] M. K. Simon, S. M. Hinedi, and W. C. Lindsey, Digital Communication Techniques- Signal
DesignandDetection.UpperSaddleRiver,NJ:PTRPrenticeHall,1995.
[23] A. J. Viterbi, Phase-locked loop dynamics in the presence of noise by Fokker-Planck
techniques,Proc.IEEE,vol.51,pp.1737-1753,1963.
[24] H.LeibandS.Pasupathy,ThephaseofavectorperturbedbyGaussiannoiseanddifferentially
coherentreceivers",IEEETrans.Inform.Theory,vol.34,pp.1491-1501,1988.
[25] F.Ayres,Jr.andE.Mendelson,TheoryandProblemsofDifferentialandIntegralCalculus,3rd
ed.,NewYork:McGraw-Hill,1990.
[26] I.S.Gradshteynand I.M.Ryzhik,TableofIntegrals,Series,andProducts,5thed.,A.Jeffrey,
Page18of27
Ed.,SanDiego,CA:Academic,1994.
[27] A. Abdi and M. Kaveh, A new velocity estimator for cellular systems based on higher order
crossings,inProc.AsilomarConf.Signals,Systems,Computers,Pacific Grove,CA,1998,pp.
1423-1427.
[28] H. Allen Barger, Dynamic characteristics of a narrowband land mobile communication
channel,IEEETrans.Vehic.Technol.,vol.47,pp.216-224,1998.
[29] G.L.Stuber,privatecommunication,1999.
[30] A. Abdi and M. Kaveh, On the utility of the gamma PDF in modeling shadow fading (slow
fading),inProc.IEEEVehic.Technol.Conf.,Houston,TX,1999,pp.2308-2312.
[31] A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Processing. Englewood Cliffs, NJ:
PrenticeHall,1989.
[32] S. H. Lin, Statistical behavior of a fading signal, Bell Syst. Tech. J., vol. 50, pp. 3211-3270,
1971.
[33] A. Abdi, H. A. Barger, and M. Kaveh, Signal modeling in wireless fading channels using
spherically invariant processes, in Proc. IEEE Int. Conf. Acoust., Speech, Signal Processing,
Istanbul,Turkey,2000,pp.2997-3000.
[34] R.GoninandA.H.Money,Nonlinear
p
L -NormEstimation.NewYork:MarcelDekker,1989.
[35] R.VonMises, berdieGanzzahligkeitder AtomgewichtundverwandteFragen,Physikal.
Z.,vol.19,pp.490-500,1918.
[36] K. W. Yip and T. S. Ng, Karhunen-Loeve expansion of the WSSUS channel output and its
applicationtoefficientsimulation,IEEEJ.Select.AreasCommun.,vol.15,pp.640-646,1997.
[37] M.Visintin,Karhunen-LoeveexpansionofafastRayleighfadingprocess,Electron.Lett.,vol.
32,pp.1712-1713,1996.
[38] P.A.BelloandB.D.Nelin,Theinfluenceoffadingspectrumonthebinaryerrorprobabilities
of incoherent and differentially coherent matched filter receivers, IRE Trans. Commun. Syst.,
vol.10,pp.160-168,1962.
[39] W.C.Jakes,Jr.,Ed.,MicrowaveMobileCommunications.NewYork:Wiley,1974.
Page19of27
[40] M.D.Yacoub,FoundationsofMobileRadioEngineering.BocaRaton,FL:CRC,1993.
[41] T. B. Welch, M. J. Walker, and R. E. Ziemer, Effects of directional antennas with realizable
beam patterns on the spaced-time correlation function, in Wireless Personal Communications:
Emerging Technologies for Enhanced Communications. W. H. Tranter, T. S. Rappaport, B. D.
Woerner,andJ.H.Reed,Eds.,Boston,MA:Kluwer,1999,pp.1-9.
Page20of27
TABLEI
ESTIMATEDVALUESFORTHEPARAMETERSOFTHESIMPLENON-ISOTROPICANDCOMPOSITENON-
ISOTROPIC/ISOTROPICMODELS,TOGETHERWITHTHEESTIMATIONERRORS
Clarkes
model
Thesimplenon-isotropic
model
Thecompositenon-isotropic/isotropic
model
Record
no.
Approximate
orientationto
transmitter
MSE
Clarke

p

min(MSE
Simple
)
p

min(MSE
Composite
)
0011 Perpendicular 1.2E-1 2.4 19.8
o
3.3E-4 38 280.8
o
0.80 3.3E-4
o
1.1E-3 29 266.4
o
0.78 2.5E-4
0013 Parallel 4.6E-2 1.2 0
o
1.5E-2 700 0
o
0.50 3.0E-3
0014 Parallel 1.6E-1 2.4 0
o
7.0E-3 12 331.2
o
0.80 2.9E-4
0015 Perpendicular 1.6E-1 2.4 0
o
1.7E-2 36 331.2
o
0.72 9.7E-4
o
9.8E-3 700 108.0
o
0.42 3.1E-3
o
8.9E-3 700 79.2
o
0.56 4.1E-4
0018 Parallel 2.5E-1 3.3 0
o
3.4E-3 88 252.0
o
0.86 7.8E-4
o
2.7E-3 700 237.6
o
0.30 7.6E-4
o
1.5E-3 59 79.20
o
0.74 1.5E-4
0021 Parallel 1.2E-1 2.1 10.8
o
2.0E-4 39 259.2
o
0.80 1.8E-4
o
1.1E-2 700 252.0
o
0.60 1.3E-3
Page21of27
Fig.1. VonMisesPDFfortheangleofarrivalofscattercomponentsatthemobilestation( 0 =
p
).
0 =
5 . 0 =
----- 3 =
----
10 =
Fig.2. Von Mises PDF in polar coordinates for the angle of arrival of scatter components at the
mobilestation( 0 =
p
).
0 =
5 . 0 =
----- 3 =
----
10 =
3 2 1 0 1 2 3
0
0.2
0.4
0.6
0.8
1
1.2
p
t
a
c
s
0 0.2 0.4 0.6 0.8 1 1.2

0.2
0.1
0
0.1
0.2
p
t
a
c
s
Page22of27
Fig.3. The PDF of angle of arrival of scatter components at the mobile station for different urban
scenarios.
Clarkesscenario: 0 =
S1scenario: 3 =
-----S2scenario: 3 =
----
S3scenario: 3 =
3 2 1 0 1 2 3
0
0.1
0.2
0.3
0.4
0.5
0.6
p
t
a
c
s
Page23of27
Fig.4. Real part of the normalized autocorrelation function of the complex envelope at the mobile
stationduetomultipathcomponents(scatterandspecular),fordifferenturbanscenarios.
Clarkesscenario: 0 = K , 0 =
S1andS2scenarios: 0 = K , 10 =
S3scenario: 0 = K , 10 =
Fig.5. Real part of the normalized autocorrelation function of the complex envelope at the mobile
stationduetomultipathcomponents(scatterandspecular),fordifferenturbanscenarios.
Clarkesscenario: 0 = K , 0 =
S1andS2scenarios: 1 = K , 3 = , 0
0
=
S3scenario: 1 = K , 3 = , 2
0
=
0 0.2 0.4 0.6 0.8 1
f
m
1
0.5
0
0.5
1
e
R

r
r
0 0.2 0.4 0.6 0.8 1

f
m
0.75
0.5
0.25
0
0.25
0.5
0.75
1
e
R

r
r
Page24of27
Fig.6. Normalized power spectrum of the complex envelope at the mobile station due to scatter
components,fordifferenturbanscenarios.
S1scenario: 10 =
S2scenario: 10 =
S3scenario: 10 =
Fig.7. Normalized power spectrum of the complex envelope at the mobile station due to scatter
components,fordifferenturbanscenarios.
S1scenario: 3 =
S2scenario: 3 =
S3scenario: 3 =
0.75 0.5 0.25 0 0.25 0.5 0.75
f f
m
0
1
2
3
4
f
m
S
r
r
t
a
c
s
0.75 0.5 0.25 0 0.25 0.5 0.75

f f
m
0
0.5
1
1.5
2
2.5
3
3.5
f
m
S
r
r
t
a
c
s
Page25of27
Fig.8. MapoftheArlingtonsiteforcollectionoftheurbandatacontainedinfiles#0019-#0022.
S.NEDDERMAN
WEST3RDST.
WESTFIRSTSTREET
WESTSECONDST
W.BORDERST.
W.SOUTHST.
W
.
N
E
D
D
E
R
M
A
N
D
R
.
C
O
O
P
E
R
S
T
.
Y
A
T
E
S
S
T
.
C
O
L
L
E
G
E
S
T
.
S
.
W
E
S
T
S
T
.
S
.
O
A
K
S
T
.
S
.
P
E
C
A
N
S
T
.
1/4MILE
FILE0019
15MPH
AREAOF
SIGNAL
<90dBm
TRANSMITTERSITE
ENGINEERINGBLDG.
FILE0020
15MPH
FILE0021
15MPH
FILE0022
15MPH
6
2
3
2
2
2
3
4
4
2
1
2
2
3
3
6
4
5
3
2
ApproximateScaleinKilometers
0
0.1 0.2 0.3 0.4 0.5 0.6
2
1
1
1
2
2
2
2
2
2
2
1
3
1
Page26of27
Fig.9. Normalizedautocovarianceoftheenvelope-squaredatthemobile,fordifferentrecords.
Empirical(obtainedbythetime-averagedestimatorofthecorrelationfunction[31,p.743])
Thesimplemodelin(19),with ) , (
p
estimatedbynonlinearleastsquares
Thecompositemodelin(20),with ) , , (
p
estimatedbynonlinearleastsquares
Clarkesmodel
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0011
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0013
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0014
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0016
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
Page27of27
Fig.9.Continued.
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0017
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0018
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0019
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m
0 0.2 0.4 0.6 0.8 1

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Record #0022
N
o
r
m
a
l
i
z
e
d

a
u
t
o
c
o
v
a
r
i
a
n
c
e

o
f

e
n
v
e
l
o
p
e
s
q
u
a
r
e
d
f
m

T VT NonIsoCorrSpec Revised

Caricato da

Informazioni sul documento

Descrizione originale:

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

T VT NonIsoCorrSpec Revised

Caricato da

Copyright:

Formati disponibili

Page1of27

(extremely non-isotropic scattering), where (.) is the Dirac delta

( ) which are approximately equal to 1 , assuming is large [21,

0 0.2 0.4 0.6 0.8 1 1.2

0 0.2 0.4 0.6 0.8 1

0.75 0.5 0.25 0 0.25 0.5 0.75

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

Potrebbero piacerti anche