Professional Documents
Culture Documents
I. INTRODUCTION
A . Previous Models
When measuringtheenvelopeofatransmittedsinusoidal
carrier as seen by a mobile radio antenna in an urban environment, it is observed that this signal is time-varying, although
the transmitted envelope is not [ l ] , [2], [SI, [ 6 ] . It is also
observed that when the receiving antenna is standing still, the
envelope is almost constant [ 6 ] .
This is a fairly strong implication that the envelope
is constant in time, but varies spatially, and a model for this spatial
variation is thus called for. Of course the spatial variation of
the envelope is caused by multipath propagation, and statistical models based on this assumption have been proposed [ 1 ] [31 ~ 5 1 .
Ossanna [ l ] was the first t o propose a statistical model in
termsofasetofverticallypolarizedhorizontallytraveling
planeinterfering
waves. His model is primarilysuitedfor
suburban areas, and the basic assumptionis that just one house
I
Fig. 1 .
A componentwave,
where
E,(t) = cos
r::
- cos act
+(
p
,
183
b
'
Fig. 3.
and c
y
, and P, have theprobabilitydensityfunctions(pdfs)
p a ( . ) and p a ( . ) , respectively.
The resulting field in a point (xo, y o , zo) can now be written
M
~ ( r =)
(4)
En([)
n=l
where
E,(t) = C,
COS
w,t
2n
- -- (xo COS a,
COS
p,
B A S E S A h D FREQUENCY IN h E R T Z
Fig. 2.
+ y o sin a, cos 0,
Comparison
between
theoretical
(dashed)
and
measured
power spectra (Clarke's model).
(5)
B. A Modified Model
The previous models have in common that the component
waves are traveling horizontally. In an urban environment this
is not true, since then no transmission between a distant transmitter and a receiver on the street would be possible.
Thenewmodelproposedhere
is simplyClarke'smodel
generalized tothe casewheretheverticallypolarizedplane
waves are not necessarily traveling horizontally. A typical component wave can be seen in Fig. 3. This is a plane wave and has
an angle of arrival a, to the x-z plane and pn to the x-y plane,
and with each wave is associated an amplitude c, and a phase
shift (P, . The parameters a,, 0, pn . and c, are all random and
statistically independent. Moreover.
+ zo sin p,) + pn
(6)
where
N
~ , ( t=
)
c, cos ( a n t
+ e,)
n=l
c, sin ( a n t + 6,)
T,(t) =
n=l
and
2nv
W,
27120
where 1V is the fixed number of waves and Eo is a positive constant. The p,, are rectangularly distributed throughout 0 to 211,
e n =-
sin Pn + (Pn
n = 1, 2, ..., A!
(8)
184
IEEE TRANSACTIONS
ON
According to the central limit theorem, T,(t) and T,(t)are approximately Gaussian if N is sufficiently large. Actually, this
approximation is very good for N 2 6 [ 7 ] . Theseprocesses
will be regarded as purely Gaussian.
Since the process E(?) is Gaussian, it is completely characterizedbythemeanfunction
E{E(t)} andautocorrelation
function E{E(t) E(t + T ) } . From ( 6 ) we have
E(E(t)} = E(T,(t)}
COS
act- E{T,(t)}
(9)
sin w,t = 0
since E{T,(t)} = E{T,(t)} = 0 and
- cos
+ 7 ) ) COS W c 7
T , ( ~ + T ) } sin W,T
W,T
- C ( T ) sin W,T
(10)
EO
= - E{ sin
3, AUGUST 1979
From (1 2 ) we have
and
the
correlational
properties
of
the
process
are
thus
depicted by U ( T ) and ~(7).
Straightforward
calculations
give (see
Appendix
where
I)
C(T)
(1 6 )
and
UT}
2nv
wo = - cos (y
Bo
- (Yo) cos 00
h
2nzo
= XsinPo + q 0 .
+ Tsd(t))sin w,t
where J o ( * ) is a Bessel function of the first kind of zero order.
The envelope R ( t ) and the phase function (At) are defined
through the transformation
= TCr(t)cos w,t
(19)
Again theenvelope
through
Rr(t) andthephase
* ( t ) aredefined
ao[rl
I
I
Eo
-7-
185
186
I --.
If I>;
O,
Eo
A
-
4 sin p,
v'
- cosp,
A
< I f 1 <-v
"=
Fig. 5 .
distribution,
r 2 0, ( 2 2 )
and again this pdf is not affected by PO(.).l o ( * )is the zerothorder modified Bessel function of the first kind.
111. RAYLEIGH FADING
A. Envelope Correlation
in thedomain
ao(7) =
3
2 Jo
(y),
elsewhere
O,
SO
IT
= -a(O)F
0,
elsewhere,
(;;
--, --,
1,
(%)2)
187
- T
- _V
X
A
188
33 d5
- _
V
(d)
Fig. 6 .
Continued.
189
- Eo
_2
1
Fig. 7.
=-
So(f> = c A o ( f ) * A o ( f )
a2(7) = - aZ(7).
Thesquaredenvelope
(29)
EO
240)
is alsoofinterest,sinceonnoisyfading
cer;K(Jq$$)
IfKy
2v
0,
elsewhere
(33)
E{R2(t)} = h ( 0 )
+~ ~ ( 7 ) )
so
(30)
scale.
are approximatelyThevalue
of SV, near f = 2vb is proportionalto
cos &,/sin2 Om, so for small values of P, most of the energy
The Power spectrum ofR(t) andR2(t) Can thus be written near f = 0 in ,Soy? has been shifted to f = 2vb. For not too
small values of P,, SV, is rather flat, and to observe, finite.
S ( f ) = FCC a2(7)} = C A ( f ) * A ( f )
(32)
B. Phase Correlation
where C equals nlEo or 4.
The two-dimensional pdf for the phaseis [8]
&Owing us that both
proportional to a2(7).
pV(t),@(t+dPl*
92)
rR
('1 and
' R 2(')
190
1 0 ~ l O g l O ~; ~Kl = C(=I
Eo 2.i
- -2va
10 dB
&
191
25 dB
- 25 dB
(4
Fig. 8. Continued.
192
4T2
4s'
-T
Of course, r9(7) changes very much for large 7 even for small
P,, so the power spectrum of the phase F{rv(7)} is probably
a(7)
significantlychangedevenforsmallvaluesof
.0, However,
p ( 7 ) = -.
a@)
(35) an expression
for found.
been
F{r9(7)}
not has
Further insightof the phasecorrelationcanbeachieved
Since the phase (At) is rectangularly distributed throughout by looking at the process
~ d= dt
~ + )7)- dt).m e pdf for
0 to 2n
from found
be Adt) can
(34) [SI
where
E{cp(t)l = n
4n2
E{p2(t)} = -.
3
(36)
and it is seen that for futed 7,p(7) determines the shape of the
pdf.Fig.10shows
(Acp) for
various 7,usingaO(7).For
small values of .4,...-7 there is no significant difference in this
pdf, using a(7) instead of ao(7).
should
and
itnoticed
be that
this
expression
is independent
of
a(0).
To beableto see anydifferencebetween r9(7), using a(7)
or u0(7), 0, has to be chosen comparatively large (see Fig. 9).
r>O
--<i,ip<-
1 Thedotindicatesderivationwithrespectto
derivation with respect to T .
OGcpG2n
(39)
r, and theprime
193
-2n
271 Aq
at weand
once have
and
1
e-(1/~)[(r2/a(~))-(rip2/a(~))~,
r>O
p p ( o ) = I1 F l
Gom
-=<ip<-m
elsewhere
fromwhichthestatisticalindependence
processes
can
easily
be
seen. Especially,
we
have
yields
of p p ( .from
) but
(12) we
194
NO.
VT-28,
3 , AUGUST 1979
- $(t +
T)}
and by integration
O<r<m
r) (49)
where
e(*)
$ ( t ) of course is independ-
A. Envelope Correlation
As was mentioned earlier, it is often more valuable to know
the autocorrelation function for the squared envelope, rather
Fig. 11. Thepdf for thephase. (a) Ed/Eo = 1 and 2 (dashed) atan instant to. (b) Ed/EO = 1 and 2 (dashed at = t , =
(ni2wo) + n.(2n/wo);n= 0, f l , *2, '...
195
196
TRANSACTIONS
IEEEVEHICULAR
ONTECHNOLOGY,
VT-28,
VOL.
thanfortheenvelopeitself.FromAppendix
Itseemsplausiblethatthe
phase $ ( t ) = vr(t)- w o t should
be a stationary process, since
111 wehave
(59)
0,
elsewhere
Now, by taking the Fourier transform of (57), the pow-er spectrum is easily found as
Previousmodels
[1]-[3]forthefadingenvelopeata
mobile radio channel have been extended to include a multipath propagation mode? often present in the streets of a city
with tall buildings. It was assumed that the
waves in such an
environment also could have a vertical component of propagawhere Scf) is thepowerspectrumofthesquaredenvelope
when no specular component is present. This spectrum is given tion direction, and the effects of this extension were studied,
with the previous models as a reference and special case.
by (3% (33).
InFig.13
we cancomparethesespectrawhen
It was found that the pdf of the envelope and the phase
pp(0) =
wereleftunchanged,whereasthecorrelationalpropertiesof
S@) or p p @ ) (59). It can be noticed that when
p p ( p ) = S@),
the spectrum is not finite f o r f = 0, I f 1 = (v/h)(1 cos (a0 - these quantities were not.
Examining the autocorrelation of the envelope by means of
y) cos Po), whereas the spectrum is always finite when pp(0)
the power spectrum, it was found that the shape of the specis given by (59).
trum was heavily affected even for
small deviations from the
purely
horizontal
mode
of
multipath
propagation. Concerning
B. Phase Correlation
theautocorrelationofthephase,thepowerspectrumcould
not be found analytically, but it was made plausible that this
From Fig. 14 where the phasors of the specular component
same way,i.e.,alower
the Rayleigh component R ( t ) ,and the resulting envelope spectrumshould be affectedinthe
frequency spectrum should be achieved.
R,(r) areshown,someinterestingfactsabouttheresulting
It is assumed that this new extended model should be used
phase cp,(t) canbeseen.
As wasnoticedearlier,thephase
function q,(t) is not stationary, since its pdf is a function of 1. in an urban environment, but of course, measurements have to
This is because the phasor of the specular component is rotat- be made to confirm whether this new model fits significantly
better.
ing with an angular frequency ofw o rad/s.
197
rRr(
15
Fig. 12. Autocorrelation for the envelope. (a) Ed/Eo = 0,pa(@)= 6(@) and p a @ )given by (59), &,= 4 5 O (Qshed). (b)
Ed/Eo = 2, pp03) = 6(8) and p a @ )given by (591, Pm = 45" (dashed). 1: QO - Y = 0,bo = 0.2: QO - = ,,/2,
= 0.
-,
198
*.)
t
.
)
(Eo12*L
G
v
20 dB
-4
199
200
IEEE TRANSACTIONS
ON
VEHICULAR
TECHNOLOGY,
VOL.
NO.
VT-28,
3, AUGUST 1979
to 2n, i.e.,
OG(PG2n
elsewhere.
E{C,C,
m,n=l
APPENDIX I
Tc(t + 7)) =
x
N
Tc(t)=
C, COS ( a n t
+ e,)
n=l
where
Eo
2N
x
N
E{ cos w,7)
=n=l
2nv
wheretheleastexpressioncomesfromcarryingoutthe
expectation over c,. Since all w, areidenticallydistributed
subscripts can be deleted yielding
a(7) = EO E{ COS U T } .
This expression cannot be further simplified without specifying Pa(*) and Pp(.).
20 1
In much
the
+ 0, + e,)) + -1 2N
2
E{c,c,
T gives
cos,
+ 7)
sin (w,(t
m,n=l
- w,t + en - e,))
(70)
and, using the stochastic variables S1 and S 2 from (67), bringing out the expectation with respect to9,, 9, gives
C(T) =
l N
E ( c , ~sin O,T)
EO
= -E(
2
n=l
sin
UT)
(71 1
y=
sin e
d(x, Y ) --
fi
d@, 0)
cos e
APPENDIX I1
POWER SPECTRUM A ( f ) WHEN ~ p @ )
(77)
COS
.
1
Adf)=
a(7) =
2nsinPm
Eo
I"
x = a c o s
I"
pm r = - G
_E,
f~Jo (h
277vr cos 0)
4 sin p,
is wanted. Since
cos P
m]
[a
MI,
U
IfPx
(73)
we have
EO
277 sin 0,
EO
2n sin p,
(74)
/m
JxU
dx
- .
V
If I Q - cosp,.
(79)
202
IEEE TRANSACTIONS
VEHICULAR
ON
Fig. 18.
Phasor diagram.
Fig. 17.
Theintegration area.
Now, since
of the
+2aE{R(t))
(83)
2x dX
=JT
E{R(t)}
E{R (t)}= 2 4 0 )
x '
If'>;
V
<If\<-
-cosp,
x
wherethesecondterm
since (see (1 3))
A 0=
'
arcsin
we have
V
IfI<-cosp,.
(81)
= 2E{Tc2(t) Tc2(t
APPENDIX 111
CALCULATION OF THE AUTOCORRELATION
FUNCTION FOR THE SQUARED ENVELOPE
AT RICIAN FADING
In Fig. 18, R ( f ) is the Rayleigh envelope, (c(f) the Rayleigh
phase,and R.(t)
. . . theresultingRicianfadineenvelone. r~ From
~ - ~ I
+ T)} + 2E2{TC2(t)}
(88)
E{TC2(f) TC2(t+ T ) }
= E{Tc (t)} E{TC2(t + 7))
- -
= aye) (89)
+ 2a2(7)
+ 2l?
ADING
X1
AULIN:
203
CHANNELS
MOBILE RADIO
whereyielding
REFERENCES
[ I ] J . F. Ossanna,Jr.,
[2]
[3]
[4]
(93)
[5]
[6]
we have
[7]
[8]
(91
[lo]
A modelformobileradiofadingdueto
buildingreflections:Theoreticalandexperimentalwaveform
powerspectra, Bell.Syst.Tech.
J., vol. 43, pp. 2935-2971,
Nov. 1964.
R. H.Clarke,Astatisticaltheoryofmobileradioreception,
BellSyst. Tech. J., vol. 47, pp. 957-1000, July 1968.
M. J.Gans, A power-spectraltheoryofpropagation
in the
mobileradioenvironment,
IEEE Trans. Veh.Technol., vol.
VT-21, pp. 27-38, Feb. 1972.
S. 0. Rice,Mathematicalanalysisofrandomnoise,
BellSyst.
Tech. J., vol. 23, pp. 282-332, July 1944; and Bell Syst. Tech.
J., vol. 24, pp. 46-156, Jan. 1945.
-, Statistical Properties of a
sine wave plus random noise,
Bell Syst. Tech. J., vol. 27, pp. 109-157, Jan. 1948.
W. C. Jakes efal., MicrowaveMobileCommunications.
New
York: Wiley, 1974.
T.Aulin,Characteristicsofadigitalmobileradiochannel,
TelecommunicationTheory,
Univ. of Lund,Lund,Sweden,
Tech. Rep. TR-93, Oct. 1977.
W. R.Bennett,Distribution
of thesum of randomlyphased
components, Quart. A p p l . Math., vol. 5 , pp. 385-393, Jan.
1948.
E. W. Ng and M . Geller, A table of integrals of the error functions,J. Res. NBS-B, vol. 73B, pp. 1-20, Jan.-Mar. 1969.
I. S. Gradshteyn and I. W. Ryzhik, Table of Integrals, Series and
Products. New York:Academic, 1965.
E. N. Gilbert,Energyreceptionformobileradio,
Bell Syst.
Tech. J., vol. 44, pp. 1779-1803, Oct. 1965.
= rl
COS
x2 = rl sin 0,
x 3 = r2 COS O 2
x 4 = r2 sin O2
= E { X l X,}
+ E{X,
X 4 } = ={x1
x31
(96)
of themobileradiochannel.