LEVEL-CROSSING RATES AND AVERAGE FADE

DURATION FOR WIRELESS CHANNELS WITH

SPATIALLY COMPLICATED MULTIPATH
Gregory D. Durgin and Theodore S. Rappaport
Mobile and Portable Radio Research Group
Bradley Department of Electrical and Computer Engineering
432 New Engineering Building Virginia Polytechnic Institute and State University
Blacksburg, VA 24061-0350 http://www.mprg.ee.vt.edu
e-mail: [email protected]

Abstract { This paper derives equations that describe paper { Eqn (9) and Eqn (11) { are exact equations
level-crossing rates and average fade durations for mo- describing the level-crossing rates and average fade
bile receivers operating in spatially complicated mul- durations of Rayleigh-fading channels, regardless of
tipath channels. The results are useful for describing the spatial complexity of impinging multipath. Fur-
fading statistics for wireless systems that operate in thermore, we describe these fading statistics in terms
realistic propagation channels or employ smart anten- of three physically-meaningful multipath spatial pa-
nas, adaptive arrays, or other technologies that lead to rameters [7, 8, 9]. The paper concludes by compar-
a non-omnidirectional distribution of multipath power ing the level-crossing rates and average fade durations
in space. to three well-documented cases of narrowband mobile
fading. This comparison demonstrates the simplicity
I. INTRODUCTION and generality of the new equations.
Characterization of level-crossing rates (how often a
received signal fades below a threshold voltage) and II. DIRECTIONAL RAYLEIGH-
average fade durations (the average interval of time FADING CHANNELS
a received signal spends below a threshold voltage)
are important for the design of mobile wireless re- One of the most popular time-varying channel mod-
ceivers that mitigate the eects of small-scale fad- els in mobile wireless communications is the Rayleigh
ing [1]. Mathematical models for these fading statis- fading channel model [3, 10, 11]. It is well known that
tics typically assume that equal amounts of multipath the sum of many small multipath waves will produce
power impinge upon the receiver uniformly from the a voltage envelope that is distributed according to the
horizon in all directions [2]. This idealized multi- Rayleigh probability density function (PDF) [12, 13]:
path channel does not exist in practice, but is useful
for approximating the statistics of narrowband fad-
 2 
ing for mobile receivers that operate in a cluttered r r
environment without a line-of-sight path to the trans- fR (r) = 2 exp , 22 (1)
mitter [3, 4]. As wireless systems evolve, spatial dis-
crimination technology such as adaptive arrays and
smart antennas will be employed at both the base sta- where r is the magnitude of the voltage envelope and
tion and mobile [5, 6]. Under these conditions, non- 22 is the mean-squared received voltage or average
omnidirectional spatial channel models and the fading power. In the past, second-order statistics (statistics
statistics they produce become important to receiver that relate to the evolution over time of a random
design. process) have relied on an omnidirectional distribution
This paper presents several equations that describe of multipath power to develop expressions for level-
crossing rates and average fade durations [1, 2, 14].
fading statistics in wireless multipath channels of any Here we present exact level-crossing rate and average
spatial complexity. The main contributions of this fade duration equations for multipath propagation of
 This work is sponsored by a Bradley Fellowship at Virginia any spatial complexity, using just three basic spatial
Tech and by the MPRG Industrial Aliates Program. channel parameters.
Three Spatial Channel Parameters  The parameter is a measure of angular constric-
tion. This parameter also ranges between 0 and
A level-crossing rate analysis requires the calcula- 1. Angular constriction does not aect the aver-
tion of the mean-square rate-of-change of the received age value of b2 in a local area. Instead, Eqn (2)
time-varying complex voltage [12, 13]. This value is shows that the range of possible b2 values in a
denoted as b2 in the literature [15]. For multipath that local area increases as angular constriction in-
arrive uniformly from all azimuthal angles, the value creases. Increasing values of angular constriction
for b2 is 2 !max
2 , where !
max is the maximum Doppler indicate that multipath power is becoming more
radian frequency (!max = 2v  , v is receiver velocity concentrated about two angles-of-arrival.
and  is the wavelength of the carrier frequency) [14].
The envelope fading statistics of omnidirectional mul-  The parameter max is the azimuthal direction of
tipath propagation are isotropic { that is, the same in maximum fading rate. This parameter is a com-
all directions. Therefore, level-crossing rate and av- panion to the constriction parameter. It indicates
erage fade duration do not depend on the direction the direction a narrowband receiver would travel
traveled by the receiver. to maximize b2 .
For the general case of non-omnidirectional propaga- Regardless of the azimuthal complexity of p(), these
tion, the expression for b2 is more complicated and three parameters suce to describe b2 exactly, under
depends on the azimuthal direction of receiver travel, the standard assumptions of local areas.
. However, it may be shown that the mean-squared
rate-of-change of complex voltage always takes the fol- Exact Analytical Level-Crossing Rates
lowing form (see Appendix for derivation):
The general expression for a level-crossing rate is given
b2 = !max
2
2 2 (1 + cos [2( , max)]) (2) by the following [3, 12, 13]:
Z1
This equation requires three spatial parameters that
are based on the angular distribution of multipath NR = rf
_ RR_ (R; r_)dr_ (7)
power, p(): s
0

 = 1 , jFF12j
2 where R is the threshold level and fRR_ (R; r_) is the
(3) joint PDF of envelope and its time derivative. For a
0
  Rayleigh-fading signal, the level-crossing rate of the
= F 2F , F12 envelope process is
 0 2 2
(4)
F0 , jF1 j r
NR =  b2 exp ,2
,

(8)
max = 12 PhasefF0 F2 , F12 g (5)
The variablep is the normalized threshold level, such
where that  = R= 22 [14].
Z2
Fn = p(0 ) exp(jn0 )d0 (6) By substituting Eqn (2) into Eqn (8), we arrive at an
0 exact expression for level-crossing rates in a Rayleigh
Fn is the nth complex Fourier coecient of the az- fading channel with any azimuthal distribution of
imuthal distribution of multipath power, p(
). multipath power and any direction of mobile receiver
travel, :
The physical meaning behind the three spatial param-
NR = !p max p ,

eters (, , and max) are discussed in detail in [7] and 1 + cos [2( , max)] exp ,2 (9)
2
[9]. A brief summary of the three parameters is in-
cluded below: Average fade duration, , is dened as the following
[3, 15]:
ZR
 The parameter  is a measure of angular spread.  = 1 f (r)dr (10)
R
This parameter ranges between 0 and 1. De- NR
0
creasing values of angular spread indicate that Substitution of the Rayleigh PDF of Eqn (1) and the
multipath power is becoming more concentrated level-crossing rate of Eqn (9) into this denition pro-
about a single direction. The mean-squared rate- duces
of-change of complex voltage in a local area de- p  , 2

creases proportional to the square of angular 2 exp  , 1
spread.  = p (11)
!max 1 + cos [2( , max)]
Eqn (9) and Eqn (11) are useful tools for studying pendicular to the ground. Due to the vertical electric-
small-scale fading statistics in the presence of non- eld polarization of the whip antenna, this propaga-
omnidirectional multipath. tion scenario is referred to as the Ez -case [15].
Eects of Non-omnidirectional Multi- The second two cases correspond to the same narrow-
band receiver in the same omnidirectional multipath
path channel, but with a small loop antenna mounted atop
Non-omnidirectional multipath propagation may have the receiver such that the plane of the loop is perpen-
dramatic eects on the fading statistics according to dicular to the ground. The antenna pattern of the
Eqn (9) and Eqn (11). As angular spread, , de- small loop antenna attenuates the arriving multipath
creases, the channel becomes more directional, caus- such that the angular distribution of power becomes
ing the level crossing rate to decrease and the average p() = A sin2  (13)
fade duration to increase. Thus, in a Rayleigh-fading
multipath channel where power is becoming increas- where A is some arbitrary gain constant. Unlike the
ingly concentrated about a single direction, a narrow- omnidirectional Ez -case, the statistics of this propaga-
band signal fades less frequently but spends longer tion scenario will depend on the direction of travel by
intervals of time below a threshold voltage. the receiver. The Hx -case will refer to a receiver trav-
The eects of angular constriction, , are also insight- eling in a direction perpendicular to the main lobes of
ful. Eqn (9) and Eqn (11) predict that fading statistics the loop antenna pattern ( = 0). The Hy -case will
may depend on the orientation of the traveling mobile refer to a receiver traveling in a direction parallel to
receiver. The discrepancy between fading statistics the main lobes ( = 2 ). Figure 1 illustrates the Ez ,
measured along dierent directions within the same Hx , and Hy cases.
local area increases as the angular constriction, , in-
creases. Three Angular Distributions of Power
Interestingly, level-crossing rates within the same local
area satisfy a type of Pythagorean relationship. If two
level-crossing rates are measured at the same mobile velocity velocity

velocity along perpendicular azimuthal directions, the Rx

Rx velocity
Rx
measurements will satisfy the following relationship:
NR ()2 + NR (  2 )2 = !max  exp ,22 (12)
2 2 2 ,

Ez Hx Hy

regardless of the value of . This Pythagorean rela- Figure 1: Three dierent multipath-induced mobile-
tionship may be veried by substituting Eqn (9) for fading scenarios.
 and  + 2 into the right-hand side of Eqn (12).
Eqn (12) demonstrates how the average level-crossing
rates in a local area are described by only the angular
spread, .
Solutions
The rst step is to calculate the three spatial param-
III. SEVERAL CLASSICAL EXAM- eters from the angular distribution of power, p(),
PLES using Eqn (3)-Eqn (5). The spatial parameters for
the Ez -case are  = 1, = 0, and max = 0. Since
As a point of comparison, this section analyzes three this case is omnidirectional, the angular spread is at a
well-known cases of propagation that have analytical maximum ( = 1) and the angular constriction is at
solutions [3]. The cases are analyzed using the three- a minimum ( = 0). For the Hx - and Hy -cases, the
parameter approach as outlined above. This approach spatial parameters are  = 1, = 12 , and max = 2 .
is shown to produce quick, comprehensive, and { most Since the impinging multipath have no clear bias in
importantly { exact solutions. one direction, the angular spread is at a maximum
just like the Ez -case. However, there is clearly a bias
The rst case corresponds to a narrowband receiver in two directions, resulting in an increased angular
operating in a local area with multipath arriving from constriction of = 21 .
all directions, such that the angular distribution of
power, p(), is a constant. The receiver antenna is After substitution of these parameters into Eqn (9)
assumed to be an omnidirectional whip, oriented per- along with the appropriate direction of mobile travel,
the level-crossing rates for the three cases become IV. CLOSING REMARKS ON NON-
OMNIDIRECTIONAL PROPA-
Ez : NR = !pmax exp ,2
,

(14)
GATION
2 The main results of this paper { Eqn (9) and Eqn (11)
! max  ,
Hx : NR = 2p exp ,2


(15) { provide insight into statistics involving the time-
r evolution of Rayleigh fading. They show that only
! max  3
Hy : NR = 2  exp ,2
,

(16) three basic channel parameters are needed to exactly
describe level-crossing rates and average fade dura-
tions. Furthermore, these three paramters correspond
to physically meaningful multipath angle-of-arrival
The corresponding average fade durations are characteristics, such as angular spread and angular
constriction.
p  ,
 Furthermore, studying the cases of Rayleigh fading
Ez :  = ! 2 exp 2 , 1 (17) produces insight into more complicated fading chan-
p
max nels. Recall that Eqn (2), the relationship between the
Hx :  = !2  exp 2 , 1 three channel parameters and the mean-squared rate-
 ,

(18) of-change of received complex voltage, is a general re-
max
r
sult that does not require Rayleigh fading conditions.
Hy :  = ! 2  3 exp 2 , 1 (19)
 ,

max
Although non-Rayleigh channels are dicult to study
analytically, Eqn (2) demonstrates that the three ba-
sic spatial parameters dominate the time-evolution of
these fading processes as well.
These expressions match the original solutions pre-
sented by Clarke in [15]. A. APPENDIX
Discussion of Three Parameter Solu- The power spectral density (PSD) of a baseband com-
plex received voltage signal is related to the angular
tions distribution of multipath power [2]:
   
The three parameter technique for nding fading p  + cos,1 !max
! + p  , cos,1 !max
!
statistics is an intuitive way to relate the physical S (!) = p
2 , !2
channel characteristics to the fading behavior. In the !max
previous examples, the spatial parameters may be cal- (20)
culated analytically or even estimated intuitively by where  is the azimuthal direction of travel and p(
)
looking at the distributions of multipath power in Fig- is the angular distribution of impinging multipath
ure 1. The use of spatial parameters to nd level cross- power. The second moment of the fading process {
ing rates and average fade durations is quite simple denoted b2 in the literature { is given by the following
when compared to the full analytical solutions of the integration [12, 13, 14]:
Ez -, Hx -, and Hy -cases presendted in [3]. The solu- !Zmax
tion is also more comprehensive. For example, once b2 = (! , !c)2 S (!) d! (21)
the spatial parameters have been found, Eqn (9) and ,!max
Eqn (11) provide statistics for all directions of receiver
travel for the Hx - and Hy - cases, and not just specic where !c is the centroid of the PSD:
directions such as  = 0 or  = 2 .
!Zmax
The solution form of Eqn (9) and Eqn (11) reveals !c = F1 ! S (!) d! (22)
an interesting property about statistics in Rayleigh- 0
,!max
fading channels. Since the three channel parame-
ters only depend on low-order Fourier coecients, F0 is dened by Eqn (6) { this is really just the average
the level-crossing rates and average fade durations of power of the process.
Rayleigh-fading channels are insensitive to the higher-
order multipath structure. The general biases of an- Now insert Eqn (22) into Eqn (21), making the change
gular spread and angular constriction truly dominate of variable 0 =   cos,1 !max
! , where the + and ,
the time evolution of these fading processes. signs correspond to the left and right terms of p(
),
respectively, of Eqn (20). After rearranging the limits [4] D.O. Reudink, \Properties of Mobile Radio
of integration, the equation for b2 becomes Propagation Above 400 MHz," IEEE Transac-
2 2 32
tions on Vehicular Technology, Nov 1974.
Z
F0 !max
2 ! 2
[5] R.B. Ertel, P. Cardieri, K.W. Sowerby, T.S. Rap-
b2 = 2 , F0 p(0 ) cos( , 0 ) d0 5
max 4
paport, and J.H. Reed, \Evolution and Appli-
2 2
0
3 cability of Spatial Channel Models for Wireless
Z Communications," IEEE Personal Communca-
+!
2
max 4
2 p(0 ) cos[2( , 0 )] d0 5 (23) tions Magazine, vol. 5, no. 1, pp. 10{12, Feb 1998.
0
[6] J.C. Liberti and T.S. Rappaport, Smart Anten-
nas for Wireless CDMA Communications, Pren-
Consider a complex Fourier expansion of b2 with re- tice Hall, New Jersey, 1999.
spect to :
( ) [7] G.D. Durgin and T.S. Rappaport, \A Basic Rela-
1
X tionship Between Multipath Angular Spread and
b2 = Real An exp(,jn) Narrowband Fading in a Wireless Channel," IEE
n=0 Electronics Letters, vol. 34, no. 25, pp. 2431{
= A0 + Real fA2 exp(,j 2)g (24) 2432, 10 Dec 1998.
[8] G.D. Durgin and T.S. Rappaport, \Eects of
All of the An are zero for odd n. This is because Multipath Angular Spread on the Spatial Cross-
b2 () = b2 ( + ); that is, a 180 change in the direc- Correlation of Received Voltage Envelopes," in
tion of mobile travel should produce identical statis- 49th IEEE Vehicular Technology Conference,
tics. Furthermore, Eqn (23) has no harmonic content Houston TX, May 1999, vol. 2, pp. 996{1000.
with respect to  for n > 2. Solving for the only two [9] G.D. Durgin and T.S. Rappaport, \Three Pa-
remaining complex coecients produces rameters for Relating Small-Scale Temporal Fad-
Z 2  
ing to Multipath Angles-of-Arrival," in PIMRC
A0 = 2 b2 d = !max
1 2 jF1 j2 = !2 2 2 '99, Osaka Japan, Sep 1999.
F 0,
2 F0 max
[10] B. Sklar, \Rayleigh Fading Channels in Mobile
0
(25) Digital Communication Systems Part I: Charac-
terization," IEEE Communications Magazine,
Z 2   pp. 90{100, July 1997.
=  b2 exp(j 2) d = !max
1 F12
2
A2 F 2,
2 F 0 [11] B. Sklar, \Rayleigh Fading Channels in Mobile
0
= A0 exp(j 2max) (26) Digital Communication Systems Part II: Miti-
gation," IEEE Communications Magazine, pp.
102{109, July 1997.
where , , and max are the three basic spatial chan- [12] S.O. Rice, \Mathematical Analysis of Random
nel parameters dened in Eqn (3)-Eqn (5). If these Noise," Bell System Technical Journal, vol. 23,
two coecients are placed back into Eqn (24), the end pp. 282{332, July 1944.
result is the relationship for b2 in Eqn (2). [13] S.O. Rice, \Mathematical Analysis of Random
REFERENCES Noise (continued)," Bell System Technical Jour-
nal, vol. 24, pp. 46{156, Jan 1945.
[1] T.S. Rappaport, Wireless Communications: [14] W.C. Jakes, \A Comparison of Specic Space Di-
Principles and Practice, Prentice-Hall Inc., New versity Techniques for Reduction of Fast Fading
Jersey, 1996. in UHF Mobile Radio Systems," IEEE Transac-
tions on Vehicular Technology, vol. VT-20, no. 4,
[2] M.J. Gans, \A Power-Spectral Theory of Propa- pp. 81{91, Nov 1971.
gation in the Mobile Radio Environment," IEEE
Transactions on Vehicular Technology, vol. VT- [15] R.H. Clarke, \A Statistical Theory of Mobile-
21, no. 1, pp. 27{38, Feb 1972. Radio Reception," Bell System Technical Jour-
nal, vol. 47, pp. 957{1000, 1968.
[3] W.C. Jakes (ed), Microwave Mobile Communi-
cations, IEEE Press, New York, 1974.