OSSC-2011


The Modeling And Simulation Of Rayleigh Flat
Fading Channels Based On SCILAB
Zhenzi Liu Ping Chen,Haofu Liao,Huichao Zhou
Beijing University of Posts and Telecommunications Beijing University of Posts and Telecommunications
Beijing, China Beijing, China
[email protected] [email protected]

Abstract—In the Wireless Communication, the multipath effects, II. JAKES EMULATOR
Doppler shift, and power loss, causing by the dissemination
environment, have always effected the high speed transportation A. Jakes Model [1],[2],[3]
of the data. On the other hand, the complexity and randomness The Fig. 1 is the generate diagram of the Jakes model.
of the wireless communication environment itself have also made 2
the study of the wireless channel’s simulation model very difficult. 2 cos 1
cos( 2  f 1 t ) N
The Jakes model is an ideal simulation model of small-scale  
Rayleigh flat fading channels, and it can be realized by the  
 T C (t )
superimposition of the limited LFO (low frequency oscillators). 2 cosN0


cos( 2f N 0 t )
So, by the simulation of Doppler shift, the analysis of the
multipath effects and taking the advantages of the Jakes model, 1 2cosN01
cos( 2 f m t )
we designed an ideal small-scale Rayleigh flat fading channel 2
block on the SCILAB. The development of this block has not only
improved the communication simulation library of the SCILAB,
but also provided an effective environment of the channel for the
closed-loop test of the wireless communication system. T (t )  TC (t )  jT S (t )
Meanwhile, the application of this channel block has also 2 sin 1
provided a reference platform for the test and construction of the cos( 2  f 1 t )
 
wireless communication simulation system. 

Keywords—Jakes model; small-scale fading channel; Rayleigh 2sin 
distribution; Doppler shift; SCILAB

cos( 2f N 0 t )
N0 
T S (t )
1 2sinN01
cos( 2 f m t )
I. INTRODUCTION 2 2
The increasingly requirement of the information by people, N
on the one hand, has made a contribution to the development of
the wireless communication services. But on the other hand, Figure 1.The Jakes model composed by N+1 low frequency osscillators
the environment of the wireless communication which brings
the multipath effects, Doppler shift, and noise, has, to some In the Fig. 1,
degree, impeded the development of the wireless
communication services. To improve this kind of condition, the
1  1    (1)
scientists have made a great deal of researches about the N 0 1
characteristics of the wireless channel, and proposed many
useful channel simulation model. The Jakes channel model is
Tc (t )  c n 1
1,n cos(n t ) (2)
just one of them, and it is an ideal simulation model for the
Rayleigh flat fading channel. N 0 1

SCILAB as a free open-source software not only has the

Ts (t )  c n 1
2,n cos(n t ) (3)
basic functions like those commercial software, such as
MATLAB, System View and so on, but also provides its user a
second chance to develop the system. This kind of And
characteristic gives the users a freedom to develop their  2
products. This paper is based on the Jakes channel model,

N 2cos  n n  1, 2, , N 0
which is developed by the SCILAB, and has studied the c1,n   (4)
statistical characteristics and transmission performances of the
2 2 cos  n  N0  1
N0 1
Jakes communication channel model.
 N


 OSSC-2011

 2 cos1t

N 2sin  n n  1, 2, , N 0  Q1c P1c




c2,n (5) 

2 2 sin  sin1t 

N0 1 n  N0  1  Q 1s  P1 s



 N 
 
 

 2 n 

 cos n  1, 2, , N 0 cos  N 0 t
n   m N  Q Nc 0 PNc0


(6)

 m n  N0  1
sin  N 0 t
 Q Ns 0  PNs 0
 n




N n  1, 2, , N 0
cos  N 0 1t
n   0 (7)  Q Nc 0 1 PNc0 1



 n  N0  1

 4 sin  N 0 1t
 QNs 0 1  PNs0 1


The signal generated by Jakes model is neither wide-sense
stationary, nor ergodic. So, we can’t replace the statistical
average by the time average. And its statistical characteristics
can’t meet the requirement of Clarke model: First, its envelope  
doesn’t strictly obey the Rayleigh distribution. Second, its cos  c t
X s (t )  90  X c (t )
autocorrelation function doesn’t approach the Bessel function.
Finally, the in-phase and quadrature component of it has a
correlation and the correlation coefficient doesn’t equal to zero. 

B. The Advancement Of Jakes Model [1],[2],[3]

Figure 2. The advanced model of Jakes
The Jakes emulator makes use of the symmetry of the
Doppler shift and reduces the number of the oscillators. But, phase component, quadrature component, and the receive
meanwhile, it makes the signal unsteady. signal respectively. (N0, v, fc, p, t) are the input parameters,
they represent the number of the LPOs (path number N =4N0 +
Marius F. Pop has proposed a new way to improve the
Jakes model. This way not only considered the symmetry of 2), speed (km/h, maximum Doppler shift f m  vfc c ), carrier
the Doppler shift, but also took the random phase into frequency (MHz), Gaussian white noise power spectral density
consideration, so that it can avoid the correlation problem, of the bilateral (W/Hz), and the sampling time (s) respectively.
which is produced by random phase, in Jakes model. The Fig. 2 Another important point of the Jakes emulator is its
is the advanced model of Jakes. statistical characteristics. The original Jakes emulator belongs
to the certainly model. The disadvantage of it is its imperfect
III. THE REALIZATION AND THE VERIFICATION IN SCILAB statistical characteristics, and will make the channel none wide-
In order to reflect the characteristic of the real wireless sense stationary. In order to make up the statistical unsteady of
channel, the Jakes model has synthesized the multipath effect, the Jakes emulator, we introduced a random phase in the
Doppler shift, and statistical characteristics. So when we advanced Jakes model which is highly consistent with the
realize the Jakes emulator in SCILAB, it must reflect those reference model. So in order to reflect the randomness of the
three characters. channel, it is important to deal with the random phase.

First, we must define the input and output of the Jakes In the reference model, the phase of each path is random
emulator. Then, in order to reflect the Doppler shift, according and average distribution. So we can define the random phase:
to the Jakes model, the Jakes emulator must provide the speed = 2*rand()
of the mobile station and the carrier frequency of the
information which determines the maximum Doppler Spread. The rand() has defined in SCILAB, it is an average
Finally, according to the principle of the Jakes model, the Jakes distribution random generator in (0, 1). So the is an average
emulator uses the limited LFOs to simulate the multipath effect distribution random number in (0, 2). Because the advanced
of the channel. So it is important to provide the number of the Jakes emulator has used the symmetry of the phase, we need to
LPOs which can both determine the number of the path of the call the rand() function respectively to realize the independence
multipath effect and the accuracy of the simulation. of different random phase. The advanced Jakes emulator
belongs to the random model because of the introduction of the
Based on the analysis above, we defined the function of the
random phase.
Jakes emulator in SCILAB:
In the real channel, there is a variety of noises. But in the
function [x, y, z] = JM(N0, v, fc, p, t) simulation, we just take the additive white Gaussian noise
[x, y, z] are the output parameters, they represent the in- (AWGN) into consideration. The power spectral density of the


 OSSC-2011
AWGN has an unlimited bandwidth, and the bilateral power
spectral density of it is usually N0/2. The statistical Rayleigh Distribution
P
characteristics of the AWGN are shown as follow: 0.6

E=0,  2  N 0 * fs 2
0.5

So in the SCILAB we can define the approximate AWGN:

0.4

n  N 0 * fs / 2 rand(1, N, ' normal' )

0.3

In the equation, the rand(1, N, ’normal’) generate a random 0.2

number which mean is 0, standard deviation is 1, and length is
N. So n is a random process which mean is 0, variance is 0.1
 2  N 0 * fs 2 .
r
1 2 3 4
From the analysis above, and the generation principle of the
advanced Jakes emulator, we can get the output when the Figure 4.Ideal Rayleigh distribution
carrier signal has passed the channel:
As shown in Fig. 3 and Fig 4, the envelope probability
z  x. * cos(2f c t )  y. * sin(2f c t )  n density of the Jakes emulator proximately matches the
Rayleigh distribution, and with the increases of number of the
After the process above, we have finished the design of the multipath, it matches the ideal Rayleigh distribution better.
core part of the Jakes emulator. Next, we will verify and When N   , its envelope probability density matches the
analysis the Jakes emulator. Rayleigh distribution completely. So this simulate channel, the
Jakes emulator, belongs to the random Rayleigh fading channel.
A. Envelope Probability Density Distribution
First, verify the statistical characteristics of the Jakes B. Doppler Shift
emulator. Because the Jakes emulator is realized by the limited In order to get the Doppler Spread obviously, we will
LPOs, the envelope probability density function of it should analysis from two points of view: the single-frequency signal
proximately be a Rayleigh distribution. When N   , it transmission and the multi-frequency signal transmission. For
should be consistent with the ideal Rayleigh distribution. the Doppler Spread of receiving the single-frequency signal,
When we verify the statistical characteristics, we need first assume that the frequency of the carrier signal
to provide a set of statistical data, and then draw the statistical is f c  900MHz and the speed of the mobile station is
diagram using the statistical function of SCILAB. For the v  108km/h , so the maximum Doppler shift of the received
selection of the statistical data, on the one hand it should reflect signal of the mobile station is
the Jakes emulator completely, and on the other hand it should v * f c 108km/h * 900MHz
be limited. So we take the data like this: First, let the single- fm    90Hz .As shown in Fig 5
frequency signal go through the channel. Then, sample the c 3 *108 m/s
output of the channel by dt = 0.001 in t  [0,1] . And finally is the spectrum of the channel characteristic in SCILAB:
take the result of the sampling as the statistical data. After the
selection, we use the histplot(a, data) function, in SCILAB, to
draw the histogram. From the histogram, we can easily get the
trends of the envelope probability density.
Fig. 3 has shown the envelope probability density curve of
different number of multipath of the Jakes emulator.

Figure 5. The Doppler shift of the quadrature component

In order to show the Doppler shift of the single-frequency
signal, we assume that the frequency of the carrier is
f c  200Hz , the maximum Doppler shift is f m  100Hz and
the number of multipath is N = 34. Then the spectrum of the
Figure 3. The envelope probability density histogram received signal in SCILAB is shown in Fig 6:


 OSSC-2011
 As shown in Fig. 7, for multi-carrier signal, the Doppler
shift will happen on every frequency, and the shift distances of
the frequencies are different. When the shift distance
f m  f 2 ( f is the interval of the neighbor frequency),
there will be an interference between neighbor frequencies.
Then the signal will be intermodulation distortion.

C. Autocorrelation Function
The study about the autocorrelation function of the Jakes
emulator is the most intuitive way that reflects whether the
channel characteristic matches the reference model. For the flat
fading channel model, its power spectrum should be the U-
spectrum, the classic power spectrum. So we can verify the
Figure 6.The spectrum of the received signal (single-carrier) channel model by watching its power spectrum.
From Fig. 5 and Fig. 6 we can find that the spectrum of the The Fig. 8 and Fig. 9 are the simulation results f the power
single-frequency signal has been expanded because of the spectral density of the received signal which carrier frequency
Doppler Effect, and we call this process the Doppler Spread. Its is f c  200Hz , and the maximum Doppler shift is f m  100Hz .
maximum Doppler shift is determined by the carrier frequency
f c and the speed of the mobile stationary v (Fig. 5). Besides
Fig. 5 and Fig. 6 reflected some characteristics of the Doppler
shift in Jakes emulator: First, the frequency of the Doppler shift
is discrete, and the number of the discrete points is determined
by the number of the LFOs. Second, these discrete frequency
points are distributed around the carrier frequency
symmetrically, and the centralization of these discrete
frequency points increases with the increasing of the offset of
the frequency.
We have known that, for the single-frequency signal, the
Doppler spread just expands its spectrum and enriches its
frequency components. While, in order to reflect the influence
of the Doppler spread to the signal, it is necessary to make a
test of the transmission of the signal under multi-carrier. Figure 8.The power specturm of the received signal (single-carrier) N=82
Assume that the frequency of the multi-carrier signal
is f c  {200,250,300,350,400} . For different carrier
frequencies, there will be different Doppler shifts. So we
assume that the maximum Doppler Spread
is f m  f c * v (108 *10) , the speed of mobile station is v  108 ,
and the number of multipath is N = 34. So as shown in Fig 7 is
the spectrum of it:

Figure 9.The power specturm of the received signal (single-carrier) N=162

As shown in Fig. 8 and Fig. 9, we can find that when

N=162, its characteristic of the power spectrum matches the
classic power spectrum closely. For the Jakes simulation
channel, the power spectrum, which correspond the
autocorrelation function of the channel, matches the classic
spectrum more closely with the number of the multipath
Figure 7.The spectrum of the received signal (multi-carrier) increases. It is consistent with the reference model.


 OSSC-2011
D. The Influence Of The Gaussian White Noise Finally, it has verified the channel which is constructed by the
The Gaussian white noise is the most influential noise Jakes emulator is a Rayleigh flat fading channel for four
among additive noise. The bandwidth of its power spectral reasons: Firstly, its power spectral density matches the classical
density is very wide, and the power spectral density is a U-spectrum. Secondly, its envelope probability density
constant in its bandwidth. So the Gaussian white noise can distribution is Rayleigh distribution. Thirdly, its statistical
seriously impact the transmission of the signal. In order to characteristic is wide-sense stationary and ergodic. Fourthly,
simulate the channel better, we must consider the influence of the signal which goes through the Jakes channel model will
the Gaussian white noise. Because of the specificity of the cause the Doppler shift and multipath effect. At last we make
Gaussian white noise, we can’t generate an ideal noise that the statement that this paper is funded by the Communication
matches the principle completely, but a proximate one. Fig. 10 Engineering National Specialty Construction Projects of
is the spectrum of the received signal which has taken the Beijing University of Posts and Telecommunications (TS2025).
influence of the Gaussian white noise into consideration.
REFERENCES
[1] Theodore S. Rappaport, “Mobile Radio Propagation: Small-Scall Fading
and Multipath”, Wireless Communications Principles and Practice, 2nd
ed., Publishing House of Electronics Industry, Beijing, pp. 177-284,
March 2004
[2] Dacheng Yang, “the modeling and simulation of the scalar channel”, the
Environment of the Mobile Communication: Basic Principles· Analysis
Methods and Modeling Techniques, 1st ed., vol. 1, China Machine Press,
Beijing, pp. 178-237, August 2003.
[3] Dacheng Yang, “mobile wireless channel”, the Advanced Technology of
the Modern Mobile Communication, 1st ed., vol. 1, China Machine
Press, Beijing, pp. 39-66, July 2005.
[4] Duo Huang, Feng Li, Weizhi Li, “the program design, script file and
function in SCILAB”, Freedom Computational Scientific Software:
Figure 10.The spectrum of the output with the AWGN SCILAB Basic Tutorial, 1st ed., vol. 1, Tsinghua University Press,
Beijing, pp. 61-86, June 2006.
As shown in Fig. 10, the noise make the spectrum [5] Gang Chuo, Wenbo Wang, Yongyu Chang, “the transmission of the
characteristic of the signal no longer smooth, and this influence mobile radio and the model of forecast transmission”, the Principles and
happens in the whole frequency band. If the power of the noise System of the Mobile Communication, 2nd ed., vol. 1, Beijing
University of Posts and Telecommunications Press, Beijing, pp. 15-44,
is too large, it will submerge the signal and cause a serious February 2009.
error in signal transmission. [6] Jiongpan Zhou, Qinhua Pang, Dawo Xu, ”the characteristics of the
reference channel and its influence on the signal transmission”,
IV. CONCLUSION Communication Theory, 3rd ed., vol. 1, Beijing University of Posts and
Telecommunications Press, Beijing, pp. 335-339, August 2008.
This paper has first introduced the process of modeling the
Jakes emulator on SCILAB, and then analyzed the statistical
characteristic and transmission performance of the Jakes model.