Performance of DS/CDMA and BFSK Signals in Rayleigh Fading Channels

By

D. Brian Epstein

And

Robert Taylor

06 Dec 99

University of Texas at Dallas

EE6390 Introduction to Wireless Communications

Supervising Professor: Dr. Murat Torlak

 ABSTRACT

This paper utilizes the Clark and Gans Fading Model to implement a Rayleigh Fading

Simulator using MATLAB. The Rayleigh fading channel is tested using BFSK and

DS/CDMA. BFSK is a frequency shift keying type of modulation and DS/CDMA is an

example of spread spectrum modulation. Spread spectrum modulation uses a pseudo-noise

sequence to spread the signal across a wide bandwidth. Cross-correlating this signal with a

locally-generated version of the pseudo-noise sequence de-spreads the signal which restores

the modulated message to the same narrow band as the original. The results of this

simulation are analyzed using the using bit error rate (BER) to compare their performance.

Other performance factors are the number of level crossings and the average fade duration.
INTRODUCTION

Cellular wireless systems experience loss of signal strength due to:

1) Doppler shift in mobile environments, and

2) Scattering due to reflections from natural and manmade obstructions.

R. H. Clarke modeled the mobile channel as a Rayleigh fading channel. Later, M. J. Gans

deduced a spectral model from Clarke’s original analysis. John I. Smith simulated the Clarke

and Gans model on a computer using the algorithm described below.

DESCRIPTION

In his model [1], R. H. Clarke considers a non-direct line of path between transmitter and

receiver in a mobile-radio environment. The signal is reflected and scattered due to

obstructions caused by man-made and natural structures. He considers scattered signals of

the originally transmitted base station signal to be the only signals received at the mobile end,

even though the original signal would be independent of whether it is being transmitted or

received by the mobile unit (also referred to as the ‘reciprocal principle’). Clarke also

considers the Doppler effect on the wave propagation due to the motion of the mobile unit.

Clarke’s considerations for his model are that, “… at any point, the received field is made

up of a number of generally horizontally traveling free-space plane waves whose azimuthal

angles of arrival occur at random for different positions of the receiver, and whose phases are

completely random such that the phase is rectangularly distributed throughout 0 to 2. The

phase and angle of arrival of each component wave will be assumed to be statistically

independent.” Also, he assumes that at any point there are a particular number, N, of carrier

waves having the same average amplitude. His model describes the carrier waves arriving at

the mobile whose phases are assumed to be Gaussian random variables. The angle of arrival

1
is assumed to be uniformly distributed on the interval (0, 2]. Because the mobile is in

motion with velocity, v, each angle, , will be associated with the Doppler shift in carrier

frequency, i.e.,

(1) f  f m cos

where,

v
(2) fm 


and  is the carrier wavelength.

According to Clarke, the E-field can be expressed as an in-phase and quadrature

component:

(3) Ez  Tc (t ) cos(2f c t )  Ts (t ) sin(2f c t )

where,

N
(4) Tc (t )  E o  C n cos(2f n t   n )
n 1

and,

N
(5) Ts (t )  E o  C n sin(2f n t   n ) .
n 1

Both Tc and Ts are Gaussian random processes. They are uncorrelated zero mean Gaussian

random variables with equal variance of:

E o2
(6) 2  .
2

The noise spectra is shown in Figure 1.

2
 Noise Spectra
1

0.9

0.8

0.7

0.6

No
0.5

0.4

0.3

0.2

0.1

0
0 20 40 60 80 100 120 140
frequency
0 fm

Figure 1

The E-field is then given by

(7) r (t )  E z (t )  Tc2 (t )  Ts2 (t )

where r(t) is Rayleigh distributed, i.e., (Figure 1)

 r  r2 

p ( r )   2
exp  2 , 0  r  
(8)  2 
0, r0


Rayleigh Fading Channel

20

10
Amplitude (dB about rms)

-10

-20

-30

-40
0 20 40 60 80 100 120 140
Elapsed Time (ms)

Figure 2

3
 For a quarter wavelength antenna and p() uniform over (0, 2], the Doppler output

spectrum is given by [1] as:

1.5
(9) S Ez ( f ) 
2
 f  fc 
f m 1   
 fm 

where p() is a fraction of the total incoming power, as shown in Figure 3.

Doppler Power Spectrum at Baseband

0.07

0.06

0.05

0.04
Se(f)

0.03

0.02

0.01

0
0 50 100 150 200 250 300
frequency

-fm 0 fm
Figure 3

There are two types of fading which are dependant upon the relationship between

Coherence Time and Symbol Period.

1) Fast Scale Fading has a high Doppler spread where the coherence time (Tc) is less

than the symbol period (Ts). It causes frequency dispersion (time selective

fading) since the channel variations are faster than the baseband signal variations.

This usually only occurs for very low data rates.

2) Slow Scale Fading has a low Doppler spread where the coherence time (Tc) is

greater than the symbol period (Ts). The channel variations are slower than the

baseband signal variations.

4
 John I. Smith [2] simulated the model on a computer with the following described

algorithm. He used a random number generator to produce two independent Gaussian noise

baseband line spectrum with a maximum frequency of fm, the Doppler shifted frequency.

There were for positive frequencies. The negative frequency components were constructed

by conjugating the positive frequency components. This signal is a purely real Gaussian

random process which is used for in-phase and quadrature components, one in each branch of

the simulator. The random spectrum is multiplied by the discrete frequency representation

S Ez ( f ) . Smith truncated the edges where S Ez ( f ) approaches infinity and the slope was

extended to the edge.

The IFFT is then taken in each branch and the quadrature branch is created by shifting the

phase –90 with the Hilbert transform. The absolute value is then squared. The square root

is taken to produce the channel response. The diagram showing the overall method is shown

in Figure 4.

Baseband Baseband
Gaussian Doppler
Noise Source Filter
cos 2f c t
Independent
sin 2f c t  s0 (t )

Baseband Baseband
Gaussian Doppler
Noise Source Filter

Figure 4

Any signal may now be tested in this channel simply by convolution with the channel

response, i.e.,

(10) r (t )   exp j st 

5
where s(t) is the signal through the channel,  is the Rayleigh random variable produced by

the above steps, and  is the random phase of incidence.

RESULTS

The Rayleigh fading channel was simulated in Matlab by writing an m-file function which

was then called for both DS/CDMA and BFSK signal simulation. DS/CDMA was written

into an m-file which used BPSK modulation with a 32 bit pseudo-random code for spreading.

An n-file was also written for BFSK modulation and a Matlab pre-written function was used.

Data
1

0.8

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140

Demodulated Signal
1

0.8

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140

Figure 5

Figure 5 shows the data and demodulated signal for DS/CDMA. Some errors can be

observed in the demodulated waveform. After demodulation, the following was obtained for

DS/CDMA:

- BER = 0.0234 = 2.3% - Velocity of Mobile = 100 kph (62 mph)

- Number of Errors = 3 - fm = 83.3 Hz
- Data Stream = 128 Bits - fc = 900 MHz

6
 Figure 6 shows the data and demodulated signal (ydemod) for BFSK. And the following

was obtained for BFSK:

- BER = 0% - Velocity of Mobile = 100 kph (62 mph)

- Number of Errors = 0 - fm = 83.3 Hz
- Data Stream = 128 Bits - fc = 900 MHz

Data
1

0.8

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140

ydemod
1

0.8

0.6

0.4

0.2

0
0 20 40 60 80 100 120 140

Figure 6
Number of Level Crossings
100

80

60

40

20

0
0 0.2 0.4 0.6 0.8 1

Average Fade Duration

0.01

0.008

0.006

0.004

0.002

0
0 0.2 0.4 0.6 0.8 1

Figure 7

7
 The number of level crossings and average fade duration is plotted versus  for the given

fm of 83.3 Hz. As can be seen in Figure 7, the maximum number of level crossings occurs at

 = 0.707.

CONCLUSION

It has been shown, through the use of the Matlab simulations, that Signals in the Rayleigh

fading environment suffer strength loss due to Doppler shift and incidence of arrival which

causes cancellation at the receiver, and, that it is clear from the tests that BFSK mitigates

these effects whereas DS/CDMA resulted in some errors.

Matlab simulation proved to be very useful for the task and reusable code was the result.

For future work, this simulation needs to be extended to a two-ray fading channel to include

multipath delay.

REFERENCES

1. R. H. Clarke, “A Statistical Theory of Mobile-Radio Reception,” The Bell Systems

Technical Journal, vol. 47, no. 6, pp. 957-1000, July-August 1968

2. John I. Smith, “A Computer Generated Multipath Fading Simulation for Mobil Radio”,
IEEE Transactions on Vehicular Technology, vol. 24, no. 3, pp. 39-40, August 1975

3. Theodore S. Rappaport, Wireless Communications: Principles and Practice, Prentice

Hall PTR, ISBN 0-13-375536-3, July 1999

4. William C. Y. Lee, Mobile Communications Design Fundamentals, John Wiley & Sons,
Inc., ISBN 0-471-57446-5, 1993

5. R. J. Holbeche, Land Mobile Radio Systems, Peter Peregrinus Ltd., ISBN 0-86341-04909,
1985

6. Khaled Ben Letaief, Khurram Muhammad, and John S. Sadowsky, “Fast Simulation of
DS/CDMA With and Without Coding in Multipath Fading Channels, IEEE Journal on
Selected Areas In Communications, vol. 15, no.4, pp. 626-639, May 1997

7. Louay M. A. Jalloul and Jack M. Holtzman, “Multipath Fading Effects on Wide-band

DS/CMDA Signals: Analysis, Simulation, and Measurements”, IEEE Transactions on
Vehicular Technology, vol. 43, no. 3, pp. 801-807, August 1994