See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/227144228

Error Probability for L -branch Coherent BPSK Equal Gain Combiners over
Generalized Rayleigh Fading Channels

Article in International Journal of Wireless Information Networks · March 2008

DOI: 10.1007/s10776-008-0070-0

CITATIONS READS

3 576

1 author:

Vidhyacharan Bhaskar
San Francisco State University
196 PUBLICATIONS 1,258 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fault diagnosis View project

IAM Challenges View project

All content following this page was uploaded by Vidhyacharan Bhaskar on 04 June 2014.

The user has requested enhancement of the downloaded file.

Int J Wireless Inf Networks
DOI 10.1007/s10776-008-0070-0

Error Probability for L-branch Coherent BPSK Equal Gain

Combiners over Generalized Rayleigh Fading Channels
Vidhyacharan Bhaskar

Springer Science+Business Media, LLC 2008

Abstract In this paper, a closed-form expression for the tool due to its improved error performance and simplicity
probability of error in a coherent BPSK system over of its implementation [3]. In [4], a closed-form solution for
Generalized Rayleigh fading channels is derived. An L- the error probability is formulated in the framework of
branch equal gain combining diversity scheme is used. statistical decision theory. The probability of error is
Theoretical results for the probability of error are plotted directly evaluated from the characteristic function (CF) of
for various values of the number of degrees of freedom (n) the two or three branch equal gain combiner output. The
and diversity order (L). A simulation is performed and the work in [4] is extended in [5] for L-branch coherent equal-
simulated results are found to match very well with the gain combining (EGC) in correlated fading channels,
theoretical results. where L [ 3. A new representation for equally correlated
Rayleigh channel gains is developed, and this enables the
Keywords Diversity
Probability of error
performance analysis of EGC in correlated fading.
Generalized Rayleigh fading
Characteristic function
Digham et al. present closed form expressions for the
Equal gain combining average probability of packet error (PPE) in [6] for no
diversity, maximal ratio combining (MRC), selection
combining (SC), and switch & stay combining (SSC)
1 Introduction diversity schemes. The average PPE is then used to obtain
a modified expression for the throughput for network
Diversity techniques are based on the notion that errors protocols.
occur in reception when the channel attenuation is large, In [7], the error probability performance of maximal
i.e., when the channel is in deep fade [1]. Diversity methods ratio combining (MRC), equal gain combining (EGC), and
require that a number of transmission paths are available, all selection combining (SC) diversity schemes with coherent
carrying the same message but having independent fading BPSK signaling on Rayleigh fading channels with Gauss-
statistics [2]. By supplying several replicas of the same ian channel estimation errors is studied. The exact
information signal transmitted over independently fading probability of error with weighting errors for MRC, EGC,
channels, the probability that all signal components will and SC diversity schemes is shown to be very similar to the
fade simultaneously is considerably reduced. case of perfect channel estimation, but with the average
Equal gain combining (EGC) results when the gains SNR per diversity branch. Bjerke et al. [8], consider the
provided in all transmission paths is set equal to one. performance of various receive antenna diversity schemes
Moreover, EGC has been accepted as an attractive diversity for use in combination with transmit diversity on the
downlink of wideband code division multiple access third
generation systems. Both open and closed loop versions of
V. Bhaskar (&) maximal ratio combining and selection combining schemes
Department of Electronics and Communication Engineering,
are studied.
SRM University, Kattankulathur, Kancheepuram Dt.,
Tamilnadu, India In [9], a generalized selection combining scheme based
e-mail: [email protected] on a threshold criterion rather than a fixed-size subset of

123
 Int J Wireless Inf Networks

the best channels is analyzed. In this scheme, only those In [16], the average signal-to-noise ratio of a generalized
diversity branches whose energy levels are above a speci- selection combining scheme, in which m diversity branches
fied threshold are combined. Closed-form analytical (m B L, where L is the total number of diversity branches
solutions for the BER performance of this scheme over available) with the largest instantaneous SNRs are selected
Nakagami fading channels are derived. Do and Wu [10] and coherently combined, is derived and a closed form
evaluate the combined effect of antenna diversity and solution is found.
spread spectrum (SS) diversity for a direct sequence code In [17], a flexible approach for the evaluation of error
division multiple access system using differential binary probability in L-branch BPSK equal gain combiners in
phase shift keying (DBPSK). Both MRC and switched Rayleigh fading channels has been presented and analyzed.
diversity (SD) techniques are considered and the BER The number of degrees of freedom, or the number of
performance for four hybrid diversity schemes are multipaths in the channel in [17] is two. This work is an
compared. extension of [17] in the sense that the number of degrees of
Annavajjala et al. [11] consider the error probability freedom is an integer greater than two.
performance of MRC, EGC, SC and generalized selection Section 2 provides a description of the system and
combining (GSC) diversity schemes with coherent BPSK channel models. Section 3 derives an expression for error
signaling on Rayleigh fading channels with Gaussian probability for an L-branch equal gain combining diversity
channel estimation errors. The exact probability for MRC, case over Rayleigh fading channels with number of degrees
EGC, SC, and GSC diversity schemes with weighting of freedom (n) greater than two, and also plots the bit error
errors is the same as that for the case with perfect channel probability for various values of n. Finally, Sect. 4 presents
estimation with average SNR per diversity branch when the the conclusions.
in-phase correlation coefficient between the actual and the
estimated channel gains is non-negative.
In [12], a compound fading model incorporating short- 2 System Description and Channel Model
term fading and shadowing is used to analyze the per-
formance of wireless systems employing microscopic The decision variable c for a coherent BPSK system can be
diversity to mitigate the effects of flat fading. This model formulated as [1]
can account for the presence of different levels of fading X
L X
L
and shadowing and provide an analytical solution for the c¼ xk þ wk ; ð1Þ
probability density function of the signal-to-noise ratio. k¼1 k¼1
Goldfeld and Wulich consider a diversity reception sys- where xk is the output signal amplitude at the kth branch
tem with majority decoding and optimal adaptive power and wk represents the complex Gaussian noise at the kth
loading applied to a Rayleigh fading channel in [13]. It branch with zero mean and variance Nk.
is shown that the proposed system has a lower average For slow fading channels, it is assumed that xk remains
bit error rate (BER) than optimal diversity reception constant within a symbol period, but changes from symbol
(ODR) with uniform power reception for the same to symbol following a Generalized Rayleigh distribution
redundancy. with probability density function (PDF) given by
In [14], a binary direct-sequence spread-spectrum mul-
2
tiple-access system with random sequences in flat Rayleigh r n1 r
pRayl
xk ðrÞ ¼ ðn2Þ
exp  ; r  0; ð2Þ
fading is considered. A new explicit closed-form expres- 2 2 rn Cðn=2Þ ck
sion is obtained for the characteristic function of the
where n is the number of degrees of freedom, ck ¼ 2r2k is
multiple-access interference signals, and it is shown that
the average received SNR at the kth branch, and r2k is the
the overall error rate can be expressed by a single integral
variance of the kth zero-mean Gaussian random variable.
whose integrand is non-negative and exponentially
Let Uc ðsÞ; URayl
xk ðsÞ; and Uwk ðsÞ be the characteristic
decaying.
function (CHF) of the parameters, c, xk, and wk respec-
Ma and Lim [15] consider the bit error probability
tively. Then from (1), and due to the independence between
(BEP) of binary and quaternary differential phase shift
the desired signals and noise in each branch, Uc(s) can be
keying (2/4 DPSK) and non-coherent frequency shift
written as [17]
keying (NCFSK) with post-detection diversity combining
in arbitrary Rician fading channels. The most significant Y
L Y
L

contribution in [15] lies in the derivation of a closed-form Uc ðsÞ ¼ URayl

xk ðsÞ Uwk ðsÞ: ð3Þ
k¼1 k¼1
cumulative distribution function (CDF) for the relevant
decision statistic expressed in a non-central Gaussian From the definition of the CHFs, and using (2), we
quadratic form. have

123
Int J Wireless Inf Networks

!
L Z Z1 Z1
1
Y
L Y X
L
Uc ðsÞ ¼ Uwk ðsÞ expð jsrk ÞpRayl
xk ðrk Þdrk r¼C ... FNORMð0;d2 Þ  rk
k¼1 k¼1 k¼1
0
!L
0
!
0 ð6Þ

2 2 X
ds 1
L
rk2 ðr1 r2 . . .rL Þn1
¼ exp   exp  dr1 . . .drL ;
2
2 Cðn=2Þ
ðn2Þ
2
c ðr1 r2 . . .rL Þn
k¼1 k

L  ð4Þ R0
P where C ¼ Lðn2Þ 1 ; and FNORMð0;d2 Þ ðxÞ ¼ pNORMð0;d2 Þ
Z1 Z1 exp js rk 2 2 ½Cðn=2ÞL
1
k¼1
 ... ðr1 r2 . . .rL Þn1 ðs þ xÞdx is the well known CDF with zero mean and vari-
ðr1 r2 . . .rL Þn
0 0
! ance d2.  
XL
rk2 pffiffi X ; where mX is
From [1], we have FX ðxÞ ¼ 12 þ 12erf xm
 exp  dr1 . . .drL ; 2 rX
c
k¼1 k the mean of the random variable X, and r2X is the variance
P
L of X. Letting mX = 0 and r2X ¼ d2 ; we have FX ðxÞ
where d2 ¼ Nk is the variance of the Gaussian random    pffiffiffi
k¼1 L
P ¼ 12 þ 12erf dpx ffiffi2 ¼ 1  Q dx since erfðxÞ ¼ 1  2Qðx 2Þ
variable, W ¼ wk .
L 
L 
k¼1 P P
[1]. So, FNORMð0;d2 Þ  rk ¼ Q 1d rk : Thus,
k¼1 k¼1
Z1 Z1 !
2
1X L Y
L n1
r k r
3 Error Probability for Generalized Rayleigh Fading Pe ¼ C ... Q rk exp  k dr1 . . .drL :
Channel d k¼1 k¼1
rnk ck
0 0
ð7Þ
3.1 Error Probability Derivation pffiffiffi
Now, UðxÞ  erfðxÞ ¼ 1  2Qðx 2Þ: So, QðxÞ ¼
  
The evaluation of error probability (ERP) for coherent 1
1  U pxffiffi2 . Substituting Q(x) in terms of U(x) in (7),
2
BPSK detection is defined as [1, 18]
we have
Z0 Z0 Z1 Z1 Z1 Y
L
2
1 C rkn1 r
Pe ¼ Probðc\0Þ ¼ pc ðsÞds ¼ Uc ðsÞ Pe ¼ Pe1  Pe2 ¼ ... n exp  k dr1 . . .drL
2p 2 k¼1
rk ck
1 1 1 0 0
Z1 Z1 !
1 C 1 X L Y
L n1
r
 expðjssÞdsds ¼ Lðn2Þ L  ... U pffiffiffi rk k
ðCðn=2ÞÞ 2p2 2 2 d 2 k¼1 k¼1
rnk
0 0
Z1 Z1 Z0 Z1
2 2
2
s d r
 ... exp  ð5Þ  exp  k dr1 . . .drL : ð8Þ
2 ck
0 0 1 1
" !#
XL Making change of variables in the integrals of (8), where
 exp js s þ ðrk Þ dsds ri0 ¼ priffifficffi and dri0 ¼ pdrffifficiffi8i ¼ 1; 2; . . .; L; and ci ¼ 2r2i ; we
k¼1 i i
! have Pe1 ¼ 12; and
X
L
rk2
ðr1 r2 . . .rL Þ n1
 exp  dr1 . . .drL : Z1 Z1
k¼1 k
c ðr1 r2 . . .rL Þn  0 n1 ðr0 Þ2 0  0 n1
 2 2

 Pe2 ¼ K r1 e 1 dr1 . . . rL U
R1 P L
By definition, exp s 2d exp js s þ ðrk Þ 0 0 ð9Þ
1
PL
 k¼1 " #
ds ¼ 2ppNORMð0;d2 Þ s þ ðrk Þ : So, 1 X L
pffiffiffiffi 0 ðrL0 Þ2 0
k¼1
 pffiffiffi ck r k e drL ;
d 2 k¼1
0 ! 1 n=2
Z1 Z1 Z0 X
L where K ¼ Cððrc11rc22...c LÞ
. The last integral in (9) can be
...rL Þn
Pe ¼ C . . . @ pNORMð0;d2 Þ s þ ðrk Þ dsA R1 0 n1   0 2
k¼1 rewritten as rL U b þ crL0 eðrL Þ drL0 ; where b ¼
0 0 1
! 0
XL
rk2 ðr1 r2 . . .rL Þn1 P
L1
pffiffiffi 0
 exp  dr1 . . .drL ck rk qffiffiffiffi
c ðr1 r2 . . .rL Þn
k¼1 k
k¼1 pffiffi ; and c ¼ 1d cL
2.
d 2

123
 Int J Wireless Inf Networks

  2
From [19], we have U b þ crL0 ¼ UðbÞ þ p2ffiffipeb 0
h   i 10
2
crL0 1  crL0 þ 13c2 rL0 ð2b2  1Þ þ
; where |e| \ 1.2 * −1
10
10-10. So, −2
10
Z1 L=2
 0 n1  

Probability of Error
0 2
rL U b þ crL0 eðrL Þ drL0 −3
10 Generalized Rayleigh fading channels
0 −4
2
 2
 10
1 ceb nþ1 2c
eb nþ1
¼ UðbÞCðn=2Þ þ pffiffiffi C þ pffiffiffi C −5
10
2 p 2 p 2
2 b2

 3 2 b2

 −6 n=4
bc e nþ2 c ð2b  1Þe nþ3 10 n=6
 pffiffiffi C þ pffiffiffi C : ð10Þ n=8
p 2 3 p 2 −7
10
Simulated results for n = 8

The derivation of the probability of error, Pe, for equal −8

10
0 2 4 6 8 10 12 14 16
gain combiners over Generalized Rayleigh fading channels
Average SNR in dB
becomes very cumbersome for L C3 because the number of
terms which need to be summed in the integrand of (9) Fig. 1 Probability of Error for Generalized Rayleigh fading channels
increases as the diversity level increases. In this paper, the with varying degrees of freedom for 2-branch EGC diversity

derivation of Pe is restricted to L = 2 only. Now, UðxÞ ¼

reference point of average SNR in the first branch. The
P
1
ð1Þk x2kþ1 error probability decreases as the number of degrees of
p2ffiffi [19].
p k!ð2kþ1Þ
k¼0 freedom (n) increases for 2-branch EGC diversity. This
Using this expression for U(x) in (10), and substituting happens because an increase in the value of n indicates that
back in (9) and simplifying, we have the effect of fading on the signal is lower. In other words,
!
KCðn=2Þ X 1
ð1Þk 2kþ1
the severity of fading is higher for a lower value of n. A
Pe2 ¼ pffiffiffi ðl  1Þ Cð2k þ n þ 1Þ
2
simulation is performed with 15000 bits for the n = 8 case
p k¼0 k!ð2k þ 1Þ
with L = 2. It is clear from Fig. 1 that the simulated results




cK n þ 1 1 n=2 cK
n þ 1 match very well with the theoretical results.
þ pffiffiffiC Cðn=2Þ þ pffiffiffi C
2 p 2 l p 2

n=2 r ffiffiffiffiffi
ffi


nþ2 1 ð 2 Þ
nþ1
1 Kc2 c1 4 Conclusions
Cðn=2Þ  C
l 2d 2p 2 l




 A comprehensive closed form expression for the evaluation
nþ3 1 ð 2 Þ nþ2
nþ2
nþ1 Kc3
C þ pffiffiffiC C of error performance in 2-branch EGC diversity over
2 3 p 2 l 2 Generalized Rayleigh fading channels is presented. A
3


n=2
Kc nþ3 1 simulation is also performed and the simulated results
 pffiffiffiC Cðn=2Þ;
6 p 2 l agree very well with the theoretical results. The error
ð11Þ performance curves are useful to the system designers to
make decisions on crucial system parameters by taking into
c1
where l ¼ 1 þ 2d 2. account the quality of service (QoS) demands.
Now, ci ¼ 2r2i is the average SNR at the ith branch
L1
8i ¼ 1; 2; . . .;L. Thus, K ¼ ðCð2n=2ÞÞL : Substituting Pe1 ¼ 12;
Appendix
and (11) in (8), the probability of error for equal gain
combiners over Generalized Rayleigh fading channels is It is interesting to see that Eq. 6 of this paper converges to
L1
Pe ¼ 12  Pe2 ; where Pe2 is shown in (11), K ¼ ½Cð2n=2ÞL ; l ¼ Eq. 11 of [17] when n = 2. Here is the proof:
qffiffiffiffi For n = 2, (6) can be written as
1 þ 2dc12 ; c ¼ 1d c2L ; and L = 2. Z1 Z1 !
X L YL
rk
L
3.2 Numerical Results Pe ¼ 2 . . . FNORMð0;d2 Þ  rk :
k¼1
c
k¼1 k
0 0
! ð12Þ
Figure 1 shows the theoretical error probability for X L 2
rk
2-branch coherent BPSK equal gain combiners over exp  dr1 dr2 . . .drL ;
k¼1 k
c
Rayleigh fading channels with varying degrees of freedom.
In this case, the SNR at each branch is measured having a where ck ¼ 2r2k Making change of variables in the integrals

123
Int J Wireless Inf Networks

of (12), where rk0 ¼ prkffiffiffi 0 drffiffiffi

ck and drk ¼ ck 8k ¼ 1; 2; . . .; L, we
p k binary signaling and gaussian weighting errors. IEEE Transac-
have tions on Wireless Communications, Vol. 4, No. 5, pp. 2267–2278,
2005.
Z1 Z1 !
XL 12. P. Shankar, Performance analysis of diversity combining algo-
L 0 pffiffiffiffi
Pe ¼ 2 . . . FNORMð0;d2 Þ  r k ck rithms in shadowed fading channels. International Journal of
k¼1 Wireless Personal Communications, Vol. 37, No. 1-2, pp. 61–72,
0 0 ð13Þ 2006.
!
Y
L X
L
2
13. L. Goldfeld and D. Wulich, Adaptive diversity reception for
 rk0 : exp  rk0 dr10 dr20 . . .drL0 : Rayleigh fading channel. European Transactions on Telecom-
k¼1 k¼1 munications, Vol. 14, No. 4, pp. 367–372, 2003.
14. J. Cheng and N. Beaulieu, Accurate DS-CDMA bit-error proba-
Now, (13) reduces to Eq. 11 of [17] when bility calculation in Rayleigh fading. IEEE Transactions on
Wireless Communications, Vol. 1, No. 1, 2002.
• the average received SNR at the kth branch, ck in this 15. Y. Ma and T. Lim, Bit error probability for MDPSK and NCFSK
paper is related to the ratio of the average bit energy to over arbitrary Rician fading channels. IEEE Journal on Selected
average noise density in the kth branch, qk of [17] by areas in Communications, Vol. 18, No. 11, pp. 2179–2189, 2000.
16. N. Kong and L. Milstein, Average SNR of a generalized diversity
ck ¼ qLk ; and
selection combining scheme. IEEE Communications Letters, Vol.
• the variance, d2, of the Gaussian random variable, W in 3, No. 3, pp. 57–59, 1999.
this paper is equal to 12. 17. S. Kotsopoulos and G. Karagiannidis, Error performance for
equal-gain combiners over Rayleigh channels. Electronic Letters,
Vol. 36, No. 10, 2000.
18. Q. Zhang, A simple approach to probability of error for equal
References gain combiners over Rayleigh channels. IEEE Transactions on
Vehicular Technology, Vol. 48, No. 4, pp. 1151–1154, 1999.
1. J. G. Proakis, Digital Communications, 4th ed. McGraw Hill, 19. M. Abramowitz and I. Stegun, Handbook of Mathematical
NY, 2001. functions with formulas, graphs and Mathematical tables.
2. W. Jakes, Microwave Mobile Communications, 2nd ed. IEEE National Bureau of Standards, Applied Mathematics Series 55,
Press, Piscataway, NJ, 1994. June 1964.
3. R. Steele and L. Hanzo, Microwave Radio Communications,
Wiley Publications, 1999. Author Biography
4. Q. Zhang, Probability of error for equal-gain combiners over
Rayleigh channels: some closed-form solutions. IEEE Transac- Vidhyacharan Bhaskar received the
tions on Communications, Vol. 45, No. 3, pp. 270–273, 1997. B.Sc. degree in Mathematics from
5. Y. Chen and C. Tellambura, Performance analysis of L-branch D.G. Vaishnav College, Chennai,
equal gain combiners in equally correlated Rayleigh fading India in 1992, M.E. degree in Elec-
channels. IEEE Communications Letters, Vol. 8, No. 3, pp. 150– trical & Communication Engineering
152, 2004. from the Indian Institute of Science,
6. F. Digham and M. Alouini, Average probability of packet error Bangalore in 1997, and the M.S.E. and
with diversity reception over arbitrarily correlated fading chan- Ph.D. degrees in Electrical Engineer-
nels. Journal of Wireless Communications and Mobile ing from the University of Alabama in
Computing, Vol. 4, No. 2, pp. 155–173, 2004. Huntsville in 2000 and 2002, respec-
7. R. Annavajjala and L. Milstein, On the performance of diversity tively. During 2002–2003, he was a
combining schemes on Rayleigh fading channels with noisy post-doc fellow with the Communi-
channel estimates. IEEE Military Communications Conference, cations research group at the University of Toronto. From September
Vol. 1, pp. 320–325, 2003. 2003 to December 2006, he was an Associate Professor in the Départ-
8. B. Bjerke, Z. Zvenar, and J. Proakis, Antenna diversity combin- ement Génie des systèmes d’information et de Télécommunication at
ing schemes for WCDMA systems in fading multipath channels. the Université de Technologie de Troyes, France. Since January 2007,
IEEE Transactions on Wireless Communications, Vol. 3, No. 1, he is a Professor and Dean of the School of Electronics and Commu-
pp. 97–106, 2004. nications Engineering at S.R.M. University, Kattankulathur, Chennai,
9. A. Sulyman and M. Kousa, Bit error rate performance analysis of India. His research interests include wireless communications, signal
a threshold based generalized selection combining scheme in processing, error control coding and queuing theory.
Nakagami fading channels. Eurasip Journal on Wireless Com-
munications and Networking, Vol. 2005, No. 2, pp. 242–248,
2005.
10. M. Do and S. Wu, Hybrid diversity combining techniques for DS-
CDMA over a multipath fading channel. International Journal of
Wireless Networks, Vol. 3, No. 2, pp. 155–158, 1997.
11. R. Annavajjala and L. Milstein, Performance analysis of linear
diversity-combining schemes on Rayleigh fading channels with

123

View publication stats