Vous êtes sur la page 1sur 5

THE PERFORMANCE OF UNBALANCED SIR-BASED SELECTION COMBINING

OVER EXPONENTIALY CORRELATED WEIBULL FADING CHANNELS


Zoran Popovi , Mihajlo Stefanovi *
* Department of Telecommunications, Faculty of Electronic Engineering, University of Ni
Department of Informatics and Telecommunications, Technical Faculty of aak

Abstract A performance analysis of a multi-branch SC multi-branch SIR-based SC with correlated fading, for both
based on the signal to interference ratio (SIR) over desired and interfered signals at the inputs.
exponentially correlated Weibull fading channels is main To simplify analysis, the diversity branches are usually
contribution of this paper. SIR based selection diversity is a assumed to be uncorrelated and independent identically
very efficient technique that reduces fading and interference distributed. However, branch correlation and no-identically
influence. In the case when the desired and interfered signal distributing will reduce the achievable diversity gain and,
envelopes experiences the exponentially correlated Weibull therefore, the uncorrelated and independent identically
fading, very useful expressions are obtained. In addition, for distributed branch assumption gives an optimistic result.
the balanced and unbalanced Weibull fading channel, the The analysis of correlative Weibull fading channel based
numerical results and their graphics represents the effects of on signal to interference ratio performance is main
fading severity and correlation on the system performance. contribution of this paper.
Keywords diversity systems, exponential correlation,
interference, multivariate Weibull distribution, SIR and II. The SIR-based analysis of Multivariate Weibull Fading
selection combining. Channel with Exponential Correlation
We consider selection combining operating in a Weibull
I. INTRODUCTION fading environment in the presentence of interference. The
There are various techniques for reducing fading effect assumption that the desired and interfered signals have to be
and influence of co-channel interference. Diversity reception a random Weibull distributed amplitudes [8] given by
is an effective and very simple technique that exploits the R
1
principle of providing the receiver with multiple faded p Z R R e

replicas of the same information bearing signal [2]. The goal Several fading correlation models have been proposed
of diversity techniques is to increase channel capacity and to and used for the performance analysis of various wireless
upgrade transmission reliability without increasing systems, corresponding to specific modulation techniques,
transmission power and bandwidth. Space diversity is an detection, and diversity combining scenarios. One of the most
efficient method for amelioration systems quality-of-service frequently used models is exponential correlation one, which
(QoS) when multiple receiver antennas are used [1]. There has been first addressed by Aalo in [5, Sec. II.B]. This model
are several principal types of combining techniques and corresponds to the scenario of multichannel reception from
division can be generally performed by their dependence on equal-spaced diversity antennas among the pairs of combined
complexity restriction put on the communication system and signals decays as the spacing between antennas increases [1].
amount of channel state information available at the receiver. In this work, these models should been considered for more
One of the least complicated combining methods is selection accurate statistical description of fading providing more
combining (SC). Combining techniques like equal gain reasonable conclusions than independent ones.
combining (EGC) and maximal ratio combining (MRC) Joint probability density function can be described similar
require all or some of the amount of the channel state [5, p. 3611] in closed form as
information of received signal.
In general, selection combining, assuming that noise p r
Z
1
( 1 ) L 1
L
l 1 l

l
rl
l 1

power is equally distributed over branches, selects the branch


L 1 r l
with the highest signal to noise ratio (SNR), which is the 1 r1 1 r L
exp ( 1 ) l L (2.1)
branch with strongest signal [1-4]. Similarly to previous, ( 1 ) 1 l 2 l L

there are type of selection combining that chooses the branch L 1 2 r l 2 r l 1 2
with highest signal and noise sum. In fading environments as I0 l l 1
1
in cellular communication systems where level of the co- l 1 i i 1
channel interference is sufficiently high as compared to the ,where I 0 is modified Bessel function
of the first kind and
thermal noise, SC selects the branch with highest signal to zero order, and correlation coefficient. Using expression
interference ratio (SIR-based selection diversity). This type
of selection combining in which the branch with the highest for expanding modified Bessel function, hence
SIR is selected, can be measured in real time both in base p r
Z
1
(1 ) L 1
L
l 1 l

l
rl
l 1

stations and in mobile stations using specific SIR estimators


as well as those for both analog and digital wireless systems L 1 r l
1 r1 1 r L
exp ( 1 ) l L (2.2)
(e.g. GSM, IS-54).
( 1 ) 1 i 2 l L

In order to study the effectiveness of any modulation
2il
scheme and type of diversity used, it is required to evaluate L 1
1 2 r l 2 r l 1 2
l l 1


the system performance over the channel conditions [3]. We i ! 2 2 2il 1
i1 ,i2 ,...,iL 1 0 l 1 l l l 1
propose a method for evaluation of the outage probability in
, and
p r
Z
1
( 1 ) L 1
L
l 1 l

l
rl
l 1
Where cl Rl2 and il rl2 are the average powers of
desired and interfered signals in l-th branch, respectively.
L 1 r l
1 r1 1 r L Hence, instantaneous values of signal-to-interference ratio
exp ( 1 ) l L (2.3)
( 1 ) 1 i 2 l L (SIR) at the diversity inputs can be defined as

i Rl2
L 1
1 l rlil l rlil 1l 1 l
i !2 rl2
1 2 il il
i1 ,i2 ,...,i L 1 0 l 1 l l l 1
The selection combiner (SC) chooses and outputs the
branch with largest SIR value
Let to view a selection diversity, if assume that desired max( 1 , 2 ,..., L )
signal envelopes transmitting in exponentially correlated
The average powers of desired and interfered signal are
Weibull fading channel, and has joint probability density
cl , il , l 1,2,..., L at l-th branch, then the average signal-to-
function defined as (2.11) and (2.14)
interference ratio (SIR) at l-th branch is
p R
R
1
( 1 ) L 1
L
l 1 l

Rl
l 1
dl
cl Sl
cl
L 1 R l R L
1 R1 1
exp ( 1 ) l L (2.4) The joint probability density function of instantaneous

( 1 ) c1
i 2 cl cL
values of SIR in L output branches of selection combiner is

L 1
i
l Rlil l i
Rl l1l 1
given in (2.20).
1
i !2 1 2

... J pR1 ,R2 ,...,RL x1
il il
i1 ,i2 ,...,iL 1 0 l 1 l cl c( l 1 ) p1 , 2 ,..., L t1 ,t2 ,...,t L t1 , x2 t2 ,..., xL t L
0
0 0

Similarly, interfering signal envelope has joint probability L fold
density pr1 ,r2 ,...,rL ( x1 , x2 ,..., xL ) x1 x2 ...x L dx1dx2 ...dx L (2.7)
p r
r
1
( 1 ) L 1

L
l 1
l l 1
r
il l
By substitution (2.4) and (2.5) into (2.6), the probability
density function p 1 , 2 ,..., L t1 , t 2 ,...,t L is given into (2.8).
L 1 r l
1 r1 1 r L Then, after L-fold integrations joint probability density
exp ( 1 ) l L (2.5)
( 1 ) i1 i 2 il iL
function into (2.9) becomes generalized expression of the

kl kl l k cumulative distribution function.
L 1
1 rl rl l1 l 1
k !2
1 2 il kl
i( l 1 ) kl
k1 ,k 2 ,...,k L 1 0 l 1 l

L L 1i1 2 i1 k1

p t
2 L
l 1 l t l1 l 1 l l
i k t 1 2
( 2 i1 k 1 ) 1 2 i1 k1 1


1 2
1 t1


( 1 ) 2 ( L 1 ) L 1 S
i1 ,i2 ,...,i L 1 0 i !2 k l !2 1 2 lL1 il kl
l 1 l 1
S1

k1 ,k 2 ,...,k L 1 0
1 il 1 il 2 il 1 il kl 1 kl
L 1 2 il 1 il kl 1 kl t l 2 l 2
1
( 2 il 1 il k l 1 k l ) l 1 tl
S
l 2 1 l
Sl

1i L 1 2 i L 1 k L 1
L 2 L 2
2 i L 1 k L 1 t L 1 t L
( 2 i L 1 k L 1 ) 1 (2.8)
S SL
L
1 2 L
P ... p t 1 , t 2 ,..., t L dt1 dt 2 ...dt L

0
0
0

L fold

1 2 L 1

L L i k 1i1

l 1 l l 1 l l 1 2 1 2
( 2 i1 k 1 ) 1 2 i1 k1 2 F1 1 i1 ,2 i1 k 1 ,2 i1 , 1 1
L 1 L S 1
i1 ,i2 ,...,iL 1 0
k1 ,k 2 ,...,k L 1 0
l 1 il !2 k l ! 2 1 2l 1il kl S1
1il 1 il
2 il 1 il kl 1 kl
L 1
1 l l 2 l 2
( 2 il 1 il k l 1 k l ) l
2 F1 1 il 1 il ,2 il 1 il k l 1 k l ,2 il 1 il ,
l 2 1 Sl S l

1i L 1
L 2 L 2
( 2 i L 1 k L 1 ) 1 2 iL 1 k L 1 2 F1 1 i L 1 ,2 i L 1 k L 1 ,2 i L 1 , L L (2.9)
SL S L

III. Outage Probability environment, Pout is defined as the probability of failing to
achieve a specified SIR value sufficient for satisfactory
One of the most frequently used performance measure reception. Therefore, the outage probability is simplifying
for diversity systems operating in fading environments is the CDF of the output SIR evaluated at .
the outage probability Pout . The outage probability is an
important statistical measure to access the quality of service Pout PR ( ) p (t )dt P ( ) (3.1)
provided by the system. In the interference limited 0
The outage probability depends of correlation
coefficient, fading parameter and average SIR per branch.
The cumulative distribution function of output SIR
which expression is given in (3.2) can be derived from (2.8)
by equating variables 1 2 ... L , and the
probability density function can be derived by
differentiating CDF as follows
dP (t )
p (t )
dt
Hence, we could use expression for the probability
density function given in (3.3).

The Fig. 1 represents the probability density function of


output SIR Triple-SC for two cases balanced and
unbalanced distributed signal.

Figure 1 The probability density function of Triple-SC

i1 i2 k1 k 2 1 i1
( 1 )2 2 i1 k1

t 2 2 t 2 2

P3 SC ( t ) 2 F1 1 i1 ,2 i1 k1 ,2 i1 ,
( 1 )2 i1 ,i2 ,k1 ,k2 0 1 i1! k1! ( 1 i1 ) S1 S
1

1i1 i2
( 2 i1 i2 k1 k2 )

t 2 2 t 2 2



2 F1 1 i1 i2 ,2 i1 i2 k1 k2 ,2 i1 i2 , S
i1! i2 ! k1! k2 !( 1 i1 i2 ) S2
2
1i2
( 2 i2 k2 )

t 3 2 t 3 2



2 F1 1 i2 ,2 i2 k2 ,2 i2 , S (3.2)
i2 ! k2 !( 1 i2 ) S3
3
1 i i2 k1 k 2 i i k1 k1
1 2 1 2 1 2 3
( 2 i1 i 2 )( 2 i1 i 2 k 1 k 2 )( 2 k 1 k 2 ) 1 t 1 2 1 t 2 2 1 t 3 2
p3 SC ( t ) t2
1 2 2S 1 S 2 S 3 i1 ! 2 i 2 ! 2 k 1 ! 2 k 2 ! 2 ( 1 i1 )( 1 i1 k 1 )( 1 k 1 ) 1 S1 S2 S3
i 1 ,i2 0
k1 ,k 2 0
2 i1 i2
t 2 2 3 2 1 2
2 F1 1 i1 k 1 ,2 i1 i 2 k 1 k 2 ,2 i1 k 1 , 2 F1 1 k 1 ,2 k 1 k 2 ,2 k 1 , t 1 i1 1 t 1
S2 S3 S1

2 i1 i2 k1 k 2
3 2
t 1 2
2 F1 1 k 1 ,2 k 1 k 2 ,2 k 1 , t

1 i1 k 1 1 t
2 2
2 F1 1 i1 ,2 i1 i 2 ,2 i1 , 2
S1 S3 S2

2 k1 k 2
3 2
t 1 2
2 F1 1 i1 k 1 ,2 i1 i2 k 1 k 2 ,2 i1 k 1 , t
2 2
1 i1 k 1 1 t

2 F1 1 i1 ,2 i1 i 2 ,2 i1 , 3 (3.3)
S1 S2 S3

A. Identically and no-identically branch distribution


The Fig. 2 and Fig. 3 represent the outage probabilities in
cases of identically and no-identically distributing the desired
signal and interfering signal for various fading parameter .
Firstly, it is evident that the outage probability decreases
rapidly as fading parameter increases. Moreover, we can to
conclude that the outage probability of no-identically branch
distributed fading parameters 1 , 2 , 3 are nearly or same as
identically distributing fading parameter 1 2 3 / 2 . It
is very important the conclusion and we will use it
hereinafter.
Weibull fading parameter 2 is the Rayleigh case,
1.75 corresponded to Nakagami-m fading parameter
m 0.76 and Hoyt q 0.53 ; 3 correspond to Nakagami
m 2.17 and Rice K 2.76 [4]. Figure 2 Pout versus S for balanced scenario
It is evident that the outage probability increases slowly
as correlation coefficient increases for both cases. Almost
it is linearly. Once again, we demonstrate the hypothesis that
overall fading parameter is average value of all branch fading
parameters. For smaller correlation coefficient it is evidently.
When the correlation coefficient uniformly increasing, the
outage probability increasing from up to 10. Generally, it is
same for both cases.

C. The balanced and unbalanced comparative analysis


To simplify analysis, the diversity branches are usually
assumed to be balanced, respecting the average SIR per
branch S l S . However, branch unbalanced case is more real
able to reduce performance of diversity gain. We are studying
two cases when average branch value of SIR successively
decreasing for S 3dB or S 6 dB . Its same as using
Figure 3 Pout versus S for no-i.d. case S l 1 S l / 2 or S l 1 S l / 4 to describe this phenomenon.
The Fig. 6 represents Pout versus fading parameter
B. The Correlation coefficient and its influence 1 2 3 and unbalanced branches S l 1 S l S in dB,
The correlation coefficient increases outage probability at
for several values of S 1 .
the output of selection diversity. The Fig. 4 and Fig. 5
presents the outage probability for predetermined values of
normalized first branch S 6 dB and S 12dB . We can to
see tendency for the outage probability for both cases, and
dependence by the fading parameter and the correlation
coefficient .

Figure 6 Pout versus and unbalanced gain S


There are tendency that characteristics of the outage
probabilities are same as characteristics of all branch average
SIR value.
The Fig. 7 and Fig. 8 represent comparative overview of
the output probability for Double-SC and Triple-SC in cases
Figure 4 Pout versus and for uncorrelated, balanced and unbalanced Weibull channel.
When increasing the outage probability slowly
decreasing for S 6 dB and rapidly for S 12dB .

Figure 7 Pout comparative overview of 2-SC and 3-SC


uncorrelated, balanced and unbalanced cases
Figure 5 Pout versus the correlation coefficient
VI. CONCLUSION
The analysis of exponentially correlated Weibull fading
channel based on signal to interference ratio performance is
main contribution of this paper. These model should been
considered for more accurate statistical description of fading
providing more reasonable conclusions than independent
ones. The outage probability depends of correlation
coefficient, fading parameter and average SIR per branch.
Numerical results and their graphical representations showed
the effects of fading severity and correlation on the system
performance. The influence of correlation between diversity
branches, fading parameter and unbalanced diversity
branches will decrease diversity gain significantly and then
reduce the performance of selection diversity systems.

REFERENCES
[1] W.C.Y. Lee Mobile Communications Engineering , 2nd
Figure 8 Pout comparative overview of 2-SC and 3-SC
ed. McGraw-Hill, New York 1998.
uncorrelated, balanced and unbalanced cases [2] W. C. Jakes Microwave Mobile Communications , 2nd
It is evident that performance of unbalanced 3-SC is ed. New York: Wiley-Interscience , 1994.
worse than performance of 2-SC for some cases. If S 3dB [3] G. Stber Principles of Mobile Communications ,
then for all values of S 6 dB are worse and if S 6 dB it is Kluwer Academic Publishers, Boston 2000.
for all values of S 12 dB . [4] M. K. Simon and M.-S. Alouini Digital Communication
over fading Channels , 2nd ed. New York: Wiley, 2005.
[5] V. A. Aalo, Performance of maximal-ratio diversity
systems in correlated Nakagami-fading environment ,
IEEE Trans. Commun. Vol. 43, No. 8, pp. 2360-2369,
Aug 1995
[6] N. C. Sagias, G. K. Karagiannidis, Gaussian Class
Multivariate Weibull Distribution: Theory and
Applications in Fading Channels, IEEE Trans.
Information Theory Vol. 51, No. 10, pp. 3608-3619, Oct
2005
[7] G. K. Karagiannidis, Performance Analysis of SIR-
based Dual Selection Diversity over Correlated
Nakagami-m Fading Channels, IEEE Trans. Veh.
Technology, Vol. 52, pp. 1207-1216, Sept 2003.
[8] M. Stefanovi, D. Milovi, The Outage Probability of
SIR-based Dual Selection Diversity over Correlated
Weibull Fading Channels, IEEE 1-4244-1468-7/07
Figure 9 The outage probability comparative overview
[9] A. Abu-Dayya and N. C. Beaulieu, Outage probabilities
for 2-SC, 3-SC and 4-SC for balanced and two
of diversity cellular systems with co-channel interference
different unbalanced scenarios
in Nakagami fading, IEEE Trans. On Veh. Technology,
Vol. 41, pp. 343-355, Nov 1992.
The correlation coefficient and fading parameter [10] M. Abramowitz, I. Stegun, Handbook of Mathematical
depends of the terrain. For urban area the correlation Functions with Formulas, Graphs, and Mathematical
coefficient is 0.10 0.06 and for the suburban Tables, Tenth Printing, Washington: NBS, 1972.
area 0.35 0.18 . Also, fading parameter is lowering for [11] A. Goldsmith, Wireless Communications, New York:
urban then for suburban fading environment [1]. Wiley, 2004.
The biggest diversity gain is obtained by going from no [12] M.D. Yacoub, D.B. da Costa, U.S. Dias, and G.
diversity to two-branch diversity. Increasing the number of Fraidenraich, Joint Statistics for Two Correlated Weibull
diversity branches from two to three, from three to four and Variates, IEEE Antennas and Wireless Propagation
further, diversity gain will be decreasing. The influence of Letters, Vol. 4, pp. 129-132, 2005.
correlation between diversity branches, fading parameter and [13] Y. Chen, and C. Tellambura, Infinite series
unbalanced diversity branches will decrease diversity gain representations of the trivariate and quadrivariate
significantly and then reduce the performance of selection Rayleigh distribution and their applications, IEEE
diversity systems. Trans. Commun., Vol. 53, pp. 2092-2101, Dec. 2005.

Vous aimerez peut-être aussi