Sie sind auf Seite 1von 5

On the MGF and BER of Linear Diversity Schemes in

Nakagami Fading Channels with Arbitrary Parameters


Valentine A. Aalo1, Senior Member, IEEE, and George P. Efthymoglou2, Member, IEEE
1
Department of Electrical Engineering, Florida Atlantic University, Boca Raton, FL 33431,USA
2
Department of Digital Systems, University of Piraeus, Piraeus 18534, Greece

real fading parameters is usually expressed in terms of the Gauss


Abstract— In this paper, we explore the relationship between hypergeometric function, 2 F1 ( a, b; c; x ) . This function is analytic
the Lauricella hypergeometric functions ( FA , FB , and FD ) for all the parameters and can be easily computed since it is in the
and the performances of linear diversity combining schemes in library of several commonly available computer software packages
Nakagami fading channels with arbitrary parameters. We such as Matlab and Mathematica. When M-ary signaling is
consider maximal ratio combining, selection combining, and employed or when linear diversity reception is used to combat
equal gain combining of an arbitrary number of independent Nakagami fading, the average error rate can usually be expressed
Nakagami faded diversity branches. Specifically, we show that in terms of the Lauricella functions. For example, Shin and Lee [3]
the moment generating function and the average bit error rate express the average symbol error rate (SER) in terms of the
of these combining schemes can all be expressed in terms of the Lauricella hypergeometric function, FD( ) (...) . Recently, the
n

Lauricella hypergeometric functions.


authors also expressed the average BER for MRC systems in a
Nakagami fading environment in terms of this function [4]. On the
Index Terms— Lauricella Hypergeometric Function, Nakagami
other hand, Ugweje [5] and Annavajjala, et al. [6] derive the
Fading Channel, Diversity Combining.
moment generating function (MGF) of the output SNR and the
average BER of a coded SC system, respectively, in terms of the
I. INTRODUCTION
Lauricella function FA( ) (...) . Furthermore, in this paper we show
n

M ultipath fading seriously degrades the performances of digital


wireless communication systems. Therefore, the
performance analysis of such systems has been extensively
for EGC, the MGF of the output SNR and the average BER can be
expressed in terms of the Lauricella function FB( ) (...) . It should
n

done by many researchers. To this end, many statistical be noted that these multivariate hypergeometric functions all
distributions are available in the literature to model the fading reduce to the Gauss hypergeometric function when the order n = 1
process in order to predict the system performance. Of these, the [7]. The main aim of this paper is to show that, in fact, several
Nakagami-m distribution [1] has attracted considerable interest as a important performance measures (such as MGF of output SNR and
fading channel model. This is because it provides a very good fit average BER) of the three main linear diversity combining schemes
to measured data in a variety of fading environments. (MRC, SC, EGC) can be expressed in terms of the Lauricella
It is well known that diversity reception can significantly functions.
improve the performances of digital wireless communication
systems in the presence of fading and co-channel interference. The II. MGF OF OUTPOUT SNR
performances of commonly used linear diversity combining
techniques such as maximal ratio combining (MRC), equal gain In a Nakagami-m fading channel, the PDF of the received SNR,
combining (EGC), and selection combining (SC) have also been γ i , on each diversity branch is given by
studied extensively for a variety of fading environments, including ( mi / γi )
mi
⎧ mi γi ⎪
⎪ ⎫
exp ⎪ ⎪
mi −1
the Nakagami fading channel [2]. In MRC, the received signals on f γi ( γi ) = γi ⎨− ⎬, γi ≥ 0 (1)
Γ( mi ) ⎪

⎩ γ i



the diversity branches are co-phased and weighted (that is, both the
amplitude and phase are used) to obtain the output of the combiner. where mi ≥ 1/ 2 is a parameter that determines the severity of the
In the EGC systems, on the other hand, the received signals are co- fading and γi is the average SNR on the i-th diversity branch [2].
phased and combined with equal weights (that is, only the signal Next, we consider the performances of three commonly employed
phase is used) to obtain the combiner output. Finally, in SC, only linear combining schemes, namely; MRC, EGC, and SC, operating
the branch with the largest instantaneous amplitude is connected to in a Nakagami fading channel with arbitrary fading parameters on
the output of the combiner. the diversity branches. We assume throughout the paper that the
The average bit error rate (BER) performance of a binary digital diversity branches are sufficiently separated from each other for
receiver operating in a Nakagami-m fading channel with arbitrary them to be statistically independent.

978-1-4244-2517-4/09/$20.00 ©2009 IEEE 1


A. Maximal Ratio Combining ⎛
⎡ N ( m / γ )mk ⎤ ⎜⎜ ∑ mi ⎟⎟

N
∞ ⎡ N ( mi ) ( −mi γ / γ i )ni ⎤
= ⎢∏ k k ⎥ γ ⎝ i=1 ⎠ ∑ ⎢∏ ⎥. (7)
ni
It is well known that the output SNR in an MRC scheme is the
sum of the SNR on each diversity branch; that is, ⎢⎣ k =1 ( k ) ⎥⎦
Γ m + 1 ⎢
n1 , n2 ,..., nN = 0 i =1 ( m i + 1) ni ni ! ⎥
⎣ ⎦
N
γ MRC = ∑ γ i . (2) The corresponding PDF is obtained by performing the term-by-
i =1 term differentiation to give
The MGF of γ MRC may be expressed as [2], [4] dFγ SC ( γ )
fγ SC ( γ ) =
⎡ N ⎛ m ⎞ mi ⎤ dγ
Mγ ( −t ) = ⎢ ∏ ⎜ i ⎟ ⎥ ⎛ N ⎞
⎞ ⎡ N (m / γ ) ⎤ ⎜⎜ ∑ mi ⎟−
mk
MRC
⎢⎣ i =1 ⎝ mi + γ i t ⎠ ⎥⎦ ⎛ N ⎟
1
= ⎜ ∑ mi ⎟ ⎢∏ k k ⎥ γ ⎝ i=1 ⎠
⎝ i =1 ⎠ ⎢⎣ k =1 Γ ( mk + 1) ⎥⎦
⎛ m1 mN ⎞
× FD ( N ) ⎜ 0, m1 ,..., mN ; β ; ,..., ⎟ (3)
m1 + γ 1t mN + γ N t ⎠ ⎛ N



⎜ 1 + ∑ mi ⎟
⎝ i =1 ⎠n1 +...+ nN ⎡⎢ N ( mi ) ni ( −mi γ / γ i ) ⎤⎥
ni
(8)
where we have used the fact that FD
(n)
( 0, β ,..., β ; c; x1 ,..., xn ) = 1 × ∑ ∏ ( m + 1) n ! ⎥.
⎢ i =1
1 n
n1 , n2 ,..., nN = 0 ⎛ ⎞
N

⎜∑ i ⎟ ⎣ ⎦
i ni i
[7]. It follows that the probability density function (PDF) of the m
output SNR is given by taking the inverse Laplace transform of (3). ⎝ i =1 ⎠ n1 +...+ nN
Therefore, the PDF of the SNR at the MRC output is given by [4] The MGF at the output of the SC may also be obtained in terms of
the Lauricella function by performing the term-by-term integration
1
fγ MRC ( γ ) = ( −t ) eγ t dt
2π j ^∫
Mγ of (8), as
MRC

Mγ SC
( −t ) = ∫ e − γ t f γ ( γ ) d γ
SC
⎛ N ⎞
mi ⎤ ⎜⎝ ∑
⎜ mi ⎟ −1 0
1 ⎡ N ⎟
= ⎢∏ ( mi / γ i ) ⎥ γ

⎞⎡ N ⎛ mk ⎞ ⎤
i =1 mk
⎛ N
⎞ ⎣ i =1 ⎛ N
1
Γ ⎜ ∑ mi ⎟ ⎦ = Γ ⎜ 1 + ∑ mi ⎟ ⎢∏ ⎜ ⎟ ⎥
⎝ i =1 ⎠ ⎝ i =1 ⎠ ⎢⎣ k =1 Γ ( mk + 1) ⎝ γ k t ⎠ ⎥⎦
⎡ N ni 1 ⎤
(4)
N ⎛ −m − mN ⎞
N
⎢∏ ( mi ) ni ( − mi γ / γ i ) ⎥ × FA( ) ⎜ 1 + ∑ mi ; m1 ,..., mN ; m1 + 1,..., mN + 1; 1 ,..., ⎟
∞ ni !⎦ γ 1t γ Nt ⎠
× ∑ ⎣ ⎝ i =1
i =1

⎛ N ⎞
⎜ ∑ mi ⎟
n1 , n2 ,..., nN = 0 ⎛ N ⎞
− ⎜1+ ∑ mi ⎟
⎞ ⎡ N (m / γ ) ⎤⎛
mk
⎝ i =1 ⎠ n1 +...+ nN ⎛ N N
m ⎞ ⎜⎝ i=1 ⎟⎠
= Γ ⎜1 + ∑ mi ⎟ ⎢∏ k k ⎥⎜t + ∑ i ⎟
⎝ i =1 ⎠ ⎣⎢ k =1 Γ ( mk + 1) ⎦⎥ ⎝ i =1 γ i ⎠
Γ (a + n)
where the Pochhammer symbol is defined as ( a )n = , ⎛ ⎞
Γ (a) ⎜ ⎟
N
m1 / γ 1 mN / γ N ⎟
with ( a )0 = 1 [8]. FA( N ) ⎜ 1 + ∑ mi ; 1,..,1
N ; m1 + 1,.., mN + 1; ⎛ ,..,
⎜ i =1 N
mi ⎞ ⎛ N
mi ⎞ ⎟
⎜⎜ ⎜t + ∑ ⎟ ⎜ t + ∑ ⎟ ⎟⎟
N − terms

B. Selection Combining ⎝ ⎝ i =1 γ i ⎠ ⎝ i =1 γ i ⎠ ⎠
In a selection combining scheme, only the branch with the largest (9)
instantaneous SNR is connected to the output of the combiner. where we have used [7, eq. A.2.1] to obtain the last line in (9),
Therefore, the output SNR is the maximum SNR on the diversity which satisfies the convergence criteria for FA(…) [7].
branches;
C. Equal Gain Combining
γ SC = max {γ 1 , γ 2 ,..., γ N } . (5)
In EGC systems, the received signals on the diversity branches
The cumulative distribution function (CDF) of the SNR at the
are co-phased and combined with equal weights to obtain the
output of the SC operating in a Nakagami fading channel with
combiner output. It can be shown that the output SNR in an equal
independent fading along the diversity branches is given by [5]
gain combining scheme is given by [10]
N
1 ⎛ mγ⎞ 2
Fγ SC ( γ ) = ∏ G ⎜ mk , k ⎟ (6) 1⎛ N ⎞
k =1 Γ ( m k ) ⎝ γk ⎠ γ EGC = ⎜ ∑
N ⎝ i =1
γi ⎟ .

(10)
z
where G (α , z ) = ∫ xα −1e − x dx is the incomplete gamma function, The MGF of Ri = γ i is given by
0 mi
2 ⎛ mi ⎞

which may also be written in terms of the confluent hypergeometric ⎧ m x2 ⎫
MRi ( −t ) = ⎜ ⎟ ∫e x
− xt 2 mi −1
exp ⎨ − i ⎬ dx. (11)
function [8, eq. (6-5.12)]. Thus the CDF is given by Γ( mi ) ⎝ γ i ⎠ 0 ⎩ γi ⎭
⎛ N ⎞

⎜∑ mi ⎟

N ⎡ ( mk / γ k )mk ⎛ −mk γ ⎞⎤ The integral in (11) may be expressed in closed form in terms of
Fγ SC ( γ ) = γ ⎝ i =1 ⎠
∏ ⎢ 1 F1 ⎜ mk ; mk + 1; ⎟⎥ the parabolic cylinder function [9], [10]. However, here, we use
k =1
⎣⎢ mk Γ ( mk ) ⎝ γk ⎠ ⎦⎥

2

(−x)
k Thus the PDF of the output SNR in EGC systems is expressed as a
the series expansion for the exponential function, e − x = ∑ , multivariate hypergeometric function [7]. The MGF of the output
k! k =0
SNR for EGC is then obtained from (16) as
and using the multiplication theorem for the gamma function [8, eq. ∞


(8.335)] Mγ EGC (−t ) = e−γ t fγ EGC (γ )d γ
22 x −1 ⎛ 1⎞
Γ (2x) = Γ ( x)Γ ⎜ x + ⎟ , (12) 0
π ⎝ 2⎠
⎡ N ⎛1 1 ⎤
mi
⎞ ⎛ mi ⎞
we rewrite (11) as
mi
π ⎢ ∏ Γ ⎜ + mi ⎟ ⎜ ⎟
⎢ i =1 ⎝ 2 ⎠ ⎝ tγ ⎠ π⎥

Γ ( mi + 1/ 2 ) ⎛ mi ⎞ ⎣ ⎦
MR ( −t ) = ⎜ 2⎟ =
i
π ⎝ γ it ⎠ ⎛1 N ⎞
(13) Γ⎜ +
⎜2
mi ⎟
⎟ ∑
( mi )k ( mi + 1/ 2 )k ( −mi / (γ i t 2 ) )
k
∞ ⎝ i =1 ⎠
×∑ .
(N) ⎛ m ⎞
k! N
1 1 1 m
∑ mi ; − γ11t ,.., − γ NNt ⎟⎟
k =0
FB ⎜ m1 ,.., mN ; m1 + ,.., mN + ; +
The series in (13) is recognized as the confluent hypergeometric ⎜ 2 2 2
function of the second kind [7], [8]. Since the diversity branches ⎝ i =1 ⎠
N (17)
are independent, the MGF of R = ∑ Ri is given by computing the in terms of Lauricella function of the second kind [7], [11].
i =1

N product of (13) as III. ERROR RATE ANALYSIS


N ⎡ Γ m + 1/ 2
( i ) ⎛ mi ⎞ i ⎤⎥
m

MR ( −t ) = ∏ ⎢ ⎜ 2⎟ The average BER is given by


i =1 ⎢
⎣ π ⎝ γ i t ⎠ ⎥⎦ ∞

PE = ∫ PE ( γ ) fγ out ( γ ) d γ (18)

⎡ N 1 ⎤
⎢∏ ( mi )ni ( mi + 1/ 2 )ni −mi / ( γ i t ) ( )
ni
× ∑
0
2

n1 , n2 ,..., nN = 0 ⎣ i =1 ni !⎦ Γ ( b, aγ )
where PE ( γ ) = is the conditional BER in an AWGN
(14) 2Γ ( b )
The PDF of R is then obtained by evaluating the inverse Laplace
transform of (14), to give channel and f γ out ( γ ) is the PDF of the output SNR, which depends
⎡ N
⎛1 ⎞ ( mi / γ i )
mi
⎤ on the type of diversity combining scheme employed [2]. The
π ⎢∏ Γ ⎜ + mi ⎟ ⎥ ⎛ N ⎞ parameter a depends on the type of binary modulation ( a = 1 for
⎢⎣ i =1 ⎝2 ⎠ π ⎥⎦ ⎜⎜⎝ 2 ∑ mi ⎟−1

fR ( r ) = r i=1 ⎠ binary phase shift keying and a = 1/ 2 for binary frequency shift
⎛ N ⎞ ⎛1 N ⎞
Γ ⎜ ∑ mi ⎟ Γ ⎜ + ∑ mi ⎟ keying), while the parameter b depends on the type of
⎝ i =1 ⎠ ⎝ 2 ⎠ (15)
i =1
demodulation ( b = 1 for noncoherent FSK or differentially
⎡ N ni 1 ⎤
coherent PSK and b = 1/ 2 for coherent PSK or FSK).
⎢∏ ( mi ) ni ( mi + 1/ 2 )ni ( −mi r / γ i )
2

∞ ni !⎦
× ∑ ⎣
i =1
. A. Maximal Ratio Combining
⎛ N ⎞ ⎛1 N ⎞
⎜∑ i ⎟ ∑ i⎟
n1 , n2 ,..., nN = 0
m ⎜ + m The average BER for MRC is obtained by substituting (4) in (18)
⎝ i =1 ⎠n1 +...+ nN ⎝ 2 i =1 ⎠ n1 +...+ nN
to obtain
It follows from (10) that the PDF of the SNR at the output of the 1 ⎡ N m ⎤
EGC is given by PE
MRC
=
⎛ N
⎞ ⎢ ∏ ( mi / γ i ) i ⎥
2Γ ( b ) Γ ⎜ ∑ mi ⎟ ⎣ i =1 ⎦
fγ EGC ( γ ) =
N fR ( Nγ ) u (γ ) ⎝ i =1 ⎠
Nγ ⎡ N ni 1 ⎤
⎢∏ ( mi )ni ( −mi / γ i ) ⎥ N
∑ ( mi + ni ) −1


⎣ ni !⎦ ∞
⎞ ( mi / γ i ) ⎤
mi
× ∑ ∫0 Γ ( b, aγ ) γ
i =1
N
⎛1 i =1

π ⎢∏ Γ ⎜ + mi ⎟ ⎥ ⎛ N ⎞
⎜∑ i ⎟
⎛ N ⎞ n1 , n2 ,..., nN = 0
⎣⎢ i =1 ⎝2 ⎠ π ⎦⎥

⎜∑ mi ⎟ −1
⎟ m
= γ ⎝ i =1 ⎠ ⎝ i =1 ⎠ n1 +...+ nN
⎛ N ⎞ ⎛1 N ⎞
Γ ⎜ ∑ mi ⎟ Γ ⎜ + ∑ mi ⎟ ⎛ N

⎝ i =1 ⎠ ⎝ 2 ⎠ Γ ⎜ b + ∑ mi ⎟ ⎡ N
⎛ mi ⎞ ⎤
i =1 mi
⎝ ⎠
⎡ N ni 1 ⎤
=
i =1
⎢ ∏⎜ ⎟ ⎥
⎢∏ ( mi )ni ( mi + 1/ 2 )ni ( −mi γ / γ i ) ⎛ ⎞ ⎢ i =1 ⎝ aγ i ⎠ ⎥
N

⎣ i =1

ni !⎦ . 2Γ ( b ) Γ ⎜1 + ∑ mi ⎟ ⎣ ⎦ (19)
× ∑ ⎛ N ⎞ ⎛1 N ⎞
⎝ i =1 ⎠
⎜ ∑ mi ⎟ ⎜ 2 + ∑ mi ⎟
n1 , n2 ,..., nN = 0
N ⎛ −m −mN ⎞
N N
⎝ i =1 ⎠ n1 +...+ nN ⎝ i =1 ⎠ n1 +...+ nN × FD ( ) ⎜ b + ∑ mi ; m1 ,..., mN ;1 + ∑ mi ; 1 ,..., ⎟.
(16) ⎝ i =1 i =1 aγ 1 aγ N ⎠

3
Thus the average BER for both coherent and noncoherent ⎛ N

Γ ⎜ b + ∑ mi ⎟ ⎡ N
/differentially coherent binary modulations may be expressed in ⎛ mi ⎞ ⎤
mi
⎝ ⎠⎢ 1
terms of the Lauricella function. Using the transformation in [8, =
i =1

2Γ ( b )
∏ ⎜ ⎟ ⎥
⎢⎣ i =1 Γ ( mi + 1) ⎝ aγ i ⎠ ⎥⎦
eq. A.2.2.1.1], we may rewrite (19) as (24)
⎛ N
⎞ ⎛ −m − mN ⎞
Γ ⎜ b + ∑ mi ⎟ ⎡ N
N
mi
⎤ FA( N ) ⎜ b + ∑ mi , m1 ,..., mN ; m1 + 1,..., mN + 1; 1 ,..., ⎟
⎝ ⎠ ⎢ ⎛ mi ⎞ ⎥ aγ aγ N
PE
MRC
=

i =1
N
⎞⎢
∏ ⎜ ⎟
m + aγ i ⎠ ⎥
⎝ i =1 1 ⎠
2Γ ( b ) Γ ⎜ 1 + ∑ mi ⎟ ⎣ i =1 ⎝ i ⎦ (20) Using the transformation in [7, eq. A.2.1.2], we may rewrite (24) as
⎝ i =1 ⎠
⎛ N

Γ ⎜ b + ∑ mi ⎟ ⎛ ⎞
N

N ⎛ m1 mN ⎞ − ⎜ b + ∑ mi ⎟
N
⎟ ⎡ N
⎛ mi ⎞ ⎤
mi
× FD ( ) ⎜1 − b; m1 ,..., mN ;1 + ∑ mi ; ,..., ⎟. ⎝ ⎛
⎠ 1+
N
m ⎞ ⎜
1
⎜ ∑
⎝ ⎠
⎢∏
i =1

⎟ ⎥
i =1
m + aγ m + aγ PESC = i
⎟ ⎜
⎝ N ⎠
2Γ ( b ) ⎢⎣ i =1 Γ ( mi + 1) ⎝ aγ i ⎠ ⎥⎦
⎝ i =1 aγ i ⎠
i =1 1 1 N

which satisfies the convergence criteria for FD(…) [7]. Note that in

the special case when the diversity branches are independent and ⎜
identically distributed, we may use the following reduction formula N
m1
× FA( N ) ⎜ b + ∑ mi ;1,...,1;
N m1 + 1,..., mN + 1; ,
of the Lauricella function [12, p. 61], ⎜ ⎛ N
mi ⎞

i =1
⎜ N − terms
aγ ⎜ 1 + ⎟
⎛ n
⎞ ⎜ 1
i =1 aγ i
FD ( n ) (α , β1 ,..., β n ; c; x,..., x ) = 2 F1 ⎜ α , ∑ βi ; c; x ⎟ , (21)to ⎝ ⎝ ⎠
⎝ i =1 ⎠ (25)

show that (20) can be easily simplified to give the more familiar ⎟
mN ⎟.
...,
result ⎛ N
mi ⎞ ⎟
aγ N ⎜1 + ∑ ⎟ ⎟⎟
Γ ( Nm + b)⎛ Nm ⎞
Nm
⎝ i =1 aγ i ⎠ ⎠
MRC
PE = ⎜ ⎟ which satisfies the convergence criteria for FA(…) [7]. Similar
2Γ (b)Γ( Nm + 1) ⎝ Nm + aγ ⎠
b
(22) results for the average BER for selection combining are given in
⎛ aγ ⎞ ⎛ Nm ⎞ [5], [6]. Note that in the case of no diversity ( N = 1) , (24) or (25)
×⎜ ⎟ F 1, Nm + b, Nm + 1;
2 1 ⎜ ⎟,
⎝ Nm + aγ ⎠ ⎝ Nm + aγ ⎠ become
m b
which agrees with [2]. SC Γ ( m + b) ⎛ m ⎞ ⎛ aγ ⎞
PE = ⎜ ⎟ ⎜ ⎟
B. Selection Combining
N =1 2Γ (b)Γ( m + 1) ⎝ m + aγ ⎠ ⎝ m + aγ ⎠
(26)
In this case, we substitute (8) in (18) to obtain ⎛ m ⎞
×2 F1 ⎜ 1, m + b, m + 1; ⎟
⎛ N ⎞ ⎝ m + aγ ⎠
⎜ ∑ mi ⎟ ⎡ N ( m / γ )mi ⎤
⎝ i =1 ⎠
⎢∏ which is the performance when there is no diversity [2].
i i
SC
PE = ⎥
2Γ ( b ) ⎢⎣ i =1 Γ ( mi + 1) ⎥⎦
C. Equal Gain Combining
⎡ L ni 1 ⎤
⎢∏ ( mi ) ni ( − mi γ / γ i ) ⎥ N For EGC, the average BER may be written as
∞ ni !⎦ ∞ ∑ ( m + ni ) −1
× ∑ ⎣ Γ ( b, aγ ) d γ
i =1 i

L ∫γ i =1

(
Γ b, ar
2
)f
n1 , n2 ,..., nN = 0
∏ i n (
i =1
m + 1) i
0 EGC
PE = ∫ 2Γ ( b )
R ( r ) dr (27)
0
(23)
With the help of [8, (6.5.37)], and using the fact that where f R ( r ) is given in (15). Thus we have [7], [11]
⎡ ⎤ ⎞ ( mi / γ i )
mi
( a )n
N
⎛1
a
= , (23) simplifies to π ⎢∏ Γ ⎜ + mi ⎟ ⎥
a + n (1 + a )n ⎢⎣ i =1 ⎝2 ⎥⎦ ⎠ π
PEEGC =
⎛ N
⎞ ⎛1 N

Γ ( b ) Γ ⎜ ∑ mi ⎟ Γ ⎜ + ∑ mi ⎟
⎛ N
⎞ ⎝ i =1 ⎠ ⎝ 2 ⎠
Γ ⎜ b + ∑ mi ⎟ ⎡ N i =1

⎛ mi ⎞ ⎤
mi
⎝ ⎠⎢ 1 ⎡ N ni 1 ⎤
PE
SC
=
i =1

2Γ ( b )
∏ ⎜ ⎟ ⎥
⎢⎣ i =1 Γ ( mi + 1) ⎝ aγ i ⎠ ⎥⎦ ∞ ⎢∏ ( mi )ni ( mi + 1/ 2 )ni ( −mi / γ i ) ⎥
ni !⎦
⎣ i =1
× ∑ ⎛ N ⎞ ⎛1 N ⎞
⎡ L ⎛ mi ⎞ 1 ⎤ ⎜ ∑ mi ⎟ ⎜ 2 + ∑ mi ⎟
ni n1 , n2 ,..., nN = 0
⎛ N

⎜ b + ∑ mi ⎟ ⎢∏ ( mi ) n ⎜ − ⎟ ⎥ ⎝ i =1 ⎠ n1 +...+ nN ⎝ i =1 ⎠ n1 +...+ nN
∞ ⎝ i =1 ⎠n1 +...+ nN ⎢⎣ i =1 i
⎝ aγ i ⎠ ni !⎥⎦

N
× ∞ 2 ∑ ( mi + ni ) −1
n1 , n2 ,..., nN = 0
L

∏ ( m + 1) i ni
× ∫ Γ b, ar ( 2
)r i =1
dr
0
i =1

4
⎞⎡ N ⎛ 1 ⎞ ( mi / ( aγ i ) )
mi

-2
10
⎛ N
π Γ ⎜ b + ∑ mi ⎟ ⎢∏ Γ ⎜ + mi ⎟ ⎥ δ = 0.5
⎝ i =1 ⎠ ⎢⎣ i =1 ⎝ 2 ⎠ π ⎥

-3
10 δ=0
=
⎛ N
⎞ ⎛1 N ⎞
2Γ ( b ) Γ ⎜ 1 + ∑ mi ⎟ Γ ⎜ + ∑ mi ⎟
-4 SC
10

⎝ i =1 ⎠ ⎝ 2 i =1 ⎠

Average BER
-5
10 EGC
⎛ N
⎞ ⎡ N ⎛1 ⎞ 1 ⎤
⎜ b + ∑ mi ⎟ ⎢∏ ( mi ) n ⎜ + mi ⎟ ( − mi / ( aγ i ) )
n

i
MRC

⎝ ⎠n1 +...+ nN ⎣ i =1 i
⎝2 ⎠n n !⎦ -6


i =1 i
10
i

⎛ N
⎞ ⎛1 N ⎞
⎜ 1 + ∑ mi ⎟ ⎜ + ∑ mi ⎟
n1 , n2 ,..., nN = 0 -7
10
⎝ i =1 ⎠n1 +...+ nN ⎝ 2 i =1 ⎠ n1 +...+ nN
-8

⎞ ( mi / ( aγ i ) )
⎡ N mi
⎤ 10
⎛1
π ⎢∏ Γ ⎜ + mi ⎟ ⎥
⎢ i =1 ⎝ 2 ⎠ π ⎥ -9

⎣ ⎦
10
=
10 15 20 25

⎛ ⎞
N Average SNR (dB) of first branch
2Γ ( b ) Γ ⎜ b + ∑ mi ⎟ Fig. 1. Average BER for four- branch SC, MRC, and EGC with exponential power
⎝ i =1 ⎠ delay profile and different fading severity.
⎛ 1 1 N
−m − mN ⎞
× FB( N ) ⎜ m1 ,..., mN ; m1 + ,..., mN + ; b + ∑ mi ; 1 ,..., ⎟ fading along the diversity branches. The MGF of the output SNR
⎝ 2 2 i =1 aγ 1 aγ N ⎠
and the BER of binary modulations were derived in terms of the
(28) well known Lauricella functions.
⎧1, if b = 1/ 2
where b =⎨ . We note that the result in (28) can
⎩ 1/ 2 if b = 1 REFERENCES
be easily expressed in terms of the simplex integrals [13] using [7, [1] M. Nakagami, “The m-distribution, a general formula of intensity distribution
eq. 2.4.6], so that the average BER for EGC may also be expressed of rapid fading”, in Statistical Methods in Radio Wave Propagation, W. G.
Hoffman, Ed., Pergamon Press, Oxford, England, 1960.
in the forms given in [14]. However, we are unable to find a [2] [2] M. K. Simon and M. S. Alouini, Digital Communications Over Fading
transformation similar to [8, eq. A.2.1.2 or eq. A.2.2.1] that Channels, John Wiley and Sons, New York, 2000.
guarantees the convergence of the Lauricella function FB (...) in [3] H. Shin and J. H. Lee,“On the Error Probability of binary and M-ary signals
in Nakagami-m fading channels,” IEEE Trans. Commun., vol. 52, pp. 536-
(28). Therefore in the numerical analysis, we restrict the fading 539, April 2004.
parameter on each diversity branch to the range ½ < mi < aγ i . [4] V. A. Aalo, T. Piboongungon, and G. P. Efthymoglou, “Another look at the
performance of MRC schemes in Nakagami-m fading channels with arbitrary
parameters,” IEEE Trans. Commun., vol. 53, no. 12, pp. 2002-2005,
IV. NUMERICAL RESULTS December 2005.
[5] O. C. Ugweje, “Selection diversity for wireless communications in
In this section, we present some selected numerical results to Nakagami-fading with arbitrary parameters,” IEEE Trans. Veh. Tech., vol.
illustrate the use of the Lauricella functions to evaluate the 50, pp. 1437-1448, November 2001.
performances of linear diversity combining schemes operating in a [6] R. Annavajjala, A. Chockalingam, and L. B. Milstein, “Performance analysis
of coded communication systems on Nakagami fading channels with
Nakagami fading environment with arbitrary fading and power
selection combining diversity,” IEEE Trans. Commun., vol. 52, pp. 1214-
parameters. Numerical results are plotted for a four-branch 1220, July 2004.
diversity system with arbitrary Nakagami fading parameters [7] H. Exton, Multiple Hypergeometric Functions and Applications, John Wiley,
{m1 , m2 , m3 , m4 } = {2.0,1.5,1.0, 0.5} and for two cases of [8]
New York, 1976.
M. Abramowitz and I. Stegun, Handbook of Mathematical Functions, Dover
exponential power profile with decay rate δ, i.e., Publications, New York, 1970.
γ i = γ 1 exp [ −δ ( i − 1)] , i = 1,..., 4 . The average BER for all [9] I. S. Gradshteyn and I. M. Ryzhik, Tables of Integrals, Series and Products,
Academic Press, New York, 1980.
combining schemes versus the average SNR of the first diversity [10] A. Annamalai, C. Telumbura, V.K. Bhargava, “Equal-gain diversity receiver
branch is plotted in Fig. 1. As expected, MRC offers the biggest performance in wireless channels,” IEEE Trans Commun., vol. 48, pp. 1732-
1744, October 2000.
diversity gain, followed by EGC and SC, while the system with [11] V. A. Aalo, T. Piboongungon, and G. P. Efthymoglou, “Performance Analysis
balanced average SNRs (δ = 0) outperforms the one with of Equal Gain Combining Systems in Nakagami Fading Channels with
unbalanced average SNRs (δ = 0.5). Finally, note that the MGF Arbitrary Fading Parameters,” Proc. 11th European Wireless Conference,
based approach can be used to compute the SER of these diversity Nicosia, Cyprus, pp. 91-95, 10-13 April 2005.
[12] H. M. Srivatava and H. L. Manocha, A Treatise on Generating Functions,
combining schemes for M-ary modulations.
John Wiley, 1984.
[13] A. H. Stroud, Approximate Calculation of Multiple Integrals, Prentice Hall,
V. CONCLUSION Englewood Cliffs, New Jersey, 1971.
[14] G. K. Karagiannidis and A. A. Kotsopoulos, “Exact evaluation of equal-gain
In this paper, we have studied the performances of maximal ratio diversity in the presence of Nakagami fading”, Electronics Letters, vol.36,
combining, selection combining, and equal gain combining systems pp. 1229-1231, 6 July, 2000.
operating over a flat-fading Nakagami-m channel with independent

Das könnte Ihnen auch gefallen