\

=
2
(1)
The erfc(x) can be expressed in an integral oI exponential Iorm as 6
( ) 0 x , d
) ( sin
x
exp
2
x erfc
0


\

=
2
2
2
(2)
Also the erfc(x) can be expressed in the Iorm oI the sum oI exponential terms 6
( )
=


\

N
1 i
2
2
1  i
sin
exp
2
i
d
x
x erfc
i
( )
2
x b  exp a
i
N
1 i
i
=
= (3)
Ior 0
0
1
.
N
/2 where
) (
a
i i
i
1
2
= (4)
) ( sin
b
i
i
2
1
= (5)
The right hand side oI (3) is numerical integration using rectangular rule. The values oI N and number oI
i
can
be increased to decrease the error between the exact Iorm and the approximated Iorm to any arbitrary value. In
6, another simple approximation is derived starting Irom (2) by applying the numerical integration using the
trapezoidal rule in the case oI N=2 Ior an arbitrary point . This resulted in the exponential approximation Ior
the QIunction as
(
+ ) x exp( )
x
exp( ) x ( Q
2
2
3
2
2 3
1
4
1
(6)
However, Ior more accuracy, we repeat the numerical integration but increase the number oI terms to be N=3
Ior arbitrary points
1
,
2
to get
( )


\

+ 
\

1
2
2
2 2 2
2
1
sin
x
exp ) x exp( x erfc
) , , x ( g
sin
x
exp
2 1
2
2
2
1
2
1
=


\


\

+ (7)
Now
1
and
2
can be chosen to minimize the integral oI the relative error in the range oI values oI interest
{ }
( ) ( )
( )
dx .
x erfc
x erfc , x, g
R
min arg
R
opt
=
0
2 1 1
(8)
We got
opt1
= 0.2 and
opt2
= 0.35 Ior R 6. Substituting Ior
1
and
2
with their optimum values
opt1
and
opt2
in (7), we get
( )
2
2 2
15 0 3 0 35 0
79 0 35 0
x
.
x
.
x
e . e . e . x erfc
+ + = (9)
Thus substituting in (1), we get
3
(
(
+ + =
2
x
58 1
x
7 0
x
2 2 2
e 15 0 e 3 0 e 35 0
2
1
x Q . . . ) (
. .
(10)
III. CLOSED FORM OF BER FOR GENERALIZED MCCDMA SYSTEMS
Recently, in 5, a generalized model Ior MCCDMA had been proposed. The generalization in this model is
presented in terms oI the normalized Irequency spacing, , between two adjacent subcarriers. Based on this
model, the MCDSCDMA scheme belongs to the Iamily oI multitone DSCDMA arrangements, iI 1, while
to the class oI orthogonal MCDSCDMA systems, iI N
p
, where N
p
T
s
/ T
c
is the spreading gain oI the DS
spread subcarrier signals. Furthermore, there exists no overlap between the main lobes oI the modulated
subcarrier signals aIter DS spreading, when 2N
p
. The generalized model in terms oI is, thereIore, playing
its role in deIining the subcarrier Irequencies which are given by
f
u
f
0
s
T
) u U (
2
1
, u0, 2, 4,...2(U1) (11)
A Transmitter Model:
The transmitted signal Ior the k
th
user is S
k
(t) is given by
=
+ =
U
u
ku u k ku k
) t f cos( ) t ( c ) t ( b P ) t ( S
1
2 2
(12)
where P is transmitted power per sub carrier, U is the number oI subcarriers b
ku
(t),c
k
(t) and
ku
are the data
stream, the direct sequence spreading code waveIorm and the phase angle oI the carrier modulation process.
B Channel Model:
It is assumed that the channel between the K
th
user transmitter and receiver is a multipath Nakagamim Iading
channel with the channel complex low pass representation oI the impulse response experienced by the subcarrier
u oI the user k is given by
=
=
1
0
Lp
lp
) k (
l , u
l , k
) k (
l , u
ku
) j exp( ) t ( ) t ( h
p
p
p
(13)
It is assumed that the phases are identically independently distributed (i.i.d) random variables, uniIormly
distributed over the interval 0,2). Also the attenuation Iactors are independent Nakagamim random variables.
The multipath intensity proIile (MIP) distribution is assumed to be a negative exponentially decaying
distribution.
C Receiver Model:
The Iirst user is assumed to be the user oI interest. A correlatorbased Rake receiver in conjunction with MRC is
used. The receiver is assumed to have perIect time domain synchronization with each path oI the reIerence
signal, also, the attenuations and phases are assumed to be perIect estimates oI the channel parameter. The
received signal is expressed by
) t ( b P ) t ( n ) t ( r
p p
p
p
kl ku
) k (
ul
K
k
U
u
L
l
+ =
= =
= 1 1
1
0
2 ) 2 cos( . ) ( .
) (k
ul u kl k
p p
t f t c + (14)
where
p p p
l , k u
) k (
l , u ku
) k (
l , u
t f 2 = is assumed to be an i.i.d. random variable having a uniIorm
distribution in 0,2), while n(t) represents the additive white Gaussian noise (AWGN) with zero mean and one
sided power spectral density oI N
o
.
4
D Average Bit Error Probability:
It was shown in 5 that the BER using BPSK modulation conditioned on a set oI attenuations
p
l ,
is given by



\

=


\

=
=
1
1
2
2
p
p
L
l
p
l b
Q
] Zv [ Var
]) Zv [ E (
Q ) ( P (15)
where
o
l ,
c l
p
p
.
2
=
,
p
l ,
is the attenuation Iactor oI the subcarrier u in the path l
p
,
o
is the average
signal strength,
( ) ( )
( )
1
1
1
3
1
1 2
(
(


\

+
+


\

=
M
p p
p p
o
b o
c
I U
N
.
L
, L q KL
N
E .
(16)
where,
b
E is the energy per bit,
o
N is the AWGN component, and the average SNR per bit
b
is given by
( )


\

=
o
b o
p b
N
E .
, L q
(17)
where is the rate oI average power decay, and ( ) , L q
p
is given by
( )
) exp(
) L exp(
, L q
p
p
=
1
1
(18)
N
p
is the spreading gain oI the subcarrier signals, and
M
I are given by
( )
( )
  ( )
  ( )
(
(
(
(
(


\



\

=
U
u
u
p
p
p M
N
u
N
u
u
N
U U
I
1 1
2 2 2
2
2
sin
1 .
2
.
1
1
(19)
In order to obtain the average BER, P
b
, we must average the conditional BER given by (15) over the joint pdI oI
the instantaneous SNR sequence
1
0
=
p
p p
L
l l
} { , thus we have
( )
1 1 0 1 1 0 ,.., ,
0
1
0
0
... ,..., , . 2 ......
1 1 0


\

=
p p
p
L
p
p
p
L L
L
l
l c b
d d d p S Q P
(20)
Where
p
l
S , l
p
= 0, 1,, L
p
1, the normalized square oI the Nakagami distributed multipath attenuations, is
expressed by
o
l , q
l
p
p
S
2
= (21)
Since the random variables
1
0
=
p
p p
L
l l
} { are assumed to be statistically independent, we have
( ) ( )
1
0
1 1 0
1 1 0
p
p
p
p
l p
p
L
L
l
l L ,.., ,
p ,.., , p
(22)
5
Substituting (21) and (22) into (20), we get
( ) ( )
p
p
p
p
p
l p
l
L
l
l l b
d p . Q P
=
1
0
0
2 (23)
Now substituting in (10), the modiIied exponential approximation, we get
( )
( )
( )











\

(
+
(
(
+
(
(
=
p
p
p
p
p
l
p
l
p
p
p
p
p
l
p
l
p
p
p
p
p
l
p
l
l
L
l
l
l
L
l
l
.
l
L
l
l
.
b
d p . e .
d p . e .
d p . e .
P
1
0
0
1
0
0
79 0
1
0
0
35 0
15 0
3 0
35 0
(24)
Since
o
l ,
c l
p
p
.
2
= and
p
l ,
obeys the Nakagamim distribution,
o
l , q
l
p
p
S
2
= is a Gamma distributed
random variable having a pdI given by
( )


\

=
p p
p l
l
m
m
l
m
l
m
exp
) m (
m
p
1
(25)
Substituting (25) into (24) will result in the Iorm oI Laplace transIorm. Thus (24) can be expressed as
\




\

+
+



\

+
=
=
=
1
0
1
0
35 0
35 0
35 0 15 0
p
p
p
p
L
l
p
L
l
p
b
m
m .
l
m .
.
m
m
l
m
. P








\

+
+
=
1
0
79 . 0
79 . 0
3 . 0
p
p
L
l
p
m
m
l
m
(26)
It was shown that the exact bit error probability P
b
, using MGF based approach, is given by 5, Eq. 47
d
) ( sin . m
) ( sin . m
P
p
p
L
l
m
p
l
b
=


\

+
=
2
0
1
0
2
2
1
(27)
This can be evaluated using numerical integration. It was shown also that the approximate bit error probability
P
b
, using the QIunction exponential approximation based approach, is given by 7, Eq. 30



\



\

+
+


\

+
=
=
=
1
0
1
0
3
1
75 0
75 0
4
1
p
p
p
p
L
l p
l
L
l p
l
b
m
m
m
m
m .
m .
P
(28)
These three results will be compared together. The comparison will be illustrated in the sense oI percentage
error Ior each approximation.
6
IV. NUMERICAL RESULTS
In this section, the bit error probability, derived in the previous section will be compared to those previously
published. The percentage error Ior both approximations will be also investigated.
The BER results versus the average SNR per bit were obtained Ior the multi tone case considering L
1
32,
N
1
128, 0.2 and U32. Fig.1 and Fig.3 describe the behavior oI the system, assuming L
c
5, with the
variation oI m Ior (m1, 2, 1000), with the later simulates the case Ior which m tends to inIinity, Ior a single user
and Ior K10 respectively. Fig.2 and Fig.4 illustrate the percentage oI approximation error Ior each case. Fig.5
describes the behavior oI the system with the variation oI L
c
Ior (L
c
1, 3, 7) Ior m1 and K10. Fig.6 illustrates
the percentage oI approximation error Ior each case. The case oI L
c
5 can be illustrated in other Iigures. Fig.7
gives illustration to the eIIect oI number oI users K on the BER Ior (K1, 10, 50) assuming L
c
5 and m1.
Fig.8 illustrates the percentage oI approximation error Ior each case. Fig.9 is an extra illustration Ior the eIIect oI
number oI users K on the BER Ior (K1 to 100) assuming L
c
5, m1 and dB
b
15 = .
From the numerical results, it is obvious that the results obtained using new modiIied approximation are very
precise and are nearly coincident with the previously published exact results.
V. CONCOLUSION
Numerical results had been presented and compared to the previously published results. From the comparison, it
is concluded that the derived approximation can be used with good accuracy, with no need to perIorm any
numerical integrations or computer simulations. The new approximation also has an advantage oI being in the
Iorm oI simple exponential terms. This Iorm gives an ability to be included in many calculations or
mathematical manipulations easily.
REFERENCES
1 R. Prasad and S. Hara, Overview oI multicarrier CDMA, IEEE Commun. Mag., pp. 126133, Dec. 1997.
2 E. A. Sourour and M. Nakagawa, PerIormance oI orthogonal multicarrier CDMA in a multipath Iading
channel, IEEE Trans. Commun., vol 44, pp. 356367, Mar. 1996.
3 M.S. Alouini and M. K. Simon, PerIormance oI coherent receivers with hybrid SC/MRC over Nakagami
m Iading channels, IEEE Trans. Veh. Technol., vol. 48, pp. 11551164, July 1999.
4 M. K. Simon and M.S. Alouini, Digital communications over generalized Iading channels: A uniIied
approach to perIormance analysis. New York: Wiley, 2000.
5 Lie Liang yang and Lajos Hanzo, "PerIormance oI generalized multicarrier DSCDMA over Nakagamim
Iading channels", IEEE transactions on communications, vol. 50, June 2002, pp.956966.
6 Marco Chiani ,David Dardari and Marvin K. Simon, new exponential bounds and approximations Ior the
computation oI error probability in Iading channel, IEEE transactions on wireless communications vol. 2,
pp.840845, July 2003.
7 Said M. Elnoubi, and A. Mabrouk Elshinnawy "New Closed Form Solution For The Bit Error Rate (BER)
oI MCCDMA systems in Generalized Nakagamim Fading Channels", National Radio ConIerence, Cairo,
Egypt, March, 2004.
7
0 2 4 6 8 10 12 14 16 18 20
10
7
10
6
10
5
10
4
10
3
10
2
10
1
10
0
Average SNR per bit (dB)
B
E
R
m = 1
m = 2
m = 1000
BER vs. SNR for different m, K=1
Exponential Approximation
Modified Approximation & MGF
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
Fig. 1 Error Probability vs. average
SNR per bit for different m, K=1
0 2 4 6 8 10 12 14 16 18 20
10
6
10
5
10
4
10
3
10
2
10
1
10
0
Average SNR per bit (dB)
B
E
R
m = 1
m = 2
m = 1000
BER vs. SNR for different m, K=10, Lc=5
Exponential Approximation
Modified Approximation & MGF
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
Fig. 3 Error Probability vs. average
SNR per bit for different m, K=10
0 2 4 6 8 10 12 14 16 18 20
10
4
10
3
10
2
10
1
10
0
Average SNR per bit(dB)
B
E
R
BER vs SNR for different no. of combined paths Lc, m=1, K=10
Lc = 1
Lc=3
Lc=7
Exponential Approximation
Modified Approximation
& MGF
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
Fig. 5 Error Probability vs. average
SNR per bit for different L
c
0 2 4 6 8 10 12 14 16 18 20
0
5
10
15
20
25
30
35
40
45
50
Average SNR per bit(dB)
E
r
r
o
r
P
e
r
c
e
n
t
a
g
e
(
%
)
Error percentage in BER vs SNR for different m
Exponential Approximation
Modified Approximation
m = 1, Exp Appr.
m = 1, Mod Appr.
m = 2, Exp Appr.
m =2 , Mod Appr.
m = 1000, Exp Appr.
m = 1000, Mod Appr.
Fig. 2 Error Percentage vs. average
SNR per bit for different m, K1
0 2 4 6 8 10 12 14 16 18 20
0
5
10
15
20
25
30
Average SNR per bit (dB)
E
r
r
o
r
P
e
r
c
e
n
t
a
g
e
(
%
)
Error percentage in BER vs SNR for different m
Exponential Approximation
Modified Approximation
m = 1, Exp Appr.
m = 1, Mod Appr.
m = 2, Exp Appr.
m =2 , Mod Appr.
m = 1000, Exp Appr.
m = 1000, Mod Appr.
Fig. 4 Error Percentage vs. average
SNR per bit for different m, K10
0 2 4 6 8 10 12 14 16 18 20
0
5
10
15
20
25
Average SNR per bit(dB)
E
r
r
o
r
P
e
r
c
e
n
t
a
g
e
(
%
)
Error percentage in BER vs SNR for different Lc
Exponential Approximation
Modified Approximation
Lc = 1, Exp Appr.
Lc = 1, Mod Appr.
Lc = 3, Exp Appr.
Lc =3 , Mod Appr.
Lc = 7, Exp Appr.
Lc = 7, Mod Appr.
Fig. 6 Error Percentage vs. average
SNR per bit for different L
c
8
0 2 4 6 8 10 12 14 16 18 20
10
6
10
5
10
4
10
3
10
2
10
1
10
0
Average SNR per bit (dB)
B
E
R
BER vs SNR for different K, m=1, Lc=5
K = 50
K=10
K=1
Exponential Approximation
Modified Approximation & MGF
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
MGF
Exp Appr.
Mod Appr.
Fig. 7 Error Probability vs. average
SNR per bit for different K
0 2 4 6 8 10 12 14 16 18 20
0
5
10
15
20
25
Average SNR per bit(dB)
E
r
r
o
r
P
e
r
c
e
n
t
a
g
e
(
%
)
Error percentage in BER vs SNR for different K
Exponential Approximation
Modified Approximation
K = 1, Exp Appr.
K = 1, Mod Appr.
K = 10, Exp Appr.
K =10 , Mod Appr.
K = 50, Exp Appr.
K = 50, Mod Appr.
Fig. 8 Error Percentage vs. average
SNR per bit for different K
0 10 20 30 40 50 60 70 80 90 100
10
5
10
4
10
3
10
2
10
1
BER vs SNR for different K, m=1, Lc=5
Number of users, K
B
E
R
MGF
Exponential.Approximation
Modified Approximation
Fig. 9 Error Probability vs. number of active users K
9
This paper introduces a closed Iorm expression Ior the bit error rate (BER) oI Generalized Multicarrier Code
Division Multiple Access (MCCDMA) system based on a new modiIied exponential approximation. The
closed Iorm oI the BER will be derived using the proposed new approximation as well as the exponential
approximation proposed previously in 6. The results using both approximations are compared to the results
that are previously published. The BER can be now easily computed using simple substitution without either the
need to use computer simulation or to perIorm a numerical integration.
10
'= '==   ~ " ,~ ;~' ;~;  '  '= '=; "
;~ , 'V  = ~ =' " '= "
. ' = = . . ,' = , . . '' ,=
,' ,' '=` '   " '= " , '' .' ,= ,' `'` = ,, , '=  ,
''  _'= =`  '==' ' ) ='  _'= ,  '+, '',= ( _' '` ' ,' =' _'= 
'`' ,=='' ',' '' .
= ' , , , '''   , '  ,''= = , ,' ,' '=` '' " '= ." ,'  = 
. `'` =  '=' '==' ' ' , '=` ='  ' ' ,=' ' `= " '=''  ." '=,' '= , '='
' ,'' ,',,'  '' ' ''=' .
' ,' ' =  ' =' '' ,` ,' ,' =' =' '='' '==' ' ' , '  ` , ,
`' =' '+,'= .= _ ' '' _'' _ ,' ,,'  `' _'' '  .  '+ '' = .
'='' ''=' =` ='='  ' ,' _ =,' ,' =' ',+ '=='  '= `  .'' =
' .