IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO.

5, SEPTEMBER 2004 1327

On the Performance and Capacity of an Asynchronous

SpaceTime Block-Coded MC-CDMA System in
the Presence of Carrier Frequency Offset
Xiaoyu Hu, Student Member, IEEE, and Yong Huat Chew, Member, IEEE

AbstractIn this paper, the bit-error rate (BER) performance Because of this, MC-CDMA can make use of this frequency di-
and capacity of asynchronous spacetime block-coded (STBC) versity to improve the system performance, whereas DS-CDMA
multicarrier code-division multiple-access (MC-CDMA) systems relies on the use of a more complicated RAKE receiver. It has
in the presence of carrier frequency offset (CFO) between the
transmitter and receiver oscillators are analyzed. The exact BER been shown in [1] that MC-CDMA outperforms DS-CDMA,
expression when using equal gain combining (EGC) and the MC-DS-CDMA, and multitone CDMA.
approximate BER expression when using maximum ratio com- Information theoretic studies have shown that the antenna di-
bining (MRC) are derived. These BER expressions are verified versity provided by multiple transmit and receive antennas al-
through simulations. Using these derived expressions, the achiev- lows for a dramatic increase in the capacity and is an effec-
able system capacity satisfying a minimum BER requirement
can be studied for the two cases when EGC and MRC are used tive technique for combatting fading in wireless communica-
and, hence, it is possible to compare the achievable capacity of tion systems [2], [3]. Spacetime (ST) coded systems, which
STBC MC-CDMA systems with that of MC-CDMA systems. It integrate the techniques of antenna array spatial diversity and
is concluded that small CFO has an insignificant effect on the channel coding to combat the destructive multipath channels
BER and capacity of STBC MC-CDMA systems and that this and interference from other users, have been a key focus of re-
range of CFO is important in transceiver design. Besides, STBC
MC-CDMA systems with multiple receive antennas can achieve search over the last few years [4][7]. The spatial nature of ST
higher capacity than that of the MC-CDMA systems; this amount codes guarantees diversity gain. The temporal nature guarantees
can be obtained analytically using the theoretical BER expressions that the diversity advantage is achieved without sacrificing the
derived. transmission rate. Commonly used ST coding schemes are the
Index TermsBit-error rate (BER), carrier frequency offset ST trellis codes and ST block codes. Since both MC-CDMA
(CFO), multicarrier code-division multiple access (MC-CDMA), and ST block-coded (STBC) systems can achieve better system
multiuser interference (MUI), spacetime block code (STBC). performance, the combination of both, which is also known as
STBC MC-CDMA, will surely obtain much better performance.
I. INTRODUCTION In fact, such systems have been attracting much attention re-
cently [8][10].

F UTURE wireless mobile systems are required to transport

multimedia traffics at much higher bit rates, which moti-
vates the communications research community to work on the
There are studies that reported on the bit-error rate (BER)
performance of MC-CDMA [1], [12] and STBC MC-CDMA
systems [8][11] using synchronous and asynchronous trans-
technologies suitable for the next generation of wireless mo- missions, but perfect carrier frequency synchronization was
bile communication systems. Multicarrier code-division mul- assumed. However, a major drawback of multicarrier modula-
tiple access (MC-CDMA) has emerged as a powerful candidate tion is that it is sensitive to the carrier frequency offset (CFO)
due to its capabilities of achieving high capacity over a fre- between the transmitter and receiver oscillator. CFO causes a
quency-selective fading channel [1]. In MC-CDMA, the same loss of orthogonality between subcarriers and, thus, inevitably
data symbol is transmitted simultaneously over all the narrow- results in intersubcarrier interference, which results in severe
band subcarriers and, at each subcarrier, it is encoded by chips performance degradation in MC-CDMA systems. In [13], the
taking a value of either 1 or 1, depending on the assigned effect of CFO on the downlink performance of MC-CDMA
spreading code. Each user in the system is assigned to a unique systems was investigated. They used a computer-simulation
code. While direct-sequence code-division multiple access (DS- approach to study the multiuser case, while the analytical
CDMA) spreads the signals in the time domain, MC-CDMA approach was only for a single user situation. In [14], the effect
applies the same spreading sequence in the frequency domain. of CFO on the downlink performance of MC-CDMA system
in the frequency-selective fading channels was studied. There
Manuscript received July 26, 2003; revised February 4, 2004 and April 19, also were only a few reports that considered the effect of CFO
2004. on asynchronous MC-CDMA systems. In [15], the impact of
X. Hu is with the Institute for Infocomm Research, Singapore 119613,
Singapore, and with the Department of Electrical and Computer Engineering, CFO in the uplink of MC-CDMA systems was investigated,
National University of Singapore, Singapore 119260, Singapore (e-mail: but only additive Gaussian white noise (AWGN) channels
[email protected]; [email protected]). were considered. In [16], although frequency-selective fading
Y. H. Chew is with Institute for Infocomm Research, Singapore, Singapore
119613 (e-mail: [email protected]). channels were studied in the uplink of MC-CDMA systems,
Digital Object Identifier 10.1109/TVT.2004.832402 only synchronous transmission was assumed. In [17], the effect
0018-9545/04$20.00 2004 IEEE
1328 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

of CFO on asynchronous MC-CDMA systems in correlated Next, every block symbol at the th
Rayleigh-fading channel was studied; however, this approach transmit antenna is then serial-to-parallel converted to par-
uses complicated multidimensional integration when com- allel data streams. Each data stream is spread and transmitted
puting the BER performance. To the authors best knowledge, on different subcarriers, where is the processing gain.
there still are no closed-form expressions to compute the Random spreading sequence is used in this paper. The symbol
BER of asynchronous MC-CDMA and asynchronous STBC duration on each subcarrier is times of the input symbol
MC-CDMA multiuser systems. duration , i.e., . Therefore, frequency-selective
In this paper, the effect of CFO on the performance and fading over each subcarrier can be avoided by increasing
capacity of asynchronous STBC MC-CDMA systems in if is less than the channel-delay spread. The MC-CDMA
Rayleigh-fading channels are studied. A method when using modulator transmits a total of chips resulting from the
equal gain combining (EGC) and an approximate approach BPSK symbols over a total of subcarriers. The frequency
when using maximum ratio combining (MRC) are presented. separation between the neighboring subcarriers is Hz
In both cases, the BER expressions obtained are verified and the passband null-to-null bandwidth of each subcarrier is
through computer simulations. Using these BER expressions, . The ideal central baseband frequency of each subcarrier
system capacities for both EGC and MRC receivers are studied. is given by
Although independent channel characteristics between neigh-
boring subcarriers are assumed, the theoretical performance (1)
analysis can be a good approximation to the practical chan-
nels when the correlation coefficients between neighboring In order to achieve frequency diversity, the assignment
subcarriers are small; for example, less than 0.5, and can be a of subcarriers to the chips is made such that the
good lower bound when the neighboring subcarriers have large frequency separation among subcarriers conveying the chips
correlation coefficients [18]. Our analysis provides a simple of the same data symbol is kept at maximum, as shown in
way to study the effect of CFO on the BER performance and Fig. 1. This means that data in the th data
capacity of asynchronous STBC MC-CDMA systems. stream is transmitted on the th subcarrier with the frequency
The remainder of this paper is organized as follows. The and the adjacent frequency
system model is described in Section II and various interfer- separation between these subcarriers is . Let be the
ence terms present in the STBC MC-CDMA systems are given null-to-null passband bandwidth of the STBC MC-CDMA
in Section III. In Section IV, the analysis to obtain the BER system with subcarriers; then .
expressions and system capacity of the STBC MC-CDMA Symbol duration in each subcarrier is assumed to be longer
systems, when either EGC or MRC is used, is presented. Nu- than channel-delay spread so that each subcarrier signal under-
merical results are presented and discussed in Section V. The gone flat fading and the fading process of a given user on dif-
conclusion is finally given in Section VI. ferent subcarriers is assumed to be independent throughout this
paper. In the literature, when using this model, no cyclic pre-
II. SYSTEM MODEL fixes (CP) need to be considered, since all subcarriers undergo
The equivalent block diagram of the transmitter and receiver flat fading [12][17]. In practical implementation, where inverse
of a STBC MC-CDMA system used in our analysis is shown in fast Fourier transform (IFFT) is used, the intersymbol interfer-
Fig. 1. The stream of binary phase-shift keying (BPSK) symbols ences (ISI) for the MC-CDMA symbol have to be considered
of the th user first goes through the ST block encoder. Without since the channel fading is modeled at the chip period interval
incurring any power or bandwidth penalty, the Alamoutis or- over the full bandwidth instead of at the symbol period interval
thogonal ST block coding scheme for two transmit antennas is over each subcarrier bandwidth. Normally, the use of CP re-
employed [6], [7] and more general ST codes for other number quires the use of additional bandwidth, which is not accounted
of transmit antennas can be found in [7]. Two successive block for in this analysis, since our system model is equivalent to the
symbols and for the th user are mapped to IFFT/fast Fourier transform (FFT) implemented system after
the following matrix: CP has been removed at the receiver. This bandwidth-expan-
sion factor needs to be compensated when applying the results
obtained in this paper to the IFFT/FFT implemented system.
Assume that there is asynchronous users, each employing
BPSK with the same transmitting power and at the same data
bit rate . The transmitted signal at the th antenna of user
can be expressed as
where denotes the th block
symbols for the th user with length , denotes the trans-
pose, and denotes the th
block symbols at the th transmit antenna for the th user with
length . The columns are transmitted in successive block time (2)
with the symbols in the upper and lower blocks sent through
the two transmit antennas simultaneously. The factor is where denotes the th data bit of the th data stream
to normalize the transmitted symbol energy. at the th antenna of user , is the th
HU AND CHEW: PERFORMANCE AND CAPACITY OF AN ASYNCHRONOUS SPACETIME BLOCK-CODED MC-CDMA SYSTEM 1329

Fig. 1. STBC MC-CDMA system model with 2Tx2Rx.

spreading chip for the th data bit of the th user, and is Consider uplink transmission: The received signal at the th
the rectangular pulse define over . The subscript receive antenna is given by

(3)

and is the ideal central baseband frequency at the th sub-

carrier given by (1) and is the CFO between the th users
transmitter and the oscillator at the receiver of the base station. (4)
1330 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

where is the transmission delay from the th transmit

antenna of user to the th receive antenna, which is indepen-
dently and uniformly distributed over for different , ,
and . , ,
and are, respectively, the amplitude and phase of the (9)
channel-fading gain of user when the signal is transmitted
from the th transmit antenna to the th receive antenna through
the th subcarrier, is Rayleigh distributed with where , , and
, and is uniformly distributed . In the following, we derive
the respective interference terms defined in (6) for a referenced
over . denotes AWGN at the th receive antenna
subcarrier at the receive antenna of user 1.
with zero mean and double-sided power spectral density (PSD)
.
Assume that user 1 is of interest and that the coherent receiver A. Self-Interference From the Other Subcarriers
is used. Without loss of generality, let . At the th Here we consider only those terms corresponding to .
subcarrier of the th parallel data stream, the coherent receiver There are, altogether, subcarriers that will interfere the
output for the th symbol of the th antenna is given by referenced subcarrier .

(5)

where is the desired signal term before the decision is

made and

(10)
(6)

where denotes the self-interference from the re-

maining subcarriers, and denote the mul-
tiuser interference (MUI) from the same subcarriers and other
subcarriers, respectively, and is the AWGN.

III. INTERFERENCE ANALYSIS

(11)
Channel fading is assumed to be quasistationary over the two
successive symbol intervals , i.e., the channel gain remains
constant over the time interval where , ,
which are uniformly distributed over .

(7) B. Multiuser Interference From the Same Subcarrier

Here we consider only terms in (5), resulting from other users
Let be the CFO normalized by the subcarrier who transmit using subcarrier .
spacing of each subcarrier for user . We assume that
and is uniformly distributed over .
This corresponds to the case where user 1 will have the worst
performance among the group of users since its offset frequency
observed by the base station receiver is the largest. The signal
and respective interference terms can be derived using (4)(6).
The desired th and th block symbols at the th sub-
carrier of the th data stream of the th antenna can be shown
to be given by

(8)
HU AND CHEW: PERFORMANCE AND CAPACITY OF AN ASYNCHRONOUS SPACETIME BLOCK-CODED MC-CDMA SYSTEM 1331

(12) (13)

where and
, are all uniformly
distributed over .

C. Multiuser Interference From the Other Subcarriers

Here, we consider those terms in (5) resulting from other

users who transmit using different subcarrier, as shown
in (14) and (15) at the bottom of the page where
,
, , 2, all are uniformly distributed over
.

(14)

(15)
1332 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

D. Noise And the MUI from other subcarriers is given by

The AWGN noise term has zero mean and variance

(16)

(23)
IV. BER PERFORMANCE AND CAPACITY ANALYSIS
For large value of , , , and
At the th subcarrier of the th receive antenna, we denote can be approximately assumed to be Gaussian
the factors used to weight the received signals and distributed with zero mean. The variance for defined in
from the two successive intervals to be ,
(21) is given by
and , , respectively. The
values are to be chosen depending on the combining scheme
used. For BPSK only, the real part of the signal is used for
the making decision. The decision variables for and
are given, respectively, by (24)
where .
Here, we assume that all the terms corresponding to different
subscripts are all independent from each other. Similarly, for
(22)
(17)

(25)
where
(18) .
is the generalized hypergeometric function and its
where denotes the complex conjugate. In the following BER definition can be found in [23]. For (23)
analysis, will be used as an example and we will simply
use to denote . Equation (17) can be expressed as

(19)
(26)
where . where
The desired signal is given by ,

, and . The
noise term can be written as

(20)

The self interference (SI) from other subcarriers is given by

(27)

which is a Gaussian random variable (RV) with zero mean and

variance

(21)
(28)
The MUI from same subcarriers is given by

Since interference items , ,

and are mutually uncorrelated, the total interference
also is a Gaussian RV with zero mean and variance

(22) (29)
HU AND CHEW: PERFORMANCE AND CAPACITY OF AN ASYNCHRONOUS SPACETIME BLOCK-CODED MC-CDMA SYSTEM 1333

A. EGC Substitute (36) and (37) into (35); then, the CHF of is given
For EGC scheme, the factors used to weight the received by
signal at the th subcarrier of the th antenna are given by

(30)

and
(38)

Let . The exact BER for the th data

stream is given by [23]
(31)

where , and can be ob-

tained by using a carrier synchronizer to estimate the phase (39)
. We assume perfect synchronization,
i.e., throughout this paper. Substitute (8), (9),
and (30) in (20); the desired signal is given by where

(32)

Assuming that a is transmitted, (19) can be rewritten as

(33)

Since BPSK is used, then the BER for the th data stream can (40)
be obtained by using
is the order of the Hermite polynomial and we found that
(34) is sufficient for good accuracy. is the th zero
of the th order Hermite polynomial and are the weight
where is the characteristic function (CHF) of the deci- factors given by
sion variable . is the imaginary part of complex number
. It can be easily shown that (41)

It is assumed that any bit can be sent via any of the data
(35)
streams with equal probability. Therefore, the system average
BER if the EGC scheme used is given by
Since and are independent Rayleigh RVs, then
RV is Nakagami- [21] distributed with pa- (42)
rameter . Its CHF is given by [22]

Under certain quality-of-service (QoS) requirements, the

BER performance with an EGC receiver has to be better than
a given BER threshold . This means that the system
BER performance needs to satisfy
(36)
(43)
where denotes the confluent hypergeometric func-
We define as the largest number of users that the
tion [23]. The CHF of Gaussian RV is given by [20]
system can accommodate while satisfying the QoS requirement
(43), i.e.,
(37)
(44)
1334 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

The system capacity (bandwidth efficiency) if EGC is used is where is the probability density function (pdf) for RV
given by . It can be seen that (53) involves computing the compli-
cated -dimensional integrations. To avoid such complexity
(45) in computing the BER, an approximate approach is proposed in
this section.
In the following, we first introduce a lemma and then use
where is the data rate of the system and is the
it to simplify the expression given in (51). The conditions for
total bandwidth of the system.
such approximation to be valid are also stated. The theoretical
justification on these conditions is presented in Appendix B.
B. MRC
Lemma 1: Given that are positive inde-
For the MRC scheme, the factors used to weigh the received pendent and identically distributed (i.i.d.) RVs with mean
signal at the th subcarrier of the th antenna are given by and variance . are positive i.i.d. RVs with
mean and variance , where ,
. Then, it can be shown that
(46)
and (54)

The conditions for the approximation in (54) to be held is given

by
(47)
Assume that a is transmitted; (33) becomes

(48) (54a)

where . Similarly, can be (54b)

obtained by using a carrier synchronizer. The variance of the
total interference can be written as where is the level of inaccuracy to be defined in Appendix B
and is the desired bound of the normalized standard devia-
(49) tion for to be defined in (91) of Appendix B. The correctness
of this lemma is also verified through simulations by arbitrarily
choosing some pdfs with appropriate values of mean and vari-
where is the interference coefficient at the th subcarrier of
ance.
the th data stream, which is given by
Since and are independent Rayleigh RVs with
(50) , , 2, can be
shown to be Gamma distributed [21]. The mean and variance
The variances of the respective terms in (50) have been given are given by
in (24)(26). The conditional signal-to-interference and noise
ratio (SINR) is, therefore, defined as (55)

Denote and
as the mean and variance of
(51)
the interference coefficients . We show
in Appendix C that for all values of

The BER for the th data stream conditioned on (56)

is given by
(52) when . We show in Appendix B that by applying
, , , and to
The BER for the th data stream is obtained via averaging Lemma 1, the two conditions given by (54a) and (54b) are both
over satisfied. Under this circumstance, the variance of in (49) can
be approximated by

(57)
(53)
HU AND CHEW: PERFORMANCE AND CAPACITY OF AN ASYNCHRONOUS SPACETIME BLOCK-CODED MC-CDMA SYSTEM 1335

Then, the conditional SINR for the th data stream can be

simplified as

(58)

To simplify the notation, (58) can be written as

(59)

where denotes

hereafter. Then, the approximate BER for the th data stream

is given by

(60)

where is the pdf for RV , which can be given by [20]

(61)

where . Therefore, the closed-form expres-

sion of BER for the th data stream is given by [20]

(62)

where . It is assumed that any bit can

be sent via any of the data streams with equal probability.
Therefore, the system average BER if MRC scheme is used is
given by

(63) Fig. 2. BER versus normalized carrier frequency offset " (Es=N o =
20 dB; P = 1); (a) EGC and (b) MRC.

It can be seen that it will be much easier to use the closed-form

expression (62) rather than (53) to obtain the BER performance
for the MRC receiver. V. NUMERICAL RESULTS
Similarly, the BER performance with the MRC receiver has
to be better than the given BER threshold to satisfy the
QoS requirement. This means that the system performance is The effect of CFO on BER performance and system capacity
required to satisfy for asynchronous STBC MC-CDMA systems can be obtained
easily by using (42), (45), (63), and (66). The results obtained
(64) are also compared with the conventional MC-CDMA systems
. For comparison, BPSK is used and the total system band-
Similarly, we define as the largest number of users that width and fading channel statistics for all subcarriers are chosen
fulfils the QoS requirement if MRC is used. to be the same. The BER threshold is set to to
satisfy the QoS requirement. The symbol energy is defined as
(65) . The normalized CFO is set to for desired user
1 and for the th user, which is assumed to be uniformly
Therefore, the system capacity (bandwidth efficiency) if the
distributed over . The simulation results are obtained by
MRC used can be obtained by
taking average over 500 sets of uniformly distributed random
for each given value of . In the following
(66) numerical results, the BER performance and system capacity of:
1336 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

Fig. 3. System capacity versus normalized carrier frequency offset Fig. 4. BER versus the number of parallel data streams P (Es=N o =
" (Es=N o = 20 dB; P = 1). 20 dB; L = 32; K = 8).

1) MC-CDMA systems with one transmitter and one receiver;

2) STBC MC-CDMA systems with two transmitters and one re-
ceiver; and 3) STBC MC-CDMA systems with two transmitters
and two receivers are investigated.
Figs. 2 and 3 show the BER performance and system capacity
as a function of the normalized CFO when dB
and with the number of data streams is set to 1. The ratio
is kept at a constant for fair comparison when in-
creases. It can be found that the BER performance and system
capacity remains almost the same when is less than 0.01
for the three systems. For is between 0.01 and 0.1, the
system performance degrades very slightly, since MUI is the
main impairment source and results in the error floor. Hence,
maintaining CFO within this range is already sufficient in trans-
ceiver design. When is larger than 0.1, BER performance
and system capacity deteriorate severely. The performance of
the MC-CDMA systems obtained by using our expressions
are in good agreement with [17], where the fading correlation
Fig. 5. BER versus Eb=No dB (L = 32; K = 8; P = 1).
coefficients between neighboring subcarriers is equal to 0.26.
In general, the performance and system capacity of STBC
MC-CDMA systems with two transmitters and one receiver are , the results we obtained are almost the same as that of
only slightly better than that of MC-CDMA systems despite due to the assumption that fading processes of a given
of the diversity-combining schemes applied. On the other user on different subcarriers are independent each other and that
hand, the BER performance and system capacity of STBC the slight change is due to the slight different in the interference.
MC-CDMA systems with two transmitters and two receivers However, the introduction of is necessary in practice because,
are much better than that of previous two systems, despite through properly choosing the value of , it can be guaranteed
the diversity-combining schemes used. The system capacity is that all subcarriers used by a given user undergo flat fading,
about 1.52 times higher than MC-CDMA systems. Besides, since the symbol duration over subcarriers used by a particular
the computed BER results using our proposed analysis agree user will be increased by times.
with the results obtained from our simulations well. This ver- Fig. 5 shows the BER performance versus given
ifies the correctness and effectiveness of our analysis for both , , and . Fig. 6 shows the system capacity versus
the asynchronous STBC MC-CDMA and MC-CDMA systems , given and . It can be observed that
with either EGC or MRC receivers. It can also be found that the BER performance and system capacity improved gradually
MRC receivers outperform EGC receivers in the presence of as increases when is small. BER performance and
CFO for these three asynchronous systems. system capacity remain almost the same when is greater
Fig. 4 shows BER versus the number of parallel data streams than 20 dB. For large , the BER performance and system ca-
given , , and dB. In the case if pacity degraded dramatically.
HU AND CHEW: PERFORMANCE AND CAPACITY OF AN ASYNCHRONOUS SPACETIME BLOCK-CODED MC-CDMA SYSTEM 1337

the diversity combining schemes used. STBC MC-CDMA sys-

tems with two receive antennas will obtain 1.52 times higher
in system capacity.

APPENDIX A
Proposition 1: are positive i.i.d. RVs
with mean and variance . Let ; then, by
central limit theorem (CLT), for large , can be approximated
by a Gaussian RV with mean and variance . The pdf
of RV is given by

(67)
However, since is a positive RV, inaccuracy occurs by
assuming that follows a Gaussian distributed RV because
. Denote
Fig. 6. System capacity versus E b=N o dB (L = 32; K = 8; P = 1).
(68)

We define as the level of inaccuracy when applying CLT. This

means that a small amount of inaccuracy exists to positive RVs
when applying CLT. is a small value and, ideally, as
. represents the
Gaussian cumulative function. Conversely, to guarantee a given
level of accuracy , it is required that

(69)

Lemma 2: are positive i.i.d. RVs with

mean and variance . Let , for arbitrary .
Where , it can be shown that will always greater
than with a level of inaccuracy if

(70)

Proof: Since can be approximated as a Gaussian RV

Fig. 7. BER versus the number of users (E =N = 20 dB; L = 32; P = 1).
with mean and variance , then

Fig. 7 shows the BER performance versus the number of users

when dB, , and . It can be seen that
BER degrades gradually as increases when is small. For
large , BER has already been large, even for a small number
of users.
(71)
VI. CONCLUSION where is the level of inaccuracy when applying CLT, de-
The BER performance and system capacity of asynchronous fined in (68). By Proposition 1, is negligibly small when is
STBC MC-CDMA systems in the presence of CFO are theo- large. Conversely, given , we can lead to (70).
retically analyzed in this paper. A closed-form BER expression
is obtained for the EGC receiver and an expression to estimate APPENDIX B
the BER for MRC receiver is also derived. These expressions In this appendix, we will prove Lamma 1, stated in the main
are verified through simulations. The BER performance and text. Define an RV as
system capacity degrade significantly as the normalized CFO is
larger than 0.1, but remain nearly the same when the normalized
CFO is less than 0.1. Our results show that the performance and (72)
system capacity of STBC MC-CDMA systems are better than
that of MC-CDMA systems without ST coding, regardless of
1338 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

where , . In since . The vari-

the following, the objective is to verify that if both the ance of is given by
mean and variance of can be shown to be or is approximately
equal to zero under the two specific conditions given in (54a)
and (54b).
By CLT, for large , and can be approximated as
Gaussian RVs with mean , and variance , , respec-
tively. Since , are i.i.d. RVs, the mean
and variance of can be computed to be

(73)
(82)
(74)

Similarly, the mean and variance of can be easily found as

and can be shown that [24]
(75)
(76)

The covariance of and is given by

(77)
(83)
To obtain the mean and variance of RV , we first obtain its
pdf. Since and are Gaussian RVs, the joint pdf is given Substitute (83) in (82), we have
by

(84)
(78)
where

(85)
where , and
with represents its determinant. As
Substitute (73)(77) into (85); (85) can be rewritten as

(86)
(79)

We will now estimate an upper bound for (86). Again,

Since , then the pdf of is given by in (72) is approximately Gaussian distributed and our objec-
tive is to have greater than some positive number. Applying
Lemma 2 with and (conversely, ),
(80)
will always be greater than for a given level of
inaccuracy under the condition
The mean of RV is given by
(87)

Then, (86) can be evaluated by letting and ex-

pressed as

(88)

It can be verified that

(81) when (89)

HU AND CHEW: PERFORMANCE AND CAPACITY OF AN ASYNCHRONOUS SPACETIME BLOCK-CODED MC-CDMA SYSTEM 1339

Combining (87) and (89) will give us condition (54a). Substi-

tuting (89) into (84), we obtain

(90)

Substituting (73) to (77) in (90)

(91)

where is the desired bound of the standard deviation of that

we want to define. For a given , it is required that

(92)

which gives the condition (54b).

When is large, and . Hence, for large , the
RV has zero mean and zero variance or . This means that
(54) holds under conditions (54a) and (54b) for a predefined
and .
In particular, when , with the upper bound of the nor-
malized standard deviation and the level of inaccu-
racy of applying CLT is defined at , then the condi-
tions given in (54a) and (54b) can be computed as

(93a)
(93b)

APPENDIX C
Fig. 8. NSV versus the spreading gain L (Es=N o = 20 dB).
Denote the mean and variance for , which is de-
fined in (50) by and
, respectively, with denoted the
spreading gain. We define the normalized standard vari-
ance for the th data stream as and
. Fig. 8 shows the
plot on versus as: 1) the number of users and 2)
when and dB, for various values of
. It shows that decreases as increases. We purposely
choose and to overestimate the normalized stan-
dard variance of a practical system; in general, . We
then investigate the relationships: 1) between and for
a given and . For , the results we obtained are
almost the same as that of and the plot is not presented
here; 2) between and for a given , , and ;
Fig. 9 shows that there is a strictly monotonely increase as
is small or decrease as is large in ; 3) between
and for a given , , and . Fig. 10 shows that when
dB, attains a constant value and below 20
Fig. 9. NSV versus total number of users K (Es=No = 20 dB).
dB, decreases as dB decreases.
From all these observations, it can be concluded that the
values of shown in Fig. 8 give the worst case bound: bound as is large, for a practical range of and when
Fig. 8(a) gives the bound as is small and Fig. 8(b) gives the . We can use Fig. 8 to find whether the conditions (93a)
1340 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 53, NO. 5, SEPTEMBER 2004

[9] X. Cai and A. N. Akansu, Multicarrier CDMA systems with transmit

diversity, in Proc. IEEE Vehicular Technology Conf. (VTC00), vol. 6,
2000, pp. 28172821.
[10] Z. Yang, B. Lu, and X. Wang, Bayesian Monte Carlo multiuser re-
ceiver for space-time coded multicarrier CDMA systems, IEEE J. Se-
lect. Areas Commun., vol. 19, pp. 16251637, Aug. 2001.
[11] P. L. Kafle and A. B. Sesay, Iterative semi-blind space-time multiuser
detection for MC-CDMA uplink system, presented at the IEEE Vehic-
ular Technology Conf. (VTC02), 2002.
[12] X. Gui and T. S. Ng, Performance of asychrounous orthogonal mul-
ticarrier CDMA system in frequency selective fading channel, IEEE
Trans. Commun., vol. 47, pp. 10841091, July 1999.
[13] L. Tomba and W. A. Krzymien, Effect of carrier phase noise and
frequency offset on the performance of multicarrer CDMA systems,
in Proc. IEEE Int. Conf. Communications (ICC96), Dallas, TX, June
1996, pp. 15131517.
[14] J. Jang and K. B. Lee, Effects of frequency offset on MC-CDMA
system performance, IEEE Commun. Lett., vol. 3, pp. 196198, July
1999.
[15] D. Kivanc and H. Liu, Uplink performance of MC-CDMA in the
presence of frequency offset, in Proc. Vehicular Technology Conf.
Fig. 10. NSV versus Eb=N o. (VTC99), vol. 5, 1999, pp. 28552859.
[16] T. Kim et al., Performance of an MC-CDMA system with frequency
offsets in correlated fading, in Proc. IEEE Int. Conf. Communications
and (93b) will be satisfied so as to apply Lemma 1 in the main (ICC00), vol. 2, pp. 10951099.
[17] Q. Shi and M. Latva-aho, Effect of frequency offset on the performance
text. For example, for all values of , we need to of asynchronous MC-CDMA systems in a correlated Rayleigh fading
fulfill channel, in Proc. Info-Tech and Info-Net Int. Conf. (ICII01), vol. 2,
Nov. 2001, pp. 448452.
[18] F. Lin and L. B. Milstein, Successive interference cancellation in mul-
(94) ticarrier DS/CDMA, IEEE Trans. Commun., vol. 48, pp. 15301540,
Sept. 2000.
[19] J. P. Imhof, Computing the distribution of quadratic forms in normal
Hence, if we let , the two conditions in (93a) and variables, Biometrika, vol. 48, no. 3/4, pp. 419426, 1961.
(93b) are satisfied from the above discussion. Similarly, if we [20] J. G. Proakis, Digital Communications. New York: McGraw-Hill,
are interested in the range where , then for any value 2001.
[21] A. Papoulis and S. U. Pillai, Probability, Random Variables and Sto-
of will satisfy chastic Processes. New York: McGraw-Hill, 2002.
[22] M. K. Simon and M. Alouini, Digital Communication Over Fading
(95) ChannelsA Unified Approach to Performance Analysis. New York:
Wiley, 2000.
[23] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Func-
tions. New York: Dover, 1972.
[24] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Prod-
ACKNOWLEDGMENT ucts. New York: Academic, 1980.
The authors would like to thank the anonymous reviewers for
their constructive comments.
Xiaoyu Hu (S03) received the B.Eng. and M.Eng.
REFERENCES degrees from Huazhong University of Science and
Technology, Wuhan, China, in 1997 and 2000, re-
[1] S. Hara and R. Prasad, Overview of multi-carrier CDMA, IEEE spectively, and is working toward the Ph.D. degree
Commun. Mag., vol. 35, pp. 126133, Dec. 1997. at the Institute for Infocomm Research, the National
[2] A. J. Paulraj and C. B. Papadias, Spacetime processing for wireless University of Singapore, Singapore.
communications, IEEE Signal Processing Mag., vol. 14, pp. 4983, From July 2000 to June 2001, he was a Telecom-
Nov. 1997. munication Engineer with Guangdong Telecommu-
[3] G. J. Foschini and M. J. Gans, On limits of wireless communications nication Co., Ltd., Guangzhou, China. His research
in a fading enviroment when using multiple antennas, Wireless Pers. interests include multicarrier CDMA, spacetime
Commun., vol. 6, pp. 311335, Mar. 1998. coding, synchronization, channel estimation, mul-
[4] A. F. Naguib, N. Seshadri, and A. R. Calderbank, Increasing data rate tiuser detection, and spectrally efficient spread-spectrum system.
over wireless channels, IEEE Signal Processing Mag., vol. 17, pp. Mr. Hu has been awarded a research scholarship since July 2001.
7692, May 2000.
[5] V. Tarokh, N. Seshadri, and A. R. Calderbank, Space-time codes for
high data rate wireless communication: Performance criterion and code
construction, IEEE Trans. Inform. Theory, vol. 44, pp. 744765, Mar.
1998. Yong Huat Chew (S85M97) received the B.Eng.,
[6] S. M. Alamouti, A simple transmit diversity technique for wire- M.Eng., and Ph.D. degrees in electrical engineering
less communications, IEEE J. Select. Areas Commun., vol. 16, pp. from the National University of Singapore (NUS),
14511458, Oct. 1998. Singapore.
[7] V. Tarokh, H. Jafarkhani, and A. R. Calderbank, Spacetime block He has been with the Institute for Infocomm
codes from orthogonal designs, IEEE Trans. Inform. Theory, vol. 45, Research (formerly known as the Center for Wire-
pp. 14561467, July 1999. less Communications, NUS, and the Institute for
[8] X. Hu and Y. H. Chew, Performance of spacetime block coded Communications Research), an institute under the
MC-CDMA system over frequency selective fading channel using Agency for Science, Technology and Research,
semi-blind channel estimation technique, presented at the IEEE Wire- where he presently is a Lead Scientist. His research
less Communications and Networking (WCNC03), vol. 1, Mar. 2003, interests are in high spectrally efficient wireless
pp. 414419. communication systems and transmission over HFC and DSL.