www.ietdl.

org
Published in IET Communications
Received on 6th October 2009
Revised on 15th March 2010
doi: 10.1049/iet-com.2009.0642

ISSN 1751-8628

Performance of multiple-input and multipleoutput orthogonal frequency and code division

multiplexing systems in fading channels
P. Li W. Hamouda
Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, Canada H3G 1M8
E-mail: [email protected]

Abstract: In broadband downlink transmission, orthogonal frequency-division multiplexing (OFDM) combined

with code-division multiple access (CDMA) is a prospective technique for high-data rate transmission in future
wireless communication systems. By adding spatial diversity, multiple-input and multiple-output orthogonal
frequency and code division multiplexing (MIMO-OFCDM) offers superior performance relative to both
traditional OFDM systems and single-input and single-output OFCDM (SISO-OFCDM) systems. In this study, the
authors present an analytical study and investigation of a MIMO-OFCDM downlink system that hires
orthogonal variable spreading factor codes to spread each transmitted symbol in both time and frequency
domains. Different gain combining schemes are employed in the frequency domain to recover the data
symbols of the desired code channels, and space time block coding is used to achieve spatial diversity. The
more general Ricean fading channel is used to model the MIMO channel. The OFCDM system employs
Alamouti transmit diversity scheme with multiple receive antennas. For systems without multi-code
interference (MCI), analytical bit-error rate results are obtained and compared with simulation results. The
authors also investigate the effect of correlation in frequency domain, where we verify that minimum meansquare error frequency combining is more robust to MCI than equal-gain combining.

Introduction

Along with the fast growing demand of information exchange,

telecommunication systems are required to provide fast and
reliable services to high-data rate applications such as video
conferences, real-time broadcast and on-line games.
Optimising the use of available bandwidth is one of the
major challenges in wireless communications. Multipleaccess systems provide a solution for optimising frequency
use and maximising exibility. For instance, frequencydivision multiple access (FDMA), time-division multiple
access (TDMA), spread spectrum multiple access (SSMA)
and space-division multiple access (SDMA) are well
investigated and implemented in different communication
systems [1]. Also, orthogonal frequency-division
multiplexing (OFDM) is one useful multiplexing scheme
that has attracted many researchers [2 4]. It divides data
into a set of parallel streams to be transmitted using
IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11
doi: 10.1049/iet-com.2009.0642

mutually orthogonal sub-carriers. Hence the total channel

bandwidth is converted into multiple slow rate channels to
avoid the effects of intersymbol interference at the receiver
side. On the other hand, code-division multiple access
(CDMA) is one form of spread-spectrum signalling that is
broadly used in telecommunication systems [5, 6]. In a
CDMA system, users are assigned orthogonal codes so that
different users can share the same frequency at the same
time with little or no interference.
The integration of OFDM and CDMA takes advantage
of both systems. Typical implementations of CDMA
into OFDM-based systems include multi-carrier directsequence CDMA (MC-DS-CDMA), multi-carrier
CDMA (MC-CDMA). MC-DS-CDMA systems spread
symbols in time domain only [7, 8], so they cannot achieve
frequency domain diversity. On the other hand,
MC-CDMA systems spread symbols in frequency domain
1

& The Institution of Engineering and Technology 2011

www.ietdl.org
only [8, 9], and therefore cannot adapt to variable
transmission rates.
In a different approach for high-data rate transmission,
orthogonal frequency and code division multiplexing
(OFCDM), a combination of OFDM and CDMA can
provide better performance than both traditional direct
sequence CDMA (DS-CDMA) and OFDM schemes
[10]. In an OFCDM system, symbols are spread in both
frequency and time domain. Therefore OFCDM systems
can deliver frequency diversity gain and can be adapted
to applications with different transmission rates, which is
preferred in the 4G downlink transmission systems [8, 10, 11].
A single-input single-output OFCDM (SISO-OFCDM)
system with two-dimensional spreading in time and frequency
domain has been studied in [10, 12, 13]. In these works, the
authors focused on implementing hybrid detection technique
in OFCDM systems. Minimum mean-square error (MMSE)
combining technique is used to gain diversity in the frequency
domain. By comparing the different detection techniques in
[14], we notice that when no multi-code interference (MCI)
is present, maximal ratio combining (MRC), equal-gain
combining (EGC) and MMSE combining can achieve
the full frequency/spatial diversity gains. However for systems
with MCI, MMSE combining becomes superior to MRC
and EGC.
As the demand to provide reliable transmission for highdata rate applications increases, multiple-input multipleoutput (MIMO) techniques such as space time trellis
coding and space time block coding (STBC) are
introduced and widely studied [15 17]. To overcome the
problem of fading through diversity, MIMO techniques
implement multiple transmit antennas for the transmission
of the same signal through independent channels to
provide reliable wireless transmission. By adding spatial
diversity, a MIMO system can provide more reliable
transmission than a SISO system. In this paper, we explain
how to combine MIMO techniques, specically STBC,
with OFCDM systems. The space time scheme used in
our study is the Alamouti scheme [18], which employs
N 2 and M antennas at the transmitter and receiver
sides, respectively. We present the system structure,
transmission and detection methods of MIMO-OFCDM

systems. Also, we derive the bit-error rate (BER)

expression for such MIMO-OFCDM systems when
MRC, EGC and MMSE combining techniques are
employed. In our simulations, the more general Ricean
fading channel is used to model the MIMO channel.
In our study, the expressions for system BER are
considered under the condition that no MCI is present.
The accuracy of the BER expressions is veried when
compared with the simulated ones for MIMO-OFCDM
systems with different combining methods. These
comparisons are carried over different channels and with
different system parameters to explore the benets of
OFCDM-based systems. Both analytical and simulation
results show the large diversity gains achieved when
incorporating STBC with OFCDM. Also, we investigate
the effect of correlation in frequency domain where we
verify that MMSE frequency combining is robust to MCI.
The rest of the paper is organised as follows. In Section 2,
we introduce the transmitter structure and the transmission
scheme for the MIMO-OFCDM system. The combining
and detection methods are explained in Section 3. In
Section 4, we discuss the error probability for different gain
combining techniques. Then, we present comparison of
analytical results and simulation results in Section 5. In
Section 6, we draw our conclusions.

System description

Combined with STBC technique, SISO-OFCDM systems

introduced in [12] can be extended to MIMO-OFCDM
systems to gain spatial diversity. The transmitter structure
is shown in Fig. 1. The system considered here employs
N 2 transmit antennas and M receive antennas. At the
beginning of each two frame durations, 2KNB user data are
modulated into binary phase-shift keying (BPSK) symbols
for transmission, where K is the total number of code
channels actually transmitted, and NB is the group of data
symbols to be spread with the same spreading code and
sent to one transmit antenna. When quadrature phase-shift
keying or higher order modulation is used, Grey coding
should be employed. Among these modulated symbols,
KNB symbols are transmitted from the rst transmit

Figure 1 Transmitter structure of MIMO-OFCDM system

2

& The Institution of Engineering and Technology 2011

IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11

doi: 10.1049/iet-com.2009.0642

www.ietdl.org
antenna, and the remaining KNB symbols from the second
transmit antenna. Before transmission from the
corresponding antenna, the modulated symbols are then
converted from a serial symbol stream into NB parallel
symbol streams. The spreading is provided by a spreading
code generator, and each group of NB symbols will be
spread with a unique code. In the OFCDM system, let the
spreading factor in frequency domain and time domain be
NF and NT , respectively, then the total length of the
spreading code, NS , is given by
NS = NT NF

(1)

Hence, a maximum of NS NB symbols can be transmitted

at the same time when each group is assigned a unique
spreading code. Since K is the total number of code
channels, then K NS . The overall spreading code C is a
combination of the time-domain spreading code, C T , and
frequency-domain spreading code, C F . Here, both C T and
C F are orthogonal variable spreading factor (OVSF) codes.
The generation and some attributes of OVSF
codes can be found in [19]. Let C(k) represent the kth
(k = 0, 1, . . . , K 1) spreading code, C T (a) represent the
ath (a = 0, 1, . . . , NT 1) time-domain OVSF spreading
code of length NT , C F (b) represent the bth
(b = 0, 1, . . . , NF 1) frequency-domain OVSF spreading
code of length NF , denoted, respectively, as

c0 (k)
cNT (k)
...

c1 (k)
cNT +1 (k)
...

...

.
..
C(k) =

...
c(NF 1)NT (k) c(NF 1)NT +1 (k) . . .

cNT 1 (k)
c2NT 1 (k)

...
cNS 1 (k)
(2)

C T (a) = c0T (a)

c1T (a) . . .

cNTT 1 (a)



C F (b) = c0F (b) c1F (b) . . . cNF F 1 (b)

(3)
(4)

When the code C(k) (k = 0, 1, . . . , K 1) is generated by

C T (a) and C F (b), then
C(k) = [C F (b)]T C T (a)

(5)

where (.)T represents transpose operation.

In our study we focus on downlink transmission, where we
assume perfect synchronisation. Also, we assume that each
sub-carrier experiences at fading, so the orthogonality of
the OVSF code in time domain is preserved. However, the
orthogonality in frequency domain is distorted because of
different fading among different sub-carriers.
In an OFCDM system, multi-code interference arises when
different users data are spread with different frequency-domain
spreading codes and the same time-domain spreading code.
To overcome this, the spreading code should be designed in
such way to take advantage of the orthogonality in time
IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11
doi: 10.1049/iet-com.2009.0642

domain. Specically, the spreading code generator should

minimise the number of frequency-domain OVSF spreading
codes used so that the level of MCI can be minimised.
For example, the rst NT spreading codes are generated
by multiplying C F (0) with C T (0), C T (1), C T (2), . . . ,
C T (NT 1), respectively, the second NT spreading codes are
generated by multiplying C F (1) with C T (0), C T (1),
C T (2), . . . , C T (NT 1), respectively, and so forth. From
this method, one can see that when the code C(k) is generated
by C T (a) and C F (b), the relationship between k, a and b is
given by

a = k mod NT = k NT

k
b=
NT

k
NT


(6)


(7)

where the operator x represents the integer portion of x

with x x.
Given the spreading factor in frequency domain, NF , then
NF sub-carriers are dedicated to transmit the same data
symbol. In order to minimise the correlation among these
NF sub-carriers, one should ensure that the sub-carriers are
separated enough using frequency interleaving. The method
of interleaving is explained as follows. In a group of NB
symbols spread with the same spreading code, the rst
symbol is spread into the zeroth, NB th, (2NB )th, . . . and
[(NF 1)NB ]th sub-carriers, the second symbol is spread
into the rst, (NB + 1)th, (2NB + 1)th, . . . and [(NF
1)NB + 1]th sub-carriers and so forth. Finally, the NB th
symbol is spread into the (NB 1)th, (2NB 1)th, . . . ,
and (L 2 1)th sub-carriers, where L = NF NB is the total
number of sub-carriers employed in the OFCDM system.
To explain how the two-dimensional spreading is done,
let us take the spreading code assigned to the jth
( j = 1, 2, . . . , NB ) symbol in the kth (k = 0, 1, . . . , K 1)
group as an example. This assigned spreading code, noted as
C(k), is presented in Fig. 2. The rst NT chips of C(k),
{c0 (k), c1 (k), . . . , cNT 1 (k)} are assigned to the ( j 2 1)th
sub-carrier; the second NT chips of C(k), {cNT (k),
cNT +1 (k), . . . , c2NT 1 (k)} are assigned to the ( j 1 + NB )th
sub-carrier; and so forth. Finally, the last NT chips of C(k)
are assigned to the ( j 1 + (NF 1)NB )th sub-carrier.
As shown in Fig. 2, a signal transmitted during one time
slot on one sub-carrier in OFCDM system is called an
OFCDM symbol. After two-dimensional spreading and
frequency interleaving for a group of NB modulated
symbols, an OFCDM frame consisting of L NT
OFCDM symbols is constructed. Assigned with different
spreading codes, in total, K OFCDM frames are
multiplexed to build a superframe, and this step is done
using a frame multiplexer, shown in Fig. 3.
3

& The Institution of Engineering and Technology 2011

www.ietdl.org

Figure 2 Two-dimensional spreading and frequency interleaving

is the user data spread by C(k) on the lf th sub-carrier
transmitted from the nth antenna during the rst frame
duration and cd (k) is the dth chip of C(k). Here, d is the
corresponding chip index of the spreading code, given by

d=

Figure 3 Multiplexing K frames into a superframe

Let the indices l0 , l1 , . . . , lNF 1 represent the NF subcarriers carrying a user data spread by the kth spreading
code C(k) after frequency interleaving, and C(k) is given by
(5). Within the OFCDM superframe transmitted from the
nth (n 1, 2) transmit antenna, the transmitted signal on
the lf th ( f = 0, 1, . . . , NF 1) sub-carrier during the tth
(t = 0, 1, . . . , NT 1) OFCDM symbol duration is a
summation of K OFCDM symbols and is given by

1
Es K
d c (k)
Sn (lf , t) =
2NS k=0 n,k d

lf
NB


NT + t

(9)

Let S1 and S2 denote, respectively, the two superframes

transmitted from antenna 1 and 2. Then as shown in
Fig. 4, the OFCDM superframes S1 , S2 , S1 and S2 are
transmitted according to the Alamouti scheme through
OFDM transmission system. During T1 , the rst frame
duration, S1 is transmitted from the rst transmit antenna
and S2 from the second antenna. In the second frame
period, S1 and S2 are transmitted with the antenna order
reversed. Here we assume that the OFCDM symbols
transmitted on the same sub-carrier experience the same
fading within a pair of OFCDM superframes duration.
That is, fading is xed for the duration of two frames and
changes independently from one frame to another.

(8)

where Es is the transmitted signal energy for each OFCDM

symbol, normalised by the number of transmit antennas, dn,k

The receiver structure is shown in Fig. 5. After

experiencing channel fading and noise distortion, the signal
is picked up by M receive antennas. Let us use Rm,1 , Rm,2
to represent the received OFCDM superframes for the two

Figure 4 Alamouti scheme in MIMO-OFCDM system

4

& The Institution of Engineering and Technology 2011

IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11

doi: 10.1049/iet-com.2009.0642

www.ietdl.org

Figure 5 Receiver structure of MIMO-OFCDM system

consecutive frame duration at the receiver side shown in

Fig. 4. Rm,1 (lf , t) and Rm,2 (lf , t) represent the received
OFCDM signal at the mth (m = 1, 2, . . . , M) receive
antenna on the lf th sub-carrier during the jth OFCDM
symbol duration within the rst and second frame duration,
respectively. Therefore Rm,1 (lf , t) and Rm,2 (lf , t) are given by

1
Es K
Rm,1 (lf , t) =
(hm,1 d + hm,2
lf d2,k )cd (k) + hm,1 (lf , t)
2NS k=0 lf 1,k

and 
Y 2 (lf , t) are given by

Y 1 (lf , t) =

m=1

(11)
where hm,n
(m = 1, 2, . . . , M, n 1, 2) is the fading
lf
coefcient on the lf th sub-carrier from the nth (n 1, 2)
transmit antenna to the mth receive antenna. hm,1 (lf , t) and
hm,2 (lf , t) are the noise terms on the lf th sub-carrier during
the t th OFCDM symbol duration at the mth receive
antenna. hm,1 (lf , t) and hm,2 (lf , t) are additive white
Gaussian noise (AWGN) samples with zero mean and
variance No /2 per dimension.


Y 2 (lf , t) =

M

m=1

(12)

m,1
[(hm,2
lf ) Rm,1 (lf , t) hlf Rm,2 (lf , t)]


Es eq
eq
=
h d c (k) + hm,2 (lf , t)
2NS lf 2,k d

(13)

eq

where hlf is the overall equivalent channel on the lf th

sub-carrier, dened as
eq

h lf =

M

m=1

m,2 2
2
(|hm,1
lf | + |hlf | )

(14)

and the equivalent noise after combining becomes

eq

h1 (lf , t) =

M

m=1

eq

h2 (lf , t) =

m,2
[(hm,1
lf ) Rm,1 (lf , t) + hlf Rm,2 (lf , t)]


Es eq
eq
h d c (k) + hm,1 (lf , t)
=
2NS lf 1,k d

(10)

1
Es K
m,2

(hm,1
Rm,2 (lf , t) =
lf d2,k + hlf d1,k )cd (k) + hm,2 (lf , t)
2NS k=0

M


M

m=1

Detection algorithm
eq

m,2
[(hm,1
lf ) hm,1 (lf , t) + hlf hm,2 (lf , t)]

(15)

m,1
[(hm,2
lf ) hm,1 (lf , t) hlf hm,2 (lf , t)]

(16)

eq

In this section, we propose the detection algorithm to recover

the transmitted data symbols. The algorithm extracts receive
diversity based on the optimal decision for the Alamouti
scheme with multiple receive antennas.

h1 (lf , t) and h2 (lf , t) are both Gaussian with zero mean

and variance

When the transmitted signals are picked up by receive

antennas, they are rst combined to decouple the data
Y 2 (lf , t) be the transformed
symbols. Let 
Y 1 (lf , t) and 
received signals after combining at the lf th sub-carrier
within the tth OFCDM symbol duration. Then 
Y 1 (lf , t)

where we considered the variance of each hm,1 (lf , t) and

hm,2 (lf , t) to be 1/2 per dimension.

IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11

doi: 10.1049/iet-com.2009.0642

eq

eq

eq

E{|h1 (lf , t)|2 } = E{|h2 (lf , t)|2 } = hlf

(17)

Note that the transmitted symbol pair have the same

spreading code. As explained before, the orthogonality in
5

& The Institution of Engineering and Technology 2011

www.ietdl.org
time domain is preserved, and therefore time-domain
despreading is carried out before frequency-domain
despreading. Now, the received signal transmitted from the
nth transmit antenna during T1 on the lf th sub-carrier
using the ath time-domain spreading code, and after timedomain despreading is given by
T
1 
[
Y n (lf , t) ctT (a)]
NT t=0

Es eq
 (b) + heq
h d cF
=
n (lf )
2NS lf n,k lf /NB

N 1

f =0

rn (lf )cF

lf /NB

 (b)weq (lf )

(19)

where weq (lf ) is the equivalent combining weight of the lf th

sub-carrier. For some commonly used gain combining
techniques, weq (lf ) is given by
weq (lf ) = (hlf ) = hlf
eq

eq


d n,k = argmin |zn,k s|2

(20)

(25)

{s}

where {s} is the set of all possible transmitted data symbols.

Performance analysis

In this section, we study the performance of MIMOOFCDM systems without MCI. In that, we obtain a
semi-analytical expression for the BER considering BPSK
transmission. The average BER of the system will then be
evaluated using Monte-Carlo approach.
Let en represent the error vector corresponding to the data
symbols sent to the nth transmit antenna, then en is given by
en = zn,k dn,k
NF 1
F
eq
f =0 rn (lf )clf /NB (b)w (lf )
=  N 1 eq
dn,k
F
Es /2NS f =0
hlf weq (lf )
NF 1

eq

hn (lf )weq (lf )

=  N 1 eq
F
Es /2NS f =0
hlf weq (lf )
f =0

for MRC [20]

(hlf )

(24)

(18)

Since the same data symbol is transmitted on the sub-carriers

with indices l0 , l1 , . . . , lNF 1 , the outputs of time-domain
despreader over these NF sub-carriers are then combined,
which is also called frequency-domain despreading. Here, a
suitable combining method is employed to extract
frequency-domain diversity over these NF sub-carriers of
interest, and the output of the frequency-domain
despreader can then be written as
N
F 1

yn,k
1
zn,k =  N 1 eq
F
Es /2NS f =0 hl weq (lf )

Finally, a hard decision based on minimum distance criteria is

made to recover the corresponding user data

rn (lf ) =

yn,k =

This output is normalised to form the decision variable,

which will be used in MCI cancellation if needed. The
decision variable is then given by

(26)

eq

eq

w (lf ) =

eq
|hlf |

=1

(21)

for EGC [20, 21] and

(hlf )
eq

weq (lf ) =

(1 +

eq
Kc )(hlf )2

+ ([Es /2NF N0 ])

(22)

for MMSE combining [14] where here we generalised the

optimisation function dened for the SISO-OFCDM
system in [12] to a MIMO-OFCDM system, with Kc
being the number of interfering codes, given by the
number of users transmitted on the same sub-carriers and
using the same time-domain spreading code but different
frequency-domain spreading codes. In a MIMO-OFCDM
system with no MCI, the combining weight of the MMSE
becomes
weq (lf ) =

eq
(hlf )

(hlf )2 + ([Es /2NF N0 ])1

eq

& The Institution of Engineering and Technology 2011

(23)

eq

where hlf is dened in (14). Conditioned on the channel,

NF 1 eq
eq
f =0 hn (lf )w (lf ) is also Gaussian with zero mean and
variance
NF 1
1 
eq
2
s
=
hlf |weq (lf )|2
h
NT

(27)

f =0

Let r be the ratio of the total transmitted signal power to the

background noise power. Since the additive noise terms
hm,1 (lf , t) and hm,2 (lf , t) are complex AWGN with
variance 0.5 per dimension, then r = ES , and the error
vector, en , becomes
NF 1

eq
eq
f =0 h n (lf )w (lf )
en =  N 1 eq
F
r/2NS f =0
hlf weq (lf )

(28)

where h eq
n (lf ) is the equivalent noise of the lf th sub-carrier at
the receiver side during Tn (n 1, 2). Since the additive
noise terms hm,1 (lf , t) and hm,2 (lf , t) are complex AWGN
with 0.5 per dimension, it is easy to show that the
IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11
doi: 10.1049/iet-com.2009.0642

www.ietdl.org
eq

equivalent noise, hn (lf ), is complex Gaussian noise with zero


m,1 2
m,2 2
mean and variance (1/2NT ) M
m=1 (|hlf | + |hlf | ) =

eq
hlf /2NT per
 NF 1 eq
f =0 h n (lf

Then, the error vector en in (26) can be simplied to

NF 1

eq

)w (lf ) is also Gaussian with zero mean and

variance

eq

hn (lf )
en =  N 1 eq
F
(r/2NS ) f =0
h lf

(35)

NF 1
1 
eq
2
s
=
h lf
h
NT

(36)

f =0

dimension. Conditioned on the channel,

where
NF 1
1 
eq
2
s
=
hlf |weq (lf )|2
h
NT

(29)

f =0

f =0

4.1 MRC
eq

With MRC, since the equivalent channel coefcient hlf is a

real number
w(lf ) = (hlf ) = hlf
eq

eq

Therefore the probability of bit error conditioned on the

channel is given by
Pe|h = P(Re(en ) . 1)

NF 1 eq
h
(l
)
n f
f =0
= P Re N 1 eq . 1
F
(r/2NS ) f =0
h lf


 NF 1
 r  eq
h
= Q
NF f =0 lf

(30)

The error vector en becomes

NF 1 eq
eq
eq
hn (lf )weq (lf )
f =0 hn (lf )hlf
=  N 1 eq 2
en =  N 1 eq
F
F
(r/2NS ) f =0
(r/2NS ) f =0
hlf weq (lf )
(hlf )
NF 1
f =0

(37)

(31)

4.3 MMSE combining

and
1
2
s
=
h
NT

N
F 1
f =0

eq

hlf |w(lf )|2 =

1
NT

N
F 1
f =0

When the MMSE is used for frequency combining

eq

(hlf )3

NF 1

hn (lf )(hlf /(hlf )2 + (r/2NF )1 )

(38)
en =  N 1 eq 2 eq 2
F
(r/2NS ) f =0
[(hlf ) /(hlf ) + (r/2NF )1 ]

(32)

When BPSK modulation is used, the probability of bit error

conditioned on the channel is given by
BER|h = P(Re(en ) . 1)

NF 1 eq
eq
h
(l
)w
(l
)
f
n f
f =0
. 1
= P Re N 1 eq
F
(
r /2NS ) f =0
hlf weq (lf )

 N
N
F 1
F 1
r
eq
eq
eq
eq
hn (lf )hlf .
h lf h lf
= P Re
2N
S
f =0
f =0


NF 1 eq 2
(r/2NS ) f =0 (hlf )
= Q
NF 1 eq 3
(hlf )
(1/2NT ) f =0


NF 1 eq 2
r
/N
)
(h
)
(
F
l
f
=0
f


= Q

NF 1 eq 3
(h
)
lf
f =0

(33)

eq

f =0

NF 1
(hlf )3
1 
2
1 2
NT f =0 [(heq
lf ) + (r/2NF ) ]

w(lf ) =

=1

IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11

doi: 10.1049/iet-com.2009.0642

(34)

(39)

and
Pe|h = P(Re(en ) . 1)
N 1
F

eq
eq
eq
[hn (lf )hlf /(hlf )2 + (r/2NF )1 ]
= P  N 1 eq 2 eq 2
. 1
F
[(hlf ) /(hlf ) + (r/2NF )1 ]
(r/2NS ) f =0

 NF 1
2
[1/(1
+
D)]
f =0

= Q
(40)

NF 1 eq
1
[h
/(D
+
2
+
D)]
lf
f =0
f =0

where

4.2 EGC

eq
(hlf )
eq
|hlf |

eq

eq

2
s
=
h

D=
With EGC, the combining weight on the lf th equivalent
channel is given by

eq

r
2NF

!1

(hlf )2
eq

(41)

Simulation results

Here we assess the performance of the MIMO-OFCDM

system described earlier over Ricean fading channels, where
both simulation and analytical results are presented.
7

& The Institution of Engineering and Technology 2011

www.ietdl.org

Figure 6 BER for MIMO-OFCDM system using MRC over Rayleigh fading channel: NF 8, NT 8, K 8
Channel state information (CSI) is assumed to be perfectly
known at the receiver side, and CSI is xed for the
duration of two consecutive OFCDM frames. Different
system parameters are used when investigating the
performance with MRC, EGC and MMSE combining.
The effect of sub-carrier correlation because of imperfect
frequency interleaving is considered.

Fig. 6 presents the performance of MIMO-OFCDM systems

with different number of receive antennas over Rayleigh fading
channels using MRC. The performance of a SISO-OFCDM
system introduced in [10] with NF = 8 is shown as a
reference. From this gure we can see that the slope of the
curves increases as the spatial diversity of the system increases.
The MIMO-OFCDM systems with 1, 2 and 3 receive

Figure 7 Effect of MCI on the BER for MIMO-OFCDM system over Rayleigh fading channel: M 3, NF 16, NT 8, K 32
8

& The Institution of Engineering and Technology 2011

IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11

doi: 10.1049/iet-com.2009.0642

www.ietdl.org

Figure 8 BER for MIMO-OFCDM system using EGC over Ricean fading channel: M 3, NF 8, NT 8, K 8
antennas achieve overall diversity orders of 16, 32 and 48
respectively, whereas the SISO-OFCDM system achieves a
diversity order of 8. At the same time, we can see the accuracy
of our analytical results when compared with the simulated ones.
In Fig. 7 we compare MRC, EGC and the MMSE-based
gain combining used in MIMO-OFCDM systems over
Rayleigh fading channels when MCI is present in the

system. The considered system is running at 25% of full load

and NF = 16. That is, every four spreading codes are
interfering with each other, and hence Kc equals 3. From
these results, and compared to the performance of the MRC
and EGC, the MMSE combiner is more robust to MCI.
We present the BER performance of two MIMOOFCDM systems over Ricean fading channels in Figs. 8

Figure 9 BER for MIMO-OFCDM system using MMSE combining over Ricean fading channel: M 1, NF 8, NT 8, K 8
IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11
doi: 10.1049/iet-com.2009.0642

& The Institution of Engineering and Technology 2011

www.ietdl.org

Figure 10 BER for MIMO-OFCDM system using EGC over Rayleigh fading channel: M 2, NF 8, NT 8, K 8
and 9, where three receive antennas with EGC and one
receive antenna with MMSE combining are used in the
two systems, respectively. The parameter, k, known as the
Ricean factor, represents the ratio of the dominant
component (line-of-sight) to the total power of scattered
waves. From these two gures we see that as k increases,
the system performance changes from a Rayleigh fading
channel towards a Gaussian channel.
In Fig. 10 we study the performance degradation caused by
correlation among sub-carriers of interest for both SISO and
MIMO-OFCDM systems when employing EGC. The
correlated Rayleigh fading channels with given correlation
coefcient are generated according to the scheme
introduced in [22]. From the gure we can see that for
SISO and MIMO-OFCDM systems, the impact on the
overall diversity order is the same when sub-carriers of
interest (i.e. assigned to the same user data symbol) are
correlated. This is because the correlation is only present
among sub-carriers from the same transmit antenna to the
same receive antenna. From this example we can see that
BER of MIMO-OFCDM systems can be easily obtained
once the fading channel model is dened.

Conclusion

We presented the transmission and detection methods for

MIMO-OFCDM downlink transmission in Ricean fading
channels. It was shown that by employing STBC, MIMOOFCDM systems provide signicant performance
improvement relative to SISO-OFCDM. Semi-analytical
expressions for the BER of MRC, EGC and MMSE
combining techniques have been obtained. It was shown
10

& The Institution of Engineering and Technology 2011

that in the absence of multi-code interference, MRC,

EGC and MMSE combining achieve the full frequency/
spatial diversity gain. However, when the system suffers
from multi-code interference, only the MMSE combiner
can be more tolerant to interference with much higher gain
than both MRC and EGC. Furthermore, we have shown
that when frequency correlation exists among sub-carriers,
both the SISO and MIMO-OFCDM systems experience
same degree of diversity loss.

References

[1] STUBER G.L.: Principles of mobile communication

(Kluwer Academic Publishers, 2001)
[2] COUCH L.W. II: Digital and analog communication
systems (Pearson, 2007)
[3] ZHANG Z. , WU J. : Study and performance simulation
analysis of OFDM systems. IET Int. Conf. on Wireless,
Mobile and Multimedia Networks, 2006
[4] SIVADASAN J.: Effectiveness of orthogonal frequency
division
multiplexing
technique
for
wireless
telecommunication systems. IET-UK Int. Conf. on
Information and Communication Technology in Electrical
Sciences (ICTES 2007), Chennai, India, 2007
[5] TANNER R., WOODARD J. : WCDMA requirements and
practical design (Wiley, 2004)
[6] OFARRELL T. , BEALE M.: Code-division multiple-access
(CDMA) techniques in optical bre LANs. Second IEE
IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11
doi: 10.1049/iet-com.2009.0642

www.ietdl.org
National Conf. on Telecommunications, University of York,
England, 2 5 April 1989, pp. 111 115

Globcom Conf. 1995, Singapore, November 1995, vol. 3,

pp. 2059 2063

[7] BARBANCHO I., TARDON L.J., BARBANCHO A.M.: Efcient MC-DSCDMA transmitter for ICI self-mitigation, Electron. Lett.,
2006, 42, (9), pp. 541 542

[15] TAROKH V., SESHADRI N., CALDERBANK A.R.: Space-time codes

for high data rate wireless communication: performance
criterion and code construction, IEEE Trans. Inf. Theor.,
1998, 44, (2), pp. 744 765

[8] ZHOU Y., WANG J., SAWAHASHI M.: Downlink transmission of

broadband OFCDM systems Part II: effect of doppler
shift, IEEE Trans. Commun., 2006, 54, (6), pp. 1097 1108
[9] MCCORMICK A.C. , AL-SUSA E.A. : Multicarrier CDMA for
future generation mobile communication, Electron.
Commun. Eng. J., 2002, 14, (2), pp. 52 60
[10] WANG J.: OFCDM a promising wireless multiple
access technique for 4G mobile communications. IET
Seminar on Wideband and Ultrawideband Systems
and Technologies: Evaluating Current Research and
Development, 6 November 2008, pp. 1 10
[11] BOUMARD S., WEISSENFELT M., CHI H., NURMI J.: A wireless
MIMO STC OFDM system implementation. IEEE 17th Int.
Symp. on Personal, Indoor and Mobile Radio Commun.,
Helsinki, Finland, September 2006, pp. 1 5
[12] ZHOU Y., WANG J., SAWAHASHI M.: Downlink transmission of
broadband OFCDM system Part I: hybrid detection, IEEE
Trans. Commun., 2005, 53, (4), pp. 718 729
[13] ZHOU Y., WANG J., SAWAHASHI M. : Turbo-coded OFCDM
systems for 4G mobile communications. Sixth IEE Int.
Conf. on 3G and Beyond, London, UK, 7 9 November
2005, pp. 1 5
[14] KAISER S.: On the performance of different detection
technique for OFDM-CDMA in fading channels. IEEE

IET Commun., 2011, Vol. 5, Iss. 1, pp. 1 11

doi: 10.1049/iet-com.2009.0642

[16] TAROKH V., JAFARKHANI H., CALDERBANK A.R.: Space time block
coding for wireless communications: performance results,
IEEE J. Select. Areas Commun., 1999, 17, (3), pp. 451 460
[17] BASAR E., AYGOLU U.: High-rate full-diversity space-time
block codes for three and four transmit antennas, IET
Commun., 2009, 3, (8), pp. 1371 1378
[18] ALAMOUTI S.M.: A simple transmit diversity technique for
wireless communications, IEEE J. Select. Areas Commun.,
1998, 16, (8), pp. 1451 1458
[19] MAEDA N., KISHIYAMA Y., ATARASHI H., SAWASHASHI M.: Variable
spreading factor-OFCDM with two dimensional spreading
that prioritizes time domain spreading for forward link
broadband wireless access. IEEE Veh. Techn. Conf., Jeju,
Korea, April 2003, vol. 1, pp. 127 132
[20] PROAKIS J.G.: Digital communications (McGraw-Hill,
2000, 4th edn.)
[21] KARAGIANNIDIS G., ZOGAS D.A., KOTSOPOULOS S.A.: Statistical
properties of the EGC output SNR over correlated
Nakagame-m fading channels, IEEE Trans. Wirel.
Commun., 2004, 3, (5), pp. 1764 1769
[22] ERTEL R.B., REED J.H.: Generation of two equal power
correlated Rayleigh fading envelopes, IEEE Commun. Lett.,
1998, 2, (10), pp. 276 278

11

& The Institution of Engineering and Technology 2011