Vous êtes sur la page 1sur 5

A Novel OSSMIC Receiver for Downlink

MIMO MC-CDMA Systems


Antonis Phasouliotis and Daniel K.C. So
School of Electrical & Electronic Engineering
The University of Manchester
Manchester, UK
email: antonis.phasouliotis@postgrad.manchester.ac.uk, d.so@manchester.ac.uk
Abstract In this paper, a chip level ordered successive spatial
and multiuser interference cancellation (OSSMIC) receiver
architecture is presented for downlink multiple-input multiple-
output multi-carrier code division multiple access (MIMO MC-
CDMA) systems. The proposed receiver performs ordered layer
space-time detection. Unlike the existing ordered successive
interference cancellation (OSIC) receiver, the proposed receiver
cancels both spatial & multiuser interference in the SIC process.
This allows detection in multiuser scenario, which is not possible
in conventional chip level OSIC receiver. Simulation results show
that the proposed receiver significantly outperforms both the
existing chip level OSIC detector and the chip level linear
receiver. As the proposed receiver only requires the knowledge of
the desired users spreading sequence, it is a viable solution for
downlink MIMO MC-CDMA communications.
I. INTRODUCTION
The introduction of new wireless communication services
including mobile internet, video telephony, and high quality
multimedia streaming increases the demand for more
bandwidth efficient wireless communication systems that can
achieve higher user capacities and faster data rates from
already existing technologies. The research towards a new
wireless platform, called fourth generation (4G), has been
initiated worldwide. One of the possible solutions for 4G is the
multiple input multiple output multi-carrier code division
multiple access (MIMO MC-CDMA) system [1]. This refers to
the combination of MIMO techniques [2] with MC-CDMA
system [3], the latter being the combined scheme of orthogonal
frequency division multiplexing (OFDM) with code division
multiple access (CDMA).
OFDM is a multicarrier modulation technique employed
for high data rate wireless applications and is suitable for
combating inter-symbol interference (ISI) that arises in
frequency selective channels. CDMA is a multiplexing
technique which permits multiple users to access the wireless
channel simultaneously by modulating and spreading their
input data signals with different spreading sequences.
In addition, MIMO system, a scheme based on multiple
transmitting and multiple receiving antennas can achieve very
high data rates in rich multipath scattering environment without
increasing the transmission bandwidth or the total transmitted
power of the system. The vertical Bell labs layered space-time
(V-BLAST) architecture (or ordered successive interference
cancellation (OSIC)) [4] is a MIMO system that has recently
gained major research attention because it can realize very high
data transfer rates by using spatial multiplexing to exploit the
multiple spatial channels. For MIMO MC-CDMA system, a
chip level OSIC receiver has been proposed and demonstrated
good performance for single user case [5]. However, its
performance is severely degraded in multiuser cases and has a
high error floor appearing in medium SNR range due to
multiple access interference (MAI) and error propagation.
Therefore, it performs even worse than the linear zero forcing
(ZF) or minimum mean square error (MMSE) receivers. In [6],
chip level and symbol level MMSE linear receivers are
presented. Although the symbol level detector has better
performance over the chip level detector, it needs high
computational complexity and also requires the knowledge of
all other users spreading sequence. A partial MMSE-OSIC
receiver based on multiuser detection [7] achieves similar
performance to the symbol level MMSE-OSIC, but also
requires all other users spreading sequence. Without a
practical receiver architecture, downlink MIMO MC-CDMA
system is not as favourable as other simpler transceiver
schemes such as MIMO OFDMA.
In this paper, a more practical receiver namely the chip
level ordered successive spatial and multiuser interference
cancellation (OSSMIC) receiver is proposed for downlink
MIMO MC-CDMA systems. This receiver performs layered
space-time processing with both spatial and multiuser
interference cancellation. It outperforms linear receivers [6]
and unlike [5], it is capable for multiuser scenarios. Moreover,
the receiver does not require the knowledge of other users
spreading sequences and is suitable for downlink
communications. It should also be noted that the proposed
scheme invalidates the perception that linear MMSE detector
performs better than iterative detectors for MIMO MC-CDMA
systems [7, 8, 9].
The rest of this paper is organized as follows. Section II
introduces the system model by presenting both the transmitted
and the received signal. In Section III, the novel chip level
OSSMIC receiver design for the downlink MIMO MC-CDMA
is proposed. In Section IV, the performance of the proposed
architecture is evaluated through simulations and is compared
with the performance of the linear detectors and the chip level
OSIC detector. Finally, section V concludes and summarizes
the paper.
978-1-4244-1645-5/08/$25.00 2008 IEEE 1271
Authorized licensed use limited to: Chulalongkorn University. Downloaded on June 11, 2009 at 22:39 from IEEE Xplore. Restrictions apply.
II. MIMOMC-CDMASIGNAL MODEL
A. Transmitted Signal Model
Consider a downlink MIMO MC-CDMA transmitter
model with N transmitting antennas for K users. The QPSK
symbols are grouped into N substreams of P symbols each, and
thus the symbol matrix for user k (k = 1,2,,K) can be
represented as

1 2 N
k k k k
=

D d d d
N P
C
where the row vector d
k
n
denotes the datastream of P symbols
that will be transmitted by the n-th antenna (n = 1,2,,N), and
is defined as
T
,1 , 2 ,
n n n n
k k k k P
d d d =

d
1 P
C .
The symbols of each user are first serial-to-parallel converted
to form the N substreams, and then spread with the
corresponding user specific spreading sequence in order to
form the chip-level transmit matrix
,1 , 2 , k k k k PG k k
= =

S s s s D c
N PG
C
where
,1 , 2 , k k k k G
c c c =

c
1 G
C

is the signature sequence of user k, C refers to the chip
alphabet, denotes the Kronecker product and G is the length
of the spreading sequence. Each user is allocated a distinct
spreading code and in order to ensure orthogonality between
the different users, Walsh-Hadamard sequences are used. The
spread signals of all users from each transmitting antenna are
then combined and all the P x G parallel data sequences are
interleaved in the frequency domain with a chip level block
interleaver of size P x G. The interleaved data are next mapped
onto N
c
= P x G available subcarriers (f = f
1
, f
2
,,
c
N
f ) and
transformed into the time domain by the inverse fast Fourier
transform (IFFT). The N
c
subcarriers can be represented by
index i and related to the p-th symbol (p = 1,2,,P) and the g-
th chip (g = 1,2,,G) by i = (p 1)G + g. The transmitted
chips of all users at the i-th subcarrier can be indicated as

T
1 2
,
1
K
N
i i i i k i
k
x x x
=
= =


x s
1 N
C (1)
where x
i
n
is the chip transmitted by the n-th antenna at the i-th
subcarrier, and can be expressed as

, , , /
1 1
K K
n n n
i k i k i k i G
k k
x s c d
= =


= =

(2)
in which
,
n
k i
s indicates the chip transmitted by the n-th antenna
for the k-th user at the i-th subcarrier, and a

denotes the
largest integer that is smaller than a. The output signal from the
IFFT is added with cyclic prefix (CP) before transmission over
the wireless multipath channel. The channel is considered to be
quasi-static frequency selective fading corrupted by additive
white Gaussian noise (AWGN). In order to avoid inter symbol
interference (ISI), the duration of CP is assumed to be longer
than the maximum delay spread of the channel.
B. Received Signal Model
The receiver of the desired user consists of M receiving
antennas. Upon receiving the signal, the cyclic prefix (CP) is
removed and the fast Fourier transform (FFT) of size N
c
is
performed. The received signal model at the i-th subcarrier,
after the FFT operation can be characterized as

i i i i
= + r H x n (3)
where the received signal, the channel and the noise vectors are
represented respectively as
T
1 2 M
i i i i
r r r =

r
1 M
C
(1,1) (1, )
1
( ,1) ( , )
, ,
N
i i
N
i i i
M M N
i i
h h
h h
= =







H h h

. . .

M N
C
T
1 2 M
i i i i
n n n =

n
1 M
C .
The channel response at the i-th subcarrier between transmit
antenna n (n = 1,2,,N) and receive antenna m (m = 1,2,,M)
is denoted by h
i
(m,n)
and the N x 1 AWGN noise vector at the i-
th subcarrier is signified by n
i
. The received signal expression
in (1) can be rearranged as

' ' ' '
, , ', ',
' ' ' '
CAI MAI 1 MAI 2
n n n n n n n n
i i k i i k i i k i i k i i
n n k k k k n
s s s s

= + + + +

r h h h h n

. (4)
Considering the right hand side of (4), the first term represents
the transmitted chips from the desired substream n of the
desired user k. The second term corresponds to the co-antenna
interference (CAI) arising from the other substreams of the
desired user. Multiple access interference (MAI) coming from
other users n-th substream and all other substreams are
expressed by the third term (MAI 1) and the fourth term (MAI
2) respectively.
III. OSSMICRECEIVER
A. Successive Interference Cancellation procedure
The chip level OSIC receiver proposed in [5] for multiuser
MIMO MC-CDMA systems has a poor performance. This is
because, after the chip level linear filter that removes CAI &
MAI 2, chip detection is made before despreading. For this
reason, the MAI 1 in (4) is not suppressed before detection and
thus causes severe degradation. Instead, we propose a novel
OSSMIC receiver design that performs chip level spatial
nulling, optimum detection ordering of the received
substreams, symbol detection after despreading, and chip level
spatial and multiuser interference cancellation. The block
diagram of the novel receiver without interleaving for the
desired user k is depicted in Figure 1. This could easily be
extended to the interleaved case. The strongest received
substream is considered initially to be the desired signal for
detection while the remaining substreams are suppressed. The
layered detection process is performed using an optimum
selection ordering which is discussed in the next section.
1272
Authorized licensed use limited to: Chulalongkorn University. Downloaded on June 11, 2009 at 22:39 from IEEE Xplore. Restrictions apply.
Without loss of generality, the first substream is considered.
For convenience the analysis of ZF filtering is discussed first
and it can be easily extended to the MMSE case. Some issues
for the MMSE case will be explained later in this section.
Hence, considering the received signal passing through a linear
ZF filter, the estimates of the transmitted chips for the first
substream at the i-th subcarrier can be expressed as
1
,
1 1 1
k i i i i i i i i
y
+ + +
= = +

H r H H x H n

1 1
, ',
'
k i k i i
k k
s s n

= + +

(5)
where H
i
+
is the pseudoinverse matrix of H
i
and [A]
1
denotes
the first row of matrix A. The symbol decision statistic is
obtained when the chip estimates in (5) are despread by the
desired users spreading sequence.
1 1 1 1
, , , , , , ', ,
1 1 ' 1 1
G G G G
k p k g k i k g k i k g k i k g i
g g k k g g
z c y c s c s c n
= = = =
= = + +



1
, ,
1
G
k p k g i
g
d c n
=
= +

(6)
where i is related to p and g as shown in section II A., and this
is used for all subsequent expressions. The p-th symbol
detection for the first substream of the k-th user is then
performed to obtain
1
,

k p
d . The next step in a layered space-time
receiver is to remove the detected substream from the received
signal. The modified channel matrix is then formed by setting
the detected substreams channel response to zero, i.e. the first
column of H
i
is zeroed [4]. This modified channel matrix is
subsequently used to compute the nulling filter for the next
stage. The channel zeroing process is important as it increases
the post detection SNR for the following detection stages.
However, if the multiuser interference coming from the first
transmit antenna is not completely removed, it will not be
suppressed by any subsequent nulling processes. This will
significantly deteriorate the performance. With the described
cancellation procedure, the interference caused by the
substream of the other users, i.e. MAI 1 in (4), is not removed
and it causes poor system performance. Therefore, the MAI 1
in the detected substream must be eliminated.
The most straight forward method to determine this MAI
term is to detect the first substream for all users. However, this
requires the knowledge of all other users spreading sequence
and will increase the computational complexity. Alternatively,
this MAI term could be computed for every i-th subcarrier by
spreading the symbol decision statistic z
1
k,p
with the desired
users spreading sequence to produce
T
1
, , ( 1) 1 , ( 1) 2 , , k p k p G k p G k pG k k p
v v v z
+ +
= =

v c . (7)
This respread signal v
k,p
contains the desired user signal and
noise. It is then subtracted from the chip estimate
1
, k i
y in (5) to
obtain the signal contribution in MAI 1 term and a second term
which enhances the noise

1 1 T
, , , ', , ,
' 1
G
k i k i k i k i k g k g i
k k g
I y v s c c n
=
= = +

. (8)
The whole cancellation procedure for the first substream from
the received signal at the i-th subcarrier can be characterized by

T
1 T
, , ,
1

( )
i i k g k p k i i
c d I

= +

r r H (9)
where
i

r is the modified received signal containing the CAI,


MAI 2 and enhanced noise terms for the next detection layer.
The above stated steps are repeated until all the remaining
substreams are detected.
It must be noted that when linear MMSE filter is used, the
interference calculated in this case, i.e. I
k,i
, also contains some
form of residual CAI and MAI. This happens because when the
received signal is filtered using the MMSE, the CAI & MAI 2
in (4) are not completely removed. Moreover, as the MMSE is
not an inversion process as in ZF case, the MAI for the n-th
substream after filtering will not have a normalized magnitude
as in (8). Hence, MAI 1 cannot be completely removed by the
despreading process. Nevertheless, as the MMSE filter
minimizes the mean-square error between the original and the
estimated data, its performance will still be superior to the ZF
OSSMIC receiver as shown in section IV.
B. Detection Ordering
The optimum detection order for the novel OSSMIC
receiver is determined according to the substream which
produces the smallest mean square error. The MSE matrix for
the MMSE receiver with respect to the i-th subcarrier is given
by

1
0
diag( )
H
i i xx i N
N

= + J H R H I (10)
where diag(A) denotes the diagonal elements of matrix A,
superscript H indicates matrix Hermitian, I
N
signifies the N x N
identity matrix and R
xx
=E
x
(K/G)I
N
with E
x
being the chip
energy. The total MSE of the n-th substream is given by

1
( )
N
n i
i
c
j n
=
=

J . (11)
The substream with the smallest total MSE is detected first.
IV. SIMULATION RESULTS
The performance of the proposed ZF / MMSE OSSMIC
receiver with and without chip level block interleaving for
downlink MC-CDMA systems is evaluated through Monte
Carlo simulations and is presented in this section. In addition,
the performance of the novel receiver is compared with the
performance of other detectors for MIMO MC-CDMA
systems. Four transmit and four receive antennas are
considered for the downlink system. Each transmitted frame is
assumed to comprise of eight uncoded QPSK modulated
symbols (P = 8) and each spreading sequence contains eight
chips (G = 8). Walsh-Hadamard spreading sequences are used
for each user. The size of the FFT is considered to be the same
as the number of the subcarriers i.e. N
c
= 64 per frame. The
MIMO channel includes an independent channel path with the
same power level for each transmit and receive antenna pair
1273
Authorized licensed use limited to: Chulalongkorn University. Downloaded on June 11, 2009 at 22:39 from IEEE Xplore. Restrictions apply.
F
F
T
F
F
T
Nulling
Filter
Nulling
Filter
Nulling
Filter
Nulling
Filter
P
/
S
&
R
E
O
R
D
E
R
Symbol
Detector
P
/
S
Interference
Canceller
Interference
Canceller
MAI
Calculation
Symbol
Mapper
Spreading
MIMO Linear
Detector
Symbol
Mapper
Spreading
Interference
Canceller
MIMO Linear
Detector
MIMO Linear
Detector
(User k)
Ant 1
Ant M
Despreading
, k G
c
,1 k
c
,1 k
c
, k G
c
1
N
c
r
M
N
c
r
1
1
r
1
M
r
N
c
r
1
r
G
r
1
,1 k
y
1
, k G
y
1
, k N
c
y
1 N G
c
+
r
1
, 1 k N G
c
y
+
1
,1 k
z
1
, k P
z
1
k
z
1

k
d
1
I
2
I
Detected
Data
1
k
y
Figure 1: The block diagram of OSSMIC receiver for downlink MIMO MC-CDMA system.
Each of the MIMO channels is assumed to possess frequency
selective Rayleigh fading with two taps at arrival times {0,1}
normalized to the chip period. It is assumed that the
maximum delay spread is shorter than the duration of the
cyclic prefix. Hence, ISI is avoided and each chip experiences
flat fading.
Figure 2 demonstrates the BER performance with respect to
different E
b
/N
0
for the linear chip level ZF & MMSE detectors
1274
Authorized licensed use limited to: Chulalongkorn University. Downloaded on June 11, 2009 at 22:39 from IEEE Xplore. Restrictions apply.
[6], the chip level OSIC detectors [5] and our novel ZF &
MMSE OSSMIC detectors for the half loaded four user (K = 4)
case. In view of the performance of ZF and MMSE OSIC
detectors, significant error floors appear at BER = 2x10
-1
for
E
b
/N
o
= 5 dB due to MAI arising to the system. It is evident
that the proposed OSSMIC for both ZF and MMSE filters
overcomes the problem of MAI resulting in a superior
performance over the OSIC detectors. Further observations
show that at BER = 10
-3
, ZF OSSMIC detector produces 5 dB
improvement over linear ZF receiver. Similarly, the MMSE
OSSMIC detector shows 9 dB improvement at BER = 10
-4
over
linear MMSE detector. In addition, when an 8 x 8 chip level
block interleaving is applied to the novel MMSE OSSMIC
receiver, frequency diversity is better exploited ensuing in a
further 4 dB improvement at BER = 10
-5
.
0 5 10 15 20 25 30
10
-7
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
Eb/No in dB
B
E
R
ZF-OSIC
MMSE-OSIC
linear ZF
linear MMSE
ZF-OSSMIC
MMSE-OSSMIC
MMSE-OSSMIC-interleaving
Figure 2: BER performance of novel OSSMIC, linear and OSIC detectors for
4 user downlink MIMO MC-CDMA system.
The same architectures are evaluated for the fully loaded (K =
8) case and their BER performances are depicted in Figure 3. It
can be clearly observed that the problem of MAI causing error
floors is still significant for OSIC receiver. Both the novel ZF
and MMSE OSSMIC detectors produce similar performance
improvement as the previous case over linear detectors. It
should be noted that when the number of users in the system
increases, the performance improvement by block interleaving
is reduced due to higher residual interference.
V. CONCLUSION
In this paper, a chip level OSSMIC receiver architecture with
ZF or MMSE filtering is proposed for downlink MIMO MC-
CDMA systems. The proposed scheme cancels both CAI and
MAI that arises within the system during multiuser
transmission. Furthermore, the proposed receiver only requires
the desired users signature sequence during the detection
process. Simulations reveal a significant advantage in the BER
0 5 10 15 20 25
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
Eb/No in dB
B
E
R
ZF-OSIC
MMSE-OSIC
linear ZF
linear MMSE
ZF-OSSMIC
MMSE-OSSMIC
MMSE-OSSMIC-interleaving
Figure 3: BER performance of novel OSSMIC, linear and OSIC detectors for
8 user downlink MIMO MC-CDMA system.
performance of the OSSMIC receiver over existing receivers.
Shattering existing perception, the proposed receiver permits
layered space-time processing with good performance for
spatially multiplexed MC-CDMA system.
REFERENCES
[1] M. Juntti, M. Vehkapera, J. Leinonen, Z. Li, D. Tujkovic, and S.
Tsumura, S. Hara, MIMO MC-CDMA communications for future
cellular systems, IEEE Comms Mag, vol. 43, issue 2, pp.118-124, Feb
2005.
[2] D. Gesbert, M. Shafi, S. Da-shan, P.J. Smith and A. Naguib, From
theory to practice: an overview of MIMO space-time coded wireless
systems, IEEE J. Select. Areas Commun, vol. 21, issue 3, pp. 281302,
Oct. 1998.
[3] C.R. Nassar, B. Natarajan, Z. Wu, D. Wiegandt, S.A. Zekavat, and S.
Shattil, Multi-carrier Technologies for Wireless Communications, USA:
Kluwer Academic Publishers, 2002.
[4] P.W. Wolniansky, G.J. Foschini, G.D. Golden, and R.A. Valenzuela,
V-Blast: An architecture for realizing very high data rates over the rich-
scattering wireless channel, in Proc IEEE ISSSE-98, pp. 295-300, Sep.
1998.
[5] L. Zhongding, P. Xiaoming, and C. Francois V-BLAST receivers for
downlink MC-CDMA systems, in Proc. VTC Fall, vol. 2, pp.866-870,
2003.
[6] M. Vehkapera D. Tujkovic, Z. Li, and M. Juntti, Receiver design for
spatially layered downlink MC-CDMA system, IEEE Trans. on
Vehicular Technology, vol. 54, no. 3, pp. 1042-1055, May 2005.
[7] Y. Lee, H. Park, Low-Complexity Detections for downlink MIMO
MC-CDMA systems, in Proc. PIMRC Fall, pp. 1-5, Sep 2006
[8] K. Kyeongyeon, H. Jaesang, L. Chungyong, and S. Seijoon,
Asymptotic analysis of downlink MIMO multicarrier CDMA systems
with a minimum mean square error receiver, in Proc.VTC Spring, vol
3, pp. 1501-1505, 2006.
[9] J. Hu, K. Kim, S. Shim, M. Kim, and C. Lee, An MMSE-nulling
partial-PIC receiver for multiuser downlink MIMO MC-CDMA
systems, IEICE Transactions on Communications, vol. E88-B, No. 4,
pp. 1725-1729, Apr 2005.
1275
Authorized licensed use limited to: Chulalongkorn University. Downloaded on June 11, 2009 at 22:39 from IEEE Xplore. Restrictions apply.