Parity Bit Selected and Permutation Spreading for

CDMA/MIMO Systems in Frequency-Nonselective

Rayleigh Fading Channels

Claude DAmours
School of Information Technology and Engineering
University of Ottawa
Ottawa, Canada
Jean-Yves Chouinard
Dpartement de gnie lectrique et informatique
Universit Laval
Quebec City, Canada
AbstractIn this paper, we propose new techniques for
CDMA/MIMO systems that employ multiple spreading codes.
Rather than assigning each transmit antenna a different
spreading waveform, the data to be transmitted on a given
signaling interval determines which spreading waveforms will be
used by each transmit antenna. Compared to a more
conventional approach, which employs fixed spreading
waveforms for each transmit antenna, our approach provides bit
error rate performance improvement at moderate and high
E
b
/N
o
.

Index TermsCDMA, MIMO, Error Control Coding.
I. INTRODUCTION
N [1], a novel method of using the spreading waveforms to
carry redundant information in a code division multiple
access (CDMA) system is presented. A block of k data bits is
input to the parity bit calculator of a systematic (n,k) linear
block encoder. Rather than append these (n-k) parity bits to
the end of the data block, the parity bits are used to select one
of the 2
(n-k)
mutually orthogonal spreading waveforms assigned
to the user. The receiver must first determine which spreading
waveform was used by the transmitter. This is done by
inputting the received signal to a bank of matched filters, each
matched to the different spreading waveforms. Once the
receiver determines which spreading waveform was employed
by the transmitter, it then compares the outputs of the
appropriate matched filter to the subset of data vectors that are
associated with the given spreading waveform.
Wireless communications systems employing multiple
transmit and receive antennas are referred to as multiple input
multiple output (MIMO) systems. MIMO systems take
advantage of fading diversity to allow spatial multiplexing
(SM) or space-time coding (STC). V-BLAST [2] is an
example of a SM system and Alamouti code [3] is an example
of a simple STC. Both perform well if the channel gains are
known to the transmitter or the receiver. In [4, 5], MIMO
channel gains are accurately measured by employing
pseudonoise (PN) waveforms. Channel gains in CDMA
systems are also accurately estimated using spread pilot
symbols in [6, 7]. This implies that the use of orthogonal
spreading waveforms at different transmit antennas will allow
the receiver to accurately estimate the channel gains in MIMO
systems. Therefore, the combination of CDMA and MIMO
techniques is a natural one.
For channel estimation, pilots would be periodically
inserted into the data streams of each transmit antenna. To
estimate the channel gain on each transmit antenna receive
antenna link, we would use orthogonal spreading waveforms
to eliminate any crosstalk between the different links. Thus it
would also seem natural to spread the data transmitted from
each transmit antenna with the same spreading waveforms.
However, from [1], we see that by selecting different
spreading waveforms depending on the data that is being
transmitted, we obtain a performance gain. We propose using
a similar approach for CDMA/MIMO systems. However, we
will show in section III that we need to modify the approach
of [1] to apply the concept to CDMA/MIMO systems.
II. CDMA/MIMO SYSTEMS EMPLOYING ORTHOGONAL
SPREADING WAVEFORMS
Consider a MIMO system that has N
t
transmit antennas and
N
r
receive antennas. The serial data, whose bit rate is R
b
, is
converted into N
t
parallel data streams, each with bit rate
R
b
/N
t
. The ith data stream of user m is spread by spreading
waveform w
mi
(t), which is an antipodal signal with chip rate R
c

and is selected from a set of mutually orthogonal spreading
waveforms {c
m1
(t), c
m2
(t), , c
mN
(t)}C
m
. Therefore

=
=

j i
j i
dt t c t c
T
mj mi
T
0
1
) ( ) (
0
1
. (1)
Different users are assigned unique sets of spreading
waveforms. Thus C
m
C
l
= for l m.
On signaling interval n, the message to be transmitted is
m
(n)
= [ ]
) ( ) (
2
) (
1
,..., ,
n
N
n n
t
m m m , where m
k
(n)
= 0 or 1 with equal
probability and is independent of the other bits in the message
vector. The data is transmitted using binary phase shift keying
(BSPK) or antipodal modulation, therefore we define b
(n)
=
[ ]
) ( ) (
2
) (
1
,..., ,
n
N
n n
t
b b b = 2m
(n)
-1.
At the receiver, the output of each receiver antenna is
correlated with each spreading waveform and the
I
1550-2252/$25.00 2007 IEEE 1475
contributions from the different antennas are combined
according to the technique under consideration. A generic
CDMA/MIMO transmitter is shown in Figure 1. The link
gains are also shown,
ij
is the complex channel gain on the
link between transmit antenna i and receive antenna j.
Depending on the spreading method used, w
i
(n)
(t) is the
spreading waveform used to spread the data transmitted by
antenna i on time interval n.

















Figure 1: CDMA/MIMO transmitter.
The CDMA/MIMO receiver is shown in Figure 2. The kth
matched filter output on receive antenna j is u
jk
. The value of
these decision variables and how they are combined depends
on the method used by the transmitter. For illustration
purposes, we will examine the performance of a
CDMA/MIMO system with 4 transmit antennas.




















Figure 2: CDMA/MIMO receiver.

A. Conventional Approach
For a conventional CDMA/MIMO system the spreading
waveforms are fixed, i.e. ) ( ) (
) (
nT t c t w
mi
n
mi
= , since they do
not depend on the data to be transmitted on the nth signaling
interval. Therefore, we only need N = N
t
= 4 spreading
waveforms per user for this system.
For this system, the decision variables of Figure 2 are:

) ( ) ( ) ( n
jk
n
kj
n
k jk
n ATb u + = (2)
where A is the carrier amplitude and
) (n
jk
n is the response of
the jth receive antennas kth matched filter to the input noise.
The receiver then estimates
) (n
i
b from the decision variables
of (2) as follows:
( )
|
|
.
|

\
|
=

j
n
ij ji
n
i
u b
*
) ( ) (
sgn

(3)

B. CDMA/MIMO System with Parity Bit Selected
Spreading
Parity bit selected spreading is discussed in [1]. A block of
data is input to the parity bit calculator of a systematic block
code. Rather than append the parity bits to the end of the
message block, the parity bits are used to select one of a set of
2
(n-k)
spreading sequences, where n-k is the number of parity
bits of the code. The result is that the system divides the
possible messages into groups that are carried by the same
spreading waveform.
Let M be the set off all possible message vectors. Let M
1
be
a subset of M that is closed under modulo-2 addition. Let M
2
,
M
3
, , M
K
be the cosets of M
1
, where K =
t
N
2 /L, where L is
the number of vectors in M
1
. Let
) ( ... ) ( ) (
) ( ) (
2
) (
1
t w t w t w
n
mN
n
m
n
m
t
= = = = c
mp
(t-nT) if m
(n)
M
p
.
Each transmit antenna uses the same spreading waveform, but
the spreading waveform used is related to the message that is
being transmitted, much like parity bits are related to the
messages in block coding.
In the system that is demonstrated in this paper, M
1
=
{0000, 1111}. This subset of M has 7 cosets, M
2
= {0001,
1110}, M
3
= {0010, 1101}, etc. If m
(n)
is either 0000 or 1111,
the output of all antennas will be spread by spreading
waveform c
m1
(t-nT). Similarly, if m
(n)
is either 0001 or 1110,
the output of all antennas will be spread by spreading
waveform c
m2
(t-nT). This strategy requires that each user be
assigned a unique set of 8 mutually orthogonal spreading
waveforms.
The decision variables input to the decision device in Figure
2 are given by:

+ +
+ +
=
used not is ) ( if ,
used is ) ( if ),
(
) (
) ( ) (
4
) (
4
) (
3
) (
3
) (
2
) (
2
) (
1
) (
1
nT t c n
nT t c n b
b b b AT
u
mk
n
jk
mk
n
jk
n
j
n
n
j
n n
j
n n
j
n
jk


(4)

N
t
transmit
antennas

N
r
receive
antennas
r
N 1

11

12

21

22

r
N 2

1
t
N

2
t
N

r t
N N

b
1
(n)

b
2
(n)

) (n
Nt
b
Serial
to
Parallel
Converter
Data
in
) (
) (
1
t w
n
m
) (
) (
2
t w
n
m
) (
) (
t w
n
mNt

N
t
transmit
antennas

N
r
receive
antennas
r
N 1

11

12

21

22

r
N 2

1
t
N

2
t
N

r t
N N

b
1
(n)

b
2
(n)

) (n
Nt
b
Serial
to
Parallel
Converter
Data
in
) (
) (
1
t w
n
m
) (
) (
2
t w
n
m
) (
) (
t w
n
mNt

Filter
matched to
c
1
(t)
Filter
matched to
c
2
(t)
Filter
matched to
c
N
(t)

u
11
u
12
u
1N
Bank of
matched
filters
u
21
u
22

u
2N
Bank of
matched
filters

1
r
N
u
2
r
N
u
N N
r
u
Combiner
And
Decision
Device

) (
1

n
b
) (
2
n
b
) (

n
N
t
b

Filter
matched to
c
1
(t)
Filter
matched to
c
2
(t)
Filter
matched to
c
N
(t)

u
11
u
12
u
1N
Bank of
matched
filters
u
21
u
22

u
2N
Bank of
matched
filters

1
r
N
u
2
r
N
u
N N
r
u
Combiner
And
Decision
Device

) (
1

n
b
) (
2
n
b
) (

n
N
t
b
1476
We use maximum likelihood (ML) detection to estimate
b
k
(n)
. Therefore we estimate b
(n)
as the vector which
minimizes:

( )
2
2
( ) ( )

min
n n

= +
b
b u bH u
B
(5)
where u is the vector made up of decision variables from the
matched filters corresponding to the spreading waveform used
by vector b, u is the vector made up of all of the decision
variables not in u, the set of all possible vectors of b is B and
H
(n)
is the N
t
N
r
channel matrix on the nth signaling interval,
which is given by:

(
(
(
(
(

=
r t t t
r
r
N N N N
N
N
n



"
# % # #
"
"
2 1
2 12 21
1 12 11
) (
H (6)

For example, if we consider the case when N
t
= N
r
= 4, then
for b
(n)
= [-1,-1,-1,-1], the spreading waveform used is c
m1
(t-
nT). Therefore, when considering b = [-1,-1,-1,-1] in (5), u =
[u
11
, u
21
, u
31
, u
41
]. The vector u in (5) is [u
12
, u
13
, u
14
, u
22
,
,u
44
]. Similarly, when considering b = [-1,-1,-1,1], which
corresponds to spreading waveform c
m2
(t), we use u = [u
12
,
u
22
, u
32
, u
42
].
C. CDMA/MIMO with Permutation Spreading
This system is similar to CDMA/MIMO with parity bit
spreading in that the possible message vectors are divided into
cosets and that each coset has a unique spreading signature. In
parity bit selected spreading, the data on different antennas are
spread by a single spreading waveform that is selected based
on the parity bits that are generated when a message is
encoded. When permutation spreading is used, depending on
which coset the message comes from, a unique permutation of
spreading waveforms are used. Each permutation employs N
t

of the N spreading waveforms and we attempt to minimize the
number of spreading waveforms that each permutation have in
common. Furthermore, if a spreading waveform is used by
antenna i in one permutation, it cannot be used by antenna i in
any other permutation. The design of the different spreading
permutations is based on t-designs [8] which are used in
permutation modulation schemes [9,10].
For the system discussed in this paper, the spreading
permutations are listed in Table 1. The system employs 8
spreading waveforms per user. Each permutation of 4
spreading waveforms has 2 spreading waveforms in common
with 5 other permutations and 1 spreading waveform in
common with the 2 other permutations. Each spreading
waveform appears in 4 permutations and is used by a given
antenna only once.
The advantage to using permutation spreading is that the
different spreading patterns create dependence between the
parallel data streams yet still maintains orthogonality between
the streams. In other words, the decision variables at the
output of each antenna will contain either noise or a signal
contribution from one transmit antenna, but not all transmit
antennas as in (4).

Table 1: Spreading permutations for CDMA/MIMO system
with 4 transmit antennas.
Coset Message
vectors
) (
) (
1
t w
n
m
) (
) (
2
t w
n
m
) (
) (
3
t w
n
m
) (
) (
4
t w
n
m

M
1
0000
1111
c
1
(t-nT) c
3
(t-nT) c
5
(t-nT) c
7
(t-nT)
M
2
0001
1110
c
8
(t-nT) c
1
(t-nT) c
4
(t-nT) c
5
(t-nT)
M
3
0010
1101
c
2
(t-nT) c
4
(t-nT) c
3
(t-nT) c
8
(t-nT)
M
4
0011
1100
c
5
(t-nT) c
2
(t-nT) c
6
(t-nT) c
3
(t-nT)
M
5
0100
1011
c
6
(t-nT) c
7
(t-nT) c
1
(t-nT) c
4
(t-nT)
M
6
0101
1010
c
3
(t-nT) c
6
(t-nT) c
8
(t-nT) c
1
(t-nT)
M
7
0110
1001
c
7
(t-nT) c
8
(t-nT) c
2
(t-nT) c
6
(t-nT)
M
8
0111
1000
c
4
(t-nT) c
5
(t-nT) c
7
(t-nT) c
2
(t-nT)

The decision variables can be expressed as:

+
=
all at used not is ) ( if ,
antenna t by transmi
used is ) ( if ,
) (
) ( ) ( ) (
nT t c n
i
nT t c n ATb
u
mk
n
jk
mk
n
jk
n
ij
n
i
jk

(7)
On each receive antenna, 4 decision variables will contain a
signal component plus a noise component, while the other 4
will contain only a noise component.
Maximum likelihood detection is used for this scheme. In
this case, the detection rule is:

2
2
( ) ( ) ( )
1

min
t
N
n n n
i i i
i
b

=
| |
= +
|
\ .

b
b u h u
B
(8)
where ] ,... , [
2 1
i r i i
k N k k i
u u u = u , k
i
corresponds to the subscript
of the spreading waveform used by transmit antenna i on the
nth signaling interval, u is the vector containing all of the
decision variables not in u
1
, u
2
,
r
N
u and h
i
(n)
is the ith row
vector of H
(n)
in (6). For example, if we are considering b
from M
8
in (8), then u
1
= [u
14
, u
24
, u
34
, u
44
], u
2
= [u
15
, u
25
, u
35
,
u
45
], u
3
= [u
17
, u
27
, u
37
, u
47
] and u
4
= [u
12
, u
22
, u
32
, u
42
]. The
reason why we are careful to ensure that a spreading
waveform is used only once by each transmit antenna is so
that when performing (8) on an incorrect vector, even though
some of the decision variables used contain a signal
component, the channel gain in the decision variable part and
the channel gain in the b
i
(n)
h
i
(n)
part of (8) are not the same. In
this manner, we do not accidentally cancel out the signal part
of the decision variable when considering an incorrect
message vector. This is assuming, of course, that the different
1477
channel gains are lowly correlated.
III. SIMULATION RESULTS
Simulations to determine the bit error rate (BER)
performance of the three systems discussed in section 2 are
presented in this section. We consider systems with 4 transmit
antennas and receivers with 1 or 4 receive antennas. The
following assumptions were made in the simulation models:
1) The fading is frequency nonselective. In other words, the
multipath spread is 0 and there is no channel induced
intersymbol interference (ISI).
2) The channel gains are slowly varying circularly-
symmetric complex Gaussian random variables with 0
mean and variance 1.
3) The channel gains of different transmit antenna receive
antenna links are uncorrelated (therefore they are
independent).
4) The receiver estimates the channel gains perfectly without
a power penalty. In other words, when detecting a
symbol or message, the receiver uses the exact channel
matrix.
Figure 3 shows the bit error rate of the three different
CDMA/MIMO methods for the case when there are 4
transmitting antennas and 1 receiving antenna. Figure 4 shows
the simulated BER of the three types of CDMA/MIMO
systems when there are 4 transmitting and 4 receiving
antennas.

1.00E-05
1.00E-04
1.00E-03
1.00E-02
1.00E-01
10 12 14 16 18 20 22 24
Eb/No (in dB)
B
i
t

E
r
r
o
r

R
a
t
e
Conventional method
Parity bit selected spreading
Permutation spreading


Figure 3: BER for the different CDMA/MIMO systems
with N
t
= 4 and N
r
= 1.
We can see from the BER results of Figures 3 and 4 that
both parity bit selected spreading and permutation spreading
provide BER improvement compared to fixed orthogonal
spreading of the different transmit antenna outputs. This is
due to the dependence created by the spreading strategies
between the parallel data streams. In the more conventional
system, if a data stream encounters a set of channels on which
the channel gain is poor, the detection of the data on that
stream is prone to high error rates. In the new strategies, data
transmitted on poor channels can benefit from the detection of
data streams on links with high channel gain.

1.00E-06
1.00E-05
1.00E-04
1.00E-03
1.00E-02
1.00E-01
4 5 6 7 8 9 10 11 12 13 14
Eb/No (in dB)
B
i
t

E
r
r
o
r

R
a
t
e
Conventional method
Parity bit selected spreading
Permutation spreading

Figure 4: BER for the different CDMA/MIMO systems
with N
t
= 4 and N
r
= 4.
It is also interesting to note that permutation spreading
provides significant BER performance gains compared to
parity bit selected spreading. This is due to the nature of the
decision variables in each scheme. In parity bit selected
spreading, one decision variable per antenna contains a signal
component which is a sum of all of the transmitted signals.
All other decision variables are noise. Although this should
make it easy to detect which spreading sequence was used, in
some cases, even when all channel gains are strong, it is
possible for the linear combination of these channel gains to
yield a results that is close to 0. Thus, there is some self-
interference similar to SM systems. In permutation spreading,
all decision variables are made up of one signal plus noise, or
noise only. There is no real self interference when the channel
gains are independent.
IV. MULTIPLE USER SYSTEMS
In the previous section, we examined the performance of
single user CDMA/MIMO systems. However, the advantage
of CDMA systems is that multiple users can simultaneously
transmit over the same range of frequencies by assigning each
user a unique spreading waveform. In CDMA/MIMO, we
need to assign each user a unique set of spreading waveforms.
Consider an asynchronous CDMA system. Let us suppose
that each users signal is received with the same average
E
b
/N
o
. In other words, power control is used to correct for the
large scale fading. Let us further assume that the power
control is not fast enough to correct for the small scale fading.
It is shown in [11] that, if we assume random signature
spreading waveforms and we have perfect power control, the
energy per bit to effective noise spectral density is:

( )( )( )
c b o b
o b
o
b
R R N E U
N E
N
E
/ 3 / 2 ) 1 ( 1 +
=

(9)
where U is the number of simultaneously transmitting users
1478
and R
c
/R
b
is the spreading factor. For CDMA/MIMO systems,
the random signature spreading waveform assumption is not
correct since an interfering user cannot use any spreading
waveform since the desired user is assigned a handful of
mutually orthogonal spreading waveforms. However, if the
spreading factor is high, there are many spreading waveforms
available, and since the system is asynchronous, even
orthogonal codes can yield poor cross-correlation properties,
thus the random signature assumption is still a good
approximation. Each interfering user produces N
t
interfering
data streams, each with bit rate R
b
/N
t
. Therefore, we have
N
t
(U-1) interfering data streams each transmitted with
spreading factor N
t
R
c
/R
b
. Thus if we modify (9) to take this
into account, the factors N
t
and 1/N
t
cancel each other out.

( ) ( ) ( ) 1 ( 1) 2/ 3 /
b b o
t b o b t c
o
E E N
N U E N R N R
N
=

+
(10)
Thus (9) is also valid for our CDMA/MIMO system.
Let us assume that we are employing spreading waveforms of
128 chips per signaling interval. The spreading factor for each
parallel data stream is 128, but since we transmit N
t
bits per
signaling interval, the actual spreading factor is 128/N
t
or 32
for our systems in section III. Suppose that each user
transmits with enough power so that his or her signal is
received with average E
b
/N
o
= 25 dB. Let us further assume
that the systems capacity is determined by the constraint that
the BER cannot be greater than 10
-3
. Using the BER results
from figures 3 and 4, we can determine the maximum number
of simultaneously transmitting users for each system under
these conditions. Table 2 shows the approximate value for

o b
N E / so that the BER is 10
-3
(obtained from Figures 2 and
3) for each system and Table 3 shows the maximum number
of users that the 3 CDMA/MIMO systems can support.
Table 2:

o b
N E / required for each system for BER = 10
-3
.

# receive
antennas
Conventional
orthogonal
spreading
Parity bit
selected
spreading
Permutation
spreading
1 23 dB 19 dB 15.7 dB
4 10.3 dB 9.6 dB 5.6 dB

Table 3: Maximum number of users supported for the
CDMA/MIMO systems when E
b
/N
o
= 25 dB, R
c
/R
b
= 32, N
t
=
4 and BER
max
= 10
-3
.

# receive
antennas
Conventional
orthogonal
spreading
Parity bit
selected
spreading
Permutation
spreading
1 1 1 2
4 5 6 14

We can see from the results in Table 3 that the improved
BER performance of parity bit selected spreading and
permutation spreading systems can be traded off against
increased capacity compared to the conventional
CDMA/MIMO system.
V. CONCLUSION
In this paper we presented new techniques for spreading
parallel data streams in CDMA/MIMO systems. The new
techniques create dependency between the streams and
provide performance improvement or improved spectral
efficiency compared to a more conventional spreading
approach.
The disadvantage of the new techniques is that each user
must be assigned additional spreading waveforms and there is
an overall increase in system complexity. However, in the
case of permutation spreading, the gains are large and may
warrant the additional complexity.
In this paper, we have only considered uncorrelated fading
links. Adding correlation to the channel model would not
affect the performance of the system which employs a
conventional spreading approach as there is no dependency
between the different data streams, nor is there any self
interference at the receiver due to the orthogonality of the
spreading waveforms. However, correlation should degrade
the performance of the two new strategies, so it would be
advantageous to determine the performance on these systems
in correlated fading channels.
REFERENCES
[1] C. DAmours, Parity Bit Selected Spreading Sequences: A Block
Coding Approach to Spread Spectrum, IEEE Commun. Letters, vol. 9,
pp 16-18, January 2005.
[2] P.W. Wolniansky, G.J. Foschini, G.D. Golden, R.A. Valenzuela, V-
BLAST: An Architecture for Realizing Very High Data Rates Over the
Rich-Scattering Wireless Channel, Proc. URSI Int. Symp. on Signals,
Systems and Electronics (ISSSE 98), pp. 295-300, Pisa, Italy, Sept. 1998.
[3] S.M. Alamouti, A Simple Transmit Diversity Technique for Wireless
Communications, J. Sel. Areas Commun., vol. 16, pp. 1451-1458, Oct.
1998.
[4] C. Squires, T. Willink, B. Caron, A Flexible Platform for MIMO
Channel Characterisation and System Evaluation, Proc. 15
th
Int. Conf.
on Wireless Comm., Calgary, Canada, July 2003.
[5] G. Downes, T. Willink, C. DAmours, Capacity Limitations of Real
MIMO Channels and Their Impact on System Performance, Proc. 17
th

Int. Conf. on Wireless Comm., pp. 100-104, Calgary, Canada, July 2005.
[6] S. Affes, N. Kandil, P. Mermelstein, Efficient Use of Pilot Signals in
Wideband CDMA Array Receivers, Proc. IEEE Int. Conf. Acoustics,
Speech and Sig. Proc. (ICASSP 2003), pp. 2526-2531, Hong Kong,
April 2003.
[7] C. DAmours, M. Moher, A. Yongacoglu, Comparison of Pilot Symbol
Assisted and Differentially Detected BPSK for DS-CDMA Systems
Employing RAKE Receivers in Rayleigh Fading Channels, IEEE
Trans. Veh. Tech., vol. 47, pp. 1258-1267, Nov. 1998.
[8] F.J. MacWilliams and N.J.A. Sloane, The Theory of Error-Correcting
Codes, North-Holland: New York, NY, 1977.
[9] D. Slepian, Permutation Modulation, Proc. IEEE, vol. 53 pp. 22-236,
March 1965.
[10] C. DAmours., A. Yongaoglu., Hybrid DS/FH-CDMA System
Employing MT-FSK Modulation for Mobile Radio, Proc. IEEE Int.
Symp. Pers. Indoor Mobile Radio Commun. (PIMRC 95), pp. 164-168,
Toronto, Canada, Sept. 1995.
[11] M.B. Pursley, Performance Evaluation for Phase-Coded Spread-
Spectrum Multiple-Access Communication Part I: System Analysis,
IEEE Trans. Commun., vol. 25, pp. 795-799, Aug. 1977.
1479