4-PSK Balanced STTC with two transmit antennas

Thi Minh Hien Ngo, Gheorghe Zaharia, St ephane Bougeard, Jean Francois H elard
Institute for Electronics and Telecommunications of Rennes (IETR) - UMR CNRS 6164
INSA - 20 avenue des Buttes de Co esmes, 35043 Rennes, France
Email: [email protected]
AbstractIn this paper we introduce a new class of space-time
trellis codes (STTC). We call them Balanced STTC (B-STTC)
because the points of the constellation are used with the same
probability. Comparing to known codes, the balanced codes offer
the best performance. Therefore, the systematic search for good
codes can be reduced to this class. We present here the design of
the 4-PSK balanced STTC with 2 transmit antennas. A complete
list of the best 4-state and several 16-state good balanced codes
are also given.
I. INTRODUCTION
Barring complexity issues, STTC have been shown to
outperform other space-time coded systems. Tarokh et al.
[1] introduced the concept of STTC as an extension to the
conventional time-convolutional coding and derived criteria to
obtain optimal codes. For quasistatic at Rayleigh or Rician
channels, performance is shown to be determined by the diver-
sity advantage quantied by the rank of certain matrices and
by the coding advantage that is quantied by the determinants
of these matrices. For rapidly changing at Rayleigh chan-
nels, performance is determined by the diversity advantage
quantied by the generalized Hamming distance of certain
sequences and by the coding advantage that is quantied by
the generalized product distance of these sequences which
are constructed from pairs of distinct codewords. In [2] Chen
introduced the trace criterion which governs the coding for
systems with a great product of the numbers of transmit and
receive antennas.
Based on above criteria, many different STTC for 2 transmit
antennas have been found by a systematic code search [3][9].
Until now, no efcient construction method of STTC has
been proposed. The main contribution of this paper is the
description of a method of construction of a new class of
Balanced STTC (B-STTC) which offers the best performance.
Therefore, the systematic search for good codes can be reduced
to this class.
The rest of the paper is organized as follows. Section
II briey describes the STTC. Their performance criteria,
depending on the channel properties, are presented in Section
III. The new class of balanced codes is introduced in Section
IV and their properties are listed in Section V. The design
of 4-PSK balanced space-time trellis codes for 2 transmit
antennas is described in Section VI. Finally, it is shown
in Section VII that the best B-STTC outperform or equal
previously known codes.
II. SPACE TIME TRELLIS CODING
We consider the case of 4-PSK space-time trellis encoder
as shown on Fig.1.
- - c
k
1,1
c
k
2,1
c
k
1,2
c
k
2,2
c
k
1,+1
c
k
2,+1
y
k
t
? ? ? ?

mod 4
-
- - - c
n
T
1,1
c
n
T
2,1
c
n
T
1,2
c
n
T
2,2
c
n
T
1,+1
c
n
T
2,+1
y
n
T
t
? ? ? ? ? ?

mod 4
-
INPUT MEMORY
b
t
1
b
t
2
b
t
1
b
t
2
- -
? ? ? ?
? ? ? ?
? ? ? ? ? ?
- - c
1
1,1
c
1
2,1
c
1
2,2
c
1
1,+1
c
1
2,+1
y
1
t
? ? ? ?

mod 4
-
b
t1
1
b
t1
2
-
? ?
-
c
1
1,2
? ?
-
? ?
? ?

-
Fig. 1. Space-time trellis encoder with 4-PSK and n
T
transmit antennas
In the general case of a 2
n
-PSK modulation, this encoder
is composed of one input block of n bits and memory
blocks of n bits. At each time t Z, all the bits of a block
are replaced by the n bits of the previous block. The i
th
bit b
tj+1
i
, i = 1 . . . n, of the j
th
block, j = 1 . . . + 1,
is associated to n
T
multiplier coefcients c
k
i,j
Z
2
n,
k = 1 . . . n
T
where n
T
is the number of transmit antennas. A
ST trellis encoder is thus classically dened by its generator
matrix C of n
T
n( + 1) coefcients:
C =
_

_
c
1
1,1
. . . c
1
n,1
. . . c
1
1,+1
. . . c
1
n,+1
.
.
. . . .
.
.
.
c
k
1,1
. . . c
k
n,1
. . . c
k
1,+1
. . . c
k
n,+1
.
.
. . . .
.
.
.
c
n
T
1,1
. . . c
n
T
n,1
. . . c
n
T
1,+1
. . . c
n
T
n,+1
_

_
(1)
The encoder outputs for the k
th
antenna are computed as
y
k
t
=
n

i=1
+1

j=1
b
tj+1
i
c
k
i,j
mod 2
n
(2)
h
a
l
-
0
0
1
7
2
7
3
3
,

v
e
r
s
i
o
n

1

-

2
0

S
e
p

2
0
0
7
Author manuscript, published in "Vehicular Technology Conference, 2007. VTC2007-Spring. IEEE 65th, Ireland (2007)"
where y
k
t
Z
2
n represents the index of the 2
n
-PSK symbol
s
k
t
= e
y
k
t

2
sent to the k
th
antenna. The modulated streams
for all antennas are then transmitted simultaneously.
At the receiver, each one of the n
R
antennas collects the
superpositions of the faded replicas of the n
T
transmitted sym-
bols. Each link between one transmit antenna and one receive
antenna is assumed to experience statistically independent at
Rayleigh fading. Therefore, we can use the following model
of the received signal:
r
l
t
=
n
T

k=1
h
kl,t
s
k
t
+ n
l
t
(3)
where r
l
t
is the signal received at antenna l at time t;
h
kl,t
is the complex path gain from transmit antenna k to
receive antenna l at time t; s
k
t
is the transmitted complex
symbol corresponding to y
k
t
; and n
t
l
is the AWGN sample
for receive antenna l at time t. The noise samples are
independent samples of a zero-mean complex Gaussian
random variable with spectral density N
0
/2 per dimension.
Maximum likelihood decoding is then used to extract the
transmitted codeword.
III. DESIGN CRITERIA
Design criteria have been proposed in [1] [2] to exploit the
n
T
n
R
spatial diversity order and to offer optimal coding gain.
Cases of slow and fast Rayleigh fading are mainly studied.
The transmitted n
T
dimension symbols s
t
=
_
s
1
t
s
2
t
. . . s
n
T
t

T
,
where []
T
denotes the transpose operator, are assumed to be
grouped in a frame of length L
f
.
A channel is said to be a slow Rayleigh fading channel
if, during a transmission of a frame, the complex path gains
h
kl
do not change with time t, but are independent from one
frame to the next one. In the case of a fast Rayleigh fading
channel, the complex path gains h
kl,t
are independent from a
n
T
dimension symbol to the next one. For each case, criteria
are derived from the minimization of the Pairwise Error
Probability (PEP), i.e. the probability of transmitting the
n
T
L
f
dimension coded frame S =
_
s
t
s
t+1
. . . s
t+L
f
1

and deciding erroneously in favour of another n

. The
n
T
n
T
product matrix A = BB

is introduced where
B

denotes the hermitian of the n

_
e
1
t
s
1
t
. . . e
1
t+q
s
1
t+q
. . . e
1
t+L
f
1
s
1
t+L
f
1
.
.
.
.
.
.
.
.
.
.
.
.
e
n
T
t
s
n
T
t
. . . e
n
T
t+q
s
n
T
t+q
. . . e
n
T
t+L
f
1
s
n
T
t+L
f
1
_

_ (4)
A. The case of slow Rayleigh fading
In the case of slow Rayleigh fading channels, two criteria
have been proposed in [1]. Firstly, in order to maximize the
diversity advantage, the product matrix A has to be full rank
over all possible pairs (E, S). Since the maximal value of
rank(A) is n
T
, the achieved spatial diversity order is then
equal to the product n
T
n
R
. Secondly, in order to maximize
the coding gain, the minimum product

rank(A)
k=1

k
, where

k
are the nonzero eigenvalues of A, has to be maximized by
choosing the generator matrix C. If rank(A) = n
T
, we can
note that
det(A) =
n
T

k=1

k
=
n
T

k=1
_
_
L
f
1

q=0

e
k
t+q
s
k
t+q

2
_
_
(5)
The codes that can achieve the best performance present the
greatest value of the minimum product

rank(A)
k=1

k
. For a
full rank product matrix, the achieved coding advantage is
then equal to det(A)
(1/n
T
)
.
B. The case of fast Rayleigh fading
In the case of fast Rayleigh fading channels, different
criteria have been obtained in [1]. Tarokh denes the Hamming
distance d
H
(E, S) between two coded frames E and S as
the number of time intervals for which |e
t
s
t
| = 0. In
order to maximize the diversity advantage over fast Rayleigh
fading channels, the minimal Hamming distance d
H
(E, S)
computed over all pairs of coded frames has to be maximized
by choosing the generator matrix C. The achieved spatial
diversity order is then equal to the product d
H
(E, S).n
R
.
In the same way, Tarokh introduces the product distance
d
2
p
(E, S) as the product of euclidian distance between the
L
f
n
T
-dimension symbols composing the coded frames S and
E. The product distance is written
d
2
p
(E, S) =
L
f
1

q=0
_
n
T

k=1

e
k
t+q
s
k
t+q

2
_
=
L
f
1

q=0
d
2
E
(e
t+q
, s
t+q
) (6)
In order to maximize the coding gain, the minimum product
distance d
2
p
(E, S) computed over all pairs (E, S) has to be
maximized by choosing the generator matrix C. The achieved
coding advantage is then equal to d
2
p
(E, S)
1/d
H
(E,S)
.
C. The case of a larger number of antennas
In [2], Chen proposes a new criterion which is valid in
the case of slow and fast Rayleigh fading channels since the
product rank(A).n
R
> 3. Under this assumption, the PEP
is minimized if the sum of all the eigenvalues of the product
matrix is maximized. For a square matrix, the sum of all the
eigenvalues is equal to the trace of the matrix. It can be written
as:
tr(A) =
n
T

k=1

k
=
n
T

k=1
_
_
L
f
1

q=0

e
k
t+q
s
k
t+q

2
_
_
(7)
For each pair of coded frames, a matrix A and then tr(A)
can be computed. The minimum trace is the minimum of all
these values tr(A). Since the product rank(A).n
R
> 3, the
minimization of the PEP amounts to use a code which has the
maximum value of the minimum trace.
h
a
l
-
0
0
1
7
2
7
3
3
,

v
e
r
s
i
o
n

1

-

2
0

S
e
p

2
0
0
7
IV. BALANCED CODES
A. Why balanced codes?
The concept of balanced codes is based on the observation
that all the good STTC proposed in the literature present the
same property: the generated symbols of the constellation are
equally probable.
Indeed, if the binary input data is generated by a me-
moryless source S = {0, 1} with equally probable symbols,
then, in the case of 4 - PSK modulation, from a given state
X=[x
1
x
2
... x
L
]
T
Z
L
2
of the shift-register realized by (+1)
blocks of n bits, the MIMO symbol Y=[y
1
y
2
... y
n
T
]
T
Z
n
T
4
generated by the STTC encoder shown in Fig.1 is:
Y = C X (8)
where C is the generator matrix (1). This is a deterministic
relation. Therefore, the STTC is dened by a map:
: Z
L
2
Z
n
T
4
(9)
which associates to the state X an unique codeword Y. Note
that (Z
L
2
) Z
n
T
4
represents the set of generated codewords
Y. A given codeword Y can be obtained for several states
X Z
L
2
. Let n(Y) be the number of occurrences of the
codeword Y.
By denition, a STTC is balanced if and only if each
generated codeword Y (Z
L
2
) has the same number of
occurrences n(Y) = n
0
1.
In addition, if (Z
L
2
) = Z
n
T
4
, then all the codewords are
generated and the STTC is fully balanced.
Due to the random source S = {0, 1}, from a given state X
the encoder can have only 4 equally probable next states. The
matrix T of the transition probabilities between these states
corresponds to a Markov chain. Due to the symmetry of the
matrix T, the steady state probabilities of the states X are
all equal. For a balanced code, by using (8), the generated
codewords Y are also equally probable. In other words, the
generated symbols of the constellation are equally probable.
B. Properties of balanced STTC
The design of the B-STTC is based on the following
properties:
Theorem 1: If a MIMO code with a L-length shift-register
is fully balanced then L L
min
= n.n
T
. One can observe
that L
min
= dim(Z
n
T
2
n ).
Theorem 2: Let us consider a balanced MIMO code with a
L-length shift-register. Then, for any additional column matrix
C
i
Z
n
T
2
n , the resulting MIMO code with a (L + 1)-length
shift-register is also balanced.
Denition 1: The vectors C
1
, C
2
, . . . , C
L
are linearly in-
dependent if the equation
x
1
C
1
+ x
2
C
2
+ + x
L
C
L
= 0 Z
n
T
2
n (10)
with x
i
{0, 1} holds if and only if x
i
= 0 for every
i = 1, 2, . . . , L.
Denition 2: A set of linearly independent vectors
C
1
, C
2
, . . . , C
m
is called a base for Z
n
T
2
n if and only if
span(C
1
, C
2
, ..., C
m
) =
_
m

i=1
x
i
C
i
/x
i
{0, 1} , i = 1, 2, ..., L
_
= Z
n
T
2
n .
In this case, m = n.n
T
is the dimension of the set Z
n
T
2
n .
V. DESIGN OF 4-PSK STTC WITH 2 TRANSMIT ANTENNAS
The design of the fully balanced codes includes 2 steps:
First step: generation of all the bases of Z
2
4
.
Second step: permutation of the column vectors of each
obtained base to generate all the fully balanced codes.
A. Properties of the bases of Z
2
4
In the following, the properties of the bases of Z
2
4
, which
characterize the fully balanced codes are listed. Further, by
using these properties, we can design all the base of Z
2
4
, then
the fully balanced codes.
Property 1: One base of Z
2
4
contains dim
_
Z
2
4
_
= 4 vectors.
Property 2: The null vector 0 Z
2
4
can not be used to form
a base.
Property 3: If the vector C
i
Z
2
4
is used to form a base,
then C
j
= C
i
does not belong to this base.
Property 4: If the vectors C
1
, C
2
, . . . , C
m
Z
2
4
with
m < 4 are linearly independant then the vectors
C
m+1
=
m

i=1

i
C
i
Z
2
4
,
i
{1, 0, 1} (11)
can not be used to obtain m+1 linearly independent vectors.
Property 5: C
0
= 2Z
2
2
is a normal subgroup of the additive
group Z
2
4
. For each element v Z
2
2
we consider the coset
C
v
= v + 2Z
2
2
, where addition is in Z
2
4
. Then,Z
2
4
is divided
into 4 cosets.
Property 6: If u
1
C
u
and v
1
C
v
then u
1
+ v
1
C
uv
where represents the addition in Z
2
2
.
Property 7: If u
1
C
u
then u
1
+C
v
= C
uv
.
Property 8: u +C
u
= C
0
= 2Z
2
2
.
Property 9: The sum of two cosets is dened by
C
u
+C
v
= {u
1
+ v
1
/u
1
C
u
and v
1
C
v
} (12)
Property 10: The direct sum of two cosets is a coset:
C
u
+C
v
= C
uv
Property 11: If u
1
C
u
then u
1
C
u
.
Property 12: If p
1
, p
2
Z
2
2
\ {0} and p
1
= p
2
, then
span(2p
1
, 2p
2
) = 2Z
2
2
= C
0
.
Property 13: If p
i
Z
2
2
then the sum of 2 different elements
of the coset C
p
i
is an element of C
0
\{2p
i
}, for i = 1, 2, 3.
Property 14: One base of Z
2
4
contains at least one vectors
in C
0
.
Property 15: One base of Z
2
4
contains at most 2 vectors in
the same coset.
Property 16: One base of Z
2
4
always contains 2 vectors
which belong to 2 cosets C
p
1
, C
p
2
different from C
0
such as the
vectors p
1
, p
2
are linearly independent. Similarly, the cosets
C
p
1
, C
p
2
are called linearly independent.
h
a
l
-
0
0
1
7
2
7
3
3
,

v
e
r
s
i
o
n

1

-

2
0

S
e
p

2
0
0
7
One base of Z
2
4
contains at least one vector in C
0
and at
most 2 vectors in C
0
(Property 14, 15). Then, there are two
types of fully balanced codes of Z
2
4
:
Type I codes which contain only 1 non-null vector in C
0
Type II codes which contain 2 non-null vectors in C
0
B. Design of fully balanced codes of type I
A base of a fully balanced code of type I contains only
one vector 2p
i
C
0
\{0} = C

0
. A second vector 2p
j
C

0
is
obtained as the sum of 2 different vectors of a coset C
p
j
: there
are 3 possibilities to choose this coset. Due to the structure
of a coset C
p
j
= {p
j
, p
j
, q
j
, q
j
}, there are 4 different
ways to choose these 2 vectors :{p
j
, q
j
}, {p
j
, q
j
}, {p
j
, q
j
},
{p
j
, q
j
}. The last vector can be chosen in a different
coset, so there are 2 4 = 8 possibilities. Therefore, there
are 3 4 8 = 96 different bases of type I in Z
2
4
.
C. Design of fully balanced codes of type II
A base of a fully balanced coset code of type II contains 2
different vectors in C

0
= 2Z
2
2
\{0}. These 2 vectors generate
C
0
= 2Z
2
2
. Therefore, it is sufcient to choose a vector u
1
in
a coset C
u
to obtain the whole coset C
u
. In the same way, if
we choose another vector v
1
in a coset C
v
, the whole coset
C
v
is also generated. Moreover, u
1
+ v
1
C
uv
. Therefore,
the sum of the vector u
1
+ v
1
with the coset C
0
gives C
uv
.
Hence, all the vectors of Z
2
4
are generated.
The algorithm to obtain a base of type II is as follows:
1) Choose 2 different vectors in C

0
(3 possibilities);
2) Choose 2 different cosets C
u
and C
v
different from C
0
(3 possibilities);
3) For each of these 2 cosets, choose one representative
(4 possibilities for each coset).
Therefore, there are 3 3 4 4 = 144 bases of type II in Z
2
4
.
Finally, the total number of the bases of Z
2
4
is: 96+144 = 240
bases.
VI. CODE PERFORMANCE
Before showing all the best codes based on the trace
criterion, we propose herein some trace properties of 4-state
4-PSK STTC:
Property P
1
: the codes C =
_
C
1
C
2
C
3
C
4
_
,
C

=
_
C
1
C
2
C
3
C
4
_
and C

=
_
C
1
C
2
C
3
C
4
_
have
the same minimum trace value.
Property P
2
: the codes C=
_
C
1
C
2
C
3
C
4
_
and
C

=
_
C
2
C
1
C
3
C
4
_
have the same minimum trace
value.
Property P
3
: the codes C=
_
C
1
C
2
C
3
C
4
_
and
C

=
_
C
3
C
4
C
1
C
2
_
achieve the same minimum trace
value.
Property P
4
: the same minimum trace value is obtained
by using a permutation between the rows of the generator
matrix C, i.e, a permutation between the indices of the
transmit antennas.
TABLE I
STRUCTURE OF 4-STATE 4-PSK FULLY BALANCED STTC WITH 2
TRANSMIT ANTENNAS AND min (TR(A)) = 10
C =
_
2 0 2 1
1 2 0 2
_
C

=
_
0 2 2
C
4
2 1 2
_
C
4

__
1
1
_
,
_
1
3
__
An exhaustive computer search is carried out to detect all
the 4-state 4-PSK STTC with 2 transmit antennas that achieve
the maximum rank and the maximum trace. A set of 80 codes
with min (rank(A)) = 2 and min (tr(A)) = 10 is found. All
these codes offer a minimum product distance d
2
p
= 4 6 = 24
which is the best product distance that can be achieved for
4-state 4-PSK STTC with 2 transmit antennas. Besides, it can
be observed that all these codes are fully balanced STTC of
type II. Since the minimum trace value has been obtained, all
these 80 codes can be generated by using Table I with the
trace properties P
1
to P
4
. As an example, if we choose from
Table I the generator matrix C, by using the properties P
1
to
P
3
, we obtain the generator matrix
_
0 2 1 2
2 3 2 0
_
(in bold in
Tab. II) proposed by Chen [2].
The distance spectrum of all these 80 codes has been also
computed for L
f
= 3. Among them, there are 16 codes
which achieve 10 as the minimum euclidian distance with
the multiplicity 4 whereas the other 64 vectors have 10 as
the minimum euclidian distance with the multiplicity 6. Then,
all these 16 4-state fully balanced codes which offer the best
performance over fast and slow Rayleigh fading channels with
two or more receive antennas is given in Tab. II below.
TABLE II
4-STATE 4-PSK FULLY BALANCED STTC WITH 2 TRANSMIT ANTENNAS
AND min (TR(A)) = 10
_
1 2 0 2
2 0 2 1
_ _
1 2 0 2
2 0 2 3
_ _
3 2 0 2
2 0 2 1
_ _
3 2 0 2
2 0 2 3
_
_
2 1 2 0
0 2 1 2
_ _
2 1 2 0
0 2 3 2
_ _
2 3 2 0
0 2 1 2
_ _
2 3 2 0
0 2 3 2
_
_
2 0 2 1
1 2 0 2
_ _
2 0 2 3
1 2 0 2
_ _
2 0 2 1
3 2 0 2
_ _
2 0 2 3
3 2 0 2
_
_
0 2 1 2
2 1 2 0
_ _
0 2 3 2
2 1 2 0
_ _
0 2 1 2
2 3 2 0
_ _
0 2 3 2
2 3 2 0
_
In this table, all the codes are related due to the trace
properties presented before. The codes of the third line are
obtained from the codes of the rst line by permuting the
indices of the antennas, i.e. the lines of the generator matrix
C. In the same way, the codes of the fourth line are obtained
from the codes of the second line of the table.
In order to conrm the utility of the fully balanced STTC,
an exhaustive computer search of all 4-state STTC has also
been carried out. The obtained results conrm that Table II
contains all the best STTC. There are not other codes with
better performance than the codes given in this table.
For the 16-state 4-PSK STTC, the construction and the
research of the best codes are made from the 4-state 4-PSK
STTC by using Theorem 2. Table III herein contains all the
h
a
l
-
0
0
1
7
2
7
3
3
,

v
e
r
s
i
o
n

1

-

2
0

S
e
p

2
0
0
7
TABLE III
16-STATE 4-PSK FULLY BALANCED STTC WITH 2 TRANSMIT ANTENNAS
AND min (TR(A)) = 16
_
2 3 2 3 2 1
0 2 2 1 0 2
_ _
2 1 2 3 2 3
0 2 2 1 0 2
_ _
2 3 2 1 2 1
0 2 2 1 0 2
_ _
2 1 2 1 2 3
0 2 2 1 0 2
_
_
0 2 2 1 0 2
2 3 2 3 2 1
_ _
0 2 2 1 0 2
2 1 2 3 2 3
_ _
0 2 2 1 0 2
2 3 2 1 2 1
_ _
0 2 2 1 0 2
2 1 2 1 2 3
_
_
2 1 2 1 2 3
0 2 2 3 0 2
_ _
2 3 2 1 2 1
0 2 2 3 0 2
_ _
2 1 2 3 2 3
0 2 2 3 0 2
_ _
2 3 2 3 2 1
0 2 2 3 0 2
_
_
0 2 2 3 0 2
2 1 2 1 2 3
_ _
0 2 2 3 0 2
2 3 2 1 2 1
_ _
0 2 2 3 0 2
2 1 2 3 2 3
_ _
0 2 2 3 0 2
2 3 2 3 2 1
_
_
3 2 3 2 1 2
2 0 1 2 2 0
_ _
1 2 3 2 3 2
2 0 1 2 2 0
_ _
3 2 1 2 1 2
2 0 1 2 2 0
_ _
1 2 1 2 3 2
2 0 1 2 2 0
_
_
2 0 1 2 2 0
3 2 3 2 1 2
_ _
2 0 1 2 2 0
3 2 1 2 1 2
_ _
2 0 1 2 2 0
1 2 3 2 3 2
_ _
2 0 1 2 2 0
1 2 1 2 3 2
_
_
1 2 1 2 3 2
2 0 3 2 2 0
_ _
3 2 1 2 1 2
2 0 3 2 2 0
_ _
1 2 3 2 3 2
2 0 3 2 2 0
_ _
3 2 3 2 1 2
2 0 3 2 2 0
_
_
2 0 3 2 2 0
1 2 1 2 3 2
_ _
2 0 3 2 2 0
3 2 1 2 1 2
_ _
2 0 3 2 2 0
1 2 3 2 3 2
_ _
2 0 3 2 2 0
3 2 3 2 1 2
_
16-state Balanced STTC which offer the best performance
over fast and slow Rayleigh fading channels with two or more
receive antennas. All these codes have min (tr(A)) = 16 and
offer a minimum product distance d
2
p
= 128. Among them,
we found the code proposed by Chen
_
1 2 1 2 3 2
2 0 3 2 2 0
_
(in bold in Tab. III).
Finally, the performance of all these 4-state and 16-state
codes for 2 transmit antennas is evaluated by simulation with
1 and 2 receive antennas over fast Rayleigh fading channels.
Each simulated frame consists of 128 symbols transmitted
from each transmit antennas. The Frame Error Rate (FER)
performance are shown in Fig. 2.
Note that all the codes given in Tab. II achieve the same
performance, i.e. the performance of the Chens code [2] : full
rank, min (tr(A)) = 10, d
2
p
= 24. These codes outperform the
codes proposed in [7].
4 6 8 10 12 14 16 18
10
3
10
2
10
1
10
0
SNR(dB)
F
r
a
m
e

E
r
r
o
r

R
a
t
e


4state 2Tx1Rx
16state 2Tx1Rx
4state 2Tx2Rx
16state 2Tx2Rx
Fig. 2. FER performance of the 4-PSK STTC with two transmit antennas
VII. CONCLUSION
In this paper, a new class of 4-PSK STTC for two transmit
antennas has been proposed. These codes generate the points
of the constellation with the same probability. It has been
shown that the best STTC belong to this class. Therefore,
the systematic search for good codes can be drastically
reduced to this class. The design of these balanced codes
has been described. A complete list of the best 4-state codes
and several 16-state codes for 2 transmit antennas have also
been given. All the fully balanced STTC listed in this paper
are equivalent, i.e. they have the same rank, trace, product
distance and distance spectrum. The simulation results have
shown that they outperform the other STTC for 2 transmit
antennas.
REFERENCES
[1] V. Tarokh, N. Seshadri, and A. Calderbank, Space-time codes for
high data rate wireless communication: Performance criterion and code
construction, IEEE Trans. Inform. Theory, vol. 44, no. 2, pp. 744765,
March 1998.
[2] Z. Chen, J. Yuan, and B. Vucetic, Improved space-time trellis coded
modulation scheme on slow fading channels, Electron. Lett., vol. 37,
no. 7, pp. 440441, March 2001.
[3] T. Liew and L. Hanzo, Space-time codes and concatenated channel codes
for wireless communications, Proceedings of the IEEE, vol. 90, no. 2,
pp. 187219, Feb 2002.
[4] W. Firmanto, B. Vucetic, and J. Yuan, Space-time TCM with improved
performance on fast fading channels, IEEE Commun. Lett., vol. 5, no. 4,
pp. 154156, April 2001.
[5] S. Baro, G. Bauch, and A. Hansmann, Improved codes for space-time
trellis-coded modulation, IEEE Commun. Lett., vol. 4, no. 1, pp. 2022,
Jan. 2000.
[6] D. Ionescu, K. Mukkavilli, Y. Zhiyuan, and J. Lilleberg, Improved 8- and
16-state space-time codes for 4PSK with two transmit antennas, IEEE
Commun. Lett., vol. 5, no. 7, pp. 301303, July 2001.
[7] Y. Hong and A. G. i Fabregas, New Space-Time Trellis Codes for Slow
Fading Channels, in Proc. IEEE VTC 2006-Spring, vol. 3, May 2006,
pp. 1492 1496.
[8] B. Rassool, F. Heliot, L. Revelly, M. Dohler, R. Nakhai, and H. Aghvami,
Fast search techniques for obtaining space-time trellis codes for Rayleigh
fading channels and its performance in CDMA systems, in Proc. IEEE
VTC 2003-Spring, vol. 1, April 2003, pp. 6669.
[9] Y. Jung and J. Lee, Improved space-time trellis codes using expanded
signal set in slow Rayleigh fading channels, in Proc. IEEE ISIT2002,
June 2002, p. 248.
h
a
l
-
0
0
1
7
2
7
3
3
,

v
e
r
s
i
o
n

1

-

2
0

S
e
p

2
0
0
7