Professional Documents
Culture Documents
1 INTRODUCTION
n past years, there have been studies about a directsequence code-division multiple-access (DS/CDMA)
communication system which employs a code-diversity
scheme as a method for suppressing multiple access interference (MAI). This is known as DS/CDMA codediversity. The original concept of a code-diversity system
is based on the use of a sum of several PN sequences as
the spreading codes for users [1]. This approach enables
diversity reception by employing several receiving
branches matched to each one of the individual PN sequences which compose the spreading code. Codediversity has been shown to be effective in suppressing
multiple access interference, especially under a near-far
problem environment, that is, when the received signal
power of any of the undesired users is much larger than
the desired users.
After the proposal of the original method, codediversity was simulated under multipath fading channels,
and showed to be effective still [2]. Afterward, a constant
amplitude composite sequence code-diversity method
was introduced, which simplified the original system by
implementing diversity reception using multivalued orthogonal cyclic shifted M (OM) sequences at the receiver
[3]. More recently, two-valued digitally-constructed
branch sequences with constant composite value have
been proposed. The application of these sequences has
the benefit of simplifying hardware implementation [4].
As with any diversity method, the calculation of appropriate weight coefficients for branch combining is critical
to the systems bit error rate (BER) performance. The original code-diversity proposes a calculation method based
on maximal ratio combining (MRC), in which each weight
coefficient is calculated as the division of the root mean
square (RMS) level of the desired signal component, by
the power of the interference-plus-noise component at
each branch [1]. Seeking further performance improvement, an adaptive weight control (AWC) method based
on the least mean squares (LMS) algorithm was proposed
in [5]. This method showed notable improvement over
the original one, nonetheless, requiring increased complexity and a considerably longer training sequence.
Motivated by the question of optimal weight coefficients for code-diversity, this paper proposes a calculation
method based on optimum combining theory, whereby
the weight coefficient vector that maximizes output signal-to-interference plus noise ratio (SNIR) can be found
by solving a linear system involving an estimated interference-plus-noise correlation matrix [6]. This approach is
based in noting that the output SINR of a DS/CDMA
code-diversity system can be written in the form of a generalized Rayleigh quotient, and therefore has the associated optimal solution presented in [7].
In Sect. 2, we present a baseband model of a
DS/CDMA code-diversity communication system. Furthermore, we introduce the concept of orthogonal cyclic
shifted M (OM) sequences and both weight coefficient
calculation methods proposed so far for code-diversity:
MRC by anti-cross correlation and adaptive weight control. In Sect. 3, we derive an expression for the signal-tointerference plus noise ratio (SNIR) and present the equation that maximizes it, according to optimum combining
theory. In Sect. 4, a direct method to calculate weight coefficients is proposed based on optimum combining. The
method requires two steps; estimating the interference-
plus-noise correlation matrix at the receiver and calculating the optimum weigh coefficients by solving a linear
system. In Sect. 5, we present numerical results of a performance evaluation based on the simulation of its average bit error rate. Finally in Sect. 6, we present the conclusions.
2 SYSTEM MODEL
2.1 OM Sequences
The branch sequences are core elements of the codediversity method. In this model, orthogonal cyclic shifted
M sequences are used. These are a set of multilevel sequences especially devised to make constant amplitude
code-diversity possible [3]. A set of base OM sequences is
generated by first building an NN matrix that contains
all the shifted versions of a maximal length (M) sequence
of code length N, and then adding and multiplying an
orthogonalization component and a normalization factor,
respectively.
Given an M sequence of code length N, the n-th chip of
the k-th OM sequence, that is omn(k) (n = 0, 1, 2, , N-1; k =
1, 2, , N), is given by the following expression
omn
(k )
= ( M n+k 1 + ) ,
(1)
1+ 1+ N
,
N
(2)
1
.
N +1
(3)
and
gk =
,
E [ I ] + E [ N k2 ]
(4)
2
k
where
is the RMS value of the desired signal component, E[Ik2] is the interference power, and E[Nk2] is the
noise power; all at the output of the k-th branch correlator.
It is worth noting that the power levels are calculated as
probabilistic averages over the random variables of noise
and data symbols.
Assuming that the desired user's power component at
the output of the k-th branch's correlator, E[Sk2], is a
known value,
can be calculated as
= E[ Sk2 ] ,
SNIR =
2
k
2
k
E[ I ] + E[ N ] = E[ Z ] E[ S ] ,
(7)
SNIR =
2
M
E g k Sk
k =1
E `[Sk Sk ' ]
k'
E [( N k + I k )( N k ' + I k ' )]
k =1 k =1
M
g g
( ss )
k ' kk '
g g
( ss )
k ' kk '
k =1 k =1
M M
R
R
k =1 k =1
g g
(6)
k'
k =1 k =1
g g
k
(5)
2
k
M M
M M
E g k g k ' Sk Sk '
k =1 k =1
gT R ssg
,
gT R nng
R11( ss )
( ss )
R
R ss = 21
(
R ss )
M 1
R12( ss ) R1(Mss )
( ss )
R22
R2( ssM) ,
( ss )
RM( ss2) RMM
(8)
2
M
(
)
E g k I k + N k
k =1k
The above powers averages are calculated as secondorder probabilistic averages (expectations) with respect to
data symbols and noise, since these are in practice slow
varying time functions that can be modeled as stationary
(10)
and
R11( nn )
( nn )
R
R ss = 21
R ( nn )
M 1
R12( nn ) R1(Mnn )
( nn )
)
R22
R2( nn
M
,
( nn )
RM( nn2) RMM
(11)
where
(12)
and
(9)
(13)
1 ~
g = R nn
V,
(14)
where
is an arbitrary constant, V is the propagation
vector, ~ denotes complex conjugate, and Rnn-1 denotes the
inverse of Rnn. In this system model, let = 1 and V = 1 =
(1, 1, , 1)T. Then (14) reduces to
1
g = R nn
1 .
(15)
4 A PROPOSED METHOD
Previous methods for calculating weight coefficients in
the code-diversity system were based on either, maximum ration combining (MRC) by anti-cross correlation,
or adaptive weight control (AWC). MRC by anti-cross
correlation is a simple and convenient method, yet, it
cannot properly maximize the SINR in a code-diversity
system as a consequence of multiple access interference
not beign independent among branches. Hence, MRC by
anti-cross correlation becomes sub-optimal in a
DS/CDMA code-diversity system.
On the other hand, adaptive weight control has shown
to achieve a notably better performance. In fact, optimum
combining is known to be achievable by using least mean
square (LMS) adaptive arrays, which are the core of the
AWC method. However, AWC has shown some drawbacks in the code-diversity system. First, it requires a
longer training sequence; around 5,000 and 200,000 bits
(as seen in [2] and [5]), whereas the original MRC method
requires only 1,000 bits [1]. Aditionally, because AWC is
an interative method, there is the issue of properly selecting the step size, which incorrectly selected causes the
algorithm no to converge.
Looking to overcome these drawbacks, in this section
we introduce a direct calculation method (non-iterative)
similar to MRC by anti-cross correlation, but based on
optimum combining.
(16)
(17)
and thus
(18)
sider that from (18), (10), (11), (12), and (13), it should be
clear that the following extended matrix relation holds
R nn = R zz R ss ,
(19)
R11( zz )
( zz )
R
R ss = 21
R ( zz )
M 1
R12( zz ) R1(Mzz )
( zz )
R22
R2( zzM)
,
( zz )
RM( zz2) RMM
(20)
where
(21)
1 K
T
zi zi
K i =1
R zz =
(22)
Sk = DATA(t ) PN 2 (t ) OM k (t )dt.
0
(23)
P b0 OM k (t )dt ,
(24)
Sk = Pb0 OM k (t )dt ,
0
1 T
= PTb0 OM k (t )dt ,
0
T
= PTb0ok ,
(25)
E [ Sk Sk ' ] = E
[(
PTb0ok
2
[ ]
)(
)]
PTb0ok ,
= PT E b0 ok ok ,
= PT 2ok ok .
(26)
R ss = PT 2OOT ,
(27)
where O is defined as the column vector holding the values o1 , o2 ,..., oM . The superscript T denotes the transpose
of a matrix/vector.
From (19), (22), and (27), the interference-plus-noise
correlation matrix can be expressed as
R nn =
1 K
T
zi zi PT 2OOT .
K i =1
(28)
R nng = 1.
(29)
5 NUMERICAL RESULTS
This section shows the results of computer simulations
carried out in order to evaluate the performance of the
proposed method. The simulation models a near-far problem environment, where the user's signal is degraded by
AWGN and a single interfering user (multiple access interference) with greater received power (SIR = -10dB).
Gold codes are used as PN sequences for both user and
interference. The asynchronous CDMA environment is
modeled using asynchronous bit timing and synchronous
chip timing. Interference delay takes all the integer values
from from 0 to 30 chips (code lenght is 31). Code-diversity
employs five receiving branches, each with one composite
OM sequence built using the order allocation algorithm.
Weight coefficients are calculated with MRC by anti-cross
correlation, adaptive weight control (AWC), and the proposed method. Further details are shown in Table 1.
Table 1 Simulation specifications
Modulation system
BPSK-DS/SS
Number of user
PN Sequences
Gold sequences
Preferred pair M[45], M[75]
G0 and G3
Code length
31
Transmission line
Eb/N0
0 ~ 8 dB
Signal-to-interference Ratio
(SIR)
-10 dB
Asynchronous
Synchronous
Interference delay
0 ~ 30 chips
Number of branches
Branch sequences
OM sequences (M[67])
Branch composition
Order allocation
Training bits
10-2
6 CONCLUSIONS
In this paper, we proposed a weight coefficient calculation method for DS/CDMA code-diversity based on optimum combining theory and evaluated its performance
through computer simulations. The results showed us
that the difference between AWC and the proposed
method is nearly negligible, hence having almost the
same performance. This result is consistent with optimum
combining theory [6], given that an optimum combiner
can also be implemented using an LMS adaptive array;
the core of adaptive weight control. Using the proposed
method, however, has two advantages over AWC. First, it
is a direct (non-iterative) calculation method not subject
to the convergence issues of numerical algorithms. Second, it needs a far smaller training sequence (only 1,000
bits). Its main drawback is requiring an estimate for the
user's received power level at the receiver, something that
is not necessary with AWC and a shared characteristic
with MRC by anti-cross correlation. The overhead caused
by finding the solution to the linear system could be of
concern, but it'll be negligible compared to the training
time if the hardware is capable of performing many floating point operations per bit period T.
REFERENCES
[1] T. Seki, M. Hamamura and S. Tachikawa, Suppression effects of multiple access interference in
DS/CDMA with code-diversity, IEICE Trans.
Fundamentals, vol. E82-A, no. 12, pp.2720-2727, Dec.
1999.
[2] R. Manzanilla, M. Hamamura and S. Tachikawa, Sequence suppression characteristics of code-diversity
DS/CDMA over multipath fading channels, IEICE
Trans. Fundamentals, vol. E84-A, no. 12, pp. 2983-2990,
Dec. 2001.
[3] S. Hasebe and S. Tachikawa, DS/SS code-diversity
communication system using a constant amplitude
composite sequence, IEICE Trans. Fundamentals, vol.
J90-A, no. 1, pp. 44-53, Jan. 2007.
[4] H. Noguchi and S. Tachikawa, A novel spreading
sequence and its performances for code-diversity in
DS/CDMA communication systems, Proc. Int. Symposium on Inf. Theory and its Applications (ISITA), pp.
347-351, Oct. 2012.
[5] S. Tachikawa and M. Takekawa, DS/CDMA codediversity communication system using adaptive
weight control, Proc. 14th Asia-Pacific Conf. on Commun. (APCC2008), Oct. 2008.
[6] J.H. Winters, Optimum combining in digital mobile
radio with cochannel interference, IEEE Journal on
Selected Areas in Commun., vol. SAC-2, no.4, pp. 528539, Jul. 1984.
[7] C. Baird and C. Zahm, Performance criteria for narrowband array processing, Proc. IEEE Conf. on Decision and Control, vol. 10, 1971.
[8] M. Pursley, Performance evaluation for phase-coded
spread-spectrum multiple-access communication
part I: system analysis, IEEE Trans. on Commun., vol.