A New Method to Design CDMA Spreading Sequences
Transcript of A New Method to Design CDMA Spreading Sequences
-
8/8/2019 A New Method to Design CDMA Spreading Sequences
1/4
A New Method to Design CDMA Spreading Sequences
*
Bi Jianxin Wang Yingmin Tian Hongxin Yi Kechu
(National Key Lab. On ISN, Xidian University, Xian 710071, China)
*
This paper is supported by National Natural Science Funds (N0. 69872027).
Abstract: It is well-known that multi-value CDMA
spreading codes can be designed by means of a pair of
mirror multi-rate filter banks based on some
optimizing criterion. This paper indicates that there
exists a theoretical bound in the performance of its
circulate correlation property, which is given by an
explicit expression. Based on this analysis, a criterion
of maximizing entropy is proposed to design such
codes. Computer simulation result suggests that the
resulted codes outperform the conventional binary
balance Gold codes for an asynchronous CDMA
system.
Key words: Multi-rate unitary filter banks; CDMA
spreading code; Maximum entropy criteria.
. INTRODUCTION
The selection of spreading codes is essential for
asynchronous CDMA systems. Recently, it is believed
to be promising that the CDMA spreading code is
designed based on multi-rate filter bank theory. For
this purpose, Akansu[1] has converted the design of
CDMA spreading code into an optimizing problem.
Simulations suggested that the resulted multi-valued
codes outperform Gold codes. Qinghua Shi[2], et al.,
give another scheme, from frequency domain, for
spreading code generation based on multi-rate filter
bank theory and a lattice structure. This paper
indicates that there exists a theoretical bound in the
performance of spreading codes circulate correlation
property, and gives an explicit expression for it. Based
on this analysis, a criterion of maximizing entropy is
proposed to design such codes. Computer simulation
result suggests that the resulted codes behave better
than the conventional binary balance Gold codes for
an asynchronous CDMA system.
. TWO-BAND MULTIRATE UNITARY
FILTER-BANK THEORY(TBFMFB)
The coefficients of two-band unitary filter-banks
only depend on )n(h0 , which satisfies[3]
=+n
00 )k()k2n(h)n(h (1)
From the viewpoint of the frequency domain, the
parameterization representation of the two-band
unitary is given as follows:
=
111JJ
1
0
z
1)(ZRZ)(ZR)(R
)z(H
)z(HL ( 2/NJ = ) (2)
=
cossinsincos)(R
= 2z0
01Z
So, we can use 2/N parameters { }J21 ,,, L to
represent )n(h0 .
. The property of circulate correlation of the
codes based on TBFMFB
From definition of circulate correlation, it is evident
that
1
=
=+
= )1N,,5,3,1i()i(R
)2N,,4,2,0i()i(R
)i(Ra
a
b L
L
(3)
2 0)i(Rab = )2N,,4,2,0i( = L (4)
where N-length normalized vector a and vector b
are the coefficients of the analysis filter bank, and
)i(Ra denotes i-shifting circulate auto-correlation of
vector a and )i(Rab denotes i-shifting circulate cross-
correlation.
Theorem 1 If N-length b (N= u2 , Iu ) is the
-
8/8/2019 A New Method to Design CDMA Spreading Sequences
2/4
mirror of a normalized vector a , then, the following
inequality is true:
1)i(R)i(R1N
0i
2ab
1N
1i
2a +
=
=
)Zi,1Ni0( (5)
where the equality is valid if and only if)2N,,4,2k(0)k(Ra == L .
Proof: Let A and B denote two circular matrices:
1N210 aaaaA = L
1N210 bbbbB = L
where ia and ib represent two column vectors
obtained respectively by circular--shifting vector a
and b for i elements. Let
y= +
=
=
1N
0i
2ab
1N
0i
2a )i(R)i(R
=22
aBaA + = TTT a)BBAA(a + (6)
Making use of the properties of circular matrix, we
have [4]
1N22
1NN1 PaPaPaIaA
++++= L
1N22
1NN1 PbPbPbIbB
++++= L (7)
where
=
00
10
0
0
01
00
0000000
00
1
0
00
10
L
L
MMM
L
MM
L
L
P
Substituting (7) into (6) and using iNTi P)P( = , we
have
y= )PaPaPaIa)(PaPaPaIa{(a 1NN23211N221NN1 ++++++++ LL
+ T1NN23211N221NN1 a)}PbPbPbIb)(PbPbPbIb( ++++++++ LL
(8)
Using the properties of matrix P ,we can simplify 8
into
y=2+ )1N(R)1N(R)1(R)1(R)1N(R)1(R baba2a2a +++++ LL
(9)
With Lemma1 we obtain
y=2+2[ 2a2
a2
a )2N(R)4(R)2(R ++ L ] (10)
So we have
1)2N(R2)4(R2)2(R21)i(R)i(R2
a2
a2
a
1N
0i
2ab
1N
1i
2a ++ +=+
=
=L
The equation is true if and only if
)2N,,4,2k(0)k(Ra == L .
It should be noticed that )0i()i(R 2a and2
ab )i(R
represent the normalized energies of all the possible
multi-path interfere and all the possible multi-access
interfere in the output of correlation receiver,
respectively. So the theorem implies :
(1)It provides the theoretical basis for designing
CDMA spreading code based on unitary filter bank.
When we design CDMA spreading code based on
TBFMFB, if we adopt the interfere energy minimum
criterion in the output of correlator, the optimization
solutions are the coefficients of unitary filter banks
(2)It gives us the lowest bound of the sum of all
possible interfere energy in the output of correlator.
That is, if we use a pair of CDMA spreading codes
based on TBFMFB for user codes, the sum of all
possible interfere energy in the output of correlator is,
on the condition of equal power transmission, greater
than or equal to its signal energy. Moreover, theequality is valid if and only if the codes are resulted
from a unitary filer bank.
. a Proposed Method for Designing CDMA
Spreading Codes Based on Unitary Filter Bank
It is of significance that we adopt the minimum of
the maximum of side lobes of absolute value of
circulate auto-correlation and cross-correlation as an
objection function. In another viewpoint, the flatter
side lobes of absolute value of circulate auto-
correlation and cross-correlation are, the better the
performance of the corresponding asynchronous
CDMA systems will be. So,we use a criterion of
maximizing entropy. The entropy of a vector
[ ]N1 p,,pp L= =
2
ab
2
ab
2
ab
2
a
2
a
2
a )1N(R,,)3(R,)1(R,)1N(R,,)3(R,)1(R LL
-
8/8/2019 A New Method to Design CDMA Spreading Sequences
3/4
is defined as
)p(logp)p(logp)p(entropy N2N121 ++= L (11)
Furthermore, we also search the spreading code from
the viewpoint of the frequency domain (Z-domain),
and get the same merits as reference [2]. It is obvious
that p is the function of J21 ,,, L and )p(entropy
is also the function of J21 ,,, L . So, the objection
function we use for optimizing is
{ }[ ]),,,p(entropymaxarg),,,( J21,,,
J21
J21
=
LL
L
(12)
To make )n(h0 suitable for DS-CDMA systems, we
must spread )n(h0 in both time and frequency as
evenly as possible. So, by using the entropy function
defined above, we add two constrains as follows
1T c)aa(entropy
22T c))a(FFT/)a(FFT)a(FFT(entropy
where ( )aFFT denotes the Fourier transform of vector
a . In experiment, these parameters
are: 46.4cc 21 == , 32N = . The experiment results are
showed in fig.1,2, 3, and 4. Gold codes of length-31
are used for comparison. The Gold codes result fromsumming (mod 2) a pair of m-sequences. The
preferred pair of primitive polynomial is 45 and 67.
Fig.1 compares the spectra of a 32-length optimized
code for the two-user case with a 31-length balanced
Gold code. The circulate auto-correlation and
circulate cross-correlation of these codes are also
displayed in Fig.2 and 3. It is obvious from these
figures that the circulate auto-correlation, circulate
cross-correlation and frequency properties of the
multi-valued optimized code outperform the
comparable duration binary balanced Gold code.
. PERFORMANCE OF THE PROPOSED
MULTIVALUED CDMA SPREADING CODES
By using Monte Carlo computer simulation method,
the bit error rate (BER) performance of a two-user
asynchronous CDMA communication system based
on the user code type presented above are investigated
in this section. Antipodal signaling and BPSK
modulation for CDMA are used in the simulations.
The channel noise is modeled by additive white
gaussian noise (AWGN).
In the process of evaluating the BER performance,
it is assumed that for a user there are a multipath
interfere and a multiuser interfere, whose delays are
random variables evenly distributed in
[ ]cc T)1N(T0 L , where cT denotes the duration
of a chip. Because circulate correlation properties of
the optimized codes outperform those of the balanced
Gold code, it is expected that the BER performance of
the optimized codes are also better than the latter. The
result showed in Fig.4 also substantiates it.
. CONCLUDING REMARKS
Multi-rate unitary filter bank provides us powerful
mathematical tool for designing CDMA spreading
code. This paper provides the theoretical basis for
designing CDMA spreading code based on unitary
filter bank, and gives us a lower bound of the sum of
all possible interfere energy in the output of correlator.
Based on this analysis, a criterion of maximizingentropy is proposed to design such codes. It is worth
while to further investigate how to extend this method
to more generalized case where user number K is
more than 2.
REFERENCES
[1] Ali N.Akansu, M.V.Tazebay and R.A.Haddad, a newlook at digital orthogonal transmultiplexers for CDMA
communications,IEEE trans. On SP-45, No.1, January
1997.
[2] Qinghua Shi and Shixin Cheng, optimal spreadingsequence design based on PR-QMF theory,
ELECTRONICS LETTERS, 18th March 1999, Vol.35,
No.6.
[3] C.S.Burrus,R.A.Gopinath,H.Guo, introduction towavelets and wavelets transforms---a primer, prentill
hall,1998.
-
8/8/2019 A New Method to Design CDMA Spreading Sequences
4/4
[4] 1999 6
[5] Verdu Sergio. Multiuser Detection. Cambridge UniversityPress, 1998.
Fig.1 Frequency spectra of the optimized 32-length Fig.2 Circulate auto-correlation functions of the
codes and 31-length balanced Gold codes. optimized 32-length codes and 31-length balanced
Gold codes.
Fig.3 Circulate cross-correlation functions of Fig.4 BER performance of two-user asynchronous CDMA
the optimized 32-length codes and 31-length system for different user code types
balanced Gold codes.
0 0.5 1 1.5 2 2.5 30
0.5
1
1.5
2
31-length balanced Gold code
frequency(rad)
magnitude
function
0 0.5 1 1.5 2 2.5 30
0.5
1
1.5
2
32-length optimized code
frequency(rad)
magnitudefunction
5 10 15 20 25 30-0.5
0
0.5
1
31-length balanced Gold code
time shiftcirculateauto-correlationfunction
5 10 15 20 25 30-0.5
0
0.5
1
32-length optimized code
time shiftcirculate
auto-correlationfunction
5 10 15 20 25 30-0.4
-0.2
0
0.2
0.4
31-length balanced Gold code
time shiftcirculate
cross-correlationfunction
5 10 15 20 25 30-0.4
-0.2
0
0.2
0.4
32-length optimized code
circulatecros
s-correlationfunction
time shift
-2 - 1 0 1 2 3 4 5 6 7 81 0
- 4
1 0- 3
1 0- 2
1 0- 1
1 00
SNR in dB
errorrate
2 - u s e r a s y n c h r o n o u s C D M A r e c e i v e r
s ing le us e r
3 1 - l e n g t h b a l a n c e d G o l d c o d e
3 2 - l e n g t h o p t i m i z e d c o d e