SPREADING CODES IN CDMA

DETECTION
by
Ayse Kortun
Submitted to
Dr.Aykut Hocann
in partial fulllment of the
requirements for the course of
EE-574 Detection & Estimation
Theory
Eastern Mediterranean University
July 2003
Contents
1 Introduction 1
1.1 Problem Denition . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Spreading Codes 4
2.1 PSEUDONOISE(PN) SEQUENCE . . . . . . . . . . . . . . . . . 4
2.2 GOLD SEQUENCES . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 KASAMI SEQUENCES . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 HADAMARD-WALSH(ORTHOGONAL) CODES . . . . . . . . 9
2.5 VARIABLE-LENGTH ORTHOGONAL CODES . . . . . . . . . 10
2.6 MULTIPLE SPREADING CODE ALLOCATION . . . . . . . . 11
3 Simulation Results 13
4 Conclusion 17
5 References 19
1
List of Figures
1.1 The baseband model of a DS-CDMA system. . . . . . . . . . . . 2
2.1 Feedback Shift Register. . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Autocorrelation function for a maximum-length sequence with
chip duration T
c
and period NT
c
. . . . . . . . . . . . . . . . . . . 6
2.3 Generation of a Gold code set. . . . . . . . . . . . . . . . . . . . 7
2.4 Generation of Kasami set . . . . . . . . . . . . . . . . . . . . . . 8
2.5 Code tree for generation of variable length orthogonal codes (SF:
spreading factor) . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.6 Application of Walsh and PN codes in the forward link of CDMA 11
2.7 Application of Walsh and PN codes in the reverse link of CDMA 12
3.1 Correlation values of [7 1]PN-code and [7 3]-[7 3 2 1]Gold code.
N=127. Initial contents of ip-ops:1000000 . . . . . . . . . . . . 14
3.2 Correlation values of [7 1]PN-code and [7 3]-[7 3 2 1]Gold code.
N=127. Initial contents of ip-ops:1000000 . . . . . . . . . . . . 15
3.3 Correlation values of PN-codes chosen from the sets of [7 1]and[7
6 5 4]. N=127. Initial contents of ip-ops:1000000 . . . . . . . . 15
3.4 Correlation values of Kasami codes. N=127. Initial contents of
ip-ops:1111 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2
Abstract
This study deals with the main characteristics of the Maximal Length, Gold,
Kasami, Walsh and variable-length orthogonal codes and their functions in
code-division multiple access(CDMA) networks.
The important properties of the sequences are examined. A method of multiple
spreading for channelization and scrambling in CDMA is described. Auto and
cross correlation properties of some codes are mentioned according to obtained
graphs. Also, importance of spreading codes on CDMA detection is mentioned.
The main focus of this study is to emphasize the importance of code properties
for detection. In other words, the eect of codes on DS-CDMA performance.
Chapter 1
Introduction
Direct Sequence Code Division Multiple Access(DS-CDMA) is a method of
multiplexing users by distinct codes and in this method all users use the same
bandwidth. In DS-CDMA, each user has its own spreading code. The selection
of a good code is important, because auto-correlation properties and length of
the code sets bound on the system capacity.
The code sets can be divided into two classes: orthogonal codes and non-
orthogonal codes. Walsh sequences fall in the rst category, while the other
sequences(PN,Gold and Kasami)are shift-register sequences. When the user
codes are orthogonal, the output of the correlator in the receiver is zero except
the desired sequence.
In synchronous DS-CDMA systems the code sequence in the receiver is exactly
same with that in the transmitter, so there is no time shift between the users.
When the user codes are orthogonal in the synchronized systems, there is no
multiple access interference between the users after despreading. In practice, it
is dicult to realize synchronized DS-CDMA systems and time shifts between
users decrease the systems capacity.
The most important measures that specify the codes performance are, the or-
derly low values of cross-correlation between codes and the rate of eect of
auto-correlation values from time shifts.
In this study, fundamental properties of auto and cross-correlation of some im-
portant codes have been examined in case of asynchronous situation. Also the
eect of code properties on DS-CDMA systems analysis and multiple spreading
technique for channelization and scrambling in CDMA and are focused. Fi-
nally, in the last section, the importance of codes for the CDMA performance
is discussed.
1.1 Problem Denition
In multi-user systems, the main reason that aects the performance is the mul-
tiple access interference. Especially in an occasion where users are mobile,
asynchronous and power imbalance problems emerges among the users. Be-
cause of this reason, the selection of spreading codes to dierentiate the users
plays an important role in the system capacity.
1
In the DS-CDMA technique, each bits of the users data are multiplied with a
code in the transmitter. The code sequence used in transmitter performs the
role of spreading code. The baseband model of a DS-CDMA system is shown in
gure 1.1. Let b
k
denote a binary data sequence, and c
k
denote a code sequence.
The waveforms b(t) and c(t) denote polar representations in terms of two levels
as 1. By multiplying the information bit by the code, each information bit is
divided into a small time increments that are called chips. The received data
r(t) consist of transmitted signal m(t) plus an interference denoted by i(t). To
recover the original signal b(t), the received signal r(t) is multiplied with the
code sequence that used in transmitter then passed through the low-pass lter.
Finally a decision is made by the receiver. As an example, we can consider an
r(t) b(t)
c(t)
m(t)
+ m(t)
i(t)
(a)
(b)
x
LPF
Decision
Device
r(t)
c(t)
z(t) v
Threshold=0
Say 1 if v > 0
Say -1 if v < 0
(c)
x
Figure 1.1: The baseband model of a DS-CDMA system.
simple CDMA system by ignoring the channel. Let us think the rst bits of
four users. By multiplying each bit with a (PN)code, users bits are represented
by seven chips(spreading) as shown below.
U
1
[1]: 1 C
1
: -1 -1 1 1 1 -1 1 -1 -1 1 1 1 -1 1
U
2
[1]:-1 C
2
: 1 -1 -1 1 1 1 -1 -1 1 1 -1 -1 -1 1
U
3
[1]: 1 C
3
: -1 1 -1 -1 1 1 1 -1 1 -1 -1 1 1 1
U
4
[1]:-1 C
4
: 1 -1 1 -1 -1 1 1 -1 1 -1 1 1 -1 -1

+
r[1]: -4 2 0 0 2 -2 2 received data
2
The received data r consists information of four users. To recover the origi-
nal bits of users from the received data, the received data should multiplied
with the code sequence in the receiver that is exactly same with that is used for
spreading the original data in transmitter(despreading). We assume that the
receiver operates in perfect synchronism with the transmitter. As a last step,
decision is made by comparing the results with a threshold value as shown be-
low.
r[1] * C
1
= 4-2+2+2+2=8 > 0 1
r[1] * C
2
= -4-2+2-2-2= -8 < 0 1
r[1] * C
3
= 4+2+2-2+2=8 > 0 1
r[1] * C
4
= -4-2-2-2+2= -8 < 0 1
As shown from the example above, in recovering the original data, the selection
of codes and their correlation with each other are of signicant importance. Also
perfect synchronism plays a very important role in the system performance.
In this study, dierent spreading codes have been examined as time shifts,
auto-correlation and cross-correlation functions points of view. The relation
of spreading codes with the direct sequence code-division multiple access(DS-
CDMA) analysis has been researched by taking into consideration the asyn-
chronous case.
3
Chapter 2
Spreading Codes
In DS-CDMA system, for despreading operation, the received data should mul-
tiplied with the same code in the receiver. So the other user codes in the
same frequency band must be uncorrelated with the desired user code. For
this reason the DS-CDMA codes have to be designed so as to posses very low
cross-correlation.
Auto-correlation shows the measure of similarity between the code and its
cyclic shifted copy. Because of this reason, the codes that have the best proper-
ties of auto-correlation have frequently been used in removing the asynchronous
in communication systems. The auto-correlation can be expressed as below[1].
R
c
(k) =
N

n=1
a
n
a
n+k
(2.1)
Cross-correlation is the measure of similarity between two dierent codes. In
other words cross-correlation describes the interference between codes a
n
and
b
n
[1].
R
c
(k) =
N

n=1
a
n
b
n+k
(2.2)
where a
n
and b
n
are the elements of two dierent codes and have period N.
2.1 PSEUDONOISE(PN) SEQUENCE
A pseudonoise(PN) sequence is a binary sequence of 1s and 0s and it is peri-
odic. Its some characteristics that are similar to random binary sequences(having
equal # of 0s and 1s),very low correlation between any two shifted version of
the sequence and low cross-correlation between any two sequences.
Pseudo-Random sequence is not random(deterministic) but it looks randomly
for the user who doesnt know the code. The larger the period of the PN spread-
ing code, will be more random binary wave and it is harder to detect it.
A PN sequence is generated by a feedback shift register which is diagrammed in
Fig 2.1. It consist of a shift register made up of m ip-ops and a logic circuit.
The ip-ops in the shift register are regulated by a single timing clock. Binary
sequences are shifted through the shift registers and the output of the various
4
Logic
1
2 m
Output
sequence
Clock
Flip-flop
Logic
Figure 2.1: Feedback Shift Register.
stages are logically combined and feedback as the input to the st stage.
The initial contents of the ip-ops determine the contents of the memory. The
generated PN sequence is determined by mainly three factors that are length
m of the shift register, ip-ops initial state and the feedback logic.
The number of possible states of the shift register is at most 2
m
for a m ip-
ops. So the generated PN sequence must become periodic with a period of at
most 2
m
.
When the feedback logic consists of exclusive-OR gates, the shift register is
called a linear and in such a case, the zero state is not permitted. Therefore
the period of a PN sequence produced by a linear m-stage shift register can
not exceed 2
m
1. When a sequence of period 2
m
1 generated, it is called a
maximal length(ML) sequence.
Maximal-length sequences satisfy the following three properties.
Balance Property: In each period of maximum-length sequence, the number
of 1s is always one more than the number of 0s. So there must be 2
m1
ones and 2
m1
1 zeros in a full period of the sequence.
Run Property: Here, the run represents a subsequence of identical sym-
bols(1s or 0s) within one period of the sequence. The length of this
subsequence is the length of the run. Among the runs of 1s and 0s
in each period of a maximum-length sequence, one-half the run of each
kind are of length one, one-fourth are length two, one-eighth are of length
three,etc. For a maximum-length sequence generated by a linear feedback
shift register of length m, the total number of runs is (N+1)/2 where
N=2
m
1.
Correlation Property: The autocorrelation function of a maximum-length
sequence is periodic, binary valued and has a period T=NT
c
where T
c
is chip duration. This property is called the correlation property. The
autocorrelation function is[2]
R
c
() =
_
1
N+1
NT
c
||, ||

1
N
, for the remainder of the period
5
1.0
NT
c
1/N
T
c
R
c
(t)
t
NT
c
-NT
c
Figure 2.2: Autocorrelation function for a maximum-length sequence with chip
duration T
c
and period NT
c
.
which is shown in gure 2.2. Since the correlation between dierent shifts of an
m-sequence is almost zero, they can be used as dierent codes with an excellent
correlation property.
As it is mentioned before, a maximum-length sequence can be generated by
using a linear feedback shift register that was shown in gure 2.1. The feed-
back logic for a desired period N can be found from the tables of the necessary
feedback connections for varying number of ip-ops. The rst number of a
feedback tap tells the length of the shift register, in other words the number of
ip-ops and the rest of the numbers tell us the location of exclusive-OR gates.
Welch obtained the following lower bound on the cross-correlation between
any pair of binary sequences of period L in a set of N sequences[1]:
R
c
(k) L

N 1
NL 1

L (2.3)
2.2 GOLD SEQUENCES
The autocorrelation properties of m-sequence can not be bettered but they dont
exhibit a good cross-correlation properties for CDMA. Its know that, a set of
spreading codes used for multiple access system should have as little mutual
interference as possible. For this reason, a particular class of PN sequences
are used that are called Gold sequences. They can be chosen such that, the
cross-correlation values between the codes over a set of codes are uniform and
bounded.
Gold codes can be generated by modulo-2 addition of two maximum-length
sequences with the same length. The code sequences are added chip by chip by
synchronous clocking. The generated codes are of the same length as the two
m-sequences which are added together.
6
One of the advantage of Gold codes is in generating large number of codes. To
dene a set of Gold codes, preferred pairs of m sequences are used. If a is an
m-sequence of length N, the second sequence a

can be obtained by sampling

every qth symbol of a. The second sequence is called decimation of the rst
sequence.
The following conditions are sucient to dene a preferred pair a and a

of
m-sequences:
1. n= 0; that is, n is odd or n=2
2. a

=a[q] where q is odd and either

q= 2
k
+ 1 or q= 2
2k
2
k
+ 1
3. gcd(n, k) =
_
1, for n odd;
2, for n=2.
The decimation of an m-sequence may or may not yield another m-sequence.
The set of Gold codes for this preferred pair of m-sequence is dened by {a,
a

, a+a, a+Da

, a+D
2
a

, ..., a+D
N1
a

} where D is the delay element. An

illustration of generating a Gold set is shown in gure 2.3[1]. The 2
5
+1=33
1 2 3 4 5
1 2
3 4 5
Gold
sequence
[ 5 2 ]
seq1: N=2
5
-1=31 chips
[5 4 3 2 ]
seq 2 : N=2
5
-1=31 chips
Figure 2.3: Generation of a Gold code set.
codes are generated by the above structure as follows:
Sequence 1: 1111100011011101010000100101100
Sequence 2: 1111100100110000101101010001110
0 shift combination: 0000000111101101111101110100010
1 shift

combination: 0000101010111100001010000110001
. . .
30 shift combination: 1000010001000101000110001101011

shift of sequence 2 to the left

The N+1 elements of a Gold codes sets have the property that any pair of
7
codes in the set have a three-valued cross-correlation. In this set, except the
sequences a and a

,the rest of the sequences are not m-sequences. Hence, their

autocorrelation functions are not two-valued, and it takes the same three values
as cross-correlation. The three-level cross-correlation values of Gold sequences
are as given in table 1[2].
Shift-Register Length, m Period(CodeLength) Cross-Correlation
-
m odd N=2
m
-1 -1/N
-(2
(m+1)/2
+ 1)/N
(2
(m+1)/2
1)/N
m even and not N=2
m
-1 -1/N
divisible by 4 -(2
(m+2)/2
+ 1)/N
(2
(m+2)/2
1)/N
-
Table 1. Three-Level Cross-Correlation Properties of Gold Sequences
2.3 KASAMI SEQUENCES
Kasami sequence sets are one of the important types of binary sequence sets
because of their very low cross-correlation. Kasami codes are based on PN
codes of length N=2
m
-1 where m is an even integer.
1 1 1 1 0 1 0 1 1 0 0 1 0 0 0
15 bit PN code
3 bit Kasami code
1 1 0, 1 1 0
Figure 2.4: Generation of Kasami set
a: 1 1 1 1 0 1 0 1 1 0 0 1 0 0 0 1 1 1 1 0 1 0 1 1 0 0 1 0 0 0
a: 1 1 0 1 1 0 1 1 0 1 1 0 1 1 0 0 1 1 0 1 1 0 1 1 0 1 1 0 1 1

0 0 1 0 1 1 1 0 1 1 1 1 1 1 0 1 0 0 1 1 0 0 0 0 0 1 0 0 1 1
There are two dierent sets of Kasami sequences. Generation of a small set
of Kasami sequences is similar to the generation of Gold sequences with M=2
n/2
binary sequences of period N=2
n
-1, where n is even. To generate a Kasami se-
quence,rst of all the sequence a is found by selecting every (2
n/2
+1)st bit of
an m-sequence a. The resulting sequence a is an m-sequence. The rst Kasami
8
sequence can be found by adding(modulo-2 addition) the sequences a and a.
Then by adding all 2
n/2
-2 cyclic shifts of the sequence a with the sequence a,
a new set of Kasami sequences can be formed. Bye including the sequence a in
the set, a set of 2
n/2
sequences can be found. For example, for the case of m=4,
we take 15 length PN code and take its every 5th bit and keep repeating it to
nd the sequence a. The rst member of the set is found by adding a with
the PN codea that is shown in gure 2.4. We then, shift the Kasami code by 1
bit and produce another member of the set.
The large set of Kasami sequences contains both the Gold sequences and the
small set of Kasami sequences as subsets.
2.4 HADAMARD-WALSH(ORTHOGONAL) CODES
The Hadamard-Walsh codes are generated in a set of N=2
n
codes with length
N=2
n
. The generating algorithm is as follows[1].
H
2N
=
_
H
N
H
N
H
N
H
N
_
where N is a power of 2 and overscore denotes the binary complement of the
bits in the matrix. The smallest set of N=0 is H
0
=[1] with the length 1. The
rows or columns of matrix H
N
are the Hadamard-Walsh codes since the matrix
H
N
is symmetric. The sets of 2 and 4 codes are shown below.
H
2
=
_
1 1
1 1
_
H
4
=
_

_
1 1 1 1
1 1 1 1
1 1 1 1
1 1 1 1
_

_
As shown above, in each set, the rst row of the matrix consist all 1s and rest
of the rows contains N/2 0s and N/2 1s. Also row N/2 starts with N/2 1s
and ends with N/2 0s.
Orthogonality is the most important property of Hadamard-Walsh codes.
Because of this orthogonality property, the cross-correlation between any two
Hadamard-Walsh codes of the same set(matrix) is zero, when system is per-
fectly synchronized.
Walsh codes are not maximal length or PN type codes for spread spectrum.
Although the members of the set are orthogonal, they do not give any spread-
ing. They are used in forward channel of IS-95 CDMA type system for their
orthogonality.
Walsh code spreading can be used if all users of the same channel are synchro-
nized in time, because the cross-correlation between dierent shifts of Walsh
codes is not zero.
9
2.5 VARIABLE-LENGTHORTHOGONAL CODES
Variable-length orthogonal codes are designed to improve the capability of the
system by using higher bit rates. Depending on the desired bit rates and spread-
ing bandwidth in the system, a range for the code length can be obtained.
Variable length orthogonal codes are generated by using tree-structure as
shown below[3]. Starting from C
1
=1, a set of 2
K
spreading codes can be gener-
c
1
(1)={1}
c
2
(1)={1,1}
c
2
(2)={1,-1}
c
4
(1)={1,1,1,1}
c
4
(4)={1,-1,-1,1}
c
4
(3)={1,-1,1,-1}
c
4
(2)={1,1,-1,-1}
c
8
(1)
c
8
(2)
c
8
(3)
c
8
(4)
c
8
(5)
c
8
(6)
c
8
(7)
c
8
(8)
SF=1 SF=2 SF=3 SF=4
Figure 2.5: Code tree for generation of variable length orthogonal codes (SF:
spreading factor)
ated at the kth layer(k=1,2...K) from the root of the tree. The code length of
the kth layer is 2
K
chips. The generated codes of the same layer from a same
layer, form a set of Walsh codes and they are orthogonal.
Denoting the set of N binary spreading codes of N-chip length by NxN size
of matrix C
N
, it can be expressed as shown below[3].
C
N
=
_

_
C
N
(1)
C
N
(2)
C
N
(3)
. . .
C
N
(N 1)
C
N
(N)
_

_
=
_

_
C
N/2
(1)C
N/2
(1)
C
N/2
(1)C
N/2
(1)
C
N/2
(2)C
N/2
(2)
C
N/2
(2)C
N/2
(2)
. . .
C
N/2
(N/2)C
N/2
(N/2)
C
N/2
(N/2)C
N/2
(N/2)
_

_
where C
N
(n) is the row vector of N elements and N=2
K
(K is a positive integer).
C
N/2
(n) is the binary complement of C
N/2
(n) and is the row vector of N/2
elements.
10
Any two codes of dierent layers shown in g 2.5 are orthogonal except for
the case that one of the two codes is a mother of the other. For example all of
C
32
(2), C
16
(1), C
8
(1), C
4
(1) and C
2
(1) are mother codes of C
64
(3) so they are
not orthogonal against C
64
(3). Furthermore, if a code of any layer is assigned
to a user, all the codes generated from this code can not be assigned to other
users of the same bandwidth requesting lower rates. This is restriction in order
to maintain orthogonality.
2.6 MULTIPLE SPREADING CODE ALLOCATION
As it is known, all users in a CDMA systems are multiplied by a code sequence
that has a chip rate is greater than the data rate. The way to orthogonalize the
users is, multiplying each users binary input by a short spread sequence which
is orthogonal to all other user of the same cell. The short orthogonal codes are
called channelization codes. One type of this binary orthogonal sequences is
the variable-length orthogonal codes that is explained in previous section.
After distinguishing the users of the same cell by using chanellization codes, the
users of dierent cells are distinguished by multiplication of data with a long
pseudorandom sequence. The long PN sequences are called scrambling codes.
Hence, each transmission channelization code is distinguished by a scrambling
code.
Modulo 2 sum Modulo 2 sum

Coded data
Modulo 2 sum Modulo 2 sum
Walsh # N
User # 1
Walsh # 1 Base station PN code
Modulation
User # N
Coded data
a) Forward Link CDMA
Modulo 2 sum Modulo 2 sum
Figure 2.6: Application of Walsh and PN codes in the forward link of CDMA
The application of dierent spreading in forward and reverse links of CDMA(IS-
95) is shown in gures 2.6 and 2.7. As it can be seen from the gure, in the
forward link, orhogonal(Walsh) spreading is used because all users are synchro-
nized, where the code channels are distinguished by dierent short spreading
11
codes. Conversely, for the reverse link all users are asynchronous and there-
fore, channelization codes can be implemented using variable-length orthogonal
codes. Also, in the forward link, all the base stations of dierent cells use the
same PN sequence as the scrambling code but in the reverse link, each base sta-
tion is identied by a unique time oset of its pseudorandom binary sequence.
Orthogonal
Modulator
Modulo 2 sum Modulator
PN code #N
Orthogonal
Modulator
Modulo 2 sum Modulator
PN code #1
User # 1
Coded data

User # N
Coded data
b) Reverse Link CDMA
Variable-length orthogonal code
Orthogonal
Modulator
Modulo 2 sum
Figure 2.7: Application of Walsh and PN codes in the reverse link of CDMA
12
Chapter 3
Simulation Results
In the third chapter, the fundamental properties and generation of some codes
are discussed. Also it was mentioned before that auto and cross-correlation
of these codes play a very important role in the CDMA systems performance.
Hence, in this chapter, these properties are examined and compared according
to the simulation results, in case of asynchronous situation.
As it can be seen from gure 3.1(a), the auto-correlation properties of PN
and Gold codes are exactly same. These properties are very similar to the
orthogonality properties since the correlation values between codes for the time
shifts which are less than 1 chip are very low(7.8x10
3
).
In gure 3.1(b), the cross-correlation values are shown for the interval[-
2T
c
,2T
c
]. It is clear that while the cross-correlation values are bounded by
three values, PN codes have higher and multi values. Since the PN codes are
chosen from the same set, the cross correlation properties are similar with the
shifted autocorrelation properties. Time shifts less than one chip are modelled
by increasing the resolution four times. In the new generation of DS-CDMA
systems, Gold codes are preferred since cross-correlation properties are required.
Especially for the case of asynchronous, the cross-correlation values of PN codes
are high which cause multiple access interference(MAI).
In gure 3.2, the correlation values of PN and Gold sequences are shown for
the case of higher time shifts[-10T
c
,10T
c
](PN codes are chosen from the same
set). The highest cross-correlation value of Gold code is 0.134.
In gure 3.3, the cross-correlation of two PN sequences are shown which
are chosen from two dierent sets that are produced by dierent feedback con-
nections. In the IS-95 standards, DS-CDMA systems, a long code is produced
(N=2
42
-1) and the parts of this code are used as spreading codes for dierent
users.In gure 3.4, auto and cross-correlation values of Kasami sequences are
shown. It is clear that, they have higher auto-correlation values for the time
shifts which are greater than one chip as compare to PN and Gold sequences.
Its maximum cross-correlation value is around 0.333(absolute value).
Mathematically it is proven that, for the large L and m odd, the maximum
value of the cross-correlation function between any pair of Gold sequences is
R
max
=

2L, for even R

max
=2

L. For the Kasami sequences, maximum cross-

correlation value is found as R
max
=

L.
13
2Tc 1.5Tc Tc 0.5Tc 0 0.5Tc Tc 1.5Tc 2Tc
0.2
0
0.2
0.4
0.6
0.8
1
1.2
Delay
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
2Tc 1.5Tc Tc 0.5Tc 0 0.5Tc Tc 1.5Tc 2Tc
0.2
0
0.2
0.4
0.6
0.8
1
1.2
Delay
C
r
o
s
s
c
o
r
r
e
l
a
t
i
o
n
PN
Gold
PN
Gold
Figure 3.1: Correlation values of [7 1]PN-code and [7 3]-[7 3 2 1]Gold code.
N=127. Initial contents of ip-ops:1000000
Given a set of N sequences of period L, a lower bound on their maximum
cross-correlation is[1]
R
max
L

N 1
NL 1
(3.1)
which for one large values of L and N, is approximated as R
max

L. Compar-
ing this lower bound with the maximum value of the cross-correlation function
between any pair of Gold sequences, it is clear that Gold sequences are slightly
suboptimal. On the other hand, it is observed that Kasami sequences satisfy
the lower bound for large values of L.
Because of all these properties, we can conclude that Gold & Kasami se-
quences are appropriate for CDMA applications.
14
10Tc 8Tc 6Tc 4Tc 2Tc 0 2Tc 4Tc 6Tc 8Tc 10Tc
0.2
0
0.2
0.4
0.6
0.8
1
1.2
Delay
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
10Tc 8Tc 6Tc 4Tc 2Tc 0 2Tc 4Tc 6Tc 8Tc 10Tc
0.2
0
0.2
0.4
0.6
0.8
1
1.2
Delay
C
r
o
s
s
c
o
r
r
e
l
a
t
i
o
n
PN
Gold
PN
Gold
Figure 3.2: Correlation values of [7 1]PN-code and [7 3]-[7 3 2 1]Gold code.
N=127. Initial contents of ip-ops:1000000
10Tc 8Tc 6Tc 4Tc 2Tc 0 2Tc 4Tc 6Tc 8Tc 10Tc
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
Delay
C
r
o
s
s
c
o
r
r
e
l
a
t
i
o
n
Figure 3.3: Correlation values of PN-codes chosen from the sets of [7 1]and[7 6
5 4]. N=127. Initial contents of ip-ops:1000000
15
5Tc 4Tc 3Tc 2Tc Tc 0 Tc 2Tc 3Tc 4Tc 5Tc
0.4
0.2
0
0.2
0.4
0.6
0.8
1
Delay
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
5Tc 4Tc 3Tc 2Tc Tc 0 Tc 2Tc 3Tc 4Tc 5Tc
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
Delay
C
r
o
s
s
c
o
r
r
e
l
a
t
i
o
n
Figure 3.4: Correlation values of Kasami codes. N=127. Initial contents of
ip-ops:1111
16
Chapter 4
Conclusion
The probability of error of a desired user in the direct sequence spread spectrum
system with K multiple access users can be dened as
P(E)=f(P
k
,
k
,
k
)
where it is a function of power of the kth user,P
k
, phase shift of the kth user
caused by modulation,
k
, and the amount of shift of kth user caused by the
asynchronous system ,
k
. The received signal will consist of sum of K dierent
transmitted data (one is desired user and K 1 undesired users). Reception
is achieved by correlating the received data with the desired code sequence to
produce a decision variable.
When the interference exists in the system, the probability of error is Q-function
that is function of signal to interference noise ratio(SINR). For the case of no
interference, Q-function is a function of signal to noise ratio(SNR). The average
bit error probability for the case of interference existence and synchronized
system is given as below[5].
P
e
= Q
_
_

_
1
1
3N

K1
k=1
P
k
P
0
+
N
0
2T
b
P
0
_
_
(4.1)
where P
0
is the power of desired user, T
b
is the bit duration, T
c
is the chip
duration, N is the # of chip and N
0
is thermal noise. In some mobile radio
environments, communication links are interference-limited. For the interfer-
ence limited case (only MAI, without noise), the average bit error probability
is given as[5]
P
e
= Q
_
_

_
3N

K1
k=1
P
k
P
0
_
_
(4.2)
In the noninterference limited case, for perfect power control, which means
power of all users are same where p
k
=P
0
for all k=1...K-1[5],
P
e
= Q
_
_

_
1
K1
3N
+
N
0
2T
b
P
0
_
_
(4.3)
17
where K is the number of user.The interference limited case with perfect power
control, the equation (4.3) can be approximated by
P
e
= Q
_
_

3N
K 1
_
_
(4.4)
As it can be seen from the equation (4.4), as the number of users increase
the probability of error increases as well, so they are directly proportional. As
a result, multiple access interference that is caused by the undesired users is
directly related with the cross-correlation properties of the codes of these users
so, since the probability of error depends on the MAI, the eect of codes and
their correlation properties play a very important role in the detection.
18
Chapter 5
References
[1] E. H. Dinan ve B. Jabbari, Spreading codes for direct sequence CDMA
and wideband CDMA cellular networks, IEEE Communications
Magazine, vol. 36, pp.48-54, September 1998
[2] R.L. Peterson, R. L Ziemer, D. E Borth, Introduction to Spread Spectrum
Communications, Upper Saddle River: NJ, Prentice Hall, 1995.
[3] F. Adachi, M. Sawahashi, and K. Okawa, Tree-Structured Generation
of Orthogonal Spreading Codes with Dierent Length for Farward Link
of DS-CDMA Mobile Radio,Elect. Lett., vol. 33, no. 1, Jan. 1997 , pp.
27-28.
[4] R. C. Dixon, Spread Spectrum Systems with Commercial Applications,
Wiley, 1994.
[5] Theodore S. Rappaport, Wireless Communications, Upper Saddle River:
Prentice Hall, 1996
19