CODE DIVISION MULTIPLE

ACCESS (CDMA)

Multiple Access

Why SS is attractive from the multiple access

point of view ?

FDM/TDM-based multi-access strategy may not be

optimal from the overall resource utilization point of
view (unused time slots or frequency bands).
Orthogonal signaling, however, is also possible by
signals that overlap both in time and frequency.

Code Division Multiple Access

(CDMA)

Multiple users can use the same channel as long

as different users are assigned different PN
sequence (code)

Several users can transmit simultaneously on the same

channel

The transmissions from other users will look like

interference.
CDMA main application is wireless
communication.

Direct Sequence - Code Division

Multiple Access (DS-CDMA)

DS-CDMA Example
Transmitter for User 1
m1(t)

m1(t)c1(t)

Wireless
Channel

Receiver for User 1

m1(t)+
m2(t)c1(t)c2(t)

m1(t)+e1(t)

TSymbol

m1(t)

m1(t)c1(t)+
m2(t)c2(t)

c1(t)

Transmitter for User 2

m2(t)

m2(t)c2(t)

c1(t)

Receiver for User 2

m2(t)+
m1(t)c1(t)c2(t)

m2(t)+e2(t)

TSymbol

m2(t)

c2(t)

Important Note:
The value of ei(t) depends on the cross
correlation properties between c1 & c2
ei(t)=0 if c1 & c2 are orthogonal

c2(t)
mi(t):
ci(t):
ei(t):
mi(t):

Information Message of User i

Spreading code of user i
Interference sensed at receiver of user I
5
Message detected at receiver

DS-CDMA Example

Consider CDMA system with 2 users

Receiver signal can be written as
Assume that the receiver is interested in
decoding user 1

Because s2(t) is not de-spreaded by c1(t), it will create

interference to s1(t)
The signal s1(t) can be decoded with little or no
degration
6

channel->>
polar sig.->
detecting A ... ->

DS-CDMA Example

User A: code
User B: code
The received pattern can be written as

Decoding the bit of user A we get

and
The bit of user A is decoded correctly

DS-CDMA Multiuser
Interference

The inherent multi-user interference (MUI) reduction by a

factor GP might not be sufficient, but more advanced MUI
mitigation/cancellation is usually needed.

Spreading Code Requirements

Good CDMA spreading codes should be

characterized by relatively low cross-correlation
properties to minimize multiple access
interference (MAI).
Good CDMA spreading codes should be
characterized by low autocorrelation properties
to minimize inter-symbol interference due to
multi-path channels
Ideally it is desirable to have both correlation
functions to approach zero

10
10

Orthogonal Codes

Orthogonal codes

All pairwise cross correlations are zero

Fixed- and variable-length codes used in CDMA systems
For CDMA application, each mobile user uses one
sequence in the set as a spreading code

Provides zero cross correlation among all users

Types

Walsh codes
Variable-Length Orthogonal codes

11

Walsh-Hadamard Codes

Walsh functions provide orthogonal spreading codes

Walsh matrices constructed recursively as follows:

H
H2n n
Hn

Hn
where H1 1
Hn
c1

1 1
H2

1 1
c1
c2

1 1 1 1

1 1 1 1
H4
1 1 1 1

1 1 1 1

c2
c3

c4

12

Orthogonal Variable Spreading Factor

(OVSF) using Walsh Codes
c11
c21

SF = 1

SF = 2

c22

SF = 4

c42

c41

c43

c44

SF = 8
SF = 16

OVSF TREE

Available system bandwidth determines the value of Tchip

TSymbol=SF x Tchip Bit rate is inversely proportional to SF

OVSF permits users to be allocated different SF (i.e., bit rates)

13

Orthogonal Variable Spreading Factor

(OVSF) using Walsh Codes
c11
c21

SF = 1

SF = 2

c22

SF = 4

c41

c42

c43

c44

SF = 8
SF = 16

OVSF TREE

If a user is allocated a certain code, then all codes that branch from such
code cannot be allocated to any other user

c21 is orthogonal to c22, c43, c44

c21 is NOT orthogonal to c41, c42

14

Characteristics of Walsh
Codes

Walsh codes are orthogonal presuming perfect

synchronization
Walsh codes suffer from poor auto-correlation
properties for time offsets that is greater than
zero
Walsh codes suffer from poor cross-correlation
properties when codes are not perfectly
synchronized (i.e., for time offsets greater than
zero)

15

Problems with m-sequences

Cross-correlations with other m-sequences

generated by different input sequences can be
quite high
Easy to guess connection setup in 2m samples
so not too secure
In practice, Gold codes or Kasami sequences
which combine the output of m-sequences are
used.

16

Gold Sequences

Gold sequences constructed by the XOR of two msequences with the same clocking
Codes have well-defined cross correlation properties
Only simple circuitry needed to generate large
number of unique codes
In following example, two shift registers generate the
two m-sequences and these are then bitwise XORed

17

Spreading Codes: Gold Codes

Sum two maximallength sequences of the same length

but using different generators

Example of Gold Code Generator of length 27-1

Sequence 1 Generator: x7+x3+1
Sequence 2 Generator: x7+x5+x4+x3+x2+x+1

R0

R1

R2

R3

R4

R5

R6

R0

R1

R2

R3

R4

R5

R6

Gold
Sequence

18

Gold Sequence Properties

For each starting state of the first generator,

there are 2m-1 potential starting states of the
second generator
Gold was able to show that for particular
choices of generator polynomials, Gold
sequences could have good cross-correlation
properties
The auto-correlation of Gold codes is
proportional to 2/sqrt(2m-1)
19

Spreading in Cellular CDMA

Systems

Cellular CDMA systems use two layers of

spreading
Channelization codes (orthogonal codes)

Provides orthogonality among users within the same cell

Long PN sequences (scrambling code)

Provides good randomness properties (low cross

correlation)
Reduces interference from other cells

20

10

Multiple Access Performance

Assume K users in the same frequency band,

Interested in user 1, other users interfere
4

6
5

1
2
21

Signal Model

Interested in signal 1, but we also get signals

from other K-1 users:

xk t 2 mk t t k ck t t k cos c t t k k
2 mk t t k ck t t k cos ct k

k k ct k

At receiver,
K

x t x1 t

xk t
k 2
22

11

Interfering Signal

After mixing and despreading (assume

t1=0)

zk t 2 mk t t k ck t t k c1 t cos ct k cos ct 1

After LPF
wk t mk t t k ck t t k c1 t cos k 1

After the integrator-sampler

T

cos k 1 b mk t t k ck t t k c1 t dt
0

Ik

23

At Receiver

m(t) =+/-1 (PSK), bit duration Tb

Interfering signal may change amplitude at tk

t
T

I k cos k 1 b1 k ck t t k c1 t dt b0 b ck t t k c1 t dt
0
tk

Tb
m t c1 t c1 t dt
0 1

At User 1: I1

Ideally, spreading codes are Orthogonal:

Tb
c t c1 t dt A
0 1

Tb
c t t k c1 t dt 0
0 k

24

12

Multiple Access Interference

(MAI)

If the users are assumed to be equal power

interferers, can be analyzed using the central
limit theorem (sum of IID RVs)

1
Pb Q
K 1 N0
3N
2 Eb

Where N is spreading factor

25

Near/Far Problem

Performance estimates derived using

assumption that all users have same power level
Reverse link (mobile to base) makes this
unrealistic since mobiles are moving
Adjust power levels constantly to keep equal

1
26

13

Near/Far Problem

Pe Q
1

3N

1
Eb( k )
N
0(1)

(1)
2 Eb
k 2 Eb
K

K interferers, one strong interfering signal

dominates performance
Can result in capacity losses of 10-30%
27

14