S-72.

2211 Mobile Communication Systems and Services

Principles of DS-CDMA

Olav Tirkkonen
TKK, Department of Communications and Networking

Spread Spectrum Techniques

• Spreading the transmission bandwidth to be >> information
symbol frequency
• introducing redundancy by occupying more frequency

• Frequency Hopped Spread Spectrum (FH-SS)

• Direct Sequence Spread Spectrum
• spreading by multiplying linear modulation symbols with a spreading code

• different spreading codes can be used to multiplex channels and/or for

multiple access
• Direct Sequence Code Division Multiple Access (DS-CDMA)

Principles of CDMA 2
 Spreading: Chips and Symbols
• A chip is the shortest modulated signal in a DS-CDMA system
• Chip rate signal bandwidth
• A symbol is spread over multiple chips
• The Spreading Factor (SF) tells how many chips are used to
transmit one symbol
Rc chip rate
spreading factor = SF = =
Rs symbol rate
• SF is a spectrum spreading factor: bit rate ⇒ chip rate
• The Spreading Code is a sequence of SF chips
• Usually the chips are +-1
• Spreading code can be understood as SF × 1 vector c
• normalization: cTc = SF  T
• Example: SF=8 spreading code c = 1 1 −1 1 −1 −1 1 −1
• Spreading
• Transmitted symbol x (any linear modulation)
 T
• Transmitted chip sequence xc = x x −x x −x −x x −x

Principles of CDMA 3

Despreading
• At receiver, despreading is performed
• Example: 1-tap channel h,constant during transmission
• equivalent baseband signal model, one sample per chip
• received signals during the SF chips of transmitting the spread symbol:
 T
y = hxc + n = hx hx −hx hx −hx −hx hx −hx + n
where n is noise and interference
• despreading by multiplying with the (transposed) spreading code:
z = cTy = hxcTc + cTn = SF hx + cTn
• the chips carrying information about symbol x are coherently
combined
• noise and interference is despread and non-coherently combined
(as long as interference is not transmitted with the same
spreading code)

Principles of CDMA 4
 Interference from Other Code Channel

• assume transmission on another code channel (e.g. another

user) with same timing and same SF
• symbol, code, channel of user of interest: x1, h1, c1

• symbol, code, channel of interfering user: x2, h2, c2

• received signals during the SF chips

y = h1x1c1 + h2x2c2 + n
where n is noise (and other interference)

• despreading:
z1 = cT1y = SF h1x1 + h2x2cT1c2 + cT1n
• interference caused by transmission using c2 on c1 determined by
cross-correlation cT
1 c2 of c1 and c2

Principles of CDMA 5

Interference form Other Channel Tap

 
• Example: 2-tap channel h1 h2
• received signals during the SF chips of transmitting the spread
symbol:
 T
y = h1x 1 1 −1 1 −1 −1 1 −1 +
 T
h2x 0 1 1 −1 1 −1 −1 1 +
 T
h2x0 −1 0 0 0 0 0 0 0 + n
 T
= h1x cT + h2 xcTshift 1 + h2(x0 − x)c1 0 0 0 0 0 0 0 + n
• x0 is the previous transmitted symbol, c1 is the first chip in c
• cshift 1 is the permuted version of c where last chip is first
• despreading with timing of first channel tap:
z = cTy ≈ SF h1x + h2x cTcshift 1 + cTn
• Inter-Path Interference is characterized by auto-correlation
cTcshift n of code with shifted versions of itself

Principles of CDMA 6
 Orthogonal and Non-orthogonal Codes I

• A family of spreading codes is a set {cj }

• different spreading codes define different code multiplexed channels
• A family of spreading codes is orthogonal if
cTi cj = SF δij

0, i = j
• Here δij is the Kronecker delta, δij =
1, i = j
• at most SF orthogonal codes with length SF
• example for SF = 4:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 1 1 1
⎢1⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
c1 = ⎢ ⎥ c2 = ⎢ −1 ⎥ c3 = ⎢ 1 ⎥ c4 = ⎢ −1 ⎥
⎣1⎦ ⎣ 1 ⎦ ⎣ −1 ⎦ ⎣ −1 ⎦
1 −1 −1 1
• these are mutually orthogonal:
cT1c2 = cT1c3 = cT1c4 = cT2c3 = cT2c4 = cT3c4 = 0
• and normalized:
cT1c1 = cT2c2 = cT3c3 = cT4c4 = SF = 4
Principles of CDMA 7

Orthogonal and Non-orthogonal Codes II

• Orthogonality is lost if codes are received with different timing

• Autocorrelation properties of orthogonal codes are poor
• for example the autocorrelation of c1 = [1 1 1 1] with any shifted version of
itself is SF = 4

• this leads to poor resistance against inter-path interference

• For random spreading codes, cross & and autocorrelation are

  √   √
E cTi cj = SF for i = j E cTc shiftn = SF for n = 0
√
• expected cross-correlation of two random codes is SF

• expected
√ auto-correlation of random code with shifted versions of itself is
SF

Principles of CDMA 8
 Multicode CDMA vs. TDMA
• With orthogonal codes, the channel can be divided into SF
orthogonal code channels, each with symbol rate 1/SF of the chip
rate

• Just as with TDMA, one can divide the chip rate into SF
orthogonal time domain channels with rate 1/SF
• in TDMA, symbols/channels are multiplexed in the time domain

• in orthogonal CDMA, symbols/channels are multiplexed in the code domain

⇒ Multicode transmission

• in orthogonal CDMA with given SF , at most SF orthogonal

channels can be designed

Principles of CDMA 9

Multicode CDMA vs. TDMA Example I

• 4 symbols on 4 signaling channels, possibly with different wireless
channels h1, h2, h3, h4: ⎡ ⎤
h1x1
⎢ h2x2 ⎥
• Time Division Multiplexed channels: y = ⎢ ⎥
⎣ h3x3 ⎦ + n
h4x4
• 4 Code Division Multiplexed channels with SF = 4:
• define spreading matrix
 using SF orthogonal codes:
C = c1 c2 c3 c4
• for orthogonal spreading, C is orthogonal: CTC = CCT = SF I
• Rx signal is sum of Rx signals of the CDM channels:
y = 12 (h1x1c1 + h2x2c2 + h3x3c3 + h4x4c4) + n
⎡ ⎤ ⎡ ⎤
h1x1 h1x1
  ⎢ h2x2 ⎥ ⎢ ⎥
= 12 c1 c2 c3 c4 ⎢ ⎥ + n = √1 C ⎢ h2x2 ⎥ + n
⎣ h3x3 ⎦ SF ⎣ h3x3 ⎦
h4x4 h4x4
1
• normalization by 2 to have same Tx power in TDM and CDM

Principles of CDMA 10
 Multicode CDMA vs. TDMA Example II
• despreading of all channels:
√ 4 despreading outputs zi
• include scaling with 1/ SF = 1/2 into despreading
• scales received ⎤ and interference + noise similarly
⎡ signals ⎡ ⎤
z1 h1x1
⎢ z2 ⎥ ⎢ ⎥
z=⎢ ⎥ T T ⎢ h2x2 ⎥ + √ 1 CTn
⎣ z3 ⎦ = √SF C y = √SF C C ⎣ h3x3
1 1
⎦ SF

z4 h4x4
⎡ ⎤
h1 x 1
⎢ h2 x 2 ⎥
• with orthogonal spreading z = ⎢ ⎥
⎣ h3x3 ⎦ + ñ
h4 x 4
• noise hase not been coloured in despreading:
     
E ññH = SF1 E CTnnHC = SF1 CT E nnH C = N0I
 
=N0 I

• In 1-tap channel TDM & multicode orthogonal CDM equivalent

• difference in Peak-to-average power Ratio (PAR), higher in CDM

Principles of CDMA 11

Multicode CDMA vs. TDMA: Conclusion

• Apart for PAR, TDM and orthogonal CDM differ in wideband
multitap channels
• in TDM Inter-Symbol Interference (ISI) caused
• in CDM, inter-chip interference results in Inter-Path Interference (IPI)
• To mitigate ISI in TDM, multipath equalizers required
• polynomic complexity in delay spread
• in CDM, IPI is mitigated by RAKE receiver
• linear complexity in delay spread
• with increasing IPI, the number of multicodes in CDM can be diminished, to
keep received signal at target level

• With fixed bandwidth & chip rate, CDMA with large SF offers less
complex receiver processing against multipath interference than
TDMA and a more graceful diminishing of transmission rate with
increasing interference

Principles of CDMA 12
 Spreading in Cellular Systems
• Downlink is intra-cell synchronous by definition
• Transmissions to all users have the same timing
• Orthogonal spreading may be used
• Including possible multicode transmissions to a user
• To synchronize intra-cell UL, accurate Timing Advance is required
• Less than a fraction of the chip rate
• If UL is intra-cell synchronized, orthogonal spreading may be used
• If UL is not synchronous, orthogonal spreading for different users cannot be
used
• Pseudo random spreading is used to randomize interference
• In WCDMA, chip rate 4 Mcps, UL synchonization not considered
• required TA accuracy would be < 10−7 s
• Orthogonal spreading can be used for multicode transmission from a user
• In CDMA systems, resources used in different cells are not
orthogonal
• Reuse 1
• Pseudo-random spreading (“scrambling”) to mitigate inter-cell interference

Principles of CDMA 13

Spreading Partitioning
• functions of spreading in Cellular system
1. to provide immunity against multipath interference
• good autocorrelation properties
2. to whiten the inter-cell interference (randomization)
3. to provide orthogonal CDM for synchronized intra-cell channels
• these somewhat conflicting targets are achieved by partitioning
spreading code into two parts
• scrambling code
• a long sequence of pseudo-random +-1:s generated by mathematically defined random sign
generator, known at BS and MS
• fulfils targets 1 and 2
• channelization code
• performs the spreading from symbol rate to chip rate
• length SF
• family of SF orthogonal codes, provides multiple orthogonal channels, if needed
• fulfils target 3

• a symbol is spread by multiplying with a channelization code of

length SF and a changing set of SF consecutive chips of the
scrambling code

Principles of CDMA 14
 Spreading and Scrambling, Example
• SF = 4
 
• channelization code c = −1 1 1 −1
• pseudo-random scrambling code
• data symbols are BPSK, one bit per symbol, values ±1
• spectrum is spread from bit rate to chip rate = 4 x bit rate

Principles of CDMA 15

Processing Gain I
• multi-user CDMA (UL), N users
• the received signal of user k after despreading
N
T
zk = ck y = SF hk xk + hixicTk ci + cTk n
i=1
i=k
N

√
≈ SF hk xk +
 SF hixi + cTk n

signal i=1 noise
i=k
 
Multiple Access Interference

• the last approximate equality uses expected cross-correlation of

pseudo-random spreading
 
• expected noise power is E cTk nnHck Rc = cTk ck N0 Rc = SF N0 Rc
• the energy of the SF noise samples combined in despreading, times chip bandwidth (chip
rate)

Principles of CDMA 16
 Processing Gain II
• expected signal power is SF 2|h|2|x|2Rc = SF 2Pk
• Pk is the received signal power for user k , the channel gain times the
transmitted symbol power (symbol energy/chip duration)
• expected interference power is
 
N N ∗ ∗
N 2 2
N
SF E i=1 hi xi j=1 hj xj Rc = SF i=1 |hi | |xi | Rc = SF i=1 Pi
i=k j=k i=k i=k
• the post-despreading SINR of user k is
SF Pk Pk
γk = N ≡ G N
i=1 Pi + N0 Rc i=1 Pi + N0 Rc
i=k i=k
• the spreading factor provides the processing gain G = SF
against noise and interference
• processing gain is 3dB per doubling of SF
• here Multiple Access Interference (MAI) treated
• similarly, spreading provides processing gain against Inter-Path Interference
and Inter-Cell Interference

Principles of CDMA 17

Near–Far Effect
G Pk
• recall SINR of user k : γk = N
i=1 Pi +N0 Rc
i=k
• With the same Tx power, received power differences of UL users
due to path loss may be up to 90 dB, due to fast fading up to 50dB.
• To overcome this, a user in a disadvantaged position would need
a processing gain of the same magnitude, i.e. G 109
• This near-far effect causes a significant reduction of the capacity
• To cope with the near-far effect, effective uplink power control is
required in DS-CDMA systems

Principles of CDMA 18
 Power Control in CDMA I

• Power control commands are given with fixed periods

• the target is to track fast fading (caused by mobilty of user)
• the better channel, the less power transmitted
Principles of CDMA 19

Power Control in CDMA II

Purpose of PC:

• Uplink
• removes excessive intra-cell (and inter-cell) interference caused by
transmissions of users close to BS

• makes Rx power of users nearly equal to serve whole coverage area

properly

• Downlink
• removes excessive inter-cell interference caused by transmissions to users
close to BS

• keeps the received signal at minimum required level properly

• Both links
• mitigate fast fading
• creates reliable channels for circuit switched traffic

Principles of CDMA 20
 Advantages of DS-CDMA
• All resources can be used in all cells
• wideband channel and large SF produce sufficient SINR for cell edge users
⇒ high system capacity
• Protection against multipath interference
• based on good autocorrelation properties of the spreading codes
• if spreading factor is large
• Multipath diversity can be utilised by the RAKE-receiver
• It is easy to multiplex different channels in the code domain
• control and transport channels, different users
• Silence periods in the transmitted signal do not consume
resources
• Narrowband interference rejection
• A narrowband signal will be spread in the correlation receiver
• Privacy and low probability of interception
• signal can be detected only if spreading code is known
• In a hostile environment good anti-jamming properties

Principles of CDMA 21

Disatvantages of DS-CDMA
• In UL, chip synchronisation between the users is overwhelming
⇒ Multiple Access Interference between users
• In DL, Inter-Path Interference reduces orthogonality of users
• Accurate power control needed to avoid near-far problems which
put distant users in an unfavourable situation
• CDMA is fundamentally an access scheme for low rates and
many users
• RAKE works well in severe multipath channel only if significant fraction of the
possible orthogonal codes are not used (DL)

• despreading in UL works well only for large SF

• When striving for high data rates with high SINR requirements, IPI and MAI
dominate performance

• More complex receivers (chip equalizers) are needed to mitigate IPI and MAI
• Simplicity of DS-CDMA is lost

Principles of CDMA 22
S-72.2211 Mobile Communication Systems and Services

CDMA Capacity

Olav Tirkkonen
TKK, Department of Communications and Networking

CDMA Uplink Capacity

• divide interfering users to own cell and other cell interference
• N users in own cell, M interfering cells, Ni users in cell i
• activity factors ρj and ρi,l of own cell and other cell users
• multipath interference can be modeled as interfering user
• post-despreading SINR of user k is
G Pk
γk =
Iown + Iother + Pk /γRF
• average own and other cell interference powers, and SNR (before
despreading) are
N
 Ni
M 
 Pk
Iown = ρj Pj Iother = ρi,l Pj γRF =
i=1 l=1
N0Rc
j=1, j=k
• other cell interference is fraction of own cell interference:
Iother ≈ f Iown
• f depends on path loss and distribution of users, typically f ≈ 0.6
• the received powers are subject to power control
CDMA Capacity 2
 UL Capacity, Similar Users
• Consider the case where all users receive the same service
• target SINR γtarget

• power control target Pj = P

• same activity factor ρj = ρ

G
γk = ≥ γtarget
(1 + f )(N − 1)ρ + 1/γRF
• this constrains the number of users that can receive this service:
 
1 G 1
N ≤ Nmax = − +1
ρ(1 + f ) γtarget γRF
• Nmax is directly proportional to G

• inversly proportional to γtarget

CDMA Capacity 3

UL Capacity, Example
200
Target SINR = 3dB
Target SINR = 6dB
150 Target SINR = 9dB
Nmax

100

50

0
−20 −15 −10 −5 0 5 10
RF SNR

• parameters: G = 256, ρ = 0.5, f = 0.6

• capacity saturates when γRF grows: interference limitation
• Nmax is significantly less than the spreading factor G

CDMA Capacity 4
 UL Capacity, Non-similar Users
• different users have different SINR targets and processing gains
• target SINR for user j is γj , spreading factor is Gj
• assume that the MAI affects all users in the same way: power control
condition is
γj G j Pj
=
γk Gk Pk
• this means that the feedback coupling in PC needs not to be taken into account (how the
user’s PC affect other user’s PC, and again the users PC)

• valid if a user consumes a small fraction of the resources in the cell

• the PC of one user does not affect the interference level in the cell

• interference can be expressed in terms of SINR target using the

power control condition:
G Pk 1
γk =  =  ργ
(1 + f ) j=k ρj Pj + γRF (1 + f ) j=k Gjj γjk + γPRFk
Pk

CDMA Capacity 5

Noise Rise
• the SINR can be calculated from the previous equation:
⎛ ⎞
 ρj γj
γk = ⎝1 − (1 + f ) ⎠ Gk γRF
Gj
j=k

• Define γ0 = Gk γRF, post-despreading SINR in absence of MAI

• the noise rise is
⎛ ⎞−1
γ0 Iown + Iother + N0Rc ⎝  ρj γj
= = 1 − (1 + f ) ⎠
γk N0Rc Gj
j=k

• the increase in disturbances over thermal noise due to MAI

 ργ
• Note: if (1 + f ) j=k Gj jj = 1, noise rise is infinite
• target SINR cannot be met at any SNR

CDMA Capacity 6
 Fractional Load
• the more users the higher load in the system
• the higher SINR targets, the higher load
• the more activitym the higher load
• the fractional load is
 ρj γj
η = (1 + f )
j
Gj
• note: all users taken into account (also user k )
• noise rise equation was derived based on the power control condition that
assumed that a user takes negligible resources
⇒ no essential difference betwwen fractional load and sum in noise rise
• relation of fractional and noise rise:
γ0 1
≈
γk 1 − η
• when number of users and their SINR requirements grow so that
η → 1, the required SNR (and the Tx power) grows to infinity
• infinitely interference limited network

CDMA Capacity 7

Pole Capacity
• fractional load 1 determines the pole capacity of the CDMA
system  ρj γj
η = (1 + f ) =1
j
Gj
• the number of users and SINR requirements that can be served if all users
have ∞ Tx power
• with finite Tx powers, capacity is less than pole capacity
• Example: similar users
• pole capacity equation:
1 ργ G
Nmax = 1 ⇒ Nmax =
1+f G (1 + f )ργ
• takle ρ = 0.5, G = 256, γ = [3, 6, 9]dB = [2, 4, 8]
• f = 0.6 = 3/5


5 1 1
Nmax = · 256 · 1, , = [160, 80, 40]
8 2 4
• compare plot for similar users above
CDMA Capacity 8
 DL Fractional Load
• In DL CDMA, users in a cell typically use orthogonal spreading
codes

• Multiple Access Interference arises from inter-path interference

• partly destroys orthogonality of spreading codes

• can be modelled by an orthogonality factor α:

 ρj γj
η = (1 − α + f )
j
Gj

• in frequency flat (single path) channel, α = 1

• no in-cell interference

• in frequency selective fading, α < 1

CDMA Capacity 9

Interference Margin I
• According to noise rise equation, post-despreading SNR γ0 is
larger than SINR γk

• an Interference Margin (IM) needs to be added to the link budget

 
1
IM = 10 log
1−η
• takes care of the multiple access interference

• when load is approaching the pole capacity, the IM becomes infinite

• interference margin, fractional load, noise rise:

• different ways to view the same phenomenon: degree of interfernce
limitation of CDMA

CDMA Capacity 10
 Interference Margin II

CDMA Capacity 11

Load vs. Coverage I

• Consider the path loss model Lp = L0 + 10n log(r) in [dB]
• L0 is the path loss at 1 km distance
• n is the path loss exponent
• r is the distance in km
• Link budget: PTx − S = L1 + Lp + IM
• PTx is the transmitter power level [dBm]
• S is the receiver sensitivity [dBm]
• IM is the interference margin due to traffic load
• L1 is the sum of system gains and losses except for Lp and IM
• determine coverage area A0 = πr02 for zero fractional load and
coverage area A = πr 2 for non-zero fractional load
• find ratio from
A
10n log r + IM = 10n log r0 ⇒ = 10−IM/5n = (1 − η)2/n
A0

CDMA Capacity 12
 Load vs. Coverage II

n
η 2 3 4 5
0 1 1 1 1
0.5 0.5 0.63 0.71 0.87
0.7 0.3 0.45 0.55 0.79
0.9 0.1 0.22 0.32 0.63

• Cell breathing: Coverage area decreases with increasing traffic

load
• price from reuse 1: increasing interference at cell edge
• breathing stronger for low path loss exponents
• cell size must be planned according to maximum load

CDMA Capacity 13