ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Revision of Lecture Eighteen

• Previous lecture has discussed equalisation using Viterbi algorithm: Note similarity
with channel decoding using maximum likelihood sequence estimation principle

• It also discusses blind equalisation based on constant modulus algorithm and

soft decision directed method: Note low complexity of algorithm, and stochastic
gradient adaptive principle

• Next two lectures we turn to important topic of multiple access techniques, with
briefly discussion on channelisation spreading codes and CDMA technology
multiplexing multiple access

Wireless
CODEC MODEM
Channel

• We will also discuss medium access control, MAC, the important sublayer linking
physical layer, PHY, and network layer

229
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Multiple Access

• Multiple access: allow many users to share simultaneously a finite amount of radio spectrum
• In terms of network protocol architecture: the medium access control which is part of data link layer
• Duplexing: mobile can simultaneously sent to and receive from base station. A duplexing channel
consists of two simplex channels
• Frequency division duplexing: employ two distinct frequency bands, one for forward channel (base
to mobile) and the other for reverse channel (mobile to base)
A duplexer in transceiver FDMA/FDD: duplex channel 1
switches between two frequency duplex channel 2
bands to operate as either reverse forward
transmitter or receiver channel channel
f
frequency split

• Time division duplexing: employ single frequency band but use time split to provide forward and
reverse slots. There is no need for a switch but time delay between forward and reverse channels
may be notable duplex channel 1
TDMA/TDD:
Transceiver operates on same duplex channel 2
frequency band as either reverse forward
transmitter or receiver channel channel
t
time split

230
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Multiple Access Techniques

• Frequency division multiple access: Total system bandwidth is divided into narrow frequency
slots. Each user is allocated a unique frequency band or channel
A user is free to transmit or receive all the time on its allocated radio channel, but the cost of
transceiver is high, as each has to be designed on a different band
• Time division multiple access: Time frame is divided into slots (channels). Each user is allocated
a particular time slot or channel
A user is limited to transmit or receive only regular bursts of a wideband signal, but it takes
advantages of digital technologies
• Code division multiple access: User data are spread by high-rate chip sequences to entire system
bandwidth. Each user is allocated a unique code
Hardware requires high-rate electronics, but this technology offers much higher capacity and many
advantages
• Space division multiple access: If there are no separation in frequency or time or code domain,
an alternative way of separating different users is spatial separation
For example, control radiated energy for each user in space or exploit different arrival angles using
multiple antennas with spatial processing. SDMA further enhances efficiency of spectrum utilisation

231
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Multiple Access Techniques (continue)

• Channels are dynamically allocated and how to make reservation to gain an access
may involve a contention process

• FDMA and TDMA have a hard capacity: no more user can access after reaching
the capacity while CDMA and SDMA have soft capacity: they allow more users
at a cost of gradual degraded quality
CDMA and SDMA are interference limited: capacity limited by MAI

• There are many combinations of multiple access/duplexing

– GSM uses TDMA/FDD: on a radio channel, time frame is divided into 8 slots to support 8 users;
and for a user, the forward and reverse channels are separated in frequency by 45 MHz
– UMTS (UTRAN) provides both FDD and TDD
– For FDD, downlink channel and uplink channel are generally different; for TDD, downlink channel
and uplink channel may be regarded as the same
– Implication: for TDD, base can identify uplink channel and send it to mobile who may then use
it as its (downlink) channel

232
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Numbers of Channels in FDMA / TDMA

• Let Btotal be the total system bandwidth, Bguard be the guard band at edge, and Bch the single
radio channel bandwidth. Then the number of channels in a FDMA system:

Btotal − 2Bguard
N =
Bch

• Give Btotal and Bguard, and let m be the maximum users supported on each Bch. Then the
number of channels in a TDMA system:

m(Btotal − 2Bguard)
N =
Bch

• Let bOH be the overhead bits per frame and bT ot the total bits per frame. Then the efficiency of
TDMA is defined by
bOH
„ «
ηf = 1 − × 100%
bT ot

• AMPS uses FDMA/FDD with Btotal = 12.5 MHz, Bguard = 10 kHz and Bch = 30 kHz

12.5 × 106 − 2(10 × 103 )

N = = 416
30 × 103

233
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Comparison of FDMA / TDMA (continue)

• In GSM forward link, Btotal = 25 MHz, Bch = 200 kHz and each Bch supports 8 speech
channels. Assuming Bguard = 0, then

8(25 × 106 )
N = 3
= 1000
200 × 10

A GSM time slot consists of 6 trailing bits, 8.25 guard bits, 26 training bits and 2 traffic bursts of
58 data bits: bT ot = 1250 b/frame, bOH = 322 b/frame and ηf = 74.24%
• FDMA is 1st generation technology: a radio channel for one user and, when not in use, becomes
wasted. It requires very few bits for overhead but system costs are high due to single channel per
carrier design and costly RF bandpass filters to minimise adjacent channel interference
• TDMA is 2nd generation technology: a single carrier is shared by several users. The burst nature
of transmission makes the handoff process simpler, since it is able to listen for other base station
during idle time slots. TDMA systems require high synchronisation overhead and guard slots to
separate users. Bandwidth on demand can be implemented
• FDMA will always be part of the system for dividing spectrum, e.g. in a TDMA system, FDMA
is used to divide the total system bandwidth into radio channels and each radio channel is then
TDMA for several users

234
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Direct sequence Spread Spectrum Modulation

• Recall that to transmit at rate fb requires certain bandwidth B , and all schemes discussed so far
try to be bandwidth efficient, as spectrum resource is extremely precious
• Image a modulation scheme which requires a bandwidth ≫ B ! Spread spectrum does this
• Original purposes of SS techniques:
security and anti-jamming original spread signal Jamming
signal white noise like where to put?
The basic idea is to spread
communication signal, which is
relatively narrowband, to a very f f
wideband white noise like one B

SS as a single user scheme will

be very bandwidth inefficient
1 1 user 1
but it turns out to be
an excellent multi-user scheme 1 2

to simultaneously transmit 1&2

multiple users on a single
carrier, and in doing so becomes 2 2 user 2
very bandwidth efficient and has 2 transmitted 1
on single
a much larger capacity orignal spread carrier despread

235
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

DS-SS Transceiver

• The modulator/demodulator are similar to standard ones, except bandwidth dealt

with is much wider: a code generator generates a chip sequence of rate fchip ≫ fd,
where fd is the data symbol rate and clock recovery is not shown
recovered
data X X X X data
Tx Rx
signal code
signal syn &
track
code carrier carrier code
generator generator recovery generator

• Typically fchip = N · fd, and data signal data signal

N is called spreading gain X
code signal code signal
• The operation of spreading X
and despreading: spread signal spread signal

• A key feature of CDMA technology: spreading sequences or codes

236
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Pseudo-Noise Sequences

• PN sequences: are periodic binary sequences with autocorrelation exist noise-like

property. Let two PN sequences of period N be {an}N −1 N −1
n=0 and {bn }n=0
N −1 N −1
1 X 1 X
autocorrelation: Raa(k) = anan+k crosscorrelation: Rab (k) = an bn+k
N n=0 N n=0

• They can be produced with shift register feedback circuit or generator polynomial
A few rules:
g(x) = 1 + x2 + x5:
– All zeros state is not allowed
– For m registers, period N ≤ 2m − 1 +
– Think binary bits 0, 1 as waveforms +1, −1

• Basic requirements of spreading codes

– A code should have good autocorrelation property (ideally, an impulse at offset k = 0). Then
code synchronisation and tracking will be easy: just repeatedly shift the code one chip and do
the correlation with the received signal until the correlation peaks
– For any two signature codes, their crosscorrelation should be very small (ideally, zero over all
values of k). Then multi-user interference (MUI) will be small

237
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

m-Sequences and Gold Codes

• Recall that, for m registers, sequence period N ≤ 2m − 1. Those PN sequences which have a
maximal period or length N = 2m − 1 are called maximal length sequences or m-sequences
• Autocorrelation of an m-sequence: 
1, k = lN
Raa (k) =
− N1 , k 6= lN

– Excellent autocorrelation property but relatively poor crosscorrelation property

– Size of codes for a m is small (usually << N ), thus small multi-user capacity
• Gold codes: given a set of m-sequences, certain pairs have three-value crosscorrelation function
(−1, −t(m), t(m) − 2), with factor N1 omitted
m−seq.1
(
2(m+1)/2 + 1, m odd
t(m) =
2(m+2)/2 + 1, m even
+
maximum of which is smaller than those of the rest +
m-sequences. These are called preferred pairs and are
used to generate Gold codes m−seq.2
– Lower peak crosscorrelation than m-sequences and +
generally lower crosscorrelation

• Repeating for all possible cyclically shifted modulo-2 additions of a preferred pair results in N
sequences, plus original preferred pair → size of Gold codes from a preferred pair is N + 2

238
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Comparison

The table compares m-sequences with Gold codes, where m is the number of registers,
N the code length, and peak level is the maximum crosscorrelation value (with the
factor N1 omitted)
m-sequences Gold codes
m N Set size Peak level Set size Peak level
3 7 2 5 9 5
4 15 2 9 17 9
5 31 6 11 33 9
6 63 6 23 65 17
7 127 18 41 129 17
8 255 16 95 257 33
9 511 48 113 513 33
10 1023 60 383 1025 65
11 2047 176 287 2049 65
12 4095 144 1407 4097 129

239
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

m-Sequence Example
clock register state output offset k The shift register feedback circuit for
0 0 1 0 code 0 1 2 3 4 5 6 the generating polynomial g(x) =
1 0 0 1 0 1 -1 -1 -1 1 1 -1 1 + x2 + x3 with the initial condition:
2 1 0 0 1 -1 1 -1 -1 -1 1 1 +
3 1 1 0 0 1 -1 1 -1 -1 -1 1
4 1 1 1 0 1 1 -1 1 -1 -1 -1 0 1 0
5 0 1 1 1 -1 1 1 -1 1 -1 -1
6 1 0 1 1 -1 -1 1 1 -1 1 -1 There are m = 3 shift register stages
7 0 1 0 1 -1 -1 -1 1 1 -1 1 and the period of m-sequences is N =
2m − 1 = 7.
Autocorrelation function: N −1
1 X
Raa (k) = an an+k with N = 7
N n=0
1 1
Raa (0) = (1 × 1 + (−1) × (−1) + 1 × 1 + 1 × 1 + (−1) × (−1) + (−1) × (−1) + (−1) × (−1)) = × 7 = 1
7 7
1
Raa (1) = (1 × (−1) + (−1) × 1 + 1 × (−1) + 1 × 1 + (−1) × 1 + (−1) × (−1) + (−1) × (−1))
7
1 1 1
= (−1 − 1 − 1 + 1 − 1 + 1 + 1) = × (−1) = −
7 7 7
..
.
1
Raa (6) = (1 × (−1) + (−1) × 1 + 1 × 1 + 1 × (−1) + (−1) × (−1) + (−1) × (−1) + (−1) × 1)
7
1 1 1
= (−1 − 1 + 1 − 1 + 1 + 1 − 1) = × (−1) = −
7 7 7

240
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Orthogonal Codes
• Orthogonal codes: these codes are orthogonal, i.e. crosscorrelation at lag or offset k = 0 is zero
• In perfect synchronisation, there is no MUI, but in practice autocorrelation and crosscorrelation of
codes at k 6= 0 are also important. Note that multipath will destroy orthogonality
• Walsh codes: apply Hadamard transform
2 3
» – » – 0 0 0 0
Hn Hn 0 0 6 0 1 0 1 7
H1 = [0] H2n = e.g. H2 = , H4 = 6 7, ···
Hn Hn 0 1 4 0 0 1 1 5
0 1 1 0

• Note that columns of Hn are orthogonal and the same for rows (remember 0, 1 → +1, −1, try
this on H4 and verify orthogonality)
• For Walsh codes, N = 2, 4, 8, 16 etc. and, for code length N , there are N orthogonal codes
(multi-user capacity similar to Gold codes)
But autocorrelation and crosscorrelation at k 6= 0 are considerably larger than Gold codes
• Orthogonal Gold codes: by carefully attaching an additional “0” to original Gold codes, the resulting
codes become orthogonal with much better autocorrelation characteristics than Walsh codes
• OVSF: codes with orthogonal variable spreading factors to support different rates
• Codes with large zero correlation zone: no need to achieve precise synchronisation

241
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Multiuser Detection

b (k)
• Consider downlink case (handset), where 1 M S1 (z) noise
bi (k) ∈ {±1} denotes k-th bit of user b (k) chip rate
2 channel
i, user i code si = [si,1 · · · si,M ]T , M S2 (z) r(k)
the received signal vector at chip rate . Σ C(z) Σ
. chip rate
T
r(k) = [r1(k) · · · rM (k)] , and K is . sampler
bK (k)
the number of users M SK (z)
code filters

• Assume that there is no multipath distortion, the codes are orthogonal and synchronisation is
perfect. Then to detect user i data one only needs to despread r(k) with the code si :
T
y(k) = si r(k) and b̂i (k) = sgn(y(k))

• This is called matched filter detection. However, when the channel distortion exists, multiuser
interference becomes serious and the matched filter detection is no longer adequate
• A widely used detector is the linear detector given by
T
y(k) = w r(k) and b̂i (k) = sgn(y(k))

Note that this is similar to linear equaliser → MMSE and MBER designs, adaptive LMS and LBER

242
ELEC6014 (EZ412/612) Radio Communications Networks and Systems S Chen

Summary

• Multiple access schemes: FDMA, TDMA, CDMA, SDMA

Basic techniques, capacity of a multiple access scheme,
CDMA (3rd generation technology), SDMA (4th generation technology?)

• Direct-sequence spread-spectrum modulation: principle and basic transceiver

structure

• Spread sequences or signature codes and multiuser capacity

PN sequences: maximal length sequences, Gold codes, how to generate them
Orthogonal codes: Walsh codes, orthogonal Gold codes, how to generate them

• Multiuser detection: matched filter detection, linear detector

243