Unit-4

Equalization, Diversity
And
Speech coding

1 Introduction

Mobile radio channel (unit 3) is particularly dynamic due to

* Multipath fading
* Doppler spread
As a result, they have a strong negative impact on BER of any modulation
technique (Digital Communication)

To improve received signal quality in hostile mobile radio environment,

Equalization, Diversity + Channel Coding can be used independently or in
tandem. (Three techniques are used independently or in tandem to improve
receiver signal quality)

Equalization compensates for ISI created by multipath with time dispersive

channels (W>BC)

Diversity also compensates for fading channel impairments, and is usually

implemented by using two or more receiving antennas
2

Equalization
Compensates for inter-symbol interference (ISI) created by
multipath within time depressive channels.
* Linear equalization, nonlinear equalization

Equalizers must be adaptive since the channel is generally

unknown and time varying.
Diversity
Usually implemented by using two or more receiving
antennas.
Is employed to reduce the depth and duration of the fades
experienced by a receiver.
* Spatial diversity
* Time diversity (RAKE receiver)
* antenna polarization diversity, frequency diversity
3

Channel coding

Improve link performance by adding redundant data bits in the

transmitted message
Baseband
Signal

Channel
coding

Modulation
Carrier

Used by the receiver to detect or correct some (or all) of errors

introduced by the channel in a particular sequence of message bits

Block codes, convolutional codes.

2.Fundamentals of equalization
since the mobile fading channel is random and time varying,
equalizers must take the time varying characteristics of the
mobile channel, and thus are called adaptive equalizers.
General operating modes of an adaptive equalizer:
* Training
* Tracking

Three factors affect the time spanning over which an

equalizer converges: equalizer algorithm, equalizer structure
and time rate of change of the multipath radio channel
TDMA wireless systems are particularly well suited for
equalizers
5

 Equalizer is usually implemented at baseband or at IF in a

receiver
y( t ) x( t ) f ( t ) n ( t )

f*(t): complex conjugate of f(t)

nb(t): baseband noise at the input of the equalizer
heq(t): impulse response of the equalizer

an equalizer is actually an inverse filter of the channel.

7

d t y t heq t
x t f t heq t mb t heq t
t
F f H eq f 1

If the channel is frequency selective, the equalizer

enhances the frequency components with small
amplitudes and attenuates the strong frequencies in the
received frequency response
For a time-varying channel, an adaptive equalizer is
needed to track the channel variations
8

3 A Generic Adaptive Equalizer

A time-varying filter which must be constantly be returned.

Uses ek to minize a cost function and updates the equalizer

weights in a manner that iteratively reduces the cost
function.
Classical equalization theory : using training sequence to
minimize the cost function
2

E[| ek | ]

Basic Structure of Adaptive Equalizer

Transversal filter with N delay elements, N+1 taps, and N+1
tunable complex weights

These weights are updated continuously by an adaptive algorithm.

The adaptive algorithm is controlled by the error signal ek
10

Solutions for Optimum Weights

Error signal ek xk ykTk xk kT yk
where yk yk yk 1 yk 2 .... yk N T
k 0 k

Square error
Expected MSE
where

ek

1k
2

2 k

.... Nk

xk2 kT yk ykT k 2 xk ykT k

Ex R 2p

E ek

2
k

yk2

y y
*
R E yk yk E k 1 k
....

yk N yk

yk yk 1

yk yk 2

yk21
....
yk N yk 1

yk 1 yk 2
....
yk N yk 2

p E xk y k E xk yk

xk yk 1 xk yk 2 .... xk yk N

.... yk yk N

.... yk 1 yk N
.... ....

2
.... yk N
T

Solutions for Optimum Weights

Optimum weight vector
R 1p

Minimum mean square error (MMSE)

E p R

E p

min

Minimizing the MSE tends to reduce the bit error rate

4 Equalizers in a communications Receiver

Two general categories - linear and nonlinear
equalization
if d(t) is not used in the feedback path to adapt the equalizer,
the equalization is linear
if d(t) is fed back to change the subsequent outputs
of the equalizer, the equalization is nonlinear

13

Classification of equalizers

14

5.Linear Equalizer Techniques

Linear transversal equalizer (LTE, made up of tapped delay
lines )

Basic linear transversal equalizer structure

Finite impulse response (FIR) filter

Infinite impulse response (IIR) filter

Equalizer Techniques

F Tapped delay line filter with both feedforward and feedback taps

Structure of a Linear Transversal Equalizer

d k

N2

n N1

*
n

T
2
E e(n)

y k n

No
2

F (e jT ) No

Minimum
mean squared
error

F(e j t ) :frequency response of the channel

N o :noise spectral density

6.Nonlinear Equalization
Used in applications where the channel distortion is too severe
Three effective methods
Decision Feedback Equalization (DFE)
Maximum Likelihood Symbol Detection
Maximum Likelihood Sequence Estimator (MLSE)

Basic idea : once an information symbol has been detected and

decided upon, the ISI that it induces on future symbols can be
estimated and subtracted out before detection of subsequent
symbols Can be realized in either the direct transversal form or as
N
N
a lattice filter
*

dk

E e(n)

n N1

min

y k n Fi d k i
3

T
exp{
2

i 1

No

ln[
F (e

jT

) No

] d}

Nonlinear Equalizer-DFE

Decision feedback equalizer (DFE)

Nonlinear Equalization- Predictive DFE

Predictive DFE (proposed by Belfiore and Park)
Consists of an FFF and an FBF, the latter is called a noise
predictor
Predictive DFE performs as well as conventional DFE as the limit
in the number of taps in FFF and the FBF approach infinity
The FBF in predictive DFE can also be realized as a lattice
structure .
The RLS algorithm can be used to yield fast convergence

Nonlinear Equalizer- Predictive DFE

Predictive decision feedback equalizer

Nonlinear Equalization--MLSE
MLSE tests all possible data sequences (rather than decoding
each received symbol by itself ), and chooses the data sequence
with the maximum probability as the output
Usually has a large computational requirement
First proposed by Forney using a basic MLSE estimator structure
and implementing it with the Viterbi algorithm

Nonlinear Equalizer-MLSE

The structure of a maximum likelihood sequence equalizer(MLSE) with an

adaptive matched filter

MLSE requires knowledge of the channel characteristics

in order to compute the metrics for making decisions
MLSE also requires knowledge of the statistical
distribution of the noise corrupting the signal

Algorithm for Adaptive Equalization

Performance measures for an algorithm
Rate of convergence
Misadjustment
Computational complexity
Numerical properties

Factors dominate the choice of an equalization structure and its

algorithm
The cost of computing platform
The power budget
The radio propagation characteristics

Algorithm for Adaptive Equalization

The speed of the mobile unit determines the channel fading rate
and the Doppler spread, which is related to the coherent time of
the channel directly
The choice of algorithm, and its corresponding rate of
convergence, depends on the channel data rate and coherent time
The number of taps used in the equalizer design depends on the
maximum expected time delay spread of the channel
The circuit complexity and processing time increases with the
number of taps and delay elements

Algorithm for Adaptive Equalization

Three classic equalizer algorithms :
* zero forcing (ZF),
* least mean squares (LMS), and
* recursive least squares (RLS) algorithms

Zero Forcing: The equalizer coefficients are chosen to

force the samples of the combined channel and equalizer
impulse response to a -function.
H ch ( f ) H eq ( f ) 1

f 1/(2T )

Algorithm for Adaptive Equalization

Least Mean Square Algorithm: Update the coefficients
according to the error.

dk (n) wTN (n)y N (n)

e (n) x (n) d (n)
k

w N (n 1) w N (n) ek* (n)y N (n)

28

Diversity Techniques
* Requires no training overhead
* Can provides significant link improvement with little added cost
* Diversity decisions are made by the Rx, and are unknown to the Tx

Diversity concept
If one radio path undergoes a deep fade, another independent path
may have a strong signal. (Diversity exploits the random nature of
radio propagation by finding independent (or at least highly
uncorrelated) signal paths for communication.)
By having more than one path to select from, both the
instantaneous and average SNRs at the receiver may be improved,
often by as much as 20 dB to 30 dB

Diversity Techniques
Microscopic diversity and Macroscopic diversity
The former is used for small-scale fading(deep fading) while the
latter for large-scale fading(selecting different base stations)
Antenna diversity (or space diversity)

Space diversity
*
*
*
*

Selection diversity
Feedback diversity
Maximal ration combining
Equal gain diversity

Diversity Techniques- Selection diversity

The receiver branch having the highest instantaneous SNR
is connected to the demodulator
The antenna signals themselves could be sampled and the

best one sent to a single demodulation

Selection diversity

Diversity Techniques- selection diversity .

To implement selection diversity
Antenna switch
Maximal ratio combining
Performance for M branch selection diversity .

Pr SNR 1 Pr 1 , .... , M 1 (1 e r/ )M
d

PM ( )
Pr SNR
(1 e / ) M 1e /
d

k 1

1
k

Diversity techniques

Graph of probability distributions of SNR= threshold for M branch

selection diversity. The term represents the mean SNR on each branch

34

Diversity Techniques-Feedback or scanning diversity

The signal, the best of M signals, is received until it falls
below threshold and the scanning process is again initiated

Basic form for scanning diversity

Diversity Techniques- Maximal Ratio Combining

Diversity
The received signal from the ith antenna
xi : Transmitted signal

ni : noise for the ith antenna with power N

M

NT N | Gi |2

rM G r
i 1

*
i i

i 1

rM2
(SNR) M
2 NT

Pr{ M } p( M )d M 1 e
0

MM 1e /
P( M ) M
( M 1)!
M

ri Gi xi ni

( / )k 1

k 1 (k 1)!
M

Diversity Techniques
The control algorithms for setting the gains and phases for maximal
ratio combining are similar to those required in equalizer and
RAKE receiver

MRC

Diversity Techniques-review
Maximal ratio combining

The signals from all of the M branches are weighted according to their
signal voltage to noise power ratios and then summed

Equal gain diversity

The branch weights are all set to unity but the signals from each are cophased to provide equal gain combining diversity
Frequency diversity
Frequency diversity transmits information on more than one carrier
frequency

Frequencies separated by more than the coherence bandwidth of the channel

will not experience the same fades

Time diversity
Time diversity repeatedly transmits information at time spacing that exceed
the coherence time of the channel

 Practical Considerations for space diversity

1
2

* For mobile units

* For base station x 10
to assure the decorrelation (narrow angle of incident fields)

Reception Methods of Space Diversity

*
*
*
*

Selection diversity
Feedback diversity (or scanning diversity)
Maximal Ratio Combining
Equal Gain Diversity

39

 Polarization Diversity

40

 Frequency Diversity
* Signal xmitted on more than one fc >= coherence bandwidth
(wont experience the same fade)

Time Diversity
* Xmit signal repeatedly >= coherence time
* RAKE Receiver for CDMA (multipath channel)

41

RAKE Receiver
RAKE Receiver
M

Z m Z m
m 1

Z m2
M

Z m2

m 1

An M-branch (M-finger) RAKE receiver implementation. Each correlator detects a time shifted
version of the original CDMA transmission, and each finger of the RAKE correlates to a portion
of the signal which is delayed by at least one chip in time from the other finger.

Speech Coding

Taxonomy of Speech Coders

Speech Coders
Waveform Coders

Time Domain:
PCM, ADPCM

Frequency Domain:
e.g. Sub-band coder,
Adaptive transform
coder

Source Coders

Linear
Predictive
Coder

Vocoder

Channel Vocoder (1940s-1960s)

Source-filter decomposition
* filterbank breaks into spectral bands
* transmit slowly-changing energy in each band
* 10-20 bands, perceptually spaced

Downsampling
Excitation with a pitch / noise model

48

LPC encoding
The classic source-filter model

Compression gains:
* filter parameters are ~slowly changing
* excitation can be represented many ways

49

Linear Predictive Code

Model speech production system as

an auto-regressive model:

Transfer function

S ( z)
H ( z)

E( z)

s ( n ) a ( k ) s ( n k ) e( n )

k 1

Model parameters are computed for

speech segment (~30 ms).
Parameters {a(k); k=1:p} are found
by solving a Toeplitz system of
equations.

periodic
pulse
train
generator

G
v/u
voiced

u[n]

1 a(k ) z k
k 1

unvoiced
random
sequence
generator

To encode speech, one may

transmit the quantized parameters
{a(k)} and G or equivalent
parameter set.
The model order is 8-10 in most
speech coding standards.

H(z) = 1
P

1 akz-k
k=1

Vocal Tract Model

50

LPC Speech Coder

Buffer

Voice/
Un-voice

Pitch
Analysis

Encoder

Channel

Decoder

Synthesizer

LPC
filter

Excitation

52

Encoding LPC filter parameters

For communications quality:
* 8 kHz sampling (4 kHz bandwidth)
* ~10th order LPC (up to 5 pole pairs)
* update every 20-30 ms 300 - 500 param/s

Representation & quantization

* {ai} - poor distribution,
cant interpolate
* reflection coefficients {ki}:
guaranteed stable
* log area ratios (LAR) - stable

Bit allocation (filter):

* GSM (13 kbps):
8 LARs x 3-6 bits / 20 ms = 1.8 Kbps
53

Excitation

Excitation as LPC residual is already better than raw

signal:
* save several bits/sample, still > 32 Kbps

Crude model: U/V flag + pitch period

* ~ 7 bits / 5 ms = 1.4 Kbps LPC10 @ 2.4 Kbps

54

CELP
Code excited linear predictive (CELP) speech coding.
White noise input does not give satisfactory results:
* the residue sequence still contains important information for
speech synthesis
* it is necessary to send the residue to receiving end too.

To save space, use vector quantization (VQ) technique to

encode the residue sequence
* Hence the name code excited.

In CELP, each code book is a linear vector containing 0 or

1
* each code word length is 60 samples
* successive code words are overlapped by 58 samples
* a linear search is performed to find the best code words as input to
56
the LPC model.

CELP

Represent excitation with codebook

e.g. 512 sparse excitation vectors
* linear search for minimum weighted error

57