You are on page 1of 29

Lecture # 8

Equalization


To compensate for channel induced ISI we use a process known as
Equalization: a technique of correcting the frequency response of
the channel
The filter used to perform such a process is called an equalizer





Since H
R
(f) is matched to H
T
(f), we usually worry about H
C
(f)
The goal is to pick the frequency response H
EQ
(f) of the equalizer
such that

where
and the phase characteristics

) (
) (
1
) ( 1 ) ( ) (
f j
C
EQ EQ c
c
e
f H
f H f H f H
u
= =
| ) ( |
1
| ) ( |
f H
f H
C
EQ
=
) ( ) ( f f
C EQ
O = O
3.4 Equalization

Equalization Process

Apply a filter that results in an equalized impulse response having
zero ISI and channel distortion.
This means that convolution of the channel impulse response and
the equalizer impulse response must equal 1 at the center tap and
have nulls at the other sample points within the filter span.
It can be difficult to determine the inverse of the channel response if
the channel response is zero at any frequency, then the inverse is
not defined at that frequency.
The receiver generally does not know what the channel response is.
Channel changes in real time so equalization must be adaptive
The equalizer can have an infinite impulse response even if the
channel has a finite impulse response
The impulse response of the equalizer must usually be truncated


Problems with Equalization

Equalization Techniques or Structures
Three Basic Equalization Structures
Linear Transversal Filter
Simple implementation using Tap Delay Line or FIR filters
FIR filter has guaranteed stability (although adaptive algorithm
which determines coefficients may still be unstable)

Decision Feedback Equalizer
Extra step in subtracting estimated residual error from signal

Maximal Likelihood Sequence Estimator (Viterbi)
Optimal performance
High complexity and implementation problem (not heavily used)

Linear Transversal Equalizer

This is simply a linear filter with adjustable parameters
The parameters are adjusted on the basis of the measurement of
the channel characteristics
A common choice for implementation is the transversal filter (Tap
Delay Line) or the FIR filter with adjustable tap coefficient









Total number of taps = 2N+1
Total delay = 2Nt
Fig. 3.26
N is chosen sufficiently large so that equalizer spans length of the ISI.

Normally the ISI is assumed to be limited to a finite number of samples

The output y
k
of the Tap Delay Line equalizer in response to the input
sequence {x
k
} is



where c
n
is the weight of the n
th
tap
Ideally, we would like the equalizer to eliminate ISI resulting in



But this cannot be achieved in practice.

=

= =
N
N n
n k n k
N N k x c y 2 ,...... 2 ,

=
=
=
0 , 0
0 , 1
k
k
y
k
However, the tap gains can be chosen such that



There are two types of such equalizer (i.e., linear equalizers)
Preset Equalizer:
Transmits a training sequence that is compared at the receiver
with a locally generated sequence
Requires an initial training sequence
Differences between sequences are used to update the
coefficient c
n
Adaptive Equalizer:
Equalizer adjust itself periodically during transmission of data
The tap weights constitute the adaptive filter coefficient

=
=
=
N k
k
y
k
,...... 2 , 1 , 0
0 , 1
The two techniques can be combined into a robust equalizer. In this
case, there are two modes of operation:
Training Mode
For the training mode, a known sequence is transmitted and a
synchronized version is generated at the receiver
Decision-directed mode
When training mode is complete, the adaptive algorithm is
switched on
The tap weights are then adjusted with info from training mode
The impulse response of the transversal filter is

=
=
=
N
N n
fn j
n eq
N
N n
n eq
e c f H
n t c t h
t t
t o
2
) (
) ( ) (
If x(t) is the signal pulse corresponding to



then the equalized output signal is




Nyquist zero ISI condition implies that



) ( ) ( ) ( ) ( f H f H f H f X
R C T
=

=
=
N
N n
n
n t x c t y ) ( ) ( t

=
=
= = =

=
N k
k
n kT x c kT y y
N
N n
n k
,...., 2 , 1 , 0
0 , 1
) ( ) ( t
Since there are 2N+1 equalizer coefficients, we may express in
matrix form as:
y=Xc
where:
X = (2N+1) x(2N+1) matrix with elements x(kT - nt)
c = (2N+1) column coefficient vector
y = (2N+1) column vector

Since this design forces the ISI to be zero at sampling instants t =
kT, the equalizer is called zero-forcing equalizer (ZFE)

Thus we obtain a set of (2N+1) linear equations for the ZFE

In Figure 3.26 t is chosen as high as T
t = T Symbol-spaced equalizer; t < T Fractional-spaced
equalizer
Zero-Forcing Solution
For N=1
)) 1 ( 0 ( )) 0 ( 0 ( )) 1 ( 0 ( ) 0 ( , 0
1 0 1
+ + = =

x c x c x c y k
)) 1 ( 1 ( )) 0 ( 1 ( )) 1 ( 1 ( ) 1 ( , 1
1 0 1
+ + = =

c x c x c y k
)) 1 ( 1 ( )) 0 ( 1 ( )) 1 ( 1 ( ) 1 ( , 1
1 0 1
+ + = =

x c x c x c y k

(
(
(

(
(
(


=
(
(
(



1
0
1
) 0 ( ) 1 ( ) 2 (
) 1 ( ) 0 ( ) 1 (
) 2 ( ) 1 ( ) 0 (
) 1 (
) 0 (
) 1 (
c
c
c
x x x
x x x
x x x
y
y
y
1 ) 1 2 ( ) 1 2 ( ) 1 2 ( + + + N N N
For N=2







Generalizing results:




(
(
(
(
(
(

(
(
(
(
(
(




=
(
(
(
(
(
(

2
1
1
2
0
) 0 ( ) 1 ( ) 2 ( ) 3 ( ) 4 (
) 1 ( ) 0 ( ) 1 ( ) 2 ( ) 3 (
) 2 ( ) 1 ( ) 0 ( ) 1 ( ) 2 (
) 3 ( ) 2 ( ) 1 ( ) 0 ( ) 1 (
) 4 ( ) 3 ( ) 2 ( ) 1 ( ) 0 (
) 2 (
) 1 (
) 0 (
) 1 (
) 2 (
c
c
c
c
c
x x x x x
x x x x x
x x x x x
x x x x x
x x x x x
y
y
y
y
y
1
0
1
0
=
(
(
(

= N f or where z
z X c
1
=
Minimum MSE Solution
A more robust equalizer can be obtained if {c
n
} tap weights are
chosen to minimize the mean square error(MSE) of all ISI terms plus
noise power at the output of equalizer
MSE is defined as:
the expected value of the squared difference between
the desired data symbol and estimated data symbol

Whereas

xc = ) ( ) ( n z n e
] | ) ( [|
2
n e E MSE =
] ) ( ) ( [
2
c x 2 xc cx
T T T
n z n z E + =
c x c x x c
T T T 2
] ) ( [ 2 ] [ )] ( [ n z E E n z E + =
c R c cR
x
T
xx z z
2
2
+ =o
0 =
c
c
= V
c
MSE
0 2 2 =
x xx
R c R
z
x
xx
R R c
z
1
=
Deterministic Case:





Example 3.6: A Minimum 7-Tap Equalizer
Consider that the tap weights of an equalizing transversal filter are
to be determined by transmitting a single impulse as a training
signal. Let the equalizer circuit be made up of 7 taps. Given a
received distorted set of pulse samples{x(k)}, with values 0.0108, -
0.0558, 0.1617,1.0000, -0.1749, 0.0227, 0.0110, use a minimum
MSE solution to find the weights {c
n
} that will minimize the ISI. With
these weights, calculate the resulting values of the equalized pulse
samples at the following times:

What is the largest magnitude sample contributing to ISI, and what is
the sum of all the ISI magnitudes?
T
x
x R ) (n z
z
=
x x R
T
xx
=
} 6 ....., , 2 , 1 , 0 { = k
Solution: For a 7-tap filter (N=3)
Dimensions for matrix x will be 4N+1 by 2N+1 = 13x7
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(

=
0108 . 0 0 0 0 0 0 0
0558 . 0 0108 . 0 0 0 0 0 0
1617 . 0 0558 . 0 0108 . 0 0 0 0 0
0000 . 1 1617 . 0 0558 . 0 0108 . 0 0 0 0
1749 . 0 0000 . 1 1617 . 0 0558 . 0 0108 . 0 0 0
0227 . 0 1749 . 0 0000 . 1 1617 . 0 0558 . 0 0108 . 0 0
0110 . 0 0227 . 0 1749 . 0 0000 . 1 1617 . 0 0558 . 0 0108 . 0
0 0110 . 0 0227 . 0 1749 . 0 0000 . 1 1617 . 0 0558 . 0
0 0 0110 . 0 0227 . 0 1749 . 0 0000 . 1 1617 . 0
0 0 0 0110 . 0 0227 . 0 1749 . 0 0000 . 1
0 0 0 0 0110 . 0 0227 . 0 1749 . 0
0 0 0 0 0 0110 . 0 0227 . 0
0 0 0 0 0 0 0110 . 0
x
Using matrix x, form autocorrelation matrix R
xx
and cross
correlation matrix R
zx.
Solution for tap weights is:




Using these weights, the 13 equalized samples {y(k)} at times
:



The largest magnitude sample contributing to ISI : 0.0095
The sum of all the ISI magnitudes : 0.0195

=

} {
, 3 , 2 , 1 , 0 , 1 , 2 , 3
c c c c c c c
0269 . 0 , 0670 . 0 , 1318 . 0 , 9495 . 0 , 1659 . 0 , 0108 . 0 , 0116 . 0
} 6 ....., , 2 , 1 , 0 { = k
0003 . 0 , 0022 . 0 , 0095 . 0 , 0015 . 0 , 0007 . 0
, 0003 . 0 , 0000 . 1 , 0000 . 0 , 0000 . 0 , 0007 . 0 , 0041 . 0 , 0001 . 0 , 0001 . 0


Steepest Descent Algorithm
Difficult to find the inverse of a large matrix.
Use gradient based iterative techniques
Cost function




Start with an initial estimate of c
0
and update it by
moving in the opposite direction of the gradient of J.
Keep on updating the old estimate till convergence is
reached.

] | ) ( [|
2
n e E MSE J = =
c R c cR
x
T
xx z z
2
2
+ =o
Steepest Descent Algorithm
Steepest Descent Algorithm
Consider the coefficients:

The steepest descent algorithm is given by:


Where


If we use instantaneous estimate of we have:


Which is called the LMS algorithm. As in the previous case z is the
desired signal and x is the received signal.

] ,........, ,......, [
0 N N
c c c

i i i
V =
+

2
1
1
c c
) ( 2
zx
i
xx
i
R c R = V
i
V
) ) ( ( 2
T T
x c xx n z
i i
= V
clear all;close all
L=3; % Signal Duration in seconds
fs=8000; % Sampling frequency
N=21; % Number of filter taps
%Training Signal
z=rand(1,L*fs);
% Impulse response of the channel
h=[1,0.7,0.2,-0.5,-0.8,-0.4,0,0.25,0.1,0.05,0,0];
x_r=filter(h,1,z);
% Intialization
% Delay line
x=zeros(1,N);
% Filter Coefficients
c=zeros(1,N);
% Step Size
mu=0.001;
Matlab Example :
% LMS Algorithm
for n=1:L*fs
x=[x(2:N) x_r(n)];
e(n)=z(n)-c*x;
c=c+mu*e(n)*x;
end

figure(1)
H=fftshift(fft(h,fs));H=abs(H);H=H/max(H);
plot(0:fs/2-1,H(fs/2+1:fs));hold on
Cf=fftshift(fft(c,fs));Cf=abs(Cf);Cf=Cf/max(Cf);
plot(0:fs/2-1,Cf(fs/2+1:fs),'r');grid;xlabel('Frequency (Hz)');
gtext('|H(f)|')
gtext('|C(f)|')
0 500 1000 1500 2000 2500 3000 3500 4000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Frequency (Hz)
|H(f)|
|H(f)|
|H(f)|
|C(f)|
Results:
Training Mode vs. Decision Directed mode
Fractionally Spaced Equalizer
The spectrum property of the baud-rate and fractionally spaced equalizer.

Decision Feedback Equalizer

A decision-feedback equalizer (DFE) is a nonlinear equalizer that
employs previous decisions to eliminate the ISI caused by
previously detected symbol
It consists of a feedforward section a feedback section and a
detector connected together as shown







The filters are usually fractionally spaced FIR with adjustable tap
coefficients
The detector is a symbol-by-symbol detector
DFE is based on the principle that once you have determined the
value of the current transmitted symbol, you can exactly remove the
ISI contribution of that symbol to future received symbols
The nonlinear feature is due to the decision device, which attempts
to determine which symbol of a set of discrete levels was actually
transmitted.
Once the current symbol has been decided, the filter structure can
calculate the ISI effect it would tend to have on subsequent received
symbols and compensate the input to the decision device for the
next samples.
This postcursor ISI removal is accomplished by the use of a
feedback filter structure.


Adaptive Equalization for Digital Cellular
Telephony

The direct sequence spreading employed by CDMA (IS-95) obviates
the need for a traditional equalizer.
The TDMA systems (for example, GSM and IS-54), on the other
hand, make great use of equalization to contend with the effects of:
multipath-induced fading,
ISI due to channel spreading,
additive received noise,
channel-induced spectral distortion, etc
Of the nonlinear equalizers, the DFE is currently the most practical
system to implement in a consumer system.
Other designs that outperform the DFE in terms of convergence or
noise performance, but these generally come at the expense of
greatly increased system complexity.