‘eye_eq’ Program Tutorial

Jungsub Byun

When we send transmissions more closely in succession to increase the data transmission
rate, interference between them is unavoidable. This phenomenon is called intersymbol
interference(ISI) and this can be a devastating factor to the proper function of the optimum
detector. So, communication engineers use equalization methods to minimize the effect of
ISI. The equalization methods are designed to turn a bandlimited channel with ISI into a
new memoryless flat channel at the receiver output. Actually, they hope to create another
new flat AWGN-like channel.
One method to form a picture of this ISI is the eye diagram. If the trigger is synchronized to
the symbol rate, the eye diagram is like the image of an oscilloscope. The eye diagram is
depicted through overlapping a few serial symbol intervals of the modulated and filtered
continuous-time waveform. In this eye diagram for binary transmission on a channel, a
distinct opening can be noticed in the center of the plot. Because the ISI results in the
spread among the path traces, the opening in the eye becomes narrower as the ISI increases.

Minimum-Mean-Square-Error Decision Feedback Equalizer(MMSE-DFE) utilizes the

previous attempts to estimate the current symbol through an SBS detector. Any trailing
intersymbol interference resulted from the former symbols is reorganized and then
subtracted in the feedback system. So it is likely that the channel output signal can be a
causal signal after going through the feedforward filter. The feedback section will then
subtract (without noise enhancement) any trailing ISI.

For showing various effects on various channels in Equalization part, the ‘eye_eq’ program
tutorial consists of
[1] Function description of eye_eq,
[2] Plotting the frequency response of various channels,
[3] Comparison between MMSE DFE and MMSE LE - FIR equalizer performance
(SNR) versus varying from 1-1*D-1 to 1+1*D-1 channels with fixed SNRmfb/noise
variance, and
[4] Plotting and Comparison among different channel’s 'EYE Diagram'- before
Equalizer (Channel output)/ After Equalizer(Filter output). Calculate probability
error with/without Equalizer – eye_eq program.
[1] Function description of eye_eq
function []= eye_eq(p,Ex,noise_var,eq_type);
%-----------------------------------------------------------------------------
%p = pulse response [a*D^-1 + 1] ==> [a 1]
% Ex = average energy of signals, Ex_bar
% noise_var = noise variance
% eq_type = Z => ZERO FORCING
% M => MMSE
% D => MMSE-DFE
% outputs: pe_(zfe/mmle_dfe/mmse_le)= probability of error with Equalizer from N input
2PAM[+1/-1] data sequences
% outputs: pe_no_eq = probability of error without Equalizer
% outputs: dfseSNR = receiver(equalizer)SNR, unbiased in dB
% outputs: pe_SNR = error probability estimation from receiver(equalizer)SNR, Pe = Q
function of sqrt(dfseSNR)sdsds
% this function shows Frequency response of the channel p and equalizer filter, eye
diagram, receiver SNR, and probability of error.
% N = 250; % # of input 2 PAM[+1/-1] data, you can increase the N of input data
sequences in order to calculate the Pe accurately
% created 1/06 by Jungsub Byun and M. Malkin EE379A
% -------------------------------------------------------------------------
function [dfseSNR,w_t]=dfecolor(l,p,nff,nbb,delay,Ex,noise); this program has come to be
used throughout the industry to compute/project equalizer performance. You will learn later.
Summary of eye_eq algorithm
For the channel p at SNRmfb=10dB
a. generate the input binary 2PAM[+1 and -1] data sequences
b. make the data signal after the channel p without Gaussian noise through filter or conv
command
c. generate Gaussian noise
d. make received signal with noise for the ISI channel p at the receiver
e. make nff taps of feedforward filter and nbb taps of feedback filter by dfecolor program
f. convert the ISI received signal into a new AWGN-like channel at the receiver output, filtering
with nff taps of feedforward filter
e. make the operation part of nbb taps feedback filtering after feedforward equalizing
f. make the detection and decision part of the equalized data
g. calculate the error probability estimation after equalizing – the decision feedback equalizer of
nff taps of feedforward filter and nbb taps of feedback filter with comparing between the
equalized data and original input binary 2PAM[+1 and -1] data sequences
h. calculate probability of error without dfe equalizer, comparing between received signal with
noise for the channel p and original input binary sequences
i. estimate the probability of error from dfseSNR(dfecolor result), Pe = Q function of
sqrt(dfseSNR)
j. create EYE Diagram by a1=(sinc(x2)); a2=(sinc(x2-1)); a3=(sinc(x2-2)); % sinc function
k. Plot and compare EYE Diagrams –original binary Input data at the transmitter, before
Equalizer (Channel output), After Equalizer(MMSE DFE output), and decision data of after
MMSE DFE at the receiver.
[2] Frequency response of different channels

A. 'Frequency response of channel 1+a*D a=0.1, 0.5, 0.9, 1'

B. ('Frequency response of channel 1-a*D a=0.1, 0.5, 0.9, 1')

-1
C. ('Frequency response of channel a*D +1 a=0.1, 0.5, 0.9, 1')
D. ('Frequency response of channel -a*D-1+1 a=0.1, 0.5, 0.9, 1')

E. ('Frequency response of channel [1+D-1], [1+D-1+D-2], [1+D-1+D-2+D-3], [1+D-1+D-2+D-

3
+D-4], [1+D-1+D-2+D-3+D-4+D-5]')

F. ('Frequency response of channel (1+D)2 , (1+D)2*(1+D)+1')

[3] Comparison between MMSE DFE and MMSE LE - FIR equalizer
performance(SNR) for 1+a*D-1 versus varying from 1-1*D-1 to 1+1*D-1 channels with
fixed SNRmfb/noise variance (comparing between MMSE DFE and MMSE LE)

A. Fix SNRmfb=10dB, Noise variance=(1+a2)/10<=[1+a*D-1] channel

'FIR equalizer performance(SNR) for 1+a*D-1 versus from 1-1*D-1 to 1+1*D-1 channels
with 13 feedforward taps, with(MMSE DFE)/without(MMSE LE) feedback tap, and
delay=12'

-1 -1 -1
qualizer preformance(SNR) for 1+a*D and versus from 1-1D to 1+D channel with 13 feedforward taps, with/without feedback tap, a
10

8
SNRof gdfecolor

5
MMSE DFE with one feedback tap
MMSE LE without feedback tap

4
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
-1
a varies from -1 to 1 in 1+a*D channel

B. Fix noise variance=0.181

'FIR equalizer performance(SNR) for 1+a*D-1 versus from 1-1D-1 to 1+D-1 channel with 13
feedforward taps, with(MMSE DFE)/without(MMSE LE) feedback tap, delay=12, and
noise variance=0.181'
-1 -1 -1
formance(SNR) for 1+a*D and versus from 1-1D to 1+D channel with 13 feedforward taps, with/without feedback tap, delay=12, an
9 MMSE DFE with one feedback tap, noise variance=0.181
MMSE LE without feedback tap, noise variance=0.181
8.5

7.5
SNRof gdfecolor

6.5

5.5

4.5

4
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
-1
a varies from -1 to 1 in 1+a*D channel
[4] function []= eye_eq(p,Ex,noise_var,eq_type);
%-----------------------------------------------------------------------------
%p = pulse response [a*D^-1 + 1] ==> [a 1]
% Ex = average energy of signals, Ex_bar
% noise_var = noise variance
% eq_type = Z => ZERO FORCING
% M => MMSE
% D => MMSE-DFE
% outputs: pe_(zfe/mmle_dfe/mmse_le)= probability of error with Equalizer from N input
2PAM[+1/-1] data sequences
% outputs: pe_no_eq = probability of error without Equalizer
% outputs: dfseSNR = receiver(equalizer)SNR, unbiased in dB
% outputs: pe_SNR = error probability estimation from receiver(equalizer)SNR, Pe = Q
function of sqrt(dfseSNR)sdsds
% this function shows Frequency response of the channel p and equalizer filter, eye
diagram, receiver SNR, and probability of error.
% N = 250; % # of input 2 PAM[+1/-1] data, you can increase the N of input data
sequences in order to calculate the Pe accurately
% created 1/06 by Jungsub Byun and M. Malkin EE379A
%-------------------------------------------------------------------------

1+ 0.9D Channel 'EYE Diagram'- Transmitter Input data

0.5

-0.5

-1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

 -1
EXAMPLE 3.4.1, p176, 1+0.9*D channel
eye_eq([.9 1],1,0.181,'z')
pe_zfe = 0.1200 probability of error with Equalizer from N input 2PAM[+1/-1] data sequences
pe_no_eq = 0.1793 probability of error without Equalizer
pe_SNR = 0.3231(error probability estimation from dfseSNR, Pe = Q function of sqrt(dfseSNR))
dfseSNR = 0.2108 receiver(equalizer)SNR, unbiased in dB
Channel Frequency Response ZF EQUALIZER RESPONSE
10 20

5 15
M agnitude (dB)

M agnitude (dB)
0 10

-5 5

-10 0

-15 -5

-20 -10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

0 70

-10 60

-20 50
Phase (degrees)

P hase (degrees)
-30 40

-40 30

-50 20

-60 10

-70 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

ZF EQUALIZED CHANNEL , SNR =0.21075 dB EYE Diagram- before Equalizer (Channel output)
1

0.5 1
M agnitude (dB)

-0.5 0.5

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample)
0

0.5 -0.5
Phase (degrees)

-0.5 -1

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Frequency (×π rad/sample)

EYE Diagram - After ZFE

ZF EQUALIZER IMPULSE RESPONSE VS # of TAPS
1

1 0.8

0.6

0.5 0.4

0.2

0 0

-0.2

-0.5
-0.4

-0.6

-1
-0.8

-1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 10 20 30 40 50 60 70 80 90 100
EXAMPLE 3.4.2, p180
eye_eq([-.5 1+.25*i -.5*i],1,0.15625,'z')
pe_zfe = 0.5175
pe_no_eq = 0.5657
dfseSNR = -107.3157

Channel Frequency Response

6 ZF EQUALIZER RESPONSE
122
4
120
Magnitude (dB)

M agnitude (dB)
118
0
116
-2
114
-4
112
-6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 110
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample)
Normalized Frequency (×π rad/sample)

-20 0
-40 -500
Phase (degrees)

-60 -1000

Phase (degrees)
-80 -1500
-100 -2000
-120 -2500
-140 -3000

-160 -3500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

ZF EQUALIZED CHANNEL , SNR =-107.3157 dB

1 EYE Diagram- before Equalizer (Channel output)

0.5
M agnitude (dB)

-0.5 0.5

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample)
0

0.5
-0.5
Phase (degrees)

-0.5 -1

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Frequency (×π rad/sample)

EYE Diagram - After ZFE

0.5

-0.5

-1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

EXAMPLE 3.5.1, p187
eye_eq([.9 1],1,0.181,'m')
pe_mmse_le = 0.0155
pe_no_eq = 0.1768
pe_SNR = 0.0086
dfseSNR = 5.6834

MMSE EQUALIZER RESPONSE

Channel Frequency Response
2
10

5
0

Magnitude (dB)
Magnitude (dB)

-5 -2

-10
-4
-15

-20 -6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

0 0

-10
-500

Phase (degrees)
Phase (degrees)

-20
-1000
-30

-40 -1500
-50
-2000
-60

-70 -2500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

EYE Diagram- before Equalizer (Channel output)

MMSE EQUALIZED CHANNEL , SNR =5.6834 dB
0

-5 1
Magnitude (dB)

-10

-15

-20 0.5

-25

-30
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) 0

-500 -0.5
Phase (degrees)

-1000

-1500
-1
-2000

-2500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Frequency (×π rad/sample)

EYE Diagram - After MMSE-LE

MMSE EQUALIZER IMPULSE RESPONSE VS # of TAPS
0.6

1 0.5

0.4

0.5 0.3

0.2

0 0.1

-0.5 -0.1

-0.2

-1
-0.3

-0.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 5 10 15 20 25 30 35 40
EXAMPLE 3.5.2, p 190
eye_eq([-.5 1+.25*i -.5*i],1,0.15625,'m')
pe_mmse_le = 0.0129
pe_no_eq = 0.5480
dfseSNR = 6.7026 - 0.0000i

MMSE EQUALIZER RESPONSE

Channel Frequency Response 2
6
1
4
0

M agnitude (dB)
Magnitude (dB)

2 -1

0 -2

-2 -3

-4
-4
-5
-6 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample)
Normalized Frequency (×π rad/sample)

500
-20

-40 0

Phase (degrees)
-60
Phase (degrees)

-500
-80
-1000
-100
-1500
-120
-2000
-140

-160 -2500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

EYE Diagram- before Equalizer (Channel output)

MMSE EQUALIZED CHANNEL , SNR =6.7026-8.3099e-016i dB
0

1
-1
Magnitude (dB)

-2

0.5
-3

-4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) 0

-500
-0.5
Phase (degrees)

-1000

-1500
-1
-2000

-2500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Frequency (×π rad/sample)

EYE Diagram - After MMSE-LE

0.5

-0.5

-1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

EXAMPLE 3.6.1, p 201
eye_eq([.9 1],1,0.181,'d')
pe_mmse_dfe = 0.0129
pe_no_eq = 0.2247
pe_SNR = 0.0019
dfseSNR = 8.3573

FEEDFORW ARD SECTION MMSE-DFE EQUALIZER RESPONSE

Channel Frequency Response 0
10

Magnitude (dB)
-5
Magnitude (dB)

-5
-10
-10

-15
-15
-20 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

0 0

-10 -500

Phase (degrees)
Phase (degrees)

-20
-1000
-30

-40 -1500
-50
-2000
-60

-70 -2500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

EYE Diagram- before Equalizer (Channel output)

MMSE-DFE EQUALIZED CHANNEL, SNR =8.3573dB

20
1
10
Magnitude (dB)

-10
0.5
-20

-30

-40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0
Normalized Frequency (×π rad/sample)

1000

0
-0.5
Phase (degrees)

-1000

-2000

-3000 -1

-4000

-5000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Frequency (×π rad/sample)

EYE Diagram - After MMSE-DFE FEEDFORW ARD SECTION MMSE-DFE EQUALIZER IMPULSE RESPONSE VS # of TAPS
0.8

1 0.6

0.4

0.5

0.2

0 0

-0.2

-0.5

-0.4

-1 -0.6

-0.8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 10 20 30 40 50 60
EXAMPLE 3.6.2, p 203
eye_eq([-.5 1+.25*i -.5*i],1,0.15625,'d')
pe_mmse_dfe = 0.0052
pe_no_eq = 0.5884
dfseSNR = 8.3665 - 0.0000i

FEEDFORW ARD SECTION MMSE-DFE EQUALIZER RESPONSE

Channel Frequency Response -1
6

4 -1.5

Magnitude (dB)
Magnitude (dB)

2
-2
0

-2 -2.5

-4
-3
-6 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Normalized Frequency (×π rad/sample)
Normalized Frequency (×π rad/sample)

500
-20

-40 0

Phase (degrees)
Phase (degrees)

-60 -500
-80
-1000
-100
-1500
-120
-2000
-140

-160 -2500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) Normalized Frequency (×π rad/sample)

EYE Diagram- before Equalizer (Channel output)

MMSE-DFE EQUALIZED CHANNEL, SNR =8.3665-4.0732e-015idB
10

1
5
M agnitude (dB)

-5
0.5

-10

-15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Frequency (×π rad/sample) 0

1000

0
-0.5
Phase (degrees)

-1000

-2000

-3000 -1
-4000

-5000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalized Frequency (×π rad/sample)

EYE Diagram - After MMSE-DFE

0.5

-0.5

-1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2