Inter symbol Interference

Definition:
In a communication system when the data is being transmitted in the
form of pulses (bits), the output produced at the receiver due to the
other bits or symbols interferes with the output produced by the
desired bit. This is called inter symbol interference (ISI). This inter
symbol interference will introduce error in the detected signal at the
receiver.

Baseband binary data transmission system

Input Transmitter Transmission Channel

binary
data (bk) Pulse X(t) Transmitting
Channel 1
Generator Filter

Output
Clock pulses White Gaussian noise
Binary
Y1(t) Y2(t) Receiving
Data Decision
device Filter
Sample at
time t=iTb

Threshold

Receiver
Consider the baseband binary PAM system. The input signal consists
of binary data sequence (bk) with a bit duration of Tb seconds. This
sequence is applied to pulse generator which produces a discrete PAM
signal (line code).
( ) ∑ ( ) (1)

Where v(t) denotes the basic pulse normalized such that v(0)=1. The
first block of the system is pulse amplitude modulator which converts
the input sequence into polar form i.e.
If bk = 1 then ak = 1
and bk = 0 then ak = -1

The PAM signal x(t) is then passed through a transmitting filter. The
output of the transmitting filter is then transmitted over transmission
channel. The impulse response of the channel is h(t). A random noise
is then added to the transmitted signal when it travels over the
transmission channel. Thus the signal received at the receiving end is
contaminated one. The channel output is applied to receiving filter.
This filter output is sampled synchronously with the transmitter. The
sampling instants are determined by clock or timing signal which is
then extracted from the receiving filter output. A sequence of samples
is obtained at the output of receiving filter which is used to reconstruct
the original data sequence with the help of a decision making device.
Each sample is compared with a predetermined threshold level in the
decision making device. If the amplitude of the sample is higher than
the threshold level then it is decided that a symbol 1 is received.
On the other hand if the signal has amplitude lower than the threshold,
then the decision is that a 0 is received. The receiving filter output can
be written as
( ) ∑ ( ) ( ) ………………….. (2)

Where is a scaling factor and the noise n(t) is the noise produced at
the output of the receiving filter due to the channel added noise. The
term ( ) represents the combined impulse response of the
receiving filter. The receiving filter output y(t) is sampled at the time
instant with i=0,1,2… .
This results in sampled version of y(t) as follows:
( ) ∑ ( ) ( )………………..(3)
( ) ∑ ( ) ( )………….(4)

This is the receiver output y(t) at instant t= .

Equation (4) has two terms:
1) The first term is produced by the ith transmitted bit.
Theoretically, only this term should be present at the receiver
output but practically it is not so.
2) The second term represents the collective residual effect of all the
transmitted bits, corresponding to the sampling time instant t=ti.
This residual effect is known as the inter symbol interference (ISI).
Causes of Inter Symbol Interference
The ISI results because the overall frequency response of the system is
never perfect and pulse spreading is bound to take place. When the
short pulse of duration Tb seconds is transmitted through a band
limited transmission system, then various frequency components
present in system then various frequency components present in the
input pulse are differentially attenuated and more importantly
differentially delayed by the system. Due to this the pulse appearing at
the output of the system will be considered over an interval which is
longer than Tb seconds. Due to this dispersion the adjacent symbols
Input Output

Transmission
Dispersed pulse
System

Duration
Tb longer than
Tb

will interfere with each in time domain other when transmitted over
the communication channel. This will result in the inter symbol
interference. The transmitted pulse of duration Tb seconds and the
dispersed pulse of duration more than Tb seconds are shown in
figure. The four causes of ISI are
1) Timing inaccuracies
2) Amplitude distortion
3) Insufficient bandwidth
4) Phase distortion
Timing inaccuracies:
If the transmitter rate of transmission is not same as the ranging
frequency of the given channel then ISI will take place.
Insufficient bandwidth:
If the transmission rate is less than the channel bandwidth then there is
a small possibility of timing error. If we reduce the channel bandwidth
then timing error will increase and ISI will also increase.
Amplitude distortion:
The frequency response of communication channels cannot be
predicted accurately. If frequency characteristics of communication
channel differ from expected value, the pulse distortion will take
place. Pulse distortion results in reduction of peaks of pulses.
Phase distortion:
Various frequency components in the input pulse undergo different
amount of time delay while travelling through the channel. Then
phase distortion will take place.
Effect of ISI:-
ISI results in the introduction of error. Due to this receiver can make
an error in deciding whether it has received logic 1 or logic 0. Another
effect of ISI is the cross talk which may take place due to overlapping
of the adjacent pulses due to spreading. If ISI and noise are absent
then transmitted bit can be decoded correctly at the receiver.
How to reduce ISI?
It is necessary to use the special filters called equalizers in order to
reduce ISI and its effect. The function which produces zero inter
symbol interference is a sinc function. Thus instead of a rectangular
pulse if we transmit the sinc pulse then Isi can be reduced. Using the
sinc pulse for transmission is known as “Nyquist Pulse Shaping.” We
know that the
WeWW
Shape of the
input for zero ISI
is sinc shape

Output
Ideal pulse shape for zero
ISI
Frequency response of
reconstruction filter

Frequency response of filter

We know that the Fourier transform of transform of a sinc pulse is a
rectangular function. Therefore to preserve all the frequency
components, the frequency response of the filter must be exactly flat
in the pass band and zero in the attenuation band. This type of filter is
practically not available. Therefore practically the frequency response
of the filter is modified with different roll off factor “ to obtain
practically achievable filter response curves.

𝛼 0

𝛼 1/2

𝛼 1

f/ B0
0 1 3/2 2

Practical Filter Characteristics