Example

What is the minimum bandwidth for transmitting data at a

rate of 33.6 kbps without ISI?

Answer:
The minimum bandwidth is equal to the Nyquist
bandwidth.
Therefore, (BW)min = W = Rb/2 = 33.6/2 = 16.8 kHz

• Note: If a 100% roll-off characteristic is used,

bandwidth = W(1+α) = 33.6 kHz

BT.67

Example
Bandwidth requirement of the T1 system
T1 system
– multiplex 24 voice inputs, based on an 8-bit PCM word.
– bandwidth of each voice input (B) = 3.1 kHz

For converting the voice signal into binary sequence,

• The minimum sampling rate = 2B = 6.2 kHz
• Sampling rate used in telephone system =8 kHz

BT.68

1
 Example
With a sampling rate of 8 kHz, each frame of the
multiplexed signal occupies a period of 125µs.

:
: No. of bits = 8 ·24+1=193

8 bit from 8 bit from …. 8 bit from 1 bit for

1st input 2nd input 24th input Synchronization

125 µs BT.69

Example
Correspondingly, the bit duration is 125 µs/193 = 0.647 µs.

For eliminating ISI, the minimum transmission bandwidth is

1 / 2Tb = 772kHz

BT.70

2
 Eye diagrams
This is a simple way to give a measure of how severe the ISI
(as well as noise) is.

This pattern is generated by overlapping the incoming signal

elements. Example: bipolar NRZ PAM
Tb
1 0 1 1 0 0

BT.71

Eye diagrams
Eye pattern is often used to monitoring the performance of
baseband signal.
– The best time to sample the received waveform is when
the eye opening is largest.
Effects of noise 1
are ignored

BT.72

3
 Eye diagrams
The maximum distortion and ISI are indicated by the vertical
width of the two branches at sampling time.

BT.73

Eye diagrams
The noise margin or immunity to noise is proportional to the
width of the eye opening.

BT.74

4
 Eye diagrams
The sensitivity of the system to timing errors is determined
by the rate of closure of the eye as the sampling time is
varied.

BT.75

Equalization
In preceding sections, raised-cosine filters were used to
eliminate ISI. In many systems, however, either the channel
characteristics are not known or they vary.

Example
The characteristics of a telephone channel may vary as a
function of a particular connection and line used.

It is advantageous in such systems to include a filter that

can be adjusted to compensate for imperfect channel
transmission characteristics, these filters are called
equalizers.

BT.76

5
 Before equalization

After equalization

BT.77

Transversal filter (zero-forcing equalizer)

xk

T is the bit duration.

BT.78

6
 Equalization

The problem of minimizing ISI is simplified by considering

only those signals at correct sample times.

The sampled input to the transversal equalizer is

x(kT ) = x k
x0
x2
The output is
y (kT ) = y k

For zero ISI, we require that

1 k = 0 x1
yk = 
0 k ≠ 0 …(*)
BT.79

The output can be expressed as aN xk − N

N
yk = ∑ an xk −n xk

n=− N

−N ≤k≤N a0 xk
a− N xk + N

There are 2N+1 independent equations in terms of an . This

limits us to 2N+1 constraints, and therefore (*) must be
modified to
1 k =0
yk = 
0 k = ±1,±2,...,± N BT.80

7
 Equalization
The 2N+1 equations becomes
 xo x−1 L x− N L x− 2 N −1 x− 2 N   a − N   0 
    
 x1 x0 L x− N +1 L x− 2 N x− 2 N +1  a− N +1  0
 M M  M   M 
    
 xN xN −1 L x0 L x− N −1 x− N   a0  = 1
 M M  M   M 
    
 x2 N −1 x2 N − 2 L xN −1 L x− 2 x−1   a N −1  0
x x2 N −1 L xN x−1 x0   a N  0
 2N L

BT.81

Example
Determine the tap weights of a three-tap, zero-forcing
equalizer for the input where
x− 2 = 0.0, x−1 = 0.2, x0 = 1.0, x1 = −0.3, x2 = 0.1 , N=1

xk = 0 for k > 2

The three equations are

a−1 + 0.2a0 =0
− 0.3a−1 + a0 + 0.2a1 =1
0.1a−1 − 0.3a0 + a1 =0

Solving, we obtain
a−1 = −0.1779, a1 = 0.2847, a0 = 0.8897
BT.82

8
 Equalization

The values of the equalized pulse are

y− 3 = 0.0, y− 2 = −0.0356,
y−1 = 0.0, y0 = 1.0, y1 = 0.0,
y2 = 0.0036, y3 = 0.0285

This pulse has the desired zeros to either side of the peak,
but ISI has been introduced at sample points farther from
the peak.

BT.83

BT.84

9
 Duobinary Signaling

Intersymbol interference is an undesirable

phenomenon that produces a degradation in system
performance.

However, by adding intersymbol interference to the

transmitted signal in a controlled manner, it is possible
to achieve a signaling rate equal to the Nyquist rate of
2W symbols per second in a channel of bandwidth W
Hz.

BT.85

Coding and decoding

Consider a binary input sequence {bk } consisting of

uncorrelated binary symbols 1 and 0, each having
duration Tb . This sequence is applied to a pulse-
amplitude modulator producing a two-level sequence
of short pulses (approximating a unit impulse), whose
amplitude is

 1 if symobl bk is 1
ak = 
− 1 if symbol bk is 0

BT.86

10
 This sequence is applied to a duobinary encoder as
shown below: {ck }
Nyquist
{ak } channel

Delay
Tb
ck = ak + ak −1

One of the effects of the duobinary encoding is to

change the input sequence {a k } of uncorrelated two-
level pulses into a sequence {ck } of correlated three-
level pulses. This correlation between the adjacent
pulses may be viewed as introducing intersymbol
interference into the transmitted signal in an artificial
manner. BT.87

Example
Consider {bk }= 0010110 where the first bit is a startup bit.
Encoding:
{bk }: 0 0 1 0 1 1 0
{ak }: -1 -1 +1 -1 +1 +1 -1
{ck }: -2 0 0 0 +2 0

{ak }
t
Tb

{ck }
t

BT.88

11
Decoding:
Using the equation a k = ck − a k −1 , or simply using

If ck = +2 , decide that a k = +1.

If ck = −2 , decide that a k = −1.
If ck = 0 , decide opposite of the previous
decision.

BT.89

Duobinary Signaling: Impulse response and frequency

spectrum

Let us now examine an equivalent model of the

duobinary encoder. The Fourier transfer of a delay can
be described as e −2πfTb , therefore, the transfer function
of the encoder is H I ( f ) is
H I ( f ) = 1 + e − j 2πfTb

The transfer function of the Nyquist channel is

1 f < 1 / 2Tb
HN ( f ) = 
0 otherwise

BT.90

12
 The overall equivalent transfer function H ( f ) of the
is then given by
H ( f ) = H I ( f )H N ( f ) for | f |< 1 / 2Tb
= (1 + e − j 2πfTb )
= (e jπfTb + e − jπfTb )e − jπfTb
= 2e − jπfTb cosπfTb

H(f) has a gradual roll-off

to the band edge which
can be easily implemented

BT.91

The corresponding impulse response h(t) is found by taking

the inverse Fourier transform of H(f)

sin(πt / Tb ) sin(π (t − Tb ) / Tb )
h(t ) = +
πt / Tb π (t − Tb ) / Tb
sin(πt / Tb ) sin(πt / Tb )
= +
πt / Tb π (t − Tb ) / Tb
Tb2 sin(πt / Tb )
=
πt (t − Tb )

BT.92

13
Notice that there are only two nonzero samples, at Tb -second
intervals, give rise to controlled ISI from the adjacent bit.
The introduced ISI is eliminated by use of the decoding
procedure.
BT.93

14