SAMPLING THEORY AND PULSE MODULATION

5.1 Introduction
 There are two types of signals, continuous time signal and discrete-time signals.
 Due to some recent advance development in digital technology the inexpensive,
light eight, programmable and easily reproducible discrete-time systems are
available. Therefore, the processing of discrete-time signals is more flexible and
is also preferable to processing of continuous-time signals.
 For this purpose we should be able to convert a continuous-time signal into
discrete-time signal. The problem is solved by a fundamental mathematical tool
known a sampling theorem.
 With the help of sampling theorem, a continuous-time signal may be completely
represented and recovered from the knowledge of samples taken uniformly. It
provides a mechanism for representing a continuous-time signal by a discrete-
time signal. Therefore sampling theorem may be viewed as a bridge between
continuous-time signals and discrete-time signals.

5.2 Sampling theorem

 In discretizing a continuous-time signal sufficient number of samples of the
signal must be taken so that the original signal is represented in its samples
completely. The number of samples to be taken depends on the maximum
frequency present in the signal. Different types of samples are also taken like
ideal samples, natural samples and flat-top samples.
 The statement of sampling theorem can be given in two parts as:
 A bandlimited signal of finite energy, which has no frequency-component
higher than fm Hz, is completely described by its sample values at uniform
intervals less than or equal to 1 2fm second apart.
 A bandlimited signal of finite energy, which has no frequency-component
higher than fm Hz, may be completely recovered from the knowledge of its
samples taken at the rate of 2fm samples per second.
 The first part represents the representation of the signal in its samples and
minimum sampling rate required to represent a continuous-time signal into its
samples whereas the second part of the theorem represents reconstruction of the
original signal from its samples.
 Combining the two parts, the sampling theorem may be stated as A continuous-
time signal may be completely represented in its samples and recovered back if
the sampling frequency is fs  2fm . Here fs is the sampling frequency and fm is
the maximum frequency present in the signal.
5.3 Proof of sampling theorem

 Let us consider a continuous time signal x(t) whose spectrum is bandlimited to

fm Hz. This means that x(t) has no frequency component beyond fm Hz. Therefore
X  ω  is zero for ω  ωm , i.e., X  ω   0 for ω  ωm where ωm  2fm , as shown.

 Sampling of x(t) at a rate of fs Hz ( fs samples per second) may be achieved by

multiplying x(t) by an impulse train δ Ts  t  , as shown. The impulse δ Ts  t 
consists of unit impulses repeating periodically every Ts seconds, where Ts  1 fs
.

 This sampled signal consists of impulses spaces every Ts seconds (the sampling
interval) and can be represented as g  t   x  t  δ Ts  t  .
 Since the impulse train δ Ts  t  is a periodic signal of period Ts , it may be
expressed as trigonometric Fourier series as
1
δTs  t   1  2cos  ωs t   2cos  2ωs t   2cos  3ωs t   
Ts 
2π
where ωs   2πfs .
Ts
1
 Therefore, g  t   x  t  1  2cos  ωs t   2cos  2ωs t   2cos  3ωs t    .
Ts
 Therefore G  ω  , the Fourier transformation of g(t), is
1
G  ω   X  ω   X  ω  ωs   X  ω  ωs   X  ω  2ωs   X  ω  2ωs   
Ts 
 
1
Or, G  ω  
Ts
 X  ω  nωs 
n

Where Fourier transform of x(t) is X  ω  ,

Fourier transform of 2x  t  cos  ωs t  is X  ω  ωs   X  ω  ωs  ,
Fourier transform of 2x  t  cos  2ωs t  is X  ω  2ωs   X  ω  2ωs  and so on.
 From the above expression it is clear that the spectrum of G  ω  consists of X  ω 
repeating periodically with period ωs  2π Ts rad/sec of fs  1 Ts , as shown.

 Now if we have to reconstruct x(t) from g(t), we must be able to recover X  ω  from
G  ω  . This is possible if there is no overlap between successive cycles of G  ω  .
The figure requires of G  ω  shows that this is possible if fs  2fm or Ts  1 2fm .
 It may be also noted from the figure that the spectrum of sampled signals extends
up to infinity and the ideal bandwidth of sampled signal is infinite.
 Further the original or desired spectrum X  ω  is centered at ω  0 and is having
bandwidth or maximum frequency equal to ωm . The desired spectrum may be
recovered by passing the sampled signals with spectrum G  ω  through a low-
pass filter with cut-off frequency ωm .

5.4 Nyquist rate and Nyquist interval

 When the sampling rate becomes exactly equal to 2fm samples per second, then
it is called Nyquist rate. Nyquist rate is also called the minimum sampling rate
and is given by fs  2fm . Similarly, maximum sampling interval is called Nyquist
interval and is given by Ts  1 2fm .
 When the continuous-time bandlimited signal is sampled at Nyquist rate, the
sampled spectrum G  ω  contains non-overlapping repeating periodically. But
the successive cycles of touches each other, as shown. Therefore, the original
spectrum X  ω  can be recovered from the sampled spectrum by using a low pass
filter with cut-off frequency ωm .
Example 5.1
An analog signal is expressed by the equation

x  t   3cos 50πt   10sin  300πt   cos 100πt  .

Calculate the Nyquist rate for this signal.

Solution

We have ω1t  50πt , or f1  50π  2π   25 Hz . Similarly, f2  300π  2π   150 Hz and

f3  100π  2π   50 Hz .

Therefore the maximum frequency present in x(t) is f2  150 Hz and the Nyquist rate is
fs  2f2  300 Hz .

Example 5.2
Find the Nyquist rate and Nyquist interval for the signal

1
x t  cos  4000πt  cos 1000πt  .
2π

Solution:

1
X(t) can be written as x  t   cos 5000πt   cos  3000πt 
4π 

Therefore, we have, ω1t  5000πt or f1  5000π  2π   2500 Hz and ω2t  3000πt or,
f2  3000π  2π   1500 Hz .

Therefore the maximum frequency present in x(t) is f1  2500 Hz and the Nyquist rate
is fs  2f1  5000 Hz  5 kHz .

1 1 1
Nyquist interval is Ts     0.2  103 Sec  0.2 mS .
2fm 2  2500 5000
Example 5.3

A continuous-time signal is given by x  t   8cos  200πt  . Determine (i) Minimum

sampling rate, i.e., Nyquist rate required to avoid aliasing. (ii) If sampling frequency is
400 Hz, what is the discrete-time signal obtained after sampling? (iii) If sampling
frequency is 150 Hz, what is the discrete-time signal obtained after sampling?

Solution

(i) The highest frequency component of continuous-time signal is 100 Hz. Hence
minimum sampling rate required to avoid aliasing is fs  2f  200 Hz .
(ii) The continuous-time signal x(t) is sampled as 400 Hz. The frequency of the
discrete-time signal will be
Frequency of continuous-time signal  f  100 1
F  
Sampling frequency  fs  400 4
 πn 
Then the discrete-time signal will be given as x n  8cos  2πFn  8cos  
 2 
.
(iii) The continuous-time signal x(t) is sampled as 150 Hz. The frequency of the
discrete-time signal will be
f 100 2
F  
fs 150 3
Then the discrete-time signal will be given as
 4πn   2π    2πn 
x n  8cos  2πFn   8cos    8cos  2π   n  8cos  .
 3   3    3 

5.5 Signal reconstruction: The interpolation formula

 The process of reconstructing a continuous-time signal x(t) from its samples is
called interpolation.
 The expression for sampled signal is written as
1
g  t   x  t  δTs  t    x  t   2x  t  cos  ωs t   2x  t  cos 2ωst   
Ts 
From the above equation, it may be observed that the sampled signal contains a
component x  t  Ts .
 To recover x(t) or X  ω  , the sampled signal must be passed through an ideal low-
pass filter of bandwidth fm Hz and gain Ts . Therefore, the reconstruction or
 ω 
interpolating filter transfer function may be expressed as H  ω   Ts  Re ct  
 4πfm 
.
  The impulse response h(t) of this filter is the inverse Fourier transform of H  ω  ,
i.e.,
  ω 
h  t   F 1 H  ω  F 1  Ts  Rect     2fm Tssinc  2πfmt 
  4πfm  
 Assuming that sampling is done at Nyquist rate, then Ts  1 2fm and 2fmTs  1 .
Therefore, h  t   1.sinc  2πfmt  , as shown below.
From the figure it may be observed that h(t)=0 at all Nyquist sampling instants
t   n 2fm . Except at t = 0.
 Now, when the samples signal g(t) is applied at the input of this filter, the output
will be x(t). Each sample in g(t), being an impulse, produces a sinc pulse of height
equal to the strength of the sample. Addition of sinc pulses produced by all the
samples results in x(t).
 For instant, the kth sample of the input g(t) is the impulse x  kTs  δ  t  kTs  . The
filter output of this impulse will be x  kTs  h  t  kTs  . Therefore, the filter output
to g(t), which is x(t), may be expressed as a sum
x  t    x  kTs  h  t  kTs    x  kTs  sinc 2πfm  t  kTs 
k k

  k 
Or, x  t    x  kTs  sinc 2πfm  t     x  kTs  sinc  2πfmt  kπ 
k   2fm  k
This equation is known as interpolation formula, which provides values of x(t)
between samples as a weighted sum of all the sample values.

5.6 Effect of under sampling: Aliasing

 When a continuous-time bandlimited signal is sampled at a rate lower than the
Nyquist rate fs  2fm , then the successive cycles of the spectrum G  ω  of the
sampled signal g(t) overlap each other. As shown below.

 The signal is under-sampled in this case and some amount of aliasing is

produced in this process. Aliasing is the phenomenon in which a high frequency
 component in the frequency-spectrum of the signal takes identity of a lower-
frequency component in the spectrum of the sampled signal.
 Because of the overlap phenomenon, it is not possible to recover original signal
x(t) from the sampled signal g(t) by low-pass filtering since the spectral
components in the overlap regions ass and hence the signal is distorted.
 Since any information signal contains a large number of frequencies, so, to decide
a sampling frequency is always a problem. Therefore, a signal is first passed
through a low-pass filter. This low-pass filter blocks all the frequencies which are
above fm Hz. This process is known as band limiting of the original signal x(t).
This low-pass filter is called pre-alias filter because it is used to prevent aliasing
effect. After band limiting, it becomes easy to decide sampling frequency since
the maximum frequency is fixed at fm Hz.

5.7 Sampling of bandpass signals

 Previously we discussed sampling theorem for low-pass signals. However, when
the given signal is a bandpass signal, then a different criteria must be used to
sample the signal.
 The bandpass signal x(t) whose maximum bandwidth is 2fm can be completely
represented into and recovered from its samples if it is sampled at the minimum
rate of twice the bandwidth. Here, fm is the maximum frequency component
present in the signal.
 Hence, if the bandwidth is 2fm , then the minimum sampling rate for bandpass
signal must be 4fm samples per second.
 The spectrum of an arbitrary bandpass signal is shown below. The spectrum is
centered around frequency fc . The bandwidth is 2fm . Thus, the frequency
present in the bandpass signal are from fc  fm to fc  fm . This means that the
highest frequency present in the bandpass signal is fc  fm . Generally the center
frequency fc  fm .

 The bandpass signal can be represented in terms of its in-phase  xI  t  and
quadrature xQ  t  components as x  t   x I  t  cos  2πfc t   x Q  t  sin  2πfc t  .
  The in-phase and quadrature components are obtained by multiplying x(t) by
cos  2πfc t  and sin  2πfc t  and then suppressing the sum frequencies by means
of low-pass filters. Thus, in-phase  xI  t  and quadrature xQ  t  components
contain only low frequency components. The spectrum of these components is
limited to fm to fm , as shown below.

 After mathematical manipulation it can be shown that x(t) can be written as


 n   n   n 
x t   x   sinc  2fmt   cos 2πfc  t  
n  4fm   2   4fm  
 Comparing this reconstruction formula with that of low-pass signal given before,
 n   n 
we observe that x(t) is replaced by x   . Here, x    x  nTs  is the sampled
 4fm   4fm 
1
version of bandpass signal and Ts  .
4fm
 Thus, if 4fm samples per second are taken, then the bandpass signal of
bandwidth 2fm can be completely recovered from its samples.

5.8 Sampling technique

 Basically, there are three types of sampling techniques as under

 Instantaneous sampling
 Natural sampling
 Flat-top sampling
 Out of these three instantaneous sampling is called ideal sampling whereas
natural sampling and flat-top sampling are called practical sampling methods.

5.8.1 Ideal sampling / Instantaneous sampling / Impulse sampling

 In this type of sampling, the sampling function is a train of impulses, as shown.

X(t) is the input signal. The circuit is called switching sampler. It simply consists
of a switch.
 If we assume that the closing time ‘t’ of the switch approaches zero, then the
output g(t) of this circuit will contain only instantaneous value of the signal x(t).

 We know that the train of impulses may be represented as δTs  t    δ  t  nTs 
n

. This is known a sampling waveform.

 The sampled signal g(t) is expressed as the multiplication of x(t) and δ Ts . Thus,
 
g  t   x  t  δTs  t   x  t  .  δ  t  nTs    x  nTs  δ  t  nTs 
n n

 The Fourier transform of the ideally sampled signal given by above equation may

be expressed as G  f   fs  X  f  nfs  .
n
  This equation gives the spectrum of ideally sampled signal. It shows that the
spectrum X(f) is periodic in fs . However, it may be noted that ideal or
instantaneous sampling is possible only in theory since it is impossible to have a
pulse whose width approaches zero.

5.8.2 Natural sampling

 The instantaneous sampling results in the samples whose width approaches to

zero. Due to this, the power content in the instantaneous sampled pulse is
negligible. Thus, this method is not suitable for transmission purpose. Natural
sampling is a practical method. In natural sampling pulse has a finite width τ .
 Let us consider an analog continuous-time signal x(t) to be sampled at the rate
fs Hz. Here it is assumed that fs is higher than Nyquist rate such that sampling
theorem is satisfied.
 Again let us consider a sampling function c(t) which is a train of periodic pulses
of width τ and frequency equal to fs Hz.
 The following figure shows a functional diagram of a natural sampler. With the
help of this natural sampler, a sampled signal g(t) is obtained by multiplication
of sampling function c(t) and input signal x(t).

 When c(t) goes high the switch ‘S’ is closed. Therefore g(t)=x(t) when c(t)=A and
g(t)=0 when c(t)=0, where A is the amplitude of c(t). The waveforms of signals x(t),
c(t) and g(t) are shown below. The sampled signal g(t) may be described
mathematically as g  t   c  t  x  t  .
 We know that the exponential Fourier series of any periodic waveform is

expressed as s  t    Cne j2πnt T0 .
n

 For the periodic pulse train c(t), we have To  Ts  1 fs = period of c(t).


Alternatively, fo  fs  1 T0  1 Ts = frequency of c(t). Therefore, c  t    Cne j2πfsnt
n

.
 Since c(t) is a rectangular pulse train, therefore Cn for this waveform will be
τA
expressed as Cn  sinc  fn τ  , where τ is the pulse width and fn is the harmonic
Ts
frequency.
 
τA τA
Therefore, c t  T sinc n 
f τ e j2πfs nt
and g    sinc  fn τ  e j2πfsnt .x  t  .
t 
n s n Ts

This is required time-domain representation for naturally sampled signal g(t).

 To get the frequency-domain representation of the naturally sampled signal g(t),
let us take its Fourier transform as
τA 
 G  f   FT g  t   
Ts n
sinc  fn τ  FT e j2πfsnt .x  t  .

 Recall the frequency-shifting property of Fourier transform which states that

τA 
e j2πfs nt .x  t   X  f  nfs  . Therefore, G  f    sinc  fn τ  X  f  nfs  .
Ts n
 Since fn  nfs =harmonic frequency, we can write,
 τA 
G f    sinc  nfs τ  X  f  nfs  . This equation shows that the spectra of x(t),
Ts n
i.e., X(f) are periodic in fs and are weighted by the sinc function. The following
figure illustrates some arbitrary spectra of x(t) and corresponding G(f).

5.8.3 Flat top sampling / rectangular pulse sampling

 Flat top sampling like natural sampling is also a practically possible sampling
method. But natural sampling is little complex whereas it is quite easy to get flat
top samples.
 In flat-top sampling or rectangular pulse sampling, the top of the samples
remains constant and is equal to the instantaneous value of the baseband signal
x(t) at the start of sampling.
 The duration or width of each sample is τ and sampling rate is equal to fs  1 Ts
.
 The functional diagram of a sample and hold circuit which is sued to generate
the flat top samples is shown below. The general waveform of flat top samples is
also shown.
 From the figure it may be noted that only starting edge of the pulse represents
instantaneous value of the baseband signal x(t). The flat top pulse g(t) is
mathematically equivalent to the convolution of instantaneous sample and a
pulse h(t) as shown below.

 The starting edge of the pulse represents the point where baseband signal is
sampled and width is determined by function h(t). Therefore g(t) will be expressed
as g  t   s  t   h  t  . This equation is explained in the following figure. On
convolution of s(t) and h(t), we get a pulse whose duration is equal to h(t) only
but amplitude is defined by s(t).
 The train of impulses may be represented mathematically as

δTs  t    δ  t  nTs  .
n

 The signal s(t) is obtained by multiplication of baseband signal x(t) and δ Ts  t  .

Thus
 
s  t   x  t  δTs  t   x  t   δ  t  nTs    x  nTs  δ t  nTs 
n n

 The sampled signal g(t) is

  
g t  s t  h t   s  τ  h  t  τ  dτ   n

x  nTs δ  t  nTs  h  t  τ  dτ
 
 
Or, g  t    x  nTs   δ  t  nTs  h  t  τ  dτ
n  

 According to shifting property of delta function, we know that

 

 f  t  δ  t  t0  dt  f  t0  . Therefore g  t    x  nTs  h  t  nTs  . This equation

 n
 represents value of g(t) in terms of sampled value x  nTs  and function h  t  nTs 
for flat top sampled signal.

 Taking Fourier transform of g  t   s  t   h  t  we have G  f   s  f  H  f  .

 
 S(f) is given as S  f   fs  X  f  nfs  . Therefore, G  f   fs  X  f  nfs  H  f  .
n n
5.9 Aperture effect

 The spectrum of flat top sampled signal is expressed as


G  f   fs  X  f  nfs  H  f  . This equation shows that the signal g(t) is obtained
n

by passing the signal s(t) through a filter having transfer function H(f). The
corresponding response h(t) in time-domain is shown below along with its
spectrum. The spectrum of a rectangular pulse is H  f   τsinc  fτ  e  jπfτ  A  1 .
This is one pulse of rectangular pulse train. Each sample of x(t) is convolved with
the pulse.

 From the figure it may be observed that by using flat top samples an amplitude
distortion is introduced in the reconstructed signal x(t) from g(t). In fact high
frequency roll-off of H(f) acts like a low-pass filter and thus attenuates the upper
portion of message spectrum. These high frequencies of x(t) are affected. This
type of effect is known as aperture effect.
 As the duration ‘ τ ’ of the pulse increases, the aperture effect is more prominent.
Hence, during reconstruction an equalizer is needed to compensate this effect.
As shown below, the receiver contains a low-pass reconstruction filter with cut-
off frequency slightly higher than the maximum frequency present in the message
signal. The equalizer compensates for the aperture effect. It also compensates for
the attenuation by the low-pass reconstruction filter.

Ke  j2πftd
 The transfer function of the equalizer is expressed as Heq  f   . Here ‘ td
Hf 
’ is known as the delay introduced by low-pass filter which is equal to τ 2 .
 jπfτ
Kee K
 Therefore, Heq  f    .
τsinc  fτ  ee  jπfτ
τsinc  fτ 
Example 5.4

The following figure shows the spectrum of an arbitrary signal x(t). This signal is
sampled at the Nyquist rate with a periodic train of rectangular pulses of duration 50/3
milliseconds. Determine the spectrum of the sampled signal for frequencies up to 50 Hz
giving relevant expression.

Solution

From the figure it may be observed that the signal is bandlimited to 10 Hz. Thus
fm  10 Hz . So the Nyquist rate is 2fm  20 Hz .

Since the signal is sampled at the Nyquist rate, the sampling frequency would be
fs  20 Hz . Given that the rectangular pulses are used for sampling, i.e., flat top
sampling is used.

50 0.05
The given value of rectangular sampling pulse duration is τ   103  sec .
3 3

0.05  0.05f   j0.05πf

Therefore, H  f   τsinc  fτ  e jπfτ  sinc  e
3
and
3  3 

0.05  0.05f   j0.05πf
G  f   20  X  f  20n  sinc  e
3

n 3  3 

1 3  0.05f   j0.05πf
Or, G  f   
3 n3
X  f  20n sinc 
 3 
e
3

This expression gives the spectrum up to 60 Hz (since n  3 ) for the sampled signal.

Example 5.5

A flat top sampling system samples a signal of maximum frequency 1 Hz with 2.5 Hz
sampling frequency. The duration of the pulse is 0.2 seconds. Compute the amplitude
distortion due to aperture effect at the highest signal frequency. Also determine the
equalization characteristics.
Solution:

The magnitude of the transfer function H(f) is H  f   τsinc  fτ   0.2sinc  0.2f  .

Aperture effect at the highest frequency 1 Hz is H1  0.2sinc 0.2  0.1871  18.71% .

K K
Equalizer characteristics is Heq  f   
τsinc  fτ  0.2sinc  0.2f 

5.10 Analog pulse modulation methods

 In pulse modulation methods, the carrier is no longer a continuous signal but

consists of a pulse train. Some parameters of which is varied according to the
instantaneous value of the modulating signal.
 There are two types of pulse modulation system
 Pulse amplitude modulation (PAM)
 Pulse time modulation (PTM)
 In pulse amplitude modulation (PAM), the amplitude of the pulses of the carrier
pulse train is varied in accordance with the modulating signal whereas in pulse
time modulation (PTM), the timing of the pulses of the carrier pulse train is varied.
 There are two types of PTM
 Pulse width modulation (PWM)
 Pulse position modulation (PPM)
 In pulse width modulation (PWM), the width of the pulses of the carrier pulse
train is varied in accordance with the modulating signal whereas in Pulse position
modulation (PPM), the position of pulses of the carrier pulse train is varied.
 All the above pulse modulation methods (i.e., PAM, PWM and PPM) are called
analog pulse modulation methods because the modulating signal is analog in
nature. Different types of pulse analog modulation methods are shown below.
  According to the sampling theorem, if a modulating signal is bandlimited to fm
Hz, the sampling frequency must be at least 2fm Hz, and, hence the frequency of
the carrier pulse train must also be at least 2fm Hz.

5.11 Pulse amplitude modulation (PAM)

 Pulse amplitude modulation is defined as that type of modulation in which
amplitudes of regularly spaced rectangular pulses vary according to
instantaneous value of the modulating or message signal.
 The pulse in a PAM signal may be of flat top or natural type or ideal type. Actually,
all the sampling methods which have been discussed are basically pulse
amplitude modulation methods.
 Out of these there pulse amplitude modulation methods the flat top PAM is most
popular and is widely used. The reason for using flat top PAM is that during the
transmission, the noise interferences with the top of the transmitted pulses and
this noise can be easily removed if the PAM pulses has flat top.
 In case of natural samples PAM signal, the pulse has varying top in accordance
with the signal variation. Now when such type of pulses is received at the receiver,
it is always contaminated by noise. Then it becomes quite difficult to determine
the shape of the top of the pulse and thus amplitude detection of the pulse is not
exact. Due to this, errors are introduced in the received signal.
 A sample and hold circuit (shown below) is used to produce flat top sampled PAM.
The circuit consists of two field effect transistors switches and a capacitor. The
sampling switch is closed for a short duration by a short pulse applied to the gate
G1 of the transistor. During this period, the capacitor ‘C’ is quickly charged up to
a voltage equal to the instantaneous sample value of the incoming signal x(t).
Now the sampling switch is opened and the capacitor ‘C’ holds the charge. The
discharge switch is then closed by a pulse applied to the gate G 2 of other
transistor. Due to this, the capacitor ‘C’ is discharged to zero volts. The discharge
switch is then opened and thus the capacitor has no voltage. Hence the output
of the sample and hold circuit consists of a sequence of flat top samples.
5.11.1 Transmission bandwidth in PAM
 In a pulse amplitude modulated (PAM) signal the pulse duration ‘ τ ’ is considered
to be very very small in comparison to time period (i.e., sampling period) Ts
between any two samples, i.e., τ Ts .
 Now, if the maximum frequency in the modulating signal x(t) is fm , then
according to sampling theorem, the sampling frequency fs must be equal to or
1 1
higher than the Nyquist rate, i.e., fs  2fm , or, Ts  . Therefore, τ Ts  .
2fm 2fm
 If the ‘On’ and ‘off’ time of the pulse amplitude modulated (PAM) pulse is same as
1
shown below, then the maximum frequency of the PAM pulse will be fmax  .
2τ

 Therefore, the bandwidth required for the transmission of a PAM signal would be
equal to the maximum frequency fmax . Thus, we have Transmission bandwidth
1
BW  fmax or BW  .
2τ
1 1
 Again since, τ , BW  fmax or, BW fmax .
2fmax 2τ

Example 5.6

For a pulse-amplitude modulated (PAM) transmission of voice signal having maximum

frequency equal to fm  3 kHz , calculate the transmission bandwidth. It is given that
the sampling frequency fs  8 kHz and the pulse duration τ  0.1Ts .

Solution

1 1
Ts   =0.125 103 S  125 μS
fs 8 103

Therefore, τ  0.1Ts  12.5 μS

 1 1
Now, BW  , Therefore, BW  or, BW  40 kHz
2τ 2  12.5  10 6

5.11.2 Demodulation of PAM signals

 For pulse amplitude modulated (PAM) signals, the demodulation is done using a
holding circuit, as shown below. In this method the received PAM signal is allowed
to pass through a holding circuit and a low-pass filter.

 The following figure illustrates a very simple holding circuit. Here ‘S’ is closed
after the arrival of the pulse and it is opened at the end of the pulse. In this way,
the capacitor C is charged to the pulse amplitude value and it holds this value
during the interval between the two pulses. Hence, the sampled values are held
as shown in the figure.

5.11.3 Drawbacks of PAM signals

 Following are the drawbacks of a PAM signal
 The bandwidth required for the transmission of a PAM signal is very very
large in comparison to the maximum frequency present in the modulating
signal.
 Since the amplitude of the PAM pulses varies in accordance with the
modulating signal therefore the interference of noise is maximum in a PAM
signal. This noise cannot be removed easily.
 Since the amplitude of the PAM signal varies, therefore, this also varies
the peak power required by the transmitted with modulating signal.
  After this the holding circuit output is smoothened in low-pass filter, as shown.
It may be observed that some kind of distortion is introduced due to the holding
circuit.

 The circuit is known as zero holding circuit and it considers only the previous
sample to decide the value between the pulses.
 A first order holding circuit considers the previous two samples whereas a second
order holding circuit considers the previous three samples and so on. However,
as the order of the holding circuit increases, the distortion decreases at the cost
of the circuit complexity. In fact, the amount of permissible distortion decides the
order of the holding circuit.

5.11.3 Transmission of PAM signals

 If the PAM signals are to be transmitted directly, i.e., over a pair of wires then no
further signal processing is necessary. However, if they are to be transmitted
through the space using an antenna, they must first be amplitude or frequency
or phase modulated by a high frequency carrier and only then they can be
transmitted. Thus, the overall system will be then known as PAM-AM or PAM-FM
or PAM-PM respectively.
 At the receiving end, AM or FM or PM detection is first employed to get the PAM
signal and then the message signal is recovered from it.

5.12 Pulse time modulation

 In pulse time modulation, amplitude of pulse is held constant, whereas position
of pulse (PPM) or width of pulse (PWM) is made proportional to the amplitude of
signal at the sampling instant. Because the amplitude is held constant and does
not carry any information, therefore amplitude limiters can be sued.
 The amplitude limiters, similar to those used in FM, will clip off the portion of the
signal corrupted by noise and hence provide a good degree of noise immunity.
5.12.1 Pulse width modulation

 Pulse width modulation is also known as pulse duration modulation (PDM).

 Three variations of pulse width modulation are possible. In one variation, the
leading edge of the pulse is held constant and change is pulse width with signal
is measured with respect to the leading edge. In other variation, the tail edge is
held constant and with respect to it, pulse width is measured. In the third
variation, venter of the pulse is held constant and pulse width changes on either
side of the center of the pulse. This is illustrated in the following figure. The
modulating signal is at its positive peak at point (A) and at its negative peak at
(B).

5.12.1.1 Frequency spectrum of PWM

 With a sinusoidal modulating signal at frequency fm , the spectrum of PWM signal

consists the modulating signal frequency fm along with several harmonics. This
is shown below.

 To have a better separation with respect to frequency, between highest frequency

of baseband signal ( fm ) and lower sidebands fs (sampling frequency), a higher
sampling frequency which is more than Nyquist rate is used; and pulse deviation
is kept small.
5.12.1.2 Modulation of PWM signal or PWM generation

 A pulse width modulator is shown below. It is basically a monostable

multivibrator with a modulating input signal applied at the control voltage input.

2
 Internally, the control voltage is adjusted to VCC . Externally applied modulating
3
signal changes the control voltage, and hence the threshold voltage level. As a
result, the time period required to charge the capacitor up to threshold voltage
level changes, giving pulse modulated signal at the output, as shown below.
5.12.1.3 Demodulation of PWM signal

 The block diagram of PWM detector is shown below. The received PWM signal is
applied to the Schmitt trigger circuit. This Schmitt trigger circuit removes the
noise in the PWM waveform. The regenerated PWM is then applied to the ramp
generator and synchronization pulse detector.
 The ramp generator produces ramps of the duration of pulses such that height
of ramps are proportional to the widths of PWM pulses. The maximum ramp
voltage is retained till next pulse. On the other hand, synchronous pulse detector
produces reference pulses with constant amplitude and pulse width. These
pulses are delayed by specific amount of delay.
 The delayed reference pulses and the output of ramp generator is added with the
help of adder. The output of adder is given to the level shifter. Here, negative
offset shifts the waveform. Then the negative part of the waveform is clipped by
rectifier.
 Finally, the output of rectifier is passed through low-pass filter to recover the
modulating signal, as shown above.
5.12.1.4 Advantages of PM

 Unlike, PAM, noise is less, since in PWM, amplitude is held constant.

 Signal and noise separation is very easy.
 PWM communication does not require synchronization between transmitter and
receiver.

5.12.1.5 Disadvantages of PWM

 In PWM, pulses are varying in width and therefore their power contents are
variable. This requires that the transmitter must be able to handle the power
contents of the pulse having maximum pulse width.
 Large bandwidth is required for the PWM communication as compared to PAM.

5.12.2 Pulse position modulation

 In pulse position modulation the amplitude and width of the pulses are kept
constant, while the position of each pulse, with reference to the position of a
reference pulse, is changed according to the instantaneous sampled value of the
modulating signal. Thus, the transmitter has to send synchronizing pulses to
keep the transmitter and receiver in synchronism.
 As the amplitude and width of the pulses are constant, the transmitter handles
constant power output, a definite advantage over the PWM.
 The disadvantage of the PPM system is the need for transmitter-receiver
synchronization.
 Pulse position modulation is obtained from pulse width modulation, as shown
below. The position of the pulse is proportional to the width of the pulse in PWM
and hence it is proportional to the instantaneous amplitude of the sampled
modulating signal.
5.12.2.1 Generation of PPM signal
 The following figure shows the PPM generator. It consists of differentiator and a
monostable multivibrator.
 The input to the differentiator is a PWM waveform. The differentiator generates
positive and negative spikes corresponding to leading and trailing edges of the
PWM waveform.
 Diode D1 is used to bypass the positive spikes. The negative spikes are used to
trigger monostable multivibrator.
 The monostable multivibrator then generate the pulses of same width and
amplitude in reference to trigger to give pulse position modulated waveform, as
shown below.
5.12.2.2 Demodulation in PPM

 In case of pulse-position demodulation, it is customary to convert the received

pulses that vary in position to pulses that vary in length. One way to achieve this
is shown below.
 The flip-flop circuit is set or turned on (giving high input) when the reference
pulse arrives. This reference pulse is generated by reference pulse generator of
the receiver with the synchronization signal from the transmitter.
 The flip-flop circuit is reset or turned off (giving low output) at the leading edge
of the position modulated pulse.
 The PWM pulses are then demodulated by PWM demodulator to get original
modulating signal.

5.12.2.3 Advantage of PPM

 Like PWM, in PPM, amplitude is held constant thus less noise interference.
 Like PPM, signal and noise separation is very easy.
 Because of constant pulse widths and amplitudes, transmission power for each
pulse is same.

5.12.2.4 Disadvantage of PPM

 Synchronization between transmitter and receiver is required

 Large bandwidth is required as compared to PAM.