Table of Contents

ADCS (Sem. 5/ ECE GTN

3.6.3 Phase Modulation (P.M.)... 3-22 3.16.2 Superheterodyne Receivers....
3.6.4 .J-44
Frequency Modulation (F.M.).*** 3-22 3.16.3 Waveforms at Various Points of a
3.6.5 Generation of F.M. using Phase Superheterodyne
Receiver ..
3-48
Modulator ...3-23 3.16.4 Frequency Spectrums at Various Points
3.6.6 Generation of P.M. using of Superheterodyne
a Frequency a
ceiver... 3-48
Modulator 3.16.5 Advantages of
...3-24
Superheterodyning
3.6.7 Squared Modulation 3-24 3.16.6 Frequency Parameters of AM
3.7 Generation of FM Waves.. 3-29 Receiver '****
*******'***°** . . J-47
47
3.7.1 Direct F.M. **********
.. **** ** ** ****** .3-29 3.17 Receiver Characteristics *****'***********'******'** 3-47
3.7.2 Varactor Diode Modulator ... .3-30 3.17.1 Sensitivity..
***********'***'**
*******.. . J47
3.7.3 Advantage of Direct FM 3.17.2 Selectivity.
Generation.. 3-31 347
3.7.4 Disadvantages of Direct Method . 3-31 3.17.3 Fidelity... 3-48
3.8 Effect of Mixing and Multiplication on FM 3.17.4 Image Frequency and Image Frequency
Wave. 3-31 Rejection Ratio(IFRR)..
3.8.1 Effect of Mixing 3-31 3.17.5 Double Spotting ***'****************** . 3-48
3.8.2 Effect of Multiplication ******** ******* .3-31 3.18 FM Receiver.. 3-49

3.9 Indirect Method (Amstrong Method) of FM 3.18.1 Difference between FM and AM

Generation ..*** ************"* ** ******..3-32 Receivers * 49

3.10 Effect of Noise in FM and PM . . .3-33 3.18.2 FM Receiver Block Diagram with
3.10.1 Advantages of FM. . 3-33 Waveforms *******.
349

3.19 Stereophonic FM Broadcast System . . 3-49

3.10.2 Disadvantages of FM .**
. ************** 3-33
3.20 Stereo FM Multiplex Reception. 3-50
3.10.3 Applications of FM .3-33
3.21 Interference in Angle Modulated
3.11 Pre-emphasis and De-emphasis ..3-33

3.11.1 Pre-emphasis... ************** *******3 - 3 3

Systems. *************************** ******************.3-51

3.21.1 Adjacent Channel Interference

nce....... J-51
3.11.2 De-emphasis.. .3-34
3.21.2 Co-Channel Interference (Capture Effect)
3.12 Basic FM Demodulators ***"**.**.***' .3-35
inFM Receivers.... *

"*** *** *
.3-52
"****
3.12.1 Principle of Slope Detection. .-36
3.21.3 Comparison of AM, FM and PM.. -
3.12.2 Simple Slope Detector . . . *******"
.3-36
3.22 Advantages and Disadvantages of Angle
3.12.3 Balanced Slope Detector....
************"*******.3-37 Modulation. .. . 3-53

3.12.4 Phase Discriminator [Foster Seeley 3.22.1 Advantages * * * * *

.3-53

Discriminator. ***'**********************'***"
3-38
53
3.22.2 Disadvantages of Angle Modulation..
3.12.5 Ratio Detector.. .3-40
3.23 University Questions and Answers
3-54

3.12.6 Comparison of FM Demodulators.341 3-53

Review Questions..
3.13 Zero Crossing Detector as Frequency
4-1 to 4-22
Demodulator * ****** *********'****"**'* ..3-42
Chapter 4: Sampling
3.14 FM Detection using PLL..***************** *** **..3-42 Syllabus: Sampling theorem, Sampling and
SIgnal

3.15 Nonlinear Effects in FM Systems . .. 43 reconstruction, Aliasing, Types of sampling.

3.15.1 Nonlinear effects on the FM system.3-43 4.1 Introduction... *********'
*'********"""

,4-2
3.15.2 Phase Nonlinearities..
*********'********i*.
.....3-444 4.1.1 Baseband Systems *******.** ***

3.16 AM Receivers 3-44 4.1.2 Formatting Textual Data (Character 4-2

.16.1 Functions of a Receiver ****************".

344
Coding) ****'****** **'****

Jenil Thakkar
echkinewled
UDIlCa tion.
 5 ADCS (Sem. 5/ ECE/ GTU)
Table of Contentss
4.1.3 Messages, Characters and Symbols
...4-3 5.3.2 PCM Transmitter (Encoder) 5-5

Information
...4-3 5.3.3 Shape of the PCM Signal .**************'***'*" 5-6
4.2 Formatting Analog .

4-3 5.3.4 PCM Transmission Path ******' *** *** 5-6

4.2.1 Sampling
4.2.2 Low Pass and Band Pass Signals *****
.4-3 5.3.5 PCM Receiver (Decoder) ''*'''*''**'''*'. 5-6
**:

4-4 5.3.6 Quantization Process -/

4.2.3 Sampling Process *** '* * '*
''*****

5.4 Derivation of Expression for the Quantization

Sampling Theorem.
4 - 5

4.3 ******* ********

4-5
Error ************* ** *********'*** * '"* 5-8
4.3.1 Proof of Sampling Theorem. * * * *

4-9 5.5 Expression for the Maximum Signal to

4.4 Recovery using Ideal Low Pass Filter ******

Quantization Noise Ratio[S/ Na] 5-9

4.5 Interpolation . **''** *** ****** '* ********'*******.4-9
5.6 Signaling Rate and Bandwidth of PCM '***** ...5-10
4.5.1 InterpolationFormula.. 4-10

5.7 Effect of Noise in PCM System

*** *'**
.5-10

4.6 Aliasing or Foldover Error .4-11

5.8 Robust Quantization ..5-11
4.6.1 How to Eliminate Aliasing ?. * * * *4 - 1 2
5.8.1 Nonuniform Quantization . . .5-11

4.6.2 Nyquist Rate and Nyquist Interval...4-12

5.9 Companding (Companded PCM) *** ** .5-12
'*
4.6.3 Effect of Non ldeal Reconstruction
Filter 4-12
5.9.1 Compressor Characteristics.. ... 5-13

5.9.2 Expander Characteristics . . 5-13

4.6.4 Examples on Sampling Theorem for Lowv
Pass Signals... .4-13
5.9.3 Compander Characteristics ***'*********
5-13

4.7 5.9.4 Compressor Characteristics ..5-15

Sampling Techniques 4-15

4.7.1 Ideal or Instantaneous or Impulse 5.10 Virtues, Limitations and Modifications of PCM...5-16
Sampling 4-15 5.10.1 Applications of PCM . -165

4.7.2 Practical Aspects of Sampling and Signal 5.10.2 Virtues of PCM .**********
. ****'****
5-16

Recovery 4-15 5.10.3 Limitations of PCM. ....5-16

4.8 Natural Sampling or
Chopper Sampling -15 5.10.4 Modifications in PCM ***************'******
5-17

4.9 Flat Top Sampling or Rectangular Pulse

5.11 Linear Delta Modulation (D.M.). .... 5-211
Sampling 4-18

4.9.1
5.11.1 Delta Modulator Transmitter . . 5-22
Aperture effect.. .4-20
5.11.2 D.M. Receiver . . .5-23
4.9.2 Comparison of Sampling Techniques...4-21
5.11.3 Features of D.M. . *************
.5-23
****
4.10 Applications of Sampling Theorem 4-22
5.11.4 Applications of D.M. ****** ***********. ... 5-23
Review Questions.. 4-22
5.11.5 Distortions in the DM System. ...5-23
Chapter 5: Analog to Digital Conversion 5-1 to 5-36
5.11.6 Advantages of Delta Modulation.5-24
Syllabus : Quantization, PCM, Companding, DPCM,
5.11.7 Disadvantages of Delta Modulation .5-25
ADPCM, Delta modulation, Adaptive delta modulation, T1
5.11.8 Condition for Avoiding the Slope
carrier system.
Overload Error.. 5-25
5.1 Introduction. .5-2 5.11.9 Examples on D.M.. 5-25
5.1.1 Advantages of Digital Representation of a 5.12 Differential Pulse Code Modulation (DPCM).5-27
Signal. * *****.**"** ******** ******
...5-2
5.12.1 Role of a Predictor . ***************** '*****.5-27
5.1.2 Disadvantages 5-2 5.12.2 DPCM Transmitter ************°*******2**** 7

5.2 A to D Conversion...
****'*** ***************************.5-2 5-28
5.12.3 DPCM Receiver . . *********'***°''

5.2.1 5-28
Quantization Process ********...5-3
5.12.4 Advantage of DPCM *** *****' *'****'*

5.3 Pulse Code Modulation 5-29

(PCM) * ***************
5-4 5.12.5 Disadvantages **************** ****

5.3.1 Pulse Code Modulation (PCM) System.5-4

Jenil Thakkar TechKnowledge
PuDIIC a t i o n s
 Chapter

4
Sampling
Syllabus
Sampling theorem, Sampling and signal reconstruction, Aliasing, Types of sampling.

Chapter ContentsS
4.1 Introduction
4.2 Formatting Analog Infomation

4.3 Sampling Theorem

Filter
4.4 Recovery using ldeal Low Pass
4.5 Interpolation
4.6 Aliasing or Foldover Error

4.7 Sampling Techniques

4.8 Natural Sampling or Chopper Sampling
4.9 Flat Top Sampling or Rectangular Pulse Sampling
4.10 Applications of Sampling Theorem

Jenil Thakkar
 ADCS (Sem. 5/ECE/ GTU) 4-2

4.1 Introduction Digifal

signal
ampling
Text
Pused
Ouantizer Encoder wavolorm
Formatting is the first
step followed in signal signals
ampler
nodulalor ransmitte
FORMAT Binary
processing. It is carried out in order to ensure that the digts

message or source signal is FORMAT(REVERSE) Chamel

compatible with digital
nalog LPF Encoder Do Receiver
processing. signals TLmodudator Receiver
Text
Binary Pulsed
Transmit Formatting is defined the
digits wavelorrm
as
process of Digital
signa
transforming the source information into the digital
symbols. (E-588) Fig. 4,1.1: Formatting and baseband pulsea
modulation
At the receiver
exactly the opposite transformation will
take place in order to recover the Textual information is converted into binary
original information form b
back. means of a coder block.

If data compression is done in addition to We have to use three different blocks

formatting, Sampling
then the process is called quantizer and encoder in order to practically carry out
as source
coding.
Sometimes formatting is treated as a special case of the task of formatting.

source coding. The binary digits present at the output of the pulse
The topics related to the term modulator are transmitted over the Baseband channel
formatting are as follows:
such as a pair of conducting wires or co-axial cable
Character coding
In order to transmit the binary digits on any channel, it
2. Sampling
is necessary to convert the binary digits to waveforms
3. Quantization in the pulsed form) that are compatible with the
4. Pulse Code Modulation (PCM) channel.
The next step after formatting is pulse modulation. This job is done by the pulse modulator in Fig. 411
It converts the formatted digital signal into the pulses Hence the pulse modulator output is in the form of à
that can be transmitted over the cable. pulse train.
These are the baseband pulses. The types of baseband After travelling
the baseband channel, these pulsed
over

pulses are as follows waveforms are received by the demodulator at nE

1. Various line codes receiver

M-ary pulse modulation PAM, PwM, PPM etc. The demodulator recovers the original binary digits
2.
from the pulsed waveform and applies them for tne
This process is called as the baseband signalling.
reverse formatting.
These waveforms are then transmitted over the cables.
he reverse y
formatting is a process which is exa
4.1.1 Baseband Systems: opposite to that of formatting
t is used for
Block diagram recovering the estimate (a signal wsignal
close to the
Fig. 4.1.1 shows the simplified block diagram of a digital original signal) of the original analog s
+.1.2
communication system, which emphasizes on the Formatting Textual Data (Charact
formatting and baseband transmission part. Coding):
be
The digital data does not need formatting. So it wil The original information is to
Signal
that
bypass the formatting section and will be directlv forms
communicated can be in the following
possible

to the pulse modulation section.

applied Analog signals
Jenil Thakkar
"DIications
TechKnoul
ADCS (Sem. 5/ECE/GTU) 4-3 Sampling
applications, such as in
This property is used in certain
Textual data
2. consist of a sequence of
data (computer to computer moving pictures, which
3. Digital an
individual frames, each of which represents
communication).

sample in time) of
a
characters is instantaneous view (i.e., a
consisting of alphanumeric
The data continuously changing scene.
known as the textual data.
viewed in sequence at a
When these samples are
encoded into standard formats using one
t is character fast rate, we perceive an accurate

codes such as ASCI code.

sufficiently
of the following representation of the original continuously moving

4.1.3 Messages,
Characters and Symbols scene.
between
contains a multiple of theorem also acts as a bridge
The textual message
The sampling
continuous-time signals and
discrete-time signals.
in the sequential form.
alphanumeric characters
easier to process discrete-
them digitally, these characters are In many applications, it is
In order to transmit is often
more flexible and
of bits called as a bit time signals because it is
first encoded into a sequence continuous-time signals.
stream or baseband signal. preferable to processing
signals obtained due to
be combined together to form a This is because the discrete
A group of "n" bits can the digital
we can have M 2" =
sampling can be easily processed by
symbol. With "n" bits per symbol, lightweight,
technology systems that inexpensive, are

symbols programmable, and easily reproducible

known as M-
A system that uses these M symbols is an

to discrete-
Due to sampling, it becomes possible use

ary system. continuous-time

time system technology to implement
Ifn 1, then M = 2" = 2 and the system is called as a
continuous-time signals.
systems and process
binary system. to convert a continuous-time
Thus, we use sampling
2, 2 4 and the system is known as a 4-
Forn M =

signal to a discrete-time signal, process the discrete-

ary or quaternary system. discrete-time system and then
time signal using a

convert back to continuous time.

4.2 Formatting Analog Information:
In this chapter, we introduce the concept of sampling
form then the
f the information is in the analog and the process of reconstructing a continuous-time

cannot be done.
Instead we have to
character coding signal from its samples.
Convert the analog information to digital signa. under which a
We will explain the conditions

In order to do so the first step

is to convert the continuous-time signal can be exactly reconstructed
a discrete
time
continuous time analog signal into from its samples and the consequences when these
called sampling.
analog signal with the help of process conditions are not satisfied.

4.2.1 Sampling: Following this, we explore the processing of

continuous-time signals that have been converted to

continuous-time signal
fcertain conditions are met, a
discrete-time signals through sampling.
recoverable from
Canbe completely represented by and
at points
knowledge of its values, or samples, obtained 4.2.2 Low Pass and Band Pass Signals:
equally spaced in time.
a basic
1. A Low Pass Signal
that follows from
nis is a surprising property
defined as the signal which has a
resut that is referred to as the sampling theorem. A low pass signal is
frequency spectrum extending right
from 0 Hz to W Hz,
his theorem is extremely important and useful.

TechPubICatfons
Knowledgë
Jenil Thakkar
 ADCS (Sem. 5/ECE/ GTU) 4-4
if a

W to
single sided spectrum is plotted and it extends from Continuous time
analog signal Sampler Discrete time Sampling
-

+ w, if
double a
sided spectrum is x(t)
analog signal
shown in Fig. 4.2.1(a). plotted, as
Samplingg
A video signal
signal which has frequency spectrum
a
from s(t)
O to 5 MHz is an
example of low pass signal.
X( (L-156) Fig. 4.2.2: Sampling process
At the input of the sampler we
apply an analog signal
and at its output we get the sampled version of the
input signal as shown.

-W This sampled signal represents the original

analog input
signal faithfully if the sampling rate is
(a) Spectrum of a low pass signal
adequately large.
Need of sampling:
X()
In the pulse modulation and digital modulation
systems,
the signal to be transmitted, needs to be in the discrete

ZA
time form.

If the message
signal is coming from a digital source
-2w 2W- (e.g. a
digital computer) then it is in the proper for a

(b) Spectrum of a band pass

digital communication system to be processed.
signal
(D-406) Fig. 4.2.1 However, this is not always the case. The message signal
2. A Band Pass can be analog in nature
Signal (e.g. speech or video signal)D
In such a case it has to be first converted into a discrete
Definition:
time signal. We use the
Thus "sampling process" to do this.
bandpass signals are the signals having their Thus using the sampling process we convert d
spectrum extending from f to f2 with f2 > f and both
continuous time signal into a discrete time
f and f, being of nonzero value. signal.
For
Inputsignal
example, the voice signal which has
spectrum a
For the
extending from 20 Hz to 3.4 kHz is a bandpass signal. sampling process to be of practical utility t

necessary to choose the sampling rate

The nature of the spectrum of a bandpass signal is properly.
The input signal x(t) should
completely different. be a band limited signal.
That means its
It is as shown in Fig. 4.2.1(b). The bandwidth is 2W Hz spectrum should exist only betwee 0
and some finite
but it is centered about a frequency f rather than frequency say fm Or W.
zero. We assume that f > W, therefore (f-W)> 0. Sampling signal
A
Sampling signal s(t) is a train of unit impulses, sppaced
4.2.3 Sampling Process: by a
period of T, seconds.
Definition: This sampling function at 3
samples the input signa n a l .

rate of"f," samples per second.

The sampling process is defined as the process of
time analog signal to a discrete Therefore "T" represents the sampling perioasuch tthat
converting a continuous

and the sampled signal is the discrete

analog signal T, Sampling period (4.2.1)
of the original analog signal.
time representation
and f,
Fig. 4.2.2 summarizes the
sampling process.
=
+ Sampling rate.

Jenil Thakkar Tech Kneulodg

UbIItations
 Sampling
ADCS (Sem. 5/ECE/GTU)
4-5

Q.6 State and prove sampling theorem

in timedomain.
Output signal: Explain aliasing and Nyquist rate
The output signal
is the sampled version of the input
(Dec. 17, Dec. 18, 7 Marks, May 18, 4 Marks)}
It is a discrete
time signal which is represented
signal. Introduction:
by xs ().
order torepresent the original message
signal
In
is obtained by multiplying the input
Output signal information), it is necessary
signal in time dimain. "faithfully" (without loss of
signal with the sampling to take as samples of the original signal
many
as
x (t)x s ()
possible
Reconstruction or recovery of samples, closer is the
the number
Higher
Reconstruction or recovery is the process of recovering
representation.
its sampled version x, (t).
the original signal x (t) from The number of samples depends
on the "sampling
rate"

For this the sampled signal x, (t) is processed through and the maximum frequency
of the signal to be
and cutoff frequency
an ideal low pass filter with gain T sampled.
greater than W. introduced in 1949 by Shannon.
Sampling theorem was

resulting output signal will be exactly equal

x(T). called as "Shannon's
The
Therefore this theorem is also

Requirements of sampling process sampling theorem".

the following sampling theorem in time domain, for
The sampling process should satisfy The statement of
follows
requirements the band limited signals of finite energy is as
should represent the original signal
1. Sampled signal Statement:
faithfully L. If a finite energy signal x (t) contains no
reconstruct the original than "W" Hz (i.e. it is a band
2. We should be able to frequencies higher
version. limited signal) then it is completely determined by
from its
signal sampled
specifying its values at the instants of time which
4.3 Sampling Theorem: are spaced (1/2W) seconds apart.

14
GTU : May 13, Dec.13, May 14, Dec. 2. If a finite energy signal x (t) contains no frequency
18, May 19
May 16, Dec.16, Dec. 17,May 18, Dec. components higher than "W" Hz then it may be
recovered from its samples which are
University Questions completely
spaced (1/2W) seconds apart.
Q1 Explain briefiy the Nyquist sampling theorem.
5
(May 13, Dec. 13, Marks) Combined statement of sampling theorem A
for low pass continuous time signal x ( can be completely
Q. 2 State and prove sampling theorem
signalsin time domain. (May 14,
May 19 7 Marks) represented in its sampled form and recovered back from
Q.3 Explain briefly Nyquise's sampling theorem. Whatis the sampled form if the sampling frequency f,2 2W where
maximum frequency of the continuous time
interpolation formula.
W is the
interpolation process? Derive
(Dec. 14, 7 Marks)| signalx().
theorem. What is 4.3.1 Proof of Sampling Theorem
4Explain briefy Nyquise's sampling
? Derive interpolation
interpolation process
(May 16,7 Marks) GTU Dec. 14, May 16, Dec. 16, Dec. 17,
formula. May 18, Dec. 18, May 19
whose spectrum is
State and prove that a signalreconstructed exactly University Questions
band-limited to B Hz can be 2B a.1 Explain briefly Nyquist's sampling theorem. What
is
rom its samples taken uniformly at rate R
>
a
interpolation process ? Derive interpolation
Hz. Also explain the practical difficulties in signal
(Dec. 16,7 Marks) formula. (Dec. 14,7 Marks)
reconstruction. Tech Knowledgë
Jenil Thakkar Pubiicatlons
 ADCS (Sem. 5/ECE/ GTU)
Q. 2 4-6
Explain briefly Nyquist's sampling theorem. Sampling
interpolation process ? Derive What is Step 1 Represent the sampling function
formula. interpolation mathematically.
(May 16, 7 Marks) Step 2 Represent
Q.3 State and the sampled
prove that a signal signal
band-limited to B Hz can be whose spectrum is mathematically.
from its samples taken reconstructed exactly Step 3: Obtain the Fourier transform of
Hz. Also uniformly at a
rate R 28 the sampled
explain the practical
difficulties in signal
signal
reconstruction. Step 4 Prove that the sampled signal
Q. 4 State and prove (Dec. 16, 7 Marks) x, () completelv
represents x (t).
sampling theorem in
Explain aliasing and Nyquist rate. time domain. Step 5: Represent x (t) as summation of sinc
(Dec. 17, Dec. 18, 7 Marks, May (interpolation).
functions
Let us now
18,4 Marks) Step 6 Graphical representation of the
prove the sampling theorem in
time domain. process. interpolation
The
assumptions made for this proof are as follows: Step 7 Actual recovery of x () using an ideal low
Assumptions filter.
pass
Let x (t) be Part 1: Obtain the
a
continuous time analog signal as shown in Spectrum of the sampled signal:
Fig. 4.3.1. Step 1:
Represent the sampling function s (t)
Continuous
üme mathematically
analog Sample values
SIgnal Fig. 4.3.1 shows the sampling function
(a) s () which is a
train of unit impulses of
period T, and frequency
The sample function
s(9 8(t-nT) s (t) can be represented
A unit mathematically as follows ang he
impulse
train used- b) s (t)= (t+2 T) +
(t
as sampling. vivmw .23T4T
+
T) +ô (t)
function
(t-T +8 (t-2 T)
****

+
**
ASTtuti
Sampled
signa
s (t)= 28t-nT) (4.3.2)
Step 2: Represent the sampled signal x, (t)
********** ************************ * wwwww.w ww **ww
ww

(D-408) Fig. 4.3.1: Sampling of a continuous time signal x (t) mathematically

Fig. 4.3.1 shows the
sampled signal x (t)
Let x (t) be a signal with finite energy and infinite
present only at the
graphically. It is
duration. sampling instants i.e. T, 2T, etc. ana
itsinstantaneous amplitude is
Let (t) be a strictly band limited equal to the amplitude or
x
signal with the original signal x (t) at the
maximum frequency of "W" Hz. sampling instants.
The encircled points in Fig. 4.3.1 show this.
Let s (t) be the sampling function as shown in Fig. 4.3.1.
Let us
represent the instantaneous of
It is a train of
equally spaced unit impulses, by a period the various
amplitude x (0
a
sampling points t
1ne =
nT, as x (n T).
of T, seconds. Hence the sampling rate or sampling theamplitude of the encircled points of Fig.
frequency is 4.3.1
Looking at the sampled signal x, (t) we can he
say tna
, Sampling rate. (4.3.1) sampled signal is obtained by multiplying x() and s ().
( t ) x (t) s (t) x (n
=
x .4.3.3)

Procedure to be followed T)x s () =

Substituting the expression for (t) from quation

We are going to follow the steps given below to prove (4.3.2)
s
Equo the
we
get the mathematical expression
thesampling theorem sampled signal x, (t) as,

Jenil Thakkar BTech Knowledg

UbITCatio"
 ADCS (Sem. 5/ECE/GTU) 4-7 Sampling

Xs (t) =
2 x(T.).8(t-nT) 4.3.4) x,( x (0.|28 (f-nf .(4.3.8)
n

the Fourier transform of the sampled where denotes convolution.

Step 3: Obtain
signal Interchanging the orders of convolution and summation

train of impulses (dirac delta results in

The Fourier transform of a

function) is given by,

2 x08 (f-nf) 4.3.9)
x, (9 =
f,
n -

x(0 fo2 -nf)

that the
n-co From the properties of delta function, we find
Therefore the Fourier transform of the sampling convolution of X () and &(f- nf) is equal to X (f- nf.).

function is given by, Hence the above equation can be simplified as follows

F.T. of the sampled signal, X, () is given by,

s = f, 2 8-nf) .(4.3.5)
n-oo

2 xf-nf) 4.3.10)
in the time domainis represented X, =
f
The sampled signal
as product of x (t) and s (t).
where X () = Fourier transform of the original signal
ie. x (t) = x (t) xs (t) 4.3.6)
x (t).
sides
Taking the Fourier transform of both the get,
we
plotted
-

The spectrum Xz () of the sampled signal is as

i.e. X, () X () * S ( .4.3.7)
shown in Fig. 4.3.2.

Substituting the value of S() from Equation (4.3.5)

we

get,

X(f
Spectrum of a bandlimited

signai x(t)

-W W
(a) Spectrum of the original signal x(t)

Second term X() www Spectrum of the

-W W sampled signal
sX (f+s Fourth tem atf 2WV
Third term
X () X(f-)

-3W -W +W= 3W 2
l.e. Sampling
b) Spectrum of the sampled
signal xg() with f, 2W, =

done exactly at Nyqulst rate

ls

(D-410) Fig. 4.3.2: Spectrum ofsampled signal

Thus depending on the value of "n" (which extends

Conclusion from Equation (4.3.10)
the
from o to + o) we will get infinite number of original
The term X (f -

nf) in Equation (4.3.10) represents spectrums X () centered at frequencies 0, f, t 2f,

shifted version of the spectrum X () of the original
t3f, + 41, .etc. In other words,
signal x (t).
TechPubilcations
Knowledge
Jenil Thakkar
 ADCS (Sem. 5/ECE/ GTU) 4-8
X (f
nf= X () at f 0, +f, t2f, t 3f, 4.3.11) We can obtain another useful expression for
the
Samping
This concept will be clear
if we open Equation transform Xs () by taking the Fourier Fourier
(4.3.10) transform
sform of both
and write the terms the sides of the equation stated above
separately as shown below. as,
Now open the summation
sign in Equation (4.3.10) to
get x,(0 2 x(nT)e 2 zn
fis
4.3.13)
Equation (4.3.10) can also be written as
This equation is the Fourier transform of a
discrete tim
time
signal x,(t.
X, (0 f, X( 2 f,X (f-nf) 4.3.12) Step 2: Obtain the FT of input signal
n0
Therefore it is called as the discrete Fourier
X, 00 t,X ((+2)+,X+)+X (0+,X - transform
(DFT).
X () shifted right by f, Compare it with the definition of Fourier transform of
Spectrum X() a
+X () shifted left by
f, continuous time signal. ie.
X () shifted left by 2,

(D-1705)
X (0 =
x (t) e2« dt
Comment
As the signal is discrete, the integration sign has
From been
Equation (4.3.12) we conclude that the process of
replaced by the summation sign and t" has been
uniform sampling of a signal in the time
domain results replaced by "nT,"
in a periodic in
spectrum the frequency domain with a Now consider
period equal to the sampling rate Equation (4.3.12),

Part 2: To prove that sampled signal x, (t) completely

represents x (t): ,X (f- nf)
n-o
n#0
Step1: Obtain the FT of sampled signal:
The sampled signal has been shown in Fig. 4.3.3.
X(x,
(0-2xt-nt
-x(3T Butin the range -Wsfswthe second term ot tne
above expression will
***... 4Ts) not be present

x-Tg ...... (4.3.14)

2T 31,
Substitute f, =
2W and X, () from Equation (4.3.1 to

x-T get,
8(t+T) TB-T
x(0)8()
(D-411) Fig. 4.3.3 : Sampled signal x, () X(nT)ej2xnfis ..4.3.15)

X (t) can be represented in the summation form as

This is the
follows (Refer Fig. 4.3.3). frequency spectrum of x(t) in terms of x (nly
ie. the
(0) 8 (
sampled signal.
x ()= (- T) 6t T,)
x + + x

Substitute T, 1/2W to
=

4.3.12(a) get,
+xT)8(t-T)+.
x ( =

x, ()=2 x(n T) 8 (t-nT)

2 2 X(n/2w). e2nf/2W

..-W sfs W .4.316)

Jenil Thakkar
T e c hK n o w e d

PUDIICalib"s
Dns
 ADCS (Sem. 5/ECE / GTU) 4-9 Sampling
Spectrum of x() Spectrum of

Note This equation shows that the spectrum of

is x (t) $ampled signal

same as the spectrum of x () in the frequency

range VWto +W.
Frequency
Hence the sampled signal represents the original signal
Froquoncy rosponso
of reconstruction filiter
x (t) successfully.

Thus if the sample values x (n/2W of the signal x (t) are -W +W

Frequency
specified for all time, then the Fourier transform X ( of x(t)
the original signal is uniquely determined by using the
Spoctrum of flter output
Equation (4.3.16).

Because x () is related to X () by the inverse Fourier +W Frequency

(D-414) Fig. 4.4.1(b) :Operation of reconstruction
transform, it follows that the signalx (t) is itself uniquely
When the sampled signal x, () is applied at the input,
determined by the sample values x (n/2W) for
this filter will allow only the shaded portion in the
- nso will
spectrum of x, (t) to pass through to the output and
In other words, the sequence of samples (x (n / 2W)}
block all other frequency components.
contains all the information of x (t). Thus the frequency components only corresponding to
Thus, we have proved second part of the sampling x (t)will be passed through to the output and the

theorem. original signal x (t) is recovered.

However, ideal filters are generally not used in practice

4.4 Recovery using ldeal Low Pass
for a variety of reasons.
Filter In any practical application, the ideal LPF in Fig. 44.1(a)
would be replaced by a non ideal filter Ho) that
Reconstruction or recovery is the process of recovering
approximates the desired frequency characteristic with
the original signal x () from its sampled version xg ().
sufficient accuracy. (ie, Ho) 1 for lo < W, and
For this the sampled signal x, () is processed through Hjo) 0 elsewhere).

an ideal low pass flter with gain T and cutoff frequency Due to such in the low pass filtering
approximation
greater than W. stage, some discrepancy is expected to exist, between

The recovery process is as shown in Fig. 4.4.1(a). the original signal and the recovered signal.
We need to select the non-ideal filter based on the
The ideal low pass filter is called as reconstruction
acceptable level of distortion for the application under
filter.
consideration.
Sampled signal Reconstruction Original signal 4.5 Interpolation
filter x() GTU Dec: 14
IdealLP.F University Questions
a.1 Explain briefly Nyquist's sampling theorem. What is
(D-413) Fig. 4.4.1a) : Recovery using an ideal LPF interpolation process? Derive interpolation formula.
Dec.14, 7 Marks)
The resulting output signal will be exactly equal x(T). Definition:

he filter used here is an ideal low pass filter with a

Interpolation, that is, the fitting of a continuous signal
to a set of sample values, is
Drickwall frequency spectrum as shown in Fig. 4.41(6).
a
commonly used
procedure for reconstructing a function, either
approximately or exactly, from samples.

Tech Knewled
Jenil Thakkar Pubiltations
ADCS (Sem. 5/ECE/ GTU) 4-10
4.5.1 Interpolation Formula:
University Questions
GTU : Dec. 14 Equation(4.5.3) provides an
interpolation fo
Sampiling
Q.1 Explain briefly Nyquist's sampling theorem. VWhat is reconstructing the
original signal x () fr. formula for
sequence of sample values x (n from
interpolation process? Derive interpolation formula. /2W). ne

(Dec. 14, 7 Marks)

The "sinc" function plays the role of an
From function. interpola
olation
Equation (4.3.16) we can obtain x (t) by taking the
inverse Fourier transform Each sample (n/ 2W) is
(IFT).
x
multiplied by a delayed
x (t) IFT {X (A} = of the interpolation function i.e. sinc version
function.
Then all these resulting waveforms are added
=
IFT
IF|2Wn x(n/2W). ein fnw
x (t).
to obtain
-o
Using the definition of
inverse Fourier transform, Graphical representation of the interpolation
W
process:
Let us re-arrange Equation (4.5.3) as follows:
x(t) =
x (n / 2W. erfn/W2 f df
-W n= -o x (t) = 2
Interchanging the order of
x (nT) sinc 2w (t- n
summation and integration
we get,
This is because
2 Ts
x (t) = 2 x 1
(n/ 2W) 2w
n x (t) = 2 x (nT) sinc 2W (t-
n-co
nT) 4.5.4)
x (t) =
x (n/ 2W)-2 Let us
expand this equation to write,
n

j2 nf(t-n/2W) W
j2t-2 x (t) = x

+x
(0) sinc 2 Wt
(t 2T)
+

sinc 2W (t
x (#T) sinc 2W
(t+T
+2T)+ 4.5.5)
Refer Fig. 8.5.4 for the
graphical representation of the
x (t) =
x (n / 2W
1 interpolation process.
1.
First term:x (0) sinc 2Wt
j4Wt-2W
CO

This will have

j2rW (t- n/2W) -j2 « W (t-n/2W)
a maximum amplitude at t 0. Ihe
maximum amplitude is equal to the sample value x0
att 0 .
j2nW (t-n/2W) -j 2 n W (t-n/2 W)
This
2xn/2w sinc function will pass through zeros a
n = -o j4 TW2 w t 1/2 W, t 1/4 W ..etc.
This is as shown in
The term inside the square bracket is a "sinc" function. Fig. 8.2.4.
2. Second term: x
(t T,) sinc 2W (t +T,):
sin (27Wt- nrc) This sinc
x (t) =
2 x (n/ 2W)
(2Wt -n) ..(4.5.1) function will have maximum amplituade at
ttT. The maximum amplitude is
OO

value x ( T) at t +
equal to the sample
We can simplify the equation above by using the
Thus sinc 2W
T, respectively.
definition of the "sinc function". The sinc function is (t + T) represents shifted sinc function
E
"sinc 2Wt by a period +T, This is as shown in Fig. * 5.1.
defined as :
3. Remaining terms:
sinc x =
sin (Tx) 4.5.2)
TTX Similarly the third term, x (t 2T) sinc 2W (t 9
Therefore Equation (4.5.1) can be written as represents shifted sinc function "sinc 2We" by a
eriod
pe
of +2T, and so on.
x (n / 2W) sinc (2 Wt - n)
x (t)=2 4.5.3) We can plot all these sinc functions along the
wi
sampled signal x () as shown in Fig.
4.5.1
Jenil Thakkar
TechKneuledg
PUDIIcationS
ADCS (Sem. 5/ECE /GTU) 4-11 Sampling
wwwww ****w***ww*

X)
Sampled signal

-2Ts * **
s
**

Reconstructed signal x() [Output of filter)

Sample values x (nT) *;

Sinc functions

-2Ts 0 21s
***w**** wsASASASM *w**A**www.w**wiw*t

(D-412) Fig.4.5.1:Reconstruction of the original signal x (t) from its samples using the interpolation
The signal x () is not strictly band limited. The spectrum
Note that the peak amplitude of any sinc function is
equal to the corresponding sample value x (nTs). of signal x (t) is shown in Fig. 4.6.1(6).

Actual reconstruction with a low pass filter X()

****

As shown in Fig. 4.5.1, the peaks of the sinc pulses

********* *

represent the amplitudes of the samples.

The signal x (t) expressed in Equation (4.5.3) ie. the fy2High frequencies in x()
(a) Spectrum of a continuous time signal x()
is then passed through an ideal low pass
sampled signal
filter to recover the original signal x (t).
Overlapping
SpectrumTS

filter is therefore called as the

This low pass
reconstruction filter.

4.6 Aliasing or Foldover Error: High frequencies in x(1) take on the identity
of lower frequency due to aliasing

(t) with f.
GTU: May 11, Dec. 12, New Syll. May 17, Dec.17, Dec.18
: (b) Spectrum of the sampled version of x « 2W

(D-415) Fig. 4.6.1

University Questions
Q.1 What is the effect of under-sampling? The spectrum X, () of the discrete time signal x () is
(May 11, 3Marks) shown in Fig. 4.6.1(b) which is nothing but the sum of
2 State, prove and explain the sampling theorem x () and infinite number of frequency shifted replicas of
What is aliasing effect ? (Dec. 12,7Marks)
(May 2017,1 Mark it as explained earlier.
a3 What is aliasing?
domain. Consider the two replicas of X () which are centered
State and prove sampling theorem intime
Explain aliasing and Nyquist rate. about the frequencies f, and -f

(Dec.(Dec.2017,7 Marks Ifwe usea reconstruction filter with its pass-band

and / if
signal x (t) is not strictly band limited
or extending from-f,/2 to + f,/2 then its output will not
the
then an error
the
sampling frequency f, is less than 2W, be an undistorted version of the original signal x (t).
called aliasing or foldover error is observed. Some distortion will be present in the filter output.
n the aliasing or foldover error, the adjacent spectrums
This distortion occurs due to the overlapping of the
if , <
n the spectrum of the sampled signal overlap adjacent spectrums as shown in Fig. 4.6.1(b).
2W. This is shown in Fig. 4.6.1(b).
Tech Kaeledge
Pubiic ations

Jenil Thakkar
ADCs (Sem. 5/ ECE/GTU 4-12
Due to this
overlapping, it is seen that the portions of A guard band is created between thee
Sampling
the frequency shifted replicas are "folded over" inside spectrums as shown in Fig. 4.6.2(b). adjacent
the desired spectrum. X(0
Spectrum of original
slgnal x(t)
Due to this "fold
over", high frequencies in X () are
reflected into low
frequencies in
-W :W
X, (f). Guard band Xg) Guard band
This can be understood
by comparing the shaded
portions of the spectra shown in Figs. 4.6.1(a) and (b).
Definition of Aliasing
--w -f+V -W
The (D-417) Fig 4.6.2(6) : Spectrum of a sampled
phenomenon of a high frequency in the spectrum
signal for f,> 2W
of the original signal x (t),
taking on the identity of lower
frequency in the spectrum of the sampled signal x, () is 4.6.2 Nyquist Rate and Nyquist Interval :
called as aliasing or fold over error. New Syll.: GTU : May 17, Dec.17, Dec. 18
Effect of aliasing: University Questions
Due to aliasing some of the information contained in a. 1 State and prove sampling theorem in time domain.
the original signal x (t) is lost in the
Explain aliasing and Nyquist rate.
process of sampling.
(Dec. 2017, 7 Marks)
4.6.1 How to Eliminate Aliasing ?
Nyquist rate
Aliasing can be completely eliminated by: The minimum sampling rate of "2W"
samples per
Using an antialiasing or prealiasing filter and second for a signal x () having maximum frequency of
"w" Hz is called as "Nyquist rate".
2 Using the sampling frequency f,> 2W.
Nyquist rate = 2W Hz
1. Using an anti aliasing filter

In order to avoid alising, use a band limiting low pass Nyquist interval
filter and The reciprocal of Nyquist rate i.e. 1/2W is called as the
pass the signal x () through it before sampling
as shown in Fig. 4.6.2(a). Nyquist interval.
Nyquist interval = 1/2W seconds
Bandlimiting
low pass filter Sampler
4.6.3 Effect of Non Ideal Reconstruction
Strictly Filter
bandlimited x()
ldeal reconstruction filter
(D-416) Fig. 4.6.2(a) : Use of a bandlimiting filter
to eliminate aliasing We have to pass the sampled signal througn a

This filter has a cutoff frequency at fe = W, therefore it reconstruction filter in order to obtain the original signat
will strictly band limit the signal x () before sampling back from the sampled version.
As mentioned earlier t h e reconstruction filter is a o
takes place.
in
This filter is also called as antialiasing filter or prealias pass filter. It is
expected to pass all the frequencie
the range of (-W to + W) Hz.
filter.
i to
2. Using the sampling frequency f,> 2W This is because the original band limite
signal x ( is

In order to avoid alising, increase the sampling "w" Hz.

reconstruction

frequencyf, to a great extent i.e. f, >>> 2W. Therefore the frequency response of a recons

Due to this, even though x (t) is not strictly band limited,

filter should be as shown in Fig. 4.6.3.
s filter

This is the ideal low

the spectrums will not overlap. frequency response of an pa
TechKnouled
PUDIICatipns

Jenil Thakkar
ADCS (Sem. 5/ECE/GTU) 4-13 Sampling
Amplitude From Equation (1) it is clear that the maximum

frequency component present in the signal x (t) is of

2500 Hz.

-W W
Frequency In other words x (t) is band limited to 2.5 kHz

response of an ideal low pass (W 2.5 kHz).

(D-418) Fig. 4.6.3 Frequency
:

filter used as a reconstruction filter . Nyquist rate = 2W = 2x 2.5 kHz = 5 kHz Ans.

Practical reconstruction filter:

and Nyquistinterval =2 5 1 0 0 . 2 msec ..Ans.
However, it is not possible to practically realize an ideal

low pass filter. (b) x (t)= sin 200 t

Therefore a practical low pass filter with a frequency

In order to calculate the Nyquist rate we need to
as shown in Fig. 4.6.4 is used.
response calculate the maximum frequency component present
Spectrum of continuous signal x(0
in its spectrum.
Frequency its
W The spectrum of x (t) can be obtained by taking
X,(0 Spectrum of sampled signal
fourier transform.

Multiply numerator and denominator of x () by 200 to

-W)- W) -W W W+Guard bandFrequency
Amplitude response of
get,
reconstruction filter
x (t) = 200 sin (200r t)
(-f+W) -W W -W 0 Frequency (2007t t)
D-419) Fig. 4.6.4: Amplitude response of a practical sin tt
But Sinct
reconstruction filter Ttt

It is possible to use the practical low pass filter without sin (200 t sinc (200 t)
(2007t t)
introducing any distortion due to the presence of the
x (t) = 200 sinc (200 t)
guard bands between the adjacent frequency spectrums
as shown in Fig. 4.6.3.
We know that A sinc 2 Wt 2w
That is why it is necessary to have f,> 2W.

200 sinc (200t) 200

4.6.4 Examples on Sampling Theorem for
Low Pass Signals
Find the Nyquist rate and Nyquist interval for
200 sinc (200 t) rect 200
Ex. 4.6.1
each of the following signals: X ( = rect [f/200]

5 cos 1000 t cos 4000 t .

(a) x (t) =
The spectrum X () have been shown in Fig. P. 4.6.1.

(b) x () = Sin 200t Variable is X(

frequenoy
Soln.
x() =1 rect 200
(a) x (t)= 5 cos 1000 rt cos 4000 rt: 100 100 Frequency
Width ls
given signal x (t) is in the form of product of
cosine Amplitude
The 200 Hz Width 200 Hz
term. So let us use the following standard expression (D-420) Fig. P. 4.6.1: Spectrum of the given signal
2 cos A cos B cos (A+B) + cos (A-B) From Fig. P. 4.6.1 the maximum frequency in the
frequency spectrum is 100 Hz.
5 cos 1000 t cos 4000 t
Nyquist rate = 2x 100 = 200 Hz Ans.
2.5 cos 5000 t +2.5 cos 3000 tt
And Nyquist interval = 1/200 = 5 mS ..Ans.
x(t) = 2.5 cos 5000 t + 2.5 cos 3000 t 1)
Tech Knowledga

Jenil Thakkar
Pubileatlons
 ADCS (Sem. 5/ECE/ GTU) 4-14

Ex. 4.6.2 The spectrum of continuous time analog

a To draw the spectrum of sampled signal X, ( :
Sampling
signal x (t) is as shown in Fig. P. 4.6.2(a). This
We know that the ideal sampling results in the
signal is sampled at a frequency f which is 3/2 spectr
ctrum
X which is expressed mathematically as,
times the maximum
frequency fM in the
spectrum of x (t). Draw the spectrum of the
sampled signal clearly showing the effect of
X,6 f 2 x(f-nf)
under sampling. If the sampled
signal is then
passed through an ideal low pass filter having
a cutoff frequency fe =
fM then draw the Xg( =
fX (f) +
n=-oo
f,X (f nf)
spectrum of signal recovered at the filter n0
output. Substitute f, = 15 to get, X, ()
X(D

= 15X () + 15 x (f- 15n)

n =-o (3)
n 0
10 fHz
10
Fig.P.4.6.2(b) shows the spectrum X (f).
(D-421) Fig. P. 4.6.2(a) : Spectrum of the
Observe the overlapping of spectrums as i, <2 f
continuous analog signal x (t)
Reconstruction of the signal
Soln.
An ideal low pass filter with cutoff
From Fig. P. 4.6.2(a) it is clear that the signal x (t) is band
a frequency
limited to 10 Hz i.e. f fM 10 Hz is being used for reconstruction.
fM 10 Hz The response of this filter is shown in P. 4.6.2(c)
.(1) as Fig.
Hence the sampling frequency f, is given by. It is going to allow only that portion of Xz () to pass

through, which lies in the frequency range of 10 Hz to

*f 15Hz. .2)
10 Hz as shown in Fig. P. 4.6.2(d).

15

(b)
30 15 10 5 10 15 30
Spectrums
overlap and aliasing H(D
takes place
Response of an ideal
low pass filter

(c)

-10 10
10Hz

Spectrum of the
reconstructed signal (d)

10 -5 5 10
(b) Spectrum of sampled signal (c) Response of an ideal low pass filter (d) Output of the filter pass fiter
(D-422) Fig. P. 4.6.2

PubiIcatlons

Jenil Thakkar
Tech
 GTU) 4-15 Sampling
5/ECE/
ADCS (Sem.
Disadvantages of ideal sampling:
4.7 Sampling Techniques:
is that due to
1. The disadvantage of ideal sampling
be divided into two categories is very
Sampling techniques
can
the transmitted power
very narrow samples,
low. Thus the ideally
namely: small and the S/N ratio is

1 ldeal sampling 2. Practical sampling get lost in the background

sampled pulses may

noise.
4.7.1 ldeal or Instantaneous or Impulse to achieve
2. Ideal sampling is not possible
Sampling to
is practically impossible
practically, because it
described in the previous zero. Therefore
The sampling technique have pulses of widths approaching
the unit impulse train as a sampling used.
section, which uses flat top sampling is
practically natural or

ideal or instantaneous or impulse the

function is called as Tdeal sampling was used only to prove

sampling. sampling theorem.

Features ofidealsampling and

4.7.2 PracticalAspects of Sampling
function s (t) is a train of impulses. The
1 The sampling Signal Recovery:
short (can be
duration of each impulse is extremely
are different from the
approximated to zero). The practical sampling techniques

2. The sampling function can be expressed mathematically ideal sampling in the following ways

the duration of the

as 1 In practical sampling methods,
of the
sampling pulses is finite and the amplitude

s ( 2 8 (t nTs)
pulses is also finite.
n-o the practical low
2. Practical sampling methods use

sampled signal x (t) is expressed as, reconstruction. Guard band between

3 The
pass filters for
the sampled signal is
the adjacent spectrums in
22 x(nTJ8(t-nT) distortions. Ideal filters are not
Xs() necessary to avoid

used.
ideally sampled signal given by,
is
4. Spectrum of the The signalx () which is to be sampled is not strictly
.
bandlimited. Due to this there are problems faced
x(f-1) .4.7.1)
x,00 2 while deciding the sampling rate f
n
consists of There are two popularly used practical sampling
. The spectrum of the ideally sampled signal
the original signal.
i.e. X () and its techniques. They are
the spectrum of
centered at the frequencies 1. Natural sampling or chopper sampling and
infinite number of replicas
tf,t2f, 3f,..etc. 2 Flat top sampling.
is shown
ideally sampled signal
as
6. The waveform of an

in Fig. 4.7.1 4.8 Natural Sampling or Chopper

Sampling:
As explained earlier, the ideal sampling can not be

implemented practically.
A more reasonable and practically feasible manner of
2T-T. T, 2T

sampling is called as "Natural Sampling", as shown in

x ()
(D-428) Fig. 4.7.1 An idealy sampled signal Fig. 4.8.1.

Jenil Thakkar TechPubIC

Knowledgë
ations
ADCs (Sem. 5 /ECE/ GTU 4-16
MOSFET acting
Sampling
Baseband as a switch
signal

Natural
RLPAM signal
Sampling
signal
(b) Natural PAM with a MOSFET
s(t) = acting as a switch
ET
c(0x()
Naturally
Sampledd
Switch Switch OFF
signal ON
*
*-.

Naturally sampled
signal
(D-429) Fig. 4.8.1 Process of natural sampling
Looking at the 3T
waveforms in Fig. 4.8.1 we note the
(e) Natural PAM signal
following important points: (D-430) Fig. 4.8.2
Here the sampling waveformc (t) consists of a train of Reconstruction:
pulses each
having a duration " and separated by the The reconstruction technique for natural
sampling time T samplingis
similar to that for the instantaneous
The baseband sampling.
or
modulating signal x () and the
A
sampled signal s (t) signal sampled at the Nyquist rate may be
= c ()
(t) are as shown in Fig. 4.8.1.
x
reconstructed exactly by passing the
The sampled signal is obtained by
multiplication of x (t) sampled signal
and c (t). through ideal low pass filter with cutoff at
an
frequency
W. where W is the highest
The sampled signal is a train of pulses of width t, whose
frequency component of the
signal x ().
amplitudes are varying. These pulses do not have flat
Low pass
tops but their tops follow the waveform of the signal x filter characteristics
Naturally Tdeal Original
(t). s i a n l o w passLanalog
signal signal
The sampling rate is greater than or equal to the filter x(t)
Nyquist rate.
Circuit arrangement for natural sampling: Pass band
(D-431) Fig. 4.8.3
Reconstruction of original signal from
Natural sampling is sometimes called as chopper naturally sampled signal
sampling because the waveform of the sampled signal
Crosstalk in naturally sampled signals
appears to be chopped off from the continuous time
With the
signal x (t).
samples of finite duration, it is not possible to

completely eliminate the crosstalk generated in 3

The chopPper arrangement is as shown in Fig. 4.8.2 channel which is
where the chopper switch is being operated by the
bandlimited.
When N number of
sampling function "c (t)*. channels are to be time division
multiplexed, the maximum sample duration
Chopper switch TN is =

The width "t" should be

as large as possible to increa
the signal power but increase in "" will increase ue
Naturally
PAM signal possibility of crosstalk.
To reduce
crosstalk, "" should be
(a) Natural PAM Thus the width "" is
as
possiD
short as

Fig. 4.8.2 (Contd..)

a
compromise between two
contradicting requirements.
Jenil Thakkar TechKnewledga
PUbIitatlon
 5/ECE/ GTU) 4-17 Sampling9
ADCS (Sem.

Spectru
of naturally sampled signal: Therefore take the fourier transform of both sides of

The naturally sampled signal s (t) of Fig. 4.8.1 is Equation (4.8.5) to write,
of the signal x (t) and
abtained by multiplication
function c (t). sinc (f, t) FT (e . x (t)) .(4.8.6)
sampling S(
s (0 x (t) c (t) .(4.8.1)
(t) can be expressed in the form of Use the frequency shifting property of fourier transform,
However c
series as follows
complexfourier which statesthat

Ce2nt e2n1 nt.x () > X(f-fn) 4.8.7)

c ( 2 22 (4.8.2)

Therefore, S () = 2 sinc (,T)X(f- nf)(4.8.8)

Because c () is a train of rectangular pulses We can
n: - oo

obtain the value of c, as:

But as f nf, the spectrum of a naturally sampled
sinc If.T
o signal is expressed as follows,
where T = Pulse width = t in this case

f Harmonic frequency of f, i.e. fh = nf, sinc (nf, t) X (f- nf) ..(4.8.9)

n
or Tnf, and T =T, Conclusions from Equation (4.8.9):
TA 1. The term X (f nf) represents the shifted version
sinc If,
-

..4.8.3)
of the frequency spectrum X (). The spectrum S (0
Substitute this value of c, into Equation (4.8.2) to get,
consists of X ( and its shifted replicas as shown in

Fig. 4.8.4(b).
c(t) Tsincf,1ea ..(4.8.4)
These shifted replicas are observed at frequencies f
2.
n=-a
Substitute this expression for c (t) into Equation (4.8.1) f y +21, +3 f...etc.

to get, 3 The spectrum of x (t) is periodic in fs and weighted

by the sinc function. (See the term sinc (nf, ) in

s0 2 sinc f,t)e «i,nt.x() ..(4.8.5)
Equation (4.8.9). Therefore the amplitude of the
=

This expression represents the naturally sampled signal spectrum of naturally sampled signal reduces on

in time domain. To the spectrum, it is necessary

to both sides of Y axis as shown in Fig. 4.8.4.
plot
take the fourier transform of the expression for s (t).

X(

-Wcontinuous time signal x()

(a) Spectrum of
S()
sinc (níst)

A -W

b) Spectrum of aturally
W

sampled signal
(D-432) Fig.4.8.4
2

Jenil Thakkar Tech Knowledgë

PubIIcations
 ADCS (Sem. 5/ ECE/GTU) 4-18

Merits and demerits of natural

sampling: Operation of the sample and hold circuit: Sampling
1. Generation is easy. The sample and hold circuit consists of
two FET.
2. We can use practical low pass filter for
and a capacitor as shown in Fig. 4.9.1(a). switches
reconstruction. The analog signal x () is applied at the
input of e
3. The amplitudes of high frequency components circuit and the sampled signal s () is this
decrease therefore some distortion is introduced. the capacitor.
ained
obtained across
a

4. Increased SNR due to finite A gate pulse will be

pulse width of the applied to ate G at the
sampling function and that of the instant of
sampled signal. sampling for a very short time.
5. For large values of "" there is a possibility off The sampling switch will turn on and the
crosstalk. capactor
charges through it to the
sample value
4.9 x (nT).
Flat Top Sampling or
Rectangular
Pulse Sampling: The switch is then turned off. Both the
sampling
will remain OFF for a duration of ""
FETs
seconds and the
The natural
sampling is rarely employed in practice. capacitor will hold the
voltage across it constant for this
Instead the other practical
sampling technique called period. Thus the is
pulse stretched to "" seconds
flat top sampling is employed in
practice. At the end of thepulse interval (), a pulse is applied to
In the flat top sampling G ie. gate terminal of discharge FET.
technique, the analog signal x
() is sampled instantaneously at the This will turn
rate f, =and the on the discharge FET and short circuit the

duration of each capacitor. The output voltage then reduces to zero. This
sample is lengthened to a duration "
is as shown in Fig. 4.9.2.
as shown in Fig. 4.9.1(b).
x{)
Thus the amplitudes of these pulses are constant and Flat top sampled signal

equal to the corresponding sampled values.

Sampling
Switch
Sampling
switch ON
Discharge
switch ON
s(t) (D-434) Fig. 4.9.2: Operation of sample and hold circuit
Sampled
signal Principle of
G generating the flat top sampled pulses
From Fig. 4.9.2 it is clear that only the rising edge o
Discharge
swltch
each pulse represents the instantaneous value of tnE

analog signal x (t).

(a) Sample and hold circuit to obtain the flat topped samples Therefore the flat top sampled pulses can be obtaind
by convolution" of the instantaneous sample and o
x(nT) pulse h (t) of duration t.
s() This is true because with
convolution of any function
the delta function
results in the same function.
ie. x (t) *& (t) = x ) .(4.9.1)
ht) h(t)

(b) Flat top sampled signal

(D-433) Fig. 4.9.1

The flat top pulses can be obtained by using the sample

(D-435) Fig. 4.9.3: Principle of
and hold circuit shown in Fig. 4.9.1(a). generating the flat t
sampled pulses
Jenil Thakkar ETechKnowled

PUbIItation
ADCS (Sem.5/ECE/GTU) 4-19 Sampling
Therefore the width of s (t) is decided by h (t) and the
This is the replication property of the delta function.
is being used for the generation of flat
amplitude of s (t) depends on x (t).
This property
is expressed
top sampled pulses. The ideally sampled signal x, ()
4.9.3. From
This principle is graphically explained in Fig. mathematically as,

we can write that,

Fig. 4.9.3
2 x (n T) 8 (t- n T (4.9.4)
x, ()
=

h () 8 () *h (t) .(4.9.2)

Actual generation offlat top pulses Nows (t) = x, () * h (t)

In Equation (4.9.2), replace

on RHS byif we (t) xô (t) i.e.
Using the definition of convolution,
the ideally sampled signal then we get the flat top

sampled signal s (t). h (t-v) dv ..(4.9.5)

x, ()
*
h (t) =
x , (V)
Flattop sampled signal, s () =
xs () *h () -00

be graphically explained as shown in deliberately instead "V"

This equation can Note that "t" has not been used
Fig. 4.9.4. is being used.
***** ****************

*** Substitute Equation (4.9.4) into Equation (4.9.5) to get,

****

Continuous
tme
x (nT) & (v- nT) h (t-v) dv
analog s()
signal
- n=-co
(a) Contimuous time signal x(t) Interchanging the order of summation and integration
and rearranging we get,
Ideally
sampled
signal
s (t) = x(nT)J 8(v-nT)h (t-v) dv4.9.6)
OT, 2
- 00

********

(b) Ideally sampled signal x,() ww********

ht) Now let us use the shifting property of delta function as,
Pulsetrain..

f (t).8(t-t) dt f (t) 4.9.7)

- 00

( Sampling function h() of constant width pulses

Apply this property to Equation (4.9.6) to write,
s(x0)*h)
Flat top sampled signal

s (t) 2 x (nT)h (t-nT) 4.9.8)

n-o

This expression represents the flat top sampled signal in

d) Flat top sampled pulses s()
=
x() h()
time domain, in terms of the instantaneous sample
(D-436) Fig. 4.9.4 values x (nT) and train of fixed duration pulses.
are obtained by
hus the flat top sampled pulses To obtain the spectrum of s (, let us use the
convolution of the ideally sampled signal x () and
convolution theorem which states that, convolution in
puise train of finite pulse width h (t). The width of each
time domain is transformed into multiplication of
pulse is t sec.
transforms in frequency domain.

Spectrum of flat top sampled signal: s (t) = xg, (t)* h (t).

The flat top sampled signal is given by. Taking Fourier transform of both the sides we get,
s (t) x, (t) * h (t) ..4.9.3)

Jenil Thakkar Tech Knowledy

Pubilatlons
ADCS (Sem. 5/ ECE/ GTU) 4-20
s (
From Equation (4.9.1),
X, () H () .4.9.10) Spectrum of flat top sampled signal: Sampling
This is the expression for the
spectrum of a faat
X (f- nf) sampled signal. top
(4.9.11)
n - 0o
H is the spectrum of h (t). As h
Therefore Equation (4.9.10) (t) is a
becomes, pulse, its spectrum is a sinc function.
ectangular
S0 2x (f- nf). H() 4.9.12) Therefore product of the spectrums X (f) and H ( is
shown in Fig. 4.9.5, because s ) is equal to the product
of X () and H ().

X(0t
Spectrum of
analog signal
x(t)
M M

Spectrum of
instantaneously
sampled signal
(1+W) - W ) -w
W (-W) (s+ W)
H(0
Spectrum of
sampling
signal h(t)
-1/t W

-Error due to aperture effect

Spectrum of flat
topped sampled
ignal
-1/t -+W)
- ) -w w -W) +W) 1/t f

(D-437) Fig. 4.9.5 Spectrum ofa flat top sampled

signal

Observations from the Fig. 4.9.5 4.9.1 Aperture effect:

1. The signal x (t) has a flat spectrum over its entire Consider the spectrum of the flat topped signal. We are
range from 0 to W. The transform of the interested in the portion of the spectrum upto
instantaneous signal Xz () has been drawn below it. frequency W.
The sampling frequency f, = 1/T, is large enough
The spectrum should have been flat in this portion or

to allow the guard band. the spectrum but it is not as shown in Fig. 4.9.5.
The spectrum of the sampling signal h (t) is a sinc
2 The shaded portion shows an error due to an e ffect

function. called "aperture effect".

is the
3. The spectrum of the flat topped signal The high frequency
roll off characteristics of H
a typica
product of these two spectrum. Due to the ) acts like a low pass filter and attenuates the uppe pper

with the sinc function, this spectrum portion (high gnal

multiplication frequency) of message sy
the
goes to zero at f =1/T. spectrum. This loss of high frequency content is ca
called

as the "aperture effect".

Jenil Thakkar TechKnowledgo

P ubIIC atio
 ADCS (Sem. 5/ECE /GTU)
4-21 Sampling
Flat top
effect is due to the finite pulse width "" of sampled
ReconstruotionH Equalizer Analog slgnal
x(1)
The aperture Sgnal
Flter

sampling signal. With increase in the width t, the

the
(D-439) Fig. 4.9.7: Reconstruction of x (t)
reduce and the will increase.
frequency 1/t will
error

be reduced by reducing Equalizer

The aperture effect the
can

pulse width t
as shown in Fig. 4.9.6 The amplitude response of the equalizer is such that the
can be corrected in reconstruction equivalent transfer function is 1.
The aperture effect
equalizer.
by including an
Aperture error
H 1x| H,(0| =1
where H, () = Transfer function of the
Spectrum Pulse width z largee
of fial top
sampled equalizer
sig na W 1/t Transfer function of the
and H) =

Spectnum Redúction in aperture effect due to

reconstruction filter
offlattop reduction in the
sampled pulse width "r"
signal H (I = H (O| T sinc (f T)

(D-438) Fig. 4.9.6: Effect of pulse width "" on the aperture

1tfT
effect T sin (r ft)

Reconstruction of original signal x (t): Merits and demerits of flat top sampling:
Due to the aperture effect discussed earlier, an
1 Better SNR due to increased signal power. This is

amplitude distortion as well as a delay is introduced in due to the finite width "t" of the pulses.
the flat top sampled signal. Generation is easy.
2
This distortion can be corrected by connecting an
3. Practical filters can be used for reconstruction.
equalizer after the reconstruction filter (low pass filter)
4. Aperture effect introduces distortion.
as shown in Fig. 4.9.7.

4.9.2 Comparison of Sampling Techniques

Table 4.9.1: Comparison of sampling techniques

Ideal sampling Natural sampling Flat top sampling

Sr. Parameter
No.
Train of impulses Train of finite duration Train of finite duration
Nature of the sampling
pulses pulses
function
Uses a multiplier Uses a chopper Uses a sample and hold
Circuit arrangement
circuit

3. Practically not Practically realizable Practically realizable

Practical realizability
realizable

Refer Fig.A Refer Fig. B Refer Fig. C

4 Waveforms
Tends to infinity Satisfies Nyquist criteria Satisfies Nyquist criteria
5.
Sampling rate
3. Mathematical representation
in time domain
(t)= s () 2 s (t)=
n - co
n-
x (nT ) 8 (t-nT) x () sinc (nf,c) e x (T) h (t-nT.)

7.
Frequency spectrum X () =f,
n - coo
s (-1.
X(f nfs) sinc (nf,) x (f- nf,) | X(f-nf,)H (0

Jenil Thakkar TechPublc

Knowledgö
atlons
 ADCS (Sem. 5/ ECE/GTU) 4-22
Sr.
No.
Parameter ldeal sampling Natural sampling Flat top sampling
Samping
8. Signal power Very low due to the use Increases with increase in
Increases with increase
of impulses the pulse width t in pulse widtht
9 Bandwidth requirement Very high Increases with the Increases with reduction
reduction in pulse width in pulse width
10. Effect of noise
Maximum Moderate Moderate
-

xg(t) Review Questions

x(t)
Q. 1 Define the sampling process and explain its
necessity in communication system.

Q. 2 State and prove the sampling theorem for a low

pass bandlimited signal.
Fig. A
Q. 3 What do you understand by the word bandlimited ?
Q. 4 Explain the term aliasing and its effects.

Q.5 How can we avoid aliasing ?

* Q.6

Q.7
State the sampling theorem for a

What is the difference between ideal and

bandpass signal.
practical
(D-440) Fig. B Fig.C samp ing
Q.8 Why can't we use the ideal sampling in practice?
4.10 Applications of Sampling Theorem . 9 Explain the process of recovering the original signal
The from its samples.
sampling theorem is extremely important in signal
Q..5 What is
analysis, processing and transmission. baseband formatting?
Q.6 What is a low pass signal ?
It allows us to replace a continuous time signal byya
discrete time signal (sequence of samples). Q.7 Write a note on sample and hold circuit.

This is useful in digital filters. Some other applications of 8 Explain the generation of flat topped samples using
sampling theorem are: sample and hold circuit.
Q. 9 State and
Pulse Amplitude Modulation (PAM) system. explain sampling theorem.
Pulse Width Modulation (PWM) system.
Q.10 State and explain the practical sampling techniques
Q.11 Derive the sampling theorem for low pass
3 Pulse Position Modulatin (PPM) system. signas
Q.12 Write a note on: Flat
Pulse Code Modulation (PCM), Delta Modulation topped sampling
(DM), Adaptive Delta Modulation (ADM) systems. Q. 13 Compare various sampling techniques.
Time Division Multiplexing (TDM)

Jenil Thakkar
echKnowledg

PubIltatipn>