Communication System Engineering

th
PET5I103 ANALOG COMMUNICATION (3-0-2) (5 Sem ECE-
ETC)
MODULE-I
1. SIGNALS AND SPECTRA: An Overview of Electronic Communication Systems,
Signal
and its Properties, Fourier series Expansion and its Use, The Fourier Transform,
Orthogonal Representation of Signal.
2. RANDOM VARIABLES AND PROCESSES: Probability, Random variables, Useful
Probability Density functions, Useful Properties and Certain Application Issues.
3. AMPLITUDE MODULATION SYSTEMS: Need for Frequency translation,
Amplitude
Modulation (Double Side Band with Carrier DSB-C), Single Sideband Modulation
(SSB) Other AM Techniques and Frequency Division Multiplexing.
MODULE-II
4. ANGLE MODULATION: Angle Modulation, Tone Modulated FM Signal, Arbitrary
Modulated FM signal, FM Modulators and Demodulators, Approximately Compatible
SSB Systems.
5. PULSE MODULATION AND DIGITAL TRANSMISSION OF ANALOG
SIGNAL:Analog
to Digital (Noisy Channel and Role of Repeater), Pulse Amplitude Modulation and
Concept of Time division multiplexing, Digital Representation of Analog Signal
MODULE-III
6. MATHEMATICAL REPRESENTATION OF NOISE: Some Sources of Noise,
Frequency-domain Representation of Noise, Superposition of Noises, Linear
Filtering of Noise.
7. NOISE IN AMPLITUDE MODULATION SYSTEM: Framework for Amplitude
Demodulation, Single Sideband Suppressed Carrier (SSB-SC), Double Sideband
Suppressed Carrier (DSB-SC), Double Sideband with Carrier (DSB-C).
MODULE-IV
8. NOISE IN FREQUENCY MODULATION SYSTEM: An FM Receiving System,
Calculation of Signal to Noise Ratio, Comparison of FM and AM, Pre emphasis and
De-emphasis and SNR Improvement, Noise in Phase Modulation and Multiplexing
Issues, The FM Demodulator using Feedback (FMFB).
Additional Module (Terminal Examination-Internal)
1. AMPLITUDE MODULATION SYSTEMS:Radio Transmitter and Receiver.
2. PULSE MODULATION: Pulse Width Modulation and Pulse Position Modulation.
3. SYSTEM NOISE IN FREQUENCY MODULATION:Threshold in Frequency Modulation,
Calculation of Threshold in an FM Discriminator.

PEL6I101 COMMUNICATION ENGINEERING (6th Sem EEE)

MODULE-I
INTRODUCTION: Elements of an Electrical Communication System, Communication
Channels and their Characteristics, Mathematical Models for Communication Channels
FREQUENCY DOMAIN ANALYSIS OF SIGNALS AND SYSTEMS: Fourier series,
Fourier Transforms, Power and Energy, Sampling and Band limited signals, Band pass
signals
MODULE-II
ANALOG SIGNAL TRANSMISSION AND RECEPTION: Introduction to modulation,
Amplitude Modulation (AM), Angle Modulation, Radio and Television broadcasting
MODULE-III
PULSE MODULATION SYSTEMS: Pulse amplitude modulation, Pulse Time Modulation
PULSE CODE MODULATION: PCM system, Intersymbol interference, Eye patterns,
Equalization, Companding, Time Division Multiplexing of PCM signals, Line codes,
Bandwidth of PCM system, Noise in PCM systems,
MODULE-IV
Delta Modulation (DM), Limitations of DM, Adaptive Delta Modulation, Noise in Delta
Modulation, Comparison between
PCM and DM, Delta or Differential PCM (DPCM), S-Ary System
 MODULE-I
Elements of an Electrical Communication System

Information Transmitter Channel Receiver Information

Source Sink
Speech, Music, Analog Additive,
Analog
Image, Video, (Point-to-point, Linear Time Digital
Instrumentation and broadcast), Low Invariant, Linear
measurement signals, Power, Medium Time varying,
EEG, ECG, Seismic, Power, High Band limited,
meteorological signals, Power, Digital Power limited Transducers
Data etc
(Speaker, Picture
(Transducers required) tube etc)

Information is obtained from real life signals through the use of transducers. For example,
speech is converted into a corresponding electrical signal by a microphone and moving
picture signals are converted into the appropriate electrical signals by various cameras. The
information so obtained is called a signal that becomes a function of time which is usually
analog in nature. Signals may be described in time domain or in frequency domain. The
frequency domain description of a signal is known as spectrum that would be covered
subsequently. Data generated by the keystroke of a computer become the information when
communication is made through e-mail.

The transmitter may operate in a point-to-point mode or in a broadcast mode wherein there is
a number of receivers corresponding to a single transmitter. It may be wired, wireless. The
transmitter may also operate at different power levels depending upon the application, range
of service and type of service. We get three distinct types of transmitters: simplex, half
duplex and full duplex. The broadcast transmitters usually meant for entertainment purpose
are simplex type as information flow is unidirectional. The receiver can not communicate
back to the transmitter. In half duplex system, information can flow between the transmitter
and the receiver in one direction only at a time, but not simultaneously. The walkie-talkie is
an example of simplex type of communication. The telephone provides an example of a full
duplex type of communication.

The channel may be modelled as

 Additive noise type: the channel introduces noise that is added to the transmitted
signal (satellite channels)
 Channel
xt 
+ r t   xt   nt 

nt 

 Linear time invariant (LTI) type: the channel behaves as a linear filter whose
impulse response (or alternatively the transfer function) does not vary with respect
to time. The transmitted signal is convolved with the impulse response to produce
the channel output. (Leased land line telephone lines or simply the telephone
channel)

xt  Linear
+ r t   xt   ht   nt 
Filter

ht  nt 

Channel

 Linear time varying (LTV) type: The channel again, here behaves as a linear filter.
However, unlike the LTI channel, the impulse response of the channel varies with
respect to time. The channel output is observed to be a convolution of the
transmitted signal and the time varying impulse response. Cellular channels
provide a bright example of this kind of channel.

Linear
xt  time
+ r t   xt   ht ;   nt 
varying
nt 
Filter

Channel
The receiver’s function is to retrieve the original transmitted signal from noisy, distorted
signals that arrive at its input. An analog receiver is entrusted with the task of replicating the
original waveform from its noise corrupt and channel induced distorted versions. A digital
receiver makes a decision (within a sampling interval) as to “which one out of M number of
symbols”.

Performance metric of receivers:

Signal to noise ratio (SNR) at receiver output fort the analog one
Probability of bit error or
Mean square error (MSE) for the digital type.

The sink is usually a speaker that reproduces speech signals from the corresponding electrical
output or a picture tube that reproduces the picture. It may be a computer also that is intended
to receive an e-mail.

Electromagnetic Spectrum

Table No. 1.1 Allocation of frequencies for existing applications

S Frequency Nomenclature Application/Usage

No Range
1 30 Hz- 300 Hz Extremely low frequencies Underwater Communication
(ELF)
2 300 Hz- 3. 0 KHz Voice Frequency (VF) Telephone
3 3.0 KHz – 30 KHz Very low frequencies (VLF) Navigation
4 30 KHz- 300 KHz Low Frequency (LF) Radio navigation
5 300 KHz - 3 MHz Medium Frequencies (MF) AM radio broadcasting
6 3 MHz- 30 MHz High Frequencies (HF) AM, Amateur radio, mobile
7 30 MHz – 300 Very High Frequencies TV, FM, Mobile communications
MHz (VHF)
8 300 MHz- 3 GHz Ultra High Frequencies TV, radar, satellite communications
(UHF)
9 3 GHz- 30 GHz Super High Frequencies Terrestrial microwave and satellite
(SHF) communications
10 105 GHz – 106 Optical Frequencies Optical communication
GHz

Signal Analysis: Fourier Series

A signal is periodic if it repeats itself after a certain time; xt   xt  T  where T is its period.


 j 2nt 
xt   x n exp 
n    T 

1 T  j 2nt 
where x n   xt exp  dt
T 0
 T 
Two signals x1 t  and x2 t  are said to be orthogonal over a period T if their inner product is zero;

 x t x t dt  0 for the case when the signals are real valued functions.
T
1 2
0

For example: V sin 2f 0 t and V cos 2f 0 t , V sin 2mf 0 t , V sin 2nf 0 t are orthogonal to each other
over the period T

1
The fundamental frequency is expressed as f 0 
T

• The signal must satisfy a set of conditions known as ‘Dirichlet’s conditions’

• These ae

• A) The signal is absolutely summable over its period

 xt  dt  
0

2nt 2nt
T T
2 2
The coefficient a n   xt cos dt and bn   xt sin dt
T0 T T0 T

1 2 b 
xn  a n  bn2 and  x n    tan 1  n 
2  an 

1
where   1
n

Fourier transform of a function is evaluated as


Xf   xt exp j 2ft dt


t
Q.1 Show that  m d  mt   u t 


Proof:

 t
mt   u t    m ut   d   m d
 

This is because u t     0 for   t .

Q.2. Find the spectrum for a signal defined as

vt   sin 2f1t cos 2f 2 t

 1
Soln: The signal is v t   sin 2f 1t cos 2f 2 t  sin2  f1  f 2 t   sin2  f1  f 2 t 
2

The spectrum corresponding to the first term is

1
  f   f1  f 2     f   f1  f 2 
2j

The spectrum corresponding to the first term is

1
  f   f1  f 2     f   f1  f 2 
2j

dg t 
Q.3. Find the spectrum of a signal defined as t
dt

Soln: We know that

d
The function g t   G  f  , then g t   j 2fG  f  and similarly,
dt

d
G  f    j 2t g t 
df

Let us differentiate the function g1 t    j 2t g t  once more with respect to time. Hence, we
obtain

d  d  d
g1 t    j 2t g t   j 2g t   j 2f G f 
dt  dt  df

From the linearity property of the Fourier transform operator, we have, corresponding to the second
term of the expression,

 j 2g t    j 2G f  (Multiplying  j 2 to both sides )

dg t 
Thus, the Fourier transform of the function t is obtained by writing
dt

dg t  1
t   j 2fG  f   1  j 2G f   G  f   fG  f 
dt j 2 j 2

Q.4 Find the Fourier transform of a function given as xt   t  2 f  2t 

From the frequency differentiation property of the Fourier transform, we have

 1 d
tx t   Xf 
 j 2 df
1 d  f
 tx 2t   X  
 j 2 .2 df  2 
1  f   f 
 2 x 2t   2. X    X  
2  2  2
1 d  f   f 
 t  2 x 2t   X    X  
 j 2 .2 df  2   2

Q.4 Find the Fourier transform of a signal given as

dx
y t   t
dt

Soln: Differentiating a function like tx t  in the time domain, we have

d
txt   t d xt   xt 
dt dt
d d
 t xt   tx t   xt 
dt dt

From the above, we note that,

1 d
tx t   Xf 
 j 2 df
d 1 d d
 txt   j 2f . Xf   f Xf 
dt  j 2 df df

However, from the linearity principle, the time differentiated function has two parts; the transform
corresponding to yt  and the other corresponding to xt  . Therefore,

d
f X  f   Y f  X  f 
df
 d 
 Y  f    X  f   f X  f 
 df 
dx  d 
t   X  f   f X  f 
dt  df 

Q.5 Find the Fourier transform of a signal defined as

yt   x1  t 

Soln: y t   x1  t   x t  1

 xt   X  f 
We know that, for x t   X *  f 
 x t  1  X *  f  exp j 2f 

This is because


 x t    x t exp j 2ft dt


Let
t  
 dt  d

   x  exp j 2f d




 x exp j 2f d


 x exp j 2f d 





 X  f 

Q.6 Find the Fourier transform of a signal given as

yt   1  t x1  t 

Soln: As in the previous problem,

tx 1  t  
1 d
j 2 df
Yf  
1 d
j 2 df

X *  f  exp j 2f  
1 d * 
  j 2X  f  exp j 2f   df X  f  exp j 2f 
*

j 2
 
 1 dX  f 
*

  X *  f  exp j 2f 
 j 2 df 

From the linearity property of the Fourier transform, we have

 1 dX *  f  
y t   1  t x1  t   X *  f  exp j 2f     X *  f  exp j 2f 
 j 2 df 
 1 dX *  f  
 2 X *  f    exp j 2f 
 j 2 df 
The impulse has no mathematical or physical meaning unless it appears under the operation of
integration. Two of the most significant integration properties are

I. Replication property

xt    t  t 0   xt  t 0 

To prove this, we write

 x  t    t d

0

let
t    t0  
   t    t0
 d   d

  xt    t   d

0


  xt    t 0   d


We know that,    attains a value of 1 at   0 . Therefore,

 xt    t   d  xt  t .1


0 0

 xt  t 0 

This is known as replication property.

II. Sampling property

We also have,

 xt  t  t dt  xt 


0 0

To prove this, we note that

t  t0  
 t    t0
 dt  d

Hence the above integral becomes

 x  t   d

0
We know that,    attains a value of 1 at   0 . Therefore, the above integral has just one value
that is nonzero occurring at   0 and this value is given as

xt 0 

This completes the proof.

III.

Further, we have

xt  t  t 0   xt 0  t  t 0 

This is because the impulse function has a value of 1 at t  t 0 . Hence, only one value of the function
xt  is retained which occurs at t  t 0 .

1
 at    t 
a

This is because

Let us evaluate

  at dt


at  x
x
Let  t 
a
dx
dt 
a

Therefore,

 
1 1
  at dt 


a 
  x dx   t 
a

IV. xt    t  t 0   xt  t 0 

To prove this, we write

xt    t  t 0 

  x  t    t d

0

Let
t    t0  
 t    t0  
 d   d

Substitution of the above in the integral gives us

 x  t    t d

0


   xt    t 0   d


 xt  t 0 

This is because the impulse function has a value of 1 at t  t 0 . From the previous problem we get
this.

V. xt  T1    t  T2   xt  T1  T2 

xt  T1    t  T2 

  x  T  t    T d

1 2

t    T2  
Let  d  d
   t    T2

 x  T  t    T d

1 2


   xt    T1  T2   d
Thus, 

  xt    T1  T2   d


 xt  T1  T2 

A little extension of the this result as applied to impulse functions give us

VI.  t  T1    t  T2    t  T1  T2 

Soln: Let us prove this using the Fourier transform properties.

We know that,
xt   X  f 
 xt  T1   X  f  exp j 2fT1 
  t  T1   exp j 2fT1 

Therefore,

 t  T1    t  T2   exp j 2fT1  exp j 2fT2 

 exp j 2f T1  T2 
  t  T1  T2 

Prove the duality theorem of Fourier transform which states that if xt   X  f  , then
X t   x f 

Proof:


xt    X  f exp j 2ft df


Hence,


x t    X  f exp j 2ft df


Let us interchange the roles of frequency and time in the above expression

Therefore,


x f    X t exp j 2ft dt   X t 


 X t   x f 

Find out the Fourier transform of x t  .

Soln: We know that,


xt    X  f exp j 2ft df


Let t  

Therefore,


x     X  f  exp j 2f df

Table 1.2 Some commonly used functions and their Fourier transforms

xt  Xf 
1 t  T sinc  fT 
rect  
T 
2 sinc 2Wt  1  f 
rect 
2W  2W 
3 exp at u t , a  0 1
a  j 2f
4 exp a t , a  0 2a
a 2  2f 
2

5 
exp  t 2  
exp  f 2

6  t T sin c  fT  2

1  , t  T
 T
0 , t  T

7 t exp at u t  1
a  j 2f 2
8 sgn(t) 1
jf
9 t 2 1

2f  2
2 2 f 2

10 u t  1 1 
    f 
2  jf 
11 exp j 2f c t    f  fc 
12 
1 
 m
  t  nT  0    f  T 
n   T0 m   0 

Prove the 7th entry of Table 1.1 from the appropriate property of the Fourier transform.

Soln: The appropriate property that we use to prove this is the frequency domain differentiation
which is

1 d
txt   Xf 
 j 2 df

As we note that,
 1
exp at u t , a  0 
a  j 2f
1 d
 t exp at u t    Xf 
j 2 df
1 d  1  1  j 2 1
     
j 2 df  a  j 2f  j 2 a  j 2f  2
a  j 2f 2
This completes the proof.

Prove the 9th entry of Table 1.1 using appropriate properties of Fourier transform

Soln: We make use of the previous result. The function under consideration may be expressed as

t t  0
t 
 t t  0

The positive going part may also be considered as a limiting case of

lim t exp at u t 
a 0

Similarly, the negative going part may also be considered as the limiting case of the previous
function however, with a reversed time

t t  0
t 
 t t  0
 limt exp at u t   t expat u  t 
a0

Let us combine the Fourier transforms of the two functions

Xf 
1

1


2 a 2   j 2f  
2

a  j 2f  2
a  j 2f  2
a 2
  j 2f  
2

For the limiting case of a  0 , we have

lim X  f  

2 a 2   j 2f 
2
  2  j 2f  2

2
a 0
a 2
  j 2f 
2 2
  j 2f  4
2f 2
Prove the 12th entry of Table 1.1.

Soln: The Fourier coefficient of this function is defined as

T0 2
1  n  1
T0   t exp  j 2 T t dt  T
T0 2 0 0

Hence, the Fourier series of an impulse train is expressed as

 1 
 nt 
xt    exp j 2 T 
T0 n    0 

The trigonometric Fourier series would consist of the coefficients a n ’ s and bn

T
2 0 2nt 2
a n    t cos dt 
T0 0 T0 T0

T
2 0 2nt
bn    t sin dt  0
T0 0 T0

From the 11th entry of this table, we note that

 n   n
exp j 2 t     f  
 T0   T0 

Therefore, the Fourier series corresponding to an impulse train is expressed as

1 
 m
Xf     f  T 
T0 m    0 

The Fourier series of some commonly used waveforms

1. Half wave rectified sine wave of amplitude A volt

1 1 2 2 2 2 
xt   A  sin t  cos 2t  cos 4t  cos 6t  cos 8t 
 2 3 15 35 63 
 A A 2m
2
    sin t   cos nt
  2 n2 n 1 
2
 
2. Full wave rectified sine wave of peak amplitude A volt
2A  2 2 2 2 
xt   1  cos 2t  cos 4t  cos 6t  cos 8t 
  3 15 35 63 
 2 A   2 A  2m 2
   2 cos nt
      n2 n  1  
3. Rectangular or square wave of peak to peak amplitude 2 A volt
4A  1 1 1 
xt    sin t  sin 3t  sin 5t  sin 7t 
  3 5 7 
4A 1
 
 n 2 k 1 n
sin nt

4. Triangular wave
8A  1 1 1 
xt   2 
sin t  sin 3t  sin 5t  sin 7t  ...
  9 25 49 
 A
0 T T
2

Fig.5

The waveform is expressed as

4A T
 T t 0t  4

 4A T 3T
xt    t  2A t
 T 4 4
4A 3T
 T t  2A 4  t  T

This function exhibits odd symmetry. Hence it contains the sine terms only. The coefficient
bn is evaluated as

2 T 4 4 A 3T 4
 4A  
bn   t dt    t  2 A  sin n 0 tdt 
T  0 T T 4
T  
2 4A T 4 4A
3T 4 3T 4


T
  t sin n 0 t dt 
T T4
t sin n 0 t dt  2 A  sin n 0 t dt 
 T 0 T 4 
Let us evaluate the above coefficient term by term. The first term gives us

2 4A 1 
 . sin n0t  n0t cos n0tT0 4 
T  T n0  2

8A  n0T n0T n T 
 2 
sin  cos 0 
n0T   4 4 4 
8A  n n n 
 2 2 
sin  cos 
4 n  2 2 2 
The second integral becomes
3T 4
2 4A

T T  t sin n t dt
T 4
0

2 4A 1 
  . sin n 0 t  n 0 t cos n 0 t T3T44 
T  T n 0 2

 3n 0T 3n 0T
8A 3n 0T n 0 T n 0 T n 0 T 
 sin 4  4 cos 4  sin 4  4 cos 4 
4n 2 2
 
8 A  3n 3n 3n n n n 
  2 2 sin  cos  sin  cos
4n   2 2 2 2 2 2 
The third integral becomes
3T 4
2
2 A  sin n 0 t dt
T T 4

4A  1 
cos n 0 t  T 4
3T 4
  
T  n 0 
4A  1  3n 0T n T 
    cos  cos 0 
T  n 0  4 4 
4A  1  3n n 
    cos  cos 
T  n 0  2 2 
The fourth integral is
T
2  4A 
.  t  2 A  sin n 0 tdt
T 3T 4  T 
2 4A 1 
  . sin n 0 t  n 0 t cos n 0 t T3T 4 
T  T n 0 2

8A  3n 0T 3n 0T 3n 0T 
 sin n 0T  n 0T cos n 0T  sin 4  4 cos 4 
n 0T  2
 
8A  3n 3n 3n 
 2 
sin 2n  2n cos 2n  sin  cos 
2n   2 2 2 
The fifth integral becomes

T
2
T 3T 4
 2 A. sin n o tdt

4A  1  3n 0T 
    cos n 0T  cos 
T  n 0  4 
Combining all the terms, we obtain
8A  n n n  8A  3n 3n 3n n n n 
2 2 
sin  cos  2 2 sin 2  2 cos 2  sin 2  2 cos 2 
4 n  2 2 2  4n   
4A  1  3n n 
    cos  cos 
T  n 0  2 2 
8A  3n 3n 3n 
  sin 2n  2n cos 2n  sin  cos 
n 0T  2
 2 2 2 
4A  1  3n 
   cos 2n  cos 
T  n 0  2 
 n n
For values of n  2m being even, the terms sin becomes  1
m
vanishes and cos
2 2
3n
The term cos  0 always. Similarly, for odd values of n  2m  1 , the term
2
cos
2m  1 0
2
This is simplified to
8A  n  8 A  3n n n n 
2 2 
sin   2 2 sin  sin  cos
4 n  2  4n   2 2 2 2 
4A  1  n  8A  3n 
     cos  2 
 2n cos 2n  sin 
T  n 0  2  n 0T   2 
4A  1 
  cos 2n 
T  n 0
2A  n n 3n 3n  2 A  n n 
 2 2  sin  sin  sin  sin  2 2  cos 
 n  2 2 2 2   n  2 2 
4A n 8A
 cos   2n   4 A
2 n 2 4 n 2 2
2 n
2A  n n 3n 3n 
 2 2  sin  sin  sin  in 
 n  2 2 2 2 
4A  n 3n 
 2 2 
sin  sin 
 n  2 2 
4 A   n  3n   n  3n 
 .2 cos  sin  
 2 n 2   4   4 
8A  n 
 .  cos n sin
2 2 
 n  2 

8A 8A
We note that, if n  1 , the above term is , for n  3 , it is  2 2 , for n  5 , the
 n
2 2
 n
8A
above term is . Hence the series amplitudes become alternately positive and negative
 2n 2
1
and vary at the rate of 2 .
n
Thus, the Fourier series expansion of the triangular waveform as shown in Fig.5 is
8A  1 1 1 
x t   2 
sin  0 t  sin 3 0 t  sin 5 0 t  sin 7 0 t  ... 
  9 25 49 
 8A  1 1 1 
6. xt   2 
sin t  sin 3t  sin 5t  sin 7t  ...
  9 25 49 

A
0 

For the above triangular pulse, the Fourier series is obtained by noting that it can be
obtained from Fig. 5 by shifting it by half a period.
  T 1  T 
sin  0  t  2   9 sin 3 0  t  2  
8A     
xt   2 
  1  T 1  T 
  sin 5 0  t    sin 7 0  t    ...
 25  2  49  2 
8A  1 1 1 
 2 
cos  0 t  cos 3 0 t  cos 5 0 t  cos 7 0 t  ... 
  9 25 49 

7. Full wave rectified triangular wave

A 4A  1 1 1 
xt    2  sin t  sin 3t  sin 5t  sin 7t  ... 
2   9 25 49 

A
0  2

This waveform exhibits even symmetry. This has an average value given as
T
1
xt dt
T 0
a0 

1 1 A
 . A.T 
T 2 2
The waveform is expressed as
2A T
 T t 0  t  2
xt   
 2 A t  2 A T  t  T
 T 2
The corresponding integrals become
2  2A 1 T 2
 . sin n  t  n  t cos n  t  0 
T  T n 0 2
0 0 0

4A  n T n T n T 
 2 
sin 0  0 cos 0 
n0T   2 2 2 
4A  n n 
 2 2 
sin n  cos 
 n  2 2 
2A n
 cos
n 2
The second integral becomes
T
2 2A
T T T2
 t sin n 0 t dt

2 2A 1 
  . sin n 0 t  n 0 t cos n 0 t TT 2 
T  T n 0 2

4A n 0 T n 0 T n 0 T 
 sin n 0T  n 0T cos n 0T  sin 2  2 cos 2 
n 2 2
 
4A  n n n 
  2 2 sin n  n cos n  sin  cos
n   2 2 2 
4A  n n n 
 2 2 
 n cos n  sin  cos
n   2 2 2 
4A 4A n 2 A n
 cos n  2 2 sin  cos
n n  2 n 2

The third integral becomes

T
2
2 A  sin n 0 tdt
T T 2

4A  1 
cos n 0 t  T 2
T
  
T  n 0 
4A  1  n T 
    cos n 0T  cos 0 
T  n 0  2 
4A  1  n 
    cos n  cos 
T  n 0  2 
 2A n 4 A 4A n 2 A n
 cos  cos n  2 2 sin  cos
n 2 n n 2 n 2
4 A  1  n 
    cos n  cos 
T  n 0  2 
4A n 4 A n 4 A 4A n 4A
 sin  cos  cos n  cos  cos n
n
2 2
2 n 2 n n 2 n
4A n
 2 2 sin
n 2
4A
 2 2  1
n

n
Hence, the Fourier series becomes

xt  
A 4A
 2 
 1 sin n t
n

2  n 2 m1 n 2
0

8. Trapezoidal waveform
The waveform is expressed as
6 A T
 T t 0t 6

A T T
t 
 6 3

 6A T 2T
xt    t  3A t
 T 3 3
 2T 5T
 A 3  t  6

 6 A t  6 A 5T  t  T
 T 6
The first integral is
T
^
2 6A
T T 0
. t sin n 0 tdt

12 A
 sin n0t  n 0t cos n0t  T0 6
n0T 2

12 A  n 0T n 0T n T 
 2 2 
sin  cos 0 
4 n  6 6 6 
3A  n n n 
  sin  cos 
 2n2  3 3 3 
The second integral is
 T 3
2
T T6
. A sin n0 tdt

2A T 3
 cos n0 t T 6
n  0T
2A  n T n T 
  cos 0  cos 0 
n 2  3 6 
A  2n n 
  cos  cos 
n  3 3 
The third integral is
2T 3
2  6A 
T   
T 3
T
t  3 A  sin n0 tdt

3  4n 4n 4n 2n 2n 2n 
 2 
sin  cos  sin  cos 
 n  2
3 3 3 3 3 3 
3A  4n 2n 
  cos  cos 
n  3 3 
We note that, the term
3A  4n 2n 
  cos  cos 
n  3 3 
3A  4n  2n   2n  4n 
 .2 sin   cos 0
n  3. 2   2.3 
The fourth integral is
5T 6
2
 . A  sin n 0 tdt
T 2T 3
A  5n 4n 
  cos  cos 
n  3 3 
The fifth integral is
6T 6
2  6A 
 
T 5T 6  T
t  6 A  sin n0 tdt

3  5n 5n 5n 
 2 
sin 2n  2n cos 2n  sin  cos 
 n 
2
3 3 3 
6A  5n 
  cos 2n  cos 
n  3 
Combining all the terms, we have
 3 A  n n n  A  2 n n 
2 2 
sin  cos   cos  cos 
 n  3 3 3  n  3 3 
3A  4n 4n 4n 2n 2n 2n 
 2 2  sin  cos  sin  cos 
 n  3 3 3 3 3 3 
A  5n 4 n 
  cos  cos 
n  3 3 
3A  5n 5n 5n 
 2 2  sin 2n  2n cos 2n  sin  cos 
 n  3 3 3 
6A  5n 
  cos 2n  cos 
n  3 
A 2 n 4 A 4n 2 A 2n A  5n 4 n 
 cos  cos  cos   cos  cos 
n 3 n 3 n 3 n  3 3 
6A 5A 5n 6 A 6 A 5n
  cos   cos
n n 3 n n 3
3A  4n 2n 
  cos  cos 
n  3 3 
3A  4n  2n 4n  2n 
  sin . sin 0
n  2.3 2.3 
3 A  n 4n 2n 5n 
 2 2 
sin  sin  sin  sin  Combining all sin  terms
 n  3 3 3 3 
3 A  n n 2n 2n 
 2 2 
sin  sin  sin  sin   sin       sin 
 n  3 3 3 3 
3. 2 A  n 2n  3.2.2 A  n  2n   n  2n 
 2 2 
sin  sin  sin   cos 
 n  3 3   2n2  3.2   3.2 
12 A  n   n 
 2 2 sin   cos  Here n can not be even, it can not be a
 n  2   6 
multiple of 3 either. The allowed values of n are1, 5 ,7, 11...
12 A
  1 3  62 32
 n
2 2
2  n
If we combine all the cosine terms, the result is zero.
The desired Fourier series is
6 3A  1 1 
xt    sin 0 t  sin 5 0 t  sin 7 0 t  
 
2
25 49 
9. A periodic impulse sequence (Impulse train)


 T t     t  nT 
n  

where T is its period.

The coefficient a 0 is evaluated as

 T 2
1 1
a0    t dt 
T T 2 T

Similarly,

T 2
2 2 2
an    t  cos n t dt 
T T 2 T T

As the delta function train is an even function of time, the coefficient bn is zero.

The Fourier series corresponding to such an impulse train is expressed as


 2  1 
 2 
 T t    x n exp jn t   exp jn T t
n    T  T n   

1
This is because the coefficient c n  for the delta train
T

The generalized Fourier series of any arbitrary periodic signal is expressed as


 j 2 
xt   x n exp nt 
n    T 
   j 2 
 xt     x n exp nt 
 n    T 

  j 2 
  x exp
n
T
nt 

n   

 j 2 
  x exp
n
T
nt 

n   

 n
  x n  f  
n    T

Here, we define just one period of the wave given as

 xt   T 2  t  T 2
xT T   
0 otherwise

Any arbitrary periodic signal may be expressed as

 
x t    xt  nT   x t     t  nT 
T
n   n  

Taking the Fourier transform of both the sides, we have

     

x t     x t  nT    xT t     t  nT 
 n     n   
 
1  m 1  m  m
 XT  f     f     X T   f    X  f 
T m    T  T m    T  T

Comparing the two results, we have


1 n
xn 
T
X
n  
T  
T 

Hence, the generalized Fourier series of any arbitrary periodic signal is expressed as, using this result


 j 2 
xt   x exp n nt 
n    T 
1 
n  j 2 
  X T   exp nt 
T n    T   T 

Taking the Fourier transform of both the sides, we have

    

x t     x t  nT    xT t     t  nT 
 n     n   
 
1  m 1  m  m
 XT  f     f     X T   f    X  f 
T m    T  T m    T  T

For the sampled signal case, when t  nT , we have

xnT   xT nT    nT 

 
x t    xt  nT   x t     t  nT 
T
n   n  

Taking the Fourier transform of both the sides,

    

x t     xt  nT    x T t     t  nT 
 n     n   
 
 Xf   xt  nT    X T  f  exp j 2fnT 
n   n  

We note that,

 t  mT   exp j 2fmT 
 
Hence, the Fourier transform of  T t     t  mT    exp j 2fmT  is
m   m  
Q. Find the spectrum of a full wave rectified sine wave from fundamentals.

A sampled signal is expressed as

 
x t   xt    t  nTs    xnTs  t  nTs 
n   n  

  
  
 x t    xt    t  nTs   X  f      t  nTs 
 n     n   
1 
 n
 Xf     f  T 
Ts n    s 

1 
n  n

Ts
 X  T   f  
Ts 
n    s 
  
   xnTs  t  nTs 
 n   

  xnT  t  nT 
n  
s s


 j 2nt 
  xnT exp  s
Ts 

n   
1 
 n

Ts
 X  f  T 
n    s 

If we compare both the sides of the transform, we note that,


1 
n
 xnTs   Ts
 X  T 
n   n    s

Find out the Fourier transform of a Gaussian pulse given as

xt   exp  t 2  
Soln: The Fourier transform is expressed as


Xf   exp t exp j 2ft dt
2



  exp t  j 2ft dt 
2



  exp t  j 2ft  f 2  f dt
2 2



By adding and subtracting a term like f 2

to the argument of the exponential function
 
Xf   exp t  j 2ft  f 2  f dt
2 2



 exp f   exp  t  jf  dt


2 2



 t  jf   u
Let
  dt  du

Substituting this in the above, we get


Xf 
2
exp  f  2
 exp u du
2

 0



2

exp  f 2

 2
 exp  f 2

Inference: The spectrum of a Gaussian pulse is also another Gaussian pulse

Second Method:

This can also be derived by another method.

Frequency domain differentiation of a given signal gives us

xt   X  f 
dX  f 
   j 2txt 
df

Suppose, we have a signal that is described by a first order differential equation expressed as

dx t 
 2tx t 
dt

Taking the transform of both the sides, we have

dX  f 
j 2fX  f    j
df
 X  f   exp  f  2

If xt  is a continuous signal bandlimited to  m radians per second, then show that

k
xt   sin ckt   xt  for k   m

Proof:
k
xt   sin ckt  becomes in the frequency domain,


k   f 
 X  f .     X  f   k    k
 k  2k  

Taking the inverse transform we note that, in the range of  k    k , the signal would be exactly
equal to xt  for a frequency range of k   m .

We note that, in order to replicate the function xt  , the condition is that k   m otherwise for
k   m , multiplication of the two functions in the frequency domain would result in spectrum
mutilation of X  f 

Hence show that,

k
sin c m t   sin c m t   sin c m t  for  n   m

Proof:

Use of the above result gives us

n   f    f    f 
   .         for  n   m
   m  2 m   n  2 n    m  2 m 

Taking the inverse transform of the above result we get

n   f    f    f 
   .         sin c m t  for  n   m
   m  2 m   n  2 n   m  2 m 

S Nomenclature Mathematical Description

No.
1 Linearity ax1 t   bx2 t   aX 1  f   bX 2  f 
2 Time Scaling 1 f
xat   X  
a a
3 Duality If xt   X  f  then X t   x f 
4 Time Shifting xt  t 0   X  f exp j 2ft 0 
5 Frequency
Shifting xt  exp j 2f c t   X  f  f c 
(Modulation
Theorem)
6 Area under 

xt   xt dt  X 0


7 Area under 

Xf   X  f df

 x0 

8 Time domain dxt 

differentiation  j 2fX  f 
dt
9 Time domain X 0 
t
1
integration x d  j 2f X  f   2   f 
10 Frequency dX  f 
domain   j 2txt 
df
differentiation
11 Complex x  t   X   f 
conjugation
Re xt 
12 1
2

X  f   X   f  
Im xt 
13 1
2j

X  f   X   f  
14 Multiplication 

in time x1 t x2 t    X  X  f   d

1 2
domain 

15 Time domain 

convolution  x t x t   d  X  f X  f 

1 2 1 2

16 Parseval’s  

 xt y t dt   X  f Y  f df

* *
Theorem
 
17 Rayleigh’s  

 xt  dt   x f  df
2 2
Theorem
 
18 Moments  n n
 j  d
Property t n
x t dt    Xf 
 2  df
n
f 0

Prove property 16 in table 1.3

 

 xt  y t dt   X  f Y  f df

* *

 

Proof:
 

 xt y t dt   X  f exp j 2ft dfy t dt

*

 

 X  f df yt exp j 2ft  dt 

 

 X  f df yt exp j 2ft dt  

*

 

  X  f dfY  f 
*


Next we show the convolution of two rectangular pulses of different amplitudes and different
durations. The result is observed to be a trapezoidal pulse having a duration equal to the sum
of the durations of the individual pulses.
 A2

T1
A2

T2
A2

-T2
A2

-T2+t t

A1A2t

t
A2

-T2+t t

A1A 2 T2 t

 A A dλ  A A T
 T2  t
1 2 1 2 2

T2

-T2+t t
A1A 2 T2
T1

A A dλ  A A T T  t
T t
1 2 1 2 1 T2
1+T2
Objective: Fourier Transform of Periodic Signals

We have a periodic signal xt   xt  T0  having a period T0 that satisfies the Dirichlet’s
conditions. As we have seen previously, this signal is expressed as a linear weighted
combinations of its Fourier series coefficients xn  as


 j 2n 
xt   x n exp t 
n    T0 

Taking the Fourier transform of both the sides we get

   j 2n  
X  f   xt     xn exp t  
 n    T0  
   j 2n  
 n
  x n exp t    x n  f  
n     T0  n   T0 

We observe the following from the above:

 That Fourier transform of a periodic signal xt  consists of a sequence of

impulses in frequency at multiples of the fundamental frequency of the periodic
signal.
 The weights of the impulses are just the Fourier series coefficients of the periodic
signal
 Thus we obtain a discrete or line spectrum corresponding to a periodic signal
 Properties of Fourier transform would be utilized to compute the Fourier series
coefficients as follows

We define a truncated signal xT0 t  as

 T0 T0
 xt    t 
xT0 t    2 2
0 otherwise
This only means that we are just considering one period of the signal and we have set other
periods to zero. The periodic signal is restored by repeating this truncated signal with a period
of T0 . Hence, we get back our signal as
 
xt    xT0 t  nT0   xt     t  nT 
0
n   n  

Now, apply Fourier transform to both the sides.

 
 

X  f   x t     xT0 t  nT0    xt     t  nT0 
n    n   
1 
 m
 X T0  f     f  T 
T0 m    0 

1 
 m

T0
 X  f   f  T 
T0
m    0 

Let us compare both the transforms. We can immediately see that


 n 1 
m  m
Xf   x   f  T
n
  X T0
   f  
n    0  T0 m    T0   T0 
1 n
xn  X T0  
T0  T0 
The following steps are followed to find out the Fourier series coefficients xn :
 Truncate the signal to just one period.
 Determine the Fourier transform of this truncated signal.
n
 Evaluate the Fourier transform of the truncated signal at a frequency f  to obtain
T0
1
the n  th harmonic and multiply it by
T0
Example: Find out the Fourier series coefficients of a triangular pulse train by this method.

Soln: From Table 2.3, we note that the truncated triangular pulse has a Fourier transform
given as
 t
1  , t  T
xT0 t    T  T sin c 2  fT 
0 , t  T

n
Hence, at a frequency of f  , this becomes
T0
n n 
X T0    T sin c 2  T 
 T0   T0 
1 1 n 1 n  T n 
Multiplying it by , we obtain X T0    .T sin c 2  T   sin c 2  T 
T0 T0  T0  T0  T0  T0  T0 
2
  nT    nT 
 sin    sin 2  
T  n  T T T T
sin c 2  T     0    0 
T0 T
 0  T    nT   T  2 2 2
n T
   
0 0

T
  0   T02
We observe from the above that, for a triangular wave, the Fourier series coefficients decay at
2
 n
the rate of   and they are always positive. This is a faster decay as compared to a similar
 T0 
duration rectangular waveform.

Objective: To learn power, energy and autocorrelation function of a given signal.

The energy and power of a signal are representative of the energy or power delivered by the
signal when the signal is interpreted as a voltage or current source feeding a 1Ω resistor. The
energy content of a signal xt  , denoted by  x is defined as


 xt 
2
x  dt and the power similarly, can be expressed as

T 2
1
xt  dt
T  T 
2
Px  lim
T 2

A signal is energy-type if  x   and is power-type if 0  Px   . A signal can not be,

therefore both an energy or a power type signal . For energy type signals, Px  0 and for
power type signals  x   . Usually all periodic signals (with the exception of xt   0 ) are
power-type and have power
 T0
1
 xt 
2
Px  dt
T0 

In the above, T0 is the period of the signal and  is any arbitrary number.
Example: Find out the average power in a periodic sine wave.

Soln: Let the sine wave be represented as

xt   Vm sin 2ft
1
Where f 
T
 T0 T
1 1
 xt  dt  Vm sin 2ft 2 dt
T 0
2
Px 
T0 

Vm2 T 2 Vm2 T 2
Vm T T
 2
Vm
Thus, 
T 0
sin 2ftdt 
2T  1  cos 4ft dt  2T  dt   cos cos 4ftdt   2T .T
0 0 0 
2
V
 m

2
T
This is because  cos 4ftdt  0
0

Energy-type Signals
The energy of a signal may be expressed as
 

 xt 
2
x  dt or



 x f 
2
x  df


This follows from the fact that the energy of a given signal can not be different whether it is
computed in the time domain or in the frequency domain. The equality of the two above
expressions is known as Rayleigh’s theorem.

Example: Find out the energy contained in a signal given as

xt   10 sin c10t
Soln: It is easier to evaluate the energy in the frequency domain. The spectrum
10  f   t 
of this signal is xt   10 sin c10t  X  f       
10  10   10 
This is a rectangular pulse in the frequency domain with unit amplitude and
bandwidth of 10 units.
 5

 x f  df   12 df  10 units
2
Therefore,  x 
 5

Relation between convolution and autocorrelation of a given function xt 

We may compute the autocorrelation of an energy-type signal as
 
R x    x   x *      xt x t   dt 

 xt   x t dt


 

This is a function of the lag  and also gives us the relationship between the autocorrelation
and convolution of a given signal. As the signal is correlated with itself for different values of
this lag parameter, it is known as autocorrelation. We are trying to find out the degree of
similarity between the original waveform and a delayed or advanced version of it.
By setting   0 in the above, we obtain
  
R x 0   x 0   x * 0    xt x t dt   xt x t dt   xt 
  2
dt   x
  

Let us find out the time-average autocorrelation function and power spectral density of the
power type signals. Let us assume that xt  is a periodic signal with period T0 that has the
Fourier series coefficients xn . The time-average autocorrelation function for such a signal is
defined as
T 2
1
Rx    lim xt x * t   dt
T  T 
T 2
kT0 2 T0 2 T0 2
1 k 1
 lim  xt x t   dt  lim
*
 xt x t   dt 
*
 xt x t   dt
*
k  kT k  kT T0
0  k T0 2 0 T0 2 T0 2
These steps were followed to eliminate the limiting term and to express the autocorrelation
function in terms of one period of the signal. The substitution of the Fourier series expansion
in the above yields
 j 2nt   t 
T0 2 T0 2
1 1 
 j 2nt   *
Rx     xt x t   dt 
*
  x n exp   xm exp  dt
T0 T0 2
T0 T0 2 n    T0  m  T0 
T0 2
1  
 j 2nt   j 2mt   j 2n 

T0   x
T0 2 n   m  
n x m* exp
 T0 
 exp 
 T0 
 exp
 T0
dt

T0 2
1  
 j 2nt   j 2mt   j 2n 

T0   x
T0 2 n   m  
n x m* exp
 T0 
 exp 
 T0 
 exp
 T0
dt


 j 2n  
 j 2n 
xx    xn exp
2
 exp
n
*
n

n    T0  n    T0 
We note that, the autocorrelation function of a periodic signal consists of discrete valued
power components located at integral multiples of the fundamental. The power components
2
are proportional to xn . Taking the Fourier transform of both the sides, we obtain
 j 2n  2   j 2n 
 
S x  f   Rx      x n exp    xn exp
2
 
n    T0  n    T0 
2   j 2n  2  n


 x n exp    x n   f  
n     T0  n    T0 
This S x  f  gives us the power spectral density of the periodic signal. Power spectral density
means the distribution of power of the signal as a function of frequency.

The total power content of the periodic signal is obtained by integrating S x  f  with respect
to frequency . When this is done, the power becomes


x
2
Px  n
n  

This relation is known as Rayleigh’s relation.

 MODULE-II

AMPLITUDE MODULATION

Modulation of a baseband signal may be viewed as a low pass to band pass conversion. This
is usually accomplished by multiplication of the baseband signal with a periodic sinusoidal
waveform of a frequency higher known as the carrier than that of the baseband signal. The
baseband signal henceforth will be called the modulating signal. Multiplication of the
modulating signal with a sinusoidal carrier in the time domain results in a shifting of the
spectrum of the modulating signal in the frequency domain. Let the modulating signal be
denoted as mt  and the sinusoidal carrier be Ac sin  c t . Multiplication of the two in the time
domain generates a signal v AM t  expressed as

v AM t   mt Ac sin  c t

If the spectrum of mt  be M  f  , then the product signal v AM t  has a spectrum given as

V AM  f  
Ac
2j
  
M f  fc c  M f  f c 

where j is the complex number equal to  1 . The above expression is because of the fact
that the spectrum of a pure sinusoid sin  c t of frequency f c consists of two impulses
1
centered at  f c with amplitude . In a similar fashion, we note that multiplication of mt 
2j
with a carrier of the form Ac cos  c t gives us

v AM t   mt Ac cos  c t

The spectrum of this signal takes the form of

V AM  f  
Ac
2
  
M f  fc c  M f  f c 
We observe that, the process of multiplication of mt  with either Ac sin  c t or Ac cos  c t
has given rise to two new frequency components in the spectrum of the output signal. These
two frequencies f  f c and f c  f are called the upper side band (USB) and the lower side
band (LSB) respectively. The process of generation of these two side bands along with the
carrier is known as double side band with carrier (DSB plus C). The expression for DSB with
full carrier is

v DSB C t   mt   mt  Ac sin  c t  mt 1  Ac sin  c t 

As we may further observe, it can be generated by a multiplier and adder circuit. This is
illustrated in Fig. L9.1.

For example: L  3 , and m  7 The terms ( L  1)  2 . The pulse qt  kT  is present for the
instants from (2) to 0 . This lasts for, hence 3 symbol intervals. However, the shifted pulse
qt    kT  lasts from k  1 to k  m  L  7  3  4 th instant which has saturated to ½ as
the pulse at the 7th signalling interval may originate at this, may have the 6th pulse as its only
or the 5th pulse may be the pulse two intervals earlier. Hence, all the values of qt    kT 
would have saturated to ½ from -2 to 4th signalling interval whereas the original pulse

Adder
mt 1 Vc sin  c t 
x
mt , f m mt Vc sin  c t

Vc sin c t

Fig. L 9.1 Conceptual generation of DSB with full carrier type of AM signal

For a sinusoidal modulating signal, the instantaneous amplitude of the carrier becomes

Vc  Vm sin  m t as the modulating signal sits atop the amplitude of the carrier. As we are
interested in the instantaneous amplitude of the carrier as it should change in accordance with
the amplitude of the modulating signal, the overall modulated signal looks like

 V 
v DSB C t   Vm sin  m t  Vc sin  c t.Vm sin  m t  Vc 1  m sin  m t  sin  c t
 Vc 

We define the modulation index or the depth of modulation of this type of AM signal is
defined as

Vm
ma 
Vc

The ratio of the peak amplitudes of the carrier and the modulating signal and it has a
maximum value of unity. Usually, the value of ma  1 , in order for an envelope detector to
work at the receiver. If m a  1 , we understand it as 100% modulated signal and for a value of
ma  1 , we realize an overmodulated signal. For standard AM broadcast, the value of
modulation index is 30%. Depending on the amplitude level of the modulating signal, a
modulator may be a low level modulator or a high level modulator. A low level modulator
may be constructed by injecting the modulating signal either to the base or the emitter of a
transistor. Let us study such a modulator.
is zero.

 Vcc

Rc

R1
AM

From Output
Carrier
Cm Re
frequency Cc
Ce
R2

Modulating
signal

Fig.L 9.2 A BJT amplifier with emitter modulation circuit to generate DSB plus C

In Fig.L. 9.2, the dc bias condition is set up by the voltage divider R1 and R2 , the emitter
resistor Re , collector resistor Rc and the supply voltage Vcc . The ac voltage gain of the BJT
amplifier depends on its quiescent emitter current. As the modulating signal has been injected
into the emitter, the instantaneous emitter current becomes

i E  I E  K 1Vm cos  m t

where I E is the quiescent value of the emitter current and K 1 is a constant. Amplitude
modulation results if K1Vm is smaller than I E . As the voltage amplification is a function of
the total emitter current, we get

Av  K 2 i E  K 2 I E  K 1V m cos  m t 

where K 2 is another constant. The input to the amplifier is the carrier voltage coupled
through a transformer, the output voltage of this circuit is

V0  AvVc cos  c t  K 2 I E  K1Vm cos  m t  cos  c t

We can observe that, amplitude modulation has been achieved. The tuned circuit present at
the collector allows the two side bands to pass through and suppresses other harmonics from
appearing at the output. This constitutes a band pass filter with center frequency around the
carrier frequency with a pass band of 2 f m .
A low level modulation is also achieved by injecting the modulating signal to the base of the
transistor. The circuit for achieving this is illustrated in Fig. L 9.3.
From
carrier
frequency
Oscillator AM Output

Modulating
mt 
Voltage
Amplifier

 Vbb  Vcc

Fig. L 9. 3 A BJT amplifier with base modulation circuit to generate DSB plus C

Another circuit to accomplish DSB plus C generation is the switching modulator illustrated in
Fig.L 9. 4.
ct   Vc cos  c t

mt  v1 t 
RL v2 t 

Fig. L 9.4 A Switching Modulator

In this circuit, we assume that the carrier applied to the diode is larger than the modulating
signal in amplitude. It is further assumed that the diode is an ideal switch which implies that
for the forward bias condition corresponding to ct   0 , it shows zero resistance. The
transfer characteristic of the diode-load resistor may be modeled as piece wise linear. This
means

v t , ct   0
v 2 t    1
0, ct   0
where v1 t   mt   Vc cos  c t . We observe from the above that, the output voltage v2 t 
varies periodically between the voltage v1 t  and zero with a frequency of f c . The output
voltage, may alternatively be expressed as

v 2 t   mt   Vc cos  c t g t 

where g t  is viewed as a periodic pulse train with unity amplitude and a duty cycle of 50%,
1
the time period being equal to T0  . The Fourier series expansion of this pulse train gives
fc
us

1 2   1
n 1
g t     cos c t 2n  1
2  n1 2n  1

Substitution of this in the above expression gives rise to two components. The first term is
Vc  4 
1  mt  cos  c t is the desired DSB plus C component. The second term that
2  Vc 
contains all harmonics are filtered out by the use of a band pass filter with a center frequency
of f c with a bandwidth of 2 f m .

Square Law Modulator

A square law modulator is shown in Fig. L 9.5. This uses the nonlinear property of an active
device like a diode, BJT etc. The modulating signal is relatively weak. The output of the
device can be related to the input as
Nonlinear
Device
mt 
v1 t  v2 t  RL
Vc cos  c t

tuned to
fc

Fig.L 9.5 A square law modulator that employs a nonlinear device

v 2 t   a1v1 t   a 2 v12 t 

where a1 and a 2 are constants. The input voltage is expressed as

v1 t   mt   Vc cos  c t
Hence, the output voltage becomes

v 2 t   a1 mt   Vc cos  c t   a 2 mt   Vc cos  c t 

2

Expansion of the second term in the above gives us

v2 t   a1 m t   Vc cos  c t   a2 mt   Vc cos  c t 

2

 
 1 
a2 m2 t  2Vc mt cosct Vc2 cos2 ct  a2 m2 t  2Vcmt cosct  1 cos2ct
2
 
LECTURE-10

L 10.1 High Level Modulator

All the transmitters employing the previous circuits are known as low level modulators. This
is because the amplitude of the modulating signal is rather small that may come from a
microp phone or a typical video camera like the vidicon. Amplification of the modulated
signal takes place after these circuits. Hence such circuits are known as low level
transmitters. For the high level modulation, the modulating signal is amplified first before it
amplitude modulates the carrier. This is usually carried out in class-C power amplifiers. This
is because, as the modulating signal has been already amplified, it can not drive linear power
amplifiers. Such a high level modulator employing a class-C power amplifier is shown in Fig.
L 10.1.
 Vcc

Class B push-
mt  pull power
amplifier

AM
Output

Vc cos  c t

Fig.L 10.1 Class B modulator and class C power amplifier

The output of the nonlinear device (the NPN transistor here biased to near cut off by the
carrier which has been injected at its base) becomes
  1 
v 2 t   a1 mt   Vc cos  c t   a 2 m 2 t   2Vc mt  cos  c t  1  cos 2 c t  (L 10.1)
 2 

considering only upto the second term in the power series expression for the output of a
nonlinear power amplifier

We observe from the above that,

a1Vc cos  c t  2a 2 mt Vc cos  c t  Vc a1  2a 2 mt  cos  c t (L 10.2)

Is the desired DSB with carrier. To realize demodulation with simple, low cost demodulators
such as an envelope detector, we have to ensure that

2a 2
1
a1
The component with cos 2ct has been rejected by the tuned circuit (band pass filter with a
centre frequency equal to the carrier frequency) connected to the collector of the power
amplifier and hence does not appear at the output of the modulator.

L 10.2 POWER IN AN AM SIGNAL

A conventional (DSB with full carrier) AM signal expressed as

v AM t   Vc 1  ma cos  m t cos  c t (L 10.3)
corresponding to a modulating signal expressed as Vm cos  m t . This is otherwise known as
tone modulation.

From (L 10.3), it is observed that,

mV
v AM t   Vc cos  c t  a c cos c   m t  cos c   m t  (L 10.4)
2
V2
The first term in (L 10.4) gives us a power of c . This is because a periodic sine wave with
2
unit amplitude such as a carrier has an time-averaged power equal to ½ W.
m 2V 2
Both the second and the third terms give us equal powers of a c as these are also
8
sinusoidal waveforms. The total power, hence becomes in a DSB plus carrier type of AM
waveform,
V 2 m 2V 2 m 2V 2 V 2 m 2V 2 V 2  m 2 
Pt  c  a c  a c  c  a c  c 1  a  (L 10.5)
2 8 8 2 4 2  2 

Vc2
We observe that from (L 10.5), out of this total power, the carrier power is Pc  whereas
2
2
m V 
the modulating signal gives us a power of  a c  .
 2 
This carrier power represents a wastage of power as it does not convey any useful
information. If the modulation index has a value of 1, then the total transmitted power is 1.5
Vc2
. If we choose not to transmit the carrier power, then we actually transmit a power of 0.5
2
which accounts for a power saving of 66%. This is so as the carrier does not contain any
useful information about the modulating signal. If the modulating signal is any arbitrary
signal mt  , then its average power becomes m 2 t  and the total power in a DSB plus carrier
 
type of AM waveform becomes Pc 1  m 2 t  . Similarly, the power in a DSBSC type of AM
waveform is Pc m 2 t  . The SSBSC type of AM waveform will have a power content of
1 / 2 Pc m 2 t  .

LECTURE-11
Objective:

To learn DSB-SC modulation/demodulation techniques

 (a) Balanced Modulator Circuit

In a DSB-SC form of amplitude modulation, carrier is suppressed as it does not convey any
information. This carrier suppression is accomplished in a number of ways. We start with a
balanced modulator circuit. This is realized by BJT/FETs or devices possessing nonlinear
characteristics. Such a circuit is shown in Fig. L 11.1.

id1
vgs1
½ em

Modulating
DSBSC output
Signal input
Carrier
½ em input vgs2
id2

Fig. L 11.1 A balanced modulator circuit realized with two FETs

Any circuit that produces the product of two input waveforms (the modulating signal and the
carrier) is a balanced modulator. The FET is used here as it has a transfer characteristic which
is nonlinear, so that the output contains a term equal to the product of the input voltages,
besides other cross terms. The transfer characteristic of the FET is almost parabolic and may
be approximated as

i d  I 0  av gs  bv gs2 (11.1)
where I 0 is the current for zero gate-source voltage, and a, b are constants. Since the drain
currents i d 1 and i d 2 flow in the opposite directions in the primary winding of the output
transformer, the effective primary current i p is


i p  i d 1  i d 2  a v gs1  v gs 2   b v gs2 1  v gs2 2  (11.2)
 a v gs1  v gs 2   bv gs1  v gs 2 v gs1  v gs 2 
This becomes equal to, upon application of Kirchoff’s law to the input loops of Fig. L 11.1,

1 1
v gs1  em  ec and v gs 2  em  ec (11.3)
2 2

We obtain,

i p  ae m   b2ec e m  (11.4)

The RF output transformer rejects the low-frequency term like em passing only the product
2bec em which is the desired DSBSC signal. However, generation of DSBSC by this circuit
requires the two FETS matched completely with respect to I 0 , a and b . Otherwise, residual
components would appear at the output which obviously is not the desired modulated
waveform. These days the BMs are available in the integrated circuit (IC) form. Motorola’s
AN531 is one such IC.

 (b) Chopper Amplifier based modulator

Fig. L 11.2 (a)

 Rs

mt 
RL E0  mt cos  c t

E0

cos c t

Fig. L 11. 2 Chopper balanced modulator

In Fig. L 11.2, chopping of the signal is accomplished by the diode bridge at a rate equal to
the carrier frequency. The signal applied to the bridge is the message signal plus the dc bias.
All four diodes of the bridge conduct during the positive half cycle of the carrier thereby
giving no output voltage and none of them conduct during the negative half cycles of the
carrier alternately which makes the signal becoming available across the load resistance. The
carrier is prevented at the output by means of a tuned circuit.

mt  cos  c t R
mt 

cos  c t

Fig. L 11.3 Demodulation of the DSBSC signal produced by Fig. L 11. 2(a)

For demodulation of the DSBSC signal, we need to multiply mt  cos  c t by a synchronously
generated carrier cos  c t . The same circuits as those used for modulation can be used for
demodulation. However, the demodulating circuit differs from the modulator in that the
output of the demodulator should contain a low pass filter whereas the modulator has a
bandpass filter at its output. The low pass filtering is provided by the RC combination as
shown in the above figure. The demodulation may be accomplished by multiplying the
modulated signal by any periodic signal of frequency  c .If  t  is any periodic signal of
frequency  c , then it has a Fourier series  f  given as


 t       f  nf 
n c
n  

It is apparent that, if the modulated signal mt  cos  c t is multiplied by this periodic signal
 t  , the corresponding spectrum becomes


1
mt  cos  c t t   M  f  f c   M  f  f c     n  f  nf c 
2 n  

1 
  n M  f  n  1 f c   M  f  n  1 f c 

2 n 
From the above, it is observed that the resultant spectrum contains a term M  f  which can
be filtered out by a low pass filter.

Another form of the balanced modulator is shown in Fig. L 11.2 (b).

cos  c t
+
mt 

Fig. L 11. 4 Another realization of a balanced modulator

Modulation is achieved by using nonlinear devices. A semiconductor diode is a nonlinear

device. A nonlinear device such as a diode may be approximated by a power series like
i  av  bv 2

Transistors and vacuum tubes also exhibit similar relationships between the input and the
output under large signal conditions. To analyze this circuit, we consider the nonlinear circuit
element in series with the resistance R as a composite nonlinear element whose terminal
voltage v and the current i are related as above. The voltages v1 and v2 are given as
v1  cos  c t  mt  and v 2  cos  c t  mt 
The currents i1 and i2 are given as
i1  av1  bv12  acosc t  mt   bcosc t  mt  and
2

i 2  acos  c t  mt   bcos  c t  m t 

2

Hence the output voltage is given by

v0  i1 R  i 2 R  2 R2bmt  cos  c t  amt 

The signal amt  in this equation can be filtered put by using a bandpass filter tuned to  c at
the output terminals. Semiconductor diodes are conveniently used for the nonlinear circuit
elements in this circuit. All of the modulators discussed above generate a suppressed-carrier
amplitude modulated signal and are known as balanced modulators.

tuned to f c

mt  Kmt  cos  c t

Square wave at frequency f c

Fig, L 11.5. Ring Modulator that uses a centre tapped transformer at input as well as the
output

The diodes in Fig.L 11.5 form a ring as they all point in the same way. They are controlled
by a square wave ct  of frequency equal to carrier frequency f c which is applied in a
longitudinal manner by means of two centre-tapped transformers. Under the assumptions of a
perfect centre tap and identical diodes, there would be no leakage of modulation frequency
into the modulator output. Let us assume the diodes to be ideal. On the positive half cycle of
the square wave serving as the carrier, the top and bottom diodes become ‘on’ and the signal
mt  passes on to the output. Similarly, during the negative half cycles of the carrier, the
diagonal diodes become ‘on’ switching off the top and bottom diodes. Hence the message
signal passes on to the output, however with a negative polarity. Let us find out the kind of
modulated waveform at the secondary output of the output transformer.
The square wave has a Fourier series given as

 1n
 2n  1 cos2f c t 2n  1
4
ct  
 n 1
The ring modulator output is, therefore
st   mt ct 

 1n
 2n  1 cos2f c t 2n  1mt 
4

 n 1
There is no output from the modulator at the carrier frequency, that is the modulator output
consists entirely of modulation products. The ring modulator sometimes is referred to as the
double-balanced modulator because it is balanced with respect to the message signal as well
as the carrier.

Under the assumption of the message signal being bandlimited to  f m  f  f m , the

spectrum of the modulator output consists of sidebands around each of the odd harmonics of
the square wave carrier as shown in Fig. Here it has been assumed that f c  f m so that
sideband overlapping is avoided which arises when sidebands belonging to adjacent
harmonic frequencies f c and 3 f c overlap with each other. A bandpass filter with a centre
frequency f c and bandwidth 2 f m at the output would select the sidebands centered around
f c and reject all other components.

LECTURE-12 DEMODULATION OF AM SIGNALS

(a) Demodulation of DSB with full carrier type of modulated signals

Demodulation is the process of recovery of the original message signal embedded in the AM
wave. This is accomplished by the demodulator circuit in the receiver. The simplest
demodulator is a rectifier followed by a low pass filter which is called diode detector.

Vc  mt cos  c t C R

Vo t 

Fig. L 12.1 Envelope detector for conventional AM systems

This circuit is called so as it responds to the envelope of the incoming AM signal. On the
positive half cycle, the diode conducts and the capacitor C charges to the peak value of the
rectified voltage. As the incoming signal falls below this value the diode becomes non
conducting. This is due to the fact that the anode side voltage of the diode is less than the
cathode side voltage. Thus, the capacitor tends to hold the previously acquired peak value.
The capacitor discharges through the resistor at a slow rate. During the next positive half
cycle, the input signal becomes greater than the capacitor voltage and the diode starts
conducting again allowing the capacitor to charge up to the immediate peak value. The
capacitor discharges slowly during the off period of the diode which results in a small change
in its output voltage. During each positive half cycle, the capacitor charges to the peak value
of the incoming signal and holds this voltage until the next positive cycle. The time constant
RC of the output circuit is adjusted in such a manner that the exponential decay of the
capacitor voltage during the discharge period will follow the envelope approximately. The
output voltage now has a ripple component at  c which is filtered out by another low pass
filter.

The instantaneous AM signal is

v AM t   Vc 1  ma cos  m t 
At any time instant, t  t 0 , the slope of the envelope is given by

dv AM t 
 ma  mVc sin  m t 0
dt t t0
At that particular time, the envelope is given as
v AM t 0   Vc 1  ma cos  m t 0 

Let t 0 be the time instant when the capacitor C starts discharging. At any subsequent time t ,
the decayed capacitor voltage becomes

 t t 
 0 
vt   V t 0  exp  RC 

At t  t 0 , the rate of change of decay is

dvt  V t 0  V 1  ma cos  m t 0 
  c
d t  t 0  t t RC RC
0

If clipping of the negative peaks of the modulating signal is to be avoided, then at t  t 0 , the
slope of the decayed capacitor voltage must be equal to or less than that of the modulated
carrier. This is equivalent to saying that,
V 1  ma cos  m t 0 
 c  ma  mVc sin  m t 0
RC

1 m  sin  m t 0
or,  a m
RC 1  ma cos  m t 0
This gives us an upper limit for the circuit time constant as
 
 
1  1 
RC  .
m  m 
a m sin  t
m 0

 
 1  ma cos  m t 0 

ma m sin  m t 0
Making an maximization of RHS, the term is maximum when
1  ma cos  m t 0
cos  m t 0  ma which implies that sin  m t 0  1  ma2

Substitution of the values of the above yields

1 1  ma2
RC  .
m ma

The above equation indicates that, for 100% modulation, the product RC should be zero
which is not practical. In practice, it is found that for
 1
RC 
 m ma
the distortion in the diode demodulator output is not excessive. The highest frequency that
can be detected by this circuit is
1
 mHigh 
RCm a

(b) Demodulation of suppressed carrier type of modulated signals

 m 2 t  
mt  cos  c t Low pass filter K mt   
X  2 
with cut off f c

cos  c t

Fig. L 12.2 A synchronous/ coherent demodulator for DSBSC signals

(i) Costas Loop

This as shown in Fig. L 12.3 consists of two coherent detectors. A voltage controlled
oscillator initially adjusted to operate at the correct suppressed carrier frequency, f c ,
assumed to be known a priori, supplies the locally generated carrier to the two coherent
detectors-to one of them directly and to the other through a -900 phase shifter. The top
 
coherent detector receives the cos  c t   directly from the voltage controlled oscillator
 
(VCO). The bottom balanced modulator has a carrier of the form sin  c t   obtained by
feeding the VCO output through a 900 phase shifter. The incoming DSBSC signal
 
E c mt  cos  c t is fed as the other input to both of the balanced modulators. Suppose the
carrier phase error is zero which means the phase offset  between the incoming carrier and
1
the locally generated carrier is zero. Then the output of the I-channel is Ec mt  and that of
2
the Q-channel is zero. The I-channel output is taken as the demodulated signal. Now under a
practical situation, there exists a finite phase offset between the two carriers. Then, for such a
E
case the I-channel produces an output proportional to c mt  cos  while that of the Q-
2
E
channel is c mt  sin  . Both of these outputs have been shown to be fed to the phase
2
discriminator which consists of a multiplier followed by a low pass filter. For values of 
quite small, we have cos  1 and sin   0 . The low pass filter used in the phase
discriminator has a cut off frequency of the order of a few Hertz, gives a dc voltage
proportional to  at its output since variations in  will be very slow as compared to the
variations in m 2 t  . Thus we have a dc voltage that has the same polarity as  and is
proportional to it. This changes the frequency of oscillation of VCO in such a way so as to
lock it to f c , thereby keeping the phase offset within very small values.
 I-channel 1
E mt  cos Detector
Low pass 2 c output
Balanced
Modulator filter


cos  c t    Voltage Phase
DSBSC signal controlled discriminator
-900 phase oscillator
E c mt  cos  c t shifter


sin  c t    dc control
voltage
Balanced Low pass
Modulator filter 1
E mt  sin 
Q-channel 2 c

Fig. L 12.3 Costas loop for demodulation of DSBSC signals

The Costas loop provides a good practical solution to achieve phase synchronism common to
coherent detection. However, it suffers from one major disadvantage-the 1800 phase
 
ambiguity of the demodulated signal. Suppose in stead of receiving E c mt  cos  c t we have
 
 E c mt  cos  c t . The output of the multiplier used in the phase discriminator produces an
output proportional to E c2 m 2 t  , it is insensitive to the polarity of the incoming signal.
Under the locked conditions of the phase discriminator, we are not certain about the polarity
of the demodulated signal; whether it is mt  or  mt  . However, for demodulating audio
signals, this does not pose a serious problem as our ears are insensitive to polarity of the
demodulated signal. For video signals, a demodulated signal with negative polarity
reproduces an inverted picture in the receiver which is obviously very objectionable.
Similarly, for polar data also this phase ambiguity issue would damage the data as ‘1’
becomes ‘0’ and vice-versa. The phase control of the loop ceases for the condition of no
modulation present at the input. However, this is not a serious problem as the loop establishes
the lockup condition very fast.

(ii) Squaring loop

Another realization of a DSBSC demodulator is shown in Fig. L 12.4. This is
known as a squaring loop.
 DSBSC signal
Multiplier LPF
E c mt  cos  c t

Kmt 
Bandpass
Square law Frequency
filter centred Limiter
device divider  2
at f c

Fig. L 12.4 A squaring loop

Single Sideband Modulation

Hilbert Transform

The Hilbert transform a time function is obtained by shifting all frequency components by
900. It is, therefore represented by a linear system having a transfer function H  f  as shown
in the figure below

900
Hf 
1

-900
Fig. L 13.1 Transfer function for Hilbert transformer

We note that the phase function is odd. The positive frequency components get a -90 0 phase
shift whereas the negative frequencies undergo a 900 phase shift. The system function is
given as
H  f    j sgn f  corresponding to an impulse response of
1
ht  
t
The SSB signal may be generated by passing a DSBSC modulated signal through a band-pass
filter of transfer function H u  f  . Let us find out this H u  f  . We know that a DSBSC signal
is expressed as
s DSBSC t   E c mt  cos 2f c t

This is a bandpass signal containing only the in phase component. The low pass complex
envelope of the DSBSC modulated signal is given as
s DSBSC t   E c mt 
~

The SSB modulated signal is also a bandpass signal. However, unlike the DBSC modulated
su t 
signal, it has a quadrature as well as an inphase component. Let the low pass signal ~
denote the complex envelope of s u t  . Hence,
s t   Re~
u s t  exp j 2f t 
u c

We next proceed to find out the low pass complex equivalent ~ su t  . To do so, the bandpass
filter transfer function is replaced by a an equivalent low pass filter of transfer function
~
H u  f  as shown in Fig. From the Fig. we observe that
1
~  1  sgn  f , 0  f  f m
Hu  f   2
0 elsewhere
The DSBSC modulated signal is replaced by its complex envelope. The spectrum of this is
~
S DSBSC  f   E c M  f 

The desired complex envelope ~ su t  is determined by evaluating the inverse Fourier

~ ~
transform of the product of H u  f S DSBSC  f  . Thus,
~ ~ E
H u  f S DSBSC  f   c M  f 1  sgn  f 
2
Let us have a signal m̂t  such that mˆ t    j sgn  f M  f 
Thus,
E
su t   c mt   jmˆ t 
~
2
Accordingly, the mathematical expression for the SSB modulated wave is
E
su t   c mt  cos 2f c t  mˆ t  sin 2f c t 
2

This equation tells us that, except for a scaling factor, a modulated wave containing only an
upper sideband has an inphase component equal to the message signal mt  and a quadrature
component equal to m̂t  , the Hilbert transform of mt  .

From the foregoing we may note that, when the objective is to retain the lower sideband only,
the transfer function of the bandpass filter needs to be modified to
1
~  1  sgn  f ,  f m  f  0
Hl  f   2
0 elsewhere

Thus, the output of this bandpass filter in response to the complex envelope of the DSBSC
modulated signal becomes
~ ~ E
H l  f S DSBSC  f   c M  f 1  sgn f 
2
that gives us
 E
sl t   c mt   jmˆ t 
~
2
Accordingly, the mathematical expression for the SSB modulated wave is that contains the
lower sideband only is
E
sl t   c mt  cos 2f c t  mˆ t  sin 2f c t 
2

mt  cos  c t
Balanced
Modulator

cos  c t

mt  

Σ
2

 Balanced
 Modulator
2 mh t sin  c t
mh ;

Fig. L 13.2 Phase shift method of generation of SSB-SC signal

The SSB signal as generated by Fig. L 13.2 has a waveform expressed as

 SSB t   mt sin  c t  m h t  cos  c t

where m h t  is the signal obtained by shifting the phase of each frequency component of

mt  by .
2

(b) Weaver’s method of SSB-SC generation

 Low pass Balanced
Balanced
filter cut off Modulator
Modulator
frequency f 0
2 cos  0 t
900 Phase
0 -
90 Phase Shifter SSB-SC
2 sin  0 t +
sin  m t Shifter Output
∑
Audio AF Carrier
2 sin  0 t RF Carrier
Input Generator +
Generator
Low pass sin  c t
Balanced
filter cut off Balanced
Modulator
frequency f 0 Modulator

Fig. L 13.3 SSBSC generation by Weaver’s method

The Weaver’s method modifies the phasing method to rid of the design issues arising in
wideband phase shifters. It uses an audio frequency sub carrier at a frequency of f 0 . Let us
find out the expression for the summer output as shown in Fig.

The top left balanced modulator produces an output which is given as

2 sin  m t cos  0 t  sin  0   m t  sin  0   m t 
The low pass filter with a cut off frequency of f 0 rejects the higher frequency term given by
 0   m  . Therefore, the output of the top right balanced modulator is expressed as
1
sin 0   m t cosc t   sin c  0   m t  sin c  0   m t 
2
The bottom left balanced modulator produces an output which is given as
2 sin  m t sin  0 t  cos 0   m t  cos 0   m t 
The low pass filter with a cut off frequency of f 0 rejects the higher frequency term given by
 0   m  . Therefore, the output of the bottom right balanced modulator is expressed as
1
cos0   m t sin c t   sin  c   0   m t  sin  c   0   m t 
2
Hence, the output of the summing amplifier becomes
1
sin c   0   m t  sin  c   0   m t   1 sin c  0   m t  sin c  0   m t 
2 2
 sin  c   0   m t  sin  c   0   m t
Hence, the modulator generates the USB-SC corresponding to a carrier frequency of
 f c  f 0  or the LSB-SC corresponding to a carrier frequency of  f c  f 0  .
The Weaver’s method has certain advantages such as:

 No need for any sideband suppression filter

 No need for any wideband phase shifter
  As the phase shifters are designed for a single frequency, they are extremely simple
and cheap.
 No need for frequent adjustments
 Easy to change from USB-SC to LSB-SC and vice versa at the summing junction
output.
Demodulation of SSB signals
Can be accomplished by any synchronous/ coherent kind of demodulators discussed earlier.

Vestigial Sideband Modulation

This is a compromise between the bandwidth conserving feature of a typical SSBSC
modulation and demodulation simplicity of the conventional AM signals. This is widely used
for transmitting television (TV) signals occupying a spectrum in the VHF and UHF band of
frequencies. In this form of amplitude modulation, one sideband is fully transmitted while a
vestige or a part of the other sideband is transmitted. The carrier is also transmitted
completely to aid the process of demodulation or picture signal recovery at the receiver
through the use of simple envelope detectors.
Television signals

The exact details of modulation format used to transmit the video signal characterizing a TV
system are influenced by two factors:

a) The video signal exhibits a large bandwidth and significant low frequency content,
which rules out the possibility of using SSB. This is because SSB would require
extremely extensive filtering to separate the two sidebands. In the presence of significant
amount of low frequency contents which are necessary to reproduce the picture signal at
the receiver, it is very difficult to suppress one sideband completely as the two sidebands
are separated from each other by a small amount. Neither DSBSC is also useful as it
requires a double bandwidth. Hence VSB becomes a choice that entails the transmission
of one sideband completely and the other sideband being used partially.

b) The circuitry used for demodulation in the receiver should be simple and therefore cheap;
this suggests the use of envelope detection, which requires the addition of a carrier to the
VSB modulated wave.

With regard to point (a), it is to be noted that although there is indeed a basic desire to
conserve bandwidth, in commercial TV broadcasting the transmitted signal is not quite VSB
modulated. The reason is that at the transmitter power levels are high, with the result that it
would be expensive to rigidly control the filtering of sidebands. Instead, a VSB filter is
inserted in each receiver where the power levels are low. The overall performance is the same
as conventional vestigial sideband modulation except for some wasted power and bandwidth.
Sideband Shaping Filter in VSB

Let us replace the sideband shaping filter by an equivalent complex lowpass filter of transfer
~ ~
function H  f  as shown in Fig. The filter H  f  may be expressed as the difference between
~ ~
two components H u  f  and H a  f  as

~ ~ ~
H  f   Hu  f  Ha  f 

The two components are described individually as follows:

~
a) The transfer function H u  f  as shown in Fig. pertains to a complex low pass filter
equivalent to a bandpass filter designed to reject the lower sideband completely.
~
b) The transfer function H a  f  shown in Fig. accounts for both the generation of a
vestige of the lower sideband and the removal of a corresponding portion from the
upper sideband.

We may redefine the transfer function of the shaping filter as

~
1
H  f   2
 ~

 1  sgn  f   2 H a  f  ,  f a  f  f m
0 elsewhere

~
The signum function sgn  f  and the transfer function H a  f  are both odd functions of the
frequency f . Hence they both have purely imaginary inverse Fourier transforms.
Accordingly, we may introduce a new transfer function as

HQ  f  
1
j
 ~

sgn  f   2 H a  f 

that has a purely real transfer function. Let hQ t  denote the inverse Fourier transform of
H Q  f  ; that is

hQ t   H Q  f 

Thus, our equivalent low pass shaping filter, in terms of the new filter becomes

~
1
 
 1  jH Q  f  ,  f a  f  f m
H  f   2
0 elsewhere

The VSB modulated signal is now derived in time domain. To do so, we write

st   Re~
s t  exp j 2f c t  (C)
where ~ s t  is the complex envelope of st  . Since ~
s t  is the output of the complex low pass
~
filter of transfer function H  f  which is produced in response to the complex envelope of
the DSBSC modulated signal, we may express the spectrum of ~ s t  as
~ ~ ~
S  f   H  f S DSBSC  f 

We have the complex DSBSC signal defined as

~
S DSBSC  f   E c M  f 

Thus, the output of the equivalent shaping low pass filter is

~ E

~
S  f   c 1  jH Q  f  M  f 
2

Taking the inverse Fourier transform of the above we get

E

s t   c mt   jmQ t 
~
2

In the above, the quadrature component of the message signal mQ t  is defined as

mQ t   mt   hQ t 

Therefore, the VSB modulated signal becomes, from (C),

Ec E
st   mt  cos 2f c t   c mQ t sin 2f c t (D)
2 2
As we observe, this is the desired representation of the VSB modulated signal containing a
E
vestige of the lower sideband. The component c mt  is the in phase component of the
2
Ec
modulated signal and the component mQ t  is the quadrature component.
2

Modulating Product + Σ Modulated signal

Modulator
signal mt  mt 
E c cos 2f c t -

Oscillator

Filter

HQ  f  -900 phase shifter

E c sin 2f c t
m t  Product
Modulator
Fig. L 14.1 A method of generating VSB signal by a sideband shaping filter

The DSBSC and SSB signals may be considered to be two special cases of the VSB
modulated signal as defined in (D). If the vestigial sideband is increased to the width of a full
sideband, the resulting signal becomes a DSBSC wave with the result that mQ t  vanishes. If,
on the other hand, the width of the vestigial sideband is reduced to zero, the resulting signal
becomes an SSB signal containing the upper sideband, with the result that mQ t   mˆ t  ,
where m̂t  is the Hilbert transform of mt  .

Crystal oscillator

Video signal 38.9 MHz

38.9 MHz
from camera Video Diode
Delay VSB
processing bridge
equalizer filter
unit modulator

Power Linear Video up

To
Amplifier Amplifier converter
antenna
Picture carrier

+38.9 MHz
Crystal Frequency
Oscillator Multiplier

Fig. L 14.2 A portion of the TV transmitter to transmit picture signal only that uses low level
modulation

A TV transmitter showing the use of VSB for transmission of video signals is illustrated in
Fig. L 14.2.

Quadrature Carrier Multiplexed System

This refers to the transmission of two independent baseband signals using the same carrier.
The baseband signals DSBSC modulate a given carrier. As two independent message signals
are transmitted simultaneously on the same carrier offset from each other in phase by 90 0, it
is known as quadrature carrier multiplexed modulation system. Both of the modulating
signals require the same amount of bandwidth for transmission.
 m1 t  Balanced E c m1 t  cos  c t
Modulator

-900 Phase Carrier +

Shifter Signal
∑ Channel
E c cos  c t
m2 t  Balanced -
Modulator
E c m 2 t sin  c t

Fig. L 15.1 A transmitter that utilizes quadrature carrier multiplexing to transmit two
independent message signals

The output at the summing junction is expressed as

y t   Ec m1 t  cos  c t  m2 t sin  c t 

Both of the modulated signals are DSBSC signals that require synchronous detection at the
receiver. A block schematic of such a receiver is shown in Fig. L 15.2. In this diagram, we
have not shown explicitly how frequency and phase synchronism is achieved between the
transmitted and the regenerated carriers. However, this is also not important for our case now.

Balanced LPF cut off m1 t 

Modulator frequency f m1

-900 Phase Carrier

E c m1 t  cos  c t Shifter Signal
 E c m2 t sin  c t
cos  c t
Balanced LPF cut off m2 t 
Modulator
frequency f m 2

Fig. L 15.2 A receiver for detecting signals on quadrature carrier multiplexing

As we may observe quadrature carrier multiplexing reduces the requirement of number of

subcarriers besides reducing the bandwidth required to transmit the multiplexed signal. A
typical application of this scheme is used in color TV transmission wherein the color signals
are transmitted simultaneously on two independent carriers.
VESTIGIAL SIDEBAND MODULATION (VSB)
From previous lectures, it is to be noted that although there is indeed a basic desire to
conserve bandwidth, in commercial TV broadcasting the transmitted signal is not quite VSB
modulated. The reason is that at the transmitter power levels are high, with the result that it
would be expensive to rigidly control the filtering of sidebands. Instead, a VSB filter is
inserted in each receiver where the power levels are low. The overall performance is the same
as conventional vestigial sideband modulation except for some wasted power and bandwidth.

Filtering of Sidebands

Let the output from a product modulator be expressed as

u t   E c mt  cos 2f c t

Let the transfer function of the bandpass filter following the product modulator be H  f  .
Thus, the spectrum of the filtered modulated signal that appears at the output of the bandpass
filter becomes

S  f   U  f H  f 


Ec
2
M  f  f c   M  f  f c H  f 
In the above, M  f  denotes the spectrum of the message signal. The problem we address
here is to design a filter transfer function required to produce a modulated signal st  with
the desired spectral characteristics such that the original message signal may be recovered
from st  by coherent detection.

Coherent detection entails the multiplication of the incoming received signal with a locally
generated carrier E c' cos 2f c t that is synchronous with the transmitted carrier both in
frequency and phase (let us ignore for the time being how this exact synchronism is
achieved). Thus the receiver makes use of another product modulator whose output becomes
vt   st Ec' cos 2f c t
S  f   U  f H  f 


Ec
2
M  f  f c   M  f  f c H  f 

Ec
2
 
M f  fc H  f  
Ec
2

M f  fc H  f  
 S f  fc   Ec
2
  
M f  2 fc H f  fc   Ec
2
  
M f  fc  fc H f  fc 
 S f  fc   Ec
2
  
H f  fc M f  2 fc  M  f   
and

 S f  fc   Ec
2

M f  fc  fc H f  fc     Ec
2
  
M f  fc  fc H f  fc 

Ec
2
  
H f  fc M f  2 fc  M  f   
Hence, we have

E c'
Vf  
2
S  f  f c   S  f  f c 
Ec Ec'

4
 
M  f  H f  fc  H f  fc   
E c E c'

4
M  f  2 f c H  f  f c   M  f  2 f c H  f  f c 
The high frequency components of vt  represented by the second term are removed by a low
pass filter that follows the product modulator. Thus, the spectrum of the low pass filtered
signal output becomes

E c E c'
V0  f  
4
 
M  f  H f  fc  H f  fc   
For a distortionless reproduction of the original message signal at the coherent detector
output, we require V0  f  to be a scaled version of M  f  which further requires that

  
H f  f c  H f  f c  2 H f c = a constant   
We know that the message signal has a spectrum such that M  f  is zero outside the interval
of  f m  f  f m . Hence we need to satisfy the above equation for values of f in this
interval only. As a further simplification, we set
  
H fc 
1
2
so that

  
H f  fc  H f  fc  1  fm  f  fm  (A)

We note that st  is a bandpass signal. Hence, the canonical form of representation of it in
terms of its inphase s I t  and quadrature components sQ t  become

st   s I t  cos 2f c t  sQ t  sin 2f c t

We observe that, the spectrum of the inphase component is related to the modulated signal as

  
S f  f c  S f  fc ,  f m  f  f m
SI  f   

0 elsewhere

This becomes

  
SI  f   U f  fc H f  fc  U f  fc H f  fc     

Ec
2
 
M  f  H f  fc  H f  fc ,     fm  f  fm

Ec
 Mf 
2

From the above, we get

Ec
s I t   mt 
2

Now let us determine the quadrature component sQ t  . To do so, we first find out S Q  f 
which is expressed as


 j S f  fc  S f  fc ,
SQ  f   
    fm  f  fm
0 elsewhere

This takes the form of

  
SQ  f   j U f  f c H f  f c  U f  f c H f  f c     
 j
Ec
2
 
M  f  H f  fc  H f  fc ,     fm  f  fm

A close look at the above expression tells us that the quadrature component sQ t  can be
generated from the message signal by passing it through a filter having a transfer function
given as
    
H Q  f   j H f  fc  H f  fc ,  fm  f  fm 
Let m' t  denote the output of this filter in response to the message signal mt  . Thus, the
quadrature component of the modulated signal becomes

Ec
sQ t   m' t 
2

Combining the inphase and the quadrature components of the modulated signal, we obtain

Ec Ec
s t   mt  cos 2f c t  m' t  sin 2f c t (B)
2 2

Two important points are made at this point:

a) The inphase component s I t  is completely independent of the transfer function

H  f  of the bandpass filter involved in the generation of the modulated signal st  so
long as it satisfies (A).
b) The spectral modification attributed to the transfer function H  f  is confined solely
to the quadrature component sQ t  .

The role of the quadrature component is only to interfere with the inphase conmponent so
as to reduce or eliminate power in one of the sidebands of the modulated signal ,
depending upon the application of interest.

Modulating Product + Modulated signal

Σ
Modulator
signal mt  st 
E c cos 2f c t -

mt  Oscillator

Filter

HQ  f 
-90 phase shifter

E c sin 2f c t

m' t  Product
Modulator

Fig. L 16.1 A schematic for implementing a VSB signal

Envelope Detection of a VSB plus carrier

Commercial analog television broadcasting makes use of VSB plus a sizeable amount of
carrier to transmit video signal that occupies a bandwidth of typically 4-5 MHz. As it is a
broadcast type of service, hence it represents a point-to-multipoint communication.
Thousands of receivers need to be low cost which calls for envelope detection to be used in
order to recover the video signal. It is therefore of interest to determine the distortion
introduced because of envelope detector. The input to the envelope detector is

 1  1
xt   Ac 1  k a mt  cos 2f c t  k a Ac mQ t sin 2f c t
 2  2

Where k a is a constant that determines the percentage modulation. The output of the
envelope detector is expressed as
1
 1 2
 1  
2 2

at   Ac 1  k a mt    k a mQ t  

 2  2  

1
  1  
2 2

   k a mQ t    
 1   
= Ac 1  k a mt   1   2 
 2    1  1 k mt   
   2   
a


The rightmost term indicates the distortion contributed by the quadrature component mQ t  of
the VSB signal. The distortion can be reduced by (i) reducing the percentage modulation to
reduce k a and (ii) increasing the width of the vestigial sideband to reduce mQ t  . Both of
these methods are used practically. In commercial TV broadcasting, the vestigial sideband
occupies a width of about 1.25 MHz which amounts to about one-quarter of full sideband.
This has been determined empirically as the width of the vestigial sideband modulation
required to keep the distortion due to mQ t  within tolerable limits when the percentage
modulation is nearly 100.

SUPERHETERDYNE RECEIVERS
The modulated signals, we learned in previous lectures are typically detected in radio
receivers known as superheterodyne receivers. Edwin Armstrong invented the concept of
super heterodyne receiver in 1918. A receiver is designed to carry out the inverse operation of
a transmitter. Modulation is an important transmitter signal processing task that is decided by
a host of factors such as the baseband signal type, the channel conditions, the simplicity and
cost of a receiver and the type of application or service. Modulation of a carrier by a
baseband signal is essentially a low pass to bandpass conversion that is effected by signal
multiplication in time domain. Multiplication of a signal by a sinusoid shifts all frequencies
up and down by the frequency of the sinusoid. Because of this, station selection can be
accomplished by building a fixed bandpass filter and shifting the input frequencies so that the
station of interest falls in the passband of the filter. This is analogous to constructing a
viewing window on the frequency axis and instead of moving this window around to view a
particular portion of the axis, we keep the window stationary and shift the entire axis. This
shifting is called heterodyning and the resulting receiver is called a superheterodyne receiver.
A typical receiver is shown in Fig. L 17.1.

Ec  mt cos 2f c t Ec  mt cos 2f IF t

RF Amplifiers kmt 
with band pass Frequency IF Audio
Detector
filters tuned to converter Amplifier Amplifier
desired f c
Speaker

Local Oscillator
f LO  f c  f IF

Fig. L 17.1 A superheterodyne radio receiver broadcast signals

Heterodyning produces both an upward and downward shift in frequency. While one of these
shifts moves the desired station into the IF window (450 to 460 KHz), the other shift moves
another station into this same window. This undesired signal is called an image and needs to
be eliminated from the receiver.

A complete bandpass system consists of the transmission channel plus tuned amplifiers and
coupling devices connected at each end. Hence, the overall frequency response has a more
complicated shape than that of a single tuned amplifier. Various physical effects result in a
loose but significant connection between the system’s bandwidth and the carrier frequency f c
. The antennas in a radio system produce significant distortion unless the frequency range is
small compared to f c , moreover, design of a reasonably distortionless bandpass amplifier
turns out to be quite difficult if the bandwidth B is either very large or very small compared
B
to f c . As a rule of thumb, the fractional bandwidth should be kept within the range
fc

B
0.01   0.1
fc
The bandwidth of the system should be within 1% to 10% of the carrier frequency. Systems
designed this way are called narrowband systems. All the communication systems that we see
or work with fall into this category of narrowband systems unless otherwise mentioned.

As an example, let us listen to the Cuttack station operating at 972 KHz carrier frequency.
The local oscillator is set to 972+455 = 1327 KHz. Multiplication by this sinusoid places the
station at 972 KHz right into the IF filter passband. But the station operating at 1327+455 =
1782 KHz also multiplies the local oscillator frequency to produce a component at 455 KHz.
This image station would be heard right on top of the desired station. The separation between
the image and the desired station is twice the IF frequency or 910 KHz. A bandpass filter
with a passband of less than 1820 KHz would accomplish the separation. This filter must
pass the desired station, while rejecting the station 910 KHz away. This filter needs to be
tunable also. But it need not be a sharp bandpass filter. A single stage of tuned circuit is
adequate.

The antenna receives a signal that is a weighted sum of all broadcast signals. After some
filtering to be examined later, the incoming signal is amplified in an RF amplifier. The
resulting signal is shifted up and down in frequency by multiplying by a sinusoidal oscillator
called the local oscillator. The output of the heterodyner is applied to sharp bandpass filter
consisting of multiple filtering stages. This filtering is combined with amplification. The
fixed band pass filter is set at 455 KHz, called the intermediate frequency (IF) and has a
bandwidth of 10 KHz matching that of the station. In most receivers, the IF filter is made of
three tuned circuits that are aligned so as to generate a Butterworth filter characteristics. The
output of the IF amplifier represents a modulated signal with a fixed carrier frequency of 455
KHz with amplification and being separated out from the other signals.

Choice of the IF frequency

 Clearly, the IF frequency must not lie in the frequency band allotted for a given
communication application. For example, commercial AM uses a frequency band
from 535-1650 KHz. Thus, IF can not be taken to be any value inside this band.
 A very high value of IF would result in poor selectivity and poor adjacent-channel
rejection unless sharp cutoff filters are used.

The incoming radio signal is given the advantage of image frequency rejection by the RF
amplifier. All broadcast signals in the standard AM broadcast band (535-1605 kHz) are
translated to a fixed frequency of 455 kHZ by the IF amplifier. The IF amplifier decides most
of the gain and bandwidth of the radio receiver. This is in fact, an ingenious combination of
amplifier and a bandpass filter. This is the greatest advantage of a superheterodyne receiver.
After the signal is amplified, it is fed to an appropriate detector. This may be a noncoherent
detector like an envelope detector that detects DSB with carrier type of signals or it may be
any of the coherent detectors discussed previously. As we may observe, the original signal is
obtained at the output of this detector which is power amplified (usually push pull
configuration) and delivered to the speaker for reproducing the original speech or voice.
These receivers are usually equipped with automatic gain control (AGC) circuits that
maintains the output from the speaker at a constant level in spite of variations occurring in the
input signal. A part of the detector is tapped and given to IF amplifier input, mixer and the RF
amplifier in a negative feedback manner so that if the demodulators output increases due to
some reason, the IF amplifier is biased towards nearer the cutoff so that the gain reduces and
vice versa otherwise.

MODULE-III

ANGLE MODULATION

Frequency modulation results when the frequency of the sinusoidal carrier is varied in
accordance with the instantaneous changes in the amplitude of the modulating signal. The
instantaneous frequency of the modulated signal becomes

 i   c  k f mt 

This gives us the concept of instantaneous frequency; i.e. the frequency as a function of time
as it is frequency of the carrier that keeps on changing in accordance with the modulating
signal. If the modulating signal is analog, the frequency change is continuous. If the baseband
signal is digital, then the frequency changes in a digital manner with respect to time. For
example, if the modulating signal is a binary waveform that takes on only two amplitude
levels, then the carrier frequency also changes in two steps; one frequency corresponding to a
binary 0 and the other corresponding to a binary 1. These are also known as mark and space
frequencies. This scheme is called binary frequency shift keying (BFSK). If we consider an
m-ary waveform as the baseband signal, then the carrier frequency will also change in m-
steps giving rise to a M-ary frequency shift keying (MFSK).

In a similar manner, if we write,

 i    k p mt 

and substitute this as the phase of a sinusoidal carrier, then we obtain

v PM t   Vc sin  i   Vc sin   k p mt 

For a sinusoidal modulating signal, this takes on the form of

v PM t   Vc sin   k pV m sin  m t 

This is called phase modulation. Because the phase of the carrier is made to vary in
accordance with the instantaneous value of the modulating signal. We note that, in adding the
modulating signal to the phase, we have to take care of the dimensionality of the signal. This
is so as we can add a phase component to another phase. Thus, the proportionality constant
k p should be defined properly. For example, in case of a binary baseband signal, the phase is
expected to change in two phases. One phase corresponds to binary zero and the other phase
corresponding to binary one. Hence, we write,

v PM t   Vc sin   k p .0   Vc sin 

for a binary zero. This means that the carrier is transmitted as such without any change in its
amplitude, frequency and phase. However, for a binary one, we write,

v PM t   Vc sin    .1  Vc sin 

in keeping with the fact that we can add one phase to another. Hence k p assumes the value of
 . The carrier is inverted in phase by 180 in correspondence with a binary one. This is called
binary phase shift keying (BPSK). If a quaternary waveform is used as the modulating signal,
the carrier phase change assumes four distinct values and this is called quaternary phase shift
keying. In phase modulation, the instantaneous frequency is expressed as

di d
i   k p mt 
dt dt

which is proportional to the derivative of the modulating signal. For example, if the
modulating signal is a sinusoid, the instantaneous frequency is proportional to a cosinusoid.
A smooth time domain signal gives a continuous kind of instantaneous frequency. A square
waveform, a trapezoidal waveform will yield abrupt changes in the instantaneous frequency.
As we may note, the constant k p has the dimension of radians per volt. It is evident that, if we
differentiate the modulating signal in the time domain and then this is used to frequency
modulate a carrier, we obtain phase modulation. From the definition of a frequency
modulated signal, we define what is called frequency deviation as

   i   c  k f mt 

We thus find that, the instantaneous frequency deviation is directly proportional to the
strength of the modulating signal. A signal with larger amplitude produces more frequency
deviation and a signal with smaller amplitude gives rise to a lower frequency deviation. The
maximum frequency deviation is, hence

 max  k f mt  max

The maximum frequency deviation is directly proportional to the peak amplitude of the
signal. For a sinusoidal modulating signal, we have

 max  k f Vm

or

k f Vm
f max 
2
Thus,
t
 i    i dt    c  k f mt dt   c t   m d


as  is a dummy variable.

Hence the expression for the modulated signal takes the following form

 t

v FM t   Vc cos i t     Vc cos  c t  k f  m d   
 

For a sinusoidal signal Vm cos  m t , the frequency modulated signal looks like the following

 k f Vm 
v FM t   Vc cos i t     Vc cos  c t  sin  m t   
 m 

Here we define the modulation index for an FM signal as

f max k f mt  max

 
fm fm

For all practical purposes, we may consider  to be zero without any loss of generality.

From the expression of the FM signal, the modulation index  is

k f Vm

2f m

We note from the above that the modulation index for an FM signal is greater than unity
unlike that in the DSB plus carrier type of AM. The modulation index here depends upon the
peak amplitude of the modulating signal and its maximum frequency content. A standard
value of the maximum frequency deviation is 75 KHz which is used for the FM broadcast
systems. The FM broadcast systems operate in the 88-108 MHz. If we examine the
expression for the FM signal, we note that it contains a term like cosine of a sine. The
expansion of this term gives a cosine of another cosine and cosine of a sine. These are
captured by Bessel’s function. We write

v FM t   Vc cos c t   sin  m t   Vc cos  c t. cos sin  m t   sin  c t. sin  sin  m t 

We write

cos sin  m t   J 0    2 J 2  sin 2 m t  2 J 4   sin 4 m t  ...  2 J 2 n   sin 2n m t

sin  sin  m t   J 1  sin  m t   2 J 3  sin 3 m t  2 J 5  sin 5 m t  ...  2 J 2 n 1  sin 2n  1 m t

Substitution of the two expressions in the expression for the FM signal gives us

cos  c t J 0    2 J 2  sin 2 m t  2 J 4  sin 4 m t  ...  2 J 2 n  sin 2n m t  ... 

v FM t   Vc  
 sin  c t.2 J 1  sin  m t   2 J 3   sin 3 m t  2 J 5  sin 5 m t  ...  2 J 2 n 1  sin 2n  1 m t... 
The bracketed term is simplified as

2 sin  c t sin  m t  cos c   m t  cos c   m t

2 sin  c t sin n m t  cos c  n m t  cos c  n m t

2 cos  c t sin n m t  sin  c  n m t  sin  c  n m t

 J 0   cos  c t  J 1  cos c   m t  cos c   m t  J 2  sin  c  2 m t  sin  c  2 m t

v FM t   Vc  
 J 3  cos c  3 m t  cos c  3 m t  J 4  sin  c  4 m t  sin  c  4 m t  ... 

It is obvious from the above equation that, the FM signal contains a carrier term whose
amplitude is J 0  Vc , two sidebands at frequencies  c   m  with amplitude J 1  Vc ,
another pair of sidebands at frequencies  c  2 m  with amplitude J 2  Vc and so on. The
sidebands occur at frequencies  c  n m  with amplitude J n  Vc . It is apparent that the
spectrum of an FM signal extends up to infinity theoretically in both positive and negative
frequency axes. And we may be led to the belief that the bandwidth of an FM signal is
consequently infinite due to the presence of infinite number of sidebands. This is correct.
Then do we require an infinite amount of bandwidth to transmit an FM signal? The answer is
no. This is due to the fact that, although there are sidebands occurring at frequencies
 c  n m  , however, their amplitudes vary as J n  Vc which assumes smaller values as n
becomes higher. Hence, it is practical to consider a few values of n in the expression for the
FM signal in order to compute its bandwidth. We may note, in passing that the spectrum of a
phase modulated signal will look identical to that of the FM signal for the same value of
modulation index.

Reproducing the expression for the FM signal, we note that, when the modulation index is
very less, i.e.   1 , sin    and cos   1 , then we call this a narrowband FM signal
which has a simplified expression of

Vc cos c t   sin  m t   Vc cos  c t. cos sin  m t   sin  c t. sin  sin  m t 

and it becomes

Vc cos c t   sin  m t   Vc cos  c t  sin  c t. sin  m t 

We expand this as
   
v NBFM t   Vc cos  c t  cos c   m t  cos c   m t 
 2 

As we may note from the above expression, this looks similar to that obtained for a DSB+C
type of AM waveform, however with certain differences. The AM waveform under
consideration looks like

 m 
v AM t   Vc cos  c t  a cos c   m t  cos c   m t 
 2 

We observe from the above figure that, the resultant of the two sidebands in an AM
waveform lies parallel to the phasor and points in the same direction as the carrier. Thus, the
total amplitude at any instant is the sum of these two.

Multiple Tone Wideband FM (Non Linear Modulation)

We consider frequency modulation of a sinusoidal carrier when the modulating signal

contains multiple sinusoids. For example, let us make

mt   Vm1 cos  m1t  Vm 2 cos  m 2 t

This signal when frequency modulates a sinusoidal carrier gives


v FM t   Vc cos  c t  k f  mt dt 
Vm1 Vm 2
Hence,  mt dt  sin  m1t  sin  m 2 t
 m1 m2

The resulting FM signal becomes

  Vm1 V 
v FM t   Vc cos  c t  k f  sin  m1t  m 2 sin  m 2 t 
   m1 m2 

Let us denote

k f Vm1 k f Vm 2
1  , 2 
 m1 m2

The FM signal becomes in terms of the modulation indices,

v FM t   Vc cos c t  1 sin  m1t   2 sin  m 2 t 

Expansion of this equation gives us

v FM t   Vc cos  c t. cos1 sin  m1t   2 sin  m 2 t

This further results in

cos  c t.cos1 sin  m1t . cos 2 sin  m 2 t   sin 1 sin  m1t . sin  2 sin  m 2 t  
v FM t   Vc  
 sin  c t.cos1 sin  m1t . cos 2 sin  m 2 t   sin 1 sin  m1t . sin  2 sin  m 2 t 

Let us expand the first curly bracketed term further and see what we get
2n 2n
cos1 sin  m1t . cos 2 sin  m 2 t    J m 1 sin m m1t. J p  2 sin m m 2 t
m 0 p 0
2n 2n
  J m 1 J p  2 sin m m1t sin p m 2 t
m0 p0

2 l 1
Use of sin  sin  m t    J l   sin l m t  gives us
l 0

2 m 12 p 1
sin  1 sin  m1t . sin  2 sin  m 2 t     J  J  sinm t sin  p t 
m 1 p 2 m1 m2
m 0 p 0

Similarly, the second curly bracketed term gives us

 2 n 2 n 1 
v FM t   Vc   J m  1 J p  2  cos c t  m m1t  p m 2 t 
m 0 p 1 

The expression contains the following terms

a) a carrier frequency  c t with an amplitude J 0 1 J 0  2 

b) a set of sidebands corresponding to the first sinusoid  m1 with amplitudes J m  1 J p  2 

occurring at frequencies  c  m m1  for m  1,2,3,..

c) a set of sidebands corresponding to the first sinusoid  m 2 with amplitudes J m  1 J p  2 

occurring at frequencies  c  p m1  for p  1,2,3,..

d) a set of cross modulation terms  c  m m1  p m 2  with amplitudes J m  1 J p  2  for

m  1,2,3,.. and p  1,2,3,..
 mt  d Frequency Phase Modulated
dt Modulator Signal

Sinusoidal
carrier

A phase modulator through the use of differentiation of the message signal

Similarly, we note that,

mt  t Phase Frequency

 m d Modulated Signal
 Modulator

Sinusoidal
carrier
A frequency modulator through the use of integration of the message signal

These two figures illustrate the relationship between frequency modulation and phase
modulation. If we integrate a signal in the time domain and then use this signal to phase
modulate a sinusoidal carrier, we obtain a frequency modulated signal. Hence, this is called
angle modulation and as we may note, is a nonlinear modulation system.

FM as a nonlinear modulation system

This follows from that fact that any sinusoid is expressed as

ReVc exp jt 

For our case, the FM signal is

   t
 
v FM t   Vc Reexp  j   c t  k f  m  d  

     

We find that,

expx1  x2   exp x1   expx 2 

  t

where either x1 or x 2 is taken as j  c t  k f  m d  .
  

This is in contrast with the DSB plus carrier type of AM system which is a linear modulation
system.

POWER IN AN FM SIGNAL

We may find that, from the definition of the FM signal, the power in the FM signal is a
constant. This is because, the power of a signal depends on its amplitude. The amplitude of
2
an FM signal is a constant and hence the power is Vc Watts. For a sinusoidal carrier of
2
peak amplitude 1V, the power of the corresponding FM signal is 0.5 Watt. The information is
contained in the frequency changes and the amplitude does not change. Hence power required
to transmit an FM signal is a constant which is again in contrast with an AM signal. Any
change in the amplitude of an FM signal is due to channel imperfections or distortions which
are removed usually by means of a limiter in the receiver. It is due to this reason that an FM
signal sounds clearer than an AM signal for the same program. The sound transmission in TV
employs FM.

Comparison of AM and FM

Both FM and PM are called angle modulation and we have looked at circuits that can be used
to transmit information. However, FM is preferred for practical systems. This is due to the
fact that, in a PM waveform, information resides in the phase of the modulated carrier. In
order to retrieve information from it at the receiver, we have to have perfect knowledge about
phase. Maintaining a coherent phase for all possible values of phase in a PM waveform is an
arduous task. This is because the phase of the carrier continually changes in response to an
analog signal. Thus, system designers prefer to work with FM as extraction of the
instantaneous frequency can be performed by a host of circuits. The FM has the following
merits over AM.

1) The amplitude of the FM is a constant as information to be transmitted resides in the

instantaneous frequency of the carrier. This amplitude is independent of the
modulation index. Hence, the transmitted power in FM is a constant unlike AM where
the modulation index decides the amount of power being transmitted. This implies
that low level modulation may be performed in an FM transmitter facilitating the use
of power efficient class C amplifiers in order to make the transmitted signal suitable
for wireless transmission. All the transmitted power in FM is useful whereas in AM
most of goes as a waste.
2) FM receivers are typically fitted with limiters that take care of amplitude fluctuation
caused due to noise and thus, amplitude variations do not affect the quality of the
reproduced signal. We say that an FM system enjoys more noise immunity.
3) It is possible to further reduce the effect of noise by increasing frequency deviation.
AM does not have this feature as we cannot go for overmodulation in AM.
4) FM broadcast takes place in the upper VHF and UHF ranges which happen to be less
noisy than the MF and HF allotted to AM broadcast.
 5) FM broadcast takes place by space wave propagation limiting the radius of operation
to slightly more than line of sight. It is possible to operate several independent
transmitters on the same frequency with considerably less interference than would be
possible with AM.

ANGLE MODULATOR CIRCUITS

In a varactor diode modulator, the junction capacitance of a reverse biased diode changes
linearly with the modulating signal. The bias is varied by modulating voltage in series with a
voltage of  Vb . This change in the junction capacitance of the diode brings about a change in
the oscillating frequency of a suitable oscillator connected to this diode. This is the simplest
reactance modulator and is often used for automatic frequency control and remote tuning.

Basic reactance modulator using FET

An FET based reactance modulator is shown in Fig. Here, the drain to gate impedance is
assumed to be very large compared to the gate-to-source impedance. Three configurations of
an FET based modulator are realized and shown in the following diagrams.

To determine z , a voltage vds is applied between the drain and source. The resulting drain
current i d is computed as follows. The gate voltage is

v ds
v g  ib R  R
R  jX c

and the gate-to-source voltage is v gs  v g  v s  v g as the source is grounded. This causes a

drain current i d given as

gm R
i d  g m v gs  v ds
R  jX c

The drain-to-source impedance thus, is

v ds v ds R  jX c 1  jX 
z    1  c 
id g m Rv gs gmR gm  R 
R  jX c

As the reactance is much larger than the resistance, we will approximate the drain-to-source
impedance as

Xc
z j
gm R
This means that the reactance is capacitive and we may write the drain-to-source impedance
as

Xc 1 1
X eq   
g m R 2 f g m RC 2 f C eq

The output impedance of the FET under these conditions is purely capacitive and is given as
C eq  g m RC . Following observations are made from this equivalent capacitance.

 The expression g m RC has the dimension of capacitance

 The impedance z would have a resistive component if the gate-to-source resistance is
not small compared to the gate-to-drain impedance.
 The equivalent capacitance depends on the device transconductance and thus can be
varied with respect to a voltage.
 The capacitance can be adjusted to any value , within limits, through judicious
selection of R and C .

In practice, the gate-to-drain impedance is made five to ten times the gate-source impedance.
Let X c  nR at the carrier frequency. The, we have

1 1 1
Xc   nR and therefore C   . Substitution of this value of
C nR 2 f nR
capacitance into the equivalent capacitance obtained earlier, we get

gmR gm
C eq  g m RC  
2 f nR 2 f n

II. FET Reactance Modulator

We refer to Fig. where the places of R and C have been swapped and we further assume
that the resistance is much larger than the reactance; i.e. R  X c . Everything else remaining
the same, we write

1 v ds 1 v ds
v g  ib .  . 
jC 1 jC 1  jRC
R
jC

g m v ds
id  g m v g 
1  jRC

And therefore, the drain-to-source impedance is expressed as

v ds 1  jRC R 1  jCR
z     jC  
id gm gm  R  gm
 The expression for z shows that it is inductive, and the equivalent inductance is given as

RC
L eq  . The other two cases of FET reactance modulators are shown in the following
gm
figures.

Reactance Frequency Class C power

Oscillator
Modulator Multiplier amplifier

f c  f mt  Frequency 
n f c  f mt  
Multiplier
Down f c  n f mt 
converter
n

Local oscillator
n  1 f c

Audio in Reactance LC oscillator Frequency Class C power

+ Modulator amplifier
+ Multiplier
-

dc bias

Very LPF Crystal

Discriminator Mixer
filter oscillator

A wide band FM transmitter

FM Demodulators

Frequency demodulation is the process that enables us to recover the original modulating
signal from a frequency-modulated signal. The objective is to produce a transfer
characteristic that is the inverse of that of the frequency modulator, which can be realized
directly or indirectly. The requirement of a FM demodulator is to produce an output voltage
that varies linearly with frequency. A direct method uses frequency discriminator that
produces an instantaneous amplitude being proportional to the instantaneous frequency of the
input FM signal. The slope detector is a very basic form of FM demodulator, though its
linearity is not good.

The frequency discriminator is a combination of a slope circuit and an envelope detector. An

ideal slope detector has an imaginary transfer characteristic that varies linearly with
frequency inside a prescribed frequency interval.

Balanced Slope Detector

This is also called round Travis detector. It has two slope detectors, one tuned to a frequency
above the carrier frequency while the other is tuned to below-the carrier frequency. The
envelope detectors that follow the two slope detectors combine to produce a differential
voltage. The output from this detector is observed to have an S shape when plotted as a
function of frequency. When the incoming signal is unmodulated, the differential output
voltage is then the incoming signal is unmodulated, the differential output voltage is zero as
both the envelope detectors give identical outputs. When the carrier frequency is towards a
higher frequency, one arm produces more voltage than the other and hence a positive voltage
is obtained. On the other hand, when the carrier frequency deviates towards a lower
frequency, it is the other arm that will produce more voltage than the other and hence the net
differential output becomes negative.

Slope detector Envelope

detector
FM signal H1  f  +
Balanced signal
Σ
Slope detector Envelope -
detector
H2 f 

Fig. Block schematic of a frequency discriminator

A simple frequency demodulator can also be realized by the following circuit.

Foster-Seeley Discriminator

This is also known as the center tuned discriminator. This is a derived form of the balanced
slope detector and widely used in FM demodulators. Here both the primary and the secondary
are tuned to the carrier frequency. This greatly simplifies the alignment problem of the
balanced slope detector and yields better linearity. The voltage applied to each diode is the
sum of the primary voltage and the corresponding half-secondary voltage. The primary and
secondary voltages are:
 i) exactly 900 out of phase for an input carrier frequency of fc
ii) less than 900 out of phase an input carrier frequency higher than fc
iii) more than 900 out of phase an input carrier frequency lower than fc

This results in individual voltages being equal only for an incoming frequency equal to the
carrier frequency. At all other values of carrier frequency, the output from one diode is higher
than the other that depends on the deviation of the carrier frequency from its original value.
The output magnitude depends on the deviation of the input frequency from the carrier
frequency.

f0  fc
+ +

Kx t 

- -
f0  fc

Phase Discriminator (Foster Seely Discriminator)

In this circuit, the individual component voltages will be the same at the diode inputs at all
frequencies, the vector sums will differ with the phase difference between primary and
secondary windings. The result is that the individual output voltages are equal only at the
carrier frequency. At all other frequencies the output of one diode is greater than that of the
other. Which diode has the larger output depends entirely on whether the incoming frequency
is below or above the carrier frequency. It is noted that they are the same as in a balanced
slope detector. Accordingly, the overall output is positive or negative according to the input
frequency. As required, the magnitude of the output depends on the deviation of the input
frequency from the carrier frequency.
 C a’
D1 +
1 b
C3
R3
C2
C1 L2 L3
V12 - Va’b’
L1 -

M C4
a R4
2
+
b’
D2

Phase Discriminator circuit implementation for an FM demodulator

The resistances forming the load are made much larger than the capacitance reactances. The
circuit composed of C , L3 and C 4 is effectively placed across the primary winding. This is
shown in Fig.2. The voltage across L3 ,VL becomes

V12 Z L 3 jL3
VL  = V12
Zc  ZC4  Z L jL3  j 1 / C  1 / C 4 

L3 is an RF choke and is deliberately made large. Thus the inductive reactance greatly
exceeds those of C and C4 , especially since the first of these is a coupling capacitor and the
second is an RF bypass capacitor. All of these imply that

V L  V12 .Hence, it is proved that the voltage across the RF choke is equal to the applied
primary voltage. The mutually coupled, double-tuned circuit has high values of primary and
secondary Q and a low mutual inductance. We may, therefore neglect the reflected resistance
from the secondary and the primary resistance. The primary current is given as

V12
IP 
jL1

The voltage induced in the secondary due to this primary current is

V s   jMI p

with the sign depending on the direction of winding. Let us work with the negative voltage.
The secondary voltage becomes
 V12 M
Vs   jMI p   jM   V12
jL1 L

The voltage across the secondary winding, Vab , can now be calculated with the aid of Fig.3.
The secondary has been redrawn here. It follows from this figure that

Z c2 jX C 2  V12 M / L1  jM V12 X C 2
Vab  Vs  
Z c2  Z L 2  R2 R2  j  X L 2  X C 2  L1 R2  jX C 2

where X 2  X L 2  X C 2

and may be positive, negative or even zero, depending on the frequency. The total voltages
applied to the two diodes may be written as

1
Vao  Vac  V L  Vab  V12
2

1
Vbo  Vbc  V L  Vac  V L   Vab  V12
2

The voltage applied to each diode is the sum of the primary voltage and the corresponding
half-secondary voltage. The dc output voltages cannot be calculated exactly because the
diode drop is not known. However, it is known that each is proportional to the peak value of
the RF voltage applied to the respective diode. Hence,

Va 'b '  Va 'o  Vb 'o  Vao  Vbo

Let us consider the case when the input frequency f in is instantaneously equal to f c . For this
condition, X 2 is zero and the voltage becomes

jM V12 X C 2 V12 X C 2 M90 0

Vab  
L1 R2 R2 L1

From the above equation, we note that, the secondary voltage Vab leads the applied primary
1 1
voltage by 900. Thus, Vab leads V12 by 900, and  Vab lags V12 by 900. Now we add up
2 2
these two diode voltages vectorially. This is shown in the following figures. It is observed
that, since Vao  Vbo , the discriminator output is zero. For any incoming frequency other than
the carrier frequency, there is a net output voltage. Let us consider the case when f in is less
than f c . Hence, X L 2 is less than X C 2 so that X 2 is negative. Hence, the output voltage
becomes

jM V12 X C 2 V X M90 0 V12 X C 2 M

90   
0
Vab   12 C 2 
L1 R2  X 2 L1 Z 2    0 L1 Z 2
 1
From this, it is observed that, Vab lags behind V12 by more than 900 so that  Vab must lead
2
0
V12 by 90 . It is apparent from the vector diagram that Vao is less than Vbo . Thus, the
discriminator output is negative when f in is less than f c . Similarly, when the incoming
frequency is greater than the carrier frequency, X 2 is positive, and the angle of the
impedance Z 2 is also positive. Thus, Vab lags V12 by less than 900 . This time Vao is greater
than Vbo and the output voltage Va 'b ' is positive.

If the frequency response is plotted for the phase discriminator, it follows the required S

shape as shown in Fig.5. As the input frequency moves farther and farther away from the
center frequency, the difference between the two diode input voltages becomes greater and
greater. The output of the discriminator will increase up to the limits of the useful range, as
shown in this figure. The limits correspond roughly to the half-power points of the
discriminator tuned transformer. Beyond these points, the diode input voltages are reduced
because of the frequency response of the transformer, so that the overall output falls.

The phase discriminator is much easier to align than the balanced slope detector. There are
only tuned circuits, and both are tuned to the same frequency. Linearity is also better, because
the circuit depends less on frequency response and more on the primary-secondary phase
relation, which is quite linear. The only less noticeable disadvantage of this circuit is that it
does not provide any amplitude limiting.

Ratio Detector

In the Foster-Seeley discriminator, changes in the magnitude of the input signal will give rise
to amplitude changes in the resulting output voltage. This makes prior limiting necessary. A
ratio detector addresses this problem by incorporating an amplitude limiter into the Foster-
Seeley discriminator circuit.

A close look at the above FM detector reveals that the sum Vao  Vbo is a constant, although
the difference keeps on changing with respect to the change in the incoming frequency.
Deviation from this ideal does not result in undue distortion in the ratio detector. It follows
that any variations in the magnitude of this sum voltage is considered undesirable. This needs
to be suppressed. A discriminator that provides this suppression remains unaffected by the
amplitude of the incoming signal. The ratio detector is obtained from the Foster-Seeley
discriminator by i) reversing one diode, ii) placing a large capacitor C 5 across the output and
iii) taking the output from elsewhere.
 C a’
D1 +
1 b
C3
R3 R5
C2
L3
V12 C1 L2 0 - Vo
0’ C5
L1 +

a R4 C4 R6
2
-
b’
D2

A ratio detector circuit

Reversing of the diode D2 makes o positive with respect to b' , so that Va 'b ' is now a sum
voltage, rather than a difference voltage. Hence it becomes possible to connect a large
capacitor between a' and b' in order to keep this voltage a constant. With the connection of
this capacitor C 5 , Va 'b ' does not represent the output voltage, rather the output voltage is taken
between o and o' . It is now necessary to ground one of these two points, and o appears to be
more convenient. In practice, R5  R6 , and hence the output voltage Vo is calculated as

Va 'b ' V  V b 'o V  V b 'o

Vo  Vb 'o '  Vb 'o   Vb 'o  a 'o  Vb 'o  a 'o
2 2 2

This equation shows that the ratio detector output voltage is equal to half the difference
between the output voltages from individual diodes. Hence the output voltage is proportional
to the difference between the individual output voltages. The ratio detector therefore behaves
identically to the discriminator for input frequency changes. The S curve applies equally to
both the circuits.

The slope detectors-single or balanced- are not used in practice. They have been included
here to gain an understanding of frequency-to-voltage conversion and help in building
practical FM demodulators. The Foster-Seeley discriminator is very widely used in both
narrowband and wideband FM radio receivers. It is also used in satellite station receivers,
especially for the reception of TV carriers. The ratio detector is a good FM demodulator
typically used in TV receivers for recovering frequency modulated audio signal. Its
advantage over the discriminator is that it provides both limiting and a voltage suitable for
AGC, while the main advantage of the discriminator is that it is very linear. Thus, the
discriminator is preferred in situations in which linearity is an important characteristic (high-
quality FM receivers), whereas the ratio detector is preferred in which linearity is not critical,
but component and price savings. Under critical noise conditions as encountered in receiving
satellite signals, the phase-locked loop is typically used.

Limiting of FM Waves

When an FM wave is transmitted through a communications channel, in general, the output is

not expected to have a constant amplitude because of channel imperfections. At the receiver,
it is essential to remove the amplitude fluctuations in the channel output prior to frequency
demodulation. This is customarily done by means of an amplitude limiter. The transfer
characteristic of an ideal hard limiter is shown in Fig.

We assume the limiter to be a memoryless device in order to analyze the operation of this
circuit. The limiter output, in general can be expressed as

  1 if x c t   0
vt   sgnxc t   
 1 if x c t   0

We also assume the amplitude fluctuations to be slow compared to the zero-crossing rate of
the FM wave xc t  . The sign changes of xc t  may be considered to be proportional to the
carrier phase shifts as given by

vt   sgncos t 

where  t  is the phase component of the carrier containing the message signal. The function
sgncos t  a function of  , is a periodic square wave when the modulation is zero. The
Fourier series representation of this function gives us

cos2n  1 
 n
4
sgncos      1
 n 1 2n  1
Use of  t  in place of  in the above expression gives us

cos2n  12f c t   t 

 n
4
vt      1
 n 1 2n  1
From the above, we note that hard limiting the FM signal produces image sidebands at odd
harmonics of the carrier frequency f c . For a very large carrier frequency, a bandpass filter
centered at f c selects the desired FM signal and rejects the higher order terms. The bandpass
filter output, therefore becomes

4
vt   cos2f c t   t 

In practice, the combination of the hard limiter and band-pass filter is implemented as a
single circuit commonly referred to as band-pass limiter.

A phase discriminator makes use of the following:

1
cost  cost     cos2t     cos 
2

Use of a low pass filter with a cut off frequency of  rad/s will eliminate the double
frequency term and the output would be proportional to cos .

The composite signal at the IF filter output, is given as

r t   vt   nt  where nt  is a band-limited version of the white noise wt  . In particular,
nt  is the sample function of a noise process N t  with the following power spectral
density:

PM Demodulators:

Phase modulators are the same as frequency modulators except that the signal is
differentiated first and then fed to the VCO. We may approximate the process of
differentiation by the following

dx
xt   xt  t 0   t 0
dt

+ Envelope
Σ
- detector

Delay t0

Fig. A phase demodulator

This leads to a demodulator as shown next. This is a phase demodulator as a time shift is
equivalent to a phase shift. Any system that has a transfer function magnitude that is
approximately linear with frequency in the range of frequencies of the FM wave changes FM
into AM. Even a sloppy band pass filter will work as a discriminator if we operate over a
limited range relative to the filter bandwidth, The linearity of a band pass filter discriminator
can be improved by adopting the principles of a balanced modulator. The characteristic is
subtracted from a shifted version of itself. The difference between the outputs of the two band
pass filters with separate center frequencies is considered.
Let us assume that the modulated signal at the input to this circuit is

xc t   Ac cos2f c t   sin 2f m t 

 Ac cos2f c t . cos sin 2f m t   Ac sin 2f c t . sin sin 2f m t 

When the modulation index is less than unity and the delay produced is sufficiently small, we
may approximate

cos2f m t 0   1 and sin 2f m t 0   2f m t 0

Under these conditions, the modulated signal is approximated as

xc t   Ac cos2f c t . cos 2f m t 0   Ac sin 2f c t . sin 2f m t 0 

 Ac cos2f c t   Ac sin 2f c t . 2f m t 0 

The delayed signal is

xc t  t 0   Ac cos2f c t  t 0    sin 2f m t  t 0 

The IF used for TV is 40 MHz.

 MODULE-IV
Sampling of Analog Signals

4.1 Sampling of a band limited analog signal is of three types:

a) Instantaneous sampling
b) Flat top sampling and
c) Natural sampling

Instantaneous sampling is achieved by multiplying a band limited signal by a periodic

impulse train. Let the periodic impulse train with period Ts be represented as


p t     t  nT  s
n  

The spectrum of this is


P f   f s    f  kf  s
k  

Multiplication of the message signal mt  and this periodic impulse train is identical to
convolution in the frequency domain. Hence, we write
 
M samp  f   M  f   P  f   M  f  f s    f  kf   f  M  f  kf 
s s s
k   k  

We observe that, after the multiplication, the resultant sampled signal becomes periodic with
the amplitude of the impulse varying in proportion to the amplitude of the baseband signal.
The spectrum is a line spectrum in the sense that the individual spectra are centered at
integral multiples of the sampling frequency with a bandwidth equal to twice that of the
original baseband signal. Hence, in this regard, an individual spectrum may be viewed as
being equivalent to a DSBSC signal spectrum. Before we proceed further to understand the
nuances of sampling, let us review a few concepts from fundamentals that we learned in
module-I of this course.

 0 m  n

Q 4.1: Show that  sin c2Bt  m  sin c2Bt  n dt   1
  2 B m  n

Soln: The function can be written as

  k  1  f   jkf
sin c2Bt  k   sin c 2B t    rect  e
2B

  2 B  2B  2B 

The above integral is expressed as

  j m  n  f
 2 B
1  f   jmf 1  f   jmf  1  f  
 2B rect 2B e df    exp
2B 2B
rect e rect  df
2B  2B   2B  2 B  B
2B 

This result follows from the identity as shown below

  

 g t g t dt   G  f G  f df   G  f G  f df


1 2

1 2

1 2


1 m  n
We know that,

 exp j n  m tdt  0 mn

Thus, the integral in our problem assumes a value equal to

 j m  n  f
B B

 Bexp 2 B df   df  2 B
B

Therefore,

 j m  n  f
2 B
 1  f   1 1
 2 B rect  2 B  
  
 exp
B
2B df  2 B 2 .2 B  2 B

mn

and zero for the case when m  n

4.2 Reconstruction of the signal from its samples

A close look at the spectrum of the sampled signal reveals that it contains the original signal
spectrum alongwith the other spectra centered at  nf s . Hence to recover the original signal it
suffices to pass the sampled spectrum through an ideal low pass filter or brick wall filter
having the following frequency domain characteristics:

2 f m f  f m
Hf  
0 otherwise

The output of this filter is

Mˆ  f   M samp  f H  f 

We observe from the above that, the rectangular filter passes only the baseband component
having a maximum frequency content of f m Hz. Other spectral components are discarded at
the output of the filter. Let us see the effect of rectangular filtering in the time domain.

sin 20t
Q 4.2: Assume that a bandlimited function, st   is sampled at 19 samples per
t
second. The sampling function is a unit height pulse train with pulse widths of 1 msec. The
sampled waveform forms the input to a low pass filter with cutoff frequency 10Hz. Find the
output of the low pass filter and compare this with the original signal st  .

Soln: We only need to know the first two coefficients in the Fourier series expansion of the
pulse train. These are given by

0.001
a0   0.019
1 / 19
510 4

 cos 2  19tdt   sin 19  10 t 

2 3
a1  38
4
510

The output time function is the inverse Fourier transform of S 0  f  , and is given by

0.019 sin 20t sin t

s 0 t    0.038 cos19t
t t
The second term represents the aliasing error.

Q 4.3: A 100Hz pulse train forms the input to the RC filter. The output of the filter is
sampled at 700 samples per second. Find the aliasing error.

Soln: The square wave can be expanded in a Fourier series to yield,

1 2 2 2
vin t    cos 2  100t  cos 2  300t  cos 2  500t...
2  3 5

1 n 3 2
    1 2 cos 2n  100t
2 n 1, n odd n

The filter transfer function is given by

1 1
Hf  
1  j 2fRC 1  j 2f 0.00167 

The output of the filter is found by modifying each term in the input Fourier series. The
amplitude is multiplied by the transfer function magnitude and the phase is shifted by the
transfer function phase. The result is

1
v0 t  
2
  
 0.45 cos 2  100t  450  0.067 cos 2  300t  71.6 0 
   
 0.025 cos 2  500t  78.7 0  0.013 cos 2  700t  81.9 0

Let us assume impulse sampling. The result is that the component at 500Hz appears at 200Hz
in the reconstructed waveform, and the component at 700Hz appears at dc (zero frequency).
We shall ignore the higher harmonics. The reconstructed waveform is therefore given by
 1
v0 t  
2
  
 0.45 cos 2  100t  450  0.067 cos 2  300t  71.6 0 
  
 0.025 cos 2  200t  78.7 0  0.013 cos  81.9 0 
The last two terms represent the aliasing error.

Q 4.4: The function st   cos 2t is sampled every ¾ second. Evaluate the aliasing error.

Soln: The impulse train of period Ts , each narrow impulse being of width dt has a Fourier
series expansion as

dt 2dt  t t 
S t     cos 2  cos 2  2  ...
Ts Ts  Ts Ts 

3
The sampling period is Ts 
4

0.001 0.004
Hence, a0    0.00133 assuming dt to be of 1 msec duration.
3/ 4 3

0.002 4 0.008 8t

a1  cos 2  t  cos  0.00266 cos2.66t 
3/ 4 3 3 3

The aliasing error is due to 0.00266 cos2.66t  .

Q 4.5: A signal mt  is band-limited to B Hz is sampled by a periodic pulse train pTS t 

made up of a rectangular pulse of width 1 8B seconds (centered at the origin) repeating at the
Nyquist rate (2B pulses per second). Show that the sampled signal mt  is given by

1 
2  k 
mt   mt    sin  mt  cos k S t  S  4B
4 k 1 k  4 

1
Soln. The period of the periodic pulse train is T0  as the pulses repeat at the rate of 2 B
2B
pulses per second. The fundamental frequency is, therefore f 0  f S  2 B . The Fourier series
of the periodic rectangular pulse train is written by computing the Fourier coefficients
T0 2 1 16 B
1 1 1
a0 
T0  mt dt  2B  dt  2B. 8B  4
T0 2 1 16 B

The coefficient a n is computed as

 T0 2 1 16 B 1 16 B
2 2nt 2nt T 2nt
an   mt  cos dt  4 B  cos dt  4 B. 0 sin
T0 T0 2
T0 1 16 B
T0 2n T0 1 16 B

1  2n1 16 B   2n1 16 B    1  2n1 16 B  

 4 B. sin  sin      .2 sin 
2 B.2n  T0   T0  
 n  T0 

We note that

 2n1 16B    n   n   n 
sin    sin   sin   sin 
 T0   8BT0   8B.1 2 B   4 

Therefore,

 2n1 16B   2 n 
  sin
1
an  .2 sin 
n  T0  n  4 

Thus, the periodic rectangular pulse train is expressed as

1  n 1  2 n
mt     sin cos n 0 t    sin cos n S t
4 n 1 4 4 n 1 n 4

The sampled signal is obtained by simply multiplying the message signal by the periodic
rectangular pulse train. We write,

1  2 n 
mt   mt . pTS t   mt .   sin cos n S t 
 4 n1 n 4 
1 
2 n
 mt    sin mt  cos n S t
4 n 1 n 4

mˆ t   m samp t   ht 

  mkT  t  kT 
k  
s s


 mt    t  kTs   ht 
k  

2 fm
 mt    t  kTs   sin c2 f m t 
k   2 fm

  mt sin c2 f t  kT 
k  
m s

We observe from the above that the reconstructed signal m̂t  is obtained by the
superposition of sinc(x) pulses.

We observe from the above the following:

  The sampled signal contains a component proportional to the message signal mt  .
n
 The signal has an envelope proportional to sin c 
4
 The message signal can be recovered from the sampled signal by passing it through a
low pass filter of cut off frequency B Hz and gain 4.

The commutator approach toward multiplexing requires that the sampling rate of the various
channels be identical. If signals with different sampling rates must be multiplexed, there are
two general approaches that can be taken. One uses a buffer to store sample values and then
intersperse these and spit them out at a fixed rate. The buffer approach is also effective if
sampling rates contain variation (jitter). This is known as asynchronous multiplexing. The
system must be designed so that the buffer always has samples to send when requested by the
channel. This might require inserting stuffing samples if the buffer gets empty. Alternately,
the buffer must be large enough so that it does not overflow with input samples.

The buffer approach is also used if the various sources are transmitting asynchronously. That
is, suppose that they are not always transmitting information. The sizing of the buffer requires
a probability analysis and the resulting multiplexer is known as a statistical multiplexer. The
statistical multiplexer represents an efficient technique for multiplexing channels since a
source only has a time slot when it needs it. On the negative side, since individual source
messages are not occurring at a regular rate, the message must be tagged with a user ID. If the
channels are synchronous with the samples occurring at a regular and continuous rate, the
statistical multiplexer approach is not the best approach.

The second general technique involves sub-and super-commutation. This requires that all
sampling rates be multiples of some basic rate. Meeting these requirements might require
sampling some of the channels at a rate higher than what you would use without
multiplexing. For example, if we have two channels with required sampling rates of 8KHz
and 15.5KHz, in order to effect that combination we might choose to sample the higher
frequency channel at 16KHz.

The concept of sub-and super-commutation is quite simple, and we illustrate it with an

example. Let us suppose that we have a commutator wheel with 32 slots. Suppose we wish to
multiplex the following 44 channels:

1 channel sampled at 80KHz 1 channel sampled at 40KHz

18 channels sampled at 10KHz 8 channels sampled at 1250Hz

16 channels sampled at 625 Hz

We note that all of the sampling rates are multiples of 625Hz. Let us choose the basic rate of
the commutator wheel to be 10,000 rotations per second. Therefore, each of the 18 channels
which must be sampled at 10KHz get one slot on the wheel. The channel that must be
sampled at 40KHz needs four equally-spaced slots on the wheel, so it is sampled four times
during each 0.1msec rotation of the wheel. Similarly, the 80KHz channel needs eight equally-
spaced slots on the wheel. These higher rates are multiplexed using supercommutation.

The channels that need to be sampled at less than 10KHz only must be sampled on selected
rotations of the wheel. For example, a 1250 channel needs to be sampled once every eight
rotations of the wheel while a 625 Hz channel needs to be sampled only once every 16
rotations. We accomplish this using subcommutation wheels. The eight 1250Hz channels are
commutated together with a wheel rotating at a rate of 1250 rotations per second. Each 0.1
msec, one of the channels is connected to a cell on the main commutator wheel. Similarly, the
sixteen 625 Hz channels are commutated with a wheel rotating at 625 rotations per second.

Binary 1’s and 0’s such as in PCM signaling may be represented in various serial-bit
signaling formats called line codes. Some of the widely used line codes are shown in the
following figure. There are two major categories of the line codes: return-to-zero (RZ) and
non-return-to zero (NRZ). With RZ coding, the waveform returns to a zero-volt level for a
fraction (usually half) of the bit interval. Before discussing more about the line codes, let us
touch upon some of the desirable aspects of a line code.

a) Self-synchronization- There is enough timing built into the code so that bit
synchronizers can be designed to extract the timing or the clock signal. A
long series of 1’s and 0’s should not cause a problem in timing recovery
required at the receiver in order to establish the operating clock.
b) Low probability of bit error: Receivers can be designed that will recover
the binary data with a low probability of bit error when the input data signal
is corrupted by the noise or ISI.
c) Spectrum matching to the channel If the channel is ac coupled, the PSD of
the line code should contain insignificant portions at frequencies near zero.
In addition the signal bandwidth need to be sufficiently small compared to
the channel bandwidth so that ISI will not be a serious issue.
d) Transmission bandwidth: It should be as small as possible.
e) Error detection capability: It should be possible to implement this feature
easily by the addition of channel encoders and decoders, or it should be
incorporated into the line code.
f) Transparency-The data protocol and line code are designed so that every
possible sequence of data is faithfully and transparently received.
A quarternary signal may be formed by grouping the message bits in blocks of two and using
four amplitude levels to represent the four possible combinations 00,01,10 and 11. Thus,
r
T  2Tb and r  b . Different assignment rules or codes may relate bk to the grouped
2
message bits. We show two such codes in Table 1.
 Table No.1 Two codes for the line codes

bk Natural code Gray code

3A 11 10
2

A 10 11
2

A 01 01
2

3A 00 00

2

The Gray code has advantages relative to noise-induced errors because only one bit changes
from going from level to level. Quaternary coding generalizes to M  ary coding in which
blocks of n message bits are represented by an M -level waveform with

M  2n

Such a pulse corresponds to n  log 2 M bits. The M-ary signaling rate is decreased to

rb
r
log 2 M

We note that the use of M  ary coding reduces the requirement of transmission bandwidth
by log 2 M as compared to binary transmission. However, increased signal power is required
to maintain the same spacing between the amplitude levels. For an M  ary signaling format,
the power associated with the signal is

 2i  1   A   3 A   5 A   2i  1 
2 2 2 2 2 2 2 2
 A   3A   5 A 
bk2          ..            ..  
2  2   2   2   2  2   2   2 

This can be expressed as

2

2i  1  A   M  1 A 2
M /2 2
1

2
b  2
k
2

M i 1 2 12

Design of a time-division multiplexer

Q 4.6: Design a time-division multiplexer that will accommodate 11 channels. Assume that
the sources have the following specifications.

Source 1. Analaog, 2-KHz bandwidth, Source 2. Analog, 4-KHz bandwidth, Source 3.

Analog, 2-KHz bandwidth, Sources 4-11. Digital, synchronous at 7200 bits/sec
We also assume that the analog sources will be converted into 4-bit PCM words and, for
simplicity, that frame sync will be provided via a separate channel and synchronous TDM
lines are used. To satisfy the Nyquist rate for the analog sources, sources 1,2 and 3 need to be
sampled at 4,8 and 4 KHz respectively. This can be accomplished by rotating the first
commutator at f1  4 KHz and sampling source 2 twice on each revolution. This produces a
16-kilosample/sec TDM PAM signal on the commutator output.Each of the analog sample
values is converted into a 4-bit PCM word, so that the rate of the TDM PCM signal on the
ADC output is 64 kbps. The digital data on the ADC output may be merged with the data
from the digital sources by using a second commutator rotating at f 2  8 KHz and wired so
that the 64-kbps PCM signal is present on 8 of the 16 terminals. This provides an effective
sampling rate of 64kbps. On the other eight terminals the digital sources are connected to
provide a data transfer rate of 8kbps for each source. Since the digital sources are supplying a
7.2-kbps data stream, pulse stuffing is used to raise the source rate to 8kbps.

The main advantage of TDM is that it can easily accommodate both analog and digital
sources. However, when analog signals are converted to digital signals without redundancy
reduction, they consume a great deal of digital system capacity.

Q 4.7: Consider a PCM TDM system in which 24 signals are to be processed. Each of the
signals is bandlimited to 3.4 KHz and 8 bits are to be used for each quantized sample.
Conventional NRZ-L encoding is used and an additional 8-bit sync word is placed in each
frame. Find out the minimum bandwidth required.

Soln. The width of the shortest possible pulse needs to be determined in order to find
out the bandwidth. The sampling rate is

2 x 3.4 = 6.8 KHz

1 1 T f 0.147
The frame time is   0.147 mS , The word time is Tw    0.00588mS
f s 6. 8 k 25

where the value of 24 represents the 24 data plus the sync word for each frame.

0.00588
The bit interval is    0.000735mS
8

0.5
Hence the minimum transmission bandwidth is  680.272 KHz
0.000735
4.4 Digital PAM Signals

The PAM pulse train can be represented as

mt    bk p t  kT  (1)
k

where the modulating amplitude bk represents the k th symbol in the message sequence, so
the amplitude belong to a set of M discrete values. The index k ranges from   to 
unless otherwise mentioned. The unmodulated pulse pt  may be rectangular or some other
shape, subject to the condition

1 t  0
p t    (2)
0 t  T ,2T ,3T ...

This condition is necessary to ensure the recovery of the message signal by sampling mt 
periodically at t  iT , i  0,1,2,... as

miT    bk p iT  kT   bk
k

The rectangular pulse pt   t /   satisfies the above equation (2) if   T , as does any
T
time limited pulse with pt   0 for t  . T represents the pulse-to-pulse interval or the
2
1
time allocated to one symbol. The signaling rate becomes r  measured in symbols per
T
second or baud. In the binary case, the bit rate becomes

1
rb 
Tb
Sampling Discharge
switch Switch

G2
mt 
C m s t 
Gl

Fig 4.1. Flat Top sampling

In order to derive the power spectrum of the binary PAM waveform, under the assumption of
independent and identically distributed (i.i.d) bits, we write
  b2 jk

E bk b j  
0 jk

As the rectangular pulse has a power spectrum of T sin c 2 fT , the power spectrum of the
binary PAM signal becomes

 b2
G f   P f 
2

This is true if the PAM waveform has a mean value of zero. For a nonzero mean, the
expression becomes

G  f    b2T sin c 2 fT  mb2

where mb is its mean value. For unipolar signal formats, the ensemble average is given by
the following autocorrelation function

Rb n   E bk bk  n 

For a digital PAM signal having a pulse spectrum P f  and amplitude autocorrelation
function Rb n  , the power spectrum becomes


1
G f   P  f   Rb n e  j 2nfT
2

T n  

For the case of uncorrelated symbols with a nonzero mean, we obtain

 

 R n e
n  
b
 j 2nfT
  b2  mb2 e
n  
 j 2nfT

Use of Poisson’s sum formula gives us

 
1  n
e
n  
 j 2nfT

T
   f  T 
n  

and hence we write

2 2
 b2 m  
n  n
G f   P f    b  
2
P    f  
T  T  n   T   T

This result shows that the power spectrum of a digital PAM signal has impulses at harmonics
of the signaling rate r , unless the mean is zero or P f   0 at all values of frequency. It is
apparent from the above discussion that a synchronization signal can be obtained by applying
mt  to a narrow BPF centered at one of these harmonic frequencies. The average power is
obtained by integrating G f  over all f . Hence, m 2
For PTM Signals, under the assumption of uniform sampling, the duration of the k th pulse
is

 k   0 1  mkTs 

in which the unmodulated duration  0 represents mkTs   0 and the modulation index 
controls the amount of duration modulation. The condition 1  mkTs  ensures no missing or
negative pulses. The PPM pulses have fixed duration and amplitude and hence, unlike PAM
and PWM, they do not suffer from the drawback of missing or negative pulses. The k th
pulse in a PPM signal begins at a time

t k  kTs  t d  t 0 mkTs 

in which the unmodulated position kTs  t d represents mkTs  0  and the constant t 0
controls the placement of the modulated pulse. Let us consider rectangular pulses with
amplitude A centered at t  kTs in order to have an informative approximation for the PWM
waveform and let us further assume that  k varies slowly from pulse to pulse. Then the
spectrum for these natural sampled waveform is

2A
m p t   Af s 0 1  mt    sin n t  cos n s t
n 1 n

where  t   f s 0 1  mt  . From this equation, we observe that the PWM signal has a dc
component in addition to the message signal and phase modulated waves at the harmonics of
the sampling frequency f s . The phase modulation has negligible overlap in the message
band when  0  Ts so that the message signal can be recovered by lowpass filtering with a
DC block.

mt  Comparator
+
PWM
-
Monoshot PPM

Sawtooth
generator

Fig 4.2. Generation of PWM and PPM

A very popular IC NE 555 has been used to generate PWM (PDM/PTM) and PPM signals is
shown in Fig.4.3. (Students are encouraged to analyze the operation of this circuit and see
how it generates a waveform whose width or duration is modified according to a message or
modulating signal).

 Vcc

8 4
R2
PWM
R1 7 3
output
NE555
6 Modulating
C1 5
signal
2
C2 1
Clock

Fig. 4.3 An IC (NE 555 timer) based PWM modulator

Once we are able to generate a PWM signal, generation of a PPM signal is rather easier as it
can be carried out by a differentiating circuit.

 Vcc

8 4
R2
PPM
R1 7 3
output
NE555
6
C1 5

2
PWM C2 1
input

Fig. 4.4 PPM Modulator from PWM modulator

Message recovery in a PWM signal can also be carried out by converting the pulse-time
modulation to pulse-amplitude modulation. To do so, we need to generate a ramp signal such
as shown in Fig.4.5. This waveform is seen to start at time kTs and stops at t k , restarts at
k  1Ts . Both the start and stop epochs can be obtained from the edges of a PWM waveform
whereas a PPM waveform must have an auxiliary synchronization signal for the start epoch.

 Vcc

R -
Q2
+
Q1 C
PWM
Signal

Fig. 4.5 A PWM Demodulator

The operation of the circuit is explained as follows. The BJT Q1 acts as an inverter. The
transistor Q2 hence remains cut off during the high going portions of the incoming PWM
signal. This allows the capacitor C to get charged towards the biasing voltage through the
resistor R . The time constant RC is so chosen that before it can charge to  Vcc , the next
pulse of the input signal arrives. If it is high, then Q2 goes to the saturation condition which
makes the capacitor discharge through the ‘ON’ transistor Q2. The output at the collector of
this transistor is, therefore a swatooth kind of waveform whose envelope follows the
modulating signal. The second order low pass filter realized by the operational amplifier
helps to recover the message or the modulating signal from this sawtooth waveform.
 MODULE-V

NOISE IN ANALOG MODULATED SYSTEMS

5.1 Noise in AM receivers with envelope detection

The received signal at the envelope detector input consists of the modulated signal mt  at IF
and the narrowband noise nt  . This narrowband noise nt  is typically expressed in terms of
its inphase component nI t  and the quadrature phase component nQ t  . Thus the received
signal r t  becomes,

r t   mt   nt   Ac 1  m a mt cos 2f c t   n I t  cos 2f c t   nQ t sin 2f c t 

=  Ac  Ac m a mt   n I t cos2f c t   nQ t sin 2f c t 

Thus, the envelope of the received signal becomes


y t   r t    Ac  Ac ma mt   n I t   nQ2 t 
2
1
2

The signal yt  represents the output of an ideal envelope detector. The phase of the received
signal is not of any interest to the envelope detector as it responds to the envelope of the
received signal only and not to the phase changes. From the expression of the envelope, we
note that it is the vector sum of two noisy components; one is the desired signal plus the
inphase noise while the second term is noise only. In order to recover the original signal, we
may immediately see that, the term y t  needs a simple manipulation as follows. Expansion
of this term as a binomial expression and subsequent dropping of higher order terms give us

1
y t   r t    Ac  Ac ma mt   n I t   Ac  Ac ma mt   nI t  12 nQ2 t   ...  nQ t 
2

When the average carrier power is large compared with the average noise power, so that the
receiver is operating satisfactorily, the signal term is usually larger than the noise terms n I t 
and nQ t  , most of the time. The second term in the RHS is usually very small compared to
the first term and so the other terms following this in the series. Thus, the envelope of the
received signal, is approximated as, to a good extent

y t   r t   Ac  Ac ma mt   n I t 

The presence of the dc or the constant term Ac in the envelope detector is due to the
demodulation of the transmitted carrier. This term may be neglected as it does not contribute
to the original message signal. This term may be removed simply by means of a blocking
capacitor. We note that, the output of the envelope detector is the original signal, except for
the scaling factor. Thus, the signal-to-noise ratio, at the envelope detector output is expressed
as

Ac2 m a2 P
SNR o, AM 
2WN 0

This expression is valid subject to the following conditions:

a) The noise at the receiver input is small compared to the carrier

b) The modulation index satisfies ma  1
The figure of merit for an AM receiver is thus,

SNR o ma2 P

SNR C 1  ma2 P

5.2 Noise in Angle Modulated Systems

An angle modulated carrier, in general, is expressed as

xc t   Ac cos c t   t 

t
where  t   k p m t  for PM and  t   k f  m d for FM


As we have noted from the discussion on noise in AM systems, the noise appearing at the
output of the IF amplifier which is also the input to the demodulator is a bandpass noise with
a PSD of G n  f  and bandwidth equal to that of the IF amplifier; i.e. 2f  B  .

This bandpass noise can be expressed in terms of the quadrature components as

nt   n I t  cos  c t  nQ t sin  c t

where nI t  and nQ t  are both low pass signals of bandwidth 2f  B  . This noise can
also be represented in terms of an envelope and phase as nt   E n t  cos c t   n t  .

Due to the nonlinear nature of angle modulation, superposition can not be applied. However,
in special cases, the noise output is calculated by assuming the signal component to be zero.
We derive the first the results for PM and extend these to the FM case.

5.3 Phase Modulation

The narrow-band modulation is assumed to be approximately linear. Therefore, we undertake

the case of wideband frequency modulation. For such wide-band modulation, the signal
changes very slowly compared to noise nt  . The modulating signal bandwidth is B , and the
noise bandwidth is 2f  B  with f  B . Thus, the phase and the frequency variations of
the modulated carrier are much slower than are the variations of nt  . The modulated carrier
appears to have constant frequency and phase over several cycles, and hence, the carrier
appears to be unmodulated. We, may therefore calculate the output noise by assuming the
message signal to be zero or a constant. This is a qualitative argument for the linearity of
wideband angle modulated signals. We outline a quantitative analysis as detailed below.

The demodulator input corresponding to phase modulation is given by

r t   mt   nt   Ac cos c t   t   nt 

 Ac cos c t   t   E n t  cos c t   n t 
 Rt  cos c t   t    t 

E n t  E n t sin  t 

R t   t 
Ac
 t 

 c t   t   n t 

Fig. 5.1. Phasor diagram of the noisy FM signal appearing at the discriminator input

Expansion of this term gives us

r t   Ac cos  c t. cos  t   sin  c t. sin  t   E n t cos  c t. cos n t   sin  c t. sin n t 

Arranging the carrier terms of the above expression, we have

r t   Ac cos  t   E n t  cos n t cos  c t   Ac sin  t   E n t sin n t sin  c t

Therefore,


R t   Ac2  E n2  2 Ac E n cos t    n t  12
,

Ac sin  t   E n t  sin n t 
tan t   
Ac cos  t   E n t  cos n t 

 t   k p m t  for phase modulation. The resultant can also be written as

12
 E2 E 
R t   Ac 1  n2  2 n cos t    n t 
 Ac Ac 
We are interested in analyzing the effect of additive noise on the phase angle of the signal
appearing at the discriminator input. The envelope term appearing in the above expression is
not of any interest to us as information resides in the phase of the modulated signal. Hence,
any change in the phase of the signal at the discriminator input due to noise is likely to bring
about a change to the original signal. As this analysis is quite involved, we seek to simplify
this by making reasonable assumptions about the SNR at the discriminator input. We
consider the case of large SNR first. Under this situation, the phasor diagram corresponding
to the actual phase, the extra phase shift introduced due to the additive noise is illustrated in
Fig.

From Fig.5.1,

sin  t   E n t sin  t  Rt 

where  t    n t    t 


For small noise case, E n t   Ac almost always,  t   for almost all t and the
2
resultant R t  is approximated as

Rt   Ac

therefore,

sin t    t   E n t sin  t  Rt 

E n t 
 t   sin n t    t 
Ac

The discriminator detects the phase of the input and gives an output proportional to

En
x0 t   k p mt   sin n t    t 
Ac

We observe from the above expression that, the noise has affected the phase of the modulated
signal by adding one term to the original phase. As we have assumed the phase corresponding
to the message signal to vary slowly than the  n t  term, we approximate  t  by a constant
 . Therefore,

En E
 t   sin  n t      n sin n t . cos   cos n t . sin  
Ac Ac


1
Ac

nQ t  cos   n I t sin  
The two quadrature components of white noise are uncorrelated to each other. Hence, the
PSD corresponding to these two terms is

cos 2  sin 2  S nQ  f 
S   f   2
S nQ  f   2
S nI   f  
Ac Ac Ac2

This is because the PSDs corresponding to the two quadrature components are assumed to be
equal.


 f  f  B
For a white channel noise the PSDs are equal to S   f    Ac2
0 otherwise


The demodulated noise bandwidth is f  B . However, the useful signal bandwidth is only
B as the demodulated output passes through a low pass filter of cutoff frequency B to
remove the out of band noise. Thus, the PSD of the low pass filter output noise is


 f B
S n0  f    Ac2
0 otherwise


The output noise power, is, therefore,

  2B
N 0  2 B. 2  2
 A
 Ac  c

The signal power observed at the output of the demodulator is

S 0  k p m 2 t 

The output SNR, of a PM receiver is therefore,

2 m t 
2
 Ac k p 
S0
N0 2B

These results are valid for small noise case and apply to both NBPM and WBPM. For PM,
the maximum frequency deviation is expressed as

 . 
f  k p m ,p where m ,p  mt 
  max

Substitution of these in the output SNR gives us

S 0  Ac f  m 2 t 
2


N0 2B .
m 2p
5.4 Noise in FM Systems

Frequency modulation can be viewed as a special kind of phase modulation, where the
t
modulating signal is  m d as illustrated in Fig.

At the receiver, we demodulate FM with

t
a PM demodulator followed by a differentiator. The PM demodulator output is k f  m d .


The subsequent differentiator gives an output of the form k f mt  , so that we have, for the
output signal power

S 0  k 2f mt 
2

The phase demodulator output noise is identical to the one derived in the previous section

with a PSD equal to 2 for white channel noise. This noise is passed through an ideal
Ac
differentiator that has a transfer function equal to j 2f . Hence the PSD of the output noise is
2
j 2f times the input PSD. We, therefore, write


 2 2f 
2
f B
S n 0  f    Ac
0 otherwise


The output noise power is, hence,

B
 8 2B 3
 A 2 2f  df
2
N0  
B c 3 Ac2

Hence, the output SNR is

S0  k 2 mt 2  A 2 2   k 2 mt 2 
3  f
  c
  3
f

N0 2 
 2B   B    2B 2 
   

As f  2k f m p , we write

 k 2 mt 2 
  3 f   mt    3 2  mt 
2 2   2 
S0 
 3
f

N0  2B 2   B   m p
2 

 m 2p



 

The transmission bandwidth is about 2f . Hence, for each doubling of the bandwidth, the
output SNR increases by 6 dB. Unlike PM, the output SNR does not increase indefinitely
because of the appearance of threshold. This is because an increase in bandwidth results in a
correspondingly increased noise power creeping into the system compared to the carrier
power resulting in threshold.

For tone modulation,

 mt 2 
   0.5
 m 2p 
 

and thus, the output SNR becomes

S0 3 2
  
N0 2

The output SNR in dB is plotted in Fig. as a function of  (also in dB) for various values of
 . The dotted portion of the curve indicates the threshold region. Although the graphs in Fig.
are valid for tone modulation only, they can be used for any other modulating signal simply
 mt 2   mt 2 
by shifting them vertically by a factor of  2  0.5  2 2  . For tone modulation, we
 mp   mp 
   
observe that FM is superior to PM by a factor of 3 dB. This does not mean that FM is
superior to PM for other modulating signals as well. In fact, PM is better than FM for most
practical signals. We write,

S 0 N 0 PM  Bm p  
2

 
S 0 N 0 FM  3m ' 2p 


Thus, we observe that PM is better than FM from the output SNR point of view under the
condition of Bm p  > 3m ' p . If the PSD of the message signal is concentrated at lower
2 2

frequencies, low frequency components predominate it and m 'p is small. This favors PM.
Thus, in general, PM is better than FM for message signals having predominant low
frequency components (like the video signal) and FM is better than PM for message signals
that have an abundance of high frequency components. This explains the better SNR of FM
than PM for tone modulation as all the signal power is concentrated in the highest frequency
band. But for most of the practical signals, the signal power is usually concentrated at lower
frequencies and this makes PM a better candidate than FM for the modulation choice.