TC 503

Digital Communication Theory

Course Teacher:
Dr. Muhammad Imran Aslam

SPECTRUM DENSITY

Dr. M. Imran Aslam

Spectral Density
The spectral density of a signal characterizes the distribution
of the signals energy or power in the frequency domain.
This concept is particularly important when considering
filtering in communication systems.
We need to be able to evaluate the signal and noise at the
filter output.
The energy spectral density (ESD) or the power spectral
density (PSD) is used in the evaluation.
Dr. M. Imran Aslam

1. Energy Spectral Density (ESD)

Energy spectral density (ESD) describes the signal energy per unit
bandwidth measured in joules/hertz.
If X(f) is Fourier transform of an energy signal x(t), then its ESD [ () ],
2
the squared magnitude spectrum i.e. x ( f ) = X ( f )
According to Parsevals theorem, the energy of x(t):

Ex =

x (t) dt =
2

Therefore: E x = x (f) df

2
|X(f)|
df

As for real signals x(t), |X(f)| is an even function of frequency.

Therefore, Energy spectral density is symmetrical in frequency about
origin and total energy of the signal x(t) can be expressed as:

E x = 2 x (f) df
0

Parsevals Theorem: The sum (or integral) of square of a signal is equal to sum (or
integral of square of its (Fourier) transform.
Dr. M. Imran Aslam

2. Power Spectral Density (PSD)

The power spectral density (PSD) describes the signal power per unit
bandwidth measured in watts/hertz.
The PSD Gx(f ) of a periodic signal x(t) is a real, even, and nonnegative
function of frequency that gives the distribution of the power of x(t) in
the frequency domain.

2
G
(f
)
=
|C
|
( f nf 0 )

x
n
PSD is represented as:
n=-

NOTE: For non-periodic case = lim | ()|2

Whereas the average power of a periodic signal x(t) is represented as:

T /2

1 0 2
Px =
x (t) dt
=

T0 T0 / 2

|C |

n=-

Using PSD, the average normalized power of a real-valued signal is

represented as:
=
Px

G (f) df 2 G
=
x

(f) df

Dr. M. Imran Aslam

Problem

Dr. M. Imran Aslam

Solution

Dr. M. Imran Aslam

AUTOCORRELATION

Dr. M. Imran Aslam

Autocorrelation

Autocorrelation of an Energy Signal

Correlation is a matching process; autocorrelation refers to the
matching of a signal with a delayed version of itself.
Autocorrelation function of a real-valued energy signal x(t) is defined
as:

R x ( ) =

x(t) x (t + ) dt

for

- < <

The autocorrelation function Rx() provides a measure of how closely

the signal matches a copy of itself as the copy is shifted units in time.
Rx() is not a function of time; it is only a function of the time
difference between the waveform and its shifted copy.

Dr. M. Imran Aslam

Properties of autocorrelation of an Energy Signal

R x ( ) =R x ( )

symmetrical about zero

R x ( ) R x (0) for all maximum value occurs at the origin

R x ( ) x (f )

autocorrelation and ESD form a

Fourier transform pair, as designated
by the double-headed arrows

=
E X R=
x (0)

x 2 (t ) dt

value at the origin is equal to

the energy of the signal

= x ( f ) df
-

Dr. M. Imran Aslam

10

Autocorrelation of a Power Signal

Autocorrelation function of a real-valued power signal x(t) is defined as:

R x ( )

T /2

lim
T

1
x(t) x (t + ) dt

T T / 2

for - < <

When the power signal x(t) is periodic with period T0, the
autocorrelation function can be expressed as

R x ( )

1
T0

T0 / 2

x(t) x (t + ) dt

for - < <

T0 / 2

Dr. M. Imran Aslam

11

Properties of Autocorrelation of a Power Signal

The autocorrelation function of a real-valued periodic signal has the
following properties similar to those of an energy signal:

R x ( ) =R x ( )

symmetrical about zero

R x ( ) R x (0) for all

maximum value occurs at the origin

R x ( ) Gx (f)
=
PX R=
x (0)

1
T0

T0 /2

T0 /2

x 2 (t ) dt

autocorrelation and PSD form a

Fourier transform pair
value at the origin is equal to the
average power of the signal

G ( f ) df
x

Dr. M. Imran Aslam

12

Random Signals
Random Variables

All useful message signals appear random; that is, the receiver does
not know, a priori, which of the possible waveform have been sent.
Let a random variable X(A) represent the functional relationship
between a random event A and a real number.
The (cumulative) distribution function FX(x) of the random variable X is
given by

FX=
( x) P( X x)

Another useful function relating to the random variable X is the

probability density function (PDF)

dFX ( x)
PX ( x) =
dx
Dr. M. Imran Aslam

13

Ensemble Averages

=
m X E=
{X }

xp

( x)dx

E{ X 2 } =

x 2 p X ( x)dx

The first moment of a probability

distribution of a random variable
X is called mean value mX, or
expected value of a random
variable X
The second moment of a
probability distribution is the
mean-square value of X

Central moments are the

moments of the difference
between X and mX and the
( x m X ) 2 p X ( x)dx second central moment is the
variance of X

=
X ) E{( X m X ) 2 }
var(

var(
=
X ) E{ X 2 } E{ X }2

Variance is equal to the

difference between the meansquare value and the square of
the mean

Dr. M. Imran Aslam

14

Random Processes
A random process X(A, t) can be viewed as a function of two variables:
a random variable (A) and time (t).
X(Ak,t) is called a
sample function
X(A, tk) is simply
gives random
variable
X(Ak,tk) is just a
number

Dr. M. Imran Aslam

15

Statistical Averages of a Random Process

A random process whose distribution functions are continuous can be
described statistically with a probability density function (PDF).
A partial description consisting of the mean and autocorrelation
function are often adequate for the needs of communication systems.
Mean of the random process X(t) :

=
E{ X (tk )}

xp ( x) dx
=

Xk

mX (tk )

Autocorrelation function of the random process X(t)

RX (t1 , t2 ) = E{ X (t1 ) X (t2 )}

Dr. M. Imran Aslam

16

Stationarity
A random process X(t) is said to be stationary in the strict
sense if none of its statistics are affected by a shift in the time
origin.
A random process is said to be wide-sense stationary (WSS) if
two of its statistics, its mean and autocorrelation function, do
not vary with a shift in the time origin.

E{ X (t=
)} m=
A constant
X
RX (t1 , t2 =) RX (t1 t2 )= RX ( )

Dr. M. Imran Aslam

17

Autocorrelation of a WSS Process

For a wide-sense stationary process, the autocorrelation function is

only a function of the time difference = t1 t2;

RX ( ) E{ X (t ) X (t + )}
=

for

< <

Properties of the autocorrelation function of a real-valued widesense stationary process are

RX (=
) RX ( )
RX ( ) RX (0) for all

RX ( ) GX ( f )

RX (0) = E{ X 2 (t )}

Symmetrical in about zero

Maximum value occurs at the origin
Autocorrelation and power spectral
density form a Fourier transform pair
Value at the origin is equal to the
average power of the signal

Dr. M. Imran Aslam

18

Time Averaging and Ergodicity

When a random process belongs to a special class, known as an
ergodic process, its time averages equal its ensemble averages.
The statistical properties of such processes can be determined by time
averaging over a single sample function of the process.
A random process is ergodic in the mean if
T /2

1
mX = lim
X (t )dt

x T
T / 2
It is ergodic in the autocorrelation function if

1
=
RX ( ) lim
x T

T /2

X (t ) X (t + )dt

T / 2
Dr. M. Imran Aslam

19

For Ergodic process, fundamental electrical parameters can be

evaluated using signal statistics. Following is the summary of
these relations:

Dr. M. Imran Aslam

20

Power Spectral Density

A random process X(t) can generally be classified as a power signal
having a power spectral density (PSD) GX(f )
PSD of a real valued X(t) is a real-valued function of f.
Principal features of PSD functions

GX ( f ) 0

GX (=
f ) GX ( f )

GX ( f ) RX ( )

=
PX R=
X (0)

( f )df

PSD is always positive

PSD is symmetric about
frequency origin
PSD and autocorrelation form a
Fourier transform pair
Relationship between average
normalized power and PSD

Dr. M. Imran Aslam

21

Example of calculation
of Autocorrelation Rx()
Rx() provides
information
regarding signal
bandwidth
If Rx() ramps down
slowly, we are
dealing with a lowbandwidth signal
If Rx() ramps down
fast, we are dealing
with a highbandwidth signal

Dr. M. Imran Aslam

22

Dr. M. Imran Aslam

23

Problem

Dr. M. Imran Aslam

24

Solution

Dr. M. Imran Aslam

25

Problem

Dr. M. Imran Aslam

26

Solution

Dr. M. Imran Aslam

27

NOISE IN COMMUNICATION
SYSTEMS

Dr. M. Imran Aslam

28

Noise in Communication Systems

The term noise refers to unwanted electrical signals that are always
present in electrical systems; e.g spark-plug ignition noise, switching
transients, and other radiating electromagnetic signals.
Noise due to thermal agitation of electrons in electronic circuits
(known as thermal noise) is the most dominant in communication
systems.
Can describe thermal noise as a zero-mean Gaussian random process.
A Gaussian process n(t) is a random function whose amplitude at any
arbitrary time t is statistically characterized by the Gaussian probability
density function
1 n 2
1
=
p ( n)
exp
2
2
NOTE: 2 represents noise power
Dr. M. Imran Aslam

29

Noise in Communication Systems

The normalized or standardized Gaussian density function of a zeromean process is obtained by assuming unit variance.

Dr. M. Imran Aslam

30

White Noise
The primary spectral characteristic of thermal noise is that its power
spectral density is the same for all frequencies of interest in most
communication systems
Power spectral density Gn(f )

N0
Gn ( f ) =
watts / hertz
2

Autocorrelation function of white noise is

N0
Rn ( ) =
( )
{Gn ( f )} =
2
1

The average power Pn of white noise is infinite

p ( n) =

N0
df =
2

Dr. M. Imran Aslam

31

Dr. M. Imran Aslam

32

 The effect on the detection process of a channel with additive white

Gaussian noise (AWGN) is that the noise affects each transmitted
symbol independently.
Such a channel is called a memoryless channel.
The term additive means that the noise is simply superimposed or
added to the signal

Dr. M. Imran Aslam

33

SIGNAL TRANSMISSION THROUGH

LINEAR SYSTEMS

Dr. M. Imran Aslam

34

Signal Transmission through Linear Systems

A system can be characterized equally well in the time domain or
the frequency domain, techniques will be developed in both
domains
The system is assumed to be linear and time invariant.
It is also assumed that there is no stored energy in the system at
the time the input is applied

Dr. M. Imran Aslam

35

Impulse Response
The linear time invariant system or network is characterized in the time
domain by its impulse response h(t )

=
y (t ) h=
(t ) when x(t ) (t )

The response of the network to an arbitrary input signal x (t )is found by

the convolution of x (t )with h (t )

y (t ) = x(t ) h(t ) =

x( )h(t )d

The system is assumed to be causal, which means that there can be no

output prior to the time, t =0,when the input is applied.
For causal systems, The convolution integral can be expressed as:

=
y (t )

x( )h(t )d
0

Dr. M. Imran Aslam

36

Frequency Transfer Function

The frequency-domain output signal Y (f )is obtained by taking the
Fourier transform

Y( f )= X( f )H( f )

Frequency transfer function or the frequency response is defined as:

H( f ) =

Y( f )
X(f )

H ( f ) = H ( f ) e j ( f )
The phase response is defined as:

( f ) = tan
NOTE:

Im{H ( f )}
Re{H ( f )}

h(t ) H ( f )
Dr. M. Imran Aslam

37

Random Processes and Linear Systems

If a random process forms the input to a timeinvariant linear system, the output will also be a
random process.
The input power spectral density GX (f )and the
output power spectral density GY (f )are related as:
GY ( f ) = GX ( f ) H ( f )

Dr. M. Imran Aslam

38

Distortionless Transmission
What is the required behavior of an ideal transmission line?
The output signal from an ideal transmission line may have some time
delay and different amplitude than the input
It must have no distortionit must have the same shape as the input.
Therefore, For ideal distortionless transmission:
Output signal in time domain

y=
(t ) Kx(t t0 )

Output signal in frequency domain

Y ( f ) = KX ( f )e j 2 ft0

Therefore required Transfer Function H

=
(f)
for Distortionless transmission is

Dr. M. Imran Aslam

Y( f )
= Ke j 2 ft0
H( f )

39

What is the required behavior of an ideal transmission line?

The overall system response must have a constant magnitude response
The phase shift must be linear with frequency
All of the signals frequency components must also arrive with identical
time delay in order to add up correctly
Time delay t0 is related to the phase shift and the radian frequency
= 2f by:
t0 (seconds) = (radians) / 2f (radians/seconds )
Another characteristic often used to measure delay distortion of a
signal is called envelope delay or group delay:

1 d ( f )
( f ) =
2 df

For distortionless transmission, group delay should be constant.

In practice a signal may be distorted in passing through some parts of a
system
Phase or amplitude correction (equalization) network may be used to
correct distortion.
Dr. M. Imran Aslam

40

FILTERS

Dr. M. Imran Aslam

41

Ideal Filters
Ideal filter passes (without
distortion) all frequency
components between
frequencies fl and fu. For ideal
filters the response exists
between two cut-off frequencies
fu (upper cutoff) and fl (lower
cutoff).

For band pass filters:

fu > 0, fl < , fu > fl
For low pass filters:
fl = 0, 0 < fu <
For high pass filters:
0 < fl < , fu
Dr. M. Imran Aslam

42

Transfer Function of Ideal Low Pass Filter

For the ideal low-pass filter transfer function with bandwidth Wf = fu
hertz can be written as:

H ( f ) = H ( f ) e j ( f )
Where

1
H( f ) =
0

and

j ( f )

=e

for | f | < fu
for | f | fu
j 2 ft0
Figure1.11 (b) Ideal low-pass filter

Dr. M. Imran Aslam

43

Impulse Response of Ideal Low Pass FIlter

The impulse response of the ideal low-pass filter:

h(t ) = 1{H ( f )}

H ( f )e j 2 ft df

fu

e j 2 ft0 e j 2 ft df

fu

fu

e j 2 f (t t0 ) df

fu

sin 2 fu (t t0 )
= 2 fu
2 fu (t t0 )

The impulse response is non-causal

This ideal filter is not realizable

2 fu sinc 2 fu (t t0 )
Dr. M. Imran Aslam

44

Dr. M. Imran Aslam

45

Realizable Filters
The simplest
example of a
realizable low-pass
filter is an RC filter

H( f ) =
H( f ) =

1
1 + j 2 f
1
1 + (2 f ) 2

e j ( f )

Dr. M. Imran Aslam

46

Dr. M. Imran Aslam

47

Realizable Filters
There are several useful approximations to the ideal low-pass filter
characteristic and one of these is the Butterworth filter

Hn ( f )
=

1
1 + ( f / fu )

2n

n 1

Where fu is upper 3dB cutoff

frequency and n is referred to
as the filter order
Butterworth filters are popular
because they are the best
approximation to the ideal, in
the sense of maximal flatness
in the filter pass band.
Filter Shape Factor is a measure of how well a realizable filter approximates
the ideal filter. It is typically defined as the ratio of filter bandwidth at 60dB
and 6dB amplitude
Dr. M. Imran Aslam

48

Problem

Dr. M. Imran Aslam

49

Solution

Dr. M. Imran Aslam

50

Signals, Circuits, and Spectra

Signals are described in terms of their spectra

Circuits are described in terms of their spectral characteristics.

Dr. M. Imran Aslam

51

Signals, Circuits, and Spectra

Wp: Input pulse duration
(Wp = 1/T)
Wf: Filter Bandwidth
[Wf=1/(2RC)]

Dr. M. Imran Aslam

52

BANDWIDTH

Dr. M. Imran Aslam

53

Bandwidth Of Digital Data

Baseband versus Bandpass

An easy way to translate the

spectrum of a low-pass or baseband
signal x(t) to a higher frequency is to
multiply or heterodyne the baseband
signal with a carrier wave cos 2fct
xc(t) is called a double-sideband
(DSB) modulated signal
xc(t) = x(t) cos 2fct
(1.70)
From the frequency shifting theorem
Xc(f) = 1/2 [X(f-fc) + X(f+fc) ]
(1.71)
Generally the carrier wave frequency
is much higher than the bandwidth of
the baseband signal fc >> fm and
therefore WDSB = 2fm
We need twice as much bandwidth
to transmit a DSB version of a
baseband signal
Dr. M. Imran Aslam

54

Bandwidth Dilemma
Theorems of
communication and
information theory are
based on the assumption of
strictly bandlimited
channels
The mathematical
description of a real signal
does not permit the signal
to be strictly duration
limited and strictly
bandlimited.

Strictly bandlimited signal is not realizable

For realizable signal bandwidth is infinite.
Therefore a bandwidth criterion is needed.
Dr. M. Imran Aslam

55

1.7.2 Bandwidth Dilemma

All bandwidth criteria have in common the attempt to specify a
measure of the width, W, of a nonnegative real-valued spectral
density defined for all frequencies | f | <
The single-sided power spectral density for a single heterodyned
pulse xc(t) takes the analytical form:

sin ( f f c )T
Gx ( f ) = T

(
)

f
f
T

Where, fc: carrier frequency

T: Pulse duration

Dr. M. Imran Aslam

56

Different Bandwidth Criteria

(a) Half-power bandwidth.
(b) Equivalent rectangular
or noise equivalent
bandwidth.
(c) Null-to-null bandwidth.
(d) Fractional power
containment bandwidth.
(e.g. 99% power)
(e) Bounded power
spectral density.
(f) Absolute bandwidth.
(infinite in this case)

Dr. M. Imran Aslam

57