UNIT- 1

CHAPTER -1

BASICS OF SIGNALS AND SYSTEMS

1.1 Introduction
1.2 Energy Signal and Power Signal

1.3 Transformations of the independent variable

1.3.1 Examples of Transformations of the Independent Variable

1.4 Periodic Signals

1.5 Even and Odd Signals

1.6 Exponential and sinusoidal signals

1.6.1 Continuous-time complex exponential and sinusoidal signals

1.6.2 Discrete-time complex exponential and sinusoidal signals

1.6.3 Sinusoidal Signals

1.7. The discrete-Time Unit Impulse and Unit Step Sequences

1.8 The Continuous-Time Unit Step and Unit Impulse Functions

1.9 Sampling property of the continuous-time unit impulse:

1.10 Continuous-Time and Discrete-Time Systems

1.10.1 Simple Examples of Systems

1.11 Interconnects of Systems

1.12 Basic System Properties

1.12.1 Systems with and without Memory

1.12.2 Invertibility and Inverse System

1.12.3 Causality

1.12.4 Stability
1.12.5 Time Invariance
1.12.6 Linearity
1.0 OBJECTIVE

• Understand continuous and discrete time signals.

• Understand continuous and discrete time systems.
• Classify the signals and Systems

1.1 INTRODUCTION
Signals are represented mathematically as functions of one or more independent variables. Here
we focus attention on signals involving a single independent variable. For convenience, this
will generally refer to the independent variable as time.

There are two types of signals: continuous-time signals and discrete-time signals.

Continuous-time signal: The variable of time is continuous. A speech signal as a function of

time is a continuous-time signal.

Figure 1.1: Graphical representation of Continuous-time signal

Discrete-time signal: the variable of time is discrete. The weekly Dow Jones stock market
index is an example of discrete-time signal.

Figure 1.2 : Graphical representation of Discrete-time signals

To distinguish between continuous-time and discrete-time signals we use symbol t to denote
the continuous variable and n to denote the discrete-time variable. And for continuous-time
signals we will enclose the independent variable in parentheses (·), for discrete-time signals we
will enclose the independent variable in bracket [·].

A discrete-time signal x [n]may represent a phenomenon for which the independent variable is
inherently discrete. A discrete-time signal x [n]may represent successive samples of an
underlying phenomenon for which the independent variable is continuous. For example, the
processing of speech on a digital computer requires the use of a discrete time sequence
representing the values of the continuous-time speech signal at discrete points of time.

1.2 Energy Signal and Power Signal

If v (t) and i(t)are respectively the voltage and current across a resistor with resistance R ,
then the instantaneous power is

1
p(t) = v(t) i(t) = R v2 (t)

The total energy expended over the time interval is t1 ≤ t ≤ t2

𝑡2 𝑡2 1
∫𝑡1 𝑝(𝑡)𝑑𝑡 = ∫𝑡1 𝑅
v2 (t) dt

and the average power over this time interval is

1 𝑡2 1 𝑡2 1
∫ 𝑝(𝑡)𝑑𝑡
𝑡2 –𝑡1 𝑡1
= 𝑡2 –𝑡1 ∫𝑡1 𝑅
v2 (t) dt

For any continuous-time signal x (t)( or any discrete-time signal x [n], the total energy over the
time interval t1 ≤ t ≤ t2 in a continuous-time signal x (t) is defined as

𝑡2
∫𝑡1 |𝑥(𝑡)| 2 dt

where |x| denotes the magnitude of the (possibly complex) number x .

The time-averaged power is

1 𝑡2
𝑡2 –𝑡1
∫𝑡1 |𝑥(𝑡)| 2 dt

Similarly the total energy in a discrete-time signal x [n] over the time interval n1 ≤ n ≤ n2 is
defined as

∑𝑛2
𝑛1|𝑥[𝑛]|
2
The average power is
1
∑𝑛2
𝑛1|𝑥[𝑛]|
2
𝑛2−𝑛1+1

In many systems, we will be interested in examining the power and energy in signals over an
infinite time interval, that is, for - ∞ ≤ t ≤ +∞ or. - ∞ ≤ n ≤ +∞
The total energy in continuous time is then defined
𝑇 ∞
E ∞= Lim (T→∞) ∫−𝑇 |𝑥(𝑡)| 2 dt = ∫−∞|𝑥(𝑡)|2 dt ,

And in discrete time,

E ∞= Lim (N→∞) ∑𝑁 ∞
−𝑁 |𝑥 [𝑛 ]| = ∑−∞ |𝑥 [𝑛 ]|
2 2

For some signals, the integral in continuous Equation or sum in discrete might not converge,
that is, if x (t) or x [n] equals a nonzero constant value for all time. Such signals have infinite
energy, while signals with E ∞ < ∞ have finite energy.
The time-averaged power over an infinite interval
1 𝑇
P∞ = Lim (T→∞) ∫ |𝑥(𝑡)| 2 dt
2𝑇 −𝑇
1
P∞ = Lim (N→∞) 2𝑁+1
∑𝑛2
𝑛1|𝑥[𝑛]|
2

Three types of signals:

Type 1: signals with finite total energy, E ∞ < ∞ and zero average power,
E∞
P∞ = Lim (T→∞) =0
2𝑇

Type 2: with finite average power P∞.

If P∞ > 0 , then E∞= ∞ .

An example is the signal x [n] = 4 ,

it has infinite energy, but has an average power of P∞ =16.

Type 3: signals for which neither P∞ and E∞ are finite. An example of this signal is x(t )= t .
1.3 Transformations of the independent variable

In many situations, it is important to consider signals related by a modification of the

independent variable. These modifications will usually lead to reflection, scaling, and shift.
1.3.1 Examples of Transformations of the Independent Variable
 Figure.1.3 Discrete-time signals related by a time shift.

Fig. 1.4 Continuous-time signals related by a time shift.

Fig. 1.5 (a) A discrete-time signal x [n]; (b) its reflection, x [-n] about n = 0

Fig. 1.6 (a) A continuous-time signal x( t) ; (b) its reflection, x (-t)about t = 0 .

 Fig. 1.7 Continuous-time signals related by time scaling.

1.4 Periodic Signals

A periodic continuous-time signal x (t) has the property that there is a positive value of T for
which x (t) = x (t + T) for all t
From Equation , we can deduce that if x (t) is periodic with period T,

then x (t) = x (t + mT) for all t and for all integers m .

Thus, x( t) is also periodic with period 2T, 3T, …. The fundamental period T0 of x( t) is the
smallest positive value of T

Fig. 1.8 Continuous-time periodic signal.

A discrete-time signal x [n] is periodic with period N ,

where N is an integer, if it is unchanged by a time shift of N,

x[ n] = x [n + N] for all values of n.
If Equation holds, then x [n] is also periodic with period 2N , 3N , …. The fundamental period
N0 is the smallest positive value of N for which Equation holds.
 Fig. 1.9 Discrete-time periodic signal.

1.5 Even and Odd Signals

In addition to their use in representing physical phenomena such as the time shift in a radar
signal and the reversal of an audio tape, transformations of the independent variable are
extremely useful in examining some of the important properties that signal may possess.
Signal with these properties can be even or odd signal, periodic signal:
An important fact is that any signal can be decomposed into a sum of two signals, one of which
is even and one of which is odd.

Fig. 1.10 An even continuous-time signal; (b) an odd continuous-time signal.

which is referred to as the even part of x( t) .

Similarly, the odd part of x (t) is given by

Exactly analogous definitions hold in the discrete-time case.

 Fig.1.11 The even-odd decomposition of a discrete-time signal

1.6 Exponential and sinusoidal signals

1.6.1 Continuous-time complex exponential and sinusoidal signals

The continuous-time complex exponential signal

x(t)= Ceat
Where C and a are in general complex numbers.
Real exponential signals

Fig. 1.12 The continuous-time complex exponential signal at x(t)= Ceat

, (a) a > 0 ; (b) a < 0 .
Periodic complex exponential and sinusoidal signals
If a is purely imaginary,

we have x( t)= ejա0t

An important property of this signal is that it is periodic. We know x (t) is periodic with period
T if

ejա0t = ejա0(t +T) = ejա0t ejա0T

For periodicity, we must have

ejա0T =1
For ա0 ≠ 0, the fundamental period T0 is
2𝜋
T0 = |ա0|

Thus, the signals ejա0t and e-jա0t have the same fundamental period.

A signal closely related to the periodic complex exponential is the sinusoidal signal
x (t) = A cos(ա0 t + Ɵ)

With seconds as the unit of t, the units of Ɵ and ա0 are radians and radians per second. It is
also known ա0=2Лf0, where f0 has the unit of circles per second or Hz.
The sinusoidal signal is also a periodic signal with a fundamental period of T0 .

Fig. 1.13 Continuous-time sinusoidal signal.

Using Euler’s relation, a complex exponential can be expressed in terms of sinusoidal signals
with the same fundamental period:

ejա0t = cos ա0t + j sin ա0t

Similarly, a sinusoidal signal can also be expressed in terms of periodic complex exponentials
with the same fundamental period:
A sinusoid can also be expresses as

And

Periodic signals, such as the sinusoidal signals provide important examples of signal with
infinite total energy, but finite average power. For example:

Since there are an infinite number of periods as t ranges from - ∞ to + ∞ , the total energy
integrated over all time is infinite. The average power is finite since

General complex Exponential signals

Consider a complex exponential at Ceat, where C= |C| e jƟ is expressed in polar and

a = r +j w0 is expressed in rectangular form.

Then

Thus, for r = 0 , the real and imaginary parts of a complex exponential are sinusoidal.
For r > 0 , sinusoidal signals multiplied by a growing exponential.

For r < 0 , sinusoidal signals multiplied by a decaying exponential.

Damped signal – Sinusoidal signals multiplied by decaying exponentials are commonly
referred to as damped signal.
 Fig. 1.14 (a) Growing sinusoidal signal; (b) decaying sinusoidal signal.

1.6.2 Discrete-time complex exponential and sinusoidal signals

A discrete complex exponential or sequence is defined by

x[n]= Cαn
where C and αare in general complex numbers. This can be alternatively expressed

x [n]= C eβn
where α =eβ
Real Exponential Signals

If C and α are real, we have the real exponential signals

Fig. 1.15 Real Exponential Signal x [n]= Cαn

: (a) α >1; (b) 0<α< 1 (c) -1< α< 0; (d) α < -1
1.6.3 Sinusoidal Signals
x[n]= e jω0n

e jω0n = cosω0 n + j sin ω0 n

Similarly, a sinusoidal signal can also be expresses in terms of periodic complex exponentials
with the same fundamental period:
𝐴 𝐴
A cos (ω0 n +ϕ ) = e jϕ e j ω0 n + e -jϕ e -j ω0 n
2 2

A sinusoid can also be expresses as

A cos (ω0 n +ϕ )= A Re{ e j(ω0 n +ϕ)}

And A sin (ω0 n +ϕ )= A Im{ e j(ω0 n +ϕ)}
The above signals are examples of discrete signals with infinite total energy, but finite average
power. For example: every sample of x[ n] = e (-j ω0 n ) contributes 1 to the signal’s energy.
Thus the total energy - ∞ < n < +∞ is infinite, while the average power is equal to 1.

Fig.1.16 Discrete-time sinusoidal signal.

1.7. The discrete-Time Unit Impulse and Unit Step Sequences

Discrete-time unit impulse is defined as

Fig. 1.17 Discrete-time unit impulse.

Discrete-time unit step is defined as

Fig. 1.18 Discrete-time unit step sequence.

The discrete-time impulse unit is the first difference of the discrete-time step

The discrete-time unit step is the running sum of the unit sample:

It can be seen that for n < 0 , the running sum is zero, and for n ≥0 , the running sum is 1.
If we change the variable of summation from m to k = n - m we have,
The unit impulse sequence can be used to sample the value of a signal at n = 0. Since it is
nonzero only for n = 0, it follows that

More generally, a unit impulse

This sampling property is very important in signal analysis.

1.8 The Continuous-Time Unit Step and Unit Impulse Functions

Continuous-time unit step is defined as

Fig. 1.19 Continuous-time unit step function

The continuous-time unit step is the running integral of the unit impulse

The continuous-time unit impulse can also be considered as the first derivative of the
continuous time unit step,
Since u (t) is discontinuous at t = 0 and consequently is formally not differentiable. This can
be interpreted, however, by considering an approximation to the unit step u Δ(t) , as illustrated
in the figure below, which rises from the value of 0 to the value 1 in a short time interval of
length Δ.

Fig. 1.20 (a) Continuous approximation to the unit step uΔ (t) ; (b) Derivative of uΔ (t) .
The derivative is

Note that It is a short pulse, of duration Δ and with unit area for any value of Δ. As Δ -> 0 ,
becomes narrower and higher, maintaining its unit area. At the limit,

And

Graphically, it is represented by an arrow pointing to infinity at t = 0 , “1” next to the arrow

represents the area of the impulse.
 Fig. 1.21 Continuous-time unit impulse

1.9 Sampling property of the continuous-time unit impulse:

Or more generally,

Example:

Consider the discontinuous signal x (t)

Fig. 1.22 The discontinuous signal and its derivative.

Note that the derivative of a unit step with a discontinuity of size of k gives rise to an impulse
of area k at the point of discontinuity.
1.10 Continuous-Time and Discrete-Time Systems
A system can be viewed as a process in which input signals are transformed by the system or
cause the system to respond in some way, resulting in other signals as outputs. Examples

Fig. 1. 23 Examples of systems. (a) A system with input voltage v s (t) and output voltagev0(t)
.
(b) A system with input equal to the force f(t ) and output equal to the velocity v( t) .

A continuous-time system is a system in which continuous-time input signals are applied and
results in continuous-time output signals.

A discrete-time system is a system in which discrete-time input signals are applied and results
in discrete-time output signals.

1.10.1 Simple Examples of Systems

Example 1: Consider the RC circuit in Fig. 23 (a).

The current i(t ) is proportional to the voltage drop across the resistor:
The current through the capacitor is

Equating the right-hand sides of both the above equations, we obtain a differential equation
describing the relationship between the input and output:

Example 2: Consider the system in Fig. 23 (b), where the force f(t ) as the input and the velocity
v( t) as the output. If we let m denote the mass of the car and pv the resistance due to friction.
Equating the acceleration with the net force divided by mass, we obtain

It is first-order linear differential equations of the form:

Example 3: Consider a simple model for the balance in a bank account from month to month.
Let y [n] denote the balance at the end of nth month, and suppose that y[n] evolves from month
to month according the equation:

y [n] = 1.01y[n -1] + x [n] ,

or

y [n] -1.01y[n -1] = x [n] ,

where x [n] is the net deposit (deposits minus withdraws) during the nth month 1.01y[n -1]
models the fact that we accrue 1% interest each month.
Some conclusions:
· Mathematical descriptions of systems have great deal in common;
· A particular class of systems is referred to as linear, time-invariant systems.
· Any model used in describing and analyzing a physical system represents an idealization of
the system.

1.11 Interconnects of Systems

Fig. 1.24 Interconnection of systems. (a) A series or cascade interconnection of two systems;
(b) A parallel interconnection of two systems;

(c) Combination of both series and parallel systems.

Fig. 1.25 Feedback interconnection.

 Fig. 1.26 A feedback electrical amplifier.
1.12 Basic System Properties

1.12.1 Systems with and without Memory

A system is memoryless if its output for each value of the independent variable as a given time
is dependent only on the input at the same time. For example:
y[n] =(2 x[n] –x2[n])2
is memoryless.

A resistor is a memoryless system, since the input current and output voltage has the
relationship,

v (t) = R i(t ) ,

where R is the resistance.

One particularly simple memoryless system is the identity system, whose output is identical to
its input, that is
 y(t)=x(t) or y[n]=x[n]
An example of a discrete-time system with memory is an accumulator or summer.

Or

y[n]-y[n-1]=x[n]
Another example is a delay
y[n]=x[n-1]

A capacitor is an example of a continuous-time system with memory

where C is the capacitance

1.12.2 Invertibility and Inverse System
A system is said to be invertible if distinct inputs leads to distinct outputs.

Fig. 1.27Concept of an inverse system.

Examples of non-invertible systems:
y [n] = 0 ,

the system produces zero output sequence for any input sequence.
y( t) = x2( t) ,

in which case, one cannot determine the sign of the input from the knowledge of the output.
Encoder in communication systems is an example of invertible system, that is, the input to the
encoder must be exactly recoverable from the output.

1.12.3 Causality
A system is causal if the output at any time depends only on the values of the input at present
time and in the past. Such a system is often referred to as being nonanticipative, as the system
output does not anticipate future values of the input.

The RC circuit in Fig. 23 (a) is causal, since the capacitor voltage responds only to the present
and past values of the source voltage. The motion of a car is causal, since it does not anticipate
future actions of the driver.
The following expressions describing systems that are not causal:

y [n] = x [n] - x[ n +1],

and

y (t) = x( t +1)
All memoryless systems are causal, since the output responds only to the current value of input.

Example: Determine the Causality of the two systems:

(1) y [n] = x [-n]

(2) y (t )= x (t) cos(t +1)

Solution: System (1) is not causal, since when n < 0 , e.g. n = -4 , we see that y [-4] = x [4] , so
that the output at this time depends on a future value of input.

System (2) is causal. The output at any time equals the input at the same time multiplied by a
number that varies with time.
1.12.4 Stability

A stable system is one in which small inputs leads to responses that do not diverge. More
formally, if the input to a stable system is bounded, then the output must be also bounded and
therefore cannot diverge.
Examples of stable systems and unstable systems:
The above two systems are stable system.

The accumulator y[n] = ∑𝑛𝐾=−∞ 𝑥[𝑘] is not stable, since the sum grows continuously even if
x [n] is bounded.

Check the stability of the two systems:

• S1; y( t) = tx (t) ;

• S2: y(t)= e x(t)

• S1 is not stable, since a constant input x (t)= 1, yields y(t ) = t , which is not bounded –
no matter what finite constant we pick,| y( t)| will exceed the constant for some t.

• S2 is stable. Assume the input is bounded |x (t)| < B , or - B < x (t) < B for all t.

We then see that y (t) is bounded e-B < y(t) < eB

1.12.5 Time Invariance

A system is time invariant if a time shift in the input signal results in an identical time shift in
the output signal. Mathematically, if the system output is y (t) when the input is x( t) , a
timeinvariant system will have an output of y(t-t0) when input is x(t-t0).
Examples: ·

The system y (t) = sin[x (t)] is time invariant.

The system y [n] = n x[n] is not time invariant. This can be demonstrated by using
counterexample. Consider the input signal x1[n] =δ[n] , which yields y1[ n] =0 . However, the
input x2[n]= δ[n-1] yields the output y2[n]=n δ[n-1]. Thus, while x2[n] is the shifted version
of x1[n] , y2[n] is not the shifted version of y1[ n] .
The system y (t) = x( 2t) is not time invariant.
To check using counter example. Consider x1(t) shown in Fig. 1.30 (a), the resulting output
y1( t) is depicted in Fig. 1.30 (b). If the input is shifted by 2, that is, consider x2 (t)= x1 (t -2) ,
as shown in Fig. 1.30 (c), we obtain the resulting output y2(t)= x2(2t) shown in Fig. 1.30 (d). It
is clearly seen that y2(t) ≠ y1(t-2), so the system is not time invariant.
 Fig. 1.28 Inputs and outputs of the system y( t) = x(2t)
1.12.6 Linearity
The system is linear if

• The response to x1(t)+ x2(t) is y1(t)+y2(t) - additivity property

• The response to ax1(t) is ay1(t) - scaling or homogeneity property.
• The two properties defining a linear system can be combined into a single statement:
• Continuous time: ax1(t)+bx2(t) → a y1(t) + b y2(t)
• Discrete time: ax1[n] +b x2[n] → ay1[n] + b y2[n]
• Here a and b are any complex constants.
• Superposition property: If xk [n], k = 1,2,3…… are a set of inputs with corresponding
outputs yk [n], k = 1,2,3…. , then the response to a linear combination of these inputs
given by

Is

which holds for linear systems in both continuous and discrete time.
For a linear system, zero input leads to zero output.
 Examples:
• The system y (t) = t x(t) is a linear system.
• The system y(t) = x2(t) is not a liner system. ·
• The system y [n][ = Re{x [n] }, is additive, but does not satisfy the homogeneity, so
it is not a linear system.
• The system y[ n][ = 2x [n] + 3 is not linear. y [n] = 3 if x [n] = 0 , the system violates
the “zeroin/zero-out” property. However, the system can be represented as the sum
of the output of a linear system and another signal equal to the zero-input response of
the system. For system y [n]= 2x[ n] + 3, the linear system is
x[n] → 2 x[n]
and the zero-input response is
y0[n]=3
as shown in Fig. 1.29.

Fig. 1.29 Structure of an incrementally linear system. y0(t) is the zero-input

response of the system.

The system represented in Fig. 1.29 is called incrementally linear system. The system
responds linearly to the changes in the input.

The overall system output consists of the superposition of the response of a linear
system with a zero-input response.

SUMMARY

Signals are represented mathematically as functions of one or more independent variables.

There are two types of signals: continuous-time signals and discrete-time signals.
The variable of time is continuous in case of Continuous-time signal.
The variable of time is discrete in case of Discrete-time signal.
In many situations, it is important to consider signals related by a modification of the
independent variable. These modifications will usually lead to reflection, scaling, and shift.

A periodic continuous-time signal x (t) has the property that there is a positive value of T for
which x (t) = x (t + T) for all t
Any signal can be decomposed into a sum of two signals, one of which is even and one of
which is odd.
The sinusoidal signal is also a periodic signal with a fundamental period of T0 .
The continuous-time unit impulse can also be considered as the first derivative of the
continuous time unit step.
The continuous-time unit step is the running integral of the unit impulse.

A continuous-time system is a system in which continuous-time input signals are applied and
results in continuous-time output signals.
A discrete-time system is a system in which discrete-time input signals are applied and results
in discrete-time output signals.
A system is memoryless if its output for each value of the independent variable as a given time
is dependent only on the input at the same time.
A system is said to be invertible if distinct inputs leads to distinct outputs.
A system is causal if the output at any time depends only on the values of the input at present
time and in the past. Such a system is often referred to as being nonanticipative, as the system
output does not anticipate future values of the input.
A stable system is one in which small inputs leads to responses that do not diverge. More
formally, if the input to a stable system is bounded, then the output must be also bounded and
therefore cannot diverge.

A system is time invariant if a time shift in the input signal results in an identical time shift in
the output signal. Mathematically, if the system output is y (t) when the input is x( t) , a time
invariant system will have an output of y(t-t0) when input is x(t-t0).

The system is linear if

The response to x1(t)+ x2(t) is y1(t)+y2(t) - additivity property

The response to ax1(t) is ay1(t) - scaling or homogeneity property.

Books

1. Digital Signal Processing by S. Salivahanan, C. Gnanapriya Second Edition, TMH

References

1. Digital Signal Processing by Sanjit K. Mitra, Third Edition, TMH

2. Signals and systems by A Anand Kumar (PHI) 2011
3. Signals and Systems by Alan V. Oppenheim and Alan S. Willsky with S. Hamid Nawab,
Second Edition, PHI (EEE)
4.Digital Signal Processing by Apte, Second Edition, Wiley India.
 UNIT-1
Chapter-2
FOURIER SERIES
2.0 Objectives

2.1 Introduction To Fourier Series

2.2 Goal - Fourier Analysis

2.3 Trigonometric Fourier Series

2.4 Fourier Series over Other Intervals

2.5 Representation of Aperiodic Signals: The Continuous-Time Fourier Transform

2.5.1 Development of The Fourier Transform Representation Of An Aperiodic Signal
2.5.2 Convergence of Fourier Transform

2.5.3 Examples of Continuous-Time Fourier Transform

2.6 The Fourier Transform for Periodic Signals

2.7 Properties of the Continuous-Time Fourier Transform

2.7.1 Linearity

2.7.2 Time Shifting

2.7.3 Conjugation and Conjugate Symmetry

2.7.4 Differentiation and Integration

2.7.5 Time and Frequency Scaling

2.7.6 Duality
2.7.7 Parseval’s Relation
2.8 The Convolution Properties

2.9 The Multiplication Property

2.10 Summary of Fourier Transform Properties And Basic Fourier Transform Pairs
2.0 OBJECTIVES
• Understand Trigonometric Fourier series components
• Periodic Fourier series components
• Understand the properties of Fourier transform

2.1INTRODUCTION TO FOURIER SERIES

We will now turn to the study of trigonometric series. You have seen that functions have series
representations as expansions in powers of x, or x − a, in the form of Maclaurin and Taylor
series. Recall that the Taylor series expansion is given by

where the expansion coefficients are determined as

From the study of the heat equation and wave equation, we have found that there are infinite
series expansions over other functions, such as sine functions. We now turn to such expansions
and in the next chapter we will find out that expansions over special sets of functions are not
uncommon in physics. But, first we turn to Fourier trigonometric series.

We will begin with the study of the Fourier trigonometric series expansion

We will find expressions useful for determining the Fourier coefficients {an, bn} given a
function f(x) defined on [−L, L]. We will also see if the resulting infinite series reproduces f(x).
However, we first begin with some basic ideas involving simple sums of sinusoidal functions.

There is a natural appearance of such sums over sinusoidal functions in music. A pure note can
be represented as
y(t) = A sin(2π f t)

where A is the amplitude, f is the frequency in hertz (Hz), and t is time in seconds. The
amplitude is related to the volume of the sound. The larger the amplitude, the louder the sound.
In Figure 2.1 we show plots of two such tones with f = 2 Hz in the top plot and f = 5 Hz in the
bottom one.

In these plots you should notice the difference due to the amplitudes and the frequencies. You
can easily reproduce these plots and others in your favorite plotting utility.
As an aside, you should be cautious when plotting functions, or sampling data. The plots you
get might not be what you expect, even for a simple sine function.
 Figure 2.1: Plots of y(t) = A sin(2π f t) on [0, 5] for f = 2 Hz and f = 5 Hz.

In Figure 2.2 we show four plots of the function y(t) = 2 sin(4πt). In the top left you see a proper
rendering of this function. However, if you use a different number of points to plot this
function, the results may be surprising. In this example we show what happens if you use N =
200, 100, 101 points instead of the 201 points used in the first plot. Such disparities are not
only possible when plotting functions, but are also present when collecting data. Typically,
when you sample a set of data, you only gather a finite amount of information at a fixed rate.
This could happen when getting data on ocean wave heights, digitizing music and other audio
to put on your computer, or any other process when you attempt to analyze a continuous signal.

Figure 2.2: Problems can occur while plotting. Here we plot the function y(t) = 2 sin 4πt
using N = 201, 200, 100, 101 points.
Next, we consider what happens when we add several pure tones. After all, most of the sounds
that we hear are in fact a combination of pure tones with different amplitudes and frequencies.
In Figure 2.3 we see what happens when we add several sinusoids. Note that as one adds more
and more tones with different characteristics, the resulting signal gets more complicated.
However, we still have a function of time.

Figure 2.3: Superposition of several sinusoids.

Given a function f(t), can we find a set of sinusoidal functions whose sum converges to f(t)?”
Looking at the superposition in Figure 2.3, we see that the sums yield functions that appear to
be periodic. This is not to be unexpected. We recall that a periodic function is one in which the
function values repeat over the domain of the function. The length of the smallest part of the
domain which repeats is called the period. We can define this more precisely: A function is
said to be periodic with period T if f(t + T) = f(t) for all t and the smallest such positive number
T is called the period.
2.2 GOAL - FOURIER ANALYSIS

Given a signal f(t), we would like to determine its frequency content by finding out what
combinations of sines and cosines of varying frequencies and amplitudes will sum to the given
function. This is called Fourier Analysis.

2.3 TRIGONOMETRIC FOURIER SERIES

As we have seen in the last section, we are interested in finding representations of functions in
terms of sines and cosines. Given a function f(x) we seek a representation in the form

Notice that we have opted to drop the references to the time-frequency form of the phase. This
will lead to a simpler discussion for now and one can always make the transformation nx = 2π
fnt when applying these ideas to applications.
The series representation in Equation is called a Fourier trigonometric series. We will simply
refer to this as a Fourier series for now.

Figure 2.4: Plot of the function f(t) defined on [0, 2π] and its periodic extension.

The set of constants a0, an, bn, n = 1, 2, . . . are called the Fourier coefficients. The constant
term is chosen in this form to make later computations simpler, though some other authors
choose to write the constant term as a0. Our goal is to find the Fourier series representation
given f(x). Having found the Fourier series representation, we will be interested in determining
when the Fourier series converges and to what function it converges.
 Figure 2.5: Superposition of several sinusoids.

Looking at the superpositions in Figure 2.5, we see that the sums yield functions that appear to
be periodic. This is not to be unexpected. We recall that a periodic function is one in which the
function values repeat over the domain of the function. The length of the smallest part of the
domain which repeats is called the period. We can define this more precisely: A function is
said to be periodic with period T if f(t + T) = f(t) for all t and the smallest such positive number
T is called the period. For example, we consider the functions used in Figure 3.3. We began
with y(t) = 2 sin(4πt). Recall from your first studies of trigonometric functions that one can
determine the period by dividing the coefficient of t into 2π to get the period. In this case we
have

From our discussion in the last section, we see that The Fourier series is periodic. The periods
of cos nx and sin nx are 2π n . Thus, the largest period, T = 2π, comes from the n = 1 terms and
the Fourier series has period 2π. This means that the series should be able to represent functions
that are periodic of period 2π. While this appears restrictive, we could also consider functions
that are defined over one period. we can show a function defined on [0, 2π]. In the same figure,
we show its periodic extension. These are just copies of the original function shifted by the
period and glued together. The extension can now be represented by a Fourier series and
restricting the Fourier series to [0, 2π] will give a representation of the original function.
Therefore, we will first consider Fourier series representations of functions defined on this
interval. Note that we could just as easily considered functions defined on [−π, π] or any
interval of length 2π. We will consider more general intervals later in the chapter.

Fourier Coefficients Theorem 2.1. The Fourier series representation of f(x) defined on [0, 2π],
when it exists, is given by equation with Fourier coefficients

These expressions for the Fourier coefficients are obtained by considering special integrations
of the Fourier series. We will now derive the an integrals in equation. We begin with the
computation of a0. Integrating the Fourier series term by term in Equation above, we have

We will assume that we can integrate the infinite sum term by term. Then
we will need to compute

From these results we see that only one term in the integrated sum does not vanish leaving

This confirms the value for a02. Next, we will find the expression for an. We multiply the
Fourier series above by cos mx for some positive integer m. This is like multiplying by cos
2x, cos 5x, etc. We are multiplying by all possible cos mx functions for different integers m all
at the same time. We will see that this will allow us to solve for the an’s.

We find the integrated sum of the series times cos mx is given by

Integrating term by term, the right side becomes

2𝜋
We have already established that ∫0 cos 𝑚𝑥 𝑑𝑥 = 0 which implies that the first term
vanishes. Next we need to compute integrals of products of sines and cosines. This requires
that we make use of some of the trigonometric identities listed . For quick reference, we list
these here.
Useful Trigonometric Identities

2𝜋
We first want to evaluate ∫0 cos 𝑛𝑥 𝑐𝑜𝑠 𝑚𝑥 𝑑𝑥. We do this by using the

There is one caveat when doing such integrals. What if one of the denominators m ± n vanishes?
For this problem m + n ≠ 0, since both m and n are positive integers. However, it is possible
for m = n. This means that the vanishing of the integral can only happen when m ≠ n. So, what
can we do about the m = n case? One way is to start from scratch with our integration. (Another
way is to compute the limit as n approaches m in our result and use L’Hopital’s Rule.)
 2𝜋
For n = m we have to compute ∫0 cos 2 𝑚𝑥 𝑑𝑥 . This can also be handled using a
trigonometric identity. Using the half angle formula, with θ = mx, we find

To summarize, we have shown that

This holds true for m, n = 0, 1, . . . . [Why did we include m, n = 0?] When we have such a
set of functions, they are said to be an orthogonal set over the integration interval. A set of
(real) functions {φn(x)} is said to be orthogonal on [a, b] if

Furthermore, if we also have that

these functions are called orthonormal.

The set of functions {𝑐𝑜𝑠𝑛𝑥)∞𝑛=0 are orthogonal on [0, 2π]. Actually, they are orthogonal on
any interval of length 2π. We can make them orthonormal by dividing each function by √ π as
indicated by Equation .
This is sometimes referred to normalization of the set of functions. The notion of
orthogonality is actually a generalization of the orthogonality of vectors in finite dimensional
𝑏
vector spaces. The integral ∫𝑎 𝑓 (𝑥 ) 𝑓(𝑥 )𝑑𝑥 is the generalization of the dot product, and is
called the scalar product of f(x) and g(x), which are thought of as vectors in an infinite
dimensional vector space spanned by a set of orthogonal functions.
2𝜋
we still have to evaluate ∫0 sin 𝑛𝑥 cos 𝑚𝑥 𝑑𝑥. We can use the trigonometric identity
involving products of sines and cosines, Setting A = nx and B = mx,
That

So,
For these integrals we also should be careful about setting n = m. In this special case, we have
the integrals

Finally, we can finish evaluating the expression in Equation. We have determined that all but
one integral vanishes. In that case, n = m. This leaves us with

Solving for am gives

Since this is true for all m = 1, 2, . . . , we have proven this part of the theorem. The only part
left is finding the bn’s This will be left as an exercise for the reader.

We now consider examples of finding Fourier coefficients for given functions. In all of these
cases we define f(x) on [0,2 П]

Example 2.1. f(x) = 3 cos 2x, x ∈ [0, 2π]. We first compute the integrals for the Fourier
coefficients.

The integrals for a0, an, n ≠ 2, and bn are the result of orthogonality. For a2, the integral can
be computed as follows:
Therefore, we have that the only nonvanishing coefficient is a2 = 3. So there is one term and
f(x) = 3 cos 2x.

Well, we should have known the answer to the last example before doing all of those integrals.
If we have a function expressed simply in terms of sums of simple sines and cosines, then it
should be easy to write down the Fourier coefficients without much work. This is seen by
writing out the Fourier series,

For the last problem, f(x) = 3 cos 2x. Comparing this to the expanded Fourier series, one can
immediately read off the Fourier coefficients without doing any integration. In the next
example we emphasize this point.

Example 2.2. f(x) = sin2 x, x ∈ [0, 2π].

We could determine the Fourier coefficients by integrating as in the last example. However,
it is easier to use trigonometric identities. We know that

There are no sine terms, so bn = 0, n = 1, 2, . . . . There is a constant term, implying a0/2 =

1/2. So, a0 = 1. There is a cos 2x term, corresponding to n = 2, so a2 = −1 2 . That leaves an =
0 for n ≠ 0, 2. So, a0 = 1, a2 = −1 2 , and all other Fourier coefficients vanish

Example 2.3. f(x) = 1, 0 < x < π, −1, π < x < 2π, .

Figure 2.6: Plot of discontinuous function in Example 2.3

We have found the Fourier coefficients for this function. Before inserting them into the Fourier
series , we note that cos nπ = (−1) n . Therefore,

So, half of the bn’s are zero. While we could write the Fourier series representation as

we could let n = 2k − 1 in order to capture the odd numbers only. The answer can be written
as
Having determined the Fourier representation of a given function, we would like to know if
the infinite series can be summed; i.e., does the series converge? Does it converge to f(x)?
We will discuss this question later in the chapter after we generalize the Fourier series to
intervals other than for x ∈ [0, 2π].

2.4 FOURIER SERIES OVER OTHER INTERVALS

In many applications we are interested in determining Fourier series representations of
functions defined on intervals other than [0, 2π]. In this section we will determine the form of
the series expansion and the Fourier coefficients in these cases. The most general type of
interval is given as [a, b]. However, this often is too general. More common intervals are of the
form [−π, π], [0, L], or
[−L/2, L/2]. The simplest generalization is to the interval [0, L]. Such intervals arise often in
applications. For example, for the problem of a one dimensional string of length L we set up
the axes with the left end at x = 0 and the right end at x = L. Similarly for the temperature
distribution along a one dimensional rod of length L we set the interval to x ∈ [0, 2π]. Such
problems naturally lead to the study of Fourier series on intervals of length L. We will see later
that symmetric intervals, [−a, a], are also useful. Given an interval [0, L], we could apply a
transformation to an interval of length 2π by simply rescaling the interval. Then we could apply
this transformation to the Fourier series representation to obtain an equivalent one useful for
functions defined on [0, L].

Figure 2.7: A sketch of the transformation between intervals x ∈ [0, 2π] and t ∈ [0, L]
We define x ∈ [0, 2π] and t ∈ [0, L]. A linear transformation relating these intervals is simply
x = 2πt L as shown in Figure 2.7. So, t = 0 maps to x = 0 and t = L maps to x = 2π. Furthermore,
this transformation maps f(x) to a new function g(t) = f(x(t)), which is defined on [0, L]. We
will determine the Fourier series representation of this function using the representation for
f(x) from the last section. Recall the form of the Fourier representation for f(x) in Equation

Inserting the transformation relating x and t, we have

This gives the form of the series expansion for g(t) with t ∈ [0, L]. But, we still need to
determine the Fourier coefficients. Recall, that
We need to make a substitution in the integral of x = 2πt L . We also will need to transform the
differential, dx = 2π L dt. Thus, the resulting form for the Fourier coefficients is

Similarly, we find that

We note first that when L = 2π we get back the series representation that we first studied. Also,
the period of cos 2nπt L is L/n, which means that the representation for g(t) has a period of L
corresponding to n = 1. At the end of this section we present the derivation of the Fourier series
representation for a general interval for the interested reader.
At this point we need to remind the reader about the integration of even and odd functions on
symmetric intervals. We first recall that f(x) is an even function if f(−x) = f(x) for all x. One
can recognize even functions as they are symmetric with respect to the y-axis as shown in
Figure 2.8

Figure 2.8: Area under an even function on a symmetric interval, [−a, a].
If one integrates an even function over a symmetric interval, then one has that

One can prove this by splitting off the integration over negative values of x, using the
substitution x = −y, and employing the evenness of f(x). Thus,
This can be visually verified by looking at Figure 2.8. A similar computation could be done for
odd functions. f(x) is an odd function if f(−x) = −f(x) for all x. The graphs of such functions
are symmetric with respect to the origin as shown in Figure 2.9. If one integrates an odd
function over a symmetric interval, then one has that

Odd Functions

Figure 2.9: Area under an odd function on a symmetric interval, [−a, a].

Example 2.4.
Let f(x) = |x| on [−π, π] We compute the coefficients, beginning as usual with a0. We have,
using the fact that |x| is an even function,

We continue with the computation of the general Fourier coefficients for f(x) = |x| on [−π, π].
We have

Here we have made use of the fact that |x| cos nx is an even function. In order to compute the
resulting integral, we need to use integration by parts ,

1
by letting u = x and dv = cos nx dx. Thus, du = dx and v = ʃ dv = 𝑛 sin nx.
Continuing with the computation, we have

Here we have used the fact that cos nπ = (−1) n for any integer n. This leads to a factor (1 −
(−1) n ). This factor can be simplified as
 4
So, an = 0 for n even and an = − for n odd. Computing the bn’s is simpler. We note that
πn 2
we have to integrate |x| sin nx from x = −π to π. The integrand is an odd function and this is a
symmetric interval. So, the result is that bn = 0 for all n. Putting this all together, the Fourier
series representation of f(x) = |x| on [−π, π] is given as

While this is correct, we can rewrite the sum over only odd n by reindexing. We let n = 2k − 1
for k = 1, 2, 3, . . . . Then we only get the odd integers. The series can then be written as

Throughout our discussion we have referred to such results as Fourier representations. We have
not looked at the convergence of these series. Here is an example of an infinite series of
functions. What does this series sum to? We show in Figure 2.10 the first few partial sums.
They appear to be converging to f(x) = |x| fairly quickly. Even though f(x) was defined on [−π,
π] we can still evaluate the Fourier series at values of x outside this interval. In Figure 2.11, we
see that the representation agrees with f(x) on the interval [−π, π]. Outside this interval we have
a periodic extension of f(x) with period 2π. Another example is the Fourier series representation
of f(x) = x on [−π, π] This is determined to be

As seen in Figure 2.12 we again obtain the periodic extension of the function. In this case we
needed many more terms. Also, the vertical parts of the

Figure 2.10: Plot of the first partial sums of the Fourier series representation for f(x) = |x|.
Figure 2.11: Plot of the first 10 terms of the Fourier series representation for f(x) = |x| on the
interval [−2π, 4π].

Figure 2.12: Plot of the first 10 terms and 200 terms of the Fourier series representation for
f(x) = x on the interval [−2π, 4π].

2.5 Representation of Aperiodic Signals: The Continuous-Time Fourier Transform

2.5.1 Development of the Fourier Transform Representation of an Aperiodic Signal
Starting from the Fourier series representation for the continuous-time periodic square wave:
The Fourier coefficients ak for this square wave are

or alternatively

where 2sin(ωT1 ) /ω represent the envelope of Tak ·

• When T increases or the fundamental frequency ω0 = 2П / T decreases, the envelope
is sampled with a closer and closer spacing. As T becomes arbitrarily large, the original
periodic square wave approaches a rectangular pulse.
• Tak becomes more and more closely spaced samples of the envelope, as T → ∞ , the
Fourier series coefficients approaches the envelope function.
This example illustrates the basic idea behind Fourier’s development of a representation for
aperiodic signals.
Based on this idea, we can derive the Fourier transform for aperiodic signals.

Suppose a signal x(t) with a finite duration, that is, x(t) = 0 for |t | > T1 , as illustrated in the
figure below.

• From this aperiodic signal, we construct a periodic signal ẋ(t) , shown in the figure
below.

• As T →∞ , ~x (t) = x( t) , for any infinite value of t .

• The Fourier series representation of ~x (t) is
Since ~x( t) = x( t) for |t| < T / 2 , and also, since x(t) = 0 outside this interval, so we have

• Define the envelope X( jw) of Tak as

we have for the coefficients ak ,

Then ~x( t) can be expressed in terms of X( jw), that is

• As T → ∞ , ~x (t) = x (t) and consequently, Equation becomes a representation of x(t).

• In addition, ω0 → 0 as T →∞ , and the right-hand side of Equation becomes an
integral.
We have the following Fourier transform:

2.5.2 Convergence of Fourier Transform

If the signal x(t) has finite energy, that is, it is square integrable,
Then we guaranteed that X( jw) is finite or Equation converges.
If e(t) =~ x (t) - x (t) , we have

An alternative set of conditions that are sufficient to ensure the convergence:

Condition1: Over any period, x(t) must be absolutely integrable, that is

Condition 2: In any finite interval of time, x(t) have a finite number of maxima and minima.
Condition 3: In any finite interval of time, there are only a finite number of discontinuities.
Furthermore, each of these discontinuities is finite.

2.5.3 Examples of Continuous-Time Fourier Transform

Example: consider signal x(t) e-at u(t) = , a > 0 .
From Equation,

If a is complex rather then real, we get the same result if Re{a}> 0

The Fourier transform can be plotted in terms of the magnitude and phase, as shown in the
figure below.

Example: Let x (t)= e –a|t| , a > 0

The signal and the Fourier transform are sketched in the figure below

Example:

That is, the impulse has a Fourier transform consisting of equal contributions at all
frequencies.
Example: Calculate the Fourier transform of the rectangular pulse signal
The Inverse Fourier transform is

Since the signal x(t) is square integrable,

xˆ(t) converges to x(t) everywhere except at the discontinuity, T1 t = ± , where xˆ(t) converges
to ½, which is the average value of x(t) on both sides of the discontinuity.
In addition, the convergence of xˆ(t) to x(t) also exhibits Gibbs phenomenon. Specifically, the
integral over a finite-length interval of frequencies

As W →∞ , this signal converges to x(t) everywhere, except at the discontinuities. More over,
the signal exhibits ripples near the discontinuities. The peak values of these ripples do not
decrease as W increases, although the ripples do become compressed toward the discontinuity,
and the energy in the ripples converges to zero.

Example: Consider the signal whose Fourier transform is

The Inverse Fourier transform is

Comparing the results in the preceding example and this example, we have
This means a square wave in the time domain, its Fourier transform is a sinc function. However,
if the signal in the time domain is a sinc function, then its Fourier transform is a square wave.
This property is referred to as Duality Property.

We also note that when the width of X( jw) increases, its inverse Fourier transform x(t) will be
compressed. When W → ∞ , X( jw) converges to an impulse. The transform pair with several
different values of W is shown in the figure below.

2.6 The Fourier Transform for Periodic Signals

The Fourier series representation of the signal x(t) is

It’s Fourier transform is

Example: If the Fourier series coefficients for the square wave below are given
The Fourier transform of this signal is

Figure : Fourier transform of a symmetric periodic square wave

Example:

The Fourier transforms for x (t ) = sin ω0t and x(t ) = cosω0t are shown in the figure below.
Example: Calculate the Fourier transform for signal

The Fourier series of this signal is

The Fourier transform is

The Fourier transform of a periodic impulse train in the time domain with period T is a periodic
impulse train in the frequency domain with period 2П /T , as sketched din the figure below.
2.7 Properties of The Continuous-Time Fourier Transform
2.7.1 Linearity

Then

2.7.2 Time Shifting

Then

Or

Thus, the effect of a time shift on a signal is to introduce into its transform a phase shift, namely,
-ω0t .

Example: To evaluate the Fourier transform of the signal x(t) shown in the figure below.
The signal x(t) can be expressed as the linear combination

x 1(t) and x2( t) are rectangular pulse signals and their Fourier transforms are

Using the linearity and time-shifting properties of the Fourier transform yields

2.7.3 Conjugation and Conjugate Symmetry

Then

Replacing ω by -ω , we see that

The right-hand side is the Fourier transform of x * (t).
If x(t) is real, from Equation we can get

We can also prove that if x(t) is both real and even, then X( jw) will also be real and even.

Similarly, if x(t) is both real and odd, then X( jw) will also be purely imaginary and odd.
A real function x(t) can be expressed in terms of the sum of an even function xe(t) =
Ev{x(t)}and an odd function xo (t) = Od{x(t)}. That is

Form the Linearity property,

From the preceding discussion, F{xe(t)} is real function and F{xo(t)} is purely imaginary. Thus
we conclude with x(t) real,

Example: Using the symmetry properties of the Fourier transform and the result

to evaluate the Fourier transform of the signal x(t)=e -|a|t , where a > 0 .
Since

So
2.7.4 Differentiation and Integration

Then

Example: Consider the Fourier transform of the unit step x(t) = u(t).

It is know that

Also note that

The Fourier transform of this function is

where G(0) = 1.

Example: Consider the Fourier transform of the function x(t) shown in the figure below.
From the above figure we can see that g(t) is the sum of a rectangular pulse and two impulses.

Note that G(0) = 0 , using the integration property, we obtain

It can be found X( jw) is purely imaginary and odd, which is consistent with the fact that x(t)
is real and odd.
2.7.5 Time and Frequency Scaling

Then

From the equation we see that the signal is compressed in the time domain, the spectrum will
be extended in the frequency domain.

Conversely, if the signal is extended, the corresponding spectrum will be compressed.

If a = -1, we get from the above equation,

That is, reversing a signal in time also reverses its Fourier transform.

2.7.6 Duality
The duality of the Fourier transform can be demonstrated using the following example.

The symmetry exhibited by these two examples extends to Fourier transform in general. For
any transform pair, there is a dual pair with the time and frequency variables interchanged.
Example: Consider using duality and the result

to find the Fourier transform G( jw) of the signal

Multiplying this equation by 2П and replacing t by - t , we have

Interchanging the names of the variables t and ω , we find that

Based on the duality property we can get some other properties of Fourier transform:

2.7.7 Parseval’s Relation

We have

Parseval’s relation states that the total energy may be determined either by computing the
energy per unit time 2 x(t) and integrating over all time or by computing the energy per unit
frequency |X(jw) |2 / 2П and integrating over all frequencies. For this reason, 2 X ( jw) is
often referred to as the energy-density spectrum.
2.8 The convolution properties

The equation shows that the Fourier transform maps the convolution of two signals into product
of their Fourier transforms.
H( jw), the transform of the impulse response, is the frequency response of the LTI system,
which also completely characterizes an LTI system.
Example: The frequency response of a differentiator.

From the differentiation property,

The frequency response of the differentiator is

Example: Consider an integrator specified by the equation:

The impulse response of an integrator is the unit step, and therefore the frequency response of
the system:

So we have

which is consistent with the integration property.

Example: Consider the response of an LTI system with impulse response

to the input signal

To calculate the Fourier transforms of the two functions:

Therefore,

using partial fraction expansion (assuming a ≠ b ), we have

The inverse transform for each of the two terms can be written directly. Using the linearity
property, we have

We should note that when a = b , the above partial fraction expansion is not valid. However,
with a = b , we have
2.9 The Multiplication Property

Multiplication of one signal by another can be thought of as one signal to scale or modulate the
amplitude of the other, and consequently, the multiplication of two signals is often referred to
as amplitude modulation.
Example: Let s(t) be a signal whose spectrum S( jw) is depicted in the figure below.

Also consider the signal

The spectrum of r(t) = s(t) p(t) is obtained by using the multiplication property,

which is sketched in the figure below.

From the figure we can see that the signal is preserved although the information has been
shifted to higher frequencies. This forms the basic for sinusoidal amplitude modulation systems
for communications.
Example: If we perform the following multiplication using the signal r(t) obtained in the
preceding example and p (t)= cosω0t ,
that is,
g(t) = r(t) p(t)
The spectrum of P( jw), R( jw) and G( jw) are plotted in the figure below

If we use a lowpass filter with frequency response H( jw) that is constant at low frequencies
and zero at high frequencies, then the output will be a scaled replica of S( jw). Then the output
will be scaled version of s(t)- the modulated signal is recovered.

2.10 Summary of Fourier Transform Properties and Basic Fourier Transform Pairs
System Characterized by Linear Constant-Coefficient Differential Equations An LTI system
described by the following differential equation:

which is commonly referred to as an Nth-order differential equation.

The frequency response of this LTI system

where X( jw), Y( jw) and H( jw) are the Fourier transforms of the input x(t), output y(t) and
the impulse response h(t), respectively.
Applying Fourier transform to both sides, we have

From the linearity property, the expression can be written as

From the differentiation property,

H( jw) is a rational function, that is, it is a ratio of polynomials in ( jw).

Example: Consider a stable LTI system characterized by the differential equation

The frequency response is

The impulse response of this system is then recognized as

Example: Consider a stable LTI system that is characterized by the differential equation
The frequency response of this system is

Then, using the method of partial-fraction expansion, we find that

The inverse Fourier transform of each term can be recognized as

Example: Consider a system with frequency response of

and suppose that the input to the system is

find the output response.

The output in the frequency domain is give as

Using partial-fraction expansion, we have

By inspection, we get directly the inverse Fourier transform:

SUMMARY
A function is said to be periodic with period T if f(t + T) = f(t) for all t and the smallest such
positive number T is called the period.
The Fourier series representation of f(x) defined on [0, 2π], when it exists, is given by equation
with Fourier coefficients

If one integrates an even function over a symmetric interval, then one has that
Over any period, x(t) must be absolutely integrable, that is

In any finite interval of time, x(t) have a finite number of maxima and minima.

In any finite interval of time, there are only a finite number of discontinuities.
The Fourier transform can be plotted in terms of the magnitude and phase, as

The impulse has a Fourier transform consisting of equal contributions at all frequencies.
The Fourier series representation of the signal x(t) is

It’s Fourier transform is

The Fourier transform of a periodic impulse train in the time domain with period T is a
periodic impulse train in the frequency domain with period 2П /T ,
Different properties of Fourier transforms are Linearity, Time Shifting, Conjugation and
Conjugate Symmetry, Differentiation and Integration, Time and Frequency Scaling, Duality,
Parseval’s Relation, convolution properties, Multiplication Property
Question:
Q1) Explain Trigonometric Fourier Series with example.
Q2) Explain Exponential Fourier Series with example.
Q3) Explain Convergence Of Fourier Transform.
Q4) Explain Fourier Transform For Periodic Signals
Q5) Explain Properties of Fourier Transform .
Q6) Explain Convolution properties of Fourier Transform.
Q7) Explain Multiplication properties of Fourier Transform.

Books
1. Digital Signal Processing by S. Salivahanan, C. Gnanapriya Second Edition, TMH
References

1. Digital Signal Processing by Sanjit K. Mitra, Third Edition, TMH

2. Signals and systems by A Anand Kumar (PHI) 2011

3. Signals and Systems by Alan V. Oppenheim and Alan S. Willsky with S. Hamid Nawab,
Second Edition, PHI (EEE)
4.Digital Signal Processing by Apte, Second Edition, Wiley India.
 Unit -2
Chapter -3
LAPLACE TRANSFORM
3.0 Objectives
3.1 Introduction
3.2 Definition of Laplace Transform
3.3 Convergence of Laplace Transform
3.4 Properties of ROC
3.5 Properties of Laplace Transform
3.5.1 Linearity
3.5.2 Time Shifting (Translation in Time Domain)
3.5.3 Shifting in s- Domain (Complex Translation)
3.5.4 Time Scaling
3.5.5 Differentiation in Time Domain
3.5.6 Differentiation in s- Domain
3.5.7 Convolution in Time Domain
3.5.8 Integration in Time domain
3.5.9 Integration in s- Domain
3.6 Examples of Laplace Transform
3.7 Unilateral Laplace Transform
3.7.1 Differentiation in Time Domain
3.7.2 Initial Value Theorem
3.7.3 Final Value Theorem
3.0 OBJECTIVES
• Understand Laplace transform of basic signals.
• To understand and apply properties of Laplace transform.
• To understand and apply unilateral Laplace transforms.

3.1 INTRODUCTION
Laplace transform represents continuous time signals in terms of complex exponentials i.e.
e –st
Continuous time systems are also analyzed more effectively using Laplace transform.
Laplace transform can be applied to the analysis of unstable systems also.
Types of Laplace Transform
i) Bilateral or two sided Laplace transform
ii) Unilateral or one sided Laplace transform
3.2 Definition of Laplace Transform
∞
X(s)= ∫−∞ 𝑥(𝑡) e-st dt

Here the independent variable ‘s’ is complex in nature and it is given as

s = σ + jω
Here σ is real part of ‘s’ It is called attenuation constant.
jω is the imaginary part of ‘s’ and it is called complex frequency.
The Laplace transform pair x(t) and X(s) is represented as,
x(t) X(s)
The unilateral Laplace transform is given as
∞
X(s)= ∫0 𝑥(𝑡) e-st dt

Laplace transform is mainly used for causal signals.

The Inverse Laplace transform is given as
1 𝜎+𝑗∞
x(t) = 2𝜋𝑗 ∫𝜎−𝑗∞ 𝑋(𝑠) est dt

3.3 Convergence of Laplace Transform

We know that Laplace Transform is basically the Fourier transform of x(t) e-σt .
If Fourier transform of x(t) e-σt exists , then Laplace transform of x(t) exits.
For the Fourier transform to exists , x(t) e-σt must be absolutely integrable.
∞
∫−∞ | x(t) e − σt| | dt < ∞

The range of values of σ for which Laplace transform converges is called region of convergence
or ROC.
Example 3.1:- Calculate the Laplace transform of following functions and plot their ROC
i) x(t)= eat u(t)
Solution:-

The shaded area is called Region of convergence

Since Re(s) is real part of ‘s’ i.e. σ .
Hence ROC : σ > a or Re(s) > a
ii) x(t) = -eat u(-t)
Solution

The shaded region shows ROC of s < a

Thus,

Example 3.2 Determine the Laplace transform of

i) x1(t) = e -2t u(t) – e2t u(-t)
Solution:
From above result Laplace transform of x1(t) will be

The figure shows ROC of -2 < s < 2

ii) x2(t) = 3 e -2t u(t) – e-t u(t)

Solution:

Therefore Laplace transform of x2(t) become

ROC : s > -2 and s > -1

Both the terms of converge for ROC of s > -1 .Hence ROC of Laplace transform will be s >-1

3.4 Properties of ROC

1. No poles lie in ROC.
2. ROC of the causal signal is right hand sided. It is of the form Re(s) > a.
3. ROC of the noncausal signal is left hand sided. It is of the form Re(s) < a.
4. The system is stable if its ROC includes jω axis of s-plane.

3.5 Properties of Laplace Transform

For all the properties we have,

3.5.1 Linearity
Statement: Laplace transform follows superposition principle ,i.e. it is linear

Proof:
Here ROC : R1 ᴖ R2 indicates the intersection of R1 and R2
3.5.2 Time Shifting (Translation in Time Domain)
Statement: A time shift in the signal introduces frequency shift in frequency domain.

Proof:

By substituting in above equation we have

3.5.3 Shifting in s- Domain (Complex Translation)

Statement: A shift in the frequency domain is equivalent to multiplying the time domain signal
by complex exponential.

Proof:
3.5.4 Time Scaling
Statement: Expansion in time domain is equivalent to compression in frequency domain and
vice versa

𝑅
ROC : 𝑎

Proof:

Similar procedure can be repeated for Laplace transform of x(-at) . We get

The above equations can be combined as follows:

As a special case with a= -1 we have

This result shows that inverting the time axis inverts frequency axis as well as ROC.

3.5.5 Differentiation in Time Domain

Statement: Differentiation in time domain adds a zero to the system.

Proof:
Differentiate both sides of above equation with respect to ‘t’ i.e.

For multiple order derivative

3.5.6 Differentiation in s- Domain

Statement: Differentiation in s-domain corresponds to multiplying the time domain sequence
by -t

Proof:

Differentiating above equation with respect to ‘s’

For multiple order differentiation in s-domain,

3.5.7 Convolution in Time Domain

Statement: The Laplace transform of convolution of two functions is equivalent to
multiplication of their Laplace transforms.

ROC : containing R1 ᴖ R2
Proof:

Taking Laplace transform of both the sides,

Changing the order of integration,

3.5.8 Integration in Time domain

Statement: Time domain integration adds a pole to the system.
𝑡 𝑋(𝑠)
∫−∞ 𝑥(𝜏)𝑑𝜏 → , ROC : R ᴖ [ Re(s) > 0 ]
𝑠

Proof:
Hence above equation becomes

Taking Laplace transform of both sides,

For multiple order of integration,

3.5.9 Integration in s- Domain

Statement:
Frequency domain integration corresponds to dividing the time domain signal by t

Proof:
Changing the order of integration and rearranging the terms,

3.6 Examples of Laplace Transform

Example 3.3 : Obtain the Laplace transform and ROC of following signals :
i) x(t) = u(t)
Solution:

Here e –s x ∞ = e -∞ = 0
If s > 0 . Then above equation will be,
 ROC
Thus

ii) x(t) = δ(t)

Solution:

Here use,

With t0 = 0 , the above equation becomes

This is convergent for all values of s.

iii) x(t)= r(t)

Solution :
Integrating by parts,

Similarly,

iv) x(t) = t e-at u(t)

Solution :

By differentiation in s-domain property,

Similarly,

Example 3.4: Obtain the Laplace transform of following signals :

i) x(t) = A sin ω0 t
Solution :

We know that ,
The above equation can be written as ,

Here s > jω can be written as σ+ jω > 0 +jω , hence σ >0 .

Therefore ROC will be Re(s) or σ >0.

ii) x(t) = A cos ω0 t u(t)

Solution:

Therefore,
3.7 Unilateral Laplace Transform
The unilateral Laplace transform is given as ,
∞
X(s) = ∫0− 𝑥(𝑡) e-st dt

The lower limit is taken as 0- to indicate that initial conditions at t=0 are also considered.
Note that unilateral Laplace transform will be always convergent since ROC will be always
R.H.S. of S-plane.
3.7.1 Differentiation in Time Domain
Let x(t) → X(S) Laplace transform pair
Then,
𝑑𝑥(𝑡)
→ s X(s ) – x(0-)
𝑑𝑡

Here x(0-) is value of x(t) at t=0- . It is initial value of x(t).

Proof :
By definition of Laplace transform,

Integrating above equation by parts,

The integration term in above equation is Laplace transform of x(t) .

Hence,

We know that e -∞ = 0 and e0 =1. Hence the above equation becomes,

This property can be further expanded for multiple differentiations as follows:

3.7.2 Initial Value Theorem
Let x(t) → X(S) Laplace transform pair
Then initial value of x(t) is given as

Provided that the first derivatives of x(t) should be Laplace transformable.

Proof: From the differentiation property of Laplace transform we know that,

Let us take limit of the above equation as s→∞ ,i.e.,

Consider L.H.S of above equation i.e.,

Therefore above equation becomes

x(0-) indicates the value of x(t) just before t=0 and x(0+) indicates value of x(t) just after t=0 .
If the function x(t) is continuous at t=0 , then its value just before and after t=0 will be same
i.e.,
This equation is used to determine the initial value of x(t) and its derivative.
3.7.3 Final Value Theorem
Let x(t) → X(S) Laplace transform pair
Then initial value of x(t) is given as

Proof : From differentiation property we know that,

Let us take limit of above equation as s→0 , i.e.

Consider L.H.S of above equation,

Hence equation can be written as

Application of Initial and Final Value Theorem

The initial voltage on the capacitor or current through an indicator can be evaluated with the
help of initial value theorem.
The final charging voltage on capacitor or saturating currents through an inductor can be
evaluated with the help of final value theorem.
Example 3.5: Find f(∞) final value of function whose Laplace transform is given by
𝟓 𝟏
F(s)= 𝒔 - 𝒔−𝟒

Solution:
Final value is given as,

Example 3.6: Use the s- domain shift property and Fourier transform pair

To derive the unilateral Laplace transform of x(t) = e-at u(t) cos ω1t u(t)
Solution:

Here

By shifting in s-domain property,

Example 3.7 : Determine initial and final values of signal x(t) whose unilateral Laplace
transform :
Solution:
Initial value is given by,

Final value is given as,

SUMMARY
Laplace transform represents continuous time signals in terms of complex exponentials i.e.
e –st
Continuous time systems are also analyzed more effectively using Laplace transform.
Laplace transform can be applied to the analysis of unstable systems also.
∞
X(s)= ∫−∞ 𝑥(𝑡) e-st dt

Here the independent variable ‘s’ is complex in nature and it is given as

s = σ + jω
Here σ is real part of ‘s’ It is called attenuation constant.
jω is the imaginary part of ‘s’ and it is called complex frequency.
Types of Laplace Transform i) Bilateral or two sided Laplace transform
ii) Unilateral or one sided Laplace transform
The Inverse Laplace transform is given as
1 𝜎+𝑗∞
x(t) = 2𝜋𝑗 ∫𝜎−𝑗∞ 𝑋(𝑠) est dt

If Fourier transform of x(t) e-σt exists , then Laplace transform of x(t) exits.
For the Fourier transform to exists , x(t) e-σt must be absolutely integrable.
∞
∫−∞ | x(t) e − σt| | dt < ∞
The range of values of σ for which Laplace transform converges is called region of convergence
or ROC.
No poles lie in ROC.
ROC of the causal signal is right hand sided. It is of the form Re(s) > a.
ROC of the noncausal signal is left hand sided. It is of the form Re(s) < a.
The system is stable if its ROC includes jω axis of s-plane.
Properties of Laplace Transform are Linearity, Time Shifting, Shifting in s- Domain, Time
Scaling, Differentiation in Time Domain, Differentiation in s- Domain, Convolution in Time
Domain, Integration in Time domain, and Integration in s- Domain.

The unilateral Laplace transform is given as ,

∞
X(s) = ∫0− 𝑥(𝑡) e-st dt

The lower limit is taken as 0- to indicate that initial conditions at t=0 are also considered.
Note that unilateral Laplace transform will be always convergent since ROC will be always
R.H.S. of S-plane.

Questions:
1) Calculate Laplace transform of e-at u(t).
1
[Ans. , ROC: s > -a ]
𝑠+𝑎
2) Calculate Laplace transform of -e-at u(t).
1
[Ans. , ROC: s > -a ]
𝑠+𝑎
3) Calculate Laplace transform of -e-3t u(-t).
1
[Ans. - 𝑠+3 , ROC: s > -3 ]
4) Obtain Laplace transform of the following signals.
i. x(t) = sin (3t) u(t)
ii. x(t) = e-2t u (t + 1)

5) Find the Laplace transform of the following with ROC:

i. x(t) = u (t -5)
ii. x(t) = e5t u (-t + 3)

6) State and prove initial value theorem of Laplace transforms.

7) State and prove final value theorem of Laplace transforms.
8) Explain properties of Laplace Transform.
9) State and prove Convolution in Time Domain property.
 10) State and prove Integration in Time domain property.

11) State and prove Integration in s- Domain property.

Books
1. Digital Signal Processing by S. Salivahanan, C. Gnanapriya Second Edition, TMH
References
1. Digital Signal Processing by Sanjit K. Mitra, Third Edition, TMH
2. Signals and systems by A Anand Kumar (PHI) 2011
3. Signals and Systems by Alan V. Oppenheim and Alan S. Willsky with S. Hamid Nawab,
Second Edition, PHI (EEE)
4.Digital Signal Processing by Apte, Second Edition, Wiley India.
Unit 3 :
Chapter 4 : Z-transform
Unit Structure
4.0 Objective
4.1 Introduction
4.2 Definition of z-transform
4.2.1.1 Region of Convergence (ROC)
4.3 Properties of z-transform
4.4 Evaluation of the Inverse of z-transform
4.5 Summary
4.6 Exercise
4.7 List of References

4.0 OBJECTIVE

By the end of this chapter, student will be able to understand Z-transform as a tool for the
solution of linear constant difference equations. Also one can analyse discrete time systems in
the frequency domain.

4.1 INTRODUCTION

Z-transform simplies signal analysis by reducing the number of poles and zeros to a finite
number in z-plane. Z-transform has real and imaginary parts, whose plot is called Z-plane.
Z-transform maps(transforms) any point 𝑠 = ±𝜎 ± 𝑗𝑤 in s-plane to a corresponding point
𝑧(𝑟|𝜃) in the z-plane using the relationship :
𝑧 = 𝑒 𝑠𝑇 , where T is the sampling period

The poles and zeros of discrete time system are plotted in the complex z-plane.
Figure 4.1 shows Mapping of s-plane into z-plane for 𝑧 = 𝑒 𝑗𝑤𝑇
 Fig 4.1 Mapping of s-plane into z-plane for 𝑧 = 𝑒 𝑗𝑤𝑇

The stability of the system can be checked using pole-zero plot. Also z-transform can be used
to analyse discrete time systems for finding system transfer function and digital network
realisation.

4.2 Definition of Z-transform

The z-transform of a discrete time signal 𝑥(𝑛) can be defined as :

∞

𝑍[𝑥(𝑛)] = 𝑋(𝑧) = ∑ 𝑥(𝑛)𝑧 −𝑛

𝑛=−∞

Where z is a complex variable. This equationis also called two sided z transform.
One sided z-transform is given as :
∞

𝑍[𝑥(𝑛)] = 𝑋(𝑧) = ∑ 𝑥(𝑛)𝑧 −𝑛

𝑛=0

Inverse Z-transform
Inverse z-transform is defined as :
𝑥(𝑛) = 𝑍 −1 [𝑋(𝑧)]
Inverse z-transform is applied to recover original time domain discrete signal from ite
frequency domain signal.

Z-transform can be denoted as:

𝑍
𝑥(𝑛) ↔ 𝑋(𝑧)
Or z-transform can also be denoted as:
𝑋 (𝑧) = 𝑍[𝑥(𝑛)]

4.2.1.1 Region of Convergence (ROC)

If the output signal magnitude of thedigital signal system, 𝑥(𝑛) is to be finite, then the
magnitude of its z-transform must be finite. The Z values in the z-plane for which the
magnitude of 𝑋(𝑧) is finite is called the Region of Convergence (ROC).
ROC for 𝑋(𝑧) is the area outside the unit circle in the z-plane.
𝑧
Z-transform of the unit step 𝑢(𝑛) is 𝑋 (𝑧) = 𝑧−1 which has a zero at 𝑧 = 0 and pole at 𝑧 = 1
and the ROC is |𝑧| > 1 extending to ∞ as shown in Fig 4.2

Fig 4.2 Pole-zero plot and ROC of the Unit-Step response 𝑢(𝑛)

Properties of ROC :
1) ROC does not contain any poles
2) System stability can be checked with ROC
3) ROC also determines the types of sequence as
a) Causal or Non-causal signal
b) Finite or Infinite signal

Table 4.1) Finite Duration causal, anti-causal and two-sided signals with their ROCs.
Finite Duration Signals and their ROCs
Table 4.2) Infinte Duration causal, anticausal and two-sided signals with their ROCs
Infinite Duration Signals and their ROCs
Table 4.3) Some z-transform pairs
Sl Signal Sequence Laplace z-transform ROC
𝑥(𝑡) 𝑥(𝑛) Transform 𝑋(𝑧)
𝑋(𝑠)
1. 𝛿(𝑡) 𝛿(𝑛) 1 All z-plane

2. 𝛿(𝑡 − 𝑘) 𝛿(𝑛 − 𝑘) 𝑒−𝑘𝑠 𝑧−𝑘 |𝑧| > 0, 𝑘>0

|𝑧| < ∞, 𝑘<0

3. 𝑢(𝑡) 𝑢(𝑛) 1 1 𝑧 |𝑧| > 1

=
𝑠 1− 𝑧−1 𝑧−1

4. −𝑢(−𝑛 − 1) 1 1 𝑧 |𝑧| < 1

=
𝑠 1− 𝑧−1 𝑧−1
5. 𝑒−𝑎𝑡 𝑒−𝑎𝑛 1 1 𝑧 |𝑧| > |𝑒−𝑎 |
=
𝑠+𝑎 1 − 𝑒−𝑎 𝑧−1 𝑧 − 𝑒−𝑎

Example 4.1) Determine z-transform of following finite duration signals.

(a) 𝑥 (𝑛) = { 4 , 2 3
↑
, , 1
, 3
}

(b) 𝑥(𝑛) = { 4 , 3
, 1
, 4
, 4 0
↑
, 2
}

(c) 𝑥(𝑛) = 𝛿(𝑛)

(d) 𝑥(𝑛) = { 4 , 2
, 3
, 1
, 3
}

Solution :

(a) 𝑥 (𝑛 ) = { 4 , 2 3
↑
, , 1
, 0
}

Taking z-transform,
𝑋(𝑧) = 4𝑧 + 2 + 3𝑧 −1 + 𝑧 −2 + 3𝑧 −3
ROC entire plane except 𝑧 = 0 and 𝑧 = ∞

(b) 𝑥 (𝑛 ) = { 4 , 3
, 1
, 4
, 4 0
↑
, 2
}

Taking z-transform,
𝑋(𝑧) = 4𝑧 4 + 3𝑧 3 + 𝑧 2 + 4𝑧 1 + 4 + 2𝑧 −2
ROC entire plane except 𝑧 = 0 and 𝑧 = ∞
(c) 𝑥(𝑛) = 𝛿(𝑛), hence 𝑋 (𝑧) = 1, ROC : Entire z-plane

(d) 𝑥 (𝑛 ) = { 4 , 2
, 3
, 1
, 3
},

since there is no bottom arrow, it is assumed to be below first element i.e. 4

𝑋(𝑧) = 4𝑧 4 + 3𝑧 3 + 𝑧 2 + 4𝑧 1 + 4 + 2𝑧 −2
ROC entire plane except 𝑧 = 0 and 𝑧 = ∞

4.3 Properties of Z-transform

Table 4.4 shows some of the important properites of z-transform

Table 4.4) Z-tranform properties
Example 4.2) Find z-tranform of the signal
𝑥(𝑛) = 𝛿 (𝑛 + 1) + 2𝛿(𝑛 − 1)
Solution : Taking z-transform of the given signal,
𝑋(𝑧) = 𝑍[𝛿(𝑛 + 1) + 2𝛿 (𝑛 − 1)]
Using linearity property,
𝑋(𝑧) = 𝑍[𝛿(𝑛 + 1)] + 2𝑍[𝛿 (𝑛 − 1)]
1
𝑋(𝑧) = 𝑧 + 2𝑧 −1 = { , 0↑ , 1
} (Ans)

Example 4.3) By applying time shifting property, determine the inverse z-transform of the
signal
𝑧 −1
𝑋(𝑧) = 1−2𝑧 −1

Solution : By applying time shifting propery, we have 𝑘 = 1 and 𝑥(𝑛) = (2)𝑛 𝑢(𝑛)
 Hence, 𝑥(𝑛) = (2)(𝑛−1) 𝑢(𝑛 − 1) (Ans)

Example 4.4) Determine the convolution of two sequences

𝑥(𝑛) = {2, 0, 1} 𝑎𝑛𝑑 ℎ(𝑛) = {3, 2, 1, 2}
Solution : Taking z-transorm of the two given signals 𝑥 (𝑛) 𝑎𝑛𝑑 ℎ(𝑛),
𝑋(𝑧) = 2 + 𝑧 −2 and H(𝑧) = 3 + 2𝑧 −1 + 𝑧 −2 + 2𝑧 −3
𝑌(𝑧) = 𝑋(𝑧)𝐻 (𝑧) = (2 + 𝑧 −2 )(3 + 2𝑧 −1 + 𝑧 −2 + 2𝑧 −3 )
= (6 + 4𝑧 −1 + 2𝑧 −2 + 4𝑧 −3 + 3𝑧 −2 + 2𝑧 −3 + 𝑧 −4 + 2𝑧 −5 )
= (6 + 4𝑧 −1 + 5𝑧 −2 + 6𝑧 −3 + 𝑧 −4 + 2𝑧 −5 )
Taking Inverse z-transform,

𝑦(𝑛) = {6↑ , 4
, 5
, 6
, 1
, 2
} (Ans)

Initial Value Theorem

If 𝑥(𝑛) is a causal sequence with z-transform 𝑋(𝑧), the initial value can be
determined as :
𝑥(0) = lim 𝑥(𝑛) = lim [ 𝑋(𝑧)]
𝑛→0 |𝑧|→∞

Final Value Theorem

If 𝑋(𝑧) = 𝑍[𝑥(𝑛)] and the poles of 𝑋(𝑧) are all inside the unit circle, then
the final value of the sequence, 𝑥(∞) can be determined as :
lim 𝑥(∞) = lim 𝑥 (1 − 𝑧 −1 ) 𝑋(𝑧)
𝑛→∞ |𝑧|→1

if 𝑥(∞) exists

Example 4.5) Find Initial and final values of 𝑥(𝑛) for 𝑋(𝑧) = 1 + 2𝑧 −1 + 3𝑧 −2
Solution :
2 3
𝑥(0) = lim [ 1 + 2𝑧 −1 + 3𝑧 −2 ] = 1 + + =1
|𝑧|→∞ ∞ ∞

𝑥(∞) = lim [(1 − 𝑧 −1 )( 1 + 2𝑧 −1 + 3𝑧 −2 )]

|𝑧|→1
 = lim [ 1 + 2𝑧 −1 + 3𝑧 −2 − 𝑧 −1 − 2𝑧 −2 − 3𝑧 −3 ]
|𝑧|→1

= lim [ 1 + 𝑧 −1 + 𝑧 −2 − 3𝑧 −3 ]
|𝑧|→1

= 1+1+1−3 = 0
(Ans)

4.4 Evaluation of Inverse Z-transform

Various methods are available to take Inverse Z-transform. We will use Partial Fraction
method for taking Inverse z-transform.

Example 4.6) Find Inverse z of the following :

−12 6
𝑋 (𝑧 ) = −1
−
1 + 4𝑧 1 + 3𝑧 −1
Solution : Taking Inverse z-transform,
𝑥(𝑛) = [−12(−4)𝑛 − 6(−3)𝑛 ]𝑢(𝑛) (Ans)

Example 4.7) Find the signal 𝑥(𝑛), whose z-transform is given as

1
𝑋 (𝑧 ) =
(1 + 𝑧 −1 )(1 − 𝑧 −1 )
Solution :
1
𝑋 (𝑧 ) =
(1 + 𝑧 −1 )(1 − 𝑧 −1 )
𝑎 𝑏
𝑋 (𝑧 ) = +
(1 + 𝑧 ) (1 − 𝑧 −1 )
−1

Equating numerators,
1 = 𝑎(1 − 𝑧 −1 ) + 𝑏(1 + 𝑧 −1 )
1 = (𝑎 − 𝑎𝑧 −1 + 𝑏 + 𝑏𝑧 −1 )
1 = 𝑎 + 𝑏 + 𝑏𝑧 −1 − 𝑎𝑧 −1
1 = (𝑎 + 𝑏) + (𝑏 − 𝑎)𝑧 −1
 Equating like terms,
(𝑎 + 𝑏) = 1, (𝑏 − 𝑎 ) = 0
Solving simultaneously,
1 1
𝑎= , 𝑏=
2 2
1 1
(2) (2)
𝑋 (𝑧 ) = +
(1 + 𝑧 −1 ) (1 − 𝑧 −1 )
Taking Inverse z-transform,
1 1
𝑥(𝑛) = [ (−1)𝑛 + (1)𝑛 ] 𝑢(𝑛)
2 2

Example 4.8) Find 𝑥(𝑛) for given 𝑋(𝑍)

2 + 3𝑧 −1
𝑋 (𝑧 ) =
5 1 1
1 + 4 𝑧 −1 + 8 𝑧 −2 − 8 𝑧 −3

Solution : Writing denominator in terms of its factors,

2 + 3𝑧 −1
𝑋 (𝑧 ) =
1 1
(1 + 𝑧 −1 )(1 + 2 𝑧 −1 )(1 − 4 𝑧 −1 )

𝑎 𝑏 𝑐
𝑋 (𝑧 ) = + +
(1 + 𝑧 ) (1 + 1 𝑧 −1 ) (1 − 1 𝑧 −1 )
−1
2 4

2 + 3𝑧 −1 8
𝑎= | =−
1 1 5
(1 + 2 𝑧 −1 )(1 − 4 𝑧 −1 )
𝑧 −1=−1

2 + 3𝑧 −1 8
𝑏= | =
1 3
(1 + 𝑧 −1 )(1 − 4 𝑧 −1 ) −1
𝑧 =−2

2 + 3𝑧 −1 14
𝑐= | =
1 1 15
(1 + 2 𝑧 −1 )(1 + 2 𝑧 −1 )
𝑧 −1 =4
 8 8 14
(− 5) (3) (15)
𝑋 (𝑧 ) = + +
(1 + 𝑧 −1 ) (1 + 1 𝑧 −1 ) (1 − 1 𝑧 −1 )
2 4

Taking Inverse z-transform,

𝑥(𝑛) = 𝑍 −1 [𝑋(𝑧)]

8 8 1 𝑛 14 1 𝑛
𝑥(𝑛) = [− (−1)𝑛 + (− ) + ( ) ] 𝑢(𝑛)
5 3 2 15 4

(Ans)

𝑧2
Example 4.9) Determine inverse z-transform of 𝑋 (𝑧) = (𝑧−𝑎)2 for ROC |𝑧| > |𝑎|

Solution : We will use Residue method.

𝑧2
𝑋 (𝑧 ) =
(𝑧 − 𝑎)2
𝑧2 𝑧 𝑛+1
𝑋 (𝑧)𝑧 𝑛−1 = 𝑧 𝑛−1
=
(𝑧 − 𝑎)2 (𝑧 − 𝑎)2
We note that, the pole is at 𝑧 = 𝑎 and order 𝑚 = 2.
The Residue of 𝑋 (𝑧)𝑧 𝑛−1 at 𝑧 = 𝑎 is calculated as :
𝑅𝑒𝑠 1 𝑑 2−1
[𝑋(𝑧)𝑧 𝑛−1 ] = lim { (𝑧 − 𝑝𝑖 )2 𝑋(𝑧)𝑧 𝑛−1 }𝑧=𝑎
𝑧=𝑎 (2 − 1)! 𝑧→𝑝𝑖 𝑑𝑧 2−1

𝑑 𝑧 𝑛+1
={ ( 𝑧 − 𝑎 )2 }
𝑑𝑧 (𝑧 − 𝑎)2 𝑧=𝑎

𝑑 𝑧 𝑛+1
={ }𝑧=𝑎
𝑑𝑧

= (𝑛 + 1)𝑎 𝑛
 𝑅𝑒𝑠
𝑥 (𝑛 ) = ∑ [𝑋(𝑧)𝑧 𝑛−1 ]
𝑧=𝑎
𝑖=1

𝑥(𝑛) = (𝑛 + 1)𝑎𝑛 𝑢(𝑛) since ROC |𝑧| > |𝑎|

(Ans)

4.5 SUMMARY

1) Z-transform is a powerful tool in digital signal analysis

2) Z-transform is used to analyze discrete-time systems for finding the transfer function,
stability and network realization of system
3) 𝑍[𝑥(𝑛)] = ∑∞
𝑛=−∞ 𝑥(𝑛)𝑧
−𝑛
, where 𝑧 = 𝑒 𝑠𝑇 , T being smapling period
4) The Z values in the z-plane for which the magnitude of 𝑋(𝑧) is finite is called the Region
of Convergence (ROC).
5) The stability of the system can be determined by using the location of poles of 𝐻(𝑧)
6) Some of the important properties of x-transform are linearity, time reversal, time shifting,
time scling, differentiation, concoulution and correlation.
7) Inverse z-transform can be obtained by either of the methods : long division method,
partial fraction method or residue method

4.6 EXERCISE

1) Define z-transform
2) Explain shift property of z-transform
3) Explain what is discrete convolution
4) Explain inverse z-transform
5) Determine z-trasnform of 𝑢(𝑛 − 5)
1 1
6) Determine the inverse z-transform of 𝑋(𝑧) = , for |𝑧| < 2
(𝑧+2)2

𝑧 2 +𝑧 1
7) Determine the causal signal 𝑥 (𝑛 ) having z-transform 𝑋(𝑧) = 1 3 1
for |𝑧| > 2
(𝑧− ) (𝑧− )
2 4

4.7 LIST OF REFERENCES

1) Digital Signal Processing, S Salivahanan, TMH

2) Digital Signal Processing, Sanjit Mitra, TMH
3) Signals and Systems, A Anand Kumar, PHI
4) Digital Signal Processing, Apte, Wiley India
Unit 4 :
Chapter 5 : Linear Time Invariant Systems
Unit Structure
5.0 Objective
5.1 Introduction
5.2 Properties of DSP System
5.3 Discrete Convolution
5.4 Solution of Linear constant coefficient Difference Equation
5.5 Frequency domain representation of Discrete Time Signals and Systems
5.6 Difference Equation and its relationship with System Functions, Impulse Response and
Frequency Response
5.7 Frequency Response
5.8 Summary
5.9 Exercise
5.10 List of References

5.0 OBJECTIVE

This chapter deals with Linear time invariant systems analysis using tool like z-transform.
Important methods like convolution, response to discrete systems by using standrd digital
inputs like impulse and unit step signals are discussed in detail. Concept of difference
equation is introduced for signal analysis.

5.1 INTRODUCTION

A typical Digital System has input 𝑥(𝑛) and has an output 𝑦(𝑛) as a response to input. The
sytem output depends on the input as well as on the system parameters. Some of the inputs
used to study digital systems are Impulse and Unit Step Signals, among others. Similarly the
Systems can also be classified as Liner Time Variant, Time Invariant and others.

5.2 PROPERTIES OF DSP SYSTEMS

Various system properties as discussed in this chapter are linearity, time invariance, causality
and stability.
Linearity
𝐹 [𝑎1 𝑥1 (𝑛) ± 𝑎2 𝑥2 (𝑛)] = 𝑎1 𝐹[𝑥1 (𝑛)] ± 𝑎2 𝐹[𝑥2 (𝑛)] = 𝑎1 𝑦1 (𝑛) ± 𝑎2 𝑦2 (𝑛)
Where F is an operator
Example 5.1) Check if the system 𝑭[𝒙(𝒏)] = 𝟑 𝒏 𝒙(𝒏) + 𝟒 is linear :
Solution :
𝐹 [𝑥1 (𝑛) + 𝑥2 (𝑛)] = 3 𝑛 [𝑥1 (𝑛) + 𝑥2 (𝑛)] + 4

𝐹[𝑥1 (𝑛)] + 𝐹[𝑥2 (𝑛)] = [3 𝑛 𝑥1 (𝑛) + 4] + [3 𝑛 𝑥2 (𝑛) + 4]

Since, 𝐹[𝑥1 (𝑛) + 𝑥2 (𝑛)] ≠ 𝐹 [𝑥1 (𝑛)] + 𝐹[𝑥2 (𝑛)], the systemin non-linear.

Time Invariance
A system is said to be time-invariant if the relationship between the input and output does not
change with time.
If 𝑦(𝑛) = 𝐹[𝑥(𝑛)], then 𝑦(𝑛 − 𝑘) = 𝐹 [𝑥(𝑛 − 𝑘)] = 𝑧 −𝑘 𝐹[𝑥(𝑛)]
𝑧 −𝑘 represents a single delay of 𝑘 samples.

Example 5.2) Check if system 𝑦(𝑛) = 𝑎 𝑛 𝑥(𝑛) is time invariant or not

Solution :
𝐹[𝑥(𝑛 − 𝑘)] = 𝑎 𝑛 𝑥(𝑛 − 𝑘)
The delayed repsonse is :

𝑦(𝑛 − 𝑘) = 𝑎 (𝑛 − 𝑘) [𝑥(𝑛 − 𝑘)]

Since, 𝐹 [𝑥(𝑛 − 𝑘)] ≠ 𝑦(𝑛 − 𝑘), the system is not time invariant. (Ans)

Causality
The systems in which changes in the output are only dependent on the changes in the present
and past values of the input and/or previous output values, and are not dependent on future
input values are called Causal Sytems.
The Causality condition for linear time invariant systems is given as :
ℎ(𝑛) = 0 𝑓𝑜𝑟 𝑛 < 0
Example 5.3) Check if system 𝑦(𝑛) = 𝑥(𝑛 − 2) + 𝑥(𝑛 − 3) is causal or not.
Solution :
In this system, the output is computed only on past sample values i.e. 𝑥(𝑛 − 2) and
𝑥(𝑛 − 3), the system is causal.

Example 5.4) Check if system 𝑦(𝑛) = 𝑥(𝑛 − 2) + 𝑥(𝑛 + 3) is causal or not.

Solution :
In this system, the output is computed on past sample values i.e. 𝑥(𝑛 − 2) and also
future values, i.e. 𝑥(𝑛 + 3), the system is non-causal.

Stability
A DSP is said to be stable, if system poles are given as :
|𝑝𝑖 | < 1 i.e. ∑∞
𝑛=−∞|ℎ(𝑛)| < ∞

The pole-zero plot in Fig. 5.1 shows pole position of stable and unstable systems.

Fig 5.1 Z-plane regions for stable and unstable systems

𝑧 2−𝑧+1
Example 5.5) Check stability of the system 𝐻 (𝑧) = 1
𝑧 2 −𝑧+
2

Solution :
 𝑧2 − 𝑧 + 1 𝑧2 − 𝑧 + 1
𝐻 (𝑧 ) = =
1 1 1 1 1
𝑧 2 − 𝑧 + 2 (𝑧 + 2 + 𝑗 2)(𝑧 + 2 − 𝑗 2)

1 1
Since, |2 ± 𝑗 2| < 1, the given system is stable. (Ans)

Bounded Input – Bunded Output (BIBO) stability :

A system is said to BIBO stable, if and only if every bounded input gives bounded output.
The impulse response of the system decides the BIBO stability of the linear time invariant
system.
The necessary and sufficient condition for the BIBO stability is :
∞

∑|ℎ(𝑘)| < ∞
𝑘=0

Example 5.6) Check BIBO stability of system 𝑦(𝑛) = 3𝑥(𝑛) + 5

Solution :
𝑦(𝑛) = 3𝑥(𝑛) + 5
If 𝑥(𝑛) = 𝛿(𝑛), then 𝑦(𝑛) = ℎ(𝑛)
Hence, impulse response is ℎ(𝑛) = 3𝛿(𝑛) + 5
When, 𝑛 = 0, ℎ(0) = 3𝛿(0) + 5 = 3 + 5 = 8
When, 𝑛 = 1, ℎ(1) = 3𝛿(1) + 5 = 0 + 5 = 5
so, ℎ(1) = ℎ(2) = . . . = ℎ(𝑘) = 5
therefore,
ℎ(𝑛) = 8, 𝑤ℎ𝑒𝑛 𝑛 = 0
ℎ(𝑛) = 5, 𝑤ℎ𝑒𝑛 𝑛 ≠ 0
The necessary and sufficient condition for BIBO stability is :
∞

∑|ℎ(𝑘)| < ∞
𝑘=0

therefore, ∑∞
𝑘=0|ℎ(𝑘)| = |ℎ(0)| + |ℎ(𝑎)|+. . . +|ℎ (𝑘 )|

= 8 + 5 + 5+ . . . +5+ .. .
This is diverging series, hence the given system is BIBO unstable. (Ans)
5.3 DISCRETE CONVOLUTION

Convoluting two signals in time domain is same as muliplying two signals in frequency
domain. Convolution is useful in studying analysing input signal response to the given
system. The concolution of the two signals is given as :
𝑦(𝑛) = 𝑥(𝑛) ∗ ℎ(𝑛)
∞

𝑦(𝑛) = ∑ 𝑥(𝑘) ℎ(𝑛 − 𝑘)

𝑘=0

Properties of Convolution :
a) Commutative law :
𝑥 (𝑛) ∗ ℎ(𝑛) = ℎ(𝑛) ∗ 𝑥(𝑛)

b) Associative law :
For 𝑦1 (𝑛) = 𝑥(𝑛) ∗ ℎ1 (𝑛)
and 𝑦(𝑛) = 𝑦1 (𝑛) ∗ ℎ2 (𝑛)
𝑦(𝑛) = [𝑥(𝑛) ∗ ℎ1 (𝑛)] ∗ ℎ2 (𝑛)
𝑦(𝑛) = 𝑥 (𝑛) ∗ [ℎ1 (𝑛) ∗ ℎ2 (𝑛)]
c) Distributive law :
𝑥(𝑛) ∗ [ℎ1 (𝑛) + ℎ2 (𝑛)] = 𝑥(𝑛) ∗ ℎ1 (𝑛) + 𝑥(𝑛) ∗ ℎ2 (𝑛)

1 1 0 1 1
Example 5.7) Plot the signal given by sequence { , , , , }
↑

Solution :
5.4 SOLUTION OF LINEAR CONSTANT COEFFICIENT DIFFERENCE
EQUATION

A discrete time system transforms an input sequence into an output sequence according to the
recursion formula that represents the solution of a difference equation.
The general formof the difference equation is :
𝑁 𝑀

𝑦(𝑛) = − ∑ 𝑎𝑘 𝑦(𝑛 − 𝑘) + ∑ 𝑏𝑘 𝑥(𝑛 − 𝑘)

𝑘=1 𝑘=0

Where N is the order of the differenc equation.

The solution of the dfference equation has two parts :
𝑦(𝑛) = 𝑦ℎ (𝑛) + 𝑦𝑝 (𝑛)

Where, 𝑦ℎ (𝑛) is the solution to the homogeneous difference equation and 𝑦𝑝 (𝑛) represents
the particular solution to the difference equation.

Table 5.1) Particular solution of several types of inputs

5.5 FREQUENCY DOMAIN REPRESENTATION OF DISCRETE TIME SIGNALS
AND SYSTEMS

In a discrete time invariant system, if the input is of the form 𝑒 𝑖𝜔𝑡 , the output is 𝐻(𝑤)( 𝑒 𝑖𝜔𝑡 ).
𝐻( 𝑒 𝑖𝜔𝑡 ) is a function of 𝑤, which denotes the frequency response of the system.

𝐻( 𝑒 𝑖𝜔𝑡 ) = 𝐻𝑟 ( 𝑒 𝑖𝜔𝑡 ) + 𝑗𝐻𝑖 ( 𝑒 𝑖𝜔𝑡 ) or

𝐻 ( 𝑒 𝑖𝜔𝑡 )
𝐻( 𝑒 𝑖𝜔𝑡 ) = |𝐻𝑟 ( 𝑒 𝑖𝜔𝑡 )|𝑒 𝑗∅ , where ∅ = 𝑡𝑎𝑛−1 𝐻 𝑖 ( 𝑒 𝑖𝜔𝑡)
𝑟

The input-output relation is :

𝑦(𝑛) = 𝐻( 𝑒 𝑖𝜔 )𝑒 𝑖𝜔𝑛

Example 5.8) Find transfer function of the system :

𝑦(𝑛) = −2𝑦(𝑛 − 1) − 3𝑦(𝑛 − 2) + 𝑥(𝑛) + 𝑥(𝑛 − 1)
Solution :
The given difference equation can be written as :
𝑦(𝑛) + 2𝑦(𝑛 − 1) + 3𝑦(𝑛 − 2) = 𝑥(𝑛) + 𝑥(𝑛 − 1)
The system transfer function can be written as :
1 + 𝑒 −𝑗𝜔
𝐻(𝑒 𝑗𝜔 ) =
1 + 2𝑒 −𝑗𝜔 + 3𝑒 −𝑗2𝜔
(Ans)
5.6 DIFFERENCE EQUATION AND ITS RELATIONCHIP WITH SYSTEM
FUNCTION, IMPULSE RESPONSE AND FREQUENCY RESPONSE

A causal LIT system is defined by a linear constant coefficient difference equation :

𝑁 𝑀

∑ 𝑎𝑘 𝑦(𝑛 − 𝑘) = ∑ 𝑏𝑘 𝑥(𝑛 − 𝑘)
𝑘=0 𝑘=0

𝑌(𝑧) ∑𝑀 𝑏 𝑧 −𝑘
The system function 𝐻 (𝑧) = 𝑋(𝑧) = ∑𝑁𝑘=0 𝑎𝑘 𝑧 −𝑘
𝑘=0 𝑘

IIR Systems – Infinite Impulse Response Systems

An LTI system is said to be an Infinite Impulse Response (IIR) system if its unit
sample response ℎ(𝑛) is of infinte duration. Recursive filter having feedback has an impulse
response that is theoretically continues for ever.

FIR Systems – Finite Impulse Response Systems

An LTI system is said to be a finite impulse response (FIR) system if its unit sample
response ℎ(𝑛) is of finite duration. Non-recursive filters can be FIR systems.

Example 5.9) A DSP is given as a difference equation :

𝑦(𝑛) = 0.2 𝑥(𝑛) − 0.5 𝑥(𝑛 − 2) + 0.4 𝑥(𝑛 − 3)
Digital input sequence {−1, 1, 0, −1} is applied to this DSP. Find output response.
Solution :
𝑦(𝑛) = 0.2 𝑥(𝑛) − 0.5 𝑥(𝑛 − 2) + 0.4 𝑥(𝑛 − 3)
Taking z-transform of the given difference equation,
𝑌(𝑧) = 0.2 𝑋(𝑧) − 0.5 𝑧 −2 𝑋(𝑧) + 0.4 𝑧 −3 𝑋(𝑧)
Therefore,
𝑌(𝑧)
𝐻 (𝑧 ) = = 0.2 − 0.5 𝑧 −2 + 0.4 𝑧 −3
𝑋(𝑧)
Input sequence is, 𝑥 (𝑛) = {−1, 1, 0, −1}, and its z-transform is
𝑋 (𝑧) = −1 + 𝑧 −1 − 𝑧 −3
Since, 𝑌(𝑍) = 𝐻 (𝑧). 𝑋(𝑧)
𝑌(𝑧) = {−0.2 + 0.2 𝑧 −1 + 0.5 𝑧 −2 − 1.1 𝑧 −3 + 0.4 𝑧 −4 + 0.5 𝑧 −5 − 0.4 𝑧 −6
Taking Inverse transform,
 𝑦(𝑛) = {−0.2, 0.2, 0.5, −1.1, 0.4, 0.5, −0.4}
(Ans)

Example 5.9) Determine the impulse response of the systems described by the difference
equation :
𝑦(𝑛) = 0.7𝑦(𝑛 − 1) − 0.1𝑦(𝑛 − 2) + 2𝑥(𝑛) − 𝑥(𝑛 − 2)
Solution :
The given difference equation is :
𝑦(𝑛) = 0.7𝑦(𝑛 − 1) − 0.1𝑦(𝑛 − 2) + 2𝑥(𝑛) − 𝑥(𝑛 − 2)
This equation canbe written as :
𝑦(𝑛) − 0.7𝑦(𝑛 − 1) + 0.1𝑦(𝑛 − 2) = 2𝑥(𝑛) − 𝑥(𝑛 − 2)
Input impulse is give as 𝑥 (𝑛) = 𝛿(𝑛).
Hence, 𝑦(𝑛) = 𝑢(𝑛)
Taking z-transform,
𝐻 (𝑧)[1 − 0.7𝑧 −1 + 0.1𝑧 −2 = (2 − 𝑧 −2 )

𝐻(𝑧) 2𝑧 2 − 1
=
𝑧 𝑧(𝑧 − 0.5)(𝑧 − 0.2)

𝐻(𝑧) 𝐴1 𝐴2 𝐴3
= + +
𝑧 𝑧 (𝑧 − 0.5) (𝑧 − 0.2)

𝐻(𝑧) 10 10 1 46 1
= − +
𝑧 𝑧 3 (𝑧 − 0.5) 3 (𝑧 − 0.2)

10 𝑧 46 𝑧
𝐻(𝑧) = 10 − +
3 (𝑧 − 0.5) 3 (𝑧 − 0.2)

Taking inverse z-transform,

10 46
ℎ(𝑛) = −10 𝛿(𝑛) − (0.5)𝑛 𝑢(𝑛) + (0.2)𝑛 𝑢(𝑛)
3 3
(Ans)
Example 5.10) Determine the unit step response of the systems described by the difference
equation :
𝑦(𝑛) = 0.6𝑦(𝑛 − 1) − 0.08𝑦(𝑛 − 2) + 𝑥(𝑛)
Solution :
The given difference quation of the system is :
𝑦(𝑛) = 0.6𝑦(𝑛 − 1) − 0.08𝑦(𝑛 − 2) + 𝑥(𝑛)
This can be written as :
𝑦(𝑛) − 0.6𝑦(𝑛 − 1) + 0.08𝑦(𝑛 − 2) = 𝑥(𝑛)
Taking z-transform,
1
𝑌 (𝑧)[1 − 0.6𝑧 −1 + 0.08𝑧 −2 ] =
1 − 𝑧 −1

1
𝑌 (𝑧 ) =
(1 − 𝑧 −1 )[1 − 0.6𝑧 −1 + 0.08𝑧 −2 ]

1
𝑌 (𝑧 ) =
(1 − 𝑧 −1 )(1 − 0.4𝑧 −1 )(1 − 0.2𝑧 −1 )

𝐴1 𝐴2 𝐴3
𝑌 (𝑧 ) = + +
(1 − 𝑧 ) (1 − 0.4𝑧 ) (1 − 0.2𝑧 −1 )
−1 −1

25 1 4 1 1 1
𝑌 (𝑧 ) = − +
12 (1 − 𝑧 ) 3 (1 − 0.4𝑧 ) 4 (1 − 0.2𝑧 −1 )
−1 −1

Taking inverse z-transform,

25 4 1
𝑠(𝑛) = 𝑦(𝑛) = 𝑢(𝑛) − (0.4)𝑛 𝑢(𝑛) + (0.2)𝑛 𝑢(𝑛)
12 3 4
(Ans)

5.7 FREQUENCY RESPONSE

Frquency response describes the magnitude and phase shift over a range of frequencies.

Properties of Frequency Response

Properties of frequency response of a real sequence ℎ(𝑛) are given as :

a) 𝐻(𝑒 𝑗𝜔 ) takes on values for all 𝜔

b) 𝐻(𝑒 𝑗𝜔 ) is periodic in 𝜔 with period 2𝜋

c) The magnitude response |𝐻(𝑒 𝑗𝜔 )| is an even function of 𝜔 and symmetric about 𝜋

d) The magnitude response |𝐻(𝑒 𝑗𝜔 )| is an odd function of 𝜔 and antisymmetric

about 𝜋

Frequency Response of an inter connection of Systems

Parallel connection:
When there are 𝐿 number of linear time invariant systems in time domain connected in
parallel, the impilse response ℎ(𝑛) of the resultant sytem is given as :
𝐿

ℎ( 𝑛 ) = ∑ ℎ𝑘 (𝑛 )
𝑘=1

Using, linearity property of z-transform, the frequency response of the complete system is :
𝐿

𝐻 (𝑧) = ∑ 𝐻𝑘 (𝑧) = 𝐻1 (𝑧) + 𝐻2 (𝑧)+ . . . +𝐻𝐿 (𝑧)

𝑘=1

Where 𝑧 = 𝑒 𝑗𝜔

Parallel interconnection of linear discrete time signal is shown in figure 5.2

Figure 5.2 Parallel interconnection of linear discrete time systems

Cascade connection :
The impulse respopnse of 𝐿 linear time invriant systems connected in cascade is given as :
ℎ(𝑛) = ℎ1 (𝑛) ∗ ℎ2 (𝑛) ∗ . . .∗ ℎ𝐿 (𝑛)
Using Convolution property, z-transform is obtained as :
𝐻 (𝑧) = 𝐻1 (𝑧)𝐻2 (𝑧) . . . 𝐻𝐿 (𝑧)

Figure 5.3 Cascade interconnection of linear discrete time systems

5.8 SUMMARY

1) Linear time invariant systems have properties like linearity, time-invariance and causality
which can be used in system analysis
2) A DSP is said to be stable, if system poles are given as :
|𝑝𝑖 | < 1 i.e. ∑∞
𝑛=−∞|ℎ(𝑛)| < ∞

3) A system is said to BIBO stable, if and only if every bounded input gives bounded output.
4) Convolution is useful in studying analysing input signal response to the given system.
5) A discrete time system transforms an input sequence into an output sequence according to
the recursion formula that represents the solution of a difference equation.
6) Frquency response describes the magnitude and phase shift over a range of frequencies.

5.9 EXERCISE

1) Explain what do you understand by discrete tim invariant systems

2) Explain conditions of causality and stability of a linear time invariant systems
3) What is BIBO stability?
4) Write a note on system transfer function
5) Find the stability region of the causal system
𝑧 −1
𝐻 (𝑧 ) =
1 − 𝑧 −1 − 𝑧 −2
6) Write a note on discrete convolution
7) Check if the system 𝑦(𝑛) = 𝑥(−𝑛) is causal or not
8) Explain how can the linearity of a discrete system be found out?

5.10 LIST OF REFERENCES

1) Digital Signal Processing, S Salivahanan, TMH

2) Digital Signal Processing, Sanjit Mitra, TMH
3) Signals and Systems, A Anand Kumar, PHI
4) Digital Signal Processing, Apte, Wiley India
 UNIT V
Discrete and Fast Fourier Transforms
5.0 Introduction
5.1 discrete Fourier series
5.2 Discrete time Fourier transform (DTFT)
5.3 Fast Fourier transforms (FFT)
5.4 Inverse DFT
5.5 Composite radix FFT
5.6 Fast (Sectioned) convolution
5.7 Correlation
5.0 Introduction
The DFT is important because it is the mathematical relation that is implemented
by the various Fast Fourier Transform (FFT) algorithms. In this section we will
discuss
• Relationships between periodic and finite-duration time functions
• The discrete Fourier series (DFS) for periodic time functions
• The discrete Fourier transform (DFT) for finite-duration time functions
The discrete Fourier series (DFS) is used to represent periodic time functions and
the DFT is used to represent finite-duration time functions. The two
representations are virtually identical mathematically, and they are closely related
because a finite-duration time function can be thought of as a single period of a
periodic time function. Conversely, a periodic time function can be easily
constructed from a finite-duration time function simply by repeating the finite-
duration sequence over and over again, ad infinitum.
Let’s adopt the notation used by OSYP to distinguish these functions: let x˜[n]
represent a periodic discrete-time sequence with period N, and let x[n](without
the tilde) represent a finite-duration sequence that is nonzero for 0<= n<=N-1.
We than note (trivially) that
5.1 Discrete Fourier series
• Definitions of the DFS

If a time function x˜[n] is periodic has period N, we can write

There are only N unique frequency components because for integer k, n, and r.

Comments:
1. Both the time function x˜[n] and the Fourier series coefficients X˜[k] are
periodic with period N, so they are represented by only N distinct (possibly
complex) numbers.
2. The series can be evaluated over any consecutive of values of n or k of length
N.
4. The frequency components are N equally-spaced samples of the frequencies
of the DTFT, or alternatively they represent N equally-spaced locations around
the unit circle of the z-plane.

Properties of the DFS

In general, the properties of the DFS are very similar to what we would expect
from the DTFT, except that the functions considered are periodic.

Note that in the previous two properties, shift in one domain corresponds to
multiplication in the other by a complex exponential, both of which may be
easier to evaluate when you apply the definition for WN.

The sum on the right hand side is very important for our work and is referred to
as periodic convolution or circular convolution. The operator for circular
convolution is normally written as an asterisk with a circle around it, possibly
with an accompanying number that indicates the size of the convolution (which
matters).

5.2 Discrete time Fourier transform (DTFT)

The discrete-time Fourier transform has essentially the same properties as the
continuous-time Fourier transform, and these properties play parallel roles in
continuous time and discrete time. As with the continuous-time Fourier
transform, the discrete-time Fourier transform is a complex-valued function
whether or not the sequence is real-valued. Furthermore, as we stressed in Lecture
10, the discrete-time Fourier transform is always a periodic function of fl. If x(n)
is real, then the Fourier transform is conjugate symmetric, which implies that the
real part and the magnitude are both even functions and the imaginary part and
phase are both odd functions. Thus for real-valued signals the Fourier transform
need only be specified for positive frequencies because of the conjugate
symmetry. Whether or not a sequence is real, specification of the Fourier
transform over a frequency range of 2π specifies it entirely. For a real-valued
sequence, specification over the frequency range from, for example, 0 to π is
sufficient because of conjugate symmetry.
The time-shifting property together with the linearity property plays a key role in
using the Fourier transform to determine the response of systems characterized
by linear constant-coefficient difference equations. As with continuous time, the
convolution property and the modulation property are of particular significance.
As a consequence of the convolution property, which states that the Fourier
transform of the convolution of two sequences is the product of their Fourier
transforms, a linear, time-it variant system is represented in the frequency domain
by its frequency response. This representation corresponds to the scale factors
applied at each frequency to the Fourier transform of the input. Once again, the
convolution property can be thought of as a direct consequence of the fact that
the Fourier transform decomposes a signal into a linear combination of complex
exponentials each of which is an Eigen function of a linear, time-invariant system.
The frequency response then corresponds to the eigenvalues. The concept of
filtering for discrete-time signals is a direct consequence of the convolution
property. The modulation property in discrete time is also very similar to that in
continuous time, the principal analytical difference being that in discrete time the
Fourier transform of a product of sequences is the periodic convolution.
 Figure 5.1: DTFT
5.3 Fast Fourier transforms (FFT)
The time taken to evaluate a DFT on a digital computer depends principally on
the number of multiplications involved, since these are the slowest operations.
With the DFT, this number is directly related to V (matrix multiplication of a
vector), where is the length of the transform. For most problems, is chosen to be
at least 256 in order to get a reasonable approximation for the spectrum of the
sequence under consideration – hence computational speed becomes a major
consideration. Highly efficient computer algorithms for estimating Discrete
Fourier Transforms have been developed since the mid-60’s. These are known as
Fast Fourier Transform (FFT) algorithms and they rely on the fact that the
standard DFT involves a lot of redundant calculations.

It is easy to realise that the same values of : are calculated many times as
the computation proceeds. Firstly, the integer product nk repeats for different
combinations of n and k; secondly, is a periodic function with N only
distinct values.
5.4 Inverse DFT
The inverse transform of

Is given as:

i.e. the inverse matrix is 1/N : times the complex conjugate of the original
(symmetric) matrix.

5.5 Composite radix FFT

It is not always possible to work with sequences whose length is a power of 2.
However, efficient computation of the DFT is still possible if the sequence
length may be written as a product of factors. For example, suppose that N may
be factored as follows: N=N1.N2.
We then decompose x(n) into N2 sequences of length NI and arrange these
sequences in an array as follows:

5.6 Fast (Sectioned) convolution

5.6.1 Fast Circular Convolution
Since

Cost
• Direct
➢ N2 complex multiplies.
➢ N (N - 1) complex adds.

5.6.2 Fast Linear Convolution

For linear convolution, we must zero-pad sequences so that circular wrap-around always
wraps over zeros.
 Figure 5.2: Fast Linear Convolution
To achieve linear convolution using fast circular convolution, we must use zero-padded DFTs
of length .

Figure 5.3
Choose shortest convenient N (usually smallest power-of-two greater than or
equal to L +M - 1).

5.6.3 Running Convolution

Suppose L = ∞, as in a real time filter application, or (L >> M). There are
efficient block methods for computing fast convolution.

5.6.3.1 Overlap-Save (OLS) Method

Note that if a length-M filter h (n) is circularly convolved with a length-N segment of a signal
x (n),
 Figure 5.4: Overlap –Save method
The Overlap-Save Method: Break long signal into successive blocks of N samples, each block
overlapping the previous block by M-1 samples. Perform circular convolution of each block
with filter h (m). Discard first M - 1 point in each output block, and concatenate the remaining
points to create y (n).
 Figure 5.5

5.7 Correlation
The concept of correlation can best be presented with an example. Figure 5.6 shows the key
elements of a radar system. A specially designed antenna transmits a short burst of radio wave
energy in a selected direction. If the propagating wave strikes an object, such as the helicopter
in this illustration, a small fraction of the energy is reflected back toward a radio receiver
located near the transmitter. The transmitted pulse is a specific shape that we have selected,
such as the triangle shown in this example. The received signal will consist of two parts: (1) a
shifted and scaled version of the transmitted pulse, and (2) random noise, resulting from
interfering radio waves, thermal noise in the electronics, etc. Since radio signals travel at a
known rate, the speed of light, the shift between the transmitted and received pulse is a direct
measure of the distance to the object being detected. This is the problem: given a signal of
some known shape, what is the best way to determine where (or if) the signal occurs in another
signal. Correlation is the answer. Correlation is a mathematical operation that is very similar
to convolution. Just as with convolution, correlation uses two signals to produce a third signal.
This third signal is called the cross-correlation of the two input signals. If a signal is correlated
with itself, the resulting signal is instead called the autocorrelation. The convolution machine
was presented in the last chapter to show how convolution is performed. Figure 5.6 is a similar
illustration of a correlation machine. The received signal, x[n], and the cross-correlation signal,
y[n], are fixed on the page. The waveform we are looking for, t[n], commonly called the target
signal, is contained within the correlation machine. Each sample in y[n] is calculated by
moving the correlation machine left or right until it points to the sample being worked on. Next,
the indicated samples from the received signal fall into the correlation machine, and are
multiplied by the corresponding points in the target signal. The sum of these products then
moves into the proper sample in the cross correlation signal.

Figure 5.6
The amplitude of each sample in the cross-correlation signal is a measure of how much the
received signal resembles the target signal, at that location. This means that a peak will occur
in the cross-correlation signal for every target signal that is present in the received signal. In
other words, the value of the cross-correlation is maximized when the target signal is aligned
with the same features in the received signal. What if the target signal contains samples with a
negative value? Nothing changes. Imagine that the correlation machine is positioned such that
the target signal is perfectly aligned with the matching waveform in the received signal. As
samples from the received signal fall into the correlation machine, they are multiplied by their
matching samples in the target signal. Neglecting noise, a positive sample will be multiplied
by itself, resulting in a positive number. Likewise, a negative sample will be multiplied by
itself, also resulting in a positive number. Even if the target signal is completely negative, the
peak in the cross-correlation will still be positive. If there is noise on the received signal, there
will also be noise on the cross correlation signal. It is an unavoidable fact that random noise
looks a certain amount like any target signal you can choose. The noise on the cross-correlation
signal is simply measuring this similarity. Except for this noise, the peak generated in the cross-
correlation signal is symmetrical between its left and right. This is true even if the target signal
isn't symmetrical. In addition, the width of the peak is twice the width of the target signal.
Remember, the cross-correlation is trying to detect the target signal, not recreate it. There is no
reason to expect that the peak will even look like the target signal. Correlation is the optimal
technique for detecting a known waveform in random noise. That is, the peak is higher above
the noise using correlation than can be produced by any other linear system. (To be perfectly
correct, it is only optimal for random white noise). Using correlation to detect a known
waveform is frequently called matched filtering.
The correlation machine and convolution machine are identical, except for one small
difference. As discussed in the last chapter, the signal inside of the convolution machine is
flipped left-for-right. This means that samples numbers: 1, 2, 3˛ run from the right to the left.
In the correlation machine this flip doesn't take place, and the samples run in the normal
direction.

References:
• Digital Signal Processing by S. Salivahanan, C. Gnanapriya Second Edition, TMH
• Digital Signal Processing by Sanjit K. Mitra, Third Edition, TMH
• Signals and systems by A Anand Kumar (PHI) 2011
• Signals and Systems by Alan V. Oppenheim and Alan S. Willsky with S. Hamid
Nawab, Second Edition, PHI (EEE)
• Digital Signal Processing by Apte, Second Edition, Wiley India
Unit 6

6.0 Objectives

6.1 Finite Impulse Response (FIR )Filters

6.1.1 Introduction

6.1.2 Magnitude response and phase response of digital filters

6.1.3 Frequency response of linear phase FIR filters

6.1.4 Design techniques of FIR filters

6.1.5 Design of optima linear phase FIR filters

6.2 Infinite Impulse Response (IIR) Filters

6.2.1 Introduction

6.2.2 IIR filter design by approximation of derivatives

6.2.3 IIR filter design by impulse invariant method

6.2.4 IIR filter design by the bilinear transformation

6.2.5 Butterworth filters

6.2.6 Chebyshev filters

6.2.7 Elliptic filters

6.2.8 Frequency transformation

6.3 Conclusion

6.4 List of References

6.5 Bibliography

6.6 Unit End Exercises

6.0 OBJECTIVES

After going through this unit you will be able to :

● Define finite impulse response as well as infinite impulse response filters

● Describe design techniques of FIR filters
● Explain IIR filter design by approximation of derivatives,impulse invariant method,bilinear
transformation
● Understand butterworth and chebyshev filters.
● Explain frequency transformation
● Describe frequency response of linear phase FIR filters.
6.1 Finite Impulse Response (FIR) Filters

6.1.1 Introduction

A filter is a frequency selective system.Digital filters are classified as finite duration unit impulse
response filters or infinite duration unit impulse response filters.In signal processing,a finite impulse
response filter is a filter whose impulse response is of finite duration. I.e it has a finite number of
non zero terms.FIR filter is a filter with no feedback in its equation.the response of FIR filter depends
only on past and present samples.FIR filters are usually implemented using non-recursive
structure,however they can be realized in both recursive as well as non recursive structures.

Following are the main advantages of FIR filters:

1. FIR filters are always stable.

2. FIR filters are free of limit cycle oscillations , when implemented on a finite word length
digital system.
3. Excellent design methods are available for various kinds of FIR filters.
Disadvantages of FIR filters are as follows:

1. Memory requirement for FIR filter is very high

2. The implementation of FIR filters is very costly, since it requires more arithmetic operations
and hardware components such as multipliers,adders and delay elements.

The basis FIR filter is characterized by two equations :

𝑦(𝑛) = ℎ(𝑘) (𝑥(𝑛 − 𝑘) (6.1a)

𝐻(𝑧) = ℎ(𝑘) (6.1b)

Where ℎ(𝑘), 𝑘 = 0,1, . . . 𝑁 − 1 , are the impulse response coefficient of the filter, 𝐻(𝑧)is the
transfer function of the filter and N is the filter length,i.e the number of filter coefficients.equation
6.1a is the FIR difference equation.

y(n) is a function only of past and present values of input x(n).FIR filters are always stable ,if they are
implemented by direct evaluation as shown in equation 6.1a.Equation 6.1b represents the transfer
function of filter,which provides a mean of analyzing the filter.

All DSP processors available have architecture suited to FIR filtering.FIR filters are very simple to
implement.

6.1.2 Magnitude response and phase response of digital filters

The magnitude response of the filter can be characterized in terms of frequency bands the filter will
pass or reject.The transfer function of a FIR causal filter is given by

𝐻(𝑧) = ∑𝑁−1
𝑛=0 ℎ(𝑛) 𝑧 −𝑛
where h(n) is the impulse response of the filter. The frequency response [Fourier transform

of h(n)] is given by

𝐻(𝜔) = ∑𝑁−1
𝑛=0 ℎ(𝑛) 𝑒 −𝑛𝜔𝑗
which is periodic in frequency with period 2 p , i.e.,

𝐻(𝜔) = 𝐻(𝜔 + 2𝜋𝑘), k = 0,1,2..

Since 𝐻(𝜔) is complex it can be expressed as

𝐻(𝜔) = ±|𝐻(𝜔 )| 𝑒 −𝑛𝜔𝑗

Where |𝐻(𝜔)| is the magnitude response and 𝜃(𝜔)is the phase response .

We define the phase delay 𝜏𝑝 and group delay 𝜏𝑔 of a filter as:

𝜏𝑝 = −𝑑𝜃(𝜔) / 𝑑(𝜔) = −𝑑𝜃(𝜔) / 𝑑(𝜔) =

6.1.3 Frequency response of linear phase FIR filters

The frequency response of the filter is the fourier transform of its impulse response.if h(n) is the
impulse response of the system, then the frequency system is denoted by 𝐻(𝜔) or 𝐻(𝑒 𝑗𝜔 ).𝐻(𝜔)is a
complex function of frequency 𝜔 and so it can be expressed as magnitude function

|𝐻(𝜔)| and phase function ∠𝐻(𝜔)

Linear phase filters have 4 possible type of impulse response, depending on N and the type of
symmetry :

1. Symmetrical impulse response when N is odd.

2. Symmetrical impulse response when N is even.
3. Asymmetrical impulse response when N is odd.
4. Asymmetrical impulse response when N is even.

6.1.3.1 Frequency response of linear phase FIR filter when impulse response is symmetrical and N
is odd.
The equation for frequency response of linear filter when impulse response is symmetrical and N is
odd is given by :

The magnitude function is given by

The phase function is given by

Figure a shows symmetrical impulse response when N is 9,wheres figure b shows

the corresponding magnitude function of frequency response.

(a) Symmetrical impulse response, N = 9 (b) Magnitude function of 𝐻(𝜔).

From the figure it can be observed that the magnitude function of h is symmetric with 𝜔 = 𝛱 , when
the impulse response is symmetric and N is an odd number.

6.1.3.2 Frequency response of linear phase FIR filter when impulse response is symmetrical and N
is even.

The expression for frequency response of linear phase FIR filter when impulse response is
symmetrical and N is even is given by
 .

The magnitude function of H( w ) is given by

The phase function of H( w ) is given by

Following figure (a) shows a symmetrical impulse response when N = 8, and figure (b)

shows the corresponding magnitude function of frequency response. From these figures it

can be observed that the magnitude function of 𝐻(𝜔)is antisymmetric with𝜔 = 𝛱 , when

impulse response is symmetric and N is even number.

Fig (a) symmetrical impulse response(N=8) Fig (b) magnitude function of frequency response

6.1.3.3 Frequency response of linear phase FIR filter when impulse response is anti symmetric and
N is odd.

This is the equation for frequency response of linear phase FIR filter when impulse response

is antisymmetric and N odd.

The magnitude function is given by

The phase function is given by

Where

Figure a) shows an antisymmetric impulse response when N = 9, and Figure (b)

shows the corresponding magnitude function of frequency response. From these figures, it

can be observed that the magnitude function is antisymmetric with 𝜔 = 𝛱 , when the impulse

response is antisymmetric and N is odd.

Fig(a) antisymmetric impulse response(N=9) Fig(b) magnitude function of frequency response

6.1.3.4 Frequency response of linear phase FIR filter when impulse response is anti symmetric and
N is even.
This is the equation for the frequency response of linear phase FIR filter when impulse

response is antisymmetric and N is even.

The magnitude function is given by

The phase function is given by

Where

Figure (a) shows an antisymmetric impulse response when N = 8, and Figure (b)

shows its corresponding magnitude function of frequency response. it can be observed that the
magnitude function of 𝐻(𝜔) is symmetric with 𝜔 = 𝛱 when the impulse response is antisymmetric
and N is an even number.
Fig (a) Antisymmetric impulse response for N = 8 Fig (b) Magnitude function of 𝐻(𝜔).

6.1.4 Design techniques of FIR filters

Design of digital filter involves :

1. Filter specification : This may include stating the type of filter, for example lowpass filter, the
desired amplitude and/or phase responses and the tolerance, the sampling frequency, and
the word length of input data.
2. Coefficient calculation to find transfer function: At this step,we determine the coefficient of
transfer function.
3. Realization function : This involves converting the transfer function into a suitable filter
network or structure.
4. Analysis of finite wordlength effects : Here, we analyze the effect of quantizing the filter
coefficients and the input data as well as the effect carrying out the filtering operations
5. Implementation : This involves producing the software and/or hardware and performing the
actual filtering.

To design FIR filters following methods are followed :

1. Fourier series method

2. Window method
3. Frequency sampling method
4. Optimum filter design

6.1.4.1 Fourier series method of design

The procedure for designing FIR filters by Fourier series method is as follows:

Step 1: Choose the desired frequency response 𝐻𝑑 (𝜔) of the filter.

Step 2: Evaluate the Fourier series coefficients of 𝐻𝑑 (𝜔𝑇)which gives the desired

impulse responseℎ𝑑 (𝑛).

Step 3: Truncate the infinite sequence ℎ𝑑 (𝑛) to a finite sequenceℎ (𝑛).

Step 4: Take Z-transform of ℎ (𝑛) to get a non-causal filter transfer function 𝐻(𝑧).

Step 5: Multiply 𝐻(𝑧) by𝑧 −(𝑁−1)/2 to convert the non-causal transfer function to a

realizable causal FIR filter transfer function.

We know that any periodic function can be expressed as a linear combination of complex
exponentials. The frequency response of a digital filter is periodic with period equal to sampling
frequency. Therefor, the desired frequency response of an FIR filter can be represented by fourier
series as :

Where the fourier coefficients hd(n) are the desired impulse response sequence of the filter.the
samples of hd(n) can be determined using the equation :

Where ws is a sampling frequency in rad/sec. Fs is sampling frequency in Hz. T = 1/Fs is a sampling

period in sec.

The impulse response from the above equation is an infinite duration sequence.

For FIR filters , we truncate this infinite impulse response to a finite duration sequence of length N ,
where N is odd.Therefore ,

Taking Z - transform of the above equation for ℎ (𝑛) , we get

This transfer function of the filter 𝐻(𝑧) represents a non-causal filter. Hence the transfer function
represented by the above equation for 𝐻(𝑧) is multiplied by 𝑧 −(𝑁−1)/2. Therefore

Since ℎ (𝑛)=h(-n) , we express 𝐻(𝑧)as :

Hence we see that casualty is brought by multiplying the transfer function by the delay factor

𝛼 = (𝑁 − 1)/2. This modification does not affect the amplitude response of the filter, however the
abrupt truncation of the fourier series results in oscillations in the pass band and stop band.these
oscillations are due to slow convergence of the fourier series.This effect is known as gibbs
phenomenon.

6.1.4.2 Design of FIR filter using windows.

A finite weighing sequence w(n) with which the infinite impulse response is multiplied to obtain a
finite impulse response is called a window.A finite weighing sequence w(n) with which the infinite
impulse response is multiplied to obtain a finite impulse response is called a window. This is
necessary because abrupt truncation of the infinite impulse response will lead to oscillations in the
pass band and stop band, and these oscillations can be reduced through the

use of less abrupt truncation of the Fourier series.

The desirable characteristics of the window :

1. The central lobe of the frequency response of the window should contain most of the energy
and should be narrow.
2. The highest side lobe level of the frequency response should be small.
3. The side lobes of the frequency response should decrease in energy rapidly as w tends to pi.

The procedure for designing FIR filters using windows is:

Step 1: For the desired frequency response𝐻(𝜔), find the impulse responseℎ𝑑 (𝑛)

using the equation:

Step 2: Multiply the infinite impulse response with a chosen window sequence

𝜔(𝑛) of length N to obtain filter coefficients ℎ (𝑛), i.e.

Step 3: Find the transfer function of the realizable filter

Some common window functions are :

1. Rectangular
2. Bartlett
3. Hanning
4. Hamming
5. Blackmann

Rectangular window:

The weighting function (window function) for an N-point rectangular window is given by

The spectrum (frequency response) of rectangular window 𝑊𝑅 (𝜔) is given by the Fourier

transform of 𝑤𝑅 (𝑛)

The characteristic features of rectangular window are

(i) The main lobe width is equal to 4𝛱/𝑁

(ii) The maximum side lobe magnitude is –13 dB.

(iii) The side lobe magnitude does not decrease significantly with increasing 𝜔.
In a rectangular window , the width of the transition region is related to the width of the main lobe
of window spectrum.Gibbs oscillations are noticed in the pass band and stop band.The attenuation
in the stop band is constant and cannot be varied.

Bartlett Window :

Bartlett window is also called a triangular window. This window has been chosen such that it has
tapered sequences from the middle on either side. The window function w T (n) is defined as

In magnitude response of triangular window, the side lobe level is smaller than that of the
rectangular window being reduced from –13 dB to –25 dB. However, the main lobe width is now
8𝛱/𝑁 or twice that of the rectangular window.

The triangular window produces a smooth magnitude response in both pass band and stop band,
but it has the following disadvantages when compared to magnitude response obtained by using
rectangular window:

1. The transition region is more.

2. The attenuation in the stop band is less.

Because of these characteristics, the triangular window is not usually a good choice

Hanning window :

The Hanning window function is given by

The width of main lobe is 8𝛱/𝑁, i.e twice that of a rectangular window which results in doubling of
the transition region of the filter. The peak of the first side lobe is –32 dB relative to the maximum
value. This results in smaller ripples in both the pass band and stop band of the low-pass filter
designed using the Hanning window. The minimum stop band attenuation of the filter is 44 dB. At
higher frequencies the stop band attenuation is even greater. When compared to a triangular
window, the main lobe width is the same, but the magnitude of the side lobe is reduced, hence the
Hanning window is preferable to the triangular Window.

Hamming window :

The Hamming window function is given by

In the magnitude response for N = 31, the magnitude of the first side lobe is down about 41

dB from the main lobe peak, an improvement of 10 dB relative to the Hanning window. But this
improvement is achieved at the expense of the side lobe magnitudes at higher frequencies, which
are almost constant with frequency. The width of the main lobe is 8 p /N. In the magnitude response
of a low-pass filter designed using the Hamming window, the first side lobe peak is –51 dB, which is –
7 dB lesser with respect to the Hanning window filter. However, at higher frequencies, the stop band
attenuation is low when compared to that of Hanning window. Because the Hamming window
generates lesser oscillations in the side lobes than the Hanning window for the same main lobe
width, the Hamming window is generally preferred.

Blackman window :
The Blackman window function is another type of cosine window and given by the equation

Or

In the magnitude response, the width of the main lobe is 12𝛱/𝑁, which is highest among windows.
The peak of the first side lobe is at –58 dB and the side lobe magnitude decreases with frequency.
This desirable feature is achieved at the expense of increased main lobe width. However, the main
lobe width can be reduced by increasing the value of N. The side lobe attenuation of a low-pass filter
using Blackman window is –78 dB.

Summary of window characteristics

6.1.4.3 Frequency sampling method

The ideal frequency response is sampled at sufficient number of points (i.e. N-points). These samples
are the DFT coefficients of the impulse response of the filter. Hence the impulse response of the
filter is determined by taking IDFT.

Let 𝐻𝑑 (𝜔) = Idea frequency response

𝐻(𝑘) = DFT sequence obtained by sampling 𝐻𝑑 (𝜔)

 ℎ(𝑛) = Impulse response of FIR filter

The impulse response ℎ(𝑛) is obtained by taking IDFT of 𝐻(𝑘) .The samples of impulse response
should be real. The terms 𝐻(𝑘)𝑒 𝑗(2𝛱𝑛𝑘/𝑁) should be matched by the 𝑒 −𝑗(2𝛱𝑛𝑘/𝑁).

Frequency sampling methods include two design techniques i.e,

1. type-I design
2. type-II design.
In the type-I design, the set of frequency samples includes the sample at frequency 𝜔 = 0

When other set of samples are used instead of 𝜔 = 0,such a design procedure is referred to as the
type-II design

Procedure for type-I design

1. Choose the ideal (desired) frequency response 𝐻𝑑 (𝜔).

2𝛱𝑘
2 Sample 𝐻𝑑 (𝜔) at N-points by taking 𝜔 = 𝜔𝑘 = 𝑁
, where 𝑘 = 0, 1, 2, 3, . . . , (𝑁 – 1) to generate
the sequence 𝐻(𝑘).

3. Compute the N samples of h(n) using the following equations:

When N is odd ,

When n is even ,

4. Take Z-transform of the impulse response h(n) to get the transfer function 𝐻(𝑧)
Procedure for type-II design

1. Choose the ideal frequency response 𝐻𝑑 (𝜔).

2𝛱(2𝑘+1)
2. Sample 𝐻𝑑 (𝜔) at 𝑁-points by taking 𝜔 = 𝜔𝑘 = 2𝑁

where 𝑘 = 0, 1, 2, . . . , (𝑁 – 1) to generate the sequence 𝐻(𝑘).

3. Compute the N samples ofℎ(𝑛) using the following equations:

When N is odd,

When N is even,

4. Take Z-transform of the impulse response ℎ(𝑛) to get the transfer function 𝐻(𝑧).

6.1.4.4 Optimum filter design

In optimum filter design method, the weighted approximation error between the desired

frequency response and the actual frequency response is spread evenly across the pass band

and evenly across the stop band of the filter. This results in the reduction of maximum error.

The resulting filter has ripples in both the pass band and the stop band. This concept of

design is called optimum equiripple design criterion.

6.1.5 Design of optimal linear phase FIR filters

The optimal method is based on the concept of equiripple passband and stopband. In the passband
, the practical response oscillates between 1-𝛿𝑝 and 1+𝛿𝑝 . in th stopband the filter response lies
between 0 and𝛿𝑠 .The difference between the ideal filter and the practical response can be viewed as
error function

𝐸(𝜔) = 𝑊(𝜔)[𝐻𝐷 (𝜔) − 𝐻(𝜔)

Where 𝐻𝐷 (𝜔 is the ideal or desired response and 𝑊(𝜔) is a weighing function that allows the
relative error of approximation between different bands to be defined in the optimal method.

The main problem in the optimal method is to find the location of the external frequencies. A
powerful technique which employs remez exchange algorithm to find the external frequencies has
been developed.For a given set of specifications( that is passband edge frequencies N and the ratio
between the passband and stopband ripples the optimal method involves the following key steps :

1. Use the remez exchange algorithm to find the optimum set of external frequencies.
2. Determine the frequency response using external frequencies
3. Obtain the impulse response coefficients.

Flow chart of the optimal method.:

 Flowchart for optimal method

The heart of the optimal method is the first step where an iterative process is used to determine the
external frequencies of a filter whose amplitude-frequency response satisfies the optimality
condition.This step relies on the alternation theorem which specifies the number of external
frequencies that can exist for a given value of N.

6.2 Infinite Impulse Response (IIR) Filters

6.2.1 Introduction

The type of filters which make use of feedback connection to get the desired filter implementation
are known as recursive filters.Their impulse response is of infinite duration. So ,they are called IIR
filters. IIR filters are designed by considering all the infinite samples of the impulse response. The
impulse response is obtained by taking the inverse Fourier transform of ideal frequency response.
There are several techniques available for the design of digital filters having an infinite duration unit
impulse response. The popular methods for such filter design uses the technique of first designing
the digital filter in analog domain and then transforming the analog filter into an equivalent digital
filter because the analog filter design techniques are well developed.

IIR filters normally require fewer coefficients than FIR filters.These filters are mainly used when
throughput and sharp cutoff is the important requirement.The physically realizable and stable IIR
filter cannot have a linear phase. For a filter to have a linear phase, the condition to be satisfied is
h(n) = h(N – 1 – n) where N is the length of the filter and the filter would have a mirror image pole
outside the unit circle for every pole inside the unit circle. This results in an unstable filter. As a
result, a causal and stable IIR filter cannot have linear phase. In the design of IIR filters, only the
desired magnitude

Important features of IIR filters:

1. The physically realizable IIR filters do not have linear phase.

2. The IIR filter specifications include the desired characteristics for the magnitude

response only.
6.2.2 IIR filter design by approximation of derivatives

The approximation of derivative method is also known as backward difference method.The analog
filter having the rational system function H(s) can also be described by the linear

constant coefficient differential equation.

𝑑𝑦(𝑡) 𝑑𝑥(𝑡)
=
𝑑𝑡 𝑑𝑡
In this method of IIR filter design by approximation of derivatives, an analog filter is converted into a
digital filter by approximating the above differential equation into an equivalent difference equation.

𝑑𝑦(𝑡)
The backward difference formula is substituted for the derivative at time 𝑡 = 𝑛𝑇
𝑑𝑡
Thus,

𝑑𝑦(𝑡) 𝑦(𝑛𝑇)−𝑦(𝑛−1)𝑇
=
𝑑𝑡 𝑇
𝑑𝑦(𝑡) 𝑦(𝑛)−𝑦(𝑛−1)
Or =
𝑑𝑡 𝑇

where T is the sampling interval and 𝑦(𝑛) = 𝑦(𝑛𝑇)

𝑑𝑦(𝑡)
The system function of an analog differentiator with an output is 𝐻(𝑠) = 𝑠 and
𝑑𝑡
the digital system which produces the outpu t[𝑦(𝑛) − 𝑦(𝑛 − 1)]/𝑇 has the system function
𝐻(𝑧) = [1 − 𝑧 −1]/𝑇.

Comparing these two, we can say that the frequency domain equivalent

𝑑𝑦(𝑡) 𝑦(𝑛)−𝑦(𝑛−1)
for the relationship = is:
𝑑𝑡 𝑇

1−𝑧 −1
𝑠=
𝑇
Thus, this is the analog domain to digital domain transformation.

Mapping of the z-plane from the s-plane

 1−𝑧 −1 1
We have 𝑠= i.e 𝑧=
𝑇 1−𝑠𝑇

Substituting 𝑠 = 𝑗𝛺 in the expression for z, we have

1
𝑧=
1−𝑗𝛺𝑇
1 𝛺𝑇
= 2 +𝑗
1+𝑗𝛺2 𝑇 1+𝑗𝛺2 𝑇 2
It can be observed that the mapping of the equation𝑠 = (1 − 𝑧 −1) )/𝑇 takes the left half plane of s-
domain into the corresponding points inside the circle of radius 0.5 and centre at z = 0.5. Also the
right half of the s-plane is mapped outside the unit circle. Because of this,mapping results in a stable
analog filter transformed into a stable digital filter. However, since the location of poles in the z-
domain are confined to smaller frequencies, this design method can be used only for transforming
analog low-pass filters and band pass filters which are having smaller resonant frequencies.

Fig : Mapping of s-plane into z-plane by the backward difference method.

6.2.3 IIR filter design by impulse invariant method

The desired impulse response of the digital filter is obtained by uniformly sampling the impulse
response of the equivalent analog filter. The main idea behind this is to preserve the frequency
response characteristics of the analog filter. For the digital filter to possess the frequency response
characteristics of the corresponding analog filter, the sampling period T should be sufficiently small
(or the sampling frequency should be sufficiently high) to minimize (or completely avoid) the effects
of aliasing.

Let ℎ𝑎 (𝑡) = Impulse response of analog filter

𝑇 = Sampling period

ℎ(𝑛) = Impulse response of digital filter

For impulse invariant transformation,

ℎ(𝑛)= ℎ𝑎 (𝑡) = ℎ𝑎 (𝑛𝑇)

Analog filter’s system function is given by

𝑁
𝐴𝑖
𝐻𝑎 (𝑠) = ∑
𝑠 − 𝑝𝑖
𝑖=1

The relationship between the transfer function f the digital filter and analog filter is given by

𝑁
𝐴𝑖
𝐻(𝑧) = ∑
1 − 𝑒 𝑝𝑖 𝑇 𝑧 −1
𝑖=1
Comparing the above expressions for 𝐻𝑎 (𝑠)and 𝐻(𝑧), we can say that the impulse invariant

transformation is accomplished by the mapping.

1 1
=
𝑠 − 𝑝𝑖 1 − 𝑒 𝑝𝑖 𝑇 𝑧 −1
The above mapping shows that the analog pole at 𝑠 = 𝑝𝑖 is mapped into a digital pole at 𝑧 =
𝑒 𝑝𝑇 . Therefore, the analog poles and the digital poles are related by the relation.
𝑧 = 𝑒 𝑠𝑇
 Fig : Mapping of (a) s-plane into (b) z-plane by impulse invariant transformation.

The mapping from the analog frequency 𝛺 to the digital frequency 𝜔 by impulse invariant
transformation is many-to-one which simply reflects the effects of aliasing due to sampling of the
impulse response.

The stability of a filter (or system) is related to the location of the poles. For a stable analog filter the
poles should lie on the left half of the s-plane. That means for a stable digital filter the poles should
lie inside the unit circle in the z-plane.

6.2.4 IIR filter design by the bilinear transformation

The IIR filter design using impulse invariant as well as approximation of derivatives methods is
appropriate only for the design of low-pass filters and band pass filters whose resonant frequencies
are small. These techniques are not suitable for high-pass or band reject filters. The limitation is
overcome in the mapping technique called the bilinear transformation. This transformation is a one-
to-one mapping from the s-domain to the z-domain. That is, the bilinear transformation is a
conformal mapping that transforms the imaginary axis of s-plane into the unit circle in the z-plane
only once, thus avoiding aliasing

of frequency components. In this mapping, all points in the left half of s-plane are mapped

inside the unit circle in the z-plane, and all points in the right half of s-plane are mapped

outside the unit circle in the z-plane. So the transformation of a stable analog filter results in

a stable digital filter. The bilinear transformation can be obtained by using the trapezoidal

formula for the numerical integration.

 𝑏
Let the system function of analog filter be 𝐻𝑎 (𝑠) =
𝑠+𝑎

The differential equation describing the above analog filter can be obtained as:

Or 𝑠𝑌(𝑠) + 𝑎𝑌(𝑠) = 𝑏 𝑋(𝑠)

Taking inverse Laplace transform on both sides, we get

Integrating the above equation between the limits (nT – T) and nT, we have

The trapezoidal rule for numeric integration is expressed as:

Therefore, we get

After taking z-transform,the system function of a digital filter is

Comparing this with the analog filter system function 𝐻𝑎 (𝑠) we get

On rearranging,
This is the relation between analog and digital poles in bilinear transformation.

So to convert an analog filter function into an equivalent digital filter function, we need to put

The general characteristic of the mapping 𝑧 = 𝑒 𝑠𝑇 may be obtained by putting𝑠 = 𝜎 + 𝑗𝛺 and

expressing the complex variable z in the polar form as 𝑧 = 𝑟𝑒 𝑗𝜔 in the above equation for s.

Thus,

Which is equal to

On the imaginary axis of s-plane 𝜎 = 0 and correspondingly in the z-plane r = 1.

Therefore, The relation between analog and digital frequencies is:

 F

Fig : Mapping between 𝛺 and 𝜔 in bilinear transformation.

The mapping is non-linear and the lower frequencies in analog domain are expanded in the digital
domain, whereas the higher frequencies are compressed. This is due to the nonlinearity of the
arctangent function and is usually known as frequency warping.

6.2.5 Butterworth filters

To design a Butterworth IIR digital filter, first an analog Butterworth filter transfer function is
determined using the given specifications. Then the analog filter transfer function is converted to a
digital filter transfer function using either impulse invariant transformation or bilinear
transformation.Infinite-duration Impulse Response (IIR) Filters

The analog Butterworth filter is designed by approximating the ideal frequency response

using an error function. The error function is selected such that the magnitude is maximally

flat in the passband and monotonically decreasing in the stopband. (Strictly speaking the

magnitude is maximally flat at the origin, i.e., at W = 0, and monotonically decreasing with

increasing W).

The magnitude response of low-pass filter obtained by this approximation is given by

where W c is the 3 dB cutoff frequency and N is the order of the filter.

6.2.5.1 Frequency response of the Butterworth filter

The frequency response of Butterworth filter depends on the order N. The magnitude

response for different values of N are shown in Figure. From Figure 8.8, It can be observed that the
approximated magnitude response approaches the ideal response as the value of N increases.

However, the phase response of the Butterworth filter becomes more nonlinear with increasing N.

Magnitude response of Butterworth low-pass filter for various values of N.

Design procedure for low-pass digital Butterworth IIR filter:

The low-pass digital Butterworth filter is designed as per the following steps:

Step 1 : Choose the type of transformation, i.e., either bilinear or impulse invariant transformation.

Step 2 : Calculate the ratio of analog edge frequencies depending upon the transformation chosen
such as bilinear or impulse

Step 3 : Decide the order N of the filter. Choose N such that it is an integer just greater than or equal
to the value obtained.

Step 4 : Calculate the analog cutoff frequency for both transformation

Step 5 : Determine the transfer function of the analog filter.

Step 6: Using the chosen transformation, transform the analog filter transfer function H a (s)

to digital filter transfer function H(z).

Step 7 : Realize the digital filter transfer function H(z) by a suitable structure.
Properties of Butterworth filters

1. The Butterworth filters are all pole designs (i.e. the zeros of the filters exist at ¥).

2. The filter order N completely specifies the filter.

The magnitude response approaches the ideal response as the value of N increases.

The magnitude is maximally flat at the origin.

The magnitude is a monotonically decreasing function of W.

At the cutoff frequency W c , the magnitude of normalized Butterworth filter is 1/

2 . Hence the dB magnitude at the cutoff frequency will be 3 dB less than the

maximum value.

6.2.6 Chebyshev filters

or designing a Chebyshev IIR digital filter, first an analog filter is designed using the given

specifications. Then the analog filter transfer function is transformed to digital filter transfer

function by using either impulse invariant transformation or bilinear transformation.

The analog Chebyshev filter is designed by approximating the ideal frequency response

using an error function. There are two types of Chebyshev approximations. In type-1

approximation, the error function is selected such that the magnitude response is equiripple

in the passband and monotonic in the stopband. In type-2 approximation, the error function

is selected such that the magnitude function is monotonic in the passband and equiripple in

the stopband. The type-2 magnitude response is also called inverse Chebyshev response. The

type-1 design is presented in this book

The magnitude response of type-1 Chebyshev low-pass filter is given by:

where e is attenuation constant given by

Frequency response of the Chebyshev filter:

The frequency response of Chebyshev filters depends on order N. The approximated

response approaches the ideal response as the order N increases. The phase response of the

Chebyshev filter is more nonlinear than that of the Butterworth filter for a given filter length

N.

Design procedure for low-pass digital Chebyshev IIR filter:

The low-pass Chebyshev IIR digital filter is designed following the steps given below.

Step 1 : Choose the type of transformation.

(Bilinear or impulse invariant transformation)

Step 2 : Calculate the attenuation constant e .

Step 3 : Calculate the ratio of analog edge frequencies W 2 /W 1 .

Step 4 : Decide the order of the filter N

Step 5 : Calculate the analog cutoff frequency W c . for both transformation.

Step 6 : Determine the analog transfer function H a (s) of the filter, wehn the order of N is odd or
even

Step 7 : Using the chosen transformation, transform H a (s) to H(z), where H(z) is the

transfer function of the digital filter.

Properties of Chebyshev filters (Type 1):

1. The magnitude response is equiripple in the passband and monotonic in the Stopband.
2. The chebyshev type-1 filters are all pole designs.
3. The normalized magnitude function has a value of 1/ 1 + F 2 at the cutoff frequency W c
4. The magnitude response approaches the ideal response as the value of N increases.
Inverse Chebyshev filters

Inverse Chebyshev filters are also called type-2 Chebyshev filters. A low-pass inverse

Chebyshev filter has a magnitude response given by

where e is a constant and W c is the 3 dB cutoff frequency. The Chebyshev polynomial c N (x)

is given by

The

magnitude response has maximally flat passband and equiripple stopband, just the opposite

of the Chebyshev filters response. That is why type-2 Chebyshev filters are called the inverse

Chebyshev filters.

6.2.7 Elliptic filters

The elliptic filter is sometimes called the Cauer filter. This filter has equiripple passband and

stopband. Among the filters discussed so far, for a given filter order, pass band and stop
band deviations, elliptic filters have the minimum transition bandwidth. The magnitude

response of an elliptic filter is given by

where 𝑈𝑁 (𝑥)is the Jacobian elliptic function of order N and 𝜀 is a constant related to the

passband ripple.

6.2.8 Frequency transformation

In the design techniques discussed so far, we have considered only low-pass filters.

This low-pass filter can be considered as a prototype filter and its system function H p (s) can

be determined. The high-pass or band pass or band stop filters are designed by designing a

low-pass filter and then transforming that low-pass transfer function into the required filter

function by frequency transformation. Frequency transformation can be accomplished in two

ways.

Basically there are four types of frequency selective filters, viz. low-pass, high-pass, band pass

and the band stopped. In Figure 8.11, the frequency response of the ideal case is shown in solid

lines and practical case in dotted lines

 Frequency response of (a) Low-pass filter, (b) High-pass filter, (c) Band pass filter and

(d) Band stop filter.

The high-pass or band pass or band stop filters are designed by designing a

low-pass filter and then transforming that low-pass transfer function into the required filter

function by frequency transformation. Frequency transformation can be accomplished in two

ways:

1. Analog frequency transformation

2. Digital frequency transformation

Analog frequency transformation:

In the analog frequency transformation, the analog system function 𝐻𝑝 (𝑠) of the prototype

filter is converted into another analog system function 𝐻(𝑠) of the desired filter (a low-pass

filter with another cutoff frequency or a high-pass filter or a band pass filter or a band stop

filter). Then using any of the mapping techniques (impulse invariant transformation or

bilinear transformation) this analog filter is converted into the digital filter with a system
function 𝐻(𝑧).

The frequency transformation formulae used to convert a prototype low-pass filter into a low-pass
(with a different cutoff frequency), high-pass, band pass or band stop are given in Table. 𝛺𝑐 is the
cutoff frequency of the low-pass prototype filter.𝛺𝑐 * cutoff frequency of new low-pass filter or
high-pass filter and 𝛺1 and 𝛺2 are the cutoff frequencies of band pass or band stop filters.

Digital Frequency Transformation

As in the analog domain, frequency transformation is possible in the digital domain also. The

frequency transformation is done in the digital domain by replacing the variable 𝑧 −1 by a function of
𝑧 −1 , i.e.,𝑓(𝑧 −1 ). This mapping must take into account the stability criterion. All the

poles lying within the unit circle must map onto itself and the unit circle must also map onto

Itself.

Following table gives the formulae for the transformation of the prototype low pass digital

filter into a digital low-pass, high-pass, band pass or band stop filters.
The frequency transformation may be accomplished in any of the available two techniques,
however, caution must be taken to which technique to use. For example, the impulse invariant
transformation is not suitable for high-pass or bandpass filters whose resonant frequencies are
higher. In such a case, suppose a low-pass prototype filter is converted into a high-pass filter using
the analog frequency transformation and transformed later to a digital filter using the impulse
invariant technique. This will result in aliasing problems. However, if the same prototype low-pass
filter is first transformed into a digital filter using the impulse invariant technique and later
converted into a high-pass filter using the digital frequency transformation, then it will not have any
aliasing problem. Whenever the bilinear transformation is used, it is of no significance whether
analog frequency transformation is used or digital frequency transformation. In this case, both
analog and digital frequency transformation techniques will give the same result
6.3 Conclusion

● Based on impulse response, filters are of two types: (i) IIR filters and (ii) FIR filters.The IIR
filters are designed using an infinite number of samples of impulse response.They are of
recursive type, whereby the present output depends on the present input, past input and
past output samples. The FIR filters are designed using only a finite number of samples of
impulse response. They are non-recursive types whereby the present output depends on the
present input and past input samples.
● The necessary and sufficient condition for the linear phase characteristic of FIR filter is that
the phase function should be a linear function of 𝜔, which in turn requires constant phase
delay or constant phase and group delay.
● The transformation of analog filter to digital filter without modifying the impulse response of
the filter is called impulse invariant transformation (i.e. in this transformation, the impulse
response of the digital filter will be the sampled version of the impulse response of the
analog filter).
● FIR filter is always stable because all its poles are at the origin.
● The two concepts that lead to the design of FIR filter by Fourier series are: (i) The frequency
response of a digital filter is periodic with period equal to sampling frequency.(ii) Any
periodic function can be expressed as a linear combination of complex exponentials.
● A finite weighing sequence w(n) with which the infinite impulse response is multiplied to
obtain a finite impulse response is called a window. This is necessary because abrupt
truncation of the infinite impulse response will lead to oscillations in the pass band and stop
band, and these oscillations can be reduced through the use of less abrupt truncation of the
Fourier series.
● Chebyshev approximation is one in which the approximation function is selected such that
the error is minimized over a prescribed band of frequencies.
● Type-1 Chebyshev approximation is one in which the error function is selected such that the
magnitude response is equiripple in the passband and monotonic in the stopband.
● Type-2 Chebyshev approximation is one in which the error function is selected such that the
magnitude response is monotonic in the passband and equiripple in the stopband. The type-
2 Chebyshev response is called inverse Chebyshev response.
6.4 List of References

1. Digital Signal Processing - A. Anand Kumar

2. Digital Signal Processing A Practical Approach by Emanuel C Ifeachor , Barrie W. Jervis
3. https://www.hindawi.com/journals/tswj/2013/320489/
4. https://www.sciencedirect.com/topics/computer-science/designed-filter
5. https://www.sciencedirect.com/topics/computer-science/bilinear-transformation
6. https://www.sciencedirect.com/topics/engineering/magnitude-response
7. https://www.vyssotski.ch/BasicsOfInstrumentation/SpikeSorting/Design_of_FIR_Filters.pdf

6.5 Bibliography

6.6 Unit End Exercises

1. What is an FIR filter? Compare an FIR filter with an IIR filter.

2. Write the steps in the design of FIR filters.

3. Explain FIR filter design using windowing method.

4. Find the frequency response of a rectangular window.

5.Design an FIR digital filter to approximate an ideal low-pass filter with pass band gain of unity,
cutoff frequency of 1 kHz and working at a sampling frequency of𝑓𝑠 = 4 kHz. The length of the
impulse response should be 11. Use the Fourier series method.

6. Compare analog and digital filters. State the advantages of digital filters over

analog filters.

7. Define infinite impulse response and finite impulse response filters and compare.
8. Justify the statement IIR filter is less stable and give reason for it.

9. Describe digital IIR filter characterization in time domain.

10. Describe digital IIR filter characterization in z-domain.

11. Discuss the impulse invariant method.

12. What are the limitations of impulse invariant method?

13 Compare impulse invariant and bilinear transformation methods.

14. Discuss the magnitude and phase responses of digital filters.