X[ m ] =

N –1
∑ x [n ]e
mn
– j2π -------
N
Signals and
Systems
n=0

Second Edition
Steven T. Karris
with MATLAB® Applications

Includes

step-by-step

procedures

for designing

analog and

digital filters

Orchard Publications
www.orchardpublications.com
 Students and working professionals will find
Signals and Systems Signals and Systems with MATLAB® Applications,
Second Edition, to be a concise and easy-to-learn
with MATLAB® Applications text. It provides complete, clear, and detailed expla-
Second Edition nations of the principal analog and digital signal
Steven T. Karris processing concepts and analog and digital filter
design illustrated with numerous practical examples.

This text includes the following chapters and appendices:

• Elementary Signals • The Laplace Transformation • The Inverse Laplace Transformation
• Circuit Analysis with Laplace Transforms • State Variables and State Equations • The Impulse
Response and Convolution • Fourier Series • The Fourier Transform • Discrete Time Systems
and the Z Transform • The DFT and The FFT Algorithm • Analog and Digital Filters
• Introduction to MATLAB • Review of Complex Numbers • Review of Matrices and Determinants
Each chapter contains numerous practical applications supplemented with detailed instructions
for using MATLAB to obtain quick solutions.

Steven T. Karris is the president and founder of Orchard Publications. He earned a bachelors
degree in electrical engineering at Christian Brothers University, Memphis, Tennessee, a mas-
ters degree in electrical engineering at Florida Institute of Technology, Melbourne, Florida, and
has done post-master work at the latter. He is a registered professional engineer in California
and Florida. He has over 30 years of professional engineering experience in industry. In addi-
tion, he has over 25 years of teaching experience that he acquired at several educational insti-
tutions as an adjunct professor. He is currently with UC Berkeley Extension.

Orchard Publications, Fremont, California

Visit us on the Internet
www.orchardpublications.com
or email us: [email protected]

ISBN 0-9709511-8-3

$39.95 U.S.A.
Signals and Systems
with MATLAB® Applications

Second Edition
Steven T. Karris

Orchard Publications
www.orchardpublications.com
Signals and Systems with MATLAB Applications, Second Edition

Copyright © 2003 Orchard Publications. All rights reserved. Printed in the United States of America. No part of this
publication may be reproduced or distributed in any form or by any means, or stored in a data base or retrieval system,
without the prior written permission of the publisher.

Direct all inquiries to Orchard Publications, 39510 Paseo Padre Parkway, Fremont, California 94538

Product and corporate names are trademarks or registered trademarks of the Microsoft™ Corporation and The
MathWorks™ Inc. They are used only for identification and explanation, without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Library of Congress Control Number: 2003091595

ISBN 0-9709511-8-3

Copyright TX 5-471-562
 Preface

This text contains a comprehensive discussion on continuous and discrete time signals and systems
with many MATLAB® examples. It is written for junior and senior electrical engineering students,
and for self-study by working professionals. The prerequisites are a basic course in differential and
integral calculus, and basic electric circuit theory.
This book can be used in a two-quarter, or one semester course. This author has taught the subject
material for many years at San Jose State University, San Jose, California, and was able to cover all
material in 16 weeks, with 2½ lecture hours per week.
To get the most out of this text, it is highly recommended that Appendix A is thoroughly reviewed.
This appendix serves as an introduction to MATLAB, and is intended for those who are not familiar
with it. The Student Edition of MATLAB is an inexpensive, and yet a very powerful software
package; it can be found in many college bookstores, or can be obtained directly from
The MathWorks™ Inc., 3 Apple Hill Drive , Natick, MA 01760-2098
Phone: 508 647-7000, Fax: 508 647-7001
http://www.mathworks.com
e-mail: [email protected]

The elementary signals are reviewed in Chapter 1 and several examples are presented. The intent of
this chapter is to enable the reader to express any waveform in terms of the unit step function, and
subsequently the derivation of the Laplace transform of it. Chapters 2 through 4 are devoted to
Laplace transformation and circuit analysis using this transform. Chapter 5 discusses the state
variable method, and Chapter 6 the impulse response. Chapters 7 and 8 are devoted to Fourier series
and transform respectively. Chapter 9 introduces discrete-time signals and the Z transform.
Considerable time was spent on Chapter 10 to present the Discrete Fourier transform and FFT with
the simplest possible explanations. Chapter 11 contains a thorough discussion to analog and digital
filters analysis and design procedures. As mentioned above, Appendix A is an introduction to
MATLAB. Appendix B contains a review of complex numbers, and Appendix C discusses matrices.
New to the Second Edition
This is an refined revision of the first edition. The most notable changes are chapter-end summaries,
and detailed solutions to all exercises. The latter is in response to many students and working
professionals who expressed a desire to obtain the author’s solutions for comparison with their own.
The author has prepared more exercises and they are available with their solutions to those
instructors who adopt this text for their class.
The chapter-end summaries will undoubtedly be a valuable aid to instructors for the preparation of
presentation material.
The last major change is the improvement of the plots generated by the latest revisions of the
MATLAB® Student Version, Release 13.
Orchard Publications
Fremont, California
www.orchardpublications.com
[email protected]

2
Table of Contents
Chapter 1

Elementary Signals
Signals Described in Math Form .................................................................................................................1-1
The Unit Step Function ................................................................................................................................1-2
The Unit Ramp Function ...........................................................................................................................1-10
The Delta Function .....................................................................................................................................1-12
Sampling Property of the Delta Function................................................................................................1-12
Sifting Property of the Delta Function.....................................................................................................1-13
Higher Order Delta Functions...................................................................................................................1-15
Summary........................................................................................................................................................1-19
Exercises........................................................................................................................................................1-20
Solutions to Exercises .................................................................................................................................1-21

Chapter 2

The Laplace Transformation

Definition of the Laplace Transformation................................................................................................. 2-1
Properties of the Laplace Transform.......................................................................................................... 2-2
The Laplace Transform of Common Functions of Time .....................................................................2-12
The Laplace Transform of Common Waveforms..................................................................................2-23
Summary........................................................................................................................................................2-29
Exercises........................................................................................................................................................2-34
Solutions to Exercises .................................................................................................................................2-37

Chapter 3

The Inverse Laplace Transformation

The Inverse Laplace Transform Integral....................................................................................................3-1
Partial Fraction Expansion ...........................................................................................................................3-1
Case where F ( s ) is Improper Rational Function ( m ≥ n )................................................................... 3-13
Alternate Method of Partial Fraction Expansion................................................................................... 3-15
Summary....................................................................................................................................................... 3-18

Signals and Systems with MATLAB Applications, Second Edition i

Orchard Publications
Exercises .......................................................................................................................................................3-20
Solutions to Exercises .................................................................................................................................3-22

Chapter 4

Circuit Analysis with Laplace Transforms

Circuit Transformation from Time to Complex Frequency ................................................................................................................... 4-1
Complex Impedance Z ( s ) ...........................................................................................................................4-8
Complex Admittance Y ( s ) ........................................................................................................................4-10
Transfer Functions ......................................................................................................................................4-13
Summary .......................................................................................................................................................4-16
Exercises .......................................................................................................................................................4-18
Solutions to Exercises .................................................................................................................................4-21

Chapter 5

State Variables and State Equations

Expressing Differential Equations in State Equation Form................................................................... 5-1
Solution of Single State Equations.............................................................................................................. 5-7
The State Transition Matrix ......................................................................................................................... 5-9
Computation of the State Transition Matrix ...........................................................................................5-11
Eigenvectors .................................................................................................................................................5-18
Circuit Analysis with State Variables ........................................................................................................5-22
Relationship between State Equations and Laplace Transform...........................................................5-28
Summary .......................................................................................................................................................5-35
Exercises .......................................................................................................................................................5-39
Solutions to Exercises .................................................................................................................................5-41

Chapter 6

The Impulse Response and Convolution

The Impulse Response in Time Domain................................................................................................... 6-1
Even and Odd Functions of Time..............................................................................................................6-5
Convolution.................................................................................................................................................... 6-7
Graphical Evaluation of the Convolution Integral ..................................................................................6-8
Circuit Analysis with the Convolution Integral...................................................................................... 6-18
Summary ...................................................................................................................................................... 6-20

ii Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Exercises....................................................................................................................................................... 6-22
Solutions to Exercises ................................................................................................................................ 6-24

Chapter 7

Fourier Series
Wave Analysis.................................................................................................................................................7-1
Evaluation of the Coefficients .....................................................................................................................7-2
Symmetry.........................................................................................................................................................7-7
Waveforms in Trigonometric Form of Fourier Series .......................................................................... 7-11
Gibbs Phenomenon.................................................................................................................................... 7-24
Alternate Forms of the Trigonometric Fourier Series .......................................................................... 7-25
Circuit Analysis with Trigonometric Fourier Series .............................................................................. 7-29
The Exponential Form of the Fourier Series ......................................................................................... 7-31
Line Spectra ................................................................................................................................................. 7-35
Computation of RMS Values from Fourier Series ................................................................................ 7-40
Computation of Average Power from Fourier Series ........................................................................... 7-42
Numerical Evaluation of Fourier Coefficients....................................................................................... 7-44
Summary....................................................................................................................................................... 7-48
Exercises....................................................................................................................................................... 7-51
Solutions to Exercises ................................................................................................................................ 7-53

Chapter 8

The Fourier Transform

Definition and Special Forms ...................................................................................................................... 8-1
Special Forms of the Fourier Transform ................................................................................................... 8-2
Properties and Theorems of the Fourier Transform................................................................................ 8-9
Fourier Transform Pairs of Common Functions ...................................................................................8-17
Finding the Fourier Transform from Laplace Transform.....................................................................8-25
Fourier Transforms of Common Waveforms.........................................................................................8-27
Using MATLAB to Compute the Fourier Transform ...........................................................................8-33
The System Function and Applications to Circuit Analysis..................................................................8-34
Summary........................................................................................................................................................8-41
Exercises........................................................................................................................................................8-47
Solutions to Exercises .................................................................................................................................8-49

Signals and Systems with MATLAB Applications, Second Edition iii

Orchard Publications
Chapter 9

Discrete Time Systems and the Z Transform

Definition and Special Forms ...................................................................................................................... 9-1
Properties and Theorems of the Z Tranform .......................................................................................... 9-3
The Z Transform of Common Discrete Time Functions....................................................................9-11
Computation of the Z transform with Contour Integration ...............................................................9-20
Transformation Between s and z Domains...........................................................................................9-22
The Inverse Z Transform..........................................................................................................................9-24
The Transfer Function of Discrete Time Systems .................................................................................9-38
State Equations for Discrete Time Systems ............................................................................................9-43
Summary .......................................................................................................................................................9-47
Exercises .......................................................................................................................................................9-52
Solutions to Exercises .................................................................................................................................9-54

Chapter 10

The DFT and the FFT Algorithm

The Discrete Fourier Transform (DFT) ..................................................................................................10-1
Even and Odd Properties of the DFT.....................................................................................................10-8
Properties and Theorems of the DFT................................................................................................... 10-10
The Sampling Theorem ........................................................................................................................... 10-13
Number of Operations Required to Compute the DFT.................................................................... 10-16
The Fast Fourier Transform (FFT) ....................................................................................................... 10-17
Summary .................................................................................................................................................... 10-28
Exercises .................................................................................................................................................... 10-31
Solutions to Exercises .............................................................................................................................. 10-33

Chapter 11

Analog and Digital Filters

Filter Types and Classifications ................................................................................................................ 11-1
Basic Analog Filters.................................................................................................................................... 11-2
Low-Pass Analog Filters............................................................................................................................ 11-7
Design of Butterworth Analog Low-Pass Filters ............................................................................... 11-11
Design of Type I Chebyshev Analog Low-Pass Filters...................................................................... 11-22
Other Low-Pass Filter Approximations................................................................................................ 11-34
High-Pass, Band-Pass, and Band-Elimination Filters......................................................................... 11-39

iv Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Digital Filters ............................................................................................................................................. 11-49
Summary..................................................................................................................................................... 11-69
Exercises..................................................................................................................................................... 11-73
Solutions to Exercises .............................................................................................................................. 11-79

Appendix A

Introduction to MATLAB®
MATLAB® and Simulink® ........................................................................................................................A-1
Command Window.......................................................................................................................................A-1
Roots of Polynomials ...................................................................................................................................A-3
Polynomial Construction from Known Roots.........................................................................................A-4
Evaluation of a Polynomial at Specified Values.......................................................................................A-6
Rational Polynomials ....................................................................................................................................A-8
Using MATLAB to Make Plots ................................................................................................................A-10
Subplots ........................................................................................................................................................A-18
Multiplication, Division and Exponentiation .........................................................................................A-18
Script and Function Files...........................................................................................................................A-25
Display Formats ..........................................................................................................................................A-30

Appendix B

Review of Complex Numbers

Definition of a Complex Number.............................................................................................................. B-1
Addition and Subtraction of Complex Numbers..................................................................................... B-2
Multiplication of Complex Numbers......................................................................................................... B-3
Division of Complex Numbers .................................................................................................................. B-4
Exponential and Polar Forms of Complex Numbers ............................................................................. B-4

Appendix C

Matrices and Determinants

Matrix Definition .......................................................................................................................................... C-1
Matrix Operations......................................................................................................................................... C-2
Special Forms of Matrices ........................................................................................................................... C-5
Determinants ................................................................................................................................................. C-9
Minors and Cofactors.................................................................................................................................C-12

Signals and Systems with MATLAB Applications, Second Edition v

Orchard Publications
Cramer’s Rule ..............................................................................................................................................C-16
Gaussian Elimination Method..................................................................................................................C-19
The Adjoint of a Matrix.............................................................................................................................C-20
Singular and Non-Singular Matrices ........................................................................................................C-21
The Inverse of a Matrix .............................................................................................................................C-21
Solution of Simultaneous Equations with Matrices ..............................................................................C-23
Exercises ......................................................................................................................................................C-30

vi Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 1
Elementary Signals

his chapter begins with a discussion of elementary signals that may be applied to electric net-
T works. The unit step, unit ramp, and delta functions are introduced. The sampling and sifting
properties of the delta function are defined and derived. Several examples for expressing a vari-
ety of waveforms in terms of these elementary signals are provided.

1.1 Signals Described in Math Form

Consider the network of Figure 1.1 where the switch is closed at time t = 0 .

vS t = 0
+
+ v out open terminals
−
−

Figure 1.1. A switched network with open terminals.

We wish to describe v out in a math form for the time interval – ∞ < t < +∞ . To do this, it is conve-
nient to divide the time interval into two parts, – ∞ < t < 0 , and 0 < t < ∞ .
For the time interval – ∞ < t < 0 , the switch is open and therefore, the output voltage v out is zero. In
other words,
v out = 0 for – ∞ < t < 0 (1.1)

For the time interval 0 < t < ∞ , the switch is closed. Then, the input voltage v S appears at the output,
i.e.,
v out = v S for 0 < t < ∞ (1.2)

Combining (1.1) and (1.2) into a single relationship, we get

⎧ 0 –∞ < t < 0
v out = ⎨ (1.3)
⎩ vS 0 < t < ∞

We can express (1.3) by the waveform shown in Figure 1.2.

Signals and Systems with MATLAB Applications, Second Edition 1-1

Orchard Publications
Chapter 1 Elementary Signals

v out
vS

0
t

Figure 1.2. Waveform for v out as defined in relation (1.3)

The waveform of Figure 1.2 is an example of a discontinuous function. A function is said to be dis-
continuous if it exhibits points of discontinuity, that is, the function jumps from one value to another
without taking on any intermediate values.

1.2 The Unit Step Function u 0 ( t )

A well-known discontinuous function is the unit step function u 0 ( t ) * that is defined as

⎧0 t<0
u0 ( t ) = ⎨ (1.4)
⎩1 t>0

It is also represented by the waveform of Figure 1.3.

u0 ( t )
1
0 t

Figure 1.3. Waveform for u 0 ( t )

In the waveform of Figure 1.3, the unit step function u 0 ( t ) changes abruptly from 0 to 1 at t = 0 .
But if it changes at t = t 0 instead, it is denoted as u 0 ( t – t 0 ) . Its waveform and definition are as
shown in Figure 1.4 and relation (1.5).

1
u0 ( t – t0 )
t
0 t0

Figure 1.4. Waveform for u 0 ( t – t 0 )

* In some books, the unit step function is denoted as u ( t ) , that is, without the subscript 0. In this text, however, we
will reserve the u ( t ) designation for any input when we discuss state variables in a later chapter.

1-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Unit Step Function

⎧0 t < t0
u0 ( t – t0 ) = ⎨ (1.5)
⎩1 t > t0

If the unit step function changes abruptly from 0 to 1 at t = – t 0 , it is denoted as u 0 ( t + t 0 ) . Its

waveform and definition are as shown in Figure 1.5 and relation (1.6).
u0 ( t + t0 )
1

−t0 0 t

Figure 1.5. Waveform for u 0 ( t + t 0 )

⎧0 t < –t0
u0 ( t + t0 ) = ⎨ (1.6)
⎩1 t > –t0

Example 1.1
Consider the network of Figure 1.6, where the switch is closed at time t = T .

vS t = T
+
+ v out open terminals
−
−

Figure 1.6. Network for Example 1.1

Express the output voltage v out as a function of the unit step function, and sketch the appropriate
waveform.
Solution:
For this example, the output voltage v out = 0 for t < T , and v out = v S for t > T . Therefore,

v out = v S u 0 ( t – T ) (1.7)

and the waveform is shown in Figure 1.7.

Signals and Systems with MATLAB Applications, Second Edition 1-3

Orchard Publications
Chapter 1 Elementary Signals

vS u0 ( t – T )
v out

t
0 T

Figure 1.7. Waveform for Example 1.1

Other forms of the unit step function are shown in Figure 1.8.

Τ −Τ
t t t
0 0 0
(a) (b) (c)
−A −A −A
–A u0 ( t ) –A u0 ( t – T ) –A u0 ( t + T )

Au 0 ( – t ) Au 0 ( – t + T ) Au 0 ( – t – T )
A A A

t t t
0 0 Τ (e) −Τ 0 (f)
(d)

Τ −Τ
t t t
0 (g) 0 0 (i)
(h)
−A −A −A
–A u0 ( –t ) –A u0 ( – t + T ) –A u0 ( – t – T )

Figure 1.8. Other forms of the unit step function

Unit step functions can be used to represent other time-varying functions such as the rectangular
pulse shown in Figure 1.9.
u0 ( t )
1
1
t t t
0 1 0 0
(c)
(a) (b)
–u0 ( t – 1 )

Figure 1.9. A rectangular pulse expressed as the sum of two unit step functions
Thus, the pulse of Figure 1.9(a) is the sum of the unit step functions of Figures 1.9(b) and 1.9(c) is
represented as u 0 ( t ) – u 0 ( t – 1 ) .

1-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Unit Step Function

The unit step function offers a convenient method of describing the sudden application of a voltage
or current source. For example, a constant voltage source of 24 V applied at t = 0 , can be denoted
as 24u 0 ( t ) V . Likewise, a sinusoidal voltage source v ( t ) = V m cos ωt V that is applied to a circuit at
t = t 0 , can be described as v ( t ) = ( V m cos ωt )u 0 ( t – t 0 ) V . Also, if the excitation in a circuit is a rect-
angular, or triangular, or sawtooth, or any other recurring pulse, it can be represented as a sum (dif-
ference) of unit step functions.

Example 1.2
Express the square waveform of Figure 1.10 as a sum of unit step functions. The vertical dotted lines
indicate the discontinuities at T, 2T, 3T and so on.

v(t)
A
{ }
T 2T 3T
t
0
–A | ~

Figure 1.10. Square waveform for Example 1.2

Solution:
Line segment { has height A , starts at t = 0 , and terminates at t = T . Then, as in Example 1.1, this
segment is expressed as
v1 ( t ) = A [ u0 ( t ) – u0 ( t – T ) ] (1.8)

Line segment | has height – A , starts at t = T and terminates at t = 2T . This segment is expressed
as
v 2 ( t ) = – A [ u 0 ( t – T ) – u 0 ( t – 2T ) ] (1.9)

Line segment } has height A , starts at t = 2T and terminates at t = 3T . This segment is expressed as
v 3 ( t ) = A [ u 0 ( t – 2T ) – u 0 ( t – 3T ) ] (1.10)

Line segment ~ has height – A , starts at t = 3T , and terminates at t = 4T . It is expressed as

v 4 ( t ) = – A [ u 0 ( t – 3T ) – u 0 ( t – 4T ) ] (1.11)

Thus, the square waveform of Figure 1.10 can be expressed as the summation of (1.8) through (1.11),
that is,

Signals and Systems with MATLAB Applications, Second Edition 1-5

Orchard Publications
Chapter 1 Elementary Signals

v ( t ) = v1 ( t ) + v2 ( t ) + v3 ( t ) + v4 ( t )
= A [ u 0 ( t ) – u 0 ( t – T ) ] – A [ u 0 ( t – T ) – u 0 ( t – 2T ) ] (1.12)
+A [ u 0 ( t – 2T ) – u 0 ( t – 3T ) ] – A [ u 0 ( t – 3T ) – u 0 ( t – 4T ) ]

Combining like terms, we get

v ( t ) = A [ u 0 ( t ) – 2u 0 ( t – T ) + 2u 0 ( t – 2T ) – 2u 0 ( t – 3T ) + … ] (1.13)

Example 1.3
Express the symmetric rectangular pulse of Figure 1.11 as a sum of unit step functions.
i(t)
A

t
–T ⁄ 2 0 T⁄2

Figure 1.11. Symmetric rectangular pulse for Example 1.3

Solution:
This pulse has height A , starts at t = – T ⁄ 2 , and terminates at t = T ⁄ 2 . Therefore, with reference to
Figures 1.5 and 1.8 (b), we get

i ( t ) = Au 0 ⎛ t + --- ⎞ – Au 0 ⎛ t – --- ⎞ = A u 0 ⎛ t + --- ⎞ – u 0 ⎛ t – --- ⎞

T T T T
⎝
(1.14)
2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠

Example 1.4
Express the symmetric triangular waveform of Figure 1.12 as a sum of unit step functions.
v(t)
1

t
–T ⁄ 2 0 T⁄2

Figure 1.12. Symmetric triangular waveform for Example 1.4

Solution:
We first derive the equations for the linear segments { and | shown in Figure 1.13.

1-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Unit Step Function

2 v(t) 2
--- t + 1 1 – --- t + 1
T T
{ |
t
–T ⁄ 2 0 T⁄2

Figure 1.13. Equations for the linear segments of Figure 1.12

For line segment {,

v 1 ( t ) = ⎛ --- t + 1⎞ u 0 ⎛ t + --- ⎞ – u 0 ( t )
2 T
⎝T ⎠ ⎝ 2⎠
(1.15)

and for line segment |,

v 2 ( t ) = ⎛ – --- t + 1⎞ u 0 ( t ) – u 0 ⎛ t – --- ⎞
2 T
⎝ T ⎠ ⎝ 2⎠
(1.16)

Combining (1.15) and (1.16), we get

v ( t ) = v1 ( t ) + v2 ( t )
(1.17)
= ⎛ --- t + 1⎞ u 0 ⎛ t + T
--- ⎞ – u 0 ( t ) + ⎛ – --2- t + 1⎞ u 0 ( t ) – u 0 ⎛ t – T
--- ⎞
2
⎝T ⎠ ⎝ 2⎠ ⎝ T ⎠ ⎝ 2⎠

Example 1.5
Express the waveform of Figure 1.14 as a sum of unit step functions.

v( t)
3

t
0 1 2 3

Figure 1.14. Waveform for Example 1.5.

Solution:
As in the previous example, we first find the equations of the linear segments { and | shown in Fig-
ure 1.15.

Signals and Systems with MATLAB Applications, Second Edition 1-7

Orchard Publications
Chapter 1 Elementary Signals

v(t)
3
{
2
2t + 1
1 –t+3
|
t
0 1 2 3

Figure 1.15. Equations for the linear segments of Figure 1.14

Following the same procedure as in the previous examples, we get
v ( t ) = ( 2t + 1 ) [ u 0 ( t ) – u 0 ( t – 1 ) ] + 3 [ u 0 ( t – 1 ) – u 0 ( t – 2 ) ]
+ ( – t + 3 ) [ u0 ( t – 2 ) – u0 ( t – 3 ) ]

Multiplying the values in parentheses by the values in the brackets, we get

v ( t ) = ( 2t + 1 )u 0 ( t ) – ( 2t + 1 )u 0 ( t – 1 ) + 3u 0 ( t – 1 )
– 3u 0 ( t – 2 ) + ( – t + 3 )u 0 ( t – 2 ) – ( – t + 3 )u 0 ( t – 3 )

or
v ( t ) = ( 2t + 1 )u 0 ( t ) + [ – ( 2t + 1 ) + 3 ]u 0 ( t – 1 )
+ [ – 3 + ( – t + 3 ) ]u 0 ( t – 2 ) – ( – t + 3 )u 0 ( t – 3 )

and combining terms inside the brackets, we get

v ( t ) = ( 2t + 1 )u 0 ( t ) – 2 ( t – 1 )u 0 ( t – 1 ) – t u 0 ( t – 2 ) + ( t – 3 )u 0 ( t – 3 ) (1.18)

Two other functions of interest are the unit ramp function, and the unit impulse or delta function. We
will introduce them with the examples that follow.

Example 1.6
In the network of Figure 1.16 i S is a constant current source and the switch is closed at time t = 0 .

R
iS t = 0
+
vC ( t )
−
C

Figure 1.16. Network for Example 1.6

1-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Unit Step Function

Express the capacitor voltage v C ( t ) as a function of the unit step.

Solution:
The current through the capacitor is i C ( t ) = i S = cons tan t , and the capacitor voltage v C ( t ) is
t
1
∫– ∞ i
v C ( t ) = ---- *
C ( τ ) dτ (1.19)
C

where τ is a dummy variable.

Since the switch closes at t = 0 , we can express the current i C ( t ) as

iC ( t ) = iS u0 ( t ) (1.20)

and assuming that v C ( t ) = 0 for t < 0 , we can write (1.19) as

i 0
1 t ---S-
∫–∞ u0 ( τ ) dτ iS t
v C ( t ) = ----
C ∫– ∞ i S u 0 ( τ ) dτ = C + ----
C ∫ 0 u 0 ( τ ) dτ (1.21)
⎧
⎪
⎪
⎨
⎪
⎪
⎩
0
or

iS
v C ( t ) = ----- tu 0 ( t ) (1.22)
C

Therefore, we see that when a capacitor is charged with a constant current, the voltage across it is a
linear function and forms a ramp with slope i S ⁄ C as shown in Figure 1.17.

vC ( t )

slope = i S ⁄ C
t
0

Figure 1.17. Voltage across a capacitor when charged with a constant current source.

* Since the initial condition for the capacitor voltage was not specified, we express this integral with – ∞ at the
lower limit of integration so that any non-zero value prior to t < 0 would be included in the integration.

Signals and Systems with MATLAB Applications, Second Edition 1-9

Orchard Publications
Chapter 1 Elementary Signals

1.3 The Unit Ramp Function u 1 ( t )

The unit ramp function, denoted as u 1 ( t ) , is defined as

t
u1 ( t ) = ∫– ∞ u 0 ( τ ) d τ (1.23)

where τ is a dummy variable.

We can evaluate the integral of (1.23) by considering the area under the unit step function u 0 ( t ) from
– ∞ to t as shown in Figure 1.18.

Area = 1 × τ = τ = t
1

t
τ

Figure 1.18. Area under the unit step function from – ∞ to t

Therefore, we define u 1 ( t ) as

⎧0 t<0
u1 ( t ) = ⎨ (1.24)
⎩t t≥0

Since u 1 ( t ) is the integral of u 0 ( t ) , then u 0 ( t ) must be the derivative of u 1 ( t ) , i.e.,

d
----- u 1 ( t ) = u 0 ( t ) (1.25)
dt

Higher order functions of t can be generated by repeated integration of the unit step function. For
example, integrating u 0 ( t ) twice and multiplying by 2, we define u 2 ( t ) as

⎧0 t<0 t
u2 ( t ) = ⎨ 2
⎩t t≥0
or u2 ( t ) = 2 ∫–∞ u1 ( τ ) dτ (1.26)

Similarly,

⎧0 t<0 t
u3 ( t ) = ⎨ 3
⎩t t≥0
or u3 ( t ) = 3 ∫–∞ u2 ( τ ) dτ (1.27)

and in general,

1-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Unit Ramp Function

⎧0 t<0 t
un ( t ) = ⎨ n
⎩t t≥0
or un ( t ) = 3 ∫–∞ un – 1 ( τ ) dτ (1.28)

Also,
1d
u n – 1 ( t ) = --- ----- u n ( t ) (1.29)
n dt

Example 1.7
In the network of Figure 1.19, the switch is closed at time t = 0 and i L ( t ) = 0 for t < 0 .

R t = 0

iS +

`
iL ( t )
L −
v (t) L

Figure 1.19. Network for Example 1.7

Express the inductor current i L ( t ) in terms of the unit step function.

Solution:
The voltage across the inductor is
di L
v L ( t ) = L ------- (1.30)
dt

and since the switch closes at t = 0 ,

iL ( t ) = iS u0 ( t ) (1.31)
Therefore, we can write (1.30) as

d
v L ( t ) = Li S ----- u 0 ( t ) (1.32)
dt

But, as we know, u 0 ( t ) is constant ( 0 or 1 ) for all time except at t = 0 where it is discontinuous.

Since the derivative of any constant is zero, the derivative of the unit step u 0 ( t ) has a non-zero value
only at t = 0 . The derivative of the unit step function is defined in the next section.

Signals and Systems with MATLAB Applications, Second Edition 1-11

Orchard Publications
Chapter 1 Elementary Signals

1.4 The Delta Function δ ( t )

The unit impulse or delta function, denoted as δ ( t ) , is the derivative of the unit step u 0 ( t ) . It is also
defined as
t

∫– ∞ δ ( τ ) d τ = u0 ( t ) (1.33)

and

δ ( t ) = 0 for all t ≠ 0 (1.34)

To better understand the delta function δ ( t ) , let us represent the unit step u 0 ( t ) as shown in Figure
1.20 (a).
1
Figure (a)
0
t
−ε ε
1
Area =1 2ε Figure (b)
0
−ε ε t

Figure 1.20. Representation of the unit step as a limit.

The function of Figure 1.20 (a) becomes the unit step as ε → 0 . Figure 1.20 (b) is the derivative of
Figure 1.20 (a), where we see that as ε → 0 , 1 ⁄ 2 ε becomes unbounded, but the area of the rectangle
remains 1 . Therefore, in the limit, we can think of δ ( t ) as approaching a very large spike or impulse
at the origin, with unbounded amplitude, zero width, and area equal to 1 .
Two useful properties of the delta function are the sampling property and the sifting property.

1.5 Sampling Property of the Delta Function δ ( t )

The sampling property of the delta function states that

f ( t )δ ( t – a ) = f ( a )δ ( t ) (1.35)

or, when a = 0 ,

f ( t )δ ( t ) = f ( 0 )δ ( t ) (1.36)

1-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Sifting Property of the Delta Function

that is, multiplication of any function f ( t ) by the delta function δ ( t ) results in sampling the function
at the time instants where the delta function is not zero. The study of discrete-time systems is based
on this property.
Proof:
Since δ ( t ) = 0 for t < 0 and t > 0 then,
f ( t )δ ( t ) = 0 for t < 0 and t > 0 (1.37)
We rewrite f ( t ) as
f(t) = f(0) + [f(t) – f(0)] (1.38)
Integrating (1.37) over the interval – ∞ to t and using (1.38), we get
t t t

∫– ∞ f ( τ )δ ( τ ) dτ = ∫– ∞ f ( 0 )δ ( τ ) dτ + ∫–∞ [ f ( τ ) – f ( 0 ) ]δ ( τ ) dτ (1.39)

The first integral on the right side of (1.39) contains the constant term f ( 0 ) ; this can be written out-
side the integral, that is,
t t

∫– ∞ f ( 0 )δ ( τ ) dτ = f ( 0 ) ∫– ∞ δ ( τ ) d τ (1.40)

The second integral of the right side of (1.39) is always zero because
δ ( t ) = 0 for t < 0 and t > 0
and
[f(τ) – f(0 ) ] τ=0
= f( 0) – f(0 ) = 0
Therefore, (1.39) reduces to
t t

∫– ∞ f ( τ )δ ( τ ) dτ = f ( 0 ) ∫– ∞ δ ( τ ) d τ (1.41)

Differentiating both sides of (1.41), and replacing τ with t , we get

f ( t )δ ( t ) = f ( 0 )δ ( t )
(1.42)
Sampling Property of δ ( t )

1.6 Sifting Property of the Delta Function δ ( t )

The sifting property of the delta function states that

Signals and Systems with MATLAB Applications, Second Edition 1-13

Orchard Publications
Chapter 1 Elementary Signals

∫–∞ f ( t )δ ( t – α ) dt = f(α) (1.43)

that is, if we multiply any function f ( t ) by δ ( t – α ) , and integrate from – ∞ to +∞ , we will obtain the
value of f ( t ) evaluated at t = α .
Proof:
Let us consider the integral
b

∫a f ( t )δ ( t – α ) dt where a < α < b (1.44)

We will use integration by parts to evaluate this integral. We recall from the derivative of products
that
d ( xy ) = xdy + ydx or xdy = d ( xy ) – ydx (1.45)
and integrating both sides we get

∫ x dy = xy – y dx ∫ (1.46)

Now, we let x = f ( t ) ; then, dx = f ′( t ) . We also let dy = δ ( t – α ) ; then, y = u 0 ( t – α ) . By substitu-

tion into (1.46), we get
b b

∫a ∫a u0 ( t – α )f ′( t ) dt
b
f ( t )δ ( t – α ) dt = f ( t )u 0 ( t – α ) – (1.47)
a

We have assumed that a < α < b ; therefore, u 0 ( t – α ) = 0 for α < a , and thus the first term of the
right side of (1.47) reduces to f ( b ) . Also, the integral on the right side is zero for α < a , and there-
fore, we can replace the lower limit of integration a by α . We can now rewrite (1.47) as
b b

∫a f ( t )δ ( t – α ) dt = f ( b ) – ∫ α f ′( t ) d t = f(b ) – f(b) + f(α)

and letting a → – ∞ and b → ∞ for any α < ∞ , we get

∞

∫–∞ f ( t )δ ( t – α ) dt = f ( α ) (1.48)
Sifting Property of δ ( t )

1-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Higher Order Delta Functions

1.7 Higher Order Delta Functions

An nth-order delta function is defined as the nth derivative of u 0 ( t ) , that is,

n
n δ
δ ( t ) = ----- [ u 0 ( t ) ] (1.49)
dt

The function δ' ( t ) is called doublet, δ'' ( t ) is called triplet, and so on. By a procedure similar to the
derivation of the sampling property of the delta function, we can show that

f ( t )δ' ( t – a ) = f ( a )δ' ( t – a ) – f ' ( a )δ ( t – a ) (1.50)

Also, the derivation of the sifting property of the delta function can be extended to show that
∞ n
nd
∫
n
f ( t )δ ( t – α ) dt = ( – 1 ) -------n- [ f ( t ) ] (1.51)
–∞ dt t=α

Example 1.8
Evaluate the following expressions:
4
a. 3t δ ( t – 1 )
∞
b. ∫–∞ tδ ( t – 2 ) dt
2
c. t δ' ( t – 3 )
Solution:
4
a. The sampling property states that f ( t )δ ( t – a ) = f ( a )δ ( t – a ) For this example, f ( t ) = 3t and
a = 1 . Then,
4 4
3t δ ( t – 1 ) = { 3t t=1
}δ ( t – 1 ) = 3δ ( t – 1 )

∞
b. The sifting property states that ∫–∞ f ( t )δ ( t – α ) dt = f ( α ) . For this example, f ( t ) = t and α = 2 .

Then,
∞

∫–∞ tδ ( t – 2 ) dt = f ( 2 ) = t t = 2 = 2
c. The given expression contains the doublet; therefore, we use the relation

Signals and Systems with MATLAB Applications, Second Edition 1-15

Orchard Publications
Chapter 1 Elementary Signals

f ( t )δ' ( t – a ) = f ( a )δ' ( t – a ) – f ' ( a )δ ( t – a )

Then, for this example,
2 2 d 2
t δ' ( t – 3 ) = t δ' ( t – 3 ) – ----- t t = 3 δ ( t – 3 )
t=3 dt
= 9δ' ( t – 3 ) – 6δ ( t – 3 )

Example 1.9
a. Express the voltage waveform v ( t ) shown in Figure 1.21 as a sum of unit step functions for the
time interval – 1 < t < 7 s .
b. Using the result of part (a), compute the derivative of v ( t ) and sketch its waveform.

v(t) (V)

1
−1 1 2 3 4 5 6 7
0
t (s)

−1

−2

Figure 1.21. Waveform for Example 1.9

Solution:
a. We first derive the equations for the linear segments of the given waveform. These are shown in
Figure 1.22.
Next, we express v ( t ) in terms of the unit step function u 0 ( t ) , and we get

v ( t ) = 2t [ u 0 ( t + 1 ) – u 0 ( t – 1 ) ] + 2 [ u 0 ( t – 1 ) – u 0 ( t – 2 ) ]
+ ( – t + 5 ) [ u0 ( t – 2 ) – u0 ( t – 4 ) ] + [ u0 ( t – 4 ) – u0 ( t – 5 ) ] (1.52)
+ ( – t + 6 ) [ u0 ( t – 5 ) – u0 ( t – 7 ) ]

Multiplying and collecting like terms in (1.52), we get

1-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Higher Order Delta Functions

v(t) (V) v(t)

–t+5
3

2 –t+6

1
−1 0
1 2 3 4 5 6 7

t (s)

−1
2t
−2

Figure 1.22. Equations for the linear segments of Figure 1.21

v ( t ) = 2tu 0 ( t + 1 ) – 2tu 0 ( t – 1 ) – 2u 0 ( t – 1 ) – 2u 0 ( t – 2 ) – tu 0 ( t – 2 )
+ 5u 0 ( t – 2 ) + tu 0 ( t – 4 ) – 5u 0 ( t – 4 ) + u 0 ( t – 4 ) – u 0 ( t – 5 )
– tu 0 ( t – 5 ) + 6u 0 ( t – 5 ) + tu 0 ( t – 7 ) – 6u 0 ( t – 7 )
or
v ( t ) = 2tu 0 ( t + 1 ) + ( – 2t + 2 )u 0 ( t – 1 ) + ( – t + 3 )u 0 ( t – 2 )
+ ( t – 4 )u 0 ( t – 4 ) + ( – t + 5 )u 0 ( t – 5 ) + ( t – 6 )u 0 ( t – 7 )

b. The derivative of v ( t ) is
dv
------ = 2u 0 ( t + 1 ) + 2tδ ( t + 1 ) – 2u 0 ( t – 1 ) + ( – 2t + 2 )δ ( t – 1 )
dt
– u 0 ( t – 2 ) + ( – t + 3 )δ ( t – 2 ) + u 0 ( t – 4 ) + ( t – 4 )δ ( t – 4 ) (1.53)
– u 0 ( t – 5 ) + ( – t + 5 )δ ( t – 5 ) + u 0 ( t – 7 ) + ( t – 6 )δ ( t – 7 )

From the given waveform, we observe that discontinuities occur only at t = – 1 , t = 2 , and
t = 7 . Therefore, δ ( t – 1 ) = 0 , δ ( t – 4 ) = 0 , and δ ( t – 5 ) = 0 , and the terms that contain these
delta functions vanish. Also, by application of the sampling property,
2tδ ( t + 1 ) = { 2t t = –1
}δ ( t + 1 ) = – 2δ ( t + 1 )
( – t + 3 )δ ( t – 2 ) = { ( – t + 3 ) t=2
}δ ( t – 2 ) = δ ( t – 2 )
( t – 6 )δ ( t – 7 ) = { ( t – 6 ) t=7
}δ ( t – 7 ) = δ ( t – 7 )

and by substitution into (1.53), we get

Signals and Systems with MATLAB Applications, Second Edition 1-17

Orchard Publications
Chapter 1 Elementary Signals

dv
------ = 2u 0 ( t + 1 ) – 2 δ ( t + 1 ) – 2u 0 ( t – 1 ) – u 0 ( t – 2 )
dt (1.54)
+ δ ( t – 2 ) + u0 ( t – 4 ) – u0 ( t – 5 ) + u0 ( t – 7 ) + δ ( t – 7 )

The plot of dv ⁄ dt is shown in Figure 1.23.

dv
------ (V ⁄ s)
dt
2

δ(t – 2) δ(t – 7)
1

−1 0 1 2 3 4 5 6 7
t (s)
−1

– 2δ ( t + 1 )

Figure 1.23. Plot of the derivative of the waveform of Figure 1.21.

We observe that a negative spike of magnitude 2 occurs at t = – 1 , and two positive spikes of
magnitude 1 occur at t = 2 , and t = 7 . These spikes occur because of the discontinuities at
these points.

MATLAB* has built-in functions for the unit step, and the delta functions. These are denoted by the
names of the mathematicians who used them in their work. The unit step function u 0 ( t ) is referred
to as Heaviside(t), and the delta function δ ( t ) is referred to as Dirac(t). Their use is illustrated with
the examples below.
syms k a t; % Define symbolic variables
u=k*sym('Heaviside(t-a)') % Create unit step function at t = a
u =
k*Heaviside(t-a)
d=diff(u) % Compute the derivative of the unit step function
d =
k*Dirac(t-a)

* An introduction to MATLAB® is given in Appendix A.

1-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

int(d) % Integrate the delta function

ans =
Heaviside(t-a)*k

1.8 Summary
• The unit step function u 0 ( t ) that is defined as

⎧0 t<0
u0 ( t ) = ⎨
⎩1 t>0

• The unit step function offers a convenient method of describing the sudden application of a volt-
age or current source.
• The unit ramp function, denoted as u 1 ( t ) , is defined as

t
u1 ( t ) = ∫– ∞ u 0 ( τ ) d τ
• The unit impulse or delta function, denoted as δ ( t ) , is the derivative of the unit step u 0 ( t ) . It is also
defined as
t

∫– ∞ δ ( τ ) d τ = u0 ( t )

and
δ ( t ) = 0 for all t ≠ 0
• The sampling property of the delta function states that

f ( t )δ ( t – a ) = f ( a )δ ( t )
or, when a = 0 ,
f ( t )δ ( t ) = f ( 0 )δ ( t )
• The sifting property of the delta function states that
∞

∫–∞ f ( t )δ ( t – α ) dt = f(α)

• The sampling property of the doublet function δ' ( t ) states that

f ( t )δ' ( t – a ) = f ( a )δ' ( t – a ) – f ' ( a )δ ( t – a )

Signals and Systems with MATLAB Applications, Second Edition 1-19

Orchard Publications
Chapter 1 Elementary Signals

1.9 Exercises
1. Evaluate the following functions:

a. sin tδ ⎛⎝ t – π
---⎞
6⎠

b. cos 2tδ ⎛⎝ t – π
---⎞
⎠ 4

c. cos t δ ⎛⎝ t – π
---⎞
2
2⎠

d. tan 2tδ ⎛⎝ t – π
---⎞
⎠ 8

∞
2 –t
e. ∫– ∞ t e δ ( t – 2 ) dt

f. sin t δ 1 ⎛⎝ t – π
---⎞
2
2⎠

2.
a. Express the voltage waveform v ( t ) shown in Figure 1.24, as a sum of unit step functions for
the time interval 0 < t < 7 s .
b. Using the result of part (a), compute the derivative of v ( t ) , and sketch its waveform.

v(t) (V) v(t)

20
– 2t
e
10

0
1 2 3 4 5 6 7 t( s)
−10

−20

Figure 1.24. Waveform for Exercise 2

1-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

1.10 Solutions to Exercises

Dear Reader:
The remaining pages on this chapter contain the solutions to the exercises.
You must, for your benefit, make an honest effort to solve the problems without first looking at the
solutions that follow. It is recommended that first you go through and solve those you feel that you
know. For the exercises that you are uncertain, review this chapter and try again. If your results do
not agree with those provided, look over your procedures for inconsistencies and computational
errors. Refer to the solutions as a last resort and rework those problems at a later date.
You should follow this practice with the exercises on all chapters of this book.

Signals and Systems with MATLAB Applications, Second Edition 1-21

Orchard Publications
Chapter 1 Elementary Signals

1. We apply the sampling property of the δ ( t ) function for all expressions except (e) where we apply
the sifting property. For part (f) we apply the sampling property of the doublet.
We recall that the sampling property states that f ( t )δ ( t – a ) = f ( a )δ ( t – a ) . Thus,

π
a. sin tδ ⎛⎝ t – π
---⎞ = sin t
π π π
δ ⎛ t – ---⎞ = sin --- δ ⎛ t – ---⎞ = 0.5δ ⎛ t – ---⎞
6⎠ t = π⁄6 ⎝ 6⎠ 6 ⎝ 6⎠ ⎝ 6⎠

π
b. cos 2tδ ⎛⎝ t – π
---⎞ = cos 2t
⎠ ⎝ 4⎠
---⎞ = cos --- δ ⎛ t – π
δ⎛ t – π --- ⎞ = 0
4 t = π⁄4 2 ⎝ 4⎠

c. cos t δ ⎛⎝ t – π
---⎞ = --- ( 1 + cos 2t )
π π π
δ ⎛ t – ---⎞ = --- ( 1 + cos π )δ ⎛ t – ---⎞ = --- ( 1 – 1 )δ ⎛ t – ---⎞ = 0
2 1 1 1
2⎠ 2 t=π⁄2
⎝ 2⎠ 2 ⎝ 2⎠ 2 ⎝ 2⎠

π
d. tan 2tδ ⎛⎝ t – π
---⎞ = tan 2t
⎠
π π π
δ ⎛ t – ---⎞ = tan --- δ ⎛ t – --- ⎞ = δ ⎛ t – ---⎞
⎝ 8⎠
8 t = π⁄8 4 ⎝ 8⎠ ⎝ 8⎠

∞
We recall that the sampling property states that ∫–∞ f ( t )δ ( t – α ) dt = f ( α ) . Thus,

∞
2 –t 2 –t –2
e. ∫– ∞ t e δ ( t – 2 ) dt = t e t=2
= 4e = 0.54

We recall that the sampling property for the doublet states that
f ( t )δ' ( t – a ) = f ( a )δ' ( t – a ) – f ' ( a )δ ( t – a )
Thus,

π π π
sin t δ ⎛ t – --- ⎞ = sin t δ ⎛ t – --- ⎞ – ----- sin t δ ⎛ t – --- ⎞
2 1 2 1 d 2
⎝ 2⎠ t = π⁄2 ⎝ 2 ⎠ dt t = π⁄2 ⎝ 2⎠

π π
δ ⎛ t – --- ⎞ – sin 2t δ ⎛ t – --- ⎞
1
= --- ( 1 – cos 2t )
1
f. t = π⁄2 ⎝ 2⎠ t = π⁄2 ⎝ 2⎠
2
π π π
= --- ( 1 + 1 )δ ⎛ t – --- ⎞ – sin πδ ⎛ t – --- ⎞ = δ ⎛ t – --- ⎞
1 1 1
2 ⎝ 2⎠ ⎝ 2⎠ ⎝ 2⎠

2.
– 2t
v( t) = e [ u 0 ( t ) – u 0 ( t – 2 ) ] + ( 10t – 30 ) [ u 0 ( t – 2 ) – u 0 ( t – 3 ) ]
a.
+ ( – 10 t + 50 ) [ u 0 ( t – 3 ) – u 0 ( t – 5 ) ] + ( 10t – 70 ) [ u 0 ( t – 5 ) – u 0 ( t – 7 ) ]

or

1-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
– 2t – 2t
v( t) = e u0 ( t ) – e u 0 ( t – 2 ) + 10tu 0 ( t – 2 ) – 30u 0 ( t – 2 ) – 10tu 0 ( t – 3 ) + 30u 0 ( t – 3 )
– 10tu 0 ( t – 3 ) + 50u 0 ( t – 3 ) + 10tu 0 ( t – 5 ) – 50u 0 ( t – 5 ) + 10tu 0 ( t – 5 )
– 70u 0 ( t – 5 ) – 10tu 0 ( t – 7 ) + 70u 0 ( t – 7 )
– 2t – 2t
= e u0 ( t ) + ( –e + 10t – 30 )u 0 ( t – 2 ) + ( – 20t + 80 )u 0 ( t – 3 ) + ( 20t – 120 )u 0 ( t – 5 )
+ ( – 10t + 70 )u 0 ( t – 7 )

b.
dv – 2t – 2t – 2t – 2t
------ = – 2e u 0 ( t ) + e δ ( t ) + ( 2e + 10 )u 0 ( t – 2 ) + ( – e + 10t – 30 )δ ( t – 2 )
dt
– 20u 0 ( t – 3 ) + ( – 20t + 80 )δ ( t – 3 ) + 20u 0 ( t – 5 ) + ( 20t – 120 )δ ( t – 5 ) (1)
– 10u 0 ( t – 7 ) + ( – 10t + 70 )δ ( t – 7 )

Referring to the given waveform we observe that discontinuities occur only at t = 2 , t = 3 ,

and t = 5 . Therefore, δ ( t ) = 0 and δ ( t – 7 ) = 0 . Also, by the sampling property of the delta
function
– 2t – 2t
( –e + 10t – 30 )δ ( t – 2 ) = ( – e + 10t – 30 ) t=2
δ ( t – 2 ) ≈ – 10δ ( t – 2 )

( – 20t + 80 )δ ( t – 3 ) = ( – 20t + 80 ) t=3

δ ( t – 3 ) = 20δ ( t – 3 )

( 20t – 120 )δ ( t – 5 ) = ( 20t – 120 ) t=5

δ ( t – 5 ) = – 20 δ ( t – 5 )

and with these simplifications (1) above reduces to

– 2t – 2t
dv ⁄ dt = – 2e u 0 ( t ) + 2e u 0 ( t – 2 ) + 10u 0 ( t – 2 ) – 10δ ( t – 2 )
– 20u 0 ( t – 3 ) + 20δ ( t – 3 ) + 20u 0 ( t – 5 ) – 20δ ( t – 5 ) – 10u 0 ( t – 7 )
– 2t
= – 2e [ u 0 ( t ) – u 0 ( t – 2 ) ] – 10δ ( t – 2 ) + 10 [ u 0 ( t – 2 ) – u 0 ( t – 3 ) ] + 20δ ( t – 3 )
– 10 [ u 0 ( t – 3 ) – u 0 ( t – 5 ) ] – 20δ ( t – 5 ) + 10 [ u 0 ( t – 5 ) – u 0 ( t – 7 ) ]

The waveform for dv ⁄ dt is shown below.

dv ⁄ dt (V ⁄ s)
20 δ ( t – 3 )
20
10

1 2 3 4 5 6 7 t (s)
– 10
– 10δ ( t – 2 )
– 20
– 2t – 20 δ ( t – 5 )
– 2e

Signals and Systems with MATLAB Applications, Second Edition 1-23

Orchard Publications
Chapter 1 Elementary Signals

NOTES

1-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 2
The Laplace Transformation

his chapter begins with an introduction to the Laplace transformation, definitions, and proper-
T ties of the Laplace transformation. The initial value and final value theorems are also discussed
and proved. It concludes with the derivation of the Laplace transform of common functions
of time, and the Laplace transforms of common waveforms.

2.1 Definition of the Laplace Transformation

The two-sided or bilateral Laplace Transform pair is defined as
∞
– st
L { f( t) }= F(s) = ∫– ∞ f ( t ) e dt (2.1)

σ + jω
–1 1
∫σ – jω
st
L { F ( s ) } = f ( t ) = -------- F ( s ) e ds (2.2)
2πj

–1
where L { f ( t ) } denotes the Laplace transform of the time function f ( t ) , L { F ( s ) } denotes the
Inverse Laplace transform, and s is a complex variable whose real part is σ , and imaginary part ω ,
that is, s = σ + jω .
In most problems, we are concerned with values of time t greater than some reference time, say
t = t 0 = 0 , and since the initial conditions are generally known, the two-sided Laplace transform
pair of (2.1) and (2.2) simplifies to the unilateral or one-sided Laplace transform defined as

∞ ∞
– st – st
L {f(t)}= F(s) = ∫t 0
f( t)e dt = ∫0 f ( t ) e dt (2.3)

σ + jω
–1 1
∫σ – jω
st
L { F ( s ) } = f ( t ) = -------- F ( s ) e ds (2.4)
2πj

The Laplace Transform of (2.3) has meaning only if the integral converges (reaches a limit), that is, if
∞
– st
∫0 f ( t ) e dt < ∞ (2.5)

To determine the conditions that will ensure us that the integral of (2.3) converges, we rewrite (2.5)

Signals and Systems with MATLAB Applications, Second Edition 2-1

Orchard Publications
Chapter 2 The Laplace Transformation

as
∞
– σt – jωt
∫0 f ( t )e e dt < ∞ (2.6)

– jωt – jωt
The term e in the integral of (2.6) has magnitude of unity, i.e., e = 1 , and thus the condition
for convergence becomes
∞
– σt
∫0 f ( t )e dt < ∞ (2.7)

Fortunately, in most engineering applications the functions f ( t ) are of exponential order*. Then, we
can express (2.7) as,
∞ ∞ σ 0 t – σt
– σt
∫0 f ( t )e dt < ∫0 ke e dt (2.8)

and we see that the integral on the right side of the inequality sign in (2.8), converges if σ > σ 0 .
Therefore, we conclude that if f ( t ) is of exponential order, L { f ( t ) } exists if
Re { s } = σ > σ 0 (2.9)

where Re { s } denotes the real part of the complex variable s .

Evaluation of the integral of (2.4) involves contour integration in the complex plane, and thus, it will
not be attempted in this chapter. We will see, in the next chapter, that many Laplace transforms can
be inverted with the use of a few standard pairs, and therefore, there is no need to use (2.4) to obtain
the Inverse Laplace transform.
In our subsequent discussion, we will denote transformation from the time domain to the complex
frequency domain, and vice versa, as
f(t) ⇔ F(s) (2.10)

2.2 Properties of the Laplace Transform

1. Linearity Property
The linearity property states that if
f 1 ( t ), f 2 ( t ), …, f n ( t )
have Laplace transforms

σ0 t
* A function f ( t ) is said to be of exponential order if f ( t ) < ke for all t ≥ 0 .

2-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties of the Laplace Transform

F 1 ( s ), F 2 ( s ), … , F n ( s )
respectively, and
c 1 , c 2 , …, c n
are arbitrary constants, then,

c1 f1 ( t ) + c2 f2 ( t ) + … + cn fn ( t ) ⇔ c1 F1 ( s ) + c2 F2 ( s ) + … + cn Fn ( s ) (2.11)

Proof:
∞
L { c1 f1 ( t ) + c2 f2 ( t ) + … + cn fn ( t ) } = ∫t 0
[ c 1 f 1 ( t ) + c 2 f 2 ( t ) + … + c n f n ( t ) ] dt

∞ ∞ ∞
– st – st – st
= c1 ∫t 0
f1 ( t ) e dt + c 2 ∫t 0
f2 ( t ) e dt + … + c n ∫t 0
fn ( t ) e dt

= c1 F1 ( s ) + c2 F2 ( s ) + … + cn Fn ( s )

Note 1:
It is desirable to multiply f ( t ) by u 0 ( t ) to eliminate any unwanted non-zero values of f ( t ) for t < 0 .

2. Time Shifting Property

The time shifting property states that a right shift in the time domain by a units, corresponds to mul-
– as
tiplication by e in the complex frequency domain. Thus,

– as
f ( t – a )u 0 ( t – a ) ⇔ e F(s) (2.12)

Proof:
a ∞
– st – st
L { f ( t – a )u 0 ( t – a ) } = ∫0 0e dt + ∫ a f(t – a)e dt (2.13)

Now, we let t – a = τ ; then, t = τ + a and dt = dτ . With these substitutions, the second integral
on the right side of (2.13) becomes
∞ ∞
–s ( τ + a ) – as – sτ – as
∫0 f(τ)e dτ = e ∫0 f ( τ ) e dτ = e F(s)

3. Frequency Shifting Property

The frequency shifting property states that if we multiply some time domain function f ( t ) by an
– at
exponential function e where a is an arbitrary positive constant, this multiplication will produce a
shift of the s variable in the complex frequency domain by a units. Thus,

Signals and Systems with MATLAB Applications, Second Edition 2-3

Orchard Publications
Chapter 2 The Laplace Transformation

– at
e f(t) ⇔ F(s + a) (2.14)

Proof:
∞ ∞
– at – at – st – ( s + a )t
L {e f(t)} = ∫0 e f(t)e dt = ∫0 f ( t ) e dt = F ( s + a )

Note 2:
A change of scale is represented by multiplication of the time variable t by a positive scaling factor
a . Thus, the function f ( t ) after scaling the time axis, becomes f ( at ) .
4. Scaling Property
Let a be an arbitrary positive constant; then, the scaling property states that

f ( at ) ⇔ --- F ⎛ --- ⎞
1 s
(2.15)
a ⎝a⎠

Proof:
∞
– st
L { f ( at ) } = ∫0 f ( at ) e dt

and letting t = τ ⁄ a , we get

∞ ∞
τ
d ⎛ ---⎞ = --- d ( τ ) = --- F ⎛ ---⎞
–s ( τ ⁄ a ) 1 –( s ⁄ a ) τ 1 s
L { f ( at ) } = ∫0 f(τ )e
⎝ a⎠ a ∫0 f ( τ ) e a ⎝ a⎠
Note 3:
Generally, the initial value of f ( t ) is taken at t = 0 − to include any discontinuity that may be present
at t = 0 . If it is known that no such discontinuity exists at t = 0− , we simply interpret f ( 0 − ) as f ( 0 ) .

5. Differentiation in Time Domain

The differentiation in time domain property states that differentiation in the time domain corresponds
to multiplication by s in the complex frequency domain, minus the initial value of f ( t ) at t = 0 − .
Thus,

d −
f ' ( t ) = ----- f ( t ) ⇔ sF ( s ) – f ( 0 ) (2.16)
dt

Proof:
∞
– st
L {f '(t)} = ∫0 f ' ( t ) e dt

2-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties of the Laplace Transform

Using integration by parts where

∫ v du = uv – u dv ∫ (2.17)

– st – st
we let du = f ' ( t ) and v = e . Then, u = f ( t ) , dv = – se , and thus
∞
– st ∞ – st – st a
L { f ' ( t ) } = f ( t )e
0
−
+s ∫0 −
f(t)e dt = lim
a→∞
f ( t )e
0
−
+ sF ( s )

– sa − −
= lim [ e f ( a ) – f ( 0 ) ] + sF ( s ) = 0 – f ( 0 ) + sF ( s )
a→∞

The time differentiation property can be extended to show that

d2
-------- f ( t ) ⇔ s 2 F ( s ) – sf ( 0 − ) – f ' ( 0 − ) (2.18)
2
dt

d3
-------- f ( t ) ⇔ s 3 F ( s ) – s 2 f ( 0 − ) – sf ' ( 0 − ) – f '' ( 0 − ) (2.19)
3
dt

and in general
n
d
-------- f ( t ) ⇔ s n F ( s ) – s n – 1 f ( 0 − ) – s n – 2 f ' ( 0 − ) – … – f n–1
(0 )
−
(2.20)
n
dt

To prove (2.18), we let

d
g ( t ) = f ' ( t ) = ----- f ( t )
dt
and as we found above,
−
L { g ' ( t ) } = sL { g ( t ) } – g ( 0 )
Then,
− − −
L { f '' ( t ) } = sL { f ' ( t ) } – f ' ( 0 ) = s [ sL [ f ( t ) ] – f ( 0 ) ] – f ' ( 0 )
− −
= s 2 F ( s ) – sf ( 0 ) – f ' ( 0 )
Relations (2.19) and (2.20) can be proved by similar procedures.

We must remember that the terms f ( 0 − ), f ' ( 0 − ), f '' ( 0 − ) , and so on, represent the initial conditions.
Therefore, when all initial conditions are zero, and we differentiate a time function f ( t ) n times,
this corresponds to F ( s ) multiplied by s to the nth power.

Signals and Systems with MATLAB Applications, Second Edition 2-5

Orchard Publications
Chapter 2 The Laplace Transformation

6. Differentiation in Complex Frequency Domain

This property states that differentiation in complex frequency domain and multiplication by minus one,
corresponds to multiplication of f ( t ) by t in the time domain. In other words,

d
tf ( t ) ⇔ – ----- F ( s ) (2.21)
ds

Proof:
∞
– st
L { f( t)} = F(s) = ∫0 f ( t ) e dt

Differentiating with respect to s, and applying Leibnitz’s rule* for differentiation under the integral, we
get
∞ ∞ ∞
d- d – st ∂ e – st f ( t )dt = – st
----
ds
F ( s ) = -----
ds ∫0 f( t )e dt = ∫0 ∂s ∫0 – t e f ( t )dt

∞
– st
= – ∫0 [ tf ( t ) ] e dt = – L [ tf ( t ) ]

In general,
n
n nd
t f ( t ) ⇔ ( – 1 ) -------n- F ( s ) (2.22)
ds

The proof for n ≥ 2 follows by taking the second and higher-order derivatives of F ( s ) with respect
to s .
7. Integration in Time Domain
This property states that integration in time domain corresponds to F ( s ) divided by s plus the initial
value of f ( t ) at t = 0 − , also divided by s . That is,

t −
F ( s -) + f------------
(0 )
∫ –∞
f ( τ ) dτ ⇔ ----------
s s
(2.23)

b
* This rule states that if a function of a parameter α is defined by the equation F ( α ) = ∫a f ( x, α ) dx where f is some

known function of integration x and the parameter α , a and b are constants independent of x and α , and the par-
dF- b
∂( x, α )
tial derivative ∂f ⁄ ∂α exists and it is continuous, then ------ =
dα ∫a ----------------
∂( α )
- dx .

2-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties of the Laplace Transform

Proof:
We express the integral of (2.23) as two integrals, that is,
t 0 t

∫– ∞ f ( τ ) dτ = ∫– ∞ f ( τ ) dτ + ∫ 0 f ( τ ) dτ (2.24)

The first integral on the right side of (2.24), represents a constant value since neither the upper, nor
the lower limits of integration are functions of time, and this constant is an initial condition denoted
as f ( 0 − ) . We will find the Laplace transform of this constant, the transform of the second integral
on the right side of (2.24), and will prove (2.23) by the linearity property. Thus,

∞ ∞ – st ∞
– st – st e
∫0 f ( 0 ) e ∫0 e
− − − −
L {f (0 )} = dt = f ( 0 ) dt = f ( 0 ) --------
–s 0 (2.25)
− −
f(0 ) f(0 )
= f ( 0 ) × 0 – ⎛ – ------------⎞ = ------------
−
⎝ s ⎠ s

This is the value of the first integral in (2.24). Next, we will show that
t
F(s)
∫0 f ( τ ) dτ ⇔ ----------
s
-

We let
t
g(t) = ∫0 f ( τ ) dτ
then,
g' ( t ) = f ( τ )
and
0
g( 0) = ∫0 f ( τ ) dτ = 0

Now,
−
L { g' ( t ) } = G ( s ) = sL { g ( t ) } – g ( 0 ) = G ( s ) – 0

sL { g ( t ) } = G ( s )

G(s)
L { g ( t ) } = -----------
s

⎧ t ⎫ F(s)
L ⎨
⎩
∫0 f ( τ ) dτ ⎬⎭ = -----------
s
(2.26)

and the proof of (2.23) follows from (2.25) and (2.26).

Signals and Systems with MATLAB Applications, Second Edition 2-7

Orchard Publications
Chapter 2 The Laplace Transformation

8. Integration in Complex Frequency Domain

This property states that integration in complex frequency domain with respect to s corresponds to
f ( t ) exists. Thus,
division of a time function f ( t ) by the variable t , provided that the limit lim -------
-
t→0 t

∞
f( t)
-------- ⇔
t ∫s F ( s ) ds (2.27)

Proof:
∞
– st
F(s) = ∫0 f ( t ) e dt

Integrating both sides from s to ∞ , we get

∞ ∞ ∞
– st
∫s F ( s ) ds = ∫s ∫0 f ( t ) e dt ds

Next, we interchange the order of integration, i.e.,

∞ ∞ ∞
– st
∫s F ( s ) ds = ∫0 ∫s e ds f ( t ) dt

and performing the inner integration on the right side integral with respect to s , we get

∞ ∞ ∞ ⎧ f ( t )⎫
– st ∞ f(t) – st
∫s F ( s ) ds = ∫0 –1
--- e
t s
f ( t ) dt = ∫0 -------t - e dt = L ⎨ --------⎬
⎩ t ⎭

9. Time Periodicity
The time periodicity property states that a periodic function of time with period T corresponds to
T
– st – sT
the integral ∫0 f ( t ) e dt divided by ( 1 – e ) in the complex frequency domain. Thus, if we let f ( t )

be a periodic function with period T , that is, f ( t ) = f ( t + nT ) , for n = 1, 2, 3, … we get the trans-
form pair

T – st
∫0 f ( t ) e dt
f ( t + nT ) ⇔ ----------------------------
– sT
- (2.28)
1–e

2-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties of the Laplace Transform

Proof:
The Laplace transform of a periodic function can be expressed as
∞ T 2T 3T
– st – st – st – st
L {f(t)} = ∫0 f( t )e dt = ∫0 f( t)e dt + ∫T f( t)e dt + ∫ 2T f ( t ) e dt + …

In the first integral of the right side, we let t = τ , in the second t = τ + T , in the third t = τ + 2T ,
and so on. The areas under each period of f ( t ) are equal, and thus the upper and lower limits of
integration are the same for each integral. Then,
T T T
– sτ –s ( τ + T ) – s ( τ + 2T )
L {f(t)} = ∫0 f(τ)e dτ + ∫0 f(τ + T )e dτ + ∫0 f ( τ + 2T ) e dτ + … (2.29)

Since the function is periodic, i.e., f ( τ ) = f ( τ + T ) = f ( τ + 2T ) = … = f ( τ + nT ) , we can write

(2.29) as
T
– sT – 2sT – sτ
L {f(τ)} = (1 + e +e + …) ∫0 f ( τ ) e dτ (2.30)

By application of the binomial theorem, that is,

2 3 1 -
1 + a + a + a + … = ----------- (2.31)
1–a
we find that expression (2.30) reduces to
T
– sτ
∫0
f ( τ ) e dτ
L { f ( τ ) } = ----------------------------------
– sT
τ–e
10. Initial Value Theorem
The initial value theorem states that the initial value f ( 0 − ) of the time function f ( t ) can be found
from its Laplace transform multiplied by s and letting s → ∞ .That is,
−
lim f ( t ) = lim sF ( s ) = f ( 0 ) (2.32)
t→0 s→∞

Proof:
From the time domain differentiation property,
d- −
---- f ( t ) ⇔ sF ( s ) – f ( 0 )
dt
or

Signals and Systems with MATLAB Applications, Second Edition 2-9

Orchard Publications
Chapter 2 The Laplace Transformation

⎧d ⎫ ∞
d
----- f ( t ) e –st dt
∫0
−
L ⎨ ----- f ( t ) ⎬ = sF ( s ) – f ( 0 ) =
⎩ dt ⎭ dt

Taking the limit of both sides by letting s → ∞ , we get

T
d – st
∫ ----- f ( t ) e
−
lim [ sF ( s ) – f ( 0 ) ] = lim lim dt
s→∞ s→∞ T → ∞ ε dt
ε→0

Interchanging the limiting process, we get

T
d – st
T → ∞ ∫ ε dt
−
lim [ sF ( s ) – f ( 0 ) ] = lim ----- f ( t ) lim e dt
s→∞ s→∞
ε→0
and since
– st
lim e = 0
s→∞
the above expression reduces to
−
lim [ sF ( s ) – f ( 0 ) ] = 0
s→∞
or
−
lim sF ( s ) = f ( 0 )
s→∞

11. Final Value Theorem

The final value theorem states that the final value f ( ∞ ) of the time function f ( t ) can be found from
its Laplace transform multiplied by s, then, letting s → 0 . That is,

lim f ( t ) = lim sF ( s ) = f ( ∞ ) (2.33)

t→∞ s→0

Proof:
From the time domain differentiation property,
d
----- f ( t ) ⇔ sF ( s ) – f ( 0 − )
dt
or
⎧d ⎫ ∞
d
----- f ( t ) e – st dt
∫0
−
L ⎨ ----- f ( t ) ⎬ = sF ( s ) – f ( 0 ) =
⎩ dt ⎭ dt

Taking the limit of both sides by letting s → 0 , we get

2-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties of the Laplace Transform

T
d – st
∫ ----- f ( t ) e
−
lim [ sF ( s ) – f ( 0 ) ] = lim lim dt
s→0 s→0 T → ∞ ε dt
ε→0

and by interchanging the limiting process, we get

T
d – st
∫ ----
−
lim [ sF ( s ) – f ( 0 ) ] = lim - f(t) lim e dt
s→0 T→∞ ε
dt s→0
ε→0
Also, since
– st
lim e = 1
s→0
the above expression reduces to
T T
d-
∫ ∫ε f ( t )
−
lim [ sF ( s ) – f ( 0 ) ] = lim ---- f ( t ) dt = lim
s→0 T→∞ ε dt T→∞
ε→0 ε→0
−
= lim [ f ( T ) – f ( ε ) ] = f ( ∞ ) – f ( 0 )
T→∞
ε→0
and therefore,
lim sF ( s ) = f ( ∞ )
s→0

12. Convolution in the Time Domain

Convolution* in the time domain corresponds to multiplication in the complex frequency domain,
that is,
f 1 ( t )*f 2 ( t ) ⇔ F 1 ( s )F 2 ( s ) (2.34)

Proof:
∞ ∞ ∞
– st
L { f 1 ( t )*f 2 ( t ) } = L ∫– ∞ f 1 ( τ )f 2 ( t – τ ) dτ = ∫0 ∫0 f1 ( τ )f2 ( t – τ ) dτ e dt
(2.35)
∞ ∞
– st
= ∫ 0 f 1 ( τ ) ∫0 f 2 ( t – τ ) e dt dτ

We let t – τ = λ ; then, t = λ + τ , and dt = dλ . By substitution into (2.35),

* Convolution is the process of overlapping two signals. The convolution of two time functions f 1 ( t ) and f 2 ( t ) is
∞
denoted as f 1 ( t )*f 2 ( t ) , and by definition, f 1 ( t )*f 2 ( t ) = ∫–∞ f1 ( τ )f2 ( t – τ ) dτ where τ is a dummy variable. We will
discuss it in detail in Chapter 6.

Signals and Systems with MATLAB Applications, Second Edition 2-11

Orchard Publications
Chapter 2 The Laplace Transformation

∞ ∞ ∞ ∞
–s ( λ + τ ) – sτ – sλ
L { f 1 ( t )*f 2 ( t ) } = ∫0 f1 ( τ ) ∫0 f2 ( λ ) e dλ dτ = ∫0 f 1 ( τ )e dτ ∫0 f 2 ( λ ) e dλ

= F 1 ( s )F 2 ( s )

13. Convolution in the Complex Frequency Domain

Convolution in the complex frequency domain divided by 1 ⁄ 2πj , corresponds to multiplication in the
time domain. That is,
1
f 1 ( t )f 2 ( t ) ⇔ -------- F 1 ( s )*F 2 ( s ) (2.36)
2πj
Proof:
∞
– st
L { f 1 ( t )f 2 ( t ) } = ∫0 f1 ( t )f2 ( t ) e dt (2.37)

and recalling that the Inverse Laplace transform from (2.2) is

σ + jω
1 µt
f 1 ( t ) = --------
2πj ∫σ – jω F 1 ( µ )e dµ

by substitution into (2.37), we get

∞ σ + jω
1 µt – st
L { f 1 ( t )f 2 ( t ) } = ∫0 --------
2πj ∫σ – jω F 1 ( µ )e dµ f 2 ( t ) e dt

σ + jω ∞
1 – ( s – µ )t
= --------
2πj ∫σ – jω F1 ( µ ) ∫0 f 2 ( t ) e dt dµ

We observe that the bracketed integral is F 2 ( s – µ ) ; therefore,

σ + jω
1 1
L { f 1 ( t )f 2 ( t ) } = --------
2πj ∫σ – jω F1 ( µ )F2 ( s – µ )dµ = -------- F 1 ( s )*F 2 ( s )
2πj

For easy reference, we have summarized the Laplace transform pairs and theorems in Table 2.1.

2.3 The Laplace Transform of Common Functions of Time

In this section, we will present several examples for finding the Laplace transform of common func-
tions of time.
Example 2.1
Find L { u 0 ( t ) }

2-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Functions of Time

TABLE 2.1 Summary of Laplace Transform Properties and Theorems

Property/Theorem Time Domain Complex Frequency Domain

1 Linearity c1 f1 ( t ) + c2 f2 ( t ) c1 F1 ( s ) + c2 F2 ( s )
+ … + cn fn ( t ) + … + cn Fn ( s )

2 Time Shifting f ( t – a )u 0 ( t – a ) – as
e F(s)
3 Frequency Shifting – as F(s + a)
e f(t)
4 Time Scaling f ( at ) 1 ⎛ s⎞
--- F ---
a ⎝ a⎠

5 Time Differentiation d −
----- f ( t ) sF ( s ) – f ( 0 )
See also (2.18) through (2.20) dt
6 Frequency Differentiation tf ( t ) d
– ----- F ( s )
See also (2.22) ds
7 Time Integration t −
∫–∞ f ( τ ) dτ F(s) f (0 )
----------- + ------------
s s
8 Frequency Integration f( t) ∞
--------
t ∫s F ( s ) d s
9 Time Periodicity f ( t + nT ) T – st
∫0 f ( t ) e dt
--------------------------------
– sT
1–e
10 Initial Value Theorem lim f ( t ) −
t→0 lim sF ( s ) = f ( 0 )
s→∞
11 Final Value Theorem lim f ( t ) lim sF ( s ) = f ( ∞ )
t→∞ s→0
12 Time Convolution f 1 ( t )*f 2 ( t ) F 1 ( s )F 2 ( s )

13 Frequency Convolution f 1 ( t )f 2 ( t ) 1
-------- F 1 ( s )*F 2 ( s )
2πj

Signals and Systems with MATLAB Applications, Second Edition 2-13

Orchard Publications
Chapter 2 The Laplace Transformation

Solution:
We start with the definition of the Laplace transform, that is,
∞
– st
L { f(t)} = F(s) = ∫0 f ( t ) e dt

For this example,

∞ st ∞
–e
= 0 – ⎛ – --- ⎞ = ---
– st 1
∫0 1 e
1
L { u0 ( t ) } = dt = ---------
s ⎝ s ⎠ s
0

Thus, we have obtained the transform pair

1
u 0 ( t ) ⇔ --- (2.38)
s

for Re { s } = σ > 0 .*

Example 2.2
Find L { u 1 ( t ) }

Solution:
We apply the definition
∞
– st
L { f( t)} = F(s) = ∫0 f ( t ) e dt

and for this example,

∞
– st
L { u1 ( t ) } = L { t } = ∫0 t e dt

We will perform integration by parts recalling that

∫ u dv = uv – v du ∫ (2.39)

We let
– st
u = t and dv = e
then,
– st
du = 1 and v = – e -
----------
s

* This condition was established in (2.9).

2-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Functions of Time

By substitution into (2.39),

∞
– st ∞ ∞ – st – st – st
–t e – – t e - – e--------
∫0
L { t } = ------------- – e - dt = ------------
---------- (2.40)
s 0 s s s
2
0

Since the upper limit of integration in (2.40) produces an indeterminate form, we apply L’ Hôpital’s
rule*, that is,
d
(t)
– st t = lim d t
= lim ----- 1
lim te - ---------------- = lim -------- = 0
t→∞ t → ∞ e st t→∞ d st t → ∞ se st
(e )
dt

Evaluating the second term of (2.40), we get L { t } = ---12-

s
Thus, we have obtained the transform pair

1
t ⇔ ---2- (2.41)
s

for σ > 0 .

Example 2.3
n
Find L { t u 0 ( t ) }

Solution:
To find the Laplace transform of this function, we must first review the gamma or generalized facto-
rial function Γ ( n ) which is defined as

∞
n – 1 –x
Γ(n) = ∫0 x e dx (2.42)

f( x)
* Often, the ratio of two functions, such as ---------- , for some value of x, say a, results in an indeterminate form. To
g(x)
f(x)
work around this problem, we consider the limit lim ---------- , and we wish to find this limit, if it exists. To find this
x→a g(x)
d d
limit, we use L’Hôpital’s rule which states that if f ( a ) = g ( a ) = 0 , and if the limit ------ f ( x ) ⁄ ------ g ( x ) as x
dx dx
f ( x ) = lim ⎛ -----
approaches a exists, then, lim ----------
d-
f ( x ) ⁄ ------ g ( x )⎞
d
x → a g(x) x → a ⎝ dx dx ⎠

Signals and Systems with MATLAB Applications, Second Edition 2-15

Orchard Publications
Chapter 2 The Laplace Transformation

The integral of (2.42) is an improper integral* but converges (approaches a limit) for all n > 0 .
We will now derive the basic properties of the gamma function, and its relation to the well known
factorial function
n! = n ( n – 1 ) ( n – 2 ) ⋅ ⋅ 3 ⋅ 2 ⋅ 1
The integral of (2.42) can be evaluated by performing integration by parts. Thus, in (2.39) we let
–x n–1
u = e and dv = x
Then,
n
–x
du = – e dx and v = x-----
n
and (2.42) is written as
n –x ∞ ∞
1 n –x
∫0 x e
x e
Γ ( n ) = ------------ + --- dx (2.43)
n x=0
n

With the condition that n > 0 , the first term on the right side of (2.43) vanishes at the lower limit
x = 0 . It also vanishes at the upper limit as x → ∞ . This can be proved with L’ Hôpital’s rule by dif-
ferentiating both numerator and denominator m times, where m ≥ n . Then,
m m–1
d n d n–1
n –x n m
x m–1
nx
x e - = lim -------
lim ----------- dx - = lim d
x - = lim ------------------ x - = …
-----------------------------------
x→∞ n x → ∞ ne x x→∞ d m x→∞ m–1 x
x d
m
ne m–1
ne
dx dx
n–m
n ( n – 1 ) ( n – 2 )… ( n – m + 1 )x
= lim ----------------------------------------------------------------------------------- ( n – 1 ) ( n – 2 )… ( n – m + 1 )
- = lim -------------------------------------------------------------------- = 0
x→∞ ne
x x→∞ m–n x
x e
Therefore, (2.43) reduces to
∞
1 n –x
Γ ( n ) = ---
n ∫0 x e dx

and with (2.42), we have

* Improper integrals are two types and these are:

b
a. ∫a f ( x ) dx where the limits of integration a or b or both are infinite

b
b. ∫a f ( x ) dx where f(x) becomes infinite at a value x between the lower and upper limits of integration inclusive.

2-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Functions of Time

∞ ∞
n– 1 –x 1 n –x
Γ(n) = ∫0 x e dx = ---
n ∫0 x e dx (2.44)

By comparing the integrals in (2.44), we observe that

Γ(n + 1)
Γ ( n ) = --------------------- (2.45)
n

or

nΓ ( n ) = Γ ( n + 1 ) (2.46)

It is convenient to use (2.45) for n < 0 , and (2.46) for n > 0 . From (2.45), we see that Γ ( n ) becomes
infinite as n → 0 .

For n = 1 , (2.42) yields

∞
–x –x ∞
Γ(1) = ∫0 e dx = – e 0
= 1 (2.47)

and thus we have the important relation,

Γ(1) = 1 (2.48)
From the recurring relation of (2.46), we obtain
Γ(2) = 1 ⋅ Γ(1) = 1
Γ ( 3 ) = 2 ⋅ Γ ( 2 ) = 2 ⋅ 1 = 2! (2.49)
Γ ( 4 ) = 3 ⋅ Γ ( 3 ) = 3 ⋅ 2 = 3!
and in general

Γ ( n + 1 ) = n! (2.50)

for n = 1, 2, 3, …

The formula of (2.50) is a noteworthy relation; it establishes the relationship between the Γ ( n )
function and the factorial n!
n
We now return to the problem of finding the Laplace transform pair for t u 0 t , that is,
∞
n – st
∫0 t
n
L { t u0 t } = e dt (2.51)

To make this integral resemble the integral of the gamma function, we let st = y , or t = y ⁄ s , and

Signals and Systems with MATLAB Applications, Second Edition 2-17

Orchard Publications
Chapter 2 The Laplace Transformation

thus dt = dy ⁄ s . Now, we rewrite (2.51) as

∞ ∞
n
⎛ y-- ⎞ e –y d ⎛ y-- ⎞ = ----------
1 n –y Γ ( n + 1 -) = ----------
∫0 ∫0 y e
n
L { t u0 t } = - dy = ------------------- n! -
⎝s⎠ ⎝s⎠ n+1 n+1 n+1
s s s
Therefore, we have obtained the transform pair

n n!
t u 0 ( t ) ⇔ ----------
n+1
- (2.52)
s

for positive integers of n and σ > 0 .

Example 2.4
Find L { δ ( t )}
Solution:
∞
– st
L { δ ( t )} = ∫0 δ ( t ) e dt

and using the sifting property of the delta function, we get

∞
– st –s ( 0 )
L { δ ( t )} = ∫0 δ ( t ) e dt = e = 1

Thus, we have the transform pair

δ(t) ⇔ 1 (2.53)

for all σ .

Example 2.5
Find L { δ ( t – a ) }
Solution:
∞
– st
L {δ(t – a)} = ∫0 δ ( t – a ) e dt

and again, using the sifting property of the delta function, we get

2-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Functions of Time

∞
– st – as
L {δ(t – a)} = ∫0 δ ( t – a ) e dt = e

Thus, we have the transform pair

– as
δ(t – a) ⇔ e (2.54)

for σ > 0 .

Example 2.6
– at
Find L { e u0 ( t ) }

Solution:
∞ ∞
– at – at – st – ( s + a )t
L {e u0 ( t ) } = ∫0 e e dt = ∫0 e dt

∞
= ⎛ – -----------⎞ e
1 – ( s + a )t 1
= -----------
⎝ s + a⎠ s+a
0

Thus, we have the transform pair

– at 1
e u 0 ( t ) ⇔ ----------- (2.55)
s+a

for σ > – a .

Example 2.7

⎧ n – at ⎫
Find L ⎨ t e u0 ( t ) ⎬
⎩ ⎭
Solution:
For this example, we will use the transform pair of (2.52), i.e.,
n n!
t u 0 ( t ) ⇔ ----------
n+1
- (2.56)
s
and the frequency shifting property of (2.14), that is,
– at
e f( t) ⇔ F( s + a) (2.57)

Signals and Systems with MATLAB Applications, Second Edition 2-19

Orchard Publications
Chapter 2 The Laplace Transformation

Then, replacing s with s + a in (2.56), we get the transform pair

n – at n!
t e u 0 ( t ) ⇔ -------------------------
n+1
(2.58)
(s + a)

where n is a positive integer, and σ > – a . Thus, for n = 1 , we get the transform pair

– at 1
te u 0 ( t ) ⇔ ------------------2 (2.59)
(s + a)

for σ > – a .
For n = 2 , we get the transform

2 – at 2!
t e u 0 ( t ) ⇔ ------------------3 (2.60)
(s + a)

and in general,

n – at n!
t e u 0 ( t ) ⇔ -------------------------
n+1
(2.61)
(s + a)

for σ > – a .

Example 2.8
Find L { sin ωt u 0 ( t ) }

Solution:
∞ a
– st – st
L { sin ωt u 0 ( t ) } = ∫0 ( sin ωt ) e dt = lim ∫
a→∞ 0
( sin ωt ) e dt

and from tables of integrals*

1 jωt – jωt – at 1
* This can also be derived from sin ωt = ----- ( e – e ) , and the use of (2.55) where e u 0 ( t ) ⇔ ----------- . By the lin-
j2 s+a
earity property, the sum of these terms corresponds to the sum of their Laplace transforms. Therefore,
1 ⎛ ------------- ω
1 ⎞ = -----------------
1 - – --------------
L [ sin ωtu 0 ( t ) ] = -----
j2 ⎝ s – jω s + jω⎠
s + ω2
2

2-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Functions of Time

ax
e ( a sin bx – b cos bx )
∫
ax
e sin bx dx = ------------------------------------------------------
2 2
a +b
Then,
– st a
( – s sin ωt – ω cos ωt )
L { sin ωt u 0 ( t ) } = lim e----------------------------------------------------------
-
a→∞ 2
s +ω
2
0

– as
= lim ( – s sin ωa – ω cos ωa ) + ----------------
e-------------------------------------------------------------- ω - = ----------------
ω -
a→∞ 2
s +ω
2 2
s +ω
2 2
s +ω
2

Thus, we have obtained the transform pair

ω
sin ωt u 0 t ⇔ ----------------
2
-
2
(2.62)
s +ω

for σ > 0 .

Example 2.9
Find L { cos ω t u 0 ( t ) }

Solution:
∞ a
– st – st
L { cos ω t u 0 ( t ) } = ∫0 ( cos ωt ) e dt = lim
a → ∞∫ 0
( cos ωt ) e dt

and from tables of integrals*

ax
( acos bx + b sin bx )-
∫ e cos bx dx = e-----------------------------------------------------
ax
2 2
a +b
Then,

1 jωt – jωt
* We can use the relation cos ωt = --- ( e + e ) and the linearity property, as in the derivation of the transform
2
d −
of sin ω t on the footnote of the previous page. We can also use the transform pair ----- f ( t ) ⇔ sF ( s ) – f ( 0 ) ; this
dt
is the time differentiation property of (2.16). Applying this transform pair for this derivation, we get
1d 1 d 1 ω s
L [ cos ω tu 0 ( t ) ] = L ---- ----- sin ω tu 0 ( t ) = ---- L ----- sin ω tu 0 ( t ) = ---- s ----------------- = -----------------
ω dt ω dt ω s2 + ω2 s2 + ω2

Signals and Systems with MATLAB Applications, Second Edition 2-21

Orchard Publications
Chapter 2 The Laplace Transformation

– st a
e ( – s cos ωt + ω sin ωt )
L { cos ω t u 0 ( t ) } = lim ----------------------------------------------------------
-
a→∞ 2
s +ω
2
0

– as
e ( – s cos ωa + ω sin ωa ) s s
= lim --------------------------------------------------------------- + ----------------- = -----------------
a→∞ 2
s +ω
2 2
s +ω
2 2
s +ω
2

Thus, we have the fransform pair

s -
cos ω t u 0 t ⇔ ----------------
2 2
(2.63)
s +ω
for σ > 0 .

Example 2.10
– at
Find L { e sin ωt u 0 ( t ) }

Solution:
Since
ω
sin ωtu 0 t ⇔ ----------------
2
-
2
s +ω
using the frequency shifting property of (2.14), that is,

– at
e f(t) ⇔ F(s + a ) (2.64)

we replace s with s + a , and we get

– at ω
e sin ωt u 0 ( t ) ⇔ -------------------------------
2 2
(2.65)
(s + a) + ω
for σ > 0 and a > 0 .

Example 2.11
– at
Find L { e cos ω t u 0 ( t ) }

Solution:
Since
s -
cos ω t u 0 ( t ) ⇔ ----------------
2 2
s +ω

2-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Waveforms

using the frequency shifting property of (2.14), we replace s with s + a , and we get

– at s+a
e cos ω t u 0 ( t ) ⇔ -------------------------------
2 2
(2.66)
(s + a) + ω
for σ > 0 and a > 0 .
For easy reference, we have summarized the above derivations in Table 2.2.

TABLE 2.2 Laplace Transform Pairs for Common Functions

f (t) F(s)
1 u0 ( t ) 1⁄s

2 t u0 ( t ) 1⁄s
2

3 n
t u0 ( t ) n!
-----------
n+1
s
4 δ(t) 1
5 δ(t – a) e
– as

6 – at 1
e u0 ( t ) -----------
s+a
7 n – at n!
t e u0 ( t ) -------------------------
n+1
(s + a)
8 sin ωt u 0 ( t ) ω
-----------------
2 2
s +ω
9 cos ω t u 0 ( t ) s -
----------------
2 2
s +ω
10 e
– at
sin ωt u 0 ( t ) ω
-------------------------------
2 2
(s + a) + ω
11 e
– at
cos ω t u 0 ( t ) s+a
-------------------------------
2 2
(s + a) + ω

2.4 The Laplace Transform of Common Waveforms

In this section, we will present some examples for deriving the Laplace transform of several wave-
forms using the transform pairs of Tables 1 and 2.

Signals and Systems with MATLAB Applications, Second Edition 2-23

Orchard Publications
Chapter 2 The Laplace Transformation

Example 2.12
Find the Laplace transform of the waveform f P ( t ) of Figure 2.1. The subscript P stands for pulse.

fP ( t )
A

t
0 a

Figure 2.1. Waveform for Example 2.12

Solution:
We first express the given waveform as a sum of unit step functions. Then,
fP ( t ) = A [ u0 ( t ) – u0 ( t – a ) ] (2.67)

Next, from Table 2.1,

– as
f ( t – a )u 0 ( t – a ) ⇔ e F(s )
and from Table 2.2,
u0 ( t ) ⇔ 1 ⁄ s
For this example,
Au 0 ( t ) ⇔ A ⁄ s
and
– as A
Au 0 ( t – a ) ⇔ e ---
s
Then, by the linearity property, the Laplace transform of the pulse of Figure 2.1 is

A –as A A – as
A [ u 0 ( t ) – u 0 ( t – a ) ] ⇔ --- – e --- = --- ( 1 – e )
s s s

Example 2.13
Find the Laplace transform for the waveform f L ( t ) of Figure 2.2. The subscript L stands for line.

fL ( t )
1

t
0 1 2
Figure 2.2. Waveform for Example 2.13

2-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Waveforms

Solution:
We must first derive the equation of the linear segment. This is shown in Figure 2.3. Then, we
express the given waveform in terms of the unit step function.
fL ( t ) t–1

1
t
0 1 2

Figure 2.3. Waveform for Example 2.13 with the equation of the linear segment

For this example,

f L ( t ) = ( t – 1 )u 0 ( t – 1 )
From Table 2.1,
– as
f ( t – a )u 0 ( t – a ) ⇔ e F(s)

and from Table 2.2,

1
tu 0 ( t ) ⇔ ---2-
s

Therefore, the Laplace transform of the linear segment of Figure 2.2 is

–s 1
( t – 1 )u 0 ( t – 1 ) ⇔ e ---2- (2.68)
s

Example 2.14
Find the Laplace transform for the triangular waveform f T ( t ) of Figure 2.4.

Solution:
We must first derive the equations of the linear segments. These are shown in Figure 2.5. Then, we
express the given waveform in terms of the unit step function.

fT ( t )
1

t
0 1 2
Figure 2.4. Waveform for Example 2.14

Signals and Systems with MATLAB Applications, Second Edition 2-25

Orchard Publications
Chapter 2 The Laplace Transformation

fT ( t )
1 t –t+2

t
0 1 2

Figure 2.5. Waveform for Example 2.13 with the equations of the linear segments

For this example,

fT ( t ) = t [ u0 ( t ) – u0 ( t – 1 ) ] + ( – t + 2 ) [ u0 ( t – 1 ) – u0 ( t – 2 ) ]
= tu 0 ( t ) – tu 0 ( t – 1 ) – tu 0 ( t – 1 ) + 2u 0 ( t – 1 ) + tu 0 ( t – 2 ) – 2u 0 ( t – 2 )

and collecting like terms,

f T ( t ) = tu 0 ( t ) – 2 ( t – 1 )u 0 ( t – 1 ) + ( t – 2 )u 0 ( t – 2 )
From Table 2.1,
– as
f ( t – a )u 0 ( t – a ) ⇔ e F(s)

and from Table 2.2,

1
tu 0 ( t ) ⇔ ---2-
s
Then,
1 –s 1 – 2s 1
tu 0 ( t ) – 2 ( t – 1 )u 0 ( t – 1 ) + ( t – 2 )u 0 ( t – 2 ) ⇔ ---- – 2e ---2- + e ---2-
2
s s s
or
1 –s – 2s
tu 0 ( t ) – 2 ( t – 1 )u 0 ( t – 1 ) + ( t – 2 )u 0 ( t – 2 ) ⇔ ---2- ( 1 – 2e + e )
s
Therefore, the Laplace transform of the triangular waveform of Figure 2.4 is

1 –s 2
f T ( t ) ⇔ ---2- ( 1 – e ) (2.69)
s

Example 2.15
Find the Laplace transform for the rectangular periodic waveform f R ( t ) of Figure 2.6.

2-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Laplace Transform of Common Waveforms

fR ( t )
A

t
0 a 2a 3a

−A
Figure 2.6. Waveform for Example 2.15
Solution:
This is a periodic waveform with period T = 2a , and thus we can apply the time periodicity prop-
erty
T
– sτ
∫0 f ( τ ) e
dτ
L { f ( τ ) } = -------------------------------
– sT
1–e

where the denominator represents the periodicity of f ( t ) . For this example,

2a a 2a
1 - – st 1 - – st – st
L { f R ( t ) } = --------------------
1–e
– 2as ∫0 fR ( t ) e dt = --------------------
1–e
– 2as ∫0 Ae dt + ∫a ( –A ) e dt

– st a– st 2a
- –
A e e
= --------------------
– 2as
----------
- + -------
-
1–e s 0 s a
or
A - ( – e – as + 1 + e – 2as – e –as )
L { f R ( t ) } = ---------------------------
– 2as
s(1 – e )
– as 2
A A(1 – e )
- ( 1 – 2e –as + e – 2as ) = -------------------------------------------------
= --------------------------- -
– 2as – as – as
s(1 – e ) s(1 + e )(1 – e )
⎛ A ( 1 – e – as ) A ⎛ e as ⁄ 2 e – as ⁄ 2 – e – as ⁄ 2 e – as ⁄ 2⎞ ⎞
- = --- ⎜ ---------------------------------------------------------------⎟ ⎟
= ⎜ --- ----------------------
⎝ s ( 1 + e –as ) s ⎝ e as ⁄ 2 e –as ⁄ 2 + e –as ⁄ 2 e –as ⁄ 2⎠ ⎠
– as ⁄ 2 ⎛ as ⁄ 2 – as ⁄ 2 ⎞
Ae e –e A sinh ( as ⁄ 2 )
- ⎜ --------------------------------- ⎟ = --- -----------------------------
= --- -------------
s e – as ⁄ 2 ⎝ e as ⁄ 2 + e –as ⁄ 2 ⎠ s cosh ( as ⁄ 2 )

or

f R ( t ) ⇔ --- tanh ⎛ -----⎞

A as
⎝ 2⎠
(2.70)
s

Signals and Systems with MATLAB Applications, Second Edition 2-27

Orchard Publications
Chapter 2 The Laplace Transformation

Example 2.16
Find the Laplace transform for the half-rectified sine wave f HW ( t ) of Figure 2.7.

1
sint f HW ( t )

π 2π 3π 4π

Figure 2.7. Waveform for Example 2.16

Solution:
This is a periodic waveform with period T = 2π . We will apply the time periodicity property

T
– sτ
∫0 f ( τ ) e dτ
L { f ( τ ) } = -------------------------------
– sT
1–e

where the denominator represents the periodicity of f ( t ) . For this example,

2π π
1 - – st 1 - – st
L { f HW ( t ) } = --------------------
1–e
– 2πs ∫0 f( t )e dt = --------------------
1–e
– 2πs ∫0 sin t e dt

π
– st – πs
1 e ( s sin t – cos t ) 1 (1 + e )
= --------------------
– 2πs
- ------------------------------------------
2
= ------------------
2
--------------------------
– 2πs
1–e s +1 (s + 1) (1 – e )
0

– πs
(1 + e )
1 -----------------------------------------------
L { f HW ( t ) } = ------------------
2 – πs – πs
(s + 1) (1 + e )(1 – e )
or

1
f HW ( t ) ⇔ ------------------------------------------
2 – πs
(2.71)
(s + 1)(1 – e )

2-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

2.5 Summary
• The two-sided or bilateral Laplace Transform pair is defined as
∞
– st
L {f(t )}= F(s ) = ∫– ∞ f ( t ) e dt

σ + jω
–1 1
∫σ – jω
st
L { F ( s ) } = f ( t ) = -------- F ( s ) e ds
2πj
–1
where L { f ( t ) } denotes the Laplace transform of the time function f ( t ) , L { F ( s ) } denotes
the Inverse Laplace transform, and s is a complex variable whose real part is σ , and imaginary
part ω , that is, s = σ + jω .
• The unilateral or one-sided Laplace transform defined as
∞ ∞
– st – st
L {f(t)}= F(s) = ∫t 0
f(t)e dt = ∫0 f ( t ) e dt

• We denote transformation from the time domain to the complex frequency domain, and vice
versa, as
f(t) ⇔ F(s)
• The linearity property states that
c1 f1 ( t ) + c2 f2 ( t ) + … + cn fn ( t ) ⇔ c1 F1 ( s ) + c2 F2 ( s ) + … + cn Fn ( s )

• The time shifting property states that

– as
f ( t – a )u 0 ( t – a ) ⇔ e F(s )

• The frequency shifting property states that

– at
e f(t) ⇔ F(s + a)

• The scaling property states that

f ( at ) ⇔ --- F ⎛ --- ⎞
1 s
a ⎝ a⎠

• The differentiation in time domain property states that

d −
f ' ( t ) = ----- f ( t ) ⇔ sF ( s ) – f ( 0 )
dt

d 2- − −
-------
2
f ( t ) ⇔ s 2 F ( s ) – sf ( 0 ) – f ' ( 0 )
dt

Signals and Systems with MATLAB Applications, Second Edition 2-29

Orchard Publications
Chapter 2 The Laplace Transformation

d3
-------- f ( t ) ⇔ s 3 F ( s ) – s 2 f ( 0 − ) – sf ' ( 0 − ) – f '' ( 0 − )
3
dt
and in general
n
d
-------- f ( t ) ⇔ s n F ( s ) – s n – 1 f ( 0 − ) – s n – 2 f ' ( 0 − ) – … – f n–1 −
(0 )
n
dt

where the terms f ( 0 − ), f ' ( 0 − ), f '' ( 0 − ) , and so on, represent the initial conditions.
• The differentiation in complex frequency domain property states that

d
tf ( t ) ⇔ – ----- F ( s )
ds
and in general,
n
n nd
t f ( t ) ⇔ ( – 1 ) -------n- F ( s )
ds

• The integration in time domain property states that

t −
F(s) f (0 )
∫ –∞
f ( τ ) dτ ⇔ ----------- + ------------
s s

• The integration in complex frequency domain property states that

∞
f ( t -)
-------
t
⇔ ∫s F ( s ) ds
( t ) exists.
provided that the limit lim f-------
-
t→0 t

• The time periodicity property states that

T – st
∫0 f ( t ) e dt
f ( t + nT ) ⇔ ----------------------------
– sT
-
1–e
• The initial value theorem states that
−
lim f ( t ) = lim sF ( s ) = f ( 0 )
t→0 s→∞

• The final value theorem states that

lim f ( t ) = lim sF ( s ) = f ( ∞ )
t→∞ s→0

2-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• Convolution in the time domain corresponds to multiplication in the complex frequency domain,
that is,
f 1 ( t )*f 2 ( t ) ⇔ F 1 ( s )F 2 ( s )

• Convolution in the complex frequency domain divided by 1 ⁄ 2πj , corresponds to multiplication

in the time domain. That is,
1
f 1 ( t )f 2 ( t ) ⇔ -------- F 1 ( s )*F 2 ( s )
2πj

• The Laplace transforms of some common functions of time are shown below.
u0 ( t ) ⇔ 1 ⁄ s

t ⇔ 1 ⁄ s2

n n! -
t u 0 ( t ) ⇔ ----------
n+1
s

δ(t) ⇔ 1
– as
δ(t – a) ⇔ e

– at 1
e u 0 ( t ) ⇔ -----------
s+a

– at 1
te u 0 ( t ) ⇔ ------------------2
(s + a)

2 – at 2!
t e u 0 ( t ) ⇔ ------------------3
(s + a)

n – at n!
t e u 0 ( t ) ⇔ -------------------------
n+1
(s + a)

ω -
sin ωt u 0 t ⇔ ----------------
2 2
s +ω

s
cos ω t u 0 t ⇔ ----------------
2
-
2
s +ω

– at ω
e sin ωt u 0 ( t ) ⇔ -------------------------------
2 2
(s + a) + ω

– at s+a
e cos ω t u 0 ( t ) ⇔ -------------------------------
2 2
(s + a) + ω

Signals and Systems with MATLAB Applications, Second Edition 2-31

Orchard Publications
Chapter 2 The Laplace Transformation

• The Laplace transforms of some common waveforms are shown below.

fP ( t )
A

t
0 a
A – as A A – as
A [ u 0 ( t ) – u 0 ( t – a ) ] ⇔ --- – e --- = --- ( 1 – e )
s s s

fL ( t )
1

t
0 1 2

–s 1
( t – 1 )u 0 ( t – 1 ) ⇔ e ---2-
s

fT ( t )
1

t
0 1 2

1 –s 2
f T ( t ) ⇔ ---2- ( 1 – e )
s

fR ( t )
A

t
0 a 2a 3a

−A

f R ( t ) ⇔ --- tanh ⎛ -----⎞

A as
s ⎝ 2⎠

2-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

1
sint f HW ( t )

π 2π 3π 4π

1
f HW ( t ) ⇔ ------------------------------------------
2 – πs
(s + 1 )(1 – e )

Signals and Systems with MATLAB Applications, Second Edition 2-33

Orchard Publications
Chapter 2 The Laplace Transformation

2.6 Exercises
1. Find the Laplace transform of the following time domain functions:
a. 12
b. 6u0 ( t )

c. 24u 0 ( t – 12 )

d. 5tu 0 ( t )
5
e. 4t u 0 ( t )

2. Find the Laplace transform of the following time domain functions:

a. j8
b. j5 ∠– 90°
– 5t
c. 5e u0 ( t )

7 – 5t
d. 8t e u0 ( t )

e. 15δ ( t – 4 )
3. Find the Laplace transform of the following time domain functions:
3 2
a. ( t + 3t + 4t + 3 )u 0 ( t )

b. 3 ( 2t – 3 )δ ( t – 3 )
c. ( 3 sin 5t )u 0 ( t )

d. ( 5 cos 3t )u 0 ( t )

e. ( 2 tan 4t )u 0 ( t ) Be careful with this! Comment and skip derivation.

4. Find the Laplace transform of the following time domain functions:

a. 3t ( sin 5t )u 0 ( t )
2
b. 2t ( cos 3t )u 0 ( t )
– 5t
c. 2e sin 5t

2-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

– 3t
d. 8e cos 4t

e. ( cos t )δ ( t – π ⁄ 4 )
5. Find the Laplace transform of the following time domain functions:
a. 5tu 0 ( t – 3 )
2
b. ( 2t – 5t + 4 )u 0 ( t – 3 )
– 2t
c. ( t – 3 )e u0 ( t – 2 )

2(t – 2)
d. ( 2t – 4 )e u0 ( t – 3 )

– 3t
e. 4te ( cos 2t )u 0 ( t )

6. Find the Laplace transform of the following time domain functions:

a. d ( sin 3t )
dt

b. d ( 3e )
– 4t
dt

c. d ( t cos 2t )
2
dt

d. d ( e sin 2t )
– 2t
dt

e. d ( t e )
2 – 2t
dt
7. Find the Laplace transform of the following time domain functions:
sin t
a. ---------
t
t
sin τ
b. ∫0 ---------
τ
- dτ

sin at
c. ------------
t
∞
cos τ
d. ∫t ----------- dτ
τ

Signals and Systems with MATLAB Applications, Second Edition 2-35

Orchard Publications
Chapter 2 The Laplace Transformation

∞ –τ
e-------
e. ∫t τ
dτ

8. Find the Laplace transform for the sawtooth waveform f ST ( t ) of Figure 2.8.

f ST ( t )
A

t
a 2a 3a

Figure 2.8. Waveform for Exercise 8.

9. Find the Laplace transform for the full rectification waveform f FR ( t ) of Figure 2.9.

f FR ( t ) Full Rectified Waveform

sint
1

a 2a 3a 4a

Figure 2.9. Waveform for Exercise 9

2-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

2.7 Solutions to Exercises

1. From the definition of the Laplace transform or from Table 2.2 we get:
– 12s 24 2 5!
a. 12 ⁄ s b. 6 ⁄ s c. e ⋅ ------ d. 5 ⁄ s e. 4 ⋅ ----6-
s s
2. From the definition of the Laplace transform or from Table 2.2 we get:
5 7! – 4s
a. j8 ⁄ s b. 5 ⁄ s c. ----------- d. 8 ⋅ ------------------8 e. 15e
s+5 (s + 5)
3.
3 × 2! 4 3
a. From Table 2.2 and the linearity property 3!
----- + -------------- + ---- + ---
4 3 2
s s s s

– 3s
b. 3 ( 2t – 3 )δ ( t – 3 ) = 3 ( 2t – 3 ) t=3
δ ( t – 3 ) = 9δ ( t – 3 ) and 9δ ( t – 3 ) ⇔ 9e

2 2
5 s sin 4t 4 ⁄ (s + 2 )
- d. 5 ⋅ ---------------- e. 2 tan 4t = 2 ⋅ ------------- ⇔ 2 ⋅ ---------------------------- = 8
c. 3 ⋅ --------------- ---
2 2 2 2 cos 4t 2 2
s +5 s +3 s ⁄ (s + 2 ) s

This answer looks suspicious because 8 ⁄ s ⇔ 8u 0 ( t ) and the Laplace transform is unilateral,
that is, there is one-to-one correspondence between the time domain and the complex fre-
quency domain. The fallacy with this procedure is that we assumed that if f 1 ( t ) ⇔ F 1 ( s ) and
f1 ( t ) F1 ( s )
f 2 ( t ) ⇔ F 2 ( s ) , we cannot conclude that ---------- ⇔ ------------- .
f2 ( t ) F2 ( s )
1
For this exercise f 1 ( t ) ⋅ f 2 ( t ) = sin 4t ⋅ ------------- and as we’ve learned multiplication in the time
cos 4t
domain corresponds to convolution in the complex frequency domain. Accordingly, we must
∞
– st
use the Laplace transform definition ∫0 ( 2 tan 4t )e dt and this requires integration by parts. We

skip this analytical derivation. The interested reader may try to find the answer with the MAT-
LAB code syms s t; 2*laplace(sin(4*t)/cos(4*t))
4. From (2.22)
n
n nd
t f ( t ) ⇔ ( – 1 ) -------n- F ( s )
ds
Then,
a.
– 5 ⋅ ( 2s )
3 ( – 1 ) ----- ⎛ ---------------
1 d 5 ⎞ 30s -
- = – 3 ------------------------ = -----------------------
⎝
ds s + 5 2 2 ⎠ 2 2 2
( s + 25 ) 2
( s + 25 )

Signals and Systems with MATLAB Applications, Second Edition 2-37

Orchard Publications
Chapter 2 The Laplace Transformation

b.
d ⎛ –s +9⎞
2 2 2 2
d s + 3 – s ( 2s )
2 ( – 1 ) -------2- ⎛ ---------------
2 d s ⎞
- = 2 ----- ----------------------------------- = 2 ----- ⎜ ---------------------⎟
⎝ 2 2⎠ ds ⎝ 2
(s + 9) ⎠
ds 2 2
ds s + 3 2
(s + 9)
2 2 2 2
( s + 9 ) ( – 2s ) – 2 ( s + 9 ) ( 2s ) ( – s + 9 -)
= 2 ⋅ -------------------------------------------------------------------------------------------------
4
2
(s + 9)
2 2 3 3
( s + 9 ) ( – 2s ) – 4s ( – s + 9 -) – 2s – 18s + 4s – 36s-
= 2 ⋅ -------------------------------------------------------------------
3
= 2 ⋅ -------------------------------------------------------
3
2 2
(s + 9) (s + 9)
3 2
2s – 54s 2s ( s – 27 ) 4s ( s 2 – 27 )
= 2 ⋅ ----------------------3 = 2 ⋅ --------------------------
3
- = ---------------------------
2 2 3
(s + 9) (s + 9) 2
(s + 9)
c.
2×5
------------------------------ 10
= ------------------------------
-
2 2 2
(s + 5) + 5 ( s + 5 ) + 25
d.
8 ( s + 3 ) - = ------------------------------
----------------------------- 8(s + 3) -
2 2 2
(s + 3) + 4 ( s + 3 ) + 16
e.
– ( π ⁄ 4 )s
cos t π⁄4
δ ( t – π ⁄ 4 ) = ( 2 ⁄ 2 )δ ( t – π ⁄ 4 ) and ( 2 ⁄ 2 )δ ( t – π ⁄ 4 ) ⇔ ( 2 ⁄ 2 )e
5.
a.
– 3s ⎛
---- + ------ ⎞ = --- e ⎛ --- + 3⎞
5 15 5 –3s 1
5tu 0 ( t – 3 ) = [ 5 ( t – 3 ) + 15 ]u 0 ( t – 3 ) ⇔ e
⎝ 2 s ⎠ s ⎝s ⎠
s
b.
2 2
( 2t – 5t + 4 )u 0 ( t – 3 ) = [ 2 ( t – 3 ) + 12t – 18 – 5t + 4 ]u 0 ( t – 3 )
2
= [ 2 ( t – 3 ) + 7t – 14 ]u 0 ( t – 3 )
2
= [ 2 ( t – 3 ) + 7 ( t – 3 ) + 21 – 14 ]u 0 ( t – 3 )
2
= [ 2 ( t – 3 ) + 7 ( t – 3 ) + 7 ]u 0 ( t – 3 ) ⇔ e
– 3s ⎛ 2 × 2!- + ---
------------- --- ⎞
7- + 7
⎝ 3 2 s⎠
s s
c.
– 2t –2 ( t – 2 ) –4
( t – 3 )e u 0 ( t – 2 ) = [ ( t – 2 ) – 1 ]e ⋅ e u0 ( t – 2 )
–4 – 2s 1 1 –4 – 2s – ( s + 1 )
⇔e ⋅e ------------------ – ---------------- = e ⋅ e -------------------
2 (s + 2) 2
(s + 2) (s + 2)

2-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

d.
2(t – 2) –2 ( t – 3 ) –2
( 2t – 4 )e u 0 ( t – 3 ) = [ 2 ( t – 3 ) + 6 – 4 ]e ⋅ e u0 ( t – 3 )
–2 – 3s 2 2 –2 – 3s s+4
⇔e ⋅e ------------------ + ---------------- = 2e ⋅ e ------------------
(s + 3)
2 ( s + 3 ) (s + 3)
2

e.
– 3t 1 d s+3 d s+3
4te ( cos 2t )u 0 ( t ) ⇔ 4 ( – 1 ) ----- -----------------------------
- = – 4 ----- -----------------------------------
ds ( s + 3 ) 2 + 2 2 ds s 2 + 6s + 9 + 4
2
d s+3 - s + 6s + 13 – ( s + 3 ) ( 2s + 6 )-
⇔ – 4 ----- ----------------------------
2
= – 4 -----------------------------------------------------------------------
2
ds s + 6s + 13 2
( s + 6s + 13 )
2 2 2
s + 6s + 13 – 2s – 6s – 6s – 18 4 ( s + 6s + 5 )
⇔ – 4 -----------------------------------------------------------------------------
2
- = -----------------------------------
-
2 2
( s + 6s + 13 ) 2
( s + 6s + 13 )

6.
a.
----- f ( t ) ⇔ sF ( s ) – f ( 0 − ) −
3 d
sin 3t ⇔ ---------------
2
-
2
f ( 0 ) = sin 3t t=0
= 0
s +3 dt
d 3 3s
( sin 3t ) ⇔ s ---------------
- – 0 = -------------
-
dt 2
s +3
2 2
s +9
b.
– 4t 3 d- − − – 4t
3e ⇔ ----------- ---- f ( t ) ⇔ sF ( s ) – f ( 0 ) f ( 0 ) = 3e = 3
s+4 dt t=0

d – 4t 3 3s 3 ( s + 4 ) – 12
( 3e ) ⇔ s ----------- – 3 = ----------- – ------------------- = -----------
dt s + 4 s+4 s+4 s+4
c.
2
s - 2 2 d s -
cos 2t ⇔ ---------------
2 2
t cos 2t ⇔ ( – 1 ) -------2- -------------
2
s +2 ds s + 4
2 2 2
d s + 4 – s ( 2s ) d –s + 4 2
( s + 4 ) ( – 2s ) – ( – s + 4 ) ( s + 4 )2 ( 2s )
2 2
----- --------------------------------- = ----- --------------------- = ------------------------------------------------------------------------------------------------
-
ds 2 2 ds 2 2 4
(s + 4) (s + 4) 2
(s + 4)
2 2 3 3 2
( s + 4 ) ( – 2s ) – ( – s + 4 ) ( 4s -) = –
= ----------------------------------------------------------------------- 2s – 8s + 4s – 16s- = 2s
---------------------------------------------------- ( s – 12 -)
--------------------------
2 3 2 3 2 3
(s + 4) (s + 4) (s + 4)
Thus,
2
2 2s ( s – 12 )
t cos 2t ⇔ --------------------------
3
-
2
(s + 4)
and

Signals and Systems with MATLAB Applications, Second Edition 2-39

Orchard Publications
Chapter 2 The Laplace Transformation

d 2 −
( t cos 2t ) ⇔ sF ( s ) – f ( 0 )
dt
2
2s ( s – 12 ) 2 2
( s – 12 )
- – 0 = 2s
⇔ s -------------------------- -----------------------------
-
2 3 3
(s + 4) 2
(s + 4)
d.
----- f ( t ) ⇔ sF ( s ) – f ( 0 − )
2 – 2t 2 d
sin 2 t ⇔ ---------------
2
-
2
e sin 2t ⇔ ---------------------------
2
-
s +2 (s + 2) + 4 dt
d – 2t 2 2s
( e sin 2t ) ⇔ s ---------------------------
- – 0 = ---------------------------
-
dt (s + 2) + 4
2
(s + 2) + 4
2

e.
2 2! 2 – 2t 2! d- −
t ⇔ ----3- t e ⇔ ------------------3 ---- f ( t ) ⇔ sF ( s ) – f ( 0 )
s (s + 2) dt
d 2 – 2t 2! 2s
( t e ) ⇔ s ------------------3 – 0 = ------------------
dt (s + 2) (s + 2)
3

7.
a.
1 ⎧ sin t ⎫ sin t exists. Since
sin t ⇔ -------------
2
- but to find L ⎨ --------- ⎬ we must show that the limit lim --------
-
s +1 ⎩ t ⎭ t→0 t

∞
sin t 1
∫s
sin x = 1 this condition is satisfied and thus --------
lim --------- -⇔ -------------- ds . From tables of integrals
- 2
x→0 x t s +1
1 1 –1 1 –1
∫ x----------------
2
+a
2
dx = --- tan ( x ⁄ a ) + C . Then,
a ∫ s-------------
2
+1
- ds = tan ( 1 ⁄ s ) + C and the constant of inte-

gration C is evaluated from the final value theorem. Thus,

–1 sin t –1
lim f ( t ) = lim sF ( s ) = lim s [ tan ( 1 ⁄ s ) + C ] = 0 and --------- ⇔ tan ( 1 ⁄ s )
t→∞ s→0 s→0 t

b.
t −
sin t –1 F( s) f ( 0 )
From (a) above --------- ⇔ tan ( 1 ⁄ s ) and since
t ∫ –∞
f ( τ ) dτ ⇔ ----------- + ------------ , it follows that
s s

t
sin τ 1 –1
∫0 ---------
τ
- dτ ⇔ --- tan
s
(1 ⁄ s)

2-40 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

c.
From (a) above --------- ⇔ tan ( 1 ⁄ s ) and since f ( at ) ⇔ --- F ⎛⎝ --- ⎞⎠ , it follows that
sin t –1 1 s
t a a
–1
1⁄s
⇔ --- tan ⎛ ---------⎞ or ------------ ⇔ tan ( a ⁄ s )
sin at
------------ 1 sin at –1
at a ⎝ a ⎠ t

d.
∞
s cos t s
cos t ⇔ -------------
2
s +1 t ∫
- , ---------- ⇔ -------------- ds , and from tables of integrals
2
s s +1

x 1 s - 1
∫ ∫
2 2 2
----------------
2 2
dx = --- ln ( x + a ) + C . Then, ------------- 2
ds = --- ln ( s + 1 ) + C and the constant of inte-
x +a 2 s +1 2
gration C is evaluated from the final value theorem. Thus,
t −
1 F( s) f ( 0 )
∫
2
lim f ( t ) = lim sF ( s ) = lim s --- ln ( s + 1 ) + C = 0 and using f ( τ ) dτ ⇔ ----------- + ------------ we
t→∞ s→0 s→0 2 –∞ s s
∞
cos τ 1
get ∫t ----------- dτ ⇔ ----- ln ( s 2 + 1 )
τ 2s

e.
–t ∞
–t 1 e 1
e ⇔ ----------- , ------ ⇔
s+1 t ∫s ----------
s+1
- ds , and from tables of integrals

1 1 1
∫ --------------
ax + b
- dx
2 s+1 ∫
= --- ln ( ax + b ) . Then, ----------- ds = ln ( s + 1 ) + C and the constant of integration C
is evaluated from the final value theorem. Thus,
t −
F ( s -) + ------------
f ( 0 ) we
lim f ( t ) = lim sF ( s ) = lim s [ ln ( s + 1 ) + C ] = 0
t→∞ s→0 s→0
and using ∫ –∞
f ( τ ) dτ ⇔ ----------
s s
∞ –τ
e 1
get ∫t ------- dτ ⇔ --- ln ( s + 1 )
τ s

8.
f ST ( t ) A
--- t
a
A

t
a 2a 3a

This is a periodic waveform with period T = a and its Laplace transform is

Signals and Systems with MATLAB Applications, Second Edition 2-41

Orchard Publications
Chapter 2 The Laplace Transformation

a a
1 - A A
--- te –st dt = ------------------------- – st
F ( s ) = -----------------
1–e
– as ∫ 0 a a(1 – e )
– as
-
∫0 te dt (1)

and from (2.40) of Example 2.2 and limits of integration 0 to a we get

a 0
a – st – st – st – st
– st
∫0 te
a te e te e
L {t} 0
= dt = – ---------- – -------- = ---------- + --------
s 2 s 2
s s
0 a

– as – as
1
1- – ae
= --- e - = ---
------------ – -------- - [ 1 – ( 1 + as )e – as ]
2 s 2 2
s s s

Adding and subtracting as we get

a 1 – as 1 – as
L {t} 0
= ---2- [ ( 1 + as ) – ( 1 + as )e – as ] = ---2- [ ( 1 + as ) ( 1 – e ) – as ]
s s
By substitution into (1) we get
A 1- – as A - ⋅ [ ( 1 + as ) ( 1 – e – as ) – as ]
F ( s ) = -------------------------
– as
- ⋅ ---
2
[ ( 1 + as ) ( 1 – e ) – as ] = ------------------------------
2 – as
a(1 – e ) s as ( 1 – e )
A ( 1 + as ) Aa A ( 1 + as ) a
= ----------------------- – ----------------------------- = ----- ------------------- – -----------------------
2 – as as s – as
as as ( 1 – e ) (1 – e )
9.
This is a periodic waveform with period T = a = π and its Laplace transform is
T π
1 - – st 1 - – st
F ( s ) = -----------------
1–e
– sT ∫0 f ( t )e dt = ----------------------
(1 – e )
– πs ∫0 sin te dt

From tables of integrals

ax
e ( asin bx – b cos bx )
∫
ax
sin bxe dx = -----------------------------------------------------
2 2
a +b
Then,
– st π – πs
e ( s sin t – cos t -)
1 - ----------------------------------------- 1+e -
1 - ------------------
F ( s ) = -----------------
– πs
⋅ 2
= -----------------
– πs
⋅ 2
1–e s +1 0
1–e s +1
– πs
1+e πs
- ⋅ ------------------- = -------------- coth ⎛ -----⎞
1 1
= -------------
2 – πs 2 ⎝ 2⎠
s +1 1–e s +1

2-42 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 3
The Inverse Laplace Transformation

his chapter is a continuation to the Laplace transformation topic of the previous chapter and
T presents several methods of finding the Inverse Laplace Transformation. The partial fraction
expansion method is explained thoroughly and it is illustrated with several examples.

3.1 The Inverse Laplace Transform Integral

The Inverse Laplace Transform Integral was stated in the previous chapter; it is repeated here for con-
venience.
σ + jω
–1 1
∫σ – jω
st
L { F ( s ) } = f ( t ) = -------- F ( s ) e ds (3.1)
2πj

This integral is difficult to evaluate because it requires contour integration using complex variables
theory. Fortunately, for most engineering problems we can refer to Tables of Properties, and Com-
mon Laplace transform pairs to lookup the Inverse Laplace transform.

3.2 Partial Fraction Expansion

Quite often the Laplace transform expressions are not in recognizable form, but in most cases appear
in a rational form of s , that is,
N(s)
F ( s ) = ----------- (3.2)
D(s)

where N ( s ) and D ( s ) are polynomials, and thus (3.2) can be expressed as

m m–1 m–2
N(s) bm s + bm – 1 s + bm – 2 s + … + b1 s + b0
F ( s ) = ----------- = ------------------------------------------------------------------------------------------------------------------
- (3.3)
D(s) n
an s + an – 1 s
n – 1
+ an – 2 s
n – 2
+ … + a1 s + a0

The coefficients a k and b k are real numbers for k = 1, 2, …, n , and if the highest power m of N ( s )
is less than the highest power n of D ( s ) , i.e., m < n , F ( s ) is said to be expressed as a proper rational
function. If m ≥ n , F ( s ) is an improper rational function.

In a proper rational function, the roots of N ( s ) in (3.3) are found by setting N ( s ) = 0 ; these are
called the zeros of F ( s ) . The roots of D ( s ) , found by setting D ( s ) = 0 , are called the poles of F ( s ) .
We assume that F ( s ) in (3.3) is a proper rational function. Then, it is customary and very convenient

Signals and Systems with MATLAB Applications, Second Edition 3-1

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

n
to make the coefficient of s unity; thus, we rewrite F ( s ) as
1- m m–1 m–2
---- ( bm s + bm – 1 s + bm – 2 s + … + b1 s + b0 )
N(s) a
F ( s ) = ----------- = -----------------------------------------------------------------------------------------------------------------------------
n
(3.4)
D(s) n an – 1 n – 1 an – 2 n – 2 a1 a
s + ----------- s + ----------- s + … + ----- s + ----0-
an an an an

The zeros and poles of (3.4) can be real and distinct, or repeated, or complex conjugates, or combina-
tions of real and complex conjugates. However, we are mostly interested in the nature of the poles, so
we will consider each case separately.
Case I: Distinct Poles
If all the poles p 1, p 2, p 3, …, p n of F ( s ) are distinct (different from each another), we can factor the
denominator of F ( s ) in the form
N(s)
F ( s ) = ------------------------------------------------------------------------------------------------- (3.5)
( s – p1 ) ⋅ ( s – p2 ) ⋅ ( s – p3 ) ⋅ … ⋅ ( s – pn )

where p k is distinct from all other poles. Next, using the partial fraction expansion method,*we can
express (3.5) as
r1 r2 r3 rn
F ( s ) = -----------------
- + ----------------- - + … + -----------------
- + ----------------- - (3.6)
( s – p1 ) ( s – p2 ) ( s – p3 ) ( s – pn )

where r 1, r 2, r 3, …, r n are the residues, and p 1, p 2, p 3, …, p n are the poles of F ( s ) .

To evaluate the residue r k , we multiply both sides of (3.6) by ( s – p k ) ; then, we let s → p k , that is,

r k = lim ( s – p k )F ( s ) = ( s – p k )F ( s ) (3.7)
s → pk s = pk

Example 3.1
Use the partial fraction expansion method to simplify F 1 ( s ) of (3.8) below, and find the time domain
function f 1 ( t ) corresponding to F 1 ( s ) .

3s + 2
F 1 ( s ) = -------------------------- (3.8)
2
s + 3s + 2

* The partial fraction expansion method applies only to proper rational functions. It is used extensively in integra-
tion, and in finding the inverses of the Laplace transform, the Fourier transform, and the z-transform. This
method allows us to decompose a rational polynomial into smaller rational polynomials with simpler denomina-
tors from which we can easily recognize their integrals and inverse transformations. This method is also being
taught in intermediate algebra and introductory calculus courses.

3-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Partial Fraction Expansion

Solution:
Using (3.6), we get

3s + 2 3s + 2 r1 r2
F 1 ( s ) = -------------------------- = --------------------------------- = ---------------
- + ---------------
- (3.9)
2
s + 3s + 2 (s + 1)(s + 2) (s + 1) (s + 2)

The residues are

3s + 2-
r 1 = lim ( s + 1 )F ( s ) = --------------- = –1 (3.10)
s → –1 (s + 2) s = –1

and
3s + 2
r 2 = lim ( s + 2 )F ( s ) = ---------------- = 4 (3.11)
s → –2 (s + 1) s = –2

Therefore, we express (3.9) as

F 1 ( s ) = ------------------------- – 1 - + ---------------
3s + 2 - = --------------- 4 - (3.12)
2
s + 3s + 2 ( s + 1 ) ( s + 2)

and from Table 2.2 of Chapter 2

– at 1
e u 0 ( t ) ⇔ ----------- (3.13)
s+a
Then,
– 1 - + ---------------
F 1 ( s ) = --------------- 4 - ⇔ ( – e – t + 4e – 2t ) u ( t ) = f ( t ) (3.14)
(s + 1) (s + 2)
0 1

The residues and poles of a rational function of polynomials such as (3.8), can be found easily using
the MATLAB residue(a,b) function. For this example, we use the code
Ns = [3, 2]; Ds = [1, 3, 2]; [r, p, k] = residue(Ns, Ds)
and MATLAB returns the values
r =
4
-1
p =
-2
-1
k =
[]

Signals and Systems with MATLAB Applications, Second Edition 3-3

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

For this MATLAB code, we defined Ns and Ds as two vectors that contain the numerator and
denominator coefficients of F ( s ) . When this code is executed, MATLAB displays the r and p vec-
tors that represent the residues and poles respectively. The first value of the vector r is associated
with the first value of the vector p, the second value of r is associated with the second value of p,
and so on.
The vector k is referred to as the direct term and it is always empty (has no value) whenever F ( s ) is
a proper rational function, that is, when the highest degree of the denominator is larger than that of
the numerator. For this example, we observe that the highest power of the denominator is s 2 ,
whereas the highest power of the numerator is s and therefore the direct term is empty.
We can also use the MATLAB ilaplace(f) function to obtain the time domain function directly from
F ( s ) . This is done with the code that follows.
syms s t; Fs=(3*s+2)/(s^2+3*s+2); ft=ilaplace(Fs); pretty(ft)
When this code is executed, MATLAB displays the expression
4 exp(-2 t)- exp(-t)
Example 3.2
Use the partial fraction expansion method to simplify F 2 ( s ) of (3.15) below, and find the time
domain function f 2 ( t ) corresponding to F 2 ( s ) .
2
3s + 2s + 5
F 2 ( s ) = ------------------------------------------------- (3.15)
3 2
s + 12s + 44s + 48
Solution:
First, we use the MATLAB factor(s) symbolic function to express the denominator polynomial of
F 2 ( s ) in factored form. For this example,

syms s; factor(s^3 + 12*s^2 + 44*s + 48)

ans =
(s+2)*(s+4)*(s+6)
Then,
2 2
3s + 2s + 5 3s + 2s + 5 r1 r2 r3
F 2 ( s ) = ------------------------------------------------- = -------------------------------------------------- = ---------------
- + ---------------
- + ---------------
- (3.16)
3 2 ( s + 2 )( s + 4 )( s + 6) (s + 2) (s + 4) (s + 6)
s + 12s + 44s + 48
The residues are
2
3s + 2s + 5 9
r 1 = --------------------------------- = --- (3.17)
(s + 4)(s + 6) s = –2
8

3-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Partial Fraction Expansion

2
3s + 2s + 5 37
r 2 = --------------------------------- = – ------ (3.18)
(s + 2)(s + 6) s = –4
4

2
3s + 2s + 5 89
r 3 = --------------------------------- = ------ (3.19)
(s + 2)(s + 4) s = –6
8

Then, by substitution into (3.16) we get

2
3s + 2s + 5 9⁄8 – 37 ⁄ 4 89 ⁄ 8
F 2 ( s ) = ------------------------------------------------- = ---------------- + ---------------- + ---------------- (3.20)
3 2 (s + 2) (s + 4) (s + 6)
s + 12s + 44s + 48
From Table 2.2 of Chapter 2
– at 1
e u 0 ( t ) ⇔ ----------- (3.21)
s+a
Then,

9⁄8 – 37 ⁄ 4 89 ⁄ 8
F 2 ( s ) = ---------------- + ---------------- + ---------------- ⇔ ⎛ --- e – ------ e + ------ e ⎞ u 0 ( t ) = f 2 ( t )
9 – 2t 37 – 4t 89 – 6t
⎝ ⎠
(3.22)
(s + 2) (s + 4) (s + 6) 8 4 8

Check with MATLAB:

syms s t; Fs = (3*s^2 + 4*s + 5) / (s^3 + 12*s^2 + 44*s + 48); ft = ilaplace(Fs)
ft =
-37/4*exp(-4*t)+9/8*exp(-2*t)+89/8*exp(-6*t)
Case II: Complex Poles

Quite often, the poles of F ( s ) are complex*, and since complex poles occur in complex conjugate
pairs, the number of complex poles is even. Thus, if p k is a complex root of D ( s ) , then, its complex
conjugate pole, denoted as p k∗ , is also a root of D ( s ) . The partial fraction expansion method can
also be used in this case, but it may be necessary to manipulate the terms of the expansion in order to
express them in a recognizable form. The procedure is illustrated with the following example.
Example 3.3
Use the partial fraction expansion method to simplify F 3 ( s ) of (3.23) below, and find the time
domain function f 3 ( t ) corresponding to F 3 ( s ) .

s+3
F 3 ( s ) = ------------------------------------------- (3.23)
3 2
s + 5s + 12s + 8

* A review of complex numbers is presented in Appendix B

Signals and Systems with MATLAB Applications, Second Edition 3-5

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

Solution:
Let us first express the denominator in factored form to identify the poles of F 3 ( s ) using the MAT-
LAB factor(s) function. Then,
syms s; factor(s^3 + 5*s^2 + 12*s + 8)
ans =
(s+1)*(s^2+4*s+8)
The factor(s) function did not factor the quadratic term. We will use the roots(p) function.
p=[1 4 8]; roots_p=roots(p)
roots_p =
-2.0000 + 2.0000i
-2.0000 - 2.0000i
Then,
s+3 s+3
F 3 ( s ) = ------------------------------------------- = ------------------------------------------------------------------------
3 2 ( s + 1 ) ( s + 2 + j2 ) ( s + 2 – j2 )
s + 5s + 12s + 8

s+3 r1 r2 r 2∗
F 3 ( s ) = ------------------------------------------
- = ---------------
- + --------------------------- + ------------------------
- (3.24)
3 2 ( s + 1 ) ( s + 2 + j2 ) ( s + 2 – j 2 )
s + 5s + 12s + 8
The residues are
s+3 2
r 1 = -------------------------- = --- (3.25)
2 5
s + 4s + 8 s = –1

s+3
r 2 = -----------------------------------------
- 1 – j2
= ------------------------------------ 1 – j2
= ------------------
( s + 1 ) ( s + 2 –j 2 ) s = – 2 – j2
( – 1 – j2 ) ( – j4 ) – 8 + j4
(3.26)
( 1 – j2 ) ( – 8 – j4 )
= ----------------------- ----------------------- = –------------------------
16 + j12 = – 1 j3-
--- + -----
( – 8 + j4 ) ( – 8 – j4 ) 80 5 20

1 j3 ∗
r 2∗ = ⎛ – --- + ------⎞ = – --- – ------
1 j3
⎝ 5 20⎠
(3.27)
5 20

By substitution into (3.24),

2⁄5 – 1 ⁄ 5 + j3 ⁄ 20 – 1 ⁄ 5 – j3 ⁄ 20
F 3 ( s ) = ---------------- + ----------------------------------- + ----------------------------------- (3.28)
(s + 1) ( s + 2 + j2 ) ( s + 2 –j 2 )
The last two terms on the right side of (3.28), do not resemble any Laplace transform pair that we
derived in Chapter 2. Therefore, we will express them in a different form. We combine them into a
single term*, and now (3.28) is written as

3-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Partial Fraction Expansion

2⁄5 1 ( 2s + 1 )
F 3 ( s ) = ---------------- – --- ⋅ ------------------------------
- (3.29)
( s + 1 ) 5 ( s 2 + 4s + 8 )

For convenience, we denote the first term on the right side of (3.29) as F 31 ( s ) , and the second as
F 32 ( s ) . Then,

2⁄5 2 –t
F 31 ( s ) = ---------------- ⇔ --- e = f 31 ( t ) (3.30)
(s + 1) 5

Next, for F 32 ( s )

1 ( 2s + 1 )
F 32 ( s ) = – --- ⋅ ------------------------------
2
- (3.31)
5 ( s + 4s + 8 )

and recalling that

– at ω
e sin ωtu 0 t ⇔ -------------------------------
2 2
(s + a) + ω
(3.32)
– at s+a
e cos ωtu 0 t ⇔ -------------------------------
2 2
(s + a) + ω

we express F 32 ( s ) as

1 3 3
⎛ s + --- + --- – --- ⎞
2 ⎜ --------------------------------
2 2 2⎟ s+2 –3 ⁄ 2
F 32 ( s ) = – --- --- ⎛ --------------------------------- + --------------------------------- ⎞
- = –2
5 ⎜ ( s + 2 ) 2 + 2 2 )⎟ 5 ⎝ ( s + 2 )2 + 22 ) ( s + 2 )2 + 22 ) ⎠
⎝ ⎠
s + 2 - ⎞ ------------ 6 ⁄ 10- ⎛ -------------------------------- (3.33)
--- ⎛ -------------------------------- -⎞
2
= –2 +
5 ⎝ ( s + 2 )2 + 22 ) ⎠ 2 ⎝ ( s + 2 )2 + 22 ) ⎠
s + 2 - ⎞ -----
= – --- ⎛ -------------------------------- + - ⎛ --------------------------------
- ⎞
2 3 2
⎝
5 (s + 2 ) + 2 ) 2 2 ⎠ 10 ⎝ 2 2 ⎠
(s + 2 ) + 2 )
Addition of (3.30) with (3.33) yields
2⁄5 s+2
F 3 ( s ) = F 31 ( s ) + F 32 ( s ) = ---------------- – --- ⎛ --------------------------------
- ⎞ + ------ ⎛ --------------------------------- ⎞
2 3 2
( s + 1 ) 5 ⎝ ( s + 2 ) 2 + 2 2 ) ⎠ 10 ⎝ ( s + 2 ) 2 + 2 2 ) ⎠
2 –t 2 – 2t 3 –2t
⇔ --- e – --- e cos 2t + ------ e sin 2t = f 3 ( t )
5 5 10
Check with MATLAB:

* Here, we used MATLAB with simple((−1/5 +3j/20)/(s+2+2j)+(−1/5 −3j/20)/(s+2−2j)). The simple func-
tion, after several simplification tools that were displayed on the screen, returned (-2*s-1)/
(5*s^2+20*s+40)

Signals and Systems with MATLAB Applications, Second Edition 3-7

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

syms a s t w; % Define several symbolic variables

Fs=(s + 3)/(s^3 + 5*s^2 + 12*s + 8); ft=ilaplace(Fs)
ft =
2/5*exp(-t)-2/5*exp(-2*t)*cos(2*t)
+3/10*exp(-2*t)*sin(2*t)
Case III: Multiple (Repeated) Poles
In this case, F ( s ) has simple poles, but one of the poles, say p 1 , has a multiplicity m . For this condi-
tion, we express it as
N(s)
F ( s ) = ------------------------------------------------------------------------------------------ (3.34)
m
( s – p 1 ) ( s – p 2 )… ( s – p n – 1 ) ( s – p n )

Denoting the m residues corresponding to multiple pole p 1 as r 11, r 12, r 13, …, r 1m , the partial frac-
tion expansion of (3.34) is written as
r 11 r 12 r 13 r 1m
F ( s ) = --------------------
- + --------------------------- - + … + -----------------
- + --------------------------- -
( s – p1 )
m
( s – p1 )
m–1
( s – p1 )
m–2 ( s – p1 )
(3.35)
r2 r3 rn
+ ----------------- - + … + -----------------
- + ----------------- -
( s – p2 ) ( s – p3 ) ( s – pn )

For the simple poles p 1, p 2, …, p n , we proceed as before, that is, we find the residues as

r k = lim ( s – p k )F ( s ) = ( s – p k )F ( s ) (3.36)
s → pk s = pk

The residues r 11, r 12, r 13, …, r 1m corresponding to the repeated poles, are found by multiplication of
m
both sides of (3.35) by ( s – p ) . Then,
m 2 m–1
( s – p 1 ) F ( s ) = r 11 + ( s – p 1 )r 12 + ( s – p 1 ) r 13 + … + ( s – p 1 ) r 1m
rn ⎞ (3.37)
+ ( s – p 1 ) ⎛ -----------------
m r2 r3
- + … + -----------------
- + ----------------- -
⎝ ( s – p2 ) ( s – p3 ) ( s – p n )⎠

Next, taking the limit as s → p 1 on both sides of (3.37), we get

m 2 m–1
lim ( s – p 1 ) F ( s ) = r 11 + lim [ ( s – p 1 )r 12 + ( s – p 1 ) r 13 + … + ( s – p 1 ) r 1m ]
s → p1 s → p1

rn ⎞
( s – p 1 ) ⎛ -----------------
m r2 r3
+ lim - + … + -----------------
- + ----------------- -
s → p1 ⎝ ( s – p2 ) ( s – p3 ) ( s – p n )⎠
or

3-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Partial Fraction Expansion

m
r 11 = lim ( s – p 1 ) F ( s ) (3.38)
s → p1

and thus (3.38) yields the residue of the first repeated pole.
The residue r 12 for the second repeated pole p 1 , is found by differentiating (3.37) with respect to s
and again, we let s → p 1 , that is,

r 12 = lim ----d [ ( s – p )m F ( s ) ] (3.39)

- 1
s → p 1 ds

In general, the residue r 1k can be found from

m 2
( s – p 1 ) F ( s ) = r 11 + r 12 ( s – p 1 ) + r 13 ( s – p 1 ) + … (3.40)

whose ( m – 1 )th derivative of both sides is

k–1
d
1 ------------- m
- k – 1- [ ( s – p 1 ) F ( s ) ]
( k – 1 )!r 1k = lim ----------------- (3.41)
s → p 1( k – 1 )! ds

or
k–1
d
1 ------------- m
r 1k - k – 1- [ ( s – p 1 ) F ( s ) ]
= lim ----------------- (3.42)
s → p 1 ( k – 1 )! ds

Example 3.4
Use the partial fraction expansion method to simplify F 4 ( s ) of (3.43) below, and find the time
domain function f 4 ( t ) corresponding to F 4 ( s ) .

s+3
F 4 ( s ) = ----------------------------------- (3.43)
2
(s + 2)(s + 1)
Solution:
We observe that there is a pole of multiplicity 2 at s = – 1 , and thus in partial fraction expansion
form, F 4 ( s ) is written as

s+3
F 4 ( s ) = -----------------------------------
r1
= ---------------
r 21
- + ------------------
r 22
+ ---------------
- (3.44)
( s + 2 )( s + 1 )
2 (s + 2) (s + 1) 2 (s + 1)

The residues are

Signals and Systems with MATLAB Applications, Second Edition 3-9

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

s+3
r 1 = ------------------ = 1
2
(s + 1) s = –2

s+3
r 21 = ----------- = 2
s+2 s = –1

d s+3 (s + 2) – (s + 3)
r 22 = ----- ⎛ ----------- ⎞ = --------------------------------------- = –1
ds ⎝ s + 2 ⎠ (s + 2)
2
s = –1 s = –1

The value of the residue r 22 can also be found without differentiation as follows:

Substitution of the already known values of r 1 and r 21 into (3.44), and letting s = 0 *, we get

s+3 1 2 r 22
----------------------------------- = ---------------- + ------------------ + ---------------
-
(s + 1) (s + 2)
2 (s + 2) s = 0 (s + 1)
2 (s + 1) s=0
s=0 s=0
or
3 1
--- = --- + 2 + r 22
2 2

from which r 22 = – 1 as before. Finally,

s+3
F 4 ( s ) = ----------------------------------- = --------------- 2
1 - + ------------------ – 1 - ⇔ e – 2t + 2te – t – e – t = f ( t )
+ --------------- (3.45)
(s + 2)(s + 1)
2 ( s + 2 ) (s + 1)
2 ( s + 1) 4

Check with MATLAB:

syms s t; Fs=(s+3)/((s+2)*(s+1)^2); ft=ilaplace(Fs)
ft = exp(-2*t)+2*t*exp(-t)-exp(-t)
We can use the following code to check the partial fraction expansion.
syms s
Ns = [1 3]; % Coefficients of the numerator N(s) of F(s)
expand((s + 1)^2); % Expands (s + 1)^2 to s^2 + 2*s + 1;
d1 = [1 2 1]; % Coefficients of (s + 1)^2 = s^2 + 2*s + 1 term in D(s)
d2 = [0 1 2]; % Coefficients of (s + 2) term in D(s)
Ds=conv(d1,d2); % Multiplies polynomials d1 and d2 to express the
% denominator D(s) of F(s) as a polynomial
[r,p,k]=residue(Ns,Ds)

* This is permissible since (3.44) is an identity.

3-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Partial Fraction Expansion

r =
1.0000
-1.0000
2.0000
p =
-2.0000
-1.0000
-1.0000
k =
[]

Example 3.5
Use the partial fraction expansion method to simplify F 5 ( s ) of (3.46) below, and find the time
domain function f 5 ( t ) corresponding to the given F 5 ( s ) .
2
s + 3s + 1 -
F 5 ( s ) = ------------------------------------- (3.46)
3 2
(s + 1) (s + 2)
Solution:
We observe that there is a pole of multiplicity 3 at s = – 1 , and a pole of multiplicity 2 at s = – 2 .
Then, in partial fraction expansion form, F 5 ( s ) is written as

r 11 r 12 r 13 r 21 r 22
F 5 ( s ) = ------------------ + ------------------ + ---------------
- + ------------------ + ---------------
- (3.47)
(s + 1)
3
(s + 1)
2 ( s + 1 ) ( s + 2 )2 ( s + 2 )

The residues are

2
+ 3 s + 1-
r 11 = s------------------------- = –1
2
(s + 2) s = –1

d ⎛ s + 3 s + 1⎞
2
r 12 = ----
- ⎜ -⎟
-------------------------
ds ⎝ ( s + 2 ) 2 ⎠
s = –1
2 2
s + 2 ) ( 2s + 3 ) – 2 ( s + 2 ) ( s + 3 s + 1 )-
= (--------------------------------------------------------------------------------------------- s+4
= ------------------ =3
4 3
(s + 2) s = –1
(s + 2) s = –1

Signals and Systems with MATLAB Applications, Second Edition 3-11

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

1 d ⎛ s + 3 s + 1⎞ 1 d d ⎛ s + 3 s + 1⎞
2 2 2
r 13 = ----- -------2- ⎜ -------------------------
-⎟ = --- ----- ----- ⎜ -------------------------
-⎟
2! ds ⎝ ( s + 2 ) 2 ⎠ 2 ds ds ⎝ ( s + 2 ) 2 ⎠
s = –1 s = –1

3 2
1d s+4 1 ( s + 2 ) – 3 ( s + 2 ) ( s + 4 )-
= --- ----- ⎛ ------------------3 ⎞ = --- ---------------------------------------------------------------
2 ds ⎝ ( s + 2 ) ⎠ 2 (s + 2)
6
s = –1 s = –1
1 s + 2 – 3s – 12 ⎞ –s–5
= --- ⎛ ----------------------------------
- = ------------------ = –4
2⎝ (s + 2)
4 ⎠
(s + 2)
4
s = –1 s = –1

Next, for the pole at s = – 2 ,

2
s + 3s + 1
r 21 = -------------------------- = 1
3
(s + 1) s = –2
and
d ⎛ s + 3 s + 1⎞
2 3 2 2
( s + 1 ) ( 2s + 3 ) – 3 ( s + 1 ) ( s + 3 s + 1 )
r 22 = ----- ⎜ -------------------------
-⎟ = ---------------------------------------------------------------------------------------------------
ds ⎝ ( s + 1 ) 3 ⎠ (s + 1)
6
s = –2 s = –2
2 2
( s + 1 ) ( 2s + 3 ) – 3 ( s + 3 s + 1 ) – s – 4s
= ----------------------------------------------------------------------------- = -------------------- =4
4 4
(s + 1) s = –2
(s + 1) s = –2

By substitution of the residues into (3.47), we get

–1 3 –4 1 4
F 5 ( s ) = ------------------ + ------------------ + ---------------- + ------------------ + ---------------- (3.48)
(s + 1)
3
(s + 1)
2 (s + 1) (s + 2) 2 (s + 2)

We will check the values of these residues with the MATLAB code below.
syms s; % The function collect(s) below multiplies (s+1)^3 by (s+2)^2
% and we use it to express the denominator D(s) as a polynomial so that we can
% we can use the coefficients of the resulting polynomial with the residue function
Ds=collect(((s+1)^3)*((s+2)^2))
Ds =
s^5+7*s^4+19*s^3+25*s^2+16*s+4
Ns=[1 3 1]; Ds=[1 7 19 25 16 4]; [r,p,k]=residue(Ns,Ds)
r =
4.0000
1.0000
-4.0000
3.0000
-1.0000

3-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Case for m ≥ n

p =
-2.0000
-2.0000
-1.0000
-1.0000
-1.0000
k =
[]
From Table 2.2 of Chapter 2
– at 1 – at 1 n – 1 – at ( n – 1 )!
e ⇔ ----------- te ⇔ ------------------2 t e ⇔ ------------------n
s+a (s + a) (s + a)

and with these, we derive f 5 ( t ) from (3.48) as

1 2 –t –t –t – 2t – 2t
f 5 ( t ) = – --- t e + 3te – 4e + te + 4e (3.49)
2
We can verify (3.49) with MATLAB as follows:
syms s t; Fs=−1/((s+1)^3) + 3/((s+1)^2) − 4/(s+1) + 1/((s+2)^2) + 4/(s+2);
ft=ilaplace(Fs)
ft = -1/2*t^2*exp(-t)+3*t*exp(-t)-4*exp(-t)
+t*exp(-2*t)+4*exp(-2*t)

3.3 Case for m ≥ n

Our discussion thus far, was based on the condition that F ( s ) is a proper rational function. However,
if F ( s ) is an improper rational function, that is, if m ≥ n , we must first divide the numerator N ( s ) by
the denominator D ( s ) to obtain an expression of the form
2
F ( s ) = k0 + k1 s + k2 s + … + km – n s
m–n N(s)
+ ----------- (3.50)
D(s)

where N ( s ) ⁄ D ( s ) is a proper rational function.

Example 3.6
Derive the Inverse Laplace transform f 6 ( t ) of
2
s + 2s + 2
F 6 ( s ) = -------------------------- (3.51)
s+1

Signals and Systems with MATLAB Applications, Second Edition 3-13

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

Solution:
For this example, F 6 ( s ) is an improper rational function. Therefore, we must express it in the form
of (3.50) before we use the partial fraction expansion method.
By long division, we get
2
s + 2s + 2- = ----------
F 6 ( s ) = ------------------------- 1 -+1+s
s+1 s+1
Now, we recognize that
1
----------- ⇔ e –t
s+1
and
1 ⇔ δ(t)
but
s⇔?
To answer that question, we recall that
u 0' ( t ) = δ ( t )
and
u 0'' ( t ) = δ' ( t )

where δ' ( t ) is the doublet of the delta function. Also, by the time differentiation property
2 2 2 1
u 0'' ( t ) = δ' ( t ) ⇔ s F ( s ) – sf ( 0 ) – f ' (0 ) = s F ( s ) = s ⋅ --- = s
s
Therefore, we have the new transform pair
s ⇔ δ' ( t ) (3.52)
and thus,
2
s + 2s + 2- = ----------
F 6 ( s ) = ------------------------- 1 - + 1 + s ⇔ e – t + δ ( t ) + δ' ( t ) = f ( t ) (3.53)
6
s+1 s+1
In general,
n
d - n
-------
n
δ(t) ⇔ s (3.54)
dt
We verify (3.53) with MATLAB as follows:
Ns = [1 2 2]; Ds = [1 1]; [r, p, k] = residue(Ns, Ds)
r =
1

3-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Alternate Method of Partial Fraction Expansion

p =
-1
k =
1 1
Here, the direct terms k= [1 1] are the coefficients of δ ( t ) and δ' ( t ) respectively.

3.4 Alternate Method of Partial Fraction Expansion

Partial fraction expansion can also be performed with the method of clearing the fractions, that is,
making the denominators of both sides the same, then equating the numerators. As before, we
assume that F ( s ) is a proper rational function. If not, we first perform a long division, and then work
with the quotient and the remainder as we did in Example 3.6. We also assume that the denominator
D ( s ) can be expressed as a product of real linear and quadratic factors. If these assumptions prevail,
m
we let ( s – a ) be a linear factor of D ( s ) , and we assume that ( s – a ) is the highest power of ( s – a )
that divides D ( s ) . Then, we can express F ( s ) as

N(s) r1 r2 rm
F ( s ) = ----------- = ---------- - + … ------------------
- + ----------------- - (3.55)
D(s) s – a (s – a) 2
(s – a)
m

2 2 n
Let s + αs + β be a quadratic factor of D ( s ) , and suppose that ( s + αs + β ) is the highest power
of this factor that divides D ( s ) . Now, we perform the following steps:
1. To this factor, we assign the sum of n partial fractions, that is,
r1 s + k1 r2 s + k2 rn s + kn
-------------------------- - + … + ---------------------------------
- + --------------------------------- -
2 2 n
s + αs + β ( s 2 + αs + β ) 2
( s + αs + β )

2. We repeat step 1 for each of the distinct linear and quadratic factors of D ( s )
3. We set the given F ( s ) equal to the sum of these partial fractions
4. We clear the resulting expression of fractions and arrange the terms in decreasing powers of s
5. We equate the coefficients of corresponding powers of s
6. We solve the resulting equations for the residues

Example 3.7
Express F 7 ( s ) of (3.56) below as a sum of partial fractions using the method of clearing the fractions.

– 2s + 4
F 7 ( s ) = ------------------------------------- (3.56)
2 2
(s + 1)(s – 1)

Signals and Systems with MATLAB Applications, Second Edition 3-15

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

Solution:
Using Steps 1 through 3 above, we get

– 2s + 4 r1 s + A r 21 r 22
F 7 ( s ) = ------------------------------------- = ------------------ + -----------------
- + ---------------
- (3.57)
2
(s + 1)(s – 1)
2 2
(s + 1) (s – 1)
2 (s – 1)

With Step 4,
2 2 2
– 2s + 4 = ( r 1 s + A ) ( s – 1 ) + r 21 ( s + 1 ) + r 22 ( s – 1 ) ( s + 1 ) (3.58)
and with Step 5,
3 2
– 2s + 4 = ( r 1 + r 22 )s + ( – 2r 1 + A – r 22 + r 21 )s
(3.59)
+ ( r 1 – 2A + r 22 ) s + ( A – r 22 + r 21 )

Relation (3.59) will be an identity is s if each power of s is the same on both sides of this relation.
Therefore, we equate like powers of s and we get
0 = r 1 + r 22
0 = – 2r 1 + A – r 22 + r 21
(3.60)
– 2 = r 1 – 2A + r 22
4 = A – r 22 + r 21

Subtracting the second equation of (3.60) from the fourth, we get

4 = 2r 1
or
r1 = 2 (3.61)

By substitution of (3.61) into the first equation of (3.60), we get

0 = 2 + r 22
or
r 22 = – 2 (3.62)

Next, substitution of (3.61) and (3.62) into the third equation of (3.60) yields
– 2 = 2 – 2A – 2
or
A = 1 (3.63)
Finally by substitution of (3.61), (3.62), and (3.63) into the fourth equation of (3.60), we get

3-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Alternate Method of Partial Fraction Expansion

4 = 1 + 2 + r 21
or
r 21 = 1 (3.64)

Substitution of these values into (3.57) yields

– 2s + 4 2s + 1 1 2
F 7 ( s ) = ------------------------------------- = ------------------ + ------------------ – ---------------- (3.65)
2 2 2 2 (s – 1)
(s + 1)(s – 1) (s + 1) (s – 1)

Example 3.8
Use partial fraction expansion to simplify F 8 ( s ) of (3.66) below, and find the time domain function
f 8 ( t ) corresponding to F 8 ( s ) .
s+3
F 8 ( s ) = ------------------------------------------
- (3.66)
3 2
s + 5s + 12s + 8
Solution:
This is the same transform as in Example 3.3, where we found that the denominator D ( s ) can be
expressed in factored form of a linear term and a quadratic. Thus, we write F 8 ( s ) as

s+3
F 8 ( s ) = ------------------------------------------------ (3.67)
2
( s + 1 ) ( s + 4s + 8 )
and using the method of clearing the fractions, we rewrite (3.67) as

s+3 r1 r2 s + r3
F 8 ( s ) = ------------------------------------------------ = ----------
- + -------------------------
- (3.68)
2
( s + 1 ) ( s + 4s + 8 ) s + 1 s + 4s + 82

As in Example 3.3,
s+3 -
r 1 = ------------------------- = 2
--- (3.69)
2 5
s + 4s + 8 s = –1

Next, to compute r 2 and r 3 , we follow the procedure of this section and we get
2
( s + 3 ) = r 1 ( s + 4s + 8 ) + ( r 2 s + r 3 ) ( s + 1 ) (3.70)

Since r 1 is already known, we only need two equations in r 2 and r 3 . Equating the coefficient of s 2
on the left side, which is zero, with the coefficients of s 2 on the right side of (3.70), we get
0 = r1 + r2 (3.71)

Signals and Systems with MATLAB Applications, Second Edition 3-17

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

and since r 1 = 2 ⁄ 5 , then r 2 = – 2 ⁄ 5 .

To obtain the third residue r 3 , we equate the constant terms of (3.70). Then, 3 = 8r 1 + r 3 or
3 = 8 × 2 ⁄ 5 + r 3 , or r 3 = – 1 ⁄ 5 . Then, by substitution into (3.68), we get

2⁄5 1 ( 2s + 1 )
F 8 ( s ) = ---------------- – --- ⋅ ------------------------------
- (3.72)
( s + 1 ) 5 ( s 2 + 4s + 8 )

as before.
The remaining steps are the same as in Example 3.3, and thus f 8 ( t ) is the same as f 3 ( t ) , that is,

f 8 ( t ) = f 3 ( t ) = ⎛ --- e – --- e cos 2t + ------ e sin 2t⎞ u 0 ( t )

2 –t 2 – 2t 3 –2t
⎝5 5 10 ⎠

3.5 Summary
• The Inverse Laplace Transform Integral defined as
σ + jω
–1 1
∫σ – jω
st
L { F ( s ) } = f ( t ) = -------- F ( s ) e ds
2πj

is difficult to evaluate because it requires contour integration using complex variables theory.
• For most engineering problems we can refer to Tables of Properties, and Common Laplace trans-
form pairs to lookup the Inverse Laplace transform.
• The partial fraction expansion method offers a convenient means of expressing Laplace trans-
forms in a recognizable form from which we can obtain the equivalent time-domain functions.
• If the highest power m of the numerator N ( s ) is less than the highest power n of the denomina-
tor D ( s ) , i.e., m < n , F ( s ) is said to be expressed as a proper rational function. If m ≥ n , F ( s ) is
an improper rational function.
• The Laplace transform F ( s ) must be expressed as a proper rational function before applying the
partial fraction expansion. If F ( s ) is an improper rational function, that is, if m ≥ n , we must first
divide the numerator N ( s ) by the denominator D ( s ) to obtain an expression of the form
2 m–n N(s)
F ( s ) = k0 + k1 s + k2 s + … + km – n s + -----------
D(s)

• In a proper rational function, the roots of numerator N ( s ) are called the zeros of F ( s ) and the
roots of the denominator D ( s ) are called the poles of F ( s ) .

3-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• The partial fraction expansion method can be applied whether the poles of F ( s ) are distinct, com-
plex conjugates, repeated, or a combination of these.
• When F ( s ) is expressed as
r1 r2 r3 rn
F ( s ) = -----------------
- + ----------------- - + … + -----------------
- + ----------------- -
( s – p1 ) ( s – p2 ) ( s – p3 ) ( s – pn )

r 1, r 2, r 3, …, r n are called the residues and p 1, p 2, p 3, …, p n are the poles of F ( s ) .

• The residues and poles of a rational function of polynomials can be found easily using the MAT-
LAB residue(a,b) function. The direct term is always empty (has no value) whenever F ( s ) is a
proper rational function.
• We can use the MATLAB factor(s) symbolic function to convert the denominator polynomial
form of F 2 ( s ) into a factored form.

• We can use the MATLAB collect(s) and expand(s) symbolic functions to convert the denomi-
nator factored form of F 2 ( s ) into a polynomial form.

• In this chapter we developed the new transform pair

s ⇔ δ' ( t )
and in general,
n
d
-------- δ ( t ) ⇔ s n
n
dt
• The method of clearing the fractions is an alternate method of partial fraction expansion.

Signals and Systems with MATLAB Applications, Second Edition 3-19

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

3.6 Exercises
1. Find the Inverse Laplace transform of the following:
4
a. -----------
s+3

4
b. ------------------2
(s + 3)

4
c. ------------------4
(s + 3)

3s + 4
d. ------------------5
(s + 3)
2
s + 6s + 3-
e. -------------------------
5
(s + 3)
2. Find the Inverse Laplace transform of the following:
3s + 4
a. ----------------------------
2
-
s + 4s + 85

4s + 5
b. --------------------------------
2
-
s + 5s + 18.5

2
s + 3s + 2
c. -----------------------------------------------
3 2
-
s + 5s + 10.5s + 9

2
s – 16
d. ---------------------------------------------
3 2
-
s + 8s + 24s + 32

s+1
e. ------------------------------------------
3 2
-
s + 6s + 11s + 6

3. Find the Inverse Laplace transform of the following:

3s + 2
a. ----------------
2
-
s + 25
2
5s + 3
b. --------------------2- (See hint on next page)
2
(s + 4)

3-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

⎧ 1 s
2
⎫
⎪ ------- ( sin αt + αt cos αt ) ⇔ -----------------------
-⎪
2
⎪ 2α (s + α ) ⎪
2 2
Hint: ⎨ ⎬
⎪ --------
1 1
-⎪
⎪ 3- ( sin αt – αt cos αt ) ⇔ -----------------------
2 2 2⎪
⎩ 2α (s + α ) ⎭

2s + 3 -
c. --------------------------------
2
s + 4.25s + 1
3 2
s + 8s + 24s + 32
d. ---------------------------------------------
2
-
s + 6s + 8

– 2s 3 -
e. e ---------------------
3
( 2s + 3 )

4. Use the Initial Value Theorem to find f ( 0 ) given that the Laplace transform of f ( t ) is
2s + 3
---------------------------------
2
s + 4.25s + 1
Compare your answer with that of Exercise 3(c).
5. It is known that the Laplace transform F ( s ) has two distinct poles, one at s = 0 , the other at
s = – 1 . It also has a single zero at s = 1 , and we know that lim f ( t ) = 10 . Find F ( s ) and f ( t ) .
t→∞

Signals and Systems with MATLAB Applications, Second Edition 3-21

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

3.7 Solutions to Exercises

1.
a.
4 4 4 4 – 3t 2 – 3t
----------- ⇔ 4e – 3t b. ------------------ ⇔ 4te – 3t c. ------------------ ⇔ ----- t 3 e = --- t 3 e
s+3 (s + 3)
2
(s + 3)
4 3! 3
d.
3s + 4 3(s + 4 ⁄ 3 + 5 ⁄ 3 – 5 ⁄ 3 ) (s + 3) – 5 ⁄ 3 1 1
------------------ = ----------------------------------------------------------- = 3 ⋅ -------------------------------
5
- = 3 ⋅ ------------------ – 5 ⋅ ------------------
4 5
(s + 3)
5
(s + 3)
5
(s + 3) (s + 3) (s + 3)

⇔ ----- t e – ----- t e = --- ⎛ t e – ------ t e ⎞

3 3 – 3t 5 4 – 3t 1 3 –3t 5 4 – 3t
3! 4! 2⎝ 12 ⎠

e.
2 2 2
+ 6s + 3- = ----------------------------------
s------------------------- ( s + 3 ) – ------------------
s + 6s + 9 – 6- = ------------------ 6 1
= ------------------
1
– 6 ⋅ ------------------5
(s + 3)
5
(s + 3)
5 5
(s + 3) (s + 3)
5
(s + 3)
3
(s + 3)

⇔ ----- t e – ----- t e = --- ⎛ t e – --- t e ⎞

1 2 –3t 6 4 –3t 1 2 – 3t 1 4 – 3t
2! 4! 2⎝ 2 ⎠

2.
a.
3s + 4 3(s + 4 ⁄ 3 + 2 ⁄ 3 – 2 ⁄ 3) (s + 2) – 2 ⁄ 3 (s + 2) 1 2×9
----------------------------- = ----------------------------------------------------------- = 3 ⋅ -------------------------------
2
- = 3 ⋅ ------------------------------ – --- ⋅ ------------------------------
2 2 2 9 2 2
2
s + 4s + 85 ( s + 2 ) + 81
2
(s + 2) + 9 (s + 2) + 9 (s + 2) + 9
(s + 2) 2 9 2
- – --- ⋅ ------------------------------ ⇔ 3e – 2t cos 9t – --- e – 2t sin 9t
= 3 ⋅ -----------------------------
2 2 9 2 2 9
(s + 2) + 9 (s + 2) + 9
b.
4s + 5
4s + 5 - = ---------------------------------------------------- 4s + 5 s+5⁄4
-------------------------------- - = --------------------------------------- = 4 ⋅ ---------------------------------------
2 2
2
s + 5s + 18.5
2
s + 5s + 6.25 + 12.25
2
( s + 2.5 ) + 3.5
2
( s + 2.5 ) + 3.5
s + 10 ⁄ 4 – 10 ⁄ 4 + 5 ⁄ 4 s + 2.5 1 5 × 3.5
= 4 ⋅ --------------------------------------------------------
2 2
- = 4 ⋅ --------------------------------------- – ------- ⋅ ---------------------------------------
2 2 3.5 2 2
( s + 2.5 ) + 3.5 ( s + 2.5 ) + 3.5 ( s + 2.5 ) + 3.5
( s + 2.5 ) 10 3.5 – 2.5t 10 – 2.5t
= 4 ⋅ ---------------------------------------
2 2
– ------ ⋅ ---------------------------------------
2 2
⇔ 4e cos 3.5t – ------ e sin 3.5t
( s + 2.5 ) + 3.5 7 ( s + 2.5 ) + 3.5 7

c. Using the MATLAB factor(s) function we get:

syms s; factor(s^2+3*s+2), factor(s^3+5*s^2+10.5*s+9)
ans = (s+2)*(s+1)
ans = 1/2*(s+2)*(2*s^2+6*s+9)

3-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

Then,
2
s + 3s + 2 (s + 1)(s + 2) (s + 1) s+1
------------------------------------------------ = ---------------------------------------------------- = ----------------------------------- = ----------------------------------------------------------------
3 2 2 2 2
s + 5s + 10.5s + 9 ( s + 2 ) ( s + 3s + 4.5 ) ( s + 3s + 4.5 ) s + 3s + 2.25 – 2.25 + 4.5
s + 1.5 – 1.5 + 1 s + 1.5 1 0.5 × 1.5
= -------------------------------------------- = -------------------------------------------- – ------- ⋅ -------------------------------------------
2
-
2
2 1.5
( s + 1.5 ) + ( 1.5 )
2 2
( s + 1.5 ) + ( 1.5 )
2
( s + 1.5 ) + ( 1.5 )
s + 1.5 1 1.5 – 1.5t 1 – 1.5t
= -------------------------------------------- – --- ⋅ ---------------------------------------
2 2
⇔e cos 1.5t – --- e sin 1.5t
2 3 3
( s + 1.5 ) + ( 1.5 )
2
( s + 2.5 ) + 3.5
d.
2
s – 16 ( s + 4 )( s – 4 ) (s – 4) s+2–2–4
---------------------------------------------- = ------------------------------------------------ = ------------------------------ = ------------------------------
3 2 2 2 2 2 2
s + 8s + 24s + 32 ( s + 4 ) ( s + 4s + 8 ) (s + 2) + 2 (s + 2) + 2
s+2 1 6×2
= ------------------------------ – --- ⋅ ------------------------------
2 2
2 2
(s + 2) + 2
2
(s + 2) + 2
s+2
= ------------------------------
2 - ⇔ e –2t cos 2t – 3e – 2t sin 2t
– 3 ⋅ -----------------------------
2 2
(s + 2) + 2
2 2
(s + 2) + 2
e.
s+1 (s + 1)
- = -------------------------------------------------
------------------------------------------ 1
- = --------------------------------
-
3
s + 6s + 11s + 6
2 ( s + 1 ) ( s + 2 ) ( s + 3 ) ( s + 2 ) (s + 3)

1 r1 r2 1- 1-
= --------------------------------
- = ----------
- + ----------- r 1 = ---------- =1 r 2 = ---------- = –1
( s + 2 )( s + 3 ) s + 2 s + 3 s+3 s = –2
s+2 s = –3

1 1 1 – 2t – 3t
= --------------------------------- = ----------- – ----------- ⇔ e – e
( s + 2 )( s + 3 ) s+2 s+3

3.
a.
3s + 2 3s 1 2×5 s 2 5 2
2
- = 3 ⋅ ---------------- + --- ⋅ ---------------- ⇔ 3 cos 5t + --- sin 5t
----------------- = ---------------- + --- ⋅ ---------------
2 2 2 2 2
2
s + 25
2
s +5
2 5 s +5 s +5 5 s +5 5

b.
2 2
5s + 3 5s 3 1 1
--------------------- = ----------------------- + ----------------------- ⇔ 5 ⋅ ------------ ( sin 2t + 2t cos 2t ) + 3 ⋅ ------------ ( sin 2t – 2t cos 2t )
2 2 2 2 2 2 2 2 2×2 2×8
(s + 4) (s + 2 ) (s + 2 )

⇔ ⎛ --- + ------⎞ sin 2t + ⎛ --- – ------⎞ 2t cos 2t = ------ sin 2t + ------ t cos 2t
5 3 5 3 23 17
⎝ 4 16⎠ ⎝ 4 16⎠ 16 8

Signals and Systems with MATLAB Applications, Second Edition 3-23

Orchard Publications
Chapter 3 The Inverse Laplace Transformation

c.
-------------------------------- 2s + 3
2s + 3 - = --------------------------------------- r1
- = ----------
r2
- + -----------------
-
2
s + 4.25s + 1 ( s + 4 ) ( s + 1 ⁄ 4 ) s + 4 s + 1⁄4
2s + 3-
r 1 = ----------------- –5 - = 4
= --------------- --- r 2 = 2s + 3-
-------------- 5 ⁄ 2- = 2
= ------------ ---
s+1⁄4 s = –4
– 15 ⁄ 4 3 s+4 s = –1 ⁄ 4
15 ⁄ 4 3
4⁄3 2⁄3 2 – 4t –t ⁄ 4
----------- + ------------------ ⇔ --- ( 2e + e )
s+4 s+1⁄4 3
d.
3 2 2 2
s + 8s + 24s + 32 ( s + 4 ) ( s + 4s + 8 ) ( s + 4s + 8 )
---------------------------------------------- = ------------------------------------------------ = ------------------------------- and by long division
2
s + 6s + 8 ( s + 2 )( s + 4 ) (s + 2)
2
s + 4s + 8 4 – 2t
-------------------------- = s + 2 + ----------- ⇔ δ' ( t ) + 2δ ( t ) + 4e
s+2 s+2
e.
– 2s 3 – 2s
e ----------------------
3
e F ( s ) ⇔ f ( t – 2 )u 0 ( t – 2 )
( 2s + 3 )
3⁄2
3
3⁄8 3⁄8
F ( s ) = ---------------------- = ------------------------------- = ---------------------------------- = -------------------------3- ⇔ --- ⎛ ----- t e
3 3 1 2 – ( 3 ⁄ 2 )t⎞ 3 2 –( 3 ⁄ 2 )t
= ------ t e
8 2!⎝ ⎠ 16
( 2s + 3 )
3
( 2s + 3 ) ⁄ 2
3 3
[ ( 2s + 3 ) ⁄ 2 ]
3
(s + 3 ⁄ 2)

---------------------- ⇔ ------ ( t – 2 ) 2 e –( 3 ⁄ 2 ) ( t – 2 ) u 0 ( t – 2 )
– 2s – 2s 3 3
e F(s)= e 3
( 2s + 3 ) 16

4. The initial value theorem states that lim f ( t ) = lim sF ( s ) . Then,

t→0 s→∞
2
2s + 3 2s + 3s
f ( 0 ) = lim s --------------------------------
- = lim --------------------------------
-
s → ∞ s 2 + 4.25s + 1 s → ∞ s 2 + 4.25s + 1

2 2 2
2s ⁄ s + 3s ⁄ s
= lim ----------------------------------------------------------- 2+3⁄s
- = lim -------------------------------------------
-=2
s → ∞ s ⁄ s + 4.25s ⁄ s + 1 ⁄ s
2 2 2 2 s → ∞ 1 + 4.25 ⁄ s + 1 ⁄ s 2

The value f ( 0 ) = 2 is the same as in the time domain expression that we found in Exercise 3(c).
A(s – 1)
- and lim f ( t ) = lim sF ( s ) = 10 . Then,
5. We are given that F ( s ) = -------------------
s(s + 1) t→∞ s→0
A(s – 1) ( s – 1 )- = – A = 10 . Therefore,
lim s -------------------- = A lim ---------------
s → 0 s(s + 1) s → 0 (s + 1)

10 ( s – 1 -) = ---
F ( s ) = –------------------------
r1 r2
- = 10
- + ---------- 20 - ⇔ ( 10 – 20e – t )u ( t ) , that is,
------ – ----------
s(s + 1) s s+1 s s+1 0

–t
f ( t ) = ( 10 – 20e )u 0 ( t ) and we see that lim f ( t ) = 10
t→∞

3-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 4
Circuit Analysis with Laplace Transforms

his chapter presents applications of the Laplace transform. Several examples are given to illus-
T trate how the Laplace transformation is applied to circuit analysis. Complex impedance, com-
plex admittance, and transfer functions are also defined.

4.1 Circuit Transformation from Time to Complex Frequency

In this section we will derive the voltage-current relationships for the three basic passive circuit
devices, i.e., resistors, inductors, and capacitors in the complex frequency domain.
a. Resistor
The time and complex frequency domains for purely resistive circuits are shown in Figure 4.1.
Time Domain Complex Frequency Domain
+
iR ( t ) v R ( t ) = Ri R ( t ) + V R ( s ) = RI R ( s )
vR ( t ) vR ( t ) IR ( s ) VR ( s )
R
i R ( t ) = -----------
- VR ( s ) I R ( s ) = -------------
-
R R
− R
−

Figure 4.1. Resistive circuit in time domain and complex frequency domain
b. Inductor
The time and complex frequency domains for purely inductive circuits is shown in Figure 4.2.
Time Domain Complex Frequency Domain
−
di L + V L ( s ) = sLI L ( s ) – Li L ( 0 )
+ v L ( t ) = L -------
dt −

`
sL IL ( s ) VL ( s ) iL ( 0 )
iL ( t )
vL ( t ) L
` I L ( s ) = ------------- + --------------
t
1
i L ( t ) = ---
L ∫–∞ vL dt VL ( s )
− −
Ls s

− + Li L ( 0 )
−
Figure 4.2. Inductive circuit in time domain and complex frequency domain
c. Capacitor
The time and complex frequency domains for purely capacitive circuits is shown in Figure 4.3.

Signals and Systems with MATLAB Applications, Second Edition 4-1

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

Time Domain Complex Frequency Domain

dv C + 1- −
+ i C ( t ) = C --------- ----- I C ( s ) = sCV C ( s ) – Cv C ( 0 )
dt sC +
C + iC ( t )
− IC ( s )
−
vC ( t ) t IC ( s ) vC ( 0 )
1 VC ( s ) V C ( s ) = -----------
- + ----------------
− v C ( t ) = ----
C ∫– ∞ i C d t −
sC s
− +
− vC ( 0 )
----------------
− s
Figure 4.3. Capacitive circuit in time domain and complex frequency domain
Note:
In the complex frequency domain, the terms sL and 1 ⁄ sC are called complex inductive impedance,
and complex capacitive impedance respectively. Likewise, the terms and sC and 1 ⁄ sL are called com-
plex capacitive admittance and complex inductive admittance respectively.
Example 4.1
Use the Laplace transform method to find the voltage v C ( t ) across the capacitor for the circuit of
Figure 4.4, given that v C ( 0 − ) = 6 V .

vS 1Ω
+
C +
− −
vC ( t )
12u 0 ( t ) V 1F

Figure 4.4. Circuit for Example 4.1

Solution:
We apply KCL at node A as shown in Figure 4.5.

R iR A
vS 1Ω
+
C + iC
− −
vC ( t )
12u 0 ( t ) V 1F

Figure 4.5. Application of KCL for the circuit of Example 4.1

Then,
iR + iC = 0
or

4-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Transformation from Time to Complex Frequency

v C ( t ) – 12u 0 ( t ) dv C
------------------------------------- + 1 ⋅ --------- = 0
1 dt
or
dv
--------C- + v C ( t ) = 12u 0 ( t ) (4.1)
dt
The Laplace transform of (4.1) is
sV C ( s ) – v C ( 0 ) + V C ( s ) = 12
−
------
s
or
12
( s + 1 )V C ( s ) = ------ + 6
s
or
6s + 12-
V C ( s ) = ------------------
s(s + 1)
By partial fraction expansion,
6s + 12 r r2
V C ( s ) = ------------------- = ----1 + ---------------
-
s(s + 1) s (s + 1)

r 1 = 6s + 12-
----------------- = 12
(s + 1) s=0

6s + 12
r 2 = ------------------ = –6
s s = –1
Therefore,
12 6 –t –t
V C ( s ) = ------ – ----------- ⇔ 12 – 6e = ( 12 – 6e )u 0 ( t ) = v C ( t )
s s+1
Example 4.2
Use the Laplace transform method to find the current i C ( t ) through the capacitor for the circuit of
Figure 4.6, given that v C ( 0 − ) = 6 V .

vS 1Ω iC ( t )
+
C +
− −
vC ( t )
12u 0 ( t ) V 1F

Figure 4.6. Circuit for Example 4.2

Signals and Systems with MATLAB Applications, Second Edition 4-3

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

Solution:
This is the same circuit as in Example 4.1. We apply KVL for the loop shown in Figure 4.7.

R 1Ω
vS
+
C +
− iC ( t ) −
vC ( t )
12u 0 ( t ) V 1F

Figure 4.7. Application of KVL for the circuit of Example 4.2

t
1
Ri C ( t ) + ----
C ∫– ∞ i C ( t ) d t = 12u 0 ( t )

and with R = 1 and C = 1 , we get

t
iC ( t ) + ∫– ∞ i C ( t ) d t = 12u 0 ( t ) (4.2)

Next, taking the Laplace transform of both sides of (4.2), we get

−
IC ( s ) vC ( 0 ) 12
I C ( s ) + -----------
- + ---------------- = ------
s s s

⎛1 + 1
--- ⎞ I C ( s ) = ------ – --- = ---
12 6 6
⎝ s⎠ s s s

s+1 ⎞
⎛ ----------
- I (s) = 6
---
⎝ s ⎠ C s
or
6 –t
I C ( s ) = ----------- ⇔ i C ( t ) = 6e u 0 ( t )
s+1
Check: From Example 4.1,
–t
v C ( t ) = ( 12 – 6e )u 0 ( t )
Then,
dv C dv d –t –t
i C ( t ) = C --------- = --------C- = ( 12 – 6e )u 0 ( t ) = 6e u 0 ( t ) + 6δ ( t ) (4.3)
dt dt dt

The presence of the delta function in (4.3) is a result of the unit step that is applied at t = 0 .

4-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Transformation from Time to Complex Frequency

Example 4.3
In the circuit of Figure 4.8, switch S 1 closes at t = 0 , while at the same time, switch S 2 opens. Use
the Laplace transform method to find v out ( t ) for t > 0 .

is ( t )
t = 0 2A
S2
t = 0 R1 L 1 0.5 H

`
1F
+
S1 2Ω i L1 ( t )

`
C + 0.5 H
R2 v out ( t )
− v ( 0− ) = 3 V
C 1Ω L 2
−
Figure 4.8. Circuit for Example 4.3
Solution:
Since the circuit contains a capacitor and an inductor, we must consider two initial conditions One
is given as v C ( 0 − ) = 3 V . The other initial condition is obtained by observing that there is an initial
current of 2 A in inductor L 1 ; this is provided by the 2 A current source just before switch S 2
opens. Therefore, our second initial condition is i L1 ( 0 − ) = 2 A .

For t > 0 , we transform the circuit of Figure 4.8 into its s-domain* equivalent shown in Figure 4.9.

1
`

+
+
−

1/s 2 0.5s 1V

+
− 3/s
1 0.5s
`V out ( s )

−
Figure 4.9. Transformed circuit of Example 4.3
In Figure 4.9 the current in L 1 has been replaced by a voltage source of 1 V . This is found from the
relation
− 1
L 1 i L1 ( 0 ) = --- × 2 = 1 V (4.4)
2
The polarity of this voltage source is as shown in Figure 4.9 so that it is consistent with the direction
of the current i L1 ( t ) in the circuit of Figure 4.8 just before switch S 2 opens.

* Henceforth, for convenience, we will refer the time domain as t-domain and the complex frequency domain as s-
domain

Signals and Systems with MATLAB Applications, Second Edition 4-5

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

The initial capacitor voltage is replaced by a voltage source equal to 3 ⁄ s .

Applying KCL at node , we get
V out ( s ) – 1 – 3 ⁄ s V out ( s ) V out ( s )
------------------------------------------ + ----------------- + ----------------- = 0 (4.5)
1⁄s+2+s⁄2 1 s⁄2
and after simplification

2s ( s + 3 ) -
V out ( s ) = ------------------------------------------ (4.6)
3 2
s + 8s + 10s + 4

We will use MATLAB to factor the denominator D ( s ) of (4.6) into a linear and a quadratic factor.
p=[1 8 10 4]; r=roots(p) % Find the roots of D(s)
r =
-6.5708
-0.7146 + 0.3132i
-0.7146 - 0.3132i
y=expand((s + 0.7146 − 0.3132j)*(s + 0.7146 + 0.3132j))% Find quadratic form
y =
s^2+3573/2500*s+3043737/5000000
3573/2500 % Find coefficient of s
ans =
1.4292
3043737/5000000 % Find constant term
ans =
0.6087
Therefore,
2s ( s + 3 ) 2s ( s + 3 )
V out ( s ) = ------------------------------------------- = ---------------------------------------------------------------------- (4.7)
3 2 2
s + 8s + 10s + 4 ( s + 6.57 ) ( s + 1.43s + 0.61 )
Now, we perform partial fraction expansion.
2s ( s + 3 ) r1 r2 s + r3
V out ( s ) = ---------------------------------------------------------------------- = ------------------
- + ----------------------------------------
- (4.8)
( s + 6.57 ) ( s + 1.43s + 0.61 ) s + 6.57 s + 1.43s + 0.61
2 2

2s ( s + 3 )
r 1 = ----------------------------------------- = 1.36 (4.9)
2
s + 1.43s + 0.61 s = – 6.57

4-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Transformation from Time to Complex Frequency

The residues r 2 and r 3 are found from the equality

2
2s ( s + 3 ) = r 1 ( s + 1.43s + 0.61 ) + ( r 2 s + r 3 ) ( s + 6.57 ) (4.10)

Equating constant terms of (4.10), we get

0 = 0.61r 1 + 6.57r 3

and by substitution of the known value of r 1 from (4.9), we get

r 3 = – 0.12

Similarly, equating coefficients of s 2 , we get

2 = r1 + r2
and using the known value of r 1 , we get

r 2 = 0.64 (4.11)
By substitution into (4.8),
0.64s – 0.12 0.64s + 0.46 – 0.58
V out ( s ) = ------------------- + ----------------------------------------- = ------------------- + ------------------------------------------------------- *
1.36 1.36
s + 6.57 s 2 + 1.43s + 0.61 s + 6.57 s 2 + 1.43s + 0.51 + 0.1
or
s + 0.715 – 0.91 -
1.36 - + ( 0.64 ) -------------------------------------------------------
V out ( s ) = ------------------
s + 6.57 ( s + 0.715 ) + ( 0.316 )
2 2

1.36 0.64 ( s + 0.715 ) 0.58

= ------------------- + -------------------------------------------------------- – -------------------------------------------------------- (4.12)
s + 6.57 ( s + 0.715 ) 2 + ( 0.316 ) 2 ( s + 0.715 ) 2 + ( 0.316 ) 2
1.36 0.64 ( s + 0.715 ) 1.84 × 0.316
= ------------------- + -------------------------------------------------------- – --------------------------------------------------------
s + 6.57 ( s + 0.715 ) + ( 0.316 ) ( s + 0.715 ) 2 + ( 0.316 ) 2
2 2

Taking the Inverse Laplace of (4.12), we get

– 6.57t – 0.715t – 0.715t
v out ( t ) = ( 1.36e + 0.64e cos 0.316t – 1.84e sin 0.316t )u 0 ( t ) (4.13)

0.64s – 0.12 -
* We perform these steps to express the term ----------------------------------------
2
in a form that resembles the transform pairs
s + 1.43s + 0.61
– at s+a ω
- and e – at sin ωtu 0 ( t ) ⇔ -------------------------------- . The remaining steps are carried out in
e cos ωtu 0 ( t ) ⇔ -------------------------------
2 2 2 2
(s + a) + ω (s + a) + ω
(4.12).

Signals and Systems with MATLAB Applications, Second Edition 4-7

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

4.2 Complex Impedance Z(s)

Consider the s – domain RLC series circuit of Figure 4.10, where the initial conditions are assumed
to be zero.
sL
R

`
+
+ V out ( s )
− I(s ) 1-
-----
VS ( s ) sC
−
Figure 4.10. Series RLC circuit in s-domain
1
For this circuit, the sum R + sL + -----
- represents the total opposition to current flow. Then,
sC

VS ( s )
I ( s ) = ----------------------------
- (4.14)
R + sL + ----- 1-
sC

and defining the ratio V s ( s ) ⁄ I ( s ) as Z ( s ) , we get

VS ( s ) 1-
Z ( s ) ≡ ------------- = R + sL + ----- (4.15)
I(s ) sC

and thus, the s – domain current I ( s ) can be found from

VS ( s )
I ( s ) = ------------
- (4.16)
Z(s )

where

1
Z ( s ) = R + sL + ------ (4.17)
sC

We recall that s = σ + j ω . Therefore, Z ( s ) is a complex quantity, and it is referred to as the complex

input impedance of an s – domain RLC series circuit. In other words, Z ( s ) is the ratio of the voltage
excitation V s ( s ) to the current response I ( s ) under zero state (zero initial conditions).

Example 4.4
Find Z ( s ) for the circuit of Figure 4.11. All values are in Ω (ohms).

4-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Complex Impedance Z(s)

+
1 1⁄s
VS ( s ) s
` s
`
−
Figure 4.11. Circuit for Example 4.4
Solution:
First Method:
We will first find I ( s ) , and we will compute Z ( s ) from (4.15). We assign the voltage V A ( s ) at node
A as shown in Figure 4.12.
1 VA ( s ) 1⁄s
+ A
I( s)
VS ( s ) s
` s
`
−
Figure 4.12. Circuit for finding I ( s ) in Example 4.4
By nodal analysis,
VA ( s ) – VS ( s ) VA ( s ) VA ( s )
--------------------------------- + -------------- + ------------------ = 0
1 s s+1⁄s

⎛1 + 1
--- + ------------------ ⎞ V A ( s ) = V S ( s )
1
⎝ s s+1⁄s ⎠

3
s +1
V A ( s ) = ------------------------------------
3 2
- ⋅ VS ( s )
s + 2s + s + 1

The current I ( s ) is now found as

VS ( s ) – VA ( s ) ⎛ 3
s +1 ⎞ 2
2s + 1 -
I ( s ) = --------------------------------
- = ⎜ 1 – ------------------------------------- ⎟ V S ( s ) = ------------------------------------ ⋅ VS ( s )
1 ⎝ s + 2s + s + 1 ⎠
3 2 3
s + 2s + s + 1
2

and thus,
VS ( s ) 3
s + 2s + s + 1
2
Z ( s ) = ------------
- = ------------------------------------- (4.18)
I(s ) 2s + 1
2

Second Method:

Signals and Systems with MATLAB Applications, Second Edition 4-9

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

We can also compute Z ( s ) by successive combinations of series and parallel impedances, as it is

done with series and parallel resistances. For this example, we denote the network devices as
Z 1, Z 2, Z 3 and Z 4 shown in Figure 4.13.

1 1⁄s
a
Z1 Z3
Z(s) s
` Z2 s
`Z 4

b
Figure 4.13. Computation of the impedance of Example 4.4 by series − parallel combinations
To find the equivalent impedance Z ( s ) , looking to the right of terminals a and b , we start on the
right side of the network and we proceed to the left combining impedances as we combine resis-
tances. Then,
Z ( s ) = [ ( Z 3 + Z 4 ) || Z 2 ] + Z 1

2 3 3 2
s(s + 1 ⁄ s ) s +1 s +s s + 2s + s + 1
Z ( s ) = -------------------------- + 1 = ---------------------------- + 1 = ----------------- + 1 = ------------------------------------- (4.19)
s+s+1⁄s 2
( 2s + 1 ) ⁄ s 2s + 1
2
2s + 1
2

We observe that (4.19) is the same as (4.18).

4.3 Complex Admittance Y(s)

Consider the s – domain GLC parallel circuit of Figure 4.14 where the initial conditions are zero.

`
1-
V(s ) G -----
sL
IS ( s ) sC
−
Figure 4.14. Parallel GLC circuit in s-domain
For this circuit,
1
GV ( s ) + ------ V ( s ) + sCV ( s ) = I ( s )
sL

⎛ G + -----
1
- + sC⎞ ( V ( s ) ) = I ( s )
⎝ sL ⎠

Defining the ratio I S ( s ) ⁄ V ( s ) as Y ( s ) , we get

I(s) 1 1
Y ( s ) ≡ ----------- = G + ------ + sC = ---------- (4.20)
V(s) sL Z(s)

4-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Complex Admittance Y(s)

and thus the s – domain voltage V ( s ) can be found from

IS ( s )
V ( s ) = ----------- (4.21)
Y(s)

where

1
Y ( s ) = G + ------ + sC (4.22)
sL

We recall that s = σ + j ω . Therefore, Y ( s ) is a complex quantity, and it is referred to as the complex

input admittance of an s – domain GLC parallel circuit. In other words, Y ( s ) is the ratio of the cur-
rent excitation I S ( s ) to the voltage response V ( s ) under zero state (zero initial conditions).

Example 4.5
Compute Z ( s ) and Y ( s ) for the circuit of Figure 4.15. All values are in Ω (ohms). Verify your
answers with MATLAB.
`

13s
8⁄s 10 20

Z(s)
Y(s)
` 5s 16 ⁄ s

Figure 4.15. Circuit for Example 4.5

Solution:
It is convenient to represent the given circuit as shown in Figure 4.16.

Z1

Z ( s ), Y ( s ) Z2 Z3

Figure 4.16. Simplified circuit for Example 4.5

where

Signals and Systems with MATLAB Applications, Second Edition 4-11

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

2
Z 1 = 13s + 8 13s + 8-
--- = -------------------
s s

Z 2 = 10 + 5s

16 4 ( 5s + 4 )
Z 3 = 20 + ------ = -----------------------
s s
Then,
4 ( 5s + 4 )
( 10 + 5s ) ⎛ -----------------------⎞
Z2 Z3 2 ⎝ ⎠
+ 8- + ---------------------------------------------------
s
Z ( s ) = Z1 + ----------------- = 13s
------------------- -
Z2 + Z3 s 4 ( 5s + 4 )
10 + 5s + -----------------------
s
4 ( 5s + 4 )
( 10 + 5s ) ⎛ -----------------------⎞
2 ⎝
13s + 8- + ----------------------------------------------------
s ⎠ 13s 2 + 8 20 ( 5s 2 + 14s + 8 )
= ------------------- - = -------------------- + -------------------------------------------
s 2 s 2
5s + 10s + 4 ( 5s + 4 )
---------------------------------------------------- 5s + 30s + 16
s
4 3 2
65s + 490s + 528s + 400s + 128
= -------------------------------------------------------------------------------------
2
s ( 5s + 30s + 16 )
Check with MATLAB:
syms s; z1 = 13*s + 8/s; z2 = 5*s + 10; z3 = 20 + 16/s; z = z1 + z2 * z3 / (z2+z3)
z =
13*s+8/s+(5*s+10)*(20+16/s)/(5*s+30+16/s)
z10 = simplify(z)
z10 =
(65*s^4+490*s^3+528*s^2+400*s+128)/s/(5*s^2+30*s+16)
pretty(z10)
4 3 2
65 s + 490 s + 528 s + 400 s + 128
-------------------------------------
2
s (5 s + 30 s + 16)
The complex input admittance Y ( s ) is found by taking the reciprocal of Z ( s ) , that is,
2
Y ( s ) = ---------- s ( 5s + 30s + 16 )
1 = ------------------------------------------------------------------------------------
- (4.23)
Z(s) 4 3
65s + 490s + 528s + 400s + 128
2

4-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Transfer Functions

4.4 Transfer Functions

In an s – domain circuit, the ratio of the output voltage V out ( s ) to the input voltage V in ( s ) under
zero state conditions, is of great interest in network analysis. This ratio is referred to as the voltage
transfer function and it is denoted as G v ( s ) , that is,

V out ( s )
G v ( s ) ≡ ----------------- (4.24)
V in ( s )

Similarly, the ratio of the output current I out ( s ) to the input current I in ( s ) under zero state condi-
tions, is called the current transfer function denoted as G i ( s ) , that is,

I out ( s )
G i ( s ) ≡ ---------------- (4.25)
I in ( s )

The current transfer function of (4.25) is rarely used; therefore, from now on, the transfer function
will have the meaning of the voltage transfer function, i.e.,

V out ( s )
G ( s ) ≡ ----------------- (4.26)
V in ( s )

Example 4.6
Derive an expression for the transfer function G ( s ) for the circuit of Figure 4.17, where R g repre-
sents the internal resistance of the applied (source) voltage V S , and R L represents the resistance of
the load that consists of R L , L , and C .

+
RL

Rg

`
L v out
vg
+ C
−

−
Figure 4.17. Circuit for Example 4.6
Solution:
No initial conditions are given, and even if they were, we would disregard them since the transfer
function was defined as the ratio of the output voltage V out ( s ) to the input voltage V in ( s ) under

Signals and Systems with MATLAB Applications, Second Edition 4-13

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

zero initial conditions. The s – domain circuit is shown in Figure 4.18.

+
RL
Rg

V in ( s )
sL
`V out ( s )

+
− 1
------
sC
−
Figure 4.18. The s-domain circuit for Example 4.6
The transfer function G ( s ) is readily found by application of the voltage division expression of the
s – domain circuit of Figure 4.18, i.e.,
R L + sL + 1 ⁄ sC
V out ( s ) = --------------------------------------------------- V in ( s )
R g + R L + sL + 1 ⁄ sC
Then,
V out ( s ) R L + Ls + 1 ⁄ sC
G ( s ) = ----------------
- = --------------------------------------------------
- (4.27)
V in ( s ) R g + R L + Ls + 1 ⁄ sC

Example 4.7
Compute the transfer function G ( s ) for the circuit of Figure 4.19 in terms of the circuit constants
R 1, R 2, R 3, C 1, and C 2 Then, replace the complex variable s with jω , and the circuit constants with
their numerical values and plot the magnitude G ( s ) = V out ( s ) ⁄ V in ( s ) versus radian frequency ω .

C2 10 nF
R2 40 K
R1 R3

200 K 50K
vin vout
C1
25 nF

Figure 4.19. Circuit for Example 4.7

Solution:
The complex frequency domain equivalent circuit is shown in Figure 4.20.

4-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Transfer Functions

1/sC2
R2
R1 R3
1 2
V1 ( s ) V2 ( s )
Vin (s) Vout (s)
1/sC1

Figure 4.20. The s-domain circuit for Example 4.7

Next, we write nodal equations at nodes 1 and 2.
At node 1,
V 1 ( s ) – V in ( s ) V1 V 1 ( s ) – V out ( s ) V 1 ( s ) – V 2 ( s )
---------------------------------- + --------------
- + ------------------------------------
- + -------------------------------- = 0 (4.28)
R1 1 ⁄ sC 1 R2 R3
At node 2,

V2 ( s ) – V1 ( s ) V out ( s )
- = -----------------
------------------------------- - (4.29)
R3 1 ⁄ sC 2

Since V 2 ( s ) = 0 (virtual ground), we express (4.29) as

V 1 ( s ) = ( – sR 3 C 2 )V out ( s ) (4.30)

and by substitution of (4.30) into (4.28), rearranging, and collecting like terms, we get:

⎛ ----
1 1
- + ---- 1
- + sC 1 ⎞ ( – sR 3 C 2 ) – ----
- + ---- 1 1
- V ( s ) = ----- V in ( s )
⎝ R1 R2 R3 ⎠ R 2 out R1
or
V out ( s ) –1
G ( s ) = -----------------
- = ----------------------------------------------------------------------------------------------------
- (4.31)
V in ( s ) ⎛ 1 1 1 ⎞ 1
R1 ----- + ----- + ----- + sC 1 ( sR 3 C 2 ) + -----
⎝ R1 R2 R3 ⎠ R2

By substitution of s with jω and the given values for resistors and capacitors, we get
–1
G ( j ω ) = -------------------------------------------------------------------------------------------------------------------------------------------------------------
-
5 ⎛ 1 –8 ⎞ –8 1
2 × 10 ------------------- + j 2.5 × 10 ω ( j 5 × 10 × 10 ω ) + ---------------4-
4
⎝ 20 × 10 3 ⎠ 4 × 10
or

Signals and Systems with MATLAB Applications, Second Edition 4-15

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

V out ( j ω ) –1
G ( j ω ) = --------------------- = ------------------------------------------------------------------------
– –3
(4.32)
V in ( j ω ) 2.5 × 10 ω – j5 × 10 ω + 5
6 2

We use MATLAB to plot the magnitude of (4.32) on a semilog scale with the following code:
w=1:10:10000; Gs=−1./(2.5.*10.^(−6).*w.^2−5.*j.*10.^(−3).*w+5);
semilogx(w,abs(Gs)); grid; hold on
xlabel('Radian Frequency w'); ylabel('|Vout/Vin|');
title('Magnitude Vout/Vin vs. Radian Frequency')
The plot is shown in Figure 4.21. We observe that the given op amp circuit is a second order low-
pass filter whose cutoff frequency ( – 3 dB ) occurs at about 700 r ⁄ s .

Figure 4.21. G ( jω ) versus ω for the circuit of Example 4.7

4-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

4.5 Summary
• The Laplace transformation provides a convenient method of analyzing electric circuits since
integrodifferential equations in the t – domain are transformed to algebraic equations in the
s – domain .

• In the s – domain the terms sL and 1 ⁄ sC are called complex inductive impedance, and com-
plex capacitive impedance respectively. Likewise, the terms and sC and 1 ⁄ sL are called complex
capacitive admittance and complex inductive admittance respectively.
• The expression
1
Z ( s ) = R + sL + ------
sC
is a complex quantity, and it is referred to as the complex input impedance of an s – domain
RLC series circuit.

• In the s – domain the current I ( s ) can be found from

VS ( s )
I ( s ) = ------------
-
Z(s)
• The expression
1
Y ( s ) = G + ------ + sC
sL
is a complex quantity, and it is referred to as the complex input admittance of an s – domain
GLC parallel circuit.

• In the s – domain the voltage V ( s ) can be found from

IS ( s )
V ( s ) = -----------
Y(s )

• In an s – domain circuit, the ratio of the output voltage V out ( s ) to the input voltage V in ( s )
under zero state conditions is referred to as the voltage transfer function and it is denoted as
G ( s ) , that is,
V out ( s )
G ( s ) ≡ -----------------
V in ( s )

Signals and Systems with MATLAB Applications, Second Edition 4-17

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

4.6 Exercises
1. In the circuit of Figure 4.22, switch S has been closed for a long time, and opens at t = 0 . Use
the Laplace transform method to compute i L ( t ) for t > 0 .

t = 0 R1

S 10 Ω
L
20 Ω
i (t) `
1 mH +
R2 −
L 32 V

Figure 4.22. Circuit for Exercise 1

2. In the circuit of Figure 4.23, switch S has been closed for a long time, and opens at t = 0 . Use
the Laplace transform method to compute v c ( t ) for t > 0 .

R1 t = 0 R3 R4

6 KΩ S 30 KΩ 20 KΩ

60 KΩ C +
+ R2 v (t) R5
− 40 − C
72 V ------ µF 10 KΩ
9

Figure 4.23. Circuit for Exercise 2

3. Use mesh analysis and the Laplace transform method, to compute i 1 ( t ) and i 2 ( t ) for the circuit
of Figure 4.24, given that i L (0 − ) = 0 and v c (0 − ) = 0 .

L1 R2
`

2H 3Ω

`1 H
R1 1Ω L2
+
− +
i1 ( t )
C +
v1 ( t ) = u0 ( t ) − i2 ( t ) −
1F
v 2 ( t ) = 2u 0 ( t )

Figure 4.24. Circuit for Exercise 3

4-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

4. For the s – domain circuit of Figure 4.25,

a. compute the admittance Y ( s ) = I 1 ( s ) ⁄ V 1 ( s )

b. compute the t – domain value of i 1 ( t ) when v 1 ( t ) = u 0 ( t ) , and all initial conditions are zero.

I1 ( s ) VC ( s )
R1 − R3

+
1Ω 1⁄s 3Ω

+ 1Ω

+
R2
− −

V1 ( s ) R4 2Ω V 2 ( s ) = 2V C ( s )

Figure 4.25. Circuit for Exercise 4

5. Derive the transfer functions for the networks (a) and (b) of Figure 4.26.

R L

+ + +
` +
C
V in ( s ) V out ( s ) V in ( s ) R V out ( s )

− (b)
− (a)
− −

Figure 4.26. Networks for Exercise 5

6. Derive the transfer functions for the networks (a) and (b) of Figure 4.27.

C R

+ + + +

V in ( s ) V out ( s ) V in ( s )
`V out ( s )
R L

(b)
− (a) − − −

Figure 4.27. Networks for Exercise 6

7. Derive the transfer functions for the networks (a) and (b) of Figure 4.28.

Signals and Systems with MATLAB Applications, Second Edition 4-19

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

`
+ L C + + R +
V in ( s ) R V out ( s ) V in ( s )
L
`V out ( s )

(b)
−
(a) − − −

Figure 4.28. Networks for Exercise 7

8. Derive the transfer function for the networks (a) and (b) of Figure 4.29.

R2
C

R1 R2 R1 C

V in ( s ) V in ( s ) V out ( s )
V out ( s )

(a) (b)
Figure 4.29. Networks for Exercise 8
9. Derive the transfer function for the network of Figure 4.30. Using MATLAB, plot G ( s ) versus
frequency in Hertz, on a semilog scale.

R4 R1 = 11.3 kΩ
R2 = 22.6 kΩ
R3 R3=R4 = 68.1 kΩ

R1 R2 C1=C2 = 0.01 µF

V out ( s )
V in ( s ) C1

C2

Figure 4.30. Network for Exercise 9

4-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

4.7 Solutions to Exercises

−
1. At t = 0 the t – domain circuit is as shown below and the 20 Ω resistor is shorted out by the
inductor.
S 10 Ω
L
20 Ω
i (t) `
1 mH +
−
L 32 V

Then,
32
iL ( t ) = ------ = 3.2 A
t=0
- 10

−
and thus the initial condition has been established as i L ( 0 ) = 3.2 A

For all t > 0 the t – domain and s – domain circuits are as shown below.

IL ( s )
–3
−
i L ( 0 ) = 3.2 A ` 10 s
20 Ω
` 20 Ω
−
1 mH +
− –3
Li L ( 0 ) = 3.2 × 10 V

From the s – domain circuit above we get

–3
3.2 × 10 3.2 – 20000t
I L ( s ) = ------------------------- = ----------------------- ⇔ 3.2e u0 ( t ) = iL ( t )
20 + 10 s s + 20000
–3

−
2. At t = 0 the t – domain circuit is as shown below.

iT ( t )
6 KΩ 30 KΩ + 20 KΩ
i2 ( t )
+ 60 KΩ vC ( t ) 10 KΩ
−
72 V
−

Then,

Signals and Systems with MATLAB Applications, Second Edition 4-21

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

− 72 V 72 V 72 V
i T ( 0 ) = ------------------------------------------------------------ = ------------------------------------- = ----------------- = 2 mA
6 KΩ + 60 KΩ || 60 KΩ 6 KΩ + 30 KΩ 36 KΩ
and
− 1 −
i 2 ( 0 ) = --- i T ( 0 ) = 1 mA
2
Therefore, the initial condition is
− −
v C ( 0 ) = ( 20 KΩ + 10 KΩ ) ⋅ i 2 ( 0 ) = ( 30 KΩ ) ⋅ ( 1 mA ) = 30 V

For all t > 0 the s – domain circuit is as shown below.

30 KΩ
VR = VC ( s )
20 KΩ 6
1 VC ( s ) 9 × 10
-------------------
---------------------------------
–6
- 40s +
60 KΩ 40 ⁄ 9 × 10 s 10 KΩ
+ VR 22.5 KΩ
− +
− −
30 ⁄ s 30 ⁄ s

( 60 KΩ + 30 KΩ ) || ( 20 KΩ + 10 KΩ ) = 22.5 KΩ

3
22.5 × 10 30 30 × 22.5 × 10
3

6
- ⋅ ------ = ------------------------------------------------------------
V C ( s ) = V R = ------------------------------------------------------------
3
-
9 × 10 ⁄ 40s + 22.5 × 10 s 6
9 × 10 ⁄ 40 + 22.5 × 10 s
3

3 3
( 30 × 22.5 × 10 ) ⁄ ( 22.5 × 10 ) 30 30
= ---------------------------------------------------------------------------- = --------------------------------------------------- = --------------
6
9 × 10 ⁄ ( 40 × 22.5 × 10 ) + s
3 6
9 × 10 ⁄ 90 × 10 + s
4 10 +s

Then,
30 – 10t
V C ( s ) = -------------- ⇔ 30e u 0 ( t ) V = v C ( t )
s + 10

3. The s – domain circuit is shown below where z 1 = 2s , z 2 = 1 + 1 ⁄ s , and z 3 = s + 3

z3
z1
`

2s 3

+
−
1
z2 `s
1⁄s + +
1⁄s I1 ( s ) I2 ( s ) −
−
2⁄s

4-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

Then,
( z 1 + z 2 )I 1 ( s ) – z 2 I 2 ( s ) = 1 ⁄ s
– z 2 I 1 ( s ) + ( z 2 + z 3 )I 2 ( s ) = – 2 ⁄ s
and in matrix form
( z1 + z2 ) –z2 I1 ( s ) 1⁄s
⋅ =
–z2 ( z2 + z3 ) I2 ( s ) –2 ⁄ s
Using MATLAB we get
Z=[z1+z2 −z2; −z2 z2+z3]; Vs=[1/s −2/s]'; Is=Z\Vs; fprintf(' \n');...
disp('Is1 = '); pretty(Is(1)); disp('Is2 = '); pretty(Is(2))
Is1 =

2
2 s - 1 + s
-------------------------------
2 3
(6 s + 3 + 9 s + 2 s ) conj(s)
Is2 =

2
4 s + s + 1
- -------------------------------
2 3
(6 s + 3 + 9 s + 2 s ) conj(s)
Therefore,
2
s + 2s – 1 - (1)
I 1 ( s ) = -------------------------------------------
3 2
2s + 9s + 6s + 3
2
4s + s + 1
I 2 ( s ) = – -------------------------------------------- (2)
3 2
2s + 9s + 6s + 3
We express the denominator of (1) as a product of a linear and quadratic term using MATLAB.
p=[2 9 6 3]; r=roots(p); fprintf(' \n'); disp('root1 ='); disp(r(1));...
disp('root2 ='); disp(r(2)); disp('root3 ='); disp(r(3)); disp('root2 + root3 =');
disp(r(2)+r(3));...
disp('root2 * root3 ='); disp(r(2)*r(3))
root1 =
-3.8170

Signals and Systems with MATLAB Applications, Second Edition 4-23

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

root2 =
-0.3415 + 0.5257i
root3 =
-0.3415 - 0.5257i
root2 + root3 =
-0.6830
root2 * root3 =
0.3930
and with these values (1) is written as
2
s + 2s – 1 r1 r2 s + r3
I 1 ( s ) = ----------------------------------------------------------------------------------- = -------------------------- - (3)
- + ---------------------------------------------------
2
( s + 3.817 ) ⋅ ( s + 0.683s + 0.393 ) ( s + 3.817 ) ( s + 0.683s + 0.393 )
2

Multiplying every term by the denominator and equating numerators we get

2 2
s + 2s – 1 = r 1 ( s + 0.683s + 0.393 ) + ( r 2 s + r 3 ) ( s + 3.817 )

2
Equating s , s , and constant terms we get
r1 + r2 = 1
0.683r 1 + 3.817r 2 + r 3 = 2
0.393r 1 + 3.817r 3 = – 1

We will use MATLAB to find these residues.

A=[1 1 0; 0.683 3.817 1; 0.393 0 3.817]; B=[1 2 −1]'; r=A\B; fprintf(' \n');...
fprintf('r1 = %5.2f \t',r(1)); fprintf('r2 = %5.2f \t',r(2)); fprintf('r3 = %5.2f',r(3))
r1 = 0.48 r2 = 0.52 r3 = -0.31
By substitution of these values into (3) we get
r1 r2 s + r3 0.52s – 0.31
0.48 - + ---------------------------------------------------
I 1 ( s ) = --------------------------
- + ---------------------------------------------------
- = -------------------------- - (4)
( s + 3.817 ) ( s + 0.683s + 0.393 )
2 ( s + 3.817 ) ( s + 0.683s + 0.393 )
2

By inspection, the Inverse Laplace of first term on the right side of (4) is
0.48
------------------------ ⇔ 0.48e – 3.82t (5)
( s + 3.82 )
The second term on the right side of (4) requires some manipulation. Therefore, we will use the
MATLAB ilaplace(s) function to find the Inverse Laplace as shown below.
syms s t

4-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

IL=ilaplace((0.52*s−0.31)/(s^2+0.68*s+0.39));
pretty(IL)
1217 17 1/2 1/2
- ---- exp(- -- t) 14 sin(7/50 14 t)
4900 50

13 17 1/2
+ -- exp(- -- t) cos(7/50 14 t)
25 50
Thus,
– 3.82t – 0.34t – 0.34t
i 1 ( t ) = 0.48e – 0.93e sin 0.53t + 0.52e cos 0.53t

Next, we will find I 2 ( s ) . We found earlier that

2
4s + s + 1 -
I 2 ( s ) = – -------------------------------------------
3 2
2s + 9s + 6s + 3
and following the same procedure we have
2
– 4s – s – 1 r1 r2 s + r3
I 2 ( s ) = ----------------------------------------------------------------------------------- = -------------------------- - (6)
- + ---------------------------------------------------
2
( s + 3.817 ) ⋅ ( s + 0.683s + 0.393 ) ( s + 3.817 ) ( s 2 + 0.683s + 0.393 )

Multiplying every term by the denominator and equating numerators we get

2 2
– 4s – s – 1 = r 1 ( s + 0.683s + 0.393 ) + ( r 2 s + r 3 ) ( s + 3.817 )

2
Equating s , s , and constant terms we get
r1 + r2 = –4
0.683r 1 + 3.817r 2 + r 3 = – 1
0.393r 1 + 3.817r 3 = – 1

We will use MATLAB to find these residues.

A=[1 1 0; 0.683 3.817 1; 0.393 0 3.817]; B=[−4 −1 −1]'; r=A\B; fprintf(' \n');...
fprintf('r1 = %5.2f \t',r(1)); fprintf('r2 = %5.2f \t',r(2)); fprintf('r3 = %5.2f',r(3))
r1 = -4.49 r2 = 0.49 r3 = 0.20
By substitution of these values into (6) we get
r1 r2 s + r3 0.49s + 0.20
– 4.49 - + ---------------------------------------------------
I 1 ( s ) = --------------------------
- + ---------------------------------------------------
- = -------------------------- - (7)
( s + 3.817 ) ( s + 0.683s + 0.393 )
2 ( s + 3.817 ) ( s + 0.683s + 0.393 )
2

Signals and Systems with MATLAB Applications, Second Edition 4-25

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

By inspection, the Inverse Laplace of first term on the right side of (7) is
0.48
------------------------ ⇔ – 4.47 e – 3.82t (8)
( s + 3.82 )
The second term on the right side of (7) requires some manipulation. Therefore, we will use the
MATLAB ilaplace(s) function to find the Inverse Laplace as shown below.
syms s t
IL=ilaplace((0.49*s+0.20)/(s^2+0.68*s+0.39)); pretty(IL)
167 17 1/2 1/2
---- exp(- -- t) 14 sin(7/50 14 t)
9800 50

49 17 1/2
+ --- exp(- -- t) cos(7/50 14 t)
100 50
Thus,
– 3.82t – 0.34t – 0.34t
i 2 ( t ) = – 4.47 e + 0.06e sin 0.53t + 0.49e cos 0.53t

4.

VC ( s )
−
+

1 3
1⁄s

+
+

1
− −
I1 ( s ) I2 ( s )
V1 ( s ) 2 V 2 ( s ) = 2V C ( s )

a. Mesh 1:
( 2 + 1 ⁄ s ) ⋅ I1 ( s ) – I2 ( s ) = V1 ( s )
or
6 ( 2 + 1 ⁄ s ) ⋅ I 1 ( s ) – 6I 2 ( s ) = 6V 1 ( s ) (1)
Mesh 2:
– I 1 ( s ) + 6I 2 ( s ) = – V 2 ( s ) = – ( 2 ⁄ s )I 1 ( s ) (2)

Addition of (1) and (2) yields

( 12 + 6 ⁄ s ) ⋅ I 1 ( s ) + ( 2 ⁄ s – 1 ) ⋅ I 1 ( s ) = 6V 1 ( s )
or

4-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

( 11 + 8 ⁄ s ) ⋅ I 1 ( s ) = 6V 1 ( s )
and thus
I1 ( s ) 6 6s
Y ( s ) = ------------
- = --------------------- = ------------------
V 1 ( s ) 11 + 8 ⁄ s 11s + 8

b. With V 1 ( s ) = 1 ⁄ s we get

6s 1 6 6 ⁄ 11 6 – ( 8 ⁄ 11 )t
I 1 ( s ) = Y ( s ) ⋅ V 1 ( s ) = ------------------ ⋅ --- = ------------------ = --------------------- ⇔ ------ e = i1 ( t )
11s + 8 s 11s + 8 s + 8 ⁄ 11 11

5.
Circuit (a):
R

+ +
1 ⁄ Cs
V in ( s ) V out ( s )

− −

1 ⁄ Cs
V out ( s ) = ----------------------- ⋅ V in ( s )
R + 1 ⁄ Cs
and
V out ( s ) 1 ⁄ Cs 1 ⁄ Cs 1 1 ⁄ RC
G ( s ) = ----------------
- = ----------------------- = ---------------------------------------- = -------------------- = -----------------------
V in ( s ) R + 1 ⁄ Cs ( RCs + 1 ) ⁄ ( Cs ) RCs + 1 s + 1 ⁄ RC

Circuit (b):
L
`

+ +

V in ( s ) R V out ( s )

− −

R
V out ( s ) = ---------------- ⋅ V in ( s )
Ls + R
and
V out ( s ) R⁄L -
R - = ------------------
G ( s ) = ----------------
- = ---------------
V in ( s ) Ls + R s+R⁄L

Both of these circuits are first-order low-pass filters.

Signals and Systems with MATLAB Applications, Second Edition 4-27

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

6.
Circuit (a):
C

+ +

V in ( s ) R V out ( s )

− −

R
V out ( s ) = ----------------------- ⋅ V in ( s )
1 ⁄ Cs + R
and
V out ( s ) R RCs s
G ( s ) = ----------------
- = ----------------------- = ------------------------ = -----------------------
V in ( s ) 1 ⁄ Cs + R ( RCs + 1 ) s + 1 ⁄ RC

Circuit (b):
R

+
+
V in ( s ) L
` V− out ( s )

−
Ls
V out ( s ) = ---------------- ⋅ V in ( s )
R + Ls
and
V out ( s ) Ls s
G ( s ) = ----------------
- = ---------------- = -------------------
V in ( s ) R + Ls s+R⁄L

Both of these circuits are first-order high-pass filters.

7.
Circuit (a):
`

+ L C +

V in ( s ) R V out ( s )

− −

R
V out ( s ) = ------------------------------------ ⋅ V in ( s )
Ls + 1 ⁄ Cs + R

4-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

and
V out ( s ) R RCs ( R ⁄ L )s
G ( s ) = ----------------
- = ------------------------------------ = --------------------------------------- = --------------------------------------------------
V in ( s ) Ls + 1 ⁄ Cs + R 2
LCs + 1 + RCs
2
s + ( R ⁄ L )s + 1 ⁄ LC
This circuit is a second-order band-pass filter.
Circuit (b):
+ R +

V in ( s )
L
`V out ( s )

− −

Ls + 1 ⁄ Cs
V out ( s ) = ------------------------------------ ⋅ V in ( s )
R + Ls + 1 ⁄ Cs
and
V out ( s ) Ls + 1 ⁄ Cs LCs + 1
2 2
s + 1 ⁄ LC
G ( s ) = ----------------
- = ------------------------------------ = --------------------------------------- = --------------------------------------------------
V in ( s ) R + Ls + 1 ⁄ Cs 2
LCs + RCs + 1
2
s + ( R ⁄ L )s + 1 ⁄ LC
This circuit is a second-order band-elimination (band-reject) filter.
8.
Circuit (a):

R1 R2

V in ( s )
V out ( s )

R × 1 ⁄ Cs V (s) z
Let z 1 = R 1 and z 2 = -------------------------
2
- and since for inverting op-amp ----------------
out
- = – ---2- , for this circuit
R 2 + 1 ⁄ Cs V in ( s ) z1

V out ( s ) – [ ( R 2 × 1 ⁄ Cs ) ⁄ ( R 2 + 1 ⁄ Cs ) ] – ( R 2 × 1 ⁄ Cs ) –R1 C
G ( s ) = ----------------
- = -------------------------------------------------------------------------
- = ----------------------------------------
- = -------------------------
-
V in ( s ) R1 R 1 ⋅ ( R 2 + 1 ⁄ Cs ) s + 1 ⁄ R2 C

This circuit is a first-order active low-pass filter.

Signals and Systems with MATLAB Applications, Second Edition 4-29

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

Circuit (b):
R2

R1 C

V in ( s ) V out ( s )

V (s) z
Let z 1 = R 1 + 1 ⁄ Cs and z 2 = R 2 and since for inverting op-amp ----------------
out
- = – ---2- , for this circuit
V in ( s ) z1

V out ( s ) –R2 – ( R 2 ⁄ R 1 )s
G ( s ) = ----------------
- = -------------------------
- = --------------------------
V in ( s ) R 1 + 1 ⁄ Cs s + 1 ⁄ R1C

This circuit is a first-order active high-pass filter.

9.
R4 R1 = 11.3 KΩ
R2 = 22.6 KΩ
R3 V1 R3=R4 = 68.1 KΩ

R1 R2 V3 C1=C2 = 0.01 µF

V2
V out ( s )
V in ( s ) C1

C2

At Node V 1 :
V 1 ( s ) V 1 ( s ) – V out ( s )
------------- + ------------------------------------- = 0
R3 R4
or
⎛ ----- 1- ⎞ V ( s ) = -----
1- + ----- 1
- V ( s ) (1)
⎝R R4 ⎠ 1
R 4 out
3
At Node V 3 :
V3 ( s ) – V2 ( s ) V3 ( s )
--------------------------------- + ---------------- = 0
R2 1 ⁄ C1 s

4-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

and since V 3 ( s ) = V 1 ( s ) we express the last relation above as

V1 ( s ) – V2 ( s )
- + C 1 sV 1 ( s ) = 0
--------------------------------
R2
or
⎛ -----
1
- + C 1 s⎞ V 1 ( s ) = ------ V 2 ( s ) (2)
1
⎝R ⎠ R 2
2

At Node V 2 :
V 2 ( s ) – V in ( s ) V 2 ( s ) – V 1 ( s ) V 2 ( s ) – V out ( s )
---------------------------------- + --------------------------------- + ------------------------------------- = 0
R1 R2 1 ⁄ C2 s
or
⎛ ----- V in ( s ) V 1 ( s )
- + ------ + C 2 s⎞ V 2 ( s ) = --------------
1 1
- + ------------- + C 2 sV out ( s ) (3)
⎝R R2 ⎠ R1 R2
1
From (1)
( 1 ⁄ R4 ) R3
V 1 ( s ) = --------------------------------------- V out ( s ) = ----------------------- V out ( s ) (4)
( R3 + R4 ) ⁄ R3 R4 ( R3 + R4 )

From (2)
V 2 ( s ) = R 2 ⎛ ------ + C 1 s⎞ V 1 ( s ) = ( 1 + R 2 C 1 s )V 1 ( s )
1
⎝R ⎠
2

and with (4)

R3 ( 1 + R2 C1 s )
V 2 ( s ) = ------------------------------------ V out ( s ) (5)
( R3 + R4 )

By substitution of (4) and (5) into (3) we get

R3 ( 1 + R2 C1 s ) V in ( s ) 1 R3
⎛ -----
1
- + ------ + C 2 s⎞ ------------------------------------ V out ( s ) = --------------
1
- + ------ ----------------------- V out ( s ) + C 2 sV out ( s )
⎝R R2 ⎠ ( R3 + R4 ) R1 R2 ( R3 + R4 )
1
or
⎛ ----- 1- + C s⎞ R
1- + ----- 3 ( 1 + R2 C1 s ) 1 R3 1
- – C 2 s V out ( s ) = ------ V in ( s )
- – ------ ----------------------
-----------------------------------
⎝R R 2 ⎠ ( R3 + R4 ) R 2 ( R 3 + R 4 ) R 1
1 2

and thus
V out ( s ) 1
G ( s ) = ----------------
- = --------------------------------------------------------------------------------------------------------------------------------------------
V in ( s ) R3 ( 1 + R2 C1 s ) 1 R3
R 1 ⎛ ------ + ------ + C 2 s⎞ ------------------------------------ – ------ ----------------------- – C 2 s
1 1
⎝R R2 ⎠ ( R3 + R4 ) R2 ( R3 + R4 )
1

By substitution of the given values and after simplification we get

Signals and Systems with MATLAB Applications, Second Edition 4-31

Orchard Publications
Chapter 4 Circuit Analysis with Laplace Transforms

7
7.83 × 10
G ( s ) = ---------------------------------------------------------------------
-
2 4 7
s + 1.77 × 10 s + 5.87 × 10
w=1:10:10000; s=j.*w; Gs=7.83.*10.^7./(s.^2+1.77.*10.^4.*s+5.87.*10.^7);...
semilogx(w,abs(Gs)); grid; hold on
xlabel('Radian Frequency w'); ylabel('|Vout/Vin|');
title('Magnitude Vout/Vin vs. Radian Frequency')

The plot above indicates that this circuit is a second-order low-pass filter.

4-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 5
State Variables and State Equations

his chapter is an introduction to state variables and state equations as they apply in circuit anal-
T ysis. The state transition matrix is defined, and the state space-to-transfer function equivalence
is presented. Several examples are given to illustrate their application.

5.1 Expressing Differential Equations in State Equation Form

As we know, when we apply KCL or KVL in networks that contain energy-storing devices, we
obtain integro-differential equations. Also, when a network contains just one such device (capacitor
or inductor), it is said to be a first order circuit. If it contains two such devices, it is said to be second-
order circuit, and so on. Thus, a first order linear, time-invariant circuit can be described by a differ-
ential equation of the form
dy
a 1 ------ + a 0 y ( t ) = x ( t ) (5.1)
dt
A second order circuit can be described by a second-order differential equation of the same form as
(5.1) where the highest order is a second derivative.
An nth-order differential equation can be resolved to n first-order simultaneous differential equa-
tions with a set of auxiliary variables called state variables. The resulting first-order differential equa-
tions are called state space equations, or simply state equations. These equations can be obtained
either from the nth-order differential equation, or directly from the network, provided that the state
variables are chosen appropriately. The state variable method offers the advantage that it can also be
used with non-linear and time-varying devices. However, our discussion will be limited to linear,
time-invariant circuits.
State equations can also be solved with numerical methods such as Taylor series and Runge-Kutta
methods, but these will not be discussed in this text*. The state variable method is best illustrated
through several examples presented in this chapter.

Example 5.1
A series RLC circuit with excitation
jωt
vS ( t ) = e (5.2)

* These are discussed in “Numerical Analysis using MATLAB and Spreadsheets” ISBN 0-9709511-1-6.

Signals and Systems with MATLAB Applications, Second Edition 5-1

Orchard Publications
Chapter 5 State Variables and State Equations

is described by the integro-differential equation

t
di 1
∫– ∞ i d t
jωt
Ri + L ----- + ---- = e (5.3)
dt C

Differentiating both sides and dividing by L we get

2
d t R di 1 1 jωt
------- + --- ----- + ------- i = --- jωe (5.4)
dt
2 L dt LC L

or
2
d t R di 1 1 jωt
------- = – --- ----- – ------- i + --- jωe (5.5)
dt
2 L dt LC L

Next, we define two state variables x 1 and x 2 such that

x1 = i (5.6)
and
dx
x 2 = di
----- = --------1 = x· 1 (5.7)
dt dt
Then,
2 2
x· 2 = d i ⁄ dt (5.8)

where x· k denotes the derivative of the state variable x k .

From (5.5) through (5.8), we obtain the state equations

x· 1 = x 2
R 1 1 jωt (5.9)
x· 2 = – --- x 2 – ------- x 1 + --- jωe
L LC L

It is convenient and customary to express the state equations in matrix* form. Thus, we write the
state equations of (5.9) as

x· 1 0 1 x 0
= 1
1
+ u (5.10)
x· 2 – ------
- –R
--- x 2 --1- j ω e jωt
LC L L

* For a review of matrix theory, please refer to Appendix C.

5-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Expressing Differential Equations in State Equation Form

We usually write (5.10) in a compact form as

x· = Ax + bu (5.11)
where

x· 0 1 x1 0
x· = 1 , A = 1 , x = , b= 1 , and u = any input (5.12)
x· 2 – ------
- –R
--- x2 --- j ω e jωt
LC L L

The output y ( t ) is expressed by the state equation

y = Cx + du (5.13)
where C is another matrix, and d is a column vector. Therefore, the state representation of a net-
work can be described by the pair of the of the state space equations

x· = Ax + bu
y = Cx + du (5.14)

The state space equations of (5.14) can be realized with the block diagram of Figure 5.1.

x· x
Σ Σ
+ +
u b ∫ dt C y

+ +

Figure 5.1. Block diagram for the realization of the state equations of (5.14)

We will learn how to solve the matrix equations of (5.14) in the subsequent sections.
Example 5.2
A fourth-order network is described by the differential equation
4 3 2
d y d y d y dy
--------- + a 3 --------3- + a 2 -------2- + a 1 ------ + a 0 y ( t ) = u ( t ) (5.15)
dt
4
dt dt dt

where y ( t ) is the output representing the voltage or current of the network, and u ( t ) is any input.
Express (5.15) as a set of state equations.

Signals and Systems with MATLAB Applications, Second Edition 5-3

Orchard Publications
Chapter 5 State Variables and State Equations

Solution:
The differential equation of (5.15) is of fourth-order; therefore, we must define four state variables
that will be used with the resulting four first-order state equations.
We denote the state variables as x 1, x 2, x 3 , and x 4 , and we relate them to the terms of the given dif-
ferential equation as
2 3
dy d y d y
x1 = y ( t ) x 2 = ------ x 3 = --------- x 4 = --------- (5.16)
dt 2 3
dt dt
We observe that
x· 1 = x 2
x· 2 = x 3
x· 3 = x 4 (5.17)
4
d y
--------- = x· 4 = – a 0 x 1 – a 1 x 2 – a 2 x 3 – a 3 x 4 + u ( t )
4
dt
and in matrix form

x· 1 0 1 0 0 x1 0
x· 2 0 0 1 0 x2
= + 0 u(t) (5.18)
x· 3 0 0 0 1 x3 0
x· 4 –a0 –a1 –a2 –a3 x4 1

In compact form, (5.18) is written as

x· = Ax + bu (5.19)
where
x· 1 0 1 0 0 x1 0
x· 2 0 0 1 0 x2
x· = , A= , x= , b= 0, and u = u ( t )
x· 3 0 0 0 1 x3 0
x· 4 –a0 –a1 –a2 –a3 x4 1

We can also obtain the state equations directly from given circuits. We choose the state variables to
represent inductor currents and capacitor voltages. In other words, we assign state variables to
energy storing devices. The examples that follow illustrate the procedure.

5-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Expressing Differential Equations in State Equation Form

Example 5.3

Write state equation(s) for the circuit of Figure 5.2, given that v C ( 0 − ) = 0 .

R
C +
+ v C ( t ) = v out ( t )
− −
vS u0 ( t )

Figure 5.2. Circuit for Example 5.3

Solution:
This circuit contains only one energy-storing device, the capacitor. Therefore, we need only one
state variable. We choose the state variable to denote the voltage across the capacitor as shown in
Figure 5.3. The output is defined as the voltage across the capacitor.
R

+ vR ( t ) −
C +
+ v C ( t ) = v out ( t ) = x
− −
i
vS u0 ( t )

Figure 5.3. Circuit for Example 5.3 with state variable x assigned to it

For this series circuit,

dv C
i R = i = i C = C --------- = Cx·
dt
and
v R ( t ) = Ri = RCx·
By KVL,
vR ( t ) + vC ( t ) = vS u0 ( t )
or
RCx· + x = v s u 0 ( t )
Therefore, the state equations are
1
x· = – -------- x + v S u 0 ( t )
RC (5.20)
y = x

Signals and Systems with MATLAB Applications, Second Edition 5-5

Orchard Publications
Chapter 5 State Variables and State Equations

Example 5.4

Write state equation(s) for the circuit of Figure 5.4 assuming i L ( 0 − ) = 0 , and the output y is defined
as y = i ( t ) .
R

+ i(t)
`
L
−
vS u0 ( t )

Figure 5.4. Circuit for Example 5.4

Solution:
This circuit contains only one energy-storing device, the inductor; therefore, we need only one state
variable. We choose the state variable to denote the current through the inductor as shown in Figure
5.5.
R

i(t) = x
+
`
L
−
vS u0 ( t )

Figure 5.5. Circuit for Example 5.4 with state variable x assigned to it

By KVL,
vR + vL = vS u0 ( t )
or
di
Ri + L ----- = v S u 0 ( t )
dt
or
Rx + Lx· = v S u 0 ( t )
Therefore, the state equations are
R 1
x· = – --- x + --- v S u 0 ( t )
L L (5.21)
y = x

5-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solution of Single State Equations

5.2 Solution of Single State Equations

If a circuit contains only one energy-storing device, the state equations are written as
x· = α x + β u
(5.22)
y = k1 x + k2 u

where α , β , k 1 , and k 2 are scalar constants, and the initial condition, if non-zero, is denoted as

x0 = x ( t0 ) (5.23)
We will now prove that the solution of the first state equation in (5.22) is
α ( t – t0 ) t
αt – ατ
x( t) = e x0 + e ∫t e
0
β u ( τ ) dτ (5.24)

Proof:
First, we must show that (5.24) satisfies the initial condition of (5.23). This is done by substitution of
t = t 0 in (5.24). Then,

t0
α ( t0 – t0 ) αt –α τ
x ( t0 ) = e x0 + e ∫t 0
e β u ( τ ) dτ (5.25)

The first term in the right side of (5.25) reduces to x 0 since

α ( t0 – t0 ) 0
e x0 = e x0 = x0 (5.26)
The second term of (5.25) is zero since the upper and lower limits of integration are the same.
Therefore, (5.25) reduces to x ( t 0 ) = x 0 and thus the initial condition is satisfied.
Next, we must prove that (5.24) satisfies also the first equation in (5.22). To prove this, we differen-
tiate (5.24) with respect to t and we get

d α ( t – t0 ) d ⎧ αt t
– ατ ⎫
x· ( t ) = ----- ( e
dt
x 0 ) + ----- ⎨ e
dt ⎩ ∫t e 0
β u ( τ ) dτ ⎬
⎭
or
α ( t – t0 ) t
αt – ατ αt – ατ
x· ( t ) = α e x0 + α e ∫t 0
e β u ( τ ) dτ + e [ e βu(τ )] τ = t

α ( t – t0 ) t
αt – ατ αt – αt
= α e x0 + e ∫t 0
e β u ( τ ) dτ + e e βu( t)

or

Signals and Systems with MATLAB Applications, Second Edition 5-7

Orchard Publications
Chapter 5 State Variables and State Equations

α ( t – t0 ) t
α(t – τ)
x· ( t ) = α e x0 + ∫t e
0
β u ( τ ) dτ + β u ( t ) (5.27)

We observe that the bracketed terms of (5.27) are the same as the right side of the assumed solution
of (5.24). Therefore,
x· = α x + β u
and this is the same as the first equation of (5.22).
In summary, if α and β are scalar constants, the solution of
x· = α x + β u (5.28)
with initial condition
x0 = x ( t0 ) (5.29)

is obtained from the relation

α ( t – t0 ) t
αt –α τ
x(t) = e x0 + e ∫t e
0
β u ( τ ) dτ (5.30)

Example 5.5
Use (5.28) through (5.30) to find the capacitor voltage v c ( t ) of the circuit of Figure 5.6 for t > 0 ,
given that the initial condition is v C ( 0 − ) = 1 V

R=2 Ω
+
+
− vC ( t )
C=0.5 F −
2u 0 ( t )

Figure 5.6. Circuit for Example 5.5

Solution:
From (5.20) of Example 5.3,
1
x· = – -------- x + v S u 0 ( t )
RC
and by comparison with (5.28),
1 –1
α = – -------- = ---------------- = – 1
RC 2 × 0.5
and
β = 2

5-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The State Transition Matrix

Then, from (5.30),

α ( t – t0 ) t t
αt –α τ –1 ( t – 0 ) –t τ
x( t) = e x0 + e ∫t 0
e β u ( τ ) dτ = e 1+e ∫0 e 2u ( τ ) dτ
t t
–t –t τ –t –t τ –t –t
∫0 e d τ = e
t
= e + 2e + 2e [ e ] 0
= e + 2e ( e – 1 )

or
–t
v C ( t ) = x ( t ) = ( 2 – e )u 0 ( t ) (5.31)

If we assume that the output y is the capacitor voltage, the output state equation is
–t
y ( t ) = x ( t ) = ( 2 – e )u 0 ( t ) (5.32)

5.3 The State Transition Matrix

In Section 5.1 we defined the state equations pair
x· = Ax + bu
(5.33)
y = Cx + du

where for two or more simultaneous differential equations, A and C are 2 × 2 or higher order
matrices, and b and d are column vectors with two or more rows. In this section we will introduce
At
the state transition matrix e , and we will prove that the solution of the matrix differential equation
x· = Ax + bu (5.34)
with initial conditions
x ( t0 ) = x0 (5.35)

is obtained from the relation

A ( t – t0 ) t
–A τ
∫t e
At
x(t) = e x0 + e bu ( τ ) dτ (5.36)
0

Proof:
Let A be any n × n matrix whose elements are constants. Then, another n × n matrix denoted as
ϕ ( t ) , is said to be the state transition matrix of (5.34), if it is related to the matrix A as the matrix
power series

At 1 22 1 33 1 nn
ϕ(t) ≡ e = I + At + ----- A t + ----- A t + … + ----- A t (5.37)
2! 3! n!

Signals and Systems with MATLAB Applications, Second Edition 5-9

Orchard Publications
Chapter 5 State Variables and State Equations

where I is the n × n identity matrix.

From (5.37), we find that
A0
ϕ(0) = e = I + A0 + … = I (5.38)
Differentiation of (5.37) with respect to t yields
d At 2 2
ϕ' ( t ) = ----- e = 0 + A ⋅ 1 + A t + … = A + A t + … (5.39)
dt
and by comparison with (5.37) we get
d
----- e At = Ae At (5.40)
dt
To prove that (5.36) is the solution of (5.34), we must prove that it satisfies both the initial condition
and the matrix differential equation. The initial condition is satisfied from the relation
t0
A ( t0 – t0 ) At 0 –A τ
∫t
A0
x ( t0 ) = e x0 + e e bu ( τ ) dτ = e x 0 + 0 = Ix 0 = x 0 (5.41)
0

where we have used (5.38) for the initial condition. The integral is zero since the upper and lower
limits of integration are the same.
To prove that (5.34) is also satisfied, we differentiate the assumed solution
A ( t – t0 ) t
–A τ
∫t e
At
x( t) = e x0 + e bu ( τ ) dτ
0

with respect to t and we use (5.40), that is,

d
----- e At = Ae At
dt
Then,
A ( t – t0 ) t
–A τ At – A t
∫t e
At
x· ( t ) = Ae x 0 + Ae bu ( τ ) dτ + e e bu ( t )
0

or
A ( t – t0 ) t
–A τ At – A t
∫t e
At
x· ( t ) = A e x0 + e bu ( τ ) dτ + e e bu ( t ) (5.42)
0

We recognize the bracketed terms in (5.42) as x ( t ) , and the last term as bu ( t ) . Thus, the expression
(5.42) reduces to
x· ( t ) = Ax + bu

5-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of the State Transition Matrix

In summary, if A is an n × n matrix whose elements are constants, n ≥ 2 , and b is a column vector

with n elements, the solution of
x· ( t ) = Ax + bu (5.43)
with initial condition
x0 = x ( t0 ) (5.44)

is

A ( t – t0 ) t
–A τ
∫t e
At
x(t) = e x0 + e bu ( τ ) dτ (5.45)
0

Therefore, the solution of second or higher order circuits using the state variable method, entails the
At
computation of the state transition matrix e , and integration of (5.45).

At
5.4 Computation of the State Transition Matrix e
Let A be an n × n matrix, and I be the n × n identity matrix. By definition, the eigenvalues λ i ,
i = 1, 2, …, n of A are the roots of the nth order polynomial

det [ A – λI ] = 0 (5.46)

We recall that expansion of a determinant produces a polynomial. The roots of the polynomial of
(5.46) can be real (unequal or equal), or complex numbers.
At
Evaluation of the state transition matrix e is based on the Cayley-Hamilton theorem. This theorem
states that a matrix can be expressed as an ( n – 1 )th degree polynomial in terms of the matrix A as
At 2 n–1
e = a0 I + a1 A + a2 A + … + an – 1 A (5.47)

where the coefficients a i are functions of the eigenvalues λ .

We accept (5.47) without proving it. The proof can be found in Linear Algebra and Matrix Theory
textbooks.
Since the coefficients a i are functions of the eigenvalues λ , we must consider the following cases:

Case I: Distinct Eigenvalues (Real or Complex)

If λ 1 ≠ λ 2 ≠ λ 3 ≠ … ≠ λ n , that is, if all eigenvalues of a given matrix A are distinct, the coefficients a i
are found from the simultaneous solution of the following system of equations:

Signals and Systems with MATLAB Applications, Second Edition 5-11

Orchard Publications
Chapter 5 State Variables and State Equations

2 n–1 λ1 t
a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 = e
2 n–1 λ2 t
a0 + a1 λ2 + a2 λ2 + … + an – 1 λ2 = e
(5.48)
…
2 n–1 λn t
a0 + a1 λn + a2 λn + … + an – 1 λn = e

Example 5.6

given that A = – 2 1
At
Compute the state transition matrix e
0 –1

Solution:
We must first find the eigenvalues λ of the given matrix A . These are found from the expansion of
det [ A – λI ] = 0
For this example,

⎧ ⎫
det [ A – λI ] = det ⎨ – 2 1 – λ 1 0 ⎬ = det – 2 – λ 1 = 0
⎩ 0 –1 0 1 ⎭ 0 –1–λ
= (– 2 – λ)(– 1 – λ) = 0
or
(λ + 1 )( λ + 2) = 0
Therefore,
λ 1 = – 1 and λ 2 = – 2 (5.49)

Next, we must find the coefficients a i of (5.47). Since A is a 2 × 2 matrix, we only need to consider
the first two terms of that relation, that is,
At
e = a0 I + a1 A (5.50)

The coefficients a 0 and a 1 are found from (5.48). For this example,
λ1 t
a0 + a1 λ1 = e
λ2 t
a0 + a1 λ2 = e
or
–t
a0 + a1 ( –1 ) = e
(5.51)
– 2t
a0 + a1 ( –2 ) = e

5-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of the State Transition Matrix

Simultaneous solution of (5.51) yields

–t – 2t
a 0 = 2e – e
(5.52)
–t – 2t
a1 = e – e

and by substitution into (5.50),

) –2 1
At –t – 2t 1 0 –t – 2t
e = ( 2e – e ) + (e – e
0 1 0 –1
or

– 2t –t – 2t
e
At
= e e –e (5.53)
–t
0 e

At
In summary, we compute the state transition matrix e for a given matrix A using the following
procedure:
1. We find the eigenvalues λ from det [ A – λI ] = 0 . We can write [ A – λI ] at once by subtracting
λ from each of the main diagonal elements of A . If the dimension of A is a 2 × 2 matrix, it will
yield two eigenvalues; if it is a 3 × 3 matrix, it will yield three eigenvalues, and so on. If the eigen-
values are distinct, we perform steps 2 through 4; otherwise we refer to Case II below.
2. If the dimension of A is a 2 × 2 matrix, we use only the first 2 terms of the right side of the state
transition matrix
At 2 n–1
e = a0 I + a1 A + a 2 A + … + an – 1 A (5.54)

If A matrix is a 3 × 3 matrix, we use the first 3 terms, and so on.

3. We obtain the a i coefficients from

2 n–1 λ1 t
a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 = e
2 n–1 λ2 t
a0 + a1 λ2 + a2 λ2 + … + an – 1 λ2 = e
…
2 n–1 λn t
a0 + a1 λn + a2 λn + … + an – 1 λn = e

We use as many equations as the number of the eigenvalues, and we solve for the coefficients a i .

4. We substitute the a i coefficients into the state transition matrix of (5.54), and we simplify.

Signals and Systems with MATLAB Applications, Second Edition 5-13

Orchard Publications
Chapter 5 State Variables and State Equations

Example 5.7
At
Compute the state transition matrix e given that

5 7 –5
A = 0 4 –1 (5.55)
2 8 –3

Solution:
1. We first compute the eigenvalues from det [ A – λI ] = 0 . We obtain [ A – λI ] at once, by subtract-
ing λ from each of the main diagonal elements of A . Then,

5–λ 7 –5
det [ A – λI ] = det 0 4–λ –1 = 0 (5.56)
2 8 –3–λ

and expansion of this determinant yields the polynomial

3 2
λ – 6λ + 11λ – 6 = 0 (5.57)
We will use MATLAB roots(p) function to obtain the roots of (5.57).
p=[1 −6 11 −6]; r=roots(p); fprintf(' \n'); fprintf('lambda1 = %5.2f \t', r(1));...
fprintf('lambda2 = %5.2f \t', r(2)); fprintf('lambda3 = %5.2f', r(3))
lambda1 = 3.00 lambda2 = 2.00 lambda3 = 1.00
and thus the eigenvalues are
λ1 = 1 λ2 = 2 λ3 = 3 (5.58)

2. Since A is a 3 × 3 matrix, we need to use the first 3 terms of (5.54), that is,
At 2
e = a0 I + a1 A + a2 A (5.59)

3. We obtain the coefficients a 0, a 1, and a 2 from

2 λ1 t
a0 + a1 λ1 + a2 λ1 = e
2 λ2 t
a0 + a1 λ2 + a2 λ2 = e
2 λ3 t
a0 + a1 λ3 + a2 λ3 = e
or

5-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of the State Transition Matrix
t
a0 + a1 + a2 = e
2t
a 0 + 2a 1 + 4a 2 = e (5.60)
3t
a 0 + 3a 1 + 9a 2 = e

We will use the following MATLAB code for the solution of (5.60).
B=sym('[1 1 1; 1 2 4; 1 3 9]'); b=sym('[exp(t); exp(2*t); exp(3*t)]'); a=B\b; fprintf(' \n');...
disp('a0 = '); disp(a(1)); disp('a1 = '); disp(a(2)); disp('a2 = '); disp(a(3))
a0 =
3*exp(t)-3*exp(2*t)+exp(3*t)
a1 =
-5/2*exp(t)+4*exp(2*t)-3/2*exp(3*t)
a2 =
1/2*exp(t)-exp(2*t)+1/2*exp(3*t)
Thus,
t 2t 3t
a 0 = 3e – 3e + e
5 t 2t 3 3t
a 1 = – --- e + 4e – --- e (5.61)
2 2
1 t 2t 1 3t
a 2 = --- e – e + --- e
2 2
4. We also use MATLAB to perform the substitution into the state transition matrix, and to per-
form the matrix multiplications. The code is shown below.
syms t; a0 = 3*exp(t)+exp(3*t)-3*exp(2*t); a1 = -5/2*exp(t)-3/2*exp(3*t)+4*exp(2*t);...
a2 = 1/2*exp(t)+1/2*exp(3*t)-exp(2*t);...
A = [5 7 −5; 0 4 −1; 2 8 −3]; eAt=a0*eye(3)+a1*A+a2*A^2
eAt =
[ -2*exp(t)+2*exp(2*t)+exp(3*t), -6*exp(t)+5*exp(2*t)+exp(3*t), 4*exp(t)-3*exp(2*t)-exp(3*t)]
[ -exp(t)+2*exp(2*t)-exp(3*t), -3*exp(t)+5*exp(2*t)-exp(3*t), 2*exp(t)-3*exp(2*t)+exp(3*t)]
[ -3*exp(t)+4*exp(2*t)-exp(3*t), -9*exp(t)+10*exp(2*t)-exp(3*t), 6*exp(t)-6*exp(2*t)+exp(3*t)]

Thus,
t 2t 3t t 2t 3t t 2t 3t
– 2e + 2e + e – 6 e + 5e + e 4e – 3e – e
At
e = t 2t
– e + 2e – e
3t t
– 3e + 5e – e
2t 3t t 2t
2e – 3e + e
3t

t 2t 3t t 2t 3t t 2t 3t
– 3e + 4e – e – 9e + 10e – e 6e – 6e + e

Case II: Multiple Eigenvalues

In this case, we will assume that the polynomial of
det [ A – λI ] = 0 (5.62)

Signals and Systems with MATLAB Applications, Second Edition 5-15

Orchard Publications
Chapter 5 State Variables and State Equations

has n roots, and m of these roots are equal. In other words, the roots are
λ1 = λ2 = λ3 … = λm , λm + 1 , λn (5.63)

The coefficients a i of the state transition matrix

At n–1
= a0 I + a1 A + a2 A + … + an – 1 A
2
e (5.64)

are found from the simultaneous solution of the system of equations of (5.65) below.

n–1 λ1 t
a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1
2
= e
d- n–1 d λ1 t
( a 0 + a 1 λ 1 + a 2 λ 1 + … + a n – 1 λ 1 ) = -------- e
2
--------
dλ 1 dλ 1
2 2
d d λt
--------2 ( a 0 + a 1 λ 1 + a 2 λ 21 + … + a n – 1 λ n1 – 1 ) = --------2 e 1
dλ 1 dλ 1
… (5.65)
m–1 m–1
d d λ t
- ( a 0 + a 1 λ 1 + a 2 λ 21 + … + a n – 1 λ n1 – 1 ) = --------------
--------------
m–1 m–1
-e 1
dλ 1 dλ 1
n–1 λ m + 1t
a0 + a1 λm + 1 + a2 λm + 1 + … + an – 1 λm + 1 = e
2

…
n–1 λn t
a 0 + a 1 λn + a 2 λ n + … + a n – 1 λ n
2
= e

Example 5.8
At
Compute the state transition matrix e given that

A = –1 0
2 –1
Solution:
1. We first find the eigenvalues λ of the matrix A and these are found from the polynomial of
det [ A – λI ] = 0 . For this example,

det [ A – λI ] = det – 1 – λ 0 = 0
2 –1–λ
= (– 1 – λ)(– 1 – λ) = 0
2
= (λ + 1) = 0

5-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of the State Transition Matrix

and thus,
λ1 = λ2 = –1

2. Since A is a 2 × 2 matrix, we only need the first two terms of the state transition matrix, that is,
At
e = a0 I + a1 A (5.66)

3. We find a 0 and a 1 from (5.65). For this example,

λ1 t
a0 + a1 λ1 = e
d d λt
--------- ( a 0 + a 1 λ 1 ) = --------- e 1
dλ 1 dλ 1
or
λ1 t
a0 + a1 λ1 = e
λ1 t
a 1 = te

and by substitution with λ 1 = λ 2 = – 1 , we get

–t
a0 – a1 = e
–t
a 1 = te

Simultaneous solution of the last two equations yields

–t –t
a 0 = e + te
(5.67)
–t
a 1 = te

4. By substitution of (5.67) into (5.66), we get

e
At –t –t
= ( e + te ) 1 0 + te – t – 1 0
0 1 2 –1
or

–t
e
At
= e 0 (5.68)
–t –t
2te e

We can use the MATLAB eig(x) function to find the eigenvalues of an n × n matrix. To find out
how it is used, we invoke the help eig command.

Signals and Systems with MATLAB Applications, Second Edition 5-17

Orchard Publications
Chapter 5 State Variables and State Equations

We will first use MATLAB to verify the values of the eigenvalues found in Examples 5.6 through 5.8,
and we will briefly discuss eigenvectors on the next section.
For Example 5.6
A= [−2 1; 0 −1]; lambda=eig(A)
lambda =
-2
-1
For Example 5.7
B = [5 7 −5; 0 4 −1; 2 8 −3]; lambda=eig(B)
lambda =
1.0000
3.0000
2.0000
For Example 5.8
C = [−1 0; 2 −1]; lambda=eig(C)
lambda =
-1
-1

5.5 Eigenvectors
Consider the relation
AX = λX (5.69)
where A is an n × n matrix, X is a column vector, and λ is a scalar number. We can express this rela-
tion in matrix form as

a 11 a 12 … a 1n x 1 x1
a 21 a 22 … a 2n x 2 x2
= λ (5.70)
… … … … … …
a n1 a n2 … a nn x n xn

We write (5.70) as
( A – λI )X = 0 (5.71)
Then, (5.71) can be written as

5-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Eigenvectors

( a 11 – λ )x 1 a 12 x 2 … a1 n xn
a 21 x 1 ( a 22 – λ )x 2 … a2 n xn
= 0 (5.72)
… … … …
an 1 x1 an 2 x2 … ( a nn – λ )x n

The equations of (5.72) will have non-trivial solutions if and only if its determinant is zero*, that is, if

( a 11 – λ ) a 12 … a1 n
a 21 ( a 22 – λ ) … a2 n
det = 0 (5.73)
… … … …
an 1 an2 … ( a nn – λ )

Expansion of the determinant of (5.73) results in a polynomial equation of degree n in λ , and it is

called the characteristic equation.
We can express (5.73) in a compact form as
det ( A – λI ) = 0 (5.74)
As we know, the roots λ of the characteristic equation are the eigenvalues of the matrix A , and cor-
responding to each eigenvalue λ , there is a non-trivial solution of the column vector X , i.e., X ≠ 0 .
This vector X is called eigenvector. Obviously, there is a different eigenvector for each eigenvalue.
Eigenvectors are generally expressed as unit eigenvectors, that is, they are normalized to unit length.
This is done by dividing each component of the eigenvector by the square root of the sum of the
squares of their components, so that the sum of the squares of their components is equal to unity.
T T
In many engineering applications the unit eigenvectors are chosen such that X ⋅ X = I where X is
the transpose of the eigenvector X , and I is the identity matrix.
Two vectors X and Y are said to be orthogonal if their inner (dot) product is zero. A set of eigenvec-
tors constitutes an orthonormal basis if the set is normalized (expressed as unit eigenvectors) and
these vector are mutually orthogonal. An orthonormal basis can be formed with the Gram-Schmidt
Orthogonalization Procedure; it is beyond the scope of this chapter to discuss this procedure, and
therefore it will not be discussed in this text. It can be found in Linear Algebra and Matrix Theory
textbooks.
The example which follows, illustrates the relationships between a matrix A , its eigenvalues, and
eigenvectors.

* This is because we want the vector X in (5.71) to be a non-zero vector and the product ( A – λI )X to be zero.

Signals and Systems with MATLAB Applications, Second Edition 5-19

Orchard Publications
Chapter 5 State Variables and State Equations

Example 5.9
Given the matrix

5 7 –5
A = 0 4 –1
2 8 –3

a. Find the eigenvalues of A

b. Find eigenvectors corresponding to each eigenvalue of A
c. Form a set of unit eigenvectors using the eigenvectors of part (b).
Solution:
a. This is the same matrix as in Example 5.7, where we found the eigenvalues to be
λ1 = 1 λ2 = 2 λ3 = 3
b. We start with
AX = λX
and we let
x1
X = x2
x3
Then,

5 7 –5 x1 x1
0 4 –1 x2 = λ x2 (5.75)
2 8 –3 x3 x3

or

5x 1 7x 2 – 5x 3 λx 1
0 4x 2 –x3 = λx 2 (5.76)
2x 1 8x 2 – 3x 3 λx 3

Equating corresponding rows and rearranging, we get

5-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Eigenvectors

( 5 – λ )x 1 7x 2 – 5x 3 0
0 ( 4 – λ )x 2 –x3 = 0 (5.77)
2x 1 8x 2 – ( 3 – λ )x 3 0

For λ = 1 , (5.77) reduces to

4x 1 + 7x 2 – 5x 3 = 0
3x 2 – x 3 = 0 (5.78)
2x 1 + 8x 2 – 4x 3 = 0

By Crame’s rule, or MATLAB, we get the indeterminate values

x1 = 0 ⁄ 0 x2 = 0 ⁄ 0 x3 = 0 ⁄ 0 (5.79)

Since the unknowns x 1, x 2, and x 3 are scalars, we can assume that one of these, say x 2 , is known,
and solve x 1 and x 3 in terms of x 2 . Then, we get x 1 = 2x2 , and x 3 = 3x 2 .

Therefore, an eigenvector for λ = 1 is

x1 2x 2 2 2
Xλ = 1 = x2 = x2 = x2 1 = 1 (5.80)
x3 3x 2 3 3

since any eigenvector is a scalar multiple of the last vector in (5.80).

Similarly, for λ = 2 , we get x 1 = x 2 , and x 3 = 2x2 . Then, an eigenvector for λ = 2 is

x1 x2 1 1
Xλ = 2 = x2 = x2 = x2 1 = 1 (5.81)
x3 2x 2 2 2

Finally, for λ = 3 , we get x 1 = – x 2 , and x 3 = x 2 . Then, an eigenvector for λ = 3 is

x1 –x2 –1 –1
Xλ = 3 = x2 = x2 = x2 1 = 1 (5.82)
x3 x2 1 1

c. We find the unit eigenvectors by dividing the components of each vector by the square root of
the sum of the squares of the components. These are:

Signals and Systems with MATLAB Applications, Second Edition 5-21

Orchard Publications
Chapter 5 State Variables and State Equations

2 2 2
2 +1 +3 = 14

2 2 2
1 +1 +2 = 6

2 2 2
( –1 ) + 1 + 1 = 3
The unit eigenvectors are

2 1 –1
---------- ------- -------
14 6 3
1
Unit X λ = 1 = ---------- 1
Unit X λ = 2 = ------- 1
Unit X λ = 3 = ------- (5.83)
14 6 3
3 2 1
---------- ------- -------
14 6 3

We observe that for the first unit eigenvector the sum of the squares is unity, that is,

⎛ ---------
2 - ⎞ 2 ⎛ ---------
+
1 - ⎞ 2 ⎛ ---------
+
3 - ⎞2 4- + -----
= ----- 9- = 1
1- + ----- (5.84)
⎝ 14 ⎠ ⎝ 14 ⎠ ⎝ 14 ⎠ 14 14 14

and the same is true for the other two unit eigenvectors in (5.83).

5.6 Circuit Analysis with State Variables

In this section, we will present two examples to illustrate how the state variable method is used in cir-
cuit analysis.

Example 5.10
For the circuit of Figure 5.7, the initial conditions are i L ( 0 − ) = 0 , and v c ( 0 − ) = 0.5 V . Use the state
variable method to compute i L ( t ) and v c ( t ) .

R L
`

+
1Ω 1⁄4 H
C
+ vC ( t )
− i(t)
vs ( t ) = u0 ( t ) 4⁄3 F
−
Figure 5.7. Circuit for Example 5.10
Solution:
For this example,
i = iL

5-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Analysis with State Variables

and
di
Ri L + L ------L- + v C = u 0 ( t )
dt
Substitution of given values and rearranging, yields
1 di L
--- ------- = ( – 1 )i L – v C + 1
4 dt
or
di
------L- = – 4i L – 4v C + 4 (5.85)
dt

Next, we define the state variables x 1 = i L and x 2 = v C . Then,

di
x· 1 = ------L- (5.86)
dt
and
dv
x· 2 = -------C-
dt
Also,
dv
i L = C -------C-
dt
and thus,
dv 4
x 1 = i L = C -------C- = Cx· 2 = --- x· 2
dt 3
or
3
x· 2 = --- x 1 (5.87)
4
Therefore, from (5.85), (5.86), and (5.87), we get the state equations
x· 1 = – 4x 1 – 4x 2 + 4
3
x· 2 = --- x 1
4
and in matrix form,

x· 1
= –4 –4 1 + 4 u0 ( t )
x
(5.88)
x· 2 3 ⁄ 4 0 x2 0

We will compute the solution of (5.88) using

Signals and Systems with MATLAB Applications, Second Edition 5-23

Orchard Publications
Chapter 5 State Variables and State Equations

A ( t – t0 ) t
–A τ
∫t e
At
x( t) = e x0 + e bu ( τ ) dτ (5.89)
0

where

–4 –4 iL ( 0 ) 0
A = x0 = = b = 4 (5.90)
3⁄4 0 vC ( 0 ) 1⁄2 0

At
First, we compute the state transition matrix e . We find the eigenvalues from
det [ A – λI ] = 0
Then,

det [ A – λI ] = det – 4 – λ – 4 = 0
3 ⁄ 4 –λ
= ( –λ ) ( – 4 – λ ) + 3 = 0
2
= λ + 4λ + 3 = 0
Therefore,
λ 1 = – 1 and λ 2 = – 3

The next step is to find the coefficients a i . Since A is a 2 × 2 matrix, we only need the first two
terms of the state transition matrix, that is,
At
e = a0 I + a1 A (5.91)

The constants a 0 and a 1 are found from

λ1 t
a0 + a1 λ1 = e
λ2 t
a0 + a1 λ2 = e

and with λ 1 = – 1 and λ 2 = – 3 , we get

–t
a0 –a1 = e
(5.92)
– 3t
a 0 – 3a 1 = e

Simultaneous solution of (5.92) yields

–t – 3t
a 0 = 1.5e – 0.5e
(5.93)
–t – 3t
a 1 = 0.5e – 0.5e

We now substitute these values into (5.91), and we get

5-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Analysis with State Variables

e
At –t
= ( 1.5e – 0.5e
– 3t
) 1 0 + ( 0.5e – t – 0.5e –2t ) – 4 – 4
0 1 3⁄4 0

–t – 3t –t – 3t
–t – 3t – 2 e + 2e – 2 e + 2e
= 1.5e – 0.5e 0 + – 3t
–t – 3t 3
--- e –t – 3
--- e
0 1.5e – 0.5e 8 8
0

or
–t – 3t –t – 3t
At – 0.5 e + 1.5e – 2 e + 2e
e = – 3t
3
--- e – t – 3 –t – 3t
--- e 1.5e – 0.5e
8 8

The initial conditions vector is the second vector in (5.90); then, the first term of (5.89) becomes

–t – 3t –t – 3t
At – 0.5 e + 1.5e – 2 e + 2e 0
e x0 = – 3t
3
--- e – t – 3 –t – 3t
1⁄2
--- e 1.5e – 0.5e
8 8
or

–t – 3t
At
e x0 = –e +e (5.94)
–t – 3t
0.75e – 0.25e

We also need to evaluate the integral on the right side of (5.89). From (5.90)

b = 4 = 1 4
0 0

and denoting this integral as Int , we have

–( t – τ ) –3 ( t – τ ) –( t – τ ) –3 ( t – τ )
t – 0.5 e + 1.5e –2 e + 2e
∫t
Int = 1 4 dτ
3 –( t – τ ) 3 –3 ( t – τ ) –( t – τ ) –3 ( t – τ )
0 --- e – --- e 1.5e – 0.5e 0
8 8

or

–( t – τ ) –3 ( t – τ )
t – 0.5 e + 1.5e
Int = ∫t 0 --- e –( t – τ ) – 3
3 --- e
–3 ( t – τ ) 4 dτ (5.95)
8 8

Signals and Systems with MATLAB Applications, Second Edition 5-25

Orchard Publications
Chapter 5 State Variables and State Equations

The integration in (5.95) is with respect to τ ; then, integrating the column vector under the integral,
we get
t
–( t – τ ) –3 ( t – τ )
Int = 4 – 0.5 e + 0.5e
–( t – τ ) –3 ( t – τ )
0.375e – 0.125e
τ=0
or
–t – 3t –t – 3t
Int = 4 – 0.5 + 0.5 – 4 – 0.5 e + 0.5e = 4 0.5e – 0.5 e
0.375 – 0.125 – t
0.375e – 0.125e
– 3t –t
0.25 – 0.375 e + 0.125e
– 3t

By substitution of these values, the solution of

A ( t – t0 ) t
–A τ
∫t e
At
x( t) = e x0 + e bu ( τ ) dτ
0

is
–t – 3t –t – 3t –t – 3t
x1
= –e +e +4 0.5e – 0.5 e = e –e
x2 –t – 3t –t – 3t –t – 3t
0.75e – 0.25e 0.25 – 0.375 e + 0.125e 1 – 0.75 e + 0.25e
Then,
–t – 3t
x1 = iL = e –e (5.96)
and
–t – 3t
x 2 = v C = 1 – 0.75e + 0.25e (5.97)

Other variables of the circuit can now be computed from (5.96) and (5.97). For example, the voltage
across the inductor is
di L 1 d – t –3t 1 – t 3 – 3t
v L = L ------- = --- ----- ( e – e ) = – --- e + --- e
dt 4 dt 4 4

Example 5.11
A circuit is described by the state equation
x· = Ax + bu (5.98)
where

A = 1 0 x0 = 1 b = –1 and u = δ ( t ) (5.99)
1 –1 0 1

5-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Analysis with State Variables

Compute the state vector

x1
x =
x2

Solution:
We compute the eigenvalues from
det [ A – λI ] = 0
For this example,

det [ A – λI ] = det 1 – λ 0 = 0
1 –1 –λ
= ( 1 –λ ) ( – 1 – λ ) = 0
Then,
λ 1 = 1 and λ 2 = – 1

Since A is a 2 × 2 matrix, we only need the first two terms of the state transition matrix to find the
coefficients a i , that is,
At
e = a0 I + a1 A (5.100)

The constants a 0 and a 1 are found from

λ1 t
a0 + a1 λ1 = e
(5.101)
λ2 t
a0 + a1 λ2 = e

and with λ 1 = 1 and λ 2 = – 1 , we get

t
a0 + a1 = e
(5.102)
–t
a0 –a1 = e

and simultaneous solution of (5.102) yields

t –t
e +e
a 0 = ---------------- = cosh t
2
t –t
e –e
a 1 = ---------------- = sinh t
2
By substitution of these values into (5.100), we get

Signals and Systems with MATLAB Applications, Second Edition 5-27

Orchard Publications
Chapter 5 State Variables and State Equations

= cosh t I + sinh t A = cosh t 1 0 + sinh t 1 0 = cosh t + sinh t 0

At
e (5.103)
0 1 1 –1 sinh t cosh t – sinh t

The values of the vector x are found from

A ( t – t0 ) t t
–A τ –A τ
∫t ∫0 e
At At At
x( t) = e x0 + e e bu ( τ ) dτ = e x 0 + e bδ ( τ ) dτ (5.104)
0

Using the sifting property of the delta function we find that (5.104) reduces to

At ⎧ ⎫
x ( t ) = e x0 + e b = e ( x0 + b ) = e ⎨ 1 + –1 ⎬ = e 0
At At At At

⎩ 0 1 ⎭ 1

= cosh t + sinh t 0 0 = x1
sinh t cosh t – sinh t 1 x2
Therefore,

x1 0 0
x = = = (5.105)
–t
x2 cosh t – sinh t e

5.7 Relationship between State Equations and Laplace Transform

In this section, we will show that the state transition matrix can be computed from the Inverse
Laplace transform. We will also show that the transfer function can be found from the coefficient
matrices of the state equations.
Consider the state equation
x· = Ax + bu (5.106)
Taking the Laplace of both sides of (5.106), we get
sX ( s ) – x ( 0 ) = AX ( s ) + bU ( s )
or
( sI – A )X ( s ) = x ( 0 ) + bU ( s ) (5.107)

Multiplying both sides of (5.107) by ( sI – A ) –1 , we get

–1 –1
X ( s ) = ( sI – A ) x ( 0 ) + ( sI – A ) bU ( s ) (5.108)
Comparing (5.108) with

5-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Relationship between State Equations and Laplace Transform
t
–A τ
∫0 e
At At
x ( t ) = e x0 + e bu ( τ ) dτ (5.109)

we observe that the right side of (5.108) is the Laplace transform of (5.109). Therefore, we can com-
from the Inverse Laplace of ( sI – A ) –1 , that is, we can use the
At
pute the state transition matrix e
relation

At –1 –1
e = L { ( sI – A ) } (5.110)
Next, we consider the output state equation
y = Cx + du (5.111)
Taking the Laplace of both sides of (5.111), we get
Y ( s ) = CX ( s ) + dU ( s ) (5.112)
and using (5.108), we get
–1 –1
Y ( s ) = C ( sI – A ) x ( 0 ) + [ C ( sI – A ) b + d ]U ( s ) (5.113)
If the initial condition x ( 0 ) = 0 , (5.113) reduces to
–1
Y ( s ) = [ C ( sI – A ) b + d ]U ( s ) (5.114)
In (5.114), U ( s ) is the Laplace transform of the input u ( t ) ; then, division of both sides by U ( s )
yields the transfer function

Y( s) –1
G ( s ) = ----------- = C ( sI – A ) b + d (5.115)
U(s)

Example 5.12
At
In the circuit of Figure 5.8, all initial conditions are zero. Compute the state transition matrix e
using the Inverse Laplace transform method.

R L
`

+
3Ω 1H
C
+ vC ( t )
− i( t)
vs ( t ) = u0 ( t ) 1⁄2 F
−

Figure 5.8. Circuit for Example 5.12

Signals and Systems with MATLAB Applications, Second Edition 5-29

Orchard Publications
Chapter 5 State Variables and State Equations

Solution:
For this circuit,
i = iL
and
di
Ri L + L ------L- + v C = u 0 ( t )
dt
Substitution of given values and rearranging,
di L
------- = – 3 i L – v C + 1 (5.116)
dt
Now, we define the state variables
x1 = iL
and
x2 = vC
Then,
di
x· 1 = ------L- = – 3 i L – v C + 1 (5.117)
dt
and
dv
x· 2 = --------C
dt
Also,
dv dv
i L = C --------C = 0.5 -------C- (5.118)
dt dt
and thus,
dv
x 1 = i L = 0.5 --------C = 0.5x· 2
dt
or
x· 2 = 2x 1 (5.119)
Therefore, from (5.117) and (5.119) we get the state equations
x· 1 = – 3x 1 – x 2 + 1
(5.120)
x· 2 = 2x 1

and in matrix form,

x· 1
= –3 –1 1 + 1 1
x
(5.121)
·x 2 2 0 x2 0

5-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Relationship between State Equations and Laplace Transform

By inspection,

A = –3 –1 (5.122)
2 0

Now, we will find the state transition matrix from

At –1 –1
e = L { ( sI – A ) } (5.123)
where

( sI – A ) = s 0 – –3 –1 = s + 3 1
0 s 2 0 –2 s
Then,
s –1
--------------------------------- ---------------------------------
–1 adj ( sI – A ) 1
- s –1 = ( s + 1 ) ( s + 2 ) ( s + 1 )( s + 2)
( sI – A ) = --------------------------- = -------------------------
det ( sI – A ) 2
s + 3s + 2 2 s+3 2 s+3
--------------------------------- ---------------------------------
( s + 1 )( s + 2) ( s + 1 )( s + 2)

We find the Inverse Laplace of each term by partial fraction expansion. Then,

–t – 2t –t – 2t
{ ( sI – A ) } = – e + 2e –e +e
At –1 –1
e = L
–t – 2t –t – 2t
2e – 2e 2e – e

Now, we can find the state variables representing the inductor current and the capacitor voltage
from
t
–A τ
∫0 e
At At
x ( t ) = e x0 + e bu ( τ ) dτ

using the procedure of Example 5.11.

MATLAB provides two very useful functions to convert state space (state equations), to transfer
function (s-domain), and vice versa. The function ss2tf (state space to transfer function) converts
the state space equations
x· = Ax + Bu *
(5.124)
y = Cx + Du
to the rational transfer function form
N(s)
G ( s ) = ----------- (5.125)
D(s)

* We have used capital letters for vectors b and c to be consistent with MATLAB’s designations.

Signals and Systems with MATLAB Applications, Second Edition 5-31

Orchard Publications
Chapter 5 State Variables and State Equations

This is used with the statement [num,den]=ss2tf(A,B,C,D,iu) where A, B, C, D are the matrices of
(5.124) and iu is 1 if there is only one input. The MATLAB help command provides the following
information:
help ss2tf
SS2TF State-space to transfer function conversion.
[NUM,DEN] = SS2TF(A,B,C,D,iu) calculates the
transfer function:

NUM(s) -1
G(s) = -------- = C(sI-A) B + D
DEN(s)
of the system:
x = Ax + Bu
y = Cx + Du
from the iu'th input. Vector DEN contains the coefficients of
the denominator in descending powers of s. The numerator coeffi-
cients are returned in matrix NUM with as many rows as there
are outputs y.
See also TF2SS
The other function, tf2ss, converts the transfer function of (5.125) to the state-space equations of
(5.124). It is used with the statement [A,B,C,D]=tf2ss(num,den) where A, B, C, and D are the
matrices of (5.124), and num, den are N ( s ) and D ( s ) of (5.125) respectively. The MATLAB help
command provides the following information:
help tf2ss
TF2SS Transfer function to state-space conversion.
[A,B,C,D] = TF2SS(NUM,DEN) calculates the state-space
representation:
x = Ax + Bu
y = Cx + Du
of the system:
NUM(s)
G(s) = --------
DEN(s)
from a single input. Vector DEN must contain the coefficients of
the denominator in descending powers of s. Matrix NUM must con-

5-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Relationship between State Equations and Laplace Transform

tain the numerator coefficients with as many rows as there are

outputs y. The A,B,C,D matrices are returned in controller canon-
ical form. This calculation also works for discrete systems. To
avoid confusion when using this function with discrete systems,
always use a numerator polynomial that has been padded with zeros
to make it the same length as the denominator. See the User's
guide for more details.
See also SS2TF.

Example 5.13
For the circuit of Figure 5.9,

R L

`
+
1Ω 1H
C
+
− v C ( t ) = v out ( t )
i(t)
vs ( t ) = u0 ( t ) 1F
−

Figure 5.9. Circuit for Example 5.13

a. Derive the state equations and express them in matrix form as

x· = Ax + Bu
y = Cx + Du
b. Derive the transfer function
N(s)
G ( s ) = -----------
D(s)
c. Verify your answers with MATLAB.
Solution:
a. The differential equation describing the circuit is
di
Ri + L ----- + v C = u 0 ( t )
dt
and with the given values,
i + di
----- + v C = u 0 ( t )
dt
or
di
----- = – i – v C + u 0 ( t )
dt

Signals and Systems with MATLAB Applications, Second Edition 5-33

Orchard Publications
Chapter 5 State Variables and State Equations

We let
x1 = iL = i
and
x 2 = v C = v out
Then,
di
x· 1 = -----
dt
and
dv
x· 2 = --------c = x 1
dt
Thus, the state equations are
x· 1 = – x 1 – x 2 + u 0 ( t )
x· 2 = x 1
y = x2
and in matrix form,

x·
x· = Ax + Bu ↔ 1 = – 1 –1 x1 + 1 u ( t )
0
x· 2 1 0 x2 0
(5.126)
x1
y = Cx + Du ↔ y = 0 1 + 0 u0 ( t )
x2

b. The s – domain circuit is

R L
`

s +
1Ω C
+
− V C ( s ) = V out ( s )
V in ( s ) 1⁄s
−

Figure 5.10. Transformed circuit for Example 5.13

By the voltage division expression,

1⁄s
V out ( s ) = --------------------------- V in ( s )
1+s+1⁄s
or

5-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Relationship between State Equations and Laplace Transform

V out ( s ) 1
- = ---------------------
---------------- -
V in ( s ) 2
s +s+1
Therefore,
V out ( s ) 1
G ( s ) = ----------------
- = ---------------------
- (5.127)
V in ( s ) 2
s +s+1
c.
A = [−1 −1; 1 0]; B = [1 0]'; C = [0 1]; D = [0];% The matrices of (5.126)
[num, den] = ss2tf(A, B, C, D, 1) % Verify coefficients of G(s) in (5.127)
num =
0 0 1
den =
1.0000 1.0000 1.0000
num = [0 0 1]; den = [1 1 1]; % The coefficients of G(s) in (5.127)
[A B C D] = tf2ss(num, den) % Verify the matrices of (5.126)
A =
-1 -1
1 0
B =
1
0
C =
0 1
D =
0

Signals and Systems with MATLAB Applications, Second Edition 5-35

Orchard Publications
Chapter 5 State Variables and State Equations

5.8 Summary
• An nth-order differential equation can be resolved to n first-order simultaneous differential equa-
tions with a set of auxiliary variables called state variables. The resulting first-order differential
equations are called state space equations, or simply state equations.
• The state space equations can be obtained either from the nth-order differential equation, or
directly from the network, provided that the state variables are chosen appropriately.
• When we obtain the state equations directly from given circuits, we choose the state variables to
represent inductor currents and capacitor voltages.
• The state variable method offers the advantage that it can also be used with non-linear and time-
varying devices.
• If a circuit contains only one energy-storing device, the state equations are written as

x· = α x + β u
y = k1 x + k2 u

where α , β , k 1 , and k 2 are scalar constants, and the initial condition, if non-zero, is denoted as

x0 = x ( t0 ) (5.128)

• If α and β are scalar constants, the solution of x· = α x + β u with initial condition x 0 = x ( t 0 ) is

obtained from the relation
α ( t – t0 ) t
αt –α τ
x( t) = e x0 + e ∫t e
0
β u ( τ ) dτ

• The solution of the state equations pair

x· = Ax + bu
y = Cx + du

where A and C are 2 × 2 or higher order matrices, and b and d are column vectors with two or
At
more rows, entails the computation of the state transition matrix e , and integration of

A ( t – t0 ) t
–A τ
∫t e
At
x( t) = e x0 + e bu ( τ ) dτ
0

• The eigenvalues λ i , where i = 1, 2, …, n , of an n × n matrix A are the roots of the nth order
polynomial
det [ A – λI ] = 0

where I is the n × n identity matrix.

5-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• We can use the MATLAB eig(x) function to find the eigenvalues of an n × n matrix.

• The Cayley-Hamilton theorem states that a matrix can be expressed as an ( n – 1 )th degree poly-
nomial in terms of the matrix A as
At 2 n–1
e = a0 I + a1 A + a2 A + … + an – 1 A

where the coefficients a i are functions of the eigenvalues λ .

• If all eigenvalues of a given matrix A are distinct, that is, if λ 1 ≠ λ 2 ≠ λ 3 ≠ … ≠ λ n , the coefficients
a i are found from the simultaneous solution of the system of equations
2 n–1 λ1 t
a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 = e
2 n–1 λ2 t
a0 + a1 λ2 + a2 λ2 + … + an – 1 λ2 = e
…
2 n–1 λn t
a0 + a1 λn + a2 λn + … + an – 1 λn = e

• If some or all eigenvalues of matrix A are repeated, that is, if λ 1 = λ 2 = λ 3 … = λ m , λ m + 1 , λ n ,

the coefficients a i of the state transition matrix are found from the simultaneous solution of the
system of equations

n–1 λ1 t
a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1
2
= e
d d λt
--------- ( a 0 + a 1 λ 1 + a 2 λ 21 + … + a n – 1 λ n1 – 1 ) = -------- e 1
dλ 1 dλ 1
2 2
d- n–1 d λ1 t
( a 0 + a 1 λ 1 + a 2 λ 1 + … + a n – 1 λ 1 ) = --------2 e
2
-------
2
dλ 1 dλ 1
…
m–1 m–1
d d λ t
- ( a 0 + a 1 λ 1 + a 2 λ 21 + … + a n – 1 λ n1 – 1 ) = --------------
--------------
m–1 m–1
-e 1
dλ 1 dλ 1
n–1 λ m + 1t
a0 + a1 λm + 1 + a2 λm + 1 + … + an – 1 λm + 1 = e
2

…
n–1 λn t
a 0 + a 1 λn + a 2 λ n + … + a n – 1 λ n
2
= e

• A column vector X that satisfies the relation

AX = λX

Signals and Systems with MATLAB Applications, Second Edition 5-37

Orchard Publications
Chapter 5 State Variables and State Equations

where A is an n × n matrix and λ is a scalar number is called an eigenvector.

det ( A – λI ) = 0 (5.129)
• There is a different eigenvector for each eigenvalue.
• Eigenvectors are generally expressed as unit eigenvectors, that is, they are normalized to unit
length. This is done by dividing each component of the eigenvector by the square root of the sum
of the squares of their components, so that the sum of the squares of their components is equal to
unity.
• Two vectors X and Y are said to be orthogonal if their inner (dot) product is zero.

• A set of eigenvectors constitutes an orthonormal basis if the set is normalized (expressed as unit
eigenvectors) and these vector are mutually orthogonal.
• The state transition matrix can be computed from the Inverse Laplace transform using the rela-
tion
At –1 –1
e = L { ( sI – A ) }

• If U ( s ) is the Laplace transform of the input u ( t ) and Y ( s ) is the Laplace transform of the out-
put y ( t ) , the transfer function can be computed using the relation
Y ( s ) = C ( sI – A ) –1 b + d
G ( s ) = -----------
U(s)
• MATLAB provides two very useful functions to convert state space (state equations), to transfer
function (s-domain), and vice versa. The function ss2tf (state space to transfer function) converts
the state space equations to the transfer function equivalent, and the function tf2ss, converts the
transfer function to state-space equations.

5-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

5.9 Exercises
1. Express the integrodifferential equation below as a matrix of state equations where
k 1, k 2, and k 3 are constants.
2 t
dv
∫0 v dt
dv
-------2- + k 3 ------ + k 2 v + k 1 = sin 3t + cos 3t
dt dt

2. Express the matrix of the state equations below as a single differential equation, and let
x( y) = y(t) .

x· 1 0 1 0 0 x1 0
x· 2 x
= 0 0 1 0 ⋅ 2 + 0 u(t)
x· 3 0 0 0 1 x3 0
x· 4 –1 –2 –3 –4 x4 1

3. For the circuit of Figure 5.11, all initial conditions are zero, and u ( t ) is any input. Write state
equations in matrix form.

R C

u( t)
+
− L
`
Figure 5.11. Circuit for Exercise 3
4. In the circuit of Figure 5.12, all initial conditions are zero. Write state equations in matrix form.
L
R
`

1Ω C1 1H C2
+
−
V p cos ωtu 0 ( t ) 2F 2F

Figure 5.12. Circuit for Exercise 4

5. In the circuit of Figure 5.13, i L ( 0 − ) = 2 A . Use the state variable method to find i L ( t ) for t > 0 .

R 2Ω

10u 0 ( t )
+
− L
`2 H
Figure 5.13. Circuit for Exercise 5

Signals and Systems with MATLAB Applications, Second Edition 5-39

Orchard Publications
Chapter 5 State Variables and State Equations

6. Compute the eigenvalues of the matrices A , B , and C below.

0 1 0
A = 1 2 B = a 0 C = 0 0 1
3 –1 –a b
– 6 – 11 – 6

Hint: One of the eigenvalues of matrix C is – 1 .

At
7. Compute e given that
0 1 0
A = 0 0 1
– 6 – 11 – 6

Observe that this is the same matrix as C of Exercise 6.

8. Find the solution of the matrix state equation x· = Ax + bu given that

A= 1 0 , b= 1 , x0 = –1 , u = δ ( t ), t0 = 0
–2 2 2 0

9. In the circuit of Figure 5.14, i L ( 0 − ) = 0 , and v C ( 0 − ) = 1 V .

a. Write state equations in matrix form.

At
b. Compute e using the Inverse Laplace transform method.
c. Find i L ( t ) and v C ( t ) for t > 0 .

C
R
3⁄4 Ω
L
` 4H
4⁄3 F

Figure 5.14. Circuit for Exercice 9

5-40 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

5.10 Solutions to Exercises

1. Differentiating the given integrodifferential equation with respect to t we get
3 2
dv dv dv
-------- + k 3 -------2- + k 2 ------ + k 1 v = 3 cos 3t – 3 sin 3t = 3 ( cos 3t – sin 3t )
dt
3
dt dt
or
3 2
dv dv dv
-------- = – k 3 -------2- – k 2 ------ – k 1 v + 3 ( cos 3t – sin 3t ) (1)
dt
3
dt dt
We let
2
dv · dv ·
v = x1 ------ = x 2 = x 1 -------- = x 3 = x 2
dt 2
dt
Then,
3
dv - = x·
-------
3 3
dt
and by substitution into (1)
·
x 3 = – k 1 x 1 – k 2 x 2 – k 3 x 3 + 3 ( cos 3t – sin 3t )

and thus the state equations are

·
x1 = x2
·
x2 = x3
·
x 3 = – k 1 x 1 – k 2 x 2 – k 3 x 3 + 3 ( cos 3t – sin 3t )

and in matrix form

·
x1 0 1 0 x1 0
· = 0 0 1 ⋅ x +
x2 2 0 ⋅ 3 ( cos 3t – sin 3t )
· –k1 –k2 –k3 x3 1
x3

2.
Expansion of the given matrix yields
· · · ·
x1 = x2 x2 = x3 x3 = x2 x 4 = – x 1 – 2x 2 – 3x 3 – 4x 4 + u ( t )

Letting x = y we get
4 3 2
dy -
dy - + 4 ------- dy dy
------- 3
+ 3 -------2- + 2 ------ + y = u ( t )
dt
4
dt dt dt

Signals and Systems with MATLAB Applications, Second Edition 5-41

Orchard Publications
Chapter 5 State Variables and State Equations

3.

R iT
iL + iC

u(t)
+
− L
` vC
−
C

We let i L = x 1 and v C = x 2 . By KCL, i T = i L + i C or

u ( t ) – vC dv C
--------------------- = i L + C ---------
R dt
or
u ( t ) – x2 ·
-------------------- = x 1 + Cx 2
R
Also,
·
x 2 = Lx 1
Then,

· 1 · 1 1 1
x 1 = --- x 2 and x 2 = – ---- x 1 – -------- x 2 + -------- u ( t )
L C RC RC
and in matrix form
·
x1
= 0 1 ⁄ L ⋅ x1 + 0 ⋅ u(t)
·
x2 – 1 ⁄ C – 1 ⁄ RC x 2 1 ⁄ RC

4.
L
R v C1 iL
`

1Ω 1H
+ +
+ v C1 v C2
− − −
V p cos ωt C1 2 F C2 2F

We let i L = x 1 , v C1 = x 2 , and v C2 = x 3 . By KCL

v C1 – V p cos ωt dv C1 ·
----------------------------------- + 2 ----------- + i L = 0 or x 2 – V p cos ωt + 2x 2 + x 1 = 0
1 dt
or
· 1 1 1
x 2 = – --- x 1 – --- x 2 + --- V p cos ωt (1)
2 2 2
By KVL

5-42 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

di L · ·
v C1 = L ------- + v C2 or x 2 = 1x 1 + x 3 or x 1 = x 2 – x 3 (2)
dt
Also,
dv C2 · · 1
i L = C ----------- or x 1 = 2x 3 or x 3 = --- x 1 (3)
dt 2
Combining (1), (2), and (3) into matrix form we get

·
x1 0 1 –1 x1 0
· =
x2 –1 ⁄ 2 –1 ⁄ 2 0 ⋅ x 2
+ 1 ⁄ 2 ⋅ V p cos ωt
· 1⁄2 0 0 x3 0
x3

5.

R 2Ω

10u 0 ( t )
+
− L
`2 H
From (5.21) of Example 5.4
R 1
x· = – --- x + --- v S u 0 ( t )
L L

For this exercise α = – R ⁄ L = – 1 and b = 10 × ( 1 ⁄ L ) = 5 . Then,

α ( t – t0 ) t
αt –α τ
x( t) = e x0 + e ∫t e 0
β u ( τ ) dτ

t t
–1 ( t – 0 ) –t τ –t –t τ
= e 2+e ∫0 e 5u 0 ( τ ) dτ = 2e + 5e ∫0 e dτ
–t –t t –t –t –t
= 2e + 5e ( e – 1 ) = 2e + 5 – 5 e = ( 5 – 3e )u 0 ( t )

and denoting the current i L as the output y we get

–t
y ( t ) = x ( t ) = ( 5 – 3e )u 0 ( t )
6.
a.
⎛ ⎞
A = 1 2 det ( A – λI ) = det ⎜ 1 2 – λ 1 0 ⎟ = det 1 – λ 2 = 0
3 –1 ⎝ 3 –1 0 1⎠ 3 –1 –λ

( 1 – λ )( – 1 – λ ) – 6 = 0

Signals and Systems with MATLAB Applications, Second Edition 5-43

Orchard Publications
Chapter 5 State Variables and State Equations

2
–1–λ+λ+λ –6 = 0
2
λ = 7
and thus
λ1 = 7 λ2 = – 7
b.
⎛ ⎞
B = a 0 det ( B – λI ) = det ⎜ a 0 – λ 1 0 ⎟ = det a – λ 0 = 0
–a b ⎝ –a b 0 1⎠ –a b – λ

(a – λ )(b – λ ) = 0
and thus
λ1 = a λ2 = b
c.

0 1 0 ⎛ 0 1 0 1 0 0 ⎞⎟
⎜
C = 0 0 1 det ( C – λI ) = det ⎜ 0 0 1 –λ 0 1 0 ⎟
⎜ ⎟
– 6 – 11 – 6 ⎝ – 6 – 11 – 6 0 0 1⎠

–λ 1 0
= det 0 – λ 1 =0
– 6 – 11 – 6 – λ

2 3 2
λ ( – 6 – λ ) – 6 – ( – 11 ) ( – λ ) = λ + 6λ + 11λ + 6 = 0

and it is given that λ 1 = – 1 . Then,

3 2
λ + 6λ + 11λ + 6 2
---------------------------------------------- = λ + 5λ + 6 ⇒ ( λ + 1 ) ( λ + 2 ) ( λ + 3 ) = 0
(λ + 1)
and thus
λ1 = –1 λ2 = –2 λ1 = –3
7.
a. Matrix A is the same as Matrix C in Exercise 6. Then,
λ1 = –1 λ2 = –2 λ1 = –3

and since A is a 3 × 3 matrix the state transition matrix is

At 2
e = a0 I + a1 A + a2 A (1)
Then,

5-44 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

2 λ1 t –t
a0 + a1 λ1 + a2 λ1 = e ⇒ a0 – a1 + a2 = e
2 λ2 t – 2t
a0 + a1 λ2 + a2 λ2 = e ⇒ a 0 – 2a 1 + 4a 2 = e
2 λ3 t – 3t
a0 + a1 λ3 + a2 λ3 = e ⇒ a 0 – 3a 1 + 9a 2 = e

syms t; A=[1 −1 1; 1 −2 4; 1 −3 9];...

a=sym('[exp(−t); exp(−2*t); exp(−3*t)]'); x=A\a; fprintf(' \n');...
disp('a0 = '); disp(x(1)); disp('a1 = '); disp(x(2)); disp('a2 = '); disp(x(3))
a0 =
3*exp(-t)-3*exp(-2*t)+exp(-3*t)
a1 =
5/2*exp(-t)-4*exp(-2*t)+3/2*exp(-3*t)
a2 =
1/2*exp(-t)-exp(-2*t)+1/2*exp(-3*t)
Thus,
–t – 2t – 3t
a 0 = 3e – 3e + 3e
–t – 2t – 3t
a 1 = 2.5e – 4e + 1.5e
–t – 2t – 3t
a 2 = 0.5e – e + 0.5e

At
Now, we compute e of (1) with the following MATLAB code:
syms t; a0=3*exp(−t)−3*exp(−2*t)+exp(−3*t); a1=5/2*exp(−t)−4*exp(−2*t)+3/2*exp(−
3*t);...
a2=1/2*exp(−t)−exp(−2*t)+1/2*exp(−3*t); A=[0 1 0; 0 0 1; −6 −11 −6]; fprintf(' \n');...
eAt=a0*eye(3)+a1*A+a2*A^2
eAt =

[ 3*exp(-t)-3*exp(-2*t)+exp(-3*t), 5/2*exp(-t)-4*exp(-2*t)+3/2*exp(-3*t), 1/2*exp(-t)-exp(-2*t)+1/2*exp(-3*t)]

[-3*exp(-t)+6*exp(-2*t)-3*exp(-3*t), -5/2*exp(-t)+8*exp(-2*t)-9/2*exp(-3*t), -1/2*exp(-t)+2*exp(-2*t)-3/2*exp(-3*t)]
[3*exp(-t)-12*exp(-2*t)+9*exp(-3*t), 5/2*exp(-t)-16*exp(-2*t)+27/2*exp(-3*t), 1/2*exp(-t)-4*exp(-2*t)+9/2*exp(-3*t)]

Then,
–t – 2t – 3t –t – 2t – 3t –t – 2t – 3t
3e – 3e +e 2.5e – 4e + 1.5e 0.5e – e + 0.5e
At
e = – 3 e – t + 6e – 2t – 3e – 3t –t
– 2.5 e + 8e
– 2t
– 4.5e
– 3t –t
– 0.5 e + 2e
– 2t
– 1.5e
– 3t

–t – 2t – 3t –t – 2t – 3t –t – 2t – 3t
3e – 12e + 9e 2.5e – 16e + 13.5e 0.5e – 4e + 4.5e

8.

A= 1 0 , b= 1 , x0 = –1 , u = δ ( t ), t0 = 0
–2 2 2 0

Signals and Systems with MATLAB Applications, Second Edition 5-45

Orchard Publications
Chapter 5 State Variables and State Equations

t t
A(t – 0) –A τ –A τ
∫0 ∫0 e
At At At
x( t) = e x0 + e e bu ( τ ) dτ = e x 0 + e bδ ( τ ) dτ
(1)
⎞ At ⎛
= e x0 + e b = e ( x0 + b ) = e ⎜ –1 + 1 ⎟ = e 0
At At At At

⎝ 0 2 ⎠ 2

We use the following MATLAB code to find the eigenvalues λ 1 and λ 2

A=[1 0; −2 2]; lambda=eig(A); fprintf(' \n');...
fprintf('lambda1 = %4.2f \t',lambda(1)); fprintf('lambda2 = %4.2f \t',lambda(2))
lambda1 = 2.00 lambda2 = 1.00
Next,
λ1 t t
a0 + a1 λ1 = e ⇒ a0 + a1 = e
λ2 t 2t
a0 + a1 λ2 = e ⇒ a 0 + 2a 1 = e
Then,
t 2t 2t t
a 0 = 2e – e a1 = e – e
and

= a 0 I + a 1 A = ( 2e – e ) 1 0 + ( e – e ) 1 0
At t 2t 2t t
e
0 1 –2 2

t 2t 2t t t
= 2e – e 0 + e –e 0 = e 0
t 2t 2t t 2t t t 2t 2t
0 2e – e – 2e + 2e 2e – 2e 2e – 2e e

By substitution into (1) we get

t
0 = e 0 0
⋅ 0 =
At
x(t) = e
t 2t 2t 2t
2 2e – 2e e 2 2e
and thus
2t
x1 = 0 x 2 = 2e

9.
−
C iL ( 0 ) = 0
iL + iC −
`4 H
R iR L vC ( 0 ) = 1 V
−
3⁄4 Ω 4⁄3 F

We let x 1 = i L x 2 = v C . Then,

5-46 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

a.
iR + iL + iC = 0

vC vC
----- + i L + C ----- = 0
R dt

x2 4·
- + x 1 + --- x 2 = 0
--------
3⁄4 3
or
· 3
x 2 = – --- x 1 – x 2 (1)
4
Also,
di L ·
v L = v C = L ------- = 4x 1 = x 2
dt
or
· 1
x 1 = --- x (2)
4 2
From (1) and (2)
·
x1
= 0 1 ⁄ 4 ⋅ x1
·
x2 –3 ⁄ 4 –1 x2

and thus

A = 0 1⁄4
–3 ⁄ 4 –1
b.
At –1 –1
e = L { [ sI – A ] }

[ sI – A ] = s 0 – 0 1⁄4 = s –1 ⁄ 4
0 s –3 ⁄ 4 –1 3⁄4 s+1

∆ = det [ sI – A ] = det s – 1 ⁄ 4 = s 2 + s + 3 ⁄ 16 = ( s + 1 ⁄ 4 ) ( s + 3 ⁄ 4 )
3⁄4 s+1

adj [ sI – A ] = adj s –1 ⁄ 4 = s+1 1⁄4

3⁄4 s+1 –3 ⁄ 4 s

Signals and Systems with MATLAB Applications, Second Edition 5-47

Orchard Publications
Chapter 5 State Variables and State Equations

[ sI – A ]
–1
= --- adj [ sI – A ] = ----------------------------------------------- s + 1
1 1 1⁄4
∆ ( s + 1 ⁄ 4 ) ( s + 3 ⁄ 4 ) –3 ⁄ 4 s

s+1 1⁄4
----------------------------------------------- -----------------------------------------------
= ( s + 1 ⁄ 4 )(s + 3 ⁄ 4 ) (s + 1 ⁄ 4)(s + 3 ⁄ 4)
–3 ⁄ 4 s
----------------------------------------------- -----------------------------------------------
(s + 1 ⁄ 4 )(s + 3 ⁄ 4 ) (s + 1 ⁄ 4)(s + 3 ⁄ 4)

At –1 –1
We use MATLAB to find e = L { [ sI – A ] } with the code below.
syms s t
Fs1=(s+1)/(s^2+s+3/16); Fs2=(1/4)/(s^2+s+3/16); Fs3=(-3/4)/(s^2+s+3/16); Fs4=s/
(s^2+s+3/16);...
fprintf(' \n'); disp('a11 = '); disp(simple(ilaplace(Fs1))); disp('a12 = '); disp(simple(ilaplace(Fs2)));...
disp('a21 = '); disp(simple(ilaplace(Fs3))); disp('a22 = '); disp(simple(ilaplace(Fs4)))
a11 =
-1/2*exp(-3/4*t)+3/2*exp(-1/4*t)
a12 =
1/2*exp(-1/4*t)-1/2*exp(-3/4*t)
a21 =
-3/2*exp(-1/4*t)+3/2*exp(-3/4*t)
a22 =
3/2*exp(-3/4*t)-1/2*exp(-1/4*t)
Thus,
– 0.25t – 0.75t – 0.25t – 0.75t
e
At
= 1.5e – 0.5e 0.5e – 0.5e
– 0.25t – 0.75t – 0.25t – 0.75t
– 1.5 e + 1.5e – 0.5 e + 1.5e
c.
At ⎛ ⎞
t
A(t – 0) –A τ
∫0 e bu ( τ ) dτ = e x 0 + 0 = e ⎜ 0 + 0 ⎟
At At
x( t) = e x0 + e
⎝ 1 0⎠

– 0.25t – 0.75t – 0.25t – 0.75t – 0.25t – 0.75t

= 1.5e – 0.5e 0.5e – 0.5e 0 = 0.5e – 0.5e
– 0.25t – 0.75t – 0.25t – 0.75t 1 – 0.25t – 0.75t
– 1.5 e + 1.5e – 0.5 e + 1.5e – 0.5 e + 1.5e

and thus for t > 0

– 0.25t – 0.75t – 0.25t – 0.75t
x 1 = i L = 0.5e – 0.5e x 2 = v C = – 0.5 e + 1.5e

5-48 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 6
The Impulse Response and Convolution

his chapter begins with the definition of the impulse response, that is, the response of a circuit
T that is subjected to the excitation of the impulse function. Then, it defines convolution and
how it is applied to circuit analysis. Evaluation of the convolution integral using graphical
methods is also presented and illustrated with several examples.

6.1 The Impulse Response in Time Domain

In this section we will discuss the impulse response of a network, that is, the output (voltage or cur-
rent) of a network when the input is the delta function. Of course, the output can be any voltage or
current that we choose as the output. The determination of the impulse response assumes zero ini-
tial conditions.
We learned in the previous chapter that the state equation
x· = Ax + bu (6.1)
has the solution

A ( t – t0 ) t
–A τ
∫0 e
At
x(t) = e x0 + e bu ( τ ) dτ (6.2)

Therefore, with initial condition x 0 = 0 , and with the input u ( t ) = δ ( t ) , the solution of (6.2)
reduces to
t
–A τ
∫0 e
At
x(t) = e bδ ( τ ) dτ (6.3)

Using the sifting property of the delta function, i.e.,

∞

∫–∞ f ( t )δ ( τ ) dτ = f(0) (6.4)

and denoting the impulse response as h ( t ) , we get

At
h ( t ) = e bu 0 ( t ) (6.5)

where u 0 ( t ) is included to indicate that this relation holds for t > 0 .

Signals and Systems with MATLAB Applications, Second Edition 6-1

Orchard Publications
Chapter 6 The Impulse Response and Convolution

Example 6.1
Compute the impulse response of the series RC circuit of Figure 6.1 in terms of the constants R and
−
C , where the response is considered to be the voltage across the capacitor, and v c ( 0 ) = 0 . Then,
compute the current through the capacitor.

R
C + h(t) = v (t) = v (t)
+
−
C out
−
δ(t)

Figure 6.1. Circuit for Example 6.1

Solution:
We assign currents i C and i R with the directions shown in Figure 6.2, and we apply KCL.

iR
R iC
C +
+ h ( t ) = v C ( t ) = v out ( t )
− −
δ(t)

Figure 6.2. Application of KCL for the circuit for Example 6.1

Then,
iR + iC = 0
or
dv vC – δ ( t )
C -------C- + --------------------
- = 0 (6.6)
dt R
We assign the state variable
vC = x
Then,
dv C
-------- = x·
dt
and (6.6) becomes
x δ(t)
Cx· + --- = ---------
R R
or

6-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Impulse Response in Time Domain

1 1
x· = – -------- x + -------- δ ( t ) (6.7)
RC RC
This equation has the form
x· = ax + bu
and as we found in (6.5),
At
h ( t ) = e bu 0 ( t )
For this example,
a = – 1 ⁄ RC
and
b = 1 ⁄ RC
Therefore,
– t ⁄ RC 1
h ( t ) = vC ( t ) = e --------
RC
or
1 – t ⁄ RC
h ( t ) = -------- e u0 ( t ) (6.8)
RC

The current i C can now be computed from

dv
i C = C -------C-
dt
Thus,
i C = C ----- h ( t ) = C ----- ⎛ -------- e u 0 ( t )⎞
d d 1 – t ⁄ RC
dt dt ⎝ RC ⎠
1 –t ⁄ RC 1 –t ⁄ RC
= – ----------
2
e + --- e δ(t)
R C R

Using the sampling property of the delta function, we get

1 1 –t ⁄ RC
i C = --- δ ( t ) – ----------
2
e (6.9)
R R C

Example 6.2
For the circuit of Figure 6.3, compute the impulse response h ( t ) = v C ( t ) given that the initial condi-
tions are zero, that is, i L ( 0 − ) = 0 , and v C ( 0 − ) = 0 .
Solution:
This is the same circuit as that of Example 5.10 of the previous chapter, where we found that

Signals and Systems with MATLAB Applications, Second Edition 6-3

Orchard Publications
Chapter 6 The Impulse Response and Convolution

R L

`
+
1Ω 1⁄4 H
C
+ h ( t ) = vC ( t )
−
δ(t) 4⁄3 F
−

Figure 6.3. Circuit for Example 6.2

b = 4
0
and
–t – 3t –t – 3t
At – 0.5 e + 1.5e – 2 e + 2e
e = – 3t
3
--- e – t – 3 –t – 3t
--- e 1.5e – 0.5e
8 8

The impulse response is obtained from (6.5), that is,

At
h ( t ) = x ( t ) = e bu 0 ( t )
then,

–t – 3t –t – 3t –t – 3t
x1 – 0.5 e + 1.5e – 2 e + 2e 4 u ( t ) = – 2 e + 6e u ( t )
h( t)= x(t) = = – 3t 0 – 3t 0 (6.10)
3
--- e – t – 3 –t – 3t 3
--- e – t – 3
x2 --- e 1.5e – 0.5e 0 --- e
8 8 2 2

In Example 5.10, we defined

x1 = iL
and
x2 = vC
Then,
–t – 3t
h ( t ) = x 2 = v C ( t ) = 1.5e – 1.5e
or
–t – 3t
h ( t ) = v C ( t ) = 1.5 ( e – e ) (6.11)
Of course, this answer is not the same as that of Example 5.10, because the inputs and initial condi-
tions were defined differently.

6-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Even and Odd Functions of Time

6.2 Even and Odd Functions of Time

A function f ( t ) is an even function of time if the following relation holds.

f ( –t ) = f ( t ) (6.12)

that is, if in an even function we replace t with – t , the function f ( t ) does not change. Thus, polyno-
mials with even exponents only, and with or without constants, are even functions. For instance, the
cosine function is an even function because it can be written as the power series
2 4 6
cos t = 1 – ---- t - – ----
t - + ---- t- + …
2! 4! 6!
Other examples of even functions are shown in Figure 6.4.
f(t) f(t) f(t)
t2 + k
t2 k

t t t
0 0 0

Figure 6.4. Examples of even functions

A function f ( t ) is an odd function of time if the following relation holds.

–f ( –t ) = f ( t ) (6.13)

that is, if in an odd function we replace t with – t , we obtain the negative of the function f ( t ) . Thus,
polynomials with odd exponents only, and no constants are odd functions. For instance, the sine
function is an odd function because it can be written as the power series
3 5 7
sin t = t – ---- t - – ----
t - + ---- t- + …
3! 5! 7!
Other examples of odd functions are shown in Figure 6.5.
f(t) f(t) f(t)
mt t3

t t t
0 0 0

Figure 6.5. Examples of odd functions

Signals and Systems with MATLAB Applications, Second Edition 6-5

Orchard Publications
Chapter 6 The Impulse Response and Convolution

We observe that for odd functions, f ( 0 ) = 0 . However, the reverse is not always true; that is, if
f ( 0 ) = 0 , we should not conclude that f ( t ) is an odd function. An example of this is the function
f ( t ) = t in Figure 6.4.
2

The product of two even or two odd functions is an even function, and the product of an even func-
tion times an odd function, is an odd function.
Henceforth, we will denote an even function with the subscript e , and an odd function with the sub-
script o . Thus, f e ( t ) and f o ( t ) will be used to represent even and odd functions of time respectively.

Also,
T T

∫– T f e ( t ) dt = 2 ∫0 f e ( t ) dt (6.14)

and
T

∫– T f o ( t ) dt = 0 (6.15)

A function f ( t ) that is neither even nor odd can be expressed as

1
f e ( t ) = --- [ f ( t ) + f ( – t ) ] (6.16)
2

or as

1
f o ( t ) = --- [ f ( t ) – f ( – t ) ] (6.17)
2

By addition of (6.16) with (6.17), we get

f ( t ) = fe ( t ) + fo ( t ) (6.18)

that is, any function of time can be expressed as the sum of an even and an odd function.

Example 6.3
Determine whether the delta function is an even or an odd function of time.
Solution:
Let f ( t ) be an arbitrary function of time that is continuous at t = t 0 . Then, by the sifting property of
the delta function

6-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Convolution
∞

∫–∞ f ( t )δ ( t – t ) dt
0 = f ( t0 )

and for t 0 = 0 ,
∞

∫–∞ f ( t )δ ( t ) dt = f(0)

Also,
∞

∫–∞ fe ( t )δ ( t ) dt = fe ( 0 )

and
∞

∫–∞ fo ( t )δ ( t ) dt = fo ( 0 )

As stated earlier, an odd function f o ( t ) evaluated at t = 0 is zero, that is, f o ( 0 ) = 0 . Therefore,

from the last relation above,
∞

∫–∞ fo ( t )δ ( t ) dt = fo ( 0 ) = 0 (6.19)

and this indicates that the product f o ( t )δ ( t ) is an odd function of t . Then, since f o ( t ) is odd, it fol-
lows that δ ( t ) must be an even function of t for (6.19) to hold.

6.3 Convolution
Consider a network whose input is δ ( t ) , and its output is the impulse response h ( t ) . We can repre-
sent the input-output relationship as the block diagram shown below.
δ(t) h(t)
Network

or in general,
δ(t – τ) h( t – τ)
Network

We let u ( t ) be any input whose value at t = τ is u ( τ ) . Then,

u ( τ )δ ( t – τ ) u ( τ )h ( t – τ )
Network

Multiplying both sides by the constant dτ , integrating from – ∞ to +∞ , and making use of the fact
that the delta function is even, i.e., δ ( t – τ ) = δ ( τ – t ) , we get

Signals and Systems with MATLAB Applications, Second Edition 6-7

Orchard Publications
Chapter 6 The Impulse Response and Convolution

∞ ⎫ ⎧ ∞

∫ –∞
u ( τ )δ ( t – τ ) dτ ⎪
⎪
⎪
⎪ ∫–∞ u ( τ )h ( t – τ ) dτ
⎬ Network ⎨ ∞
∞
⎪ ⎪
∫ u ( τ )δ ( τ – t ) dτ ⎪ ⎪
⎩
∫–∞ u ( t – τ )h ( τ ) dτ
–∞ ⎭

Using the sifting property of the delta function, we find that the second integral on the left side
reduces to u ( t ) and thus

⎧ ∞
⎪
⎪
∫–∞ u ( τ )h ( t – τ ) dτ
u( t) Network ⎨ ∞
⎪
⎪
⎩
∫–∞ u ( t – τ )h ( τ ) dτ
The integral
∞ ∞

∫– ∞ u ( τ )h ( t – τ ) dτ or ∫–∞ u ( t – τ )h ( τ ) dτ (6.20)

is known as the convolution integral; it states that if we know the impulse response of a network, we
can compute the response to any input u ( t ) using either of the integrals of (6.20).
The convolution integral is usually denoted as u ( t )*h ( t ) or h ( t )*u ( t ) , where the asterisk (*) denotes
convolution.
At
In Section 6.1, we found that the impulse response for a single input is h ( t ) = e b . Therefore, if we
know h ( t ) , we can use the convolution integral to compute the response y ( t ) of any input u ( t ) using
the relation
∞ ∞
A(t – τ) –A τ
∫– ∞ ∫– ∞ e
At
y( t) = e bu ( τ ) dτ = e bu ( τ ) dτ (6.21)

6.4 Graphical Evaluation of the Convolution Integral

The convolution integral is more conveniently evaluated by the graphical evaluation. The procedure
is best illustrated with the following examples.

Example 6.4
The signals h ( t ) and u ( t ) are as shown in Figure 6.6. Compute h ( t )*u ( t ) using the graphical evalua-
tion.

6-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Graphical Evaluation of the Convolution Integral

u ( t ) = u0 ( t ) – u0 ( t – 1 )
1 h(t) = – t + 1 1

t t
0 1 0 1

Figure 6.6. Signals for Example 6.4

Solution:
The convolution integral states that
∞
h ( t )∗ u ( t ) = ∫–∞ u ( t – τ )h ( τ ) dτ (6.22)

where τ is a dummy variable, that is, u ( τ ) and h ( τ ) , are considered to be the same as u ( t ) and h ( t ) .
We form u ( t – τ ) by first constructing the image of u ( τ ) ; this is shown as u ( – τ ) in Figure 6.7.

u ( –τ )
1

−1 0
τ

Figure 6.7. Construction of u ( – τ ) for Example 6.4

Next, we form u ( t – τ ) by shifting u ( – τ ) to the right by some value t as shown in Figure 6.8.

1 u ( t –τ )

τ
0 t

Figure 6.8. Formation of u ( t – τ ) for Example 6.4

Now, evaluation of the convolution integral

∞
h ( t )∗ u ( t ) = ∫–∞ u ( t – τ )h ( τ ) dτ
entails multiplication of u ( t – τ ) by h ( τ ) for each value of t , and computation of the area from
– ∞ to +∞ . Figure 6.9 shows the product u ( t – τ )h ( τ ) as point A moves to the right.

Signals and Systems with MATLAB Applications, Second Edition 6-9

Orchard Publications
Chapter 6 The Impulse Response and Convolution

u ( t – τ ), t = 0 u ( t – τ )*h ( τ ) = 0 for t = 0
1

h(τ)

A
τ
−1 0

Figure 6.9. Formation of the product u ( t – τ )*h ( τ ) for Example 6.4

We observe that u ( t – τ ) t=0

= u ( –τ )

Shifting u ( t – τ ) to the right so that t > 0 , we get the sketch of Figure 6.10 where the integral of the
product is denoted by the shaded area, and it increases as point A moves further to the right.

u ( t – τ ), t > 0
1

h(τ)
A τ
0 t 1

Figure 6.10. Shift of u ( t – τ ) for Example 6.4

The maximum area is obtained when point A reaches t = 1 as shown in Figure 6.11.
u ( t – τ ), t = 0
1
h(τ)

A
τ
0 1

Figure 6.11. Signals for Example 6.4 when t = 1

Using the convolution integral, we find the area as a function of time t is

∞ t t 2 t
τ
2

∫– ∞ ∫0 ∫
t
u ( t – τ )h ( τ ) dτ = u ( t – τ )h ( τ ) dτ = ( 1 ) ( – τ + 1 ) dτ = τ – ---- = t – --- (6.23)
0 2 0
2

Figure 6.12 shows how u ( τ )*h ( τ ) increases during the interval 0 < t < 1 . This is not an exponential
increase; it is the function t – t 2 ⁄ 2 in (6.23), and each point on the curve of Figure 6.12 represents
the area under the convolution integral.

6-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Graphical Evaluation of the Convolution Integral

u(t)*h(t)

Figure 6.12. Curve for the convolution of u ( τ )*h ( τ ) for 0 < t < 1 in Example 6.4

Evaluating (6.23) at t = 1 , we get

2
t 1
t – --- = --- (6.24)
2 t=1
2

The plot for the interval 0 ≤ t ≤ 1 is shown in Figure 6.13.

As we continue shifting u ( t – τ ) to the right, the area starts decreasing, and it becomes zero at t = 2 ,
as shown in Figure 6.14.

0.5

0.4
2
t–t ⁄2
0.3

0.2

0.1

0
0.0 0.5 1.0 1.5 2.0

Figure 6.13. Convolution of u ( τ )*h ( τ ) at t = 1 for Example 6.4

u ( t – τ ), 1 < t < 2 u ( t – τ ), t = 2

1 1
h(τ) h(τ)

A A
τ τ
0 t−1 1 t 0 1 2

Figure 6.14. Convolution for interval 1 < t < 2 of Example 6.4

Using the convolution integral, we find that the area for the interval 1 < t < 2 is

Signals and Systems with MATLAB Applications, Second Edition 6-11

Orchard Publications
Chapter 6 The Impulse Response and Convolution

∞ 1 1 2 1
τ
∫– ∞ u ( t – τ )h ( τ ) dτ = ∫t – 1 u ( t – τ )h ( τ ) dτ = ∫ t–1
( 1 ) ( – τ + 1 ) dτ = τ – ----
2 t–1 (6.25)
2 2
1 t – 2t + 1 t
= 1 – --- – ( t – 1 ) + ------------------------ = --- – 2t + 2
2 2 2

2
Thus, for 1 < t < 2 , the area decreases in accordance with t ⁄ 2 – 2t + 2 .
Evaluating (6.25) at t = 2 , we find that u ( τ )*h ( τ ) = 0 . For t > 2 , the product u ( t – τ )h ( τ ) is zero
since there is no overlap between these two signals. The convolution of these signals for 0 ≤ t ≤ 2 , is
shown in Figure 6.15.

0.5
2
t ⁄ 2 – 2t + 2
0.4

0.3
2
t–t ⁄2
0.2

0.1

0
0.0 0.5 1.0 1.5 2.0

Figure 6.15. Convolution for 0 ≤ τ ≤ 2 of the signals of Example 6.4

Example 6.5
The signals h ( t ) and u ( t ) are as shown in Figure 6.16. Compute h ( t )*u ( t ) using the graphical evalu-
ation method.
u ( t ) = u0 ( t ) – u0 ( t – 1 )
1 1
–t
h(t) = e

t t
0 0 1

Figure 6.16. Signals for Example 6.5

Solution:
Following the same procedure as in the previous example, we form u ( t – τ ) by first constructing the
image of u ( τ ) . This is shown as u ( – τ ) in Figure 6.17.

6-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Graphical Evaluation of the Convolution Integral

u ( –τ )
1

τ
−1 0

Figure 6.17. Construction of u ( – τ ) for Example 6.5

Next, we form u ( t – τ ) by shifting u ( – τ ) to the right by some value t as shown in Figure 6.18.

1 u ( t –τ )

τ
0 t

Figure 6.18. Formation of u ( t – τ ) for Example 6.5

As in the previous example, evaluation of the convolution integral

∞
h ( t )∗ u ( t ) = ∫–∞ u ( t – τ )h ( τ ) dτ
entails multiplication of u ( t – τ ) by h ( τ ) for each value ot t , and computation of the area from
– ∞ to +∞ . Figure 6.19 shows the product u ( t – τ )h ( τ ) as point A moves to the right.

u ( t – τ ), t = 0
1 u ( t – τ )*h ( τ ) = 0 for t = 0

h(τ)
A
τ
−1 0

Figure 6.19. Formation of the product u ( t – τ )*h ( τ ) for Example 6.5

We observe that u ( t – τ ) t=0

= u ( –τ )

Shifting u ( t – τ ) to the right so that t > 0 , we get the sketch of Figure 6.20 where the integral of the
product is denoted by the shaded area, and it increases as point A moves further to the right.

Signals and Systems with MATLAB Applications, Second Edition 6-13

Orchard Publications
Chapter 6 The Impulse Response and Convolution

u ( t – τ ), t > 0 1

h(τ)
A τ
0 t

Figure 6.20. Shift of u ( t – τ ) for Example 6.5

The maximum area is obtained when point A reaches t = 1 as shown in Figure 6.21.
u ( t – τ ), t = 1
1
h(τ)
A
τ
0 1

Figure 6.21. Convolution of u ( τ )*h ( τ ) at t = 1 for Example 6.5

Its value for 0 < t < 1 is

∞ t t
–τ –τ t –τ 0 –t
∫– ∞ u ( t – τ )h ( τ ) dτ = ∫0 u ( t – τ )h ( τ ) dτ = ∫0 ( 1 ) ( e ) dτ = – e 0
= e t
= 1–e (6.26)

Evaluating (6.26) at t = 1 , we get

–t –1
1–e t=1
= 1–e = 0.632 (6.27)

The plot for the interval 0 ≤ t ≤ 1 is shown in Figure 6.22.

u(t)*h(t)

1.0
0.8 –t
1–e
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0

Figure 6.22. Convolution of u ( τ )*h ( τ ) for 0 ≤ t ≤ 1 in Example 6.5

6-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Graphical Evaluation of the Convolution Integral

As we continue shifting u ( t – τ ) to the right, the area starts decreasing. As shown in Figure 6.23, it
approaches zero as t becomes large but never reaches the value of zero.
u ( t – τ ), t = 1
1

h(τ)
A
τ
0 t−1 t

Figure 6.23. Convolution for interval 1 < t < 2 of Example 6.5

Therefore, for the time interval t > 1 , we have

–τ t – 1
t t
–τ –τ t –( t – 1 ) –t –t
∫t – 1 u ( t – τ )h ( τ ) dτ = ∫t – 1 ( 1 ) ( e ) dτ = – e t–1
= e t
= e –e = e (e – 1) (6.28)

Evaluating (6.28) at t = 2 , we find that u ( τ )*h ( τ ) = 0.233 .

For t > 2 , the product u ( t – τ )h ( τ ) approaches zero as t → ∞ . The convolution of these signals for
0 ≤ t ≤ 2 , is shown in Figure 6.24.

u(t)*h(t)

1.0
–t
0.8 1–e
–t
e (e – 1)
0.6
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0

Figure 6.24. Convolution for 0 ≤ t ≤ 2 of the signals of Example 6.5

Example 6.6
Perform the convolution v 1 ( t )*v 2 ( t ) where v 1 ( t ) and v 2 ( t ) are as shown in Figure 6.25.

Solution:
We will use the convolution integral

Signals and Systems with MATLAB Applications, Second Edition 6-15

Orchard Publications
Chapter 6 The Impulse Response and Convolution

v1 ( t )
2
v2 ( t )
1

t t
1 2

Figure 6.25. Signals for Example 6.6

∞
v 1 ( t )∗ v 2 ( t ) = ∫–∞ v ( τ )v ( t – τ ) dτ
1 2 (6.29)

The computation steps are as in the two previous examples, and are evident from the sketches of Fig-
ures 6.26 through 6.29.
Figure 6.26 shows the formation of v 2 ( – τ ) .

v1 ( t )
2
v2 ( –τ )
1

τ τ
−2 1

Figure 6.26. Formation of v 2 ( – τ ) for Example 6.6

Figure 6.27 shows the formation of v 2 ( t – τ ) and convolution with v 1 ( t ) for 0 < t < 1 .

v1 ( t )
2
v2 ( t – τ ) v 1 ( t )*v 2 ( t ) = 2 × 1 × t = 2t
1

τ
t 1

Figure 6.27. Formation of v 2 ( t – τ ) and convolution with v 1 ( t )

For 0 < t < 1 ,

t
v 1 ( t )∗ v 2 ( t ) = ∫ 0 ( 1 ) ( 2 ) dτ
t
= 2τ 0
= 2t (6.30)

6-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Graphical Evaluation of the Convolution Integral

Figure 6.28 shows the convolution of v 2 ( t – τ ) with v 1 ( t ) for 1 < t < 2 .

v1 ( t )
2
v2 ( t – τ ) 1

τ
1t
Figure 6.28. Convolution of v 2 ( t – τ ) with v 1 ( t ) for 1 < t < 2

For 1 < t < 2 ,

1
v 1 ( t )∗ v 2 ( t ) = ∫0
1
( 1 ) ( 2 ) dτ = 2τ 0
= 2 (6.31)

Figure 6.29 shows the convolution of v 2 ( t – τ ) with v 1 ( t ) for 2 < t < 3 .

v1 ( t )
2

v2 ( t – τ )
1

1
τ
t–2 t

Figure 6.29. Convolution of v 2 ( t – τ ) with v 1 ( t ) for 2 < t < 3

For 2 < t < 3

1
v 1 ( t )∗ v 2 ( t ) = ∫t – 2 ( 1 ) ( 2 ) d τ
1
= 2τ t–2
= – 2t + 6 (6.32)

From (6.30), (6.31), and (6.32), we obtain the waveform of Figure 6.30 that represents the convolu-
tion of the signals v 1 ( t ) and v 2 ( t – τ ) .

( v 1 ( t ) )∗ v 2 ( t )
2

t
0 1 2 3
Figure 6.30. Convolution of v 1 ( t ) with v 2 ( t ) for 0 < t < 3

Signals and Systems with MATLAB Applications, Second Edition 6-17

Orchard Publications
Chapter 6 The Impulse Response and Convolution

In summary, the procedure for the graphical evaluation of the convolution integral, is as follows:
1. We substitute u ( t ) and h ( t ) with u ( τ ) and h ( τ ) respectively.
2. We fold (form the mirror image of) u ( τ ) or h ( τ ) about the vertical axis to obtain u ( – τ ) or h ( – τ ) .
3. We slide u ( – τ ) or h ( – τ ) to the right a distance t to obtain u ( t – τ ) or h ( t – τ ) .
4. We multiply the two functions to obtain the product u ( t – τ ) h ( τ ) , or u ( τ ) h ( t – τ ) .
5. We integrate this product by varying t from – ∞ to +∞ .

6.5 Circuit Analysis with the Convolution Integral

We can use the convolution integral in circuit analysis as illustrated by the following example.

Example 6.7
For the circuit of Figure 6.31, use the convolution integral to find the capacitor voltage when the
input is the unit step function u 0 ( t ) , and v C ( 0 − ) = 0 .

R 1Ω
C +
+ vC ( t )
− −
1F
u0 ( t )

Figure 6.31. Circuit for Example 6.7

Solution:
Before we apply the convolution integral, we must know the impulse response h ( t ) of this circuit.
The circuit of Figure 6.31 was analyzed in Example 6.1 where we found that
1 – t ⁄ RC
h ( t ) = -------- e u0 ( t ) (6.33)
RC
With the given values, (6.33) reduces to
–t
h ( t ) = e u0 ( t ) (6.34)

Next, we use the graphical evaluation of the convolution integral as shown in Figures 6.32 through
6.34.
The formation of u 0 ( – τ ) is shown in Figure 6.32.

6-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Analysis with the Convolution Integral

u0 ( –τ )
1

τ
0

Figure 6.32. Formation of u 0 ( – τ ) for Example 6.7

Figure 6.33 shows the formation of u 0 ( t – τ ) .

u0 ( t –τ ) 1

τ
0 t

Figure 6.33. Formation of u 0 ( t – τ ) for Example 6.7

Figure 6.34 shows the convolution ( u 0 ( t ) )∗ h ( t ) .

h(τ)

τ
0 t

Figure 6.34. Convolution of u 0 ( t )*h ( t ) for Example 6.7

Therefore, for the interval 0 < t < ∞ , we get
∞ t
–t –t t –t 0 –t
u 0 ( t )*h ( t ) = ∫– ∞ u 0 ( t – τ )h ( τ ) dτ = ∫0 ( 1 )e dτ = – e 0
= e t
= ( 1 – e )u 0 ( t ) (6.35)

and the convolution u 0 ( t )*h ( t ) is shown in Figure 6.35.

1.0
0.8
0.6
0.4 –t
0.2
( 1 – e )u 0 ( t )
0.0
0.0 1.0 2.0 3.0 4.0

Figure 6.35. Convolution of u 0 ( t )∗ h ( t ) for Example 6.7

Signals and Systems with MATLAB Applications, Second Edition 6-19

Orchard Publications
Chapter 6 The Impulse Response and Convolution

6.6 Summary
• The impulse response is the output (voltage or current) of a network when the input is the delta
function.
• The determination of the impulse response assumes zero initial conditions.

• A function f ( t ) is an even function of time if the following relation holds.

f ( –t ) = f ( t )

• A function f ( t ) is an odd function of time if the following relation holds.

–f ( –t ) = f ( t )
• The product of two even or two odd functions is an even function, and the product of an even
function times an odd function, is an odd function.
• A function f ( t ) that is neither even nor odd, can be expressed as

1
f e ( t ) = --- [ f ( t ) + f ( – t ) ]
2
or as
1
f o ( t ) = --- [ f ( t ) – f ( – t ) ]
2

where f e ( t ) denotes an even function and f o ( t ) denotes an odd function.

• Any function of time can be expressed as the sum of an even and an odd function, that is,

f ( t ) = fe ( t ) + fo ( t )

• The delta function is an even function of time.

• The integral
∞

∫–∞ u ( τ )h ( t – τ ) dτ
or
∞

∫–∞ u ( t – τ )h ( τ ) dτ
is known as the convolution integral.
• If we know the impulse response of a network, we can compute the response to any input u ( t )
with the use of the convolution integral.

6-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• The convolution integral is usually denoted as u ( t )*h ( t ) or h ( t )*u ( t ) , where the asterisk (*)
denotes convolution.
• The convolution integral is more conveniently evaluated by the graphical evaluation method.

Signals and Systems with MATLAB Applications, Second Edition 6-21

Orchard Publications
Chapter 6 The Impulse Response and Convolution

6.7 Exercises
1. Compute the impulse response h ( t ) = i L ( t ) in terms of R and L for the circuit of Figure 6.36.
Then, compute the voltage v L ( t ) across the inductor.

iL ( t )

`
+ L
−
δ(t)

Figure 6.36. Circuit for Exercise 1

2. Repeat Example 6.4 by forming h ( t – τ ) instead of u ( t – τ ) , that is, use the convolution integral
∞

∫–∞ u( τ )h ( t – τ ) dτ
3. Repeat Example 6.5 by forming h ( t – τ ) instead of u ( t – τ ) .

4. Compute v 1 ( t )*v 2 ( t ) given that

⎧ 4t t≥0 ⎧ e – 2t t≥0
v1 ( t ) = ⎨ v2 ( t ) = ⎨
⎩0 t<0 ⎩ 0 t<0

5. For the series RL circuit shown in Figure 6.37, the response is the current i L ( t ) . Use the convolu-
tion integral to find the response when the input is the unit step u 0 ( t ) .

R
iL ( t )
1Ω

u0 ( t )
+
-
i(t)
L
` 1H
Figure 6.37. Circuit for Exercise 5

6. Compute v out ( t ) for the network of Figure 6.38 using the convolution integral, given that
v in ( t ) = u 0 ( t ) – u 0 ( t – 1 ) .

6-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

`
+ 1H +

v in ( t ) R v out ( t )
1Ω
− −

Figure 6.38. Network for Exercise 6

7. Compute v out ( t ) for the circuit of Figure 6.39 given that v in ( t ) = u 0 ( t ) – u 0 ( t – 1 ) .

R
+
1Ω +

v in ( t ) L
` 1H
v out ( t )

− −

Figure 6.39. Network for Exercise 7

Hint: Use the result of Exercise 6.

Signals and Systems with MATLAB Applications, Second Edition 6-23

Orchard Publications
Chapter 6 The Impulse Response and Convolution

6.8 Solutions to Exercises

1.
R

`
+ L
−
δ(t) iL ( t )

di
Ri + L ----- = δ ( t )
dt

Letting state variable x = i , the above relation is written as

x· = ( – R ⁄ L )x + ( 1 ⁄ L )δ ( t )

and this has the form x· = Ax + bu where A = – R ⁄ L , b = 1 ⁄ L , and u = δ ( t ) . Its solution is

A ( t – t0 ) t
–A τ
∫0 e
At
x( t) = e x0 + e bu ( τ ) dτ

and from (6.5)

At – ( R ⁄ L )t – ( R ⁄ L )t
h ( t ) = i ( t ) = e bu 0 ( t ) = e ⋅ 1 ⁄ L ⋅ u 0 ( t ) = ( 1 ⁄ L )e u0 ( t )

The voltage v L across the inductor is found from

v L = L ----- i ( t ) = L ----- h ( t ) = L ----- ⎛ --- e u 0 ( t ) ⎞ = ( – R ⁄ L )e

d d d 1 – ( R ⁄ L )t – ( R ⁄ L )t – ( R ⁄ L )t
u0 ( t ) + e δ(t)
dt dt dt L ⎝ ⎠

and using the sampling property of the delta function the above reduces to
– ( R ⁄ L )t
v L = ( – R ⁄ L )e u0 ( t ) + δ ( t )
2.

1 h(t) 1 1
1 1
h ( –τ ) h ( t –τ )

0 1 −1 0 τ 0 t τ 0 1 τ 0 t–1 1 t τ

From the plots above we observe that the area reaches the maximum value of 1 ⁄ 2 at t = 1 , and
then decreases to zero at t = 2 . Alternately, using the convolution integral we get

6-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
∞
u ( t )*h ( t ) = ∫–∞ u ( τ )h ( t – τ ) dτ
where h ( t ) = – t + 1 , h ( τ ) = – τ + 1 , h ( – τ ) = τ + 1 , and h ( t – τ ) = – ( t – τ ) + 1 = 1 – t + τ .
Then, for 0 < t < 1

2 t 2 2
t
τ - + ( 1 – t )τ
∫ = t---- + ( 1 – t )t = t – ---
Area 1 = [ ( 1 – t ) + τ ] dτ = ---- t-
0 2 0
2 2

and we observe that at t = 1 , Area 1 = 1 ⁄ 2 square units

Next, for 1 < t < 2

2 1
1
τ
Area 2 =
t–1
∫
[ ( 1 – t ) + τ ] dτ = ----- + ( 1 – t )τ
2 t–1
2 2
1 (t – 1) t
= --- + 1 – t – ----------------- – ( 1 – t ) ⋅ ( t – 1 ) = ---- – 2t + 2
2 2 2

and we observe that at t = 2 , Area 2 = 0

3.

1 1 1 1
h(t) h ( –τ )
h ( t –τ )
t τ τ
0 0 τ 0 t 1 0 1 t

From the plots above we observe that the area reaches its maximum value at t = 1 , and then
decreases exponentially to zero as t → ∞ . Alternately, using the convolution integral we get
∞
u ( t )*h ( t ) = ∫–∞ u ( τ )h ( t – τ ) dτ
–t –τ τ –( t – τ )
where h ( t ) = e , h ( τ ) = e , h ( – τ ) = e , and h ( t – τ ) = e . Then, for 0 < t < 1
t t
–( t – τ ) –t τ –t –t
∫0 ∫0 e dτ
t 0
Area 1 = (1 ⋅ e ) dτ = e = e (e – e ) = 1 – e

For t > 1

Signals and Systems with MATLAB Applications, Second Edition 6-25

Orchard Publications
Chapter 6 The Impulse Response and Convolution

1 1
–( t – τ ) –t τ –t τ 1 –t –t
∫0 ∫0
1 0
Area 2 = (1 ⋅ e ) dτ = e e dτ = e e 0
= e (e – e ) = e (e – 1)

4.
v2 ( t –τ ) v 1 ( t )*v 2 ( t )
1
1 v2 ( t )
v2 ( –τ ) 4t

t τ τ
0 0 τ 0 0 t

t t t
–2 ( t – τ ) – 2t
∫0 ∫0 ∫0 τe
2τ
v 1 ( t )*v 2 ( t ) = v 1 ( τ )v 2 ( t – τ ) dτ = 4τe dτ = 4e dτ

From tables of integrals,

ax
e
∫
ax
xe dx = ------2- ( ax – 1 )
a
and thus
2τ t
– 2t e ( 2τ – 1 ) – 2t 2t 0
v 1 ( t )*v 2 ( t ) = ( 4e ) --------------------------- = e [ e ( 2t – 1 ) – e ( – 1 ) ]
4 0
0 – 2t – 2t
= e ( 2t – 1 + e ) = 2t + e –1
Check:
2
v 1 ( t )*v 2 ( t ) ⇔ V 1 ( s ) ⋅ V 2 ( s ) , V1 ( s ) = 4 ⁄ s , V2 ( s ) = 1 ⁄ ( s + 2 )

4 4
V 1 ( s ) ⋅ V 2 ( s ) = --------------------- = -------------------
2 3 2
s (s + 2) s + 2s
syms s t; ilaplace(4/(s^3+2*s^2))
ans =
2*t-1+exp(-2*t)
5.
To use the convolution integral we must first find the impulse response. It was found in Exercise
1 as
– ( R ⁄ L )t
h ( t ) = i ( t ) = ( 1 ⁄ L )e u0 ( t )

and with the given values,

–t
h ( t ) = e u0 ( t )

6-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

1Ω
iL ( t )

u0 ( t )
+
-
i(t)
L
` 1H
When the input is the unit step u 0 ( t ) ,
∞
i( t ) v = u (t) =
in 0
∫–∞ u0 ( t – τ )h ( τ ) dτ
h(t) 1
1 1
u0 ( –τ ) 1 u0 ( t –τ )
–t
e

t
0 τ 0 τ τ
0 0 t
h ( t –τ )

t
–τ –τ t –τ 0 –t
i( t ) v = u (t) =
in 0
∫0 ( 1 ) ⋅ e dτ = – e 0
= e t
= ( 1 – e ) ( u0 ( t ) )

1
u 0 ( t )*h ( t )

6.
We will first compute the impulse response, that is, the output when the input is the delta func-
tion, i.e., v in ( t ) = δ ( t ) . Then, by KVL

di L
L ------- + Ri L = δ ( t )
dt

and with i L = x
1 ⋅ x· + 1 ⋅ x = δ ( t )
or
x· = – x + δ ( t )

By comparison with x· = Ax + bu we see that A = – 1 and b = 1 .

Signals and Systems with MATLAB Applications, Second Edition 6-27

Orchard Publications
Chapter 6 The Impulse Response and Convolution

From (6.5)
At –t –t
h ( t ) = e bu 0 ( t ) = e ⋅ 1 = e

Now, we compute v out ( t ) when v in ( t ) = u 0 ( t ) – u 0 ( t – 1 ) by convolving the impulse response

h ( t ) with this input v in ( t ) , that is, v out ( t ) = v in ( t )*h ( t ) . The remaining steps are as in Example
6.5 shown below.
1 0< t<1
t
–τ –τ t –τ 0 –t
h(τ)
∫0 ( 1 ) ( e ) dτ = – e 0
= e t
= 1–e
τ
0 t

t>1
1
–τ t – 1
t
–τ –τ t –t
h(τ) ∫t – 1 ( 1 ) ( e ) dτ = – e t–1
= e t
= e (e – 1)

τ
0 t−1 t

7.
R
+
1Ω +
v in ( t ) = u 0 ( t ) – u 0 ( t – 1 )
v in ( t ) L
` 1H
v out ( t )

− −

v out ( t ) = v L = v in – v R
where from Exercise 6,
–t
⎧1 – e 0< t<1
vR = ⎨
⎩ e –t ( e – 1 ) t>1
Then, for this circuit,
–t –t
⎧ (1 – (1 – e ) = e ) 0< t<1
vR = ⎨
–t –t
⎩ 0 – e ( e – 1 ) = ( 1 – e )e t>1

6-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 7
Fourier Series

his chapter is an introduction to Fourier series. We begin with the definition of sinusoids that
T are harmonically related and the procedure for determining the coefficients of the trigonomet-
ric form of the series. Then, we discuss the different types of symmetry and how they can be
used to predict the terms that may be present. Several examples are presented to illustrate the
approach. The alternate trigonometric and the exponential forms are also presented.

7.1 Wave Analysis

The French mathematician Fourier found that any periodic waveform, that is, a waveform that
repeats itself after some time, can be expressed as a series of harmonically related sinusoids, i.e., sinu-
soids whose frequencies are multiples of a fundamental frequency (or first harmonic). For example, a
series of sinusoids with frequencies 1 MHz , 2 MHz , 3 MHz , and so on, contains the fundamental
frequency of 1 MHz , a second harmonic of 2 MHz , a third harmonic of 3 MHz , and so on. In gen-
eral, any periodic waveform f ( t ) can be expressed as
1
f ( t ) = --- a 0 + a 1 cos ωt + a 2 cos 2ωt + a 3 cos 3ωt + a 4 cos 4ωt + …
2 (7.1)
+ b 1 sin ωt + b 2 sin 2ωt + b 3 sin 3ωt + b 4 sin 4ωt + …

or
∞
1
f ( t ) = --- a 0 +
2 ∑ ( a cos nωt + b sin nωt )
n n (7.2)
n=1

where the first term a 0 ⁄ 2 is a constant, and represents the DC (average) component of f ( t ) . Thus,
if f ( t ) represents some voltage v ( t ) , or current i ( t ) , the term a 0 ⁄ 2 is the average value of v ( t ) or
i(t) .

The terms with the coefficients a 1 and b 1 together, represent the fundamental frequency compo-
nent ω *. Likewise, the terms with the coefficients a 2 and b 2 together, represent the second har-
monic component 2ω , and so on.
Since any periodic waveform f ( t ) ) can be expressed as a Fourier series, it follows that the sum of the

* We recall that k 1 cos ωt + k 2 sin ωt = k cos ( ωt + θ ) where θ is a constant.

Signals and Systems with MATLAB Applications, Second Edition 7-1

Orchard Publications
Chapter 7 Fourier Series

DC , the fundamental, the second harmonic, and so on, must produce the waveform f ( t ) . Generally,
the sum of two or more sinusoids of different frequencies produce a waveform that is not a sinusoid
as shown in Figure 7.1.

Total
2nd Harmonic
Fundamental
3rd Harmonic

Figure 7.1. Summation of a fundamental, second and third harmonic

7.2 Evaluation of the Coefficients

Evaluations of a i and b i coefficients of (7.1) is not a difficult task because the sine and cosine are
orthogonal functions, that is, the product of the sine and cosine functions under the integral evalu-
ated from 0 to 2π is zero. This will be shown shortly.
Let us consider the functions sin mt and cos m t where m and n are any integers, and for conve-
nience, we have assumed that ω = 1 . Then,
2π

∫0 sin mt dt = 0 (7.3)

2π

∫0 cos m t dt = 0 (7.4)

2π

∫0 ( sin mt ) ( cos nt ) dt = 0 (7.5)

The integrals of (7.3) and (7.4) are zero since the net area over the 0 to 2π area is zero. The integral
of (7.5) is also is zero since
1
sin x cos y = --- [ sin ( x + y ) + sin ( x – y ) ]
2
This is also obvious from the plot of Figure 7.2, where we observe that the net shaded area above
and below the time axis is zero.

7-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Evaluation of the Coefficients

sin x cos x

sin x ⋅ cos x

2π
Figure 7.2. Graphical proof of ∫0 ( sin mt ) ( cos nt ) dt = 0

Moreover, if m and n are different integers, then,

2π

∫0 ( sin mt ) ( sin nt ) dt = 0 (7.6)

since
1
( sin x ) ( sin y ) = --- [ cos ( x – y ) – cos ( x – y ) ]
2

The integral of (7.6) can also be confirmed graphically as shown in Figure 7.3, where m = 2 and
n = 3 . We observe that the net shaded area above and below the time axis is zero.

sin 2x sin 3x sin 2x ⋅ sin 3x

2π
Figure 7.3. Graphical proof of ∫0 ( sin mt ) ( sin nt ) dt = 0 for m = 2 and n = 3

Also, if m and n are different integers, then,

Signals and Systems with MATLAB Applications, Second Edition 7-3

Orchard Publications
Chapter 7 Fourier Series

2π

∫0 ( cos m t ) ( cos nt ) dt = 0 (7.7)

since
1
( cos x ) ( cos y ) = --- [ cos ( x + y ) + cos ( x – y ) ]
2

The integral of (7.7) can also be confirmed graphically as shown in Figure 7.4, where m = 2 and
n = 3 . We observe that the net shaded area above and below the time axis is zero.

cos 3x cos 2x cos 2x ⋅ cos 3x

2π
Figure 7.4. Graphical proof of ∫0 ( cos m t ) ( cos nt ) dt = 0 for m = 2 and n = 3

However, if in (7.6) and (7.7), m = n , then,

2π

∫0
2
( sin mt ) dt = π (7.8)

and
2π

∫0
2
( cos m t ) dt = π (7.9)

The integrals of (7.8) and (7.9) can also be seen to be true graphically with the plots of Figures 7.5
and 7.6.
It was stated earlier that the sine and cosine functions are orthogonal to each other. The simplifica-
tion obtained by application of the orthogonality properties of the sine and cosine functions,
becomes apparent in the discussion that follows.

7-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Evaluation of the Coefficients

2
sin x
sin x

2π

∫0
2
Figure 7.5. Graphical proof of ( sin mt ) dt = π

2
cos x

cos x

2π

∫0
2
Figure 7.6. Graphical proof of ( cos m t ) dt = π

In (7.1), for simplicity, we let ω = 1 . Then,

1
f ( t ) = --- a 0 + a 1 cos t + a 2 cos 2t + a 3 cos 3t + a 4 cos 4t + …
2 (7.10)
+ b 1 sin t + b 2 sin 2t + b 3 sin 3t + b 4 sin 4t + …

To evaluate any coefficient, say b 2 , we multiply both sides of (7.10) by sin 2t . Then,

1
f ( t ) sin 2t = --- a 0 sin 2t + a 1 cos t sin 2t + a 2 cos 2t sin 2t + a 3 cos 3t sin 2t + a 4 cos 4t sin 2t + …
2
2
b 1 sin t sin 2t + b 2 ( sin 2t ) + b 3 sin 3t sin 2t + b 4 sin 4t sin 2t + …

Next, we multiply both sides of the above expression by dt , and we integrate over the period 0 to
2π . Then,

Signals and Systems with MATLAB Applications, Second Edition 7-5

Orchard Publications
Chapter 7 Fourier Series

2π 2π 2π 2π
1
∫0 f ( t ) sin 2t dt = --- a 0
2 ∫0 sin 2t dt + a 1 ∫0 cos t sin 2t dt + a 2 ∫0 cos 2t sin 2t dt

2π 2π
+ a3 ∫0 cos 3t sin 2t dt + a 4 ∫0 cos 4t sin 2t dt + …
(7.11)
2π 2π 2π

∫0 ∫0 ∫0
2
+ b1 sin t sin 2t dt + b 2 ( sin 2t ) dt + b 3 sin 3t sin 2t dt

2π
+ b4 ∫0 sin 4t sin 2t dt + …

We observe that every term on the right side of (7.11) except the term
2π

∫0 ( sin 2t ) dt
2
b2

is zero as we found in (7.6) and (7.7). Therefore, (7.11) reduces to

2π 2π

∫0 f ( t ) sin 2t dt = b 2 ∫0 ( sin 2t ) dt = b 2 π
2

or
2π
1
b 2 = ---
π ∫0 f ( t ) sin 2t dt

and thus we can evaluate this integral for any given function f ( t ) . The remaining coefficients can be
evaluated similarly.
The coefficients a 0 , a n , and b n are found from the following relations.

2π
1 1-
--- a 0 = -----
2 2π ∫0 f ( t ) dt (7.12)

2π
1
a n = ---
π ∫0 f ( t ) cos nt dt (7.13)

2π
1
b n = ---
π ∫0 f ( t ) sin nt dt (7.14)

The integral of (7.12) yields the average ( DC ) value of f ( t ) .

7-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Symmetry

7.3 Symmetry
With a few exceptions such as the waveform of Example 7.6, the most common waveforms that are
used in science and engineering, do not have the average, cosine, and sine terms all present. Some
waveforms have cosine terms only, while others have sine terms only. Still other waveforms have or
have not DC components. Fortunately, it is possible to predict which terms will be present in the
trigonometric Fourier series, by observing whether or not the given waveform possesses some kind
of symmetry.
We will discuss three types of symmetry that can be used to facilitate the computation of the trigo-
nometric Fourier series form. These are:
1. Odd symmetry − If a waveform has odd symmetry, that is, if it is an odd function, the series will
consist of sine terms only. In other words, if f ( t ) is an odd function, all the a i coefficients includ-
ing a 0 , will be zero.
2. Even symmetry − If a waveform has even symmetry, that is, if it is an even function, the series will
consist of cosine terms only, and a 0 may or may not be zero. In other words, if f ( t ) is an even
function, all the b i coefficients will be zero.

3. Half-wave symmetry − If a waveform has half-wave symmetry (to be defined shortly), only odd
(odd cosine and odd sine) harmonics will be present. In other words, all even (even cosine and
even sine) harmonics will be zero.
We defined odd and even functions in Chapter 6. We recall that odd functions are those for which
–f ( –t ) = f ( t ) (7.15)
and even functions are those for which
f ( –t ) = f ( t ) (7.16)
Examples of odd and even functions were given in Chapter 6. Generally, an odd function has odd
powers of the independent variable t , and an even function has even powers of the independent
variable t . Thus, the product of two odd functions or the product of two even functions will result
in an even function, whereas the product of an odd function and an even function will result in an
odd function. However, the sum (or difference) of an odd and an even function will yield a function
which is neither odd nor even.
To understand half-wave symmetry, we recall that any periodic function with period T , is expressed
as
f(t) = f(t + T) (7.17)
that is, the function with value f ( t ) at any time t , will have the same value again at a later time t + T .

Signals and Systems with MATLAB Applications, Second Edition 7-7

Orchard Publications
Chapter 7 Fourier Series

A periodic waveform with period T , has half-wave symmetry if

–f ( t + T ⁄ 2 ) = f ( t ) (7.18)
that is, the shape of the negative half-cycle of the waveform is the same as that of the positive half-
cycle, but inverted.
We will test the waveforms of Figures 7.7 through 7.11 for any of the three types of symmetry.
1. Square waveform
For the waveform of Figure 7.7, the average value over one period T is zero, and therefore,
a 0 = 0 . It is also an odd function and has half-wave symmetry since – f ( – t ) = f ( t ) and
–f ( t + T ⁄ 2 ) = f ( t ) .

T
A

T/2
f(b) π 2π
ωt
0
T/2
f(a)
−A

Figure 7.7. Square waveform test for symmetry

Note
An easy method to test for half-wave symmetry is to choose any half-period T ⁄ 2 length on the time
axis as shown in Figure 7.7, and observe the values of f ( t ) at the left and right points on the time
axis, such as f ( a ) and f ( b ) . If there is half-wave symmetry, these will always be equal but will have
opposite signs as we slide the half-period T ⁄ 2 length to the left or to the right on the time axis at
non-zero values of f ( t ) .
2. Square waveform with ordinate axis shifted
If we shift the ordinate axis π ⁄ 2 radians to the right, as shown in Figure 7.8, we see that the
square waveform now becomes an even function and has half-wave symmetry since f ( – t ) = f ( t )
and – f ( t + T ⁄ 2 ) = f ( t ) . Also, a 0 = 0 .

Obviously, if the ordinate axis is shifted by any other value other than an odd multiple of π ⁄ 2 ,
the waveform will have neither odd nor even symmetry.

7-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Symmetry

−π/2 π/2 2π
−2π −π π ωt
0

T/2 T/2
−A

Figure 7.8. Square waveform with ordinate shifted by π ⁄ 2

3. Sawtooth waveform
T
A

−2π −π π 2π
ωt
0
T/2 T/2
−A

Figure 7.9. Sawtooth waveform test for symmetry

For the sawtooth waveform of Figure 7.9, the average value over one period T is zero and there-
fore, a 0 = 0 . It is also an odd function because – f ( – t ) = f ( t ) , but has no half-wave symmetry
since – f ( t + T ⁄ 2 ) ≠ f ( t )
4. Triangular waveform

−2π
−π ωt
0 π 2π
T/2
−A T/2

Figure 7.10. Triangular waveform test for symmetry

Signals and Systems with MATLAB Applications, Second Edition 7-9

Orchard Publications
Chapter 7 Fourier Series

For this triangular waveform of Figure 7.10, the average value over one period T is zero and
therefore, a 0 = 0 . It is also an odd function since – f ( – t ) = f ( t ) . Moreover, it has half-wave
symmetry because – f ( t + T ⁄ 2 ) = f ( t )
5. Fundamental, Second and Third Harmonics of a Sinusoid
Figure 7.11 shows a fundamental, second, and third harmonic of a typical sinewave.

Τ/2 Τ/2 Τ/2

c
a
b b

−a
−c
Figure 7.11. Fundamental, second, and third harmonic test for symmetry

In Figure 7.11, the half period T ⁄ 2 , is chosen as the half period of the period of the fundamental
frequency. This is necessary in order to test the fundamental, second, and third harmonics for half-
wave symmetry. The fundamental has half-wave symmetry since the a and – a values, when sepa-
rated by T ⁄ 2 , are equal and opposite. The second harmonic has no half-wave symmetry because the
ordinates b on the left and b on the right, although are equal, there are not opposite in sign. The
third harmonic has half-wave symmetry since the c and – c values, when separated by T ⁄ 2 are
equal and opposite. These waveforms can be either odd or even depending on the position of the
ordinate. Also, all three waveforms have zero average value unless the abscissa axis is shifted up or
down.
In the expressions of the integrals in (7.12) through (7.14), the limits of integration for the coeffi-
cients a n and b n are given as 0 to 2π , that is, one period T . Of course, we can choose the limits of
integration as – π to +π . Also, if the given waveform is an odd function, or an even function, or has
half-wave symmetry, we can compute the non-zero coefficients a n and b n by integrating from 0 to
π only, and multiply the integral by 2 . Moreover, if the waveform has half-wave symmetry and is
also an odd or an even function, we can choose the limits of integration from 0 to π ⁄ 2 and multiply
the integral by 4 . The proof is based on the fact that, the product of two even functions is another
even function, and also that the product of two odd functions results also in an even function. How-
ever, it is important to remember that when using these shortcuts, we must evaluate the coefficients
a n and b n for the integer values of n that will result in non-zero coefficients. This point will be
illustrated in Example 7.2.

7-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series

We will now derive the trigonometric Fourier series of the most common periodic waveforms.

7.4 Waveforms in Trigonometric Form of Fourier Series

Example 7.1
Compute the trigonometric Fourier series of the square waveform of Figure 7.12.

T
A

π 2π
ωt
0

−A

Figure 7.12. Square waveform for Example 7.1

Solution:
The trigonometric series will consist of sine terms only because, as we already know, this waveform
is an odd function. Moreover, only odd harmonics will be present since this waveform has half-wave
symmetry. However, we will compute all coefficients to verify this. Also, for brevity, we will assume
that ω = 1 .
The a i coefficients are found from
2π π 2π
1 1 A π
∫0 ∫0 ∫π
2π
a n = --- f ( t ) cos nt dt = --- A cos nt dt + ( – A ) cos nt dt = ------ ( sin nt 0 – sin nt )
π π nπ π
(7.19)
A A
= ------ ( sin nπ – 0 – sin n2π + sin nπ ) = ------ ( 2 sin nπ – sin n2π )
nπ nπ

and since n is an integer (positive or negative) or zero, the terms inside the parentheses on the sec-
ond line of (7.19) are zero and therefore, all a i coefficients are zero, as expected since the square
waveform has odd symmetry. Also, by inspection, the average ( DC ) value is zero, but if we attempt
to verify this using (7.19), we will get the indeterminate form 0 ⁄ 0 . To work around this problem, we
will evaluate a 0 directly from (7.12). Then,
π 2π
1 A
a 0 = ---
π ∫0 A dt + ∫π ( – A ) dt = --- ( π – 0 – 2π + π ) = 0
π
(7.20)

Signals and Systems with MATLAB Applications, Second Edition 7-11

Orchard Publications
Chapter 7 Fourier Series

The b i coefficients are found from (7.14), that is,

2π π 2π
1 1 A π
∫0 ∫0 ∫π
2π
b n = --- f ( t ) sin nt dt = --- A sin nt dt + ( – A ) sin nt dt = ------ ( – cos n t 0 + cos nt )
π π nπ π
(7.21)
A A
= ------ ( – cos nπ + 1 + cos 2nπ – cos nπ ) = ------ ( 1 – 2 cos nπ + cos 2nπ )
nπ nπ

For n = even , (7.21) yields

A
b n = ------ ( 1 – 2 + 1 ) = 0
nπ
as expected, since the square waveform has half-wave symmetry.
For n = odd , (7.21) reduces to
A 4A
b n = ------ ( 1 + 2 + 1 ) = -------
nπ nπ
and thus
b 1 = 4A
-------
π

4A
b 3 = -------
3π
4A
b 5 = -------
5π
and so on.
Therefore, the trigonometric Fourier series for the square waveform with odd symmetry is

f ( t ) = ------- ⎛ sin ωt + --- sin 3ωt + --- sin 5ωt + …⎞ = -------

4A 1 1 4A 1
π ⎝ 3 5 ⎠ π ∑ --- sin nωt
n
(7.22)
n = odd

It was stated above that, if the given waveform has half-wave symmetry, and it is also an odd or an
even function, we can integrate from 0 to π ⁄ 2 , and multiply the integral by 4 . We will apply this
property to the following example.

Example 7.2
Compute the trigonometric Fourier series of the square waveform of Example 1 by integrating from
0 to π ⁄ 2 , and multiplying the result by 4 .
Solution:
Since the waveform is an odd function and has half-wave symmetry, we are only concerned with the
odd b n coefficients. Then,

7-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series
π⁄2
π
) = ------- ⎛ – cos n --- + 1⎞
1 4A π⁄2 4A
b n = 4 ---
π ∫0 f ( t ) sin nt dt = ------- ( – cos n t
nπ 0 nπ ⎝ 2 ⎠
(7.23)

For n = odd , (7.23) becomes

4A
b n = ------- ( – 0 + 1 ) = 4A
------- (7.24)
nπ nπ
as before, and thus the series is the same as in Example 1.

Example 7.3
Compute the trigonometric Fourier series of the square waveform of Figure 7.13.
Solution:
This is the same waveform as in Example 7.1, except that the ordinate has been shifted to the right
by π ⁄ 2 radians, and has become an even function. However, it still has half-wave symmetry. There-
fore, the trigonometric Fourier series will consist of odd cosine terms only.
Since the waveform has half-wave symmetry and is an even function, it will suffice to integrate from
0 to π ⁄ 2 , and multiply the integral by 4 . The a n coefficients are found from

π⁄2 π⁄2
π
= ------- ⎛ sin n ---⎞
1 4 4A π⁄2 4A
a n = 4 ---
π ∫0 f ( t ) cos nt dt = ---
π ∫0 A cos nt dt = ------- ( sin nt
nπ 0
)
nπ ⎝ 2⎠
(7.25)

We observe that for n = even , all a n coefficients are zero, and thus all even harmonics are zero as
expected. Also, by inspection, the average ( DC ) value is zero.

T
A

π/2 3π / 2
π
ωt
0 2π

−A

Figure 7.13. Waveform for Example 7.3

π
For n = odd , we observe from (7.25) that sin n --- , will alternate between +1 and – 1 depending on
2
the odd integer assigned to n . Thus,

Signals and Systems with MATLAB Applications, Second Edition 7-13

Orchard Publications
Chapter 7 Fourier Series

4A
a n = ± ------- (7.26)
nπ

For n = 1, 5, 9, 13 , and so on, (7.26) becomes

4A
a n = -------
nπ
and for n = 3, 7, 11, 15 , and so on, it becomes
– 4A
a n = ----------
nπ
Then, the trigonometric Fourier series for the square waveform with even symmetry is

(n – 1)
----------------
f ( t ) = ------- ⎛ cos ω t – --- cos 3ωt + --- cos 5ωt – …⎞ = -------
4A 1 4A 2 1
∑
1
( –1 ) --- cos n ωt (7.27)
π⎝ 3 5 ⎠ π n
n = odd

Alternate Solution:
Since the waveform of Example 7.3 is the same as of Example 7.1, but shifted to the right by π ⁄ 2
radians, we can use the result of Example 7.11, i.e.,

f ( t ) = ------- ⎛ sin ωt + --- sin 3ωt + --- sin 5ωt + …⎞

4A 1 1
(7.28)
π ⎝ 3 5 ⎠

and substitute ωt with ωt + π ⁄ 2 , that is, we let ωt = ωτ + π ⁄ 2 . With this substitution, relation
(7.28) becomes

---⎞ + --- sin 5 ⎛ ωτ + π

---⎞ + --- sin 3 ⎛ ωτ + π
f ( τ ) = ------- sin ⎛ ωτ + π ---⎞ + …
4A 1 1
π ⎝ 2 ⎠ 3 ⎝ 2 ⎠ 5 ⎝ 2⎠
(7.29)
= ------- sin ⎛ ωτ + π ------⎞ + --- sin ⎛ 5ωτ + 5π
---⎞ + --- sin ⎛ 3ωτ + 3π ------⎞ + …
4A 1 1
π ⎝ 2⎠ 3 ⎝ 2⎠ 5 ⎝ 2⎠

and using the identities sin ( x + π ⁄ 2 ) = cos x , sin ( x + 3π ⁄ 2 ) = – cos x , and so on, we rewrite (7.29)
as

4A 1 1
f ( τ ) = ------- cos ωτ – --- cos 3ωτ + --- cos 5ωτ – … (7.30)
π 3 5

and this is the same as (7.22).

Therefore, if we compute the trigonometric Fourier series with reference to one ordinate, and after-
wards we want to recompute the series with reference to a different ordinate, we can use the above
procedure to save time.

7-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series

Example 7.4
Compute the trigonometric Fourier series of the sawtooth waveform of Figure 7.14.
T

−π π
ωt
−2π 0 2π

−A

Figure 7.14. Sawtooth waveform

Solution:
This waveform is an odd function but has no half-wave symmetry; therefore, it contains sine terms
only with both odd and even harmonics. Accordingly, we only need to evaluate the b n coefficients.

By inspection, the DC component is zero. As before, we will assume that ω = 1 .

If we choose the limits of integration from 0 to 2π , we will need to perform two integrations since

⎧ A --- t 0<t<π
⎪ π
f(t) = ⎨
⎪A--- t – 2A π < t < 2π
⎩π

However, we can choose the limits from – π to +π , and thus we will only need one integration since
A
f ( t ) = --- t –π < t < π
π

Better yet, since the waveform is an odd function, we can integrate from 0 to π , and multiply the
integral by 2 ; this is what we will do.
From tables of integrals,
1
∫ x sin ax dx = ----2- sin a x – --x- cos ax
a a
(7.31)

Then,

Signals and Systems with MATLAB Applications, Second Edition 7-15

Orchard Publications
Chapter 7 Fourier Series

π π π
t sin nt dt = ------2- ⎛ ----2- sin nt – --- cos nt⎞
2 A
--- t sin nt dt = 2A 2A 1 t
b n = ---
π ∫0 π
-------
π2 ∫0 π ⎝n n ⎠
0 (7.32)
2A π 2A
- ( sin nt – nt cos nt )
= ---------- - ( sin nπ – nπ cos nπ )
= ----------
n2 π2 0 n2 π2
We observe that:
1. If n = even , sin nπ = 0 and cos nπ = 1 . Then, (7.32) reduces to
2A
- ( – nπ ) = – 2A
b n = ---------- -------
n2 π2 nπ
that is, the even harmonics have negative coefficients.
2. If n = odd , sin nπ = 0 , cos nπ = – 1 . Then,
2A
- ( nπ ) = 2A
b n = ---------- -------
n2 π2 nπ
that is, the odd harmonics have positive coefficients.
Thus, the trigonometric Fourier series for the sawtooth waveform with odd symmetry is

f ( t ) = ------- ⎛ sin ωt – --- sin 2ωt + --- sin 3ωt – --- sin 4ωt + …⎞ = -------
2A 1 1 1 2A n–1 1
π⎝ 2 3 4 ⎠ π ∑ ( –1 ) --- sin nωt
n
(7.33)

Example 7.5
Find the trigonometric Fourier series of the triangular waveform of Figure 7.15. Assume ω = 1 .

T
A

−2π
−π ωt
0 π/2 π 2π
−A

Figure 7.15. Triangular waveform for Example 7.5

Solution:
This waveform is an odd function and has half-wave symmetry; then, the trigonometric Fourier
series will contain sine terms only with odd harmonics. Accordingly, we only need to evaluate the b n
coefficients. We will choose the limits of integration from 0 to π ⁄ 2 , and will multiply the integral
by 4 .

7-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series

By inspection, the DC component is zero. From tables of integrals,

1
∫ x sin ax dx
x
= -----2 sin a x – --- cos ax (7.34)
a a
Then,
π⁄2 π⁄2 π⁄2
t sin nt dt = ------2- ⎛ -----2 sin nt – --- cos nt⎞
4 2A 8A 8A 1 t
b n = ---
π ∫0 ------- t sin nt dt = ------2-
π π ∫0 π n ⎝ n ⎠
0
(7.35)
π π π
= ----------2 ⎛ sin n --- – n --- cos n --- ⎞
8A π⁄2 8A
= ----------2 ( sin nt – nt cos nt ) 0 ⎝ 2 2 2⎠
n π n π
2 2

We are only interested in the odd integers of n , and we observe that:

π
cos n --- = 0
2

For odd integers of n , the sine term yields

⎧ 1 for n = 1, 5, 9, … then, b = ----------

8A
-
⎪ n
2 2
π ⎪ n π
sin n --- = ⎨
2 ⎪ – 1 for n = 3, 7, 11, … then, b = – ----------
8A -
⎪ n
2 2
⎩ n π
Thus, the trigonometric Fourier series for the triangular waveform with odd symmetry is

(n – 1)
----------------
f ( t ) = ------2- ⎛ sin ω t – 1
--- sin 3ωt + ------ sin 5ωt – ------ sin 7ωt + …⎞ = ------2-
8A 1 1 8A 1-
∑
2
( –1 ) ---- sin n ωt (7.36)
⎝ 25 49 ⎠ 2
π 9 π n = odd n

Example 7.6
The circuit of Figure 7.16 is a half-wave rectifier whose input is the sinusoid v in ( t ) = sin ωt , and its
output v out ( t ) is defined as

⎧ sin ωt 0 < ωt < π

v out ( t ) = ⎨ (7.37)
⎩ 0 π < ωt < 2π

Express v out ( t ) as a trigonometric Fourier series. Assume ω = 1 .

Solution:
The input and output waveforms for this example are shown in Figures 7.17 and 7.18.

Signals and Systems with MATLAB Applications, Second Edition 7-17

Orchard Publications
Chapter 7 Fourier Series

+
+ v out ( t )
v in ( t ) R
− −

Figure 7.16. Circuit for half-wave rectifier

Figure 7.17. Input v in ( t ) for the circuit of Figure 7.16

Figure 7.18. Output v out ( t ) for the circuit of Figure 7.16

We choose the ordinate as shown in Figure 7.19.

−2π −π 0 π 2π

Figure 7.19. Half-wave rectifier waveform for the circuit of Figure 7.16

By inspection, the average is a non-zero value, and the waveform has neither odd nor even symme-
try. Therefore, we expect all terms to be present.

7-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series

The a n coefficients are found from

2π
1
a n = ---
π ∫0 f ( t ) cos nt dt

For this example,

π 2π
A A
a n = ---
π ∫0 sin t cos nt dt + ---
π ∫π 0 cos nt dt

and from tables of integrals

cos ( m – n )x cos ( m + n )x
∫ ( sin mx ) ( cos nx ) dx
2 2
= – ------------------------------ – ------------------------------ ( m ≠ n )
2(m – n) 2(m + n)

Then,
π
A ⎧ 1 cos ( 1 – n )t cos ( 1 + n )t ⎫
a n = --- ⎨ – --- --------------------------- + ---------------------------- ⎬
π ⎩ 2 1–n 1+n
0⎭
(7.38)
A- ⎧ ----------------------------
= – ----- cos ( π + nπ )- – -----------
cos ( π – nπ -) + ----------------------------- 1 - ⎫
1 - + -----------
⎨ ⎬
2π ⎩ 1–n 1+n 1–n n+1 ⎭

Using the trigonometric identities

cos ( x – y ) = cos x cos y + sin xsiny
and
cos ( x + y ) = cos x cos y – sin x sin y
we get
cos ( π – nπ ) = cos π cos nπ + sin π sin nπ = – cos nπ
and
cos ( π + nπ ) = cos π cos nπ – sin π sin nπ = – cos nπ

Then, by substitution into (7.38),

A ⎧ – cos nπ – cos nπ 2 ⎫ A ⎧ cos nπ cos nπ 2 ⎫

a n = – ------ ⎨ ------------------ + ------------------ – --------------2 ⎬ = ------ ⎨ --------------- + --------------- + --------------2 ⎬
2π ⎩ 1 – n 1+n 1–n ⎭ 2π ⎩ 1 – n 1+n 1–n ⎭
(7.39)
A cos nπ + n cos nπ + cos nπ – n cos nπ cos nπ + 1- ⎞
= ------ ⎛ --------------------------------------------------------------------------------------- 2 -⎞ = A
+ ------------- --- ⎛ ------------------------ n≠1
2π ⎝ 1–n
2
1–n
2⎠ π ⎝ (1 – n ) ⎠
2

Next, we can evaluate all the a n coefficients, except a 1 , from (7.39).

First, we will evaluate a 0 to obtain the DC value. By substitution of n = 0 , we get

Signals and Systems with MATLAB Applications, Second Edition 7-19

Orchard Publications
Chapter 7 Fourier Series

a 0 = 2A ⁄ π
Therefore, the DC value is
1
--- a 0 = A
--- (7.40)
2 π

We cannot use (7.39) to obtain the value of a 1 ; therefore, we will evaluate the integral
π
A
a 1 = ---
π ∫0 sin t cos t dt
From tables of integrals,
1
∫ ( sin ax ) ( cos ax ) dx
2
= ------ ( sin ax )
2a
and thus,
π
A 2
a 1 = ------ ( sin t ) = 0 (7.41)
2π 0

From (7.39) with n = 2, 3, 4, 5, … , we get

A cos 2π + 1 ⎞
a 2 = --- ⎛ ------------------------
- = – 2A
------- (7.42)
π ⎝ ( 1 – 22 ) ⎠ 3π

A ( cos 3π + 1 ) = 0
a 3 = --------------------------------- (7.43)
2
π(1 – 3 )

We see that for odd integers of n, a n = 0 . However, for n = even , we get

A ( cos 4π + 1 ) 2A
a 4 = --------------------------------- = – --------- (7.44)
2 15π
π(1 – 4 )

A ( cos 6π + 1 ) 2A
a 6 = --------------------------------- = – --------- (7.45)
2 35π
π(1 – 6 )

A ( cos 8π + 1 ) 2A
a 8 = --------------------------------- = – --------- (7.46)
2 63π
π(1 – 8 )
and so on.
Now, we need to evaluate the bn coefficients. For this example,
2π π 2π
1 A A
b n = A ---
π ∫0 f ( t ) sin nt dt = ---
π ∫0 sin t sin nt dt + ---
π ∫π 0 sin nt dt

7-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series

and from tables of integrals,

sin ( m + n )x
sin ( m – n )x- – ----------------------------
∫ ( sin mx ) ( sin nx ) dx = ----------------------------
2( m – n) 2( m + n)
- ( m2 ≠ n2 )

Therefore,
π
A 1 ⎧ sin ( 1 – n )t sin ( 1 + n )t ⎫
b n = --- ⋅ --- ⎨ -------------------------- – --------------------------- ⎬
π 2⎩ 1–n 1+n
0⎭

A sin ( 1 – n )π sin ( 1 + n )π
= ------ ---------------------------- – ---------------------------- – 0 + 0 = 0 ( n ≠ 1 )
2π 1–n 1+n

that is, all the b n coefficients, except b 1 , are zero.

We will find b 1 by direct substitution into (7.14) for n = 1 . Thus,

π π
A A t sin 2t A π sin 2π
∫0
2 A
b 1 = --- ( sin t ) dt = --- --- – ------------ = --- --- – -------------- = --- (7.47)
π π 2 4 π 2 4 2
0

Combining (7.40), and (7.42) through (7.47), we find that the trigonometric Fourier series for the
half-wave rectifier with no symmetry is

A A A cos 2t cos 4t cos 6t cos 8t

f ( t ) = --- + --- sin t – --- ------------- + ------------- + ------------- + ------------- + … (7.48)
π 2 π 3 15 35 63

Example 7.7
Figure 7.20 shows a full-wave rectifier circuit with input the sinusoid v in ( t ) = A sin ωt . The output
of that circuit is v out ( t ) = A sin ωt . Express v out ( t ) as a trigonometric Fourier series. Assume
ω = 1.

+ R
v in ( t ) − +
− +
v out ( t )
−

Figure 7.20. Full-wave rectifier circuit

Solution:
The input and output waveforms are shown in Figures 7.21 and 7.22. We choose the ordinate as
shown in Figure 7.23. By inspection, the average is a non-zero value. We choose the period of the
input sinusoid so that the output will be expressed in terms of the fundamental frequency. We also
choose the limits of integration as – π and +π , we observe that the waveform has even symmetry.

Signals and Systems with MATLAB Applications, Second Edition 7-21

Orchard Publications
Chapter 7 Fourier Series

Therefore, we expect only cosine terms to be present. The a n coefficients are found from

vIN(t)

π 2π

-A

Figure 7.21. Input sinusoid for the circuit of Example 7.7

A
vOUT(t)

π 2π

Figure 7.22. Output waveform for the circuit of Example 7.7

−2π −π 0 π 2π

Figure 7.23. Full-wave rectified waveform with even symmetry

2π
1
a n = ---
π ∫0 f ( t ) cos nt dt

where for this example,

π π
1 2A
a n = ---
π ∫ –π
A sin t cos nt dt = -------
π ∫0 sin t cos nt dt (7.49)

and from tables of integrals,

7-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Waveforms in Trigonometric Form of Fourier Series

cos ( m + n )x
cos ( m – n )x- – -----------------------------
∫ ( sin mx ) ( cos nx ) dx = -----------------------------
2(n – m) 2(m + n)
- ( m2 ≠ n2 )

Since
cos ( x – y ) = cos ( y – x ) = cos x cos y + sin xsiny
we express (7.49) as
π
2A 1 ⎧ cos ( n – 1 )t cos ( n + 1 )t ⎫
a n = ------- ⋅ --- ⎨ --------------------------- – ---------------------------- ⎬
π 2⎩ n–1 n+1
0⎭

A ⎧ cos ( n – 1 )π cos ( n + 1 )π 1 1 ⎫ (7.50)

= --- ⎨ ----------------------------- – ----------------------------- – ------------ – ------------ ⎬
π⎩ n–1 n+1 n–1 n+1 ⎭

A 1 – cos ( nπ + π ) cos ( nπ – π ) – 1
= --- --------------------------------------- + --------------------------------------
π n+1 n–1

To simplify the last expression in (7.50), we make use of the trigonometric identities
cos ( nπ + π ) = cos nπ cos π – sin nπsinπ = – cos nπ
and
cos ( nπ – π ) = cos nπ cos π + sin nπsinπ = – cos nπ
Then, (7.50) simplifies to
A 1 + cos nπ 1 + cos nπ A – 2 + ( n – 1 ) cos nπ – ( n + 1 ) cos nπ
- – ------------------------- = --- -------------------------------------------------------------------------------------
a n = --- ------------------------
π n+1 n–1 π 2
n –1
(7.51)
– 2A ( cos nπ + 1 -)
= ---------------------------------------
2
n≠1
π(n – 1)

Now, we can evaluate all the a n coefficients, except a 1 , from (7.51). First, we will evaluate a 0 to
obtain the DC value. By substitution of n = 0 , we get
4A
a 0 = -------
π
Therefore, the DC value is
1
--- a 0 = 2A
------- (7.52)
2 π

From (7.51) we observe that for all n = odd , other than n = 1 , a n = 0 .

To obtain the value of a 1 , we must evaluate the integral

Signals and Systems with MATLAB Applications, Second Edition 7-23

Orchard Publications
Chapter 7 Fourier Series

π
1
a 1 = ---
π ∫0 sin t cos t dt
From tables of integrals,
1
∫ ( sin ax ) ( cos ax ) dx
2
= ------ ( sin ax )
2a
and thus,
π
1 2
a 1 = ------ ( sin t ) = 0 (7.53)
2π 0

For n = even , from (7.51) we get

– 2A ( cos 2π + 1 ) 4A
a 2 = ---------------------------------------- = – ------- (7.54)
2 3π
π(2 – 1)

2A ( cos 4π + 1 )- = – --------
a 4 = –--------------------------------------- 4A- (7.55)
2 15π
π(4 – 1)

2A ( cos 6π + 1 )- = – --------
a 6 = –--------------------------------------- 4A- (7.56)
2 35π
π(6 – 1)

– 2A ( cos 8π + 1 ) 4A
a 8 = ---------------------------------------- = – --------- (7.57)
2 63π
π(8 – 1)
and so on. Then, combining the terms of (7.52) and (7.54) through (7.57) we get

4A ⎧ cos 2ωt cos 4ωt cos 6ωt cos 8ωt ⎫

f ( t ) = 2A - + ------------------ + ------------------ + ------------------ + … ⎬
------- – ------- ⎨ ----------------- (7.58)
π π⎩ 3 15 35 63 ⎭

Therefore, the trigonometric form of the Fourier series for the full-wave rectifier with even symmetry
is
∞
4A 1 -
f ( t ) = 2A
------- – -------
π π ∑ ------------------
( n
2
– 1 )
cos nωt (7.59)
n = 2 , 4, 6 , …

This series of (7.59) shows that there is no component of the input (fundamental) frequency. This is
because we chose the period to be from – π and +π . Generally, the period is defined as the shortest
period of repetition. In any waveform where the period is chosen appropriately, it is very unlikely
that a Fourier series will consist of even harmonic terms only.

7.5 Gibbs Phenomenon

In Example 7.1, we found that the trigonometric form of the Fourier series of the square waveform

7-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Alternate Forms of the Trigonometric Fourier Series

is
f ( t ) = ------- ⎛ sin ωt + --- sin 3ωt + --- sin 5ωt + …⎞ = -------
4A 1 1 4A 1
π⎝ 3 5 ⎠ π ∑ --- sin nωt
n
n = odd

Figure 7.24 shows the first 11 harmonics and their sum. We observe that as we add more and more
harmonics, the sum looks more and more like the square waveform. However, the crests do not get
flattened; this is known as Gibbs phenomenon and it occurs because of the discontinuity of the per-
fect square waveform as it changes from +A to – A .
Crest (Gibbs Phenomenon)

Sum of first 11 harmonics

Figure 7.24. Gibbs phenomenon

7.6 Alternate Forms of the Trigonometric Fourier Series

We recall that the trigonometric Fourier series is expressed as
1
f ( t ) = --- a 0 + a 1 cos ωt + a 2 cos 2ωt + a 3 cos 3ωt + a 4 cos 4ωt + …
2 (7.60)
+ b 1 sin ωt + b 2 sin 2ωt + b 3 sin 3ωt + b 4 sin 4ωt + …

If a given waveform does not have any kind of symmetry, it may be advantageous of using the alter-
nate form of the trigonometric Fourier series where the cosine and sine terms of the same frequency
are grouped together, and the sum is combined to a single term, either cosine or sine. However, we
still need to compute the a n and b n coefficients separately.

We use the triangle shown in Figure 7.25 for the derivation of the alternate forms.

cn = an + bn
ϕn
cn an a bn b
bn cos θ n = --------------------
- = ----n- sin θ n = --------------------
- = ----n-
an + bn cn an + bn cn
θn
an
b a
cos θ n = sin ϕ n θ n = atan ----n- ϕ n = atan ----n-
an bn

Figure 7.25. Derivation of the alternate form of the trigonometric Fourier series

Signals and Systems with MATLAB Applications, Second Edition 7-25

Orchard Publications
Chapter 7 Fourier Series

We assume ω = 1 , and for n = 1, 2, 3, … , we rewrite (7.60) as

a b a b
f ( t ) = --- a 0 + c 1 ⎛ -----1 cos t + -----1 sin t⎞ + c 2 ⎛ -----2 cos 2t + -----2 sin 2t⎞ + …
1
2 ⎝ c1 c1 ⎠ ⎝ c2 c2 ⎠
an bn
+ c n ⎛ ----- cos nt + ----- sin nt⎞
⎝ cn cn ⎠

1 cos θ 1 cos t + sin θ 1 sin t cos θ 2 cos 2t + sin θ 2 sin 2t

= --- a 0 + c 1 ⎛ ⎧
⎪ ⎞ + c2 ⎛ ⎞ +…
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩

⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
2 ⎝ cos ( t – θ 1 ) ⎠ ⎝ cos ( 2t – θ 2 ) ⎠
cos θ n cos nt + sin θ n sin nt
+ cn ⎛ ⎞
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
⎝ cos ( nt – θ n ) ⎠

and, in general, for ω ≠ 1 , we get

∞ ∞ bn
∑ cn cos ⎛⎝ nωt – atan ----
-⎞
1 1
f ( t ) = --- a 0 +
2 ∑ cn cos ( nωt – θn ) = --- a 0 +
2 a n⎠
(7.61)
n=1 n=1

Similarly,
1 sin ϕ 1 cos t + cos ϕ 1 sin t
f ( t ) = --- a 0 + c 1 ⎛ ⎞
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩

2 ⎝ sin ( t + ϕ 1 ) ⎠
sin ϕ 2 cos 2t + cos ϕ 2 sin 2t sin ϕ n cos nt + cos ϕ n sin nt
c2 ⎛ ⎞ + … + cn ⎛ ⎞
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩

⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
⎝ sin ( 2t + ϕ 2 ) ⎠ ⎝ sin ( nt + ϕ n ) ⎠

and, in general, where ω ≠ 1 , we get

∞ ∞ an
∑ cn sin ⎛⎝ nωt + atan ----
-⎞
1 1
f ( t ) = --- a 0 +
2 ∑ c n sin ( nωt + ϕ n ) = --- a 0 +
2 b n⎠
(7.62)
n=1 n=1

When used in circuit analysis, (7.61) and (7.62) can be expressed as phasors. Since it is customary to
use the cosine function in the time domain to phasor transformation, we choose to use the transfor-
mation of (7.63) below.

∞ b ∞ bn
c n cos ⎛ nωt – atan ----n-⎞ ⇔ --- a 0 +
1 1
--- a 0 +
2 ∑ ⎝ an ⎠ 2 ∑ cn ∠– atan ----
an
- (7.63)
n=1 n=1

Example 7.8
Find the first 5 terms of the alternate form of the trigonometric Fourier series for the waveform of
Figure 7.26.

7-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Alternate Forms of the Trigonometric Fourier Series

f( t)
3

1
ω = 1
t
π/2 π 3π/2 2π

Figure 7.26. Waveform for Example 7.8

Solution:
The given waveform has no symmetry; thus, we expect both cosine and sine functions with odd and
even terms present. Also, by inspection the DC value is not zero.
We will compute the a n and b n coefficients, the DC value, and we will combine them to get an
expression in the form of (7.63). Then,
π⁄2 2π π⁄2 2π
1 1 3 1
a n = ---
π ∫0 ( 3 ) cos nt dt + ---
π ∫ π⁄2
( 1 ) cos nt dt = ------ sin nt
nπ 0
+ ------ sin nt
nπ π⁄2 (7.64)
3 π 1 1 π 2 π
= ------ sin n --- + ------ sin n2π – ------ sin n --- = ------ sin n ---
nπ 2 nπ nπ 2 nπ 2

We observe that for n = even , a n = 0 .

For n = odd ,

2
a 1 = --- (7.65)
π
and

2
a 3 = – ------ (7.66)
3π

The DC value is
π⁄2 2π
1 1- 1 1 π⁄2
∫0 ∫π ⁄ 2 ( 1 ) d t
2π
--- a 0 = ----- ( 3 ) dt + ------ = ------ ( 3t 0
+t π⁄2
)
2 2π 2π 2π
(7.67)
π
= ------ ⎛ ------ + 2π – ---⎞ = ------ ( π + 2π ) = ---
1 3π 1 3
2π ⎝ 2 2⎠ 2π 2

The b n coefficients are

Signals and Systems with MATLAB Applications, Second Edition 7-27

Orchard Publications
Chapter 7 Fourier Series

π⁄2 2π π⁄2 2π
1 1 –3 –1
b n = ---
π ∫0 ( 3 ) sin nt dt + ---
π ∫ π⁄2
( 1 ) sin nt dt = ------ cos nt
nπ 0
+ ------ cos nt
nπ π⁄2 (7.68)
–3 π 3 –1 1 π 1 2
= ------ cos n --- + ------ + ------ cos n2π + ------ cos n --- = ------ ( 3 – cos n2π ) = ------
nπ 2 nπ nπ nπ 2 nπ nπ
Then,
b1 = 2 ⁄ π (7.69)

b2 = 1 ⁄ π (7.70)

b 3 = 2 ⁄ 3π (7.71)

b 4 = 1 ⁄ 2π (7.72)
From (7.63),
∞ b ∞ bn
c n cos ⎛ nωt – atan ----n-⎞ ⇔ --- a 0 +
1 1
--- a 0 +
2 ∑ ⎝ an ⎠ 2 ∑ cn ∠– atan ----
an
-
n=1 n=1
where
b 2 2 b 2 2
c n ∠– atan ----n- = a n + b n ∠– atan ----n- = a n + b n ∠– θ n = a n – jb n (7.73)
an an

Thus, for n = 1, 2, 3, and 4 , we get:

2 2
2 2
a 1 – jb 1 = --- – j --- = ⎛ --2-⎞ + ⎛ --2-⎞ ∠– 45°
π π ⎝ π⎠ ⎝ π⎠
(7.74)
8 2 2 2 2
= ------ ∠– 45° = ---------- ∠– 45° ⇔ ---------- cos ( ωt – 45° )
2 π π
π
Similarly,
1 1 1
a 2 – jb 2 = 0 – j --- = --- ∠– 90° ⇔ --- cos ( 2ωt – 90° ) (7.75)
π π π

2 2 2 2 2 2
a 3 – jb 3 = – ------ – j ------ = ---------- ∠– 135° ⇔ ---------- cos ( 3ωt – 135° ) (7.76)
3π 3π 3π 3π
and
1 1 1
a 4 – jb 4 = 0 – j ------ = ------ ∠– 90° ⇔ ------ cos ( 4ωt – 90° ) (7.77)
2π 2π 2π
Combining the terms of (7.67) and (7.74) through (7.77), we find that the alternate form of the trig-
onometric Fourier series representing the waveform of this example is

7-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Circuit Analysis with Trigonometric Fourier Series

3 1
f ( t ) = --- + ---
2 π
[2 2 cos ( ωt – 45° ) + cos ( 2ωt – 90° )
(7.78)
2 2 1
+ ---------- cos ( 3ωt – 135° ) + --- cos ( 4ωt – 90° ) + …
3 2
]

7.7 Circuit Analysis with Trigonometric Fourier Series

When the excitation of an electric circuit is a non-sinusoidal waveform such as those we discussed
thus far, we can use Fouries series to determine the response of a circuit. The procedure is illustrated
with the examples that follow.

Example 7.9
The input to the series RC circuit of Figure 7.27, is the square waveform of Figure 7.28. Compute
the voltage v c ( t ) across the capacitor. Consider only the first three terms of the series, and assume
ω = 1.

R=1Ω
+
+ vC ( t )
− −
v in ( t ) C =1F

Figure 7.27. Circuit for Example 7.9

v in ( t ) T
A

π 2π ωt
0

−A

Figure 7.28. Input waveform for the circuit of Figure 7.27

Solution:
In Example 7.1, we found that the above square waveform can be represented by the trigonometric
Fourier series as

f ( t ) = ------- ⎛ sin ωt + --- sin 3ωt + --- sin 5ωt + …⎞

4A 1 1
⎝ ⎠
(7.79)
π 3 5

Signals and Systems with MATLAB Applications, Second Edition 7-29

Orchard Publications
Chapter 7 Fourier Series

Since this series is the sum of sinusoids, we will use phasor analysis to obtain the solution.
The equivalent phasor circuit is shown in Figure 7.29.

R=1
C +
+ VC
− 1- −
-----
V in jω

Figure 7.29. Phasor circuit for Example 7.9

We let n represent the number of terms in the Fourier series. For this example, we are only inter-
ested in the first three terms, and thus n = 1, 2, and 3
By the voltage division expression,

1 ⁄ ( jnω ) 1
V C n = ------------------------------ V in n = -------------- V in n (7.80)
1 + 1 ⁄ ( jnω ) 1 + jn

With reference to (7.79) the phasors of the first 3 terms of (7.80) are
4A 4A 4A
------- sin t = ------- cos ( t – 90° ) ⇔ V in1 = ------- ∠– 90° (7.81)
π π π

4A 1 4A 1 4A 1
------- ⋅ --- sin 2t = ------- ⋅ --- cos ( 2t – 90° ) ⇔ V = ------- ⋅ --- ∠– 90° (7.82)
π 3 π 3 in 2 π 3

------- ⋅ 1
4A ------- ⋅ 1
--- sin 3t = 4A --- cos ( 3t – 90° ) ⇔ V 4A 1
= ------- ⋅ --- ∠– 90° (7.83)
π 5 π 5 in 3 π 5
By substitution of (7.81) through (7.83) into (7.80), we obtain the phasor and time domain voltages
indicated in (7.84) through (7.86) below.

1 4A 1 4A
V C 1 = ----------- ⋅ ------- ∠– 90° = -------------------- ⋅ ------- ∠– 90°
1+j π 2 ∠45° π
(7.84)
4A 2 4A 2
= ------- ⋅ ------- ∠– 135° ⇔ ------- ⋅ ------- cos ( t – 135° )
π 2 π 2

1 4A 1 1 4A 1
V C 2 = -------------- ⋅ ------- ⋅ --- ∠– 90° = ------------------------- ------- ⋅ --- ∠– 90°
1 + j2 π 3 5 ∠63.4° π 3
(7.85)
4A 5 4A 5
= ------- ⋅ ------- ∠– 153.4° ⇔ ------- ⋅ ------- cos ( 2t – 153.4° )
π 15 π 15

7-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Exponential Form of the Fourier Series

1 4A 1 1 4A 1
V C 3 = -------------- ⋅ ------- ⋅ --- ∠– 90° = ---------------------------- ------- ⋅ --- ∠– 90°
1 + j3 π 5 10 ∠71.6° π 5
(7.86)
4A 10 4A 10
= ------- ⋅ ---------- ∠– 161.6° ⇔ ------- ⋅ ---------- cos ( 3t – 161.6° )
π 50 π 50
Thus, the capacitor voltage in the time domain is

4A 2 5 10
v C ( t ) = ------- ------- cos ( t – 135° ) + ------- cos ( 2t – 153.4° ) + ---------- cos ( 3t – 161.6° ) + … (7.87)
π 2 15 50

7.8 The Exponential Form of the Fourier Series

The Fourier series are often expressed in exponential form. The advantage of the exponential form
is that we only need to perform one integration rather than two, one for the a n , and another for the
b n coefficients in the trigonometric form of the series. Moreover, in most cases the integration is
simpler.
The exponential form is derived from the trigonometric form by substitution of
jωt – jωt
+e -
cos ωt = e--------------------------- (7.88)
2
jωt – jωt
e –e
sin ωt = --------------------------- (7.89)
j2

into f ( t ) . Thus,
jωt – jωt j2ωt – j2ωt
+e +e
f ( t ) = --- a 0 + a 1 ⎛ ---------------------------- ⎞ + a 2 ⎛ --------------------------------- ⎞ +
1 e e
⎝ ⎠ ⎝ ⎠
(7.90)
2 2 2
jωt – jωt j2ωt – j2ωt
–e –e
… + b 1 ⎛ ---------------------------⎞ + b 2 ⎛ --------------------------------⎞ + …
e e
⎝ j2 ⎠ ⎝ j2 ⎠

and grouping terms with same exponents, we get

f ( t ) = … + ⎛ -----2 – -----2 ⎞ e
– j2ωt ⎛ a 1 b 1 ⎞ – jωt 1
+ --- a 0 + ⎛ -----1 + -----1⎞ e + ⎛ -----2 + -----2⎞ e
a b a b jωt a b j2ωt
+ ----- – ----- e (7.91)
⎝ 2 j2 ⎠ ⎝ 2 j2 ⎠ 2 ⎝ 2 j2 ⎠ ⎝ 2 j2⎠

The terms of (7.91) in parentheses are usually denoted as

b
C – n = --- ⎛ a n – ----n-⎞ = --- ( a n + jb n )
1 1
(7.92)
2⎝ j⎠ 2

b
C n = --- ⎛ a n + ----n-⎞ = --- ( a n – j b n )
1 1
(7.93)
2⎝ j⎠ 2

Signals and Systems with MATLAB Applications, Second Edition 7-31

Orchard Publications
Chapter 7 Fourier Series

1
C 0 = --- a 0 (7.94)
2
Then, (7.91) is written as

– j2ωt – jωt jωt j2ωt

f ( t ) = … + C –2 e + C –1 e + C0 + C1 e + C2 e +… (7.95)

We must remember that the C i coefficients, except C 0 , are complex and occur in complex conju-
gate pairs, that is,
C –n = C n∗ (7.96)
We can derive a general expression for the complex coefficients C n , by multiplying both sides of
– jnωt
(7.95) by e and integrating over one period, as we did in the derivation of the a n and b n coeffi-
cients of the trigonometric form. Then, with ω = 1 ,
2π 2π 2π
– jnt – j2t – jnt – jt – jnt
∫0 f ( t )e dt = … + ∫0 C–2 e e dt + ∫0 C–1 e e dt (7.97)
2π 2π
– jnt jt – jnt
+ ∫0 C0 e dt + ∫0 C1 e e dt
2π 2π
j2t – jnt jnt – jnt
+ ∫0 C2 e e dt + … + ∫0 Cn e e dt

We observe that all the integrals on the right side of (7.97) are zero except the last one. Therefore,
2π 2π 2π
– jnt jnt – jnt
∫0 f ( t )e dt = ∫0 Cn e e dt = ∫0 C n dt = 2πC n

or
2π
1 – jnt
C n = ------
2π ∫0 f ( t )e dt

and, in general, for ω ≠ 1 ,

2π
1 – jnωt
C n = ------
2π ∫0 f ( t )e d( ωt ) (7.98)

or

T
1 – jnωt
C n = ---
T ∫0 f ( t )e d( ωt ) (7.99)

We can derive the trigonometric Fourier series from the exponential series by addition and subtrac-

7-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Exponential Form of the Fourier Series

tion of the exponential form coefficients C n and C –n . Thus, from (7.92) and (7.93),

1
C n + C – n = --- ( a n – jb n + a n + jb n )
2
or

a n = C n + C –n (7.100)

Similarly,
1
C n – C – n = --- ( a n – jb n – a n – j b n ) (7.101)
2
or

b n = j ( Cn – C–n ) (7.102)

Symmetry in Exponential Series

1. For even functions, all coefficients C i are real

We recall from (7.92) and (7.93) that

b
C – n = --- ⎛ a n – ----n-⎞ = --- ( a n + jb n )
1 1
(7.103)
2⎝ j ⎠ 2
and
b
C n = --- ⎛ a n + ----n-⎞ = --- ( a n – j b n )
1 1
(7.104)
2⎝ j⎠ 2
Since even functions have no sine terms, the b n coefficients in (7.103) and (7.104) are zero.
Therefore, both C –n and C n are real.
2. For odd functions, all coefficients C i are imaginary

Since odd functions have no cosine terms, the a n coefficients in (7.103) and (7.104) are zero.
Therefore, both C –n and C n are imaginary.

3. If there is half-wave symmetry, C n = 0 for n = even

We recall from the trigonometric Fourier series that if there is half-wave symmetry, all even har-
monics are zero. Therefore, in (7.103) and (7.104) the coefficients a n and b n are both zero for
n = even , and thus, both C – n and C n are also zero for n = even .

4. If there is no symmetry, f ( t ) is complex.

Signals and Systems with MATLAB Applications, Second Edition 7-33

Orchard Publications
Chapter 7 Fourier Series

5. C –n = C n∗ always

This can be seen in (7.103) and (7.104)

Example 7.10
Compute the exponential Fourier series for the square waveform of Figure 7.30 below. Assume that
ω = 1.
T
A

π 2π
ωt
0

−A

Figure 7.30. Waveform for Example 7.10

Solution:
This is the same waveform as in Example 7.1, and as we know, it is an odd function, has half-wave
symmetry, and its DC component is zero. Therefore, the C n coefficients will be imaginary, C n = 0
for n = even , and C 0 = 0 . Using (7.98) with ω = 1 , we get
2π π 2π
1 – jnt 1 – jnt 1 – jnt
C n = ------
2π ∫0 f ( t )e dt = ------
2π ∫0 Ae dt + ------
2π ∫π –A e dt

and for n = 0 ,
π 2π
1 –0 –0 A
C 0 = ------
2π ∫0 Ae dt + ∫π ( – A )e dt = ------ ( π – 2π + π ) = 0
2π
as expected.
For n ≠ 0 ,

π 2π π 2π
1 – jnt – jnt 1 A – jnt – A – jnt
C n = ------
2π ∫0 Ae dt + ∫π –A e dt = ------ -------- e
2π – jn 0
+ -------- e
– jn π

1 A – jnπ A –jn2π – jnπ A – jnπ – jn2π – jnπ (7.105)

= ------ -------- ( e – 1 ) + ----- ( e –e ) = ------------ ( 1 – e +e –e )
2π – jn jn 2jπn
A – jn2π – jnπ A – jnπ 2
= ------------ ( 1 + e – 2e ) = ------------ ( e – 1)
2jπn 2jπn

7-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Line Spectra
– jnπ
For n = even , e = 1 ; then,

Cn A – jnπ 2 A 2
= ------------ ( e – 1 ) = ------------ ( 1 – 1 ) = 0 (7.106)
n = even 2jπn 2jπn
as expected.
– jnπ
For n = odd , e = – 1 . Therefore,

A – jnπ 2 A 2 A 2 2A
Cn = ------------ ( e – 1 ) = ------------ ( – 1 – 1 ) = ------------ ( – 2 ) = -------- (7.107)
n = odd 2jπn 2jπn 2jπn jπn
Using (7.95), that is,
– j2ωt – jωt jωt j2ωt
f ( t ) = … + C–2 e + C–1 e + C0 + C1 e + C2 e +…

we obtain the exponential Fourier series for the square waveform with odd symmetry as

f ( t ) = ------- ⎛ … – --- e
2A 1 – j3ωt – jωt jωt 1 j3ωt ⎞ 2A 1
∑
jnωt
–e +e + --- e = ------- --- e (7.108)
jπ ⎝ 3 3 ⎠ jπ n
n = odd

The minus ( −) sign of the first two terms within the parentheses results from the fact that
C –n = C n∗ . For instance, since C 3 = 2A ⁄ j3π , it follows that C – 3 = C 3∗ = – 2A ⁄ j3π . We observe
that f ( t ) is purely imaginary, as expected, since the waveform is an odd function.
To prove that (7.108) and (7.22) are the same, we group the two terms inside the parentheses of
(7.108) for which n = 1 ; this will produce the fundamental frequency sin ωt . Then, we group the
two terms for which n = 3 , and this will produce the third harmonic sin 3ωt , and so on.

7.9 Line Spectra

When the Fourier series are known, it is useful to plot the amplitudes of the harmonics on a fre-
quency scale that shows the first (fundamental frequency) harmonic, and the higher harmonics
times the amplitude of the fundamental. Such a plot is known as line spectrum and shows the spec-
tral lines that would be displayed by a spectrum analyzer*.
Figure 7.31 shows the line spectrum of the square waveform of Example 7.1.
Figure 7.32 shows the line spectrum for the half-wave rectifier of Example 7.6.
The line spectra of other waveforms can be easily constructed from the Fourier series.

* An instrument that displays the spectral lines of a waveform.

Signals and Systems with MATLAB Applications, Second Edition 7-35

Orchard Publications
Chapter 7 Fourier Series

bn
4/π

nωt
0 1 2 3 4 5 6 7 8 9

Figure 7.31. Line spectrum for square waveform of Example 7.1

A/ 2
A/π DC

2 4 6 8
nωt
0 1
Figure 7.32. Line spectrum for half-wave rectifier of Example 7.6

Example 7.11
Compute the exponential Fourier series for the waveform of Figure 7.33, and plot its line spectra.
Assume ω = 1 .
Solution:
This recurrent rectangular pulse is used extensively in digital communications systems. To determine
how faithfully such pulses will be transmitted, it is necessary to know the frequency components.

T
A

T/κ

0
−π/κ ωt
−2π −π π/κ π 2π

Figure 7.33. Waveform for Example 7.11

As shown in Figure 7.33, the pulse duration is T ⁄ k . Thus, the recurrence interval (period) T, is k
times the pulse duration. In other words, k is the ratio of the pulse repetition time to the duration of
each pulse.
For this example, the components of the exponential Fourier series are found from

7-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Line Spectra
π π⁄k
1 – jnt A – jnt
C n = ------
2π ∫– π Ae dt = ------
2π ∫– π ⁄ k e dt (7.109)

The value of the average ( DC component) is found by letting n = 0 . Then, from (7.109) we get
π⁄k
A π π
= ------ ⎛ --- + --- ⎞
A
C 0 = ------ t
2π –π ⁄ k
2π ⎝ k k ⎠
or
A
C 0 = --- (7.110)
k

For the values for n ≠ 0 , integration of (7.109) yields

sin ⎛ ------⎞
nπ
A - – jnt π ⁄ k
jnπ ⁄ k
–e
– jnπ ⁄ k ⎝ k⎠
= ------ ⋅ ------------------------------------- = ------ ⋅ sin ⎛ ------⎞ = A --------------------
A e A nπ
Cn = -------------- e
– jn2 π –π ⁄ k nπ j2 nπ ⎝ k⎠ nπ
or
A sin ( nπ ⁄ k )
C n = --- ⋅ -------------------------- (7.111)
k nπ ⁄ k
and thus,

∞
A sin ( nπ ⁄ k )
f(t) = ∑ --- ⋅ --------------------------
k nπ ⁄ k
(7.112)
n = –∞

The relation of (7.112) has the sin x ⁄ x form, and the line spectrum is shown in Figures 7.34 through
7.36, for k = 2 , k = 5 and k = 10 .

s in (n π /k )/(n π /k )

k = 2

-1 0 -8 -6 -4 -2 0 2 4 6 8 10

Figure 7.34. Line spectrum of (7.112) for k = 2

Signals and Systems with MATLAB Applications, Second Edition 7-37

Orchard Publications
Chapter 7 Fourier Series

sin(nπ/k)/(nπ/k)

k=5

-10 -5 0 5 10

Figure 7.35. Line spectrum of (7.112) for k = 5

sin(nπ/k)/(nπ/k)

k = 10

-10 -5 0 5 10

Figure 7.36. Line spectrum of (7.112) for k = 10

The spectral lines are separated by the distance 1 ⁄ k and thus, as k gets larger, the lines get closer
together while the lines are further apart as k gets smaller. Although the space between lines seems
to be the same in each case, we should observe that the number of lines between line crossings, are
different.

Example 7.12
Use the result of Example 7.11 to compute the exponential Fourier series of the unit impulse train
Aδ ( t ± 2πn ) shown in Figure 7.37.

Solution:
From Example 7.11,
A sin ( nπ ⁄ k )
C n = --- ⋅ -------------------------- (7.113)
k nπ ⁄ k

7-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Line Spectra

... ...

4π
ωt
−8π −6π −4π −2π 0 2π 6π 8π

Figure 7.37. Impulse train for Example 7.12

and the pulse width was defined as T ⁄ k , that is,
T 2π
--- = ------ (7.114)
k k
Next, let us represent the impulse train of Figure 7.37, as a recurrent pulse with amplitude

1 1 k
A = ---------- = ------------- = ------ (7.115)
T⁄k 2π ⁄ k 2π
as shown in Figure 7.38.
1 - T
A = ------------ A
2π ⁄ k

2π/κ

0
−π/κ
ωt
−2π −π π/κ π 2π

1
Figure 7.38. Recurrent pulse with amplitude A = ------------
-
2π ⁄ k
By substitution of (7.115) into (7.113), we get
k ⁄ 2π sin ( nπ ⁄ k ) 1 sin ( nπ ⁄ k )
C n = ------------- ⋅ -------------------------- = ------ -------------------------- (7.116)
k nπ ⁄ k 2π nπ ⁄ k

and as π ⁄ k → 0 , we observe from Figure 7.38, that each recurrent pulse becomes a unit impulse,
and the total number of the pulses reduce to a unit impulse train. Moreover, recalling that
1
sin x = 1 , we see that (7.116) reduces to C = -----
lim ---------
- n - , that is, all coefficients of the exponential
x→0 x 2π
Fourier series have the same amplitude and thus,
∞
1
∑
jnωt
f ( t ) = ------ e (7.117)
2π
n = –∞

Signals and Systems with MATLAB Applications, Second Edition 7-39

Orchard Publications
Chapter 7 Fourier Series

The series of (7.117) reveals that the line spectrum of the impulse train of Figure 7.38, consists of a
train of equal amplitude, and are equally spaced harmonics as shown in Figure 7.39.
Since these spectral lines extend from – ∞ to + ∞ , the bandwidth approaches infinity.

1/ 2π

... ...

n
−4 −3 −2 −1 0 1 2 3 4

Figure 7.39. Line spectrum for Example 7.12

Let us reconsider the train of recurrent pulses shown in Figure 7.40.

T
A

T/κ

0
ωt
−2π −π −π/κ π/κ π 2π

Figure 7.40. Recurrent pulse with T → ∞

Now, let us suppose that the pulses to the left and right of the pulse centered around zero, become
less and less frequent; or in other words, the period T approaches infinity. In this case, there is only
one pulse left (the one centered around zero). As T → ∞ , the fundamental frequency approaches
zero, that is, ω → 0 as T approaches infinity. Accordingly, the frequency difference between consec-
utive harmonics becomes smaller. In this case, the lines in the line spectrum come closer together,
and the line spectrum becomes a continuous spectrum. This forms the basis of the Fourier trans-
form that we will study in the next chapter.

7.10 Computation of RMS Values from Fourier Series

The RMS value of a waveform consisting of sinusoids of different frequencies, is equal to the
square root of the sum of the squares of the RMS values of each sinusoid. Thus, if
i = I 0 + I 1 cos ( ω 1 t ± θ 1 ) + I 2 cos ( ω 2 t ± θ 2 ) + … + I N cos ( ω N t ± θ N ) (7.118)

where I 0 represents a constant current, and I 1, I 2, …, I N represent the amplitudes of the sinusoids,
the RMS value of i is found from

7-40 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of RMS Values from Fourier Series

I 0 + I 1 RMS + I 2 RMS + … + I N RMS

2 2 2 2
I RMS = (7.119)

or

1 2 1 2 1 2
I 0 + --- I + --- I + … + --- I
2
I RMS = (7.120)
2 1m 2 2m 2 Nm

The proof of (7.119) is based on Parseval’s theorem; we will discuss this theorem on the next chap-
ter. A brief description of the proof follows.
We recall that the RMS (effective) value of a function, such as current i ( t ) , is defined as
T
--1-
∫0 i d t
2
I RMS = (7.121)
T

Substitution of (7.118) into (7.121), will produce the terms I 02 , I 1m [ cos ( ω 1 t – θ 1 ) ] , and other
2 2

similar terms representing higher order harmonics. The result will also contain products of cosine
functions multiplied by a constant, or other cosine terms of different harmonic frequencies. But as
we know, from the orthogonality principle, the integration of (7.121), will produce all zero terms
except the cosine squared terms which, for each harmonic, will be

2T
I m ---
2 1 2
--------- = --- I m (7.122)
T 2
as in (7.120).

Example 7.13
Find the I RMS value of the square waveform shown in Figure 7.41 by application of

a. relation (7.121)
b. relation (7.120)

ωt

−1

Figure 7.41. Waveform for Example 7.13

Signals and Systems with MATLAB Applications, Second Edition 7-41

Orchard Publications
Chapter 7 Fourier Series

Solution:
a. By inspection, the period is T = 2π as shown in Figure 7.42.

π 2π ωt

−1

Figure 7.42. Waveform of Example 7.13 showing period T = 2π

Then,
T 2π π 2π
1 1 1
∫0 ∫0 ∫0 ∫π
2
i d( ωt ) = ------ 1 d( ωt ) + ( – 1 ) d( ωt )
2 2 2 2
I RMS = --- i dt = ------
T 2π 2π
1 π 2π 1
= ------ [ ωt 0 + ωt π
] = ------ [ 2π ] = 1
2π 2π

or
I RMS = 1

b. In Example 7.1, we found that the given waveform may be written as

i ( t ) = --- ⎛ sin ωt + --- sin 3ωt + --- sin 5ωt + …⎞

4 1 1
(7.123)
π⎝ 3 5 ⎠

and as we know, the RMS value of a sinusoid is a real number independent of the frequency and
the phase angle, and it is equal to 0.707 times its maximum value, that is, I RMS = 0.707I max .
Then, from (7.120) and (7.123),

1 1 2 1 1 2
I RMS = --- 0 + --- ( 1 ) + --- ⎛ --- ⎞ + --- ⎛ --- ⎞ + … = 0.97
4 1 2
(7.124)
π 2 2⎝3⎠ 2⎝5⎠

This is a good approximation to unity, considering that higher harmonics have been neglected.

7.11 Computation of Average Power from Fourier Series

We can compute the average power of a Fourier series from the relation
P ave = P dc + P 1ave + P 2ave + …
(7.125)
= V dc I dc + V 1RMS I 1RMS cos θ 1 + V 2RMS I 2RMS cos θ 2 + …

The proof is obtained from the definition of average power, i.e.,

7-42 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of Average Power from Fourier Series

T T
1 1
P ave = ---
T ∫0 p dt = ---
T ∫0 vi dt (7.126)

and the expression for the alternate trigonometric Fourier series, that is,

∞
1
f ( t ) = --- a 0 +
2 ∑ cn cos ( nωt – θn ) (7.127)
n=1

where f ( t ) can represent voltages and currents. Then, by substitution of these series for v and i into
(7.126), we will find that the products of v and i that have different frequencies, will be zero, and
only the products of the same frequency terms will have non-zero values. The non-zero values will
represent the average power for each harmonic in (7.125).

Example 7.14
For the circuit of Figure 7.43, compute:

a. The current i c ( t ) given that v in ( t ) = 6 ⎛⎝ cos ωt – --- cos 3ωt⎞⎠ V where ω = 1000 r ⁄ s .
1
3

b. The average power P ave delivered by the voltage source.

Solution:
a. We will use the subscripts 1 and 3 to represent the quantities due to the fundamental and third
harmonic frequencies respectively. Since the excitation consists of two sinusoids of different fre-
quencies, we can use phasor quantities, and we will denote them with capital letters.
R

1Ω C
+ iC ( t )
+
− –3 −
10 -
-----------
v in ( t ) F
3

Figure 7.43. Circuit for Example 7.14

Then,
v in1 ( t ) = 6 cos ωt ⇔ V in1 = 6 ∠0° V

–j –j
---------- = -------------------------------- = – j3
ω1 C 3
10 × 10 ⁄ 3
–3

Z 1 = 1 – j3 = 10 ∠– 71.6°

Signals and Systems with MATLAB Applications, Second Edition 7-43

Orchard Publications
Chapter 7 Fourier Series

V in1 6 ∠0°
- = ------------------------------- = 1.90 ∠71.6° ⇔ i C1 ( t ) = 1.90 cos ( ωt + 71.6° ) A
I C1 = --------- (7.128)
Z1 10 ∠– 71.6°
Next,
v in3 ( t ) = – 2 cos 3ωt = 2 cos ( 3ωt + 180° ) ⇔ V in3 = 2 ∠180° V

–j –j
---------- = ----------------------------------------- = – j1
ω3 C 3 × 10 × 10 ⁄ 3
3 –3

Z 3 = 1 – j1 = 2 ∠– 45°

V in3 2 ∠180°
- = ------------------------ = 1.41 ∠225° = 1.41 ∠( 225 – 135 )°
I C3 = ---------
Z3 2 ∠– 45° (7.129)
⇔ i C3 ( t ) = 1.41 cos ( 3ωt – 135 ° ) A

From (7.128) and (7.129),

i c ( t ) = i c1 ( t ) + i c3 ( t ) = 1.90 cos ( ωt + 71.6° ) + 1.41 cos ( 3ωt – 135 ° ) (7.130)

b. The average power delivered by the voltage source is

P ave = V 1RMS I 1RMS cos θ 1 + V 3RMS I 3RMS cos θ 3
6 1.90 2 1.41 (7.131)
= ------- ⋅ ---------- cos ( 71.6° ) + ------- ⋅ ---------- cos ( – 135 ° )
2 2 2 2
or
P ave = 0.8 w (7.132)
Check:
The average power absorbed by the capacitor is zero, and therefore, the average power absorbed by
the resistor, must be equal to the average power delivered by the source. The average power
absorbed by the resistor is
1 2 1 2 1
P ave = --- I max R = --- ( I 1max – I 3max ) = --- ( 1.90 – 1.41 ) = 0.8 w
2 2 2
2 2 2

7.12 Numerical Evaluation of Fourier Coefficients

The use of Fourier series is not restricted to electric circuit analysis. It is also applied in the analysis
of the behavior of physical systems subjected to periodic disturbances. Examples include cable
stress analysis in suspension bridges, and mechanical vibrations.
Quite often, it is necessary to construct the Fourier expansion of a function based on observed val-
ues instead of an analytic expression. Examples are meteorological or economic quantities whose

7-44 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Numerical Evaluation of Fourier Coefficients

period may be a day, a week, a month or even a year. In these situations, we need to evaluate the inte-
gral(s) using numerical integration.
The procedure presented here, will work for both the waveforms that have an analytical solution and
those that do not. Even though we may already know the Fourier series from analytical methods, we
can use this procedure to check our results.
Consider the waveform of f ( x ) shown in Figure 7.44, were we have divided it into small pulses of
width ∆x . Obviously, the more pulses we use, the better the approximation.
If the time axis is in degrees, we can choose ∆x to be 2.5° and it is convenient to start at the zero
point of the waveform. Then, using a spreadsheet, such as Microsoft Excel, we can divide the period
0° to 360° in 2.5° intervals, and enter these values in Column A of the spreadsheet.

f(x)

Figure 7.44. Waveform whose analytical expression is unknown

Since the arguments of the sine and the cosine are in radians, we multiply degrees by π (3.1459...)
and divide by 180 to perform the conversion. We enter these in Column B and we denote them as
x . In Column C we enter the corresponding values of y = f ( x ) as measured from the waveform.
In Columns D and E we enter the values of cos x and the product y cos x respectively. Similarly, we
enter the values of sin x and y sin x in Columns F and G respectively.
Next, we form the sums of y cos x and y sin x , we multiply these by ∆x , and we divide by π to
obtain the coefficients a 1 and b 1 . To compute the coefficients of the higher order harmonics, we
form the products y cos 2x , y sin 2x , y cos 3x , y sin 3x , and so on, and we enter these in subsequent
columns of the spreadsheet.
Figure 7.45 is a partial table showing the computation of the coefficients of the square waveform,
and Figure 7.46 is a partial table showing the computation of the coefficients of a clipped sine wave-
form. The complete tables extend to the seventh harmonic to the right and to 360° down.

Signals and Systems with MATLAB Applications, Second Edition 7-45

Orchard Publications
 Analytical:

7-46
1.5 Square waveform f(t)=4(sinwt/p+sin3wt/3p+sin5wt/5p+ ….)
1.0
Numerical:
0.5

0.0 Average= DC= 0.000

a1= 0.000 b1= 1.273
-0.5 a2= 0.000 b2= 0.000
-1.0 a3= 0.000 b3= 0.424
a4= 0.000 b4= 0.000
-1.5
a5= 0.000 b5= 0.254
Chapter 7 Fourier Series

0.0 2.0 4.0 6.0 8.0

a6= 0.000 b6= 0.000
a7= 0.000 b7= 0.180

x(deg) x(rad) y=f(x) 0.5*a0 cosx ycosx sinx ysinx cos2x ycox2x sin2x ysin2x cos3x ycos3x sin3x ysin3x

0.0 0.000 0.000 0.000 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000
2.5 0.044 1.000 0.044 0.999 0.999 0.044 0.044 0.996 0.996 0.087 0.087 0.991 0.991 0.131 0.131
5.0 0.087 1.000 0.044 0.996 0.996 0.087 0.087 0.985 0.985 0.174 0.174 0.966 0.966 0.259 0.259
7.5 0.131 1.000 0.044 0.991 0.991 0.131 0.131 0.966 0.966 0.259 0.259 0.924 0.924 0.383 0.383
10.0 0.175 1.000 0.044 0.985 0.985 0.174 0.174 0.940 0.940 0.342 0.342 0.866 0.866 0.500 0.500
12.5 0.218 1.000 0.044 0.976 0.976 0.216 0.216 0.906 0.906 0.423 0.423 0.793 0.793 0.609 0.609
15.0 0.262 1.000 0.044 0.966 0.966 0.259 0.259 0.866 0.866 0.500 0.500 0.707 0.707 0.707 0.707
17.5 0.305 1.000 0.044 0.954 0.954 0.301 0.301 0.819 0.819 0.574 0.574 0.609 0.609 0.793 0.793
20.0 0.349 1.000 0.044 0.940 0.940 0.342 0.342 0.766 0.766 0.643 0.643 0.500 0.500 0.866 0.866
22.5 0.393 1.000 0.044 0.924 0.924 0.383 0.383 0.707 0.707 0.707 0.707 0.383 0.383 0.924 0.924
25.0 0.436 1.000 0.044 0.906 0.906 0.423 0.423 0.643 0.643 0.766 0.766 0.259 0.259 0.966 0.966
27.5 0.480 1.000 0.044 0.887 0.887 0.462 0.462 0.574 0.574 0.819 0.819 0.131 0.131 0.991 0.991
30.0 0.524 1.000 0.044 0.866 0.866 0.500 0.500 0.500 0.500 0.866 0.866 0.000 0.000 1.000 1.000
32.5 0.567 1.000 0.044 0.843 0.843 0.537 0.537 0.423 0.423 0.906 0.906 -0.131 -0.131 0.991 0.991
35.0 0.611 1.000 0.044 0.819 0.819 0.574 0.574 0.342 0.342 0.940 0.940 -0.259 -0.259 0.966 0.966
37.5 0.654 1.000 0.044 0.793 0.793 0.609 0.609 0.259 0.259 0.966 0.966 -0.383 -0.383 0.924 0.924
40.0 0.698 1.000 0.044 0.766 0.766 0.643 0.643 0.174 0.174 0.985 0.985 -0.500 -0.500 0.866 0.866
42.5 0.742 1.000 0.044 0.737 0.737 0.676 0.676 0.087 0.087 0.996 0.996 -0.609 -0.609 0.793 0.793
45.0 0.785 1.000 0.044 0.707 0.707 0.707 0.707 0.000 0.000 1.000 1.000 -0.707 -0.707 0.707 0.707

Figure 7.45. Numerical computation of the coefficients of the square waveform (partial listing)
47.5 0.829 1.000 0.044 0.676 0.676 0.737 0.737 -0.087 -0.087 0.996 0.996 -0.793 -0.793 0.609 0.609

Orchard Publications
Signals and Systems with MATLAB Applications, Second Edition
50.0 0.873 1.000 0.044 0.643 0.643 0.766 0.766 -0.174 -0.174 0.985 0.985 -0.866 -0.866 0.500 0.500
 Analytical:

1.5 f(t)=unknown
Sine wave clipped at π/6, 5π/6 etc.
1.0
Numerical:
0.5
DC= 0.000
0.0

Orchard Publications
a1= 0.000 b1= 0.609
-0.5 a2= 0.000 b2= 0.000
a3= 0.000 b3= 0.138
-1.0 a4= 0.000 b4= 0.000
-1.5 a5= 0.000 b5= 0.028
0.0 2.0 4.0 6.0 8.0 a6= 0.000 b6= 0.000
a7= 0.000 b7= -0.010

x(deg) x(rad) y=f(x) 0.5*a0 cosx ycosx sinx ysinx cos2x ycox2x sin2x ysin2x cos3x ycos3x sin3x ysin3x

0.0 0.000 0.000 0.000 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000
2.5 0.044 0.044 0.002 0.999 0.044 0.044 0.002 0.996 0.043 0.087 0.004 0.991 0.043 0.131 0.006
5.0 0.087 0.087 0.004 0.996 0.087 0.087 0.008 0.985 0.086 0.174 0.015 0.966 0.084 0.259 0.023
7.5 0.131 0.131 0.006 0.991 0.129 0.131 0.017 0.966 0.126 0.259 0.034 0.924 0.121 0.383 0.050
10.0 0.175 0.174 0.008 0.985 0.171 0.174 0.030 0.940 0.163 0.342 0.059 0.866 0.150 0.500 0.087
12.5 0.218 0.216 0.009 0.976 0.211 0.216 0.047 0.906 0.196 0.423 0.091 0.793 0.172 0.609 0.132

Signals and Systems with MATLAB Applications, Second Edition

15.0 0.262 0.259 0.011 0.966 0.250 0.259 0.067 0.866 0.224 0.500 0.129 0.707 0.183 0.707 0.183
17.5 0.305 0.301 0.013 0.954 0.287 0.301 0.090 0.819 0.246 0.574 0.172 0.609 0.183 0.793 0.239
20.0 0.349 0.342 0.015 0.940 0.321 0.342 0.117 0.766 0.262 0.643 0.220 0.500 0.171 0.866 0.296
22.5 0.393 0.383 0.017 0.924 0.354 0.383 0.146 0.707 0.271 0.707 0.271 0.383 0.146 0.924 0.354
25.0 0.436 0.423 0.018 0.906 0.383 0.423 0.179 0.643 0.272 0.766 0.324 0.259 0.109 0.966 0.408
27.5 0.480 0.462 0.020 0.887 0.410 0.462 0.213 0.574 0.265 0.819 0.378 0.131 0.060 0.991 0.458
30.0 0.524 0.500 0.022 0.866 0.433 0.500 0.250 0.500 0.250 0.866 0.433 0.000 0.000 1.000 0.500
32.5 0.567 0.500 0.022 0.843 0.422 0.537 0.269 0.423 0.211 0.906 0.453 -0.131 -0.065 0.991 0.496
35.0 0.611 0.500 0.022 0.819 0.410 0.574 0.287 0.342 0.171 0.940 0.470 -0.259 -0.129 0.966 0.483
37.5 0.654 0.500 0.022 0.793 0.397 0.609 0.304 0.259 0.129 0.966 0.483 -0.383 -0.191 0.924 0.462
40.0 0.698 0.500 0.022 0.766 0.383 0.643 0.321 0.174 0.087 0.985 0.492 -0.500 -0.250 0.866 0.433
42.5 0.742 0.500 0.022 0.737 0.369 0.676 0.338 0.087 0.044 0.996 0.498 -0.609 -0.304 0.793 0.397
45.0 0.785 0.500 0.022 0.707 0.354 0.707 0.354 0.000 0.000 1.000 0.500 -0.707 -0.354 0.707 0.354
47.5 0.829 0.500 0.022 0.676 0.338 0.737 0.369 -0.087 -0.044 0.996 0.498 -0.793 -0.397 0.609 0.304

Figure 7.46. Numerical computation of the coefficients of a clipped sine waveform (partial listing)
50.0 0.873 0.500 0.022 0.643 0.321 0.766 0.383 -0.174 -0.087 0.985 0.492 -0.866 -0.433 0.500 0.250
Numerical Evaluation of Fourier Coefficients

7-47
Chapter 7 Fourier Series

7.13 Summary
• Any periodic waveform f ( t ) can be expressed as

∞
1
f ( t ) = --- a 0 +
2 ∑ ( a cos nωt + b sin nωt )
n n
n=1

where the first term a 0 ⁄ 2 is a constant, and represents the DC (average) component of f ( t ) .
The terms with the coefficients a 1 and b 1 together, represent the fundamental frequency com-
ponent ω . Likewise, the terms with the coefficients a 2 and b 2 together, represent the second
harmonic component 2ω , and so on. The coefficients a 0 , a n , and b n are found from the follow-
ing relations:
2π
1 1
--- a 0 = ------
2 2π ∫0 f ( t ) dt

2π
1
a n = ---
π ∫0 f ( t ) cos nt dt

2π
1
b n = ---
π ∫0 f ( t ) sin nt dt

• If a waveform has odd symmetry, that is, if it is an odd function, the series will consist of sine
terms only. We recall that odd functions are those for which – f ( – t ) = f ( t ) .
• If a waveform has even symmetry, that is, if it is an even function, the series will consist of cosine
terms only, and a 0 may or may not be zero. We recall that even functions are those for which
f( –t ) = f ( t )

• A periodic waveform with period T , has half-wave symmetry if

–f ( t + T ⁄ 2 ) = f ( t )
that is, the shape of the negative half-cycle of the waveform is the same as that of the positive
half-cycle, but inverted. If a waveform has half-wave symmetry only odd (odd cosine and odd
sine) harmonics will be present. In other words, all even (even cosine and even sine) harmonics
will be zero.
• The trigonometric Fourier series for the square waveform with odd symmetry is

f ( t ) = ------- ⎛ sin ωt + --- sin 3ωt + --- sin 5ωt + …⎞ = -------

4A 1 1 4A 1
π⎝ 3 5 ⎠ π ∑ --- sin nωt
n
n = odd

7-48 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• The trigonometric Fourier series for the square waveform with even symmetry is
(n – 1)
----------------
f ( t ) = ------- ⎛ cos ω t – --- cos 3ωt + --- cos 5ωt – …⎞ = -------
4A 1 4A 2 1
∑
1
( –1 ) --- cos n ωt
π⎝ 3 5 ⎠ π n
n = odd

• The trigonometric Fourier series for the sawtooth waveform with odd symmetry is

f ( t ) = ------- ⎛ sin ωt – --- sin 2ωt + --- sin 3ωt – --- sin 4ωt + …⎞ = -------
2A 1 1 1 2A n–1 1
π ⎝ 2 3 4 ⎠ π ∑ ( –1 ) --- sin nωt
n

• The trigonometric Fourier series for the triangular waveform with odd symmetry is
(n – 1)
----------------
f ( t ) = ------2- ⎛ sin ω t – --- sin 3ωt + ------ sin 5ωt – ------ sin 7ωt + …⎞ = ------2-
8A 1 1 8A 1
∑
1 2
( –1 ) ----- sin n ωt
⎝ 25 49 ⎠ 2
π 9 π n = odd n

• The trigonometric Fourier series for the half-wave rectifier with no symmetry is

A A cos 2t cos 4t cos 6t cos 8t

f( t) = A
--- + --- sin t – --- ------------- + ------------- + ------------- + ------------- + …
π 2 π 3 15 35 63

• The trigonometric form of the Fourier series for the full-wave rectifier with even symmetry is
∞
2A 4A 1
f ( t ) = ------- – -------
π π ∑ ------------------
( n
2
– 1
- cos nωt
)
n = 2, 4, 6, …

• The Fourier series are often expressed in exponential form as

– j2 ω t –j ω t jωt j2 ω t
f ( t ) = … + C –2 e + C –1 e + C0 + C1 e + C2 e +…

where the C i coefficients are related to the trigonometric form coefficients as

b
C –n = --- ⎛ a n – ----n-⎞ = --- ( a n + jb n )
1 1
2 ⎝ j ⎠ 2

b
C n = --- ⎛ a n + ----n-⎞ = --- ( a n – j b n )
1 1
2 ⎝ j ⎠ 2

1
C 0 = --- a 0
2

• The C i coefficients, except C 0 , are complex, and appear as complex conjugate pairs, that is,

C – n = C n∗

Signals and Systems with MATLAB Applications, Second Edition 7-49

Orchard Publications
Chapter 7 Fourier Series

• In general, for ω ≠ 1 ,
T 2π
1 – jn ω t 1 – jn ω t
C n = ---
T ∫0 f ( t )e d( ωt ) = ------
2π ∫0 f ( t )e d( ωt )

• We can derive the trigonometric Fourier series from the exponential series from the relations

an = Cn + C–n
and
bn = j ( Cn – C–n )

• For even functions, all coefficients C i are real

• For odd functions, all coefficients C i are imaginary

• If there is half-wave symmetry, C n = 0 for n = even

• C –n = C n∗ always

• A line spectrum is a plot that shows the amplitudes of the harmonics on a frequency scale.

• The frequency components of a recurrent rectangular pulse follow a sin x ⁄ x form.

• Τhe line spectrum of an impulse train consists of a train of equal amplitude, and are equally
spaced harmonics.
• Τhe RMS value of a waveform consisting of sinusoids of different frequencies, is equal to the
square root of the sum of the squares of the RMS values of each sinusoid. Thus,

I 0 + I 1 RMS + I 2 RMS + … + I N RMS

2 2 2 2
I RMS =
or
1 2 1 2 1 2
I 0 + --- I + --- I + … + --- I
2
I RMS =
2 1m 2 2m 2 Nm

• We can compute the average power of a Fourier series from the relation
P ave = P dc + P 1ave + P 2ave + …
= V dc I dc + V 1RMS I 1RMS cos θ 1 + V 2RMS I 2RMS cos θ 2 + …

• We can evaluate the Fourier coefficients of a function based on observed values instead of an
analytic expression using numerical evaluations with the aid of a spreadsheet.

7-50 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

7.14 Exercises
1. Compute the first 5 components of the trigonometric Fourier series for the waveform of Figure
7.47. Assume ω = 1 .
f(t)
A

ωt
0
Figure 7.47. Waveform for Exercise 1
2. Compute the first 5 components of the trigonometric Fourier series for the waveform of Figure
7.48. Assume ω = 1 .
f(t)
A

ωt
0
Figure 7.48. Waveform for Exercise 2
3. Compute the first 5 components of the exponential Fourier series for the waveform of Figure
7.49. Assume ω = 1 .

f(t)
A

ωt
0
Figure 7.49. Waveform for Exercise 3
4. Compute the first 5 components of the exponential Fourier series for the waveform of Figure
7.50. Assume ω = 1 .
f( t)
A⁄2

0 ωt
–A ⁄ 2

Figure 7.50. Waveform for Exercise 4

Signals and Systems with MATLAB Applications, Second Edition 7-51

Orchard Publications
Chapter 7 Fourier Series

5. Compute the first 5 components of the exponential Fourier series for the waveform of Figure
7.51. Assume ω = 1 .
f( t)
A

ωt
0
Figure 7.51. Waveform for Exercise 5

6. Compute the first 5 components of the exponential Fourier series for the waveform of Figure
7.52. Assume ω = 1 .

f(t)
A

ωt
0
−A

Figure 7.52. Waveform for Exercise 6

7-52 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

7.15 Solutions to Exercises

1.
f(t) A
A --- t
π

ωt
– 2π –π 0 π 2π

This is an even function; therefore, the series consists of cosine terms only. There is no half-wave
symmetry and the average ( DC component) is not zero. We will integrate from 0 to π and mul-
tiply by 2 . Then,
π π
2 A 2A
a n = ---
π ∫0 --- t cos nt dt = -------
π π
2 ∫0 t cos nt dt (1)

From tables of integrals,

1 x
∫ x cos ax dx = ----2- cos ax + --- sin a x
a a
and thus (1) becomes
π
a n = ------2- ⎛ ----2- cos nt + --- sin nt ⎞ = ------2- ⎛ ----2- cos nπ + --- sin ntπ – ----- – 0 ⎞
2A 1 t 2A 1 t 1
⎝ n ⎠ ⎝ n ⎠
π n 0
π n n
2

and since sin ntπ = 0 for all integer n ,

a n = ------2- ⎛ ----2- cos nπ – ----- ⎞ = ----------2- ( cos nπ – 1 ) (2)

2A 1 1 2A
⎝ 2⎠
π n n n π
2

We cannot evaluate the average ( 1 ⁄ 2 ) ⁄ a 0 from (2); we must use (1). Then, for n = 0 ,

π 2 π 2
1 2A A t A π
--- a 0 = ---------
2 2π
2 ∫0 t dt = ----2- ⋅ ----
π 2
= ----2- ⋅ -----
π 2
0
or
( 1 ⁄ 2 ) ⁄ a0 = A ⁄ 2

We observe from (2) that for n = even , a n = even = 0 . Then,

– 4A – 4A-
for n = 1, a 1 = – 4A 4A -, for n = 7, a = ----------
-------, for n = 3, a 3 = ----------2-, for n = 5, a 5 = – ---------- 3
2 2 2 2 2
π 3 π
2
5 π 7 π

Signals and Systems with MATLAB Applications, Second Edition 7-53

Orchard Publications
Chapter 7 Fourier Series

and so on.
Therefore,
∞
f ( t ) = --- a 0 – ------2- ⎛ cos t + --- cos 3t + ------ cos 5t + ------ cos 7t + …⎞ = A
1 4A 1 1 1 4A 1
2 π
⎝ 9 25 49 ⎠
--- – -------
2 π ∑ ----- cos nt
n
2
n = odd

2.
f( t) 2A
------- t
A π

ωt
0 π⁄2 π 3π ⁄ 2 π
This is an even function; therefore, the series consists of cosine terms only. There is no half-wave
symmetry and the average ( DC component) is not zero.
1 Area 2 × [ ( A ⁄ 2 ) ⋅ ( π ⁄ 2 ) ] + Aπ 3A ⋅ ( π ⁄ 2 ) 3A
Average = --- a 0 = ------------------ = --------------------------------------------------------------- = -------------------------- = -------
2 Period 2π 2π 4
π⁄2 π
2 2A
------- t cos nt dt + --2-
a n = ---
π ∫0 π π ∫π ⁄ 2 A cos nt dt (1)

and with
1 x 1
∫ x cos ax dx = ----2- cos ax + --- sin a x = ----2- ( cos ax + ax sin ax )
a a a
(1) simplifies to
π⁄2
4A 1 2A π
a n = ------2- ----2- ( cos nt + nt sin nt ) + ------- sin nt π⁄2
π n nπ
0

= ----------2- ⎛ cos nπ ------ – 1 – 0⎞ + ------- ⎛ sin nπ – sin nπ

------⎞
4A nπ 2A
------ + ------ sin nπ
2 ⎝ 2 2 2 ⎠ nπ ⎝ 2⎠
n π

and since sin ntπ = 0 for all integer n ,

------ + ------- sin ------ – ----------- – ------- sin ------ = ----------2- ⎛⎝ cos ------ – 1⎞⎠
4A 2A nπ 4A 2A nπ 4A
a n = ----------2- cos nπ nπ
2 2 nπ 2 2 2 nπ 2 2 2
n π n π n π

4A 4A
4A-, for n = 2, a = -------- 2A-
for n = 1, a 1 = ------2- ( 0 – 1 ) = – ------ 2 - ( – 1 – 1 ) = – ------
2
π π
2
4π π
2

4A 4A – 4A
for n = 3, a 3 = --------2- ( 0 – 1 ) = – ---------, for n = 4, a 4 = ----------2- ( 1 – 1 ) = 0
2 2
9π 9π 7 π

7-54 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

We observe that the fourth harmonic and all its multiples are zero. Therefore,

------- – ------2- ⎛ cos t + --- cos 2t + --- cos 3t + …⎞

4A 1 1
f ( t ) = 3A
4 π ⎝ 2 9 ⎠

3.
f(t)
A

ωt
0 π 2π

This is neither an even nor an odd function and has no half-wave symmetry; therefore, the series
consists of both cosine and sine terms. The average ( DC component) is not zero. Then,
2π
1 – jn ω t
C n = ------
2π ∫0 f ( t )e d( ωt )

and with ω = 1
2π π 2π π
1 – jnt 1 – jnt – jnt A – jnt
C n = ------
2π ∫0 f ( t )e dt = ------
2π ∫0 Ae dt + ∫π 0e dt = ------
2π ∫0 e dt

The DC value is
π π
A A
∫0
0 A
C 0 = ------ e dt = ------ t = ---
2π 2π 0 2
For n ≠ 0
π π
A – jnt A – jnt A – jn π
C n = ------
2π ∫0 e dt = --------------- e
– j2 nπ 0
= ------------ ( 1 – e
j2nπ
)

Recalling that
– jn π
e = cos nπ – j sin nπ
– jn π – jn π
for n = even , e = 1 and for n = odd , e = – 1 . Then,

A
C n = even = ------------ ( 1 – 1 ) = 0
j2nπ
and
A A-
C n = odd = ------------ [ 1 – ( – 1 ) ] = -------
j2nπ jnπ
By substitution into

Signals and Systems with MATLAB Applications, Second Edition 7-55

Orchard Publications
Chapter 7 Fourier Series

– j2 ω t –j ω t jωt j2 ω t
f ( t ) = … + C –2 e + C –1 e + C0 + C1 e + C2 e +…
we find that
--- + ----- ⎛ … – --- e + …⎞
A 1 – j3 ω t – j ω t j ω t 1 j3 ω t
f( t) = A –e + e + --- e
2 jπ ⎝ 3 3 ⎠

The minus (−) sign of the first two terms within the parentheses results from the fact that
C – n = C n∗ . For instance, since C 1 = 2A ⁄ jπ , it follows that C – 1 = C 1∗ = – 2A ⁄ jπ . We observe
that f ( t ) is complex, as expected, since there is no symmetry.
4.
f(t)
A⁄2

0 ωt
–A ⁄ 2

This is the same waveform as in Exercise 3 where the DC component has been removed. Then,

f ( t ) = ----- ⎛ … – --- e + …⎞
A 1 – j3 ω t – j ω t j ω t 1 j3 ω t
–e + e + --- e
jπ ⎝ 3 3 ⎠

It is also the same waveform as in Example 7.10 except that the amplitude is halved. This wave-
form is an odd function and thus the expression for f ( t ) is imaginary.
5.
f( t)
A

–π 0 π
ωt
–π ⁄ 2 π⁄2

This is the same waveform as in Exercise 3 where the vertical axis has been shifted to make the
waveform an even function. Therefore, for this waveform C n is real. Then,
π π⁄2
1 – jnt A – jnt
C n = ------
2π ∫– π f ( t )e dt = ------
2π ∫– π ⁄ 2 e dt

The DC value is

7-56 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
π⁄2
A π π
= ------ ⎛ --- + ---⎞ = ---
A A
C 0 = ------ t
2π –π ⁄ 2
2π ⎝ 2 2⎠ 2
For n ≠ 0
π⁄2 π⁄2
A – jnt A – jnt A – jn π ⁄ 2 jn π ⁄ 2
C n = ------
2π ∫– π ⁄ 2 e dt = --------------- e
– j2 nπ –π ⁄ 2
= --------------- ( e
– j2 nπ
–e )

jn π ⁄ 2 – jn π ⁄ 2
–e
) = ------ ⎛ --------------------------------------⎞ = ------ sin nπ
A jn π ⁄ 2 – jn π ⁄ 2 A e A
= ------------ ( e –e ------
j2nπ nπ ⎝ j2 ⎠ nπ 2

and we observe that for n = even , C n = 0

For n = odd , C n alternates in plus (+) and minus (−) signs, that is,

A
C n = ------ if n = 1, 5, 9, …
nπ
A
C n = – ------ if n = 3, 7, 11, …
nπ
Thus,
⎛ ± -----
A- jn ω t⎞
f(t) = A
--- +
2 ∑ ⎝ nπ
e
⎠
n = odd

where the plus (+) sign is used with n = 1, 5, 9, … and the minus (−) sign is used with
n = 3, 7, 11, … . We can express f ( t ) in a more compact form as

(n – 1) ⁄ 2 A- jn ω t
f( t) = A
--- +
2 ∑ ( –1 ) -----
nπ
e
n = odd

6.
f(t) 2A
------- t – 1
A π

–π ⁄ 2 π⁄2
ωt
–π 0 π
−A

We will find the exponential form coefficients C n from

π
1 – jnt
C n = ------
2π ∫–π f ( t )e dt

From tables of integrals

Signals and Systems with MATLAB Applications, Second Edition 7-57

Orchard Publications
Chapter 7 Fourier Series

ax
e
∫
ax
xe dx = ------2- ( ax – 1 )
a
Then,
0 π
⎛ – 2A
------- t – 1⎞ e dt + ∫0 ⎛⎝ ------
- t – 1⎞ e
1 – jnt 2A – jnt
C n = ------
2π ∫ –π
⎝ π ⎠ π ⎠
dt

Integrating and rearranging terms we get

jn π – jn π jn π – jn π jn π – jn π
4A 4A- ⎛ –e +e –e
- ⎞ – ------- ⋅ ----------------------------
1 e e 2A e
C n = ------ – --------- + -------- nπ ⋅ ---------------------------- + ---------------------------
2π n π n π 2 2 ⎝ j2 2 ⎠ n j2

4A ⎛
– 1 + nπ sin nπ + cos nπ – ------ sin nπ⎞
nπ
= --------------
2 2⎝ 2 ⎠
2n π

and since sin nπ = 0 for all integer n ,

2A -
C n = ----------
2 2
( cos nπ – 1 )
n π

–4 A
For n = even , C n = 0 and for n = odd , cos nπ = – 1 , and C n = ----------
2 2
-
n π

Also, by inspection, the DC component C 0 = 0 . Then,

f ( t ) = – ------- ⎛ … + --- e + …⎞
4A 1 – j3 ω t –j ω t j ω t 1 j3 ω t
+e + e + --- e
2⎝ 9 9 ⎠
π
– j3 ω t –j ω t
The coefficients of the terms e and e are positive because all coefficients of C n are real.
This is to be expected since f ( t ) is an even function. It also has half-wave symmetry and thus
C n = 0 for n = even as we’ve found.

7-58 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Chapter 8
The Fourier Transform

his chapter introduces the Fourier Transform, also known as the Fourier Integral. The defini-
T tion, theorems, and properties are presented and proved. The Fourier transforms of the most
common functions are derived, the system function is defined, and several examples are given
to illustrate its application in circuit analysis.

8.1 Definition and Special Forms

We recall that the Fourier series for periodic functions of time, such as those we discussed on the pre-
vious chapter, produce discrete line spectra with non-zero values only at specific frequencies referred
to as harmonics. However, other functions of interest such as the unit step, the unit impulse, the unit
ramp, and a single rectangular pulse are not periodic functions. The frequency spectra for these func-
tions are continuous as we will see later in this chapter.
We may think of a non-periodic signal as one arising from a periodic signal in which the period
extents from – ∞ to + ∞ . Then, for a signal that is a function of time with period from – ∞ to + ∞ ,
we form the integral
∞
– jωt
F(ω) = ∫–∞ f ( t )e dt (8.1)

and assuming that it exists for every value of the radian frequency ω , we call the function F ( ω ) the
Fourier transform or the Fourier integral.
The Fourier transform is, in general, a complex function. We can express it as the sum of its real and
imaginary components, or in exponential form, that is, as
jϕ ( ω )
F ( ω ) = Re { F ( ω ) } + jIm { F ( ω ) } = F ( ω ) e (8.2)
The Inverse Fourier transform is defined as
∞
1
∫–∞ F ( ω )e
jωt
f ( t ) = ------ dω (8.3)
2π

We will often use the following notation to express the Fourier transform and its inverse.

F {f(t)} = F(ω) (8.4)

and
–1
F { F(ω )} = f( t) (8.5)

Signals and Systems with MATLAB Applications, Second Edition 8-1

Orchard Publications
Chapter 8 The Fourier Transform

8.2 Special Forms of the Fourier Transform

The time function f ( t ) is, in general, complex, and thus we can express it as the sum of its real and
imaginary parts, that is, as
f ( t ) = f Re ( t ) + j f Im ( t ) (8.6)

The subscripts Re and Im will be used often to denote the real and imaginary parts respectively.
These notations have the same meaning as Re { f ( t ) } and Im { f ( t ) } .
By substitution of (8.6) into the Fourier integral of (8.1), we get
∞ ∞
– jωt – jωt
F(ω) = ∫– ∞ f Re ( t )e dt + j ∫–∞ fIm ( t )e dt (8.7)

and by Euler’s identity

∞ ∞
F(ω) = ∫– ∞ [ f Re ( t ) cos ωt + f Im ( t ) sin ωt ] dt – j ∫–∞ [ fRe ( t ) sin ωt –fIm ( t ) cos ω t ] dt (8.8)

From (8.2), we see that the real and imaginary parts of F ( ω ) are
∞
F Re ( ω ) = ∫–∞ [ fRe ( t ) cos ωt + fIm ( t ) sin ωt ] dt (8.9)

and
∞
F Im ( ω ) = – ∫–∞ [ fRe ( t ) sin ωt –fIm ( t ) cos ω t ] dt (8.10)

We can derive similar forms for the Inverse Fourier transform as follows:
Substitution of (8.2) into (8.3) yields
∞
1
∫–∞ [ FRe ( ω ) + jFIm ( ω ) ]e
jωt
f ( t ) = ------ dω (8.11)
2π
and by Euler’s identity,
∞
1
f ( t ) = ------
2π ∫–∞ [ FRe ( ω ) cos ωt –FIm ( ω ) sin ωt ] dω (8.12)
∞
1
+ j ------
2π ∫–∞ [ FRe ( ω ) sin ωt + FIm ( ω ) cos ω t ] dω
Therefore, the real and imaginary parts of f ( t ) are
∞
1
f Re ( t ) = ------
2π ∫–∞ [ FRe ( ω ) cos ωt –FIm ( ω ) sin ωt ] dω (8.13)

8-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Special Forms of the Fourier Transform

and
∞
1
f Im ( t ) = ------
2π ∫–∞ [ FRe ( ω ) sin ωt + FIm ( ω ) cos ω t ] dω (8.14)

Now, we will use the above relations to determine the time to frequency domain correspondence for
real, imaginary, even, and odd functions in both the time and the frequency domains. We will show
these in tabular form, as indicated in Table 8.1.

TABLE 8.1 Time Domain and Frequency Domain Correspondence (Refer to Tables 8.2 - 8.7)

f(t) F(ω)

Real Imaginary Complex Even Odd

Real
Real and Even
Real and Odd
Imaginary
Imaginary and Even
Imaginary and Odd

1. Real Time Functions

If f ( t ) is real, (8.9) and (8.10) reduce to
∞
F Re ( ω ) = ∫–∞ fRe ( t ) cos ωt dt (8.15)

and
∞
F Im ( ω ) = – ∫–∞ fRe ( t ) sin ωt dt (8.16)

Conclusion: If f ( t ) is real, F ( ω ) is, in general, complex. We indicate this result with a check mark
in Table 8.2.
We know that any function f ( t ) , can be expressed as the sum of an even and an odd function.
Therefore, we will also consider the cases when f ( t ) is real and even, and when f ( t ) is real and
odd*.

* In our subsequent discussion, we will make use of the fact that the cosine is an even function, while the sine is an
odd function. Also, the product of two odd functions or the product of two even functions will result in an even
function, whereas the product of an odd function and an even function will result in an odd function.

Signals and Systems with MATLAB Applications, Second Edition 8-3

Orchard Publications
Chapter 8 The Fourier Transform

TABLE 8.2 Time Domain and Frequency Domain Correspondence (Refer also to Tables 8.3 - 8.7)

f( t) F(ω)

Real Imaginary Complex Even Odd

Real 
Real and Even
Real and Odd
Imaginary
Imaginary and Even
Imaginary and Odd

a. f Re ( t ) is even

If f Re ( – t ) = f Re ( t ) , the product f Re ( t ) cos ωt , with respect to t , is even, while the product

f Re ( t ) sin ω t is odd. In this case, (8.15) and (8.16) reduce to:

∞
F Re ( ω ) = 2 ∫0 f Re ( t ) cos ωt dt f Re ( t ) = even (8.17)

and
∞
F Im ( ω ) = – ∫–∞ fRe ( t ) sin ωt dt = 0 f Re ( t ) = even (8.18)

Therefore, if f Re ( t ) = even , F ( ω ) is real as seen in (8.17).

To determine whether F ( ω ) is even or odd when f Re ( t ) = even , we must perform a test for
evenness or oddness with respect to ω . Thus, substitution of – ω for ω in (8.17), yields
∞ ∞
F Re ( – ω ) = 2 ∫0 f Re ( t ) cos ( – ω )t dt = 2 ∫0 f Re ( t ) cos ωt dt = F Re ( ω ) (8.19)

Conclusion: If f ( t ) is real and even, F ( ω ) is also real and even. We indicate this result in Table 8.3.

b. f Re ( t ) is odd

If – f Re ( – t ) = f Re ( t ) , the product f Re ( t ) cos ωt , with respect to t , is odd, while the product

f Re ( t ) ( sin ω t ) is even. In this case, (8.15) and (8.16) reduce to:

8-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Special Forms of the Fourier Transform

TABLE 8.3 Time Domain and Frequency Domain Correspondence (Refer also to Tables 8.4 - 8.7)

f(t) F(ω)

Real Imaginary Complex Even Odd

Real 
Real and Even  
Real and Odd
Imaginary
Imaginary and Even
Imaginary and Odd

∞
F Re ( ω ) = ∫–∞ fRe ( t ) cos ωt dt = 0 f Re ( t ) = odd (8.20)

and
∞
F Im ( ω ) = – 2 ∫0 f Re ( t ) sin ωt dt f Re ( t ) = odd (8.21)

Therefore, if f Re ( t ) = odd , F ( ω ) is imaginary.

To determine whether F ( ω ) is even or odd when f Re ( t ) = odd , we perform a test for evenness
or oddness with respect to ω . Thus, substitution of – ω for ω in (8.21), yields
∞ ∞
F Im ( – ω ) = – 2 ∫0 f Re ( t ) sin ( – ω )t dt = 2 ∫0 f Re ( t ) sin ωt dt = – F Im ( ω ) (8.22)

Conclusion: If f ( t ) is real and odd, F ( ω ) is imaginary and odd. We indicate this result in Table 8.4.
2. Imaginary Time Functions
If f ( t ) is imaginary, (8.9) and (8.10) reduce to
∞
F Re ( ω ) = ∫–∞ fIm ( t ) sin ω t dt (8.23)

and
∞
F Im ( ω ) = ∫–∞ fIm ( t ) cos ω t dt (8.24)

Signals and Systems with MATLAB Applications, Second Edition 8-5

Orchard Publications
Chapter 8 The Fourier Transform

TABLE 8.4 Time Domain and Frequency Domain Correspondence (Refer also to Tables 8.5 - 8.7)

f( t) F(ω)

Real Imaginary Complex Even Odd

Real 
Real and Even  
Real and Odd  
Imaginary
Imaginary and Even
Imaginary and Odd

Conclusion: If f ( t ) is imaginary, F ( ω ) is, in general, complex. We indicate this result in Table 8.5.

TABLE 8.5 Time Domain and Frequency Domain Correspondence (Refer also to Tables 8.6 - 8.7)

f( t) F( ω )

Real Imaginary Complex Even Odd

Real 
Real and Even  
Real and Odd  
Imaginary 
Imaginary and Even
Imaginary and Odd

Next, we will consider the cases where f ( t ) is imaginary and even, and f ( t ) is imaginary and odd.
a. f Im ( t ) is even

If f Im ( – t ) = f Im ( t ) , the product f Im ( t ) cos ωt , with respect to t , is even while the product

f Im ( t ) sin ω t is odd. In this case, (8.23) and (8.24) reduce to:

∞
F Re ( ω ) = ∫–∞ fIm ( t ) sin ω t dt = 0 f Im ( t ) = even (8.25)

and
∞
F Im ( ω ) = 2 ∫0 f Im ( t ) cos ω t dt f Im ( t ) = even (8.26)

8-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Special Forms of the Fourier Transform

Therefore, if f Im ( t ) = even , F ( ω ) is imaginary.

To determine whether F ( ω ) is even or odd when f Im ( t ) = even , we perform a test for evenness or
oddness with respect to ω . Thus, substitution of – ω for ω in (8.26) yields
∞
F Im ( – ω ) = 2 ∫0 f Im ( t ) cos ( – ω )t dt
(8.27)
∞
= 2 ∫0 f Im ( t ) cos ωt dt = F Im ( ω )

Conclusion: If f ( t ) is imaginary and even, F ( ω ) is also imaginary and even. We indicate this result in
Table 8.6.
TABLE 8.6 Time Domain and Frequency Domain Correspondence (Refer also to Table 8.7)

f(t) F(ω)

Real Imaginary Complex Even Odd

Real 
Real and Even  
Real and Odd  
Imaginary 
Imaginary and Even  
Imaginary and Odd

b. f Im ( t ) is odd

If – f Im ( – t ) = f Im ( t ) , the product f Im ( t ) cos ωt , with respect to t , is odd, while the product

f Im ( t ) sin ω t is even. In this case, (8.23) and (8.24) reduce to

∞ ∞
F Re ( ω ) = ∫– ∞ f Im ( t ) sin ω t dt = 2 ∫0 f Im ( t ) sin ω t dt f Im ( t ) = odd (8.28)

and
∞
F Im ( ω ) = ∫–∞ fIm ( t ) cos ω t dt = 0 f Im ( t ) = odd (8.29)

Therefore, if f Im ( t ) = odd , F ( ω ) is real.

Signals and Systems with MATLAB Applications, Second Edition 8-7

Orchard Publications
Chapter 8 The Fourier Transform

To determine whether F ( ω ) is even or odd when f Im ( t ) = odd , we perform a test for evenness
or oddness with respect to ω . Thus, substitution of – ω for ω in (8.28) yields
∞ ∞
F Re ( – ω ) = 2 ∫0 f Im ( t ) sin ( – ω )t dt = – 2 ∫0 f Im ( t ) sin ωt dt = – F Re ( ω ) (8.30)

Conclusion: If f ( t ) is imaginary and odd, F ( ω ) is real and odd. We indicate this result in Table
8.7.

TABLE 8.7 Time Domain and Frequency Domain Correspondence (Completed Table)

f(t) F(ω)

Real Imaginary Complex Even Odd

Real 
Real and Even  
Real and Odd  
Imaginary 
Imaginary and Even  
Imaginary and Odd  

Table 8.7 is now complete and shows that if f ( t ) is real (even or odd), the real part of F ( ω ) is even,
and the imaginary part is odd. Then,
F Re ( – ω ) = F Re ( ω ) f ( t ) = Real (8.31)
and
F Im ( – ω ) = – F Im ( ω ) f ( t ) = Real (8.32)
Since,
F ( ω ) = F Re ( ω ) + jF Im ( ω ) (8.33)
it follows that
F ( – ω ) = F Re ( – ω ) + jF Im ( – ω ) = F Re ( ω ) – jF Im ( ω )
or
F ( – ω ) = F∗ ( ω ) f ( t ) = Real (8.34)

Now, if F ( ω ) of some function of time f ( t ) is known, and F ( ω ) is such that F ( – ω ) = F∗ ( ω ) , can

we conclude that f ( t ) is real? The answer is yes; we can verify this with (8.14) which is repeated
below for convenience.

8-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Fourier Transform
∞
1
f Im ( t ) = ------
2π ∫–∞ [ FRe ( ω ) sin ωt + FIm ( ω ) cos ω t ] dω (8.35)

We observe that the integrand of (8.35) is zero since it is an odd function with respect to ω because
both products inside the brackets are odd functions*.
Therefore, f Im ( t ) = 0 , that is, f ( t ) is real.

We can state then, that a necessary and sufficient condition for f ( t ) to be real, is that F ( – ω ) = F∗ ( ω ) .
Also, if it is known that f ( t ) is real, the Inverse Fourier transform of (8.3) can be simplified as fol-
lows:
From (8.13),
∞
1
f Re ( t ) = ------
2π ∫–∞ [ FRe ( ω ) cos ωt –FIm ( ω ) sin ωt ] dω (8.36)

and since the integrand is an even function with respect to ω , we rewrite (8.36) as
∞
1
f Re ( t ) = 2 ------
2π ∫0 [ F Re ( ω ) cos ωt – F Im ( ω ) sin ωt ] dω
(8.37)
∞ ∞
1 1 j [ ωt + ϕ ( ω ) ]
= ---
π ∫0 A ( ω ) cos [ ωt + ϕ ( ω ) ] dω = --- Re
π ∫0 F ( ω )e dω

8.3 Properties and Theorems of the Fourier Transform

1. Linearity
If F 1 ( ω ) is the Fourier transform of f 1 ( t ) , F 2 ( ω ) is the transform of f 2 ( t ) , and so on, the lin-
earity property of the Fourier transform states that

a1 f1 ( t ) + a2 f2 ( t ) + … + an fn ( t ) ⇔ a1 F1 ( ω ) + a2 F2 ( ω ) + … + an Fn ( ω ) (8.38)

where a i is some arbitrary real constant.

Proof:
The proof is easily obtained from (8.1), that is, the definition of the Fourier transform. The proce-
dure is the same as for the linearity property of the Laplace transform in Chapter 2.

* In (8.31) and (8.32), we determined that F Re ( ω ) is even and F Im ( ω ) is odd.

Signals and Systems with MATLAB Applications, Second Edition 8-9

Orchard Publications
Chapter 8 The Fourier Transform

2. Symmetry
If F ( ω ) is the Fourier transform of f ( t ) , the symmetry property of the Fourier transform states that

F ( t ) ⇔ 2πf ( – ω ) (8.39)

that is, if in F ( ω ) , we replace ω with t , we get the Fourier transform pair of (8.39).

Proof:

Since
∞
1
∫–∞ F ( ω )e
jωt
f ( t ) = ------ dω
2π
then,
∞
– j ωt
2πf ( – t ) = ∫–∞ F ( ω )e dω

Interchanging t and ω , we get

∞
– j ωt
2πf ( – ω ) = ∫–∞ F ( t )e dt

and (8.39) follows.

3. Time Scaling
If a is a real constant, and F ( ω ) is the Fourier transform of f ( t ) , then,

ω
f ( at ) ⇔ ----- F ⎛ ---- ⎞
1
a ⎝ a ⎠ (8.40)

that is, the time scaling property of the Fourier transform states that if we replace the variable t in
the time domain by at , we must replace the variable ω in the frequency domain by ω ⁄ a , and
divide F ( ω ⁄ a ) by the absolute value of a .

Proof:
We must consider both cases a > 0 and a < 0 .
For a > 0 ,
∞
– jωt
F { f ( at ) } = ∫–∞ f ( at )e dt (8.41)

We let at = τ ; then, t = τ ⁄ a , and (8.41) becomes

8-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Fourier Transform

τ ω
∞ – jω ⎛ --- ⎞ ∞ – j ⎛ ----⎞ τ
⎝a⎠ τ ⎝ a⎠ 1 ω
d ⎛ --- ⎞ = --- dτ = --- F ⎛ ----⎞
1
F {f(τ)} = ∫–∞ f ( τ )e ⎝ a⎠ a ∫–∞ f ( τ )e a ⎝ a⎠
For a < 0 ,
∞
– jωt
F { f ( –at ) } = ∫–∞ f ( –at )e dt

and making the above substitutions, we find that the multiplying factor is – 1 ⁄ a . Therefore, for
1 ⁄ a we obtain (8.40).
4. Time Shifting
If F ( ω ) is the Fourier transform of f ( t ) , then,

– jωt 0
f ( t – t 0 ) ⇔ F ( ω )e (8.42)

that is, the time shifting property of the Fourier transform states that if we shift the time function
f ( t ) by a constant t 0 , the Fourier transform magnitude does not change, but the term ωt 0 is
added to its phase angle.
Proof:
∞
– jωt
F { f ( t – t0 ) } = ∫–∞ f ( t – t )e 0 dt

We let t – t 0 = τ ; then, t = τ + t 0 , dt = dτ , and thus

∞ – jω ( τ + t 0 ) – jωt 0 ∞
– jω ( τ )
F { f ( t – t0 ) } = ∫– ∞ f ( τ )e dτ = e ∫–∞ f ( τ )e dτ

or
– jωt 0
F { f ( t – t0 ) } = e F(ω)

5. Frequency Shifting
If F ( ω ) is the Fourier transform of f ( t ) , then,

jω 0 t
e f ( t ) ⇔ F ( ω – ω0 ) (8.43)

jω 0 t
that is, multiplication of the time function f ( t ) by e , where ω 0 is a constant, results in shifting
the Fourier transform by ω 0 .

Signals and Systems with MATLAB Applications, Second Edition 8-11

Orchard Publications
Chapter 8 The Fourier Transform

Proof:
∞
F { e jω t f ( t ) }0
= ∫– ∞ e
jω 0 t
f ( t )e
– jωt
dt

or
jω 0 t ∞ –j ( ω – ω0 )
F {e f(t)} = ∫–∞ f ( t )e dt = F ( ω – ω 0 )

Also, from (8.40) and (8.43)

ω–ω
f ( at ) ⇔ ----- F ⎛ ----------------0 ⎞
jω 0 t 1
e
⎝ ⎠
(8.44)
a a

Property 5, that is, (8.43) is also used to derive the Fourier transform of the modulated signals
f ( t ) cos ωt and f ( t ) sin ωt . Thus, from
jω 0 t
e f ( t ) ⇔ F ( ω – ω0 )
and
jω 0 t –j ω0 t
e +e
cos ω 0 t = -------------------------------
2
we get

F ( ω – ω0 ) + F ( ω + ω0 )
f ( t ) cos ω 0 t ⇔ ----------------------------------------------------------
- (8.45)
2

Similarly,

F ( ω – ω0 ) –F ( ω + ω0 )
f ( t ) sin ω 0 t ⇔ -------------------------------------------------------
- (8.46)
j2

6. Time Differentiation
If F ( ω ) is the Fourier transform of f ( t ) , then,
n
d
-------- f ( t ) ⇔ ( jω ) n F ( ω ) (8.47)
n
dt
n
d n
that is, the Fourier transform of -------n- f ( t ) , if it exists, is ( jω ) F ( ω ) .
dt
Proof:
Differentiating the Inverse Fourier transform, we get

8-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Fourier Transform
n n ∞ ∞ n
-------- f ( t ) = -------- ⎛ ------ dω⎞ = ------
d d 1 1 d
∫– ∞ ∫
jωt jωt
n ⎝ 2π
F ( ω )e F ( ω ) -------n- e dω
n ⎠ 2π –∞
dt dt dt
∞ ∞
dω = ( jω ) ⎛ ------ dω⎞
1 n 1
∫– ∞ ∫–∞ F ( ω )e
n jωt jωt
= ------ F ( ω ) ( jω ) e
2π ⎝ 2π ⎠

and (8.47) follows.

7. Frequency Differentiation
If F ( ω ) is the Fourier transform of f ( t ) , then,
n
n d
( – j t ) f ( t ) ⇔ ---------n F ( ω ) (8.48)
dω
Proof:
Using the Fourier transform definition, we get
n n ∞ ∞ n
--------- F ( ω ) = --------- ⎛ – j ωt
dt⎞ =
d d d – j ωt
dω
n
dω
n⎝ ∫– ∞ f ( t )e
⎠ ∫ –∞
f ( t ) ---------n e
dω
dt

∞ ∞
n – j ωt – j ωt
∫– ∞ ∫–∞ f ( t )e
n
= f ( t ) ( –j t ) e dt = ( – j t ) dt

and (8.48) follows.

8. Time Integration
If F ( ω ) is the Fourier transform of f ( t ) , then,
t
F(ω )
∫–∞ f ( τ ) dτ ⇔ ------------
jω
+ πF ( 0 )δ ( ω ) (8.49)

Proof:
We postpone the proof of this property until we derive the Fourier transform of the unit step
function u 0 ( t ) on the next section. In the special case where in (8.49), F ( 0 ) = 0 , then,

t
F(ω)
∫–∞ f ( τ ) dτ ⇔ ------------
jω
(8.50)

and this is easily proved by integrating both sides of the Inverse Fourier transform.
9. Conjugate Time and Frequency Functions
If F ( ω ) is the Fourier transform of the complex function f ( t ) , then,

Signals and Systems with MATLAB Applications, Second Edition 8-13

Orchard Publications
Chapter 8 The Fourier Transform

f∗ ( t ) ⇔ F ∗ ( – ω ) (8.51)

that is, if the Fourier transform of f ( t ) = f Re ( t ) + f Im ( t ) is F ( ω ) , then, the Fourier transform of

f∗ ( t ) = f Re ( t ) – f Im ( t ) is F∗ ( – ω ) .

Proof:
∞ ∞
– jωt – jωt
F(ω) = ∫– ∞ f ( t )e dt = ∫–∞ [ fRe ( t ) + jfIm ( t ) ]e dt

∞ ∞
– jωt – jωt
= ∫– ∞ f Re ( t )e dt + j ∫–∞ fIm ( t )e dt

Then,
∞ ∞
F∗ ( ω ) = ∫– ∞ ∫–∞ fIm ( t )e
jωt jωt
f Re ( t )e dt – j dt

Replacing ω with – ω , we get

∞
– j ωt
F∗ ( – ω ) = ∫–∞ [ fRe ( t ) – jfIm ( t ) ]e dt

and (8.51) follows.

10. Time Convolution
If F 1 ( ω ) is the Fourier transform of f 1 ( t ) , and F 2 ( ω ) is the Fourier transform of f 2 ( t ) , then,

f 1 ( t )∗ f 2 ( t ) ⇔ F 1 ( ω )F 2 ( ω ) (8.52)

that is, convolution in the time domain, corresponds to multiplication in the frequency domain.
Proof:
∞ ∞
– jωt
F { f1 ( t )∗ f2 ( t ) } = ∫–∞ ∫–∞ f ( τ )f2 ( t – τ ) dτ
1 e dt
(8.53)
∞ ∞
– jωt
= ∫–∞ f ( τ ) ∫–∞ f ( t – τ )e
1 2 dt dτ

and letting t – τ = σ , then, dt = dσ , and by substitution into (8.53),

∞ ∞ ∞ ∞
– jωτ – jωσ – jωτ – jωσ
F { f 1 ( t )∗ f 2 ( t ) } = ∫– ∞ f1 ( τ ) ∫– ∞ f 2 ( σ )e e dσ dτ = ∫– ∞ f 1 ( τ )e dτ ∫–∞ f ( σ )e
2 dσ

The first integral above is F 1 ( ω ) while the second is F 2 ( ω ) , and thus (8.52) follows.

8-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Fourier Transform

Alternate Proof:
– jωt 0
We can apply the time shifting property f ( t – t 0 ) ⇔ F ( ω )e into the bracketed integral of
– jωt 0
(8.53); then, replacing it with F 2 ( ω )e , we get
∞ ∞ ∞ – jωt 0
– jωt
F { f 1 ( t )∗ f 2 ( t ) } = ∫– ∞ f1 ( τ ) ∫– ∞ f 2 ( t – τ )e dt dτ = ∫–∞ f ( τ ) dτF
1 2 ( ω )e

∞
– jωt
= ∫–∞ f ( τ )e
1 dτF 2 ( ω ) = F 1 ( ω )F 2 ( ω )

11. Frequency Convolution

If F 1 ( ω ) is the Fourier transform of f 1 ( t ) , and F 2 ( ω ) is the Fourier transform of f 2 ( t ) , then,

1
f 1 ( t )f 2 ( t ) ⇔ ------ F 1 ( ω )∗ F 2 ( ω ) (8.54)
2π

that is, multiplication in the time domain, corresponds to convolution in the frequency domain
divided by the constant 1 ⁄ 2π .
Proof:
∞ ∞ ∞
– jωt 1- – jωt
∫– ∞ ∫ ∫–∞ F ( χ )e
jχt
F { f1 ( t )f2 ( t ) } = [ f 1 ( t ) f 2 ( t ) ]e dt = -----
2π 1 dχ f 2 ( t )e dt
–∞
∞ ∞ ∞
1 – j ( ω – χ )t 1
= ------
2π ∫– ∞ F1 ( χ ) ∫– ∞ f 2 ( t )e dt dχ = ------
2π ∫–∞ F ( χ )F ( ω – χ ) dχ
1 2

and (8.54) follows.

12. Area under f ( t )
If F ( ω ) is the Fourier transform of f ( t ) , then,
∞
F( 0) = ∫– ∞ f ( t ) d t (8.55)

that is, the area under a time function f ( t ) is equal to the value of its Fourier transform evaluated
at ω = 0 .
Proof:
– jωt
Using the definition of F ( ω ) and that e ω=0
= 1 , we see that (8.55) follows.

Signals and Systems with MATLAB Applications, Second Edition 8-15

Orchard Publications
Chapter 8 The Fourier Transform

13. Area under F ( ω )

If F ( ω ) is the Fourier transform of f ( t ) , then,
∞
1
f ( 0 ) = ------
2π ∫–∞ F ( ω ) dω (8.56)

that is, the value of the time function f ( t ) , evaluated at t = 0 , is equal to the area under its Fou-
rier transform F ( ω ) times 1 ⁄ 2π .
Proof:
jωt
In the Inverse Fourier transform of (8.3), we let e t=0
= 1 , and (8.56) follows.

14. Parseval’s Theorem

If F ( ω ) is the Fourier transform of f ( t ) , Parseval’s theorem states that
∞ ∞
1
∫– ∞ ∫– ∞ F ( ω )
2 2
f ( t ) dt = ------ dω (8.57)
2π

that is, if the time function f ( t ) represents the voltage across, or the current through an 1 Ω resis-
tor, the instantaneous power absorbed by this resistor is either v 2 ⁄ R , v 2 ⁄ 1 , v 2 , or i 2 R , i 2 . Then,
the integral of the magnitude squared, represents the energy (in watt-seconds or joules) dissipated
by the resistor. For this reason, the integral is called the energy of the signal. Relation (8.57) then,
states that if we do not know the energy of a time function f ( t ) , but we know the Fourier trans-
form of this function, we can compute the energy without the need to evaluate the Inverse Fou-
rier transform.
Proof:
From the frequency convolution property,
1
f 1 ( t )f 2 ( t ) ⇔ ------ F 1 ( ω )∗ F 2 ( ω )
2π
or
∞ ∞
– jωt 1
F { f1 ( t )f2 ( t ) } = ∫– ∞ [ f 1 ( t )f 2 ( t ) ]e dt = ------
2π ∫–∞ F ( χ )F ( ω – χ ) dχ
1 2 (8.58)

Since (8.58) must hold for all values of ω , it must also be true for ω = 0 , and under this condi-
tion, it reduces to
∞ ∞
1
∫ –∞
[ f 1 ( t )f 2 ( t ) ] dt = ------
2π ∫–∞ F ( χ )F ( –χ ) dχ
1 2 (8.59)

8-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transform Pairs of Common Functions

For the special case where f 2 ( t ) = f 1∗ ( t ) , and the conjugate functions property f∗ ( t ) ⇔ F∗ ( – ω ) ,
by substitution into (8.59), we get:
∞ ∞ ∞
1 1
∫ –∞
[ f ( t )f∗ ( t ) ] dt = ------
2π ∫ –∞
F ( ω )F∗ [ – ( – ω ) ] dω = ------
2π ∫–∞ F ( ω )F∗ ( ω ) dω
Since f ( t )f∗ ( t ) = f ( t ) and F ( ω )F∗ ( ω ) = F ( ω ) , Parseval’s theorem is proved.
2 2

The Fourier transform properties and theorems are summarized in Table 8.8.

8.4 Fourier Transform Pairs of Common Functions

In this section, we will derive the Fourier transforms of common time functions.
1.
δ(t) ⇔ 1 (8.60)
Proof:
The sifting theorem of the delta function states that
∞

∫–∞ f ( t )δ ( t – t ) dt
0 = f ( t0 )

and if f ( t ) is defined at t = 0 , then,

∞

∫–∞ f ( t )δ ( t ) dt = f( 0)

By the definition of the Fourier transform

∞
– jωt – jωt
F(ω) = ∫–∞ δ ( t )e dt = e t=0
= 1

and (8.60) follows.

We will use the notation f ( t ) ↔ F ( ω ) to show the time domain to frequency domain correspon-
dence. Thus, (8.60) may also be denoted as in Figure 8.1.
f(t) F(ω) 1
δ(t)

0 t 0 ω

Figure 8.1. The Fourier transform of the delta function

Signals and Systems with MATLAB Applications, Second Edition 8-17

Orchard Publications
Chapter 8 The Fourier Transform

TABLE 8.8 Fourier Transform Properties and Theorems

Property f( t) F(ω)
Linearity a1 f1 ( t ) + a2 f2 ( t ) + … a1 F1 ( ω ) + a2 F2 ( ω ) + …
Symmetry F(t) 2πf ( – ω )
Time Scaling f ( at ) ω
----- F ⎛ ----⎞
1
a ⎝ a⎠
Time Shifting f ( t – t0 ) – jωt 0
F ( ω )e
Frequency Shifting jω 0 t F ( ω – ω0 )
e f(t)
Time Differentiation n n
d ( jω ) F ( ω )
--------- f ( t )
n
dt
Frequency Differentiation n n
( – jt ) f ( t ) d
----------- F ( ω )
n
dω
Time Integration t F ( ω ) + πF ( 0 )δ ( ω )
∫– ∞
------------
f ( τ ) dτ jω

Conjugate Functions f∗ ( t ) F∗ ( – ω )
Time Convolution f 1 ( t )∗ f 2 ( t ) F1 ( ω ) ⋅ F2 ( ω )

Frequency Convolution f1 ( t ) ⋅ f2 ( t ) 1
------ F 1 ( ω )∗ F 2 ( ω )
2π
Area under f(t) ∞
F(0) = ∫–∞ f ( t ) dt
Area under F(w) ∞
∫
1
f ( 0 ) = ------ F ( ω ) dω
2π – ∞

Parseval’s Theorem ∞ ∞
∫ ∫
2 1 2
f ( t ) dt = ------ F ( ω ) dω
–∞ 2π –∞

Likewise, the Fourier transform for the shifted delta function δ ( t – t 0 ) is

– jωt 0
δ ( t – t0 ) ⇔ e (8.61)

8-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transform Pairs of Common Functions

2.
1 ⇔ 2πδ ( ω ) (8.62)
Proof:
–1 ∞ ∞
1
∫– ∞ ∫–∞ δ ( ω )e
jωt jωt jωt
F { 2πδ ( ω ) } = ------ 2πδ ( ω )e dω = dω = e ω=0
= 1
2π

and (8.62) follows.

The f ( t ) ↔ F ( ω ) correspondence is also shown in Figure 8.2.

f(t) F(ω)
1
2πδ ( ω )

0 t 0 ω

Figure 8.2. The Fourier transform of unity

Also, by direct application of the Inverse Fourier transform, or the frequency shifting property
and (8.62), we derive the transform
jω 0 t
e ⇔ 2πδ ( ω – ω 0 ) (8.63)

The transform pairs of (8.62) and (8.63) can also be derived from (8.60) and (8.61) by using the
symmetry property F ( t ) ⇔ 2πf ( – ω )
3.

1 jω 0 t –j ω0 t
cos ω 0 t = --- ( e +e ) ⇔ πδ ( ω – ω 0 ) + πδ ( ω + ω 0 ) (8.64)
2

Proof:
This transform pair follows directly from (8.63). The f ( t ) ↔ F ( ω ) correspondence is also shown
in Figure 8.3.
cosωω
cos 0 t0
t F Re ( ω )

π π
t t

−ω0 0 ω0 ω

Figure 8.3. The Fourier transform of f ( t ) = cos ω 0 t

Signals and Systems with MATLAB Applications, Second Edition 8-19

Orchard Publications
Chapter 8 The Fourier Transform

We know that cos ω 0 t is real and even function of time, and we found out that its Fourier trans-
form is a real and even function of frequency. This is consistent with the result in Table 8.7.
4.

1 jω0 t –j ω0 t
sin ω 0 t = ----- ( e –e ) ⇔ jπδ ( ω – ω 0 ) – jπ δ ( ω + ω 0 ) (8.65)
j2

Proof:
This transform pair also follows directly from (8.63). The f ( t ) ↔ F ( ω ) correspondence is also
shown in Figure 8.4.

sin ω 0 t F Im ( ω )
π
tt −ω0
0 ω0 ω
−π

Figure 8.4. The Fourier transform of f ( t ) = sin ω 0 t

We know that sin ω 0 t is real and odd function of time, and we found out that its Fourier trans-
form is an imaginary and odd function of frequency. This is consistent with the result in Table
8.7.
5.
2
sgn ( t ) = u 0 ( t ) – u 0 ( – t ) ⇔ ------ (8.66)
jω

where sgn ( t ) denotes the signum function shown in Figure 8.5.

f(t)
1

−1

Figure 8.5. The signum function

Proof:
To derive the Fourier transform of the sgn ( t ) function, it is convenient to express it as an expo-
nential that approaches a limit as shown in Figure 8.6.

8-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transform Pairs of Common Functions

f(t)
1
– at
e u0 ( t )
0

– at −1
–e u0 ( –t )

Figure 8.6. The signum function as an exponential approaching a limit

Then,
– at at
sgn ( t ) = lim [ e u0 ( t ) – e u0 ( –t ) ] (8.67)
a→0
and
0 ∞
at – j ωt – a t – j ωt
F { sgn ( t ) } = lim
a→0 ∫– ∞ –e e dt + ∫0 e e dt

0 ∞
( a – j ω )t – ( a + jω ) t
= lim
a→0 ∫– ∞ –e dt + ∫0 e dt (8.68)

1 1 –1 1 2
= lim --------------- + --------------- = --------- + ------ = ------
a→0 a – jω a + jω – jω jω jω

The f ( t ) ↔ F ( ω ) correspondence is also shown in Figure 8.7.

f( t)
F Im ( ω )
1
0
t ω
0
−1

Figure 8.7. The Fourier transform of sgn ( t )

We now know that sgn ( t ) is real and odd function of time, and we found out that its Fourier
transform is an imaginary and odd function of frequency. This is consistent with the result in
Table 8.7.
6.
1-
u 0 ( t ) ⇔ πδ ( ω ) + ----- (8.69)
jω

Signals and Systems with MATLAB Applications, Second Edition 8-21

Orchard Publications
Chapter 8 The Fourier Transform

Proof:
If we attempt to verify the transform pair of (8.69) by direct application of the Fourier transform
definition, we will find that

∞ ∞ – jωt ∞
– jωt – jωt
F(ω ) = ∫–∞ f ( t )e dt = F ( ω ) = ∫0 e dt = e-----------
– jω 0
(8.70)

– jωt – jωt
but we cannot say that e approaches 0 as t → ∞ , because e = 1 ∠– ωt , that is, the magni-
– jωt
tude of e is always unity, and its angle changes continuously as t assumes greater and greater
values. Since the upper limit cannot be evaluated, the integral of (8.70) does not converge.
To work around this problem, we will make use of the sgn ( t ) function which we express as

sgn ( t ) = 2u 0 ( t ) – 1 (8.71)

This expression is derived from the waveform of Figure 8.8 below.

f(t) 2

0
t

−1
Figure 8.8. Alternate expression for the signum function

We rewrite (8.71) as
1 1 1
u 0 ( t ) = --- ( 1 + sgn ( t ) ) = --- + --- sgn ( t ) (8.72)
2 2 2

and since we know that 1 ⇔ 2π δ ( ω ) and sgn ( t ) ⇔ 2 ⁄ ( jω ) , by substitution of these into (8.72)
we get
1
u 0 ( t ) ⇔ πδ ( ω ) + ------
jω
and this is the same as(8.69). This is a complex function in the frequency domain whose real part
is π δ ( ω ) and imaginary part – 1 ⁄ ω .
The f ( t ) ↔ F ( ω ) correspondence is also shown in Figure 8.9.

8-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transform Pairs of Common Functions

f(t)
F Im ( ω )
1 π
F Re ( ω )
t ω

F Im ( ω )

Figure 8.9. The Fourier transform of the unit step function

Since u 0 ( t ) is real but neither even nor odd function of time, its Fourier transform is a complex
function of frequency as shown in (8.69). This is consistent with the result in Table 8.7.
Now, we will prove the time integration property of (8.49), that is,
t
F(ω )
∫–∞ f ( τ ) dτ ⇔ ------------
jω
+ πF ( 0 )δ ( ω )

as follows:
By the convolution integral,
t
u 0 ( t )∗ f ( t ) = ∫–∞ f ( τ )u ( t – τ ) dτ
0

and since u 0 ( t – τ ) = 1 for t > τ , and it is zero otherwise, the above integral reduces to
t
u 0 ( t )∗ f ( t ) = ∫– ∞ f ( τ ) d τ
Next, by the time convolution property,
u 0 ( t )∗ f ( t ) ⇔ U 0 ( ω ) ⋅ F ( ω )
and since
1
U 0 ( ω ) = πδ ( ω ) + ------
jω
using these results and the sampling property of the delta function, we get

U 0 ( ω ) ⋅ F ( ω ) = ⎛ πδ ( ω ) + -----
1-⎞ F ( ω ) = = πδ ( ω )F ( ω ) + F ( ω ) = πF ( 0 )δ ( ω ) + ------------
------------ F(ω)
⎝ jω ⎠ jω jω

Thus, the time integration property is proved.

Signals and Systems with MATLAB Applications, Second Edition 8-23

Orchard Publications
Chapter 8 The Fourier Transform

7.

– jω 0 t 1
e u 0 ( t ) ⇔ πδ ( ω – ω 0 ) + ---------------------- (8.73)
j ( ω – ω0 )

Proof:
From the Fourier transform of the unit step function,
1-
u 0 ( t ) ⇔ πδ ( ω ) + -----
jω
and the frequency shifting property,
jω 0 t
e f ( t ) ⇔ F ( ω – ω0 )
we obtain (8.73).

8.
π 1 1
u 0 ( t ) cos ω 0 t ⇔ --- [ δ ( ω – ω 0 ) + δ ( ω + ω 0 ) ] + -------------------------- + ---------------------------
2 2 j ( ω – ω 0 ) 2j ( ω + ω 0 )
(8.74)
π jω -
⇔ --- [ δ ( ω – ω 0 ) + δ ( ω + ω 0 ) ] + -----------------
2 ω –ω
2 2
0

Proof:
We first express the cosine function as
1 jω0 t –j ω0 t
cos ω 0 t = --- ( e +e )
2
From (8.73),
– jω 0 t 1 -
e u 0 ( t ) ⇔ πδ ( ω – ω 0 ) + ----------------------
j ( ω – ω0 )
and
jω 0 t 1
e u 0 ( t ) ⇔ πδ ( ω + ω 0 ) + -----------------------
j ( ω + ω0 )
Now, using
1
u 0 ( t ) ⇔ πδ ( ω ) + ------
jω
we get (8.74).
9.
π ω
2
u 0 ( t ) sin ω 0 t ⇔ ----- [ δ ( ω – ω 0 ) + δ ( ω + ω 0 ) ] + -----------------
-2 (8.75)
j2 ω –ω
2
0

8-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Finding the Fourier Transform from Laplace Transform

Proof:
We first express the sine function as
1 jω 0 t – j ω 0 t
sin ω 0 t = ----- ( e –e )
j2
From (8.73),
– jω 0 t 1
e u 0 ( t ) ⇔ πδ ( ω – ω 0 ) + -----------------------
j ( ω – ω0 )
and
jω 0 t 1
e u 0 ( t ) ⇔ πδ ( ω + ω 0 ) + -----------------------
j ( ω + ω0 )
Using
1-
u 0 ( t ) ⇔ πδ ( ω ) + -----
jω

we obtain (8.75).

8.5 Finding the Fourier Transform from Laplace Transform

If a time function f ( t ) is zero for t ≤ 0 , we can obtain the Fourier transform of f ( t ) from the one-
sided Laplace transform of f ( t ) by substitution of s with jω .

Example 8.1
– αt
It is known that L [ e 1
u 0 ( t ) ] = ------------ . Compute F { e–αt u0 ( t ) } .
s+α
Solution:

F { e–αt u0 ( t ) } = L [e
– αt
u0 ( t ) ]
s = jω
1
= ------------
s+α
1
= ----------------
jω + α
s = jω

Thus, we have obtained the following Fourier transform pair.

– αt 1
e u 0 ( t ) ⇔ ---------------- (8.76)
jω + α

Example 8.2
It is known that
– αt s+α
L [(e cos ω 0 t )u 0 ( t ) ] = --------------------------------
2 2
( s + α ) + ω0

Signals and Systems with MATLAB Applications, Second Edition 8-25

Orchard Publications
Chapter 8 The Fourier Transform

Compute F { ( e–αt cos ω0 t )u0 ( t ) }

Solution:

F { ( e–αt cos ω0 t )u0 ( t ) } = L [(e

– αt
cos ω 0 t )u 0 ( t ) ]
s = jω
s+α jω + α
= -------------------------------- = ------------------------------------
2 2 2 2
( s + α ) + ω0 ( jω + α ) + ω 0
s = jω

Thus, we have obtained the following Fourier transform pair.

– αt jω + α
(e cos ω 0 t )u 0 ( t ) ⇔ -----------------------------------
2
-
2
(8.77)
( jω + α ) + ω 0

We can also find the Fourier transform of a time function f ( t ) that has non-zero values for t < 0 ,
and it is zero for all t > 0 . But because the one-sided Laplace transform does not exist for t < 0 , we
must first express the negative time function in the t > 0 domain, and compute the one-sided
Laplace transform. Then, the Fourier transform of f ( t ) can be found by substituting s with – jω . In
other words, when f ( t ) = 0 for t ≥ 0 , and f ( t ) ≠ 0 for t < 0 , we use the substitution

F { f(t) } = L [ f ( –t ) ] s = –j ω
(8.78)

Example 8.3
–a t
Compute the Fourier transform of f ( t ) = e
a. using the Fourier transform definition
b. by substitution into the Laplace transform equivalent
Solution:
a. Using the Fourier transform definition, we get
0 ∞ 0 ∞
( a – jω )t – ( a + jω ) t
F { e –a t } = ∫– ∞ e e
at – jωt
dt + ∫0 e
– a t – jωt
e dt = ∫– ∞ e dt + ∫0 e dt

1 1 2
= --------------- + --------------- = ------------------
a – jω a + jω ω 2 + a 2

and thus we have the transform pair

8-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transforms of Common Waveforms

–a t 2
e ⇔ -----------------
2
-
2
(8.79)
ω +a

b. By substitution into the Laplace transform equivalent, we get

F { e–a t } = L [e
– at
] s = jω
+ L [e ]
at
s = –j ω
1-
= ----------
s+a
1-
+ ----------
s+a
s = jω s = –j ω
1 1 2
= --------------- + -------------------- = ------------------
jω + a – jω + a ω 2 + a 2

and this result is the same as (8.79).

We observe that since f ( t ) is real and even, F ( ω ) is also real and even.

8.6 Fourier Transforms of Common Waveforms

In this section, we will derive the Fourier transform of some common time domain waveforms.

Example 8.4
Derive the Fourier transform of the pulse
f ( t ) = A [ u0 ( t + T ) – u0 ( t – T ) ] (8.80)
Solution:
The pulse of (8.80) is shown in Figure 8.10.

f(t)
A

t
−T 0 T
Figure 8.10. Pulse for Example 8.4

Using the definition of the Fourier transform, we get

∞ T – jωt T
– jωt – jωt
F(ω) = ∫–∞ f ( t )e dt = ∫–T Ae dt = Ae
---------------
– jω –T

– jωt – T – jωt
(e
jωt
sin ωT sin ωT
Ae -
= -------------- =A –e -) = 2A --------------- = 2AT ---------------
-----------------------------------
jω T
jω ω ωT

Signals and Systems with MATLAB Applications, Second Edition 8-27

Orchard Publications
Chapter 8 The Fourier Transform

We observe that the transform of this pulse has the ( sin x ) ⁄ x form, and has its maximum value 2AT
at ωT = 0 *.
Thus, we have the waveform pair

sin ωT
A [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ 2AT --------------- (8.81)
ωT

The f ( t ) ↔ F ( ω ) correspondence is also shown in Figure 8.11, where we observe that the ω axis
crossings occur at values of ωT = ± nπ where n is an integer.

F(ω )
f(t)
A

−π π
t −2π 2π ωT
−T 0 T 0

Figure 8.11. Fourier transform of f ( t ) = A [ u 0 ( t + T ) – u 0 ( t – T ) ]

We also observe that since f ( t ) is real and even, F ( ω ) is also real and even.

Example 8.5
Derive the Fourier transform of the pulse of Figure 8.12.

f(t)
A

t
0 2T
Figure 8.12. Pulse for Example 8.5

Solution:
The expression for the given pulse is
f ( t ) = A [ u 0 ( t ) – u 0 ( t – 2T ) ] (8.82)

sin x- = 1
* We recall that lim ---------
x→0 x

8-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transforms of Common Waveforms

Using the definition of the Fourier transform, we get

∞ 2T – jωt 2T – jωt 0 – jω2T

– jωt – jωt A(1 – e )
∫– ∞ ∫0
Ae Ae
F(ω) = f ( t )e dt = Ae dt = --------------- = --------------- = --------------------------------
– jω 0
jω 2T
jω

and making the substitutions

jωT – jωT
1 = e ⋅e
(8.83)
– jω2T – j ωT – jωT
e = e ⋅e
we get
– jωT jωT – jωT
–e – jωT ⎛ sin ωT ⎞ sin ωT- ⎞
F ( ω ) = ---------------- ⎛ ----------------------------⎞ = 2Ae --------------- = 2ATe –jωT ⎛ --------------
Ae e
⎝ ⎠ ⎝ ⎠ ⎝
(8.84)
ω j ω ωT ⎠

Alternate Solution:
We can obtain the Fourier transform of (8.82) using the time shifting property, i.e,
– jωt 0
f ( t – t 0 ) ⇔ F ( ω )e

sin ωT – jωT
and the result of Example 8.4. Thus, multiplying 2AT --------------- by e , we obtain (8.84).
ωT

We observe that F ( ω ) is complex* since f ( t ) is neither an even nor an odd function.

Example 8.6
Derive the Fourier transform of the waveform of Figure 8.13.
f(t) 2A

t
−T 0 T 2T
Figure 8.13. Waveform for Example 8.6
Solution:
The given waveform can be expressed as
f ( t ) = A [ u 0 ( t + T ) + u 0 ( t ) – u 0 ( t – T ) – u 0 ( t – 2T ) ] (8.85)

* We recall that e –j ω T consists of a real and an imaginary part.

Signals and Systems with MATLAB Applications, Second Edition 8-29

Orchard Publications
Chapter 8 The Fourier Transform

and this is precisely the sum of the waveforms of Examples 8.4 and 8.5. We also observe that this
waveform is obtained by the graphical addition of the waveforms of Figures 8.10 and 8.12. There-
fore, we will apply the linearity property to obtain the Fourier transform of this waveform.
We denote the transforms of Examples 8.4 and 8.5 as F 1 ( ω ) and F 2 ( ω ) respectively, and we get

sin ωT – jωT ⎛ sin ωT⎞

F ( ω ) = F 1 ( ω ) + F 2 ( ω ) = 2AT --------------- + 2ATe ---------------
ωT ⎝ ωT ⎠
ωT ωT ωT
–j ------- ⎛ j ------- – j -------⎞
– jωT sin ωT 2 2 2 sin ωT
= 2AT ( 1 + e ) --------------- = 2ATe ⎜e +e ⎟ --------------- (8.86)
ωT ⎝ ⎠ ωT
ωT
– j -------
ωT sin ωT
cos ⎛ ------- ⎞ ---------------
2
= 4ATe
⎝ 2 ⎠ ωT

We observe that F ( ω ) is complex since f ( t ) of (8.85) is neither an even nor an odd function.

Example 8.7
Derive the Fourier transform of
f ( t ) = A cos ω 0 t [ u 0 ( t + T ) – u 0 ( t – T ) ] (8.87)

Solution:
From (8.45),
F ( ω – ω0 ) + F ( ω + ω0 )
f ( t ) cos ω 0 t ⇔ ----------------------------------------------------------
-
2
and from (8.81),
sin ωT
A [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ 2AT ---------------
ωT
Then,

sin [ ( ω – ω 0 )T ] sin [ ( ω + ω 0 )T ]
A cos ω 0 t [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ AT ------------------------------------
- + -------------------------------------- (8.88)
( ω – ω 0 )T ( ω + ω 0 )T

We also observe that since f ( t ) is real and even, F ( ω ) is also real and even*.

Example 8.8
Derive the Fourier transform of a periodic time function with period T .

* The sin x ⁄ x is an even function.

8-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Fourier Transforms of Common Waveforms

Solution:
From the definition of the exponential Fourier series,
∞ jnω 0 t
f(t) = ∑ Cn e
n = –∞
(8.89)

where ω 0 = 2π ⁄ T , and recalling that

jω 0 t
e ⇔ 2πδ ( ω – ω 0 )
we get
jω 0 t
Cn e ⇔ 2πC 1 δ ( ω – ω 0 )
j2ω 0 t
Cn e ⇔ 2πC 2 δ ( ω – 2ω 0 )
(8.90)
…
jnω 0 t
Cn e ⇔ 2πC n δ ( ω – nω 0 )

Taking the Fourier transform of (8.89), and applying the linearity property for the transforms of
(8.90), we get
∞ ∞ ∞
F ⎧⎨ ∑
jnω 0 t ⎫ jnω 0 t
F { f(t)} =
⎩
Cn e ⎬ =
⎭
∑ Cn F { e } = 2π ∑
n = –∞
C n δ ( ω – nω )
0 (8.91)
n = –∞ n = –∞

The line spectrum of the Fourier transform of (8.91) is shown in Figure 8.14.

F( ω)
2πC –2 2πC 0
2πC 2
2πC – 4 2πC – 1 2πC 1
2πC 4
2πC – 3 2πC 3
..... .....
ω
– 4ω 0 – 3 ω 0 – 2 ω 0 – ω 0 0 ω0 2 ω0 3 ω0 4 ω0

Figure 8.14. Line spectrum for relation (8.91)

The line spectrum of Figure 8.14 reveals that the Fourier transform of a periodic time function, con-
sists of a train of equally spaced delta functions. The strength of each δ ( ω – nω 0 ) is equal to 2π
times the coefficient C n .

Example 8.9
Derive the Fourier transform of the periodic time function

Signals and Systems with MATLAB Applications, Second Edition 8-31

Orchard Publications
Chapter 8 The Fourier Transform

∞
f( t) = A ∑
n = –∞
δ ( t – nT ) (8.92)

Solution:
This function consists of a train of equally spaced delta functions in the time domain, and each has
the same strength A , as shown in Figure 8.15.

f( t)
Α

..... .....
t
−4Τ −3Τ −2Τ −Τ 0 Τ 2Τ 3Τ 4Τ

Figure 8.15. Waveform for Example 8.9

Since this is a periodic function of time, its Fourier transform is as derived in the previous example,
that is, expression (8.91). Then,
∞
F ( ω ) = 2π ∑
n = –∞
C n δ ( ω – nω 0 ) (8.93)

where ω 0 = 2π ⁄ T , and C n is found from the exponential Fourier series

T⁄2 – jnω 0 t
1
C n = ---
T ∫–T ⁄ 2 f ( t )e dt (8.94)

From the waveform of Figure 8.15, we observe that, within the limits of integration from – T ⁄ 2 to
+T ⁄ 2 , there is only the impulse δ ( t ) at the origin. Therefore, replacing f ( t ) with δ ( t ) and using the
sifting property of the delta function, we get
T⁄2 – jnω 0 t
1
∫–T ⁄ 2 δ ( t )e
1
C n = --- dt = --- (8.95)
T T

Thus, we see that all C n coefficients are equal to 1 ⁄ T , and (8.93) with ω 0 = 2π ⁄ T reduces to

2π ∞
F ( ω ) = ------ ∑
T n = –∞
δ ( ω – nω 0 ) (8.96)

The Fourier transform of the waveform of Figure 8.15 is shown in Figure 8.16.

8-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Using MATLAB to Compute the Fourier Transform

F(ω)

..... .....
ω
– 4ω 0 – 3ω 0 – 2ω 0 – ω 0 0 ω0 2ω 0 3ω 0 4ω 0

Figure 8.16. The Fourier transform of a train of equally spaced delta functions

Figure 8.16 shows that the Fourier transform of a periodic train of equidistant delta functions in the
time domain, is a periodic train of equally spaced delta functions in the frequency domain. This
result is the basis for the proof of the sampling theorem which states that a time function f ( t ) can be
uniquely determined from its values at a sequence of equidistant points in time.

8.7 Using MATLAB to Compute the Fourier Transform

MATLAB has the built-in fourier and ifourier functions to compute the Fourier transform and its
inverse. Their descriptions and examples, can be displayed with the help fourier and help ifourier
commands. In examples 8.10 through 8.13 we present some Fourier transform pairs, and how they
are verified with MATLAB.

Example 8.10
1 2 1 2
– --- t – --- ω
2 2
e ⇔ 2πe (8.97)
This time function, like the time function of Example 8.9, is its own Fourier transform multiplied by
the constant 2π .
syms t v w x; ft=exp(−t^2/2); Fw=fourier(ft)
Fw =
2^(1/2)*pi^(1/2)*exp(-1/2*w^2)
pretty(Fw)
1/2 1/2 2
2 pi exp(- 1/2 w )
% Check answer by computing the Inverse using "ifourier"
ft=ifourier(Fw)
ft =
exp(-1/2*x^2)

Signals and Systems with MATLAB Applications, Second Edition 8-33

Orchard Publications
Chapter 8 The Fourier Transform

Example 8.11
1 2
2 – ω ---
–t 1 4
te ⇔ --- j πωe (8.98)
2
syms t v w x; ft=t*exp(−t^2); Fw=fourier (ft)
Fw =
-1/2*i*pi^(1/2)*w*exp(-1/4*w^2)
pretty(Fw)
1/2 2
- 1/2 i pi w exp(- 1/4 w )

Example 8.12
–t 1
– e u 0 ( t ) + 3δ ( t ) ⇔ – --------------- + 3 (8.99)
jω + 1
syms t v w x; fourier(sym('−exp(−t)*Heaviside(t)+3*Dirac(t)'))
ans =
-1/(1+i*w)+3

Example 8.13
1-
u 0 ( t ) ⇔ πδ ( ω ) + ----- (8.100)
jω
syms t v w x; u0=sym('Heaviside(t)'); Fw=fourier(u0)
Fw =
pi*Dirac(w)-i/w
We summarize the most common Fourier transform pairs in Table 8.9.

8.8 The System Function and Applications to Circuit Analysis

We recall from Chapter 6 that, by definition, the convolution integral is
∞
h ( t )*u(t) = ∫–∞ u ( t – τ )h ( τ ) dτ (8.101)

We let
g ( t ) = f ( t )∗ h ( t ) (8.102)
and recalling that convolution in the time domain corresponds to multiplication in the frequency
domain, we get

8-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The System Function and Applications to Circuit Analysis

TABLE 8.9 Common Fourier transform pairs

f( t) F(ω)
δ(t) 1
δ ( t – t0 ) – jωt 0
e
1 2π δ ( ω )

e
– jωt 0 2π δ ( ω – ω 0 )

sgn ( t ) 2 ⁄ ( jω )
u0 ( t ) 1
------ + π δ ( ω )
jω
cos ω 0 t π [ δ ( ω – ω0 ) + δ ( ω + ω0 ) ]

sin ω 0 t jπ [ δ ( ω – ω 0 ) – δ ( ω + ω 0 ) ]

– at 1
e u0 ( t ) ---------------
jω + a
a>0 a>0

– at 1
te u0 ( t ) ----------------------
2
( jω + a )
a>0
a>0

e
– at
cos ω 0 tu 0 ( t ) jω + a
-----------------------------------
2 2
a>0 ( jω + a ) + ω
a>0

– at ω
e sin ω 0 tu 0 ( t ) ----------------------------------
2
-
2
( jω + a ) + ω
a>0
a>0
A [ u0 ( t + T ) – u0 ( t – T ) ] sin ωT
2AT ---------------
ωT

f ( t )∗ h ( t ) = g ( t ) ⇔ G ( ω ) = F ( ω ) ⋅ H ( ω ) (8.103)

We call H ( ω ) the system function. From (8.103), we see that the system function H ( ω ) and the
impulse response h ( t ) form the Fourier transform pair

Signals and Systems with MATLAB Applications, Second Edition 8-35

Orchard Publications
Chapter 8 The Fourier Transform

h(t) ⇔ H(ω) (8.104)

Therefore, if we know the impulse response h ( t ) , we can compute the response g ( t ) of any input
f ( t ) , by multiplication of the Fourier transforms H ( ω ) and F ( ω ) to obtain G ( ω ) . Then, we take the
Inverse Fourier transform of G ( ω ) to obtain the response g ( t ) .

Example 8.14
For the linear circuit of Figure 8.17 (a) below, it is known that the impulse response is as shown in
(b). Use the Fourier transform method to compute the response g ( t ) when the input f ( t ) is as
shown in (c).

f ( t ) = 2 [ u0 ( t ) – u0 ( t – 3 ) ]
3
+ Linear + – 2t
f(t) g(t) h ( t ) = 3e 2
− Circuit −
t t
(a) 0 (b) 0 1 (c) 2 3

Figure 8.17. Figure for Example 8.14.

Solution:
To facilitate the computations, we denote the input as f ( t ) = f 1 ( t ) + f 2 ( t ) where
f 1 ( t ) = 2u 0 ( t )
and
f 2 ( t ) = – 2u 0 ( t – 3 )

The system function H ( ω ) is the Fourier transform of the impulse response h ( t ) . Thus,
3
F {h(t)} = H ( ω ) = ---------------
jω + 2

Let F 1 ( ω ) be the Fourier transform of f 1 ( t ) , that is,

= F 1 ( ω ) = 2 ⎛ πδ ( ω ) + ------⎞
1
F { f1 ( t ) } ⎝ jω⎠
Then,
G 1 ( ω ) = H ( ω ) ⋅ F 1 ( ω ) = --------------- ⋅ 2 ⎛ πδ ( ω ) + ------⎞
3 1
jω + 2 ⎝ jω⎠
or

8-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The System Function and Applications to Circuit Analysis

6π 3
G 1 ( ω ) = --------------- δ ( ω ) + -------------------------- (8.105)
jω + 2 jω ( jω + 2 )

To evaluate the first term of (8.105), we apply the sampling property of the delta function, i.e.,
X(ω) ⋅ δ(ω) = X(0) ⋅ δ(ω)

and this term reduces to 3πδ ( ω ) or 1.5 [ 2πδ ( ω ) ] . Since 1 ⇔ 2πδ ( ω ) , the Inverse Fourier trans-
form of this term is
–1
F { 1.5 [ 2πδ ( ω ) ] } = 1.5 (8.106)
To find the Inverse Fourier transform of the second term in (8.105), we use partial fraction expan-
sion. Thus,
3
-------------------------- = 1.5 1.5 -
------- – -------------------
jω ( jω + 2 ) jω ( jω + 2 )
and therefore,
– 1 ⎧ 1.5 1.5 ⎫ –1 ⎧ 2 1.5 ⎫
g 1 ( t ) = 1.5 + F ⎨ ------- – -------------------- ⎬ = 1.5 + F - – -------------------
⎨ 0.75 ----- -⎬
⎩ jω ( jω + 2) ⎭ ⎩ jω ( jω + 2) ⎭
or
– 2t
g 1 ( t ) = 1.5 + 0.75 sgn ( t ) – 1.5e u0 ( t )
(8.107)
– 2t – 2t
= 1.5 + 0.75 [ 2u 0 ( t ) – 1 ] – 1.5e u 0 ( t ) = 0.75 + 1.5 ( 1 – e )u 0 ( t )

Next, we denote the response due to the second term of the input as g 2 ( t ) , and replacing u 0 ( t ) in
(8.107) with u 0 ( t – 3 ) , we get
–2 ( t – 3 )
g 2 ( t ) = 0.75 + 1.5 ( 1 – e )u 0 ( t – 3 ) (8.108)
Now, we combine (8.107) with (8.108), and we get
g ( t ) = g1 ( t ) –g2 ( t )
or
– 2t –2 ( t – 3 )
g ( t ) = 1.5 { ( 1 – e )u 0 ( t ) – ( 1 – e )u 0 ( t – 3 ) }

Example 8.15
For the circuit of Figure 8.18, use the Fourier transform method, and the system function H ( ω ) to
compute v L ( t ) . Assume i L ( 0 − ) .

Signals and Systems with MATLAB Applications, Second Edition 8-37

Orchard Publications
Chapter 8 The Fourier Transform

4Ω L
+
vL ( t )
+
−
v in ( t )
`
2H −

– 3t
v in ( t ) = 5e u0 ( t )

Figure 8.18. Circuit for Example 8.15

Solution:
We will find the system function H ( ω ) from the phasor equivalent circuit shown in Figure 8.19.

4Ω L
+

+
− `
j2ω −
V (ω) = V L out ( ω )

V in ( ω )

Figure 8.19. Phasor circuit for Example 8.15

From the phasor circuit of Figure 8.19,

j2ω jω
V out ( ω ) = ------------------ V in ( ω ) = --------------- V in ( ω )
4 + j2ω jω + 2
and the system function is
V out ( ω ) jω
H ( ω ) = -------------------- = ---------------
V in ( ω ) jω + 2
Also,
– 3t 5 -
v in ( t ) = 5e u 0 ( t ) ⇔ V in ( ω ) = --------------
jω + 3
Then,
jω 5 r1 r2
V out ( ω ) = H ( ω )V in ( ω ) = --------------- ⋅ --------------- = --------------
- + --------------
-
jω + 2 jω + 3 jω + 2 jω + 3

and by partial fraction expansion, we find that r 1 = – 10 and r 2 = 15 . Thus,

15 10
V out ( ω ) = --------------- – ---------------
jω + 3 jω + 2
and

8-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The System Function and Applications to Circuit Analysis

–1 ⎧ 15 – 10 ⎫ – 3t – 2t
v L ( t ) = v out ( t ) = F ⎨ --------------- – --------------- ⎬ = 15e – 10e
⎩ jω + 3 jω + 2 ⎭
or
– 3t – 2t
v L ( t ) = 5 ( 3e – 2e )u 0 ( t ) (8.109)

Example 8.16
For the linear circuit of Figure 8.20, the input-output relationship is
d-
---- v ( t ) + 4v out ( t ) = 10v in ( t ) (8.110)
dt out
where v in ( t ) is as shown in Figure 8.20. Use the Fourier transform method, and the system function
H ( ω ) to compute the output v out ( t ) .

3
+ – 2t
v in ( t )
Linear +
v out ( t ) v in ( t ) = 3e
− Circuit −
t
0

Figure 8.20. Network for Example 8.16

Solution:
Taking the Fourier transform of both sides of (8.110), and recalling that
n
d- n
------
n
f ( t ) ⇔ ( jω ) F ( ω )
dt
we get,
jωV out ( ω ) + 4V out ( ω ) = 10V in ( ω )
or
( jω + 4 )V out ( ω ) = 10V in ( ω )
and thus,
V out ( ω ) 10
H ( ω ) = ------------------
- = --------------- (8.111)
V in ( ω ) jω + 4
Also,
– 2t 3
V in ( ω ) = F { v in ( t ) } = F ( 3e u 0 ( t ) ) = --------------- (8.112)
jω + 2
and

Signals and Systems with MATLAB Applications, Second Edition 8-39

Orchard Publications
Chapter 8 The Fourier Transform

10 3 r1 r2
V out ( ω ) = H ( ω ) ⋅ V in ( ω ) = --------------- ⋅ --------------- = --------------
- + --------------
-
jω + 4 jω + 2 jω + 4 jω + 2

By partial fraction expansion, we find that r 1 = – 15 and r 2 = 15 . Then,

15 – 15
V out ( ω ) = --------------- + --------------- (8.113)
jω + 2 jω + 4
Therefore,

–1 ⎧ 15 – 15 ⎫ – 2t – 4t
v out ( t ) = F ⎨ --------------- + --------------- ⎬ = 15 ( e – e )u 0 ( t ) (8.114)
⎩ jω + 2 jω + 4 ⎭

Example 8.17
– 2t
The voltage across an 1 Ω resistor is known to be v R ( t ) = 3e u 0 ( t ) . Compute the energy dissi-
pated in this resistor for 0 < t < ∞ , and verify the result by application of Parseval’s theorem.
Solution:
The instantaneous power absorbed by the resistor is
2 2 – 4t
p R = v R ⁄ 1 = v R = 9e u0 ( t ) (8.115)
and thus, the energy is

∞ ∞ – 4t ∞ 0
– 4t e 9 –4t
∫0 ∫0
2
WR = v R dt = 9e dt = 9 -------- = --- e ∞
= 2.25 joules (8.116)
–4 0
4

Parseval’s theorem states that

∞ ∞
1
∫– ∞ ∫– ∞ F ( ω )
2 2
f ( t ) dt = ------ dω (8.117)
2π

Since

F(ω) = F { 3e–2t u0 ( t ) } 3
= --------------- (8.118)
jω + 2
and

9
= F ( ω ) ⋅ F∗ ( ω ) = ------------------
2
F(ω) (8.119)
2 2
ω +2

by substitution into the right integral of (8.117) we get

8-40 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The System Function and Applications to Circuit Analysis
∞
1 9
W R = ------
2π ∫– ∞ ------------------ dω
ω +2
2 2
(8.120)

We observe that the integrand of (8.120) is an even function of ω ; therefore, we can multiply the
integral by 2 , and integrate from 0 to ∞ . Then,
∞ ∞
2 9 9 1
W R = ------
2π ∫0 ------------------ dω = ---
ω +2
2 2 π ∫0 ------------------ dω
ω +2
2 2
(8.121)

From tables of integrals,

1 1
∫ a----------------
2
+x
2
dx = --- atan --x- + C
a a
Thus,
∞
9 π
W R = --- ⎛ --- atan ω
----⎞⎠
9 1
= ------ ⋅ --- = 2.25 joules (8.122)
π ⎝2 2 2π 2
0

We observe that (8.122) is in agreement with (8.116).

Signals and Systems with MATLAB Applications, Second Edition 8-41

Orchard Publications
Chapter 8 The Fourier Transform

8.9 Summary
• The Fourier transform is defined as
∞
– jωt
F(ω) = ∫–∞ f ( t )e dt

• The Inverse Fourier transform is defined as

∞
1
∫–∞ F ( ω )e
jωt
f ( t ) = ------ dω
2π

• The Fourier transform is, in general, a complex function. We can express it as the sum of its real
and imaginary components, or in exponential form as
jϕ ( ω )
F ( ω ) = Re { F ( ω ) } + jIm { F ( ω ) } = F ( ω ) e
• We often use the following notations to express the Fourier transform and its inverse.

F {f(t)} = F(ω)
–1
F {F(ω)} = f(t)

• If f ( t ) is real, F ( ω ) is, in general, complex.

• If f ( t ) is real and even, F ( ω ) is also real and even.

• If f ( t ) is real and odd, F ( ω ) is imaginary and odd.

• If f ( t ) is imaginary, F ( ω ) is, in general, complex.

• If f ( t ) is imaginary and even, F ( ω ) is also imaginary and even.

• If f ( t ) is imaginary and odd, F ( ω ) is real and odd.

• If F ( – ω ) = F∗ ( ω ) , f ( t ) is real.
• The linearity property states that

a1 f1 ( t ) + a2 f2 ( t ) + … + an fn ( t ) ⇔ a1 F1 ( ω ) + a2 F2 ( ω ) + … + an Fn ( ω )

• The symmetry property states that

F ( t ) ⇔ 2πf ( – ω )
• The scaling property states that
ω
f ( at ) ⇔ ----- F ⎛ ---- ⎞
1
a ⎝ a ⎠

8-42 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• The time shifting property states that

– jωt 0
f ( t – t 0 ) ⇔ F ( ω )e

• The frequency shifting property states that

jω 0 t
e f ( t ) ⇔ F ( ω – ω0 )

• The Fourier transforms of the modulated signals f ( t ) cos ωt and f ( t ) sin ωt are

F ( ω – ω0 ) + F ( ω + ω0 )
f ( t ) cos ω 0 t ⇔ ----------------------------------------------------------
-
2

F ( ω – ω0 ) –F ( ω + ω0 )
f ( t ) sin ω 0 t ⇔ -------------------------------------------------------
-
j2
• The time differentiation property states that
n
d
-------- f ( t ) ⇔ ( jω ) n F ( ω )
n
dt
• The frequency differentiation property states that
n
n d
( – j t ) f ( t ) ⇔ ---------n F ( ω )
dω
• The time integration property states that
t
F(ω )
∫–∞ f ( τ ) dτ ⇔ ------------
jω
+ πF ( 0 )δ ( ω )

• If F ( ω ) is the Fourier transform of the complex function f ( t ) , then,

f∗ ( t ) ⇔ F∗ ( – ω )
• The time convolution property states that

f 1 ( t )∗ f 2 ( t ) ⇔ F 1 ( ω )F 2 ( ω )

• The frequency convolution property states that

1
f 1 ( t )f 2 ( t ) ⇔ ------ F 1 ( ω )∗ F 2 ( ω )
2π

• The area under a time function f ( t ) is equal to the value of its Fourier transform evaluated at
ω = 0 . In other words,

Signals and Systems with MATLAB Applications, Second Edition 8-43

Orchard Publications
Chapter 8 The Fourier Transform

∞
F(0) = ∫–∞ f ( t ) dt
• The value of a time function f ( t ) , evaluated at t = 0 , is equal to the area under its Fourier trans-

form F ( ω ) times 1 ⁄ 2π . In other words,

∞
1
f ( 0 ) = ------
2π ∫– ∞ F ( ω ) d ω
• Parseval’s theorem states that
∞ ∞
1
∫– ∞ ∫– ∞ F ( ω )
2 2
f ( t ) dt = ------ dω
2π

• The delta function and its Fourier transform are as shown below.

f(t) F(ω) 1
δ(t)

0 t 0 ω

• The unity time function and its Fourier transform are as shown below.

f(t) F(ω)
1
2πδ ( ω )

0 t 0 ω

jω 0 t
• The Fourier transform of the complex time function e is as indicated below.
jω 0 t
e ⇔ 2πδ ( ω – ω 0 )

• The Fourier transforms of the time functions cos ω 0 t , and sin ω 0 t are as shown below.

cosωω
cos 0 t0
t F Re ( ω )

π π
t t

−ω0 0 ω0 ω

8-44 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

sin ω 0 t F Im ( ω )
π
tt −ω0
0 ω0 ω
−π

• The signum function and its Fourier transform are as shown below.

f( t)
F Im ( ω )
1
0
t ω
0
−1

• The unit step function and its Fourier transform are as shown below.
.
f(t)
F Im ( ω )
1 π
F Re ( ω )
t ω

F Im ( ω )

– jω 0 t
• The Fourier transforms of e u 0 ( t ) , u 0 ( t ) cos ω 0 t , and u 0 ( t ) sin ω 0 t are as shown below.

– jω 0 t 1
e u 0 ( t ) ⇔ πδ ( ω – ω 0 ) + ----------------------
j ( ω – ω0 )

π jω
u 0 ( t ) cos ω 0 t ⇔ --- [ δ ( ω – ω 0 ) + δ ( ω + ω 0 ) ] + -----------------
-2
2 ω –ω
2
0

π ω -
2
u 0 ( t ) sin ω 0 t ⇔ ----- [ δ ( ω – ω 0 ) + δ ( ω + ω 0 ) ] + -----------------
j2 ω –ω
2 2
0

• If a time function f ( t ) is zero for t ≤ 0 , we can obtain the Fourier transform of f ( t ) from the one-
sided Laplace transform of f ( t ) by substitution of s with jω .

Signals and Systems with MATLAB Applications, Second Edition 8-45

Orchard Publications
Chapter 8 The Fourier Transform

• If a time function f ( t ) = 0 for t ≥ 0 , and f ( t ) ≠ 0 for t < 0 , we use the substitution

F {f(t)} = L [ f ( –t ) ] s = –j ω

to obtain the Fourier transform of f ( t ) from the one-sided Laplace transform of f ( t ) .

• The pulse function f ( t ) = A [ u 0 ( t + T ) – u 0 ( t – T ) ] and its Fourier transform are as shown below.

F(ω )
f(t)
A

−π π
t −2π 2π ωT
−T 0 T 0

• The Fourier transform of a periodic time function with period T is as shown below.
∞ ∞ ∞
F {f(t)} = F ⎧⎨ ∑ Cn e
jnω 0 t ⎫
⎬ = ∑ Cn F { e
jnω 0 t
} = 2π ∑ C n δ ( ω – nω )
0
⎩ ⎭ n = –∞
n = –∞ n = –∞

• The Fourier transform of a periodic train of equidistant delta functions in the time domain, is a
periodic train of equally spaced delta functions in the frequency domain.
• The system function H ( ω ) and the impulse response h ( t ) form the Fourier transform pair

h(t) ⇔ H(ω)
and
f ( t )∗ h ( t ) = g ( t ) ⇔ G ( ω ) = F ( ω ) ⋅ H ( ω )

8-46 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

8.10 Exercises
1. Show that
∞

∫–∞ u ( t )δ ( t ) dt
0 = 1⁄2

2. Compute
– at
F { te u0 ( t ) } a > 0

3. Sketch the time and frequency waveforms of

f ( t ) = cos ω 0 t [ u 0 ( t + T ) – u 0 ( t – T ) ]

4. Derive the Fourier transform of

f ( t ) = A [ u 0 ( t + 3T ) – u 0 ( t + T ) + u 0 ( t – T ) – u 0 ( t – 3T ) ]

5. Derive the Fourier transform of

A
f ( t ) = --- t [ u 0 ( t + T ) – u 0 ( t – T ) ]
T
6. Derive the Fourier transform of

f ( t ) = ⎛ --- t + A⎞ [ u 0 ( t + T ) – u 0 ( t ) ] + ⎛ – --- t + A⎞ u 0 ( t ) – u 0 ⎛ t – --- ⎞

A A T
⎝T ⎠ ⎝ T ⎠ ⎝ 2⎠

7. For the circuit of Figure 8.21, use the Fourier transform method to compute v C ( t ) .

R1

1Ω
C +
+ vC ( t ) R2
− −
v in ( t ) 1F
0.5 Ω

v in ( t ) = 50 cos 4tu 0 ( t )

Figure 8.21. Circuit for Exercise 7

8. The input-output relationship in a certain network is

d2 d
------- v out ( t ) + 5 ----- v out ( t ) + 6v out ( t ) = 10v in ( t )
2 dt
dt
–t
Use the Fourier transform method to compute v out ( t ) given that v in ( t ) = 2e u 0 ( t ) .

Signals and Systems with MATLAB Applications, Second Edition 8-47

Orchard Publications
Chapter 8 The Fourier Transform

9. In a bandpass filter, the lower and upper cutoff frequencies are f 1 = 2 Hz , and f 2 = 6 Hz respec-
tively. Compute the 1 Ω energy of the input, and the percentage that appears at the output, if the
– 2t
input signal is v in ( t ) = 3e u 0 ( t ) volts.

10. In Example 8.4, we derived the Fourier transform pair

sin ωT
A [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ 2AT ---------------
ωT

F(ω)
f( t)
A

−π π
t −2π 2π ωT
−T 0 T 0

Figure 8.22. Figure for Exercise 10

Compute the percentage of the 1 Ω energy of f ( t ) contained in the interval – π ⁄ T ≤ ω ≤ π ⁄ T of

F(ω) .

8-48 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

8.11 Solutions to Exercises

1.
∞ ∞ ∞
d
∫– ∞ u 0 ( t )δ ( t ) dt = ∫ –∞
u 0 ( t ) u 0 ( t ) dt =
d t ∫–∞ u ( t ) d( u ( t ) )
0 0

and since at t = +∞ , u 0 ( t ) = 1 whereas at t = – ∞ , u 0 ( t ) = 0 , we replace the limits of integra-

tion with 1 and 0 . Then,
2 1
1
u0 ( t )
∫ 0
u 0 ( t ) d( u 0 ( t ) ) = -----------
2
-
0
= 1⁄2

2.
∞ ∞ ∞
– jωt – at – jωt – ( jω + a )t
F(ω) = ∫– ∞ f ( t )e dt = ∫0 te e dt = ∫0 te dt

From tables of integrals

ax
e
∫
ax
xe dx = ------2- ( ax – 1 )
a
Then,
– ( jω + a )t ∞ 0
⋅ [ – ( jω + a )t – 1 -]
F ( ω ) = e------------------------------------------------------------------ [ ( jω + a )t + 1 ] -
= ---------------------------------------------
2 ( jω + a )t 2
( jω + a ) 0
e ⋅ ( jω + a ) ∞

With the upper limit of integration we get

F(ω) 1
= ----------------------
t=0 2
( jω + a )
To evaluate the lower limit of integration, we apply L’Hôpital’s rule, i.e.,
d
----- [ ( jω + a )t + 1 ]
[ ( jω + a )t + 1 ] dt ( jω + a )
---------------------------------------------- = lim --------------------------------------------------------- = lim ------------------------------------------------------------------
- = 0
( jω + a )t t→∞d t → ∞ ( jω + a )e ( jω + a )t ⋅ ( jω + a ) 2
----- [ e ( jω + a )t ⋅ ( jω + a ) 2 ]
2
e ⋅ ( jω + a ) ∞
dt
and thus
1
F ( ω ) = ----------------------
2
( jω + a )
Check:
F(ω) = F(s) s = jω
and since

Signals and Systems with MATLAB Applications, Second Edition 8-49

Orchard Publications
Chapter 8 The Fourier Transform

– at 1
te u 0 ( t ) ⇔ ------------------2
(s + a)
it follows that
1 1
F ( ω ) = ------------------ = ----------------------
2 2
(s + a) s = jω
( jω + a )

3.
From Example 8.7,
sin [ ( ω – ω 0 )T ] sin [ ( ω + ω 0 )T ]
A cos ω 0 t [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ AT ------------------------------------
- + --------------------------------------
( ω – ω 0 )T ( ω + ω 0 )T

and using the MATLAB statements

fplot('cos(x)',[−2*pi 2*pi −1.2 1.2])
fplot('sin(x)./x',[−20 20 −0.4 1.2])
we obtain the plots below.

F( ω)
AT

A f( t)

t ω
–T 0 T –ω0 0 ω0
2π
------
T

4.
f ( t ) = A [ u 0 ( t + 3T ) – u 0 ( t + T ) + u 0 ( t – T ) – u 0 ( t – 3T ) ]

f( t)

t
– 3T – 2T –T 0 T 2T 3T

From Example 8.4.

sin ωT
A [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ 2AT ---------------
ωT
and from the time shifting property

8-50 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
– jωt 0
f ( t – t 0 ) ⇔ F ( ω )e

sin ωT j2ωt – j 2ωt

f ( t ) = A [ u 0 ( t + 3T ) – u 0 ( t + T ) + u 0 ( t – T ) – u 0 ( t – 3T ) ] ⇔ F ( ω ) = 2AT --------------- ⋅ ( e +e )
ωT
or
j2ωt – j 2ωt
sin ωT e +e sin ωT
F ( ω ) = 4AT --------------- ⋅ ⎛ -------------------------------⎞ = 4AT cos 2ωT ---------------
ωT ⎝ 2 ⎠ ωT
5.
A
A f ( t ) = --- t
f ( t ) = --- t [ u 0 ( t + T ) – u 0 ( t – T ) ] T
T A

–T t
0 T

–A

∞ T T
– jωt A – jωt – jωt
∫– ∞ ∫– T ∫– T t e
A
F(ω) = f ( t )e dt = --- t e
T
dt = --T- dt

From tables of integrals

ax
e
∫
ax
xe dx = ------2- ( ax – 1 )
a
Then,
– jωt T – jωt T – jωt T
A e A e A e
F(ω) = --- ⋅ ------------2- ⋅ ( – jωt – 1 ) = --- ⋅ ----------2- ⋅ ( – jωt – 1 ) = --- ⋅ ----------2- ⋅ [ – ( jωt + 1 ) ]
T T T
( jω ) –T
–ω –T
–ω –T
A - – jωT jωT
= ----------
2
[e ⋅ ( jωT + 1 ) – e ⋅ ( – j ωT + 1 ) ]
ω T
A- – jωT – jωT jωT jωT
= ----------
2
( jωT ⋅ e +e + jωT ⋅ e –e )
ω T
A- jωT – jωT jωT – jωT
= ----------
2
[ jωT ⋅ ( e +e ) – (e –e )]
ω T

and multiplying both the numerator and denominator by j2 we get

jωT – jωT jωT – jωT
j2 A- jωT ( e )- – (--------------------------------
+e –e -) = ----------
e j2 A-
F ( ω ) = ----------
2
------------------------------------------- 2
( ωT cos ωT – sin ω T )
ω T j2 j2 ω T

We observe that since f ( t ) is real and odd, F ( ω ) is imaginary and odd.

Signals and Systems with MATLAB Applications, Second Edition 8-51

Orchard Publications
Chapter 8 The Fourier Transform

Alternate Solution:
A
f 1 ( t ) = --- t f2 ( t )
T A⁄T
A

–T t
0 T
t
−T 0 T
–A

The waveform of f 2 ( t ) is the derivative of the waveform of f 1 ( t ) and thus f 1 ( t ) = tf 2 ( t ) . From

Example 8.4,
A sin ωT sin ωT
f 2 ( t ) ⇔ 2 --- T --------------- = 2A ---------------
T ωT ωT
From the frequency differentiation property
n
n d
( – j t ) f ( t ) ⇔ ---------n F ( ω )
dω
or
n
n n d
t f ( t ) ⇔ j ---------n F ( ω )
dω
Then,
sin ωT d sin ωT
{ f 1 ( t ) = tf 2 ( t ) } = ⎛ j ------- ⎛ 2A ---------------⎞ = j2A ------- ⎛ ---------------⎞ ⎞
d
F ⎝ dω ⎝ ωT ⎠ dω ⎝ ωT ⎠ ⎠
( ωT )T cos ωT – T ( sin ω T ) j2A
= j2A --------------------------------------------------------------- - ( ωT cos ωT – sin ωT )
= ---------
2 2
( ωT ) ω T
6.
We denote the given waveform as f 1 ( t ) , that is,

f 1 ( t ) = ⎛ --- t + A⎞ [ u 0 ( t + T ) – u 0 ( t ) ] + ⎛ – A
--- t + A⎞ u 0 ( t ) – u 0 ⎛ t – T
--- ⎞
A
⎝T ⎠ ⎝ T ⎠ ⎝ 2⎠

A
f 1 ( t ) = --- t f2 ( t ) f3( t )
A
T A⁄T
A⁄T
a
t T
t
–T 0 T −T 0 –T ⁄2 T ⁄2
b t
–A ⁄ T

8-52 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

We observe that f 1 ( t ) is the integral of f 2 ( t ) . Therefore, we will find F 2 ( ω ) and by integration we

will find F 1 ( ω ) .

We begin by finding the Fourier Transform of the pulse denoted as f 3 ( t ) , and using F 3 ( ω ) and
the time shifting and linearity properties, we will find F 2 ( ω ) .

– jωt T⁄2 – jωt –T ⁄ 2

T⁄2
--- ⎛ e----------- ⎞
--- e –jωt dt = A
A
= --- ⎛ ----------- ⎞
A e
F3 ( ω ) = ∫
–T ⁄ 2 T T ⎝ – jω ⎠ –T ⁄ 2
T ⎝ jω ⎠ T⁄2
jω ( T ⁄ 2 ) – jω ( T ⁄ 2 )
–e 2A ωT sin ( ωT ⁄ 2 )
= --- ⎛ ----------------------------------------------- ⎞ = ------- sin -------
A e
= A ---------------------------
T⎝ jω ⎠ ωT 2 ωT ⁄ 2

Using the time shifting property

– jωt 0
f ( t – t 0 ) ⇔ F ( ω )e

the Fourier transform of the left pulse a of f 2 ( t ) is

sin ( ωT ⁄ 2 ) jω ( T ⁄ 2 )
F 2a ( ω ) = A --------------------------- ⋅ e
ωT ⁄ 2

Likewise, the Fourier transform of the right pulse b of f 2 ( t ) is

sin ( ωT ⁄ 2 ) –j ω ( T ⁄ 2 )
F 2b ( ω ) = – A --------------------------- ⋅ e
ωT ⁄ 2
and using the linearity property we get
sin ( ωT ⁄ 2 ) jω ( T ⁄ 2 ) –j ω ( T ⁄ 2 )
F 2 ( ω ) = F 2a ( ω ) + F 2b ( ω ) = A --------------------------- ⋅ ( e –e )
ωT ⁄ 2
jω ( T ⁄ 2 ) –j ω ( T ⁄ 2 ) 2
sin ( ωT ⁄ 2 ) e –e sin ( ωT ⁄ 2 )
= j2A --------------------------- ⋅ ⎛ ------------------------------------------------ ⎞ = j2A -----------------------------
ωT ⁄ 2 ⎝ j2 ⎠ ωT ⁄ 2
2
This sin ( x ) ⁄ x curve is shown below and it was created with the following MATLAB code.
fplot('sin(x./2).^2./x',[0 16*pi 0 0.5])

ω
0

Signals and Systems with MATLAB Applications, Second Edition 8-53

Orchard Publications
Chapter 8 The Fourier Transform

Now, we find F 1 ( ω ) of the triangular waveform of f 1 ( t ) with the use of the integration property
by multiplying F 2 ( ω ) by 1 ⁄ jω . Thus,
2 2
1 sin ( ωT ⁄ 2 ) 2A sin ( ωT ⁄ 2 ) 2A
- ⋅ sin 2( ωT ⁄ 2 )
F 1 ( ω ) = ( 1 ⁄ jω ) ⋅ F 2 ( ω ) = ------ ⋅ j2A ----------------------------- = ------- ⋅ ----------------------------- = ----------------
jω ωT ⁄ 2 ω ωT ⁄ 2 2
ω T⁄2
2
We can plot F 1 ( ω ) by letting T ⁄ 2 = 1 . Then, F 1 ( ω ) simplifies to the form K [ ( sin x ) ⁄ x ]

This curve is shown below and it was created with the following MATLAB code.
fplot('(sin(x)./x).^2',[-8*pi 8*pi 0 1])

ω
0

7.

1Ω
+
+ vC ( t )
− −
v in ( t ) 1F
0.5 Ω

By KCL
v C ( t ) – v in ( t ) dv C v C ( t )
- + 1 ⋅ --------- + -----------
------------------------------- - = 0
1 dt 0.5

dv C
--------- + 3v C = v in
dt
Taking the Fourier transform of both sides we get
jωV C ( ω ) + 3V C ( ω ) = V in ( ω )

( jω + 3 )V C ( ω ) = V in ( ω )
and since V out ( ω ) = V C ( ω )

8-54 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

( jω + 3 )V out ( ω ) = V in ( ω )
and
V out ( ω ) 1
H ( ω ) = ------------------
- = ---------------
V in ( ω ) jω + 3
where
v in ( t ) = 50 cos 4tu 0 ( t ) ⇔ V in ( ω ) = 50π [ δ ( ω – 4 ) + δ ( ω + 4 ) ]
Then,
1
V out ( ω ) = V C ( ω ) = H ( ω ) ⋅ V in ( ω ) = --------------- ⋅ 50π [ δ ( ω – 4 ) + δ ( ω + 4 ) ]
jω + 3
and
∞
–1 1 50π [ δ ( ω – 4 ) + δ ( ω + 4 ) ]
∫–∞ ---------------------------------------------------------------
jωt
vC ( t ) = F { V C ( ω ) } = ------ -⋅e dω
2π jω + 3
∞ ∞
δ ( ω – 4 ) jωt δ(ω + 4)
∫ ∫–∞ --------------------
jωt
= 25 --------------------- ⋅ e dω + 25 -⋅e dω
– ∞ jω + 3 jω + 3

Next, using the sifting property of δ ( ω ) we simplify the above to

j4t – j 4t ⎛ e j4t – j 4t ⎞
v C ( t ) = 25 ⎛ -------------- + -------------- ⎞ = 25 ⎜ ------------------ + -------------------- ⎟ = 5 ( e ⋅ e
e e e j4t – j 53.1° – j 4t j53.1°
+e ⋅e )
⎝ j4 + 3 j4 + 3 ⎠ ⎝ 5e j53.1° – j 53.1°
⎠
5e
j ( 4t – 53.1° ) – j ( 4t – 53.1° )
e +e
= 10 ------------------------------------------------------------- = 10 cos ( 4t – 53.1° )
2
8.
d2 d
------- v out ( t ) + 5 ----- v out ( t ) + 6v out ( t ) = 10v in ( t ) = 2e – t u 0 ( t )
2 dt
dt
Taking the Fourier transform of both sides we get
2 2
[ ( jω ) + 5jω + 6 ]V out ( ω ) = 10V in ( ω ) = 10 ⋅ ---------------
jω + 1

20 -
[ ( jω + 2 ) ⋅ ( jω + 3 ) ]V out ( ω ) = 10V in ( ω ) = --------------
jω + 1

20
V out ( ω ) = ----------------------------------------------------------------------
( jω + 1 ) ⋅ ( jω + 2 ) ⋅ ( jω + 3 )

We use the following MATLAB code for partial fraction expansion where we let jω = s .
syms s; collect((s+1)*(s+2)*(s+3))

Signals and Systems with MATLAB Applications, Second Edition 8-55

Orchard Publications
Chapter 8 The Fourier Transform

ans =
s^3+6*s^2+11*s+6
num=[0 0 0 20]; den=[1 6 11 6]; [num,den]=residue(num,den); fprintf(' \n');...
fprintf('r1 = %4.2f \t', num(1)); fprintf('p1 = %4.2f', den(1)); fprintf(' \n');...
fprintf('r2 = %4.2f \t', num(2)); fprintf('p2 = %4.2f', den(2)); fprintf(' \n');...
fprintf('r3 = %4.2f \t', num(3)); fprintf('p3 = %4.2f', den(3))
r1 = 10.00 p1 = -3.00
r2 = -20.00 p2 = -2.00
r3 = 10.00 p3 = -1.00
Then,
20
V out ( ω ) = ---------------------------------------------------------------------
- + ------------------- – 20 - + -------------------
10 - + ------------------- 10 -
( jω + 1 ) ⋅ ( jω + 2 ) ⋅ ( jω + 3 ) ( jω + 1 ) ( jω + 2 ) ( jω + 3 )
and thus
–1 –t – 2t – 3t
v out ( t ) = F { V out ( ω ) } = 10e – 20e + 10e
–t – 2t – 3t
= 10 ( e – 2e +e )u 0 ( t )

9.
The input energy in joules is
∞ ∞ ∞ ∞
– 2t 2 – 2t 2 – 4t
∫– ∞ ∫0 ∫0 ∫0 9e
2
W in = v in ( t ) dt = 3e dt = 3e dt = dt

– 4t ∞ – 4t 0
9e 9e 9
= ------------ = ------------ = --- = 2.25 J
–4 0
4 ∞
4

and the Fourier transform F in ( ω ) of the input v in ( t ) is

– 2t 3
F { v in ( t ) } = F { 3e u 0 ( t ) } = ---------------
jω + 2

The energy at the output for the frequency interval 2 Hz ≤ f ≤ 6 Hz or 4π rad ≤ ω ≤ 12π rad is
∞ ∞ 12π
1 1 1-
3 - 2 dω = ----- 9 -
∫– ∞ ∫– ∞ ∫4π
2
W out = ------ F ( ω ) dω = ------ -------------- ----------------- dω
2π 2π jω + 2 2π ω +2
2 2

and from tables of integrals

1 1
∫ ----------------
2
x +a
2
dx = --- atan --x-
a a
Then,

8-56 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
12π
W out = ------ ⋅ --- atan ω = ------ ⎛ atan 12π ------⎞⎠ = ------ ( atan 6π – atan 2π )
9 1 9 9
---- --------- – atan 4π
2π 2 2 4π
4π ⎝ 2 2 4π

fprintf(' \n'); fprintf('atan(6*pi) = %4.2f \t', atan(6*pi)); fprintf('atan(2*pi) = %4.2f', atan(2*pi))

atan(6*pi) = 1.52 atan(2*pi) = 1.41
and thus
9
W out = ------ ( 1.52 – 1.41 ) = 0.08 J
4π
Therefore, the percentage of the input appearing at the output is
W out
----------- × 100 = 0.08
---------- = 3.56%
W in 2.25

10.
sin ωT
A [ u 0 ( t + T ) – u 0 ( t – T ) ] ⇔ 2AT ---------------
ωT

F(ω)
f(t)
A

−π π
t −2π 2π ωT
−T 0 T 0

First, we compute the total energy of the pulse in terms of f ( t ) .

∞ T

∫– ∞ ∫ – T A dt
2 2
W total = f ( t ) dt =

and since f ( t ) is an even function

T
2 T
∫0
2 2
W total = 2 A dt = 2A t 0
= 2A T

Next, we denote the energy in the frequency interval – π ⁄ T rad ≤ ω ≤ π ⁄ T rad as W out in the fre-

quency domain we get

∞ π⁄T
1 1 sin ωT 2
∫– ∞ ∫– π ⁄ T
2
W out = ------ F ( ω ) dω = ------ 2AT --------------- dω
2π 2π ωT

Signals and Systems with MATLAB Applications, Second Edition 8-57

Orchard Publications
Chapter 8 The Fourier Transform

and since F ( ω ) is an even function

π⁄T 2 2 2 π⁄T 2
1 sin ωT 4A T sin ωT
∫0 ∫0
2 2
W out = 2 ------ 4A T ----------------- dω = ---------------- ----------------- dω (1)
2π ( ωT )
2 π ( ωT )
2

For simplicity, we let ωT = y . Then, ω = y ⁄ T and dω = ( 1 ⁄ T )dy . Also, when ω = 0 , y = 0 ,

and when ω = π ⁄ T or ωT = π , y = π . With these substitutions we express (1) as
2 2 π 2 2 2 π 2 2 π 2
4A T sin y 4A T sin y 4A T sin y
W out = ----------------
π ∫0 -------------------------
( ( y ⁄ T )T )
2
dy = ----------------
πT ∫0 ------------
y
2
dy = -------------
π ∫0 ------------
y
2
dy (2)

But the last integral in (2) is an improper integral and does not appear in tables of integrals*.
We will attempt to simplify (2) using integration by parts. We start with the familiar
d ( uv ) = udv + vdu
from which

∫ d ( uv ) = ∫ u dv + ∫ vdu
or

∫ u dv = uv – vdu ∫
2 2
Letting u = sin y and dv = 1 ⁄ y it follows that du = 2 cos y sin y = sin 2y and v = – 1 ⁄ y .
With these substitutions (2) is written as
π
2 2 π 2 π
4A T sin y –1 4A T sin 2y
W out = ------------- ------------
π –y
– ∫0 ------ sin 2y dy = ------------- 0 +
y π ∫0 ------------- dy
y
0
(3)
2 π 2 π
4A T sin 2y 8A T sin 2y
= 2 ⋅ -------------
π ∫0 -------------
2y
dy = -------------
π ∫0 -------------
2y
dy

The last integral in (3) is also an improper integral. Fortunately, some handbooks of mathematical
tables include numerical values of the integral
π
sin x
∫0 ---------
x
- dx

* It is shown in Advanced Calculus textbooks that if the upper limit is ∞ , then

∞ ∞ 2
sin x- sin x π
∫0 ---------
x
dx = ∫0 ------------
x
2
dx = ---
2
but for other finite limits are not equal.

8-58 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

for arguments of πx in the interval 0 ≤ x ≤ 10 . Then, replacing 2y with w we get w = 2y ,

y = w ⁄ 2 , dy = ( 1 ⁄ 2 ) ⁄ dw , and for y = 0 , w = 0 whereas for y = π , w = 2π . Then, by sub-
stitution into (3) we get
2 2π 2 2π
8A T sin w ⎛ 1 --- dw⎞ = 4A T- sin w
W out = -------------
π ∫0 -----------
w ⎝2 ⎠
------------
π ∫0 -----------
w
dw (4)

From Table 5.3 of Handbook of Mathematical Functions, 1972 Edition, Dover Publications, with
πx = 2π or x = 2 we get
2π
sin w
∫0 ----------- dw
w
= 1.418
x=2
and thus (4) reduces to
2
4A T
W out = ------------- 1.418
π

Therefore, the percentage of the output for the frequency interval – π ⁄ T rad ≤ ω ≤ π ⁄ T rad is
2
W out ( 4A T ⁄ π ) ⋅ 1.418 2 × 1.418
- × 100% = ------------------------------------------- × 100% = ---------------------- × 100% = 90.3%
-------------
W total 2A T
2 π

Since this computation involves numerical integration, we can obtain the same result much faster
and easier with MATLAB as follows:
First, we define the function fourierxfm1 and we save it as an .m file as shown below. This file
must be created with MATLAB’s editor (or any other editor) and saved as an .m file in a drive that
has been added to MATLAB’s path.
function y1=fourierxfm1(x)
x=x+(x==0)*eps;% This statement avoids the sin(0)/0 value.
% It says that if x=0, then (x==0) = 1
% but if x is not zero, then (x==0) = 0
% and eps is approximately equal to 2.2e-16
% It is used to avoid division by zero.
y1=sin(x)./x;
Then, using MATLAB’s command window, we write and execute the program below.
% The quad function below performs numerical integration from 0 to 2*pi
% using a form of Simpson's rule of numerical integration.
value1=quad('fourierxfm1',0,2*pi)
value1 =
1.4182
We could also have used numerical integration with the integral

Signals and Systems with MATLAB Applications, Second Edition 8-59

Orchard Publications
Chapter 8 The Fourier Transform

π 2
sin x
∫0 ------------ dx
x
2

thereby avoiding the integration by parts procedure. Shown below are the function fourierxfm2
which is saved as an .m file and the program execution using this function.
function y2=fourierxfm2(x)
x=x+(x==0)*eps;
y2=(sin(x)./x).^2;
and after this file is saved, we execute the statement below observing that the limits of integration
are from 0 to π .
value2=quad('fourierxfm2',0,pi)
value2 =
1.4182

8-60 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Chapter 9
Discrete Time Systems and the Z Transform

his chapter is devoted to discrete time systems, and introduces the one-sided Z Transform.
T The definition, theorems, and properties are discussed, and the Z transforms of the most
common discrete time functions are derived. The discrete transfer function is also defined,
and several examples are given to illustrate its application. The Inverse Z transform, and the meth-
ods available for finding it, are also discussed.

9.1 Definition and Special Forms

The Z transform performs the transformation from the domain of discrete time signals, to another
domain which we call z – domain . It is used with discrete time signals, the same way the Laplace and
Fourier transforms are used with continuous time signals. The Z transform yields a frequency
domain description for discrete time signals, and forms the basis for the design of digital systems,
such as digital filters. Like the Laplace transform, there is the one-sided, and the two-sided Z trans-
form. We will restrict our discussion to the one-sided Z transform F ( z ) of a discrete time function
f [ n ] defined as

∞ –n
F(z) = ∑ f [ n ]z (9.1)
n=0

and the Inverse Z transform is defined as

1 k–1
j2π °∫
f [ n ] = -------- F ( z )z dz (9.2)

We can obtain a discrete time waveform from an analog (continuous or with a finite number of dis-
continuities) signal, by multiplying it by a train of impulses. We denote the continuous signal as f ( t ) ,
and the impulses as
∞
δ[n] = ∑
n=0
δ [ t – nT ] (9.3)

Multiplication of f ( t ) by δ [ n ] produces the signal g ( t ) defined as

∞
g( t) = f(t) ⋅ δ[n] = f(t) ∑ δ [ t – nT ]
n=0
(9.4)

These signals are shown in Figure 9.1.

Signals and Systems with MATLAB Applications, Second Edition 9-1

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

f(t)

t
0 (a)
δ[n]

n
0 (b)

f ( t )δ [ n ]

t
0 (c)

Figure 9.1. Formation of discrete time signals

Of course, after multiplication by δ [ n ] , the only values of f ( t ) which are not zero, are those for
which t = nT , and thus we can express (9.4) as
∞ ∞
g ( t ) = f [ nT ] ∑
n=0
δ [ t – nT ] = ∑
n=0
f [ nT ]δ [ t – nT ] (9.5)

Next, we recall from Chapter 2, that the t – domain to s – domain transform pairs for the delta
– sT
function are δ ( t ) ⇔ 1 and δ ( t – T ) ⇔ e . Therefore, taking the Laplace transform of both sides of
(9.5), and, for simplicity, letting f [ nT ] = f [ n ] , we get

⎧ ∞ ⎫ ∞
– nsT
∞
– nsT
G(s) = L ⎨ f [n]
⎩ n=0
∑
δ [ t – nT ] ⎬ = f [ n ]
⎭ n=0
e ∑= ∑ f [n]e
n=0
(9.6)

9-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Z Transform
sT
Relation (9.6), with the substitution z = e , becomes the same as (9.1), and like s , z is also a com-
plex variable.
The Z and Inverse Z transforms are denoted as

F(z) = Z {f [n]} (9.7)

and

–1
f [n] = Z {F(z)} (9.8)

The function F ( z ) , as defined in (9.1), is a series of complex numbers and converges outside the cir-
cle of radius R , that is, it converges (approaches a limit) when z > R . In complex variables theory,
the radius R is known as the radius of absolute convergence.
In the complex z – plane the region of convergence is the set of z for which the magnitude of F ( z )
is finite, and the region of divergence is the set of z for which the magnitude of F ( z ) is infinite.

9.2 Properties and Theorems of the Z Transform

The properties and theorems of the Z transform are similar to those of the Laplace transform. In
this section, we will state and prove the most common Z transforms.
1. Linearity

af 1 [ n ] + bf 2 [ n ] + cf 3 [ n ] + … ⇔ aF 1 ( z ) + bF 2 ( z ) + cF 3 ( z ) + … (9.9)

where a, b, c, … are arbitrary real or complex constants.

Proof:
The proof is easily obtained by application of the definition of the Z transform to each term on
the left side.
In our subsequent discussion, we will denote the discrete unit step function as u 0 [ n ] .

2. Shift of f [ n ]u 0 [ n ] in the Discrete Time Domain

–m
f [ n – m ]u 0 [ n – m ] ⇔ z F(z) (9.10)

Signals and Systems with MATLAB Applications, Second Edition 9-3

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

Proof:
Applying the definition of the Z transform, we get
∞ –n
Z { f [ n – m ]u 0 [ n – m ] } = ∑ f [ n – m ]u0 [ n – m ]z
n=0

and since u 0 [ n – m ] = 0 for n < m and u 0 [ n – m ] = 1 for n > m , the above expression reduces to
∞ –n
Z { f [n – m]} = ∑ f [ n – m ]z
n=0

Now, we let n – m = k ; then, n = k + m , and when n – m = 0 or n = m , k = 0 . Therefore,

∞ ∞ ∞
–( k + m ) –k –m –m –k –m
Z { f [n – m]} = ∑ f [k] z = ∑ f [ k ]z z = z ∑ f [ k ]z = z F(z)
k=0 k=0 k=0

3. Right Shift in the Discrete Time Domain

This property is a generalization of the previous property, and allows use of non-zero values for
n < 0 . The transform pair is

m–1
–m –n
f [n – m] ⇔ z F(z ) + ∑ f [ n – m ]z (9.11)
n=0

Proof:
By application of the definition of the Z transform, we get
∞ –n
Z { f [n – m]} = ∑ f [ n – m ]z
n=0

We let n – m = k ; then, n = k + m , and when n = 0 , k = – m . Therefore,

∞ ∞ ∞
–( k + m ) –k –m –m –k
Z { f [n – m]} = ∑
k = –m
f [ k ]z = ∑
k = –m
f [ k ]z z = z ∑ f [ k ]z
k = –m
–1 ∞ –1
–m –k –k –m –k
= z ∑ f [ k ]z + ∑ f [ k ]z =z F(z ) + ∑ f [ k ]z
k = –m k=0 k = –m

When k = – m , n = 0 , and when k = – 1 , n = m – 1 . Then, by substitution into the last summa-

tion term above, we get

9-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Z Transform

m–1 m–1
–m m–n –m –n
Z { f [n – m]} = z F(z) + ∑ f [ n – m ]z = z F(z ) + ∑ f [ n – m ]z
n=0 n=0
and this is the same as (9.11).
For m = 1 , (9.11) reduces to

–1
f [ n – 1 ] ⇔ z F ( z ) + f [ –1 ] (9.12)

and for m = 2 , reduces to

–2 –1
f [ n – 2 ] ⇔ z F ( z ) + f [ –2 ] + z f [ –1 ] (9.13)

4. Left Shift in the Discrete Time Domain

–1
–n
∑
m
f [n + m] ⇔ z F( z) + f [ n + m ]z (9.14)
n = –m

that is, if f [ n ] is a discrete time signal, and m is a positive integer, the mth left shift of f [ n ] is
f [n + m] .

Proof:
∞
–n
Z { f [n + m]} = ∑ f [ n + m ]z
n=0

We let n + m = k ; then, n = k – m , and when n = 0 , k = m . Then,

∞ ∞
–( k – m ) –k m
Z { f [n + m]} = ∑ f [k] z = ∑ f [ k ]z z
k=m k=m
∞ m–1 m–1
–k –k
∑ f [k] z ∑ f [k] z ∑ f [k] z
m m k
= z – =z F(z ) –
k=0 k=0 k=0

When k = 0 , n = – m , and when k = m – 1 , n = – 1 . Then, by substitution into the last sum-

mation term of the above expression, we get
–1 –1
–( n + m ) –n
∑ ∑
m m
Z {f [n + m]} = z F(z) + f [n + m] z = z F( z) + f [ n + m ]z
n = –m n = –m

and this is the same as (9.14).

Signals and Systems with MATLAB Applications, Second Edition 9-5

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

For m = 1 , the above expression reduces to

Z { f [ n + 1 ] } = zF ( z ) – f [ 0 ]z (9.15)

and for m = 2 , reduces to

2 2
Z { f [ n + 2 ] } = z F ( z ) – f [ 0 ]z – f [ 1 ]z (9.16)

n
5. Multiplication by a in the Discrete Time Domain

a f [ n ] ⇔ F ⎛ --- ⎞
n z
⎝a⎠
(9.17)

Proof:
∞ ∞ ∞ z –n
f [ n ] ⎛ --- ⎞ = F ⎛ --- ⎞
–n 1 z
------- f [ n ]z – n =
∑ ∑ ∑
n n
Z {a f [n]} = a f [ n ]z = –n ⎝ a⎠ ⎝ a⎠
a k=0
k=0 k=0

– naT
6. Multiplication by e in the Discrete Time Domain

– naT aT
e f [n ] ⇔ F( e z) (9.18)

Proof:
– naT ∞ – naT –n ∞ –n
∑e ∑ f [n](e
aT aT
Z {e f [n]} = f [n] z = z) = F(e z )
k=0 k=0

7. Multiplication by n and n 2 in the Discrete Time Domain

d
nf [ n ] ⇔ – z ----- F ( z )
dz
2 (9.19)
2 d 2d
n f [ n ] ⇔ z ----- F ( z ) + z -------2- F ( z )
dz dz

Proof:
By definition,
∞ –n
F(z) = ∑ f [ n ]z
n=0

9-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Z Transform

and taking the first derivative of both sides with respect to z we get
d ∞ –n–1 –1 ∞ –n
----- F ( z ) =
dz ∑ ( – n )f [ n ]z = –z ∑ nf [ n ]z
n=0 n=0

Multiplication of both sides by – z yields

∞ –n d
∑ nf [ n ]z = – z ----- F ( z )
dz
n=0

Differentiating one more time, we get the second pair in (9.19).

8. Summation in the Discrete Time Domain
n
f [ m ] ⇔ ⎛ ----------- ⎞ F ( z )
z
∑
m=0
⎝z–1 ⎠
(9.20)

that is, the Z transform of the sum of the values of a signal, is equal to z ⁄ ( z – 1 ) times the Z
transform of the signal. This property is equivalent to time integration in the continuous time
domain since integration in the discrete time domain is summation. We will see on the next sec-
tion that the term z ⁄ ( z – 1 ) is the Z transform of the discrete unit step function u 0 [ n ] , and
recalling that in the s – domain
1
u 0 ( t ) ⇔ ---
s
and
t
F(s)
∫0 f ( τ ) dτ ⇔ ----------
s
-

then, the similarity of the Laplace and Z transforms becomes apparent.

Proof:
Let
n
y[ n] = ∑
m=0
x[m] (9.21)

and let us express (9.21) as

n–1
y[n] =
m=0
∑ x[m] + x[n] (9.22)

Since the summation symbol in (9.21) is y [ n ] , then the summation symbol in (9.22) is y [ n – 1 ] ,
and thus we can write (9.22) as

Signals and Systems with MATLAB Applications, Second Edition 9-7

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

y[ n] = y[ n – 1] + x[n] (9.23)
Next, we take the Z transform of both sides of (9.23), and using the property
–m
x [ n – m ]u 0 [ n – m ] ⇔ z X(z)
we get
–1
Y(z) = z Y(z) + X(z)
or
1 - z
Y ( z ) = --------------- X ( z ) = ----------- X ( z )
1–z
–1 z–1

and this relation is the same as (9.20).

9. Convolution in the Discrete Time Domain
Let the impulse response of a discrete time system be denoted as h [ n ] , that is, an impulse δ [ n ] ,
produces a response h [ n ] . Likewise, a delayed impulse δ [ n – m ] produces a delayed response
h [ n – m ] , and so on. Therefore, any discrete time input signal can be considered as an impulse
train, in which each impulse has a weight equal to its corresponding sampled value. Then, for any
other input x [ 0 ], x [ 1 ], x [ 2 ], …, x [ m ] , we get
x [ 0 ]δ [ 0 ] → x [ 0 ]h [ n ]
x [ 1 ]δ [ n – 1 ] → x [ 1 ]h [ n – 1 ]
x [ 2 ]δ [ n – 2 ] → x [ 2 ]h [ n – 2 ]
…
x [ m ]δ [ n – m ] → x [ m ]h [ n – m ]
and the response at any arbitrary value m, is obtained by summing all the components that have
occurred up to that point, that is, if y [ n ] is the output due to the input x [ m ] convolved with
h [ n ] , then,
n
y[ n] = ∑ x [ m ]h [ n – m ] (9.24)
m=0
or
n
y[ n] = ∑ h [ n – m ]x [ m ] (9.25)
m=0

We will now prove that convolution in the discrete time domain corresponds to multiplication in
the Z domain, that is,

f 1 [ n ]∗ f 2 [ n ] ⇔ F 1 ( z ) ⋅ F 2 ( z ) (9.26)

9-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the Z Transform

Proof:
Taking the Z transform of both sides of (9.24), we get

⎧ ∞ ⎫ ∞ ∞
–n
Y(z) = Z ⎨
⎩m=0
∑
x [ m ]h [ n – m ] ⎬ =
⎭
∑ ∑ x [ m ]h [ n – m ] z
n=0 m=0

and interchanging the order of the summation, we get

∞ ∞ ∞ ∞
–n –n
Y(z) = ∑ ∑ x [ m ]h [ n – m ]z = ∑ x[m] ∑ h[n – m] z
m=0 n=0 m=0 n=0

Next, we let k = n – m , then, n = k + m , and thus,

∞ ∞ ∞ ∞
–( k + m ) –m –k
Y(z) = ∑ x[m] ∑ h[ k]z = ∑ x [ m ]z ∑ h[ k]z
m=0 n=0 m=0 n=0
or
Y( z) = X(z ) ⋅ H(z) (9.27)
10. Convolution in the Discrete Frequency Domain
If f 1 [ n ] and f 2 [ n ] are two sequences with Z transforms F 1 ( z ) and F 2 ( z ) respectively, then,

°∫ xF ( v )F ⎝ -v-⎠ v
1
f 1 [ n ] ⋅ f 2 [ n ] ⇔ -------- ⎛ ⎞ z –1
dv (9.28)
1 2
j 2π

where v is a dummy variable, and

°∫ is a closed contour inside the overlap convergence regions
for X 1 ( v ) and X 2 ( z ⁄ v ) . The proof requires contour integration; it will not be provided here.

11. Initial Value Theorem

f [ 0 ] = lim X ( z ) (9.29)
z→∞

Proof:
For all n ≥ 1 , as z → ∞
–n 1
z = ---- → 0
n
z

–n
and under these conditions f [ n ]z → 0 also. Taking the limit as z → ∞ in the expression

Signals and Systems with MATLAB Applications, Second Edition 9-9

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

∞ –n
F(z) = ∑ f [n] z
n=0

we observe that the only non-zero value in the summation is that of n = 0 . Then,
∞ –n –0
∑ f [ n ]z = f [0] z = f [0]
n=0
Therefore,
lim F ( z ) = f [ 0 ]
z→∞

12. Final Value Theorem

This theorem states that if f [ n ] approaches a limit as n → ∞ , we can find that limit, if it exists, by
multiplying the Z transform of f [ n ] by ( z – 1 ) , and taking the limit of the product as z → 1 .
That is,

lim f [ n ] = lim ( z – 1 ) F ( z ) (9.30)

n→∞ z→1

Proof:
Let us consider the Z transform of the sequence f [ n + 1 ] – f [ n ] , i.e.,
∞ –n
Z { f [n + 1] – f [n]} = ∑ ( f [ n + 1 ] – f [ n ] )z
n=0

We replace the upper limit of the summation with k, and we let k → ∞ . Then,
k
–n
Z { f [ n + 1 ] – f [ n ] } = lim
k→∞ ∑ ( f [ n + 1 ] – f [ n ] )z (9.31)
n=0

From (9.15),
Z { f [ n + 1 ] } = zF ( z ) – f [ 0 ]z (9.32)
and by substitution of (9.32) into (9.31), we get
k
–n
zF ( z ) – f [ 0 ]z – F ( z ) = lim
k→∞ ∑ ( f [ n + 1 ] – f [ n ] )z
n=0

Taking the limit as z → 1 on both sides, we get

9-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Z Transform of Common Discrete Time Functions

⎧ ⎫
k
–n
lim { ( z – 1 )F ( z ) – f [0 ]z } = lim ⎨ lim
z→1 z → 1 ⎩k → ∞ ∑ ( f [ n + 1 ] – f [ n ] )z ⎬
⎭
n=0
k
–n
= lim
k→∞ ∑ zlim
→1
( f [ n + 1 ] – f [ n ] )z
n=0

k
–n
lim ( z – 1 )F ( z ) – lim f [ 0 ]z = lim
z→1 z→1 k→∞ ∑ zlim
→1
{ f [n + 1] – f [n]} z
n=0

k
lim ( z – 1 )F ( z ) – f [ 0 ] = lim
z→1 k→∞ ∑
n=0
{f [n + 1]} – f [n]

= lim { f [ k ] – f [ 0 ] } = lim f [ k ] – f [ 0 ]
k→∞ k→∞

lim f [ k ] = lim ( z – 1 )F ( z )
k→∞ z→1

We must remember, however, that if the sequence f [ n ] does not approach a limit, the final value
theorem is invalid. The right side of (9.30) may exist even though f [ n ] does not approach a limit.
In instances where we cannot determine whether f [ n ] exists or not, we can be certain that it
exists, if X ( z ) can be expressed in a proper rational form as

X(z) = A ( z )-
----------
B(z)

where A ( z ) and B ( z ) are polynomials with real coefficients.

We summarize the properties and theorems of the Z transform in Table 9.1.

9.3 The Z Transform of Common Discrete Time Functions

In this section we will provide several examples to find the Z transform of some discrete time func-
tions.

Example 9.1
Find the Z transform of the geometric sequence defined as

⎧0 n = – 1 , – 2, – 3, …
f [n] = ⎨ n (9.33)
⎩a n = 0, 1, 2, 3, …

Signals and Systems with MATLAB Applications, Second Edition 9-11

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

TABLE 9.1 Properties and Theorems of the Z transform

Property / Theorem Time Domain Z transform
Linearity af 1 [ n ] + bf 2 [ n ] + … aF 1 ( z ) + bF 2 ( z ) + …
Shift of x [ n ]u 0 [ n ] f [ n – m ]u 0 [ n – m ] z
–m
F(z)
Right Shift f [n – m] –m
m–1
–n
z F(z) + ∑ f [ n – m ]z
n=0

Left Shift f [n + m] –1
–n
∑
m
z F(z ) + f [ n + m ]z
n = –m

Multiplication by a n n
a f [n] F ⎛ --- ⎞
z
⎝a ⎠
– naT – naT aT
Multiplication by e e f [n] F( e z)
Multiplication by n nf [ n ] d
– z ----- F ( z )
dz
Multiplication by n 2 2
n f [n] d 2d
2
z ----- F ( z ) + z -------2- F ( z )
dz dz
Summation in Time n
⎛ ----------
z ⎞
∑ f [m]
m=0
⎝z – 1⎠
- F(z)

Time Convolution f 1 [ n ]∗ f 2 [ n ] F1 ( z ) ⋅ F2 ( z )

Frequency Convolution f1 [ n ] ⋅ f2 [ n ]
xF 1 ( v )F 2 ⎛ --⎞ v dv
1 z –1
∫
j 2π °
--------
⎝ v⎠
Initial Value Theorem f [ 0 ] = lim F ( z )
z→∞
Final Value Theorem lim f [ n ] = lim ( z – 1 ) F ( z )
n→∞ z→1

Solution:
From the definition of the Z transform,

∞ –n ∞ n –n ∞ –1 n
F( z) = ∑ f [ n ]z = ∑ a z = ∑ ( az ) (9.34)
n=0 n=0 n=0

9-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Z Transform of Common Discrete Time Functions

To evaluate this infinite summation, we form a truncated version of F ( z ) which contains the first k
terms of the series. We denote this truncated version as F k ( z ) . Then,
k–1
n –n –1 2 –2 k – 1 –( k – 1 )
Fk ( z ) = ∑a z = 1 + az +a z +…+a z (9.35)
n=0

and we observe that as k → ∞ , (9.35) becomes the same as (9.34).

–1
To express (9.35) in a closed form, we multiply both sides by az . Then,
–1 –1 2 –2 3 –3 k –k
az F k ( z ) = az +a z +a z +…+a z (9.36)

Subtracting (9.36) from (9.35), we get

–1 k –k
F k ( z ) – az F k ( z ) = 1 – a z
or

k –k –1 k
1–a z 1 – ( az )
F k ( z ) = --------------------- = -------------------------- (9.37)
–1 –1
1 – az 1 – az

–1
for az ≠ 1
–1 k
To determine F ( z ) from F k ( z ) , we examine the behavior of the term ( az ) in the numerator of
–1 –1 k –1 –1 jθ
(9.37). We write the terms az and ( az ) in polar form, that is, az = az e and

–1 k – 1 k jkθ
( az ) = az e (9.38)
–1
From (9.38) we observe that, for the values of z for which az < 1 , the magnitude of the complex
–1 k
number ( az ) → 0 as k → ∞ and therefore,
1 z
F ( z ) = lim F k ( z ) = ------------------- = ----------- (9.39)
k→∞ 1 – az
– 1 z–a
–1
for az <1
–1 –1 k
For the values of z for which az > 1 , the magnitude of the complex number ( az ) becomes
–1
unbounded as k → ∞ , and therefore, F ( z ) = lim F k ( z ) is unbounded for az > 1.
k→∞

In summary,
∞ –1 n
F( z) = ∑ ( az )
n=0

Signals and Systems with MATLAB Applications, Second Edition 9-13

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

–1 –1
converges to the complex number z ⁄ ( z – a ) for az < 1 , and diverges for az >1.
Also, since
–1 a a
az = --- = -----
z z

–1 –1
then, az < 1 implies that z > a , while az > 1 implies z < a and thus,

⎧ z
∞ n –n ⎪ ----------- for z > a
∑a z
n
Z { a u0 [ n ] } = = ⎨ z – a (9.40)
n=0 ⎪ unbounded for z < a
⎩

The regions of convergence and divergence for the sequence of (9.40) are shown in Figure 9.2.

Im[z]
Region of
Convergence
z
F ( z ) = -----------
z–a
Region of |a|
Divergence Re[z]
F(z) → ∞

n
Figure 9.2. Regions of convergence and divergence for the geometric sequence a
To determine whether the circumference of the circle, where z = a |, lies in the region of conver-
gence or divergence, we evaluate the sequence F k ( z ) at z = a . Then,
k–1
n –n –1 2 –2 k – 1 –( k – 1 )
Fk ( z ) = ∑a z
n=0
= 1 + az +a z +…+a z
(9.41)
z=a
= 1+1+1+…+1 = k

We see that this sequence becomes unbounded as k → ∞ , and therefore, the circumference of the
circle lies in the region of divergence.

Example 9.2
Find the Z transform of the discrete unit step function u 0 [ n ] shown in Figure 9.3.

9-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Z Transform of Common Discrete Time Functions

u0 [ n ]
⎧0 n<0 1
u0 [ n ] = ⎨ ....
⎩1 n≥0
0
n

Figure 9.3. The discrete unit step function u 0 [ n ]

Solution:
From the definition of the Z transform,
∞ –n ∞ –n
F(z) = ∑ f [ n ]z = ∑ [ 1 ]z (9.42)
n=0 n=0

As in the previous example, to evaluate this infinite summation, we form a truncated version of
F ( z ) which contains the first k terms of the series, and we denote this truncated version as F k ( z ) .
Then,
k–1
–n –1 –2 –( k – 1 )
Fk ( z ) = ∑z = 1+z +z +…+z (9.43)
n=0

and we observe that as k → ∞ , (9.43) becomes the same as (9.42). To express (9.43) in a closed
–1
form, we multiply both sides by z and we get
–1 –1 –2 –3 –k
z Fk ( z ) = z +z +z +…+z (9.44)

Subtracting (9.44) from (9.43), we get

–1 –k
Fk ( z ) – z Fk ( z ) = 1 – z
or
–k –1 k
1–z 1 – (z )
F k ( z ) = ---------------- = ----------------------- (9.45)
–1 –1
1–z 1–z

–1
for z ≠ 1
–1 k – 1 k jkθ –1 k
Since ( z ) = z e , as k → ∞ , ( z ) → 0 . Therefore,
1 z
F ( z ) = lim F k ( z ) = ---------------- = ----------- (9.46)
k→∞ 1–z
– 1 z–1

for z > 1 , and the region of convergence lies outside the unit circle.

Signals and Systems with MATLAB Applications, Second Edition 9-15

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

Alternate Solution:
n n
The discrete unit step u 0 [ n ] is a special case of the sequence a with a = 1 , and since 1 = 1 , by
substitution into (9.40) we get

⎧ z
∞ –n ⎪ ----------- for z > 1
Z { u0 [ n ] } = ∑ [ 1 ]z = ⎨ z–1
⎪ unbounded for z < 1
(9.47)
n=0
⎩

Example 9.3
– naT
Find the Z transform of the discrete exponential sequence f [ n ] = e
Solution:
∞
– naT – n – aT – 1 – 2aT – 2 – 3aT – 3
F( z) = ∑e
n=0
z = 1+e z +e z +e z +…

and this is a geometric sequence which can be expressed in closed form as

– naT 1 z
Z [e ] = -------------------------- = -----------------
– aT
- (9.48)
– aT – 1
1–e z z – e

– aT – 1
for e z <1

Example 9.4
Find the Z transform of the discrete time functions f 1 [ n ] = cos n aT and f 2 [ n ] = sin naT
Solution:
From (9.48) of Example 9.3,
– naT z
e ⇔ -----------------
– aT
-
z–e
and replacing – naT with jnaT we get
jnaT z
Z [e ] = Z [ cos naT + j sin naT ] = -------------------
jaT
z–e
– j aT
z z–e
= Z [ cos naT ] + jZ [ sin n aT ] = ------------------
jaT
- ⋅ --------------------
– j aT
z–e z–e
2
z – z cos aT + jz sin aT
= Z [ cos naT ] + jZ [ sin n aT ] = ------------------------------------------------------
2
z – 2z cos aT + 1

9-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Z Transform of Common Discrete Time Functions

Equating real and imaginary parts, we get the transform pairs

2
z – z cos aT
cos naT ⇔ -----------------------------------------
2
(9.49)
z – 2z cos aT + 1
and
z cos aT
sin n aT ⇔ -----------------------------------------
2
(9.50)
z – 2z cos aT + 1
To define the regions of convergence and divergence, we express the denominator of (9.49) or
(9.50) as
jaT – j aT
(z – e ) ⋅ (z – e ) (9.51)
jaT
We see that both pairs of (9.49) and (9.50) have two poles, one at z = e and the other at
– j aT
z = e , that is, the poles lie on the unity circle as shown in Figure 9.4.

Im[z] Region of
Convergence

×Pole
1
Re[z]
Region of
Divergence
×Pole

Figure 9.4. Regions of convergence and divergence for cos n aT and sin n aT
From Figure 9.4, we see that the poles separate the regions of convergence and divergence. Also,
since the circumference of the circle lies on the region of divergence, as we have seen before, the
poles lie on the region of divergence. Therefore, for the discrete time cosine and sine functions we
have the pairs

z 2 – z cos aT
cos naT ⇔ ----------------------------------------- for z > 1 (9.52)
z 2 – 2z cos aT + 1

and

z sin a T
sin n aT ⇔ -----------------------------------------
2
for z > 1 (9.53)
z – 2z cos aT + 1

Signals and Systems with MATLAB Applications, Second Edition 9-17

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

It is shown in complex variables theory that if F ( z ) is a proper rational function*, all poles lie outside
the region of convergence, but the zeros can lie anywhere on the z -plane.
Example 9.5
Find the Z transform of discrete unit ramp f [ n ] = nu 0 [ n ] .

Solution:
∞ –n –1 –2 –3
Z { nu 0 [ n ] } = ∑ nz = 0+z + 2z + 3z +… (9.54)
n=0

We can express (9.54) in closed form using the discrete unit step function transform pair
∞ –n z
Z { u0 [ n ] } = ∑ ( 1 )z = ----------- for
z–1
z > 1 (9.55)
n=0

Differentiating both sides of (9.55) with respect to z , we get

∞
d ⎛ – n⎞ d ⎛ ----------
z ⎞
dz ⎝ ∑ ( 1 )z ⎠
=
dz z – 1 ⎠
⎝
-
n=0
or
∞ –n–1 –1 -
∑ –n z = -----------------
(z – 1)
2
n=0
Multiplication by – z yields
∞ –n ∞ –n
∑ ∑ (1) z
nz = n z -
= -----------------
2
n=0 n=0 (z – 1)

and thus we have the transform pair

z
nu 0 [ n ] ⇔ -----------------2- (9.56)
(z – 1)

We summarize the transform pairs we have derived, and others, given as exercises at the end of this
chapter, in Table 9.2.

* This was defined in Chapter 3, page 3-1.

9-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Z Transform of Common Discrete Time Functions

TABLE 9.2 The Z Transform of Common Discrete Time Functions

f[n] F(z)
δ[n] 1
δ[n – m ] z
–m

n z
a u0 [ n ] ----------- z >a
z–a
u0 [ n ] z
----------- z >1
z–1
– naT z
(e )u 0 [ n ] ------------------ e
– aT – 1
z <1
– aT
z–e
( cos naT )u 0 [ n ] 2
z – z cos aT
-----------------------------------------
2
z >1
z – 2z cos aT + 1
( sin n aT )u 0 [ n ] z sin a T
----------------------------------------- z > 1
2
z – 2z cos aT + 1
n 2
( a cos naT )u 0 [ n ] z – az cos aT
-----------------------------------------------
2 2
z >a
z – 2az cos aT + a
n az sin a T
( a sin naT )u 0 [ n ] ----------------------------------------------- z >a
2 2
z – 2az cos aT + a
u0 [ n ] –u0 [ n – m ] m
z –1
----------------------------
m–1
z (z – 1)
nu 0 [ n ] z ⁄ (z – 1)
2

3
2
n u0 [ n ] z(z + 1) ⁄ (z – 1)

[ n + 1 ]u 0 [ n ] 2
z ⁄ (z – 1)
2

n 2
a nu 0 [ n ] ( az ) ⁄ ( z – a )
n 2 3
a n u0 [ n ] az ( z + a ) ⁄ ( z – a )
n 2 3
a n [ n + 1 ]u 0 [ n ] 2az ⁄ ( z – a )

Signals and Systems with MATLAB Applications, Second Edition 9-19

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

9.4 Computation of the Z Transform with Contour Integration*

Let F ( s ) be the Laplace transform of a continuous time function f ( t ) and F∗ ( s ) the transform of
the sampled time function f ( t ) which we denote as f∗ ( t ) . It is shown in complex variables theory
that F∗ ( s ) can be derived from F ( s ) by the use of the contour integral
1 F( v)
F∗ ( s ) = --------
j2π °∫C -------------------------- dv
1–e e
– sT vT
(9.57)

where C is a contour enclosing all singularities (poles) of F ( s ) , and v is a dummy variable for s . We
can compute the Z transform of a discrete time function f [ n ] using the transformation

F ( z ) = F∗ ( s ) sT
(9.58)
z=e

By substitution of (9.58) into (9.57), and replacing v with s , we get

1 F( s)
F ( z ) = --------
j2π °∫C ----------------------- ds
1–z e
– 1 sT
(9.59)

Next, we use Cauchy’s Residue Theorem to express (9.59) as

F(s) F(s)
F(z) = ∑ Res 1-----------------------
–z e
– 1 sT
= lim ( s – p k ) -----------------------
s → pk 1–z e
– 1 sT
(9.60)
k s = pk

Example 9.6
Derive the Z transform of the discrete unit step function u 0 [ n ] using the residue theorem.
Solution:
We learned in Chapter 2, that
L [ u0 ( t ) ] = 1 ⁄ s
Then, by residue theorem of (9.60),
F(s) 1⁄s
F ( z ) = lim ( s – p k ) -----------------------
– 1 sT
= lim ( s – 0 ) -----------------------
– 1 sT
s → pk 1–z e s→0 1–z e
1⁄s 1 1 z
= lim s ----------------------- = lim ------------------------
- = ---------------- = -----------
s → 0 1 – z – 1 e sT s → 0 1 – z – 1 e sT 1–z
–1 z – 1

* This section may be skipped without loss of continuity. It is intended for readers who have prior knowledge of
complex variables theory. However, the following examples will show that this procedure is not difficult.

9-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Computation of the Z Transform with Contour Integration

for z > 1 , and this is the same as (9.47).

Example 9.7
– naT
Derive the Z transform of the discrete exponential function e u 0 [ n ] using the residue theorem.

Solution:
From Chapter 2,
– at 1-
L [e u 0 ( t ) ] = ----------
s+a
Then, by residue theorem of (9.60),
F(s) 1 ⁄ (s + a)
F ( z ) = lim ( s – p k ) -----------------------
– 1 sT
= lim ( s + a ) -----------------------
– 1 sT
s → pk 1–z e s → – a 1–z e
1
= lim ------------------------ 1 z
- = -------------------------- = ------------------
s → –a 1 – z e – 1 sT – 1 – a T –a T
1–z e z–e

for z > 1 and this is the same as (9.48).

Example 9.8
Derive the Z transform of the discrete unit ramp function nu 0 [ n ] using the residue theorem.
Solution:
From Chapter 2,
2
L [ tu 0 ( t ) ] = 1 ⁄ s

Since F ( s ) has a second order pole at s = 0 , we need to apply the residue theorem applicable to a
pole of order n. This theorem states that

n–1
F( s)
F ( z ) = lim ⎛ ------------------⎞ ( s – p k ) -------------
1 d
- ----------------------- (9.61)
s → pk⎝ ( n – 1 )! ⎠ n – 1 – 1 sT
ds 1–z e

Thus, for this example,

2
1⁄s
d s 2 ----------------------- 1
d ----------------------- z -
F ( z ) = lim ----
- – 1 sT
= lim ----
- – 1 sT
= -----------------
s → 0 ds 1–z e s → 0 ds 1–z e (z – 1)
2

for z > 1 , and this is the same as (9.56).

Signals and Systems with MATLAB Applications, Second Edition 9-21

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

9.5 Transformation Between s and z Domains

It is shown in complex variables textbooks that every function of a complex variable maps (trans-
forms) a plane xy to another plane uv . In this section, we will investigate the mapping of the plane
of the complex variable s , into the plane of the complex variable z .
Let us reconsider expressions (9.6) and (9.1) which are repeated here for convenience.
∞
– nsT
G(s) = ∑
n=0
f [n]e (9.62)

and
∞ –n
F( z) = ∑ f [ n ]z (9.63)
n=0

By comparison of (9.62) with (9.63),

G(s ) = F( z) sT
(9.64)
z=e

Thus, the variables s and z are related as

sT
z = e (9.65)

and

1
s = --- ln z (9.66)
T

Therefore,

F(z ) = G(s) 1
(9.67)
s = --- ln z
T

Since s , and z are both complex variables, relation (9.67) allows the mapping (transformation) of
regions of the s -plane into the z -plane. We find this transformation by recalling that s = σ + jω and
therefore, expressing z in magnitude-phase form and using (9.65), we get
jθ σT jωT
z = z ∠θ = z e = e e (9.68)
where,
σT
z = e (9.69)
and

9-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Transformation Between s and z Domains

θ = ωT (9.70)
Since
T = 1 ⁄ fs

the period T defines the sampling frequency f s . Then, ω s = 2πf s or f s = ω s ⁄ 2π , and

T = ( 2π ) ⁄ ω s
Therefore, we express (9.70) as
2π ω
θ = ω ------ = 2π ------ (9.71)
ωs ωs

and by substitution of (9.69) and (9.71) into (9.68), we get

σT j2π ( ω ⁄ ω s )
z = e e (9.72)

The quantity e j2π ( ω ⁄ ωs ) in (9.72), defines the unity circle; therefore, let us examine the behavior of
z when σ is negative, zero, or positive.

Case I σ < 0 : When σ is negative, from (9.69), we see that z < 1 , and thus the left half of the s -
plane maps inside the unit circle of the z -plane, and for different negative values of
σ , we get concentric circles with radius less than unity.

Case II σ > 0 : When σ is positive, from (9.69), we see that z > 1 , and thus the right half of the s -
plane maps outside the unit circle of the z -plane, and for different positive values of
σ we get concentric circles with radius greater than unity.

Case III σ = 0 : When σ is zero, from (9.72), we see that z = e j2π ( ω ⁄ ωs ) and all values of ω lie
on the circumference of the unit circle. For illustration purposes, we have mapped
several fractional values of the sampling radian frequency ω s , and these are shown
in Table 9.3.
From Table 9.3, we see that the portion of the jω axis for the interval 0 ≤ ω ≤ ω s in the s -plane,
maps on the circumference of the unit circle in the z -plane as shown in Figure 9.5.
The mapping from the z -plane to the s -plane is a multi-valued transformation since, as we have
1
seen, s = --- ln z and it is shown in complex variables textbooks that
T

ln z = ln z + j2nπ (9.73)

Signals and Systems with MATLAB Applications, Second Edition 9-23

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

TABLE 9.3 Mapping of multiples of sampling frequency

ω z θ
0 1 0
ωs ⁄ 8 1 π⁄4

ωs ⁄ 4 1 π⁄2

3ω s ⁄ 8 1 3π ⁄ 4

ωs ⁄ 2 1 π

5ω s ⁄ 8 1 5π ⁄ 4

3ω s ⁄ 4 1 3π ⁄ 2

7ω s ⁄ 8 1 7π ⁄ 4

ωs 1 2π

jω s-plane Im [ z ] z-plane

ω = 0.25ω s
ωs x

0.875ω s x x 0.125ω s
0.375ω s
0.75ω s z =1
0.625ω s ω = 0.5ω s ω = 0
x x Re [ z ]
0.5ω s 0 ω = ωs
0.375ω s
x x
0.25ω s 0.625ω s 0.875ω s
0.125ω s x

σ ω = 0.75ω s
σ<0 σ = 0 σ>0

Figure 9.5. Mapping of the s-plane to z-plane

9.6 The Inverse Z Transform

The Inverse Z transform enables us to extract f [ n ] from F ( z ) . It can be found by any of the follow-
ing three methods:
a. Partial Fraction Expansion
b. The Inversion Integral
c. Long Division of polynomials

9-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform

We will discuss these methods, and we will illustrate with examples.

a. Partial Fraction Expansion
This method is very similar to the partial fraction expansion method that we used in finding the
Inverse Laplace transform, that is, we expand F ( z ) into a summation of terms whose inverse is
known. These terms have the form
r1 z r2 z r3 z
k, ------------
-, -------------------
- , ------------
-, … (9.74)
z – p1 ( z – p ) 2 z – p2
1

where k is a constant, and r i and p i represent the residues and poles respectively; these can be real
or complex.
Before we expand F ( z ) into partial fractions, we must express it as a proper rational function. This
is done by expanding F ( z ) ⁄ z instead of F ( z ) , that is, we expand it as

F ( z )- = k-- + ------------
----------
r1 r2
-+…
- + ------------ (9.75)
z z z – p1 z – p2

and after the residues are found from

F(z ) F(z)
r k = lim ( z – p k ) ----------- = ( z – p k ) ----------- (9.76)
z → pk z z z = pk

we rewrite (9.75) as
r1 z r2 z
F ( z ) = k + ------------ -+…
- + ------------ (9.77)
z – p1 z – p2

Example 9.9
Use the partial fraction expansion method to compute the Inverse Z transform of
1
F ( z ) = ---------------------------------------------------------------------------------- (9.78)
–1 –1 –1
( 1 – 0.5 z ) ( 1 – 0.75 z ) ( 1 – z )

Solution:
We multiply both numerator and denominator by z 3 to eliminate the negative powers of z . Then,
3
z
F ( z ) = -------------------------------------------------------------
( z – 0.5 ) ( z – 0.75 ) ( z – 1 )

Next, we form F ( z ) ⁄ z , and we expand in partial fractions as

Signals and Systems with MATLAB Applications, Second Edition 9-25

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

2
F ( z )- = ------------------------------------------------------------
---------- z r1
- = -------------------
r2
- + ----------------------
r3
- + ---------------
-
z ( z – 0.5 ) ( z – 0.75 ) ( z – 1 ) ( z – 0.5 ) ( z – 0.75 ) ( z – 1 )
The residues are
2 2
z ( 0.5 )
r 1 = ---------------------------------------- = -------------------------------------------------- = 2
( z – 0.75 ) ( z – 1 ) z = 0.5
( 0.5 – 0.75 ) ( 0.5 – 1 )
2 2
z
r 2 = ------------------------------------
- ( 0.75 )
= ----------------------------------------------------
- = –9
( z – 0.5 ) ( z – 1 ) z = 0.75
( 0.75 – 0.5 ) ( 0.75 – 1 )
2 2
z 1
r 3 = -------------------------------------------- = ---------------------------------------------- = 8
( z – 0.5 ) ( z – 0.75 ) z=1
( 1 – 0.5 ) ( 1 – 0.25 )
Then,
2
F ( z )- = ------------------------------------------------------------
---------- z – 9 - + ---------------
2 - + ----------------------
- = ------------------- 8 -
z ( z – 0.5 ) ( z – 0.75 ) ( z – 1 ) ( z – 0.5 ) ( z – 0.75 ) ( z – 1 )

and multiplication of both sides by z yields

3
z 2z – 9z 8z
F ( z ) = ------------------------------------------------------------- = -------------------- + ----------------------- + ---------------- (9.79)
( z – 0.5 ) ( z – 0.75 ) ( z – 1 ) ( z – 0.5 ) ( z – 0.75 ) ( z – 1 )
To find the Inverse Z transform of (9.79), we recall that
n z
a ⇔ -----------
z–a

for z > a . Therefore, the discrete time sequence is

n n
f [ n ] = 2 ( 0.5 ) – 9 ( 0.75 ) + 8 (9.80)
Check with MATLAB:
Dz=(z−0.5)*(z−0.75)*(z−1) % The denominator of F(z)
collect(Dz); % Multiply the three factors of F(z) to get a polynomial
ans =
z^3-9/4*z^2+13/8*z-3/8
num=[0 1 0 0]; % The coefficients of the numerator
den=[1 −9/4 13/8 −3/8]; % The coefficients of the denominator
fprintf(' \n');
[num,den]=residue(num,den); % Verify the residues in (9.79)
fprintf('r1 = %4.2f \t', num(1)); fprintf('p1 = %4.2f \t', den(1));...
fprintf('r2 = %4.2f \t', num(2)); fprintf('p2 = %4.2f \t', den(2));...
fprintf('r3 = %4.2f \t', num(3)); fprintf('p3 = %4.2f \t', den(3))

9-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform

r1 = 8.00 p1 = 1.00 r2 = -9.00 p2 = 0.75 r3 = 2.00 p3 = 0.50

syms n z
fn=2*(0.5)^n−9*(0.75)^n+8; % This is the answer in (9.80)
Fz=ztrans(fn,n,z); simple(Fz) % Verify answer by first taking Z transform of f[n]
ans =
8*z^3/(2*z-1)/(4*z-3)/(z-1)
iztrans(Fz) % Now, verify that Inverse of F(z) gives back f[n]
ans =
2*(1/2)^n-9*(3/4)^n+8
We can use Microsoft Excel to obtain and plot the values of f [ n ] . The spreadsheet of Figure 9.6
shows the first 25 values of n but only part of the spreadsheet is shown.

n f[n] n n
Discrete Time Sequence f[n] = 2(0.5) − 9(0.75) + 8
0.000 1.0000 8
1.000 2.2500
2.000 3.438 7
3.000 4.453 6
4.000 5.277
5.000 5.927 5
6.000 6.429 4
7.000 6.814
8.000 7.107 3
9.000 7.328
2
10.000 7.495
11.000 7.621 1
12.000 7.715
0
13.000 7.786
1

11

13

15

17

19

21

23

25

14.000 7.84
15.000 7.88
n n
Figure 9.6. The discrete time sequence f [ n ] = 2 ( 0.5 ) – 9 ( 0.75 ) + 8

Example 9.10
Use the partial fraction expansion method to compute the Inverse Z transform of
12z
F ( z ) = ----------------------------------- (9.81)
2
( z + 1 )( z – 1 )
Solution:
Division of both sides by z and partial expansion yields

Signals and Systems with MATLAB Applications, Second Edition 9-27

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

F ( z -) = ----------------------------------
---------- 12 r1
- = ---------------
r2
- + -----------------
r3
- + ---------------
-
z (z + 1)(z – 1)
2 (z + 1) (z – 1) 2 (z – 1)

The residues are

12 12
r 1 = ------------------ = ----------------------- = 3
2 2
(z – 1) z = –1
(– 1 – 1)
12 -
r 2 = --------------- 12 - = 6
= ----------------
(z + 1) z=1
(1 + 1)

– 12
r 3 = ----- ⎛ ----------- ⎞
d 12
= ------------------ = – 3
dz ⎝ z + 1 ⎠ (z + 1)
2
z=1
Then,
F(z) 12 3 6 –3
----------- = ----------------------------------- = ---------------- + ------------------ + ----------------
z (z + 1)(z – 1)
2 ( z + 1 ) (z – 1)
2 ( z – 1)
or
12z 3z 6z – 3z
F ( z ) = ----------------------------------- = ------------------------ + ------------------ + ----------------
( z + 1 )( z – 1 )
2 ( z – ( – 1 ) ) (z – 1)
2 ( z – 1)
Now, we recall that
z
u 0 [ n ] ⇔ -----------
z–1
and
z
nu 0 [ n ] ⇔ -----------------2-
(z – 1)
for z > 1 .
Therefore, the discrete time sequence is
n
f [ n ] = 3 ( – 1 ) + 6n – 3 (9.82)
Check with MATLAB:
syms n z; fn=3*(−1)^n+6*n−3; Fz=ztrans(fn); simple(Fz)
ans =
12*z/(z+1)/(z-1)^2
We can also use the MATLAB dimpulse function to compute and display f [ n ] for any range of val-
ues of n . The following code will display the first 20 values of f [ n ] in (9.82).
% First, we must express the denominator of F(z) as a polynomial
denpol=collect((z+1)*((z−1)^2))

9-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform

denpol =
z^3-z^2-z+1
num=[12 0]; % The coefficients of the numerator of F(z) in (9.81)
den=[1 −1 −1 1]; % The coefficients of the denominator in polynomial form
fn=dimpulse(num,den,20) % Compute the first 20 values of f[n]
fn =
0
0
12
12
24
24
36
36
48
48
60
60
72
72
84
84
96
96
108
108
The MATLAB function dimpulse(num,den) plots the impulse response of the polynomial trans-
fer function G ( z ) = num ( z ) ⁄ den ( z ) where num ( z ) and den ( z ) contain the polynomial coeffi-
cients in descending powers of z . Thus, the MATLAB code
num=[0 0 12 0]; den=[1 −1 −1 1]; dimpulse(num,den)
displays the plot of Figure 9.7.

Example 9.11
Use the partial fraction expansion method to compute the Inverse Z transform of
z+1
F ( z ) = ----------------------------------------------- (9.83)
2
( z – 1 ) ( z + 2z + 2 )

Signals and Systems with MATLAB Applications, Second Edition 9-29

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

Figure 9.7. The impulse response for Example 9.10

Solution:
Dividing both sides by z and performing partial fraction expansion, we get

F(z) z+1 r r2 r3 r4
----------- = -------------------------------------------------- = ----1 + ----------
- + -----------------------
- + -----------------------
- (9.84)
z 2
z ( z – 1 ) ( z + 2z + 2 ) z z – 1 (z + 1 – j) (z + 1 + j)

The residues are

z+1
r 1 = ----------------------------------------------
- 1- = – 0.5
= -----
2
( z – 1 ) ( z + 2z + 2 ) –2
z=0
z+1 2
r 2 = -------------------------------------- = --- = 0.4
2 5
( z ) ( z + 2z + 2 ) z=1
z+1 j
r 3 = ------------------------------------------------ = --------------------------------------------------- = 0.05 + j0.15
(z)(z – 1)(z + 1 + j) z = –1+j
( – 1 + j ) ( – 2 + j ) ( j2 )
r 4 = r∗ 3 = 0.05 – j0.15
Then,
F ( z -) = -------------------------------------------------
---------- z+1 - = – 0.5- + ----------
--------- 0.05 + j0.15- + ----------------------------
0.4- + ---------------------------- 0.05 – j0.15-
z 2
z ( z – 1 ) ( z + 2z + 2 ) z z – 1 ( z + 1 – j ) (z + 1 + j)
or

F ( z ) = – 0.5 + ---------- 0.05 + j0.15 )z + (-----------------------------------

0.4z- + (------------------------------------ 0.05 – j0.15 )z-
z–1 (z + 1 – j) (z + 1 + j)
0.4z ( 0.05 + j0.15 )z ( 0.05 – j0.15 )z
= – 0.5 + ----------- + ------------------------------------ + ------------------------------------
z–1 z – (– 1 + j) z – (– 1 – j)
( 0.05 + j0.15 )z + (-----------------------------------
0.4z- + ------------------------------------
= – 0.5 + ---------- 0.05 – j0.15 )z-
z–1 z – 2e
j135°
z – 2e
– j 135°

9-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform

Recalling that
δ[n] ⇔ 1
and
n z
a u 0 [ n ] ⇔ -----------
z–a

for z > a , we find that the discrete time sequence is

n j135° n
f [ n ] = – 0.5δ [ n ] + 0.4 ( 1 ) + ( 0.05 + j0.15 ) ( 2e )
– j 135° n
+ ( 0.05 – j 0.15 ) ( 2e )
n jn135° n – j n135°
= – 0.5δ [ n ] + 0.4 + 0.05 ( 2 e ) + 0.05 ( 2 e )
n jn135° n – j n135°
+ j0.15 ( 2 e ) – j0.15 ( 2 e )
or
n n
2 3 2
f [ n ] = – 0.5δ [ n ] + 0.4 + --------- cos n135° – ------------ sin n135° (9.85)
10 10

We will use the MATLAB dimpulse function to display the first 8 values of f [ n ] in (9.85). We
recall that his function requires that F ( z ) is expressed as a ratio of polynomials in descending order.
syms n z
collect((z−1)*(z^2+2*z+2)) % First, expand denominator of given F(z)
ans =
z^3+z^2-2
The following code displays the first 10 values of f [ n ] and plots the impulse response shown in Fig-
ure 9.8.
num=[0 0 1 1]; den=[1 1 0 −2]; fn=dimpulse(num,den,11), dimpulse(num,den,16)
fn =
0
0
1
0
0
2
-2
2
-6
10

Signals and Systems with MATLAB Applications, Second Edition 9-31

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

Figure 9.8. The impulse response for Example 9.11

b. The Inversion Integral*
The inversion integral states that

1 n–1
j2π °C∫
f [ n ] = -------- F ( z )z dz (9.86)

where C is a closed curve that encloses all poles of the integrant, and by Cauchy’s residue theorem,
this integral can be expressed as

n–1
f [n] = ∑ Res [ F ( z )z ] (9.87)
k z = pk

n–1 n–1
where p k represents a pole of [ F ( z )z ] and Res [ F ( z )z ] represents a residue at z = p k .

Example 9.12
Use the inversion integral method to find the Inverse Z transform of
–1 –3
1 + 2z + z
F ( z ) = ----------------------------------------------------
- (9.88)
–1 –1
( 1 – z ) ( 1 – 0.75z )
Solution:
Multiplication of the numerator and denominator by z 3 yields

* This section may be skipped without loss of continuity. It is intended for readers who have prior knowledge of
complex variables theory.

9-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform

3 2
z + 2z + 1
------------------------------------------ (9.89)
z ( z – 1 ) ( z – 0.75 )

and by application of (9.87),

3 2 n–1 3 2 n–2
( z + 2z + 1 )z ( z + 2z + 1 )z
f [n] = ∑ Res ---------------------------------------------
z ( z – 1 ) ( z – 0.75 )
= ∑ Res ---------------------------------------------
( z – 1 ) ( z – 0.75 )
(9.90)
k z = pk k z = pk

We are interested in the values f [ 0 ], f [ 1 ], f [ 2 ], … , that is, values of n = 0, 1, 2, … .

For n = 0 , (9.90) becomes

3 2
( z + 2z + 1 )
f[ 0] = ∑ Res --------------------------------------------
2
z ( z – 1 ) ( z – 0.75 )
-
k z = pk

3 2 3 2
( z + 2z + 1 ) ( z + 2z + 1 )
= Res --------------------------------------------
2
- + Res --------------------------------------------
2
- (9.91)
z ( z – 1 ) ( z – 0.75 ) z ( z – 1 ) ( z – 0.75 )
z=0 z=1

3 2
( z + 2z + 1 )
+ Res --------------------------------------------
2
-
z ( z – 1 ) ( z – 0.75 )
z = 0.75

The first term on the right side of (9.91) has a pole of order 2 at z = 0 ; therefore, we must evaluate
the first derivative of
3 2
( z + 2z + 1 )
----------------------------------------
( z – 1 ) ( z – 0.75 )

at z = 0 . Thus, for n = 0 , (9.91) reduces to

3 2 3 2 3 2
d ( z + 2z + 1 ) ( z + 2z + 1 ) ( z + 2z + 1 )
f [ 0 ] = ----- ---------------------------------------- + --------------------------------
- + --------------------------------
-
dz ( z – 1 ) ( z – 0.75 ) z=0
2
z ( z – 0.75 )
2
z (z – 1)
z=1 z = 0.75 (9.92)
28 163
= ------ + 16 – --------- = 1
9 9

For n = 1 , (9.90) becomes

Signals and Systems with MATLAB Applications, Second Edition 9-33

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

3 2
( z + 2z + 1 )
f [1] = ∑ Res ------------------------------------------
z ( z – 1 ) ( z – 0.75 )
k z = pk
3 2 3 2
( z + 2z + 1 ) ( z + 2z + 1 )
= Res ------------------------------------------ + Res ------------------------------------------
z ( z – 1 ) ( z – 0.75 ) z=0
z ( z – 1 ) ( z – 0.75 ) z=1
3 2
( z + 2z + 1 ) (9.93)
+ Res ------------------------------------------
z ( z – 1 ) ( z – 0.75 ) z = 0.75
3 2
3
( z + 2z + 1 )
2 3
( z + 2z + 1 )
2 ( z + 2z + 1 )
= ---------------------------------------- + --------------------------------- + ----------------------------------
-
( z – 1 ) ( z – 0.75 ) z=0
z ( z – 0.75 ) z=1
z ( z – 1 )
z = 0.75

--------- = 15
= 4--- + 16 – 163 ------
3 12 4

For n ≥ 2 there are no poles at z = 0 , that is, the only poles are at z = 1 and z = 0.75 . Therefore,
3 2 n–2
( z + 2z + 1 )z
f[ n] = ∑ Res ----------------------------------------------
( z – 1 ) ( z – 0.75 )
k z = pk

3 2 n–2 3 2 n–2
( z + 2z + 1 )z ( z + 2z + 1 )z (9.94)
= Res ---------------------------------------------- + Res ----------------------------------------------
( z – 1 ) ( z – 0.75 ) ( z – 1 ) ( z – 0.75 )
z=1 z = 0.75

3 2 n–2 3 2 n–2
( z + 2z + 1 )z ( z + 2z + 1 )z
= ---------------------------------------------- + ----------------------------------------------
( z – 0.75 ) (z – 1)
z=1 z = 0.75

for n ≥ 2 .
n–2
From (9.94), we observe that for all values of n ≥ 2 , the exponential factor z is always unity for
z = 1 , but varies for values z ≠ 1 . Then,
3 2 3 2 n–2
( z + 2z + 1 ) ( z + 2z + 1 )z
f [ n ] = --------------------------------- + ---------------------------------------------
( z – 0.75 ) z=1
(z – 1) z = 0.75
3 2 n–2
4 - + [------------------------------------------------------------------------------
= --------- 0.75 + 2 ( 0.75 ) + 1 ] ( 0.75 ) - (9.95)
0.25 – 0.25
n
( 163 ⁄ 64 ) ( 0.75 ) - = 16 – 163
= 16 + ----------------------------------------- --------- ( 0.75 ) n
( – 0.25 ) ( 0.75 )
2 9

for n ≥ 2 .

9-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform

We can express f [ n ] for all n ≥ 0 as

28 4 163 n
f [ n ] = ------ δ [ n ] + --- δ [ n – 1 ] + 16 – --------- ( 0.75 ) (9.96)
9 3 9

where the coefficients of δ [ n ] and δ [ n – 1 ] are the residues that were found in (9.92) and (9.93) for
n = 0 and n = 1 respectively at z = 0 . The coefficient 28 ⁄ 9 is multiplied by δ [ n ] to emphasize
that this value exists only for n = 0 and coefficient 4 ⁄ 3 is multiplied by δ [ n – 1 ] to emphasize that
this value exists only for n = 1 .
Check with MATLAB:
syms z n; Fz=(z^3+2*z^2+1)/(z*(z−1)*(z−0.75)); iztrans(Fz)
ans =
4/3*charfcn[1](n)+28/9*charfcn[0](n)+16-163/9*(3/4)^n
We evaluate and plot f [ n ] for the first 20 values. This is shown on the spreadsheet of Figure 9.9.

n f[n]
0 1.000
Discrete Time Sequence of f[0] = 1, f[1] = 3.75,
1 3.750 and f[n] = 16−(163/9)*(3/4)n for n ≥ 2
2 5.813 16
3 8.359
4 10.270
12
5 11.702
6 12.777
7 13.582 8
8 14.187
9 14.640 4
10 14.980
11 15.235
12 15.426 0
13 15.570 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Figure 9.9. The discrete time sequence for Example 9.12

Example 9.13
Use the inversion integral method to find the Inverse Z transform of
1
F ( z ) = ----------------------------------------------------- (9.97)
–1 –1
( 1 – z ) ( 1 – 0.75z )
Solution:

Multiplication of the numerator and denominator by z 2 yields

Signals and Systems with MATLAB Applications, Second Edition 9-35

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

2
z
---------------------------------------
- (9.98)
( z – 1 ) ( z – 0.75 )

This function has no poles at z = 0 . The poles are at z = 1 and z = 0.75 . Then by (9.87),
2 n–1 n+1
z z z
f[n] = ∑ Res ----------------------------------
( z – 1 ) ( z – 0.75 )
= ∑ Res ----------------------------------------
( z – 1 ) ( z – 0.75 )
k z = pk k z = pk
n+1 n+1
z z
= Res ---------------------------------------- + Res ----------------------------------------
( z – 1 ) ( z – 0.75 ) z=1
( z – 1 ) ( z – 0.75 ) z = 0.75 (9.99)
n+1 n+1 n+1 n+1
z
= -----------------------
z
+ ---------------- ( 0.75 ) -
1 - + -----------------------
= -----------
( z – 0.75 ) z=1
( z – 1) z = 0.75 0.25 ( – 0.25 )
n
( 0.75 ) 16 n
= 4 – ------------------------------- = 4 – ------ ( 0.75 )
( 0.25 ) ( 0.75 ) 3

c. Long Division of Polynomials

To apply this method, F ( z ) must be a rational function, and the numerator and denominator must be
polynomials arranged in descending powers of z .

Example 9.14
Use the long division method to determine f [ n ] for n = 0, 1, and 2 , given that
–1 –2 –3
1 + z + 2z + 3z
F ( z ) = --------------------------------------------------------------------------------------------- (9.100)
–1 –1 –1
( 1 – 0.25z ) ( 1 – 0.5z ) ( 1 – 0.75z )
Solution:

First, we multiply numerator and denominator by z 3 , expand the denominator to a polynomial, and
arrange the numerator and denominator polynomials in descending powers of z . Then,
3 2
z + z + 2z + 3
F ( z ) = --------------------------------------------------------------------
( z – 0.25 ) ( z – 0.5 ) ( z – 0.75 )
Next, we use the MATLAB collect function to expand the denominator to a polynomial.
syms z; den=collect((z−0.25)*(z−0.5)*(z−0.75))
den =
z^3-3/2*z^2+11/16*z-3/32
Thus,

9-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Inverse Z Transform
3 2
z + z + 2z + 3
F ( z ) = ---------------------------------------------- (9.101)
3 3 2 11 3-
z – --- z + ------ z – -----
2 16 32

Now we perform long division as shown in Figure 9.10.

5 –1 81 –2
Divisor 1 + --- z + ------ z + … Quotient
2 16
3 3 2 11 3- 3 2
z – --- z + ------ z – ----- z + z + 2z + 3 Dividend
2 16 32
3 3 2 11 3-
z – --- z + ------ z – -----
2 16 32
5
--- z 2 + 21
------ z + 35 1st Remainder
------
2 16 32
5 2 15 15 – 1
--- z – ------ z + 55
------ – ------ z
2 4 32 64
81 2nd Remainder
------ z – … + …
16

Figure 9.10. Long division for the polynomials of Example 9.14

We find that the quotient Q ( z ) is
5 – 1 81 –2
Q ( z ) = 1 + --- z + ------ z + … (9.102)
2 16
By definition of the Z transform,
∞ –n –1 –2
F(z) = ∑ f [n] z = f [ 0 ] + f [ 1 ]z + f [ 2 ]z +… (9.103)
n=0

Equating like terms in (9.102) and (9.103), we get

f [ 0 ] = 1, f [ 1 ] = 5 ⁄ 2 and f [ 2 ] = 81 ⁄ 16 (9.104)
We will use the MATLAB dimpulse function to verify the answers, and to obtain the sequence of
the first 15 values of f [ n ] .
num=[1 1 2 3]; den=[1 −3/2 11/16 −3/32]; fn=dimpulse(num,den,15),...
dimpulse(num,den,16)
fn =
1.0000
2.5000
5.0625

Signals and Systems with MATLAB Applications, Second Edition 9-37

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

8.9688
10.2070
9.6191
8.2522
6.7220
5.3115
4.1195
3.1577
2.4024
1.8189
1.3727
1.0338

Figure 9.11. Impulse response for Example 9.14

Table 9.4 lists the advantages and disadvantages of the three methods of evaluating the Inverse Z
transform.

9.7 The Transfer Function of Discrete Time Systems

The discrete time system of Figure 9.12, can be described by the linear difference equation

x[n] y[n]
Linear Discrete Time System

Figure 9.12. Block diagram for discrete time system

y [ n ] + b1 y [ n – 1 ] + b2 y [ n – 2 ] + … + bk y [ n – k ] (9.105)
= a0 x [ n ] + a1 x [ n – 1 ] + a2 x [ n – 2 ] + … + ak x [ n – k ]

9-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Transfer Function of Discrete Time Systems

TABLE 9.4 Methods of Evaluation of the Inverse Z transform

Method Advantages Disadvantages

Partial Fraction Expansion • Most familiar • Requires that F ( z ) is a
• Can use the MATLAB proper rational function
residue function
Inversion Integral • Can be used whether F ( z ) • Requires familiarity with
is a rational function or not the Residues theorem
Long Division • Practical when only a small • Requires that F ( z ) is a
sequence of numbers is proper rational function
desired.
• Division may be endless
• Useful when Inverse Z
has no closed form solution
• Can use the MATLAB
dimpulse function for
large sequence of numbers

where a i and b i are constant coefficients. In a compact form,

k k
y [n] = ∑ ai x [ n – i ] – ∑ bi y [ n – i ] (9.106)
i=0 i=0

Assuming that all initial conditions are zero, taking the Z transform of both sides of (9.106), and
using the Z transform pair
–m
[f [n – m]] ⇔ z F( z)
we get
–1 –2 –k
Y ( z ) + b1 z Y ( z ) + b2 z Y ( z ) + … + bk z Y ( z )
(9.107)
–1 –2 –k
= a0 X ( z ) + a1 z X ( z ) + a2 z X ( z ) + … + ak z X ( z )

–1 –2 –k
( 1 + b1 z + b2 z + … + b k z )Y ( z )
(9.108)
–1 –2 –k
= ( a0 + a1 z + a2 z + … + ak z ) X ( z )

–1 –2 –k
a0 + a1 z + a2 z + … + ak z
–1 –2 –k
- X(z )
Y ( z ) = -------------------------------------------------------------------------- (9.109)
1 + b1 z + b2 z + … + bk z

Signals and Systems with MATLAB Applications, Second Edition 9-39

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

We define the discrete time system transfer function H ( z ) as

–1 –2 –k
N(z) a0 + a1 z + a2 z + … + ak z
H ( z ) = ----------- = --------------------------------------------------------------------------
- (9.110)
D(z) –1
1 + b1 z + b2 z + … + bk z
–2 –k

and by substitution of (9.110) into (9.109), we get

Y ( z ) = H ( z )X ( z ) (9.111)

The discrete impulse response h [ n ] is the response to the input x [ n ] = δ [ n ] , and since
∞ –n
Z {δ[n]} = ∑ δ [ n ]z = 1
n=0

we can find the discrete time impulse response h [ n ] by taking the Inverse Z transform of the dis-
crete transfer function H ( z ) , that is,

–1
h[ n] = Z {H(z)} (9.112)

Example 9.15
The difference equation describing the input-output relationship of a discrete time system with zero
initial conditions, is
y [ n ] – 0.5y [ n – 1 ] + 0.125y [ n – 2 ] = x [ n ] + x [ n – 1 ] (9.113)
Compute:
a. The transfer function H ( z )
b. The discrete time impulse response h [ n ]
c. The response when the input is the discrete unit step u 0 [ n ]

Solution:
a. Taking the Z transform of both sides of (9.113), we get
–1 –2 –1
Y ( z ) – 0.5 z Y ( z ) + 0.125z Y ( z ) = X ( z ) + z X ( z )
and thus
–1 2
Y(z ) 1+z z +z
H ( z ) = ----------- = ------------------------------------------------ = ---------------------------------------- (9.114)
X(z ) –1
1 – 0.5 z + 0.125z
–2 2
z – 0.5z + 0.125

9-40 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Transfer Function of Discrete Time Systems

b. To obtain the discrete time impulse response h [ n ] , we need to compute the Inverse Z transform
of (9.114). We first divide both sides by z and we get:
H(z) z+1
----------- = ---------------------------------------- (9.115)
z 2
z – 0.5z + 0.125
Using the MATLAB residue function, we obtain the residues and the poles of (9.115) as follows:

num=[0 1 1]; den=[1 −0.5 0.125]; [num,den]=residue(num,den);fprintf(' \n');...

disp('r1 = '); disp(num(1)); disp('p1 = '); disp(den(1));...
disp('r2 = '); disp(num(2)); disp('p2 = '); disp(den(2))
r1 =
0.5000 - 2.5000i
p1 =
0.2500 + 0.2500i
r2 =
0.5000 + 2.5000i
p2 =
0.2500 - 0.2500i
and thus,
H ( z ) = ------------------------------------
----------- 0.5 + j2.5 -
0.5 – j2.5 - + ------------------------------------
z z – 0.25 – j0.25 z – 0.25 + j0.25
or
( 0.5 – j2.5 )z ( 0.5 + j2.5 )z ( 0.5 – j2.5 )z ( 0.5 + j2.5 )z
H ( z ) = ------------------------------------------ + ------------------------------------------ = --------------------------------------- + ---------------------------------------- (9.116)
z – ( 0.25 + j0.25 ) z – ( 0.25 – j0.25 ) z – 0.25 2e
j45°
z – 0.25 2e
– j 45°

Recalling that
n z
a u 0 [ n ] ⇔ -----------
z–a

for z > a , the discrete impulse response sequence h [ n ] is

j45° n – j 45° n
h [ n ] = ( 0.5 – j2.5 ) ( 0.25 2e ) + ( 0.5 + j2.5 ) ( 0.25 2e )
n jn45° n – j n45°
= 0.5 [ ( 0.25 2 ) e ] + 0.5 [ ( 0.25 2 ) e ]
n jn45° n – j n45°
– j2.5 [ ( 0.25 2 ) e ] + j2.5 [ ( 0.25 2 ) e ]
n jn45° – j n45° n jn45° – j n45°
= 0.5 [ ( 0.25 2 ) ( e +e ) ] – j2.5 ( 0.25 2 ) ( e –e )
or

Signals and Systems with MATLAB Applications, Second Edition 9-41

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

2 n
h [ n ] = ⎛ ------- ⎞ ( cos n45° + 5 sin n45° ) (9.117)
⎝ 4 ⎠

z
c. From Y ( z ) = H ( z )X ( z ) , the transform u 0 [ n ] ⇔ ----------- , and using the result of part (a) we get:
z–1
2 2
z +z
Y ( z ) = ---------------------------------------- ⋅
z z(z + z)
----------- = -------------------------------------------------------------
-
z – 0.5z + 0.125 z – 1
2 2
( z – 0.5z + 0.125 ) ( z – 1 )
or
2
Y ( z ) = -------------------------------------------------------------
---------- (z + z) - (9.118)
z 2
( z – 0.5z + 0.125 ) ( z – 1 )
We will use the MATLAB residue function to compute the residues and poles of expression
(9.117). First, we must express the denominator as a polynomial.
syms z; denom=(z^2−0.5*z+0.125)*(z−1); collect(denom)
ans =
z^3-3/2*z^2+5/8*z-1/8
Then,
2
Y( z) z +z
---------- = ---------------------------------------------------------------------- (9.119)
z z – ( 3 ⁄ 2 )z + ( 5 ⁄ 8 )z – 1 ⁄ 8
3 2

Now, we compute the residues and poles.

num=[0 1 1 0]; den=[1 −3/2 5/8 −1/8]; [num,den]=residue(num,den); fprintf(' \n');...
disp('r1 = '); disp(num(1)); disp('p1 = '); disp(den(1));...
disp('r2 = '); disp(num(2)); disp('p2 = '); disp(den(2));...
disp('r3 = '); disp(num(3)); disp('p3 = '); disp(den(3))
r1 =
3.2000
p1 =
1.0000
r2 =
-1.1000 + 0.3000i
p2 =
0.2500 + 0.2500i
r3 =
-1.1000 - 0.3000i

9-42 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 State Equations for Discrete Time Systems

p3 =
0.2500 - 0.2500i
With these values, we express (9.119) as
2
Y ( z ) = ----------------------------------------------------------------------
---------- z +z 3.2- + -------------------------------
= ---------- – 1.1 – j 0.3 -
– 1.1 + j0.3 - + ---------------------------------- (9.120)
z z 3 – ( 3 ⁄ 2 )z 2 + ( 5 ⁄ 8 )z – 1 ⁄ 8 z – 1 z – 0.25 – j 0.25 z – 0.25 + j0.25

or
3.2z ( – 1.1 + j0.3 )z ( – 1.1 – j 0.3 )z
Y ( z ) = ----------- + ----------------------------------- + -----------------------------------
z – 1 z – 0.25 – j 0.25 z – 0.25 + j0.25
(9.121)
( – 1.1 + j0.3 )z- + ---------------------------------------
3.2z- + --------------------------------------
= ---------- ( – 1.1 – j 0.3 )z -
z – 1 z – 0.25 2e j45° z – 0.25 2e – j 45°

Recalling that
n z
a u 0 [ n ] ⇔ -----------
z–a

for z > a , we find that the discrete output response sequence is

j45° n – j 45° n
y [ n ] = 3.2 + ( – 1.1 + j0.3 ) ( 0.25 2e ) – ( 1.1 + j0.3 ) ( 0.25 2e )
n jn45° – j n45° n jn45° – j n45°
= 3.2 – 1.1 [ ( 0.25 2 ) ( e +e ) ] + j0.3 [ ( 0.25 2 ) ( e –e )]
or
n n
y [ n ] = 3.2 – 2.2 ⎛ ------- ⎞ cos n45° – 0.6 ⎛ ------- ⎞ sin n45°
2 2
⎝ 4 ⎠ ⎝ 4 ⎠
(9.122)
n
= 3.2 – ⎛ ------- ⎞ ( 2.2 cos n45° + 0.6 sin n45° )
2
⎝ 4 ⎠

The plots for the discrete time sequences h [ n ] and y [ n ] are shown in Figure 9.13.

9.8 State Equations for Discrete Time Systems

As with continuous time systems, we select state variables for discrete time systems, either from block
diagrams that relate the input-output information, or directly from a difference equation.
Consider the block diagram of Figure 9.14. We learned in Chapter 5 that the state equations repre-
senting this continuous time system are
x· = Ax + bu
(9.123)
y = Cx + du

Signals and Systems with MATLAB Applications, Second Edition 9-43

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

In a discrete time block diagram, the integrator is replaced by a delay device. The analogy between an
integrator and a unit delay device is shown in Figure 9.15.

n h[n] y[n]
0.000 1.000 1.000
1.000 1.500 2.500 h[n] and y[n] for Example 9.15
2.000 0.625 3.125 3.5
3.000 0.125 3.250
3.0
4.000 -0.02 3.234
5.000 -0.02 3.211 2.5
6.000 -0.010 3.201 2.0
7.000 -0 3.199 1.5
8.000 0.000 3.199
1.0
9.000 0.000 3.200
10.000 0.000 3.200 0.5
11.000 3E-05 3.200 0.0
12.000 -0 3.200 -0.5
13.000 -0 3.200 1 2 3 4 5 6 7 8 9 10 11
14.000 -0 3.200

Figure 9.13. Plots of h [ n ] and y [ n ] for Example 9.15

x· x
Σ Σ
+ +
u b ∫ dt C y

+ +

Figure 9.14. Block diagram for a continuous time system

Example 9.16
The input-output relation for a discrete time system is
y [ n + 3 ] + 2y [ n + 2 ] + 5y [ n + 1 ] + y [ n ] = u [ n ] (9.124)
where u [ n ] is any input, and y [ n ] is the output. Write the discrete time state equations for this sys-
tem.
Solution:
We choose our state variables as the output and the output advanced by one and by two time steps.
Thus, we choose the discrete state variables as

9-44 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 State Equations for Discrete Time Systems

x· ( t ) x(t) x[n + 1] x[ n]
∫ dt Delay

Continuous Time Discrete Time

L [ x· ( t ) ] L [x(t)] Z{x[n + 1]} Z{x[n]}

–1
1⁄s z

s-domain z-domain

Figure 9.15. Analogy between integration and delay devices

x1 [ n ] = y [ n ] x2 [ n ] = y [ n + 1 ] x3 [ n ] = y [ n + 2 ] (9.125)
Then,
x3 [ n + 1 ] = y [ n + 3 ]

x2 [ n + 1 ] = y [ n + 2 ] = x3 [ n ]

x1 [ n + 1 ] = y [ n + 1 ] = x2 [ n ]

Thus, the state equations are

x1 [ n + 1 ] = x2 [ n ]
x2 [ n + 1 ] = x3 [ n ]
x 3 [ n + 1 ] = – 2x 3 [ n ] – 5x 2 [ n ] – x 1 [ n ] = u [ n ]
and in matrix form,

x1 [ n + 1 ] 0 1 0 x1 [ n ] 0
x2 [ n + 1 ] = 0 0 1 ⋅ x2 [ n ] + 0 u[n] (9.126)
x3 [ n + 1 ] –1 –5 –2 x3 [ n ] 1

The general form of the solution is

n–1
n–1–i
∑A
n
x[ n] = A x[0] + b [ i ]u [ i ] (9.127)
i=0

The discrete time state equations are written in a more compact form as
x [ n + 1 ] = Ax [ n ] + bu [ n ]
(9.128)
y [ n ] = Cx [ n ] + du [ n ]

Signals and Systems with MATLAB Applications, Second Edition 9-45

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

We can use the MATLAB c2d function to convert the continuous time state space equation
x· ( t ) = Ax ( t ) + bu ( t ) (9.129)
to the discrete time state space equation
x [ n + 1 ] = A disc x [ n ] + b disc u [ n ] (9.130)

where the subscript disc stands for discrete, n indicates the present sample, and n + 1 the next.

Example 9.17
Use the MATLAB c2d function to convert the continuous time state space equation
x· ( t ) = Ax ( t ) + bu ( t )
where

A = 0 1 and b = 0 (9.131)
–3 –4 1

to discrete time state space equation with sampling period T s = 0.1 s .

Solution:
Adisc=[0 1; −3 −4]; bdisc=[0 1]'; [Adisc,bdisc]=c2d(Adisc,bdisc,0.1)
Adisc =
0.9868 0.0820
-0.2460 0.6588
bdisc =
0.0044
0.0820
and therefore, the equivalent discrete time state space equation is

x1 [ n + 1 ] 0.9868 0.0820 ⋅ x 1 [ n ] + 0.0044 u [ n ]

= (9.132)
x2 [ n + 1 ] – 0.2460 0.6588 x2 [ n ] 0.0820

The MATLAB d2c function converts the discrete-time state equation

x [ n + 1 ] = A disc x [ n ] + b disc u [ n ]

to the continuous time state equation

x· ( t ) = Ax ( t ) + bu ( t )

We can invoke the MATLAB command help d2c to obtain a detailed description of this function.

9-46 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

9.9 Summary
• The Z transform performs the transformation from the domain of discrete time signals, to
another domain which we call z – domain . It is used with discrete time signals, the same way the
Laplace and Fourier transforms are used with continuous time signals.
• The one-sided Z transform F ( z ) of a discrete time function f [ n ] defined as
∞ –n
F( z) = ∑ f [ n ]z
n=0

and it is denoted as
F(z) = Z {f [n]}
• The Inverse Z transform is defined as

1 k–1
j2π °∫
f [ n ] = -------- F ( z )z dz

and it is denoted as
–1
f [n] = Z {F(z)}

• The linearity property of the Z transform states that

af 1 [ n ] + bf 2 [ n ] + cf 3 [ n ] + … ⇔ aF 1 ( z ) + bF 2 ( z ) + cF 3 ( z ) + …

• The shifting of f [ n ]u 0 [ n ] where u 0 [ n ] is the discrete unit step function, produces the Z trans-
form pair
–m
f [ n – m ]u 0 [ n – m ] ⇔ z F(z )

• The right shifting of f [ n ] allows use of non-zero values for n < 0 and produces the Z transform
pair
m–1
–m –n
f [n – m] ⇔ z F(z) + ∑ f [ n – m ]z
n=0

For m = 1 , this transform pair reduces to

–1
f [ n – 1 ] ⇔ z F ( z ) + f [ –1 ]
and for m = 2 , reduces to
–2 –1
f [ n – 2 ] ⇔ z F ( z ) + f [ –2 ] + z f [ –1 ]

Signals and Systems with MATLAB Applications, Second Edition 9-47

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

• The mth left shifting of f [ n ] where m is a positive integer, produces the Z transform pair
–1
–n
∑
m
f [n + m] ⇔ z F(z) + f [ n + m ]z
n = –m

For m = 1 , the above expression reduces to

Z { f [ n + 1 ] } = zF ( z ) – f [ 0 ]z
and for m = 2 , reduces to
2 2
Z { f [ n + 2 ] } = z F ( z ) – f [ 0 ]z – f [ 1 ]z
n
• Multiplication by a produces the Z transform pair

a f [ n ] ⇔ F ⎛ --- ⎞
n z
⎝a⎠
– naT
• Multiplication by e produces the Z transform pair
– naT aT
e f [n] ⇔ F(e z)

• Multiplications by n and n 2 produce the Z transform pairs respectively

d
nf [ n ] ⇔ – z ----- F ( z )
dz
2
2 d 2d
n f [ n ] ⇔ z ----- F ( z ) + z -------2- F ( z )
dz dz
• The summation property of the Z transform states that
n
f [ m ] ⇔ ⎛ ----------- ⎞ F ( z )
z
∑
m=0
⎝z–1 ⎠

• Convolution in the discrete time domain corresponds to multiplication in the z -domain, that is,

f 1 [ n ]∗ f 2 [ n ] ⇔ F 1 ( z ) ⋅ F 2 ( z )

• Multiplication in the discrete time domain corresponds to convolution in the z -domain, that is,

1 ⎛ ⎞ z –1
f 1 [ n ] ⋅ f 2 [ n ] ⇔ --------
j 2π °∫ xF ( v )F ⎝ -v-⎠ v
1 2 dv

• The initial value theorem of the Z transform states that

f [ 0 ] = lim X ( z )
z→∞

9-48 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• The final value theorem of the Z transform states that

lim f [ n ] = lim ( z – 1 ) F ( z )
n→∞ z→1

• The Z transform of the geometric sequence

⎧0 n = – 1, – 2, – 3, …
f [n] = ⎨ n
⎩a n = 0, 1, 2, 3, …
is
⎧ z
∞ n –n ⎪ ----------- for z > a
∑a z
n
Z [a ] = = ⎨ z–a
n=0 ⎪ unbounded for z < a
⎩

• The Z transform of the discrete unit step function u 0 [ n ] shown below

u0 [ n ]
⎧0 n<0 1
u0 [ n ] = ⎨ ....
⎩1 n≥0
0
n
is
⎧ z -
∞ –n ⎪ ---------- for z > 1
Z [ u0 [ n ] ] = ∑ [ 1 ]z = ⎨ z – 1
⎪ unbounded for z < 1
n=0
⎩
• The Z transform of the discrete exponential sequence
– naT
f [n] = e
is
z
Z [ e–naT ] 1
= -------------------------- = -----------------
– aT – 1 – aT
-
1–e z z – e
– aT – 1
for e z <1

• The Z transforms of the discrete time functions f 1 [ n ] = cos n aT and f 2 [ n ] = sin naT are respec-
tively
z 2 – z cos aT
cos naT ⇔ ----------------------------------------- for z > 1
z 2 – 2z cos aT + 1

z sin a T
sin n aT ⇔ -----------------------------------------
2
for z > 1
z – 2z cos aT + 1

Signals and Systems with MATLAB Applications, Second Edition 9-49

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

• The Z transform of the discrete unit ramp f [ n ] = nu 0 [ n ] is

z
nu 0 [ n ] ⇔ -----------------2-
(z – 1)
• The Z transform can also be found by means of the contour integral

1 F(v)
F∗ ( s ) = --------
j2π °∫C -------------------------- dv
1–e e
– sT vT

and the residue theorem.

• The variables s and z are related as
sT
z = e
and
1
s = --- ln z
T
• The relation
F(z) = G(s) 1
s = --- ln z
T

allows the mapping (transformation) of regions of s -plane to z -plane.

• The Inverse Z transform can be found by partial fraction expansion, the inversion integral, and
long division of polynomials.
• The discrete time system transfer function H ( z ) is defined as
–1 –2 –k
0 + a1 z + a2 z + … + ak z
N ( z ) = a--------------------------------------------------------------------------
H ( z ) = ----------- -
D(z) –1
1 + b1 z + b2 z + … + bk z
–2 –k

• The input X ( z ) and output Y ( z ) are related by the system transfer function H ( z ) as

Y ( z ) = H ( z )X ( z )

• The discrete time impulse response h [ n ] and the discrete transfer function H ( z ) are related as
–1
h[ n] = Z {H(z)}

• The discrete time state equations are

x [ n + 1 ] = Ax [ n ] + bu [ n ]
y [ n ] = Cx [ n ] + du [ n ]
and the general form of the solution is

9-50 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary
n–1
n–1–i
∑A
n
x[ n] = A x[0] + b [ i ]u [ i ]
i=0

• The MATLAB c2d function converts the continuous time state space equation

x· ( t ) = Ax ( t ) + bu ( t )
to the discrete time state space equation
x [ n + 1 ] = A disc x [ n ] + b disc u [ n ]

• The MATLAB d2c function converts the discrete-time state equation

x [ n + 1 ] = A disc x [ n ] + b disc u [ n ]

to the continuous time state equation

x· ( t ) = Ax ( t ) + bu ( t )

Signals and Systems with MATLAB Applications, Second Edition 9-51

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

9.10 Exercises
1. Find the Z transform of the discrete time pulse p [ n ] defined as

⎧1 n = 0, 1, 2, …, m – 1
p[ n] = ⎨
⎩0 otherwise

n
2. Find the Z transform of a p [ n ] where p [ n ] is defined as in Exercise 1.
3. Prove the following Z transform pairs:
a. δ [ n ] ⇔ 1
–m
b. δ [ n – 1 ] ⇔ z
n az
c. na u 0 [ n ] ⇔ -----------------2-
(z – a)

2 n az ( z + a )
d. n a u 0 [ n ] ⇔ ---------------------
3
-
(z – a)
2
z
e. [ n + 1 ]u 0 [ n ] ⇔ -----------------2-
(z – 1)
–1
4. Use the partial fraction expansion to find f [ n ] = Z [ F ( z ) ] given that

A
F ( z ) = -------------------------------------------------
-
–1 –1
( 1 – z ) ( 1 – 0.5z )
5. Use the partial fraction expansion method to compute the Inverse Z transform of
2
z
F ( z ) = -------------------------------------------
2
( z + 1 ) ( z – 0.75 )
6. Use the Inversion Integral to compute the Inverse Z transform of
–1 –3
1 + 2z + z
F ( z ) = --------------------------------------------------
–1 –1
( 1 – z ) ( 1 – 0.5z )
7. Use the long division method to compute the first 5 terms of the discrete time sequence whose Z
transform is
–1 –2 –3
z +z –z
F ( z ) = ----------------------------------------------
–1 –2 –3
1 + z + z + 4z

9-52 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

8. a. Compute the transfer function of the difference equation

y [ n ] – y [ n – 1 ] = Tx [ n – 1 ]
– naT
b. Compute the response y [ n ] when the input is x [ n ] = e
9. Given the difference equation
T
y [ n ] – y [ n – 1 ] = --- { x [ n ] + x [ n – 1 ] }
2

a. Compute the discrete transfer function H ( z )

– naT
b. Compute the response to the input x [ n ] = e
10. A discrete time system is described by the difference equation
y[n] + y[n – 1] = x[n]
where
y [ n ] = 0 for n < 0

a. Compute the transfer function H ( z )

b. Compute the impulse response h [ n ]
c. Compute the response when the input is x [ n ] = 10 for n ≥ 0
11. Given the discrete transfer function
z+2
H ( z ) = ----------------------------
2
8z – 2z – 3

write the difference equation that relates the output y [ n ] to the input x [ n ] .

Signals and Systems with MATLAB Applications, Second Edition 9-53

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

9.11 Solutions to Exercises

1.
⎧1 n = 0, 1, 2, …, m – 1 1 p [ n ] = u0 [ n ] – u0 [ n – m ]
p[n] = ⎨
⎩0 z
otherwise u 0 [ n ] ⇔ -----------
z–1
0 1 2 3 m–1 –m z
u 0 [ n – m ] ⇔ z -----------
z–1

By the linearity property

–m –m
z –m z z(1 – z ) 1–z
Z{p[ n] } = ----------- – z ----------- = ------------------------ = -----------------
z–1 z–1 z–1 1–z
–1

2.
a f [ n ] ⇔ F ⎛ --- ⎞
n z
⎝a⎠
and from Exercise 1,
–m
1–z -
p [ n ] ⇔ ----------------
–1
1–z
Then,
–m –m –m –m –1 –m –m m –m –m m –m
n 1 – (z ⁄ a) (a – z ) ⁄ a a (a – z ) a (a – z ) 1–a z
a p [ n ] ⇔ ---------------------------- = ---------------------------------------- = ----------------------------------- = ---------------------------------- = -----------------------
–1 –1 –1 –1 –m –1 –1 –1 –1 –1
1 – (z ⁄ a) (a – z ) ⁄ a a (a – z ) a(a – z ) 1 – az
or
m –m
n z(1 – a z )
a p [ n ] ⇔ -------------------------------
z–a
3.
a.
∞
–n –0
Z{δ[n]} = ∑ δ [ n ]z = δ [ 0 ]z = 1
n=0

b.
∞
–n
Z{δ[n – m]} = ∑ δ [ n – m ]z
n=0

and since δ [ n – m ] is zero for all n except n = m , it follows that

9-54 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
∞
–m –m
Z{δ[n – m]} = ∑ δ [ 0 ]z = z
n=0

c.
From Example 9.1,
n z
f [ n ] = a u 0 [ n ] ⇔ F ( z ) = ----------- (1)
z–a

Differentiating (1) with respect to z and multiplying by – z we get

d z–a–z az
– z ----- F ( z ) = – z ------------------2- = ------------------ (2)
dz (z – a) (z – a)
2

Also, from the multiplication by n property

n d
nf [ n ] = n ( a u 0 [ n ] ) ⇔ – z ----- F ( z ) (3)
dz
and from (2) and (3)
n az
n ( a u 0 [ n ] ) ⇔ -----------------2- (4)
(z – a)

we observe that for a = 1 (4) above reduces to

z
nu 0 [ n ] ⇔ -----------------2-
(z – 1)
d.
From Example 9.1,
n z
f [ n ] = a u 0 [ n ] ⇔ F ( z ) = ----------- (1)
z–a

and taking the second derivative of (1) with respect to z we get

2 2
-------- F ( z ) = -------- ⎛ ----------- ⎞ = ----- ----- ⎛ ----------- ⎞
d d z d d z d –a 2a ( z – a ) 2a
2⎝ z – a ⎠
= ----- -----------------2- = ---------------------- = ------------------ (2)
2 dz dz ⎝z – a⎠ dz ( z – a ) 4 3
dz dz (z – a) (z – a)
2
Also, from the multiplication by n property
2
2 2 n d 2d
n f [ n ] = n ( a u 0 [ n ] ) ⇔ z ----- F ( z ) + z -------2- F ( z ) (3)
dz dz
From Exercise 9.3(c), relation (2)

Signals and Systems with MATLAB Applications, Second Edition 9-55

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

d –a z
z ----- F ( z ) = ------------------ (4)
dz (z – a)
2

and by substitution of (2) and (4) into (3) we get

2 2 2 2
2 n – a z - + z 2 -----------------
n ( a u 0 [ n ] ) ⇔ -----------------
2a -
= 2az – az ( z – a -) = 2az
-------------------------------------- --------------------------------------- az ( z + a -)
– az + a z- = ---------------------
3
(z – a)
2
(z – a) (z – a)
3
(z – a)
3
(z – a)
3

We observe that for a = 1 the above reduces to

2 z(z + 1)
n u 0 [ n ] ⇔ ------------------3-
(z – 1)
e.
Let f [ n ] = u 0 [ n ] and we know that
z
u 0 [ n ] ⇔ ----------- (1)
z–1

The term ( n + 1 )u 0 [ n ] represents the sum of the first n values, including n = 0 , of u 0 [ n ] and
thus it can be written as the summation
n
g [ n ] = ( n + 1 )u 0 [ n ] = ∑ u0 [ k ]
k=0

Since summation in the discrete time domain corresponds to integration in the continuous time
domain, it follows that
u 1 [ n ] = ( n + 1 )u 0 [ n ]

where u 1 [ n ] represents the discrete unit ramp. Now, from the summation in time property,
n

∑ f [ k ] ⇔ ⎛⎝ ----------
- ⎞ F(z)
z
z–1⎠
k=0

and with (1) above

2
G ( z ) = ⎛ ----------- ⎞ F ( z ) = ----------- ⋅ ----------- = -----------------
z z z z -
⎝z–1 ⎠ z–1 z–1 2
(z – 1)
and thus
2
z
[ n + 1 ]u 0 [ n ] ⇔ -----------------2-
(z – 1)

9-56 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

4.
2
We first multiply the numerator and denominator by z to eliminate the negative exponents of z .
2
A Az
F ( z ) = -------------------------------------------------- = -------------------------------------
–1
( 1 – z ) ( 1 – 0.5z )
–1 ( z – 1 ) ( z – 0.5 )
or
F ( z )- = ------------------------------------
---------- Az r1
- = ----------
r2
- + ---------------
z ( z – 1 ) ( z – 0.5 ) z – 1 z – 0.5

Az Az
r 1 = --------------- = 2A r 2 = ----------- = –A
z – 0.5 z=1
z–1 z = 0.5

F(z) 2A A
----------- = ----------- – ---------------
z z – 1 z – 0.5
or
2Az Az
F ( z ) = ----------- – ---------------
z – 1 z – 0.5
Since
z z
----------- ⇔ 1 ----------- ⇔ a n
z–1 z–a
it follows that
1 n 1 n
[ F ( z ) ] = f [ n ] = 2A – A ⎛ --- ⎞ = A 2 – ⎛ --- ⎞
–1
Z ⎝ 2⎠ ⎝ 2⎠

5.
F ( z )- = ----------
----------
r1 r2
- + ------------------
r3
+ ------------------------- z - (1)
- = ------------------------------------------
z z + 1 z – 0.75 ( z – 0.75 ) 2
( z + 1 ) ( z – 0.75 )
2

and clearing of fractions yields

2
r 1 ( z – 0.75 ) + r 2 ( z + 1 ) ( z – 0.75 ) + r 3 ( z + 1 ) = z (2)

With z = 0.75 (2) reduces to 1.75r3 = 0.75 from which r 3 = 3 ⁄ 7

2
With z = – 1 (2) reduces to ( – 1.75 ) r 1 = – 1 from which r 1 = – 16 ⁄ 49
2
With z = 0 (2) reduces to ( – 0.75 ) r 1 – 0.75r 2 + r 3 = 0
2
or ( 3 ⁄ 4 ) × ( – 16 ⁄ 49 ) – ( 3 ⁄ 4 )r 2 + 3 ⁄ 7 = 0 from which r 2 = 16 ⁄ 49

By substitution into (1) and multiplication by z we get

Signals and Systems with MATLAB Applications, Second Edition 9-57

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

( – 16 ⁄ 49 )z ( 16 ⁄ 49 )z ( 4 ⁄ 7 ) ⋅ ( 0.75z )
F ( z ) = -------------------------- + ----------------------- + -------------------------------------
z – ( –1 ) z – 0.75 ( z – 0.75 )
2

z n z az n
Using the transforms u 0 [ n ] ⇔ ----------- , a u 0 [ n ] = ----------
- , and na u 0 [ n ] = ------------------ we get
z–1 z–a 2
(z – a)

f [ n ] = ⎛ – ------ ⎞ ( – 1 ) + ⎛ ------ ⎞ ( 0.75 ) + --- n ( 0.75 )

16 n 16 n 4 n
⎝ 49 ⎠ ⎝ 49 ⎠ 7

Check with MATLAB:

syms z n; Fz=z^2/((z+1)*(z−0.75)^2); iztrans(Fz)
ans =
-16/49*(-1)^n+16/49*(3/4)^n+4/7*(3/4)^n*n
6.
3
Multiplication by z yields
3 2
z + 2z + 1
F ( z ) = ---------------------------------------
z ( z – 1 ) ( z – 0.5 )
By (9.86)
n–1
f [n] = ∑ Res [ F ( z )z ]
k z = pk

and for this exercise

3 2 n–2
( z + 2z + 1 )z
f [n] = ∑ Res ---------------------------------------------
( z – 1 ) ( z – 0.5 )
k z = pk

n–2
Next, we examine z to find out if there are any values of n for which there is a pole at the ori-
gin. We observe that for n = 0 there is a second order pole at z = 0 because
n–2 –2 1
z n=0
= z = ----
2
z

Also, for n = 1 there is a simple pole at z = 0 . But for n ≥ 2 the only poles are z = 1 and
z = 0.5 . Then, following the same procedure as in Example 9.12 we get:

For n = 0

9-58 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

3 2
( z + 2z + 1 )
f[ 0] = ∑ Res -----------------------------------------
2
z ( z – 1 ) ( z – 0.5 )
-
k z = pk

3 2 3 2
( z + 2z + 1 ) ( z + 2z + 1 )
= Res -----------------------------------------
2
- + Res -----------------------------------------
2
-
z ( z – 1 ) ( z – 0.5 ) z ( z – 1 ) ( z – 0.5 )
z=0 z=1

3 2
( z + 2z + 1 )
+ Res -----------------------------------------
2
-
z ( z – 1 ) ( z – 0.5 )
z = 0.5

The first term on the right side of the above expression has a pole of order 2 at z = 0 ; therefore,
we must evaluate the first derivative of
3 2
( z + 2z + 1 )
-------------------------------------
( z – 1 ) ( z – 0.5 )

at z = 0 . Thus, for n = 0 , it reduces to

3 2 3 2 3 2
d ( z + 2z + 1 ) ( z + 2z + 1 ) ( z + 2z + 1 )
f [ 0 ] = ----- ------------------------------------- + --------------------------------
- + --------------------------------
-
dz ( z – 1 ) ( z – 0.5 ) z=0
2
z ( z – 0.5 )
2
z (z – 1)
z=1 z = 0.5
= 6 + 8 – 13 = 1

For n = 1 , it reduces to
3 2
( z + 2z + 1 )
f[1] = ∑ Res ---------------------------------------
z ( z – 1 ) ( z – 0.5 )
k z = pk
3 2 3 2 3 2
( z + 2z + 1 ) ( z + 2z + 1 ) ( z + 2z + 1 )
= Res --------------------------------------- + Res --------------------------------------- + Res ------------------------------------------
z ( z – 1 ) ( z – 0.5 ) z=0
z ( z – 1 ) ( z – 0.5 ) z=1
z ( z – 1 ) ( z – 0.75 ) z = 0.5
or
3 2
3
( z + 2z + 1 )
2
( z + 2z + 1 )
3 2 ( z + 2z + 1 )
f [ 1 ] = ---------------------------------------- + --------------------------------- + -----------------------------------
( z – 1 ) ( z – 0.75 ) z=0
z ( z – 0.75 ) z=1
z(z – 1)
z = 0.5
= 2 + 8 – 13 ⋅ ( 0.5 ) = 3.5

For n ≥ 2 there are no poles at z = 0 , that is, the only poles are at z = 1 and z = 0.5 . There-
fore,

Signals and Systems with MATLAB Applications, Second Edition 9-59

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

3 2 n–2
( z + 2z + 1 )z
f[n] = ∑ Res ----------------------------------------------
( z – 1 ) ( z – 0.5 )
k z = pk

3 2 n–2 3 2 n–2
( z + 2z + 1 )z ( z + 2z + 1 )z
= Res ---------------------------------------------- + Res ----------------------------------------------
( z – 1 ) ( z – 0.5 ) ( z – 1 ) ( z – 0.5 )
z=1 z = 0.5

3 2 n–2 3 2 n–2
( z + 2z + 1 )z ( z + 2z + 1 )z
= ---------------------------------------------- + ----------------------------------------------
( z – 0.5 ) (z – 1)
z=1 z = 0.5

for n ≥ 2 .
We can express f [ n ] for all n ≥ 0 as
n
f [ n ] = 6δ [ n ] + 2δ [ n – 1 ] + 8 – 13 ( 0.5 )

where the coefficients of δ [ n ] and δ [ n – 1 ] are the residues that were found for n = 0 and
n = 1 at z = 0 .The coefficient 6 is multiplied by δ [ n ] to emphasize that this value exists only
for n = 0 and coefficient 2 is multiplied by δ [ n ] to emphasize that this value exists only for
n = 1.
Check with MATLAB:
syms z n; Fz=(z^3+2*z^2+1)/(z*(z−1)*(z−0.5)); iztrans(Fz)
ans =
2*charfcn[1](n)+6*charfcn[0](n)+8-13*(1/2)^n
7.
3
Multiplication of each term by z yields
–1 –2 –3 2
z +z –z
F ( z ) = ---------------------------------------------- z +z–1 -
= ---------------------------------
–1 –2 –3 3 2
1 + z + z + 4z z +z +z+1
The long division of the numerator by the denominator is shown below.

9-60 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

–1 –3 –4
Divisor z –2 z + z +… Quotient
3 2 2
z +z +z+1 z +z–1 Dividend
2 –1
z +z+1+ z
–1 1st Remainder
–2– z
–1 –2 –3
–2–2z –2z –2z
–1 –2 –3
z + 2z + 2z 2nd Remainder
–1
z +…
… … …

Therefore,
–1 –3 –4
F ( z ) = z –2 z + z + … (1)
Also,
∞
–n –1 –2 –3 –4
F(z) = ∑ f [ n ]z = f[0 ] + f[1 ] z + f[2] z + f[ 3] z + f[ 4] z + … (2)
0

Equating like terms on the right sides of (1) and (2) we get
f[0] = 0 f[1] = 1 f[2] = 0 f [ 3 ] = –2 f[4] = 1
8.
a.
y [ n ] – y [ n – 1 ] = Tx [ n – 1 ]
Taking the Z transform of both sides we get
–1 –1
Y ( z ) – z Y ( z ) = Tz X ( z )
and thus
–1
Y(z ) Tz T
H ( z ) = ----------- = ---------------- = -----------
X(z ) 1–z
–1 z – 1

b.
– naT z -
x[ n] = e ⇔ X ( z ) = -----------------
– aT
z–e
Then,
T z Tz
Y ( z ) = H ( z )X ( z ) = ----------- ⋅ -----------------
- = -------------------------------------------
-
z–1 z–e – aT
(z – 1) ⋅ (z – e )
– aT

Signals and Systems with MATLAB Applications, Second Edition 9-61

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

or
Y ( z ) = -------------------------------------------
---------- T r1
- = ----------
r2
- (1)
- + -----------------
z (z – 1) ⋅ (z – e )
– aT z – 1 z – e – aT

T T T –T
r 1 = ------------------ = ------------------- r 2 = ----------- = -------------------
z–e
– aT
1–e
– aT z–1 – aT 1–e
– aT
z=1 z=e

By substitution into (1) and multiplication by z we get

– aT – aT
Tz ⁄ ( 1 – e -) – Tz
Y ( z ) = --------------------------------- ⁄ ( 1 – e -)
---------------------------------
(z – 1) (z – e )
– aT

z z n
Recalling that ----------- ⇔ u 0 [ n ] and ----------- ⇔ a u 0 [ n ] we get
z–1 z–a
– naT
T
T Te - ( 1 – e –naT )u 0 [ n ]
y [ n ] = ------------------- – ------------------- = ------------------
– aT – aT – aT
1–e 1–e 1–e

9.
a.
T
y [ n ] – y [ n – 1 ] = --- { x [ n ] + x [ n – 1 ] }
2
Taking the Z transform of both sides we get
–1 T –1
Y ( z ) – z Y ( z ) = --- [ X ( z ) + z X ( z ) ]
2
or
–1 T –1
( 1 – z )Y ( z ) = --- ( 1 + z )X ( z )
2
and thus
–1
Y(z ) T 1+z T z+1
- = --- ⋅ -----------
H ( z ) = ----------- = --- ⋅ ---------------
X(z ) 2 1–z – 1 2 z–1

b.
– naT z
x[ n] = e ⇔ X ( z ) = ------------------
– aT
z–e
Then,

T z+1 z Tz ( z + 1 )
Y ( z ) = H ( z )X ( z ) = --- ⋅ ----------- ⋅ -----------------
- = ----------------------------------------------
-
2 z – 1 z – e – aT 2( z – 1) ⋅ (z – e )
– aT

9-62 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

or
Y(z) T(z + 1) r1 r2
---------- = ----------------------------------------------- = ---------- - (1)
- + -----------------
z 2(z – 1) ⋅ (z – e )
– aT z – 1 z – e –aT

T z+1- T 2 - T -
r 1 = --- ⋅ ----------------- = --- ⋅ ------------------ = ------------------
2 z – e – aT 2 1 – e –aT 1–e
– aT
z=1

– aT
T z+1 T e +1 T -
r 2 = --- ⋅ ----------- = --- ⋅ ------------------
- = ------------------
2 z–1 – aT
2 e – aT – 1 1–e
– aT
z=e

By substitution into (1) and multiplication by z we get

– aT – aT – aT
Tz ⁄ ( 1 – e ) [ ( Tz ⁄ 2 ) ⋅ ( e + 1)] ⁄ (1 – e )
Y ( z ) = ---------------------------------- + -------------------------------------------------------------------------------
(z – 1) (z – e )
– aT

z z n
Recalling that ----------- ⇔ u 0 [ n ] and ----------- ⇔ a u 0 [ n ] we get
z–1 z–a
– aT
+ 1- –naT
T -+T – --- coth ⎛ ------ ⎞ e
e T - T aT – naT
y [ n ] = ------------------ --- ⋅ ------------------ e = ------------------
– aT 2 – aT – aT 2 ⎝ 2⎠
1–e e –1 1–e

10.
a.
y[n] + y[n – 1] = x[n]

y [ n ] = 0 for n < 0
Taking the Z transform of both sides we get
–1
( 1 + z )Y ( z ) = X ( z )
and thus
Y ( z )- = ---------------
H ( z ) = ---------- 1 - = ---------- z -
z - = ------------------
X(z ) 1+z
–1 z + 1 z – ( –1 )

b.
–1 –1 ⎧ z ⎫ n
h [ n ] = Z { H ( z ) } = Z ⎨ ------------------- ⎬ = ( – 1 )
⎩ z – ( –1 ) ⎭
c.
x [ n ] = 10 for n ≥ 0
z
X ( z ) = 10 ⋅ -----------
z–1

Signals and Systems with MATLAB Applications, Second Edition 9-63

Orchard Publications
Chapter 9 Discrete Time Systems and the Z Transform

2
z 10z 10z
Y ( z ) = H ( z )X ( z ) = ----------- ⋅ ----------- = --------------------------------
-
z+1 z–1 ( z + 1 )( z – 1 )

Y ( z ) = --------------------------------
---------- 10z r1
- = ----------
r2 5 - + ----------
- = ----------
- + ---------- 5-
z (z + 1)(z – 1) z+1 z–1 z+1 z–1

5z - + ----------
Y ( z ) = ------------------ 5z - ⇔ f [ n ] = 5 ( – 1 ) n + 5
z – ( –1 ) z – 1

11.
z+2
H ( z ) = ----------------------------
2
8z – 2z – 3
2
Multiplication of each term by 1 ⁄ 8z yields
–1 –2
Y ( z )- = -----------------------------------------------------------
H ( z ) = ---------- 1 ⁄ 8 ⋅ ( z + 2z ) -
X(z ) 1 – (1 ⁄ 4) z – (3 ⁄ 8) z
–1 –2

1 –1 3 –2 1 –1 –2
1 – --- z – --- z ⋅ Y ( z ) = --- ⋅ ( z + 2z ) ⋅ X ( z )
4 8 8

and taking the Inverse Z transform, we get

1 3 1 1
y [ n ] – --- y [ n – 1 ] – --- y [ n – 2 ] = --- x [ n – 1 ] + --- x [ n – 2 ]
4 8 8 4

9-64 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 10
The DFT and the FFT Algorithm

his chapter begins with the actual computation of frequency spectra for discrete time systems.
T For brevity, we will use the acronyms DFT for the Discrete Fourier Transform and FFT for
Fast Fourier Transform algorithm respectively. The definition, theorems, and properties are
also discussed, and several examples are given to illustrate their uses.

10.1 The Discrete Fourier Transform (DFT)

In the Fourier series topic, Chapter 7, we learned that a periodic and continuous time function,
results in a non periodic and discrete frequency function. Next, in the Fourier transform topic,
Chapter 8, we saw that a non-periodic and continuous time function, produces a non-periodic and
continuous frequency function. In Chapter 9 we learned that the Z and Inverse Z transforms pro-
duce a periodic and continuous frequency function, since these transforms are evaluated on the unit
circle. This is because the frequency spectrum of a discrete time sequence f [ n ] is obtained from its
sT jωT
Z transform by the substitution of z = e = e as we saw from the mapping of the s -plane to
the z -plane in Chapter 9. It is continuous because there is an infinite number of points in the inter-
val 0 to 2π , although, in practice, we compute only a finite number of equally spaced points.
In this chapter we will see that a periodic and discrete time function results in a periodic and discrete
frequency function. For convenience, we summarize these facts in Table 10.1.

TABLE 10.1 Characteristics of Fourier and Z transforms

Topic Time Function Frequency Function

Fourier Series Continuous-Periodic Discrete-Non-periodic
Fourier Transform Continuous-Non-Periodic Continuous-Non-Periodic
Z transform Discrete-Non-Periodic Continuous-Periodic
Discrete Fourier Transform Discrete-Periodic Discrete-Periodic

In our subsequent discussion we will denote a discrete time signal as x [ n ] , and its discrete frequency
transform as X [ m ] .
Let us consider again the definition of the Z transform, that is,
∞ –n
F(z) = ∑ f [ n ]z (10.1)
n=0

Signals and Systems with MATLAB Applications, Second Edition 10-1

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

Its value on the unit circle of the z -plane, is

∞ – nj ωT
∑ f [ n ]e
jωT
F(e ) = (10.2)
n=0

This represents an infinite summation; it must be truncated before it can be computed. Let this trun-
cated version be represented by
mn
N–1 – j2π -------

∑ x [ n ]e
N
X[m] = (10.3)
n=0

where N represents the number of points that are equally spaced in the interval 0 to 2π on the unit
circle of the z -plane, and

ω = ⎛ ------- ⎞ m
2π
⎝ NT ⎠

for m = 0, 1, 2, …, N – 1 . We refer to (10.3) as the N – point DFT of X [ m ] .

The Inverse DFT is defined as

mn
1 N–1 j2π -------

∑
N
x [ n ] = ---- X [ m ]e (10.4)
N
m=0

for n = 0, 1, 2, …, N – 1 .
In general, the discrete frequency transform X [ m ] is complex, and thus we can express it as
X [ m ] = Re { X [ m ] } + Im { X [ m ] } (10.5)
for m = 0, 1, 2, …, N – 1
Since
mn
– j2π -------
2πmn
= cos -------------- – j sin 2πmn
N
e -------------- (10.6)
N N
we can write (10.3) as
mn
N–1 – j2π ------- N–1 N–1
∑ ∑ x [ n ] cos 2πmn ∑x [ n ] sin 2πmn
N
X [m] = x [ n ]e = -------------- – j -------------- (10.7)
N N
n=0 n=0 n=0

For n = 0 , (10.7) reduces to X [ m ] = x [ 0 ] . Then, the real part Re { X [ m ] } can be computed from

10-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Discrete Fourier Transform (DFT)

N–1
2πmn
Re { X [ m ] } = x [ 0 ] + ∑ x [ n ] cos -------------
N
- for m = 0, 1, 2, …, N – 1 (10.8)
n=1

and the imaginary part Im { X [ m ] } from

N–1
∑ x [ n ] sin -------------
Im { X [ m ] } = – 2πmn for m = 0, 1, 2, …, N – 1
- (10.9)
N
n=1

We observe that the summation in (10.8) and (10.9) is from n = 1 to n = N – 1 since x [ 0 ] appears
in (10.8).
Example 10.1
A discrete time signal is defined by the sequence
x [ 0 ] = 1, x [ 1 ] = 2, x [ 2 ] = 2, and x [ 3 ] = 1

and x [ n ] = 0 for all other n . Compute the frequency components X [ m ] .

Solution:
Since we are given four discrete values of x [ n ] , we will use a 4 -point DFT, that is, for this example,
N = 4 . Using (10.8) with n = 0, 1, 2 , and 3 we get
3

∑ x [ n ] cos -------------
Re { X [ m ] } = x [ 0 ] + 2πmn -
N
n=1
2πm ( 1 ) + 2 cos 2πm ( 2 ) + cos 2πm ( 3 )
= 1 + 2 cos ------------------ (10.10)
- ------------------- -------------------
4 4 4
------- + 2 cos mπ + cos 3mπ
= 1 + 2 cos mπ -----------
2 2

Next, for m = 0, 1, 2 , and 3 we get:

m = 0 Re { X [ 0 ] } = 1 + 2 ⋅ ( 1 ) + 2 ⋅ ( 1 ) + 1 ⋅ ( 1 ) = 6
m = 1 Re { X [ 1 ] } = 1 + 2 ⋅ ( 0 ) + 2 ⋅ ( – 1 ) + 1 ⋅ ( 0 ) = – 1
(10.11)
m = 2 Re { X [ 2 ] } = 1 + 2 ⋅ ( – 1 ) + 2 ⋅ ( 1 ) + 1 ⋅ ( – 1 ) = 0
m = 3 Re { X [ 3 ] } = 1 + 2 ⋅ ( 0 ) + 2 ⋅ ( – 1 ) + 1 ⋅ ( 0 ) = – 1
Now, we compute the imaginary components using (10.9). For this example,
3

∑ x [ n ] sin -------------
Im { X [ m ] } = – 2πmn = – 2 sin mπ – 2 sin mπ – sin 3mπ
- ------- -----------
N 2 2
n=1

Signals and Systems with MATLAB Applications, Second Edition 10-3

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

and for m = 0, 1, 2, 3 , we get:

m = 0 Im { X [ 0 ] } = – 2 ⋅ ( 0 ) – 2 ⋅ ( 0 ) – 1 ⋅ ( 0 ) = 0
m = 1 Im { X [ 1 ] } = – 2 ⋅ ( 1 ) – 2 ⋅ ( 0 ) – 1 ⋅ ( – 1 ) = – 1
(10.12)
m = 2 Im { X [ 2 ] } = – 2 ⋅ ( 0 ) – 2 ⋅ ( 0 ) – 1 ⋅ ( 0 ) = 0
m = 3 Im { X [ 3 ] } = – 2 ⋅ ( – 1 ) – 2 ⋅ ( 0 ) – 1 ⋅ ( 1 ) = 1

The discrete frequency components X [ m ] for m = 0, 1, 2 , and 3 are found by addition of the real
and imaginary components X re [ i ] of (10.11) and X im [ i ] of (10.12). Thus,

X [ 0 ] = 6 + j0 = 6
X [1] = – 1 – j
(10.13)
X [ 2 ] = 0 + j0 = 0
X [3] = – 1 + j

Example 10.2
Use the Inverse DFT, i.e., relation (10.4), and the results of Example 10.1, to compute the values of
the discrete time sequence x [ n ] .
Solution:
Since we are given four discrete values of x [ n ] , we will use a 4 -point DFT, that is, for this example,
N = 4 . Then, (10.4) for m = 0, 1, 2 , and 3 reduces to
mn mn
3 j2π ------- 3 jπ -------
1 1
∑ X [ m ]e ∑ X [ m ]e
4 2
x [ n ] = --- = ---
4 4
m=0 m=0
(10.14)
n 3n
jπ --- jπ ------
1 2 jπn 2
= --- X [ 0 ] + X [ 1 ]e + X [ 2 ]e + X [ 3 ]e
4

The discrete frequency components x [ n ] for n = 0, 1, 2 , and 3 are:

1
x [ 0 ] = --- { X [ 0 ] + X [ 1 ] + X [ 2 ] + X [ 3 ] }
4
1
= --- { 6 + ( – 1 – j ) + 0 + ( – 1 + j ) }= 1
4

1
x [ 1 ] = --- { X [ 0 ] + X [ 1 ] ⋅ j + X [ 2 ] ⋅ ( – 1 ) + X [ 3 ] ⋅ ( – j ) }
4
1
= --- { 6 + ( – 1 – j ) ⋅ j + 0 ⋅ ( – 1 ) + ( – 1 + j ) ⋅ ( – j ) } = 2
4

10-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Discrete Fourier Transform (DFT)

1
x [ 2 ] = --- { X [ 0 ] + X [ 1 ] ⋅ ( – 1 ) + X [ 2 ] ⋅ 1 + X [ 3 ] ⋅ ( – 1 ) }
4
1
= --- { 6 + ( – 1 – j ) ⋅ ( – 1 ) + 0 ⋅ 1 + ( – 1 + j ) ⋅ ( – 1 ) } = 2
4

1
x [ 3 ] = --- { X [ 0 ] + X [ 1 ] ⋅ ( – j ) + X [ 2 ] ⋅ ( – 1 ) + X [ 3 ] ⋅ j }
4
1
= --- { 6 + ( – 1 – j ) ⋅ ( – j ) + 0 ⋅ ( – 1 ) + ( – 1 + j ) ⋅ j } = 1
4
We observe that these are the same values as in Example 10.1. We will check the answers of Exam-
ples 10.1 and 10.2 with MATLAB and Excel.
With MATLAB, we use the fft(x) function to compute the DFT, and the ifft(x) function to compute
the Inverse DFT.
xn=[1 2 2 1]; % The discrete time sequence of Example 10.1
Xm=fft(xn) % Compute the FFT of this discrete time sequence
Xm =
6.0000 -1.0000-1.0000i 0 -1.0000+1.0000i
Xm = [6 −1−j 0 −1+j]; % The discrete frequency components of Example 10.2
xn=ifft(Xm) % Compute the Inverse FFT
xn =
1.0000 2.0000+0.0000i 2.0000 1.0000-0.0000i
To use Excel for the computation of the DFT, the Analysis ToolPak must have been installed. If not,
we can install it by clicking Add-Ins on the Tools menu, and follow the instructions on the screen.
With Excel’s Fourier Analysis tool, we get the spreadsheet shown in Figure 10.1. The instructions on
how to use it, are given on the spreadsheet.
The term e –( j2π ) ⁄ N is a rotating vector, where the range 0 ≤ θ ≤ 2π is divided into 360 ⁄ N equal seg-
ments. Therefore, it is convenient to represent it as W N , that is, we let

j2π
– --------
WN = e N (10.15)

and consequently
j2π
–1 --------
WN = e N (10.16)

Signals and Systems with MATLAB Applications, Second Edition 10-5

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

A B C D E
1 Input data x(n) are same as in Example 10.1
2 and are entered in cells A11 through A14
3
4 From the Tools drop down menu, we select
5 Data Analysis and from it, Fourier Analysis
6
7 The Input Range is A11 through A14 (A11:A14)
8 and the Output Range is B11 through B14 (B11:B14)
9
10 x(n) X(m)
11 1 6
12 2 -1-i
13 2 0
14 1 -1+i
15
16 To obtain the discrete time sequence, we select
17 Inverse from the Fourier Analysis menu
18
19 Input data are the same as in Example 10.2
20
21 The Input Range is A25 through A28 (A25:A28)
22 and the Output Range is B25 through B28 (B25:B28)
23
24 X(m) x(n)
25 6 1
26 -1-j 2
27 0 2
28 -1+j 1

Figure 10.1. Using Excel to find the DFT and Inverse DFT

Henceforth, the DFT pair will be denoted as

N–1
∑ x ( n )WN
mn
X[m] = (10.17)
n=0

and

1N–1 –m n
x [ n ] = ----
N ∑
X ( m )W N (10.18)
n=0

10-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Discrete Fourier Transform (DFT)

Also, the correspondence between x [ n ] and X [ m ] will be denoted as

x [n] ⇔ X [m] (10.19)
In Example 10.1, we found that, although the discrete time sequence x [ n ] is real, the discrete fre-
quency sequence X [ m ] is complex. However, in most applications we are interested in X [ m ] , that
is, the magnitude of X [ m ] .

Example 10.3
Use MATLAB to compute the magnitude of the frequency components of the following discrete
time function. Then, use Excel to display the time and frequency values.

x[0] x[1] x[2] x[3] x[4] x[5] x[6] x[7] x[8] x[9] x[10] x[11] x[12] x[13] x[14] x[15]
1.0 1.5 2.0 2.3 2.7 3.0 3.4 4.1 4.7 4.2 3.8 3.6 3.2 2.9 2.5 1.8

Solution:
We compute the magnitude of the frequency spectrum with the MATLAB code below.
xn=[1 1.5 2 2.3 2.7 3 3.4 4.1 4.7 4.2 3.8 3.6 3.2 2.9 2.5 1.8]; magXm=abs(fft(xn));...
fprintf(' \n'); fprintf('magXm1 = %4.2f \t', magXm(1)); fprintf('magXm2 = %4.2f \t', magXm(2));...
fprintf('magXm3 = %4.2f \t', magXm(3)); fprintf(' \n'); fprintf('magXm4 = %4.2f \t', magXm(4));...
fprintf('magXm5 = %4.2f \t', magXm(5)); fprintf('magXm6 = %4.2f \t', magXm(6)); fprintf(' \n');...
fprintf('magXm7 = %4.2f \t', magXm(7)); fprintf('magXm8 = %4.2f \t', magXm(8));...
fprintf('magXm9 = %4.2f \t', magXm(9)); fprintf(' \n');...
fprintf('magXm10 = %4.2f \t', magXm(10)); fprintf('magXm11 = %4.2f \t', magXm(11)); ...
fprintf('magXm12 = %4.2f \t', magXm(12)); fprintf(' \n');...
fprintf('magXm13 = %4.2f \t', magXm(13)); fprintf('magXm14 = %4.2f \t', magXm(14));...
fprintf('magXm15 = %4.2f \t', magXm(15))
magXm1 = 46.70 magXm2 = 11.03 magXm3 = 0.42
magXm4 = 2.41 magXm5 = 0.22 magXm6 = 1.19
magXm7 = 0.07 magXm8 = 0.47 magXm9 = 0.10
magXm10 = 0.47 magXm11 = 0.07 magXm12 = 1.19
magXm13 = 0.22 magXm14 = 2.41 magXm15 = 0.42
Now, we use Excel to plot the x [ n ] and X [ m ] values. These are shown in Figure 10.2.
On the plot of X [ m ] in Figure 10.2, the first value X [ 0 ] = 46.70 represents the DC component.
We observe that after the X [ 8 ] = 0.10 value, the magnitude of the frequency components for the
range 9 ≤ m ≤ 15 , are the mirror image of the components in the range 1 ≤ m ≤ 7 . This is not a coin-
cidence; it is a fact that if x [ n ] is an N -point real discrete time function, only N ⁄ 2 of the frequency
components of X [ m ] are unique.

Signals and Systems with MATLAB Applications, Second Edition 10-7

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

A B C D E F G H I J
1 n x(n) m |X(m)|
x(n)
2 0 1.0 0 46.70
3 1 1.5 1 11.03 4.7
4.1 4.2
4 2 2.0 2 0.42 3.8 3.6
5 3 2.3 3 2.41 3.4
3.2
3.0 2.9
6 4 2.7 4 0.22 2.7
2.5
2.3
7 5 3.0 5 1.19 2.0
1.8
8 6 3.4 6 0.07 1.5
9 7 4.1 7 0.47 1.0
10 8 4.7 8 0.10
11 9 4.2 9 0.47
12 10 3.8 10 0.07
13 11 3.6 11 1.19
|X(m)|
14 12 3.2 12 0.22
46.70

15 13 2.9 13 2.41
16 14 2.5 14 0.42
17 15 1.8 15 11.03
18
19
20
11.03

11.03
21
22
2.41

2.41
1.19

1.19
23
0.47

0.47
0.42

0.42
0.22

0.22
0.10
0.07

0.07
24
25

Figure 10.2. Plots of x [ n ] and X [ m ] values for Example 10.3

Figure 10.3 shows typical discrete time and frequency magnitude waveforms, for a N = 16 point
DFT.
Next, we will examine the even and odd properties of the DFT.

10.2 Even and Odd Properties of the DFT

The discrete time and frequency functions are defined as even or odd, in accordance with the follow-
ing relations:
Even Time Function f[ N – n] = f [ n ] (10.20)
Odd Time Function f [ N – n ] = –f [ n ] (10.21)
Even Frequency Function F[N – m] = F[ m] (10.22)
Odd Frequency Function F [ N – m ] = –F [ m ] (10.23)

10-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Even and Odd Properties of the DFT

x[n]

n
0

N
(Start of New Period)
N-1
(End of Period)
X[m]

m
0

(N/2)-1 N/2 (N/2)+1

N-1 N
(End of Period) (Start of New Period)

Figure 10.3. x [ n ] and X [ m ] for a N = 16 point DFT.

In Chapter 8, we developed Table 8-7 showing the even and odd properties of the Fourier trans-
form. Table 10.2 shows the even and odd properties of the DFT.
These can be proved by methods similar to those that we have used for the continuous Fourier
transform. To prove the first entry, for example, we expand
N–1
∑ x [ n ]WN
mn
X [m] =
n=0

into its real and imaginary parts using Euler’s identity, and we equate these with the real and imagi-
nary parts of X [ m ] = X re [ m ] + jX im [ m ] . Now, since the real part contains the cosine, and the imag-
inary contains the sine function, and cos ( – m ) = cos m while sin ( – m ) = – sin m , this entry is
proved.

Signals and Systems with MATLAB Applications, Second Edition 10-9

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

TABLE 10.2 Even and Odd Properties of the DFT

Discrete Time Sequence f [ n ] Discrete Frequency Sequence F [ m ]
Real Complex
Real part is Even
Imaginary Part is Odd
Real and Even Real and Even
Real and Odd Imaginary and Odd
Imaginary Complex
Real part is Odd
Imaginary Part is Even
Imaginary and Even Imaginary and Even
Imaginary and Odd Real and Odd

10.3 Properties and Theorems of the DFT

For the proofs of the following properties and theorems of the DFT, we will denote the DFT and
Inverse DFT as follows:

X[ m] = D{x [n]} (10.24)

and
–1
x[ n] = D {X [m]} (10.25)
1. Linearity

ax 1 [ n ] + bx 2 [ n ] + … ⇔ aX 1 [ m ] + bX 2 [ m ] + … (10.26)

where x 1 [ n ] ⇔ X 1 [ m ] , x 2 [ n ] ⇔ X 2 [ m ] , and a and b are arbitrary constants.

Proof:
The proof is readily obtained by using the definition of the DFT.
2. Time Shift

km
x [ n – k ] ⇔ WN X [ m ] (10.27)

Proof:
By definition,

10-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Properties and Theorems of the DFT
N–1
∑ x [ n ]WN
mn
D{x [n]} =
n=0

and if x [ n ] is shifted to the right by k sampled points for k > 0 , we must change the lower and
upper limits of the summation from 0 to k , and from N – 1 to N + k – 1 respectively. Then,
replacing x [ n ] with x [ n – k ] in the definition above, we get
N+k–1
∑
mn
D{x[n – k]} = x [ n – k ]W N (10.28)
n=k

Now, we let n – k = µ ; then n = k + µ , and when n = k , µ = 0 . Therefore, the above relation

becomes
N–1 N–1
m(k + µ) µm
∑ ∑ x [µ]WN
km km km
D{x [µ]} = x [ µ ]W N = W = WN X [ µ ] = WN X [ m ] (10.29)
µ=0 µ=0

We must remember, however, that although the magnitudes of the frequency components are not
affected by the shift, a phase shift of 2πkm ⁄ N radians is introduced as a result of the time shift.
To prove this, let us consider the relation y [ n ] = x [ n – k ] . Taking the DFT of both sides of this
relation, we get
2πkm
km – j -------------- – 2 πkm
Y [ m ] = W N X [ m ] = X [ m ]e N = X [ m ] ∠----------------- (10.30)
N

Since both X [ m ] and Y [ m ] are complex quantities, they can be expressed in magnitude and phase
angle form as
X [ m ] = X [ m ] ∠θ
and
Y [ m ] = Y [ m ] ∠ϕ
By substitution of these into (10.30), we get
– 2 πkm
Y [ m ] ∠ϕ = X [ m ] ∠θ ∠----------------- (10.31)
N

and since Y [ m ] = X [ m ] , it follows that

ϕ = θ – 2πkm
-------------- (10.32)
N

Signals and Systems with MATLAB Applications, Second Edition 10-11

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

3. Frequency Shift

–k m
WN x [ n ] ⇔ X [ m –k ] (10.33)

Proof:
N–1 N–1
( m – k )n
D { WN–kn x [ n ] } = ∑ WN
–k m mn
x [ n ]W N = ∑ x [ n ]WN (10.34)
n=0 n=0

and we observe that the last term on the right side of (10.34) is the same as D { x [ n ] } except that
m is replaced with m – k . Therefore,

D { WN–k m x [ n ] } = X[m – k] (10.35)

4. Time Convolution

x [ n ]∗ h [ n ] ⇔ X [ m ] ⋅ H [ m ] (10.36)

Proof:
Since
N–1
x [ n ]∗ h [ n ] = ∑ x [ n ]h [ n – k ]
k=0
then,
N–1 N–1
D { x [ n ]∗ h [ n ] } ∑ ∑ x [ k ]h [ n – k ]
mn
= WN (10.37)
n=0 k=0

Next, interchanging the order of the indices n and k in the lower limit of the summation, and also
changing the range of summation from N – 1 to N + k – 1 for the bracketed term on the right
side of (10.37), we get
N–1 N+k–1
D { x [ n ]∗ h [ n ] } ∑ ∑
kn
= x[k] h [ n – k ] WN (10.38)
k=0 n=k

Now, from the time shifting theorem,

N+k–1

∑
kn kn
h [ n – k ] WN = WN H [ m ] (10.39)
n=k

10-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Sampling Theorem

and by substitution into (10.38),

N–1
D { x [ n ]∗ h [ n ] } ∑ x [ k ] [ WN
kn
= ] H[m] = X[k ] ⋅ H[m] = X[m ] ⋅ H[m] (10.40)
k=0

5. Frequency Convolution

N–1
1 1-
x [ n ] ⋅ y [ n ] ⇔ ----
N ∑ X [ k ]Y [ m – k ] = ---
N
X [ m ]∗ Y [ m ] (10.41)
k=0
Proof:
The proof is obtained by taking the Inverse DFT, changing the order of the summation, and let-
ting m – k = µ .

10.4 The Sampling Theorem

The sampling theorem, also known as Shannon’s Sampling Theorem, states that if a continuous time
function f ( t ) is band-limited with its highest frequency component less than W , then f ( t ) can be com-
pletely recovered from its sampled values, if the sampling frequency if equal to or greater than 2W .

For example, if we assume that the highest frequency component in a signal is 18 KHz , this signal
must be sampled at 2 × 18 KHz = 36 KHz or higher so that it can be completed specified by its
sampled values. If the sampled frequency remains the same, i.e., 36 KHz , and the highest frequency
of this signal is increased to, say 25 KHz , this signal cannot be recovered by any digital-to-analog
converter.
A typical digital signal processing system contains a low-pass analog filter, often called pre-sampling
filter, to ensure that the highest frequency allowed into the system, will be equal or less the sampling
rate so that the signal can be recovered. The highest frequency allowed by the pre-sampling filter is
referred to as the Nyquist frequency, and it is denoted as f n .

If a signal is not band-limited, or if the sampling rate is too low, the spectral components of the sig-
nal will overlap each another and this condition is called aliasing. To avoid aliasing, we must increase
the sampling rate.
A discrete time signal may have an infinite length; in this case, it must be limited to a finite interval
before it is sampled. We can terminate the signal at the desired finite number of terms, by multiply-
ing it by a window function. There are several window functions such as the rectangular, triangular,
Hamming, Hanning, Kaiser, etc. To obtain a truncated sequence, we multiply an infinite sequence by
one of these window functions. However, we must choose a suitable window function; otherwise,
the sequence will be terminated abruptly producing the effect of leakage.

Signals and Systems with MATLAB Applications, Second Edition 10-13

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

As an example of how leakage can occur, let us review Example 8.7, and Exercise 8.3, where the infi-
nite sequence cos ω 0 t , or cos naT is multiplied by the window function A [ u 0 ( t + T ) – u 0 ( t – T ) ] or
A [ u 0 ( n + m ) – u 0 ( n – m ) ] . We can see that the spectrum spreads or leaks over the neighborhood of
± ω 0 . Selection of an appropriate window function to avoid leakage is beyond the scope of this book,
and will not be discussed here.
A third problem that may arise in using the DFT, results from the fact the spectrum of the DFT is
not continuous; it is a discrete function where the spectrum consists of integer multiples of the fun-
damental frequency. It is possible, however, that some significant frequency component lies between
two spectral lines and goes undetected. This is called picket-fence effect since we can only see discrete
values of the spectrum. This problem can be alleviated by adding zeros at the end of the discrete sig-
nal, thereby changing the period, which in turn changes the location of the spectral lines.We should
remember that the sampling theorem states that the original time sequence can be completely recov-
ered if the sampling frequency is adequate, but does not guarantee that all frequency components will
be detected.
To get a better understanding of the sampling frequency f s , Nyquist frequency f n , number of sam-
ples N , and the periods in the time and frequency domains, we will adopt the following notations:
N = number of samples in time or frequency period
f s = sampling frequency = samples per sec ond
T t = period of a periodic discrete time function
t t = interval between the N samples in time period T t
T f = period of a periodic discrete frequency function
t f = interval between the N samples in frequency period T f

These notations are shown in Figure 10.4. Thus, we have the relations
T T
t t = ----t f s = --1- t f = ----f 1-
t t = --- 1-
t f = --- (10.42)
N tt N Tf Tt

Example 10.4
The period of a periodic discrete time function is 0.125 millisecond, and it is sampled at 1024
equally spaced points. It is assumed that with this number of samples, the sampling theorem is satis-
fied and thus there will be no aliasing.
a. Compute the period of the frequency spectrum in KHz .
b. Compute the interval between frequency components in KHz .
c. Compute the sampling frequency f s

10-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Sampling Theorem

d. Compute the Nyquist frequency f n

x[n] Tt

tt

n
0
N
(Start of New Period)
N− 1
(End of Period)
X[m]
Tf

tf

m
0
(N/2)−1 N/2 (N/2)+1

N −1 N
(End of Period) (Start of New Period)
Figure 10.4. Intervals between samples and periods in discrete time and frequency domains

Solution:
For this example, T t = 0.125 ms and N = 1024 points . Therefore, the time between successive
time components is
–3
T
t t = ----t = 0.125 × 10 - = 0.122 µs
-----------------------------
N 1024
Then,
a. the period T f of the frequency spectrum is

Signals and Systems with MATLAB Applications, Second Edition 10-15

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

1 1
T f = --- = ------------------------------ = 8192 kHz ≈ 8.2 MHz
tt –6
0.122 × 10

b. the interval t f between frequency components is

Tf 8192 kHz
t f = ---- = ------------------------ = 8 kHz
N 1024
c. the sampling frequency f s is
1 1
f s = --- = ------------------------------ = 8.2 MHz
tt –6
0.122 × 10
d. the Nyquist frequency must be equal or less than half the sampling frequency, that is,
1 1
f n ≤ --- f s ≤ --- × 8.2 MHz ≤ 4.1 MHz
2 2

10.5 Number of Operations Required to Compute the DFT

Let us consider a signal whose highest (Nyquist) frequency is 18 KHz , the sampling frequency is
50 KHz , and 1024 samples are taken, i.e., N = 1024 . The time required to compute the entire DFT
would be
1024 samples
t = ---------------------------------------------------------------------------- = 20.48 ms (10.43)
3
50 × 10 samples per sec ond

To compute the number of operations required to complete this task, let us expand the N -point
DFT defined as
N–1
∑ x [ n ]WN
mn
X[m] = (10.44)
n=0
Then,
0 0 0 0
X [ 0 ] = x [ 0 ]W N + x [ 1 ]W N + x [ 2 ]W N + … + x [ N – 1 ]W N
0 1 2 N–1
X [ 1 ] = x [ 0 ]W N + x [ 1 ]W N + x [ 2 ]W N + … + x [ N – 1 ]W N
0 2 4 2(N – 1)
X [ 2 ] = x [ 0 ]W N + x [ 1 ]W N + x [ 2 ]W N + … + x [ N – 1 ]W N (10.45)
…
2
0 N–1 2(N – 1) (N – 1)
X [ N – 1 ] = x [ 0 ]W N + x [ 1 ]W N + x [ 2 ]W N + … + x [ N – 1 ]W N

and it is worth remembering that

⎛ – j -----
2π⎞
- (0)
0 ⎝ N⎠
WN = e = 1 (10.46)

10-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Fast Fourier Transform (FFT)
i
Since W N is a complex number, the computation of any frequency component X [ k ] , requires N
complex multiplications and N complex additions, that is, 2N complex arithmetic operations are
required to compute any frequency component of X [ k ] . If we assume that x [ n ] is real, then only
2
N ⁄ 2 of the | X [ m ] components are unique. Therefore, we would require 2N × N ⁄ 2 = N com-
plex operations to compute the entire frequency spectrum. Thus, for an N = 1024 point DFT, such
2 2
as the one with 18 KHz signal, we would require N = 1024 = 1048576 complex operations, and
these would have to be completed within 20.48 ms as we found in (10.43). Although the means of
doing this task may be possible with today’s technology, it seems impractical.
Fortunately, many of the W N terms in (10.45) are unity. Moreover, because of some symmetry prop-
erties, the number of complex operations can be reduced considerably. This is possible with the
algorithm known as FFT (Fast Fourier Transform) that was developed by Cooley and Tukey, and it is
very well documented. This algorithm is the subject of the next section.

10.6 The Fast Fourier Transform (FFT)

In this section, we will be making extensive use of the complex rotating vector
j2π
– --------
WN = e N (10.47)

Some additional properties of W N are given in (10.48) below.

– ⎛ --------⎞ N
j2π
N ⎝ N⎠ – j2π
WN = e = e = 1

– ⎛ --------⎞ -----
j2π N
N⁄2 ⎝ N⎠ 2 – jπ
WN = e = e = –1

– ⎛ --------⎞ -----
j2π N
N⁄4 ⎝ N⎠ 4 – jπ ⁄ 2
WN = e = e = –j

– ⎛ --------⎞ -------
j2π 3N
3N ⁄ 4 ⎝ N⎠ 4 – j3π ⁄ 2 (10.48)
W N = e = e = j

– ⎛ --------⎞ kN
j2π
kN ⎝ N⎠ – j2kπ
W N = e = e = 1 k = 0, 1, 2, …
– ⎛ --------⎞ ( kN + r ) – ⎛ --------⎞ kN – ⎛ --------⎞ r
j2π j2π j2π
kN + r ⎝ N⎠ ⎝ N ⎠ ⎝ N⎠ r r
WN = e = e e = 1 ⋅ WN = WN

– ⎛ --------⎞ ---
j2π k
– ⎛ --------⎞ k
j2π
k ⎝ 2N ⎠ ⎝ N ⎠2 k⁄2
W 2N = e = e = WN

We rewrite the array of (10.45) in matrix form as shown in (10.49) below.

Signals and Systems with MATLAB Applications, Second Edition 10-17

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

0 0 0 0
WN WN WN … WN
X [0] 0 1 2 N–1
x [0]
X [1] WN WN WN … WN x [1]
X [2] = W0 W2 WN
4
… WN
2(N – 1 ) ⋅ x [2] (10.49)
N N
… … … … … … …
X [N – 1] 0 N–1 2(N – 1) (N – 1)
2 x [N – 1]
WN WN WN … WN

This is a complex Vandermonde matrix and it is expressed in a more compact form as

X [ m ] = WN ⋅ x [ n ] (10.50)

The algorithm that was developed by Cooley and Tukey, is based on matrix decomposition methods,
where the matrix W in (10.50) is factored into L smaller matrices, that is,

WN = W1 ⋅ W2 ⋅ … ⋅ WL (10.51)

L
where L is chosen as L = log 2 N or N = 2
Each row of the matrices on the right side of (10.51) contains only two non-zero terms, unity and
k
W N . Then, the vector X [ m ] , is obtained from

X [ m ] = W1 ⋅ W2 ⋅ … ⋅ WL ⋅ x [ n ] (10.52)

The FFT computation begins with matrix W L in (10.52). This matrix operates on vector x [ n ] pro-
ducing a new vector, and each component of this new vector, is obtained by one multiplication and
one addition. This is because there are only two non-zero elements on a given row, and one of these
elements is unity. Since there are N components on x [ n ] , there will be N complex additions and N
complex multiplications. This new vector is then operated on by the [ W L – 1 ] matrix, then on
[ W L – 2 ] , and so on, until the entire computation is completed. It appears that the entire computa-
tion would require NL = Nlog 2 N complex multiplications, and also Nlog 2 N additions for a total of
0 N⁄2
2Nlog 2 N arithmetic operations. However, since W N = 1 , W N = – 1 , and other reductions, it is
estimated that only about half of these, that is, Nlog 2 N total arithmetic operations are required by the
2
FFT versus the N computations required by the DFT.

10-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Fast Fourier Transform (FFT)

Under those assumptions, we construct Table 10.3 to compare the percentage of computations
achieved by the use of FFT versus the DFT.

TABLE 10.3 DFT and FFT Computations

DFT FFT FFT/DFT
N N
2 N log 2N %
8 64 24 37.5
16 256 64 25
32 1024 160 15.6
64 4096 384 9.4
128 16384 896 5.5
256 65536 2048 3.1
512 262144 4608 1.8
1024 1048576 10240 1
2048 4194304 22528 0.5

A plethora of FFT algorithms has been developed and published. They are divided into two main
categories:
Category I
a. Decimation in Time
b. Decimation in Frequency
Category II
a. In−Place
b. Natural Input−Output (Double-Memory Technique)
To define Category I, we need to refer to the DFT and Inverse DFT definitions. They are repeated
below for convenience.
N–1
∑ x [ n ]WN
mn
X [m] = (10.53)
n=0

and

1N–1 –m n
N ∑
x [ n ] = ---- X [ m ]W N (10.54)
n=0

Signals and Systems with MATLAB Applications, Second Edition 10-19

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

We observe that (10.53) and (10.54) differ only by the factor 1 ⁄ N , and the replacement of W N with
–1
W N . If the DFT algorithm is developed in terms of the direct DFT of (10.53), it is referred to as
decimation in time, and if it is developed in terms of the Inverse DFT of (10.54), it is referred to as
decimation in frequency. In the latter case, the vector
2π
– j ------
WN = e N

is replaced by its complex conjugate

2π
–1 j ------
N
WN = e

that is, the sine terms are reversed in sign, and the multiplication by the factor 1 ⁄ N can be done
either at the input or output.
The Category II algorithm schemes are described in the Table 10.4 along with their advantages and
disadvantages.

TABLE 10.4 In-Place and Natural Input-Output algorithms

Category II Description Advantages Disadvantages
algorithm
In-Place The process where the result Eliminates the need for The output appears in an
of a computation of a new intermediate storage unnatural order and must
vector is stored in the same and memory require- be re-ordered.
memory location as the result ments
of the previous computation
Natural The process where the output No reordering is Requires more
Input-Output appears in same (natural) required internal memory to pre-
(Double Memory) order as the input serve the natural order

Now, we will explain how the unnatural order occurs and how it can be re-ordered.
Consider the discrete time sequence f [ n ] ; its DFT is found from
N–1
∑ f [ n ]WN
mn
F[m] = (10.55)
n=0

We assume that N is a power of 2 and thus, it is divisible by 2 . Then, we can decompose the
sequence f [ n ] into two subsequences, f even [ n ] which contains the even components, and f odd [ n ]

which contains the odd components. In other words, we choose these as

f even [ n ] = f [ 2n ]

10-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Fast Fourier Transform (FFT)

and
f odd [ n ] = f [ 2n + 1 ]
for n = 0, 1, 2, …N ⁄ 2 – 1
Each of these subsequences has a length of N ⁄ 2 and thus, their DFTs are, respectively,
N⁄2–1 N⁄2–1
∑ ∑
mn mn
F even [ m ] = f even [ n ]W N ⁄ 2 = f [ 2n ]W N ⁄ 2 (10.56)
n=0 n=0

and
N⁄2–1 N⁄2–1
∑ ∑
mn mn
F odd [ m ] = f odd [ n ]W N ⁄ 2 = f [ 2n + 1 ]W N ⁄ 2 (10.57)
n=0 n=0

where
2π
– j ----------- ⎛ – j -----
2π⎞
- 2
N⁄2 = e⎝ N⎠ 2
WN ⁄ 2 = e = WN (10.58)

For an 8 -point DFT, N = 8 . Expanding (10.55) for n = 1, 2, 3, …, 7 we get

∑ f [ n ]WN
mn
F[m] =
n=0
m 2m 3m (10.59)
= f [ 0 ] + f [ 1 ]W N + f [ 2 ]W N + f [ 3 ]W N
4m 5m 6m 7m
+ f [ 4 ]W N + f [ 5 ]W N + f [ 6 ]W N + f [ 7 ]W N

0
Expanding (10.56) for n = 0, 1, 2, and 3 and recalling that W N = 1 , we get
3

∑ f [ 2n ]WN
2mn
F even [ m ] =
n=0
0 2m 4m 6m (10.60)
= f [ 0 ]W N + f [ 2 ]W N + f [ 4 ]W N + f [ 6 ]W N
2m 4m 6m
= f [ 0 ] + f [ 2 ]W N + f [ 4 ]W N + f [ 6 ]W N

0
Expanding also (10.57) for n = 0, 1, 2, and 3 and using W N = 1 , we get
3

∑ f [ 2n + 1 ]WN
2 mn 0 2m 4m 6m
F odd [ m ] = = f [ 1 ]W N + f [ 3 ]W N + f [ 5 ]W N + f [ 7 ]W N
n=0 (10.61)
2m 4m 6m
= f [ 1 ] + f [ 3 ]W N + f [ 5 ]W N + f [ 7 ]W N

Signals and Systems with MATLAB Applications, Second Edition 10-21

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

The vector W N is the same in (10.59), (10.60) and (10.61), and N = 8 . Then,
2π 2π
– j ------ – j ------
WN = W8 = e N = e 8

m
Multiplying both sides of (10.61) by W N , we get
m m 3m 5m 7m
W N F odd [ m ] = f [ 1 ]W N + f [ 3 ]W N + f [ 5 ]W N + f [ 7 ]W N (10.62)

and from (10.59), (10.60) and (10.62), we observe that

m
F [ m ] = F even [ m ] + W N F odd [ m ] (10.63)

for n = 1, 2, 3, …, 7 .
Continuing the process, we decompose { f [ 0 ], f [ 2 ], f [ 4 ], and f [ 6 ] } into { f [ 0 ], f [ 4 ] } and
{ f [ 2 ], f [ 6 ] } . These are sequences of length N ⁄ 4 = 2 .

Denoting their DFTs as F even1 [ m ] and F even2 [ m ] , and using the relation

2π ⎛ – j -----
2π⎞
- 4
– j ----------- ⎝ N⎠
N⁄4 4
WN ⁄ 4 = e =e = WN (10.64)

for n = 0, 1 , we get
4m
F even1 [ m ] = f [ 0 ] + f [ 4 ]W N (10.65)

and
4m
F even2 [ m ] = f [ 2 ] + f [ 6 ]W N (10.66)

The sequences of (10.65) and (10.66) cannot be decomposed further. They justify the statement
made earlier, that each computation produces a vector where each component of this vector, for
n = 1, 2, 3, …, 7 is obtained by one multiplication and one addition. This is often referred to as a
butterfly operation.
Substitution of (10.65) and (10.66) into (10.60), yields
2m
F even [ m ] = F even1 [ m ] + W N F even2 [ m ] (10.67)

Likewise, F odd [ m ] can be decomposed into DFTs of length 2; then, F [ m ] can be computed from

m
F ( m ) = F even ( m ) + W N F odd ( m ) m = 0, 1, 2, …, 7 (10.68)

10-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Fast Fourier Transform (FFT)

for N = 8 . Of course, this procedure can be extended for any N that is divisible by 2.

Figure 10.5 shows the signal flow graph of a decimation in time, in-place FFT algorithm for N = 8 ,
where the input is shuffled in accordance with the above procedure. The subscript N in W has been
omitted for clarity.

Column 0 (x[n]) Column 1 (N /4 ) Column 2 (N / 2) Column 3 (N) X[m]

Row 0 x[0] W0 W0 W0 X[0]

Row 1 x[4] W4 W2 W1 X[1]

Row 2 x[2] W0 W4 W2 X[ 2]

Row 3 x[6] W4 W6 W3 X[3]

Row 4 x[1] W0 W0 W4 X[ 4]

Row 5 x[5] W4 W2 W5 X[5 ]

Row 6 x[3] W0 W4 W6 X[6]

Row 7 x[7] W4 W6 W7 X[ 7]

Figure 10.5. Signal flow graph of a decimation in time, in-place FFT algorithm, for N = 8

In the signal flow graph of Figure 10.5, the input x [ n ] appears in Column 0 . The N ⁄ 4 , N ⁄ 2 , and
N -point FFTs are in Columns 1 , 2 , and 3 respectively.

In simplified form, the output of the node associated with row R and column C , indicated as
Y ( R, C ) , is found from

Y [ Ri , C – 1 ] m
W Y [ Rj , C – 1 ]
Y [ R, C ] = + (10.69)
⎧
⎪
⎪
⎨
⎪
⎪
⎩

⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩

Dash line Solid line

m
where 0 ≤ R ≤ 7 , 0 ≤ C ≤ 3 , and the exponent m in W is indicated on the signal flow graph.

Signals and Systems with MATLAB Applications, Second Edition 10-23

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

The binary input words and the bit-reversed words applicable to the this signal flow graph, are
shown in Table 10.5.

TABLE 10.5 Binary words for the signal flow graph of Figure 10.5
n Binary Word Reversed-Bit Word Input Order
0 000 000 x[0]
1 001 100 x[4]
2 010 010 x[2]
3 011 110 x[6]
4 100 001 x[1]
5 101 101 x[5]
6 110 011 x[3]
7 111 111 x[7]

We will illustrate the use of (10.69) with the following example.

Example 10.5
Using (10.69) with the signal flow graph of Figure 10.5, compute the spectral component X [ 3 ] in
j
terms of the inputs x [ i ] and vectors W . Then, verify that the result is the same as that obtained by
direct application of the DFT.
Solution:
The N ⁄ 4 point FFT appears in Column 1. Using (10.69) we get:
0 0
Y [ 0, 1 ] = x [ 0 ] + W x [ 4 ] = Y [ 0, 0 ] + W Y [ 1, 0 ]
4 4
Y [ 1, 1 ] = x [ 0 ] + W x [ 4 ] = Y [ 0, 0 ] + W Y [ 1, 0 ]
0 0
Y [ 2, 1 ] = x [ 2 ] + W x [ 6 ] = Y [ 2, 0 ] + W Y [ 3, 0 ]
4 4
Y [ 3, 1 ] = x [ 2 ] + W x [ 6 ] = Y [ 2, 0 ] + W Y [ 3, 0 ]
(10.70)
0 0
Y [ 4, 1 ] = x [ 1 ] + W x [ 5 ] = Y [ 4, 0 ] + W Y [ 5, 0 ]
4 4
Y [ 5, 1 ] = x [ 1 ] + W x [ 5 ] = Y [ 4, 0 ] + W Y [ 5, 0 ]
0 0
Y [ 6, 1 ] = x [ 3 ] + W x [ 7 ] = Y [ 6, 0 ] + W Y [ 7, 0 ]
0 4
Y [ 7, 1 ] = x [ 3 ] + W x [ 7 ] = Y [ 6, 0 ] + W Y [ 7, 0 ]

The N ⁄ 2 point FFT appears in Column 2. Using (10.69) we get:

10-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 The Fast Fourier Transform (FFT)

0
Y [ 0, 2 ] = Y [ 0, 1 ] + W Y [ 2, 1 ]
2
Y [ 1, 2 ] = Y [ 1, 1 ] + W Y [ 3, 1 ]
4
Y [ 2, 2 ] = Y [ 0, 1 ] + W Y [ 2, 1 ]
6
Y [ 3, 2 ] = Y [ 1, 1 ] + W Y [ 3, 1 ]
(10.71)
0
Y [ 4, 2 ] = Y [ 4, 1 ] + W Y [ 6, 1 ]
2
Y [ 5, 2 ] = Y [ 5, 1 ] + W Y[ 7, 1 ]
4
Y [ 6, 2 ] = Y [ 4, 1 ] + W Y [ 6, 1 ]
6
Y [ 7, 2 ] = Y [ 5, 1 ] + W Y [ 7, 1 ]

The N point FFT appears at the outputs of Column 3, where

0
Y [ 0, 3 ] = Y [ 0, 2 ] + W Y [ 4, 2 ]
1
Y [ 1, 3 ] = Y [ 1, 2 ] + W Y [ 5, 2 ]
2
Y [ 2, 3 ] = Y [ 2, 2 ] + W Y [ 6, 2 ]
3
Y [ 3, 3 ] = Y [ 3, 2 ] + W Y [ 7, 2 ]
(10.72)
4
Y [ 4, 3 ] = Y [ 0, 2 ] + W Y [ 4, 2 ]
5
Y [ 5, 3 ] = Y [ 1, 2 ] + W Y [ 5, 2 ]
6
Y[ 6, 3 ] = Y [ 2, 2 ] + W Y [ 6, 2 ]
7
Y [ 7, 3 ] = Y [ 3, 2 ] + W Y [ 7, 2 ]

With the equations of (10.70), (10.71), and (10.72), we can determine any of the 8 spectral compo-
nents. Our example calls for X [ 3 ] ; then, with reference to the signal flow chart of Figure 10.5 and
the fourth of the equations in (10.72),
3
X [ 3 ] = Y [ 3, 3 ] = Y [ 3, 2 ] + W Y [ 7, 2 ] (10.73)
Also, from (10.71)
6
Y [ 3, 2 ] = Y [ 1, 1 ] + W Y [ 3, 1 ] (10.74)
and
6
Y [ 7, 2 ] = Y [ 5, 1 ] + W Y [ 7, 1 ] (10.75)
Finally, from (10.70)

Signals and Systems with MATLAB Applications, Second Edition 10-25

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

4
Y [ 1, 1 ] = Y [ 0, 0 ] + W Y [ 1, 0 ]
4
Y [ 3, 1 ] = Y [ 2, 0 ] + W Y [ 3, 0 ]
(10.76)
4
Y [ 5, 1 ] = Y [ 4, 0 ] + W Y [ 5, 0 ]
4
Y [ 7, 1 ] = Y [ 6, 0 ] + W Y [ 7, 0 ]
and making these substitutions into (10.73), we get
3
Y [ 3, 3 ] = Y [ 3, 2 ] + W Y [ 7, 2 ]
6 3 6
= Y [ 1, 1 ] + W Y [ 3, 1 ] + W { Y [ 5, 1 ] + W Y [ 7, 1 ] }
4 6 4
= Y [ 0, 0 ] + W Y [ 1, 0 ] + W { Y [ 2, 0 ] + W Y [ 3, 0 ] }
3 4 6 4
+ W { ( Y [ 4, 0 ] + W Y [ 5, 0 ] ) + W ( Y [ 6, 0 ] + W Y [ 7, 0 ] ) }
4 6 10
= Y [ 0, 0 ] + W Y [ 1, 0 ] + W Y [ 2, 0 ] + W Y [ 3, 0 ]
3 7 9 13
+ W Y [ 4, 0 ] + W Y [ 5, 0 ] + W Y [ 6, 0 ] + W Y [ 7, 0 ]

Rearranging in ascending powers of W , we get

3 4 6
Y [ 3, 3 ] = Y [ 0, 0 ] + W Y [ 4, 0 ] + W Y [ 1, 0 ] + W Y [ 2, 0 ]
(10.77)
7 9 10 13
+ W Y [ 5, 0 ] + W Y [ 6, 0 ] + W Y [ 3, 0 ] + W Y [ 7, 0 ]
From the signal flow graph of Figure 10.5, we observe that
Y [ 0, 0 ] = x [ 0 ]
Y [ 4, 0 ] = x [ 1 ]
Y [ 1, 0 ] = x [ 4 ]
Y [ 2, 0 ] = x [ 2 ]
Y [ 5, 0 ] = x [ 5 ]
Y[ 6, 0 ] = x [ 3 ]
Y [ 3, 0 ] = x [ 6 ]
Y[ 7, 0 ] = x [ 7 ]
and
Y [ 3, 3 ] = X [ 3 ]
Therefore, we express (10.77) as

10-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

3 4 6
X [ 3 ] = x [ 0 ] + x [ 1 ]W + x [ 4 ]W + x [ 2 ]W
7 9 10 13
+ x [ 5 ]W + x [ 3 ]W + x [ 6 ]W + x [ 7 ]W
(10.78)
3 6 9
= x [ 0 ] + x [ 1 ]W + x [ 2 ]W + x [ 3 ]W
4 7 10 13
+ x [ 4 ]W + x [ 5 ]W + x [ 6 ]W + x [ 7 ]W

We will verify that this expression is correct by the use of the direct DFT of (10.17), that is,
N–1
∑ x ( n )WN
mn
X[m] =
n=0

For m = 3 and n = 0, 1, 2, …, 7 , we get

3 6 9
X [ 3 ] = x [ 0 ] + x [ 1 ]W + x [ 2 ]W + x [ 3 ]W
(10.79)
12 15 18 21
+ x [ 4 ]W + x [ 5 ]W + x [ 6 ]W + x [ 7 ]W
and from (10.48),
kN + r r
W = W

Now, with N = 8 , and k = 1 , we get kN = 8 . Then,

12 4 15 7 18 10 21 13
W = W , W =W , W =W and W = W
and by substitution into (10.79),
3 6 9
X [ 3 ] = x [ 0 ] + x [ 1 ]W + x [ 2 ]W + x [ 3 ]W
(10.80)
4 7 10 13
+ x [ 4 ]W + x [ 5 ]W + x [ 6 ]W + x [ 7 ]W
We observe that (10.80) is the same as (10.78).
The signal flow graph of Figure 10.5, represents a shuffled input, natural output algorithm. Others
are natural input, natural output, or natural input, shuffled output. These combinations occur in
both decimation in time, and decimation in frequency algorithms.

10.7 Summary
• The N -point DFT is defined as
mn
N–1 – j2π -------

∑ x [ n ]e
N
X[m] =
n=0

where N represents the number of points that are equally spaced in the interval 0 to 2π on the
unit circle of the z -plane, and m = 0, 1, 2, …, N – 1 .

Signals and Systems with MATLAB Applications, Second Edition 10-27

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

• The N -point Inverse DFT is defined as

mn
1 N–1 j2π -------

∑
N
x [ n ] = ---- X [ m ]e
N
m=0
for n = 0, 1, 2, …, N – 1 .
• In general, the discrete frequency transform X [ m ] is complex, and it is expressed as

X [ m ] = Re { X [ m ] } + Im { X [ m ] }

The real part Re { X [ m ] } can be computed from

N–1
2πmn
Re { X [ m ] } = x [ 0 ] + ∑ x [ n ] cos -------------
N
- for m = 0, 1, 2, …, N – 1
n=1

and the imaginary part from

N–1
∑ x [ n ] sin -------------
Im { X [ m ] } = – 2πmn for m = 0, 1, 2, …, N – 1
-
N
n=1

• We can use the fft(x) function to compute the DFT, and the ifft(x) function to compute the
Inverse DFT.
• The term e – ( j2π ) ⁄ N is a rotating vector, where the range 0 ≤ θ ≤ 2π is divided into 360 ⁄ N equal
segments and it is denoted as represent it as W N , that is,

j2π
– --------
WN = e N

and consequently
j2π
–1 --------
WN = e N

Accordingly, the DFT pair is normally denoted as

N–1
∑ x ( n )WN
mn
X[ m] =
n=0
and
1N–1 –m n
x [ n ] = ----
N ∑
X ( m )W N
n=0

• The correspondence between x [ n ] and X [ m ] is denoted as

x [n] ⇔ X [m]

10-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

• If x [ n ] is an N -point real discrete time function, only N ⁄ 2 of the frequency components of

X [ m ] are unique.
• The discrete time and frequency functions are defined as even or odd, in accordance with the rela-
tions
Even Time Function f[ N – n] = f [n ]

Odd Time Function f [ N – n ] = –f [ n ]

Even Frequency Function F[N – m] = F[m]

Odd Frequency Function F [ N – m ] = –F [ m ]

• The even and odd properties of the DFT are similar to those of the continuous Fourier transform
and are listed in Table 10.2.
• The linearity property of the DFT states that

ax 1 [ n ] + bx 2 [ n ] + … ⇔ aX 1 [ m ] + bX 2 [ m ] + …

• The time shift property of the DFT states that

km
x [ n – k ] ⇔ WN X [ m ]

• The frequency shift property of the DFT states that

–k m
WN x [ n ] ⇔ X [ m –k ]

• The time convolution property of the DFT states that

x [ n ]∗ h [ n ] ⇔ X [ m ] ⋅ H [ m ]
• The frequency convolution property of the DFT states that
N–1
1 1
x [ n ] ⋅ y [ n ] ⇔ ----
N ∑ X [ k ]Y [ m – k ] = ---- X [ m ]∗ Y [ m ]
N
k=0

• The sampling theorem, also known as Shannon’s Sampling Theorem, states that if a continuous
time function f ( t ) is band-limited with its highest frequency component less than W , then f ( t )
can be completely recovered from its sampled values, if the sampling frequency if equal to or
greater than 2W .
• A typical digital signal processing system contains a low-pass analog filter, often called pre-sam-
pling filter, to ensure that the highest frequency allowed into the system, will be equal or less the
sampling rate so that the signal can be recovered. The highest frequency allowed by the pre-sam-
pling filter is referred to as the Nyquist frequency, and it is denoted as f n .

Signals and Systems with MATLAB Applications, Second Edition 10-29

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

• If a signal is not band-limited, or if the sampling rate is too low, the spectral components of the sig-
nal will overlap each another and this condition is called aliasing.
• If a discrete time signal has an infinite length, we can terminate the signal at a desired finite num-
ber of terms, by multiplying it by a window function. However, we must choose a suitable window
function; otherwise, the sequence will be terminated abruptly producing the effect of leakage
• If in a discrete time signal some significant frequency component that lies between two spectral
lines and goes undetected, the picket-fence effect is produced. This effect can be alleviated by add-
ing zeros at the end of the discrete signal, thereby changing the period, which in turn changes the
location of the spectral lines.
• The number of operations required to compute the DFT can be significantly reduced by the FFT
algorithm.
• The Category I FFT algorithms are classified either as decimation it time or decimation in fre-
quency. Decimation in time implies that DFT algorithm is developed in terms of the direct DFT,
whereas decimation in frequency implies that the DFT is developed in terms of the Inverse DFT.
• The Category II FFT algorithms are classified either as in-place or natural input−output. In-place
refers to the process where the result of a computation of a new vector is stored in the same mem-
ory location as the result of the previous computation. Natural input−output refers to the process
where the output appears in same (natural) order as the input.
• The FFT algorithms are usually shown in a signal flow graph. In some signal flow graphs the input
is shuffled and the output is natural, and in others the input is natural and the output is shuffled.
These combinations occur in both decimation in time, and decimation in frequency algorithms.

10-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

10.8 Exercises
1. Compute the DFT of the sequence x [ 0 ] = x [ 1 ] = 1 , x [ 2 ] = x [ 3 ] = – 1
2. A square waveform is represented by the discrete time sequence
x [ 0 ] = x [ 1 ] = x [ 2 ] = x [ 3 ] = 1 and x [ 4 ] = x [ 5 ] = x [ 6 ] = x [ 7 ] = – 1

Use MATLAB to compute and plot the magnitude X [ m ] of this sequence.

3. Prove that
1
a. x [ n ] cos 2πkn
------------- ⇔ --- { X [ m – k ] + X [ m + k ] }
N 2

1
b. x [ n ] sin 2πkn
------------- ⇔ ----- { X [ m – k ] + X [ m + k ] }
N j2

4. The signal flow graph of Figure 10.6 is a decimation in time, natural-input, shuffled-output type
FFT algorithm. Using this graph and relation (10.69), compute the frequency component X [ 3 ] .
Verify that this is the same as that found in Example 10.5.

x[0] W0 W0 W0 X[ 0]

x[1] W0 W0 W4 X[4 ]

x[2] W0 W4 W2 X[2]

x[3] W0 W4 W6 X[ 6]

x[4] W4 W2 W1 X[1]

x[5] W4 W2 W5 X[5 ]

x[6] W4 W6 W3 X[ 3]

x[7] W4 W6 W7 X[7]

Figure 10.6. Signal flow graph for Exercise 4

5. The signal flow graph of Figure 10.7 is a decimation in frequency, natural input, shuffled output
type FFT algorithm. There are two equations that relate successive columns. The first is
Y dash ( R, C ) = Y dash ( R i , C – 1 ) + Y dash ( R j , C – 1 )

and it is used with the nodes where two dashed lines terminate on them.

Signals and Systems with MATLAB Applications, Second Edition 10-31

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

The second equation is

m
Y sol ( R, C ) = W [ Y sol ( R i , C – 1 ) – Y sol ( R j , C – 1 ) ]

and it is used with the nodes where two solid lines terminate on them. The number inside the cir-
cles denote the power of W N , and the minus (−) sign below serves as a reminder that the brack-
eted term of the second equation involves a subtraction. Using this graph and the above equa-
tions, compute the frequency component X [ 3 ] . Verify that this is the same as in Example 10.5.

x[ 0] X[ 0]

x[1] − W0 X[4]

x[2] −W
0 X[2]

x[ 3] − W −W
2 0 X[ 6]

x[4] 0 X[1]
−W
x[5] 1 X[5]
−W
0
− W
x[ 6] −W0 X[ 3]
−W
2

x[7] −W
3
− W −W
2 0
X[7]

Figure 10.7. Signal flow graph for Exercise 5

10-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

10.9 Solutions to Exercises

1.
N–1
∑ f ( n )WN
mn
F[ m] =
n=0

where N = 4 and f [ 0 ] = f [ 1 ] = 1 , f [ 2 ] = f [ 3 ] = – 1 . Then,

3
m[0] m[1] m[2] m[3]
∑ f ( n )W4
mn
F[m] = = f [ 0 ]W 4 + f [ 1 ]W 4 + f [ 2 ]W 4 + f [ 3 ]W 4
n=0

for m = 0, 1, 2, and 3
m = 0:
0 0 0 0
F ( 0 ) = ( 1 ) ⋅ ( e ) + ( 1 ) ⋅ ( e ) + ( –1 ) ⋅ ( e ) + ( –1 ) ⋅ ( e )
= ( 1 ) ⋅ ( 1 ) + ( 1 ) ⋅ ( 1 ) + ( –1 ) ⋅ ( 1 ) + ( –1 ) ⋅ ( 1 ) = 0
m = 1:
0 ( – j2π ⁄ 4 ) × 1 ( – j2π ⁄ 4 ) × 2 ( – j2π ⁄ 4 ) × 3
F(1) = (1) ⋅ (e ) + (1) ⋅ (e ) + ( –1 ) ⋅ ( e ) + ( –1 ) ⋅ ( e )
--- – j sin π
= 1 + cos π ------ + j sin 3π
--- – cos π + j sin π – cos 3π ------
2 2 2 2
= 1 + 0 – j + 1 + 0 – 0 – j = 2 – j2
m = 2:
0 ( – j2π ⁄ 4 ) × 2 ( – j2π ⁄ 4 ) × 4 ( – j2π ⁄ 4 ) × 6
F(2) = (1) ⋅ (e ) + (1) ⋅ (e ) + ( –1 ) ⋅ ( e ) + ( –1 ) ⋅ ( e )
= 1 + cos π – j sin π – cos 2π + j sin 2π – cos 3π + j sin 3π
= 1–1–0–1+0+1+0 = 0
m = 3:
0 ( – j2π ⁄ 4 ) × 3 ( – j2π ⁄ 4 ) × 6 ( – j2π ⁄ 4 ) × 9
F(3) = (1) ⋅ (e ) + (1) ⋅ (e ) + ( –1 ) ⋅ ( e ) + ( –1 ) ⋅ ( e )
= 1 + cos 3π 3π 3π 3π
------ – j sin ------ – cos 3π + j sin 3π – cos ------ + j sin ------
2 2 2 2
= 1 + 0 + j + 1 + 0 – 0 + j = 2 + j2
Check with MATLAB:
fn=[1 1 −1 −1]; Fm=fft(fn)
Fm =
0 2.0000 - 2.0000i 0 2.0000 + 2.0000i

Signals and Systems with MATLAB Applications, Second Edition 10-33

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

2.
x [ 0 ] = x [ 1 ] = x [ 2 ] = x [ 3 ] = 1 and x [ 4 ] = x [ 5 ] = x [ 6 ] = x [ 7 ] = – 1
fn=[1 1 1 1 −1 −1 −1 −1]; magXm=abs(fft(fn)); fprintf(' \n');...
fprintf('magXm0 = %4.2f \t', magXm(1)); fprintf('magXm1 = %4.2f \t', magXm(2));...
fprintf('magXm2 = %4.2f \t', magXm(3)); fprintf('magXm3 = %4.2f \t', magXm(4));...
fprintf(' \n');...
fprintf('magXm4 = %4.2f \t', magXm(5)); fprintf('magXm5 = %4.2f \t', magXm(6));...
fprintf('magXm6 = %4.2f \t', magXm(7)); fprintf('magXm7 = %4.2f \t', magXm(8))
magXm0 = 0.00 magXm1 = 5.23 magXm2 = 0.00 magXm3 = 2.16
magXm4 = 0.00 magXm5 = 2.16 magXm6 = 0.00 magXm7 = 5.23
The MATLAB stem command can be used to plot discrete sequence data. For this exercise we
use the code
fn=[1 1 1 1 −1 −1 −1 −1]; stem(abs(fft(fn)))
and we obtain the plot below.

3.
2πkn- ⇔ 1
x [ n ] cos ------------ --- { X [ m – k ] + X [ m + k ] } x [ n ] sin 2πkn
1
------------- ⇔ ----- { X [ m – k ] + X [ m + k ] }
N 2 N j2

From the frequency shift property of the DFT

–k m
WN x [ n ] ⇔ X [ m – k ] (1)

Then,
km
W N x [ n ] ⇔ X [ m + k ] (2)

Adding (1) and (2) and multiplying the sum by 1 ⁄ 2 we get

10-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises
–k m km j2πkn ⁄ N – j 2πkn ⁄ N
( WN + WN ) ( x [ n ] ) (-----------------------------------------------------
e +e )-
----------------------------------------------------- = x [ n ] = x [ n ] cos 2πkn
-------------
2 2 N
and thus
1
x [ n ] cos 2πkn
------------- ⇔ --- [ X [ m – k ] + X [ m + k ] ]
N 2

Likewise, subtracting (2) from (1) and multiplying the difference by 1 ⁄ j2 we get
–k m km j2πkn ⁄ N – j 2πkn ⁄ N
( WN – WN ) ( x [ n ] ) (e –e )
- = ----------------------------------------------------- x [ n ] = x [ n ] sin 2πkn
---------------------------------------------------- -------------
j2 j2 N
4.
x[0] W0 W0 W0 X[ 0]

x[1] W0 W0 W4 X[4 ]

x[2] W0 W4 W2 X[2]

x[3] W0 W4 W6 X[ 6]

x[4] W4 W2 W1 X[1]

x[5] W4 W2 W5 X[5 ]

x[6] W4 W6 W3 X[ 3]

x[7] W4 W6 W7 X[7]

3
F ( 3 ) = X ( 3 ) = Y ( 6, 3 ) = Y ( 6, 2 ) + W N Y ( 7, 2 ) (1)
where
6
Y ( 6, 2 ) = Y ( 4, 1 ) + W N Y ( 6, 1 )
and
6
Y ( 7, 2 ) = Y ( 5, 1 ) + W N Y ( 7, 1 )

Going backwards (to the left) we find that

4
Y ( 4, 1 ) = Y ( 0, 0 ) + W N Y ( 4, 0 )
4
Y ( 6, 1 ) = Y ( 2, 0 ) + W N Y ( 6, 0 )
4
Y ( 5, 1 ) = Y ( 1, 0 ) + W N Y ( 5, 0 )
4
Y ( 7, 1 ) = Y ( 3, 0 ) + W N Y ( 7, 0 )

Signals and Systems with MATLAB Applications, Second Edition 10-35

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

and by substitution into (1)

6 3 6
F ( 3 ) = X ( 3 ) = Y ( 6, 3 ) = Y ( 4, 1 ) + W N Y ( 6, 1 ) + W N [ Y ( 5, 1 ) + W N Y ( 7, 1 ) ]
4 6 4
= Y ( 0, 0 ) + W N Y ( 4, 0 ) + W N [ Y ( 2, 0 ) + W N Y ( 6, 0 ) ]
3 4 6
+W N { Y ( 1, 0 ) + W N Y ( 5, 0 ) + W N [ Y ( 3, 0 ) + W N Y ( 7, 0 ) ] }
4
(2)
3 4 6
= Y ( 0, 0 ) + W N Y ( 1, 0 ) + W N Y ( 4, 0 ) + W N Y ( 2, 0 )
7 9 10 13
+W N Y ( 5, 0 ) + W N Y ( 3, 0 ) + W N Y ( 6, 0 ) + W N Y ( 7, 0 )

From the DFT definition

3 6 9
F ( 3 ) = X ( 3 ) = x ( 0 ) + WN x ( 1 ) + WN x ( 2 ) + WN x ( 3 )
(3)
12 15 18 21
+W N x ( 4 ) + W N x ( 5 ) + W N x ( 6 ) + W N x ( 7 )

By comparison, we see that the first 4 terms of (3) are the same the first, second, fourth, and sixth
terms of (2) since Y ( k, 0 ) = x ( k ) , that is, Y ( 0, 0 ) = x ( 0 ) , Y ( 1, 0 ) = x ( 1 ) , and so on. The
kN + r r 12 4
remaining terms in (2) and (3) are also the same since W 8 = W 8 and thus W 8 = W 8 ,
15 7 18 10 21 13
W 8 = W 8 , W 8 = W 8 , and W 8 = W 8 .

5.
Y dash ( R, C ) = Y dash ( R i , C – 1 ) + Y dash ( R j , C – 1 )

m
Y sol ( R, C ) = W [ Y sol ( R i , C – 1 ) – Y sol ( R j , C – 1 ) ]

x[ 0] X[ 0]

x[1] − W0 X[4]

x[2] −W
0 X[2]

x[ 3] − W −W
2 0 X[ 6]

x[4] 0 X[1]
−W
x[5] 1 X[5]
−W
0
− W
x[ 6] −W0 X[ 3]
− W2
x[7] −W
3
− W −W
2 0 X[7]

10-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

We are asked to compute F ( 3 ) only. However, we will derive all equations as we did in Example
10.5.
Column 1 (C=1):
Y ( 0, 1 ) = Y ( 0, 0 ) + Y ( 4, 0 )
Y ( 1, 1 ) = Y ( 1, 0 ) + Y ( 5, 0 )
Y ( 2, 1 ) = Y ( 2, 0 ) + Y ( 6, 0 )
Y ( 3, 1 ) = Y ( 3, 0 ) + Y ( 7, 0 )
0
Y [ 4, 1 ] = W N [ Y ( 0, 0 ) – Y ( 4, 0 ) ] (1)
1
Y [ 5, 1 ] = W N [ Y ( 1, 0 ) – Y ( 5, 0 ) ]
2
Y [ 6, 1 ] = W N [ Y ( 2, 0 ) – Y ( 6, 0 ) ]
3
Y [ 7, 1 ] = W N [ Y ( 3, 0 ) – Y ( 7, 0 ) ]

Column 2 (C=2):
Y ( 0, 2 ) = Y ( 0, 1 ) + Y ( 2, 1 )
Y ( 1, 2 ) = Y ( 1, 1 ) + Y ( 3, 1 )
0
Y ( 2, 2 ) = W N [ Y ( 0, 1 ) – Y ( 2, 1 ) ]
2
Y ( 3, 2 ) = W N [ Y ( 1, 1 ) – Y ( 3, 1 ) ]
(2)
Y [ 4, 2 ] = Y ( 4, 1 ) + Y ( 6, 1 )
Y [ 5, 2 ] = Y ( 5, 1 ) + Y ( 7, 1 )
0
Y [ 6, 2 ] = W N [ Y ( 4, 1 ) – Y ( 6, 1 ) ]
2
Y [ 7, 2 ] = W N [ Y ( 5, 1 ) – Y ( 7, 1 ) ]

Column 3 (C=3):

Y ( 0, 3 ) = Y ( 0, 2 ) + Y ( 1, 2 )
0
Y ( 1, 3 ) = W N [ Y ( 0, 2 ) – Y ( 1, 2 ) ]
Y ( 2, 3 ) = Y ( 2, 2 ) + Y ( 3, 2 )
0
Y ( 3, 3 ) = W N [ Y ( 2, 2 ) – Y ( 3, 2 ) ]
(3)
Y [ 4, 3 ] = Y ( 4, 2 ) + Y ( 5, 2 )
0
Y [ 5, 3 ] = W N [ Y ( 4, 2 ) – Y ( 5, 2 ) ]
Y [ 6, 3 ] = Y ( 6, 2 ) + Y ( 7, 2 )
0
Y [ 7, 3 ] = W N [ Y ( 6, 2 ) – Y ( 7, 2 ) ]

Signals and Systems with MATLAB Applications, Second Edition 10-37

Orchard Publications
Chapter 10 The DFT and the FFT Algorithm

F ( 3 ) = X ( 3 ) = Y ( 6, 3 ) = Y ( 6, 2 ) + Y ( 7, 2 ) (4)
where
0
Y ( 6, 2 ) = W N [ Y ( 4, 1 ) – Y ( 6, 1 ) ]
and
2
Y ( 7, 2 ) = W N [ Y ( 5, 1 ) – Y ( 7, 1 ) ]

From (1)
0
Y [ 4, 1 ] = W N [ Y ( 0, 0 ) – Y ( 4, 0 ) ]
1
Y [ 5, 1 ] = W N [ Y ( 1, 0 ) – Y ( 5, 0 ) ]
2
Y [ 6, 1 ] = W N [ Y ( 2, 0 ) – Y ( 6, 0 ) ]
3
Y [ 7, 1 ] = W N [ Y ( 3, 0 ) – Y ( 7, 0 ) ]

and by substitution into (4)

0 3 2 5
F ( 3 ) = X ( 3 ) = W N Y ( 0, 0 ) + W N Y ( 1, 0 ) – W N Y ( 2, 0 ) – W N Y ( 3, 0 )
(5)
0 3 2 5
– W N Y ( 4, 0 ) – W N Y ( 5, 0) + W N Y ( 6, 0) + W N Y ( 7, 0)

From Exercise 4,
3 6 9
F ( 3 ) = X ( 3 ) = x ( 0 ) + WN x ( 1 ) + WN x ( 2 ) + WN x ( 3 )
(6)
12 15 18 21
+W N x ( 4 ) + W N x ( 5 ) + W N x ( 6 ) + W N x ( 7 )

8i + n n n±4 n
Since Y ( k, 0 ) = x ( k ) , W 8 = W 8 and W 8 = – W 8 (see proof below), we see that
6 2 9 5 12 0 15 3 18 2 21 5
W8 = –W8 , W8 = –W8 , W 8 = – W 8 , W 8 = – W 8 , W 8 = W 8 , and W 8 = W N . Therefore, (5)
and (6) are the same.

n±4 n
Proof that W 8 = –W8 :

n±4 ±4 – j2π ⁄ 8 ⋅ ( ± 4 )
= W 8 ⋅ ( cos π −
n n n n
W8 = W8 ⋅ W8 = W8 ⋅ e + sin π ) = – W 8

10-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Chapter 11
Analog and Digital Filters

his chapter is an introduction to analog and digital filters. It begins with the basic analog filters,
T transfer functions, and frequency response. The amplitude characteristics of Butterworth and
Chebychev filters and conversion of analog to equivalent digital filters using the bilinear trans-
formation is presented. It concludes with design examples using MATLAB.

11.1 Filter Types and Classifications

Analog filters are defined over a continuous range of frequencies. They are classified as low-pass,
high-pass, band-pass and band-elimination (stop-band). The ideal amplitude characteristics of each are
shown in Figure 11.1. The ideal characteristics are not physically realizable; we will see that practical
filters can be designed to approximate these characteristics.

V out V out
---------- ----------
V in V in
1 STOP 1
PASS (CUTOFF) STOP PASS
BAND BAND BAND BAND
ω ω
ωc ωc
Ideal low− pass filter Ideal high− pass filter

V out V out
---------- ----------
V in V in
1 1
STOP PASS STOP PASS STOP PASS
BAND BAND BAND BAND BAND BAND
ω ω
ω1 ω2 ω1 ω2
Ideal band− pass Filter Ideal band − elimination filter

Figure 11.1. Amplitude characteristics of the types of filters

Another, less frequently mentioned filter, is the all-pass or phase shift filter. It has a constant ampli-
tude response but is phase varies with frequency. Please refer to Exercise 4.
A digital filter, in general, is a computational process, or algorithm that converts one sequence of
numbers representing the input signal into another sequence representing the output signal. Accord-

Signals and Systems with MATLAB Applications, Second Edition 11-1

Orchard Publications
Chapter 11 Analog and Digital Filters
ingly, a digital filter can perform functions as differentiation, integration, estimation, and, of course,
like an analog filter, it can filter out unwanted bands of frequency.
Analog filter functions have been used extensively as prototype models for designing digital filters
and, therefore, we will present them first.

11.2 Basic Analog Filters

An analog filter can also be classified as passive or active. Passive filters consist of passive devices
such as resistors, capacitors and inductors. Active filters are, generally, operational amplifiers with
resistors and capacitors connected to them externally. We can find out whether a filter, passive or
active, is a low-pass, high-pass, etc., from its the frequency response that can be obtained from its
transfer function. The procedure is illustrated with the examples that follow.

Example 11.1
Derive expressions for the magnitude and phase responses of the series RC network of Figure 11.2,
and sketch their characteristics.

+ R +
V in C V out
−
−

Figure 11.2. Series RC network for Example 11.1

Solution:
1 ⁄ jωC
V out = --------------------------- V in
R + 1 ⁄ jωC

or
V out 1 1
G ( jω ) = ---------
- = ----------------------- = ------------------------------------------------------------------------
V in 1 + jωRC 2 2 2
( 1 + ω R C ) ∠ atan ( ωRC )
(11.1)
1
= --------------------------------- ∠– atan ( ωRC )
2 2 2
1+ω R C
The magnitude of (11.1) is

V out 1
G ( jω ) = ---------
- = --------------------------------- (11.2)
V in 1+ω R C
2 2 2

and the phase angle, also known as the argument, is

11-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Basic Analog Filters

V out
θ = arg { G ( jω ) } = arg ⎛ ----------⎞ = – atan ( ωRC ) (11.3)
⎝ V in ⎠

We can obtain a quick sketch for the magnitude G ( jω ) versus ω by evaluating (11.2) at ω = 0 ,
ω = 1 ⁄ RC , and ω → ∞ . Thus,

as ω → 0 ,
G(jω) ≅ 1
for ω = 1 ⁄ RC ,
G ( j ω ) = 1 ⁄ 2 = 0.707
and as ω → ∞ ,
G(jω) ≅ 0

We will use the MATLAB code below to plot G ( jω ) versus radian frequency ω . This is shown in
Figure 11.3 where, for convenience, we let RC = 1 .
w=0:0.02:10; RC=1; magGs=1./sqrt(1+w.*RC); semilogx(w,magGs); grid

Figure 11.3. Amplitude characteristics of a series RC low-pass filter with RC = 1

We can also obtain a quick sketch for the phase angle, i.e., θ = arg { G ( j ω ) } versus ω by evaluating
of (11.3) at ω = 0 , ω = 1 ⁄ RC , ω = – 1 ⁄ RC , ω → – ∞ and ω → ∞ . Thus,
As ω → 0 ,
θ ≅ – atan 0 ≅ 0°

Signals and Systems with MATLAB Applications, Second Edition 11-3

Orchard Publications
Chapter 11 Analog and Digital Filters

For ω = 1 ⁄ RC ,
θ = – atan 1 = – 45°
For ω = – 1 ⁄ RC ,
θ = – atan ( – 1 ) = 45°
As ω → – ∞ ,
θ = – atan ( – ∞ ) = 90°
and as ω → ∞ ,
θ = – atan ( ∞ ) = – 90 °

We will use the MATLAB code below to plot the phase angle θ versus radian frequency ω . This is
shown in Figure 11.3 where, for convenience, we let RC = 1 .
w=-8:0.02:8; RC=1; argGs=−atan(w.*RC).*180./pi; plot(w,argGs); grid

Figure 11.4. Phase characteristics of a series RC low-pass filter with RC = 1

Example 11.2
The network of Figure 11.5 is also a series RC circuit, where the positions of the resistor and capaci-
tor have been interchanged. Derive expressions for the magnitude and phase responses, and sketch
their characteristics.

+ C +
V in R V out
− −

Figure 11.5. RC network for Example 11.2

11-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Basic Analog Filters

Solution:
R
V out = --------------------------- V in
R + 1 ⁄ j ωC
or
2 2 2
V out
G ( j ω ) = ---------
j ωRC j ωRC + ω R C ωRC ( j + ωRC )
- = ------------------------ = ---------------------------------------- = --------------------------------------
V in 1 + j ωRC 1+ω R C
2 2 2
1+ω R C
2 2 2

(11.4)
2 2 2
ωRC 1 + ω R C ∠ atan ( 1 ⁄ ( ωRC ) -) = --------------------------------------------
= -------------------------------------------------------------------------------------------
1 - ∠ atan ( 1 ⁄ ( ωRC ) )
2 2 2
1+ω R C 2 2 2
1 + 1 ⁄ (ω R C )
The magnitude of (11.4) is

1
G ( j ω ) = --------------------------------------------- (11.5)
2 2 2
1 + 1 ⁄ (ω R C )

and the phase angle or argument, is

θ = arg { G ( j ω ) } = atan ( 1 ⁄ ωRC ) (11.6)

We can obtain a quick sketch for the magnitude G ( j ω ) versus ω by evaluating (11.5) at ω = 0 ,
ω = 1 ⁄ RC , and ω → ∞ . Thus,

As ω → 0 ,
G(jω) ≅ 0

For ω = 1 ⁄ RC ,
G ( j ω ) = 1 ⁄ 2 = 0.707
and as ω → ∞ ,
G(jω) ≅ 1

We will use the MATLAB code below to plot G ( jω ) versus radian frequency ω . This is shown in
Figure 11.5 where, for convenience, we let RC = 1 .

w=0:0.02:100; RC=1; magGs=1./sqrt(1+1./(w.*RC).^2); semilogx(w,magGs); grid

We can also obtain a quick sketch for the phase angle, i.e., θ = arg { G ( j ω ) } versus ω , by evaluating
(11.6) at ω = 0 , ω = 1 ⁄ RC , ω = – 1 ⁄ RC , ω → – ∞ , and ω → ∞ . Thus,
As ω → 0 ,
θ ≅ – atan 0 ≅ 0°

Signals and Systems with MATLAB Applications, Second Edition 11-5

Orchard Publications
Chapter 11 Analog and Digital Filters

Figure 11.6. Amplitude characteristics of a series RC high-pass filter with RC = 1

For ω = 1 ⁄ RC ,
θ = – atan 1 = – 45°
For ω = – 1 ⁄ RC ,
θ = – atan ( – 1 ) = 45°
As ω → – ∞ ,
θ = – atan ( – ∞ ) = 90°
and as ω → ∞ ,
θ = – atan ( ∞ ) = – 90 °

We will use the MATLAB code below to plot the phase angle θ versus radian frequency ω . This is
shown in Figure 11.7 where, for convenience, we let RC = 1 .
w=−8:0.02:8; RC=1; argGs=atan(1./(w.*RC)).*180./pi; plot(w,argGs); grid
Other low-pass, high-pass, band-pass, and band-elimination passive filters consist of combinations
of resistors, inductors, and capacitors. They are given as exercises in Chapter 4.

Example 11.3
The circuit of Figure 11.8 is an active low-pass filter and its magnitude G ( j ω ) versus ω was derived
and plotted in Example 4.7 of Chapter 4.

11-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Low-Pass Analog Filters

Figure 11.7. Phase characteristics of an RC high-pass filter with RC = 1

C2 10 nF
R2 40 K
R1 R3

200 K 50K
vin vout
C1
25 nF

Figure 11.8. Low-pass filter for Example 11.3

Using the derivation procedures of the examples we discussed thus far, we can analyze any circuit to
determine its behavior. However, as we found out in Example 4.7, the derivations are tedious. Fortu-
nately, several books containing filters and their characteristics, have been published. We will discuss
these later.
In our subsequent discussion we will be concerned with filter design.

11.3 Low-Pass Analog Filters

In this section, we will use the analog low-pass filter as a basis. We will see later that, using transfor-
mations, we can derive high-pass and the other types of filters from a basic low-pass filter. We will
discuss the Butterworth, Chebyshev and Cauer (elliptic) filters.

Signals and Systems with MATLAB Applications, Second Edition 11-7

Orchard Publications
Chapter 11 Analog and Digital Filters
The first step in the design of an analog low-pass filter is to find a suitable amplitude-squared function
2
A ( ω ) , and from it derive a G ( s ) function such that
2
A ( ω ) = G ( s ) ⋅ G ( –s ) s = jω
(11.7)

Since ( jω )∗ = ( – jω ) , the square of the magnitude of a complex number can be expressed as that
number and its complex conjugate. Thus, if the magnitude is A , then

A ( ω ) = G ( jω ) = G ( jω )G∗ ( jω ) = G ( jω ) ⋅ G ( – j ω )
2
(11.8)
Now, G ( jω ) can be considered as G ( s ) evaluated at s = jω , and thus (11.7) is justified. Also, since
A is understood to represent the magnitude, it needs not be enclosed in vertical lines.

Not all amplitude-squared functions can be decomposed to G ( s ) and G ( – s ) rational functions; only
even functions of ω , positive for all ω , and proper rational functions* can satisfy (11.7).

Example 11.4
It is given that
2
3s + 5s + 7
G ( s ) = ----------------------------
2
s + 4s + 6
2
Compute A ( ω ) .
Solution:
Since
2
3s + 5s + 7
G ( s ) = ----------------------------
2
s + 4s + 6
it follows that
2
3s – 5 s + 7
G ( – s ) = --------------------------
2
s –4 s + 6
and
2 2 4 2
3s + 5s + 7 3s – 5 s + 7 + 17s + 49
- ⋅ -------------------------- = 9s
G ( s ) ⋅ G ( – s ) = ---------------------------- -------------------------------------
-
2 2 4 2
s + 4s + 6 s – 4 s + 6 s – 4s + 36
Therefore,

* It was stated earlier, that a rational function is said to be proper if the largest power in the denominator is equal
to or larger than that of the numerator.

11-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Low-Pass Analog Filters
4 2 4 2
2 9s + 17s + 49 9ω – 17 ω + 49
A ( ω ) = G ( s ) ⋅ G ( –s ) s = jω
= -------------------------------------- = --------------------------------------
4 2 4 2
s – 4s + 36 s = jω
ω + 4ω + 36

2
The general form of the amplitude square function A ( ω ) is
2k 2k – 2
2 C ( bk ω + bk – 1 ω + … + b0 )
A ( ω ) = --------------------------------------------------------------------------------
- (11.9)
2k 2k – 2
ak ω + ak – 1 ω + … + a0

where C is the DC gain, a and b are constant coefficients, and k is a positive integer denoting the
2
order of the filter. Once the amplitude square function A ( ω ) is known, we can find G ( s ) from
2 2 2 2 2
(11.9) with the substitution ( jω ) = – ω = s or ω = – s , that is,
2
G ( s ) ⋅ G ( –s ) = A ( ω ) 2 2
(11.10)
ω = –s

In the simplest low-pass filter, the DC gain of the amplitude square function is unity. In this case
(11.9) reduces to
2 b0
A ( ω ) = -----------------------------------------------------------------------
- (11.11)
2k 2k – 2
ak ω + ak – 1 ω + … + a0

and at high frequencies reduces to

2 b0 ⁄ ak
A ( ω ) ≈ -------------
2k
- (11.12)
ω

The attenuation rate of this approximation is 6k dB ⁄ octave or 20k dB ⁄ decade . To understand

this, let us review the definitions of octave and decade.
Consider two frequencies u 1 and u 2 . Let

ω2
u 2 – u 1 = log 10 ω 2 – log 10 ω 1 = log 10 ------ (11.13)
ω1

If these frequencies are such that ω 2 = 2ω 1 , we say that these frequencies are separated by one
octave, and if ω 2 = 10ω 1 , they are separated by one decade.
To compute the attenuation rate of (11.12), we take the square root of both sides. Then,

b0 ⁄ ak Cons tan t B
A ( ω ) = -----------------
- = ------------------------ = ------ (11.14)
k k k
ω ω ω
Taking the common log of both sides of (11.14) and multiplying by 20, we get

Signals and Systems with MATLAB Applications, Second Edition 11-9

Orchard Publications
Chapter 11 Analog and Digital Filters
k
20log 10 A ( ω ) = 20log 10 B – 20log 10 ω = – 20klog 10 ω + 20log 10 B (11.15)
or
A ( ω ) dB = – 20klog 10 ω + 20log 10 B (11.16)

This is an equation of a straight line with slope = – 20k dB ⁄ decade , and intercept = B as shown
in Figure 11.9.

A ( ω ) dB
40

30 – 20 dB ⁄ decade = – 6 dB ⁄ octave
B
20

10
0 log 10 ω
1 10 100 1000

Figure 11.9. The – 20 dB ⁄ decade = – 6 dB ⁄ octave line

2
The procedure of finding the transfer function G ( s ) from the amplitude square function A ( ω ) , is
best illustrated with the following example.

Example 11.5
Given the amplitude square function
2
2 16 ( – ω + 1 )
A ( ω ) = ----------------------------------------- (11.17)
2 2
(ω + 4 )( ω + 9)

find a suitable transfer function G ( s )

Solution:
From (11.10),
2
2
G ( s )G ( – s ) = A ( ω ) 16 ( s + 1 )
= -------------------------------------------- (11.18)
2 2
ω = –s 2 2
( –s + 4 ) ( –s + 9 )

This function has zeros at s = ± j1 , and poles at s = ± 2 and s = ± 3 .

11-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Butterworth Analog Low-Pass Filters

There is no restriction on the zeros but, for stability*, we select the left-half s -plane poles.
We must also select the gain constant such that G ( 0 ) = A ( 0 ) .
Let
2
K( s + 1) -
G ( s ) = -------------------------------- (11.19)
(s + 2)(s + 3)

We must find K such that G ( 0 ) = A ( 0 ) . From (11.17),

2
A ( 0 ) = 16 ⁄ 36 = 4 ⁄ 9
or
A(0) = 2 ⁄ 3
From (11.19),
G(0) = K ⁄ 6
and for G ( 0 ) = A ( 0 ) we must have,
K⁄6 = 2⁄3
or
K = 12 ⁄ 3 = 4
By substitution into (11.19),
2
(s + 1)
G ( s ) = 4 ---------------------------------
( s + 2 )( s + 3 )

11.4 Design of Butterworth Analog Low-Pass Filters

We will consider the Butterworth low-pass filter whose amplitude-squared function is
2 1
A ( ω ) = -----------------------------------
- (11.20)
2k
( ω ⁄ ωC ) + 1

where k is a positive integer, and ω C is the cutoff ( 3 dB ) frequency. Figure 11.10 shows relation
(11.20) for k = 1, 2, 4, and 8 . The plot was created with the following MATLAB code.
w_w0=0:0.02:3; Aw2k1=sqrt(1./(w_w0.^2+1)); Aw2k2=sqrt(1./(w_w0.^4+1));...
Aw2k4=sqrt(1./(w_w0.^8+1)); Aw2k8=sqrt(1./(w_w0.^16+1));...
plot(w_w0,Aw2k1,w_w0,Aw2k2,w_w0,Aw2k4,w_w0,Aw2k8); grid

* Generally, a system is said to be stable if a finite input produces a finite output. Alternately, a system is stable if
the impulse response h ( t ) vanishes after a sufficiently long time. Stability is discussed in Feedback and Control
Systems textbooks.

Signals and Systems with MATLAB Applications, Second Edition 11-11

Orchard Publications
Chapter 11 Analog and Digital Filters

Figure 11.10. Butterworth low-pass filter amplitude characteristics

All Butterworth filters have the property that all poles of the transfer functions that describes them,
lie on a circumference of a circle of radius ω C , and they are 2π ⁄ 2k radians apart. Thus, if k = odd ,
the poles start at zero radians, and if k = even , they start at 2π ⁄ 2k . But regardless whether k is odd
or even, the poles are distributed in symmetry with respect to the jω axis. For stability, we choose
the left half-plane poles to form G ( s ) .
We can find the nth roots of a the complex number s by DeMoivre’s theorem. It states that

θ + 2kπ
j ⎛ -------------------⎞
jθ ⎝ n ⎠
n
re = n re k = 0, ± 1, ± 2, … (11.21)

Example 11.6
Derive the transfer function G ( s ) for the third order ( k = 3 ) Butterworth low-pass filter with nor-
malized cutoff frequency ω C = 1 rad ⁄ s .

Solution:
With k = 3 and ω c = 1 rad ⁄ s , (11.20) reduces to

2 1
A ( ω ) = --------------- (11.22)
6
ω +1

11-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Butterworth Analog Low-Pass Filters
2 2
With the substitution ω = – s , (11.22) becomes
1 -
G ( s ) ⋅ G ( – s ) = ----------------- (11.23)
6
–s +1

Then, s = 6 1 ∠0° and by DeMoivre’s theorem, with n = 6 ,

0 + 2kπ
j ⎛ -------------------⎞
j0 ⎝ 6 ⎠
6
1e = 6 1e k = 0, 1, 2, 3, 4 , 5
Thus,
s 1 = 1 ∠0° = 1
3
s 2 = 1 ∠60° = 1
--- + j -------
2 2
1 3
s 3 = 1 ∠120° = – --- + j -------
2 2
s 4 = 1 ∠180° = – 1
3
s 5 = 1 ∠240° = – 1
--- – j -------
2 2
1 3
s 6 = 1 ∠300° = --- – j -------
2 2

As expected, these six poles lie on the circumference of the circle with radius ω c = 1 as shown in
Figure 11.11.
Im { s }
s3 s
× ×2

s4 s1
× × Re { s }

s 5× ×
s6

Figure 11.11. Location of the poles for the transfer function of Example 11.6

The transfer function G ( s ) is formed with the left half-plane poles s 3 , s 4 , and s 5 . Then,

K
G ( s ) = ----------------------------------------------------------------------------------- (11.24)
⎛s + 1 --- – j -------⎞ ( s + 1 ) ⎛ s + 1
3
--- + j -------⎞
3
⎝ 2 2 ⎠ ⎝ 2 2 ⎠

Signals and Systems with MATLAB Applications, Second Edition 11-13

Orchard Publications
Chapter 11 Analog and Digital Filters
We use MATLAB to express the denominator as a polynomial.
syms s; den=(s+1/2−sqrt(3)*j/2)*(s+1)*(s+1/2+sqrt(3)*j/2)
den =
(s+1/2-1/2*i*3^(1/2))*(s+1)*(s+1/2+1/2*i*3^(1/2))
expand(den)
ans =
s^3+2*s^2+2*s+1
Therefore, (11.24) simplifies to
K
G ( s ) = ---------------------------------------- (11.25)
3 2
s + 2s + 2s + 1
2
The gain K is found from A ( 0 ) = 1 or A ( 0 ) = 1 and G ( 0 ) = K . Thus, K = 1 and
1
G ( s ) = ---------------------------------------- (11.26)
3 2
s + 2s + 2s + 1

and this is the transfer function G ( s ) for the third order ( k = 3 ) Butterworth low-pass filter with
normalized cutoff frequency ω C = 1 rad ⁄ s .

The general form of any analog low-pass (Butterworth, Chebyshev, Elliptic, etc.) filter is

b0
G(s) lp
= ---------------------------------------------------------------- (11.27)
k 2
ak s + … + a2 s + a1 s + a0

The pole locations and the coefficients of the corresponding denominator polynomials, have been
derived and tabulated by Weinberg in Network Analysis and Synthesis, McGraw-Hill.
Table 11.1 shows the first through the fifth order coefficients for Butterworth analog low-pass filter
denominator polynomials with normalized frequency ω C = 1 rad ⁄ s .

TABLE 11.1 Values for the coefficients a i in (11.27)

Coefficients of Denominator Polynomial for Butterworth Low-Pass Filters
Order a5 a4 a3 a2 a1 a0
1 1
2 1 1.4142136 1
3 1 2 2 1
4 2.6131259 3.1442136 2.6131259 1 1
5 1 3.2360680 5.2360680 5.2360680 3.2360680 1

11-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Butterworth Analog Low-Pass Filters

We can also use the MATLAB buttap and zp2tf functions. The first returns the zeros, poles, and
gain for an N – th order normalized prototype Butterworth analog low-pass filter. The resulting fil-
ter has N poles around the unit circle in the left half plane, and no zeros. The second performs the
zero-pole to transfer function conversion.

Example 11.7
Use MATLAB to find the numerator b and denominator a coefficients for the third-order Butter-
worth low-pass filter prototype with normalized cutoff frequency*.
Solution:
[z,p,k]=buttap(3); [b,a]=zp2tf(z,p,k)
b =
0 0 0 1
a =
1.0000 2.0000 2.0000 1.0000
We observe that the denominator coefficients are the same as in Table 11.1.
Table 11.2 shows factored forms of the denominator polynomials in terms of linear and quadratic
factors with normalized frequency ω C = 1 rad ⁄ s .

TABLE 11.2 Factored forms for Butterworth low-pass filters

Denominator in Factored form for Butterworth Low-Pass Filters with ω C = 1 rad ⁄ s
k Denominator of Equation (11.27)
1 s+1
2 2
s + 1.4142s + 1
3 2
(s + 1 )(s + s + 1)
4 2 2
( s + 0.7654s + 1 ) ( s + 1.8478s + 1 )
5 2 2
( s + 1 ) ( s + 0.6180s + 1 ) ( s + 1.6180s + 1 )
6 2 2 2
( s + 05176s + 1 ) ( s + 1.4142s + 1 ) ( s + 1.9318s + 1 )
7 2 2 2
( s + 1 ) ( s + 0.4449s + 1 ) ( s + 1.2465s + 1 ) ( s + 1.8022s + 1 )
8 2 2 2 2
( s + 0.3896s + 1 ) ( s + 1.1110s + 1 ) ( s + 1.6630s + 1 ) ( s + 1.9622s + 1 )

* Henceforth, normalized cutoff frequency will be understood to be ω C = 1 rad ⁄ s

Signals and Systems with MATLAB Applications, Second Edition 11-15

Orchard Publications
Chapter 11 Analog and Digital Filters
The equations shown in Table 11.2 can be derived from
1
G ( s ) = ----------------------------------------
- (11.28)
n–1

∏ ⎛⎝ --s-i – 1⎞⎠
( –1 )
n s

i=0

n
where the factor ( – 1 ) is to ensure that G ( 0 ) = 1 , and s i denotes the poles on the left half of the
s -plane. They can be found from

2i + 1 )π + j cos ( 2i + 1 )π⎞
s i = ω C ⎛ – sin (---------------------- ----------------------⎠ (11.29)
⎝ 2k 2k

We must remember that the factors in Table 11.2 apply only when the cutoff frequency is normal-
ized to ω C = 1 rad ⁄ s . If ω C ≠ 1 , we must scale the transfer function appropriately.

We can convert to the actual transfer function using the relation

ω norm × s
G ( s ) actual = G ⎛ ----------------------- ⎞
⎝ ω actual ⎠

and since, usually ω norm = 1 rad ⁄ s ,

G ( s ) actual = G ⎛ ---------------- ⎞
s
⎝ ω actual ⎠
(11.30)

that is, we replace s with s ⁄ ω actual

Quite often, we require that ω ≥ ω C , that is, in the stop band of the low-pass filter, the attenuation to
be larger than – 20 dB ⁄ decade , i.e., we require a sharper cutoff. As we have seen from the plots of
Figure 11.10, the Butterworth low-pass filter cutoff becomes sharper for larger values of k . Accord-
ingly, we generate the plot for different values of k shown in Figure 11.12 using the following MAT-
LAB code.
w_w0=1:0.02:10; dBk1=20.*log10(sqrt(1./(w_w0.^2+1)));...
dBk2=20.*log10(sqrt(1./(w_w0.^4+1))); dBk3=20.*log10(sqrt(1./(w_w0.^6+1)));...
dBk4=20.*log10(sqrt(1./(w_w0.^8+1))); dBk5=20.*log10(sqrt(1./(w_w0.^10+1)));...
dBk6=20.*log10(sqrt(1./(w_w0.^12+1))); dBk7=20.*log10(sqrt(1./(w_w0.^14+1)));...
dBk8=20.*log10(sqrt(1./(w_w0.^16+1))); semilogx(w_w0,dBk1,w_w0,dBk2,w_w0,dBk3,...
w_w0,dBk4,w_w0,dBk5,w_w0,dBk6,w_w0,dBk7,w_w0,dBk8); grid

11-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Butterworth Analog Low-Pass Filters

Figure 11.12. Attenuation for different values of k

Example 11.8
Using the attenuation curves of Figure 11.12, derive the transfer function of a Butterworth low-pass
analog filter with pass band bandwidth of 5 rad ⁄ s , and attenuation in the stop band at least
30 dB ⁄ decade for frequencies larger than 15 rad ⁄ s .
Solution:
We refer to Figure 11.12 and at ω ⁄ ω C = 15 ⁄ 5 = 3 , we see that the vertical line at this value crosses
the k = 3 curve at approximately – 28 dB , and the k = 4 curve at approximately – 37 dB . Since we
require that the attenuation be at least – 30 dB , we use the attenuation corresponding to the k = 4
curve. Accordingly, we choose a fourth order Butterworth low-pass filter whose normalized transfer
function, from Table 11.2, is
1
G ( s ) norm = ----------------------------------------------------------------------------------------- (11.31)
2 2
( s + 0.7654s + 1 ) ( s + 1.8478s + 1 )

and since ω C = 5 rad ⁄ s , we replace s with s ⁄ 5 . Then,

1
G ( s ) actual = -------------------------------------------------------------------------------------------------
2 2
⎛ -----
s
- + ------------------- + 1 ⎞ ⎛ ------ + ------------------- + 1⎞
0.7654s s 1.8478s
⎝ 25 5 ⎠ ⎝ 25 5 ⎠

Signals and Systems with MATLAB Applications, Second Edition 11-17

Orchard Publications
Chapter 11 Analog and Digital Filters
or
625
G ( s ) actual = --------------------------------------------------------------------------------------------------
2 2
( s + 3.8270s + 25 ) ( s + 9.2390s + 25 )
(11.32)
625
= ------------------------------------------------------------------------------------------------------------
4 3 2
s + 13.066s + 85.358s + 326.650s + 625
As a last step, we wish to know how to design a circuit (passive or active), that will satisfy a transfer
function such as the one above. Fortunately, the work for us has been done by others who have
developed analog filter prototypes, both passive and active.
Some good references are:
Electronic Filter Design Handbook, Williams and Taylor, McGraw-Hill
Electronic Engineers’ Handbook, Fink and Christiansen, McGraw-Hill
Reference Data for Engineers Handbook, Van Valkenburgh, Howard Sams
As an example, the Reference Data for Engineers Handbook provides the circuit of Figure 11.13 that is
known as Second Order Voltage Controlled Voltage Source (VCVS) low-pass filter.
R4

R3

C1 vout
R1 R2

vin C2

Figure 11.13. VCVS low-pass filter

The transfer function of the second order VCVS low-pass filter of Figure 11.13 is given as
2
Kbω c
G ( s ) = --------------------------------------
- (11.33)
2 2
s + aω c s + bω c

This is referred to as a second order all-pole* approximation to the ideal low-pass filter with cutoff
frequency ω C , where K is the gain, and the coefficients a and b are provided by tables.
For a non-inverting positive gain K , the circuit of Figure 11.13 satisfies the transfer function of
(11.33) with the conditions that

* The terminology “all-pole” stems from the fact that the s−plane contains poles only and the zeros are at ± ∞ ,
that is, the s−plane is all poles and no zeros.

11-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Butterworth Analog Low-Pass Filters

2
R 1 = ----------------------------------------------------------------------------------------------------------------- (11.34)
⎧ 2 2 ⎫
⎨ aC 2 + ( [ a + 4b ( K – 1 ) ]C 2 – 4bC 1 C 2 ) ⎬ ω c
⎩ ⎭

1
R 2 = ----------------------------
- (11.35)
2
bC 1 C 2 R 1 ω c

K ( R1 + R2 )
R 3 = --------------------------
- K≠1 (11.36)
(K – 1)

R4 = K ( R1 + R2 ) (11.37)

From (11.36) and (11.37), we observe that K = 1 + R 4 ⁄ R 3 .

A fourth-order all-pole low-pass filter transfer function is a ratio of a constant to a fourth degree
polynomial. A practical method of obtaining a fourth order transfer function, is to factor it into two
second-order transfer functions of the form of (11.33), i.e.,
2 2
K1 b1 ωc K2 b2 ωc
G ( s ) = -------------------------------------------
2
- ⋅ -------------------------------------------
2 2
-
2
(11.38)
s + a1 ωc s + b1 ωc s + a2 ωc s + b2 ωc

Each factor in (11.38) can be realized by a stage (circuit). Then, the two stages can be cascaded as
shown in Figure 11.14.

Stage 1 Stage 2
vin vout

Figure 11.14. Cascaded stages

Table 11.3 lists the Butterworth low-pass coefficients for second and fourth-order designs, where a
and b apply to the transfer functions of (11.33) and (11.38).
For a practical design of a second-order VCVS circuit, we select standard values for capacitors C 1
and C 2 of the circuit of Figure 11.13, we substitute the appropriate values for the coefficients a and
b from Table 11.3, we choose desired values for the gain K and cutoff frequency ω C , and we sub-
stitute these in (11.34) through (11.37) to find the values of the resistors R 1 through R 4 .

Signals and Systems with MATLAB Applications, Second Edition 11-19

Orchard Publications
Chapter 11 Analog and Digital Filters

TABLE 11.3 Coefficients for Butterworth low-pass filter designs

Coefficients for Second and Fourth Order Butterworth Low-Pass Filter Designs
Order
a 1.41421
2
b 1.0000
a1 0.76537
b1 1.0000
4
a2 1.84776
b2 1.0000

Example 11.9
Design a second-order VCVS Butterworth low-pass filter with gain K = 2 and cutoff frequency
f C = 1 kHz .

Solution:
We will use the second order VCVS prototype op amp circuit of Figure 11.13, with capacitance val-
–8
ues C 1 = C2 = 0.01 µF = 10 F . From Table 11.3, a = 1.41421 = 2 and b = 1 .

We substitute these values into (11.34) through (11.37), to find the values of the resistors.
We use MATLAB to do the calculations as follows:
C1=10^(-8); C2=C1; a=sqrt(2); b=1; K=2; wc=2*pi*10^3;
% and from (11.34) through (11.37)
R1=2/((a*C2+sqrt((a^2+4*b*(K−1))*C2^2-4*b*C1*C2))*wc);
R2=1/(b*C1*C2*R1*wc^2); R3=K*(R1+R2)/(K-1); R4=K*(R1+R2); fprintf(' \n');...
fprintf('R1 = %6.0f \t',R1); fprintf('R2 = %6.0f \t',R2);...
fprintf('R3 = %6.0f \t',R3); fprintf('R4 = %6.0f \t',R4)
R1 = 11254 R2 = 22508 R3 = 67524 R4 = 67524
These are the calculated values but they are not standard resistor values; we must select standard
resistor values as close as possible to the calculated values.
It will be interesting to find out what the frequency response of this filter looks like, with capacitors
C 1 = C 2 = 0.01 µF a n d s t a n d a r d 1% t o l e r a n c e r e s i s t o r s w i t h va l u e s R 1 = 11.3 KΩ ,
R 2 = 2 × R 1 = 22.6 KΩ , and R 3 = R 4 = 68.1 KΩ .

We now substitute these values into the equations of (11.34) through (11.37), and we solve the first

11-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Butterworth Analog Low-Pass Filters

equation of this set for the cutoff frequency ω C . Then, we use ω C with the transfer function of
(11.33). We do this with the following MATLAB code which produces the plot of Figure 11.15.
f=1:10:5000; R1=11300; R2=22600; R3=68100; R4=R3; C1=10^(−8); C2=C1;
a=sqrt(2); b=1; w=2*pi*f; fc=sqrt(1/(b*R1*R2*C1*C2))/(2*pi); wc=2*pi*fc;
K=1+R3/R4; s=w*j; Gw=(K.*b.*wc.^2)./(s.^2+a.*wc.*s+b.*wc.^2); magGw=abs(Gw);
semilogx(f,magGw); grid; hold on; xlabel('Frequency Hz'); ylabel('|Vout/Vin|');
title ('2nd Order Butterworth Low-Pass Filter Response')
The frequency response of this low-pass filter is shown in Figure 11.15.

Figure 11.15. Plot for the VCVS low-pass filter of Example 11.9

We see that the cutoff frequency occurs at about 1 KHz .

We have used the MATLAB buttap function earlier to aid us in the design of Butterworth filters
with the cutoff frequency normalized to 1 rad ⁄ s . We can also use the bode function to display
both the (asymptotic) magnitude and phase plots. The following code will produce the Bode magni-
tude and phase plots for a two-pole Butterworth low-pass filter.
[z,p,k]= buttap(2); % Specify a two-pole filter
[b,a]=zp2tf(z,p,k); % Display in polynomial rational form
w=0:0.01:4; [mag,phase]=bode(b,a,w);
b % Display b coefficients
b =
0 0 1

Signals and Systems with MATLAB Applications, Second Edition 11-21

Orchard Publications
Chapter 11 Analog and Digital Filters
a % Display a coefficients
a =
1.0000 1.4142 1.0000
num=[0 0 1]; den=[1 sqrt(2) 1];
bode(num,den); title('Butterworth 2nd Order Low-Pass Filter')
The Bode plots are shown in Figure 11.16.

Figure 11.16. Bode plots for example 11.9

11.5 Design of Type I Chebyshev Analog Low-Pass Filters

The Type I Chebyshev filters are based on approximations derived from the Chebyshev polynomials
C k ( x ) which constitute a set of orthogonal functions. The coefficients of these polynomials are tab-
ulated in math tables. See, for example, the Handbook of Mathematical Functions, Dover Publications.
These polynomials are derived from the equations
–1
C k ( x ) = cos ( kcos x ) ( x ≤ 1) (11.39)

and
–1
C k ( x ) = cos h ( kcosh x ) x >1 (11.40)

From (11.39) with k = 0 , we get

–1
C 0 ( x ) = cos ( 0cos x ) = 1 (11.41)

11-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Type I Chebyshev Analog Low-Pass Filters

With k = 1 ,
–1
C 1 ( x ) = cos ( 1cos x ) = x * (11.42)

With k = 2
–1 2
C 2 ( x ) = cos ( 2cos x ) = 2x – 1 (11.43)
–1
and this is shown by letting cos x = α . Then,
2 2 –1
C 2 ( x ) = cos ( 2α ) = 2cos α – 1 = 2cos ( cos x ) – 1
–1 –1
cos ( cos x ) cos ( cos x ) 2
= 2 – 1 = 2x – 1
⎧
⎪
⎨
⎪
⎩
⎧
⎪
⎨
⎪
⎩
x x

We can also use MATLAB to convert these trigonometric functions to algebraic polynomials. For
example,
syms x; expand(cos(2*acos(x)))
ans =
2*x^2-1
Using this iterated procedure we can show that with k = 3, 4, and 5 , we get
3
C 3 ( x ) = 4x – 3x
4 2
C 4 ( x ) = 8x – 8x + 1 (11.44)
5 3
C 5 ( x ) = 16x – 20x + 5x

and so on.
We observe that for k = even , C k ( x ) = even , and for k = odd , C k ( x ) = odd .

The curves representing these polynomials are shown in Figure 11.17.

The Type I Chebyshev low-pass filter amplitude-square function is defined as
2 α
A ( ω ) = --------------------------------------------
- (11.45)
2 2
1 + ε Ck ( ω ⁄ ωC )

2
The quantity ε is a parameter chosen to provide the desired passband ripple and α is a constant
chosen to determine the desired DC gain.

–1
* We recall that if x = cos y , then y = cos x , and cos y = x .

Signals and Systems with MATLAB Applications, Second Edition 11-23

Orchard Publications
Chapter 11 Analog and Digital Filters

Type I Chebyshev Polynomials, k=0 through k=5

6
k=4
5
k=5 k=3 k=2
4

3
Ck(x)

k=1
2

1
k=0
0

-1
0.0 0.5 1.0 1.5 2.0
x

Figure 11.17. Type I Chebyshev polynomials

The parameter α in (11.45) is a constant representing the DC gain, ε is a parameter used in deter-
mining the ripple in the pass-band, the subscript k denotes both the degree of the Type I Chebyshev
polynomial and the order of the transfer function, and ω C is the cutoff frequency. This filter pro-
duces a sharp cutoff rate in the transition band.
Figure 11.18 shows Type I Chebyshev amplitude frequency responses for k = 3 and k = 4 .

Pass-Band in Type I Chebyshev Filters

α

α/(1+ε2)1/2
Amplitude

k=odd k=even
(k=3) (k=4)

0 1 2
ω/ωc

Figure 11.18. Chebyshev Type I Chebyshev low-pass filter for even and odd values of k .

2
The magnitude at ω = 0 is α when k = odd and α ⁄ 1 + ε when k = even . This is shown in
Figure 11.18. The cutoff frequency is the largest value of ω C for which

11-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Type I Chebyshev Analog Low-Pass Filters

1
A ( ω C ) = ------------------ (11.46)
2
1+ε
Stated in other words, the pass-band is the range over which the ripple oscillates with constant
bounds; this is the range from DC to ω C . From (11.46), we observe that only when ε = 1 the mag-
nitude at the cutoff frequency is 0.707 i.e., the same as in other types of filters. But when 0 < ε < 1 ,
the cutoff frequency is greater than the conventional 3 dB cutoff frequency ω C .

Table 11.4 gives the ratio of the conventional cutoff frequency f 3 dB to the ripple width frequency
f C of a Type I Chebyshev low-pass filter.

TABLE 11.4 Ratio of conventional cutoff frequency to ripple width frequency

Ratio of Conventional f3 dB Cutoff Frequency to Ripple
Width for Low-Pass Chebyshev Filters
Ripple Width f3 dB ⁄ fc

dB k=2 k=4
0.1 1.943 1.213
0.5 1.390 1.093
1.0 1.218 1.053

The pass-band ripple r in dB , is defined as

2
A max A max
r dB = 10log 10 -----------
2
= 20log 10 ----------- (11.47)
A min A min

where A max and A min are the maximum and minimum values respectively of the amplitude A in the
pass-band interval. From (11.45),
2 α
A ( ω ) = --------------------------------------------
- (11.48)
2 2
1 + ε Ck ( ω ⁄ ωC )

2 2 2
and A max occurs when ε C k ( ω ⁄ ω C ) = 0 . Then,
2
A max = α (11.49)
2
To find A min , we must first confirm that
2
Ck ( ω ⁄ ωC ) ≤ 1

Signals and Systems with MATLAB Applications, Second Edition 11-25

Orchard Publications
Chapter 11 Analog and Digital Filters
This can be shown to be true by (11.39), that is,
–1
C k ( x ) = cos ( kcos x ) x ≤1
or
C k ( x ) ≤ 1 for – 1 ≤ x ≤ 1
Therefore,
2
C k ( ω ⁄ ω C ) max = 1
and
2 α -
A min = ------------- (11.50)
2
1+ε
Substitution of (11.49) and (11.50) into (11.47) yields
2
A max α 2
r dB = 10log 10 -----------
2
= 10log 10 --------------------------
2
- = 10log 10 ( 1 + ε ) (11.51)
A min α ⁄ (1 + ε )

or
2 r dB
log 10 ( 1 + ε ) = ------
-
10
or
2 r dB ⁄ 10
1 + ε = 10
or

2 r dB ⁄ 10
ε = 10 –1 (11.52)

We have seen that when k = odd , there is a maximum at ω = 0 . At this frequency, (11.45) reduces
to
2
A (0) = α (11.53)
and for a unity gain, α = 1 when k = odd .
2
However, for unity gain when k = even , we must have α = 1 + ε . This is because at ω = 0 , we
must have C k ( 0 ) = 1 in accordance with(11.41). Then, the relation

2 α
A ( ω ) = ---------------------------------------------
2 2
1 + ε Ck ( ω ⁄ ωC )
reduces to
2 α
A ( 0 ) = ---------------------------- α -=1
- = -------------
2 2 2
1 + ε Ck ( 0 ) 1 + ε

11-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Type I Chebyshev Analog Low-Pass Filters

or
2
α = 1+ε

For this choice of α , the amplitude response at maxima, corresponds to

2
2 1+ε
A ( ω max ) = ----------------------------------------------------
-
2 2
1 + ε C k ( ω max ⁄ ω C )

and this will be maximum when

2
C k ( ω max ⁄ ω C ) = 0
resulting in
2
2 1+ε 2
A ( ω max ) = -------------- = 1 + ε
1
or
2
A ( ω max ) = 1+ε

Example 11.10
Derive the transfer function G ( s ) for the k = 2 , Type I Chebyshev function that has pass-band rip-
ple r dB = 1 dB , unity DC gain, and normalized cutoff frequency at ω C = 1 rad ⁄ s .

Solution:
From (11.45),
2 α
A ( ω ) = --------------------------------------------- (11.54)
2 2
1 + ε Ck ( ω ⁄ ωC )

2
and since k = even , for unity DC gain, we must have α = 1 + ε . Then, (11.54) becomes
2
2 1+ε
A ( ω ) = ---------------------------------------------
2 2
1 + ε Ck ( ω ⁄ ωC )
For k = 2
2
C 2 ( x ) = 2x – 1
and
2 2 2 2 4
C k ( ω ⁄ ω C ) = C k ( ω ) = ( 2ω – 1 ) = 4ω – 4ω + 1
Also, from (11.52),
2 r dB ⁄ 10 1 ⁄ 10
ε = 10 – 1 = 10 – 1 = 1.259 – 1 = 0.259
Then,

Signals and Systems with MATLAB Applications, Second Edition 11-27

Orchard Publications
Chapter 11 Analog and Digital Filters

2 1 + 0.259
A ( ω ) = ------------------------------------------------------------ 1.259
= -----------------------------------------------------------------
-
4 4 2
1 + 0.259 ( 4ω – 4ω + 1 ) 1.036ω – 1.036ω + 1.259
2 2
and with ω = – s ,
1.259
G ( s )G ( – s ) = --------------------------------------------------------------
-
4 2
1.036s + 1.036s + 1.259
We find the poles from the roots of the denominator using MATLAB.
d=[1.036 0 1.036 0 1.259]; p=roots(d); fprintf(' \n'); disp('p1 = '); disp(p(1));...
disp('p2 = '); disp(p(2)); disp('p3 = '); disp(p(3)); disp('p4 = '); disp(p(4))
p1 =
-0.5488 + 0.8951i
p2 =
-0.5488 - 0.8951i
p3 =
0.5488 + 0.8951i
p4 =
0.5488 - 0.8951i
We now form the transfer function from the left half-plane poles p 1 = – 0.5488 + j0.8951 and
p 2 = – 0.5488 – j 0.8951 . Then,

K
G ( s ) = ------------------------------------ K
- = -------------------------------------------------------------------------------------------------------------
-
( s – p1 ) ( s – p2 ) ( s + 0.5488 – j0.8951 ) ( s + 0.5488 + j0.8951 )

We will use MATLAB to multiply the factors of the denominator.

syms s; den=(s+0.5488-0.8951*j)*(s+0.5488+0.8951*j); simple(expand(den))
ans =
s^2+686/625*s+22047709/20000000
686/625
ans =
1.0976
22047709/20000000
ans =
1.1024

11-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Type I Chebyshev Analog Low-Pass Filters

Thus,
K
G ( s ) = -----------------------------------------------------
2
s + 1.0976s + 1.1024
and at s = 0 ,
K -
G ( 0 ) = ---------------
1.1024
2
Also, A ( 0 ) = 1 , A ( 0 ) = 1
Then,
K
G ( 0 ) = A ( 0 ) = ---------------
- = 1
1.1024
or
K = 1.1024
Therefore, the transfer function for Example 11.10 is
1.1024
G ( s ) = ----------------------------------------------------
-
2
s + 1.0976s + 1.1024
We can plot the attenuation band for Type I Chebyshev filters, as we did with the Butterworth filters
in Figure 11.12, but we need to construct one for each value of dB in the ripple region. Instead, we
will develop the following procedure.
We begin with the Chebyshev approximation
2 α
A ( ω ) = --------------------------------------------- (11.55)
2 2
1 + ε Ck ( ω ⁄ ωC )

and, for convenience, we let α = 1 . If we want the magnitude of this to be less than some value β
2
for ω ≥ ω C , we should choose the value of k in C k ( ω ⁄ ω C ) so that

1
--------------------------------------------- ≤ β 2 (11.56)
2 2
1 + ε Ck ( ω ⁄ ωC )

that is, we need to find a suitable value of the integer k so that (11.56) will be satisfied. As we have
already seen from (11.52), the value of ε can be determined from
2 r dB ⁄ 10
ε = 10 –1
once the band-pass ripple has been specified.
Next, we need to find G ( s ) from
2
A ( ω ) = G ( s )G ( – s ) s = jω

Signals and Systems with MATLAB Applications, Second Edition 11-29

Orchard Publications
Chapter 11 Analog and Digital Filters

and if we replace ω by s ⁄ j in (11.55) where α = 1 , we get

2 1
G ( s ) = --------------------------------------------- (11.57)
2 2
1 + ε C k ( s ⁄ jω C )

It can be shown that the poles of the left half of the s -plane are given by

2i + 1 )π + jc cos (----------------------
s i = ω C – b sin (---------------------- 2i + 1 )π (11.58)
2k 2k

for i = 0, 1, 2, …, 2k – 1
The constants b and c in (11.58) can be evaluated from
–1
m–m
b = ------------------- (11.59)
2
and
–1
c = m +m
------------------- (11.60)
2
where
–2 –1 1 ⁄ k
m = ( 1+ε +ε ) (11.61)
The transfer function is then computed from
k
( –1 )
G ( s ) = ---------------------------
k–1
(11.62)
⎛ --s- – 1⎞
∏ ⎝s
i
⎠
i=0

Example 11.11
Design a Type I Chebyshev analog low-pass filter with 3 dB band-pass ripple and ω C = 5 rad ⁄ s .
The attenuation for ω ≥ 15 rad ⁄ s must be at least 30 dB ⁄ decade .
Solution:
From (11.52),
2 r dB ⁄ 10 3 ⁄ 10
ε = 10 – 1 = 10 – 1 = 1.9953 – 1 ≈ 1

and the integer k must be chosen such that

1 - ≤ – 30
10log 10 ------------------------------------
2
1 + C k ( 15 ⁄ 5 )

11-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Type I Chebyshev Analog Low-Pass Filters

or
2
– 10 log 10 ( 1 + C k ( 3 ) ) ≤ – 30

2
– log 10 ( 1 + C k ( 3 ) ) ≤ – 3

2 3
1 + C k ( 3 ) ≥ 10

To find the minimum value of k which satisfies this inequality, we compute the Chebyshev polyno-
mials for k = 0, 1, 2, 3, … . From (11.41) through (11.44), we get
2
C0 ( 3 ) = 1

2 2
C1 ( 3 ) = 3 = 9

2 2 2 2
C 2 ( 3 ) = ( 2 ⋅ 3 – 1 ) = 17 = 289

2 3 2 2
C 3 ( 3 ) = ( 4 ⋅ 3 – 3 ⋅ 3 ) = 99 = 9801

2 2 3
and since C k ( 3 ) must be such that 1 + Ck ( 3 ) ≥ 10 , we choose k = 3 . Next, to find the poles of
left half of the s -plane we first need to compute m , b , and c . From (11.61),

⎛ 1⁄3
–2 –1 1 ⁄ k 1 1 ⎞ 1⁄3
m = ( 1+ε +ε ) = ⎜ 1 + ----- + ---------⎟ = ( 2 + 1)
⎝ ε
2
ε⎠
2

or
m = 1.3415
and
–1
m = 0.7454
Then, from (11.59) and (11.60),
1.3415 – 0.7454
b = --------------------------------------- = 0.298
2

1.3415 + 0.7454
c = --------------------------------------- = 1.043
2

and the poles for i = 0, 1, and 2 are found from (11.58), that is

s i = ω C – b sin (----------------------
2i + 1 )π + jc cos ( 2i + 1 )π
----------------------
2k 2k

Thus, the poles for this example are

Signals and Systems with MATLAB Applications, Second Edition 11-31

Orchard Publications
Chapter 11 Analog and Digital Filters

s 0 = 5 ⎛ – 0.298 sin π π
--- + j1.043 cos ---⎞⎠ = – 0.745 + j4.516
⎝ 6 6

s 1 = 5 ⎛ – 0.298 sin π π
--- + j1.043 cos ---⎞⎠ = – 1.49
⎝ 2 2

s 2 = 5 ⎛ – 0.298 sin 5π
------ + j1.043 cos ------⎞⎠ = – 0.745 – j4.516
5π
⎝ 6 6

Therefore, by substitution into (11.62) we get

3
( –1 ) – ( – 1.49 ) ( – 0.745 + j4.516 ) ( – 0.745 – j4.516 )
G ( s ) = -------------------------------------------------------------------------- = --------------------------------------------------------------------------------------------------------------------------
( s ⁄ s0 – 1 ) ( s ⁄ s1 – 1 ) ( s ⁄ s2 – 1 ) ( s + 1.49 ) ( s + 0.745 – j4.516 ) ( s + 0.745 + j4.516 )

We will use MATLAB to do these computations.

−(−1.49)*(−0.745+j*4.516)*(−0.745−j*4.516)
ans =
31.2144
syms s; den=(s+1.49)*(s+0.745−j*4.516)*(s+0.745+j*4.516); simple(expand(den))
ans =
s^3+149/50*s^2+23169381/1000000*s+3121442869/100000000
Then,
31.214
G ( s ) = -----------------------------------------------------------------------------
- (11.63)
3 2
s + 2.980s + 23.169s + 31.214

To verify that the derived transfer function G ( s ) of (11.63) satisfies the filter specifications, we use
the MATLAB code below to plot G ( jω ) .

w=0:0.01:100; s=j*w; Gs=31.214./(s.^3+2.98.*s.^2+23.169.*s+31.214);...

magGs=abs(Gs); dB=20.*log10(magGs); semilogx(w,dB); grid; hold on;
plot(w,magGs); grid; hold on; xlabel('Radian Frequency w rad/s');
ylabel('|G(w)| in dB'); title('Magnitude of G(w) versus Radian Frequency')
The plot is shown in Figure 11.19.

We can use the MATLAB cheb1ap function to design a Type I Chebyshev analog low-pass filter.
Thus, the [z,p,k] = cheb1ap(N,Rp) statement where N denotes the order of the filter, returns the
zeros, poles, and gain of an N – th order normalized prototype Type I Chebyshev analog lowpass fil-
ter with ripple Rp decibels in the pass band.

11-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Design of Type I Chebyshev Analog Low-Pass Filters

Figure 11.19. Plot for Example 11.11

Example 11.12
Use the MATLAB cheb1ap function to design a second order Type I Chebyshev low-pass filter
with 3 dB ripple in the pass band.
Solution:
We use the code
w=0:0.05:400; % Define range to plot
[z,p,k]=cheb1ap(2,3);
[b,a]=zp2tf(z,p,k); % Convert zeros and poles of G(s) to polynomial form
[mag,phase]=bode(b,a,w); hold on
b % Display the b coefficients
b =
0 0 0.5012
a % Display the a coefficients
a =
1.0000 0.6449 0.7079
Now, with the known values of a and b we use the bode function to produce the Bode plots as fol-
lows.

Signals and Systems with MATLAB Applications, Second Edition 11-33

Orchard Publications
Chapter 11 Analog and Digital Filters
bode(b,a), title('Bode Plot for Type 1 Chebyshev Low-Pass Filter')
This is shown in Figure 11.20.

Figure 11.20. Bode plots for the filter of Example 11.12

On the Bode plots shown in Figure 11.20, the ripple is not so obvious. The reason is that this is a
Bode plot with straight line approximations. To see the ripple, we use the following code:
w=0:0.01:10; [z,p,k]=cheb1ap(2,3); [b,a]=zp2tf(z,p,k); Gs=freqs(b,a,w);...
xlabel('Frequency in rad/s'), ylabel('Magnitude of G(s)'),...
semilogx(w,abs(Gs)); title('Type 1 Chebyshev Low-Pass Filter'), grid
The generated plot is shown in Figure 11.21.

11.6 Other Low-Pass Filter Approximations

We will briefly discuss two other filter types, the Inverted Chebyshev, and the Cauer or Elliptic.
The Inverted Chebyshev, also known as Type II Chebyshev, is characterized by the following ampli-
tude-square approximation.
2 2
2 ε C k ( ω C ⁄ ω)
A ( ω ) = -----------------------------------------
- (11.64)
2 2
1 + ε C k ( ω C ⁄ ω)

and has the ripple in the stop-band as opposed to Type I which has the ripple in the pass-band. In
(11.64), the frequency ω C defines the beginning of the stop band.

11-34 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Other Low-Pass Filter Approximations

Figure 11.21. Amplitude characteristics for the filter of Example 11.12

The characteristics of a typical Type II Chebyshev low-pass filter are shown in Figure 11.22.

A(ω)
k=4

1 dB ripple
in stop band

Figure 11.22. Type II Chebyshev low-pass filter

We can design Type II Chebyshev low-pass filters with the MATLAB cheb2ap function. Thus, the
statement [z,p,k] = cheb2ap(N,Rs) where N denotes the order of the filter, returns the zeros,
poles, and gain of an N – th order normalized prototype Type II Chebyshev analog lowpass filter
with ripple Rs decibels in the stop band.

Example 11.13
Using the MATLAB cheb2ap function, design a third order Type II Chebyshev analog filter with
3 dB ripple in the stop band.

Signals and Systems with MATLAB Applications, Second Edition 11-35

Orchard Publications
Chapter 11 Analog and Digital Filters
Solution:
We use the code
w=0:0.01:1000; [z,p,k]=cheb2ap(3,3); [b,a]=zp2tf(z,p,k); Gs=freqs(b,a,w);...
semilogx(w,abs(Gs)); xlabel('Frequency in rad/sec'); ylabel('Magnitude of G(s)');
title('Type 2 Chebyshev Low-Pass Filter, k=3, 3 dB ripple in stop band'); grid
The plot for this filter is shown in Figure 11.23.

Figure 11.23. Plot for the filter of Example 11.13

The elliptic (Cauer) filters are characterized by the low-pass amplitude-squared function
2 1
A ( ω ) = ------------------------------------ (11.65)
2
1 + Rk ( ω ⁄ ωC )

where R k ( x ) represents a rational elliptic function used with elliptic integrals. Elliptic filters have rip-
ple in both the pass-band and the stop-band as shown in Figure 11.24.
We can design elliptic low-pass filters with the MATLAB ellip function. The statement [b,a] =
ellip(N,Rp,Rs,Wn,’s’) where N is the order of the filter, designs an N – th order low-pass filter with
ripple Rp decibels in the pass band, ripple Rs decibels in the stop band, Wn is the cutoff frequency,
and ’s’ is used to specify analog elliptic filters. If ’s’ is not included in the above statement, MAT-
LAB designs a digital filter. The plot of Figure 11.24 was obtained with the following MATLAB
code:
w=0: 0.05: 500; [z,p,k]=ellip(5, 0.6, 20, 200, 's'); [b,a]=zp2tf(z,p,k);...
Gs=freqs(b,a,w); plot(w,abs(Gs)), title('5-pole Elliptic Low Pass Filter'); grid

11-36 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Other Low-Pass Filter Approximations

Figure 11.24. Characteristics of an elliptic low-pass filter

Example 11.14
Use MATLAB to design a four-pole elliptic analog low-pass filter with 0.5 dB maximum ripple in
the pass-band and 20 dB minimum attenuation in the stop-band with cutoff frequency at
200 rad ⁄ s .
Solution:
The solution is obtained with the following MATLAB code.
w=0: 0.05: 500; [z,p,k]=ellip(4, 0.5, 20, 200, 's'); [b,a]=zp2tf(z,p,k);...
Gs=freqs(b,a,w); plot(w,abs(Gs)), title('4-pole Elliptic Low Pass Filter'); grid
The plot for this example is shown in Figure 11.25.
To form the transfer function G ( s ) , we need to know the coefficients a i and b i of the denominator
and numerator respectively, of G ( s ) in descending order. Because these are large numbers, we use
the format long MATLAB command, and we get
format long
a

Signals and Systems with MATLAB Applications, Second Edition 11-37

Orchard Publications
Chapter 11 Analog and Digital Filters

Figure 11.25. Plot for filter of Example 11.14

a =
1.0e+009 *
Columns 1 through 4
0.00000000100000 0.00000033979343 0.00010586590805
0.01618902078998
Column 5
2.07245309396647
b
b =
1.0e+009 *
Columns 1 through 4
0.00000000010003 0 0.00003630258930 0
Column 5
2.04872991957189
Thus, the transfer function for this filter is
9
2.0487 × 10
G ( s ) = ----------------------------------------------------------------------------------------------------------------------------------------- (11.66)
4 3 2 6 9
s + 339.793s + 105866s + 16.189 × 10 + 2.072 × 10

11-38 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 High-Pass, Band-Pass, and Band-Elimination Filters

11.7 High-Pass, Band-Pass, and Band-Elimination Filters

Transformation methods have been developed where a low-pass filter can be converted to another
type of filter simply by transforming the complex variable s . These transformations are listed in
Table 11.5 where ω C is the cutoff frequency of a low-pass filter. The procedure is best illustrated
with the following examples.

TABLE 11.5 Filter transformations

Analog Filter Frequency Transformations
Filter Type, Frequency Replace s in G ( s ) with
Low-Pass Filter, 3 dB pass-band, Normalized Frequency ω C No Change

Low-Pass Filter, 3 dB pass-band, Non-Normalized Frequency ω LP sω

--------C-
ω LP

High-Pass Filter, 3 dB pass-band from ω = ω 2 to ω = ∞ ω LP ⋅ ω 2

--------------------
s
Band-Pass Filter, 3 dB pass-band from ω = ω LP to ω = ω 2 2
s + ω LP ⋅ ω 2
ω C ⋅ -------------------------------
s ( ω 2 – ω LP )

Band-Elimination Filter, 3 dB pass-band from ω = 0 to ω = ω LP , s ( ω 2 – ω LP )

ω C ⋅ -------------------------------
and from ω = ω 2 to ω = ∞ 2
s + ω LP ⋅ ω 2

Example 11.15
Compute the transfer function for a third-order band-pass Butterworth filter with 3 dB pass-band
from 3 KHz to 5 KHz , from a third-order low-pass Butterworth filter with cutoff frequency
f C = 1 KHz .

Solution:
We first find the transfer function for a third-order Butterworth low-pass filter with normalized fre-
quency ω C = 1 rad ⁄ s . Using the MATLAB function buttap we write and execute the following
code:
[z, p, k]=buttap(3); [b,a]=zp2tf(z,p,k)
b =
0 0 0 1
a =
1.0000 2.0000 2.0000 1.0000

Signals and Systems with MATLAB Applications, Second Edition 11-39

Orchard Publications
Chapter 11 Analog and Digital Filters
Thus, the transfer function for the third-order Butterworth low-pass filter with normalized cutoff
frequency ω c = 1 rad ⁄ s is

1
G ( s ) = ---------------------------------------
- (11.67)
3 2
s + 2s + 2s + 1
3
Next, the actual cutoff frequency is given as f C = 1 KHz or ω C = 2π × 10 rad ⁄ s . Accordingly, in
accordance with Table 11.5, we replace s with
sω s -
---------C = --------------------
ω LP 2π × 10
3

and we get

G ⎛ --------------------3- ⎞ = G' ( s ) = -------------------------------------------------------------------------------------------------------------------------------------------------

s 1 -
⎝ ⎠
2π × 10 3 3
( s ⁄ ( 2π × 10 ) ) + 2 ( s ⁄ ( 2π × 10 ) ) + 2 ( s ⁄ ( 2π × 10 ) ) + 1
3 2 3

(11.68)
11
2.48 × 10
= ------------------------------------------------------------------------------------------------------------
-
3 4 2 7 11
s + 1.26 × 10 s + 7.89 × 10 s + 2.48 × 10

Now, we replace s in the last expression of (11.68) with

2
s + ω LP ⋅ ω 2
ω C ⋅ ------------------------------- (11.69)
s ( ω 2 – ω LP )

or
2 3 3
s + 2π × 10 × 3 × 2π × 10 2
+ 12 × π × 10
2 6 2
s + 1.844 × 10
8
- = s-------------------------------------------
1 ⋅ -------------------------------------------------------------------- = --------------------------------------
-
3 3
s ( 3 × 2π × 10 – 2π × 10 ) s ( 4π × 10 )
3
1.257 × 10 s
4

Then,
11
2.48 × 10
G'' ( s ) = ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
3 2
⎛ s + 1.844 × 10 ⎞
2 8 ⎛ s + 1.844 × 10 8 ⎞
2 2
s + 1.844 × 10
8
11
⎜ --------------------------------------- ⎟ + ⎜ --------------------------------------- ⎟ + -------------------------------------- - + 2.48 × 10
⎝ 1.257 × 10 s ⎠ 4
⎝ 1.257 × 10 s ⎠ 4
1.257 × 10 s
4

We see that the computations, using the transformations of Table 11.5 become quite tedious. Fortu-
nately, we can use the MATLAB lp2lp, lp2hp, lp2bp, and lp2bs functions to transform a low pass
filter with normalized cutoff frequency, to another low-pass filter with any other specified frequency,
or to a high-pass filter, or to a band-pass filter, or to a band elimination filter respectively.

Example 11.16
Use the MATLAB buttap and lp2lp functions to find the transfer function of a third-order Butter-
worth low-pass filter with cutoff frequency f C = 2 KHz .

11-40 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 High-Pass, Band-Pass, and Band-Elimination Filters

Solution:
We will use the buttap command to find the transfer function G ( s ) of the filter with normalized
cutoff frequency at ω C = 1 rad ⁄ s . Then, we will use the command lp2lp to transform G ( s ) to
G' ( s ) with cutoff frequency at f C = 2 KHz , or ω C = 2π × 2 × 10 rad ⁄ s .
3

format short e
% Design 3 pole Butterworth low-pass filter (wcn=1 rad/s)
[z,p,k]=buttap(3);
[b,a]=zp2tf(z,p,k); % Compute num, den coefficients of this filter (wcn=1rad/s)
f=1000:1500/50:10000; % Define frequency range to plot
w=2*pi*f; % Convert to rads/sec
fc=2000; % Define actual cutoff frequency at 2 KHz
wc=2*pi*fc; % Convert desired cutoff frequency to rads/sec
[bn,an]=lp2lp(b,a,wc); % Compute num, den of filter with fc = 2 kHz
Gsn=freqs(bn,an,w); % Compute transfer function of filter with fc = 2 kHz
semilogx(w,abs(Gsn)); grid; hold on; xlabel('Radian Frequency w (rad/sec)'),...
ylabel('Magnitude of Transfer Function'),...
title('3-pole Butterworth low-pass filter with fc=2 kHz or wc = 12.57 kr/s')
The plot for the magnitude of this transfer function is shown in Figure 11.26.

Figure 11.26. Magnitude for the transfer function of (11.71)

The coefficients of the numerator and denominator of the transfer function are as follows:
b

Signals and Systems with MATLAB Applications, Second Edition 11-41

Orchard Publications
Chapter 11 Analog and Digital Filters
b =
0 0 0 1
a
a =
1.0000e+000 2.0000e+000 2.0000e+000 1.0000e+000
bn
bn =
1.9844e+012
an
an =
1.0000e+000 2.5133e+004 3.1583e+008 1.9844e+012

Thus, the transfer function with normalized cutoff frequency ω Cn = 1 rad ⁄ s is

1
G ( s ) = ---------------------------------------
- (11.70)
3 2
s + 2s + 2s + 1
4
and with actual cutoff frequency ω Cn = 2 π × 2000 rad ⁄ s = 1.2566 × 10 is
12
1.9844 × 10
G' ( s ) = ------------------------------------------------------------------------------------------------------------------------------- (11.71)
3 4 2 8 12
s + 2.5133 × 10 s + 3.1583 × 10 s + 1.9844 × 10

Example 11.17
Use the MATLAB commands cheb1ap and lp2hp to find the transfer function of a 3-pole Type I
Chebyshev high-pass analog filter with cutoff frequency f C = 5 KHz .

Solution:
We will use the cheb1ap command to find the transfer function G ( s ) of the low-pass filter with
normalized cutoff frequency at ω C = 1 rad ⁄ s . Then, we will use the command lp2hp to transform
3
G ( s ) to another G' ( s ) with cutoff frequency at f C = 5 KHz or ω C = 2π × 5 × 10 rad ⁄ s

% Design 3 pole Type 1 Chebyshev low-pass filter, wcn=1 rad/s

[z,p,k]=cheb1ap(3,3);
[b,a]=zp2tf(z,p,k); % Compute num, den coef. with wcn=1 rad/s
f=1000:100:100000; % Define frequency range to plot
fc=5000; % Define actual cutoff frequency at 5 KHz
wc=2*pi*fc; % Convert desired cutoff frequency to rads/sec
[bn,an]=lp2hp(b,a,wc); % Compute num, den of high-pass filter with fc = 5 KHz
Gsn=freqs(bn,an,2*pi*f); % Compute and plot transfer function of filter with fc = 5 KHz
semilogx(f,abs(Gsn)); grid; hold on

11-42 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 High-Pass, Band-Pass, and Band-Elimination Filters

xlabel('Frequency (Hz)'); ylabel('Magnitude of Transfer Function')

title('3-pole Type 1 Chebyshev high-pass filter with fc=5 KHz ')
The magnitude of this transfer function is plotted as shown in Figure 11.27.

Figure 11.27. Magnitude of the transfer function of (11.73) for Example 11.17

The coefficients of the numerator and denominator of the transfer function are as follows:
b
b =
0 0 0 2.5059e-001
a
a =
1.0000e+000 5.9724e-001 9.2835e-001 2.5059e-001
bn
bn =
1.0000e+000 2.2496e-011 -1.4346e-002 -6.8973e-003
an
an =
1.0000e+000 1.1638e+005 2.3522e+009 1.2373e+014
Therefore, the transfer function with normalized cutoff frequency ω Cn = 1 rad ⁄ s is

Signals and Systems with MATLAB Applications, Second Edition 11-43

Orchard Publications
Chapter 11 Analog and Digital Filters

0.2506
G ( s ) = --------------------------------------------------------------------------------
- (11.72)
3 2
s + 0.5972s + 0.9284s + 0.2506
4
and with actual cutoff frequency ω Cn = 2π × 5000 rad ⁄ s = 3.1416 × 10 , is
3
s
G' ( s ) = --------------------------------------------------------------------------------------------------------------------------------- (11.73)
3 5 2 9 14
s + 1.1638 × 10 s + 2.3522 × 10 s + 1.2373 × 10

Example 11.18
Use the MATLAB functions buttap and lp2bp to find the transfer function of a 3-pole Butterworth
analog band-pass filter with the pass band frequency centered at f 0 = 4 KHz , and bandwidth
BW = 2 KHz .
Solution:
We will use the buttap function to find the transfer function G ( s ) of the low-pass filter with nor-
malized cutoff frequency at ω C = 1 rad ⁄ s . We found this transfer function in Example 11.15 as
given by (11.67). However, to maintain a similar MATLAB code as in the previous examples, we will
include it in the code that follows. Then, we will use the command lp2bp to transform G ( s ) to
another G' ( s ) with centered frequency at f 0 = 4 KHz or ω 0 = 2π × 4 × 10 rad ⁄ s , and bandwidth
3

3
BW = 2 KHz or BW = 2π × 2 × 10 rad ⁄ s
format short e
[z,p,k]=buttap(3); % Design 3 pole Butterworth low-pass filter with wcn=1 rad/s
[b,a]=zp2tf(z,p,k); % Compute numerator and denominator coefficients for wcn=1
rad/s
f=100:100:100000; % Define frequency range to plot
f0=4000; % Define centered frequency at 4 KHz
W0=2*pi*f0; % Convert desired centered frequency to rads/sec
fbw=2000; % Define bandwidth
Bw=2*pi*fbw; % Convert desired bandwidth to rads/sec
[bn,an]=lp2bp(b,a,W0,Bw);% Compute num, den of band-pass filter
% Compute and plot the magnitude of the transfer function of the band-pass filter
Gsn=freqs(bn,an,2*pi*f); semilogx(f,abs(Gsn)); grid; hold on
xlabel('Frequency f (Hz)'); ylabel('Magnitude of Transfer Function');
title('3-pole Butterworth band-pass filter with f0 = 4 KHz, BW = 2KHz')
The plot for this band-pass filter is shown in Figure 11.28.
The coefficients b n and a n are as follows:
bn
bn =

11-44 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 High-Pass, Band-Pass, and Band-Elimination Filters

1.9844e+012 -4.6156e+001 -1.6501e+005 -2.5456e+009

Figure 11.28. Plot for the band-pass filter of Example 11.18

an
an =
1.0000e+000 2.5133e+004 2.2108e+009 3.3735e+013 1.3965e+018
1.0028e+022 2.5202e+026
Since the numerator b n and denominator a n coefficients are too large to be written in a one line
equation, we have listed them in tabular form as shown below.

Power of s Numerator b n Denominator a n

s6 0 1
s5 0 2.5133 × 10
4

s4 0 2.2108 × 10
9

s3 1.9844 × 10
12
3.3735 × 10
13

s2 – 4.6156 × 10
1
1.3965 × 10
18

s 5 22
– 1.6501 × 10 1.0028 × 10
Cons tan t 26
– 2.5456 × 10
9
2.5202 × 10

Signals and Systems with MATLAB Applications, Second Edition 11-45

Orchard Publications
Chapter 11 Analog and Digital Filters
Example 11.19
Use the MATLAB functions buttap and lp2bs to find the transfer function of a 3-pole Butterworth
band-elimination (band-stop) filter with the stop band frequency centered at f 0 = 5 KHz , and band-
width BW = 2 KHz .
Solution:
We will use the buttap function to find the transfer function G ( s ) of the low-pass filter with nor-
malized cutoff frequency at ω C = 1 rad ⁄ s . We found this transfer function as (11.67) in Example
11.15. However, to maintain a similar MATLAB code as in the previous examples, we will include it
in the code which follows. Accordingly, we will use the lp2bs function to transform G ( s ) to another
transfer function G' ( s ) with centered frequency at f 0 = 5 KHz , or ω 0 = 2π × 5 × 10 rad ⁄ s , and
3

3
bandwidth BW = 2 KHz or BW = 2π × 2 × 10 rad ⁄ s .

[z,p,k]=buttap(3); % Design 3-pole Butterworth low-pass filter, wcn = 1 r/s

[b,a]=zp2tf(z,p,k); % Compute num, den coefficients of this filter, wcn=1 r/s
f=100:100:100000; % Define frequency range to plot
f0=5000; % Define centered frequency at 5 kHz
W0=2*pi*f0; % Convert centered frequency to r/s
fbw=2000; % Define bandwidth
Bw=2*pi*fbw; % Convert bandwidth to r/s
% Compute numerator and denominator coefficients of desired band stop filter
[bn,an]=lp2bs(b,a,W0,Bw);
% Compute and plot magnitude of the transfer function of the band stop filter
Gsn=freqs(bn,an,2*pi*f); semilogx(f,abs(Gsn)); grid; hold on
xlabel('Frequency in Hz'); ylabel('Magnitude of Transfer Function');
title('3-pole Butterworth band-elimination filter with f0=5 KHz, BW = 2 KHz')
The amplitude response of this band elimination filter is shown in Figure 11.29.
The coefficients b n and a n are as follows:
bn
bn =
1.0000e+000 -7.6352e-012 2.9609e+009 -1.5071e-002
2.9223e+018 -7.4374e+006 9.6139e+026
an
an =
1.0000e+000 2.5133e+004 3.2767e+009 5.1594e+013
3.2340e+018 2.4482e+022 9.6139e+026

11-46 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 High-Pass, Band-Pass, and Band-Elimination Filters

Figure 11.29. Amplitude response for the band-elimination filter of Example 11.19

As in the previous example, we list the numerator b n and denominator a n coefficients in tabular
form as shown below.

Power of s Numerator b n Denominator a n

s6 1 1
s5 – 12 4
– 7.6352 × 10 2.5133 × 10
s4 –6 9
2.9609 × 10 3.2767 × 10
s3 –2 13
– 1.5071 × 10 5.1594 × 10
s2 18 18
2.9223 × 10 3.2340 × 10
s 6 22
– 7.4374 × 10 2.4482 × 10
26 26
Cons tan t 9.6139 × 10 9.6139 × 10

In all of the above examples, we have shown the magnitude, but not the phase response of each fil-
ter type. However, we can use the MATLAB function bode(num,den) to generate both the magni-
tude and phase responses of any transfer function describing the filter type, as shown by the follow-
ing example.

Signals and Systems with MATLAB Applications, Second Edition 11-47

Orchard Publications
Chapter 11 Analog and Digital Filters
Example 11.20
Use the MATLAB bode function to plot the magnitude and phase characteristics of the 3-pole But-
terworth low-pass filter with unity gain and normalized frequency at ω C = 1 rad ⁄ s .

Solution:
We know, from Example 11.15, that the transfer function for this type of filter is
1
G ( s ) = ---------------------------------------
-
3 2
s + 2s + 2s + 1
We can obtain the magnitude and phase characteristics with the following MATLAB code:
num=[0 0 0 1]; den=[1 2 2 1]; bode(num,den),...
title('Bode Plot for 3-pole Butterworth Low-Pass Filter'); grid
The magnitude and phase characteristics are shown in Figure 11.30.

Figure 11.30. Bode plots for 3-pole Butterworth low-pass filter

We conclude the discussion on analog filters with Table 11.6 listing the advantages and disadvantages
of each type.

11-48 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

TABLE 11.6 Advantages / Disadvantages of different types of filters

Filter Type Advantages Disadvantages
Butterworth • Simplest design • Slow rate of attenuation
• Flat pass band for order 4 or less
Chebyshev Type 1 • Sharp cutoff rate in transition • Ripple in pass band
(pass to stop) band • Bad (non-linear) phase
response
Chebyshev Type II • Sharp cutoff rate in transition • Ripple in stop band
(pass to stop) band • Bad (non-linear) phase
response
Elliptic (Cauer) • Sharpest cutoff rate among • Ripple in both pass band
all other types of filters and stop band
• Worst (most non-linear)
phase response among
the other types of filters.

11.8 Digital Filters

A digital filter is essentially a computational process (algorithm) that converts one sequence of num-
bers x [ n ] representing the input, to another sequence y [ n ] that represents the output. Thus, a digi-
tal filter, in addition of filtering out desired bands of frequency, can also be used as a computational
means of performing other functions such as integration, differentiation, and estimation.
The input-output difference equation that relates the output to the input can be expressed in the dis-
crete time domain as a summation of the form
k k
y[ n] = ∑ ai x [ n – i ] – ∑ bi y [ n – i ] (11.74)
i=0 i=0

or, in the z -domain as

k
–i
∑ ai z
N(z) i=0
G ( z ) = ----------- = ----------------------------- (11.75)
D(z) k
–i
1+ ∑ bi z
i=0

Therefore, the design of a digital filter to perform a desired function, entails the determination of
the coefficients a i and b i .

Digital filters are classified in terms of the duration of the impulse response, and in forms of realiza-
tion.

Signals and Systems with MATLAB Applications, Second Edition 11-49

Orchard Publications
Chapter 11 Analog and Digital Filters
1. Impulse Response Duration
a. An Infinite Impulse Response (IIR) digital filter has infinite number of samples in its impulse
response h [ n ]
b. A Finite Impulse Response (FIR) digital filter has a finite number of samples in its impulse
response h [ n ]
2. Realization
a. In a Recursive Realization digital filter the output is dependent on the input and the previous
values of the output. In a recursive digital filter, both the coefficients a i and b i are present.

b. In a Non-Recursive Realization digital filter the output depends on present and past values of
the input only. In a non-recursive digital filter, only the coefficients a i are present, that is,
bi = 0 .

Figure 11.31 shows third-order (3-delay element) recursive and non-recursive realizations.
Generally, IIR filters are implemented by recursive realization, whereas FIR filters are implemented
by non-recursive realization.
Filter design methods have been established, and prototype circuits have been published. Thus, we
can choose the appropriate prototype to satisfy the requirements. Transformation methods are also
available to map an analog prototype to an equivalent digital filter. Three well known methods are
the following:
1. The Impulse Invariant Method that produces a digital filter whose impulse response consists of the
sampled values of the impulse response of an analog filter.
2. The Step Invariant Method that produces a digital filter whose step response consists of the sam-
pled values of the step response of an analog filter.
3. The Bilinear Transformation that uses the transformation
z–1
s = ----------- (11.76)
z+1
or, alternately, the transformation
2 z–1
s = ----- ⋅ ----------- * (11.77)
Ts z + 1

to transform the left-half of the s -plane into the interior of the unit circle in the z -plane.
We will discuss only the bilinear transformation.

* T s is the sampling period, that is, the reciprocal of the sampling frequency fs

11-50 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

a0

a1

a2
x[n] y[n]
–1 –1 –1
+ Z Z Z a3
+

−b1

−b2
−b3
Recursive Digital Filter Realization

a0

a1

a2
x[n] y[ n]
–1 –1 –1
Z Z Z a3
+
Non-Recursive Digital Filter Realization

x[n] y[n]
+ –1
Z x[n] A y[n]
± v[n]
Unit Delay Constant Multiplier
Adder/Subtractor
y[n] = x[n – 1] y [ n ] = Ax [ n ]
y[ n] = x[ n] + v[n]

Figure 11.31. Recursive and non-recursive digital filters

To understand the bilinear transformation, we rewrite (11.76) as

z = 1 + s-
---------- (11.78)
1–s

and by substitution of s = α + jβ , we get

z = (----------------------------
1 + α ) + jβ- (11.79)
( 1 – α ) – jβ

for α ≤ 0 , 1 – α ≥ 1 + α and thus z ≤ 1 .

Signals and Systems with MATLAB Applications, Second Edition 11-51

Orchard Publications
Chapter 11 Analog and Digital Filters

In the special case where s = jω , the jω axis of the s -plane is transformed to the circumference of
the unit circle in the z -plane, as we have seen in Chapter 9, Figure 9.5.
To derive a relationship between the frequencies of the analog and digital filters, we let ω a and ω d
jω d T s
denote the analog and digital frequencies respectively. Substitution of s = jω a and z = e into
(11.76) yields
jω T jω T ⁄ 2 jω T ⁄ 2 –j ω T ⁄ 2
e
d s
–1 e
d s
e
d s
–e
d s
2j sin ( ω d T s ⁄ 2 )
- ⋅ ---------------------------------------------- = ------------------------------------
jω a = ----------------------- = ------------------ -
jω d T s jω d T s ⁄ 2 jω d T s ⁄ 2
+e
–j ωd T s ⁄ 2 2 cos ( ω d T s ⁄ 2 )
e +1 e e
or

ωd Ts
ω a = tan -----------
- (11.80)
2

We see that the analog frequency to digital frequency transformation results in a non-linear mapping;
this condition is known as warping. For instance, the frequency range 0 ≤ ω a ≤ ∞ in the analog fre-
quency is warped into the frequency range 0 ≤ ω d ≤ π ⁄ T s in the digital frequency.

Another form of the analog to digital frequency transformation can be derived from the alternate
form of the bilinear transformation of (11.77), that is,
2 z–1
s = ----- ⋅ ----------- (11.81)
Ts z + 1

We recall from (9.67) of Chapter 9 that

F(z) = G(z) = G(s) 1
(11.82)
s = ----- ln z
Ts

1
But the relation s = ----- ln z is a multi valued transformation and, as such, cannot be used to derive a
Ts
rational form in z .
1
It is shown in complex variables theory that ----- ln z can be approximated as
Ts

1 2 z–1
----- ln z ≈ ----- ⋅ ----------- (11.83)
Ts Ts z + 1

Then, by substitution into (11.82),

2 z–1
G ( z ) = G ⎛ ----- ⋅ ----------- ⎞ (11.84)
⎝ Ts z + 1 ⎠

11-52 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters
jω d
and with the substitution z = e , we get
jω
jω d ⎛ 2 e d–1⎞
G(e ) = G ⎜ ----- ⋅ -------------------
-⎟ (11.85)
⎝ T s e jω d + 1 ⎠

Since the z → s transformation maps the unit circle into the jω axis on the s -plane, the quantity
jω d
e – 1-
2- -------------------
---- ⋅ and jω must be equal to some point ω = ω a on the jω axis, that is,
T s jωd
e +1
jω d
2 e –1
jω a = ----- ⋅ -------------------
jω d
-
Ts
e +1
or
ωd ωd ω
jω d j ------
2
j ------
2 sin -----d-
1 2 e – 1 2 1 ⁄ (j2) e –e 2 2
ω a = --- ⋅ ----- ⋅ -------------------
- = ----- ⋅ ----------------- ⋅ --------------------------- = ----- ---------------
j Ts jω Ts 1 ⁄ 2 ω ω Ts ω
+1
d d d
e j ------
2
j ------
2 cos -----d-
e +e 2
or

2 ω
ω a = ----- ⋅ tan -----d- (11.86)
Ts 2

To express ω d in terms of ω a , we rewrite (11.86) as

ω ωa Ts
tan-----d- = -----------
-
2 2
Then,
–1 ωa T s
ω d = 2tan ------------
2
and for small ω a T s ⁄ 2 ,
–1 ωa T s ωa T s
tan ------------ ≈ ------------
2 2
Therefore,
ωa Ts
ω d ≈ 2 ------------ ≈ ω a T s (11.87)
2

In MATLAB z is a function of normalized frequency and thus the range of frequencies in G ( z ) is

from 0 to π . Then (11.87), when used with MATLAB, becomes

Signals and Systems with MATLAB Applications, Second Edition 11-53

Orchard Publications
Chapter 11 Analog and Digital Filters

ωa Ts
ω d ≈ ------------ (11.88)
π

The effect of warping can be eliminated by pre-warping the analog filter prior to application of the
bilinear transformation. This is accomplished by the use of (11.80) or (11.86).

Example 11.21
Compute the transfer function G ( z ) of a low-pass digital filter with 3 dB cutoff frequency at
20 Hz , and attenuation of at least 10 dB for frequencies greater than 40 Hz . The sampling fre-
quency f s is 200 Hz .

Solution:
We will apply the bilinear transformation and, arbitrarily we choose a second order Butterworth low-
pass filter which, as we see from the curves of Figure 11.12, meets the stop-band specification.
The transfer function G ( s ) of the analog low-pass filter with nor malized frequency at
ω C = 1 rad ⁄ s is found with the MATLAB buttap function as follows:

[z,p,k]=buttap(2); [b,a] = zp2tf(z,p,k)

b =
0 0 1
a =
1.0000 1.4142 1.0000
Thus, the transfer function with normalized frequency, denoted as G n ( s ) is

1
G n ( s ) = -----------------------------------
- (11.89)
2
s + 1.414s + 1

Now, we must transform this transfer function to another with the actual cutoff frequency at 20 Hz .
We denote it as G a ( s ) .

We will first pre-warp the analog frequency which, by (11.80), is related to the digital frequency as
ωd Ts
ω a = tan -----------
-
2
where
1 1
T s = --- = ---------
fs 200

Denoting the analog cutoff ( 3 dB ) frequency as ω ac , and the attenuation frequency as ω aa , we get

11-54 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

ω ac = tan 2π × 20
------------------ = tan 0.3142 = 0.325 (11.90)
2 × 200
and

ω aa = tan 2π × 40
------------------ = tan 0.6283 = 0.7265 (11.91)
2 × 200

Next, to find G a ( s ) from G n ( s ) , we replace s with s ⁄ 0.325 in (11.89) and we get

1
G a ( s ) = ---------------------------------------------------------------------------------
2
( s ⁄ 0.325 ) + ( 1.414s ) ⁄ 0.325 + 1
We will use MATLAB to simplify this expression.
syms s; simple(1/((s/0.325)^2+1.414*s/0.325+1))
1/(1600/169*s^2+1414/325*s+1)
1600/169
ans =
9.4675
1414/325
ans =
4.3508
Then,
1 01056
G a ( s ) = -------------------------------------------------------- = --------------------------------------------------- (11.92)
2 2
9.4675s + 4.3508s + 1 s + 0.4596s + 01056

z – 1- , we get
and making the substitution of s = ----------
z+1

0.1056
G ( z ) = -------------------------------------------------------------------------------------
-
– 2
( –
⎛ ----------- ⎞ + --------------------------------- + 0.1056
z 1 0.4596 z 1 )
⎝ z+1 ⎠ (z + 1)

We use MATLAB to simplify this expression.

syms z; simple(0.1056/(((z−1)/(z+1))^2+0.4596*(z−1)/(z+1)+0.1056))
ans =
264*(z+1)^2/(3913*z^2-4472*z+1615)
expand(264*(z+1)^2)

Signals and Systems with MATLAB Applications, Second Edition 11-55

Orchard Publications
Chapter 11 Analog and Digital Filters
ans =
264*z^2+528*z+264
and thus
2
264z + 528z + 264
G ( z ) = -------------------------------------------------------- (11.93)
2
3913z – 4472z + 1615

We will use the MATLAB freqz function to plot the magnitude of G ( z ) , but we must first express it
2
in negative powers of z . Dividing each term of (11.93) by 3913z we get
–1 –2
G ( z ) = 0.0675 + 0.1349z + 0.0675z -
---------------------------------------------------------------------------- (11.94)
–1 –2
1 – 1.1429z + 0.4127z

The MATLAB code below will generate G ( z ) and will plot the magnitude of this transfer function.
bz=[0.0675 0.1349 0.0675]; az=[1 −1.1429 0.4127]; [Gz, wT]=freqz(bz,az,20,200);...
semilogx(wT,abs(Gz)), axis([0.1 1000 0 1]), hold on;...
title('Digital Low-Pass Filter'), xlabel('Frequency in Hz'), ylabel('Magnitude'),grid
The magnitude is shown on the plot of Figure 11.32.

Figure 11.32. Frequency response for the digital low-pass filter of Example 11.21

Let us now plot the analog equivalent to compare the digital to the analog frequency response. The
MATLAB code below produces the desired plot.

11-56 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

[z,p,k]=buttap(2); [b,a]=zp2tf(z,p,k); w=0:0.05*2*pi:100*2*pi; f=w/(2*pi);...

fc=20; wc=2*pi*fc; [bn,an]=lp2lp(b,a,wc); Gs=freqs(bn,an,w);...
semilogx(f,abs(Gs)), axis([0.1 1000 0 1]), hold on, title('Analog Low-Pass Filter'),...
xlabel('Frequency in Hz'), ylabel('Magnitude'), grid
The frequency response for the analog low-pass equivalent is shown in Figure 11.33.

Figure 11.33. Frequency response for analog low-pass filter equivalent

Comparing the digital filter plot with the analog, we see that they are almost identical. Also, from
both plots, we see that the amplitude is greater than 0.707 ( – 3 dB ) for frequencies less than 20Hz ,
and is smaller than 0.316 (– 10 dB) for frequencies larger than 40Hz . Therefore, the filter meets the
required specifications.
An analog filter transfer function can be mapped to a digital filter transfer function directly with the
MATLAB bilinear function. The procedure is illustrated with the following example.

Example 11.22
Use the MATLAB bilinear function to derive the low-pass digital filter transfer function G ( z ) from
a second-order Butterworth analog filter with a 3 dB cutoff frequency at 50 Hz , and sampling rate
f s = 500 Hz .

Solution:
We will use the following MATLAB code to produce the desired digital filter transfer function.

Signals and Systems with MATLAB Applications, Second Edition 11-57

Orchard Publications
Chapter 11 Analog and Digital Filters
[z,p,k]=buttap(2); [num,den]=zp2tf(z,p,k); wc=2*pi*50;...
[num1,den1]=lp2lp(num,den,wc); T=1/500; [numd,dend]=bilinear(num1,den1,1/T)
numd =
0.0640 0.1279 0.0640
dend =
1.0000 -1.1683 0.4241
Therefore, the transfer function G ( z ) for this filter is
2
0.0640z + 0.1279z + 0.0640
G ( z ) = ---------------------------------------------------------------------- (11.95)
z 2 – 1.1683z + 0.4241

MATLAB provides us with all the functions that we need to design digital filters using analog proto-
types. These are listed below with the indicated notations.
N = order of the filter
Wn = normalized cutoff frequency
Rp = pass band ripple
Rs = stop band ripple
B = B(z), i.e., the numerator of the discrete transfer function G ( z ) = B ( z ) ⁄ A ( z )
A = A(z), i.e., the denominator of the discrete transfer function G ( z )
For Low-Pass Filters
[B,A] = butter(N,Wn)
[B,A] = cheb1(N,Rp,Wn)
[B,A] = cheb2(N,Rs,Wn)
[B,A] = ellip(N,Rp,Rs,Wn)
For High-Pass Filters
[B,A] = butter(N,Wn,'high')
[B,A] = cheb1(N,Rp,Wn,'high')
[B,A] = cheb2(N,Rs,Wn,'high')
[B,A] = ellip(N,Rp,Rs,Wn,'high')
Band-Pass Filters
[B,A] = butter(N,[Wn1,Wn2])
[B,A] = cheb1(N,Rp,[Wn1,Wn2])
[B,A] = cheb2(N,Rs,[Wn1,Wn2])
[B,A] = ellip(N,Rp,Rs,[Wn1,Wn2])
Band-Elimination Filters
[B,A] = butter(N,[Wn1,Wn2],'stop')
[B,A] = cheb1(N,Rp,[Wn1,Wn2],'stop')
[B,A] = cheb2(N,Rs,[Wn1,Wn2],'stop')
[B,A] = ellip(N,Rp,Rs,[Wn1,Wn2],'stop')

11-58 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

Example 11.23
The transfer functions of (11.96) through (11.99) below, describe different types of digital filters.
Use the MATLAB freqz command to plot the magnitude versus radian frequency.
–1 –2 –3 –3
G 1 ( z ) = (------------------------------------------------------------------------------------------------------------------------------
2.8982 + 8.6946 z + 8.6946z + 2.8982z ) ⋅ 10 - (11.96)
–1 –2 –3
1 – 2.3741z + 1.9294z – 0.5321z
–1 –2 –3
0.5276 – 1.5828 z + 1.5828z – 0.5276 z
G 2 ( z ) = -------------------------------------------------------------------------------------------------------- (11.97)
–1 –2 –3
1 – 1.7600z + 1.1829z – 0.2781z
–2 –4 –4
( 6.8482 – 13.6964 z + 6.8482z ) ⋅ 10
G 3 ( z ) = ---------------------------------------------------------------------------------------------------------------------------- (11.98)
–1 –2 –3 –4
1 + 3.2033z + 4.5244z + 3.1390z + 0.9603z
–1 –2
0.9270 – 1.2079 z + 0.9270z
G 4 ( z ) = ---------------------------------------------------------------------------- (11.99)
–1 –2
1 – 1.2079z + 0.8541z
Solution:
The MATLAB code to compute and plot each of the transfer functions of (11.96) through (11.99),
is given below.
% N=512 % Default
b1=[2.8982 8.6946 8.6946 2.8982]*10^(−3); a1=[1 −2.3741 1.9294 −0.5321];...
[G1z,w1T]=freqz(b1,a1);
%
b2=[0.5276 −1.5828 1.5828 −0.5276]; a2=[1 −1.7600 1.1829 −0.2781];...
[G2z,w2T]=freqz(b2,a2);
%
b3=[6.8482 0 -13.6964 0 6.8482]*10^(−4); a3=[1 3.2033 4.5244 3.1390 0.9603];...
[G3z,w3T]=freqz(b3,a3);
%
b4=[0.9270 −1.2079 0.9270]; a4=[1 −1.2079 0.8541];...
[G4z,w4T]=freqz(b4,a4);
clf; % clear current figure
%
subplot(221), semilogx(w1T,abs(G1z)), axis([0.1 1 0 1]), title('Filter for G1(z)')
xlabel(''),ylabel('Magnitude'),grid;
%
subplot(222), semilogx(w2T,abs(G2z)), axis([0.1 10 0 1]), title('Filter for G2(z)')
xlabel(''),ylabel('Magnitude'),grid;
%
subplot(223), semilogx(w3T,abs(G3z)), axis([1 10 0 1]), title('Filter for G3(z)')
xlabel(''),ylabel('Magnitude'),grid;
%

Signals and Systems with MATLAB Applications, Second Edition 11-59

Orchard Publications
Chapter 11 Analog and Digital Filters
subplot(224), semilogx(w4T,abs(G4z)), axis([0.1 10 0 1]), title('Filter for G4(z)')
xlabel(''),ylabel('Magnitude'),grid;
The plots are shown in Figure 11.34. We see that the given transfer functions are for low-pass, high-
pass, band-pass, and band-stop digital filters.

Figure 11.34. Plot for the transfer functions of Example 11.23

Example 11.24
We are given a 165 KHz total bandwidth, and within this bandwidth we must accommodate four
different signals. Each of these signals requires 25 KHz bandwidth. We are asked to define the types
of filters and cutoff frequencies to avoid interference among these signals.
Solution:
We will use Butterworth filters up to order 12 to obtain sharp cutoffs, and the following types and
bandwidths for each.
1. Low-pass filter with bandwidth 0 to 25 KHz , ( 3 dB cutoff at 25 KHz )
2. Band-pass filter with bandwidth from 40 KHz to 65 KHz
3. High-pass filter with 3 dB frequency at 90 KHz
4. Band-elimination filter with stop-band from 115 KHz to 140 KHz
The highest (Nyquist) frequency is 165 kHz so we choose a sampling frequency of 330 kHz.
The MATLAB freqz function in the code below normalizes the frequencies from 0 to π where
π = Nyquist frequency .

11-60 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

% N=512 % Default
fs=330000 % Chosen sampling frequency
Ts=1/fs; % Sampling period
fn=fs/2; % Nyquist frequency
%
f1=25000/fn % Low-pass filter cutoff frequency (Signal 1 End)
f2=40000/fn % Band-pass left cutoff frequency (Signal 2 Start)
f3=65000/fn % Band-pass right cutoff frequency (Signal 2 End)
f4=90000/fn % High-pass filter cutoff frequency (Signal 3 Start)
f5=115000/fn % Band-stop filter left cutoff frequency (Signal 3 End)
f6=140000/fn % Band-stop filter right cutoff frequency (Signal 4 Start)
% Signal 4 will terminate at 165 kHz
[b1,a1]=butter(12,f1);
[b2,a2]=butter(12,[f2,f3]);
[b3,a3]=butter(12,f4,'high');
[b4,a4]=butter(12,[f5,f6],'stop');
%
[G1z,wT]=freqz(b1,a1);
[G2z,wT]=freqz(b2,a2);
[G3z,wT]=freqz(b3,a3);
[G4z,wT]=freqz(b4,a4);
%
Hz=wT/(2*pi*Ts);
%
clf; % clear current figure
%
plot(Hz,abs(G1z),Hz,abs(G2z),Hz,abs(G3z),Hz,abs(G4z)), axis([0 16*10^4 0 1])
title('Four signals separated by four digital filters');
xlabel('Hz'),ylabel('Magnitude'),grid;
The plot of Figure 11.35 shows the frequency separations for these four signals.
In the following example, we will demonstrate the MATLAB filter function that is being used to
remove unwanted frequency components from a function. But before we use the filter function, we
must design a filter that is capable of removing those unwanted components.
Example 11.25
In Chapter 7, Example 7.6, we found that the half-wave rectifier can be represented by the trigono-
metric series
A A A cos 2t cos 4t cos 6t cos 8t
f ( t ) = --- + --- sin t – --- ------------- + ------------- + ------------- + ------------- + …
π 2 π 3 15 35 63

In this example, we want to filter out just the first 2 terms, or, in other words, to remove all cosine
terms. To simplify this expression, we let A = 3π and we truncate it by eliminating all terms beyond
the third. Then,

Signals and Systems with MATLAB Applications, Second Edition 11-61

Orchard Publications
Chapter 11 Analog and Digital Filters

Figure 11.35. Frequency separations for the signals of Example 11.24

g ( t ) = 3 + 1.5 sin t – cos 2t (11.100)

The problem now reduces to design a low-pass digital filter, and use the filter command to remove
the cosine term in (11.100).
Solution:
We will use a 6 – pole digital low-pass Butterworth filter because we must have a sharp transition
between the 1 and 2 rad ⁄ s frequency range. Also, since the highest frequency component is
2 rad ⁄ s , we must specify a sampling frequency of ω s = 4 rad ⁄ s to avoid aliasing. Thus, the sam-
pling frequency must be f s = ω s ⁄ 2π = 2 ⁄ π and the sampling period will be T s = 1 ⁄ f s = π ⁄ 2 .We
choose T = 0.5 ; this is sufficiently small. Also, we choose the cutoff frequency of the filter to be
ω C = 1.5 rad ⁄ s in order to attenuate the cosine terms.

The MATLAB code below will perform the following steps:

1. Compute coefficients of the numerator and denominator of the transfer function with normal-
ized cutoff frequency.
2. Will recompute the coefficients for the desired frequency.
3. Use the bilinear function to map the analog transfer function to a digital transfer function, and
will plot the frequency response of the digital filter.
4. Will recompute the digital filter transfer function to account for the warping effect.

11-62 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

5. Will use the filter function to remove the cosine terms

% Step 1
%
[z,p,k]=buttap(6);
[b,a]=zp2tf(z,p,k);
%
% Step 2
%
wc=1.5; % Chosen cutoff frequency
[b1,a1]=lp2lp(b,a,wc); % Convert to actual cutoff frequency
%
% Step 3
%
T=0.5; % Define sampling period
[Nz,Dz]=bilinear(b1,a1,1/T); % Map to digital filter using the bilinear transformation
w=0:2*pi/300:pi; % Define range for plot
Gz=freqz(Nz,Dz,w); % The digital filter transfer function
%
clf
%
semilogx(w,abs(Gz)); grid; hold on
% We must remember that when z is used as a function of
% normalized frequency, the range of frequencies of G(z) are
% from zero to pi and the normalized cutoff frequency on the
% plot is wc*T=1.5*0.5=0.75 r/s.
%
xlabel('Radian Frequency w in rads/sec'),...
ylabel('Magnitude of G(z)'),...
title('Digital Filter Response in Normalized Frequency')
%
fprintf('Press any key to continue \n');
%
pause;
%
% Step 4
%
p=6; T=0.5; % Number of poles and Sampling period
wc=1.5; % Analog cutoff frequency in rad/sec
wd=wc*T/pi; % Normalized digital filter cutoff frequency by (11.81)
[Nzp,Dzp]=butter(p,wd);
fprintf('Summary: \n\n');
fprintf('WITHOUT PREWARPING: \n\n');
%
fprintf('The num N(z) coefficients in descending order of z are: \n\n');

Signals and Systems with MATLAB Applications, Second Edition 11-63

Orchard Publications
Chapter 11 Analog and Digital Filters
fprintf('%8.4f \t',[Nz]);
fprintf('\n\n');
fprintf('The den D(z) coefficients in descending order of z are: \n\n');
fprintf('%8.4f \t',[Dz]);
fprintf('\n\n');
fprintf('WITH PREWARPING: \n\n');
%
fprintf('The num N(z) coefficients in descending order of z are: \n\n');
fprintf('%8.4f \t',[Nzp]);
fprintf('\n\n');
fprintf('The den D(z) coefficients in descending order of z are: \n\n');
fprintf('%8.4f \t',[Dzp]);
fprintf('\n\n');
The plot of the low-pass filter that will remove the cosine terms is shown in Figure 11.36.

Figure 11.36. Plot of the low-pass filter of Example 11.23

Therefore,
WITHOUT PREWARPING:
The num N(z) coefficients in descending order of z are:
0.0007 0.0040 0.0100 0.0134 0.0100 0.0040 0.0007
The den D(z) coefficients in descending order of z are:
1.0000 -3.2379 4.7566 -3.9273 1.8999 -0.5064 0.0578

11-64 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

WITH PREWARPING:
The num N(z) coefficients in descending order of z are:
0.0008 0.0050 0.0125 0.0167 0.0125 0.0050 0.0008
The den D(z) coefficients in descending order of z are:
1.0000 -3.1138 4.4528 -3.5957 1.7075
%
% Step 5
%
Nzp=[0.0008 0.0050 0.0125 0.0167 0.0125 0.0050 0.0008];
Dzp=[1.0000 -3.1138 4.4528 -3.5957 1.7075 -0.4479 0.0504];
n=0:150;
T=0.5;
gt=3+1.5*sin(n*T)-cos(2*n*T);
yt=filter(Nzp,Dzp,gt);
%
% We will plot the unfiltered analog signal gta
%
t=0:0.1:12;
gta=3+1.5*sin(t)-cos(2*t);
subplot(211), plot(t,gta), axis([0,12, 0, 6]); hold on
xlabel('Continuous Time t'); ylabel('Function g(t)');
%
% We will plot the filtered analog signal y(t)
%
subplot(212), plot(n*T,yt), axis([0,12, 0, 6]); hold on
xlabel('Continuous Time t'); ylabel('Filtered Output y(t)');
%
fprintf('Press any key to continue \n'); pause;
%
% We will plot the unfiltered discrete time signal g(n*T)
%
subplot(211), stem(n*T,gt), axis([0,12, 0, 6]); hold on
xlabel('Discrete Time nT'); ylabel('Discrete Function g(n*T)');
%
% We will plot the filtered discrete time signal y(n*T)
subplot(212), stem(n*T,yt), axis([0,12, 0, 6]); hold on
xlabel('Discrete Time nT'); ylabel('Filtered Output y(n*T)');
The analog and digital inputs and outputs are shown in Figure 11.37.

Signals and Systems with MATLAB Applications, Second Edition 11-65

Orchard Publications
Chapter 11 Analog and Digital Filters

Figure 11.37. Input and output waveforms for Example 11.23

We will conclude this chapter with one more example to illustrate the use of the MATLAB find
function. This function displays the subscripts of an array of numbers where a relational expression
is true. For example,
x=−2:5; % Display the integers in the range −2 <= x < =5

11-66 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Digital Filters

x =
-2 -1 0 1 2 3 4 5
k=find(x>0.8); % Find the subscripts of the numbers for which x > 0.8
k =
4 5 6 7 8
y=x(k); % Create a new array y using the indices in k
y =
1 2 3 4 5

Example 11.26
Given the function f ( t ) = 5 sin 2t – 10 cos 5t , use the MATLAB randn function to add random
(Gaussian) noise to f ( t ) and plot this signal plus noise waveform which we denote as
x ( t ) = f ( t ) + randn ( N ) = 5 sin 2t – 10 cos 5t + randn ( size ( t ) ) (11.101)
where 0 ≤ t ≤ 512 . Next, use the fft function to compute the frequency components of the 512-
point FFT and plot the spectrum of this noisy signal. Finally, use the find function to restrict the fre-
quency range of the spectrum identify the frequency components of the signal f ( t ) .
Solution:
The MATLAB code is shown below.
t=linspace(0, 10, 512); x=10*sin(2*t)−5*cos(5*t)+15*randn(size(t));
% We plot the signal to see what it looks like
%
subplot(221); plot(t,x),title('x(t)=Signal plus Noise')
%
% The input signal x is shown in the upper left corner of the graph
%
% Next, we will compute the frequency domain of the signal x
%
X=fft(x);
%
% The sampling period of x is found by the time difference of two samples
%
Ts=t(2)−t(1);
%
% and the sampling frequency is
%
Ws=2*pi/Ts;
%

Signals and Systems with MATLAB Applications, Second Edition 11-67

Orchard Publications
Chapter 11 Analog and Digital Filters
% As we know, the Nyquist frequency Wn is half the sampling frequency
%
Wn=Ws/2;
%
% Now, we will define the frequency domain axis
%
w=linspace(0,Wn,length(t)/2);
%
% The magnitude of the positive frequency components Xp are found from:
%
Xp=abs(X(1:length(t)/2));
% We want now to plot Xp versus radian frequency w
%
subplot(222); plot(w,Xp),title('Spectrum of Signal & Noise in Wide Range')
%
% We will select the frequencies of interest with the "find" function:
%
k=find(w<=20);
%
% Now we will plot this restricted range
%
subplot(212); plot(w(k), Xp(k)),title('Spectrum of Signal & Noise in Narrow Range')
%
% The last plot will have grid, labels and title
%
grid; xlabel('Frequency, rads/sec'); ylabel('Frequency Components');
title('Spectrum of Signal & Noise in Narrow Range')
The signal is shown in Figure 11.38.
We observe the appearance of the sinusoids at 2 and 5 rad ⁄ s in the lower plot. They were undistin-
guished in the time-domain of the upper left plot. The upper right plot indicates that the signal f ( t )
has frequency components in the lower range of frequencies, but these cannot be identified precisely.

11-68 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

Figure 11.38. Waveforms for Example 11.26

11.9 Summary
• Analog filters are defined over a continuous range of frequencies. They are classified as low-pass,
high-pass, band-pass and band-elimination (stop-band).
• An all-pass or phase shift filter has a constant amplitude response but is phase varies with fre-
quency.
• A digital filter, in general, is a computational process, or algorithm that converts one sequence of
numbers representing the input signal into another sequence representing the output signal.
• A digital filter, besides filtering out unwanted bands of frequency, can perform functions of differ-
entiation, integration, and estimation.
• Analog filter functions have been used extensively as prototype models for designing digital filters.
• An analog filter can also be classified as passive or active. Passive filters consist of passive devices
such as resistors, capacitors and inductors. Active filters are, generally, operational amplifiers with
resistors and capacitors connected to them externally.
• If two frequencies ω 1 and ω 2 are such that ω 2 = 2ω 1 , we say that these frequencies are sepa-
rated by one octave, and if ω 2 = 10ω 1 , they are separated by one decade.
• Τhe analog low-pass filter is used as a basis. Using transformations, we can derive high-pass and
the other types of filters from a basic low-pass filter.

Signals and Systems with MATLAB Applications, Second Edition 11-69

Orchard Publications
Chapter 11 Analog and Digital Filters
• In this chapter we discussed the Butterworth, Chebyshev Type I & II, and Cauer (elliptic) filters.
• The first step in the design of an analog low-pass filter is to find a suitable amplitude-squared func-
2
tion A ( ω ) , and from it derive a G ( s ) function such that
2
A ( ω ) = G ( s ) ⋅ G ( –s ) s = jω

2
• The general form of the amplitude square function A ( ω ) is
2k 2k – 2
2 C ( bk ω + bk – 1 ω + … + b0 )
A ( ω ) = --------------------------------------------------------------------------------
-
2k 2k – 2
ak ω + ak – 1 ω + … + a0

where C is the DC gain, a and b are constant coefficients, and k is a positive integer denoting
the order of the filter.
• The amplitude-squared function of a Butterworth analog low-pass filter is
2 1
A ( ω ) = ------------------------------------
2k
( ω ⁄ ωC ) + 1

where k is a positive integer indicating the order of the filter, and ω C is the cutoff ( 3 dB ) fre-
quency.
• All Butterworth filters have the property that all poles of the transfer functions that describes
them, lie on a circumference of a circle of radius ω C , and they are 2π ⁄ 2k radians apart. Thus, if
k = odd , the poles start at zero radians, and if k = even , they start at 2π ⁄ 2k . But regardless
whether k is odd or even, the poles are distributed in symmetry with respect to the jω axis. For
stability, we choose the poles of the left half of the s -plane to form G ( s ) .
• The general form of any analog low-pass (Butterworth, Chebyshev, Elliptic, etc.) filter is

b0
G(s) lp
= ----------------------------------------------------------------
k 2
ak s + … + a2 s + a1 s + a0

• The MATLAB buttap and zp2tf functions are very useful functions in the design of Butterworth
filters. The first returns the zeros, poles, and gain for an N – th order normalized prototype But-
terworth analog low-pass filter. The resulting filter has N poles around the unit circle in the left
half plane, and no zeros. The second performs the zero-pole to transfer function conversion.
• The Type I Chebyshev filters are based on approximations derived from the Chebyshev polynomi-
als C k ( x ) that constitute a set of orthogonal functions. The coefficients of these polynomials are
tabulated in math tables.
• We can use the MATLAB cheb1ap function to design a Type I Chebyshev analog low-pass filter.
Thus, the [z,p,k] = cheb1ap(N,Rp) statement where N denotes the order of the filter, returns

11-70 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Summary

the zeros, poles, and gain of an N – th order normalized prototype Type I Chebyshev analog low-
pass filter with ripple Rp decibels in the pass band.
• The Inverted Chebyshev, also known as Type II Chebyshev, is characterized by the following
amplitude-square approximation.
2 2
2 ε C k ( ω C ⁄ ω)
A ( ω ) = -----------------------------------------
-
2 2
1 + ε C k ( ω C ⁄ ω)

and has the ripple in the stop-band as opposed to Type I which has the ripple in the pass-band.
The frequency ω C defines the beginning of the stop band.
• The elliptic (Cauer) filters are characterized by the low-pass amplitude-squared function
2 1
A ( ω ) = -----------------------------------
-
2
1 + Rk ( ω ⁄ ωC )

where R k ( x ) represents a rational elliptic function used with elliptic integrals. Elliptic filters have
ripple in both the pass-band and the stop-band.
• We can design elliptic low-pass filters with the MATLAB ellip function. The statement [b,a] =
ellip(N,Rp,Rs,Wn,’s’) where N is the order of the filter, designs an N – th order low-pass filter
with ripple Rp decibels in the pass band, a stop band with ripple Rs decibels, Wn is the cutoff fre-
quency, and ’s’ is used to specify analog elliptic filters. If ’s’ is not included in the above statement,
MATLAB designs a digital filter.
• Transformation methods have been developed where a low-pass filter can be converted to
another type of filter simply by transforming the complex variable s . These transformations are
listed in Table 11.5 where ω C is the cutoff frequency of a low-pass filter.
• We can use the MATLAB lp2lp, lp2hp, lp2bp, and lp2bs functions to transform a low pass fil-
ter with normalized cutoff frequency, to another low-pass filter with any other specified fre-
quency, or to a high-pass filter, or to a band-pass filter, or to a band elimination filter respectively
• We can use the MATLAB function bode(num,den) to generate both the magnitude and phase
responses of any transfer function describing the filter type.
• Digital filters are classified in terms of the duration of the impulse response, and in forms of real-
ization.
• An Infinite Impulse Response (IIR) digital filter has infinite number of samples in its impulse
response h [ n ]
• A Finite Impulse Response (FIR) digital filter has a finite number of samples in its impulse
response h [ n ]

Signals and Systems with MATLAB Applications, Second Edition 11-71

Orchard Publications
Chapter 11 Analog and Digital Filters
• In a Recursive Realization digital filter the output is dependent on the input and the previous val-
ues of the output. In a recursive digital filter, both the coefficients a i and b i are present.

• In a Non-Recursive Realization digital filter the output depends on present and past values of the
input only. In a non-recursive digital filter, only the coefficients a i are present, that is, b i = 0 .

• Generally, IIR filters are implemented by recursive realization, whereas FIR filters are imple-
mented by non-recursive realization.
• Transformation methods are also available to map an analog prototype to an equivalent digital fil-
ter. Three well known methods are the following:
1. The Impulse Invariant Method that produces a digital filter whose impulse response consists
of the sampled values of the impulse response of an analog filter.
2. The Step Invariant Method that produces a digital filter whose step response consists of the
sampled values of the step response of an analog filter.
3. The Bilinear Transformation that uses the transformation
z – 1-
s = ----------
z+1
or, alternately, the transformation
2 z–1
s = ----- ⋅ -----------
Ts z + 1

• The analog frequency to digital frequency transformation results in a non-linear mapping; this
condition is known as warping.
• The effect of warping can be eliminated by pre-warping the analog filter prior to application of
the bilinear transformation.
• We can use the MATLAB freqz(b,a,N) function to plot the magnitude of G ( z )
• An analog filter transfer function can be mapped to a digital filter transfer function directly with
the MATLAB bilinear(b,a,Fs) function.
• The MATLAB filter(b,a,X) function can be used to remove unwanted frequency components
from a function.
• We can use the MATLAB find(X) function to restrict the frequency range of the spectrum in
order to identify the frequency components of the signal f ( t ) .

11-72 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

11.10 Exercises
1. The circuit of Figure 11.39 is a VCVS second-order high-pass filter whose transfer function is

V out ( s ) Ks
2
G ( s ) = ----------------
- = --------------------------------------------------------------
-
V in ( s ) 2
s + ( a ⁄ b )ω C s + ( 1 ⁄ b )ω C
2

and for given values of a , b , and desired cutoff frequency ω C , we can calculate the values of
C 1, C 2, R 1, R 2, R 3, and R 4 to achieve the desired cutoff frequency ω C .

R1
C1 C2

vin
vout
R2

R4
R3

Figure 11.39. Circuit for Exercise 1

For this circuit,
4b
R 2 = ---------------------------------------------------------------------------
⎧ 2 ⎫
C 1 ⎨ a + [ a + 8b ( K – 1 ) ] ⎬ω C
⎩ ⎭
b
R 1 = --------------------
2
C 1 R 2 ω 2C
KR 2
R 3 = -------------, K≠1
K–1
R 4 = KR 2
and the gain K is
K = 1 + R4 ⁄ R3

Using these relations, compute the appropriate values of the resistors to achieve the cutoff fre-
quency f C = 1 KHz . Choose the capacitors as C 1 = 10 ⁄ f C µF and C 2 = C 1 . Plot G ( s ) versus
frequency.
Solution using MATLAB is highly recommended.

Signals and Systems with MATLAB Applications, Second Edition 11-73

Orchard Publications
Chapter 11 Analog and Digital Filters
2. The circuit of Figure 11.40 is a VCVS second-order band-pass filter whose transfer function is
V out ( s ) K [ BW ]s -
G ( s ) = ----------------
- = ---------------------------------------
V in ( s ) 2
s + [ BW ]s + ω 0
2

R2
C2
R1

vin
vout
C1 R3

R5
R4

Figure 11.40. Circuit for Exercise 2

Let ω 0 = center frequency , ω 2 = upper cutoff frequency , ω 1 = lower cutoff frequency ,

Bandwidth BW = ω 2 – ω 1 , and Quality Factor Q = ω 0 ⁄ BW

We can calculate the values of C 1, C 2, R 1, R 2, R 3, and R 4 to achieve the desired centered frequency
ω 0 and bandwidth BW . For this circuit,

2Q
R 1 = -----------------
C1 ω0 K
2Q
R 2 = --------------------------------------------------------------------------
⎧ 2 2⎫
C 1 ω 0 ⎨ – 1 + ( K – 1 ) + 8Q ⎬
⎩ ⎭
1 -⎛ 1 1⎞
R 3 = ------------
2 2⎝R
----- + -----
⎠
C1 ω0 1 R 2

R 4 = R 5 = 2R 3

Using these relations, compute the appropriate values of the resistors to achieve center frequency
f 0 = 1 KHz , Gain K = 10 , and Q = 10 .

Choose the capacitors as C 1 = C 2 = 0.1 µF . Plot G ( s ) versus frequency.

Solution using MATLAB is highly recommended.

11-74 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

3. The circuit of Figure 11.41 is a VCVS second-order band elimination filter whose transfer func-
tion is
V out ( s ) K ( s 2 + ω0 )
2
G ( s ) = ----------------
- = ---------------------------------------
-
V in ( s ) 2
s + [ BW ]s + ω 0
2

R3
C1 C2

vin
vout
R1 R2
C3

Figure 11.41. Circuit for Exercise 3

Let ω 0 = center frequency , ω 2 = upper cutoff frequency , ω 1 = lower cutoff frequency ,

Bandwidth BW = ω 2 – ω 1 , Quality Factor Q = ω 0 ⁄ BW , and gain K = 1

We can calculate the values of C 1, C 2, R 1, R 2, R 3, and R 4 to achieve the desired centered fre-
quency ω 0 and bandwidth BW . For this circuit,

1
R 1 = --------------------
2ω 0 QC 1
2Q
R 2 = ------------
ω0 C1
2Q
R 3 = ------------------------------------
2
-
C 1 ω 0 ( 4Q + 1 )

The gain K must be unity, but Q can be up to 10. Using these relations, compute the appropriate
values of the resistors to achieve center frequency f 0 = 1 KHz , Gain K = 1 and Q = 10 .

Choose the capacitors as C 1 = C 2 = 0.1 µF and C 3 = 2C1 . Plot G ( s ) versus frequency.

Solution using MATLAB is highly recommended.

Signals and Systems with MATLAB Applications, Second Edition 11-75

Orchard Publications
Chapter 11 Analog and Digital Filters
4. The circuit of Figure 11.42 is a MFB second-order all-pass filter whose transfer function is
2
V out ( s ) K ( s 2 – a ω0 s + b ω0 )
G ( s ) = ----------------
- = ---------------------------------------------
-
V in ( s ) 2
s + aω 0 s + bω 0
2

where the gain K = cons tan t , ( 0 < K < 1 ) , and the phase is given by
aω 0 ω ⎞
φ ( ω ) = – 2tan ⎛ ---------------------
–1
-
⎝ 2 2⎠
bω 0 – ω

C2

C1 R2
R1
vin
vout
R3
R4

Figure 11.42. Circuit for Exercise 4

The coefficients a and b can be found from

φ 0 = φ ( ω 0 ) = – 2tan ⎛ ------------ ⎞
–1 a
⎝ b – 1⎠

For arbitrary values of C 1 = C 2 , we can compute the resistances from

2
R 2 = ----------------
aω 0 C 1
( 1 – K )R
R 1 = ------------------------2
4K
R
R 3 = -----2
K
R2
R 4 = ------------
-
1–K

For 0 < φ 0 < 180° , we compute the coefficient a from

1–K 4K 2
a = --------------------------------- – 1 + 1 + ------------- ⋅ tan ( φ 0 ⁄ 2 )
2K tan ( φ 0 ⁄ 2 ) 1–K

11-76 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exercises

and for – 180° < φ 0 < 0° , from

1–K 4K 2
a = --------------------------------- – 1 – 1 + ------------- ⋅ tan ( φ 0 ⁄ 2 )
2K tan ( φ 0 ⁄ 2 ) 1–K

Using these relations, compute the appropriate values of the resistors to achieve a phase shift
φ 0 = – 90° at f 0 = 1 KHz with K = 0.75 .

Choose the capacitors as C 1 = C 2 = 0.01 µF and plot phase versus frequency.

Solution using MATLAB is highly recommended.
5. The Bessel filter of Figure 11.43 has the same configuration as the low-pass filter of Example
11.3, and achieves a relatively constant time delay over a range 0 < ω < ω 0 . The second-order
transfer function of this filter is
V out ( s ) 3Kω 0
2
G ( s ) = ----------------
- = --------------------------------------
-
V in ( s ) 2
s + 3ω 0 s + 3ω 0
2

R2 C2
R1 R3

vin C1 vout

Figure 11.43. Circuit for Exercise 5

where K is the gain and the time delay T 0 at ω 0 = 2πf 0 is given as

12
T 0 = T ( ω 0 ) = ------------ sec onds
13ω 0

We recognize the transfer function G ( s ) above as that of a low-pass filter where a = b = 3

and the substitution of ω 0 = ω C . Therefore, we can use a low-pass filter circuit such as that of
Figure 11.43, to achieve a constant delay T 0 by specifying the resistor and capacitor values of the
circuit.
The resistor values are computed from

Signals and Systems with MATLAB Applications, Second Edition 11-77

Orchard Publications
Chapter 11 Analog and Digital Filters

2(K + 1)
R 2 = ------------------------------------------------------------------------------------
2 2
( aC 1 + a C 1 – 4 bC 1 C 2 ( K – 1 ) )ω 0
R
R 1 = -----2
K
1
R 3 = ----------------------------
-
2
bC 1 C 2 R 2 ω 0

Using these relations, compute the appropriate values of the resistors to achieve a time delay
T 0 = 100 µs with K = 2 . Use capacitors C 1 = 0.01 µF and C 2 = 0.002 µF . Plot G ( s ) versus
frequency.
Solution using MATLAB is highly recommended.
6. Derive the transfer function of a fourth-order Butterworth filter with ω C = 1 rad ⁄ s .

7. Derive the amplitude-squared function for a third-order Type I Chebyshev low-pass filter with
1.5 dB pass band ripple and cutoff frequency ω C = 1 rad ⁄ s .

8. Use MATLAB to derive the transfer function G ( z ) and plot G ( z ) versus ω for a two-pole, Type
I Chebyshev high-pass digital filter with sampling period T S = 0.25 s . The equivalent analog filter
cutoff frequency is ω C = 4 rad ⁄ s and has 3 dB pass band ripple. Compute the coefficients of
the numerator and denominator and plot G ( z ) with and without pre-warping.

11-78 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

11.11 Solutions to Exercises

1. We will use MATLAB for all computations.
% PART I − Find Resistor values for second order Butterworth filter, a=sqrt(2), b = 1
a=sqrt(2); b =1; C1=10^(−8); C2=C1; fc=1000; wc=2*pi*fc; K=2;
R2=(4*b)/(C1*sqrt(a^2+8*b*(K−1))*wc);
R1=b/(C1^2*R2*wc^2); R3=(K*R2)/(K−1); R4=K*R2; fprintf(' \n');...
fprintf('R1 = %5.0f Ohms \t',R1); fprintf('R2 = %5.0f Ohms \t',R2);...
fprintf('R3 = %5.0f Ohms \t',R3); fprintf('R4 = %5.0f Ohms \t',R4)
R1=12582 Ohms R2=20132 Ohms R3=40263 Ohms R4=40263 Ohms
We choose standard resistors as close as possible to those found above. These are shown in the
MATLAB code below. Part II of the code is as follows:
f=10:10:20000; w=2*pi*f; R1=12700; R2=20000; R3=40200; R4=R3; K=1+R4/R3;...
wc=(4*b)/(C1*sqrt(a^2+8*b*(K−1))*R2); s=w*j; Gw=(K.*s.^2)./(s.^2+a.*wc.*s./
b+wc.^2./b); semilogx(f,abs(Gw)); grid; hold on;...
xlabel('Frequency, Hz'), ylabel('|Vout/Vin|');...
title('2nd Order Butterworth High-Pass Filter Response')

Signals and Systems with MATLAB Applications, Second Edition 11-79

Orchard Publications
Chapter 11 Analog and Digital Filters
2. We will use MATLAB for all computations.
% PART I − Find Resistor values for second order band-pass filter f0 = 1 KHz
Q=10; K=10; C1=10^(−7); C2=C1; f0=1000; w0=2*pi*f0; R1=(2*Q)/(C1*w0*K);...
R2=(2*Q)/(C1*w0*(−1+sqrt((K−1)^2+8*Q^2))); R3=(1/(C1^2*w0^2))*(1/R1+1/R2);
R4=2*R3;...
R5=R4; fprintf(' \n'); fprintf('R1 = %5.0f Ohms \t',R1); fprintf('R2 = %5.0f Ohms \t',R2);...
fprintf('R3 = %5.0f Ohms \t',R3); fprintf('R4 = %5.0f Ohms \t',R4);...
fprintf('R5 = %5.0f Ohms \t',R5)
R1=3183 Ohms R2=1110 Ohms R3=3078 Ohms R4=6156 Ohms R5=6156 Ohms
We choose standard resistors as close as possible to those found above. These are shown in the
MATLAB code below. Part II of the code is as follows:
K=10; Q=10; f=10:10:10000; w=2*pi*f; R1=3160; R2=1100; R3=3090; R4=6190;
R5=R4;...
w0=(2*Q)/(C1*R1*K); B=w0/Q; s=w*j; Gw=(K.*B.*s)./(s.^2+B.*s+w0.^2);...
semilogx(f,abs(Gw)); axis([500 2000 0 10]); grid; hold on; xlabel('Frequency, Hz'),...
ylabel('|Vout/Vin|'); title('2nd Order Butterworth Band-Pass Filter Response')

11-80 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

3. We will use MATLAB for all computations.

% PART I − Find Resistor values for second order Butterworth band-pass filter f0 = 1 KHz
Q=10; K=1; C1=10^(−7); C2=C1; C3=2*C1; f0=1000; w0=2*pi*f0; ...
R1=1/(2*w0*Q*C1); R2=(2*Q)/(w0*C1); R3=(2*Q)/(C1*w0*(4*Q^2+1)); fprintf(' \n');...
fprintf('R1 = %5.0f Ohms \t',R1); fprintf('R2 = %5.0f Ohms \t',R2);...
fprintf('R3 = %5.0f Ohms \t',R3)
R1=80 Ohms R2=31831 Ohms R3=79 Ohms
We choose standard resistors as close as possible to those found above. These are shown in the
MATLAB code below. Part II of the code is as follows:
K=1; Q=10; f=10:10:10000; w=2*pi*f; R1=80.6; R2=31600; R3=78.7;...
w0=1/(2*R1*Q*C1); B=w0/Q; s=w*j; Gw=(K.*(s.^2+w0.^2))./(s.^2+B.*s+w0.^2);...
semilogx(f,abs(Gw)); axis([500 2000 0 1]); grid; hold on; xlabel('Frequency, Hz'),...
ylabel('|Vout/Vin|'), title('2nd Order Butterworth Band-Stop Filter Response')

Signals and Systems with MATLAB Applications, Second Edition 11-81

Orchard Publications
Chapter 11 Analog and Digital Filters
4. Let us first solve the relation
φ 0 = φ ( ω 0 ) = – 2tan ⎛ ------------ ⎞
–1 a
⎝ b – 1⎠

for b in terms of φ 0 and a so that we can find its value from the MATLAB code below. We
rewrite the above relation as
φ
tan ⎛ ------------ ⎞ = – -----0
–1 a
⎝b – 1⎠ 2
or
φ
tan ⎛ – -----0 ⎞ = ------------
a
⎝ 2⎠ b–1

φ φ
btan ⎛ – -----0 ⎞ = a + tan ⎛ – -----0 ⎞
⎝ 2⎠ ⎝ 2⎠

φ φ
b = a + tan ⎛ – -----0 ⎞ ⁄ tan ⎛ – -----0 ⎞
⎝ 2⎠ ⎝ 2⎠

% MFB 2nd order all-pass filter, f0=1 KHz

% phase shift phi=-pi/2 and gain K=3/4. Gain must be 0<K<1
% PART I − Find Resistor, a and b values
phi0=−pi/2; K=0.75; C1=10^(−8); C2=C1; f0=1000; w0=2*pi*f0;...
a=((1−K)/(2*K*tan(phi0/2)))*(−1−sqrt((1+4*K/(1−K))*(tan(phi0/2))^2));...
b=(a+tan(−phi0/2))/tan(−phi0/2); R2=2/(a*w0*C1); R1=(1−K)*R2/(4*K); ...
R3=R2/K; R4=R2/(1−K); fprintf(' \n'); fprintf('R1 = %6.0f Ohms \t',R1);...
fprintf('R2 = %6.0f Ohms \t',R2); fprintf('R3 = %6.0f Ohms \t',R3);...
fprintf('R4 = %6.0f Ohms \t',R4); fprintf(' \n');...
fprintf('a = %5.3f \t', a); fprintf('b = %5.3f \t', b)
R1=3456 Ohms R2=41469 Ohms R3=55292 Ohms R4=165875 Ohms
a = 0.768 b = 1.768
We choose standard resistors as close as possible to those found above. These are shown in the
MATLAB code below. Part II of the code is as follows:
K=3/4; a=0.768; b=1.768; C1=10^(−8); C2=C1; f=10:10:100000; w=2*pi*f;...
R1=3480; R2=41200; R3=54900; R4=165000; w0=2/(a*R2*C1); s=w*j;
Gw=(K.*(s.^2−a.*w0.*s+b.*w0.^2))./(s.^2+a.*w0.*s+b.*w0.^2);...
semilogx(f,angle(Gw)); grid; hold on; xlabel('Frequency, Hz'), ylabel('Phase Angle');...
title('2nd Order All-Pass Filter Response')
The plot shown below is the phase (not magnitude) response.

11-82 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

5. We will use MATLAB for all computations.

% MFB 2nd order Bessel filter, T0=100 microseconds, K=2
% PART I − Find resistor values
T0=100*10^(−6); K=2; C1=10^(−8); C2=2*10^(−9); a=3; b=3; w0=12/(13*T0);...
R2=(2*(K+1))/((a*C1+sqrt(a^2*C1^2−4*b*C1*C2*(K+1)))*w0); R1=R2/K;...
R3=1/(b*C1*C2*R2*w0^2); fprintf(' \n'); fprintf('R1 = %5.0f Ohms \t',R1);...
fprintf('R2 = %5.0f Ohms \t',R2); fprintf('R3 = %5.0f Ohms \t',R3)
R1 = 7486 Ohms R2 = 14971 Ohms R3 = 13065 Ohms
We choose standard resistors as close as possible to those found above. These are shown in the
MATLAB code below. Part II of the code is as follows:
K=2; a=3; b=3; C1=10^(−8); C2=2*10^(−9); f=1:10:100000; w=2*pi*f; R1=7500;...
R2=15000; R3=13000;...
w0=(2*(K+1))/((a*C1+sqrt(a^2*C1^2-4*b*C1*C2*(K+1)))*R2);...
s=w*j; Gw=(3.*K.*w0.^2)./(s.^2+a.*w0.*s+b.*w0.^2);...
semilogx(f,angle(Gw)); grid; hold on;...
xlabel('Frequency, Hz'), ylabel('Phase Angle'); title('2nd Order Bessel Filter Response')

Signals and Systems with MATLAB Applications, Second Edition 11-83

Orchard Publications
Chapter 11 Analog and Digital Filters
The plot shown below is the phase (not magnitude) response. This filter has very good phase
response but poor amplitude response. The group delay (the slope at a particular frequency) is
practically flat at frequencies near DC.

6. From (11.20)
2 1
A ( ω ) = ----------------------------------
-
2k
( ω ⁄ ωC ) + 1

and with ω C = 1 rad ⁄ s , and k = 4 we get

2 1
A ( ω ) = ---------------
8
ω +1
Then,
1
G ( s ) ⋅ G ( – s ) = --------------
8
s +1
8
We can use DeMoivre’s theorem to find the roots of s + 1 but we will use MATLAB instead.

11-84 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

syms s; y=solve('s^8+1=0'); fprintf(' \n'); disp('s1 = '); disp(simple(y(1)));...

disp('s2 = '); disp(simple(y(2))); disp('s3 = '); disp(simple(y(3)));...
disp('s4 = '); disp(simple(y(4))); disp('s5 = '); disp(simple(y(5)));...
disp('s6 = '); disp(simple(y(6))); disp('s7 = '); disp(simple(y(7)));...
disp('s8 = '); disp(simple(y(8)))
3
s1 = 1/2*2^(3/4)*(1+i)^(1/2) s1 = ( 1 ⁄ 2 ) ⋅ 4 2 ⋅ 1 + j
3
s2 = -1/2*2^(3/4)*(1+i)^(1/2) s2 = –( 1 ⁄ 2 ) ⋅ 4 2 ⋅ 1 + j
3
s3 = 1/2*i*2^(3/4)*(1+i)^(1/2) s 3 = ( 1 ⁄ 2 )j ⋅ 4 2 ⋅ 1 + j
3
s4 = -1/2*i*2^(3/4)*(1+i)^(1/2) s 4 = – ( 1 ⁄ 2 )j ⋅ 4 2 ⋅ 1 + j
3
s5 = 1/2*i*2^(3/4)*(-1+i)^(1/2) s 5 = ( 1 ⁄ 2 )j ⋅ 4 2 ⋅ – 1 + j
3
s6 = -1/2*i*2^(3/4)*(-1+i)^(1/2) s 6 = – ( 1 ⁄ 2 )j ⋅ 4 2 ⋅ – 1 + j
3
s7 = 1/2*2^(3/4)*(-1+i)^(1/2) s7 = ( 1 ⁄ 2 ) ⋅ 4 2 ⋅ – 1 + j
3
s8 = -1/2*2^(3/4)*(-1+i)^(1/2) s8 = –( 1 ⁄ 2 ) ⋅ 4 2 ⋅ – 1 + j

Since we are only interested in the poles of the left half of the s -plane, we choose the roots s 2 , s 4 ,
s 6 , and s 8 . To express the denominator in polynomial form we use the following MATLAB code:

denGs=(s-s2)*(s-s4)*(s-s6)*(s-s8); r=vpa(denGs,4)
r = (s+.9240+.3827*i)*(s+.3827-.9240*i)*(s+.9240-.3827*i)*(s+.3827+.9240*i)
expand(r)
ans =
s^4+2.6134*s^3+3.41492978*s^2+2.614014906886*s
+1.0004706353613841
and thus
1
G ( s ) = ----------------------------------------------------------------------------
-
4 3 2
s + 2.61s + 3.41s + 2.61s + 1
7. From (11.45),
2 α
A ( ω ) = --------------------------------------------- (1)
2 2
1 + ε Ck ( ω ⁄ ωC )

and with ω C = 1 and k = 3 , we find from (11.44) that

2 2 3 2
C k = C 3 = ( 4ω – 3ω )

Signals and Systems with MATLAB Applications, Second Edition 11-85

Orchard Publications
Chapter 11 Analog and Digital Filters

Also, from (11.52) with 1.5 dB ripple,

2 ( 1.5 ⁄ 10 )
ε = 10 – 1 = 0.4125
and with these values (1) is written as
2 α
A ( ω ) = ------------------------------------------------------------
-
3 2
1 + 0.4125 ⋅ ( 4ω – 3ω )
To express the denominator in polynomial form, we use the following MATLAB code.
syms w; denA=1+0.4125*expand((4*w^3-3*w)^2);...
denA = 1+33/5*w^6-99/10*w^4+297/80*w^2
and thus
2 α
A ( ω ) = -------------------------------------------------------------------------
6 4 2
6.6ω – 9.9ω + 3.7125ω + 1

8. With the approximation of (11.87) with ω C = ω a we find that ω C = ω a ⋅ T S = 4 × 0.25 = 1 .

However, using the exact relation of (11.86) and solving for ω d we find that
–1
ω d = 2 tan ( ω a T S ) ⁄ 2 = 0.9273

and this value is not very close to unity. Therefore we will compute G 1 ( z ) with pre-warping using
the following MATLAB code.
% This code designs a 2-pole Chebyshev Type 1 high-pass digital filter with
% analog cutoff frequency wc=4 rads/sec, sampling period Ts=0.25 sec., with
% pass band % ripple of 3 dB.
%
N=2; % # of poles
Rp=3; % Passband ripple in dB
Ts=0.25; % Sampling period
wc=4; % Analog cutoff frequency
% Let wd be the discrete time radian frequency. This frequency is related to
% the continuous time radian frequency wc by wd=Ts*wc with no pre-warping.
% With prewarping it is related to wc by wdp=2*arctan(wc*Ts/2).
% We divide by pi to normalize the digital cutoff frequency.
wdp=2*atan(wc*Ts/2)/pi;
% To obtain the digital cutoff frequency without prewarping we use the relation
% wd=(wc*Ts)/pi;
[Nz,Dz]=cheby1(N,Rp,wdp,'high');
%
fprintf('The numerator N(z) coefficients in descending powers of z are: \n\n');
fprintf('%8.4f \t',[Nz]); fprintf(' \n');
fprintf('The denominator D(z) coefficients in descending powers of z are: \n\n');

11-86 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

fprintf('%8.4f \t',[Dz]); fprintf(' \n');

%
fprintf('Press any key to see the graph \n');
pause;
%
w=0:2*pi/300:pi; Gz=freqz(Nz,Dz,w); plot(w,abs(Gz)); grid; xlabel('Frequency (rads/sec)');
ylabel('|H|'); title('High-Pass Digital Filter with prewarping')
The numerator N(z) coefficients in descending powers of z are:
0.3914 -0.7829 0.3914
The denominator D(z) coefficients in descending powers of z
are:
1.0000 -0.7153 0.4963
and thus the transfer function and the plot with pre-warping are as shown below.
2
0.3914z – 0.7829z + 0.3914
G 1 ( z ) = ---------------------------------------------------------------------
2
z – 0.7153z + 0.4963

Signals and Systems with MATLAB Applications, Second Edition 11-87

Orchard Publications
Chapter 11 Analog and Digital Filters

Next, we will compute G 2 ( z ) without pre-warping using the following MATLAB code:
N=2; % # of poles
Rp=3; % Pass band ripple in dB
Ts=0.25; % Sampling period
wc=4; % Analog cutoff frequency
wd=(wc*Ts)/pi;
[Nz,Dz]=cheby1(N,Rp,wd,'high');
%
fprintf('The numerator N(z) coefficients in descending powers of z are: \n\n');
fprintf('%8.4f \t',[Nz]); fprintf(' \n');
fprintf('The denominator D(z) coefficients in descending powers of z are: \n\n');
fprintf('%8.4f \t',[Dz]); fprintf(' \n');
%
fprintf('Press any key to see the graph \n');
pause;
%
w=0:2*pi/300:pi; Gz=freqz(Nz,Dz,w); plot(w,abs(Gz)); grid; xlabel('Frequency (rads/sec)');
ylabel('|H|'); title('High-Pass Digital Filter without prewarping')
The numerator N(z) coefficients in descending powers of z are:
0.3689 -0.7377 0.3689
The denominator D(z) coefficients in descending powers of z
are:
1.0000 -0.6028 0.4814
and thus the transfer function and the plot without pre-warping are as shown below.
2
0.3689z – 0.7377z + 0.3689
G 2 ( z ) = ---------------------------------------------------------------------
2
z – 0.6028z + 0.4814

11-88 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solutions to Exercises

Signals and Systems with MATLAB Applications, Second Edition 11-89

Orchard Publications
Chapter 11 Analog and Digital Filters
NOTES

11-90 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Appendix A
Introduction to MATLAB®

his appendix serves as an introduction to the basic MATLAB commands and functions, proce-
T dures for naming and saving the user generated files, comment lines, access to MATLAB’s Edi-
tor/Debugger, finding the roots of a polynomial, and making plots. Several examples are pro-
vided with detailed explanations.

A.1 MATLAB® and Simulink®

MATLAB and Simulink are products of The MathWorks™ Inc. These are two outstanding software
packages for scientific and engineering computations and are used in educational institutions and in
industries including automotive, aerospace, electronics, telecommunications, and environmental
applications. MATLAB enables us to solve many advanced numerical problems fast and efficiently.
Simulink is a block diagram tool used for modeling and simulating dynamic systems such as controls,
signal processing, and communications. In this appendix we will discuss MATLAB only.

A.2 Command Window

To distinguish the screen displays from the user commands, important terms, and MATLAB func-
tions, we will use the following conventions:
Click: Click the left button of the mouse

Courier Font: Screen displays

Helvetica Font: User inputs at MATLAB’s command window prompt >> or EDU>>*

Helvetica Bold: MATLAB functions

Times Bold Italic: Important terms and facts, notes and file names

When we first start MATLAB, we see the toolbar on top of the command screen and the prompt
EDU>>. This prompt is displayed also after execution of a command; MATLAB now waits for a new
command from the user. It is highly recommended that we use the Editor/Debugger to write our pro-
gram, save it, and return to the command screen to execute the program as explained below.
To use the Editor/Debugger:

* EDU>> is the MATLAB prompt in the Student Version

Signals and Systems with MATLAB Applications, Second Edition A-1

Orchard Publications
 Introduction to MATLAB®

1. From the File menu on the toolbar, we choose New and click on M-File. This takes us to the Editor
Window where we can type our code (list of statements) for a new file, or open a previously saved
file. We must save our program with a file name which starts with a letter. Important! MATLAB is
case sensitive, that is, it distinguishes between upper- and lower-case letters. Thus, t and T are two
different characters in MATLAB language. The files that we create are saved with the file name we
use and the extension .m; for example, myfile01.m. It is a good practice to save the code in a file
name that is descriptive of our code content. For instance, if the code performs some matrix oper-
ations, we ought to name and save that file as matrices01.m or any other similar name. We should
also use a floppy disk to backup our files.
2. Once the code is written and saved as an m-file, we may exit the Editor/Debugger window by click-
ing on Exit Editor/Debugger of the File menu. MATLAB then returns to the command window.
3. To execute a program, we type the file name without the .m extension at the >> prompt; then, we
press <enter> and observe the execution and the values obtained from it. If we have saved our
file in drive a or any other drive, we must make sure that it is added it to the desired directory in
MATLAB’s search path. The MATLAB User’s Guide provides more information on this topic.
Henceforth, it will be understood that each input command is typed after the >> prompt and fol-
lowed by the <enter> key.
The command help matlab\iofun will display input/output information. To get help with other
MATLAB topics, we can type help followed by any topic from the displayed menu. For example, to
get information on graphics, we type help matlab\graphics. The MATLAB User’s Guide contains
numerous help topics.
To appreciate MATLAB’s capabilities, we type demo and we see the MATLAB Demos menu. We
can do this periodically to become familiar with them. Whenever we want to return to the command
window, we click on the Close button.
When we are done and want to leave MATLAB, we type quit or exit. But if we want to clear all pre-
vious values, variables, and equations without exiting, we should use the command clear. This com-
mand erases everything; it is like exiting MATLAB and starting it again. The command clc clears the
screen but MATLAB still remembers all values, variables and equations that we have already used. In
other words, if we want to clear all previously entered commands, leaving only the >> prompt on the
upper left of the screen, we use the clc command.
All text after the % (percent) symbol is interpreted as a comment line by MATLAB, and thus it is
ignored during the execution of a program. A comment can be typed on the same line as the function
or command or as a separate line. For instance,
conv(p,q) % performs multiplication of polynomials p and q.
% The next statement performs partial fraction expansion of p(x) / q(x)
are both correct.

A-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Roots of Polynomials

One of the most powerful features of MATLAB is the ability to do computations involving complex
numbers. We can use either i , or j to denote the imaginary part of a complex number, such as 3-4i
or 3-4j. For example, the statement
z=3−4j
displays
z = 3.0000-4.0000i
In the above example, a multiplication (*) sign between 4 and j was not necessary because the com-
plex number consists of numerical constants. However, if the imaginary part is a function, or variable
such as cos ( x ) , we must use the multiplication sign, that is, we must type cos(x)*j or j*cos(x) for the
imaginary part of the complex number.

A.3 Roots of Polynomials

In MATLAB, a polynomial is expressed as a row vector of the form [ a n a n – 1 … a 2 a 1 a 0 ] . These
are the coefficients of the polynomial in descending order. We must include terms whose coefficients
are zero.

We find the roots of any polynomial with the roots(p) function; p is a row vector containing the
polynomial coefficients in descending order.
Example A.1
Find the roots of the polynomial
4 3 2
p 1 ( x ) = x – 10x + 35x – 50x + 24

Solution:
The roots are found with the following two statements where we have denoted the polynomial as p1,
and the roots as roots_ p1.
p1=[1 −10 35 −50 24] % Specify and display the coefficients of p1(x)
p1 =
1 -10 35 -50 24
roots_ p1=roots(p1) % Find the roots of p1(x)
roots_p1 =
4.0000
3.0000

Signals and Systems with MATLAB Applications, Second Edition A-3

Orchard Publications
 Introduction to MATLAB®

2.0000
1.0000
We observe that MATLAB displays the polynomial coefficients as a row vector, and the roots as a
column vector.
Example A.2
Find the roots of the polynomial
5 4 2
p 2 ( x ) = x – 7x + 16x + 25x + 52

Solution:
There is no cube term; therefore, we must enter zero as its coefficient. The roots are found with the
statements below, where we have defined the polynomial as p2, and the roots of this polynomial as
roots_ p2. The result indicates that this polynomial has three real roots, and two complex roots. Of
course, complex roots always occur in complex conjugate* pairs.
p2=[1 −7 0 16 25 52]
p2 =
1 -7 0 16 25 52
roots_ p2=roots(p2)
roots_ p2 =
6.5014
2.7428
-1.5711
-0.3366+ 1.3202i
-0.3366- 1.3202i

A.4 Polynomial Construction from Known Roots

We can compute the coefficients of a polynomial, from a given set of roots, with the poly(r) function
where r is a row vector containing the roots.

* By definition, the conjugate of a complex number A = a + jb is A∗ = a – jb . Refer also to Appendix B.

A-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Polynomial Construction from Known Roots

Example A.3
It is known that the roots of a polynomial are 1, 2, 3, and 4 . Compute the coefficients of this poly-
nomial.
Solution:
We first define a row vector, say r3 , with the given roots as elements of this vector; then, we find the
coefficients with the poly(r) function as shown below.
r3=[1 2 3 4] % Specify the roots of the polynomial
r3 =
1 2 3 4
poly_r3=poly(r3) % Find the polynomial coefficients
poly_r3 =
1 -10 35 -50 24
We observe that these are the coefficients of the polynomial p 1 ( x ) of Example A.1.

Example A.4
It is known that the roots of a polynomial are – 1, – 2, – 3, 4 + j5 and 4 – j5 . Find the coefficients
of this polynomial.
Solution:
We form a row vector, say r4 , with the given roots, and we find the polynomial coefficients with the
poly(r) function as shown below.
r4=[ −1 −2 −3 −4+5j −4−5j ]
r4 =
Columns 1 through 4
-1.0000 -2.0000 -3.0000 -4.0000+ 5.0000i
Column 5
-4.0000- 5.0000i
poly_r4=poly(r4)
poly_r4 =
1 14 100 340 499 246

Signals and Systems with MATLAB Applications, Second Edition A-5

Orchard Publications
 Introduction to MATLAB®

Therefore, the polynomial is

5 4 3 2
p 4 ( x ) = x + 14x + 100x + 340x + 499x + 246

A.5 Evaluation of a Polynomial at Specified Values

The polyval(p,x) function evaluates a polynomial p ( x ) at some specified value of the independent
variable x.
Example A.5
Evaluate the polynomial
6 5 3 2
p 5 ( x ) = x – 3x + 5x – 4x + 3x + 2 (A.1)
at x = – 3 .
Solution:
p5=[1 −3 0 5 −4 3 2]; % These are the coefficients
% The semicolon (;) after the right bracket suppresses the display of the row vector
% that contains the coefficients of p5.
%
val_minus3=polyval(p5, −3) % Evaluate p5 at x=−3; no semicolon is used here
% because we want the answer to be displayed
val_minus3 =
1280
Other MATLAB functions used with polynomials are the following:
conv(a,b) − multiplies two polynomials a and b

[q,r]=deconv(c,d) −divides polynomial c by polynomial d and displays the quotient q and remain-
der r.
polyder(p) − produces the coefficients of the derivative of a polynomial p.

Example A.6
Let
5 4 2
p 1 = x – 3x + 5x + 7x + 9

and

A-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Evaluation of a Polynomial at Specified Values
6 4 2
p 2 = 2x – 8x + 4x + 10x + 12

Compute the product p 1 ⋅ p 2 using the conv(a,b) function.

Solution:
p1=[1 −3 0 5 7 9]; % The coefficients of p1
p2=[2 0 −8 0 4 10 12]; % The coefficients of p2
p1p2=conv(p1,p2) % Multiply p1 by p2 to compute coefficients of the product p1p2
p1p2 =
2 -6 -8 34 18 -24 -74 -88 78 166 174 108
Therefore,
11 10 9 8 7 6
p 1 ⋅ p 2 = 2x – 6x – 8x + 34x + 18x – 24x
5 4 3 2
– 74x – 88x + 78x + 166x + 174x + 108

Example A.7
Let
7 5 3
p 3 = x – 3x + 5x + 7x + 9

and
6 5 2
p 4 = 2x – 8x + 4x + 10x + 12

Compute the quotient p 3 ⁄ p 4 using the [q,r]=deconv(c,d) function.

Solution:
% It is permissible to write two or more statements in one line separated by semicolons
p3=[1 0 −3 0 5 7 9]; p4=[2 −8 0 0 4 10 12]; [q,r]=deconv(p3,p4)
q =
0.5000
r =
0 4 -3 0 3 2 3
Therefore,
5 4 2
q = 0.5 r = 4x – 3x + 3x + 2x + 3

Signals and Systems with MATLAB Applications, Second Edition A-7

Orchard Publications
 Introduction to MATLAB®

Example A.8
Let
6 4 2
p 5 = 2x – 8x + 4x + 10x + 12

d
Compute the derivative ------ p 5 using the polyder(p) function.
dx

Solution:
p5=[2 0 −8 0 4 10 12]; % The coefficients of p5
der_p5=polyder(p5) % Compute the coefficients of the derivative of p5
der_p5 =
12 0 -32 0 8 10
Therefore,
d
------ p 5 = 12x 5 – 32x 3 + 4x 2 + 8x + 10
dx

A.6 Rational Polynomials

Rational Polynomials are those which can be expressed in ratio form, that is, as

n n–1 n–2
bn x + bn – 1 x + bn – 2 x + … + b1 x + b0
( x ) = ----------------------------------------------------------------------------------------------------------------------
R ( x ) = Num
------------------- - (A.2)
Den ( x ) m
am x + am – 1 x
m–1
+ am – 2 x
m–2
+ … + a1 x + a0

where some of the terms in the numerator and/or denominator may be zero. We can find the roots
of the numerator and denominator with the roots(p) function as before.
As noted in the comment line of Example A.7, we can write MATLAB statements in one line, if we
separate them by commas or semicolons. Commas will display the results whereas semicolons will
suppress the display.
Example A.9
Let

p num 5 4 2
x – 3x + 5x + 7x + 9
R ( x ) = -----------
- = --------------------------------------------------------
p den 6 4 2
x – 4x + 2x + 5x + 6

Express the numerator and denominator in factored form, using the roots(p) function.

A-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Rational Polynomials

Solution:
num=[1 −3 0 5 7 9]; den=[1 0 −4 0 2 5 6]; % Do not display num and den coefficients
roots_num=roots(num), roots_den=roots(den) % Display num and den roots

roots_num =
2.4186+ 1.0712i 2.4186- 1.0712i -1.1633
-0.3370+ 0.9961i -0.3370- 0.9961i
roots_den =
1.6760+0.4922i 1.6760-0.4922i -1.9304
-0.2108+0.9870i -0.2108-0.9870i -1.0000
As expected, the complex roots occur in complex conjugate pairs.
For the numerator, we have the factored form
p num = ( x – 2.4186 – j1.0712 ) ( x – 2.4186 + j1.0712 ) ( x + 1.1633 )
( x + 0.3370 – j0.9961 ) ( x + 0.3370 + j0.9961 )

and for the denominator, we have

p den = ( x – 1.6760 – j0.4922 ) ( x – 1.6760 + j0.4922 ) ( x + 1.9304 )
( x + 0.2108 – j 0.9870 ) ( x + 0.2108 + j0.9870 ) ( x + 1.0000 )

We can also express the numerator and denominator of this rational function as a combination of lin-
ear and quadratic factors. We recall that, in a quadratic equation of the form x 2 + bx + c = 0 whose
roots are x 1 and x 2 , the negative sum of the roots is equal to the coefficient b of the x term, that is,
– ( x 1 + x 2 ) = b , while the product of the roots is equal to the constant term c , that is, x 1 ⋅ x 2 = c .
Accordingly, we form the coefficient b by addition of the complex conjugate roots and this is done
by inspection; then we multiply the complex conjugate roots to obtain the constant term c using
MATLAB as follows:
(2.4186 + 1.0712i)*(2.4186 −1.0712i)
ans = 6.9971
(−0.3370+ 0.9961i)*(−0.3370−0.9961i)
ans = 1.1058
(1.6760+ 0.4922i)*(1.6760−0.4922i)
ans = 3.0512

Signals and Systems with MATLAB Applications, Second Edition A-9

Orchard Publications
 Introduction to MATLAB®

(−0.2108+ 0.9870i)*(−0.2108−0.9870i)
ans = 1.0186
Thus,
p num ( x – 4.8372x + 6.9971 ) ( x + 0.6740x + 1.1058 ) ( x + 1.1633 )
2 2
R ( x ) = -----------
- = ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
( x – 3.3520x + 3.0512 ) ( x + 0.4216x + 1.0186 ) ( x + 1.0000 ) ( x + 1.9304 )
p den 2 2

We can check this result with MATLAB’s Symbolic Math Toolbox which is a collection of tools (func-
tions) used in solving symbolic expressions. They are discussed in detail in MATLAB’s Users Manual.
For the present, our interest is in using the collect(s) function that is used to multiply two or more
symbolic expressions to obtain the result in polynomial form. We must remember that the
conv(p,q) function is used with numeric expressions only, that is, polynomial coefficients.

Before using a symbolic expression, we must create one or more symbolic variables such as x, y, t, and
so on. For our example, we use the following code:
syms x % Define a symbolic variable and use collect(s) to express numerator in polynomial
form
collect((x^2−4.8372*x+6.9971)*(x^2+0.6740*x+1.1058)*(x+1.1633))
ans =
x^5-29999/10000*x^4-1323/3125000*x^3+7813277909/
1562500000*x^2+1750276323053/250000000000*x+4500454743147/
500000000000
and if we simplify this, we find that is the same as the numerator of the given rational expression in
polynomial form. We can use the same procedure to verify the denominator.

A.7 Using MATLAB to Make Plots

Quite often, we want to plot a set of ordered pairs. This is a very easy task with the MATLAB
plot(x,y) command that plots y versus x. Here, x is the horizontal axis (abscissa) and y is the vertical
axis (ordinate).
Example A.10
Consider the electric circuit of Figure A.1, where the radian frequency ω (radians/second) of the
applied voltage was varied from 300 to 3000 in steps of 100 radians/second, while the amplitude
was held constant. The ammeter readings were then recorded for each frequency. The magnitude of
the impedance Z was computed as Z = V ⁄ A and the data were tabulated on Table A.1.

A-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Using MATLAB to Make Plots

<
1 V
<R C
`L F
Figure A.1. Electric circuit for Example A.10

TABLE A.1 Table for Example A.10

ω (rads/s) |Z| Ohms ω (rads/s) |Z| Ohms

300 39.339 1700 90.603
400 52.589 1800 81.088
500 71.184 1900 73.588
600 97.665 2000 67.513
700 140.437 2100 62.481
800 222.182 2200 58.240
900 436.056 2300 54.611
1000 1014.938 2400 51.428
1100 469.83 2500 48.717
1200 266.032 2600 46.286
1300 187.052 2700 44.122
1400 145.751 2800 42.182
1500 120.353 2900 40.432
1600 103.111 3000 38.845

Plot the magnitude of the impedance, that is, Z versus radian frequency ω .
Solution:
We cannot type ω (omega) in the MATLAB command window, so we will use the English letter w
instead.
If a statement, or a row vector is too long to fit in one line, it can be continued to the next line by typ-
ing three or more periods, then pressing <enter> to start a new line, and continue to enter data. This
is illustrated below for the data of w and z. Also, as mentioned before, we use the semicolon (;) to
suppress the display of numbers that we do not care to see on the screen.
The data are entered as follows:

Signals and Systems with MATLAB Applications, Second Edition A-11

Orchard Publications
 Introduction to MATLAB®

w=[300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900....
2000 2100 2200 2300 2400 2500 2600 2700 2800 2900 3000];
%
z=[39.339 52.789 71.104 97.665 140.437 222.182 436.056....
1014.938 469.830 266.032 187.052 145.751 120.353 103.111....
90.603 81.088 73.588 67.513 62.481 58.240 54.611 51.468....
48.717 46.286 44.122 42.182 40.432 38.845];
Of course, if we want to see the values of w or z or both, we simply type w or z, and we press
<enter>. To plot z (y-axis) versus w (x-axis), we use the plot(x,y) command. For this example, we
use plot(w,z). When this command is executed, MATLAB displays the plot on MATLAB’s graph
screen. This plot is shown in Figure A.2.

Figure A.2. Plot of impedance z versus frequency ω for Example A.10

This plot is referred to as the amplitude frequency response of the circuit.

To return to the command window, we press any key, or from the Window pull-down menu, we select
MATLAB Command Window. To see the graph again, we click on the Window pull-down menu, and
we select Figure.
We can make the above, or any plot, more presentable with the following commands:
grid on: This command adds grid lines to the plot. The grid off command removes the grid. The
command grid toggles them, that is, changes from off to on or vice versa. The default* is off.
box off: This command removes the box (the solid lines which enclose the plot), and box on
restores the box. The command box toggles them. The default is on.

A-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Using MATLAB to Make Plots

title(‘string’): This command adds a line of the text string (label) at the top of the plot.

xlabel(‘string’) and ylabel(‘string’) are used to label the x- and y-axis respectively.

The amplitude frequency response is usually represented with the x-axis in a logarithmic scale. We
can use the semilogx(x,y) command that is similar to the plot(x,y) command, except that the x-axis
is represented as a log scale, and the y-axis as a linear scale. Likewise, the semilogy(x,y) command is
similar to the plot(x,y) command, except that the y-axis is represented as a log scale, and the x-axis as
a linear scale. The loglog(x,y) command uses logarithmic scales for both axes.
Throughout this text it will be understood that log is the common (base 10) logarithm, and ln is the
natural (base e) logarithm. We must remember, however, the function log(x) in MATLAB is the nat-
ural logarithm, whereas the common logarithm is expressed as log10(x), and the logarithm to the
base 2 as log2(x).
Let us now redraw the plot with the above options by adding the following statements:
semilogx(w,z); grid; % Replaces the plot(w,z) command
title('Magnitude of Impedance vs. Radian Frequency');
xlabel('w in rads/sec'); ylabel('|Z| in Ohms')
After execution of these commands, our plot is as shown in Figure A.3.
If the y-axis represents power, voltage or current, the x-axis of the frequency response is more often
shown in a logarithmic scale, and the y-axis in dB (decibels).
To display the voltage v in dB units on the y-axis, we add the relation dB=20*log10(v), and we
replace the semilogx(w,z) command with semilogx(w,dB).

The command gtext(‘string’)* switches to the current Figure Window, and displays a cross-hair that
can be moved around with the mouse. For instance, we can use the command gtext(‘Impedance |Z|
versus Frequency’), and this will place a cross-hair in the Figure window. Then, using the mouse, we
can move the cross-hair to the position where we want our label to begin, and we press <enter>.

* A default is a particular value for a variable that is assigned automatically by an operating system and remains in
effect unless canceled or overridden by the operator.
* With MATLAB Versions 6 and higher we can add text, lines and arrows directly into the graph using the tools pro-
vided on the Figure Window.

Signals and Systems with MATLAB Applications, Second Edition A-13

Orchard Publications
 Introduction to MATLAB®

Figure A.3. Modified frequency response plot of Figure A.2.

The command text(x,y,’string’) is similar to gtext(‘string’). It places a label on a plot in some spe-
cific location specified by x and y, and string is the label which we want to place at that location. We
will illustrate its use with the following example that plots a 3-phase sinusoidal waveform.
The first line of the code below has the form
linspace(first_value, last_value, number_of_values)

This function specifies the number of data points but not the increments between data points. An
alternate function is
x=first: increment: last
and this specifies the increments between points but not the number of data points.
The code for the 3-phase plot is as follows:
x=linspace(0, 2*pi, 60); % pi is a built-in function in MATLAB;
% we could have used x=0:0.02*pi:2*pi or x = (0: 0.02: 2)*pi instead;
y=sin(x); u=sin(x+2*pi/3); v=sin(x+4*pi/3);
plot(x,y,x,u,x,v); % The x-axis must be specified for each function
grid on, box on, % turn grid and axes box on
text(0.75, 0.65, 'sin(x)'); text(2.85, 0.65, 'sin(x+2*pi/3)'); text(4.95, 0.65, 'sin(x+4*pi/3)')
These three waveforms are shown on the same plot of Figure A.4.

A-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Using MATLAB to Make Plots

Figure A.4. Three-phase waveforms

In our previous examples, we did not specify line styles, markers, and colors for our plots. However,
MATLAB allows us to specify various line types, plot symbols, and colors. These, or a combination
of these, can be added with the plot(x,y,s) command, where s is a character string containing one or
more characters shown on the three columns of Table A.2. MATLAB has no default color; it starts
with blue and cycles through the first seven colors listed in Table A.2 for each additional line in the
plot. Also, there is no default marker; no markers are drawn unless they are selected. The default line
is the solid line.
For example, plot(x,y,'m*:') plots a magenta dotted line with a star at each data point, and
plot(x,y,'rs') plots a red square at each data point, but does not draw any line because no line was
selected. If we want to connect the data points with a solid line, we must type plot(x,y,'rs−'). For
additional information we can type help plot in MATLAB’s command screen.
The plots we have discussed thus far are two-dimensional, that is, they are drawn on two axes. MAT-
LAB has also a three-dimensional (three-axes) capability and this is discussed next.
The plot3(x,y,z) command plots a line in 3-space through the points whose coordinates are the ele-
ments of x, y, and z, where x, y and z are three vectors of the same length.
The general format is plot3(x1,y1,z1,s1,x2,y2,z2,s2,x3,y3,z3,s3,...) where xn, yn and zn are vectors
or matrices, and sn are strings specifying color, marker symbol, or line style. These strings are the
same as those of the two-dimensional plots.
Example A.11
Plot the function
3 2
z = – 2x + x + 3y – 1 (A.3)

Signals and Systems with MATLAB Applications, Second Edition A-15

Orchard Publications
 Introduction to MATLAB®

TABLE A.2 Styles, colors, and markets used in MATLAB

Symbol Color Symbol Marker Symbol Line Style

b blue . point − solid line
g green o circle : dotted line
r red x x-mark −. dash-dot line
c cyan + plus −− dashed line
m magenta * star
y yellow s square
k black d diamond
w white ∨ triangle down
∧ triangle up
< triangle left
> triangle right
p pentagram
h hexagram

Solution:
We arbitrarily choose the interval (length) shown on the code below.
x= -10: 0.5: 10; % Length of vector x
y= x; % Length of vector y must be same as x
z= -2.*x.^3+x+3.*y.^2-1;% Vector z is function of both x and y*
plot3(x,y,z); grid
The three-dimensional plot is shown in Figure A.5
In a two-dimensional plot, we can set the limits of the x- and y-axes with the axis([xmin xmax ymin
ymax]) command. Likewise, in a three-dimensional plot we can set the limits of all three axes with
the axis([xmin xmax ymin ymax zmin zmax]) command. It must be placed after the plot(x,y) or
plot3(x,y,z) commands, or on the same line without first executing the plot command. This must be
done for each plot. The three-dimensional text(x,y,z,’string’) command will place string beginning
at the co-ordinate (x,y,z) on the plot.
For three-dimensional plots, grid on and box off are the default states.

* This statement uses the so called dot multiplication, dot division, and dot exponentiation where the multiplication,
division, and exponential operators are preceded by a dot. These operations will be explained in Section A.9.

A-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Using MATLAB to Make Plots

Figure A.5. Three dimensional plot for Example A.11

We can also use the mesh(x,y,z) command with two vector arguments. These must be defined as
length ( x ) = n and length ( y ) = m where [ m, n ] = size ( Z ) . In this case, the vertices of the mesh
lines are the triples { x ( j ), y ( i ), Z ( i, j ) } . We observe that x corresponds to the columns of Z , and y
corresponds to the rows.
To produce a mesh plot of a function of two variables, say z = f ( x, y ) , we must first generate the X
and Y matrices that consist of repeated rows and columns over the range of the variables x and y. We
can generate the matrices X and Y with the [X,Y]=meshgrid(x,y) function that creates the matrix
X whose rows are copies of the vector x, and the matrix Y whose columns are copies of the vector y.

Example A.12
The volume V of a right circular cone of radius r and height h is given by
2
V = ( 1 ⁄ 3 )πr h (A.4)

Plot the volume of the cone as r and h vary on the intervals 0 ≤ r ≤ 4 and 0 ≤ h ≤ 6 meters.
Solution:
The volume of the cone is a function of both the radius r and the height h, that is,
V = f ( r, h )

The three-dimensional plot is created with the following MATLAB code where, as in the previous
example, in the second line we have used the dot multiplication, dot division, and dot exponentiation.
As mentioned earlier, this will be explained in Section A.9.

Signals and Systems with MATLAB Applications, Second Edition A-17

Orchard Publications
 Introduction to MATLAB®

[R,H]=meshgrid(0: 4, 0: 6); % Creates R and H matrices from vectors r and h

V=(pi .* R .^ 2 .* H) ./ 3; mesh(R, H, V)
xlabel('x-axis, radius r (meters)'); ylabel('y-axis, altitude h (meters)');
zlabel('z-axis, volume (cubic meters)'); title('Volume of Right Circular Cone'); box on
The three-dimensional plot of Figure A.6, shows how the volume of the cone increases as the radius
and height are increased.

Figure A.6. Volume of a right circular cone.

This, and the plot of Figure A.5, are rudimentary; MATLAB can generate very sophisticated three-
dimensional plots. The MATLAB User’s manual contains more examples.

A.8 Subplots
MATLAB can display up to four windows of different plots on the Figure window using the com-
mand subplot(m,n,p). This command divides the window into an m × n matrix of plotting areas
and chooses the pth area to be active. No spaces or commas are required between the three integers
m, n and p. The possible combinations are shown in Figure A.7.

We will illustrate the use of the subplot(m,n,p) command following the discussion on multiplica-
tion, division and exponentiation that follows.

A.9 Multiplication, Division and Exponentiation

MATLAB recognizes two types of multiplication, division, and exponentiation. These are the matrix
multiplication, division, and exponentiation, and the element-by-element multiplication, division, and
exponentiation. They are explained in the following paragraphs.

A-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Multiplication, Division and Exponentiation

111
Default
Full Screen

211 221 222

121 122
212 223 224

221 222 211 221 222

122 121
212 223 224 223 224

Figure A.7. Possible subplot arrangements in MATLAB

In Section A.2, the arrays [ a b c … ] , such a those that contained the coefficients of polynomials,
consisted of one row and multiple columns, and thus are called row vectors. If an array has one col-
umn and multiple rows, it is called a column vector. We recall that the elements of a row vector are
separated by spaces. To distinguish between row and column vectors, the elements of a column vec-
tor must be separated by semicolons. An easier way to construct a column vector, is to write it first as
a row vector, and then transpose it into a column vector. MATLAB uses the single quotation charac-
ter (′) to transpose a vector. Thus, a column vector can be written either as b=[−1; 3; 6; 11] or as
b=[−1 3 6 11]'. MATLAB produces the same display with either format as shown below.
b=[−1; 3; 6; 11]
b =
-1
3
6
11
b=[−1 3 6 11]'
b =
-1
3
6
11
We will now define Matrix Multiplication and Element-by-Element multiplication.

Signals and Systems with MATLAB Applications, Second Edition A-19

Orchard Publications
 Introduction to MATLAB®

1. Matrix Multiplication (multiplication of row by column vectors)

Let
A = [ a1 a2 a3 … an ]
and
B = [ b 1 b 2 b 3 … b n ]'

be two vectors. We observe that A is defined as a row vector whereas B is defined as a column vec-
tor, as indicated by the transpose operator (′). Here, multiplication of the row vector A by the col-
umn vector B , is performed with the matrix multiplication operator (*). Then,

A*B = [ a 1 b 1 + a 2 b 2 + a 3 b 3 + … + a n b n ] = sin gle value (A.5)

For example, if
A = [1 2 3 4 5]
and
B = [ – 2 6 – 3 8 7 ]'

the matrix multiplication A*B produces the single value 68, that is,

A∗ B = 1 × ( – 2 ) + 2 × 6 + 3 × ( – 3 ) + 4 × 8 + 5 × 7 = 68

and this is verified with MATLAB as

A=[1 2 3 4 5]; B=[ −2 6 −3 8 7]'; A*B
ans =
68
Now, let us suppose that both A and B are row vectors, and we attempt to perform a row-by-row
multiplication using the following MATLAB statements
A=[1 2 3 4 5]; B=[−2 6 −3 8 7]; A*B
When these statements are executed, MATLAB displays the following message:
??? Error using ==> *
Inner matrix dimensions must agree.
Here, because we have used the matrix multiplication operator (*) in A*B, MATLAB expects vector
B to be a column vector, not a row vector. It recognizes that B is a row vector, and warns us that we
cannot perform this multiplication using the matrix multiplication operator (*). Accordingly, we must
perform this type of multiplication with a different operator. This operator is defined below.

A-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Multiplication, Division and Exponentiation

2.Element-by-Element Multiplication (multiplication of a row vector by another row vector)

Let
C = [ c1 c2 c3 … cn ]
and
D = [ d1 d2 d3 … dn ]

be two row vectors. Here, multiplication of the row vector C by the row vector D is performed with
the dot multiplication operator (.*). There is no space between the dot and the multiplication symbol.
Thus,

C.∗ D = [ c 1 d 1 c2 d2 c3 d3 … cn dn ] (A.6)
This product is another row vector with the same number of elements, as the elements of C and D .
As an example, let
C = [1 2 3 4 5]
and
D = [ –2 6 –3 8 7 ]

Dot multiplication of these two row vectors produce the following result.

C.∗ D = 1 × ( – 2 ) 2 × 6 3 × ( – 3 ) 4 × 8 5 × 7 = – 2 12 – 9 32 35
Check with MATLAB:
C=[1 2 3 4 5]; % Vectors C and D must have
D=[−2 6 −3 8 7]; % same number of elements
C.*D % We observe that this is a dot multiplication
ans =
-2 12 -9 32 35
Similarly, the division (/) and exponentiation (^) operators, are used for matrix division and exponen-
tiation, whereas dot division (./) and dot exponentiation (.^) are used for element-by-element divi-
sion and exponentiation.
We must remember that no space is allowed between the dot (.) and the multiplication, division, and
exponentiation operators.
Note: A dot (.) is never required with the plus (+) and minus (−) operators.
Example A.13
Write the MATLAB code that produces a simple plot for the waveform defined as
2
–4 t –3 t t
y = f ( t ) = 3e cos 5t – 2e sin 2t + ----------- (A.7)
t+1

Signals and Systems with MATLAB Applications, Second Edition A-21

Orchard Publications
 Introduction to MATLAB®

in the 0 ≤ t ≤ 5 seconds interval.

Solution:
The MATLAB code for this example is as follows:
t=0: 0.01: 5 % Define t-axis in 0.01 increments
y=3 .* exp(−4 .* t) .* cos(5 .* t)−2 .* exp(−3 .* t) .* sin(2 .* t) + t .^2 ./ (t+1);
plot(t,y); grid; xlabel('t'); ylabel('y=f(t)'); title('Plot for Example A.13')
Figure A.8 shows the plot for this example.

Figure A.8. Plot for Example A.13

Had we, in this example, defined the time interval starting with a negative value equal to or less than
– 1 , say as – 3 ≤ t ≤ 3 , MATLAB would have displayed the following message:

Warning: Divide by zero.

This is because the last term (the rational fraction) of the given expression, is divided by zero when
t = – 1 . To avoid division by zero, we use the special MATLAB function eps. This is a number
– 16
approximately equal to 2.2 × 10 . It will be used with the next example.
The command axis([xmin xmax ymin ymax]) scales the current plot to the values specified by the
arguments xmin, xmax, ymin and ymax. There are no commas between these four arguments. This
command must be placed after the plot command and must be repeated for each plot.
The following example illustrates the use of the dot multiplication, division, and exponentiation, the
eps number, the axis([xmin xmax ymin ymax]) command, and also MATLAB’s capability of dis-
playing up to four windows of different plots.

A-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Multiplication, Division and Exponentiation

Example A.14
Plot the functions

y = sin 2x, z = cos 2x, w = sin 2x ⋅ cos 2x, v = sin 2x ⁄ cos 2x

in the interval 0 ≤ x ≤ 2π using 100 data points. Use the subplot command to display these func-
tions as four windows on the same graph.
Solution:
The MATLAB code to produce the four subplots is as follows:
x=linspace(0,2*pi,100); % Interval with 100 data points
y=(sin(x).^ 2); z=(cos(x).^ 2);
w=y.* z;
v=y./ (z+eps); % add eps to avoid division by zero
subplot(221); % upper left of four subplots
plot(x,y); axis([0 2*pi 0 1]);
title('y=(sinx)^2');
subplot(222); % upper right of four subplots
plot(x,z); axis([0 2*pi 0 1]);
title('z=(cosx)^2');
subplot(223); % lower left of four subplots
plot(x,w); axis([0 2*pi 0 0.3]);
title('w=(sinx)^2*(cosx)^2');
subplot(224); % lower right of four subplots
plot(x,v); axis([0 2*pi 0 400]);
title('v=(sinx)^2/(cosx)^2');
These subplots are shown in Figure A.9.
The next example illustrates MATLAB’s capabilities with imaginary numbers. We will introduce the
real(z) and imag(z) functions that display the real and imaginary parts of the complex quantity z = x
+ iy, the abs(z), and the angle(z) functions that compute the absolute value (magnitude) and phase
angle of the complex quantity z = x + iy = r∠θ. We will also use the polar(theta,r) function that pro-
duces a plot in polar coordinates, where r is the magnitude, theta is the angle in radians, and the
round(n) function that rounds a number to its nearest integer.
Example A.15
Consider the electric circuit of Figure A.10.
With the given values of resistance, inductance, and capacitance, the impedance Z ab as a function of
the radian frequency ω can be computed from the following expression:

Signals and Systems with MATLAB Applications, Second Edition A-23

Orchard Publications
 Introduction to MATLAB®

Figure A.9. Subplots for the functions of Example A.14

a
<
10 Ω

Z ab
< 10 µF
10 Ω
0.1 H
` F
b

Figure A.10. Electric circuit for Example A.15

4 6
10 – j ( 10 ⁄ ω )
Z ab = Z = 10 + -------------------------------------------------------- (A.8)
5
10 + j ( 0.1ω – 10 ⁄ ω )

a. Plot Re { Z } (the real part of the impedance Z ) versus frequency ω .

b. Plot Im { Z } (the imaginary part of the impedance Z ) versus frequency ω .

c. Plot the impedance Z versus frequency ω in polar coordinates.
Solution:
The MATLAB code below computes the real and imaginary parts of Z ab that is, for simplicity,
denoted as z , and plots these as two separate graphs (parts a & b). It also displays a polar plot (part
c).

A-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Script and Function Files

w=0: 1: 2000; % Define interval with one radian interval

z=(10+(10 .^ 4 −j .* 10 .^ 6 ./ (w+eps)) ./ (10 + j .* (0.1 .* w −10.^5./ (w+eps))));
%
% The first five statements (next two lines) compute and plot Re{z}
real_part=real(z); plot(w,real_part); grid;
xlabel('radian frequency w'); ylabel('Real part of Z');
%
% The next five statements (next two lines) compute and plot Im{z}
imag_part=imag(z); plot(w,imag_part); grid;
xlabel('radian frequency w'); ylabel('Imaginary part of Z');
% The last six statements (next six lines) below produce the polar plot of z
mag=abs(z); % Computes |Z|
rndz=round(abs(z)); % Rounds |Z| to read polar plot easier
theta=angle(z); % Computes the phase angle of impedance Z
polar(theta,rndz); % Angle is the first argument
grid;
ylabel('Polar Plot of Z');
The real, imaginary, and polar plots are shown in Figures A.11, A.12, and A.13 respectively.

Figure A.11. Plot for the real part of the impedance in Example A.15

Example A.15 clearly illustrates how powerful, fast, accurate, and flexible MATLAB is.

A.10 Script and Function Files

MATLAB recognizes two types of files: script files and function files. Both types are referred to as m-
files since both require the .m extension.

Signals and Systems with MATLAB Applications, Second Edition A-25

Orchard Publications
 Introduction to MATLAB®

Figure A.12. Plot for the imaginary part of the impedance in Example A.15

Figure A.13. Polar plot of the impedance in Example A.15

A script file consists of two or more built-in functions such as those we have discussed thus far. Thus,
the code for each of the examples we discussed earlier, make up a script file. Generally, a script file is
one which was generated and saved as an m-file with an editor such as the MATLAB’s Editor/
Debugger.
A function file is a user-defined function using MATLAB. We use function files for repetitive tasks.
The first line of a function file must contain the word function, followed by the output argument, the
equal sign ( = ), and the input argument enclosed in parentheses. The function name and file name
must be the same, but the file name must have the extension .m. For example, the function file con-

A-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Script and Function Files

sisting of the two lines below

function y = myfunction(x)
y=x.^ 3 + cos(3.* x)
is a function file and must be saved as myfunction.m
For the next example, we will use the following MATLAB functions.
fzero(f,x) tries to find a zero of a function of one variable, where f is a string containing the name of
a real-valued function of a single real variable. MATLAB searches for a value near a point where the
function f changes sign, and returns that value, or returns NaN if the search fails.
Important: We must remember that we use roots(p) to find the roots of polynomials only, such as
those in Examples A.1 and A.2.

fmin(f,x1,x2)* minimizes a function of one variable. It attempts to return a value of x where f ( x ) is

minimum in the interval x 1 < x < x2 . The string f contains the name of the function to be minimized.

Note: MATLAB does not have a function to maximize a function of one variable, that is, there is no
fmax(f,x1,x2) function in MATLAB; but since a maximum of f ( x ) is equal to a minimum of – f ( x ) ,
we can use fmin(f,x1,x2) to find both minimum and maximum values of a function.
fplot(fcn,lims) plots the function specified by the string fcn between the x-axis limits specified by
lims = [xmin xmax]. Using lims = [xmin xmax ymin ymax] also controls the y-axis limits. The
string fcn must be the name of an m-file function or a string with variable x .
Note: NaN (Not-a-Number) is not a function; it is MATLAB’s response to an undefined expression
such as 0 ⁄ 0 , ∞ ⁄ ∞ , or inability to produce a result as described on the next paragraph. We can avoid
division by zero using the eps number, that we mentioned earlier.
Example A.16
Find the zeros, maxima and minima of the function
1
f ( x ) = --------------------------------------- 1
- + ---------------------------------------
- – 10
2 2
( x – 0.1 ) + 0.01 ( x – 1.2 ) + 0.04
Solution:
We first plot this function to observe the approximate zeros, maxima, and minima using the follow-
ing code.
x=−1.5: 0.01: 1.5; y=1./ ((x−0.1).^ 2 + 0.01) −1./ ((x−1.2).^ 2 + 0.04) −10; plot(x,y); grid

* This function is being replaced with the function x = fminbnd(fun,x1,x2). This function starts at X0 and finds a
local minimizer x of the function fun in the interval x1 < x < x2.

Signals and Systems with MATLAB Applications, Second Edition A-27

Orchard Publications
 Introduction to MATLAB®

The plot is shown in Figure A.14.

Figure A.14. Plot for Example A.16 using the plot command

The roots (zeros) of this function appear to be in the neighborhood of x = – 0.2 and x = 0.3 . The
maximum occurs at approximately x = 0.1 where, approximately, y max = 90 , and the minimum
occurs at approximately x = 1.2 where, approximately, y min = – 34 .

Next, we define and save f ( x ) as the funczero01.m function m-file with the following code:
function y=funczero01(x)
% Finding the zeros of the function shown below
y=1/((x−0.1)^2+0.01)−1/((x−1.2)^2+0.04)-10;
Now, we can use the fplot(fcn,lims) command to plot f ( x ) as follows.
fplot('funczero01', [−1.5 1.5]); grid
This plot is shown in Figure A.15. As expected, this plot is identical to the plot of Figure A.14 that
was obtained with the plot(x,y) command.
We will use the fzero(f,x) function to compute the roots of f ( x ) in (A.20) more precisely. The code
below must be saved with a file name, and then invoked with that file name.
x1= fzero('funczero01', −0.2);
x2= fzero('funczero01', 0.3);
fprintf('The roots (zeros) of this function are r1= %3.4f', x1);
fprintf(' and r2= %3.4f \n', x2)

A-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Script and Function Files

Figure A.15. Plot for Example A.16 using the fplot command
MATLAB displays the following:
The roots (zeros) of this function are r1= -0.1919 and r2= 0.3788
Whenever we use the fmin(f,x1,x2) or the fminbnd(f,x1,x2) function, we must remember that this
function searches for a minimum and it may display the values of local minima* , if any, before dis-
playing the function minimum. We should, therefore, plot the function with either the plot(x,y) or
the fplot(fcn,lims) command to find the smallest possible interval within which the function mini-
mum lies. For this example, we specify the range 0 ≤ x ≤ 1.5 rather than the interval – 1.5 ≤ x ≤ 1.5 .
The minimum of f ( x ) is found with the fmin(f,x1,x2) function as follows.
min_val=fmin('funczero01', 0, 1.5)
min_val = 1.2012
This is the value of x at which y = f ( x ) is minimum. To find the value of y corresponding to this
value of x, we substitute it into f ( x ) , that is,
x=1.2012; y=1 / ((x−0.1) ^ 2 + 0.01) −1 / ((x−1.2) ^ 2 + 0.04) −10
y = -34.1812
To find the maximum value, we must first define a new function m-file that will produce – f ( x ) . We
define it as follows:

* Local maxima or local minima, are the maximum or minimum values of a function within a restricted range of
values in the independent variable. When the entire range is considered, the maxima and minima are considered
be to the maximum and minimum values in the entire range in which the function is defined.

Signals and Systems with MATLAB Applications, Second Edition A-29

Orchard Publications
 Introduction to MATLAB®

function y=minusfunczero01(x)
% It is used to find maximum value from −f(x)
y=−(1/((x−0.1)^2+0.01)−1/((x−1.2)^2+0.04)−10);
We have placed the minus (−) sign in front of the right side of the last expression above, so that the
maximum value will be displayed. Of course, this is equivalent to the negative of the funczero01
function.
Now, we execute the following code to get the value of x where the maximum y = f ( x ) occurs.
max_val=fmin('minusfunczero01', 0,1)

max_val = 0.0999
x=0.0999; % Using this value find the corresponding value of y
y=1 / ((x−0.1) ^ 2 + 0.01) −1 / ((x−1.2) ^ 2 + 0.04) −10
y = 89.2000

A.11 Display Formats

MATLAB displays the results on the screen in integer format without decimals if the result is an inte-
ger number, or in short floating point format with four decimals if it a fractional number. The format
displayed has nothing to do with the accuracy in the computations. MATLAB performs all computa-
tions with accuracy up to 16 decimal places.
The output format can changed with the format command. The available formats can be displayed
with the help format command as follows:
help format

FORMAT Set output format.

All computations in MATLAB are done in double precision.
FORMAT may be used to switch between different output display formats as follows:
FORMAT Default. Same as SHORT.
FORMAT SHORT Scaled fixed point format with 5 digits.
FORMAT LONG Scaled fixed point format with 15 digits.
FORMAT SHORT E Floating point format with 5 digits.
FORMAT LONG E Floating point format with 15 digits.
FORMAT SHORT G Best of fixed or floating point format with 5 digits.

A-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Display Formats

FORMAT LONG G Best of fixed or floating point format with 15 digits.

FORMAT HEX Hexadecimal format.
FORMAT + The symbols +, - and blank are printed for positive, negative and zero elements.
Imaginary parts are ignored.
FORMAT BANK Fixed format for dollars and cents.
FORMAT RAT Approximation by ratio of small integers.
Spacing:
FORMAT COMPACT Suppress extra line-feeds.
FORMAT LOOSE Puts the extra line-feeds back in.
Some examples with different format displays age given below.
format short 33.3335 Four decimal digits (default)
format long 33.33333333333334 16 digits

format short e 3.3333e+01 Four decimal digits plus exponent

format short g 33.333 Better of format short or format short e

format bank 33.33 two decimal digits

format + only + or − or zero are printed

format rat 100/3 rational approximation

The disp(X) command displays the array X without printing the array name. If X is a string, the text
is displayed.
The fprintf(format,array) command displays and prints both text and arrays. It uses specifiers to
indicate where and in which format the values would be displayed and printed. Thus, if %f is used,
the values will be displayed and printed in fixed decimal format, and if %e is used, the values will be
displayed and printed in scientific notation format. With this command only the real part of each
parameter is processed.

Signals and Systems with MATLAB Applications, Second Edition A-31

Orchard Publications
 Introduction to MATLAB®

NOTES

A-32 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Appendix B
A Review of Complex Numbers

his appendix is a review of the algebra of complex numbers. The basic operations are defined
T and illustrated by several examples. Applications using Euler’s identities are presented, and the
exponential and polar forms are discussed and illustrated with examples.

B.1 Definition of a Complex Number

In the language of mathematics, the square root of minus one is denoted as i , that is, i = – 1 . In
the electrical engineering field, we denote i as j to avoid confusion with current i . Essentially, j is
an operator that produces a 90-degree counterclockwise rotation to any vector to which it is applied
as a multiplying factor. Thus, if it is given that a vector A has the direction along the right side of the
x-axis as shown in Figure B.1, multiplication of this vector by the operator j will result in a new vec-
tor jA whose magnitude remains the same, but it has been rotated counterclockwise by 90 ° . Also,
another multiplication of the new vector jA by j will produce another 90 ° counterclockwise direc-
tion. In this case, the vector A has rotated 180 ° and its new value now is – A . When this vector is
rotated by another 90 ° for a total of 270 ° , its value becomes j ( – A ) = – j A . A fourth 90 ° rotation
returns the vector to its original position, and thus its value is again A . Therefore, we conclude that
2 3 4
j = – 1 , j = – j , and j = 1 .

y
jA

j ( j A ) = j2 A = –A A
x
j ( –j A ) = –j A = A
2

j ( –A ) = j 3 A = –j A

Figure B.1. The j operator

Signals and Systems with MATLAB Applications, Second Edition B-1

Orchard Publications
 A Review of Complex Numbers

Note: In our subsequent discussion, we will designate the x-axis (abscissa) as the real axis, and the y-
axis (ordinate) as the imaginary axis with the understanding that the “imaginary” axis is just as “real”
as the real axis. In other words, the imaginary axis is just as important as the real axis.*
An imaginary number is the product of a real number, say r , by the operator j . Thus, r is a real num-
ber and jr is an imaginary number.
A complex number is the sum (or difference) of a real number and an imaginary number. For example,
the number A = a + jb where a and b are both real numbers, is a complex number. Then,
a = Re { A } and b = Im { A } where Re { A } denotes real part of A, and b = Im { A } the imagi-
nary part of A .

By definition, two complex numbers A and B where A = a + jb and B = c + jd , are equal if and
only if their real parts are equal, and also their imaginary parts are equal. Thus, A = B if and only if
a = c and b = d .

B.2 Addition and Subtraction of Complex Numbers

The sum of two complex numbers has a real component equal to the sum of the real components,
and an imaginary component equal to the sum of the imaginary components. For subtraction, we
change the signs of the components of the subtrahend and we perform addition. Thus, if
A = a + jb and B = c + jd
then
A + B = (a + c) + j(b + d)
and
A – B = (a – c) + j(b – d)

Example B.1
It is given that A = 3 + j 4 , and B = 4 – j 2 . Find A + B and A – B
Solution:
A + B = (3 + j4) + (4 – j2) = (3 + 4) + j(4 – 2) = 7 + j2
and
A – B = (3 + j4) – (4 – j2) = (3 – 4) + j(4 + 2) = – 1 + j6

* We may think the real axis as the cosine axis and the imaginary axis as the sine axis.

B-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Multiplication of Complex Numbers

B.3 Multiplication of Complex Numbers

Complex numbers are multiplied using the rules of elementary algebra, and making use of the fact
that j 2 = – 1 . Thus, if

A = a + jb and B = c + jd
then
A ⋅ B = ( a + jb ) ⋅ ( c + jd ) = ac + jad + jbc + j 2 bd

and since j 2 = – 1 , it follows that

A ⋅ B = ac + jad + jbc – b d
(B.1)
= ( ac – bd ) + j ( ad + bc )
Example B.2
It is given that A = 3 + j 4 and B = 4 – j 2 . Find A ⋅ B
Solution:

A ⋅ B = ( 3 + j 4 ) ⋅ ( 4 – j 2 ) = 12 – j 6 + j 16 – j 2 8 = 20 + j 10

The conjugate of a complex number, denoted as A∗ , is another complex number with the same real
component, and with an imaginary component of opposite sign. Thus, if A = a + jb , then
A∗ = a – j b .

Example B.3
It is given that A = 3 + j 5 . Find A∗
Solution:
The conjugate of the complex number A has the same real component, but the imaginary compo-
nent has opposite sign. Then, A∗ = 3 – j 5
If a complex number A is multiplied by its conjugate, the result is a real number. Thus, if A = a + jb ,
then

A ⋅ A∗ = ( a + jb ) ( a – jb ) = a 2 – jab + jab – j 2 b 2 = a + b
2 2

Example B.4
It is given that A = 3 + j 5 . Find A ⋅ A∗
Solution:
A ⋅ A∗ = ( 3 + j 5 ) ( 3 – j 5 ) = 3 + 5 = 9 + 25 = 34
2 2

Signals and Systems with MATLAB Applications, Second Edition B-3

Orchard Publications
 A Review of Complex Numbers

B.4 Division of Complex Numbers

When performing division of complex numbers, it is desirable to obtain the quotient separated into a
real part and an imaginary part. This procedure is called rationalization of the quotient, and it is done
by multiplying the denominator by its conjugate. Thus, if A = a + jb and B = c + jd , then,

A a + jb ( a + jb ) ( c – jd ) A B∗ ( ac + bd ) + j ( bc – ad )
--- = -------------- = -------------------------------------- = --- ⋅ ------ = -------------------------------------------------------
B c + jd ( c + jd ) ( c – jd ) B B∗ 2
c +d
2
(B.2)
( ac + bd -) j (----------------------
bc – ad )
= ----------------------
2 2
+ 2 2
-
c +d c + d
In (B.2), we multiplied both the numerator and denominator by the conjugate of the denominator to
eliminate the j operator from the denominator of the quotient. Using this procedure, we see that the
quotient is easily separated into a real and an imaginary part.

Example B.5
It is given that A = 3 + j 4 , and B = 4 + j 3 . Find A ⁄ B
Solution:
Using the procedure of expression (B.2), we get

A 3 + j 4 ( 3 + j 4 ) ( 4 – j 3 ) 12 – j 9 + j 16 + 12 24 + j 7 7
--- = -------------- = -------------------------------------- = -------------------------------------------- = ----------------- = 24
------ + j ------ = 0.96 + j 0.28
B 4 + j3 (4 + j3)(4 – j3) 4 +3
2 2 25 25 25

B.5 Exponential and Polar Forms of Complex Numbers

The relations

jθ
e = cos θ + j sin θ (B.3)

and

– jθ
e = cos θ – j sin θ (B.4)

are known as the Euler’s identities.

Multiplying (B.3) by the real positive constant C we get:

jθ
Ce = C cos θ + j C sin θ (B.5)

B-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exponential and Polar Forms of Complex Numbers

This expression represents a complex number, say a + jb , and thus

jθ
Ce = a + jb (B.6)
where the left side of (B.6) is the exponential form, and the right side is the rectangular form.
Equating real and imaginary parts in (B.5) and (B.6), we get
a = C cos θ and b = C sin θ (B.7)
Squaring and adding the expressions in (B.7), we get
2 2
a + b = ( C cos θ ) + ( C sin θ ) = C ( cos θ + sin θ ) = C
2 2 2 2 2 2

Then,
2 2 2
C = a +b
or

2 2
C = a +b (B.8)

Also, from (B.7)

b C sin θ
--- = --------------- = tan θ
a C cos θ
or

θ = tan ⎛ --- ⎞
–1 b
(B.9)
⎝a⎠

To convert a complex number from rectangular to exponential form, we use the expression

j ⎛ tan
–1 b⎞
---
2 2 ⎝ a⎠
a + jb = a +b e (B.10)

To convert a complex number from exponential to rectangular form, we use the expressions

jθ
Ce = C cos θ + j C sin θ
– jθ
(B.11)
Ce = C cos θ – j C sin θ

The polar form is essentially the same as the exponential form but the notation is different, that is,

jθ
Ce = C ∠θ (B.12)

Signals and Systems with MATLAB Applications, Second Edition B-5

Orchard Publications
 A Review of Complex Numbers

where the left side of (B.12) is the exponential form, and the right side is the polar form.
We must remember that the phase angle θ is always measured with respect to the positive real axis, and
rotates in the counterclockwise direction.

Example B.6
Convert the following complex numbers to exponential and polar forms:
a. 3 + j 4

b. – 1 + j 2

c. – 2 – j

d. 4 – j 3
Solution:
a. The real and imaginary components of this complex number are shown in Figure B.2.

Im
4

53.1°
Re
3

Figure B.2. The components of 3 + j 4

Then,
–1
j ⎛ tan 4 --- ⎞⎠
3 + j4 =
2
3 +4 e ⎝ 2
3 = 5e j53.1 ° = 5 ∠53.1°

Check with MATLAB:

x=3+j*4; magx=abs(x); thetax=angle(x)*180/pi; disp(magx); disp(thetax)
5
53.1301
b. The real and imaginary components of this complex number are shown in Figure B.3.

Then,
–1
j ⎛ tan -----
2⎞
- j116.6 °
– 1 + j2 =
2
1 +2 e ⎝
2 – 1⎠ = 5e = 5 ∠116.6°

B-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exponential and Polar Forms of Complex Numbers

Im
2
5

116.6°
63.4°
Re
−1

Figure B.3. The components of – 1 + j 2

Check with MATLAB:
y=−1+j*2; magy=abs(y); thetay=angle(y)*180/pi; disp(magy); disp(thetay)
2.2361
116.5651
c. The real and imaginary components of this complex number are shown in Figure B.4.

Im
206.6°
−2 Re
26.6°
−153.4°(Measured
5 Clockwise)
−1

Figure B.4. The components of – 2 – j

Then,
j ⎛ tan –
–1 1 ⎞
------
2 2 ⎝ –2 ⎠ j206.6° j ( – 153.4 )°
–2 –j 1 = 2 +1 e = 5e = 5 ∠206.6° = 5e = 5 ∠– 153.4 °
Check with MATLAB:
v=−2−j*1; magv=abs(v); thetav=angle(v)*180/pi; disp(magv); disp(thetav)
2.2361
-153.4349
d. The real and imaginary components of this complex number are shown in Figure B.5.

Then,
–1
⎛tan – -----3-⎞⎠
4 = 5e j323.1 ° = 5 ∠323.1° = 5e –j36.9 ° = 5 ∠– 36.9 °
j
4 –j 3 =
2 2
4 +3 e ⎝

Signals and Systems with MATLAB Applications, Second Edition B-7

Orchard Publications
 A Review of Complex Numbers

Im
323.1° 4
Re
−36.9°

5
−3

Figure B.5. The components of 4 – j 3

Check with MATLAB:
w=4−j*3; magw=abs(w); thetaw=angle(w)*180/pi; disp(magw); disp(thetaw)
5
-36.8699

Example B.7
Express the complex number – 2 ∠30° in exponential and in rectangular forms.
Solution:

We recall that – 1 = j 2 . Since each j rotates a vector by 90 ° counterclockwise, then – 2 ∠30° is the
same as 2 ∠30° rotated counterclockwise by 180 ° . Therefore,
– 2 ∠30° = 2 ∠( 30° + 180° ) = 2 ∠210° = 2 ∠– 150°
The components of this complex number are shown in Figure B.6.
Im
210°
−1.73 Re
30°
−150°(Measured
2 Clockwise)
−1

Figure B.6. The components of 2 ∠– 150°

Then,
– j 150 °
2 ∠– 150 ° = 2e = 2 ( cos 150° – j sin 150° ) = 2 ( – 0.866 – j0.5 ) = – 1.73 – j
Note: The rectangular form is most useful when we add or subtract complex numbers; however, the
exponential and polar forms are most convenient when we multiply or divide complex numbers.
To multiply two complex numbers in exponential (or polar) form, we multiply the magnitudes and
we add the phase angles, that is, if
A = M ∠θ and B = N ∠φ

B-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Exponential and Polar Forms of Complex Numbers

then,

jθ jφ j(θ + φ)
AB = MN ∠( θ + φ ) = M e Ne = MN e (B.13)

Example B.8
Multiply A = 10 ∠53.1° by B = 5 ∠– 36.9°
Solution:
Multiplication in polar form yields
AB = ( 10 × 5 ) ∠[ 53.1° + ( – 36.9° ) ] = 50 ∠16.2°
and multiplication in exponential form yields
j53.1° – j 36.9° j ( 53.1° – 36.9° ) j16.2°
AB = ( 10 e ) ( 5e ) = 50 e = 50 e
To divide one complex number by another when both are expressed in exponential or polar form,
we divide the magnitude of the dividend by the magnitude of the divisor, and we subtract the phase
angle of the divisor from the phase angle of the dividend, that is, if
A = M ∠θ and B = N ∠φ
then,

jθ
A M Me M j(θ – φ )
--- = ----- ∠( θ – φ ) = ------------- = ---- e (B.14)
B N jφ N
Ne

Example B.9
Divide A = 10 ∠53.1° by B = 5 ∠– 36.9°
Solution:
Division in polar form yields
A 10 ∠53.1°
--- = ------------------------ = 2 ∠[ 53.1° – ( – 36.9° ) ] = 2 ∠90°
B 5 ∠– 36.9°
Division in exponential form yields
j53.1°
A 10 e j53.1° j36.9° j90°
--- = --------------------- = 2e e = 2e
B – j36.9°
5e

Signals and Systems with MATLAB Applications, Second Edition B-9

Orchard Publications
A Review of Complex Numbers

NOTES

B-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Appendix C
Matrices and Determinants

his appendix is an introduction to matrices and matrix operations. Determinants, Cramer’s

T rule, and Gauss’s elimination method are reviewed. Some definitions and examples are not
applicable to subsequent material presented in this text, but are included for subject continuity,
and reference to more advance topics in matrix theory. These are denoted with a dagger (†) and may
be skipped.

C.1 Matrix Definition

A matrix is a rectangular array of numbers such as those shown below.

1 3 1
2 3 7 or –2 1 –5
1 –1 5
4 –7 6

In general form, a matrix A is denoted as

a 11 a 12 a 13 … a 1 n
a 21 a 22 a 23 … a 2 n
A = a 31 a 32 a 33 … a 3 n (C.1)
… … … … …
a m 1 a m 2 a m 3 … a mn

The numbers a ij are the elements of the matrix where the index i indicates the row, and j indicates
the column in which each element is positioned. For instance, a 43 indicates the element positioned
in the fourth row and third column.
A matrix of m rows and n columns is said to be of m × n order matrix.
If m = n , the matrix is said to be a square matrix of order m (or n ). Thus, if a matrix has five rows
and five columns, it is said to be a square matrix of order 5.
In a square matrix, the elements a 11, a 22, a 33, …, a nn are called the main diagonal elements. Alter-
nately, we say that the matrix elements a 11, a 22, a 33, …, a nn , are located on the main diagonal.

Signals and Systems with MATLAB Applications, Second Edition C-1

Orchard Publications
 Matrices and Determinants

† The sum of the diagonal elements of a square matrix A is called the trace* of A .
† A matrix in which every element is zero, is called a zero matrix.

C.2 Matrix Operations

Two matrices A = a ij and B = b ij are equal, that is, A = B , if and only if

a ij = b ij i = 1 , 2 , 3 , …, m j = 1, 2, 3, …, n (C.2)

Two matrices are said to be conformable for addition (subtraction), if they are of the same order m × n .

If A = a ij and B = b ij are conformable for addition (subtraction), their sum (difference) will be
another matrix C with the same order as A and B , where each element of C is the sum (difference)
of the corresponding elements of A and B , that is,
C = A ± B = [ a ij ± b ij ] (C.3)

Example C.1
Compute A + B and A – B given that

A = 1 2 3 and B = 2 3 0
0 1 4 –1 2 5

Solution:

A+B = 1+2 2+3 3+0 = 3 5 3

0–1 1+2 4+5 –1 3 9
and

A–B = 1–2 2 – 3 3 – 0 = –1 –1 3
0+1 1–2 4–5 1 –1 –1
Check with MATLAB:
A=[1 2 3; 0 1 4]; B=[2 3 0; −1 2 5]; % Define matrices A and B
A+B % Add A and B

* Henceforth, all paragraphs and topics preceded by a dagger ( † ) may be skipped. These are discussed in matrix
theory textbooks.

C-2 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Matrix Operations

ans =
3 5 3
-1 3 9

A−B % Subtract B from A

ans =
-1 -1 3
1 -1 -1

If k is any scalar (a positive or negative number), and not [k] which is a 1 × 1 matrix, then multipli-
cation of a matrix A by the scalar k is the multiplication of every element of A by k .

Example C.2
Multiply the matrix

A = 1 –2
2 3
by
a. k 1 = 5

b. k 2 = – 3 + j2

Solution:
a.

k 1 ⋅ A = 5 × 1 – 2 = 5 × 1 5 × ( – 2 ) = 5 – 10
2 3 5×2 5×3 10 15

b.

k 2 ⋅ A = ( – 3 + j2 ) × 1 – 2 = ( – 3 + j2 ) × 1 ( – 3 + j2 ) × ( – 2 ) = – 3 + j2 6 – j4
2 3 ( – 3 + j2 ) × 2 ( – 3 + j2 ) × 3 – 6 + j4 – 9 + j6

Check with MATLAB:

k1=5; k2=(−3 + 2*j); % Define scalars k1 and k2
A=[1 −2; 2 3]; % Define matrix A
k1*A % Multiply matrix A by constant k1
ans =
5 -10

Signals and Systems with MATLAB Applications, Second Edition C-3

Orchard Publications
 Matrices and Determinants

10 15
k2*A %Multiply matrix A by constant k2

ans =
-3.0000+ 2.0000i 6.0000- 4.0000i
-6.0000+ 4.0000i -9.0000+ 6.0000i

Two matrices A and B are said to be conformable for multiplication A ⋅ B in that order, only when the
number of columns of matrix A is equal to the number of rows of matrix B . That is, the product
A ⋅ B (but not B ⋅ A ) is conformable for multiplication only if A is an m × p matrix and matrix B is
an p × n matrix. The product A ⋅ B will then be an m × n matrix. A convenient way to determine if
two matrices are conformable for multiplication is to write the dimensions of the two matrices side-
by-side as shown below.
Shows that A and B are conformable for multiplication

A B
m×p p×n

Indicates the dimension of the product A ⋅ B

For the product B ⋅ A we have:

Here, B and A are not conformable for multiplication

B A
p×n m×p

For matrix multiplication, the operation is row by column. Thus, to obtain the product A ⋅ B , we
multiply each element of a row of A by the corresponding element of a column of B ; then, we add
these products.

Example C.3
Matrices C and D are defined as

1
C = 2 3 4 and D = – 1
2
Compute the products C ⋅ D and D ⋅ C

C-4 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Special Forms of Matrices

Solution:
The dimensions of matrices C and D are respectively 1 × 3 3 × 1 ; therefore the product C ⋅ D is
feasible, and will result in a 1 × 1 , that is,

1
C ⋅ D = 2 3 4 –1 = ( 2 ) ⋅ ( 1 ) + ( 3 ) ⋅ ( –1 ) + ( 4 ) ⋅ ( 2 ) = 7
2

The dimensions for D and C are respectively 3 × 1 1 × 3 and therefore, the product D ⋅ C is also
feasible. Multiplication of these will produce a 3 × 3 matrix as follows:

1 (1) ⋅ (2) (1) ⋅ (3) (1) ⋅ (4) 2 3 4

D ⋅ C = –1 2 3 4 = ( –1 ) ⋅ ( 2 ) ( –1 ) ⋅ ( 3 ) ( –1 ) ⋅ ( 4 ) = –2 –3 –4
2 (2) ⋅ (2) (2) ⋅ (3) (2) ⋅ (4) 4 6 8

Check with MATLAB:

C=[2 3 4]; D=[1; −1; 2]; % Define matrices C and D
C*D % Multiply C by D
ans =
7
D*C % Multiply D by C
ans =
2 3 4
-2 -3 -4
4 6 8
Division of one matrix by another, is not defined. However, an equivalent operation exists, and it
will become apparent later in this chapter, when we discuss the inverse of a matrix.

C.3 Special Forms of Matrices

† A square matrix is said to be upper triangular when all the elements below the diagonal are zero.
The matrix A of (C.4) is an upper triangular matrix.
In an upper triangular matrix, not all elements above the diagonal need to be non-zero.
† A square matrix is said to be lower triangular, when all the elements above the diagonal are zero.
The matrix B of (C.5) is a lower triangular matrix.

Signals and Systems with MATLAB Applications, Second Edition C-5

Orchard Publications
 Matrices and Determinants

a 11 a 12 a 13 … a 1 n
0 a 22 a 23 … a 2 n
A = 0 0 … … … (C.4)
… … 0 … …
0 0 0 … a mn

a 11 0 0 … 0
a 21 a 22 0 … 0
B = … … … 0 0 (C.5)
… … … … 0
am1 am 2 a m 3 … a mn

In a lower triangular matrix, not all elements below the diagonal need to be non-zero.
† A square matrix is said to be diagonal, if all elements are zero, except those in the diagonal. The
matrix C of (C.6) is a diagonal matrix.

a 11 0 0 … 0
0 a 22 0 … 0
C = 0 0 … 0 0 (C.6)
0 0 0 … 0
0 0 0 … a mn

† A diagonal matrix is called a scalar matrix, if a 11 = a 22 = a 33 = … = a nn = k where k is a sca-

lar. The matrix D of (C.7) is a scalar matrix with k = 4 .

4 0 0 0
D = 0 4 0 0 (C.7)
0 0 4 0
0 0 0 4

A scalar matrix with k = 1 , is called an identity matrix I . Shown below are 2 × 2 , 3 × 3 , and 4 × 4
identity matrices.

C-6 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Special Forms of Matrices

1 0 0 0
1 0 0
1 0 0 1 0 0 (C.8)
0 1 0
0 1 0 0 1 0
0 0 1
0 0 0 1

The MATLAB eye(n) function displays an n × n identity matrix. For example,

eye(4) % Display a 4 by 4 identity matrix
ans =

1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
Likewise, the eye(size(A)) function, produces an identity matrix whose size is the same as matrix
A . For example, let matrix A be defined as
A=[1 3 1; −2 1 −5; 4 −7 6] % Define matrix A

A =
1 3 1
-2 1 -5
4 -7 6
then,
eye(size(A))
displays
ans =

1 0 0
0 1 0
0 0 1

† The transpose of a matrix A , denoted as A T , is the matrix that is obtained when the rows and col-
umns of matrix A are interchanged. For example, if

1 4
A=
1 2 3 then A T = (C.9)
2 5
4 5 6
3 6

Signals and Systems with MATLAB Applications, Second Edition C-7

Orchard Publications
 Matrices and Determinants

In MATLAB we use the apostrophe (′) symbol to denote and obtain the transpose of a matrix. Thus,
for the above example,
A=[1 2 3; 4 5 6] % Define matrix A
A =
1 2 3
4 5 6
A' % Display the transpose of A
ans =
1 4
2 5
3 6

† A symmetric matrix A is a matrix such that A T = A , that is, the transpose of a matrix A is the same
as A . An example of a symmetric matrix is shown below.

1 2 3 1 2 3
A = 2 4 –5 A = 2 4 –5 = A
T
(C.10)
3 –5 6 3 –5 6

† If a matrix A has complex numbers as elements, the matrix obtained from A by replacing each
element by its conjugate, is called the conjugate of A , and it is denoted as A∗
An example is shown below.

A = 1 + j2 j A∗ = 1 – j 2 –j
3 2 – j3 3 2 + j3

MATLAB has two built-in functions which compute the complex conjugate of a number. The
first, conj(x), computes the complex conjugate of any complex number, and the second, conj(A),
computes the conjugate of a matrix A . Using MATLAB with the matrix A defined as above, we
get
A = [1+2j j; 3 2−3j] % Define and display matrix A
A =
1.0000+ 2.0000i 0+ 1.0000i
3.0000 2.0000- 3.0000i
conj_A=conj(A) % Compute and display the conjugate of A
conj_A =

C-8 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Determinants

1.0000- 2.0000i 0- 1.0000i

3.0000 2.0000+ 3.0000i

† A square matrix A such that A T = – A is called skew-symmetric. For example,

0 2 –3 0 –2 3
T
A = –2 0 –4 A = 2 0 4 = –A
3 4 0 –3 –4 0

Therefore, matrix A above is skew symmetric.

† A square matrix A such that A T∗ = A is called Hermitian. For example,

1 1–j 2 1 1+j 2 1 1+j 2

T T*
A = 1+j 3 j A = 1–j 3 –j A = 1–j 3 –j = A
2 –j 0 2 j 0 2 j 0

Therefore, matrix A above is Hermitian.

† A square matrix A such that A T∗ = – A is called skew−Hermitian. For example,

j 1–j 2 j –1–j –2 –j –1+j –2

T T*
A = –1–j 3j j A = 1–j 3j j A = 1+j – 3j –j = –A
–2 j 0 2 j 0 2 –j 0

Therefore, matrix A above is skew-Hermitian.

C.4 Determinants
Let matrix A be defined as the square matrix

a 11 a 12 a 13 … a 1 n
a 21 a 22 a 23 … a 2 n
A = a a a … a (C.11)
31 32 33 3n
… … … … …
a n 1 a n 2 a n 3 … a nn

then, the determinant of A , denoted as detA , is defined as

Signals and Systems with MATLAB Applications, Second Edition C-9

Orchard Publications
 Matrices and Determinants

detA = a 11 a 22 a 33 …a nn + a 12 a 23 a 34 …a n 1 + a 13 a 24 a 35 …a n 2 + … (C.12)
– a n 1 …a 22 a 13 … – a n 2 …a 23 a 14 – a n 3 …a 24 a 15 – …

The determinant of a square matrix of order n is referred to as determinant of order n.

Let A be a determinant of order 2 , that is,

a 11 a 12
A = (C.13)
a 21 a 22

Then,
detA = a 11 a 22 – a 21 a 12 (C.14)

Example C.4
Matrices A and B are defined as

A = 1 2 and B = 2 – 1
3 4 2 0
Compute detA and detB .
Solution:
detA = 1 ⋅ 4 – 3 ⋅ 2 = 4 – 6 = – 2
detB = 2 ⋅ 0 – 2 ⋅ ( – 1 ) = 0 – ( – 2 ) = 2
Check with MATLAB:
A=[1 2; 3 4]; B=[2 −1; 2 0]; % Define matrices A and B
det(A) % Compute the determinant of A
ans =
-2
det(B) % Compute the determinant of B
ans =
2
Let A be a matrix of order 3 , that is,

a 11 a 12 a 13
A = a 21 a 22 a 23 (C.15)
a 31 a 32 a 33

C-10 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Determinants

then, detA is found from

detA = a 11 a 22 a 33 + a 12 a 23 a 31 + a 11 a 22 a 33
(C.16)
– a 11 a 22 a 33 – a 11 a 22 a 33 – a 11 a 22 a 33

A convenient method to evaluate the determinant of order 3 , is to write the first two columns to
the right of the 3 × 3 matrix, and add the products formed by the diagonals from upper left to lower
right; then subtract the products formed by the diagonals from lower left to upper right as shown
on the diagram of the next page. When this is done properly, we obtain (C.16) above.
−
a 11 a 12 a 13 a 11 a 12
a 21 a 22 a 23 a 21 a 22
a 31 a 32 a 33 a 31 a 32 +
This method works only with second and third order determinants. To evaluate higher order deter-
minants, we must first compute the cofactors; these will be defined shortly.

Example C.5
Compute detA and detB if matrices A and B are defined as

2 3 5 2 –3 –4
A = 1 0 1 and B = 1 0 –2
2 1 0 0 –5 –6

Solution:

2 3 5 2 3
detA = 1 0 1 1 0
2 1 0 2 1

or
detA = ( 2 × 0 × 0 ) + ( 3 × 1 × 1 ) + ( 5 × 1 × 1 )
– ( 2 × 0 × 5 ) – ( 1 × 1 × 2 ) – ( 0 × 1 × 3 ) = 11 – 2 = 9
Likewise,
2 –3 –4 2 –3
detB = 1 0 – 2 1 – 2
0 –5 –6 2 –6
or
detB = [ 2 × 0 × ( – 6 ) ] + [ ( – 3 ) × ( – 2 ) × 0 ] + [ ( – 4 ) × 1 × ( – 5 ) ]
– [ 0 × 0 × ( – 4 ) ] – [ ( – 5 ) × ( – 2 ) × 2 ] – [ ( – 6 ) × 1 × ( – 3 ) ] = 20 – 38 = – 18

Signals and Systems with MATLAB Applications, Second Edition C-11

Orchard Publications
 Matrices and Determinants

Check with MATLAB:

A=[2 3 5; 1 0 1; 2 1 0]; det(A) % Define matrix A and compute detA
ans =
9
B=[2 −3 −4; 1 0 −2; 0 −5 −6];det(B) % Define matrix B and compute detB
ans =
-18

C.5 Minors and Cofactors

Let matrix A be defined as the square matrix of order n as shown below.

a 11 a 12 a 13 … a 1 n
a 21 a 22 a 23 … a 2 n
A = a a a … a (C.17)
31 32 33 3n
… … … … …
a n 1 a n 2 a n 3 … a nn

If we remove the elements of its ith row, and jth column, the remaining n – 1 square matrix is called
the minor of A , and it is denoted as M ij .

i+j
The signed minor ( – 1 ) M ij is called the cofactor of a ij and it is denoted as α ij .

Example C.6
Matrix A is defined as

a 11 a 12 a 13
A = a 21 a 22 a 23 (C.18)
a 31 a 32 a 33

Compute the minors M 11 , M 12 , M 13 and the cofactors α 11 , α 12 and α 13 .

C-12 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Minors and Cofactors

Solution:

a 22 a 23 a 21 a 23 a 21 a 22
M 11 = M 12 = M 11 =
a 32 a 33 a 31 a 33 a 31 a 32
and
1+1 1+2 1+3
α 11 = ( – 1 ) M 11 = M 11 α 12 = ( – 1 ) M 12 = – M 12 α 13 = M 13 = ( – 1 ) M 13

The remaining minors

M 21 , M 22 , M 23 , M 31 , M 32 , M 33
and cofactors
α 21, α 22, α 23, α 31, α 32, and α 33
are defined similarly.

Example C.7
Compute the cofactors of matrix A defined as

1 2 –3
A = 2 –4 2 (C.19)
–1 2 –6

Solution:

α 11 = ( – 1 )
1+1 – 4 2 = 20 α 12 = ( – 1 )
1+2 2 2 = 10 (C.20)
2 –6 –1 –6

α 13 = ( – 1 )
1+3 2 –4 = 0 α 21 = ( – 1 )
2+1 2 –3 = 6 (C.21)
–1 2 2 –6

α 22 = ( – 1 )
2+2 1 –3 = –9 α 23 = ( – 1 )
2+3 1 2 = –4 (C.22)
–1 –6 –1 2

α 31 = ( – 1 )
3+1 2 – 3 = – 8, α 32 = ( – 1 )
3+2 1 –3 = –8 (C.23)
–4 2 2 2

3+3 1 2 = –8
α 33 = ( – 1 ) (C.24)
2 –4

Signals and Systems with MATLAB Applications, Second Edition C-13

Orchard Publications
 Matrices and Determinants

It is useful to remember that the signs of the cofactors follow the pattern

+ − + − +
− + − + −
+ − + − +
− + − + −
+ − + − +

that is, the cofactors on the diagonals have the same sign as their minors.
Let A be a square matrix of any size; the value of the determinant of A is the sum of the products
obtained by multiplying each element of any row or any column by its cofactor.

Example C.8
Matrix A is defined as

1 2 –3
A = 2 –4 2 (C.25)
–1 2 –6

Compute the determinant of A using the elements of the first row.

Solution:

detA = 1 – 4 2 – 2 2 2 – 3 2 – 4 = 1 × 20 – 2 × ( – 10 ) – 3 × 0 = 40
2 –6 –1 –6 –1 2

Check with MATLAB:

A=[1 2 −3; 2 −4 2; −1 2 −6];det(A) % Define matrix A and compute detA
ans =
40
We must use the above procedure to find the determinant of a matrix A of order 4 or higher. Thus,
a fourth-order determinant can first be expressed as the sum of the products of the elements of its
first row by its cofactor as shown below.

Determinants of order five or higher can be evaluated similarly.

C-14 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Minors and Cofactors

a 11 a 12 a 13 a 14
a 22 a 23 a 24 a 12 a 13 a 14
a 21 a 22 a 23 a 24
A = = a 11 a 32 a 33 a 34 – a 21 a 32 a 33 a 34 (C.26)
a 31 a 32 a 33 a 34
a 42 a 43 a 44 a 42 a 43 a 44
a 41 a 42 a 43 a 44

a 12 a 13 a 14 a 12 a 13 a 14
+a 31 a 22 a 23 a 24 – a 41 a 22 a 23 a 24
a 42 a 43 a 44 a 32 a 33 a 34

Example C.9
Compute the value of the determinant of the matrix A defined as

2 –1 0 –3
A = –1 1 0 –1 (C.27)
4 0 3 –2
–3 0 0 1

Solution:
Using the above procedure, we will multiply each element of the first column by its cofactor. Then,

1 0 –1 –1 0 –3 –1 0 –3 –1 0 –3
A=2 0 3 – 2 –( –1 ) 0 3 –2 +4 1 0 – 1 –( –3 ) 1 0 –1
0 0 1 0 0 1 0 0 1 0 3 –2
⎧
⎪
⎪
⎨
⎪
⎪
⎩
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
⎧
⎪
⎪
⎨
⎪
⎪
⎩
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
[a] [b] [c] [d]
Next, using the procedure of Example C.5 or Example C.8, we find

[ a ] = 6 , [ b ] = – 3 , [ c ] = 0 , [ d ] = – 36
and thus
detA = [ a ] + [ b ] + [ c ] + [ d ] = 6 – 3 + 0 – 36 = – 33

We can verify our answer with MATLAB as follows:

A=[ 2 −1 0 −3; −1 1 0 −1; 4 0 3 −2; −3 0 0 1]; delta = det(A)
delta =
-33
Some useful properties of determinants are given below.

Signals and Systems with MATLAB Applications, Second Edition C-15

Orchard Publications
 Matrices and Determinants

Property 1: If all elements of one row or one column are zero, the determinant is zero. An example of
this is the determinant of the cofactor [ c ] above.
Property 2: If all the elements of one row or column are m times the corresponding elements of another
row or column, the determinant is zero. For example, if

2 4 1
A = 3 6 1 (C.28)
1 2 1

then,

2 4 1 2 4
detA = 3 6 1 3 6 = 12 + 4 + 6 – 6 – 4 – 12 = 0 (C.29)
1 2 1 1 2

Here, detA is zero because the second column in A is 2 times the first column.
Check with MATLAB:
A=[2 4 1; 3 6 1; 1 2 1];det(A)
ans =
0
Property 3: If two rows or two columns of a matrix are identical, the determinant is zero. This follows
from Property 2 with m = 1 .

C.6 Cramer’s Rule

Let us consider the systems of the three equations below
a 11 x + a 12 y + a 13 z = A
a 21 x + a 22 y + a 23 z = B (C.30)
a 31 x + a 32 y + a 33 z = C

and let
a 11 a 12 a 13 A a 11 a 13 a 11 A a 13 a 11 a 12 A
∆ = a 21 a 22 a 23 D1 = B a 21 a 23 D2 = a 21 B a 23 D3 = a 21 a 22 B
a 31 a 32 a 33 C a 31 a 33 a 31 C a 33 a 31 a 32 C

Cramer’s rule states that the unknowns x, y, and z can be found from the relations

C-16 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Cramer’s Rule

D D D
x = -----1- y = -----2- z = -----3- (C.31)
∆ ∆ ∆
provided that the determinant ∆ (delta) is not zero.
We observe that the numerators of (C.31) are determinants that are formed from ∆ by the substitu-
tion of the known values A , B , and C , for the coefficients of the desired unknown.
Cramer’s rule applies to systems of two or more equations.
If (C.30) is a homogeneous set of equations, that is, if A = B = C = 0 , then, D 1, D 2, and D 3 are
all zero as we found in Property 1 above. Then, x = y = z = 0 also.

Example C.10
Use Cramer’s rule to find v 1 , v 2 , and v 3 if

2v 1 – 5 – v 2 + 3v 3 = 0
– 2v 3 – 3v 2 – 4v 1 = 8 (C.32)
v 2 + 3v 1 – 4 – v 3 = 0

and verify your answers with MATLAB.

Solution:
Rearranging the unknowns v , and transferring known values to the right side, we get

2v 1 – v 2 + 3v 3 = 5
– 4v 1 – 3v 2 – 2v 3 = 8 (C.33)
3v 1 + v 2 – v 3 = 4

Now, by Cramer’s rule,

2 –1 3 2 –1
∆ = –4 –3 –2 – 4 – 3 = 6 + 6 – 12 + 27 + 4 + 4 = 35
3 1 –1 3 1

5 –1 3 5 –1
D1 = 8 –3 –2 8 – 3 = 15 + 8 + 24 + 36 + 10 – 8 = 85
4 1 –1 4 1

Signals and Systems with MATLAB Applications, Second Edition C-17

Orchard Publications
 Matrices and Determinants

2 5 3 2 5
D2 = –4 8 –2 –4 8 = – 16 – 30 – 48 – 72 + 16 – 20 = – 170
3 4 –1 3 4

2 –1 5 2 –1
D3 = –4 –3 8 – 4 – 3 = – 24 – 24 – 20 + 45 – 16 – 16 = – 55
3 1 4 3 1

Then, using (C.31) we get

D D D
------ = 17
x 1 = -----1- = 85 ------ --------- = – 34
x 2 = -----2- = – 170 ------ ------ = – 11
x 3 = -----3- = – 55 ------ (C.34)
∆ 35 7 ∆ 35 7 ∆ 35 7

We will verify with MATLAB as follows.

% The following code will compute and display the values of v1, v2 and v3.
format rat % Express answers in ratio form
B=[2 −1 3; −4 −3 −2; 3 1 −1]; % The elements of the determinant D of matrix B
delta=det(B); % Compute the determinant D of matrix B
d1=[5 -1 3; 8 -3 -2; 4 1 -1]; % The elements of D1
detd1=det(d1); % Compute the determinant of D1
d2=[2 5 3; -4 8 -2; 3 4 -1]; % The elements of D2
detd2=det(d2); % Compute the determinant of D2
d3=[2 -1 5; -4 -3 8; 3 1 4]; % The elements of D3
detd3=det(d3); % Compute he determinant of D3
v1=detd1/delta; % Compute the value of v1
v2=detd2/delta; % Compute the value of v2
v3=detd3/delta; % Compute the value of v3
%
disp('v1=');disp(v1); % Display the value of v1
disp('v2=');disp(v2); % Display the value of v2
disp('v3=');disp(v3); % Display the value of v3

v1=
17/7
v2=
-34/7
v3=
-11/7
These are the same values as in (C.34)

C-18 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Gaussian Elimination Method

C.7 Gaussian Elimination Method

We can find the unknowns in a system of two or more equations also by the Gaussian elimination
method. With this method, the objective is to eliminate one unknown at a time. This can be done by
multiplying the terms of any of the equations of the system by a number such that we can add (or
subtract) this equation to another equation in the system so that one of the unknowns will be elimi-
nated. Then, by substitution to another equation with two unknowns, we can find the second
unknown. Subsequently, substitution of the two values found can be made into an equation with
three unknowns from which we can find the value of the third unknown. This procedure is repeated
until all unknowns are found. This method is best illustrated with the following example which con-
sists of the same equations as the previous example.
Example C.11
Use the Gaussian elimination method to find v 1 , v 2 , and v 3 of the system of equations

2v 1 – v 2 + 3v 3 = 5
– 4v 1 – 3v 2 – 2v 3 = 8 (C.35)
3v 1 + v 2 – v 3 = 4

Solution:
As a first step, we add the first equation of (C.35) with the third to eliminate the unknown v2 and we
obtain the following equation.
5v 1 + 2v 3 = 9 (C.36)

Next, we multiply the third equation of (C.35) by 3, and we add it with the second to eliminate v 2 .
Then, we obtain the following equation.
5v 1 – 5v 3 = 20 (C.37)

Subtraction of (C.37) from (C.36) yields

11
7v 3 = – 11 or v 3 = – ------ (C.38)
7

Now, we can find the unknown v 1 from either (C.36) or (C.37). By substitution of (C.38) into (C.36)
we get

5v 1 + 2 ⋅ ⎛ – ------⎞ = 9 or v 1 = ------
11 17
⎝ 7⎠
(C.39)
7

Finally, we can find the last unknown v 2 from any of the three equations of (C.35). By substitution
into the first equation we get

Signals and Systems with MATLAB Applications, Second Edition C-19

Orchard Publications
 Matrices and Determinants

34 33 35 34
v 2 = 2v 1 + 3v 3 – 5 = ------ – ------ – ------ = – ------ (C.40)
7 7 7 7
These are the same values as those we found in Example C.10.
The Gaussian elimination method works well if the coefficients of the unknowns are small integers,
as in Example C.11. However, it becomes impractical if the coefficients are large or fractional num-
bers.

C.8 The Adjoint of a Matrix

Let us assume that A is an n square matrix and α ij is the cofactor of a ij . Then the adjoint of A ,
denoted as adjA , is defined as the n square matrix below.

α 11 α 21 α 31 … α n 1
α 12 α 22 α 32 … α n 2
adjA = α α α … α (C.41)
13 23 33 n3
… … … … …
α 1 n α 2 n α 3 n … α nn

We observe that the cofactors of the elements of the ith row (column) of A are the elements of the
ith column (row) of adjA .

Example C.12
Compute adjA if Matrix A is defined as

1 2 3
A = 1 3 4 (C.42)
1 4 3

Solution:

3 4 – 2 3 2 3
4 3 4 3 3 4
–7 6 –1
adjA = – 1 4 1 3 – 2 3 = 1 0 –1
1 3 1 3 3 4 1 –2 1
1 3 – 1 2 1 2
1 4 1 4 1 3

C-20 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Singular and Non-Singular Matrices

C.9 Singular and Non-Singular Matrices

An n square matrix A is called singular if detA = 0 ; if detA ≠ 0 , A is called non-singular.

Example C.13
Matrix A is defined as

1 2 3
A = 2 3 4 (C.43)
3 5 7

Determine whether this matrix is singular or non-singular.

Solution:

1 2 3 1 2
detA = 2 3 4 2 3 = 21 + 24 + 30 – 27 – 20 – 28 = 0
3 5 7 3 5

Therefore, matrix A is singular.

C.10 The Inverse of a Matrix

If A and B are n square matrices such that AB = BA = I , where I is the identity matrix, B is
called the inverse of A , denoted as B = A –1 , and likewise, A is called the inverse of B , that is,
–1
A = B

If a matrix A is non-singular, we can compute its inverse A –1 from the relation

–1 1
A = ------------ adjA (C.44)
detA

Example C.14
Matrix A is defined as

1 2 3
A = 1 3 4 (C.45)
1 4 3

Signals and Systems with MATLAB Applications, Second Edition C-21

Orchard Publications
 Matrices and Determinants

Compute its inverse, that is, find A –1

Solution:
Here, detA = 9 + 8 + 12 – 9 – 16 – 6 = – 2 , and since this is a non-zero value, it is possible to com-
pute the inverse of A using (C.44).
From Example C.12,
–7 6 –1
adjA = 1 0 – 1
1 –2 1
Then,

–7 6 –1 3.5 – 3 0.5
–1 1 - 1-
A = ----------- adjA = ----- = – 0.5 0 0.5 (C.46)
detA –2 1 0 –1
1 –2 1 – 0.5 1 – 0.5

Check with MATLAB:

A=[1 2 3; 1 3 4; 1 4 3], invA=inv(A) % Define matrix A and compute its inverse
A =
1 2 3
1 3 4
1 4 3
invA =
3.5000 -3.0000 0.5000
-0.5000 0 0.5000
-0.5000 1.0000 -0.5000

Multiplication of a matrix A by its inverse A –1 produces the identity matrix I , that is,
–1 –1
AA = I or A A = I (C.47)

Example C.15
Prove the validity of (C.47) for the Matrix A defined as

A = 4 3
2 2

C-22 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solution of Simultaneous Equations with Matrices

Proof:

detA = 8 – 6 = 2 and adjA = 2 –3

–2 4
Then,

A
–1
= ------------ adjA = --- 2 – 3 =
1 1 1 –3 ⁄ 2
detA 2 –2 4 –1 2
and

AA
–1
= 4 3 1 –3 ⁄ 2 = 4 – 3 –6+6 = 1 0 = I
2 2 –1 2 2–2 –3+4 0 1

C.11 Solution of Simultaneous Equations with Matrices

Consider the relation
AX = B (C.48)
where A and B are matrices whose elements are known, and X is a matrix (a column vector) whose
elements are the unknowns. We assume that A and X are conformable for multiplication. Multipli-
cation of both sides of (C.48) by A –1 yields:
–1 –1 –1
A AX = A B = IX = A B (C.49)
or
–1
X=A B (C.50)

Therefore, we can use (C.50) to solve any set of simultaneous equations that have solutions. We will
refer to this method as the inverse matrix method of solution of simultaneous equations.

Example C.16
For the system of the equations

⎧ 2x 1 + 3x 2 + x 3 = 9 ⎫
⎪ ⎪
⎨ x 1 + 2x 2 + 3x 3 = 6 ⎬ (C.51)
⎪ ⎪
⎩ 3x 1 + x 2 + 2x 3 = 8 ⎭

compute the unknowns x 1, x 2, and x 3 using the inverse matrix method.

Signals and Systems with MATLAB Applications, Second Edition C-23

Orchard Publications
 Matrices and Determinants

Solution:
In matrix form, the given set of equations is AX = B where

2 3 1 x1 9
A= 1 2 3 , X = x 2
, B = 6 (C.52)
3 1 2 x3 8

Then,
–1
X = A B (C.53)
or

–1
x1 2 3 1 9
x2 = 1 2 3 6 (C.54)
x3 3 1 2 8

Next, we find the determinant detA , and the adjoint adjA

1 –5 7
detA = 18 and adjA = 7 1 –5
–5 7 1
Therefore,
1 –5 7
–1 1 1
A = ------------ adjA = ------ 7 1 – 5
detA 18
–5 7 1

and by (C.53) we obtain the solution as follows.

x1 1 –5 7 9 35 35 ⁄ 18 1.94
1 1
X = x2 = -----
- = -----
- = 29 ⁄ 18 = 1.61 (C.55)
18 7 1 – 5 6 18 29
x3 –5 7 1 8 5 5 ⁄ 18 0.28

To verify our results, we could use the MATLAB’s inv(A) function, and then multiply A –1 by B .
However, it is easier to use the matrix left division operation X = A \ B ; this is MATLAB’s solution
of A –1 B for the matrix equation A ⋅ X = B , where matrix X is the same size as matrix B . For this
example,

C-24 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solution of Simultaneous Equations with Matrices

A=[2 3 1; 1 2 3; 3 1 2]; B=[9 6 8]';

X=A \ B
X =
1.9444
1.6111
0.2778

Example C.17
For the electric circuit of Figure C.1,

1Ω 2Ω 2Ω

V = 100 v
+ 9Ω 4Ω
− 9Ω
I1 I3
I2

Figure C.1. Circuit for Example C.17

the loop equations are

10I 1 – 9I 2 = 100
– 9I 1 + 20I 2 – 9I 3 = 0 (C.56)
– 9I 2 + 15I 3 = 0

Use the inverse matrix method to compute the values of the currents I 1 , I 2 , and I 3

Solution:

For this example, the matrix equation is RI = V or I = R –1 V , where

10 – 9 0 100 I1
R = – 9 20 – 9 , V = 0 and I = I 2
0 – 9 15 0 I3

The next step is to find R – 1 . This is found from the relation

–1 1
R = ------------ adjR (C.57)
detR

Signals and Systems with MATLAB Applications, Second Edition C-25

Orchard Publications
 Matrices and Determinants

Therefore, we find the determinant and the adjoint of R . For this example, we find that

219 135 81
detR = 975, adjR = 135 150 90 (C.58)
81 90 119

Then,
219 135 81
–1 1 1
R = ------------ adjR = --------- 135 150 90
detR 975
81 90 119
and
I1 219 135 81 100 219 22.46
1- 100
I = I2 = -------- 0 = 975 135 = 13.85
--------
-
975 135 150 90
I3 81 90 119 0 81 8.31

Check with MATLAB:

R=[10 −9 0; −9 20 −9; 0 −9 15]; V=[100 0 0]'; I=R\V; fprintf(' \n');...
fprintf('I1 = %4.2f \t', I(1)); fprintf('I2 = %4.2f \t', I(2)); fprintf('I3 = %4.2f \t', I(3)); fprintf(' \n')
I1 = 22.46 I2 = 13.85 I3 = 8.31
We can also use subscripts to address the individual elements of the matrix. Accordingly, the above
code could also have been written as:
R(1,1)=10; R(1,2)=−9; % No need to make entry for A(1,3) since it is zero.
R(2,1)=−9; R(2,2)=20; R(2,3)=−9; R(3,2)=−9; R(3,3)=15; V=[100 0 0]'; I=R\V; fprintf(' \n');...
fprintf('I1 = %4.2f \t', I(1)); fprintf('I2 = %4.2f \t', I(2)); fprintf('I3 = %4.2f \t', I(3)); fprintf(' \n')
I1 = 22.46 I2 = 13.85 I3 = 8.31
Spreadsheets also have the capability of solving simultaneous equations with real coefficients using
the inverse matrix method. For instance, we can use Microsoft Excel’s MINVERSE (Matrix Inver-
sion) and MMULT (Matrix Multiplication) functions, to obtain the values of the three currents in
Example C.17.
The procedure is as follows:
1. We start with a blank spreadsheet and in a block of cells, say B3:D5, we enter the elements of
matrix R as shown in Figure C.2. Then, we enter the elements of matrix V in G3:G5.
2. Next, we compute and display the inverse of R , that is, R –1 . We choose B7:D9 for the elements of
this inverted matrix. We format this block for number display with three decimal places. With this
range highlighted and making sure that the cell marker is in B7, we type the formula

C-26 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solution of Simultaneous Equations with Matrices

=MININVERSE(B3:D5)
and we press the Crtl-Shift-Enter keys simultaneously.
We observe that R – 1 appears in these cells.
3. Now, we choose the block of cells G7:G9 for the values of the current I . As before, we highlight
them, and with the cell marker positioned in G7, we type the formula
=MMULT(B7:D9,G3:G5)

and we press the Crtl-Shift-Enter keys simultaneously. The values of I then appear in G7:G9.

A B C D E F G H
1 Spreadsheet for Matrix Inversion and Matrix Multiplication
2
3 10 -9 0 100
4 R= -9 20 -9 V= 0
5 0 -9 15 0
6
7 0.225 0.138 0.083 22.462
-1
8 R = 0.138 0.154 0.092 I= 13.846
9 0.083 0.092 0.122 8.3077
10

Figure C.2. Solution of Example C.17 with a spreadsheet

Example C.18
For the phasor circuit of Figure C.18

85 Ω C
R1 −j100 Ω
170∠0° +
IX
V1 V2
− R3 = 100 Ω
VS
`

L R2 50 Ω
j200 Ω

Figure C.3. Circuit for Example C.18

the current I X can be found from the relation

Signals and Systems with MATLAB Applications, Second Edition C-27

Orchard Publications
 Matrices and Determinants

V1 – V2
I X = -----------------
- (C.59)
R3

and the voltages V 1 and V 2 can be computed from the nodal equations

V 1 – 170 ∠0° V 1 – V 2 V 1 – 0
------------------------------- + ------------------ + --------------- = 0 (C.60)
85 100 j200
and
V 2 – 170 ∠0° V 2 – V 1 V 2 – 0
------------------------------- + ------------------ + --------------- = 0 (C.61)
– j100 100 50

Compute, and express the current I x in both rectangular and polar forms by first simplifying like
terms, collecting, and then writing the above relations in matrix form as YV = I , where
Y = Admit tan ce , V = Voltage , and I = Current

Solution:
The Y matrix elements are the coefficients of V 1 and V 2 . Simplifying and rearranging the nodal
equations of (C.60) and (C.61), we get
( 0.0218 – j0.005 )V 1 – 0.01V 2 = 2
(C.62)
– 0.01 V 1 + ( 0.03 + j0.01 )V 2 = j1.7

Next, we write (C.62) in matrix form as

0.0218 – j0.005 – 0.01 V1 2

– 0.01 0.03 + j0.01 V2 = j1.7 (C.63)
⎧
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎩

⎧
⎨
⎩
⎧
⎨
⎩

Y V I

where the matrices Y , V , and I are as indicated.

We will use MATLAB to compute the voltages V 1 and V 2 , and to do all other computations. The
code is shown below.
Y=[0.0218−0.005j −0.01; −0.01 0.03+0.01j]; I=[2; 1.7j]; V=Y\I;% Define Y, I, and find V
fprintf('\n'); % Insert a line
disp('V1 = '); disp(V(1)); disp('V2 = '); disp(V(2)); % Display values of V1 and V2

V1 =
1.0490e+002 + 4.9448e+001i

C-28 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
 Solution of Simultaneous Equations with Matrices

V2 =
53.4162 + 55.3439i
Next, we find I X from
R3=100; IX=(V(1)−V(2))/R3 % Compute the value of IX

IX =
0.5149- 0.0590i
This is the rectangular form of I X . For the polar form we use
magIX=abs(IX) % Compute the magnitude of IX

magIX =
0.5183
thetaIX=angle(IX)*180/pi % Compute angle theta in degrees
thetaIX =
-6.5326
Therefore, in polar form
I X = 0.518 ∠– 6.53°

Spreadsheets have limited capabilities with complex numbers, and thus we cannot use them to com-
pute matrices that include complex numbers in their elements as in Example C.18

Signals and Systems with MATLAB Applications, Second Edition C-29

Orchard Publications
 Matrices and Determinants

C.12 Exercises
For Exercises 1, 2, and 3 below, the matrices A , B , C , and D are defined as:

1 –1 –4 5 9 –3 4 6
A = 5 7 –2 B = –2 8 2 C= – 3 8 D = 1 –2 3
–3 6 –4
3 –5 6 7 –4 6 5 –2

1. Perform the following computations, if possible. Verify your answers with MATLAB.
a. A + B b. A + C c. B + D d. C + D
e. A – B f. A – C g. B – D h. C – D
2. Perform the following computations, if possible. Verify your answers with MATLAB.
a. A ⋅ B b. A ⋅ C c. B ⋅ D d. C ⋅ D
e. B ⋅ A f. C ⋅ A g. D ⋅ A h. D· ⋅ C
3. Perform the following computations, if possible. Verify your answers with MATLAB.
a. detA b. detB c. detC d. detD
e. det ( A ⋅ B ) f. det ( A ⋅ C )
4. Solve the following systems of equations using Cramer’s rule. Verify your answers with MATLAB.
– x 1 + 2x 2 – 3x 3 + 5x 4 = 14
x 1 – 2x 2 + x 3 = – 4
x 1 + 3x 2 + 2x 3 – x 4 = 9
a. – 2x 1 + 3x 2 + x 3 = 9 b.
3x 1 – 3 x 2 + 2x 3 + 4x 4 = 19
3x 1 + 4x 2 – 5x 3 = 0
4x 1 + 2x 2 + 5x 3 + x 4 = 27

5. Repeat Exercise 4 using the Gaussian elimination method.

6. Solve the following systems of equations using the inverse matrix method. Verify your answers
with MATLAB.

2 4 3 –2 x1 1
1 3 4 x1 –3
a. 3 1 – 2 ⋅ x 2 = – 2 b. 2 – 4 1 3 ⋅ x 2 = 10
–1 3 –4 2 x3 – 14
2 3 5 x3 0
2 –2 2 1 x4 7

C-30 Signals and Systems with MATLAB Applications, Second Edition

Orchard Publications
Index

Symbols clc MATLAB function A-2

clear MATLAB function A-2
% (percent) symbol in MATLAB A-2 code in MATLAB A-2
cofactor of a matrix - see matrix
A collect MATLAB function 3-12
column vector in MATLAB A-19
abs MATLAB function A-23 command screen in MATLAB A-1
adjoint of a matrix - see matrix command window in MATLAB A-1
admittance 4-2, 4-10 commas in MATLAB A-8
capacitive 4-2 comment line in MATLAB A-2
inductive 4-2 complex conjugate A-4, B-3
aliasing 10-13 complex number A-3, B-2
all-pass filter - see filter complex poles 3-5
all-pole filter -see filter conj MATLAB function C-8
alternate form of the trigonometric conjugate of a complex number - see complex conjugate
Fourier series - see Fourier series conjugate of a matrix - see matrix
alternate method of partial fraction conjugate time and frequency functions 8-13
expansion - see partial fraction expansion contour integral 2-2, 9-20
amplitude-squared function 11-8 contour integration 2-2, 9-20
angle MATLAB function A-23 conv MATLAB function A-6
area under f(t) 8-15 convolution
area under F( Z ) 8-16 in the complex frequency domain 2-12
attenuation rate 11-9 in the continuous time domain 2-11
axis MATLAB function A-16, A-22 in the discrete frequency domain 9-9
in the discrete time domain 9-8
B in the frequency (jZ) domain 8-15
convolution integral 6-8
band-elimination filter - see filter graphical evaluation of 6-8
band-stop filter - see filter Cooley and Tukey algorithm 10-18
Bessel filter - see filter Cramer’s rule C-16
bilateral Laplace Transform 2-1
bilinear MATLAB function 11-57 D
bilinear transformation 11-50
bode MATLAB function 11-21, 11-48 d2c MATLAB function 9-46
box MATLAB function A-12 data points in MATLAB A-14
buttap MATLAB function 11-15, 11-39, 11-40, 11-46 dB/decade 11-9
buttefly operation 10-22 dB/octave 11-9
Butterworth low-pass filter - see filter DC (average) component in Fourier series - see Fourier series
decade 11-9
C decibel scale in MATLAB A-13
decimation in frequency 10-19, 10-20
c2d MATLAB function 9-46 decimation in time 10-19, 10-20
Cauer filter - see filter deconv MATLAB function A-6
Cayley-Hamilton theorem 5-11 default in MATLAB A-13
cheb1ap MATLAB function 11-32, 11-42 default color in MATLAB A-15
cheb2ap MATLAB function 11-35 default line in MATLAB A-15
Chebyshev filters - see filter default marker in MATLAB A-15
delta function 1-12 FFT 10-17
demo in MATLAB A-2 FFT Category I and II algorithms 10-19
DeMoivre’s theorem 11-12 fft MATLAB function 10-5, 11-67
determinant of a matrix - see matrix figure window in MATLAB A-13
differentiation filter
in complex frequency (s) domain 2-6 all-pass 11-1
in frequency (jZ) domain 8-13 all-pole 11-18
in time domain 2-4 band-elimination
digital filter - see filter analog 4-29, 11-1, 11-39, 11-46
dimpulse MATLAB function 9-28 digital 11-58
dirac MATLAB function 1-18 band-pass
direct term in MATLAB 3-4 analog 4-29, 11-1, 11-39, 11-44
discontinuous function 1-2 digital 11-58
display formats in MATLAB A-30 band-stop - see filter, band-elimination
distinct poles 3-2 Bessel 11-77
Division in MATLAB Butterworth
element-by-element operator A-21 band-elimination 11-39, 11-46
matrix operator A-20 band-pass 11-39, 11-44
matrix left division C-24 high-pass 11-39
division of complex numbers B-4 low-pass 11-7, 11-17, 11-20, 11-39
dot multiplication operator in MATLAB A-21 Cauer - see filter, elliptic
double-memory technique 10-19 Chebyshev
doublet function 1-15 Type I 11-22
Type II 11-34, 11-35
E elliptic 11-34, 11-36
Inverted Chebyshev - see filter, Chebyshev Type II
editor window in MATLAB A-2 high-pass
editor/debugger in MATLAB A-1, A-2 analog 4-28, 11-1, 11-4, 11-39, 11-42
eigenvalues 5-11 digital 11-58
eigenvector 5-19 low-pass
elements of the matrix C-1 analog 4-27, 11-1, 11-2, 11-7, 11-39, 11-42
ellip MATLAB function 11-36 digital 11-58
elliptic filter - see filter phase-shift - see filter, all-pass
eps MATLAB function A-22 filter MATLAB function 11-62
Euler’s identities B-4 final value theorem
even function 6-5, 7-33 in Laplace transform 2-10
even symmetry 7-7 in Z transform 9-10
exit MATLAB command A-2 find MATLAB function 11-67
exponential form of complex numbers B-4 Finite Impulse Response (FIR) digital filter 11-50
exponential form of the Fourier series 7-31 FIR digital filter 11-50
exponential order function 2-2 first harmonic 7-1
exponentiation in MATLAB first order circuit 5-1
element-by-element operator A-21 fmax in MATLAB - using fmin to find max A-27
matrix operator A-21 fmin in MATLAB A-27
eye MATLAB function C-7 format MATLAB command A-30
fourier MATLAB function 8-33
F Fourier integral - see Fourier Transform
Fourier series
factor MATLAB function 3-4 alternate trigonometric form 7-25
Fast Fourier Transform (FFT) 10-17 DC (average) component 7-1
 exponential form 7-31 ilaplace MATLAB function 3-4
trigonometric form 7-1 imag MATLAB function A-23
Fourier transform imaginary axis B-2
definition of 8-1 imaginary number B-2
derived from the Laplace Transform 8-25 impedance 4-2, 4-8
of the cosine function 8-19 capacitive 4-2
of the delta function 8-17 inductive 4-2
of the signum (sgn) function 8-20 improper integral 2-16
of the sine function 8-20 improper rational function 3-1, 3-13
of the unit step function 8-21 impulse invariant method 11-50
of unity 8-19 impulse response
Fourier transforms of common waveforms 8-27 in continuous time systems 6-1
fplot MATLAB command A-27 in discrete time systems 9-40
frequency response A-12 increments between points in MATLAB A-14
frequency shift 10-12 inductive admittance - see admittance
freqz MATLAB function 11-60 inductive impedance - see impedance
full rectification waveform 2-36 IIR digital filter 11-50
full-wave rectifier 7-21 Infinite Impulse Response (IIR) digital filter 11-50
full-wave rectifier with even symmetry 7-24 initial value theorem
function file in MATLAB A-26 in Laplace transform 2-9
fundamental frequency 7-1 in Z transform 9-9
fzero MATLAB function A-27 in-place algorithm 10-20
integration in complex frequency 2-8
G integration in time 2-6
Inverse Fourier transform 8-1
gamma function 2-15 Inverse Laplace transform 2-1, 3-1
Gaussian elimination method C-19 inverse matrix method of solution of equations C-23
generalized factorial function 2-15 inverse of a matrix - see matrix
geometric sequence in Z-transform 9-11 Inverse Z transform 9-1
Gibbs phenomenon 7-24 Inversion integral 9-32
Gram-Schmidt Orthogonalization Procedure 5-19 Inverted Chebyshev filter - see filter
grid MATLAB command A-12 iztrans MATLAB function 9-27
gtext MATLAB command A-13
J
H
j operator B-1
half-rectified sine wave 2-28
half-wave rectifier 7-17 L
half-wave rectifier with no symmetry 7-21
half-wave symmetry 7-7, 7-33 L’ Hôpital’s rule 2-15, 2-16
Heavyside(t) MATLAB function 1-18 Laplace transform of common functions 2-12
help in MATLAB A-2 Laplace transform of common waveforms 2-23
Hermitian matrix - see matrix Laplace Transformation 2-1
high-pass filter - see filter leakage 10-13
left shift in in discrete time domain 9-5
I Leibnitz’s rule 2-6
lims= in MATLAB A-27
identity matrix C-6 line spectrum 7-35
ifft MATLAB function10-5 linear difference equation 9-38
ifourier MATLAB function 8-33 linear factor A-9
linearity property mesh MATLAB command A-17
in Fourier transform 8-9 meshgrid MATLAB command A-17
in Laplace transform 2-2 method of clearing the fractions 3-15
in Z transform 9-3 m-file in MATLAB A-2, A-25
linspace MATLAB command A-14 minor of determinant C-12
ln in MATLAB A-13 MINVERSE Excel function C-26
log MATLAB function A-13 MMULT Excel function C-26
log10 MATLAB function A-13 modulated signals 8-12
log2 MATLAB function A-13 multiple (repeated) poles 3-8
loglog scale in MATLAB A-13 multiplication
lower triangular matrix C-5 in continuous time domain 2-12, 8-11
low-pass filter - see filter in discrete time systems 9-6
lp2bp MATLAB function 11-40, 11-44 see also convolution
lp2bs MATLAB function 11-40, 11-46 multiplication in MATLAB
lp2hp MATLAB function 11-40, 11-42 element-by-element A-21
lp2lp MATLAB function 11-40, 11-41 matrix A-20
multiplication of complex numbers B-3
M
N
main diagonal elements of a matrix - see matrix
main diagonal of a matrix - see matrix NaN in MATLAB A-27
matrices natural input-output 10-19
conformable for addition and subtraction C-2 non-recursive realization digital filter 11-50
conformable for multiplication C-4 normalized cutoff frequency 11-12
matrix - see also determinant N-point DFT 10-2
adjoint of C-20 numerical evaluation of Fourier coefficients 7-44
cofactor of C-12 Nyquist frequency 10-13
conformable for addition and subtraction - see matrices
conformable for multiplication - see matrices O
conjugate of C-8
definition of C-1 octave 11-9
determinant of C-9 odd function 6-5, 7-33
Hermitian, skew-Hermitian C-9 odd symmetry 7-7
inverse of C-21 orthogonal functions 7-2
lower triangular C-5 orthogonal vectors 5-19
main diagonal of C-1
minor of C-12 P
non-singular C-21
scalar C-6 Parseval’s theorem 8-16
singular C-21 partial fraction expansion method 3-1, 3-15
size of C-7 phase shift filter - see filter
square C-1 picket-fence effect 10-14
symmetric, skew-symmetric C-8, C-9 plot MATLAB command A-10
trace of C-2 plot3 MATLAB command A-15
transpose of C-7 polar form of complex numbers B-5
upper triangular C-5 polar in MATLAB A-23
zero C-2 polar plot in MATLAB A-24
poles of rational functions 3-1
matrix left division in MATLAB C-24 poly MATLAB function A-4
matrix multiplication in MATLAB A-18 polyder MATLAB function A-8
polyval MATLAB function A-6 left shift 9-5
pre-sampling filter 10-13 right shift 9-4
pre-warping 11-54 sifting property of the delta function 1-13
proper rational function 3-1 signal flow graph 10-23
signum function 8-20
Q spectrum analyzer 7-35
square waveform with even symmetry 7-14
quadratic factor A-9 square waveform with odd symmetry 7-12
quit MATLAB function A-2 ss2tf MATLAB function 5-31
state equations 5-1, 9-43
R state transition matrix 5-9
state variables
radius of absolute convergence 9-3 for continuous time systems 5-1
randn MATLAB function 11-67 for discrete time systems 9-43
rational polynomials A-8 step invariant method 11-50
real axis B-2 subplot MATLAB command A-18
real MATLAB function A-23 symmetric rectangular pulse 1-6
real number B-2 symmetric triangular waveform 1-6
rectangular form B-5 symmetry
recursive realization 11-50 in trigonometric Fourier series 7-7
region of convergence 9-3, 9-14, 9-17 in exponential Fourier series 7-33
region of divergence 9-3, 9-14, 9-17 property of Fourier transform 8-10
repeated poles - see multiple poles system function 8-35
residue MATLAB function 3-3 see also transfer function
residue theorem 9-20
residues 3-2 T
roots MATLAB function 3-6, A-3
round MATLAB function A-23 Taylor series 5-1
row vector A-3, A-19 text MATLAB command A-14
Runge-Kutta method 5-1 tf2ss MATLAB function 5-32
third harmonic 7-1
S time periodicity 2-8
transfer function 4-15
sampling property of the delta function 1-12 in continuous time systems 4-13
sampling theorem 10-13 in discrete time systems 9-40
sawtooth waveform with odd symmetry 7-16 see also system function
scaling property 2-4, 8-10 triangular waveform 7-9, 7-17
script file in MATLAB A-25 trigonometric form of Fourier series - See Fourier series
second harmonic 7-1 triplet function 1-15
secord-order circuit 5-1
semicolons in MATLAB A-8 U
semilogx axis in MATLAB A-13
semilogy axis in MATLAB A-13 unit impulse 1-8, 1-12
sgn function - see signum function unit ramp function
Shannon’s sampling theorem - see sampling theorem in continuous time 1-8, 1-10
shifting property in discrete time 9-18
in complex frequency (s) domain 2-3 unit step function
in frequency (jZ) domain 8-11 in continuous time 1-2
in continuous time domain 2-3, 8-11 in discrete time 9-3
in discrete time domain
V

Vandermonde matrix 10-18

warping 11-52
window function 10-13

xlabel MATLAB function A-13

ylabel MATLAB function A-13

Z transform 9-1
zlabel MATLAB function A-18
ztrans MATLAB function 9-27