1.

Dedication of "A First Course in Electrical and Computer

Engineering"
2. Preface to "A First Course in Electrical and Computer
Engineering"
3. Complex Numbers
1. Complex Numbers: Introduction
2. Complex Numbers: Geometry of Complex Numbers
3. Complex Numbers: Algebra of Complex Numbers
4. Complex Numbers: Roots of Quadratic Equations
5. Complex Numbers: Representing Complex Numbers in a
Vector Space
6. Complex Numbers: An Electric Field Computation
7. Complex Numbers: Numerical Experiment (Quadratic
Roots)
4. The Functions e^x and e^jΘ
1. The Functions e^x and e^jθ: Introduction
2. The Functions e^x and e^jθ: The Function e^x
3. The Functions e^x and e^jθ: The Function e^jθ and the
Unit Circle
4. The Functions e^x and e^jθ: The Euler and De Moivre
Identities
5. The Functions e^x and e^jθ: Roots of Unity and Related
Topics
6. The Functions e^x and e^jθ: Second-Order Differential
and Difference Equations
7. The Functions e^x and e^jθ: Numerical Experiment
(Approximating e^jθ)
5. Phasors
1. Phasors: Introduction
2. Phasors: Phasor Representation of Signals
3. Phasors: Beating between Tones
 4. Phasors: Multiphase Power
5. Phasors: Lissajous Figures
6. Phasors: Sinusoidal Steady State and the Series RLC
Circuit
7. Phasors: Light Scattering by a Slit
8. Phasors: Numerical Experiment (Interference Patterns)
6. Linear Algebra
1. Linear Algebra: Introduction
2. Linear Algebra: Vectors
3. Linear Algebra: Inner Product and Euclidean Norm
4. Linear Algebra: Direction Cosines
5. Linear Algebra: Projections
6. Linear Algebra: Other Norms
7. Linear Algebra: Matrices
8. Linear Algebra: Solving Linear Systems of Equations
9. Linear Algebra: Circuit Analysis
10. Linear Algebra: Numerical Experiment (Circuit Design)
7. Vector Graphics
1. Vector Graphics: Introduction
2. Vector Graphics: Two-Dimensional Image Representation
3. Vector Graphics: Two-Dimensional Image
Transformations
4. Vector Graphics: Composition of Transformations
5. Vector Graphics: Homogeneous Coordinates
6. Vector Graphics: Three-Dimensional Homogeneous
Coordinates
7. Vector Graphics: Projections
8. Vector Graphics: Numerical Experiment (Star Field)
8. Filtering
1. Filtering: Introduction
2. Filtering: Simple Averages
3. Filtering: Weighted Averages
 4. Filtering: Moving Averages
5. Filtering: Exponential Averages and Recursive Filters
6. Filtering: Test Sequences
7. Filtering: Numerical Experiment (Frequency Response of
First-Order Filter)
9. Binary Codes
1. Binary Codes: Introduction
2. Binary Codes: The Communication Paradigm
3. Binary Codes: From Symbols to Binary Codes
4. Binary Codes: Huffman Codes for Source Coding
5. Binary Codes: Hamming Codes for Channel Coding
6. Binary Codes: Numerical Experiment (Huffman Codes)
10. An Introduction to MATLAB
1. An Introduction to MATLAB: Introduction
2. An Introduction to MATLAB: Running MATLAB
(Macintosh)
3. An Introduction to MATLAB: Running MATLAB (PC)
4. An Introduction to MATLAB: Interactive Mode
5. An Introduction to MATLAB: Variables
6. An Introduction to MATLAB: Complex Variables
7. An Introduction to MATLAB: Vectors and Matrices
8. An Introduction to MATLAB: The Colon
9. An Introduction to MATLAB: Graphics
10. An Introduction to MATLAB: Editing Files and Creating
Functions (Macintosh)
11. An Introduction to MATLAB: Editing Files and Creating
Functions (PC)
12. An Introduction to MATLAB: Loops and Control
11. The Edix Editor
12. Useful Mathematical Identities
Dedication of "A First Course in Electrical and Computer Engineering"

Louis Scharf dedicates this book to

his wife Carol, son Greg, and daughter Heidi,
for love and inspiration;
his parents Louis Sr. and Ann,
in celebration of their 50th wedding anniversary.

Richard Behrens dedicates this book to

his wife Debbie, and child as yet unborn,
for love and encouragement;
his parents Richard and Elsie,
in gratitude for a good education.
Preface to "A First Course in Electrical and Computer Engineering"

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

This book was written for an experimental freshman course at the

University of Colorado. The course is now an elective that the majority of
our electrical and computer engineering students take in the second
semester of their freshman year, just before their first circuits course. Our
department decided to offer this course for several reasons:

1. we wanted to pique student' interest in engineering by acquainting

them with engineering teachers early in their university careers and by
providing with exposure to the types of problems that electrical and
computer engineers are asked to solve;
2. we wanted students entering the electrical and computer engineering
programs to be prepared in complex analysis, phasors, and linear
algebra, topics that are of fundamental importance in our discipline;
3. we wanted students to have an introduction to a software application
tool, such as MATLAB, to complete their preparation for practical and
efficient computing in their subsequent courses and in their
professional careers;
4. we wanted students to make early contact with advanced topics like
vector graphics, filtering, and binary coding so that they would gain a
more rounded picture of modern electrical and computer engineering.

In order to introduce this course, we had to sacrifice a second semester of

Pascal programming. We concluded that the sacrifice was worth making
because we found that most of our students were prepared for high-level
language computing after just one semester of programming.
We believe engineering educators elsewhere are reaching similar
conclusions about their own students and curriculums. We hope this book
helps create a much needed dialogue about curriculum revision and that it
leads to the development of similar introductory courses that encourage
students to enter and practice our craft.

Students electing to take this course have completed one semester of

calculus, computer programming, chemistry, and humanities. Concurrently
with this course, students take physics and a second semester of calculus, as
well as a second semester in the humanities. By omitting the advanced
topics marked by asterisks, we are able to cover Complex Numbers through
Linear Algebra, plus two of the three remaining chapters. The book is
organized so that the instructor can select any two of the three. If every
chapter of this book is covered, including the advanced topics, then enough
material exists for a two-semester course.

The first three chapters of this book provide a fairly complete coverage of
complex numbers, the functions ex and e , and phasors. Our department
jθ

philosophy is that these topics must be understood if a student is to succeed

in electrical and computer engineering. These three chapters may also be
used as a supplement to a circuits course. A measured pace of presentation,
taking between sixteen and eighteen lectures, is sufficient to cover all but
the advanced sections in Complex Numbers through Phasors.

The chapter on "linear algebra" is prerequisite for all subsequent chapters.

We use eight to ten lectures to cover it. We devote twelve to sixteen lectures
to cover topics from Vector Graphics through Binary Codes. (We assume a
semester consisting of 42 lectures and three exams.) The chapter on vector
graphics applies the linear algebra learned in the previous chapter to the
problem of translating, scaling, and rotating images. "Filtering" introduces
the student to basic ideas in averaging and filtering. The chapter on "Binary
Codes" covers the rudiments of binary coding, including Huffman codes
and Hamming codes.

If the users of this book find "Vector Graphics" through "Binary Codes" too
confining, we encourage them to supplement the essential material in
"Complex Numbers" through "Linear Algebra" with their own course notes
on additional topics. Within electrical and computer engineering there are
endless possibilities. Practically any set of topics that can be taught with
conviction and enthusiasm will whet the student's appetite. We encourage
you to write to us or to our editor, Tom Robbins, about your ideas for
additional topics. We would like to think that our book and its subsequent
editions will have an open architecture that enables us to accommodate a
wide range of student and faculty interests.

Throughout this book we have used MATLAB programs to illustrate key

ideas. MATLAB is an interactive, matrix-oriented language that is ideally
suited to circuit analysis, linear systems, control theory, communications,
linear algebra, and numerical analysis. MATLAB is rapidly becoming a
standard software tool in universities and engineering companies. (For more
information about MATLAB, return the attached card in the back of this
book to The MathWorks, Inc.) MATLAB programs are designed to develop
the student's ability to solve meaningful problems, compute, and plot in a
high-level applications language. Our students get started in MATLAB by
working through “An Introduction to MATLAB,” while seated at an IBM
PC (or look-alike) or an Apple Macintosh. We also have them run through
the demonstration programs in "Complex Numbers". Each week we give
three classroom lectures and conduct a one-hour computer lab session.
Students use this lab session to hone MATLAB skills, to write programs, or
to conduct the numerical experiments that are given at the end of each
chapter. We require that these experiments be carried out and then reported
in a short lab report that contains (i) introduction, (ii) analytical
computations, (iii) computer code, (iv) experimental results, and (v)
conclusions. The quality of the numerical results and the computer graphics
astonishes students. Solutions to the chapter problems are available from
the publisher for instructors who adopt this text for classroom use.

We wish to acknowledge our late colleague Richard Roberts, who

encouraged us to publish this book, and Michael Lightner and Ruth
Ravenel, who taught "Linear Algebra" and "Vector Graphics" and offered
helpful suggestions on the manuscript. We thank C. T. Mullis for allowing
us to use his notes on binary codes to guide our writing of "Binary Codes".
We thank Cédric Demeure and Peter Massey for their contributions to the
writing of "An Introduction to MATLAB" and "The Edix Editor". We thank
Tom Robbins, our editor at Addison-Wesley, for his encouragement,
patience, and many suggestions. We are especially grateful to Julie
Fredlund, who composed this text through many drafts and improved it in
many ways. We thank her for preparing an excellent manuscript for
production.

L. L. Scharf
R. T. Behrens
Boulder, Colorado

To the Teacher:
An incomplete understanding of complex numbers and phasors handicaps
students in circuits and electronics courses, and even more so in advanced
courses such as electromagnetics. optics, linear systems, control, and
communication systems. Our faculty has decided to address this problem as
early as possible in the curriculum by designing a course that drills complex
numbers and phasors into the minds of beginning engineering students. We
have used power signals, musical tones, Lissajous figures, light scattering,
and RLC circuits to illustrate the usefulness of phasor calculus. "Linear
Algebra" through "Binary Codes" introduce students to a handful of modern
ideas in electrical and computer engineering. The motivation is to whet
students' appetites for more advanced problems. The topics we have chosen
– linear algebra, vector graphics, filtering, and binary codes – are only
representative.

At first glance, many of the equations in this book look intimidating to

beginning students. For this reason, we proceed at a very measured pace. In
our lectures, we write out in agonizing detail every equation that involves a
N −1

sequence or series. For example, the sum ∑ z is written out as

n

n=0

Equation:
2 N −1
1 + z + z + ⋯ + z ,
and then it is evaluated for some specific value of z before we derive the
N
1 − z
analytical result Similarly, an infinite sequence like
1 − z
n
x
lim (1 + ) is written out as
n→∞ n
Equation:
2 3 100
x x x
(1 + x), (1 + ) , (1 + ) , ..., (1 + ) , ...,
2 3 100

and then it is evaluated for some specific x and for several values of n
before the limit is derived. We try to preserve this practice of pedantic
excess until it is clear that every student is comfortable with an idea and the
notation for coding the idea.

To the Student:
These are exciting times for electrical and computer engineering. To
celebrate its silver anniversary, the National Academy of Engineering
announced in February of 1990 the top ten engineering feats of the previous
twenty-five years. The Apollo moon landing, a truly Olympian and protean
achievement, ranked number one. However, a number of other
achievements in the top ten were also readily identifiable as the products of
electrical and computer engineers:

1. communication and remote sensing satellites,

2. the microprocessor,
3. computer-aided design and manufacturing (CADCAM),
4. computerized axial tomography (CAT scan),
5. lasers, and
6. fiber optic communication.

As engineering students, you recognize these achievements to be important

milestones for humanity; you take pride in the role that engineers have
played in the technological revolution of the twentieth century.
So how do we harness your enthusiasm for the grand enterprise of
engineering? Historically, we have enrolled you in a freshman curriculum
of mathematics, science, and humanities. If you succeeded, we enrolled you
in an engineering curriculum. We then taught you the details of your
profession and encouraged your faith that what you were studying is what
you must study to be creative and productive engineers. The longer your
faith held, the more likely you were to complete your studies. This seems
like an imperious approach to engineering education, even though
mathematics, physics, and the humanities are the foundation of engineering,
and details are what form the structure of engineering. It seems to us that a
better way to stimulate your enthusiasm and encourage your faith is to
introduce you early in your studies to engineering teachers who will share
their insights about some of the fascinating advanced topics in engineering,
while teaching you the mathematical and physical principles of engineering.
But you must match the teacher's commitment with your own commitment
to study. This means that you must attend lectures, read texts, and work
problems. You must be inquisitive and skeptical. Ask yourself how an idea
is limited in scope and how it might be extended to apply to a wider range
of problems. For, after all, one of the great themes of engineering is that a
few fundamental ideas from mathematics and science, coupled with a few
principles of design, may be applied to a wide range of engineering
problems. Good luck with your studies.
Complex Numbers: Introduction

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Notes to Teachers and Students:

When we teach complex numbers to beginning engineering students, we
encourage a geometrical picture supported by an algebraic structure. Every
algebraic manipulation carried out in a lecture is accompanied by a care-
fully drawn picture in order to fix the idea that geometry and algebra go
hand-in-glove to complete our understanding of complex numbers. We
assign essentially every problem for homework.

We use the MATLAB programs in this chapter to illustrate the theory of

complex numbers and to develop skill with the MATLAB language. The
numerical experiment introduces students to the basic quadratic equation of
electrical and computer engineering and shows how the roots of this
quadratic equation depend on the coefficients of the equation.

“Representing Complex Numbers in a Vector Space,” is a little demanding

for freshmen but easily accessible to sophomores. It may be covered for
additional insight, skipped without consequence, or covered after Chapter 4.
“An Electric Field Computation,” is well beyond most freshmen, and it is
demanding for sophomores. Nonetheless, an expert in electromagnetics
might want to cover the section "An electric Field Computation" for the
insight it brings to the use of complex numbers for representing two-
dimensional real quantities.

Introduction
It is hard to overestimate the value of complex numbers. They first arose in
the study of roots for quadratic equations. But, as with so many other great
discoveries, complex numbers have found widespread application well
outside their original domain of discovery. They are now used throughout
mathematics, applied science, and engineering to represent the harmonic
nature of vibrating systems and oscillating fields. For example, complex
numbers may be used to study

i. traveling waves on a sea surface;

ii. standing waves on a violin string;
iii. the pure tone of a Kurzweil piano;
iv. the acoustic field in a concert hall;
v. the light of a He-Ne laser;
vi. the electromagnetic field in a light show;
vii. the vibrations in a robot arm;
viii. the oscillations of a suspension system;
ix. the carrier signal used to transmit AM or FM radio;
x. the carrier signal used to transmit digital data over telephone lines; and
xi. the 60 Hz signal used to deliver power to a home.

In this chapter we develop the algebra and geometry of complex numbers.

In Chapter 3 we will show how complex numbers are used to build phasor
representations of power and communication signals.
Complex Numbers: Geometry of Complex Numbers

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The most fundamental new idea in the study of complex numbers is the
“imaginary number” j. This imaginary number is defined to be the square
root of −1:
Equation:

j = √ −1

Equation:
2
j = −1.

The imaginary number j is used to build complex numbers x and y in the

following way:
Equation:

z = x + jy.

We say that the complex number z has “real part” x and “imaginary part” y:
Equation:

z = Re[z] + j Im[z]

Equation:

Re[z] = x; Im[z] = y.
In MATLAB, the variable x is denoted by real(z), and the variable y is
denoted by imag(z). In communication theory, x is called the “in-phase”
component of z, and y is called the “quadrature” component. We call z =
x + jy the Cartesian representation of z, with real component x and

imaginary component y. We say that the Cartesian pair (x, y)codes the
complex number z.

We may plot the complex number z on the plane as in [link]. We call the
horizontal axis the “real axis” and the vertical axis the “imaginary axis.”
The plane is called the “complex plane.” The radius and angle of the line to
the point z = x + jy are
Equation:

2 2
r = √x + y

Equation:
y
−1
θ =tan ( ).
x

See [link]. In MATLAB, r is denoted by abs(z), and θ is denoted by

angle(z).

Cartesian and Polar Representations of the Complex Number z

The original Cartesian representation is obtained from the radius r and
angle θ as follows:
Equation:

x = rcosθ

Equation:

y = r sin θ.

The complex number z may therefore be written as

Equation:

z = x + jy

= rcosθ + jrsinθ

= r ( cos θ + j sin θ).

The complex number cosθ + jsinθ is, itself, a number that may be
represented on the complex plane and coded with the Cartesian pair
(cos θ, sinθ). This is illustrated in [link]. The radius and angle to the point

z =cos θ + jsinθ are 1 and θ. Can you see why?

The Complex Number cos θ + j sin θ

The complex number cosθ + jsinθ is of such fundamental importance to
our study of complex numbers that we give it the special symbol e :
jθ

Equation:
jθ
e =cos θ + jsinθ.

As illustrated in [link], the complex number e has radius 1 and angle θ.

jθ

With the symbol e , we may write the complex number z as

jθ

Equation:
jθ
z = re .

We call z = re a polar representation for the complex number z. We say

jθ

that the polar pair r∠θcodes the complex number z. In this polar
representation, we define |z| = r to be the magnitude of z and arg (z) = θ
to be the angle, or phase, of z:
Equation:

|z| = r

Equation:

arg (z) = θ.

With these definitions of magnitude and phase, we can write the complex
number z as
Equation:
jarg(z)
z = |z|e .

Let's summarize our ways of writing the complex number z and record the
corresponding geometric codes:
Equation:

( )
e
jθ

Exercise:
Problem:

Prove e
e

m is an integer.)

Exercise:
Problem:
z =

= (−1)

j[(π/2)+m2π]
n

= j, e
x + jy

(x, y)
=

2n+1

j[(3π/2)+m2π]
re

↓
jθ

r∠θ

= (−1) j
=

In "Roots of Quadratic Equations" we show that the definition

∣
, and
= −j, e
z e

= cosθ + jsinθ is more than symbolic. We show, in fact, that e

jθ

imaginary argument.
Exercise:
Problem:

Prove (j) 2n
and (j)
j arg(z)

is just
the familiar function e evaluated at the imaginary argument x = jθ. We
x

call e a “complex exponential,” meaning that it is an exponential with an

. Evaluate j

j(0+m2π)

= −1. Plot these identities on the complex plane. (Assume

j(π+m2π)

Find the polar representation z = re for each of the following

complex numbers:

a. z = 1 + j0;
b. z = 0 + j1;
c. z = 1 + j1;
d. z = −1 − j1.

Plot the points on the complex plane.

Exercise:
jθ
.

= 1
3 4
,j ,j
5
.
jθ
 Problem:

Find the Cartesian representation z = x + jy for each of the following

complex numbers:

a. z = √2e ;
jπ/2

b. z = √2e ;jπ/4

c. z = e ;
j3π/4

d. z = √2e .
j3π/2

Plot the points on the complex plane.

Exercise:
Problem:

The following geometric codes represent complex numbers. Decode

each by writing down the corresponding complex number z:

a. (0. 7, −0. 1) z = ?
b. (−1. 0, 0. 5) z = ?
c. 1. 6∠π/8 z =?
d. 0. 4∠7π/8 z =?

Exercise:
Problem:

Show that Im[jz] = Re[z] and Re[−jz] = Im[z]. Demo 1.1

(MATLAB). Run the following MATLAB program in order to
compute and plot the complex number e for
jθ

θ = i2π/360, i = 1, 2, ..., 360: j=sqrt(-1) n=360 for

i=1:n,circle(i)=exp(j*2*pi*i/n);end;
axis('square') plot(circle) Replace the explicit for loop
of line 3 by the implicit loop circle=exp(j*2*pi*[1:n]/n);
to speed up the calculation. You can see from [link] that the complex
number e , evaluated at angles θ = 2π/360, 2(2π/360), ..., turns out
jθ

complex numbers that lie at angle θ and radius 1. We say that e is a

jθ

complex number that lies on the unit circle. We will have much more
to say about the unit circle in Chapter 2.
The Complex Numbers e for 0 ≤ θ ≤ 2π (Demo 1.1)
jθ
Complex Numbers: Algebra of Complex Numbers

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The complex numbers form a mathematical “field” on which the usual

operations of addition and multiplication are defined. Each of these
operations has a simple geometric interpretation.

Addition and Multiplication.

The complex numbers z and z are 1 2

added according to the rule

Equation:

z1 + z2 = (x 1 + jy 1 ) + (x 2 + jy 2 )

= (x 1 + x 2 ) + j(y 1 + y 2 ).

We say that the real parts add and the imaginary parts add. As illustrated in
[link], the complex number z + z is computed from a “parallelogram
1 2

rule,” wherein z + z lies on the node of a parallelogram formed from z

1 2 1

and z .
2

Exercise:
Problem:

Let z 1 = r1 e
jθ 1
and z = r e . Find a polar formula z =r e for
2 2
jθ 2
3 3
jθ 3

z3 = z1 + z2 that involves only the variables r , r , θ , and θ . The

1 2 1 2

formula for r 3 is the “law of cosines.”

The product of z1 and z2 is
Equation:

z1 z2 = (x 1 + jy 1 )(x 2 + jy 2 )

= (x 1 x 2 − y 1 y 2 ) + j(y 1 x 2 + x 1 y 2 ).

Adding Complex Numbers

If the polar representations for z and z are used, then the product may be
1 2

written as [footnote]
We have used the trigonometric identities cos (θ + θ ) =cos θ cos θ −
1 2 1 2

sin θ1 sin θ2 and sin (θ + θ ) =sin θ cos θ + cos θ sin θ2 to derive this
1 2 1 2 1

result.
Equation:
 jθ 1 jθ 2
z1 z2 = r1 e r2 e

= (r 1 cos θ 1 + jr 1 sin θ 1 )(r 2 cos θ 2 + jr 2 sin θ 2 )

= ( r 1 cos θ 1 r 2 cos θ 2 − r 1 sin θ 1 r 2 sin θ 2 )

+ j( r 1 sin θ 1 r 2 cos θ 2 + r 1 cos θ 1 r 2 sin θ 2 )

= r 1 r 2 cos (θ 1 + θ 2 ) + jr 1 r 2 sin (θ 1 + θ 2 )

j(θ 1 +θ 2 )
= r1 r2 e .

We say that the magnitudes multiply and the angles add. As illustrated in
[link], the product z z lies at the angle (θ + θ ).
1 2 1 2

Multiplying Complex Numbers

Rotation. There is a special case of complex multiplication that will become

very important in our study of phasors in the chapter on Phasors. When z 1

is the complex number z = r e and z is the complex number

1 1
jθ 1
2

z = e
2 , then the product of z and z is
jθ 2
1 2
Equation:
jθ 2 j(θ 1 +θ 2 )
z1 z2 = z1 e = r1 e .

As illustrated in [link], z 1 z2 is just a rotation of z through the angle θ .

1 2

Rotation of Complex Numbers

Exercise:
Problem:

Begin with the complex number z = x + jy = re . Compute the

1
jθ

complex number z = jz in its Cartesian and polar forms. The

2 1

complex number z is sometimes called perp(z ). Explain why by

2 1

writing perp(z ) as z e . What is θ ? Repeat this problem for

1 1
jθ 2
2

z = −jz .
3 1
Powers. If the complex number z multiplies itself N times, then the result
1

is
Equation:
N N jN θ 1
(z 1 ) = r1 e .

This result may be proved with a simple induction argument. Assume

k
z = r e
1
k
1
. (The assumption is true for k = 1.) Then use the recursion
jkθ 1

z
k+1

1
k
= z z = r
1 1 e
k+1

1
. Iterate this recursion (or induction) until
j(k+1)θ 1

k + 1 = N . Can you see that, as n ranges from n = 1, ..., N , the angle of

z§ranges from θ1 to 2θ , ..., to N θ and the radius ranges from r1 to r , ...,

1 1
2
1

to r ? This result is explored more fully in Problem 1.19.

N
1

Complex Conjugate. Corresponding to every complex number z =

x + jy = re is the complex conjugate
jθ

Equation:
* −jθ
z = x − jy = re .

The complex number z and its complex conjugate are illustrated in [link].
The recipe for finding complex conjugates is to “change jto − j. This
changes the sign of the imaginary part of the complex number.
 A Complex Variable and Its Complex Conjugate

Magnitude Squared. The product of z and its complex conjugate is called

the magnitude squared of z and is denoted by |z| :
2

Equation:
2 * 2 2 −jθ jθ 2
|z| = z z = (x − jy) (x + jy) = x + y = re re = r .

Note that |z| = r is the radius, or magnitude, that we defined in "Geometry

of Complex Numbers".
Exercise:
Problem:

Write z as z * *
= zw . Find w in its Cartesian and polar forms.
Exercise:

*
Problem: Prove that angle (z 2 z1 ) = θ2 − θ1 .
Exercise:
 Problem:

Show that the real and imaginary parts of z = x + jy may be written

as
Equation:

1 *
Re [z] = z + z
2

Equation:

*
Im [z] = 2j z − z .

Commutativity, Associativity, and Distributivity. The complex numbers

commute, associate, and distribute under addition and multiplication as
follows:
Equation:

z1 + z2 = z2 + z1

z1 z2 = z2 z1

Equation:

(z 1 + z 2 ) + z 3 = z 1 + (z 2 + z 3 )

z 1 (z 2 z 3 ) = (z 1 z 2 )z 3

z 1 (z 2 + z 3 ) = z1 z2 + z1 z3 .

Identities and Inverses. In the field of complex numbers, the complex

number 0 + j0 (denoted by 0) plays the role of an additive identity, and the
complex number 1 + j0 (denoted by 1) plays the role of a multiplicative
identity:
Equation:
 z + 0 = z = 0 + z

z1 = z = 1z.

In this field, the complex number −z = −x + j(−y) is the additive inverse

x −y
of z, and the complex number z −1
=
2 2
+ j
2 2
is the
x + y x + y

multiplicative inverse:
Equation:

z + (−z) = 0

−1
zz = 1.

Exercise:
Problem:

Show that the additive inverse of z = re may be written as re jθ j(θ+π)

.

Exercise:

Problem: Show that the multiplicative inverse of z may be written as

Equation:

−1
1 *
1
z = z = (x − jy).
* 2 2
z z x + y

Show that z *
z is real. Show that z −1
may also be written as
Equation:
−1 −1 −jθ
z = r e .

Plot z and z −1
for a representative z.
Exercise:
 Problem: Prove (j) −1
= −j .
Exercise:

Problem: Find z −1
when z = 1 + j1.
Exercise:
Problem:
−1
* 1 *
Prove (z −1
) = z
*
= r
−1
e
jθ
= z . Plot z and (z −1
) for a
*
z z
representative z.
Exercise:
Problem:

Find all of the complex numbers z with the property that jz = −z . *

Illustrate these complex numbers on the complex plane.

Demo 1.2 (MATLAB). Create and run the following script file (name it
Complex Numbers)[footnote]
If you are using PC-MATLAB, you will need to name your file cmplxnos.m.

clear, clg
j=sqrt(-1)
z1=1+j*.5,z2=2+j*1.5
z3=z1+z2,z4=z1*z2
z5=conj(z1),z6=j*z2
avis([-4 4 -4 4]),axis('square'),plot(z1,'0')
hold on
plot(z2,'0'),plot(z3,'+'),plot(z4,'*'),
plot(z2,'0'),plot(z3,'+'),plot(z4,'*'),
plot(z5,'x'),plot(z6,'x')
 Complex Numbers (Demo 1.2)

With the help of Appendix 1, you should be able to annotate each line of
this program. View your graphics display to verify the rules for add,
multiply, conjugate, and perp. See [link].
Exercise:

Problem: Prove that z 0

= 1 .
Exercise:
Problem:

(MATLAB) Choose z = 1.05e

1
j2π/16
and z = 0.95e
2 . Write a
j2π/16

MATLAB program to compute and plot z and z for n = 1, 2, ..., 32.

n
1
n
2

You should observe a figure like [link].

Powers of z
Complex Numbers: Roots of Quadratic Equations

You probably first encountered complex numbers when you studied values
of z (called roots or zeros) for which the following equation is satisfied:
Equation:
2
az + bz + c = 0.

For a ≠ 0 (as we will assume), this equation may be written as

Equation:

2
b c
z + z + = 0.
a a

Let's denote the second-degree polynomial on the left-hand side of this

equation by p(z):
Equation:

2
b c
p (z) = z + z + .
a a

This is called a monic polynomial because the coefficient of the highest-

power term (z ) is 1. When looking for solutions to the quadratic equation
2

b c
z
2
+ z+ = 0 , we are really looking for roots (or zeros) of the
a a
polynomial p(z). The fundamental theorem of algebra says that there are
two such roots. When we have found them, we may factor the polynomial
p(z) as follows:

Equation:

2
b c
p (z) = z + z + = (z − z 1 ) (z − z 2 ).
a a

In this equation, z1 and z2 are the roots we seek. The factored form p(z) =
(z − z ) (z − z ) shows clearly that p (z ) = p (z ) = 0, meaning that the
1 2 1 2
quadratic equation p(z) = 0 is solved for z = z and z = z . In the process 1 2

of factoring the polynomial p(z), we solve the quadratic equation and vice
versa.

By equating the coefficients of z , z , and z0 on the left-and right-hand

2 1

sides of [link], we find that the sum and the product of the roots z and z 1 2

obey the equations

Equation:
b
z1 + z2 = −
a
.
c
z1 z2 =
a

You should always check your solutions with these equations.

Completing the Square. In order to solve the quadratic equation

b c
z
2
+ z + = 0 (or, equivalently, to find the roots of the polynomial
a a
b c
z +
2
z + ) , we “complete the square” on the left-hand side of [link]:
a a

Equation:
2 2
b b c
(z + ) − ( ) + = 0.
2a 2a a

This equation may be rewritten as

Equation:
2 2
b 1
2
(z + ) = ( ) (b − 4ac).
2a 2a

We may take the square root of each side to find the solutions
Equation:

b
 b 1
z 1,2 = − ± √ b 2 − 4ac.
2a 2a

Exercise:
Problem:

With the roots z and z defined in Equation 1.29, prove that

1 2

b c
(z − z 1 ) (z − z 2 ) is, indeed, equal to the polynomial z 2
+ z + .
a a
b c
Check that z 1 + z2 = − and z 1 z2 =
a a

In the equation that defines the roots z and z , the term b − 4ac is critical
1 2
2

because it determines the nature of the solutions for z and z . In fact, we

1 2

may define three classes of solutions depending on b − 4ac. 2

(i) Overdamped (b 2
− 4ac > 0) . In this case, the roots z and z are
1 2

Equation:
b 1
z 1,2 = − ± √ b 2 − 4ac.
2a 2a

These two roots are real, and they are located symmetrically about the point
b
− . When b = 0, they are located symmetrically about 0 at the points
2a
1
± √ −4ac . (In this case, −4ac > 0.) Typical solutions are illustrated in
2a
[link].

Typical Roots in the Overdamped Case; (a)

 b/2a > 0, 4ac > 0 , (b) b/2a > 0, 4ac < 0 , and
(c) b/2a = 0, 4ac < 0

Exercise:
Problem:

Compute and plot the roots of the following quadratic equations:

1
a. z 2
+ 2z + = 0 ;
2
1
b. z 2
+ 2a − = 0 ;
2
1
c. z 2
− = 0 .
2

b c
For each equation, check that z 1 + z2 = − and z 1 z2 =
a a

(ii) Critically Damped (b − 4ac = 0). In this case, the roots z and z
2
1 2

are equal (we say they are repeated):

Equation:

b
z1 = z2 = − .
2a

These solutions are illustrated in [link].

 Roots in the Critically Damped Case; (a)
b/2a > 0, and (b) b/2a < 0

Exercise:
Problem:

Compute and plot the roots of the following quadratic equations:

a. z 2
+ 2z + 1 = 0;
b. z 2
− 2z + 1 = 0;

c. z 2
= 0.

b c
For each equation, check that z 1 + z2 = − and z 1 z2 =
a a

(iii) Underdamped (b − 4ac < 0). The underdamped case is, by far, the
2

most fascinating case. When b − 4ac < 0, then the square root in the
2

solutions for z and z ([link]) produces an imaginary number. We may

1 2

write b − 4ac as −(4ac − b ) and write z as

2 2
1,2

Equation:
b 1
z 1,2 = − ± √ −(4ac − b 2 )
2a 2a

b 1
= − ± j √ 4ac − b 2 .
2a 2a

These complex roots are illustrated in [link]. Note that the roots are

purely imaginary when b = 0, producing the result

Equation:

c
z 1,2 = ±j√ .
a
 Figure 1.12: Typical Roots in the Underdamped Case; (a)
b/2a > 0, (b)b/2a < 0, and (c) b/2a = 0

In this underdamped case, the roots z and z are complex conjugates:

1 2

Equation:
*
z2 = z1 .

b c
Thus the polynomial p (z) = z 2
+ z + = (z − z 1 ) (z − z 2 ) also
a a
takes the form
Equation:
*
p(z) = (z − z 1 )(z − z )
1

2 2
= z − 2 Re [z 1 ]z + |z 1 | .

and |z
Re[z 1 ] 1|
2
are related to the original coefficients of the polynomial as
follows:
Equation:

b
2 Re [z 1 ] = −
a
Equation:

Always check these equations.

[link] is now
Equation:

Equation:

Exercise:
p(z)

θ
=

∣
2r cos θ

=
r
z1

(z − re

2
2

These equations may be used to locate z

± cos
2

jθ

√
1,2

−1
c

Let's explore these connections further by using the polar representations

for z and z :
1

Equation:
2

z 1,2 = re
±jθ

(
c

a
−
.

Then [link] for the polynomial p(z) may be written in the “standard form”
Equation:

)(z − re

− 2r cos θz + r .

= re
b

−b

√ 4ac
−jθ

±jθ

).
2
)
 Problem:

Prove that p(z) may be written as p (z) = z 2

− 2r cos θz + r in the 2

underdamped case.
Exercise:
Problem:

Prove the relations in [link]. Outline a graphical procedure for locating

b c
z 1 = re
jθ
and z 2 = re
−jθ
from the polynomial z 2
+ z + .
a a

Exercise:
Problem:

Compute and plot the roots of the following quadratic equations:

a. z 2
+ 2z + 2 = 0;
b. z 2
− 2z + 2 = 0;

c. z 2
+ 2 = 0.

2
b c
For each equation, check that 2 Re [z 1,2 ] = − and z 1,2 = .
a a
Complex Numbers: Representing Complex Numbers in a Vector Space

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

So far we have coded the complex number z = x + jy with the Cartesian

pair (x, y) and with the polar pair (r∠θ). We now show how the complex
number z may be coded with a two-dimensional vector z and show how this
new code may be used to gain insight about complex numbers.

Coding a Complex Number as a Vector. We code the complex number

x
z = x + jy with the two-dimensional vector z = [ ] :
y

Equation:

x
x + jy = z ⇔ z = [ ].
y

We plot this vector as in [link]. We say that the vector z belongs to a

“vector space.” This means that vectors may be added and scaled according
to the rules
Equation:

x1 + x2
z1 + z2 = [ ]
y1 + y2

Equation:
 ax
az = [ ].
ay

The Vector z Coding the Complex Number z

Furthermore, it means that an additive inverse −z, an additive identity 0,

and a multiplicative identity 1 all exist:
Equation:

z + (−z) = 0

Equation:

lz = z.

0
The vector 0 is 0 = [ ] .
0

Prove that vector addition and scalar multiplication satisfy these properties
of commutation, association, and distribution:
Equation:
z1 + z2 = z2 + z1

Equation:
 (z 1 + z 2 ) + z 3 = z 1 + (z 2 + z 3 )

Equation:

a(bz) = (ab)z

Equation:

a (z 1 + z 2 ) = az 1 + az 2 .

Inner Product and Norm. The inner product between two vectors z and 1

z is defined to be the real number

2

Equation:

(z 1 , z 2 ) = x 1 x 2 + y 1 y 2 .

We sometimes write this inner product as the vector product (more on this
in Linear Algebra)
Equation:
T
(z 1 , z 2 ) = z z2
1

x2
= [x 1 y 1 ][ ] = (x 1 x 2 + y 1 y 2 ).
y2

Exercise:

Problem: Prove (z 1
, z 2 ) = (z 2 , z 1 ).

When z = z
1 2 = z , then the inner product between z and itself is the norm
squared of z:
Equation:
2 2 2
||z|| = (z, z) = x + y .
These properties of vectors seem abstract. However, as we now show, they
may be used to develop a vector calculus for doing complex arithmetic.

A Vector Calculus for Complex Arithmetic. The addition of two complex

numbers z and z corresponds to the addition of the vectors z and z :
1 2 1 2

Equation:

x1 + x2
z1 + z2 ⇔ z1 + z2 = [ ]
y1 + y2

The scalar multiplication of the complex number z by the real number x

2 1

corresponds to scalar multiplication of the vector z by x :

2 1

Equation:

x2 x1 x2
x1 z2 ⇔ x1 [ ] = [ ].
y2 x1 y2

Similarly, the multiplication of the complex number z by the real number

2

y is
1

Equation:

x2 y1 x2
y1 z2 ↔ y1 [ ] = [ ].
y2 y1 y2

The complex product z 1 z2 = (x 1 + jy 1 )z 2 is therefore represented as

Equation:

x1 x2 − y1 y2
z1 z2 ↔ [ ].
x1 y2 + y1 x2

This representation may be written as the inner product

Equation:
 (v, z 1 )
z1 z2 = z2 z1 ↔ [ ]
(w, z 1 )

x2 y2
where v and w are the vectors v = [ ] and w = [ ] . By defining the
−y 2 x2

matrix
Equation:

x2 −y 2
[ ],
y2 x2

we can represent the complex product z 1

z2 as a matrix-vector multiply
(more on this in Linear Algebra):
Equation:

x2 −y 2 x1
z1 z2 = z2 z1 ↔ [ ][ ].
y2 x2 y1

With this representation, we can represent rotation as

Equation:

cos θ − sin θ x1
jθ jθ
ze = e z ↔ [ ][ ].
sin θ cos θ x2

cos θ − sin θ
We call the matrix [ ] a “rotation matrix.”
sin θ cos θ

Exercise:

Problem: Call R(θ) the rotation matrix:

Equation:

cos θ − sin θ
R (θ) = [ ].
sin θ cos θ
 Show that R(−θ) rotates by (−θ). What can you say about R(−θ)w
when w = R(θ)z?
Exercise:

Problem: Represent the complex conjugate of z as

Equation:

a b x
*
z ↔ [ ] [ ]
c d y

and find the elements a, b, c, and d of the matrix.

Inner Product and Polar Representation. From the norm of a vector, we

derive a formula for the magnitude of z in the polar representation z = re jθ

:
Equation:
1/2 1/2
2 2
r = x + y = ||z|| = (z, z) .

1 0
If we define the coordinate vectors e 1 = [ ] and e 2 = [ ] , then we can
0 1

represent the vector z as

Equation:

z = (z, e 1 )e 1 + (z, e 2 )e 2 .

See [link]. From the figure it is clear that the cosine and sine of the angle θ
are
Equation:

(z, e 1 ) (z, e 2 )
cos θ = ; sin θ =
||z|| ||z||
 Representation of z in its Natural Basis

This gives us another representation for any vector z:

Equation:

z = ||z|| cos θe 1 + ||z|| sin θe 2 .

The inner product between two vectors z and z is now 1 2

Equation:

(z 2 , e 1 )e 1
T T
(z 1 , z 2 ) = (z 1 , e 1 )e (z 1 , e 2 )e [ ]
1 2
(z 2 , e 2 )e 2

= (z 1 , e 1 ) (z 2 , e 1 ) + (z 1 , e 2 ) (z 2 , e 2 )

= ||z 1 || cos θ 1 ||z 2 || cos θ 2 + ||z 1 || sin θ 1 ||z 2 || sin θ 2 .

It follows that cos (θ 2 − θ 1 ) =cos θ 2 cos θ 1+ sin θ 1 sin θ 2 may be

written as
Equation:
 (z 1 , z 2 )
cos (θ 2 − θ 1 ) =
||z 1 || ||z 2 ||

This formula shows that the cosine of the angle between two vectors z and 1
*
z , which is, of course, the cosine of the angle of z z , is the ratio of the
2 2 1

inner product to the norms.

Exercise:
Problem:

Prove the Schwarz and triangle inequalities and interpret them:

Equation:
2 2 2
(Schwarz) (z 1 , z 2 ) ≤ ||z 1 || ||z 2 ||

Equation:

(triangle) I z 1 − z 2 || ≤ ||z 1 − z 3 || + ||z 2 − z3 ||.

Complex Numbers: An Electric Field Computation

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

We have established that vectors may be used to code complex numbers.

Conversely, complex numbers may be used to code or represent the
orthogonal components of any two-dimensional vector. This makes them
invaluable in electromagnetic field theory, where they are used to represent
the components of electric and magnetic fields.

The basic problem in electromagnetic field theory is to determine the

electric or magnetic field that is generated by a static or dynamic
distribution of charge. The key idea is to isolate an infinitesimal charge,
determine the field set up by this charge, and then to sum the fields
contributed by all such infinitesimal charges. This idea is illustrated in
[link], where the charge λ, uniformly distributed over a line segment of
length dx at point −x, produces a field dE(x) at the test point (0, h). The
field dE(x) is a “vector” field (as opposed to a “scalar” field), with
components E (x) and E (x). The intensity or field strength of the field
1 2

dE(x) is

Equation:

λdx
|dE (x)| =
2 2
4πϵ 0 (h + x )
 Infinitesimal Charge λdx Producing Field dE(x)

But the field strength is directed at angle θ(x), as illustrated in [link]. The
field dE(x) is real with components dE (x) and dE (x), but we code it
1 2

as a complex field. We say that the “complex” field at test point (0, h) is
Equation:

λdx jθ(x)
dE (x) = e
2 2
4πϵ 0 (h + x )

with components dE 1
(x) and dE 2
(x) . That is,
Equation:

dE (x) = dE 1 (x) + jdE 2 (x)

Equation:

λdx
dE 1 (x) = cos θ (x)
2 2
4πϵ 0 (h + x )

Equation:

λdx
dE 2 (x) = sin θ (x).
2 2
4πϵ 0 (h + x )

For charge uniformally distributed with density λ along the x-axis, the total
field at the test point (0, h) is obtained by integrating dE:
Equation:
∞ ∞
λ
∫ dE (x) = ∫ [cos θ (x) + jsinθ (x)]dx.
2 2
−∞ −∞ 4πϵ 0 (h + x )

The functions cos θ(x) and sin θ(x) are

Equation:

x h
cos θ (x) = ; sin θ (x) =
1/2 1/2
2 2 2 2
(x + h ) (x + h )

We leave it as a problem to show that the real component E of the field is 1

zero. The imaginary component E is 2

Equation:
∞
λh dx
E = jE 2 = j ∫
3/2
4πϵ 0 2 2
−∞ (x + h )

Equation:
∞

λh x
= j
1/2
4πϵ 0 2 2 2
h (x + h ) −∞
Equation:

λh 1 1 λ
= j + = j
2 2
4πϵ 0 h h 2πϵ 0 h

Equation:

λ
E2 = .
2πϵ 0 h

We emphasize that the field at (0, h) is a real field. Our imaginary answer
simply says that the real field is oriented in the vertical direction because
we have used the imaginary part of the complex field to code the vertical
component of the real field.
Exercise:
Problem:

Show that the horizontal component of the field E is zero. Interpret

this finding physically.

From the symmetry of this problem, we conclude that the field around the
infinitely long wire of [link] is radially symmetric. So, in polar coordinates,
we could say
Equation:

λ
E (r, θ) =
2πϵ 0 r

which is independent of θ. If we integrated the field along a radial line

perpendicular to the wire, we would measure the voltage difference
Equation:
r1
λ λ
V (r 1 ) − V (r 0 ) = ∫ dr = log r 1 − log r 0 .
r0
2πϵ 0 r 2πϵ 0
An electric field has units of volts/meter, a charge density λ has units of
coulombs/meter, and ϵ has units of coulombs/volt-meter; voltage has units
0

of volts (of course).

Complex Numbers: Numerical Experiment (Quadratic Roots)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

There is a version of the quadratic equation that will arise over and over
again in your study of electrical and mechanical systems:
Equation:
2 2
s + 2ξω 0 s + ω 0 = 0.

For reasons that can only become clear as you continue your study of
engineering, the parameter ω is called a resonant frequency, and the
0

parameter ξ ≥ 0 is called a damping factor. In this experiment, you will

begin by

1. finding the “underdamped” range of values ξ ≥ 0 for which the roots

s1 and s2 are complex;
2. finding the “critically damped” value of ξ ≥ 0 that makes the roots s1
and s2 equal; and
3. finding the “overdamped” range of values ξ ≥ 0 for which s1 and s2
are real.
4. For each of these ranges, find the analytical solution for s as a
1,2

function of ω0 and ξ; write your solutions in Cartesian and polar forms

and present your results as
Equation:

, 0 ≤ ξ ≤ ξc

s 1,2 = , ξ = ξc

, ξ ≥ ξc
 where ξc is the critically damped value of ξ. Write a MATLAB
program that computes and plots s for ω0 fixed at ω = 1 and ξ
1,2 0

variable between 0.0 and 2.0 in steps of 0.1. Interpret all of your
findings.

Now organize the coefficients of the polynomial s 2

+ 2ξs + 1 into the
array [12ξ1]. Imbed the MATLAB instructions

r=roots([1 2*e 1]);

plot(real(r(1)),imag(r(1)),'o')
plot(real(r(2)),imag(r(2)),'o')

in a for loop to compute and plot the roots of s + 2ξs + 1 as ξ ranges

2

from 0.0 to 2.0. Note that r is a 1 × 2 array of complex numbers. You

should observe [link]. We call this “half circle and line” the locus of roots
for the quadratic equation or the “root locus” in shorthand.
Roots of Quadratic Equation
The Functions e^x and e^jθ: Introduction

Note: This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Notes to Teachers and Students

It is essential to write out, term-by-term, every sequence and sum in this
chapter. This demystifies the seemingly mysterious notation. The example
on compound interest shows the value of limiting arguments in everyday
life and gives e some real meaning. The function e , covered in the
x jθ

section "The Function of ejθ and the Unit Circle and "Numerical
Experiment (Approximating ejθ, must be understood by all students before
proceeding to "Phasors" . The Euler and De Moivre identities provide every
tool that students need to derive trigonometric formulas. The properties of
roots of unity are invaluable for the study of phasors in "Phasors" .

The MATLAB programs in this chapter are used to illustrate sequences and
series and to explore approximations to sin θ and cos θ. The numerical
experiment in "Numerical Experiment (Approximating ejθ illustrates,
geometrically and algebraically, how approximations to e converge.
jθ

“Second-Order Differential and Difference Equations” is a little demanding

for freshmen, but we give it a once-over-lightly to illustrate the power of
quadratic equations and the functions ex and e . This section also gives a
jθ

sneak preview of more advanced courses in circuits and systems.

Introduction
It is probably not too strong a statement to say that the function e is the
x

most important function in engineering and applied science. In this chapter

we study the function e and extend its definition to the function e . This
x jθ

study clarifies our definition of e from "Complex Numbers" and leads us

jθ

to an investigation of sequences and series. We use the function e to

jθ

derive the Euler and De Moivre identities and to produce a number of

important trigonometric identities. We define the complex roots of unity
and study their partial sums. The results of this chapter will be used in
"Phasors" when we study the phasor representation of sinusoidal signals.
The Functions e^x and e^jθ: The Function e^x

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Many of you know the number e as the base of the natural logarithm, which
has the value 2.718281828459045. . . . What you may not know is that this
number is actually defined as the limit of a sequence of approximating
numbers. That is,
Equation:
e = lim fn
n→∞

Equation:
n
1
f n = (1 + ) , n = 1, 2, ...
n

This means, simply, that the sequence of numbers

2 3
1 1
, . . . , gets arbitrarily close to
1
(1 + 1) , (1 + ) , (1 + )
2 3

2.718281828459045. . . . But why should such a sequence of numbers be so

important? In the next several paragraphs we answer this question.
Exercise:
 Problem:

(MATLAB) Write a MATLAB program to evaluate the expression

n
1
f n = (1 + ) for n = 1, 2, 4, 8, 16, 32, 64 to show that f n ≈ e
n

for large n.

n
1
Derivatives and the Number e. The number f n = (1 + ) arises in
n

the study of derivatives in the following way. Consider the function

Equation:
x
f (x) = a , a > 1

and ask yourself when the derivative of f (x) equals f (x). The function
f (x) is plotted in [link] for a > 1. The slope of the function at point x is

Equation:
df (x) a
x+Δx
−a
x

= lim
dx Δx
Δx→0

Δx
x α −1
= α lim .
Δx
Δx→0
 The Function f (x) = a x

If there is a special value for a such that

Equation:
Δx
a − 1
lim = 1,
Δx→0 Δx

d
then f (x) would equal f (x). We call this value of a the special (or
dx
exceptional) number e and write
Equation:
x
f (x) = e

d x
f (x) = e .
dx

The number e would then be e = f (1). Let's write our condition that
Δx
a − 1
converges to 1 as
Δx
Equation:
Δx
e − 1 ≅ Δx, Δxsmall

or as
Equation:
1/Δx
e ≅ (1 + Δx) .

1/n
1 1
Our definition of e = lim (1 + ) amounts to defining Δx =
n→∞
n n

and allowing n → ∞ in order to make Δx → 0. With this definition for e,

it is clear that the function e is defined to be (e) :
x x

Equation:
x x/Δx
e = lim (1 + Δx) .
Δx→0

x
By letting Δx = we can write this definition in the more familiar form
n
Equation:
n
x
x
e = lim (1 + )
n→∞
n

This is our fundamental definition for the function e . When evaluated at x

x = 1, it produces the definition of e given in [link].

The derivative of e is, of course,

x

Equation:
n−1
d x
x 1 x
e = lim n(1 + ) = e .
n→∞
dx n n

This means that Taylor's theorem[footnote] may be used to find another

characterization for e :x

Taylor's theorem says that a function may be completely characterized by

all of its derivatives (provided they all exist).
Equation:
n
x ∞ d x 1 n
e = ∑ [ n
e ] x
n=0 dx x=0 n!
n
∞ x
= ∑ .
n=0 n!

When this series expansion for e is evaluated at x = 1, it produces the

x

following series for e:

Equation:
 ∞
1
e = ∑ .
n!
n=0

In this formula, n! is the product n(n − 1)(n − 2) ⋯ (2)1. Read n! as "n

factorial.”
Exercise:

Problem: (MATLAB) Write a MATLAB program to evaluate the sum

Equation:
N
1
SN = ∑
n!
n=0

for N = 1, 2, 4, 8, 16, 32, 64 to show that S ≅ e for large N .

N

Compare S with f from [link]. Which approximation do you

64 64

prefer?

Compound Interest and the Function 𝕖 𝕩 . There is an example from your

everyday life that shows even more dramatically how the function e arises.
x

Suppose you invest V dollars in a savings account that offers 100x%

0

annual interest. (When x = 0. 01, this is 1%; when x = 0. 10, this is 10%
interest.) If interest is compounded only once per year, you have the simple
interest formula for V , the value of your savings account after 1 compound
1

(in this case, 1 year):

V 1 = (1 + x)V 0 . This result is illustrated in the block diagram of Figure

2.2(a). In this diagram, your input fortune V0 is processed by the “interest
block” to produce your output fortune V1. If interest is compounded
monthly, then the annual interest is divided into 12 equal parts and applied
12 times. The compounding formula for V12, the value of your savings after
12 compounds (also 1 year) is
Equation:
 12
x
V 12 = (1 + ) V0 .
12

This result is illustrated in [link]. Can you read the block diagram? The
general formula for the value of an account that is compounded n times per
year is
Equation:
n
x
V n = (1 + ) V0 .
n

Vn is the value of your account after n compounds in a year, when the

annual interest rate is 100x%.

Block Diagram for Interest Computations; (a) Simple Annual Interest,

and (b) Monthly Compounding

Exercise:
Problem:

Verify in [link] that a recursion is at work that terminates at V . That n

x
is, show that V i+1 = (1 + )V 1 for i = 0, 1, ..., n − 1 produces the
n
n
x
result V n = (1 + ) V0 .
n
Bankers have discovered the (apparent) appeal of infinite, or continuous,
compounding:
Equation:
n
x
V ∞ = lim (1 + ) V0 .
n→∞ n

We know that this is just

Equation:
x
V∞ = e V0 .

So, when deciding between 100x % interest compounded daily and

1

100x % interest compounded continuously, we need only compare

2

Equation:
365
x1 x2
(1 + ) versus e .
365

We suggest that daily compounding is about as good as continuous

compounding. What do you think? How about monthly compounding?
Exercise:
Problem:

(MATLAB) Write a MATLAB program to compute and plot simple

interest, monthly interest, daily interest, and continuous interest versus
interest rate 100x. Use the curves to develop a strategy for saving
money.
The Functions e^x and e^jθ: The Function e^jθ and the Unit Circle

Note:This module is part of the collection, A First Course in Electrical and

Computer Engineering. The LaTeX source files for this collection were created
using an optical character recognition technology, and because of this process
there may be more errors than usual. Please contact us if you discover any
errors.

Let's try to extend our definitions of the function e to the argument x = jΘ. x

Then e is the function

jΘ

Equation:
n

jθ
θ
e = lim (1 + j )
n→∞ n

θ
The complex number 1 + j is illustrated in [link]. The radius to the point
n
1/2
2
θ θ θ
1 + j is r = (1 + 2
) and the angle is φ =tan −1
This means that
n n n
n/2
θ 2

the n th
power of 1 + j has radius r n
= (1 +
θ

n
2
) and angle
n
θ
nφ = n tan
−1
(Recall our study of powers of z.) Therefore the complex
n
n
θ
number (1 + j ) may be written as
n

Equation:
n n/2
2
θ θ −1
θ −1
θ
(1 + j ) = (1 + ) [cos (n tan ) + j sin (n tan )].
2
n n n n
 n/2
2
θ θ θ
For n large, (1 + 2
) 2 ≅1 , and n tan −1
≅n = θ . Therefore
n n n
n
θ
(1 + j ) is approximately
n

Equation:
n
θ
(1 + j ) = 1(cos θ + j sin θ).
n

n
θ
(1 + j ) = 1 ( cos θ + j sin θ).
n

This finding is consistent with our previous definition of e ! jθ

The Complex Number 1 + j n

θ

The series expansion for e is obtained by evaluating Taylor's formula at

jθ

x = jθ:

Equation:
∞
1 n
jθ
e = ∑ (jθ) .
n!
n=0

When this series expansion for e is written out, we have the formula
jθ

Equation:
 jθ ∞ 1 2n ∞ 1 2n+1
e = ∑ (jθ) + ∑ (jθ)
n=0 (2n)! n=0 (2n+1)!

n n
∞ (−1) 2n ∞ (−1) 2n+1
= ∑ θ + j∑ θ .
n=0 (2n)! n=0 (2n+1)!

It is now clear that cos θ and sin θ have the series expansions
Equation:
n
∞ (−1)
2n
cos θ = ∑ θ
n=0
(2n)!

Equation:
∞ n
(−1)
2n+1
sin θ = ∑ θ .
(2n + 1)!
n=0

When these infinite sums are truncated at N − 1 , then we say that we have N-
term approximations for cos θ and sin θ:
Equation:
n
N −1 (−1)
2n
cos θ ≅ ∑ θ
n=0
(2n)!

Equation:
N −1 n
(−1)
2n+1
sin θ ≅ ∑ θ .
(2n + 1)!
n=0

The ten-term approximations to cos θ and sin θ are plotted over exact
expressions for cos θ and sin θ in [link]. The approximations are very good
over one period (0 ≤ θ ≤ 2π), but they diverge outside this interval. For more
accurate approximations over a larger range of θ s, we would need to use more
′

terms. Or, better yet, we could use the fact that cos θ and sin θ are periodic in θ
. Then we could subtract as many multiples of 2π as we needed from θ to bring
the result into the range [0, 2π] and use the ten-term approximations on this new
variable. The new variable is called θ-modulo 2π.
 Ten-Term Approximations to cos θ and sin θ

Exercise:
Problem:

Write out the first several terms in the series expansions for cos θ and
sin θ.

Demo 2.1 (MATLAB). Create a MATLAB file containing the following demo
MATLAB program that computes and plots two cycles of cos θ and sin θ
versus θ. You should observe [link]. Note that two cycles take in 2(2π) radians,
which is approximately 12 radians.

clg;
j = sqrt(-1);
theta = 0:2*pi/50:4*pi;
s = sin(theta);
c = cos(theta);
plot(theta,s);
elabel('theta in radians');
ylabel('sine and cosine');
hold on
plot(theta,c);
hold off
 The Functions cos θ and sin θ

Exercise:
Problem:

(MATLAB) Write a MATLAB program to compute and plot the ten-term

approximations to cos θ and sin θ for θ running from 0 to 2(2π) in steps
of 2π/50. Compute and overplot exact expressions for cos θ and sin θ.
You should observe a result like [link].

The Unit Circle. The unit circle is defined to be the set of all complex numbers
z whose magnitudes are 1. This means that all the numbers on the unit circle

may be written as z = e . We say that the unit circle consists of all numbers
jθ

generated by the function z = e as θ varies from 0 to 2π. See [link].

jθ

A Fundamental Symmetry. Let's consider the two complex numbers z and 1

1 1
, illustrated in [link]. We call the “reflection of z through the unit circle”
* *
z1 z1

1 1
(and vice versa). Note that z 1 = r1 e
jθ 1
and = e
jθ 1
. The complex
*
r1
z1

1
numbers z 1 − e
jθ
and − e
jθ
are illustrated in [link]. The magnitude squared
*
z
1

of each is
Equation:
Equation:

Equation:

the unit circle:

Equation:

∣
1

*
z1

*
z1
z1 − e

− e

β
2
jθ
jθ

The ratio of these magnitudes squared is

β
2
=

=
2

= (
= (z 1 − e

(z 1 − e

(
1

z
*

1
− e

zi
1

z1

(e

(1−e

2
*
− e

jθ

jθ
)(

This ratio may be manipulated to show that it is independent of θ, meaning that

the points z and
1
maintain a constant relative distance from every point on

e
jθ −jθ
jθ

jθ

) (z 1 − e

z 1 −1)(z e

jθ

|z 1 | , independent of θ!

This result will be of paramount importance to you when you study digital
filtering, antenna design, and communication theory.
*
) (z 1 − e

)(

z1

z 1 )(1−z 1 e
*

1
− e

jθ
*

z1

−1)e

−jθ
−jθ

−jθ

)
− e

−jθ

z
)

1
−jθ

)
)

−jθ
).
 The Unit Circle

Exercise:

Problem: Write the complex number z − e as re . What are r and φ?

jθ jφ
The Functions e^x and e^jθ: The Euler and De Moivre Identities

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The Euler and De Moivre identities are the fundamental identities for
deriving trigonometric formulas. From the identity e =cos θ + jsin θ and
jθ

*
the conjugate identity e −jθ
= (e ) =cos θ − jsin θ, we have the Euler
jθ

identities for cos θ and sin θ:

Equation:
jθ −jθ
e +e
cos θ =
2

jθ −jθ
e −e
sin θ =
2j

These identities are illustrated in [link].

 Euler's Identities

The identity e jθ
=cos θ + j sin θ also produces the De Moivre identity:
Equation:
n n
jθ jnθ
(cos θ + j sin θ) = (e ) = e
.
= cos nθ + j sin nθ

When the left-hand side of this equation is expanded with the binomial
expansion, we obtain the identity
Equation:
n
n n−k k
∑( )(cos θ) (j sin θ) =cos nθ + j sin nθ.
k
k=0

Binomial Coefficients and Pascal's Triangle. The binomial coefficients

( ) in [link] are shorthand for the number
n
k
Equation:

n n!
( ) = k = 0, 1, ..., n.
k (n − k)!k!′

This number gives the coefficient of x y in the expansion of (x + y) .

n−k k n

How do we know that there are ( ) terms of the form x y ? One way to
n
k
n−k k

answer this question is to use Pascal's triangle, illustrated in [link]. Each

node on Pascal's triangle shows the number of routes that terminate at that
node. This number is always the sum of the number of routes that terminate
at the nodes just above the node in question. If we think of a left-hand path
as an occurrence of an x and a right-hand path as an occurrence of a y, then
we see that Pascal's triangle keeps track of the number of occurrences of
y .
n−k k
x

Pascal's Triangle and the Binomial Coefficients

Exercise:

Problem: Prove ( n
k
) = (
n
n−k
) .
Exercise:
 Problem: Find an identity for ( n−1

k−1
) + (
k
n−1
) .
Exercise:

Problem: Find “half-angle” formulas for cos 2θ and sin 2θ.

Exercise:

Problem: Show that

A. cos 3θ =cos θ − 3cos θsin θ;

2 2

B. sin 3θ = 34 cos θsin θ− sin θ. 2 3

Exercise:
Problem:

Use e j(θ 1 +θ 2 )
= e
jθ 1
e
jθ 2
= (cos θ 1 + j sin θ 1 ) (cos θ 2 + j sin θ 2 ) to
prove

a. cos (θ 1 + θ 2 ) =cos θ 1 cos θ 2 sin θ 1 sin θ 2 ;

b. sin (θ 1 + θ 2 ) =sin θ 1 cos θ 2 + sin θ 2 cos θ 1 .
The Functions e^x and e^jθ: Roots of Unity and Related Topics

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The complex number z = e is illustrated in [link]. It lies on the unit

j2π/N

circle at angle θ = 2π/N . When this number is raised to the n power, the
th

result is z = e
n j2πn/N
. This number is also illustrated in [link]. When one
of the complex numbers e j2πn/N
is raised to the N power, the result is
th

Equation:
N
j2πn/N j2πn
(e ) = e = 1.
 The Complex Numbers e j2π/N
and e j2πn/N

We say that e j2πn/N

is one of the N th
roots of unity, meaning that e j2πn/N

is one of the values of z for which

Equation:
N
z − 1 = 0.

There are N such roots, namely,

Equation:
j2πn/N
e , n = 0, 1, ..., N − 1.
As illustrated in [link], the 12 roots of unity are uniformly distributed
th

around the unit circle at angles 2πn/12. The sum of all of the N roots ofth

unity is zero:
Equation:
N −1

j2πn/N
SN = ∑ e = 0.

n=0

This property, which is obvious from [link], is illustrated in [link], where

k−1

the partial sums S k = ∑e

j2πn/N
are plotted for k = 1, 2, ..., N .
n=0

Roots of Unity

These partial sums will become important to us in our study of phasors and
light diffraction in "Phasors" and in our discussion of filters in "Filtering".
 Partial Sums of the Roots of Unity

Geometric Sum Formula. It is natural to ask whether there is an analytical

expression for the partial sums of roots of unity:
Equation:
k−1

j2πn/N
Sk = ∑ e .

n=0

We can imbed this question in the more general question, is there an

analytical solution for the “geometric sum”
Equation:
 k−1

n
Sk = ∑ z ?

n=0

The answer is yes, and here is how we find it. If z = 1, the answer is
S = k. If z ≠ 1, we can premultiply S by z and proceed as follows:
k k

Equation:
k−1 n+1 k m
zS k = ∑ z = ∑ z
n=0 m=1

k−1 m k
= ∑ z + z − 1
m=0

k
= Sk + z − 1.

From this formula we solve for the geometric sum:

Equation:
k
1−z
z ≠ 1
1−z
Sk = {
k, z = 1.

This basic formula for the geometric sum Sk is used throughout

electromagnetic theory and system theory to solve problems in antenna
design and spectrum analysis. Never forget it.
Exercise:

k−1 k−1

Problem: Find formulas for S k = ∑e

jnθ
and for S k = ∑e
j2π/N n
.

n=0 n=0

Exercise:

N −1

Problem: Prove ∑ e j2πn/N

= 0 .
n=0

Exercise:
 Problem:

Find formulas for the magnitude and phase of the partial sum
k−1

Sk = ∑ e
j2πn/N
.
n=0

Exercise:
Problem:

(MATLAB) Write a MATLAB program to compute and plot the partial

k−1

sum S k = ∑e
j2πn/N
for k = 1, 2, ..., N . You should observe [link].
n=0

Exercise:

Problem: Solve the equation (z + 1) 3

= z
3
.
Exercise:

Problem: Find all roots of the equation z 3

+ z
2
+ 3z − 15 = 0 .
Exercise:
Problem:

Find c so that (1 + j) is a root of the equation z 17

+ 2z
15
− c = 0 .
The Functions e^x and e^jθ: Second-Order Differential and Difference
Equations

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

With our understanding of the functions e , e , and the quadratic equation

x jΘ

= 0, we can undertake a rudimentary study of differential

2 b c
z + z +
a a

and difference equations.

Differential Equations. In your study of circuits and systems you will

encounter the homogeneous differential equation
Equation:
2
d d
x (t) + a 1 x (t) + a 2 = 0.
2
dt dt

Because the function e reproduces itself under differentiation, it is

st

plausible to assume that x (t) = e is a solution to the differential

st

equation. Let's try it:

Equation:
2
d st d st st
2
(e ) + a1 (e ) + a 2 (e ) = 0
dt dt

2 st
(s + a 1 s + a 2 )e = 0.

If this equation is to be satisfied for all t, then the polynomial in s must be

zero. Therefore we require
Equation:
 2
s + a 1 s + a 2 = 0.

As we know from our study of this quadratic equation, the solutions are
Equation:
a1 1
s 1,2 = − ± √ a 2 − 4a .
1 2
2 2

This means that our assumed solution works, provided s = s or s . It is a 1 2

fundamental result from the theory of differential equations that the most
general solution for x(t) is a linear combination of these assumed solutions:
Equation:
s1 t s2 t
x (t) = A 1 e + A2 e .

If a − 4a is less than zero, then the roots s and s are complex:

2
1 2 1 2

Equation:

a1 1
2
s 1,2 = − ± j √ 4a − a .
2 1
2 2

Let's rewrite this solution as

Equation:
s 1,2 = σ ± jω

where σ and ω are the constants

Equation:
a1
σ = −
2

Equation:
 1
2
ω = √ 4a − a .
2 1
2

With this notation, the solution for x(t) is

Equation:
σt jωt σt −jtωt
x (t) = A 1 e e + A2 e e .

If this solution is to be real, then the two terms on the right-hand side must
*
be complex conjugates. This means that A = A and the solution for x(t)
2 1

is
Equation:
*
σt jωt σt −J ωt
x (t) = A1 e e + A e e
1

σt jωt
= 2 Re{A 1 e e }.

The constant A may be written as A

1 1 = |A|e
jφ
. Then the solution for x(t)
is
Equation:
σt
x (t) = 2|A|e cos (ωt + φ).

This “damped cosinusoidal solution” is illustrated in [link].

 The Solution to a Second-Order Differential Equation

Exercise:
Problem:

Find the general solutions to the following differential equations:

2
d d
a. 2
X (t) + 2 x (t) + 2 = 0 ;
dt dt
2
d d
b. 2
x (t) + 2 x (t) − 2 = 0 ;
dt dt
2
d
c. 2
x (t) + 2 = 0 .
dt

Difference Equations. In your study of digital filters you will encounter

homogeneous difference equations of the form
Equation:
 x n + a 1 x n−1 + a 2 x n−2 = 0.

What this means is that the sequence{x n} obeys a homogeneous recursion:

Equation:
x n = −a 1 x n−1 − a 2 x n−2 .

A plausible guess at a solution is the geometric sequence x n = z

n
. With
this guess, the difference equation produces the result
Equation:
n n−1 n−2
z + a1 z + a2 z = 0

−1 −2 n
(1 + a 1 z + a2 z )z = 0.

If this guess is to work, then the second-order polynomial on the left-hand

side must equal zero:
−1 −2
1 + a1 z + a2 z = 0

Equation:
−1 −2
1 + a1 z + a2 z = 0

2
z + a 1 z + a 2 = 0.

The solutions are

Equation:

a1 1 2
z 1,2 = − ± j √ 4a 2 − a
2 2 1

jθ
= re .

The general solution to the difference equation is a linear combination of

the assumed solutions:
Equation:
*
 * n
n
xn = A1 z + A 2 (z 1 )
1

* * n
n
= A1 z + A 1 (z 1 )
1

n
= 2 Re{A 1 z }
1

n
= 2|A|r cos (θn + φ).

This general solution is illustrated in [link].

The Solution to a Second-Order Difference Equation

Exercise:
Problem:

Find the general solutions to the following difference equations:

a. x n + 2x n−1 + 2 = 0 ;
b. x n − 2x n−1 + 2 = 0;

c. x n
+ 2x n−2 = 0.
The Functions e^x and e^jθ: Numerical Experiment (Approximating e^jθ)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

We have demonstrated that the function e has two representations: jθ

n
jθ
1. e jθ
= lim (1 + ) ; and
n→∞ n
n k
(jθ)
2. e jθ
= lim ∑ .
n→∞
k!
k=0

In this experiment, you will write a MATLAB program to evaluate the two
functions f and S for twenty values of n:
n n

n
jθ
1. f n = (1 +
n
) , n = 1, 2, ..., 20 ; and
k

2. S .
n (jθ)
n
= ∑ , n = 1, 2, ..., 20k
k=0 k!

Choose θ = π/4(= pi/4). Use an implicit for loop to draw and plot a
circle of radius 1. Then use an implicit for loop to compute and plot f and
n

an explicit for loop to compute and plot S for n = 1, 2, ..., 100. You
n

should observe plots like those illustrated in [link]. Interpret them.

Plots for Convergence;(a) of fn, and (b) of Sn
Phasors: Introduction

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Notes to Teachers and Students:

Phasors! For those who understand them, they are of incomparable value
for the study of elementary and advanced topics. For those who
misunderstand them, they are a constant source of confusion and are of no
apparent use. So, let's understand them.

The conceptual leap from the complex number e to the phasor e

jθ j(ωt+θ)

comes in "Phasor Representation of Signals". Take as long as necessary to

understand every geometrical and algebraic nuance. Write the MATLAB
program in "Exercise 6" to fix the key ideas once and for all. Then use
phasors to study beating between tones, multiphase power, and Lissajous
figures in "Beating between Tones" through "Lissajous Figures". We
usually conduct a classroom demonstration of beating between tones using
two phase-locked sources, an oscilloscope, and a speaker. We also
demonstrate Lissajous figures with this hardware.

"Sinusoidal Steady State and the Series RLC Circuit" and Light Scattering
by a Slit" on sinusoidal steady state and light scattering are too demanding
for freshmen but are right on target for sophomores. These sections may be
covered in a sophomore course (or a supplement to a sophomore course) or
skipped in a freshman course without consequence.

In the numerical experiment in "Numerical Experiment (Interference

Patterns)", students compute and plot interference patterns for two sinusoids
that are out of phase.
Introduction
In engineering and applied science, three test signals form the basis for our
study of electrical and mechanical systems. The impulse is an idealized
signal that models very short excitations (like current pulses, hammer
blows, pile drives, and light flashes). The step is an idealized signal that
models excitations that are switched on and stay on (like current in a relay
that closes or a transistor that switches). The sinusoid is an idealized signal
that models excitations that oscillate with a regular frequency (like AC
power, AM radio, pure musical tones, and harmonic vibrations). All three
signals are used in the laboratory to design and analyze electrical and
mechanical circuits, control systems, radio antennas, and the like. The
sinusoidal signal is particularly important because it may be used to
determine the frequency selectivity of a circuit (like a superheterodyne
radio receiver) to excitations of different frequencies. For this reason, every
manufacturer of electronics test equipment builds sinusoidal oscillators that
may be swept through many octaves of Orequency. (Hewlett-Packard was
started in 1940 with the famous HP audio oscillator.)

In this chapter we use what we have learned about complex numbers and
the function e to develop a phasor calculus for representing and ma-
jθ

nipulating sinusoids. This calculus operates very much like the calculus we
developed in "Complex Numbers" and "The Functions ex and ejθ" for
manipulating complex numbers. We ap- ply our calculus to the study of
beating phenomena, multiphase power, series RLC circuits, and light
scattering by a slit.
Phasors: Phasor Representation of Signals

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

There are two key ideas behind the phasor representation of a signal:

i. a real, time-varying signal may be represented by a complex, time-

varying signal; and
ii. a complex, time-varying signal may be represented as the product of a
complex number that is independent of time and a complex signal that
is dependent on time.

Let's be concrete. The signal

Equation:

x(t) = A cos (ωt + φ),

illustrated in [link], is a cosinusoidal signal with amplitude A, frequency ω,

and phase φ. The amplitude A characterizes the peak-to-peak swing of 2A,
2π
the angular frequency ω characterizes the period T = between
ω
negative- to-positive zero crossings (or positive peaks or negative peaks),
−φ
and the phase φ characterizes the time τ = when the signal reaches its
ω
first peak. With τ so defined, the signal x(t) may also be written as
Equation:

x(t) = A cos ω(t − τ ).

 A Cosinusoidal Signal

When τ is positive, then τ is a “time delay” that describes the time (greater
than zero) when the first peak is achieved. When τ is negative, then τ is a
“time advance” that describes the time (less than zero) when the last peak
2π
was achieved. With the substitution ω = we obtain a third way of
T
writing x(t):
Equation:

2π
x(t) = A cos (t − τ ).
T

In this form the signal is easy to plot. Simply draw a cosinusoidal wave
with amplitude A and period T; then strike the origin (t = 0) so that the
signal reaches its peak at τ. In summary, the parameters that determine a
cosinusoidal signal have the following units:

A, arbitrary (e.g., volts or meters/sec, depending upon the application)

ω, in radians/sec (rad/sec)

T, in seconds (sec)

φ, in radians (rad)

τ, in seconds (sec)
Exercise:
Problem:

2π
Show that x (t) = A cos (t − τ ) is “periodic with period T ,"
T
meaning that x(t + mT ) = x(t) for all integer m.
Exercise:
Problem:

The inverse of the period T is called the “temporal frequency” of the

1
cosinusoidal signal and is given the symbol f ; the units of f = are
T

(seconds)
−1
or hertz (Hz). Write x(t) in terms of f . How is f related
to ω? Explain why f gives the number of cycles of x(t) per second.
Exercise:
Problem:
π
Sketch the function x (t) = 110 cos [2π (60)t − ] versus t. Repeat
8
π
for x (t) = 5 cos [2π (16 × 10 6
)t + ] and
4
−3
2π 10
x (t) = 2 cos [ (t − )] . For each function, determine
−3
10 8

A, ω, T , f , φ , and τ . Label your sketches carefully.

The signal x(t) = A cos (ωt + φ) can be represented as the real part of a
complex number:
Equation:
 j(ωt+φ)
x (t) = Re [Ae ]

jφ jωt
= Re[Ae e ].

We call Ae jφ
e
jωt
the complex representation of x(t) and write
Equation:
jφ jωt
x (t) ↔ Ae e ,

meaning that the signal x(t) may be reconstructed by taking the real part of
Ae
jφ
e . In this representation, we call Ae the phasor or complex
jωt jφ

amplitude representation of x(t) and write

Equation:
jφ
x (t) ↔ Ae ,

meaning that the signal x(t) may be reconstructed from Ae by jφ

multiplying with e and taking the real part. In communication theory, we

jωt

call Ae the baseband representation of the signal x(t).

jφ

Exercise:
Problem:

For each of the signals in Problem 3.3, give the corresponding phasor
representation Ae . jφ

Geometric Interpretation. Let's call

Equation:
jφ jωt
Ae e

the complex representation of the real signal A cos (ωt + φ). At t = 0, the
complex representation produces the phasor
Equation:
 jφ
Ae .

−π
This phasor is illustrated in [link]. In the figure, φ is approximately If
10
we let t increase to time t , then the complex representation produces the
1

phasor

Rotating Phasor

We know from our study of complex numbers that e just rotates thejωt 1

phasor Ae through an angle of ωt ! See [link]. Therefore, as we run t

jφ
1

from 0, indefinitely, we rotate the phasor Ae indefinitely, turning out the

jφ

2π
circular trajectory of [link]. When t = then e jωt
= e
j2π
= 1 .
ω
2π
Therefore, every ( ) seconds, the phasor revisits any given position on
ω

the circle of radius A. We sometimes call Ae jφ

e
jωt
a rotating phasor whose
rotation rate is the frequency ω:
Equation:

d
 d
ωt = ω.
dt

This rotation rate is also the frequency of the cosinusoidal signal

A cos (ωt + φ).

In summary, Ae e is the complex, or rotating phasor, representation of

jφ jωt

the signal A cos (ωt + φ). In this representation, e rotates the phasor
jωt

Ae
jφ
through angles ωt at the rate ω. The real part of the complex
representation is the desired signal Acos (ωt + φ). This real part is read off
the rotating phasor diagram as illustrated in [link]. In the figure, the angle φ
2π
is about − . As we become more facile with phasor representations, we
10

will write x (t) = Re [Xe ] and call Xe the complex representation

jωt jωt

and X the phasor representation. The phasor X is, of course, just the phasor
Ae
jφ
.
 Reading a Real Signal from a Complex,
Rotating Phasor

Exercise:
Problem:

Sketch the imaginary part of Ae e to show that this is

jφ jωt

A sin (ωt + φ). What do we mean when we say that the real and

imaginary parts of Ae e are "90 out of phase"?

jφ jωt ∘

Exercise:
 Problem:

(MATLAB) Modify Demo 2.1 in "The Function ex and ejθ" so that

θ = ωt, with ω an input frequency variable and t a time variable that

2π 2π 2π
ranges from −2( )to + 2( ) in steps of 0.02 ( ) . In your
ω ω ω

modified program, compute and plot e jωt

, Re [e
jωt
] , and Im[e jωt
] for
2π 2π 2π
−2 ( ) ≤ t < 2( ) in steps of 0.02 ( ) . Plot e jωt
in a
ω ω ω

two-dimensional plot to get a picture like [link] and plot Re[e ] and jωt

] versus t to get signals like those of [link]. You should

jωt
Im[e

observe something like [link] using the subplot features discussed in

An Introduction to MATLAB. (In the figure, w represents Greek ω.)

The Functions e jωt

, Re[e jωt
, and I m[e
]
jωt
]

Positive and Negative Frequencies. There is an alternative phasor

representation for the signal x(t) = A cos (ωt + φ). We obtain it by using
 1
the Euler formula of "The Function ex sjθ", namely, cos θ = (e
jθ
+ e
−jθ
)
2
. When this formula is applied to x(t), we obtain the result
Equation:
A j(ωt+φ) −j(ωt+φ)
x (t) = [e + e ]
2

A jφ jωt A −jφ −jωt

= e e + e e .
2 2

A
In this formula, the term e
jφ
e
jωt
is a rotating phasor that begins at the
2
phasor value Ae jφ
(for t = 0) and rotates counterclockwise with frequency
A
ω. The term e
−jφ
e
−jωt
is a rotating phasor that begins at the (complex
2
A
conjugate) phasor value e
−jφ
(for t = 0) and rotates clockwise with
2
(negative) ffequency ω. The physically meaningful frequency for a cosine
is ω, a positive number like 2π(60) for 60 Hz power. There is no such thing
as a negative frequency. The so-called negative frequency of the term
A
e
−jφ
e
−jωt
just indicates that the direction of rotation for the rotating
2
phasor is clock-wise and not counterclockwise. The notion of a negative
frequency is just an artifact of the two-phasor representation of A
cos (ωt + φ). In the one-phasor representation, when we take the “real

part,” the artifact does not arise. In your study of circuits, systems theory,
electromagnetics, solid-state devices, signal processing, control, and
communications, you will encounter both the one- and two-phasor
representations. Become facile with them.
Exercise:
Problem:

Sketch the two-phasor representation of A cos (ωt + φ). Show clearly

how this representation works by discussing the counterclockwise
rotation of the positive frequency part and the clockwise rotation of the
negative frequency part.
Adding Phasors. The sum of two signals with common frequencies but
different amplitudes and phases is
Equation:

A 1 cos (ωt + φ 1 ) + A 2 cos (ωt + φ 2 ).

The rotating phasor representation for this sum is

Equation:
jφ 1 jφ 2 jωt
(A 1 e + A2 e )e .

The new phasor is A 1e

jφ 1
+ A2 e
jφ 2
, and the corresponding real signal is
]. The new phasor is illustrated in
jφ 1 jφ 2 jωt
x (t) = Re[(A 1 e + A2 e )e

[link].

Adding Phasors

Exercise:
 Problem:

Write the phasor A e + A e as A e ; determine A and φ in

1
jφ 1
2
jφ 2
3
jφ 3
3 3

terms of A , A , φ , and φ . What is the corresponding real signal?

1 2 1 2

Differentiating and Integrating Phasors. The derivative of the signal

A cos (ωt + φ) is the signal

Equation:
d
A cos (ωt + φ) = −ωA sin (ωt + φ)
dt

jφ jωt
= − Im[ωAe e ]

jφ jωt
= Re[jωAe e ]

jπ/2 jφ jωt
= Re[ωe Ae e ].

This finding is very important. It says that the derivative of A cos (ωt + φ)
has the phasor representation
Equation:
d jφ
A cos (ωt + φ) ↔ jωAe
dt

jπ/2 jφ
↔ ωe Ae .

These two phasor representations are entirely equivalent. The first says that
the phasor Ae is complex scaled by jω to produce the phasor for
jφ

d
A cos (ωt+φ) , and the second says that it is scaled by ω and phased by
dt
+π/2 . The phasor representations of A cos (ωt + φ) and
d
A cos (ωt + φ) are illustrated in Figure 3.6. Note that the derivative
dt
“leads by π/2 radians (90)."
 Differentiating and Integrating Phasors

The integral of A cos (ωt + φ) is

Equation:
A
∫ A cos (ωt + φ)dt = sin (ωt + φ)
ω

A jφ jωt
= Im[ e e ]
ω

A jφ jωt
= Re[−j e e ]
ω

A jφ jωt
= Re[ e e ]
jω

1 −jπ/2 jφ jωt
= Re[ e Ae e ].
ω

This finding shows that the integral of A cos (ωt + φ) has the phasor
representation
Equation:
 1 jφ
∫ A cos (ωt + φ)dt ↔ Ae
jω

1 −jπ/2 jφ
↔ e Ae .
ω

1 1
The phasor Ae jφ
is complex scaled by or scaled by and phased by
jω ω

e
−jπ/2
to produce the phasor for ∫ A cos (ωt + φ)dt . This is illustrated
in [link]. Note that the integral “lags by π/2 radians (90). Keep these
geometrical pictures of leading and lagging by π/2 in your mind at all
times as you continue your more advanced study of engineering.

An Aside: The Harmonic Oscillator. The signal A cos (ωt + φ) stands

on its own as an interesting signal. But the fact that it reproduces itself (with
scaling and phasing) under differentiation means that it obeys the second-
order differential equation of the simple harmonic oscillator.[footnote] That
is, the differential equation
This means, also, that we have an easy way to synthesize cosines with
circuits that obey the equation of a simple harmonic oscillator!
Equation:
2
d x (t)
2
+ ω x (t) = 0
2
dt

has the solution

Equation:

x(t) = A cos (ωt + φ).

Try it:
Equation:
2
d d
2
x (t) = [−Aω sin (ωt + φ)] = −ω A cos (ωt + φ).
2
dt dt
The constants A and φ are determined from the initial conditions
Equation:
2 2 π 2
x (0) = A cos φ x (0) + x ( ) = A
2ω

π x(π/2ω)
x( ) = −A sin φ − =tan φ.
2ω x(0)

Exercise:
Problem:

Show how to compute A and φ in the equation

x(t) = A cos (ωt + φ) from the initial conditions x(O) and

d
x (t)|
t=0
.
dt
Phasors: Beating between Tones

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Perhaps you have heard two slightly mistuned musical instruments play
pure tones whose frequencies are close but not equal. If so, you have sensed
a beating phenomenon wherein a pure tone seems to wax and wane. This
waxing and waning tone is, in fact, a tone whose frequency is the average
of the two mismatched frequencies, amplitude modulated by a tone whose
“beat” frequency is half the difference between the two mismatched
frequencies. The effect is illustrated in [link]. Let's see if we can derive a
mathematical model for the beating of tones.

We begin with two pure tones whose frequencies are ω + ν and ω − ν

0 0

(for example, ω = 2π × 10 rad/ sec and ν = 2π rad / sec). The

0
3

average frequency is ω0, and the difference frequency is 2ν . What you hear
is the sum of the two tones:
Equation:

x (t) = A 1 cos [(ω 0 + ν)t + φ 1 ] + A 2 cos [(ω 0 − ν)t + φ 2 ].

The first tone has amplitude A and phase φ ; the second has amplitude
1 1

A and phase φ . We will assume that the two amplitudes are equal to A.
2 2

Furthermore, whatever the phases, we may write them as

Equation:

φ1 = φ + ψ and φ2 = φ − ψ

1 1
φ = (φ 1 + φ 2 ) and ψ = (φ 1 − φ 2 ).
2 2
 Beating between Tones

Recall our trick for representing x(t) as a complex phasor:

Equation:
j[(ω 0 +ν)t+φ+ψ] j[(ω 0 −ν)t+φ−ψ]
x (t) = A Re e , +, e

j(ω 0 t+φ) j(νt+ψ) −j(νt+ψ)

= A Re e , e + e

j(ω 0 t+φ)
= 2A Re e , cos, (νt + ψ)

= 2A cos (ω 0 t + φ) cos (νt + ψ).

This is an amplitude modulated wave, wherein a low frequency signal with

beat frequency ν rad/sec modulates a high frequency signal with carrier
frequency ω rad/sec. Over short periods of time, the modulating term
0

cos (νt+ψ) remains essentially constant while the carrier term

cos (ω t + φ) turns out many cycles of its tone. For example, if t runs
0
 2π
from 0 to (about 0.1 seconds in our example), then the modulating
10ν

wave turns out just 1/10 cycle while the carrier turns out 10ν ω cycles
Δ

(about 100 in our example). Every time νt changes by 2π radians, then the
modulating term goes from a maximum (a wax) through a minimum (a
wane) and back to a maximum. This cycle takes
Equation:

2π
νt = 2π ⇔ t = seconds,
ν

which is 1 second in our example. In this 1 second the carrier turns out
1000 cycles.
Exercise:
Problem:

Find out the frequency of A above middle C on a piano. Assume two

pianos are mistuned by ±1Hz(±2πrad/sec). Find their beat
frequency ν and their carrier frequency ω . 0

Exercise:

Problem: (MATLAB) Write a MATLAB program to compute and plot

A cos [(ω 0 + ν)t + φ 1 ], A cos [(ω 0 , and their sum. Then

− ν)t + φ 2 ]

compute and plot 2A cos (ω 0 t + φ) cos (νt + ψ).

Verify that the sum equals this latter signal.

Phasors: Multiphase Power

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The electrical service to your home is a two-phase service.[footnote] This

means that two 110 volt, 60 Hz lines, plus neutral, terminate in the panel.
The lines are π radians (180) out of phase, so we can write them as
It really is, although it is said to be “single phase” because of the way it is
picked off a single phase of a primary source. You will hear more about this
in circuits and power courses.
Equation:
j[2π(60)t+φ]
x 1 (t) = 110 cos [2π (60)t + φ] = Re {110e }

j2π(60)t
= Re{X 1 e }

jφ
X 1 = 110e

j[2π(60)t+φ+π]
x 2 (t) = 110 cos [2π (60)t + φ + π] = Re {110e }

j2π(60)t
= Re{X 2 e }

j(φ+π)
X 2 = 110e .

These two voltages are illustrated as the phasors X1 and X2 in [link].

 Phasors in Two-Phase Power

You may use x (t) to drive your clock radio or your toaster and the
1

difference between x (t) and x (t) to drive your range or dryer:

1 2

Equation:

x 1 (t) − x 2 (t) = 220 cos [2π(60)t + φ].

The phasor representation of this difference is

Equation:
jφ
X 1 − X 2 = 220e .

The breakers in a breaker box span the x 1

− -neutral bus for 110 volts and
to

the x -to-x buses for 220 volts.

1 2

Exercise:

Problem: Sketch the phasor X 1 − X2 on [link].

Most industrial installations use a three-phase service consisting of the

signals x (t), x (t), and x (t):
1 2 3

Equation:
 j[ω 0 t+n(2π/3)] jn(2π/3)
x n (t) = 110 Re {e } ↔ X n = 110e , n = 1, 2, 3.

The phasors for three-phase power are illustrated in [link].

Exercise:
Problem:

Sketch the phasor X − X corresponding to x (t) − x (t) on

2 1 2 1

[link]. Compute the voltage you can get with x (t) − x (t). This
2 1

answer explains why you do not get 220 volts in three-phase circuits.
What do you get?

Three-Phase Power

Constant Power. Two- and three-phase power generalizes in an obvious

way to N-phase power. In such a scheme, the N signals x (n = n

0, 1, ..., N − 1) are

Equation:

( )
 2π
x n (t) = A cos (ωt + n)
N

j2πn/N jωt j2πn/N

= Re [Ae e ] ↔ X n = Ae .

The phasors X are Ae

n
j2π(n/N )
. The sum of all N signals is zero:
Equation:

N −1 N −1 j2πn/N jωt
∑ x n (t) = Re {A ∑ e e }
n=0 n=0

j2π
1−e jωt
= Re{A j2π/N
e }
1−e

= 0.

But what about the sum of the instantaneous powers? Define the
instantaneous power of the n signal to be
th

Equation:
2 2 2 2π
p n (t) = x (t) = A cos (ωt + n)
n N
2 2
A A 2π
= + cos (2ωt + 2 n)
2 2 N

2 2
A A j(2π/N )2n j2ωt
= + Re { e e }.
2 2

The sum of all instantaneous powers is (see [link])

Equation:
N −1 2
A
P = ∑ p n (t) = N ,
2
n=0

and this is independent of time!

Exercise:
Problem:

Carry out the computations of Equation 3.33 to prove that

instantaneous power P is constant in the N-phase power scheme.
Phasors: Lissajous Figures

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Lissajous figures are figures that are turned out on the face of an
oscilloscope when sinusoidal signals with different amplitudes and different
phases are applied to the time base (real axis) and deflection plate
(imaginary axis) of the scope. The electron beam that strikes the
phosphorous face then had position
Equation:

z (t) = A x cos (ωt + φ x ) + jA y cos (ωt + φ y ).

In this representation, A cos (ωt + φ ) is the “x-coordinate of the point,”

x x

and A cos (ωt + φ) is the “y-coordinate of the point.” As time runs from
y

0 to infinity, the point z(t) turns out a trajectory like that of [link]. The
2π
figure keeps overwriting itself because z(t) repeats itself every
ω
seconds. Do you see why?
 Lissajous Figure on Oscilloscope Screen

Exercise:
Problem:

Find the intercepts that the Lissajous figure makes with the real and
imaginary axes in [link]. At what values of time are these intercepts
made?
Exercise:
Problem:

Show that the Lissajous figure z (t) = A cos (ωt + φ )+

x x

jA cos (ωt + φ ) is just the rotating phasor Ae when

j(ωt+φ)
y y
π
,
Ax = Ay = A φx = φ , and φ y = φ + .
2

Two-Phasor Representation. We gain insight into the shape of the

Lissajous figure if we use Euler's formulas to write z(t) as follows:
Equation:
 Ax Ay
j(ωt+φ x ) −j(ωt+φ x ) j(ωt+φ y ) −j(ωt+φ y )
z (t) = e + e + j e + e
2 2

jφ x jφ y −jφ x −jφ y
Ax e +jA y e Ax e +jA y e
jωt −jωt
= e + e .
2 2

This representation is illustrated in [link]. It consists of two rotating

phasors, with respective phasors B and B : 1 2

Equation:
jωt −jωt
z (t) = B 1 e + B2 e
jφ x jφ y
Ax e +jA y e
B1 =
2

−jφ x −jφ y
Ax e +jA y e
B2 =
2

Two-Phasor Representation of a Lissajous Figure

As t increases, the phasors rotate past each other where they constructively
add to produce large excursions of z(t) from the origin, and then they rotate
to antipodal positions where they destructively add to produce near
approaches of z(t) to the origin.

In electromagnetics and optics, the representations of z(t) given in [link]

and [link] are called, respectively, linear and circular representations of
elliptical polarization. In the linear representation, the x- and y-components
of z vary along the horizontal and vertical lines. In the circular
representation, two phasors rotate in opposite directions to turn out circular
trajectories whose sum produces the same effect.
Exercise:
Problem:

(MATLAB) Write a MATLAB program to compute and plot the

Lissajous figure z(t) when A = 1/2, A = 1, φ = 0, and
x y x

φ = π/6. Discretize t appropriately and choose an appropriate range

y

of values for t.
Phasors: Sinusoidal Steady State and the Series RLC Circuit

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Phasors may be used to analyze the behavior of electrical and mechanical

systems that have reached a kind of equilibrium called sinusoidal steady
state. In the sinusoidal steady state, every voltage and current (or force and
velocity) in a system is sinusoidal with angular frequency ω. However, the
amplitudes and phases of these sinusoidal voltages and currents are all
different. For example, the voltage across a resistor might lead the voltage
π
across a capacitor by 90 ( ∘
radians) and lag the voltage across an
2
π
inductor by 90 ∘
( radians) .
2

In order to make our application of phasors to electrical systems concrete,

we consider the series RLC circuit illustrated in [link]. The arrow labeled
i(t) denotes a current that flows in response to the voltage applied,and the +

and - on the voltage source indicate that the polarity of the applied voltage
is positive on the top and negative on the bottom. Our convention is that
current flows from positive to negative, in this case clockwise in the circuit.
 Series RLC Circuit

We will assume that the voltage source is an audio oscillator that pro- duces
the voltage
Equation:

V (t) = A cos (ωt + φ).

We represent this voltage as the complex signal

Equation:
jφ jωt
V (t) ↔ Ae e

and give it the phasor representation

Equation:
jφ
V (t) ↔ V ; V = Ae .

We then describe the voltage source by the phasor V and remember that we
can always compute the actual voltage by multiplying by e and taking
jωt

the real part:

Equation:
jωt
V (t) = Re {V e }.

Exercise:

Problem: Show that Re [V e jωt

] = A cos (ωt + φ) when V = Ae
jφ
.

Circuit Laws. In your circuits classes you will study the Kirchhoff laws
that govern the low frequency behavior of circuits built from resistors (R),
inductors (L), and capacitors (C). In your study you will learn that the
voltage dropped across a resistor is related to the current that flows through
it by the equation
Equation:

V R (t) = Ri (t).

You will learn that the voltage dropped across an inductor is proportional to
the derivative of the current that flows through it, and the voltage dropped
across a capacitor is proportional to the integral of the current that flows
through it:
Equation:

di
V L (t) = L (t)
dt
1
V C (t) = ∫ i (t)dt.
C

Phasors and Complex Impedance. Now suppose that the current in the
preceding equations is sinusoidal, of the form
Equation:

i(t) = B cos (ωt + θ).

We may rewrite i(t) as

Equation:
jωt
i (t) = Re {I e }

where I is the phasor representation of i(t).

Exercise:

Problem: Find the phasor I in terms of B and θ in [link].

The voltage dropped across the resistor is

Equation:

V R (t) = Ri (t)

jωt
= R Re{I e }

jωt
= Re{RI e }.

Thus the phasor representation for V R (t) is

Equation:

V R (t) ↔ V R ; V R = RI .

We call R the impedance of the resistor because R is the scale constant that
relates the “phasor voltage VR' to the “phasor current I.”

The voltage dropped across the inductor is

Equation:

di d
jωt
V L (t) = L (t) = L Re {I e }.
dt dt

The derivative may be moved through the Re[ ] operator (see [link]) to
produce the result
Equation:
 jωt
V L (t) = L Re {jωI e }

jωt
= Re{jωLI e }.

Thus the phasor representation of V L (t)

Equation:

V L (t) ↔ V L ; V L = jωLI .

We call jωL the impedance of the inductor because jωL is the complex
scale constant that relates “phasor voltage V ' to “phasor current I .”
L

Exercise:

d
Problem: Prove that the operators and Re[] commute:
dt
Equation:

d d
jωt jωt
Re {e } = Re { e }.
dt dt

The voltage dropped across the capacitor is

Equation:

1 1 jωt
V C (t) = ∫ i (t)dt = ∫ Re {I e }dt.
C C

The integral may be moved through the Re[ ] operator to produce the result
Equation:

1 I jωt
V C (t) = Re { e }
C jω

I jωt
= Re{ e }.
jωC
Thus the phasor representation of V C (t) is
Equation:

I
V C (t) ↔ V C ; V C =
jωC

1 1
We call the impedance of the capacitor because is the complex
jωC jωC

scale constant that relates “phasor voltage V " to “phasor current I .”

C

Kirchhoff's Voltage Law. Kirchhoff's voltage law says that the voltage
dropped in the series combination of R, L, and C illustrated in [link] equals
the voltage generated by the source (this is one of two fundamental
conservation laws in circuit theory, the other being a conservation law for
current):
Equation:

V (t) = V R (t) + V L (t) + V C (t).

If we replace all of these voltages by their complex representations, we

have
Equation:
jωt jωt
Re{V e } = Re{(V R + V L + V C )e }.

An obvious solution is
Equation:

V = VR + VL + VC

1
= (R + jωL + )I
jωC

where I is the phasor representation for the current that flows in the circuit.
This solution is illustrated in [link], where the phasor voltages RI , jωLI ,
 1
and I are forced to add up to the phasor voltage V .
jωC

Phasor Addition to Satisfy Kirchhoff's Law

Exercise:

1
Problem: Redraw [link] for R = ωL = = 1 .
ωC

1
Impedance. We call the complex number R + jωL + the complex
jωC

impedance for the series RLC network because it is the complex number
that relates the phasor voltage V to the phasor current I :
Equation:

V = ZI

1
Z = R + jωL + .
jωC
The complex number Z depends on the numerical values of resistance (R),
inductance (L), and capacitance (C), but it also depends on the angular
frequency (ω) used for the sinusoidal source. This impedance may be
manipulated as follows to put it into an illuminating form:
Equation:
1
Z = R + j(ωL − )
ωC

L 1
= R + j√ (ω√ LC − ).
C
ω√ LC

The parameter ω 0 =
1

√ LC
is a parameter that you will learn to call an
"undamped natural frequency" in your more advanced circuits courses.
With it, we may write the impedance as
Equation:

ω ω0
Z = R + jω 0 L ( − ).
ω0 ω

The frequency ω

ω0
is a normalized frequency that we denote by ν . Then the
impedence, as a function of normalized frequency, is
Equation:

1
Z (ν) = R + jω 0 L (ν − ).
ν

When the normalized frequency equals one (ν = 1), then the impedance is
entirely real and Z = R. The circuit looks like it is a single resistor.
Equation:
ω0 L 2 1 2 1/2
Z (ν) = R[1 + ( ) (ν − ) ] .
R ν

ω0 L 1
−1
arg Z (ν) =tan (ν − ).
R ν
The impedance obeys the following symmetries around ν = 1 :
Equation:
* 1
Z (ν) = Z ( )
ν

1
Z(ν) = Z( )
ν

1
arg Z (ν) = −arg Z ( ).
ν

In the next paragraph we show how this impedance function influences the
current that flows in the circuit.

Resonance. The phasor representation for the current that flows the current
that flows in the series RLC circuit is
Equation:
V
I =
Z(ν)

1 −jargZ(ν)
= e V
|Z(ν)|

The function H (ν) = 1

Z(ν)
displays a "resonance phenomenon." that is,
peaks at ν
|H (ν)| = 1 and decreases to zero and ν = 0 and ν = ∞ :
Equation:

0, ν = 0

1
|H (ν)| = ν = 1
R

0, ν = ∞.

When |H (ν)| = 0, no current flows.

ω
The function |H (ν)| is plotted against the normalized frequency ν = in
ω0
1
Figure 3.14. The resonance peak occurs at ν = 1 , where |H (ν)| =
R
meaning that the circuit looks purely resistive. Resonance phenomena
underlie the frequency selectivity of all electrical and mechanical networks.

Resonance in a Series RLC Circuit

Exercise:
Problem:

(MATLAB) Write a MATLAB program to compute and plot |H (ν)|

and arg H (ν) versus ν for ν ranging from 0.1 to 10 in steps of 0.1.
L
Carry out your computations for ω 0 = 10, 1, 0. 1 , and 0.01, and
R
overplot your results.

Circle Criterion and Power Factor. Our study of the impedance Z(ν) and
1
the function H (ν) = brings insight into the resonance of an RLC
Z(ν)

circuit and illustrates the ffequency selectivity of the circuit. But there is
more that we can do to illuminate the behavior of the circuit.
Equation:

1
V = RI + j(ωL − )I .
ωC

This equation shows how voltage is divided between resistor voltage RI and
inductor-capacitor voltage j(ωL − 1
)I .
ωC
Equation:

ω ω0
V = RI + jω 0 L ( − )I
ω0 ω

or
Equation:

jω 0 L 1
V = RI + (ν − )RI .
R ν

In order to simplify our notation, we can write this equation as

Equation:

V = V R + jk (ν)V R

where V is the phasor voltage RI and k(ν) is the real variable

R

Equation:

ω0 L 1
k(ν) = (ν − ).
R ν

[link] brings very important geometrical insights. First, even though the
phasor voltage V in the RLC circuit is complex, the terms V and
R R
π
jk (ν)V R are out of phase by radians. This means that, for every
2
allowable value of V , the corresponding jk (ν)V must add in a right
R R

triangle to produce the source voltage V . This is illustrated in [link]. As the

frequency ν changes, then k(ν) changes, producing other values of V and R

jk (ν)V R
that sum to V . Several such solutions for V and jk (ν)V are
R R

illustrated in Figure 3.15(b). From the figure we gain the clear impression
V V
that the phasor voltage V >R lies on a circle of radius centered at
2 2
Let's try this solution,
Equation:
V V jψ
VR = + e
2 2

V jψ
= (1 + e ),
2

and explore its consequences. When this solution is substituted into [link],
the result is
Equation:
V V
jψ jψ
V = (1 + e ) + jk (ν) (1 + e )
2 2

or
Equation:
jψ
2 = (1 + e ) [1 + jk (ν)].

The Components of V ; (a) Addition of VR and jk (ν)V R

to Produce V, and (b) Several Values of VR and jk (ν)V R

that Produce V

If we multiply the left-hand side by its complex conjugate and the right-
hand side by its complex conjugate, we obtain the identity
Equation:
 2
4 = 2 (1+ cos ψ)[1 + k (ν)].

This equation tells us how the angle ψ depends on k(ν) and, conversely,
how k(ν) depends on ψ:
Equation:
2
1 − k (ν)
cos ψ =
2
1 + k (ν)

Equation:

1− cos ψ
2
k (ν) =
1+ cos ψ

The number cos ψ lies between −1 and +1, so a circular solution does
indeed work.
Exercise:
Problem:

Check −1 ≤cos ψ ≤ 1 for −∞ < k < ∞ and −∞ < k < ∞ for

−π ≤ ψ ≤ π. Sketch k versus ψ and ψ versus k.

V
The equation V R = (1 + e
jψ
) is illustrated in [link]. The angle that V R
2
makes with V is determined from the equation
Equation:
ψ
2φ + π − ψ = π ⇒ φ =
2
 The Voltages V and VR, and the Power Factor
cos φ

In the study of power systems, cos φ is a "power factor" that determines

how much power is delivered to the resistor. We may denote the power
factor as
Equation:
ψ
η =cos φ =cos .
2

But cos ψ may be written as

Equation:
ψ
η =cos φ =cos
2

But cos ψ may be written as

Equation:
 2 2
cos ψ =cos (φ + φ) = cos φ − − sin φ

2 2
= cos φ − (1− cos φ)

2
= 2 cos φ − 1

2
= 2η − 1.

Therefore the square of the power factor η is

Equation:

cos ψ + 1 1
2
η = =
2
2 1 + k (ν)

The power factor is a maximum of 1 for k(ν) = 0, corresponding to ν = 1

(ω = ω ). It is a minimum of 0 for k(ν) = ±∞, corresponding to

0

ν = 0, ∞(ω = 0, ∞).

Exercise:
Problem:

ω0 L 1
With k defined as k (ν) = (ν − ) , plot k 2
,
(ν) cos ψ , and η 2

R ν

versus ν .
Exercise:

Problem: Find the value of ν that makes the power factor η = 0. 707.
Phasors: Light Scattering by a Slit

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

One of the most spectacular successes for phasor analysis arises in the study
of light diffraction by a narrow slit. The experiment is to shine laser light
through a slit in an otherwise opaque sheet and observe the pattern of light
that falls on a distant screen. When the slit is very narrow compared with
the wavelength of the light, then the light that falls on the screen is nearly
uniform in intensity. However, when the width of the slit is comparable to
the wavelength of the light, the pattern of light that falls on the screen is
scalloped in intensity, showing alternating light and dark bands. The
experiment, and the observed results, are illustrated in [link].
 Light Diffraction by a Slit

Why should this experiment produce this result? Phasor analysis illuminates
the question and produces an elegant mathematical description of a
profoundly important optical experiment.

Huygens's Principle. We will assume, as Christiaan Huygens did, that the

light incident on the slit sets up a light field in the slit that may be modeled
by N discrete sources, each of which radiates a “spherical wave of light.”
This model is illustrated in [link]. The distance between sources is d, and
N d = L is the width of the slit. Each source is indexed by n, and n runs

from 0 to N − 1. The 0 source is located at the origin of our coordinate

th

system.

The spherical wave radiated by the n th

source is described by the equation
Equation:

A
j[ωt−(2π/λ)r n ]
E (r n , t) = Re { e }.
N

The function E(r , t) describes the “electric field” at time t and distance rn
n

from the n source. The field is constant as long as the variable

th

ωt − (2π/λ)r is constant. Therefore, if we freeze time at t = t , the field

n 0

will be constant on a sphere of radius r . This is illustrated in [link].

n
 Huygens's Model for Light Diffraction

Exercise:
Problem:

Fix r in E(r , t) and show that E(r , t) varies cosinusoidally with

n n n

time t. Sketch the function and interpret it. What is its period?
Exercise:
Problem:

Fix t in E(r , t) and show that E(r , t) varies cosinusoidally with

n n

radius r . Sketch the function and interpret it. Call the “period in r "
n n

the “wavelength.” Show that the wavelength is λ.

Exercise:
Problem:

The “crest of the wave E(r , t)" occurs when ωt − (2π/λ)r = 0.

n n

Show that the crest moves through space at velocity v = ωλ/2π.

Geometry. If we now pick a point P on a distant screen, that point will be

at distance r from source 0, ..., r from source n, ..., and so on. If we
0 n
isolate the sources 0 and n, then we have the geometric picture of [link].
The angle φ is the angle that point P makes with the horizontal axis. The
Pythagorean theorem says that the connection between distances r and r 0 n

is
Equation:
2 2 2
(r n − nd sin φ) + (nd cos φ) = r0 .

Geometry of the Experiment

Let's try the solution

Equation:
r n = r 0 + nd sin φ.

This solution produces the approximate identity

Equation:
2 2 2
r 0 + (nd cos φ) ≅ r0
Equation:
2 2 2
r + (nd cos φ) ≅r
0 0

2
nd
1 + ( cos φ) ≅ 1.
r0

nd
This will be close for << 1 . We will assume that the slit width L is
r0

small compared to the distance to any point on the screen. Then

Nd L
= << 1 , in which case the approximate solution for r is valid n
r0 r0

for all n. This means that, for any point P on the distant screen, the light
contributed by the n source is approximately
th

Equation:
A j[ωt−(2π/λ)(r 0 +ndsinφ)]
E n (φ, t) = Re { e }
N

A −j(2π/λ)r 0 −j(2π/λ)ndsinφ jωt

= Re{ e e e }.
N

The phasor representation for this function is just

Equation:

A −j(2π/λ)r 0 −j(2π/λ)ndsinφ
E n (φ) = e e .
N

Note that E 0 (φ) , the phasor associated with the 0 th

source, is
A
e
−j(2π/λ)r 0
. Therefore we may write the phasor representation for the
N
light contributed by the n th
source to be
Equation:
−j(2π/λ)ndsinφ
E n (φ) = E 0 (φ)e .
This result is very important because it shows the light arriving at point P
from different sources to be “out of phase” by an amount that depends on
nd sin φ
the ratio
λ

Phasors and Interference. The phasor representation for the field observed
at point P on the screen is the sum of the phasors contributed by each
source:
Equation:
N −1 N −1

−j(2π/λ)ndsinφ
E (φ) = ∑ E n (φ) = E 0 (φ) ∑ e .

n=0 n=0

This is a sum of the form

Equation:
N −1

jnθ
E (φ) = E 0 (φ) ∑ e

n=0

where the angle θ is (2π/λ)d sin φ. This sum is illustrated in [link] for
several representative values of θ. Note that for small θ, meaning small φ,
the sum has large magnitude, whereas for θ on the order of 2π/N , the sum
is small. This simple geometric interpretation shows that for some values of
φ, corresponding to some points P on the screen, there will be constructive

interference between the phasors, while for other values of φ there will be
destructive interference. Constructive interference produces bright light,
and destructive interference produces darkness.
 Phasor Sums for Diffraction

The geometry of [link] is illuminating. However, we already know from our

study of complex numbers and geometric sums that the phasor sum of
Equation 3.85 may be written as
Equation:
−j(2π/λ)N dsinφ
1 − e
E (φ) = E 0 (φ) .
−j(2π/λ)dsinφ
1 − e

This result may be manipulated to produce the form

Equation:
d
A sin (N π sin φ)
−j(2π/λ)r 0 −j(π/λ)(N −1)dsinφ λ
E (φ) = e e
d
N sin (π sin φ)
λ

The magnitude is the intensity of the light at angle φ from horizontal:

Equation:
d
 d
A sin (N π sin φ)
λ
|E (φ)| =| |.
d
N sin (π sin φ)
λ

Exercise:

Problem: Derive [link] from [link].

Limiting Form. Huygens's model is exact when d shrinks to 0 and N

increases to infinity in such a way that N d → L, the slit width. Then
Equation:
πL
A sin ( sin φ)
λ
|E (φ)| →| |.
πL
sin φ
λ

L
This function is plotted in [link] for two values of the width of the slit
λ
L λ
measured in wavelengths. When << 1 (i.e., >> 1 ), then the light is
λ L
L λ
uniformly distributed on the screen. However, when > 1( < 1) ,
λ L

then the function has many zeros for | sin φ| < 1, as illustrated in the
figure. These zeros correspond to dark spots on the screen where the fields
radiated from the infinity of points within the slit interfere destructively.
 Interference Pattern for a Slit

Exercise:

Problem: Derive [link] from [link].

Exercise:

Problem: (MATLAB) Plot the discrete approximation

Equation:
A d
sin (N π sin φ)
N λ

d
sin (π sin φ)
λ

L Nd
versus sin φ for = = 10 and N = 2, 4, 8, 16, 32 . Compare
λ λ
with the continuous, limiting form
Equation:
∣
A sin (π

π
L

λ
L

λ
sin φ)

sin φ
Phasors: Numerical Experiment (Interference Patterns)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Let's add two sinusoids whose amplitudes and frequencies are identical and
whose phases are different:
Equation:

x(t) = A cos (ωt + φ) + A cos (ωt + φ + ψ).

Show analytically that this sum has the phasor representation

Equation:

ψ j[φ+(ψ/2)]
X = 2A cos ( )e .
2

Interpret this finding. Then write a MATLAB program that computes and
plots complex X on the complex plane as ψ varies from 0 to 2π and that
plots magnitude, |X|, and phase, arg X, versus the phase angle ψ. (You
2π
will have to choose ψ = n , n = 0, 1, ..., N − 1 , for a suitable N .)
N
When do you get constructive interfelence and when do you get destructive
interference? Now compute and plot x(t) versus t (you will need to
discretize t) for several interesting values of ψ. Explain your interference
results in terms of the amplitude and phase of x(t) and the magnitude and
phase of X. Use the subplots discussed in "An Introduction to MATLAB"
to plot all of your results together.
Linear Algebra: Introduction

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

We use this chapter to introduce students to the algebraic structure of

vectors and matrices and to introduce them to matrix computations. These
matrix computations are used in the chapters "Vector Graphics", "Filtering",
and "Binary Codes" to solve problems in vector graphics, filtering, and
binary coding.

Vectors are introduced in the section "Vectors", along with algebraic and
geometric interpretations of some fundamental vector operations and
properties. The section "Inner Product and Euclidean Norm", the section
"Direction Cosines", and the section "Projections" introduce inner products
and their applications, including norm, direction cosines, orthogonality, and
projections. Some important alternatives to the Euclidean norm are
introduced in the section "Other Norms". Matrices are motivated and
introduced in this section. The notation in these sections can be daunting to
the beginner, so we proceed very carefully, using example after example. In
the section "Solving Linear Systems of Equations" we codify the
elimination procedures that students have used in high school to solve linear
systems of equations. The MATLAB demonstration in Demo 2 shows how
to use MATLAB to solve linear equations. The section "Circuit Analysis"
shows how linear algebra and MATLAB can be used to analyze dc circuits.
The numerical experiment "Circuit Design" gives students practice in
building function files in MATLAB and shows how to solve a sequence of
linear equations in order to design a circuit with desired properties.

Occasionally we have placed important results in the problems. We feel that

students should not miss the material in Exercise 3 in "Vectors", Exercise 3
in "Inner Product and Euclidean Norm", Exercise 3 in "Projections",
Exercise 1 in "Matrices", and Exercise 4 in "Matrices".

Introduction
Linear algebra is a branch of mathematics that is used by engineers and
applied scientists to design and analyze complex systems. Civil engineers
use linear algebra to design and analyze load-bearing structures such as
bridges. Mechanical engineers use linear algebra to design and analyze
suspension systems, and electrical engineers use it to design and analyze
electrical circuits. Electrical, biomedical, and aerospace engineers use linear
algebra to enhance X rays, tomographs, and images from space. In this
chapter and the next we study two common problems from electrical
engineering and use linear algebra to solve them. The two problems are (i)
electrical circuit analysis and (ii) coordinate transformations for computer
graphics. The first of these applications requires us to understand the
solution of linear systems of equations, and the second requires us to
understand the representation of mathematical operators with matrices.

Much of linear algebra is concerned with systematic techniques for

organizing and solving simultaneous linear equations by elimination and
substitution. The following example illustrates the basic ideas that we
intend to develop.

Example:
A woman steps onto a moving sidewalk at a large airport and stands while
the moving sidewalk moves her forward at 1.2 meters/seconds. At the
same time, a man begins walking against the motion of the sidewalk from
the opposite end at 1.5 meters/second (relative to the sidewalk). If the
moving sidewalk is 85 meters long, how far does each person travel
(relative to the ground) before they pass each other?
To solve this problem, we first assign a variable to each unknown quantity.
Let x be the distance traveled by the woman, and let x be the distance
1 2

traveled by the man. The sum of the two distances is 85 meters, giving us
one equation:
Equation:

x 1 + x 2 = 85.

Our second equation is based on the time required before they pass. Time
equals distance divided by rate, and the time is the same for both people:
Equation:
x1 x2
= ⇒ 0. 3x 1 − 1. 2x 2 = 0.
1. 2 1. 5 − 1. 2

1. 2
We may substitute [link] into [link] to obtain the result x 2 + x 2 = 85 ,
0. 3
or
Equation:

5x 2 = 85 ⇒ x 2 = 17.

Combining the result from [link] with that of [link], we find that
Equation:

x 1 = 68.

So the man travels 17 meters, and the woman travels 68 meters.

[link] and [link] are the key equations of [link]. They may be organized
into the “matrix equation”
Equation:

1 1 x1 85
[ ][ ] = [ ].
0. 3 −1. 2 x2 0

The rules for matrix-vector multiplication are evidently

Equation:

(1) x 1 + (1)x 2 = 85

Equation:

(0. 3)x 1 + (−1. 2)x 2 = 0.

[link] and [link] may be organized into the matrix equation
Equation:

0 5 x1 85
[ ][ ] = [ ].
0. 3 −1. 2 x2 0

This equation represents one partially solved form of [link], wherein we

have used the so-called Gauss elimination procedure to introduce a zero
into the matrix equation in order to isolate one variable. The MATLAB
software contains built-in procedures to implement Gauss elimination on
much larger matrices. Thus MATLAB may be used to solve large systems
of linear equations.
Before we can apply linear algebra to more interesting physical problems,
we need to introduce the mathematical tools we will use.
Linear Algebra: Vectors

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

For our purposes, a vector is a collection of real numbers in a one-

dimensional array.[footnote] We usually think of the array as being
arranged in a column and write
In a formal development of linear algebra, the abstract concept of a vector
space plays a fundamental role. We will leave such concepts to a complete
course in linear algebra and introduce only the functional techniques
necessary to solve the problems at hand.

x1

x2

x = x3 .
|

xn

Notice that we indicate a vector with boldface and the constituent elements
with subscripts. A real number by itself is called a scalar, in distinction
from a vector or a matrix. We say that x is an n-vector, meaning that x has
n elements. To indicate that x is a real number, we write
1

Equation:

x1 ∈ ,

meaning that x is contained in , the set of real numbers. To indicate that

1

x is a vector of n real numbers, we write

Equation:
n
x ∈ ,

meaning that x is contained in R , the set of real n-tuples. Geometrically,

n

R is n-dimensional space, and the notation x ∈ means that x is a point

n n

in that space, specified by the n coordinates x , x , ..., x . [link] shows a

1 2 n

vector in R , drawn as an arrow from the origin to the point x. Our

3

geometric intuition begins to fail above three dimensions, but the linear
algebra is completely general.

A Vector in R3

We sometimes find it useful to sketch vectors with more than three

dimensions in the same way as the three-dimensional vector of [link]. We
then consider each axis to represent more than one dimension, a hyperplane,
in our n-dimensional space. We cannot show all the details of what is
happening in n-space on a three-dimensional figure, but we can often show
important features and gain geometrical insight.
Vector Addition. Vectors with the same number of elements can be added
and subtracted in a very natural way:
Equation:

x1 + y1 x1− y1

x2 + y2 x2 − y2

x + y = x3 + y3 ;x − y = x3 − y3 .

| |

xn + yn xn − yn

Example:
1 0

The difference between the vector x = 1 and the vector y = 0 is the

1 1

vector z = x − y = 1 . These vectors are illustrated in [link]. You can

0

see that this result is consistent with the definition of vector subtraction in
[link]. You can also picture the subtraction in [link] by mentally reversing
the direction of vector y to get −y and then adding it to x by sliding it to
the position where its tail coincides with the head of vector x. (The head is
the end with the arrow.) When you slide a vector to a new position for
adding to another vector, you must not change its length or direction.
 Subtraction of Vectors

Exercise:
Problem:

Compute and plot x + y and x − y for each of the following cases:

1 1

a. x = 3 , y = 2 ;
2 3

−1 1

b. x = 3 , y = 2 ;
−2 3

1 1

c. x = −3 , y = 3 .
2 2
Scalar Product. Several different kinds of vector multiplication are
defined.[footnote] We begin with the scalar product. Scalar multiplication
is defined for scalar a and vector x as
The division of two vectors is undefined, although three different
“divisions” are defined in MATLAB.
Equation:

ax1

ax2

ax = ax3 .

axn

If |a| < 1, then the vector ax is "shorter" than the vector x; if |a| > 1, then
the vector ax is ‚"longer" than x. This is illustrated for a 2-vector in [link].

The Scalar Product ax

Exercise:
 Problem:

Compute and plot the scalar product ax when x = 1/2 for each of
l/4

the following scalars:

a. a = 1;
b. a = −1;
c. a = −1/4;
d. a = 2.

Exercise:
Problem:

Given vectors , , ∈
n
and the scalar a ∈ , prove the following
identities:

a. x + y = y + x. Is vector addition commutative?

b. (x + y) + z = x + (y + z). Is vector addition associative?
c. a(x + y) = ax + ay. Is scalar multiplication distributive over
vector addition?
Linear Algebra: Inner Product and Euclidean Norm

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The inner product (x, y) between vectors x and y is a scalar consisting of

the following sum of products:
Equation:

(x, y) = x 1 y 1 + x 2 y 2 + x 3 y 3 + ⋯ + x n y n .

This definition seems so arbitrary that we wonder what uses it could

possibly have. We will show that the inner product has three main uses:

i. computing length or “norm”,

ii. finding angles between vectors and checking for “orthogonality”, and
iii. computing the “component of one vector along another” (projection).

Since the inner product is so useful, we need to know what algebraic

operations are permitted when we are working with inner products. The
following are some properties of the inner product. Given x, y, z ∈ R and
n

a ∈ R,

i. (x, y) = (y, x);

ii. (ax, y) = a(x, y) = (x, ay); and
iii. (x, y + z) = (x, y) + (x, z).

Exercise:
 Problem:

Prove the three preceding properties by using the definition of inner

product. Is the equation x(y, z) = (x, y)z also a true property? Prove
or give a counterexample.

Euclidean Norm. Sometimes we want to measure the length of a vector,

namely, the distance from the origin to the point specified by the vector's
coordinates. A vector's length is called the norm of the vector. Recall from
Euclidean geometry that the distance between two points is the square root
of the sum of the squares of the distances in each dimension. Since we are
measuring from the origin, this implies that the norm of the vector x is
Equation:

2 2 2
∥ x ∥= √ x + x + +x n .
1 2

Notice the use of the double vertical bars to indicate the norm. An
equivalent definition of the norm, and of the norm squared, can be given in
terms of the inner product:
Equation:

∥ x ∥= √ (x, x)

or
Equation:
2
∥ x ∥ = (x, x).

Example:
The Euclidean norm of the vector
Equation:
 1

3
x = ,
5

−2

is ∥ x ∥= √1 2
+ 3
2
+ 5
2
+ (−2)
2
= √ 39.

An important property of the norm and scalar product is that, for any
x ∈ R and a ∈ R ,
n

Equation:

||ax|| = |a|||x||.

So we can take a scalar multiplier outside of the norm if we take its absolute
value.
Exercise:

Problem: Prove [link].

Cauchy-Schwarz Inequality. Inequalities can be useful engineering tools.

They can often be used to find the best possible performance of a system,
thereby telling you when to quit trying to make improvements (or proving
to your boss that it can't be done any better). The most fundamental
inequality in linear algebra is the Cauchy-Schwarz inequality. This
inequality says that the inner product between two vectors x and y is less
than or equal (in absolute value) to the norm of x times the norm of y, with
equality if and only if y = αx:
Equation:

|(x, y)| ≤ ||x||||y||.

To prove this theorem, we construct a third vector z = λx − y and measure
its norm squared:
Equation:
2 2 2 2
||λx − y|| = (λx − y, λx − y) = λ ||x|| − 2λ (x, y) + ||y|| ≥ 0.

So we have a polynomial in λ that is always greater than or equal to 0

(because every norm squared is greater than or equal to 0). Let's assume
that x and y are given and minimize this norm squared with respect to λ. To
do this, we take the derivative with respect to λ and equate it to 0:
Equation:

2
(x, y)
2λ||x|| − 2 (x, y) = 0 ⇒ λ =
2
||x||

When this solution is substituted into the formula for the norm squared in
[link], we obtain
Equation:
2

(x, y) 2
2(x, y) 2
||x|| − (x, y) + ||y|| ≥ 0,
2 2
||x|| ||x||

which simplifies to
Equation:
2
(x, y) 2 2 2 2
− + ||y|| ≥ 0 ⇒ (x, y) ≤ ||x|| ||y|| .
2
||x||

The proof of the Cauchy-Schwarz inequality is completed by taking the

positive square root on both sides of [link]. When y = αx, then
Equation:
 2 2 2 2 2
(x, y) = (x, αx) = [|α| (x, x)] = (|α|||x|| )

2 2 2
= (|α| ||x|| )||x||

2
= (αx, αx)||x||

2
= (y, y)||x||

2 2
= ||y|| ||x|| ,

which shows that equality holds in [link] when y is a scalar multiple of x.

Exercise:
Problem:

Use the Cauchy-Schwarz inequality to prove the triangle inequality,

which states
Equation:

||x + y|| ≤ ||x|| + ||y||.

Explain why this is called the triangle inequality.

Linear Algebra: Direction Cosines

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Unit Vectors. Corresponding to every vector x is a unit vector u pointing

x

in the same direction as x. The term unit vector means that the norm of the
vector is 1:
Equation:

||u x || = 1.

The question is, given x, how can we find u ? The first part of the answer
x

is that u will have to be a positive scalar multiple of x in order to point in

x

the same direction as x, as shown in [link]. Thus

Equation:
u x = αx.

The Unit Vector ux

But what is α? We need to choose α so that the norm of u will be 1:
x

Equation:

||u x || = 1

Equation:

|αx|| = 1

Equation:

|α|||x|| = 1

Equation:

1
α =
||x||

We have dropped the absolute value bars on α because ||x|| is positive. The
α that does the job is 1 over the norm of x. Now we can write formulas for

u in terms of x and x in terms of u :

x x

Equation:

1
ux = x,
||x||

Equation:

x = ||x||u x .

So the vector x is its direction vector u , scaled by its Euclidean norm.

x

Unit Coordinate Vectors. There is a special set of unit vectors called the
unit coordinate vectors. The unit coordinate vectore is a unit vector in ∇
i
n
that points in the positive direction of the i th
coordinate axis. [link] shows
the three unit coordinate vectors in ∇ . 3

Unit Coordinate Vectors in R3

For three-dimensional space,R , the unit coordinate vectors are

3

Equation:

1 0 0

e1 = 0 , e2 = 1 , e3 = 0 .

0 0 1

More generally, in n-dimensional space, there are n unit coordinate vectors

given by
You should satisfy yourself that this definition results in vectors with a
norm of l.
Exercise:

Problem: Find the norm of the vector au where u is a unit vector.

Exercise:

Problem: Find the unit vector u in the direction of

x

3
a. x = [ ] ;
4

b. x = 6 ;
−2

−1
c. x = .
1

−1
Direction Cosines. We often say that vectors “have magnitude and
direction.” This is more or less obvious from "Linear Algebra: Vectors",
where the three-dimensional vector x has length √x + x + x and2
1
2
2
2
3

points in a direction defined by the components x , x , and x . It is

1 2 3

perfectly obvious from [link] where x is written as x = ||x||u . But perhaps

x

there is another representation for a vector that places the notion of

magnitude and direction in even clearer evidence.

[link] shows an arbitrary vector x ∈ R and the three-dimensional unit

3

coordinate vectorse , e , e . The inner product between the vector x and

1 2 3

the unit vector e just reads out the k component of x:

k
th

Equation:

(x, e k ) = (e k , x) = x k .

Since this is true even in R , any vector x ∈ R has the following

n n

representation in terms of unit vectors:

Equation:

x = (x, e 1 )e 1 + (x, e 2 )e 2 + ⋯ + (x, e n )e n .

The Three-Dimensional Unit Vectors

Let us now generalize our notion of an angle θ between two vectors to R n

as follows:
Equation:

(x, y)
cos θ =
||x||||y||

The celebrated Cauchy-Schwarz inequality establishes that −1− ≤ cos

θ ≤ 1. With this definition of angle, we may define the angle θ that a k

vector makes with the unit vector e to be k

Equation:

(x, e k )
cos θ k =
||x||||e k ||

But the norm of e is 1, so

k

Equation:

(x, e k ) xk
cos θ k = =
||x|| ||x||

When this result is substituted into the representation of x in [link], we

obtain the formula
Equation:

x = ||x|| cos θ 1 e 1 + ||x|| cos θ 2 e 2 + ⋯ + ||x|| cos θ n e n

= ||x||(cos θ 1 e 1 + cos θ 2 e 2 + ⋯ + cos θ n e n ).

This formula really shows that the vector x has “magnitude” ||x|| and
direction (θ , θ , ..., θ ) and that the magnitude and direction are sufficient
1 2 n

to determine x. We call ( cos θ , cos θ , ..., cos θ ) the direction cosines for
1 2 n

the vector x. In the three-dimensional case, they are illustrated in [link].

Exercise:
Problem:

Show that [link] agrees with the usual definition of an angle in two
dimensions.
Exercise:

Problem: Find the cosine of the angle between x and y where

1 2

a. x = 0 , y = 2 ;
0 2

1 −1

−1 0
b. x = , y = ;
1 1

−1 0

2 4

c. x = 1 , y = 2

−2 −4

If we compare [link] and [link] we see that the direction vector u is x

composed of direction cosines:

Equation:

cosθ 1

cosθ 2
u x =cos θ 1 e 1 + cos θ 2 e 2 + ⋯ + cos θ n e n = .

cosθ n

With this definition we can write [link] compactly as

Here x is written as the product of its magnitude ||x|| and its direction
vector u . Now we can give an easy procedure to find a vector's direction
x

angles:

(i) find ||x||;

x
(ii) calculate u x = ; and
||x||

(iii) take the arc cosines of the elements of u . x

Step (iii) is often unnecessary; we are usually more interested in the

direction vector (unit vector) ux. Direction vectors are used in materials
science in order to study the orientation of crystal lattices and in
electromagnetic field theory to characterize the direction of propagation for
radar and microwaves. You will find them of inestimable value in your
courses on electromagnetic fields and antenna design.
Exercise:
Problem:

Sketch an arbitrary unit vector u ∈ R . Label the direction cosines

3

and the components of u. S

Exercise:
Problem:

Compute the norm and the direction cosines for the vector x = 2 .

Exercise:
Problem:

Prove that the direction cosines for any vector satisfy the equality
Equation:
2 2 2
cos θ 1 + cos θ 2 + ⋯ + cos θ n = 1.
 What does this imply about the number of scalars needed to determine
a vector x ∈ R ? n

Exercise:
Problem:

Astronomers use right ascension, declination, and distance to locate

π
stars. On [link] these are, respectively, −φ, − θ , and ||x||. 3
2
Represent x = (x , x , x ) in terms of φ, θ , and ||x|| only. (That is,
1 2 3 3

find equations that give φ, θ , and ||x|| in terms of x , x , and x3, and
3 1 2

find equations that give x , x , and x3 in terms of φ, θ , and ||x||. )

1 2 3

Exercise:
Problem:

(MATLAB) Write a MATLAB function that accepts any vector

x ∈ R
n
and computes its norm and direction cosines. Make x an input
variable, and make ||x|| and u output variables.
x

Exercise:
Problem:

Let x and y denote two vectors in R with respective direction cosines

n

( cos θ , cos θ , ..., cos θ ) and ( cos φ , cos φ , ..., cos φ ). Prove
1 2 n 1 2 n

that ψ, the angle between x and y, is

Equation:

cos ψ =cos θ 1 cos φ 1 + cos θ 2 cos φ 2 + ⋯ + cos θ n cos φn .

Specialize this result to R for insight.

2

Exercise:
Problem:

2 2

Compute the angle between the vectors x = 3 and 3 . Sketch the

0 1

result.
Linear Algebra: Projections

Note:This module is part of the collection, A First Course in Electrical and

Computer Engineering. The LaTeX source files for this collection were
created using an optical character recognition technology, and because of
this process there may be more errors than usual. Please contact us if you
discover any errors.

Orthogonality. When the angle between two vectors is π/2(90 ), we say 0

that the vectors are orthogonal. A quick look at the definition of angle
(Equation 12 from "Linear Algebra: Direction Cosines") leads to this
equivalent definition for orthogonality:
Equation:

(x, y) = 0 ⇔ x and y are orthogonal.

3 −2
For example, in Figure 1(a), the vectors x = [ ] and y = [ ] are clearly
1 6

orthogonal, and their inner product is zero:

Equation:

(x, y) = 3(−2) + 1(6) = 0.

3 −2 0

In Figure 1(b), the vectors x = 1 ,y = 6 , and z = 0 are

0 0 4

mutually orthogonal, and the inner product between each pair is zero:
Equation:

(x, y) = 3(−2) + 1(6) + 0(0) = 0

Equation:
 (x, z) = 3(0) + 1(0) + 0(4) = 0

Equation:

(y, z) = −2(0) + 6(0) + 0(4) = 0.

Orthogonality of Vectors

Exercise:

Problem: Which of the following pairs of vectors is orthogonal:

1 0

a. x = 0 , y = 1 ;
1 0

1 1

b. x = 1 , y = 1 ;
0 1

c. x = e 1, y = e3 ;
a −b
d. x = [ ] ,y=[ ]?
b a

We can use the inner product to find the projection of one vector onto
another as illustrated in Figure 2. Geometrically we find the projection of x
onto y by dropping a perpendicular from the head of x onto the line
containing y. The perpendicular is the dashed line in the figure. The point
where the perpendicular intersects y (or an extension of y) is the projection
of x onto y, or the component of x along y. Let's call it z.

Component of One Vector along

Another

Exercise:
Problem:

Draw a figure like Figure 2 showing the projection of y onto x.

The vector z lies along y, so we may write it as the product of its norm ||z||
and its direction vector u :
y

Equation:
y
z = ||z||u y = ||z|| .
||y||

But what is norm ||z||? From Figure 2 we see that the vector x is just z, plus
a vector v that is orthogonal to y:
Equation:

x = z + v, (v, y) = 0.

Therefore we may write the inner product between x and y as

Equation:
 (x, y) = (z + v, y) = (z, y) + (v, y) = (z, y).

But because z and y both lie along y, we may write the inner product (x, y)
as
Equation:

(x, y) = (z, y) = (||z||u y , ||y||u y ) = ||z||||y|| (u y , u y )

2
= ||z||||y||||u y || = ||z||||y||.

(x, y)
From this equation we may solve for ||z|| = and substitute ||z|| into
||y||

Equation 6 to write z as
Equation:
y
z = ||z||
||y||

(x,y) y (x,y)
= = y.
||y|| ||y|| (y,y)

Equation 10 is what we wanted–an expression for the projection of x onto y

in terms of x and y.
Exercise:
Problem:

Show that ||z|| and z may be written in terms of cos θ for θ as

illustrated in Figure 2:
Equation:

||z|| = ||x|| cos θ

Equation:

||x|| cos θ
z = y.
||y||
Orthogonal Decomposition. You already know how to decompose a vector
in terms of the unit coordinate vectors,
Equation:

x = (x, e 1 )e 1 + (x, e 2 )e 2 + ⋯ + (x, e n )e n .

In this equation, (x, e )e is the component of x along e , or the projection

k k k

of x onto e , but the set of unit coordinate vectors is not the only possible
k

basis for decomposing a vector. Let's consider an arbitrary pair of orthogonal

vectors x and y:
Equation:

(x, y) = 0.

The sum of x and y produces a new vector w, illustrated in Figure 3, where

we have used a two-dimensional drawing to represent n dimensions. The
norm squared of w is
2
||w|| = (w, w) = [(x + y), (x + y)] = (x, x) + (x, y) + (y, x) + (y, y)

2 2
= ||x|| + ||y|| .

This is the Pythagorean theorem in n dimensions! The length squared of w

is just the sum of the squares of the lengths of its two orthogonal
components.

Sum of Orthogonal Vectors

The projection of w onto x is x, and the projection of w onto y is y:
Equation:

w = (1)x + (1)y.

If we scale w by a to produce the vector z = aw, the orthogonal

decomposition of z is
Equation:

z = aw = (a)x + (a)y.

Let's turn this argument around. Instead of building w from orthogonal

vectors x and y, let's begin with arbitrary w and x and see whether we can
compute an orthogonal decomposition. The projection of w onto x is found
from Equation 10:
Equation:

(w, x)
wx = x.
(x, x)

But there must be another component of w such that w is equal to the sum of
the components. Let's call the unknown component w . Then y

Equation:
w = wx + wy .

Now, since we know w and w already, we find w to be

x y

Equation:

(w, x)
wy = w − wx = w − x.
(x, x)
Interestingly, the way we have decomposed w will always produce w and x

w>y orthogonal to each other. Let's check this:

Equation:
(w,x) (w,x)
(w x , w y ) = x, w − x
(x,x) (x,x)

2
(w,x) (w,x)
= (x, w) − 2
(x, x)
(x,x) (x,x)

2 2
(w,x) (w,x)
= −
(x,x) (x,x)

= 0.

To summarize, we have taken two arbitrary vectors, w and x, and

decomposed w into a component in the direction of x and a component
orthogonal to x.
Linear Algebra: Other Norms

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Sometimes we find it useful to use a different definition of distance,

corresponding to an alternate norm for vectors. For example, consider the l-
norm defined as
Equation:

||x|| = (|x 1 |+|x 2 |+ ⋯ +|x n |),

1

where |x | is the magnitude of component xi. There is also the sup-norm,

i

the “supremum” or maximum of the components x , ..., x : 1 n

Equation:

||x|| sup =max (|x 1 |, |x 2 |, ..., |x n |).

The following examples illustrate what the Euclidean norm, the l-norm, and
the sup-norm look like for typical vectors.

Example:
−3

Consider the vector x = l . Then

2
 1/2

i. ||x|| = [(−3) ;
2 2 2 1/2
+ (1) + (2) ] = (14)

ii. ||x|| 1
= (| − 3| + |1| + |2|) = 6 ; and
iii. ||x|| sup
=max (| − 3|, |1|, |2|) = 3.

Example:
x1
Figure 1 shows the locus of two-component vectors x = [ ] with the
x2

property that ||x|| = 1, ||x|| 1

= 1 , or ||x|| sup
= 1.

Locus of Two-Dimensional Vectors Whose

Various Norms Are 1

The next example shows how the l-norm is an important part of city life.
Example:
The city of Metroville was laid out by mathematicians as shown in Figure
2. A person at the intersection of Avenue 0 and Street −2 (point A) is
clearly two blocks from the center of town (point C). This is consistent
with both the Euclidean norm
Equation:

2
||A|| = √ 0 + (−2) = √ 4 = 2
2

and the l-norm

Equation:

||A|| = (|0| + | − 2|) = 2.

1

But how far from the center of town is point B at the intersection of
Avenue-2 and Street 1? According to the Euclidean norm, the distance is
Equation:

2 2
||B|| = √ (−2) + (1) = √ 5.
 Metroville, U.S.A.

While it is true that point B is √5 blocks from C , it is also clear that the
trip would be three blocks by any of the three shortest routes on roads. The
appropriate norm is the l-norm:
Equation:
B
1 || = (| − 2| + |1|) = 3.
1

Even more generally, we can define a norm for each value of p from 1 to
infinity. The so-called p-norm is
Equation:
p p
p 1/p
|Ix|| p = (|x 1 +|x 2 + ⋯ +|x n | ) .

Exercise:
Problem:

Show that the Euclidean norm is the same as the p-norm with p = 2
and that the 1-norm is the p-norm with p = 1. (It can also be shown
that the sup-norm is like a p-norm with p = ∞.)

DEMO 4.1 (MATLAB). From the command level of MATLAB, type the
following lines:

>> x = [1;3;-2;4]
>> y = [0;1;2;-0.5]
>> x - y

Check to see whether the answer agrees with the definition of vector
subtraction. Now type
>> a = -1.5
>> a * x

Check the answer to see whether it agrees with the definition of scalar
multiplication. Now type

>> x' * y

This is how MATLAB does the inner product. Check the result. Type

>> norm(y)
>> sqrt(y' * y)

>> norm(y,1)
>> norm(y' * y)

Now type your own MATLAB expression to find the cosine of the angle
between vectors x and y. Put the result in variable t. Then find the angle θ
by typing

>> theta = acos(t)

The angle θ is in radians. You may convert it to degrees if you wish by

multiplying it by 180/π:

>> theta = theta * (180/pi)

Linear Algebra: Matrices

Note:This module is part of the collection, A First Course in Electrical and Computer Engineering. The LaTeX
source files for this collection were created using an optical character recognition technology, and because of this
process there may be more errors than usual. Please contact us if you discover any errors.

The word matrix dates at least to the thirteenth century, when it was used to describe the rectangular copper tray, or
matrix, that held individual leaden letters that were packed into the matrix to form a page of composed text. Each
letter in the matrix, call it a , occupied a unique position (ij) in the matrix. In modern day mathematical parlance,
ij

a matrix is a collection of numbers arranged in a two-dimensional array (a rectangle). We indicate a matrix with a
boldfaced capital letter and the constituent elements with double subscripts for the row and column:

Matrix

In this equation A is an m x n matrix, meaning that A has m horizontal rows and n vertical columns. As an
extension of the previously used notation, we write A ∈ R to show that A is a matrix of size m x n with
mxn

∈ R . The scalar element a is located in the matrix at the i row and the j column. For example, a is
th th
aij ij 23

located in the second row and the third column as illustrated in [link].

The main diagonal of any matrix consists of the elements a . (The two subscripts are equal.) The main diagonal
ii

runs southeast from the top left corner of the matrix, but it does not end in the lower right corner unless the matrix
is square (∈ R m×m
).

The transpose of a matrix A ∈ R is another matrix B whose element in row j and column i is b = a for
m×n
ji ij

1 ≤ i ≤ m and 1 ≤ j ≤ n. We write B = A to indicate that B is the transpose of A. In MATLAB, transpose is

T

denoted by A'. A more intuitive way of describing the transpose operation is to say that it flips the matrix about its
main diagonal so that rows become columns and columns become rows.
Exercise:

2 1

Problem: If A ∈ R m×n
, then A T
∈?
–
. Find the transpose of the matrix A = 5 4 .

7 9

Matrix Addition and Scalar Multiplication. Two matrices of the same size (in both dimensions) may be added
or subtracted in the same way as vectors, by adding or subtracting the corresponding elements. The equation
C = A ± B means that for each i and j, c = a ± b . Scalar multiplication of a matrix multiplies each element
ij ij ij

of the matrix by the scalar:

Equation:
 ax 11 ax 12 ax 13 ... ax 1n

ax 21 ax 22 ax 23 ... ax 2n

aX = ax 31 ax 32 ax 33 ... ax 3n .

⠇ ⠇ ⠇ ⠇
ax ml ax m2 ax m3 ... ax mn

Matrix Multiplication. A vector can be considered a matrix with only one column. Thus we intentionally blur the
distinction between R n×1
and Rn. Also a matrix can be viewed as a collection of vectors, each column of the
matrix being a vector:
Equation:

a 11 a l2 a 1n
| | |
a 21 a 22 a 2n
A = al a2 ... an = .
| | |
| | |
a m1 a m2 a mn

In the transpose operation, columns become rows and vice versa. The transpose of an n × 1 matrix, a column
vector, is a 1 ×n matrix, a row vector:
Equation:

x1

x2
T
x = ;x = [x 1 x 2 x n ].
|

xn

Now we can define matrix-matrix multiplication in terms of inner products of vectors. Let's begin with matrices
A ∈ R
m×n
and B ∈ R . To find the product AB, first divide each matrix into column vectors and row vectors
n×p

as follows:

Thus a is the i column of A and α is the j row of A. For matrix multiplication to be defined, the width of the
i
th T
j
th

first matrix must match the length of the second one so that all rows α and columns b have the same number of
T
i i

elements n. The matrix product, C = AB, is an m × p matrix defined by its elements as c = (α , b ). In words, ij i j

each element of the product matrix, c , is the inner product of the i row of the first matrix and the j column of
ij
th th

the second matrix.

For n-vectors x and y, the matrix product x y takes on a special significance. The product is, of course, a 1 × 1
T

matrix (a scalar). The special significance is that x y is the inner product of x and y:
T

Thus the notation x y is often used in place of (x, y). Recall from Demo 1 from "Linear Algebra: Other Norms"
T

that MATLAB uses x'*y to denote inner product.

Another special case of matrix multiplication is the outer product. Like the inner product, it involves two vectors,
but this time the result is a matrix:

In the outer product, the inner products that define its elements are between one-dimensional row vectors of x and
one-dimensional column vectors of y , meaning the (i, j) element of A is x y .
T
i j

Exercise:

Problem: Find C = AB where A and B are given by

0 −2 1 −3
1 −1 2
a. A = [ ], B = 4 2 2 0 ;
3 0 5
2 −2 3 1

1 0 1 2 3 4
b. A = [ ], B = [ ] ;
0 1 5 6 7 8

1 −1 −1 0 3 6

c. A = 1 −1 1 , B = 1 4 7 .
1 1 1 2 5 8

There are several other equivalent ways to define matrix multiplication, and a careful study of the following
discussion should improve your understanding of matrix multiplication. Consider A ∈ R ,B ∈ R , and m×n n×p

C = AB so that C ∈ R . In pictures, we have

m×p

Equation:
p n
p
m [C] = m [A][ ]n.
B

In our definition, we represent C on an entry-by-entry basis as

Equation:
n

c ij = (α i , b j ) = Σ a ik b kj .
k=1
In pictures,

You will prove in Exercise 3 that we can also represent C on a column basis:
Equation:
n

cj = Σ a k b kj .
k=1

Finally, C can be represented as a sum of matrices, each matrix being an outer product:
Equation:
n
T
C = Σ ai βi
i=1

A numerical example should help clarify these three methods.

Example:
Let
Equation:

1 2 1
1 2 1 3
2 2 1
A = 2 1 2 4 , B = .
1 3 2
3 3 2 1
2 1 1

Using the first method of matrix multiplication, on an entry-by-entry basis, we have

Equation:
4

c ij = Σ a ik b kj
k=1

or
Equation:
 (1 ⋅ 1 + 2 ⋅ 2 + 1 ⋅ 1 + 3 ⋅ 2) (1 ⋅ 2 + 2 ⋅ 2 + 1 ⋅ 3 + 3 ⋅ 1) (1 ⋅ 1 + 2 ⋅ 1 + 1 ⋅ 2 + 3 ⋅ 1)

C = (2 ⋅ 1 + 1 ⋅ 2 + 2 ⋅ 1 + 4 ⋅ 2) (2 ⋅ 2 + 1 ⋅ 2 + 2 ⋅ 3 + 4 ⋅ 1) (2 ⋅ 1 + 1 ⋅ 1 + 2 ⋅ 2 + 4. 1)

(3 ⋅ 1 + 3 ⋅ 2 + 2 ⋅ 1 + 1 ⋅ 2) (3 ⋅ 2 + 3 ⋅ 2 + 2 ⋅ 3 + 1 ⋅ 1) (3 ⋅ 1 + 3 ⋅ 1 + 2 ⋅ 2 + 1 ⋅ 1)

or
Equation:

12 12 8

C = 14 16 11 .

13 l9 l1

On a column basis,
Equation:
4

cj = Σ a k b kj
k=1

1 2 1 3 1 2 1 3 1 2 1

c1 = 2 1 + 1 2 + 2 1 + 4 2; c 2 = 2 2 + 1 2 + 2 3 + 4 1; c 3 = 2 1 + 1 1 + 2 2 +

3 3 2 1 3 3 2 1 3 3 2

Collecting terms together, we have

Equation:

C = [c 1 c 2 c 3 ]

(1 ⋅ 1 + 2 ⋅ 2 + 1 ⋅ 1 + 3 ⋅ 2) (1 ⋅ 2 + 2 ⋅ 2 + 1 ⋅ 3 + 3 ⋅ 1) (1 ⋅ 1 + 2 ⋅ 1 + 1 ⋅ 2 + 3 ⋅ 1)

C = (2 ⋅ 1 + 1 ⋅ 2 + 2 ⋅ 1 + 4 ⋅ 2) (2 ⋅ 2 + 1 ⋅ 2 + 2 ⋅ 3 + 4 ⋅ 1) (2 ⋅ 1 + 1 ⋅ 1 + 2 ⋅ 2 + 4. 1)

(3 ⋅ 1 + 3 ⋅ 2 + 2 ⋅ 1 + 1 ⋅ 2) (3 ⋅ 2 + 3 ⋅ 2 + 2 ⋅ 3 + 1 ⋅ 1) (3 ⋅ 1 + 3 ⋅ 1 + 2 ⋅ 2 + 1 ⋅ 1)

On a matrix-by-matrix basis,
Equation:
4
T
C = Σ ai βi
i=1

Equation:

1 2 1 3

C = 2 1 2 1 + 1 2 2 1 + 1 3 2 + 2 1 3 2 + 4 1 1 1

3 3 2 1

Equation:

1 ⋅ 1 1 ⋅ 2 1 ⋅ 1 2 ⋅ 2 2 ⋅ 2 2 ⋅ 1 1 ⋅ 1 1 ⋅ 3 1 ⋅ 2 3 ⋅ 2 3 ⋅ 1 3 ⋅ 1

= 2 ⋅ 1 2 ⋅ 2 2 ⋅ 1 + 1 ⋅ 2 1 ⋅ 2 1 ⋅ 1 + 2 ⋅ 1 2 ⋅ 3 2 ⋅ 2 + 4 ⋅ 2 4 ⋅ 1 4 ⋅ 1

3 ⋅ 1 3 ⋅ 2 3 ⋅ 1 3 ⋅ 2 3 ⋅ 2 3 ⋅ 1 2 ⋅ 1 2 ⋅ 3 2 ⋅ 2 1 ⋅ 2 1 ⋅ 1 1 ⋅ 1

Equation:

(1 ⋅ 1 + 2 ⋅ 2 + 1 ⋅ 1 + 3 ⋅ 2) (1 ⋅ 2 + 2 ⋅ 2 + 1 ⋅ 3 + 3 ⋅ 1) (1 ⋅ 1 + 2 ⋅ 1 + 1 ⋅ 2 + 3 ⋅ 1)

C = (2 ⋅ 1 + 1 ⋅ 2 + 2 ⋅ 1 + 4 ⋅ 2) (2 ⋅ 2 + 1 ⋅ 2 + 2 ⋅ 3 + 4 ⋅ 1) (2 ⋅ 1 + 1 ⋅ 1 + 2 ⋅ 2 + 4. 1)

(3 ⋅ 1 + 3 ⋅ 2 + 2 ⋅ 1 + 1 ⋅ 2) (3 ⋅ 2 + 3 ⋅ 2 + 2 ⋅ 3 + 1 ⋅ 1) (3 ⋅ 1 + 3 ⋅ 1 + 2 ⋅ 2 + 1 ⋅ 1)
as we had in each of the other cases. Thus we see that the methods are equivalent-simply different ways of
organizing the same computation!

Exercise:
Problem:

Prove that Equations 9, 11, and 13 are equivalent definitions of matrix multiplication. That is, if C = AB
where A ∈ R and B ∈ R
m×n
, show that the matrix-matrix product can also be defined by
n×p

Equation:
n

c ij = ∑ a ik b kj ,

k=1

and, if c is the k
k
th
column of C and a is the k
k
th
column of A, then
Equation:
n

c j = ∑ a k b kj .

k=1

Show that the matrix C may also be written as the ‚ "sum of outer products"
Equation:
n

T
C = ∑ ak βk .

k=1

Write out the elements in a typical outer product a kβ

T
k
.
Exercise:
Problem:

Given A ∈ R m×n
,B ∈ R , and C ∈ R , for each of the following postulates, either prove that it is true
p×q r×s

or give a counterexample showing that it is false:

T
a. (A ) = A.
T

b. AB = BA when n = p and m = q. Is matrix multiplication commutative?

c. A(B + C) = AB + AC when n = p = r and q = s. Is matrix multiplication distributive over
addition?
d. (AB ) = B A when n = p.
T T T

e. (AB)C = A(BC) when n = p and q = r. Is matrix multiplication associative?

Example:
Rotation
We know from the chapter on complex numbers that a complex number z 1
= x 1 + jy 1 may be rotated by angle θ
in the complex plane by forming the product
Equation:
jθ
z2 = e z1 .
When written out, the real and imaginary parts of z are 2

Equation:

z2 = (cos θ + j sin θ) (x 1 + jy 1 )

= (cos θ)x 1 − (sin θ)y 1 + j [(sin θ)x 1 + (cos θ)y 1 ].

If the real and imaginary parts of z and z are organized into vectors z and z as in the chapter on complex
1 2 1 2

numbers, then rotation may be carried out with the matrix-vector multiply
Equation:

x2 cosθ −sinθ x1
z2 = [ ] = [ ] [ ].
y2 sinθ cosθ y1

cosθ −sinθ
We call the matrix R (θ) = [ ] a rotation matrix.
sinθ cosθ

Exercise:

Problem: Let R(θ) denote a 2 × 2 rotation matrix. Prove and interpret the following two properties:

a. R T
(θ) = R (−θ) ;
1 0
b. R T
(θ)R (θ) = R (θ)R
T
(θ) = [ ].
0 1
Linear Algebra: Solving Linear Systems of Equations

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

We are now equipped to set up systems of linear equations as matrix- vector

equations so that they can be solved in a standard way on a computer.
Suppose we want to solve the equations from Example 1 from "Linear
Algebra: Introduction" for x and x using a computer. Recall that
1 2

Equations 1 and 2 from Linear Algebra: Introduction are

Equation:

x 1 + x 2 = 85
x1 x2
=
1.2 1.5−1.2

The first step is to arrange each equation with all references to x in the
1

first column, all references to x in the second column, etc., and all
2

constants on the right-hand side:

Equation:

x 1 + x 2 = 85

0. 3x 1 − 1. 2x 2 = 0.

Then the equations can be converted to a single matrix-vector equation. The

coefficients form a matrix, keeping their same relative positions, and the
variables and constants each form a vector:
Equation:
 1 1 x1 85
[ ][ ] = [ ].
0. 3 −1. 2 x2 0

Exercise:
Problem:

Verify by the rules of matrix multiplication that the system of

equations in 2 in equivalent to the matrix equation in 3.

Equation 3 is of the general form

Equation:

Ax = b

where in this case

Equation:

1 1 xl 85
A = [ ], x = [ ], b = [ ].
0. 3 −1. 2 x2 0

Given any A ∈ R n×n

and b ∈ R , MATLAB can solve Equation 4 for x
n

(as long as a solution exists). Key ideas in the solution process are the
identity matrix and the inverse of a matrix.

When the matrix A is the 1 × 1 matrix a, the vector x is the l-vector x, and
the vector b is the l-vector b, then the matrix equation Ax = b becomes the
scalar equation
Equation:

ax = b.

The scalar a is the inverse of the scalar a, so we may multiply on both

−1

sides of Equation 6 to obtain the result

Equation:
 −1 −1
a (ax) = a b

−1
1x = a b.

We would like to generalize this simple idea to the matrix equation Ax = b

so that we can find an inverse of the matrix A, called A , and write −1

Equation:
−1 −1
A (Ax) = A b

−1
Ix = A b.

In this equation the matrix I is the identity matrix

Equation:

1 0 0 ... 0

0 1 0 ... 0

0 0 1 ... 0
I = .

⋮ ⋮ ⋮ ⋱ ⋮

0 0 0 ... 1

It is clear that the identity matrix I and the inverse of a matrix, A , play a−1

key role in the solution of linear equations. Let's study them in more detail.

The Matrix Identity. When we multiply a scalar by 1, we get back that

same scalar. For this reason, the number 1 is called the
multiplicativeidentity element. This may seem trivial, but the generalization
to matrices is more interesting. The question is, is there a matrix that, when
it multiplies another matrix, does not change the other matrix? The answer
is yes. The matrix is called the identity matrix and is denoted by I . The
identity matrix is always square, with whatever size is appropriate for the
matrix multiplication. The identity matrix has l's on its main diagonal and
0 s everywhere else. For example,
′

Equation:
 1 0 0 0 0

0 1 0 0 0

I5 = 0 0 1 0 0 .

0 0 0 1 0

0 0 0 0 1

The subscript 5 indicates the size. In terms of the unit coordinate vectors e ,
i

we can also write the n × n identity matrix as

Equation:

| | |

In = e1 e2 ... en .

| | |

For any matrix A ∈ R n×p

, we have
Equation:

A = I n A.

For p = 1, we obtain the following special form for any vector x ∈ R : n

Equation:

x = I n x.

This last equation can be written as the sum

Equation:
n

x = ∑ xi ei .

i=1

This is a special case of one of the results you proved in Exercise 3 from
"Linear Algebra: Matrices". Figure 1 illustrates Equation 14 and shows how
the vector x can be broken into component vectors x e , each of which lies
i i

in the direction of one coordinate axis. The values of the x s are the
′
i

coordinates of x, and their magnitudes are also the lengths of the

component vectors.

Breaking a Vector into Components

Exercise:
Problem:

Use Equation 13 and the rules for matrix multiplication to show that
x ∈ R may also be written as
n

Equation:
n

x = ∑ (x, e i )e i .

i=1

This verifies Equation 11 from "Linear Algebra: Direction Cosines".

When the product of two numbers is 1 (the identity element), we say that
they are inverses of each other, like 2 and 0.5. Likewise, we say that two
square matrices are inverses of each other if their product is the identity
matrix:
Equation:
−1
AB = I ⇔ B = A .

An interesting and useful result is implied by this definition. Take the first
form of Equation 16 and multiply by B from the left:
Equation:

AB = I

⇒ B(AB) = BI

⇒ (BA)B = B

⇒ BA = I.

We emphasize that the inference made in the last step here is only valid
when B and A are square matrices. The result means that, even though
matrix multiplication is not commutative in general, we have a special case
where it is commutative. If A and B are inverses of each other, then
Equation:

AB = BA = I.

Exercise:
Problem:

Prove that the inverse of the 2 ×2 rotation matrix R(θ) is R T

(θ) .

Matrices that are not square do not have inverses. In fact, not all square
matrices have inverses. So it becomes an important issue to determine
which matrices do have inverses. If a matrix has an inverse, the inverse is
unique. If a square matrix has no inverse, it is called a singular matrix. The
determinant of a square matrix is a scalar computed from the numbers in
the matrix. It tells us whether a matrix will have an inverse or not. A matrix
is singular if and only if its determinant is zero.[footnote] In MATLAB, the
notation det(A) is used to compute the determinant. Whenever the matrix
A in Equation 4 is singular, it means one of two things about the system of
equations represented: either the equations are inconsistent and there is no
solution, or the equations are dependent and there are infinitely many
solutions.
It is not important now to understand how the determinant is defined and
computed from the elements of a matrix. In your more advanced courses
you will study the determinant in some detail.

Solving Ax = b. Let's now study ways to solve the matrix equation

Ax = b. We will assume that a unique solution for x exists. Thus a unique

matrix A exists with the property A A = I. The trick is to find it. Here
−1 −1

is one procedure.

For convenience, rewrite the matrix equation Ax = b as

Equation:

x
[Ab][ ] = 0.
−1

The matrix [Ab] ∈ R is called the augmented matrix for the system
nx(n+1)

of equations. The augmented matrix may be viewed as a systematic way of

writing all the information necessary to solve the equations.

The advantage of Equation 19 is that we may premultiply both sides by any

matrix C (of compatible size), and the right-hand side remains zero
1

(although this is equivalent to multiplying on both sides of Equation 4,

which some may prefer). We can repeat this operation as often as we like
with matrices C , C , etc. The general result is
2 3

Equation:

x
[C m ⋯ C 2 C 1 AC m ⋯ C 2 C 1 b][ ] = 0.
−1

Now suppose that we have found a sequence of matrices C 1, ... , C that

m

transforms the matrix A to the identity matrix:

C m ⋯ C 2 C 1 A = I. The augmented matrix equation in 20 simplifies to

Equation:

x
[I C m ⋯ C 2 C 1 b][ ] = 0,
−1

which can be multiplied out to reveal that the solution for x is

Equation:

x = C m ⋯ C 2 C 1 b.

The method may be summarized as follows:

i. form the augmented matrix [Ab] for the system;

ii. premultiply [Ab] by a sequence of selected matrices C , designed to
i

take A to I ; and
iii. when A is transformed to I , b will be transformed to x, so the solution
will appear as the last column of the transformed augmented matrix.

We may also conclude that the product of the matrices Ci must be the
inverse of A since A is the unique matrix for which A A = I. In
−1 −1

solving for x by this method, we have found A implicitly as the product

−1

C m ⋯C C .
2 1

Example:
Consider the equation
Equation:

3 1 x1 6
[ ] [ ] = [ ]
2 4 x2 5

Equation:

A x = b.

The augmented matrix for this equation is

Equation:

3 1 6
[A b] = [ ].
2 4 5

Now if we could add −2/3 times the first row to the second row, we
would get 0 in the lower left corner. This is the first step in transforming A
to the identity I. We can accomplish this row operation with the matrix
Equation:

1 0
C1 = [ ]
−2/3 1

Equation:

3 1 6
C1 [A b] = [ ].
0 10/3 1

Now adding −3/10 times the new second row to the first row will
introduce 0 in the (1, 2) position, bringing us closer still to the identity.
Thus
Equation:

1 −3/10
C2 = [ ]
0 1

Equation:

3 0 57/10
C2 C1 [A b] = [ ].
0 10/3 1

We now complete the transformation to identity by normalizing each row

to get the needed l's:
Equation:

1/3 0
C3 = [ ]
0 3/10
Equation:

1 0 19/10
C3 C2 C1 [A b] = [ ].
0 1 3/10

According to the last column, the solution is

Equation:

19/10
x = [ ].
3/10

We note in passing that the inverse of A is the product of the C s, so ′

1/3 0 1 −3/10 1 0
−1
A = C3 C2 C1 = [ ][ ][ ]
0 3/10 0 1 −2/3 1

1/3 −1/10 1 0 2/5 −1/10

= [ ] [ ] = [ ]⋅
0 3/l0 −2/3 1 −1/5 3/10

The method we have just used, combined with a particular way of choosing
the C matrices, is called Gauss elimination. Gauss elimination requires less
i

computation than finding the inverse of A because only the effect of A −1

on the specific vector b is computed. MATLAB can solve for x by either

method, as shown in Demo 4.2. For hand computations, we suggest
choosing the C matrices so that C produces O s everywhere below the
i 1
′

diagonal in the first column, C produces O s below the diagonal in the

2
′

second column, and so on up to C . Then C produces O s above the

n−1 n
′

diagonal in the n column, C

th
produces O s above the diagonal in
n+1
′

column n − 1, etc. The last one, C , normalizes the diagonal elements

2n−1

to l's. We have assumed for simplicity that no O s on the diagonal will be

′

encountered in hand computations.

Exercise:

Problem: Check that A −1

A = I in Example 1
Exercise:
 Problem:

Augment Equation 3 as in Equation 19 and use the technique of Gauss

elimination to solve for x.

Demo 1 (MATLAB). From the command level of MATLAB, solve the

matrix equation of Example 1 from "Linear Algebra: Introduction" by
typing

>> A = [1 1;0.3 -1.2]

>> b = [85;0]

You have entered the matrices A and b, which describe the problem. You
can now solve for x by finding the inverse of A and multiplying b:

>> x = inv(A) * b

In this example the inverse is computed quickly because A is a small

matrix. For a large (say, 30 × 30) matrix, the answer would take longer to
compute, and the more efficient method of Gauss elimination would reduce
waiting time. You can use Gauss elimination in MATLAB by typing

>> x = A \ b

You should get the same answer. Now type the MATLAB code required to
compute Ax and to show Ax − b = 0.
Exercise:
Problem:

(MATLAB) Write the following system of equations as Ax = b and

solve using MATLAB:
Equation:

3 (x 1 − x 3 ) + 2 (x 2 − 1) − 6 = x 3

Equation:
4x 3 = 7x 2 − 5
Equation:

6 (x 1 + x 2 + 2) = x 3 + 1.
Linear Algebra: Circuit Analysis

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In this section we use the linear algebra we have developed to find the
voltages and currents in a simple electrical circuit, such as the one shown in
Figure 1. There are many reasons why this might be necessary; in this
example we need to know the current flowing through the lamp to tell
whether or not it will glow. Too little current will have no visible effect; too
much current will cause the lamp to burn out. We will apply a few physical
laws relating the voltages and currents in a circuit, turn these laws into
systems of linear equations, and solve the equations for the voltages and
currents.

A Simple Electrical Circuit

Current, Voltage, and Resistance. We will use three physical quantities in

our analysis of electrical circuits: current, voltage, and resistance. Current is
the flow of electrical charge from one place to another. Electrons flowing
through a wire or through some other electronic device comprise a current.
Voltage is a difference in electric potential that makes electrons flow.
Voltage is sometimes called electromotive force because it is like a “force”
that moves electrons. Resistance is a property of the device through which
the electron current flows. The lower the resistance of a device, the more
easily current can flow through the device.

The analogy of water flowing through pipes can help you develop intuition
about electrical circuits. In this analogy, electrical current corresponds to the
flow rate of water. Voltage corresponds to the pressure that forces the water
to flow, and resistance is the friction of flow. A small pipe would impede
the flow of water more than would a large pipe, so the small pipe would
correspond to a higher resistance. While this analogy can be helpful, keep
in mind that electricity is not water. All analogies break down at some
point.

We measure electrical current in amperes. The standard symbol for current

is i, and the direction of positive flow is indicated by an arrow on the circuit
diagram. The arrow is for reference only; if the true current is in the
opposite direction, we get negative values for i. Because electrons are
negatively charged, current is defined as flowing in the opposite direction as
electron motion. But to reduce confusion, you should learn to think in terms
of current rather than electron motion.

A point in a circuit where several devices are connected together is called a

node. The conservation law for current says that “what flows in must flow
out of a node,” a principle known as Kirchhoff's current law. Kirchhoff's
current law states that the sum of all currents leaving a node is zero. In this
law, a current entering the node is considered to be a negative current
leaving the node.

Voltage is measured in volts and is usually written as v (or e). Since voltage
is a difference in potential between two points (nodes), we can show it on a
circuit diagram with a +and a –sign to indicate which two nodes we are
comparing and which one of the nodes is considered negative. As with
current, the markings are for reference only and we may end up with a
negative value of v.

In an electrical circuit, one node is usually chosen as a reference node and is

considered to have a voltage of zero. Then the voltage at every other node is
measured with respect to the reference node. This saves us the trouble of
always specifying pairs of nodes for voltage measurements and marking +
and – signs for each voltage. Other names for the reference node are
common and ground.

A constant voltage source is a device that always forces the voltage

between its two terminals to be a constant value. In Figure 1 the circle at the
left represents a constant voltage source of 5 volts, so that the voltage at the
upper (+) end is always exactly 5 volts higher than the voltage at the lower
(-) end. A voltage source is something like a battery, but idealized. Real
batteries do not maintain a constant output voltage under all conditions.

Resistance is measured in ohms and is denoted by R. A resistor is shown as

a zig-zag line in circuit diagrams and labeled with the value of its resistance
in ohms. In this chapter we will consider only devices whose resistance is
positive and the same in both directions. Ohm's law, also called the resistor
law, relates the voltage and current in a resistor. For the resistor shown in
Figure 2, with reference directions assigned to v and i as shown, Ohm's law
is
Equation:

v = iR.

Note that current flows from + to - through the resistor.

Ohm's Law
Example:
Ohm's law and Kirchhoff's current law are the only principles we need to
write equations that will allow us to find the voltages and currents in the
resistive circuit of Figure 1. We begin by choosing a reference node and
assigning variables to the voltages at every other node (with respect to the
reference node). These choices are shown in Figure 3.

Assigning Node Voltages

The constant voltage source forces v to be exactly 5 volts higher than the
1

reference node. Thus

Equation:
v 1 = 5.

Next we write equations by applying Kirchhoff's current law to each node

in the circuit (except the reference node and v , whose voltages we already
1

know). At the node labeled v are three paths for leaving current. The
2

current leaving through the 50 ohm resistor can be found by Ohm's law,
where the voltage across that resistor is v − v :
2 1

Equation:

v (v 2 − v 1 )
i 50 = =
R 50

For current leaving through the 300 ohm resistor, the voltage is v . Pay
2

careful attention to the sign; since we are interested in the current leaving
the node labeled v , Figure 4.14 indicates that to apply Ohm's law we
2

should take the voltage as +v − reference = v − 0 = v . So

2 2 2
Equation:
v2
i 300 =
300

For the 100 ohm resistor, we can write

Equation:

(v 2 − v 3 )
i 100 =
100

According to Kirchhoff's current law, the sum of these three leaving

currents is zero:
Equation:
(v 2 −v 1 ) v2 (v 2 −v 3 )
+ + = 0
50 300 100

⇒ 6 (v 2 − v 1 ) + v 2 + 3 (v 2 − v 3 ) = 0

⇒ −6v 1 + 10v 2 − 3v 3 = 0.

Notice that when we wrote the equation for the node labeled v , the 2

variable v had a + sign each time it occurred in the equation, while the
2

others had a -sign. This is always the case, and watching for it can help you
avoid sign errors. Now we apply Kirchhoff's current law at the node
labeled v to get the equation
3

Equation:
(v 3 −v 2 ) v3
+ = 0
100 2

⇒ (v 3 − v 2 ) + 50v 3 = 0

⇒ 0v 1 − 1v 2 + 51v 3 = 0.

Note that this time it is v that always shows up with a + sign.

3

Equations 2, 6, and 7 give us a system of three equations in the three

unknown variables v , v , and v . We now write them in matrix form as
1 2 3

Equation:
 1 0 0 vl 5

−6 10 −3 v2 = 0 ⋅

0 −1 51 v3 0

Exercise:
Problem:

(MATLAB) Use MATLAB to solve Equation 8. You should find

Equation:

v 1 = 5. 0000 volts

Equation:

v 2 = 3. 0178 volts

Equation:

v 3 = 0. 0592 volt.

What is the determinant of the coefficient matrix A? Is the solution

unique?

We can determine the current flowing through the lamp from v to ground
3

in Example 1 by Ohm's law:

Equation:
v v3
i = = = 0. 0296 ampere.
R 2

The visible effect will, of course, depend on the lamp. Let us assume that
the specifications for our lamp indicate that 0.05 ampere or more is required
before it will glow, and more than 0.075 ampere will cause it to burn out. In
this case, our circuit would not make the lamp glow.
Exercise:
Problem:

(MATLAB) Write and solve equations for the circuit in Figure 4. What
are the voltages at the nodes labeled v through v ? What is the
1 4

current labeled i ? And i ?

1 2

A Resistive Network
Linear Algebra: Numerical Experiment (Circuit Design)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Our analysis in Example 1 from "Linear Algebra: Circuit Analysis" and

Problem 1 from "Linear Algebra: Circuit Analysis" indicates that not
enough current will flow through the lamp to make it glow. We now wish to
change the resistance of the 100 ohm resistor to a new value so that the
lamp will glow. We replace 100 in the equations with an unknown
resistance R. Equation 2 is unchanged, but Equation 6 becomes
Equation:
(v2−v1) v2 (v2−v3)
+ + = 0
50 300 R

⇒ 6R (v2 − v1) + Rv2 + 300 (v2 − v3) = 0

⇒ −6Rv1 + (7R + 300)v2 − 300v3 = 0.

Equation 7 from "Linear Algebra: Circuit Analysis" becomes

Equation:
(v3−v2) v3
+ = 0
R 2

⇒ 2 (v3 − v2) + Rv3 = 0

⇒ 0v1 − 2v2 + (R + 2)v3 = 0.

The matrix form of these equations is

Equation:
 1 0 0 v1 5

−6R 7R +300 −300 v2 = 0 .

0 −2 R + 2 v3 0

Write a MATLAB function file called builda to accept R as an input and

return the matrix A in Equation 3 as an output. The first line of your
function file should be

function A = builda(R);

Now choose several values for R. For each choice, use your function
builda and solve the resulting matrix equation Av = b for the voltages.

Each time you choose a different R to build a different matrix A, check the
determinant of A to be sure the equations have a unique solution:

>> det(A)

Make a table of R and the corresponding values of v :

3

Now add a column to your table for the current through the lamp i = v /2.
3

Add rows to your table until you have found a value of R for which the
lamp will glow. ( i needs to be between 0.05 and 0.075 ampere.)
Vector Graphics: Introduction

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Acknowledgements
Fundamentals of Interactive Computer Graphics by J. D. Foley and A. Van
Dam, ©1982 Addison-Wesley Publishing Company, Inc., Reading,
Massachusetts, was used extensively as a reference book during
development of this chapter. Star locations were obtained from the share-
ware program “Deep Space” by David Chandler, who obtained them from
the “Skymap” database of the National Space Science Data Center.

Notes to Teachers and Students:

In this chapter we introduce matrix data structures that may be used to
represent two- and three-dimensional images. The demonstration program
shows students how to create a function file for creating images from these
data structures. We then show how to use matrix transformations for
translating, scaling, and rotating images. Projections are used to project
three-dimensional images onto two-dimensional planes placed at arbitrary
locations. It is precisely such projections that we use to get perspective
drawings on a two-dimensional surface of three-dimensional objects. The
numerical experiment encourages students to manipulate a star field and
view it from several points in space.

Once again we consider certain problems essential to the chapter

development. For this chapter be sure not to miss the following exercises:
Exercise 2 in "Two-Dimensional Image Transformations", Exercise 1 in
"Homogeneous Coordinates", Exercise 2 in "Homogeneous Coordinates",
Exercise 5 in "Three-Dimensional Homogeneous Coordinates", and
Exercise 2 in "Projections".

Introduction
Pictures play a vital role in human communication, in robotic
manufacturing, and in digital imaging. In a typical application of digital
imaging, a CCD camera records a digital picture frame that is read into the
memory of a digital computer. The digital computer then manipulates this
frame (or array) of data in order to crop, enlarge or reduce, enhance or
smooth, translateor rotate the original picture. These procedures are called
digital picture processing or computer graphics. When a sequence of
picture frames is processed and displayed at video frame rates (30 frames
per second), then we have an animated picture.

In this chapter we use the linear algebra we developed in The chapter on

Linear Algebra to develop a rudimentary set of tools for doing computer
graphics on line drawings. We begin with an example: the rotation of a
single point in the (x, y) plane.

Example:
Point P has coordinates (3, 1) in the (x, y) plane as shown in Figure 1.
π
Find the coordinates of the point P ′, which is rotated radians from P .
6
 Rotating a Single Point in the (x, y) Plane

To solve this problem, we can begin by converting the point P from

rectangular coordinates to polar coordinates. We have
Equation:

2 2
r = √ x + y = √ 10

−1 y
θ =tan ( ) ≈ 0. 3217 radian.
x

π
The rotated point P ′ has the same radius r, and its angle is θ + . We
6
now convert back to rectangular coordinates to find x′ and y′ for point P ′:
Equation:
′ π
x = r cos (θ + ) ≈ √ 10 cos (0. 8453) ≈ 2. 10
6

′ π
y = r sin (θ + ) ≈ √ 10 sin (0. 8453) ≈ 2. 37.
6

So the rotated point P ′ has coordinates (2.10, 2.37).

Now imagine trying to rotate the graphical image of some complex object
like an airplane. You could try to rotate all 10,000 (or so) points in the same
way as the single point was just rotated. However, a much easier way to
rotate all the points together is provided by linear algebra. In fact, with a
single linear algebraic operation we can rotate and scale an entire object and
project it from three dimensions to two for display on a flat screen or sheet
of paper.

In this chapter we study vector graphics, a linear algebraic method of

storing and manipulating computer images. Vector graphics is especially
suited to moving, rotating, and scaling (enlarging and reducing) images and
objects within images. Cropping is often necessary too, although it is a little
more difficult with vector graphics. Vector graphics also allows us to store
objects in three dimensions and then view the objects from various
locations in space by using projections.

In vector graphics, pictures are drawn from straight lines.[footnote] A curve

can be approximated as closely as desired by a series of short, straight lines.
Clearly some pictures are better suited to representation by straight lines
than are others. For example, we can achieve a fairly good representation of
a building or an airplane in vector graphics, while a photograph of a forest
would be extremely difficult to convert to straight lines. Many computer-
aided design (CAD) programs use vector graphics to manipulate
mechanical drawings.
It is possible to extend these techniques to deal with some types of curves,
but we will consider only straight lines for the sake of simplicity.

When the time comes to actually display a vector graphics image, it may be
necessary to alter the representation to match the display device. Personal
computer display screens are divided into thousands of tiny rectangles
called picture elements, or pixels. Each pixel is either off (black) or on
(perhaps with variable intensity and/or color). With a CRT display, the
electron beam scans the rows of pixels in a raster pattern. To draw a line on
a pixel display device, we must first convert the line into a list of pixels to
be illuminated. Dot matrix and laser printers are also pixel display devices,
while pen plotters and a few specialized CRT devices can display vector
graphics directly. We will let MATLAB do the conversion to pixels and
automatically handle cropping when necessary.

We begin our study of vector graphics by representing each point in an

image by a vector. These vectors are arranged side-by-side into a matrix G
containing all the points in the image. Other matrices will be used as
operators to perform the desired transformations on the image points. For
example, we will find a matrix R, which functions as a rotation: the matrix
product RG represents a rotated version of the original image G.
Vector Graphics: Two-Dimensional Image Representation

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Point Matrix. To represent a straight-line image in computer memory, we

must store a list of all the endpoints of the line segments that comprise the
image. If the point P = (x , y ) is such an endpoint, we write it as the
i i i

column vector
Equation:

xi
pi = [ ]
yi

Suppose there are n such endpoints in the entire image. Each point is
included only once, even if several lines end at the same point. We can
arrange the vectors P into a point matrix:
i

Equation:

G = [p1p2p3. . . pn]

x1 x2 x3 xn
= [ ].
y1 y2 y3 yn

We then store the point matrix G ∈ R 2xn

as a two-dimensional array in
computer memory.

Example:
Consider the list of points
Equation:

P1 = (0, 0)

P2 = (−1. 5, 5)

P3 = (4, 2. 3)

P4 = (4, −1).

The corresponding point matrix is

Equation:

0 −1. 5 4 4
G = [ ]⋅
0 5 2. 3 −1

Line Matrix. The next thing we need to know is which pairs of points to
connect with lines. To store this information for m lines, we will use a line
matrix, H ∈ R 2×m
. The line matrix does not store line locations directly.
Rather, it contains references to the points stored in G. To indicate a line
between points p and p , we store the indices i and j as a pair. For the k
i j
th

line in the image, we have the pair

Equation:

ik
hk = [ ].
jk

The order of i and j does not really matter since a line from p to p is the
i j

same as a line from p to p . Next we collect all the lines h into a line
j i k

matrix H :
Equation:

i1 i2 i3 ... im
H = [ ].
j1 j2 j3 ... jm
All the numbers in the line matrix H will be positive integers since they
point to columns of G. To find the actual endpoints of a line, we look at
columns i and j of the point matrix G.

Example:
To specify line segments connecting the four points of Example 1 into a
quadrilateral, we use the line matrix
Equation:

1 2 3 4
H1 = [ ].
2 3 4 1

Alternatively, we can specify line segments to form a triangle from the first
three points plus a line from P to P :
3 4

Equation:

1 2 3 3
H2 = [ ].
2 3 1 4

Figure 1 shows the points G connected first by H and then by H

1 2.

Two Sets of Lines

Demo 1 (MATLAB). Use your editor to enter the following MATLAB
function file. Save it as vgraphl.m.

function vgraphl (points, lines);

% vgraphl (points, lines) plots the points as *'s
and
% connects the points with specified lines. The po
ints
% matrix should be 2xN, and the lines matrix shoul
d be 2xM.
% The field of view is preset to (-50,50) on both
axes.
%
% Written by Richard T. Behrens, October 1989
%
m=length(lines); % find the num
ber of
% lines.
axis([-50 50 -50 50]) % set the axis
scales
axis('square')
plot(points(1,:),points(2,:),'*') % plot the poi
nts as *
hold on % keep the poi
nts...
for i=i:m % while plotti
ng the
% lines
plot([points(1,lines(1,i)) points(1,lines(2,i)
)],..
[points(2,lines(2,lines(1,i)) points(2,lin
es(2,i))],'-')

end

hold off
After you have saved the function file, run MATLAB and type the
following to enter the point and line matrices. (We begin with the
transposes of the matrices to make them easier to enter.)

>> G = [
0.6052 -0.4728;
-0.4366 3.5555;
-2.6644 7.9629;
-7.2541 10.7547;
-12.5091 11.5633;
-12.5895 15.1372;
-6.5602 13.7536;
-31.2815 -7.7994;
-38.8314 -9.9874;
-44.0593 -1.1537;
-38.8314 2.5453;
-39.4017 9.4594;
-39.3192 15.0932;
-45.9561 23.4158]

>> G = G'

>> H = [
1 2;
2 3;
3 4;
4 5;
4 7;
5 6;
8 9;
9 10;
10 11;
11 12;
12 13;
12 14]

>> H = H'
At this point you should use MATLAB's “save” command to save these
matrices to a disk file. Type

>> save dippers

After you have saved the matrices, use the function VGRAPH1 to draw the
image by typing

≫ vgraph1(G,H)

The advantage of storing points and lines separately is that an object can be
moved and scaled by operating only on the point matrix G. The line
information in H remains the same since the same pairs of points are
connected no matter where we put the points themselves.

Surfaces and Objects. To describe a surface in three dimensions is a fairly

complex task, especially if the surface is curved. For this reason, we will be
satisfied with points and lines, sometimes visualizing flat surfaces based on
the lines. On the other hand, it is a fairly simple matter to group the points
and lines into distinct objects. We can define an object matrix K with one
column for each object giving the ranges of points and lines associated with
that object. Each column is defined as
Equation:

f irst point

last point
ki = .
f irst line

last line

As with the line matrix H , the elements of K are integers.

Example:
Consider again Demo 1. We could group the points in G and the lines in H
into two objects with the matrix
Equation:

1 8

7 l4
K = .
1 7

6 12

The first column of K specifies that the first object (Ursa Minor) is made
up of points 1 through 7 and lines 1 through 6, and the second column of
K defines the second object (Ursa Major) as points 8 through 14 and lines

7 through 12.
Vector Graphics: Two-Dimensional Image Transformations

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

We now turn our attention to operating on the point matrix G to produce the
desired transformations. We will consider rotation scaling and translation
(moving) of objects. Rotation and scaling are done by matrix multiplication
with a square transformation matrix A. If we call the transformed point
matrix G , we have
new

Equation:

[G new ] = [A] [G].

We call A a matrix operator because it “operates” on G through matrix

multiplication. In contrast, translation must be done by matrix addition.

In a later section you will see that it is advantageous to perform all

operations by matrix operators and that we can modify our image
representation to allow translation to be done with a matrix operator like
rotation and scaling. We will call the modified representation homogeneous
coordinates.

Rotation. We saw in the chapter on linear algebra that the matrix that
rotates points by an angle θ is
Equation:

cosθ −sinθ
A = R (θ) = [ ].
sinθ cosθ
When applied to the point matrix G, this matrix operator rotates each point
by the angle θ, regardless of the number of points.

Example:
We can use the rotation matrix to do the single point rotation of the
example from "Vector Graphics: Introduction". We have a point matrix
consisting of only the point (3, 1):
Equation:

3
G = [ ].
1

π
The necessary transformation matrix is R(θ) with θ = Then the rotated
6
point is given by
Equation:
π π
π cos −sin 3 2. 10
6 6
G new = R ( )G = [ ] ≈ [ ]⋅
π π
6 sin cos 1 2. 37
6 6

Scaling. An object can be enlarged or reduced in each dimension inde-

pendently. The matrix operator that scales an image by a factor of s along x

the x-axis and s along the y-axis is

y

Equation:

sx 0
A = S (s x , s y ) = [ ].
0 sy

Most often we take s x = sy to scale an image by the same amount in both

dimensions.
Exercise:
 Problem:

Write out the following matrices. Simplify and give numerical answers
to two decimal places:
π
a. R( ) ;
2
b. S(3, 2);
π
c. R(− ) ;
4
d. S(−1, 1).

Exercise:
Reflections

Problem: What does S(−1, 1) do? S(1, −1)?S(−1, −1)? S(1, 1)?
Exercise:
Problem:

0 −1. 5 4 4 π
Given G = [ ] and θ = find G new
.
=R(θ)G
0 5 2. 3 −1 3

Give numerical answers to two decimal places.

Exercise:
Problem:

Apply each of the transformations in Exercise 1 and 2 to the image

Equation:

1 1 2 2 1 2 3 4
G = [ ] ; H = [ ].
1 2 2 1 2 3 4 1

Sketch the original image and each transformation of it.

Translation. An object can be moved by adding a constant vector b to
20
every point in the object. For example, b = [ ] will move an object 20
−5

units to the right and 5 units down. We can write this in terms of the point
matrix as
Equation:
T
G new = G + b1

where 1 (read “the one-vector”) is a vector of n l's:

Equation:

1
1 = .
⋮

In MATLAB, 1 may be obtained by ones(n,1). The outer product of b

with 1 in Equation 7 simply serves to make n copies of b so that one copy
can be added to each point in G.
Vector Graphics: Composition of Transformations

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Often we will want to perform several operations on an object before we

π
display the result. For example, suppose we want to rotate by and reduce
3
1
to size in each dimension:
2
Equation:
π
G1 = R ( )G
3
1 1
G new = S ( , )G 1 .
2 2

If there are n points in the matrix G, it will require 4n multiplications to

perform each of these operations, for a total of 8n multiplications.
However, we can save some multiplications by noting that
Equation:

1 1 π
G new = S ( , ) [R ( )G] = AG
2 2 3

where
Equation:
 1 1 π
A = S( , )R ( )
2 2 3

1 π 1 π
cos ( ) − sin( )
2 3 2 3
= [ ] .
1 π l π
sin ( ) cos ( )
2 3 2 3

In other words, we take advantage of the fact that matrix multiplication is

associative to combine S and R into a single operation A, which requires
only 8 multiplications. Then we operate on G with A, which requires 4n
multiplications. By “composing” the two operations, we have reduced the
total from 8n to 4n + 8 multiplications. Furthermore, we can now build
operators with complex actions by combining simple actions.

Example:
We can build an operator that stretches objects along a diagonal line by
composing scaling and rotation. We must

i. rotate the diagonal line to the x-axis with R(−θ);

ii. scale with S(s, 1); and
iii. rotate back to the original orientation with R(θ).

Figure 1 shows a square being stretched along a 45o line. The composite
operator that performs this directional stretching is
Equation:

A(θ, s) = R(θ)S(s, 1)R(−θ)

cosθ −sinθ s 0 cosθ sinθ

= [ ] [ ] [ ]
sinθ cosθ 0 1 −sinθ cosθ

2 2
scos θ + sin θ (s − 1)sinθcosθ
= [ ].
2 2
(s − 1)sinθcosθ cos θ + ssin θ

Note that the rightmost operator in a product of operators is applied first.

Rotating and Scaling for Directional Stretching

Vector Graphics: Homogeneous Coordinates

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In the previous section we were able to combine rotation and scaling into a
single composite operation by matrix multiplication. Unfortunately,
translation cannot yet be included in the composite operator since we do it
by addition rather than by multiplication.

Example:
π
Suppose we wish to rotate the image G by about the point (−10, 10).
3
Our rotation matrix R(θ) always rotates about the origin, so we must
combine three transformations to accomplish this:

i. translate the point (−10, 10) to the origin;

π
ii. rotate radians about the origin; and
3
iii. translate the origin back to (−10, 10).

l0
For step (i), we have b 0 = [ ] and
−10

Equation:
T
G1 = G + b0 1 .

For step (ii),

Equation:
 π
G2 = R( )G 1
3

π T
= R( ) [G + b 0 1 ]
3

π π T
= R( )G + R ( )b 0 1 .
3 3

For step (iii), we can use − b from step (i):

0

Equation:
T
G new = G2 − b0 1

π π T T
= R( )G + R ( )b 0 1 − b0 1
3 3

π π T
= R( )G + [(R ( ) − I)b 0 ]1 .
3 3

In this example we were unable to find a single matrix operator A to do the

entire job. The total transformation took the form

Equation:
T
G new = AG + b1 .

This is called an affine transformation because it involves both

multiplication by A and addition of a constant matrix. This is in contrast to
the more desirable linear transformation, which involves only
multiplication by A.

We will now move toward a modified representation of the image and the
operators by rewriting the last equation as
Equation:

G
G new = [Ab][ ]
T
1

π π
where in the example we had A = R( ) and b = (R ( ) − I)b 0 .
3 3
Exercise:
 Problem: Show that, for any matrices A, B, C, D of compatible sizes,
Equation:

B
AB + CD = [A C] [ ].
D

G
The matrix [ T
] looks like
1

Equation:

x1 x2 xn

y1 y2 yn ,

1 1 1

and the points (x , y , 1) are called homogeneous coordinates. We can

i i

modify Equation 5 so that the new point matrix is also in homogeneous

coordinates:
Equation:

G new A b G
[ ] = [ ] [ ].
T T T
1 O 1 1

In the new representation, each point in the image has a third coordinate,
which is always a 1. The homogeneous transformation is a 3 × 3 matrix,
Equation:

A b
Ah = [ ] ,
T
0 1

which is capable of translation, rotation, and scaling all by matrix multi-

plication. Thus, using homogeneous coordinates, we can build composite
transformations that include translation.

In homogeneous coordinates, we have

Equation:

cosθ −sinθ 0

R (θ) = sinθ cosθ 0

0 0 1

sx 0 0

S (s x , s y ) = 0 sy 0

0 0 1

1 0 tx

T (t x , t y ) = 0 1 ty .

0 0 1

Example:
The composite transformation to triple the size of an image and then move
it 2 units to the left is
Equation:

1 0 −2 3 0 0 3 0 −2

A = 0 1 0 0 3 0 = 0 3 0 .

0 0 l 0 0 1 0 0 1

On the other hand, the composite transformation to move an image 2 units

to the left and then triple its size is
Equation:

3 0 0 1 0 −2 3 0 −6

B = 0 3 0 0 1 0 = 0 3 0 .

0 0 1 0 0 1 0 0 1
In the latter case, the distance of the translation is also tripled.

Exercise:
Problem:

Find a single composite transformation in homogeneous coordinates

that rotates an image by angle θ about point (x , y ) as in Example 1.
i i
Vector Graphics: Three-Dimensional Homogeneous Coordinates

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

We now consider the storage and manipulation of three-dimensional

objects. We will continue to use homogeneous coordinates so that
translation can be included in composite operators. Homogeneous
coordinates in three dimensions will also allow us to do perspective
projections so that we can view a three-dimensional object from any point
in space.

Image Representation. The three-dimensional form of the point matrix in

homogeneous coordinates is
Equation:

x1 x2 x3 ... xn

yl y2 y3 ... yn
4×n
G = ∈ R .
z1 z2 z3 ... zn

1 1 1 ... 1

The line matrix H is exactly as before, pointing to pairs of columns in G to

connect with lines. Any other grouping matrices for objects, etc., are also
unchanged.

Image manipulations are done by a 4 × 4 matrix A. To ensure that the

fourth coordinate remains a 1, the operator A must have the structure
Equation:
 a 11 a 12 a 13 a 14

a 21 a 22 a 23 a 24
A = .
a 31 a 32 a 33 a 34

0 0 0 1

The new image has point matrix

Equation:

G new = AG.

Exercise:
Problem:

If the coordinates of the i point in G are (x , y , z , 1), what are the

th
i i i

coordinates of the i point in G

th
= AG when A is as given in
new

Equation 2?
Exercise:
Problem:

Write down the point matrix G for the unit cube (the cube with sides
of length 1, with one corner at the origin and extending in the positive
direction along each axis). Draw a sketch of the cube, numbering the
vertices according to their order in your point matrix. Now write down
the line matrix H to complete the representation of the cube.

Left- and Right-Handed Coordinate Systems. In this book we work

exclusively with right-handed coordinate systems. However, it is worth
pointing out that there are two ways to arrange the axes in three dimensions.
Figure 1(a) shows the usual right-handed coordinates, and the left-handed
variation is shown in Figure 1(b). All possible choices of labels x, y, and z
for the three coordinate axes can be rotated to fit one of these two figures,
but no rotation will go from one to the other.
 Three-Dimensional Coordinate Systems; (a)
Right-handed, and (b) Left-handed

Be careful to sketch a right-handed coordinate system when you are solving

problems in this chapter. Some answers will not be the same for a left-
handed system.

Image Transformation. Three-dimensional operations are a little more

complicated than their two-dimensional counterparts. For scaling and
translation we now have three independent directions, so we generalize the
operators of Equation 10 from "Vector Graphics: Homogeneous
Coordinates" as
Equation:

sx 0 0 0

0 sy 0 0
S (s x , s y , s z ) =
0 0 sz 0

0 0 0 1

1 0 0 tx

0 1 0 ty
T (t x , t y , t z ) = .
0 0 1 tz

0 0 0 1

Exercise:
 Problem:

Show that T(−t , −t , −t

x y z) is the inverse of T (t x, t y , t z ). T
−1

undoes the work of T .

Rotation may be done about any arbitrary line in three dimensions. We will
build up to the general case by first presenting the operators that rotate
about the three coordinate axes. R (θ) rotates by angle θ about the x-axis,
x

with positive θ going from the y-axis to the z-axis, as shown in Figure 2. In
a similar fashion, positive rotation about the y-axis using R (θ) goes from y

z to x, and positive rotation about the z-axis goes from x to y, just as in two

dimensions. We have the fundamental rotations

Equation:

1 0 0 0

0 cosθ −sinθ 0
R x (θ) =
0 sinθ cosθ 0

0 0 0 1

cosθ 0 sinθ 0

0 1 0 0
R y (θ) =
−sinθ 0 cosθ 0

0 0 0 1

cosθ −sinθ 0 0

sinθ cosθ 0 0
R z (θ) = .
0 0 1 0

0 0 0 1

A more general rotation about any line through the origin can be
constructed by composition of the three fundamental rotations. Finally, by
composing translation with the fundamental rotations, we can build an
operator that rotates about any arbitrary line in space.
 Directions of Positive Rotation

Example:
To rotate by angle φ about the line L , which lies in the x − y plane in
Figure 3, we would

i. rotate L to the x-axis with R (−θ);

z

ii. rotate by φ about the x-axis with R (φ); and

x

iii. rotate back to L with R (θ).

z

The composite operation would be

Equation:

A (θ, φ) = R z (θ)R x (φ)R z (−θ).

 Composition of Rotations

Direction Cosines. As discussed in the chapter on Linear Algebra, a vector

v may be specified by its coordinates (x, y, z) or by its length and direction.

The length of v is ||v||, and the direction can be specified in terms of the
three direction cosines ( cos θ , cos θ , cos θ ). The angle θ is measured
x y z x

between the vector v and the x-axis or, equivalently, between the vector v
and the vector e =[100] . We have
x
T

Equation:

(v, e x ) x
cos θ x = =
||v||||e x || ||v||

Likewise, θ is measured between v and e

y y
= [010]
T
, and θ is measured
z

between v and e = [001] . Thus

z
T

Equation:
 y
cos θ y =
||v||

z
cos θz =
||v||

The vector
Equation:

cos θ x

u = cos θ y

cos θ z

is a unit vector in the direction of v, so we have

Equation:

cos θ x

v = ||v||u = ||v|| cos θ y .

cos θ z

Exercise:

Problem: Show that u is a unit vector (i.e. 1 u|| = 1).

The direction cosines are useful for specifying a line in space. Instead of
giving two points P and P on the line, we can give one point P plus the
1 2 1

direction cosines of any vector that points along the line. One such vector is
the vector from P to P .
1 2

x
Arc tangents. Consider a vector v = { } in two dimensions. We know
y

that
Equation:
y
tan θ =
x
for the angle θ shown in Figure 4. If we know x and y, we can find θ using
the arc tangent function
Equation:

−1
y
θ =tan ( ).
x

Tangent and Arc Tangent

In MATLAB,

theta = atan(y/x)
−π π
Unfortunately, the arc tangent always gives answers between and
2 2
corresponding to points v in quadrants I and IV. The problem is that the
ratio−y

−x
is the same as the ratio − − −xA so quadrant III cannot be
distinguished from quadrant I by the ratio Likewise, quadrants II and IV
y

are indistinguishable.

The solution is the two-argument arc tangent function. In MATLAB,

theta = atan2(y,x)
will give the true angle θ −π and π in any of the four quadrants.

Example:
cosθ x

Consider the direction vector u = cosθ y , as shown in Figure 5. What

cosθ z

is the angle φ between the projection of u into the x − y plane and the y-
y

axis? This is important because it is R (φ ) that will put u in the y − z

z y

plane, and we need to know the angle φ in order to carry out this rotation.
y

First note that it is not the same as θ . From the geometry of the figure, we
y

can write
Equation:

cos θ x
tan φ y =
cos θ y

This gives us a formula for φ in terms of the direction cosines of u. With

y

the two-argument arc tangent, we can write

Equation:
−1
φ y =tan (cos θ x , cos θ y ).
 Angles in Three Dimensions

Exercise:
Problem:

a. Suppose point p′ is in the y − z plane in three dimensions,

p = (0, y , , 1). Find θ so that R (θ) will rotate p′ to the
′ ′ ′′
x

positive z-axis. (Hint: Use the two-argument arc tangent. θ will be

a function of y′ and z .) ′

b. Let p be any point in three-dimensional space, p = (x, y, z, 1).

Find ψ so that R (ψ) rotates p into the y − z plane. (Hint: Sketch
z

the situation in three-dimensions, then sketch a two-dimensional

view looking down at the x − y plane from the positive z-axis.
Compare with Example 2.) Your answers to parts (a) and (b) can
be composed into an operator Z(p) that rotates a given point p to
the positive z-axis, Z (p) = R (θ)R (ψ).
x z

c. Let L be a line in three-dimensional space specified by a point 1

=(x, y, z, 1) and the direction cosines ( cos θ , cos θ , cos θ ).
x y z

Use the following procedure to derive a composite operator

R(φ, L ) that rotates by angle φ about the line L :

i. translate 1 to the origin;

ii. let u =( cos θ , cos θ , cos θ
x y z, 1 ) and use Z(u) to align L
with the z-axis;
iii. rotate by φ about the z-axis;
iv. undo step (ii); and
v. undo step (i).
Vector Graphics: Projections

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Computer screens and printers are two-dimensional display devices. We

must somehow convert three-dimensional images to two dimensions in
order to display them. This task is done by another kind of matrix operator
called a projection.

To build a projection, we first choose a projection plane in the three-

dimensional space of the object we wish to view. All points of the object are
then projected onto the plane. There are many different kinds of projections,
corresponding to various geometric rules for mapping points in space onto a
plane. We begin with the parallel projection illustrated in Figure 1, wherein
the dotted lines between the points and their projections in the plane are all
parallel to one another. These dotted lines are called projectors.

Range. The projection plane is called the range of the projection. We will
assume for now that the projection plane passes through the origin. If it
does not, we may later compose the necessary translations with our
projection. Three points, not all in a line, are required to determine a plane.
We will take the origin as one of the points and suppose that the other two
points are the vectors r and r . From these vectors, we form the matrix
1 2

Equation:

R = [r 1 r 2 ],

which determines the range of the projection.

 Oblique Parallel Projection

Null-Space. To complete the specification of the parallel projection, we

must specify the direction of the projectors. The line through the origin in
this direction is called the null-space of the projection. Note that any line in
the three-dimensional object that is parallel to the null-space will disappear
to a point when projected. The null-space may be specified by a vector n
pointing in the direction of the projectors.

Orthogonal Projection. If the null-space is perpendicular to the range, we

say that the projection is an orthogonal projection. The matrix operator for
orthogonal projection is a function of the range. If we assume that R is
specified in three-dimensional coordinates (not homogeneous!), we have
the following definition for the orthogonal projection onto the range of R:
Equation:
−1
T T 3×3
P (R) = R(R R) R ∈ R .

In homogeneous coordinates, we embed this 3 × 3 projection matrix in the

general 4 × 4 transformation:
Equation:

P(R) 0
P h (R) = [ ].
T
0 1

Example:
Suppose we choose the x − y plane as our projection plane. Vectors r and 1

r can be any two points in the plane (besides the origin), so let's take
2

Equation:

1 1

r1 = 0 , r2 = 1

0 0

Then the range matrix is

Equation:

1 1

R = 0 1 .

0 0

The orthogonal projection looking down on the x − y plane from the z-

axis is
Equation:
 −1
1 1 1 1
1 0 0 1 0 0
P (R) = 0 1 [ ] 0 1 [ ]
1 1 0 1 1 0
0 0 0 0

1 0 0

= 0 1 0 .

0 0 0

In homogeneous coordinates,
Equation:

1 0 0 0

0 1 0 0
P h (R) = ⋅
0 0 0 0

0 0 0 1

Exercise:
Problem:

2 0

Let the vectors r 1 = 0 and r 2 = 3 specify the range of an

0 0

orthogonal projection. Find P(R) and P h (R) . Compare with Example

1 and explain any similarities.

Example:
In the chapter on linear algebra you learned that the projection of w onto x
is given by
Equation:
 (x, w)x
z =
(x, x)

This is an orthogonal projection of w onto a line, but it is closely related to

the projection into the plane just described. To see the similarity, let's work
on the expression for z a little:
Equation:
x(x,w)
z =
(x,x)

−1
= x(x, x) (x, w)

−1
T T
= x(x x) (x w)

−1
T T
= [x(x x) x ]w

= P(x)w.

So we see that it can be written in the same way as the projection onto the
plane. The only difference is that the range is now one-dimensional and
specified by x in place of R.

Oblique Projection. More generally, we may have a null-space n that is

not perpendicular to the range R. The projection shown in Figure 1 is an
oblique projection. Once again we start with nonhomogeneous coordinates
for n and R and write the 3 × 3 oblique projection as
Equation:
T −1 T
E (R, n) = R{R [I − −P (n)]R} R [I − −P (n)]

where
Equation:
T
−l nn
T T
P (n) = n(n n) n =
2
||n||
As with the orthogonal projection, we can return to homogeneous
coordinates by
Equation:

E(R n) 0
E h (R, n) = [ ].
T
O 1

Exercise:
Problem:

Prove and interpret the following properties of paralle1 projections

(both orthogonal and oblique):

a. P = P; E = E.
2 2

b. P R = R; ER = R.
c. En = 0; P n = 0. (First show that R T
n = 0 when n is
orthogonal to R.)

Display from Projections. Even after we have used a projection, the image
points are in three-dimensional homogeneous coordinates. How then do we
display them on a two-dimensional display? If we had chosen the x − y
plane as the range of our projection, we could let the output device
represent the x − y plane and ignore the z-coordinate of each point. The z-
coordinates of the projected image would all be 0 anyway since the
projected points would lie in the x − y plane. The fourth coordinate of 1
may also be ignored for display. But even if the projection plane is not the
x − y plane, we may perform a rotation to align the two planes and again

let the output device represent the x − y plane.

Perspective Projections. To obtain a perspective projection, we first

choose a projection plane as we did for parallel projections. Instead of
choosing a null-space parallel to all projectors, we now choose a viewpoint.
The projectors all pass through the viewpoint in a perspective projection, as
shown in Figure 2. For a viewpoint at z = d on the z-axis and a projection
plane coinciding with the x − y plane, the three-dimensional homogeneous
operator for perspective projection is
Equation:

1 0 0 0

0 1 0 0
M (d) = .
0 0 0 0

−1
0 0 1
d

The first thing you should notice about the perspective projection M(d) is
that it violates the structure given in Equation 2 from "Vector Graphics:
Three-Dimensional Homogeneous Coordinates" by having the 4,3 position
1
equal to − rather than to 0. This means that the fourth coordinates will
d
not remain 1 in the new point matrix
Equation:

G new = M (d)G.
 A Perspective Projection

This situation is interpreted to mean that the new point (x, y, z, w) must be
x y z
renormalized to ( , 1) before the operation is considered complete.
w w w
Renormalization adds another computational step to the image
transformation process. This is undesirable, but we are stuck with it if we
wish to include perspective projections in our repertoire of transformations.
Note that renormalization is a point-by-point process since w may be
different for each point.
How does renormalization affect composition of operators? We might
expect that, each time we perform an operation requiring renormalization,
we will be forced to stop and do the renormalization before going on with
other operations. In this respect we are fortunate: we may put off the
renormalization until all transformations are complete and renormalize once
just before displaying the image. Thus we are able to compose perspective
projections at will with other transformations.
Exercise:
Problem:

What is the perspective projection matrix for a viewpoint at infinity,

M(∞)? Interpret the result.

Generalization of Projections. “Projection” is a technical term in linear

algebra. A square matrix of any size is considered a projection if it obeys
the property of Exercise 2(a), P = P. All of our geometric projection
2

matrices have this property. We have considered only three-dimensional

spaces with a two-dimensional range and a one-dimensional null-space. In
general, the dimensionality of the space may be split in any way between
range and null-space. If we reinterpret homogeneous coordinates as four-
dimensional space, we have projections with three-dimensional range and
one-dimensional null-space, but the perspective projection is actually not
quite a projection in the technical sense even though M (d) = M (d)
2

because perspective projection also includes renormalization.

Vector Graphics: Numerical Experiment (Star Field)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

With Earth at the origin, we may specify the star positions for the Big and
Little Dippers in three-dimensional homogeneous coordinates. With light
years as our units, we have G =
T

1.5441 -1.2064 153.0875 1.0000

-1.0386 8.4588 142.7458 1.0000
-8.7955 26.2870 198.0698 1.0000
-12.8807 19.0964 106.5383 1.0000
-18.8926 17.461 90.6185 1.0000
-45.1364 54.2706 215.1148 1.0000
-9.6222 20.1734 88.0062 1.0000
-33.7097 -8.4048 64.6574 1.0000
-33.7144 -8.7191 52.3806 1.0000
-43.8531 -1.1483 59.7193 1.0000
-36.1086 2.3667 55.7927 1.0000
-34.2309 8.2181 52.1260 1.0000
-30.7876 11.8182 46.9810 1.0000
-61.8581 31.5183 80.7615 1.0000

To make use of this data, we need a function to display it. Enter and save
the following generalization of the function from Demo 1 in "Vector
Graphics: Two Dimensional Image Representation". Call it vhgraph.m.

function vhgraph (P,L,A,ps,ls);

% vhgraph(P,L,A,PS,LS) graphs images whose points
are
% stored in P and whose lines are stored in L. The
points
% in P must be in homogeneous coordinates in eithe
r 2 or 3
% dimensions, with each column of P representing a
point.
% The lines are coded in L, with each column of L
containing
% 2 integers pointing to a pair of points in P to
be
% connected by a line segment. If A is present, th
e points
% in P are transformed to A*P beforre graphing. Fo
r 3D data
% points, only the first two coordinates are graph
ed, so A
% should include the desired projection from 3D to
2D. The
% point symbol may be specified in PS and the line
type in
% LS, if desired
%
% Richard T. Behrens, October 1989.
%
% The first dection of the program determines the
sizes of
% all the input matrices and chacks that they make
sense.
[mp,np] = size(P);
if (nargin > 1)
[ml,nl] = size(L);
else
ml = 2; nl = 0;
end
if (nargin <=2)
A = eye(mp);
end
[mA,nA] = size(A);
if ((mp~= nA) | (ml~=2))
error('Incompatible sizes of input matrices.")
end
if (nargin <= 3);
ps = '*';
end

P = A*P; % Performs the transformation A on the

points
% (effect is only local to this functi
on).

% The next section contains a loop that renormaliz

es the
% homogeneous coordinates of the points if necessa
ry.
renorm = find((P(mA,:)~= 1));
if~isempty(renorm)
for i = 1:length(renorm)
P(:,renorm(i)) = P(:,renorm(i))/P(mA,renor
m(i));
end
end

% The next program line sets a fixed scale output

window
% from -50 to 50 in both x and y directions on the
screen.
% For automatic scaling to include all points of t
he
% image, we could use instead the line q = min(min
(P));
% r = max(max(P));
q = -50; r = 50;
axis([q r q r])
axis('square')
plot(P(1,:),P(2,:),ps) % Plots the points with sy
mbol ps.
hold on % Saves the points while w
e plot
for i = 1:nl % lines with line type LS.
plot([P(1,L(1,i))P(1,L(2,i))],..
[P(2,L(1,i))P(2,L(2,i))],ls)
end
hold off

Enter the point matrix given at rhe beginning of this section (and take its
transpose to put it in the usual form). Also enter the line matrix from Demo
1 in "Vector Graphics: Two Dimensional Image Representation". Save these
two matrices and try looking at the image

≫ save dip3d
≫ vhgraph(G,H)

No dippers in sight, right? Without specifying a transformation matrix A,

we have defaulted to looking down on the x − y plane from z = ∞ (a
parallel projection). This is how the constellations would look from a
distant galaxy (say, a billion light years north of here) through an enormous
telescope. We need a perspective view from the origin (Earth), but first we
need a set of functions to give us the fundamental operators with which we
can build the desired projection.

Take R y (θ) as an example. The function to build it looks like

function Ry = vhry (theta) ;

% Rotation matrix about y-axis for 3-D homogeneous
% coordinates.
Ry = eye(4);
Ry(1,1) = cos(theta);
Ry(3,3) = cos(theta);
Ry(3,1) = -sin(theta);
Ry(1,3) = sin(theta);

Enter and save hry.m as given. Then write functions for

Equation:

Rx (θ) vhrx. m

Rz (θ) vhrz. m

S(sx, sy, sz) vhs. m

T(tx, ty, tz) vht. m

M(d) vhm. m

Useful MATLAB functions for this task include zeros, eye, and diag.

Now build and use a perspective projection with viewpoint Earth and
projection plane at z = 1000 behind the dippers:

1. translate Earth to z = −1000 so that the projection plane coincides

with the x − y plane: T(O, 0, −1000);
2. use the fundamental perspective projection: M(−1000); and
3. translate back: T(O, 0, 1000).

≫ A = vht(0,0,1000) * vhm(-1000) * vht(0,0,-1000)

≫ vhgraph(G,H,A)

Oops! Now the image is too big; it's mostly off the screen. Scale it down
and have another look:

≫ A = vhs(.06,.06,.06) * A
≫ vhgraph(G,H,A)

Now the view should look familiar. Leave A as it is now:

≫ save dip3d

Experiment with scale and rotation about the z-axis. For example, try

≫ vhgraph(G,H,vhrz(pi/2) * A)
The two-dimensional star positions given in Demo 1 in "Vector Graphics:
Two Dimensional Image Representation" were obtained from the three-
dimensional positions with the composite operator A you are now using. To
compare the two, type

≫ Gnew = A * G

≫ for i = 1:14
Gnew(;,i) = Gnew(:,i)/Gnew(4,i);
end

≫ Gnew

and compare the x and y coordinates with those of Demo 1 in "Vector

Graphics: Two Dimensional Image Representation".

Astronomers give star positions in equatorial coordinates using right

ascension, declination, and distance. The following function converts
equatorial coordinates, which are spherical, to Cartesian coordinates with
the z-axis pointing north, the x-axis pointing at the vernal (Spring) equinox
in the constellation Pisces, and the y-axis pointing toward the Winter
solstice in the constellation Opheuchus.

function v = starxyz(rah,ram,decd,decm,dist)
% starxyz is the cartesian coordinates of a star w
hose
% spherical coordinates ( e.g. from a star catalog
) are
%
% rah right ascension hours
% ram right ascension minutes
% decd declination degrees
% decm declination minutes (should be
negative
% if decd is negative)
% dist distance (light years)
%
phi = (pi/180) * (decd + decm/60);
theta = (pi/12) * (rah + ram/60);
r = dist;
v = [r * cos(phi) * cos(theta); -
r * cos(phi) * sin(theta);
r * sin(phi)];

Have you ever wondered what the constellations would look like from other
places in the galaxy? We will soon see the answer. First we will view the
dippers from Alpha Centauri, the nearest star, whose coordinates are

-1.5680 1.3157 -3.6675.

We will look toward the centroid of the fourteen stars in the dippers, located
at

-26.3632 12.8709 100.4714.

To get the desired view, we must

1. translate the viewpoint to the origin;

2. rotate the centroid (direction to look) to the z-axis–note that the
centroid will have new coordinates after step (1); and
3. apply the composite
A=S(.06,.06,.06)*T(0,0,1000)*M(-1000)*T(0,0,-10
00) (as used to view from Earth).

Write a function vhz.m based on Exercise 5 from "Vector Graphics: Three-

Dimensional Homogeneous Coordinates" to accomplish step (2). Test it on
several random points to make sure it works right. Now write a general
perspective projection function called vhview.m. The function vhview
should accept as inputs two vectors, the first specifying the viewpoint and
the second the point to look toward. Its output should be a composite
operator that performs all three of the preceding steps.

Now we want to look toward the centroid of the dippers from Alpha
Centauri. To do so, enter the vectors for the two points of interest and
construct the view like this:
≫ acentauri = [-1.5680; 1.3157; -3.6675]
≫ centroid = [-26.3632; 12.8709; 100.4714]
≫ A = vhview(acentauri,centroid)
≫ vhgraph(G,H,A)
≫ title('From Alpha Centauri')

The farther we move from Earth, the more distorted the dippers will look in
general. It should be easy now to view them from any desired location. Just
choose a viewpoint, recalculate the composite operator for that viewpoint
using vhview, and use vhgraph. Follow this procedure to view the
dippers from each of the stars in the following list. You will need to use
starxyz first to convert their coordinates.

Right Distance
Star Ascension Declination (ly)

Alpha
14h 40m -60° 50' 4.2
Centauri

Sirius 6h 45m -16°43' 9.5

Arcturus 14h 16m 19°11' 16.6

Pollux 7h 45m 28°02' 35.9

Betelgeuse 5h 55m 7°24' 313.5

Table 1 is a six-row and four-column table with the first row as a header,
the first column describing different stars, and the values in the second
through fourth columns describing the corresponding right ascension,
declination, and distance for each star.
Of course, star viewing is not the only application of vector graphics. Do
some experiments with the unit cube (see Exercise 2 from "Vector
Graphics: Three-Dimensional Homogeneous Coordinates"). View the cube
from location (4,3,2) looking toward the origin using the procedure just
outlined for stars. You may need to adjust the scaling to get a meaningful
view.
Filtering: Introduction

Note:This module is part of the collection, A First Course in Electrical and Computer Engineering.
The LaTeX source files for this collection were created using an optical character recognition
technology, and because of this process there may be more errors than usual. Please contact us if you
discover any errors.

Notes to Teachers and Students:

Filtering is one of the most important things that electrical and computer engineers do. In this chapter
we extend everyday understanding of filters to numerical filters. We then study weighted moving
averages and exponential averages. We define the important test signals for electrical and computer
engineering and show how filters respond to them. The idea that filters are characterized by their
response to simple test signals is fundamental. In the numerical experiment, students explore the
frequency response of a simple filter, a concept that forms the basis of circuit theory, electronics, optics
and lasers, solid-state devices, communications, and control.

Introduction
A filter is any device that passes material, light, sound, current, velocity, or information according to
some rule of selectivity. Material (or mechanical) filters are commonplace in your everyday life:

i. coffee filters pass flavored water while filtering out coffee grounds;
ii. Goretex fibers pass small, warm perspiration droplets while filtering out large, cool droplets of
rain or snow;
iii. fiberglass strands in a furnace filter pass warm air while filtering out particles of dirt and dust;
iv. a centrifuge retains material of low density while spinning out (or filtering out) material of high
density; and
v. an electrostatic precipitator filters out dust and other effluents by attaching charge to them and
using an electric field to move the charged particles to a high potential drain.

The first three of these examples selectively pass material according to size; the last two selectively
pass material according to its mass density.

Typical filters for light are

i. UV filters on camera lenses and eyeglasses that pass light in the range of visible wavelengths
while blocking light in the invisible (but damaging) ultraviolet range;
ii. polaroid lenses that pass light that is randomly polarized while blocking out glare that is linearly
polarized;
iii. green fabrics that reflect green light and absorb other colors;
iv. red taillights that pass light in the long wavelength red range and reflect light in the short
wavelength violet range (look at the inside of your taillights to see violet); and
v. glacial ice that absorbs all but the blue wavelengths so that it appears blue.

Exercise:
 Problem: List as many examples of natural and man-made sound filters as you can.

Satellite Television. Among current filters, the tuner in a super-heterodyne receiver is, perhaps, the
first example that comes to mind. But satellite TV filters are another fascinating example. A typical C-
band satellite has twelve transponders (or repeaters), each of which transmits microwave radiation in a
personalized 36 MHz band. (The abbreviation MHz stands for megahertz, or 10 Hz, or 10 cycles per
6 6

second. Other common abbreviations are Hz for 1 Hz, kHz for 10 Hz, and GHz for 10 Hz.) Each
3 9

transponder actually transmits two channels of information, one vertically polarized and one
horizontally polarized. There is an 8 MHz guard band between each band, and the vertical and
horizontal channels are offset by 20 MHz. The transmission scheme for the 24 channels is illustrated in
Figure 1. The entire transmission band extends over 540 MHz, from 3. 7 × 10 Hz to 4. 24 × 10 Hz.
9 9

The satellite receiver has two different microwave detectors, one for vertical and one for horizontal
polarization, and a microwave tuner to tune into the microwave band of interest.
Exercise:

Problem: Check that the transmission scheme of Figure 1 consumes 540 MHz of bandwidth.
Exercise:
Problem:

List as many examples of natural and man-made devices for velocity filtering as you can.

Satellite TV– Vi, Vertically Polarized Channel i; H ,

j

Horizontally Polarized Channel j

An Aside on Hertz and Seconds. The abbreviation Hz stands for hertz, or cycles/second. It is used to
describe the frequency of a sinusoidal signal. For example, house current is 60 Hz, meaning that it has
60 cycles each second. The inverse of Hz is seconds or, more precisely, seconds/cycle, the period of 1
cycle. For example, the period of 1 cycle for house current is 1/60 second. When we are dealing with
sound, electricity, and electromagnetic radiation, we need a concise language for dealing with signals
and waves whose frequencies range from 0 Hz (called DC or direct current) to 10 Hz (visible light).
18

Table 1 summarizes the terms and symbols used to describe the frequency and period of signals that
range in frequency from 0 Hz to 10 Hz.
12
 Frequency Period

Hz Term Units Seconds Term Units Example

battery current:
0 Hz
Hz hertz 1 Hz sec second 1 sec
house current:
60 Hz

103 10-3 midfrequency

kHz kilohertz msec millisecond
Hz sec sound

clock
106
MHz megahertz µsec microsecond 10-6sec frequencies in
Hz
microcomputers

microwave
109 radiation for
GHz gigahertz nsec nanosecond 10-9sec
Hz satellite
communication

1012 10-12 infrared

THz terahertz psec picosecond
Hz sec radiation

Terms and symbols for Sinusoidal Signals

Numerical Filters. Rather amazingly, these ideas extend to the domain of numerical filters, the topic of
this chapter. Numerical filters are just schemes for weighting and summing strings of numbers. Stock
prices are typically averaged with numerical filters. The curves in Figure 2 illustrate the daily closing
average for Kellogg's common stock and two moving averages. The 50-day moving average is obtained
by passing the daily closing average through a numerical filter that averages the most current 50 days'
worth of closing averages. The 200-day moving average for the stock price is obtained by passing the
daily closing prices through a numerical filter that averages the most current 200 days' worth of daily
closing averages. The daily closing averages show fine-grained variation but tend to conceal trends.
The 50-day and 200-day averages show less fine-grained variation but give a clearer picture of trends.
In fact, this is one of the key ideas in numerical filtering: by selecting our method of averaging, we can
filter out fine-grained variations and pass long-term trends (or vice versa), or we can filter out periodic
variations and pass nonperiodic variations (or vice versa). Figure 2 illustrates that moving averages
typically lag increasing sequences of numbers and lead decreasing sequences. Can you explain why?

We will call any algorithm or procedure for transforming one set of numbers into another set of
numbers a numerical filter or digital filter. Digital filters, consisting of memories and arithmetic logic
units (ALUs), are implemented in VLSI circuits and used for communication, control, and
instrumentation. They are also implemented in random–or semicustom–logic circuits and in
programmable microcomputer systems. The inputs to a digital filter are typically electronic
measurements that are produced by A/D (analog-to-digital) conversion of the output of an electrical or
mechanical sensor. The outputs of the filter are “processed,” “filtered,” or “smoothed” versions of the
measurements. In your more advanced courses in electrical and computer engineering you will study
signal processing and system theory, assembly language programming, microprocessor system
development, and computer design. In these courses you will study the design and programming of
hardware that may be used for digital filtering.

Dow-Jones Averages (Adapted from the New York Stock Exchange, Daily Graphs,
William O'Neil and Co., Inc., Los Angeles, California)
Filtering: Simple Averages

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The simplest numerical filter is the simple averaging filter. This filter is
defined by the equation
Equation:
N
1
x = ∑ un .
N
n=1

The filter output x is the average of the N filter inputs u , u , ... , uN. These
1 2

inputs may be real or complex numbers, and x may be real or complex. This
simple averaging filter is illustrated in Figure 1.

A Simple Averaging Filter

Example:
If the averaging filter is excited by the constant sequence
u = u = ⋯ = u
1 2 N = u, then the output is

Equation:
N
1
x = ∑ u = u.
N
n=1

The output is, truly, the average of the inputs. Now suppose the filter is
excited by the linearly increasing sequence
Equation:

u n = n, n = 1, 2, ..., N .

This sequence is plotted in Figure 2. How do we sum such a sequence in

order to produce the average x? For N even, the average may be written as
Equation:

1 1 1
x = (x 1 + x N ) + (x 2 + x N −1 ) + ⋯ + (x N /2 + x (N /2)+1 ).
N N N

N
Each pair-sum in parentheses equals N + 1 , and there are such pair-
2
sums, so the average is
Equation:

1 N N + 1
x = (N + 1) = .
N 2 2

This is certainly a reasonable answer for the average of a linearly

increasing sequence. See Figure 2.
 Linearly Increasing Sequence

Exercise:
Problem:
N
1
Write x = ∑n as a sum of pair-sums for N odd. What does x
N
n=1

equal?

General Sum Formula. Suppose the input to the simple averaging filter is
the polynomial sequence
Equation:
k
un = n , n = 1, 2, ..., N
where k is a non-negative integer such as k = 0, 1, 2, .... The output of the
filter is
Equation:
N
(k) 1 k
x = ∑n .
N
N
n=1

(k)
We rewrite x as x to remind ourselves that we are averaging N numbers,
N

each of which is n . For example, when N = 8 and k = 2,

k

Equation:
8
(2) 1 2
1
x = ∑n = (1 + 4 + 9 + ⋯ + 64).
8
8 8
n=1

(k) (k)
Rather than study the average x N
, we will study the sum N x N
and divide
by N at the very end:
Equation:
N
(k) (k) k
S = Nx = ∑n .
N N

n=1

(k)
The sum S N
may be rewritten as the sum
Equation:
(k) N −1 k k
S = ∑ n + N
N n=1

(k) k
= S + N .
N −1

(k)
This result is very important because it tells us that the sum S N
, viewed as
(k)
a function of N , obeys a recursion in which S N
is just the sum using one
 (k)
less input, namely, S N −1
, plus N . Now, since polynomials are the most
k

(k)
general functions that obey such recursions, we know that s N
must be a
polynomial of order k + 1 in the variable N :
Equation:
(k) 2 k+1
s = a0 + a1 N + a2 N + ⋯ + a k+1 N .
N

Let's check to see that this polynomial really can obey the required
(k)
recursion. First note that S is the following polynomial:
N −1

Equation:
(k) k+1
s = a 0 + a 1 (N − 1) + ⋯ + a k+1 (N − 1) .
N −1

The term (N − 1) produces ( )N

k+1
(−1) + (
k+1

0
)N (−1) + ⋯
k+1 0 k+1

1
k 1

. (Remember the binomial expansion?) Therefore the difference between

(k) (k)
S
N
and S is
N −1

Equation:
(k) (k) k
S − S = c0 + c1 N + ⋯ + ck N .
N N −1

This recursion is general enough to produce the difference Nk provided we

can solve for a , a , ... , a
0 1 to make c = c = ⋯ = c
k+1 = 0 and
0 1 k−1
(k)
c = 1. We know that S
k = 0 for N = 0, so we know that a = 0, 0
N
(k)
meaning that the polynomial for S N
can really be written as
Equation:
(k) 2 k+1
S = a1 N + a2 N + ⋯ + a k+1 N .
N

In order to solve for the coefficients of this polynomial, we propose to write

(k)
out our equation for s as follows:
N

Equation:
( )
 (k)
(N = 1) S = a 1 + ⋯ + a k+1
1

(k) k+1
(N = 2) S = 2a 1 + ⋯ + 2 a k+1
2

(k) k+1
(N = 3) s = 3a 1 + ⋯ + 3 a k+1
3

⋮ ⋮

(k) k+1
(N = k) S = ka 1 + ⋯ + k a k+1
k

(k) k+1
(N = k + 1) S = (k + 1)a 1 + ⋯ + (k + 1) a k+1 .
k+1

Using the linear algebra we learned earlier, we may write these equations as
the matrix equation
Equation:
(k)
l 1 ... 1 ⎡s ⎤
⎡ ⎤ ⎡a 1 ⎤
1

k+1 (k)
2 4 ... 2 a2 S
2

⋮ ⋮ ⋮ = .
⋮ ⋮
2 k+1
k k ... k

⎣ 2 k+1 ⎦⎣ ⎦
(k + 1) (k + 1) ... (k + 1) a k+1 (k)
⎣S ⎦
k+1

The terms on the right-hand side of the equal sign are “initial conditions”
(k)
that tell us how the sum S begins for N = 1, 2, ..., k + 1. These initial
N
(k)
conditions must be computed directly. (For example, S = 1 + 2 .) 2
k k

Then the linear system of (k + 1) equations in (k + 1) unknowns may be

solved for a , a , ..., a . The solution for S is then complete, and we
1 2 k+1
k
N

may use it to solve for S for arbitrary N .

k
N

Example:
When k = 2, we have the following equation for the coefficients a 1, a2 ,
(2)
and a in the polynomial S = a N + a N + a N :
3 N 1 2
2
3
3
Equation:

Exercise:
Problem:

Solve for a
S

S
(2)

N
(2)

Exercise:
Problem:
− S

⎢⎥
⎡

1,

= a1 N + a2
(2)

N −1
1

3
1

9
1

27

a2 , a3

= N
N
⎦⎣
a
⎤⎡ 1 ⎤

a2

a3
⎦
=
⎡

⎣
1
2
1

+2
2

in the linear equation of Example 2. Show that

2
.
2
+ a N 3obeys the recursion
3

(MATLAB) Write a MATLAB program to determine the coefficients

a , a , ..., a
1 2 for the polynomial S . Generate a table of formulas
k+1

(k)
for the averages x for k = 1, 2, ..., 5. Evaluate these formulas for
N = 2, 4, 8, and 16.
N
(k)

Exponential Sums. When the input to an averaging filter is the sequence

Equation:

un = a
n
, n = 0, 1, 2, ..., N − 1,

we say that the input is exponential (or geometric). Typical sequences are
illustrated in Figure 6.5 for a = 0. 9, a = 1, and a = 1. 1. Don't let it throw
you that we have changed the index to run from 0 to N − 1 rather than
from 1 to N . This change is not fundamentally important, but it simplifies
our study. The sum of the inputs is
Equation:
+ 2

2
2

+ 3
⎤

2⎦
=
⎡

⎣
1

14
⎦
⎤

⋅
 N −1

n
SN = ∑ a .

n=0

Exponential Sequences

How do we evaluate this sum? Well, we note that the sum aS is N

Equation:
N −1 n+1 N k
aS N = ∑ a = ∑ a
n=0 k=1

N −1 N
= ∑ ak + a − 1
k=0

N
= SN + a − 1.

Therefore, provided a ≠ 1, the sum S is N

Equation:
N
1 − a
SN = , a ≠ 1.
1 − a
This formula, discovered already in the chapter covering the functions e x

and e , works for a ≠ 1. When a = 1, then S = N :

jθ
N

Equation:
N
1−a
, a ≠ 1
1−a
SN = {
N, a = 1.

When |a| < 1, then a N

→ 0 for N → ∞ , and we have the asymptotic
formula
Equation:
1
lim SN = , |a| < 1.
N →∞ 1 − a

Exercise:
Problem:
N −1
1
Evaluate S N = ∑ a
n
and X N = SN for a = 0. 9, 1, and 1.1 and
N
n=0

for N = 1, 2, 4, 8, 16 , and 32.

Exercise:

N −1

Problem: Prove that S N = ∑ a

n
obeys the recursion
n=0

Equation:
N −1
S N = S N −1 + a .

Prove that S N = N obeys this recursion for a = 1 and that

N
1 − a
SN = obeys it for a ≠ 1.
1 − a
Recursive Computation. Every sum of the form
Equation:
N −1

SN = ∑ un

n=0

obeys the recursion

Equation:

S N = S N −1 + u N −1 .

This means that when summing numbers you may “use them and discard
them.” That is, you do not need to read them, store them, and sum them.

You may read u to form S and discard u ; add u to S and discard u ;

0 1 0 1 1 1

add u to S ; and continue.

2 2

The Recursion S n+1 = Sn + un

This is very important for hardware and software implementations of
running sums. You need only store the current sum, not the measurements
that produced it. Two illustrations of the recursion S
n+1 = S + u are
n n

provided in Figure 4. The diagram on the left is self-explanatory. The

diagram on the right says that the sum S is stored in a memory location, to
n

be added to u to produce S , which is then stored back in the memory

n n+1

location to be added to u , and so on.

n+1
Filtering: Weighted Averages

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Weighted, tapered, or windowed averages are straightforward

generalizations of simple averages. They take the form
Equation:
N

x = ∑ wn un

n=1

with the constraint that the “weights in the window,” w , sum to 1:

n

Equation:
N

∑ w n = 1.

n=1

1
When w n = then x is the simple average studied in the section on
N
"Simple Averages".

Example:
There are many windows that are commonly used in engineering practice.
For N odd, the standard triangular window is
Equation:
 2

N + 1
wn =
N + 1
2
(1−

and the inputs u and u by (

Equation:

Exercise:
1

triangular window takes the form

w n = α(1 − β|
N

N + 1

Triangular Window
2

N + 1

∣
2

N + 1
N + 1

This window, illustrated in Figure 1, weights the input u

− n |); α, β > 0,
2
− n |).

(N +1)/2
by

The most general

N odd.
 Problem:

Determine the constraints on α and β to make the general triangular

N

window a valid window (i.e., ∑ w n = 1 ). Show that

n=1

2
α = = β is a valid solution. Propose another solution that you
N + 1

like.
Exercise:
Problem:

You are taking three 3-credit courses, one 5-credit course, and one 2-
credit course. Write down the weighted average for computing your
GPA in a system that awards 4.0 points for an A, 3. 0 points for a B, ...
, and (horrors!) 0 points for an F .
Filtering: Moving Averages

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Moving averages are generalizations of weighted averages. They are

designed to “run along an input sequence, computing weighted averages as
they go.” A typical moving average over N inputs takes the form
Equation:
N −1
xn = ∑ w k u n−k
k=0

= w 0 u n + w 1 u n−1 + ⋯ + w N −1 u n−(N −1) .

The most current input, u , is weighted by w ; the next most current input,
n 0

un−1 , is weighted by w ; and so on. This weighting is illustrated in Figure

1

1. The sequence of weights, w through w

0 , is called a “window,” a
N −1

“weighting sequence,” or a “filter.” In the example illustrated in Figure 6.8,

the current value u is weighted more heavily than the least current value.
n

This is typical (but not essential) because we usually want x to reflect

n

more of the recent past than the distant past.

 Moving Average

Example:
1
When the weights w 0
, w 1 , ..., w N −1 are all equal to , then the moving
N
average x is a “simple moving average”:
n

Equation:

1
xn = [u n + u n−1 + ⋯ + u N −1 ].
N

This is the same as the simple average, but now the simple average moves
along the sequence of inputs, averaging the N most current values.

Exercise:
Problem:
N −1
1
Evaluate the moving average x n = u n−k for the inputs
N
k=0
 0, n < 0
a. u n =
u, n ≥ 0;

0, n ≤ 0
b. u n =
n, n > 0.

Interpret your findings.

Exercise:
Problem:
N −1
1
Evaluate the simple moving average x n = u n−k when u isn
N
k=0

the sequence
Equation:

0, n < 0
un =
n
a , n ≥ 0.

Interpret your result.

Example:
When the weights w equal w a for n = 0, 1, ..., N
n 0
n
− 1 , then the
moving average x takes the form
n

N −1

xn = w0
k
a u n−k .
k=0

When a < 1, then un is weighted more heavily than u when a > 1 n−(N −1);

,un−(N −1)
is weighted more heavily than un ; when a = 1, u is weighted n

the same as u .
n−(N −1)

Exercise:
 Problem:

Evaluate w so that the exponential weighting sequence w

0 n =
N −1

w0 a
n
(n = 0, 1, ..., N − 1) is a valid window (i.e., wn = 1 ).
n=0

Exercise:
Problem:
N −1

Compute the moving average x n =

k
w 0 a u n−k when the input
k=0

sequence u is n

Equation:
n
b , n ≥ 0
un =
0, n < 0.

What happens when b = a? Can you explain this?

Filtering: Exponential Averages and Recursive Filters

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Suppose we try to extend our method for computing finite moving averages
to infinite moving averages of the form
Equation:
∞
xn = ∑ w k u n−k
k=0

= w 0 u n + w 1 u n−1 + ⋯ + w 1000 u n−1000 + ⋯

In general, this moving average would require infinite memory for the
weighting coefficients w , w , ... and for the inputs u , u , ....
0 1 n n−1

Furthermore, the hardware for multiplying w u k would have to be

n−k

infinitely fast to compute the infinite moving average in finite time. All of
this is clearly fanciful and implausible (not to mention impossible). But
what if the weights take the exponential form
Equation:

0, k < 0
wk = {
k
w0 a , k ≥ 0?

Does any simplification result? There is hope because the weighting

sequence obeys the recursion
Equation:
 0, k < 0

wk = w0 , k = 0

aw k−1 k ≥ 1.

This recursion may be rewritten as follows, for k ≥ 1:

Equation:

w k − aw k−1 = 0, k ≥ 1.

Let's now manipulate the infinite moving average and use the recursion for
the weights to see what happens. You must follow every step:
Equation:
∞
xn = ∑ w k u n−k
k=0

∞
= ∑ w k u n−k + w 0 u n
k=1

∞
= ∑ aw k−1 u n−k + w 0 u n
k=1

∞
= a∑ w m u n−1−m + w 0 u n
m=0

= ax n−1 + w 0 u n .

This result is fundamentally important because it says that the output of the
infinite exponential moving average may be computed by scaling the
previous output x n−1 by the constant a, scaling the new input u by w ,
n 0

and adding. Only three memory locations must be allocated: one for w , 0

one for a, and one for x . Only two multiplies must be implemented: one
n−1

for axn−1 and one for w u . A diagram of the recursion is given in Figure
0 n

1. In this recursion, the old value of the exponential moving average, x , n−1

is scaled by a and added to w u to produce the new exponential moving

0 n

average x . This new value is stored in memory, where it becomes x

n in
n−1

the next step of the recursion, and so on.

 Recursive Implementation of an Exponential Moving
Average

Exercise:
Problem:

Try to extend the recursion of the previous paragraphs to the weighted

average
Equation:
N −1

k
xn = a u n−k .

k=0

What goes wrong?

Exercise:
Problem:

Compute the output of the exponential moving average x n =

axn−1 + w u when the input is

0 n

Equation:

0, n < 0
un = {
u, n ≥ 0.
 Plot your result versus n.
Exercise:

Problem: Compute w in the exponential weighting sequence

0

Equation:

0, n < 0
wn = {
n
a w0 , n ≥ 0

to make the weighting sequence a valid window. (This is a special case

of Exercise 3 from Filtering: Moving Averages.) Assume −1 < a < 1
Filtering: Test Sequences

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

When we design a filter, we design it for a purpose. For example, a moving

average filter is often designed to pass relatively constant data while
averaging out relatively variable data. In an effort to clarify the behavior of
a filter, we typically analyze its response to a standard set of test signals.
We will call the impulse, the step, and the complex exponential the standard
test signals.

Unit Pulse Sequence. The unit pulse sequence is the sequence

Equation:

1, n = 0
un = δn = {
0, n ≠ 0.

This sequence, illustrated in Figure 1, consists of all zeros except for a

single one at n = 0. If the unit pulse sequence is passed through a moving
average filter (whether finite or not), then the output is called the unit pulse
response:
Equation:
∞
hn = ∑ w k δ n−k
k=0

= wn .
 Unit Pulse Sequence

(Note that δ = 0 unless n = k.) So the unit pulse sequence may be used
n−k

to read out the weights of a moving average filter. It is common practice to

use w (the k weight) and h (the k impulse response) interchangeably.
k
th
k
th

Exercise:
Problem:

Find the unit pulse response for the finite moving average
N −1

x n = ∑ w k u n−k . Caution: You must consider

k=0

n < 0, 0 ≤ n ≤ N − 1 , and n ≥ N .
Exercise:
Problem:

Find the unit pulse response for the recursive filter

x n= ax + w u .
n−1 0 n

Unit Step Sequence. The unit step sequence is the sequence

Equation:
 1, n ≥ 0
un = ξn = {
0, n < 0.

This sequence is illustrated in Figure 2. When this sequence is applied to a

moving average filter, the result is the unit step response
Equation:
n
gn = ∑ wk
k=0

n
= ∑ hk .
k=0

The unit step response is just the sequence of partial sums of the unit pulse
response.

Unit Step Sequence

Exercise:
 Problem:

Find the unit step response for the finite moving average filter
N −1

x n = ∑ w k u n−k . Specialize your general result to the special case

k=0

1
where w k = for k = 0, 1, ..., N − 1 .
N

Exercise:
Problem:

Find the unit step response for the recursive filter

x n= ax n−1 + w u . 0 n

Complex Exponential Sequence. The complex exponential sequence is the

sequence
Equation:
jkθ
uk = e , k = 0, ±1, ±2, ...

This sequence, illustrated in Figure 3, is a “discrete-time phasor” that

“ratchets” counterclockwise (CCW) as k moves to k + 1 and clockwise
(CW) as k moves to k − 1. Each time the phasor ratchets, it turns out an
angle of θ. Why should such a sequence be a useful test sequence? There
are two reasons.
 Discrete-Time Phasor

(i) e represents (or codes) cos kθ. The real part of the sequence e
jkθ jkθ
is
the cosinusoidal sequence cos kθ:
Equation:
jkθ
Re[e ] =cos kθ.

Therefore the discrete-time phasor e represents (or codes) cos kθ in the

jkθ

same way that the continuous-time phasor e codes cos ωt. If the moving
jωt

average filter
Equation:
∞

x n = ∑ h k u n−k

k=0
has real coefficients, we can get the response to a cosinusoidal sequence by
taking the real part of the following sum:
Equation:
∞
xn = ∑ h k cos (n − k)θ
k=0

∞ j(n−k)θ
= Re[∑ hk e ]
k=0

jnθ ∞ −jkθ
= Re[e ∑ hk e ].
k=0

In this formula, the sum

Equation:
∞

−jkθ
∑ hk e

k=0

is called the complex frequency response of the filter and is given the
symbol
Equation:
∞

jθ −jkθ
H (e ) = ∑ hk e .

k=0

This complex frequency response is just a complex number, with a

magnitude H (e ) and a phase arg H (e ). Therefore the output of the
jθ jθ

moving average filter is

Equation:
jnθ jθ
xn = Re [e H (e )]

jθ
jnθ jθ jargH (e )
= Re[e H (e ) e ]

jθ jθ
= H (e )| cos [nθ+ arg H (e )].
T

∣
This remarkable result says that the output is also cosinusoidal, but its
amplitude is H (e ) rather than 1, and its phase is argH (e ) rather than
jθ

0. In the examples to follow, we will show that the complex “gain” H (e )

can be highly selective in θ, meaning that cosines of some angular
frequencies are passed with little attenuation while cosines of other
frequencies are dramatically attenuated. By choosing the filter coefficients,
we can design the frequency selectivity we would like to have.

(ii) e is a sampled data version of e . The discrete-time phasor e

jkθ

the periodic sampling instants t = kT :

Equation:

e
jkθ
k

= e
jωt
jωt

can be produced physically by sampling the continuous-time phasor e

t=kT

θ = ωT .
= e
jωkT

The dimensions of θ are radians, the dimensions of ω are radians/second,

and the dimensions of T are seconds. We call T the sampling interval and
1
the sampling rate or sampling frequency. If the original angular

frequency of the phasor e

discrete-time phasor remains e

Equation:

e
jωt

j[ω+m(2π/T )]t
is increased to ω + m(
jkθ

t=kT
:

= e
j(ωkT +km2π)

This means that all continuous-time phasors of the form e

2π

= e

“hide
under the same alias” when viewed through the sampling operation. That is,
the sampled-data phasor cannot distinguish the frequency ω from the
frequency ω + m
2π

T
)

jkθ

. In your subsequent courses you will study aliasing in

more detail and study the Nyquist rule for sampling:
Equation:
jθ

, then the

j[ω+m(2π/T )]t
jωt
jθ

jkθ

at
 2π 1 Ω
T ≤ ; ≥
Ω T 2π

1
This rule says that you must sample signals at a rate ( ) that exceeds the
T

Ω
bandwidth of the signal.
2π

Example:
Let's pass the cosinusoidal sequence u k = cos kθ through the finite
moving average filter
Equation:
N −1
xn = ∑ h k u n−k
k=0

1
hk = k = 0, 1, ..., N − 1.
N

We know from our previous result that the output is

Equation:
jθ jθ
xn = H (e ) cos [nθ+ arg H (e )].

The complex frequency response for this example is

Equation:
jθ N −1 1 −jkθ
H (e ) = ∑ e
k=0 N

−jN θ
1 1−e
= −jθ
N 1−e

(Do you see your old friend, the finite sum formula, at work?) Let's try to
manipulate the result into a more elegant form:
Equation:
Equation:
H (e
jθ

|H (e
)

The magnitude of the function H (e

=

jθ
1

)|=

At θ = 0, corresponding to a “DC phasor,” H (e

H (e
jθ
e

jθ

N
−j(N /2)θ

∣
−j(θ/2)

−j[(N −1)/2]θ

is

sin (
[e

[e

sin (

. The magnitude of the complex frequency response is

) = 0

plotted in Figure 4.

Frequency Selectivity of a Moving Average Filter

This result shows that the moving average filter is frequency selective,
passing low frequencies with gain near 1 and high frequencies with gain
near 0.

Exercise:

Problem: Compute the phase of the complex frequency response

Equation:
j(N /2)θ

j(θ/2)

2
−e

θ)
−e

sin(

sin(

θ)

jθ
−j(N /2)θ

−j(θ/2)

)
N

.
θ)

θ)
]
]

equals 1; at θ =
2π

N
 N
1 sin ( θ)
jθ −j[(N −1)/2]θ 2
H (e ) = e .
1
N sin ( θ)
2

Exercise:
Problem:
1
Choose the filter length N for the filter h k = , k = 0, 1, ..., N − 1 ,
N
1
so that a 60 Hz cosine, sampled at the rate = 180 , is perfectly
T
zeroed out as it comes through the filter.
Exercise:
Problem:

(MATLAB) Write a MATLAB program to compute and plot the

magnitude H (e ) and the phase arg H (e ) versus −π < θ < π
jθ jθ

when
Equation:
N
1 sin ( θ)
jθ −j[(N −1)/2]θ 2
H (e ) = e .
1
N sin ( θ)
2

Choose suitable increments for θ.

Exercise:
Problem:

Compute the complex frequency response H (e jθ

) for the recursive
filter x = ax
n + w u .
n−1 0 n
Filtering: Numerical Experiment (Frequency Response of First-Order
Filter)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Consider the exponential moving average filter

Equation:
∞

k
xn = a u n−k; a = 0. 98.

k=0

1. Write out a few terms of the sum to show how the filter works.
2. Write xn as a recursion and discuss the computer memory required to
implement the filter.
3. Compute the complex frequency response H e for the filter. jθ

4. Write a MATLAB program to plot the magnitude and phase of the

complex frequency response H e versus θ for θ = −πto + π in
jθ

2π
steps of Do this for two values of a, namely, a = 0. 98 and
64
a = −0. 98 . Explain your findings.
5. Write a MATLAB program to pass the following signals through the
filter when a = 0. 98:

a. u n = δn

b. u n = ξn
2π
c. u n = ξ n cos n
64
 2π
d. u n = ξ n cos n
32
2π
e. u n = ξ n cos n
16
2π
f. u n = ξ n cos n
8
2π
g. u n = ξ n cos n
4
2π
h. u n = ξ n cos n .
2

Plot the outputs for each case and interpret your findings in terms of the
complex frequency response H e . Repeat step 5 for a = −0. 98.
jθ

Interpret your findings.

Binary Codes: Introduction

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Acknowledgment: Richard Hamming's book, Information Theory

andCoding, Prentice-Hall, New York (1985) and C. T. Mullis's unpublished
notes have influenced our treatment of binary codes. The numerical
experiment was developed by Mullis.

We use this chapter to introduce students to the communication paradigm

and to show how arbitrary symbols may be represented by binary codes.
These symbols and their corresponding binary codes may be computer
instructions, integer data, approximations to real data, and so on.

We develop some ad hoc tree codes for representing information and then
develop Huffman codes for optimizing the use of bits. Hamming codes add
check bits to a binary word so that errors may be detected and corrected.
The numerical experiment has the students design a Huffman code for
coding Lincoln's Gettysburg Address.

Introduction
It would be stretching our imagination to suggest that Sir Francis had digital
audio on his minde (sic) when he wrote the prophetic words

"Sir Francis Bacon, 1623 ...a man may express and signifie the intentions of
his minde, at any distance... by... objects... capable of a twofold difference
onely. "
Nonetheless, this basic idea forms the basis of everything we do in digital
computing, digital communications, and digital audio/video. In 1832,
Samuel F. B. Morse used the very same idea to propose that telegram words
be coded into binary addresses or binary codes that could be transmitted
over telegraph lines and decoded at the receiving end to unravel the
telegram. Morse abandoned his scheme, illustrated in Figure 1, as too
complicated and, in 1838, proposed his fabled Morse code for coding letters
(instead of words) into objects (dots, dashes, spaces) capable of a threefold
difference onely (sic).

Generalized Coder-Decoder

The basic idea of Figure 1 is used today in cryptographic systems, where

the “address a " is an encyphered version of a message w ; in vector
i i

quantizers, where the “address a " is the address of a close approximation to

i

data w ; in coded satellite transmissions, where the “address a " is a data

i i

word w plus parity check bits for detecting and correcting errors; in digital
i

audio systems, where the “address a " is a stretch of digitized and coded
i

music; and in computer memories, where a is an address (a coded version

i

of a word of memory) and w is a word in memory.

i
In this chapter we study three fundamental questions in the construction of
binary addresses or binary codes. First, what are plausible schemes for
mapping symbols (such as words, letters, computer instructions, voltages,
pressures, etc.) into binary codes? Second, what are plausible schemes for
coding likely symbols with short binary words and unlikely symbols with
long words in order to minimize the number of binary digits (bits) required
to represent a message? Third, what are plausible schemes for “coding”
binary words into longer binary words that contain “redundant bits” that
may be used to detect and correct errors? These are not new questions.
They have occupied the minds of many great thinkers. Sir Francis
recognized that arbitrary messages had binary representations. Alan Turing,
Alonzo Church, and Kurt Goedel studied binary codes for computations in
their study of computable numbers and algorithms. Claude Shannon, R. C.
Bose, Irving Reed, Richard Hamming, and many others have studied error
control codes. Shan- non, David Huffman, and many others have studied
the problem of efficiently coding information.

In this chapter we outline the main ideas in binary coding and illustrate the
role that binary coding plays in digital communications. In your subsequent
courses in electrical and computer engineering you will study integrated
circuits for building coders and decoders and mathematical models for
designing good codes.
Binary Codes: The Communication Paradigm

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

A paradigm is a pattern of ideas that form the foundation for a body of

knowledge. A paradigm for (tele-) communication theory is a pattern of
basic building blocks that may be applied to the dual problems of (i)
reliably transmitting information from source to receiver at high speed or
(ii) reliably storing information from source to memory at high density.
High-speed communication permits us to accommodate many low-rate
sources (such as audio) or one high-rate source (such as video). High-
density storage permits us to store a large amount of information in a small
space. For example, a typical 1.2 Mbyte floppy disc stores 9. 6 × 10 bits of
6

information, whereas a typical CD stores about 2 × 10 bits, enough for

9

one hour's worth of high-quality sound.

Basic Building Blocks in a (Tele-) Communication

System
Figure 1 illustrates the basic building blocks that apply to any problem in
the theory of (tele-) communication. The source is an arbitrary source of
information. It can be the time-varying voltage at the output of a vibration
sensor (such as an integrating accelerometer for measuring motion or a
microphone for measuring sound pressure); it can be the charges stored in
the CCD array of a solid-state camera; it can be the addresses generated
from a sequence of keystrokes at a computer terminal; it can be a sequence
of instructions in a computer program. The source coder is a device for
turning primitive source outputs into more efficient representations. For
example, in a recording studio, the source coder would convert analog
voltages into digital approximations using an A/D converter; a fancy
source coder would use a fancy A/D converter that finely quantized likely
analog values and crudely quantized unlikely values. If the source is a
source of discrete symbols like letters and numbers, then a fancy source
code would assign short binary sequences to likely symbols (such as e) and
long binary sequences to unlikely symbols (such as z). The channel coder
adds “redundant bits” to the binary output of the source coder so that errors
of transmission or storage may be detected and corrected. In the simplest
example, a binary string of the form 01001001 would have an extra bit of 1
added to give even parity (an even number of l's) to the string; the string
10110111 would have an extra bit of 0 added to preserve the even parity. If
one bit error is introduced in the channel, then the parity is odd and the
receiver knows that an error has occurred. The modulator takes outputs of
the channel coder, a stream of binary digits, and constructs an analog
waveform that represents a block of bits. For example, in a 9600 baud
Modem, five bits are used to determine one of 2 = 32 phases that are used
5

to modulate the signal A cos (ωt + φ). Each possible string of five bits has
its own personalized phase, φ, and this phase can be determined at the
receiver. The signal A cos (ωt + φ) is an analog signal that may be
transmitted over a channel (such as a telephone line, a microwave link, or a
fiber-optic cable). The channel has a finite bandwidth, meaning that it
distorts signals, and it is subject to noise or interference from other
electromagnetic radiation. Therefore transmitted information arrives at the
demodulator in imperfect form. The demodulator uses filters matched to the
modulated signals to demodulate the phase and look up the corresponding
bit stream. The channel decoder converts the coded bit stream into the
information bit stream, and the source decoder looks up the corresponding
symbol. This sequence of steps is illustrated symbolically in Figure 7.3.

Symbolic Representation of Communication

In your subsequent courses on communication theory you will study each

block of Figure 1 in detail. You will find that every source of information
has a characteristic complexity, called entropy, that determines the
minimum rate at which bits must be generated in order to represent the
source. You will also find that every communication channel has a
characteristic tolerance for bits, called channel capacity. This capacity
depends on signal-to-noise ratio and bandwidth. When the channel capacity
exceeds the source entropy, then you can transmit information reliably; if it
does not, then you cannot.
Binary Codes: From Symbols to Binary Codes

Note:This module is part of the collection, A First Course in Electrical and

Computer Engineering. The LaTeX source files for this collection were
created using an optical character recognition technology, and because of this
process there may be more errors than usual. Please contact us if you
discover any errors.

Perhaps the most fundamental idea in communication theory is that arbitrary

symbols may be represented by strings of binary digits. These strings are
called binary words, binary addresses, or binary codes. In the simplest of
cases, a finite alphabet consisting of the letters or symbols s , s , ..., s
0 1 M −1 is
represented by binary codes. The obvious way to implement the
representation is to let the i binary code be the binary representation for the
th

subscript i:
Equation:

s 0 ∼ 000 = a0

s 1 ∼ 001 = a1

s 6 ∼ 110 = a6

s 7 ∼ 111 = a7 .

The number of bits required for the binary code is N where

Equation:
N −1 N
2 < M ≤ 2 .

We say, roughly, that N =log 2 M .

Octal Codes. When the number of symbols is large and the corresponding
binary codes contain many bits, then we typically group the bits into groups
of three and replace the binary code by its corresponding octal code. For
example, a seven-bit binary code maps into a three-digit octal code as
follows:
Equation:

0000000 ∼ 000

0000001 ∼ 001

0100110 ∼ 046

101111 ∼ 137

1111111 ∼ 177.

The octal ASCII codes for representing letters, numbers, and special
characters are tabulated in Table 1.
Exercise:

Problem: Write out the seven-bit ASCII codes for A, q, 7, and{.

'0 '1 '2 '3 '4 '5 '6 '7

'00x ␀ ␁ ␂ ␃ ␄ ␅ ␆ ␇

'01x ␈ ␉ ␊ ␋ ␌ ␍ ␎ ␏

'02x ␐ ␑ ␒ ␓ ␔ ␕ ␖ ␗

'03x ␘ ␙ ␚ ␛ ␜ ␝ ␞ ␟
 '04x ␠ ! " # $ % & '

'05x ( ) * + , - . /

'06x 0 1 2 3 4 5 6 7

'07x 8 9 : ; < = > ?

'10x @ A B C D E F G

'11x H I J K L M N O

'12x P Q R S T U V W

'13x X Y Z [ \ ] ^ _

'14x ` a b c d e f g

'15x h i j k l m n o

'16x p q r s t u v w

'17x x y z { | } ~ ␡

Octal ASCII Codes (from Donald E. Knuth, The TEXbook, ©1986 by the
American Mathematical Society, Providence, Rhode Island p. 367, published
by Addison-Wesley Publishing Co.)

Exercise:
Problem:

Add a 1 or a 0 to the most significant (left-most) position of the seven-bit

ASCII code to produce an eight-bit code that has even parity (even
number of 1's). Give the resulting eight-bit ASCII codes and the
corresponding three-digit octal codes for %, u, f , 8, and +.
Quantizers and A/D Converters. What if the source alphabet is infinite?
Our only hope is to approximate it with a finite collection of finite binary
words. For example, suppose the output of the source is an analog voltage
that lies between −V and +V . We might break this peak-to-peak range up
0 0

2V
into little voltage cells of size A and approximate the voltage in each cell
M
by its midpoint. This scheme is illustrated in Figure 1. In the figure, the cell
2 V pV Δ
C is defined to be the set of voltages that fall between i and –
i −
M M

V
2V
i– MΔ + M
A
:
Equation:

2V 0 V0 2V 0 V0
C i = {V : i. − < V ≤ i + }.
M M M M

The mapping from continuous values of V to a finite set of approximations is

Equation:

2V 0
Q (V ) = i , if V ∈ Ci .
M

2 V0
That is, V is replaced by the quantized approximation i whenever V lies in
–
M
2V
cell C . We may represent the quantized values i 0 with binary codes by
i
–
M

simply representing the subscript of the cell by a binary word. In a subsequent

course on digital electronics and microprocessors you will study A/D
(analog-to-digital) converters for quantizing variables.
 A Quantizer

Example:
If M = 8, corresponding to a three-bit quantizer, we may associate quantizer
cells and quantized levels with binary codes as follows:
Equation:
 2V 0
V ∈ C −3 ⇒ V −3 = (−3) ∼ 111
8

2V 0
V ∈ C −2 ⇒ V −2 = (−2) ∼ 110
8

2V 0
V ∈ C −1 ⇒ V −1 = (−1) ∼ 101
8

V ∈ C0 ⇒ V 0 = 0 ∼ 000

2V 0
V ∈ C1 ⇒ V 1 = (1) ∼ 001
8

2V 0
V ∈ C2 ⇒ V 2 = (2) ∼ 010
8

2V 0
V ∈ C3 ⇒ V 3 = (3) ∼ 011.
8

This particular code is called a sign-magnitude code, wherein the most

significant bit is a sign bit and the remaining bits are magnitude bits (e.g.,
110 ∼ −2 and 010 ∼ 2). One of the defects of the sign-magnitude code is

that it wastes one code by using 000 for 0 and 100 for-O. An alternative code
that has many other advantages is the 2's complement code. The 2 s ′

complement codes for positive numbers are the same as the sign-magnitude
codes, but the codes for negative numbers are generated by complementing
all bits for the corresponding positive number and adding 1:
Equation:

−4 ∼ 100

−3 ∼ 101 (100 + 1)

−2 ∼ 110 (101 + 1)

−1 ∼ 111 (110 + 1)

0 ∼ 000

1 ∼ 001

2 ∼ 010

3 ∼ 011.

Exercise:
 Problem:

Generate the four-bit sign-magnitude and four-bit 2's complement binary

codes for the numbers −8, −7, ..., −1, 0, 1, 2, ..., 7.
Exercise:
Problem:

Prove that, in the 2 s complement representation, the binary codes for

′

−nand + n sum to zero. For example,

Equation:

101 + 011 = 000

(−3) (3) (0).

In your courses on computer arithmetic you will learn how to do arithmetic in

various binary-coded systems. The following problem illustrates how easy
arithmetic is in 2's complement.
Exercise:
Problem:

Generate a table of sums for all 2 s complement numbers between −4

′

and +3. Show that the sums are correct. Use

0 + 0 = 0, 0 + 1 = 1, 1 + 0 = 1, and 1 + 1 = 0 with a carry into the

next bit. For example, 001 + 001 = 010.

Binary Trees and Variable-Length Codes. The codes we have constructed

so far are constant-length codes for finite alphabets that contain exactly
M = 2
N
symbols. In the case where M = 8 and N = 3, then the eight
possible three-bit codes may be represented as leaves on the branching tree
illustrated in Figure 2(a). The tree grows a left branch for a 0 and a right
branch for a 1, until it terminates after three branchings. The three-bit codes
we have studied so far reside at the terminating leaves of the binary tree. But
what if our source alphabet contains just five symbols or letters? We can
represent these five symbols as the three-bit symbols 000 through 100 on the
binary tree. This generates a constant-length code with three unused, or
illegal, symbols 101 through 111. These are marked with an "x" in Figure
2(a). These unused leaves and the branches leading to them may be pruned to
produce the binary tree of Figure 2(b).

If we admit variable-length codes, then we have several other options for

using a binary tree to construct binary codes. Two of these codes and their
corresponding binary trees are illustrated in Figure 3. If we disabuse ourselves
of the notion that each code word must contain three or fewer bits, then we
may construct binary trees like those of Figure 4 and generate their
corresponding binary codes. In Figure 4(a), we grow a right branch after each
left branch and label each leaf with a code word. In Figure 4(b), we prune off
the last right branch and associatea code word with the leaf on the last left
branch.

Binary Trees and Constant-Length Codes; (a) Binary Tree, and (b)
Pruned Binary Tree

All of the codes we have generated so far are organized in Table 2. For each
code, the average number of bits/symbol is tabulated. This average ranges
from 2.4 to 3.0. If all symbols are equally likely to appear, then the best
variable-length code would be code 2.

Binary Trees and Variable-Length Codes; (a) Binary Tree for Variable-
length Code, and (b) Another Binary Tree for Variable-length Code
All of the codes we have constructed have a common characteristic: each
code word is a terminating leaf on a binary tree, meaning that no code word
lies along a limb of branches to another code word. We say that no code word
is a prefix to another code word. This property makes each of the codes
instantaneously decodable, meaning that each bit in a string of bits may be
processed instantaneously (or independently) without dependence on
subsequent bits.
Exercise:

Problem: Decode the following sequence of bits using code 2:

Equation:

0111001111000000101100111.

Left-Handed Binary Trees for Variable-Length Codes; (a) Left-handed

Binary Tree, and (b) Pruned Binary Tree
 Code Average
# S0 S1 S2 S3 S4 Bits/Symbol

1 000 001 010 011 100 15/5 = 3.0

2 000 001 01 10 11 12/5 = 2.4

3 000 001 010 011 1 13/5 = 2.6

4 1 01 001 0001 00001 15/5 = 3.0

5 1 01 001 0001 0000 14/5 = 2.8

Variable Length Codes

Exercise:
Problem:

Illustrate the following codes on a binary tree. Which of them are

instantaneously decodable? Which can be pruned and remain
instantaneously decodable?
Equation:

S0 S1 S2 S3 S4

011 100 00 11 101

011 100 00 0 01

010 000 100 101 111.

Code #2 generated in Table 2 seems like a better code than code #5 because
its average number of bits/symbol (2.4) is smaller. But what if symbol S is a
0

very likely symbol and symbol S is a very unlikely one? Then it may well
4

turn out that the average number of bits used by code #5 is less than the
average number used by code #2. So what is the best code? The answer
depends on the relative frequency of use for each symbol. We explore this
question in the next section.
Binary Codes: Huffman Codes for Source Coding

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In 1838, Samuel Morse was struggling with the problem of designing an

efficient code for transmitting information over telegraph lines. He reasoned
that an efficient code would use short code words for common letters and
long code words for uncommon letters. (Can you see the profit motive at
work?) In order to turn this reasoned principle into workable practice,
Morse rummaged around in the composition trays for typeface in a
printshop. He discovered that typesetters use many more e s than s's. He
′

then formed a table that showed the relative frequency with which each
letter was used. His ingenious, variable-length Morse code assigned short
codes to likely letters (like “dot” for e ) and long codes to unlikely letters
′

(like “dash dash dot dot” for z ). We now know that Morse came within
′

about 15% of the theoretical minimum for the average code word length for
English language text.

A Variation on Morse's Experiment. In order to set the stage for our study
of efficient source codes, let's run a variation on Morse's experiment to see
if we can independently arrive at a way of designing codes. Instead of
giving ourselves a composition tray, let's start with a communication source
that generates five symbols or letters S , S , S , S , S . We run the source
0 1 2 3 4

for 100 transmissions and observe the following numbers of transmissions

for each symbol:
Equation:

′
 ′
50 S s
0

′
20 S s
1

′
20S s
2

′
5 S s
3

′
5 S s.
4

We will assume that these “source statistics” are typical, meaning that 1000
transmissions would yield 500 S s and so on.
′
0

The most primitive binary code we could build for our source would use
three bits for each symbol:
Equation:

S 0 ∼ 000

S 1 ∼ 001

S 2 ∼ 010

S 3 ∼ 011

S 4 ∼ 100

x ∼ 101

x ∼ 110

x ∼ 111.

This code is inefficient in two ways. First, it leaves three illegal code words
that correspond to no source symbol. Second, it uses the same code word
length for an unlikely symbol (like S ) that it uses for a likely symbol (like
4

S ). The first defect we can correct by concatenating consecutive symbols

0

into symbol blocks, or composite symbols. If we form a composite symbol

consisting of M source symbols, then a typical composite symbol is
S S S S S S S S S . The number of such composite symbols that can
1 0 1 4 2 3 1 2 0

be generated is 5 . The binary code for these 5 composite symbols must

M M

contain N binary digits where

Equation:
 N −1 M N
2 < 5 < 2 (N ≅ M log 2 5).

The number of bits per source symbol is

Equation:

N
≅log 2 5 = 2. 32.
M

This scheme improves on the best variable length code of Table 2 from
"Binary Codes: From Symbols to Binary Codes" by 0.08 bits/symbol.
Exercise:
Problem:

Suppose your source of information generates the 26 lowercase roman

letters used in English language text. These letters are to be
concatenated into blocks of length M. Complete the following table of
N (number of bits) versus M (number of letters in a block) and show
N
that approaches log 2
26 .
M

1 2 3 4 5 6

N 5 10

N /M 5 5
Now let's reason, as Morse did, that an efficient code would use short codes
for likely symbols and long codes for unlikely symbols. Let's pick code #5
from Table 2 from "Binary Codes: From Symbols to Binary Codes" for this
purpose:
Equation:

S0 S1 S2 S3 S4

1 01 001 0001 0000.

This is a variable-length code. If we use this code on the 100 symbols that
generated our source statistic, the average number of bits/symbol is
Equation:

1
[50 (1) + 20 (2) + 20 (3) + 5 (4) + 5 (4)] = 1. 90bits/symbol.
100

Exercise:
Problem:

Use the source statistics of Equation 1 to determine the average

number of bits/symbol for each code in Table 2 from "Binary Codes:
From Symbols to Binary Codes".

Entropy. So far, each ad hoc scheme we have tried has produced an

improvement in the average number of bits/symbol. How far can this go?
The answer is given by Shannon's source coding theorem, which says that
the minimum number of bits/symbol is
Equation:
M
N
≥ − p i log 2 p i
M
i=1

where pi is the probability that symbol Si is generated and − p log p

i 2 i

is a fundamental property of the source called entropy. For our five-symbol

example, the table of pi and − log p is given in Table 2. The entropy is
i

1.861, and the bound on bits/symbol is

Equation:

N
≥ 1. 861.
M

Code #5 comes within 0.039 of this lower bound. As we will see in the
next paragraphs, this is as close as we can come without coding composite
symbols.

Symbol Probability ~ Log Probability

S0 0.5 1

S1 0.2 2.32

S2 02 2.32

S3 0.05 4.32

S4 0.05 4.32

Source statistics for Five-Symbol Source

Exercise:
 Problem:

Select an arbitrary page of English text. Build a table of source

statistics containing p (relative frequencies) and-log p for a through z
i i

. (Ignore distinction between upper and lower case and ignore

punctuation and other special symbols.) Compute the entropy
26

− p i log 2 p i .
i=1

Huffman Codes. In the late 1950s, David Huffman discovered an

algorithm for designing variable-length codes that minimize the average
number of bits/symbol. Huffman's algorithm uses a principle of optimality
that says, “the optimal code for M letters has imbedded in it the optimal
code for the M − 1 letters that result from aggregating the two least likely
symbols.” When this principle is iterated, then we have an algorithm for
generating the binary tree for a Huffman code:

i. label all symbols as “children”;

ii. “twin” the two least probable children and give the twin the sum of the
probabilities:

iii. (regard the twin as a child; and

iv. repeat steps (ii) and (iii) until all children are accounted for.

This tree is now labeled with 1's and 0's to obtain the Huffman code. The
labeling procedure is to label each right branch with a 1 and each left
branch with a 0. The procedure for laying out symbols and constructing
Huffman trees and codes is illustrated in the following examples.
Example:
Consider the source statistics
Equation:

Symbol S0 S1 S2 S3 S4

Probability 0.5 0.2 0.2 0.05 0.05

for which the Huffman algorithm produces the following binary tree and its
corresponding code:

Example:
The Huffman code for the source statistics
Equation:

Symbol S0 S1 S2 S3 S4

Probability 0.75 0.075 0.075 0.05 0.05

is illustrated next:

Exercise:
Problem:

Generate binary trees and Huffman codes for the following source
statistics:
Equation:

Symbol S0 S1 S2 S3 S4 S5 S6 S7

Probability1 0.20 0.20 0.15 0.15 0.1 0.1 0.05 0.05

Probability2 0.3 0.25 0.1 0.1 0.075 0.075 0.05 0.05.

Coding a FAX Machine. Symbols can arise in unusual ways and be

defined quite arbitrarily. To illustrate this point, we consider the design of a
hypothetical FAX machine. For our design we will assume that a laser
scanner reads a page of black and white text or pictures, producing a high
voltage for a black spot and a low voltage for a white spot. We will also
assume that the laser scanner resolves the page at 1024 lines, with 1024
spots/line. This means that each page is represented by a two-dimensional
array, or matrix, of pixels (picture elements), each pixel being 1 or 0. If we
simply transmitted these l's and O s, then we would need
′

1024 × 1024 = 1, 059, 576 bits. If these were transmitted over a 9600

baud phone line, then it would take almost 2 minutes to transmit the FAX.
This is a long time.

Let's think about a typical scan line for a printed page. It will contain long
runs of 0's, corresponding to long runs of white, interrupted by short bursts
of l's, corresponding to short runs of black where the scanner encounters a
line or a part of a letter. So why not try to define a symbol to be “a run of k
0's" and code these runs? The resulting code is called a “run length code.”
Let's define eight symbols, corresponding to run lengths from 0 to 7 (a run
length of 0 is a 1):
Equation:

S0 = run length of 0 zeros (a 1)

S1 = run length of 1 zero

S7 = run legnth of 7 zeros.

If we simply used a simple three-bit binary code for these eight symbols,
then for each scan line we would generate anywhere from 3 × 1024 bits
(for a scan line consisting of all 1's) to 3 × 1024/7 ≅ 400 bits (for a scan
line consisting of all 0's). But what if we ran an experiment to determine the
relative frequency of the run lengths S through S and used a Huffman
0 7

code to “run length encode” the run lengths? The following problem
explores this possibility and produces an efficient FAX code.
Exercise:
Problem:

An experiment conducted on FAXed documents produces the

following statistics for run lengths of white ranging from 0 to 7:
Equation:

Symbol S0 S1 S2 S3 S4 S5 S6 S7

Probability 0.01 0.06 0.1 0.1 0.2 0.15 0.15 0.2

These statistics indicate that only 1% of a typical page is black.

Construct the Huffman code for this source. Use your Huffman code to
code and decode these scan lines:
Binary Codes: Hamming Codes for Channel Coding

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The idea behind Hamming codes is to intersperse, or append, extra binary

digits to a binary code so that errors in transmission of the code over a
channel may be detected and corrected. For example, suppose we transmit
the code 01101001, and it is received as 01001001. In this transmission, the
third most significant bit is received erroneously. Let's define the following
“modulo-2 addition” of binary numbers:
Equation:

0 ⊕ 0 = 0

0 ⊕ 1 = 1

1 ⊕ 0 = 1

1 ⊕ 1 = 0.

Multiplication in modulo-2 arithmetic is simply 0 ⋅ 0 = 0 ⋅ 1 = 1 ⋅ 0 = 0

and 1 ⋅1 = 1. Then we can say that the error sequence 00100000 is “added”
to the transmission 01101001 to produce the erroneous reception:
Equation:

01101001 transmitted

⊕ 00100000 error

01001001 received.
Hamming error correcting codes will permit us to receive the erroneous
transmission and to detect and correct the error. This is obviously of great
value in transmitting and storing information. (Imagine how upset you
would be to have the binary code for your checking account confused with
that of Mrs. Joan Kroc.)

Choosing the Number of Check Bits. Let's suppose we have N bits of

information that we wish to transmit and that we wish to intersperse “check
bits” that will enable us to detect and correct any single bit error in the
transmission. If we use N information bits and n check bits, then we will
transmit a code word containing N + n bits. The n check bits can code 2 n

events, and we want these events to indicate whether or not any errors
occurred and, if so, where they occurred. Therefore we require

where (N + n) is the number of single error events that can occur and +1 is
the number of no-error events. For example, when N = 4, we require
n = 3 so that 2 ≥ (4 + 3) + 1.
3

How many check bits do you require to code seven bits of information for
single error correction?

Code Construction. Let's suppose we have constructed an (N , n)

Hamming code consisting of N information bits and n check bits (or parity
bits). We denote the information bits by x , x , ..., x and the check bits by
1 2 N

c , c , ..., c . These bits may be interspersed. When N = 4 and n = 3, then

1 2 n

a typical array of bits within a code word would be one of the following:
Equation:

c1 x1

c2 x2

x1 x3

c3 or x4 .

x2 c1

x3 c2

x4 c3
The first ordering is “natural” (as we will see), and the second is
“systematic” (a term that is used to describe any code whose head is
information and whose tail is check). If a single error occurs in an (N , n)
code, then the received code word will be the modulo-2 sum of the code
word and the error word that contains a 1 in its i position: th

Equation:

c1 0

c2 0

xl 0

c3 ⊕ l .

x2 0

x3 0

x4 0

We would like to operate on this received code word in such a way that the
location of the error bit can be determined. If there were no code word, then
an obvious solution would be to premultiply the error word by the parity
check matrix
Equation:

1 0 1 0 1 0 1

0 1 1 0 0 1 1
T
A = .
0 0 0 1 1 1 1

[(1) (2) (3) (4) (5) (6) (7)]

The i column of A is just the binary code for i. When A premultiplies

th T T

an error word, the error bit picks out the column that codes the error
position:
If the error word contains no error bits, then the product is 0, indicating no
errors.

This seems like a good idea, but what about the effect of the code word? In
Exercise 1, you are asked to show that the effect of the parity check matrix
A applied to the modulo-2 sum of a code word x and an error word e is
T

Equation:
T T T
A (x ⊕ e) = A x ⊕ A e.

In this equation all sums and products obey the rules of modulo-2
arithmetic.
Exercise:
Problem:

Let y = x ⊕ e denote the modulo-2 sum of a code word x and an error

word e; A is a parity check matrix. Show that
T

Equation:
T T T
A y = A x ⊕ A e.

We have designed the parity check matrix A so that the syndromeA e

T T

produces a binary code for the error location. (The location of the error is
 ′
th e syndrome for the error word.) The product A x will interfere with this T

syndrome unless A x = 0. Therefore we will require that the code word x

T

satisfy the constraint

Equation:
T
A x = 0.

This constraint actually defines the Hamming code. Let's illustrate this point
by applying the constraint to a code word in its “natural format” x = T

(c c x c x x x ).
1 2 1 3 2 3 4

Natural Codes. When the information bits and the check bits are coded in
their natural order (c c x c x x x ), then we may determine the check bits
1 2 1 3 2 3 4

by writing A x as follows:
T

Equation:

c1

c2

1 0 1 0 1 0 1 x1 0

0 1 1 0 0 1 1 c3 = 0

0 0 0 1 1 1 1 x2 0

x3

x4

We use the rules of modulo-2 arithmetic to write these constraints as

Equation:

c1 ⊕ x1 ⊕ x2 ⊕ x4 = 0

c2 ⊕ x1 ⊕ x3 ⊕ x4 = 0

c3 ⊕ x2 ⊕ x3 ⊕ x4 = 0.

Therefore the check bits c 1, c2 , and c are simply the following modulo-2
3

sums
Equation:

c1 = x1 ⊕ x2 ⊕ x4

c2 = x1 ⊕ x3 ⊕ x4

c3 = x2 ⊕ x3 ⊕ x4.

This finding may be organized into the matrix equation

Equation:

c1 1101

c2 1011
x1
xl 1000
x2
c3 = 0111 .
x3
x2 0100
x4
x3 0010

x4 0001

This equation shows how the code word x is built from the information bits
(x , x , x , x ). We call the matrix that defines the construction a coder
1 2 3 4

matrix and write it as H :

Equation:

x = HΘ

T T
x = (c1c2x1c3x2x3x4) Θ = (x1x2x3x4)

Equation:
 1101

1011

1000

H = 0111

0100

0010

0001

This summarizes the construction of a Hamming code x.

Exercise:
Problem:

Check to see that the product of the parity check matrix A and the
T

coder matrix H is A H = 0. Interpret this result.

T

Exercise:
Problem:

Fill in the following table to show what the Hamming (4, 3) code is:
Equation:
 x1 x2 x3 x4 c1 c2 x1 c3 x2 x3 x4

0 0 0 0 0 0 0 0 0 0 0

0 0 0 1 1 1 0 1 0 0 1

0 0 1 0

0 0 1 1

0 1 0 0

0 1 0 1

0 1 1 0

0 1 1 1

1 0 0 0

1 0 0 1

1 0 1 0

1 0 1 1

1 1 0 0

1 1 0 1

1 1 1 0

1 1 1 1

Exercise:
Problem:

Design a Hamming (11, n) code for coding eleven information bits

against single errors. Show your equations for c , c , ..., c and write
1 2 n

out the coder matrix H for x = HΘ.

Decoding. To decode a Hamming code, we form the syndrome A y for the T

received (and possibly erroneous) code word y = x ⊕ e. Because A x = 0, T

the syndrome is
Equation:
T
s = A e.
Convert this binary number into its corresponding integer location and
change the bit of y in that location. If the location is zero, do nothing. Now
strip off the information bits. This is the decoding algorithm.
Exercise:
Problem:

Use the table of Hamming (4, 3) codes from Exercise 3 to construct a

table of received codes that contain either no bit errors or exactly one
bit error. Apply the decoding algorithm to construct (x , x , x , x )
1 2 3 4

and show that all received code words with one or fewer errors are
correctly decoded.

Digital Hardware. The tables you have constructed in Exercise 3 and 5 for
coding and decoding Hamming (4, 3) codes may be stored in digital logic
chips. Their functionality is illustrated in Figure 1. The coder chip accepts
(x x x x ) as its address and generates a coded word. The decoder chip
1 2 3 4

accepts (c c x c x x x ) as its address and generates a decoded word. In

1 2 1 3 2 3 4

your courses on digital logic you will study circuits for implementing
coders and decoders.

Digital Logic for Hamming Code

Exercise:
Problem:

Discuss the possibility of detecting a received (4, 3) code word that is

neither a valid code word nor a code word with a single error. How
would you use such a detector?
Exercise:
Problem:

What fraction of received seven-bit words can be correctly decoded as

Hamming (4, 3) codes?

Systematic Codes. Systematic Hamming codes are codes whose

information bits lead and whose check bits trail. The format for a (4, 3)
code is then (x x x x c c c ). The construction of a (4, 3) code word from
1 2 3 4 1 2 3

the information bits may be written as

Equation:

x1 1 0 0 0

x2 0 1 0 0
x1
x3 0 0 1 0
x2
x4 = 0 0 0 1
x3
c1 c11 c12 c13 c14
x4
c2 c21 c22 c23 c24

c3 c31 c32 c33 c34

The coder matrix takes the form

Equation:

1 0 0 0

0 1 0 0

0 0 1 0 I

H = 0 0 0 1 = _____ .

c11 c12 c13 c14 C

c21 c22 c23 c24

c31 c32 c33 c34

The problem is to find the matrix C that defines the construction of check
bits. The constraint A x = 0 produces the constraint A H = 0 so that
T T

A HΘ = 0. The constraints A H = 0 may be written out as

T T

Equation:

1 0 0 0

0 1 0 0

1 0 1 0 1 0 1 0 0 1 0 0 0 0 0

0 1 1 0 0 1 1 = 0 0 0 1 = 0 0 0 0 .

0 0 0 1 1 1 1 c11 c12 c13 c14 0 0 0 0

c21 c22 c23 c24

c31 c32 c33 c34

These constraints produce all the equations we need (twelve equations in

twelve unknowns) to determine the c . ij

Exercise:
Problem:

Solve Equation 19 for the c . Show that the coder matrix for a
ij

systematic Hamming (4, 3) code is

Equation:

1 0 0 0

0 1 0 0

0 0 1 0

H = 0 0 0 1 .

0 1 1 1

1 0 1 1

1 1 0 1

Exercise:
 Problem:

Show that the coder matrix of Exercise 7 is a permutation of the coder

matrix in Equation 14. (That is, the rows are reordered.)
Exercise:
Problem:

(MATLAB) Write a MATLAB program that builds Hamming (4, 3)

codes from information bits (x x x x ) and decodes Hamming (4, 3)
1 2 3 4

codes (c c x c x x x ) to obtain information bits (x x x x ).

1 2 1 3 2 3 4 1 2 3 4

Synthesize all seven-bit binary codes and show that your decoder
correctly decodes correct codes and one-bit error codes.
Binary Codes: Numerical Experiment (Huffman Codes)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The paragraph below contains Lincoln's Gettysburg Address. Ignore special

symbols like periods and ignore the distinction between lowercase and
uppercase letters. Compute the relative frequency of occurrence for each of
the 26 letters of the alphabet. Write a MATLAB program to generate a
binary tree for the Huffman code of the Gettysburg Address. When you
have generated the code, compute the average code word length
Equation:
26
ni
L = ∑ li
N
i=1

where is the relative frequency for symbol i and l is the code word
ni
i
N

length for symbol i. Compare L with the entropy

Equation:
26 26
ni ni ni N
H = −∑ log 2 = ∑ log 2
N N N ni
i=1 i=1

and compare L to log 2

26 . Interpret your findings.

Gettysburg Address
Fourscore and seven years ago, our fathers brought
forth upon this
continent a new nation, conceived in liberty and
dedicated to the proposition
that all men are created equal. Now we are engaged
in a great civil war,
testing whether that nation or any nation so
conceived and so dedicated can
long endure. We are met on a great battlefield of
that war. We have come
to dedicate a portion of that field as a final
resting-place for those who have
given their lives that that nation might live. It
is altogether fitting and proper
that we should do this. But in a larger sense, we
cannot dedicate, we cannot
consecrate, we cannot hallow this ground. The
brave men, living and dead,
who struggled here have consecrated it far above
our power to add or detract.
The world will little note nor long remember what
we say here, but it can
never forget what they did here. It is for us the
living rather to be dedicated
here to the unfinished work which they who fought
here have thus far so nobly
advanced. It is rather for us to be here dedicated
to the great task remaining
before us–that from these honored dead we take
increased devotion to that
cause for which they gave the last full measure of
devotion–that we here
highly resolve that these dead should not have
died in vain, that this nation
under God shall have a new birth of freedom, and
that government of the
people, by the people, for the people shall not
perish from the Earth.
-Abraham Lincoln, November 19, 1863
An Introduction to MATLAB: Introduction

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Acknowledgment: This appendix, written with assistance from Cédric J.

Demeure and Peter Massey, was inspired by the MATLAB user's manual
from The MATHWORKS, Inc. The MA TLAB Primer, available through
the MATLAB User's Group, is a useful learning aid for teachers and
students. To join the MATLAB User's Group, send your request via E-mail
to matlab-users [email protected].

Introduction
MATLAB stands for “Matrix Laboratory.” It is a computing environment
specifically designed for matrix computations. The program is ideally
suited to circuit analysis, signal processing, filter design, control system
analysis, and much more. Beyond that, its versatility with complex numbers
and graphics makes it an attractive choice for many other programming
tasks. MATLAB can be thought of as a programming language like
PASCAL, FORTRAN, C, or BASIC. Like most versions of BASIC,
MATLAB can be used in an interactive mode wherein statements are
executed immediately as they are typed. Alternatively, a program can be
written in advance and saved to a disc file using an editor and then executed
in MATLAB. You will find both modes of operation useful.
An Introduction to MATLAB: Running MATLAB (Macintosh)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In order to run MATLAB on a Macintosh SE or PLUS computer, you need

the program called EDU-MATLAB. The program requires at least 1 Mbyte
of memory, System 3.0 or above, Finder version 3.0 or above, and an 800K
drive. A hard disc drive is highly recommended. In order to run MATLAB
on a Macintosh II, IIx, IIcx, or SE/30, you need the program called MacII-
MATLAB, and the same system requirements apply.

To start MATLAB, you may need to open the folder containing the
MATLAB program. Then just “double-click” the program icon or the
program name (for example, EDU-MATLAB). Figure A.1 shows a typical
organization of the folder containing Mac II-MATLAB. It contains the main
program, the settings file, the demonstrations folder, and any toolbox
folders. The double-click on Mac II-MATLAB produces the Command
window as shown in Figure A.2. You will also see a Graph window
partially hidden behind it. (The fact that the window is not in ffont means
that it is opened but not currently active.) If you do not know what
“clicking,” “dragging,” “pop-up menu,” and “trash” mean, you should stop
reading now and familiarize yourself with the Macintosh.
 The MATLAB Folder
(©Apple Computer, Inc.,
used with permission.)

In the command window, you should see the prompt ≫. The program
interpreter is waiting for you to enter instructions. At this point it is a good
idea to run the demonstration programs that are available in the “About
MATLAB” menu under the Apple menu. Just click on the “demos” button
and select a demo. During pauses, strike any key to continue. Whenever
you have a MATLAB file in any folder, then you may double-click the file
to launch the program. This allows you to have your own folder containing
your own MATLAB files, separated from the MATLAB folder.

The Command Window

(©Apple Computer, Inc.,
used with permission.)

MATLAB has four types of windows:

i. Command for computing, programming, and designing input/output

displays;
ii. Graph for displaying plots and graphs;
iii. Edit for creating and modifying your own files; and
iv. Help for getting on-line help and for running demos.
All windows follow the traditional behavior of Macintosh windows. You
can resize them (actually the help window has a fixed size) or move them.
For more details on menus and windows, see the Macintosh and MATLAB
manuals.
An Introduction to MATLAB: Running MATLAB (PC)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In order to run MATLAB (Version 3.5) on an IBM or compatible personal

computer, you must have a floating point math coprocessor
installed and at least 512 kbytes of memory. The program is called PCMAT-
LAB.EXE, but it is usually invoked via the batch file MATLAB.BAT in the
MATLAB subdirectory. If you are using a menu system and MATLAB is
one of your choices, just choose it. Otherwise, go to the MATLAB
suiUirectory and type MATLAB.

You may be able to usea more powerful implementation of MATLAB if you

have an 80286 or 80386 machine. AT-MATLAB runs on an 80286 with at
least 1 Mbyte of extended memory. AT-MATLAB is distributed with PC-
MATLAB. 386-MATLAB, a special version for 80386 or 80486 machines
with virtual memory support and no limits on variable size, is sold
separately.

When you run MATLAB, you should see the prompt . The program
interpreter is waiting for you to enter instructions. Some MATLAB
instructions, such as plot, are graphics-type instructions which plot results
and data. Execution of one of these graphics instructions puts the PC screen
into the graphics mode, which displays the resulting plot. No instructions
can be executed in the graphics mode other than a screen-dump function.
Striking any other key will return the PC to the command mode, but the
graphics are temporarily stored (like variables) and can be recalled by the
shg (show graphics) instruction. If you wish, you may run some of the
demonstration programs now by entering demo and following the on-screen
instructions.
An Introduction to MATLAB: Interactive Mode

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In command mode, MATLAB displays a prompt and waits for your

input. You may type any legal mathematical expression for immediate
evaluation. Try the following three examples (press “enter” or “return” at
the end of each line):

≫ 2+2
≫ 5^2
≫ 2*sin(pi/4)

The variable pi = 3.14 is built into MATLAB, as are the sin function
and hundreds of other functions. When you entered each of the preceding
lines, MATLAB stored the results in a variable called ans for answer. The
value of ans was then displayed. The last line should have produced the
square root of 2. We can manipulate ans to find out

≫ ans^2

The new answer is very close to 2, as expected. Let's see what the roundoff
error is:

≫ ans-2
An Introduction to MATLAB: Variables

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

Any result you wish to keep for a while may be assigned to a variable other
than ans:

≫ x = pi/7
≫ cos(x)
≫ y = sin(x)^2+cos(x)^2;
≫ y

A semicolon (;) at the end of the line suppresses printing of the result, as
when we calculated y in the next-to-last line just shown. This feature is
especially useful when writing MATLAB programs where intermediate
results are not of interest and when working with large matrices.

MATLAB supports the dynamic creation of variables. You can create your
own variables by just assigning a value to a variable. For example, type x
= 3.5+4.2. Then the real variable x contains the value 7.7. Variable
names must start with an alphabetical character and be less than nineteen
characters long. If you type x = -3*4.0, the content 7.7 is replaced by
the value -12. Some commands allow you to keep track of all the variables
that you have already created in your session. Type who or whos to get the
list and names of the variables currently in memory (whos gives more
information than who). To clear all the variables, type in clear. To clear a
single variable (or several) from the list, follow the command clear by
the name of the variable you want to delete or by a list of variable names
separated by spaces. Try it now.
MATLAB is case sensitive. In other words, x and X are two different
variables. You can control the case sensitivity of MATLAB by entering the
command casesen, which toggles the sensitivity. The command
casesen on enforces case sensitivity, and casesen off cancels it.

If one line is not enough to enter your command, then finish the first line
with two dots (. . ) and continue on the next line. You can enter more than
one command per line by separating them with commas if you want the
result displayed or with semicolons if you do not want the result displayed.
For example, type

≫ theta = pi/7; x = cos(theta); y = sin(theta);

≫ x,y

to first compute theta,cos(theta), and sin(theta) and then to

print x and y.
An Introduction to MATLAB: Complex Variables

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

The number √−1 is predefined in MATLAB and stored in the two variable
locations denoted by i and j. This double definition comes from the
preference of mathematicians for using i and the preference of engineers for
using j (with i denoting electrical current). i and j are variables, and their
contents may be changed. If you type j = 5, then this is the value for j
and j no longer contains √−1. Type in j = sqrt(-1) to restore the
original value. Note the way a complex variable is displayed. If you type i,
you should get the answer

i =
0+1.0000i.

The same value will be displayed for j. Try it. Using j, you can now enter
complex variables. For example, enter z1 = 1+2*j and z2 =
2+1.5*j. As j is a variable, you have to use the multiplication sign *.
Otherwise, you will get an error message. MATLAB does not differentiate
(except in storage) between a real and a complex variable. Therefore
variables may be added, subtracted, multiplied, or even divided. For
example, type in x = 2, z = 4.5*j, and z/x. The real and
imaginary parts of z are both divided by x. MATLAB just treats the real
variable x as a complex variable with a zero imaginary part. A complex
variable that happens to have a zero imaginary part is treated like a real
variable. Subtract 2*j from z1 and display the result.
MATLAB contains several built-in functions to manipulate complex
numbers. For example, real (z) extracts the real part of the complex
number z. Type

≫ z = 2+1.5*j, real(z)

to get the result

z =
2.000+1.500i

ans =
2

Similarly, imag(z) extracts the imaginary part of the complex number z.

The functions abs(z) and angle(z) compute the absolute value
(magnitude) of the complex number z and its angle (in radians). For
example, type

≫ z = 2+2*j;
≫ r = abs(z)
≫ theta = angle(z)
≫ z = r*exp(j*theta)

The last command shows how to get back the original complex number
from its magnitude and angle. This is clarified in Chapter 1: Complex
Numbers.

Another useful function, conj (z), returns the complex conjugate of the
complex number z. If z = x+j*y where x and y are real, then conj
(z) is equal to x-j*y. Verify this for several complex numbers by using
the function conj (z).
An Introduction to MATLAB: Vectors and Matrices

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

As its name indicates, MATLAB is especially designed to handle matrices.

The simplest way to enter a matrix is to use an explicit list. In the list, the
elements are separated by blanks or commas, and the semicolon (;) is used
to indicate the end of a row. The list is surrounded by square brackets [ ].
For example, the statement

≫ A = [1 2 3;4 5 6;7 8 9]

results in the output

A =
1 2 3
4 5 6
7 8 9

The variable A is a matrix of size 3 × 3. Matrix elements can be any

MATLAB expression. For example, the command

≫ x = [-1.3 sqrt(3) (1+2+3)*4/5]

results in the matrix

x =
-1.3000 1.7321 4.8000
We call a matrix with just one row or one column a vector, and a 1 × 1
matrix is a scalar. Individual matrix elements can be referenced with indices
that are placed inside parentheses. Type x(5) = abs(x(1)) to produce
the new vector

x =
-1.3000 1.7321 4.8000 0.000 1.3000

Note that the size of x has been automatically adjusted to accommodate the
new element and that elements not referenced are set equal to 0 (here x(4)).
New rows or columns can be added very easily. Try typing r = [10 11
12],A = [A;r]. Dimensions in the command must coincide. Try r =
[13 14],A = [A;r].

The command size(A) gives the number of rows and the number of
columns of A. The output from size(A) is itself a matrix of size 1 × 2.
These numbers can be stored if necessary by the command [m n] =
size(A). In our previous example, A = [A;r] is a 4 × 3 matrix, so the
variable m will contain the number 4 and n will contain the number 3. A
vector is a matrix for which either m or n is equal to 1. If m is equal to 1, the
matrix is a row vector; if n is equal to 1, the matrix is a column vector.
Matrices and vectors may contain complex numbers. For example, the
statement

≫ A = [1 2;3 4]+j*[5 6;7 8]

and the statement

≫ A = [1+5*j 2+6*j;3+7*j 4+8*j]

are equivalent, and they both produce the matrix

A =
1.0000+5.0000i 2.0000+6.0000i
3.0000+7.0000i 4.0000+8.0000i

Note that blanks must be avoided in the second expression for A. Try typing
≫A = [1 + 5*j 2 + 6*j 2 + 6*j;3 +7*j 4 + 8*j]

What is the size of A now?

MATLAB has several built-in functions to manipulate matrices. The special

character, ', for prime denotes the transpose of a matrix. The statement A =
[ 1 2 3;4 5 6;7 8 9]' produces the matrix

A =
1 4 7
2 5 8
3 6 9

The rows of A' are the column of A, and vice versa. If A is a complex
matrix, then A is its complex conjugate transpose or hermitian transpose.
For an “unconjugate” transpose, use the two-character operator dot-prime (.
'). Matrix and vector variables can be added, subtracted, and multiplied as
regular variables if the sizes match. Only matrices of the same size can be
added or subtracted. There is, however, an easy way to add or subtract a
common scalar from each element of a matrix. For example, x = [1 2 3
4],x = x-1 produces the output

x =
1 2 3 4

x =
0 1 2 3

As discussed in the chapter on linear algebra, multiplication of two matrices

is only valid if the inner sizes of the matrices are equal. In other words, A*B
is valid if the second size of A (number of columns) is the same as the first
size of B (number of rows). Let a represent the element of A in the i
i,j
th

row and the j column. Then the matrix A*B consists of elements
th

Equation:
n

(AB) = a i,k b k,j

i,j

k=1
where n is the number of columns of A and the number of rows of B. Try
typing A = [1 2 3;4 5 6];B = [7;8;9]; A*B. You should get
the result

ans =
50
112

The inner product between two column vectors x and y is the scalar defined
as the product x'*y or equivalently as y'*x For example, x =
[1;2],y = [3;4],x'*y, leads to the result

ans =
11

Similarly, for row vectors the inner product is defined as x*y'. The
Euclidean norm of a vector is defined as the square root of the inner product
between a vector and itself. Try to compute the norm of the vector [1 2 3
4]. You should get 5.4772. The outer product of two column (row)
vectors is the matrix x*y' (x'*y).

Any scalar can multiply or be multiplied by a matrix. The multiplication is

then performed element by element. Try A = [1 2 3;4 5 6;7 8
9];A*2. You should get

ans =
2 4 6
8 10 12
14 26 28

Verify that 2*A gives the same result.

The inverse of a matrix is computed by using the function inv(A) and is

only valid if A is square. If the matrix is singular, meaning that it has no
inverse, a message will appear. Try typing inv(A). You should get
Warning: Matrix is close to singular or badly scal
ed.
Results may be inaccurate. RCOND=2.937385e-18

ans =
1.0e+16*
0.3152 -0.6304 0.3152
-0.6304 1.2609 -0.6304
0.3152 -0.6304 0.3152

The inverse of a matrix may be used to solve a linear system of equations.

For example, to solve the system

you could type A = [1 2 3;1 -2 4;0 -2 1]; b = [2;7;3];

inv(A)*b and get

ans =
1
-1
1

Equation:

1 2 3 x1 2

1 −2 4 x2 = 7 ,

0 −2 1 x3 3

Check to see that this is the correct answer by typing A*[1;-1;1]. What
do you see?

MATLAB offers another way to solve linear systems, based on Gauss

elimination, that is faster than computing the inverse. The syntax is A\b
and is valid whenever A has the same number of rows as b. Try it.

The “Dot” Operator. Sometimes you may want to perform an operation

element by element. In MATLAB, these element-by-element operations are
called array operations. Of course, matrix addition and subtraction are
already element-by-element operations. The operation A. *B denotes the
multiplication, element by element, of the matrices A and B. Make two 3
×3 matrices, A and B, and try

≫ A*B
≫ A.*B

Suppose we want to find the square of each number in A. The proper way to
specify this calculation is

≫ A_squared=A.^2

where the period (dot) indicates an “array operation” to be performed on

each element of A. Without the dot, A is multiplied by A according to the
rules of matrix multiplication described in Chapter 4, giving a totally
different result:

≫ A^2

Subtleties. Because MATLAB can do so many different mathematical

functions with just a few keystrokes, there are times when a very slight
change in what you type will lead to a different result. Using the matrix A
entered earlier, type the following two lines:

≫ 2.^A
≫ 2 .^A %with a space after
the 2.

In the first case, the dot is “absorbed” by the 2 as a decimal point and the ^
is taken as a matrix exponential. But, when the dot is separated from the 2
by a space, it becomes part of the operator (. ^) and specifies that 2 should
be raised to the power of each element in A. The point is, you should be
very careful to type what you mean in an unambiguous way until you are
familiar enough with MATLAB to know how the subtle situations wiIl be
interpreted. An unambiguous way of typing the preceding lines is
≫ (2.)^A %for matrix
exponential
≫ (2).^A %for array
exponential.
An Introduction to MATLAB: The Colon

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

You can use the colon several ways in MATLAB (see help :). Its basic
meaning is a vector of sequential values. For example, type

≫ x = 3:9

to get

x =
3 4 5 6 7 8 9

For increments other than 1, use statements like

≫ x = 1:0.5:4
≫ x = 6:-1:0

Most MATLAB functions will accept vector inputs and produce vector out-
puts. The statement

≫ y = sqrt(1:10)

builds a vector of integers from 1 to 10 and takes the square root of each of
those numbers. Try it.

Now for another subtlety–what is the effect of each of the following

statements and why?
≫ 1+1:5
≫ 1+(1:5)

Appending to a Matrix or Vector. A matrix or vector can be enlarged in

size by appending new values to the old values. Let x=[1 3 5]:
Equation:

x = [x6810] ⇒ x = [1356810]

1 3 5 6 8 10
y = [x; 1 : 6] ⇒ y = [ ].
1 2 3 4 5 6
An Introduction to MATLAB: Graphics

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

In "Complex Numbers", you learn that complex variables can be

represented as points in the plane. MATLAB makes it easy for you to plot
complex variables in a graph. Type z1 = 1+.5*j;plot(z1,'o').
The graph window should be activated and the point z1 displayed by a 'o'.
You must specify the symbol for display, and the authorized symbols for
point display are '.', 'o', '+', and '*'. When you are displaying a
curve (to come later), no type is necessary. MATLAB automatically adjusts
the scale on a graph to accommodate the value of the point being plotted. In
this case, the range is [0, 2] for the real part and [0, 1] for the imaginary
part.

Let's now plot a second complex number by typing z2 = 2+1.5*j;

plot(za,'o'). Note that the second plot command erases the first
plot and changes the scaling to [0, 4] and [0, 3]. Sometimes you may want
to have the points plotted on the same graph. To achieve this, you have to
use the command hold on after the first plot. Try the following:

≫ plot(z2,'o')
≫ hold on
≫ plot(z1,'o')
≫ hold off

The advantage in using the hold command is that there is no limit to the
number of plot commands you can type before the hold is turned off, and
these plots may involve the same variable plotted over a range of values.
You can also use different point displays. A disadvantage of the hold
command is that the scaling is enforced by the first plot and is not adjusted
for subsequent plots. This is why we plot the point z2 first. Try reversing
the order of the plots and see what happens. This means that points outside
the scaling will not be displayed. The command hold off permits
erasing the current graph for the next plot command.

You can freeze the scaling of the graph by using the command axis.
MATLAB gives you the message

Axis scales frozen

ans =
0 4 0 3

This freezes the current axis scaling for subsequent plots. Similarly, if you
type axis a second time, MATLAB resumes the automatic scaling feature
and prints the message

Axis scales auto-ranged

ans =
0 4 0 3

The axis scaling can also be manually enforced by using the command

≫ axis([xmin xmax ymin ymax])

where (xmin, ymin) is the lower left corner and (xmax, ymax) is
the upper right corner of the graph. This scaling remains in effect until the
next axis command is entered (with or without arguments).

Another way to plot several complex numbers on the same graph is to

display them as a curve. For this purpose, you have to store the numbers in
a vector. For example, type z1 = z1,z(2) = z2,plot(z). Note that
the two points are at the two extremes of the line plotted on the graph. If
you specify a symbol, then no line is drawn, just the extreme points. Try
plot(z,'o') .
If you examine your current graph carefully, you will notice that the unit
lengths on the x and y axes are not quite the same. In fact, MATLAB adjusts
the length of an axis to conform to the overall size of the graph window.
What this means is that a 45o line will actually be displayed at an angle
depending on the overall aspect ratio of the graph window. To ensure that
the aspect ratio is equal to 1, you may enter the command
axis('square'). MATLAB will then enforce an aspect ratio equal to 1,
regardless of the aspect ratio for the outside graph window. This ensures
that circles appear as circles and not as ellipses. MATLAB will make the
square graph as large as possible to fit within the graph window. To go back
to the default ratio, just type in axis('normal').

To add labels to your graph, the functions xlabel('text'),

ylabel('text'), and title('text') are useful and self-
explanatory. The argument text contains a string of characters. Add the
label Real on the horizontal axis and the label Imaginary on the vertical
axis of your graph. The command grid draws a grid on your graph. The
grid does not remain in effect for the next plots. Try it.

The plot Instruction. The plot instruction in MATLAB is very versatile. It

can be used to plot several different types of data. Its syntax is
plot(x,y,' symbol and/or color') or plot(y,'symbol
and/or color'). The instruction will plot a vector of data versus
another vector of data. The first vector is referenced to the horizontal axis
and the second to the vertical axis. If only one vector is used, then it is
plotted with reference to the vertical axis while the horizontal axis is
automatically forced to be the index of the vector for the corresponding data
point. The notation inside the apostrophes is optionally used to designate
whether each element of the vector is to be plotted as a single point with a
certain symbol or as a curve with a straight line drawn between each data
value. The colors can also be specified. Possible symbols are (*,o,+,.),
and colors are (r,g,b,w) (red, green, blue, and white). Complex valued
vectors are plotted by making the horizontal axis the real part of the vector
and the vertical axis the imaginary part. Warning: A complex valued vector
will automatically be plotted correctly on the complex plane (instead of real
versus imaginary) only if every element of the vector is complex valued.
Try plot (x,y,'*') , plot (x,'*'), plot (y,'*'), plot
(y,x,'*') and plot (z,'*'), plot (Real (z), Imag
(z),'*') to clarify your understanding of plot. Use
x = [1357], y = [2468], z =[1 + j, 2 + 2j, 3 + 3j].

We may summarize as follows:

(red star—points with the

values of

plot(x,'*r') x

on vertical and indicies

on horizontal)

(line—connected curve
of the value of

plot(y) on vertical and the value

of

on horizontal)

(line—connected curve
plot(x,y,'og') of the value of

on vertical and the value

of
 x

on horizontal)

(circle—points of the
value of

plot(x,y,'og') on vertical and the value

of

on horizontal)

(line—connectedplot of

plot(real(z),imag(z)) z

on the complex plane)

(blue plus—points of

plot(real(z),imag(z),'+b') z

on the complex plane).

Subplots. It is possible to split the graphics screen up into several separate
smaller graphs rather than just one large graph. As many as four subplots
can be created. The MATLAB instruction subplot (mnp) signifies
which of the smaller graphs is to be accessed with the next plot statement.
The mnp argument consists of three digits. The m and n are the numbers of
rows (m) and columns (n) into which the screen should be divided. The p
designates which of the matrix elements is to be used. For example,

Help and Demos. MATLAB has on-line help and a collection of

demonstrations. For a list of available functions, type

≫ help

For help on a specific function, sin for example, type

≫ help sin

To learn how to use colon (;, a very important and versatile character) in
MATLAB, type

≫ help :

The demos will also help you become more familiar with MATLAB and its
capabilities. To run them, type

≫ demo
An Introduction to MATLAB: Editing Files and Creating Functions
(Macintosh)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

If you quit MATLAB now, all the commands you have typed will be lost.
This is where the Edit window is useful. If you choose new from the File
menu, a new window appears with the title Edit 1: Untitled. In this window,
you should type in all the commands you would like MATLAB to execute
at once. When you are finished typing, you may save the file by choosing
save or save as in the file menu and by entering a name for the file. If the
edit window is active (that is, if it appears in front), then choose save and
go from the file menu to save the file and execute it. If the command
window is active, then you can execute the file by entering its name.

Editing Files. To test your understanding of file editing, enter the following
commands in a file named myfile:

clear, clg %Clear variables a

nd graphics
j=sqrt(-1); %To be sure
z1=1+.5*j,z2=2+1.5*j %Enter variables
z3=z1+z2,z4=z1*z2 %Compute sum and p
roduct
axis([0 4 0 4]), %First plot
axis('square'),plot(z1,'o')
hold on %Allow overplot
plot(z2,'o'),plot(z3,'+'), %Other plots
 plot(z4,'*')
hold off

You do not have to type the % sign and the text that follows it. These are
simply comments in a file. They are ignored by the MATLAB interpreter.
You should, however, make a habit of adding comments (preceded by %) to
your file if you want to be able to understand programs that have been
written long ago.

Do not forget to save your file. Such a file is called a script file. It contains
MATLAB commands that could have been entered one by one in the
command window. You have three ways to execute a script file:

i. with the edit window active, choose save and go from the file menu;
ii. with the command window active, enter the file name; or
iii. with the command window active, choose run script... from the M-file
menu. In this case, a menu pops up to ask you which file you want to
execute.

Try each of these three methods in order to get used to their differences.
Figure 1 shows the plot that you should get.

Creating Functions. MATLAB puts many commands at your disposal, and

you just have to enter their names (with or without arguments) to execute
them. Some commands are built in to MATLAB, and others are contained
in files to which you have access (not to modify them, but to see how things
are done). You can try to display the contents of a command file in the
command window by entering the command type filename. Enter type plot.
You should get the message

??? Built-in function.

 A Typical Graph (©Apple
Computer, Inc., used with
permission.)

This means that the command plot is a build-in function and thus cannot
be typed by the user (it is compiled with the program). Similarly, try to type
the function file containing the command sinh by entering the command
type sinh. You should get

function y=sinh(x)
%SINH SINH(X) is the hyperbolic sine of the elemen
ts of X.
y=(exp(x)-exp(-x))/2;

This is a typical example of a function file. It differs from a script file by the
fact that the first line contains the word function. What it does is make
the command y = sinh(x) equivalent to the command y=(exp(x)-
exp(-x))/2 . The variable x is the input argument (there could be more
than one), and y is the output argument (there also could be more than one).
The second line contains comments about the function and its arguments.
They can spread to several lines. Display them by typing the command
help sinh. The third line contains the actual function commands (there
could also be several lines).

As an exercise, enter the function perp(x):

function y=perp(x)
%PERP PERP(X) is a complex value perpendicular to
X.
j=sqrt(-1);
y=j*real(x)-imag(x);

Evaluate perp on various complex numbers. Replace the last line by y =

x*j;. Do you get the same result? Why?

Note that local variables are just that–local. Only the input and out- put
arguments are kept in memory after the function is called and executed. For
more on functions and multiple arguments, see the MATLAB manual. In
the previous example, the variable j is local. If, before using the function
perp, you use the same variable j, its contents will not be affected by the
command perp(z). Verify this.

Normally, while a file (script or function) is executed, the commands are

not displayed on the command window. Just the results are displayed. The
command echo allows you to view all the instructions. This is useful for
debugging and demonstrations. It is turned off by typing echo off.
An Introduction to MATLAB: Editing Files and Creating Functions (PC)

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

You should write a MATLAB program whenever you anticipate exe- cuting
some sequence of statements several times or again in a later session. On an
IBM PC, you may use any text editor to write a program, as long as the file
can be saved in ASCII format without the control codes used by most word
processors. Applicable text editors include Edix, Wordstar, XTree Pro, and
Turbo Pascal's editor.

Editing Files. If you have enough memory, you can run your editor without
leaving MATLAB by using the exclamation point (!), like this:

≫ !EDIX

The exclamation point may be used to execute any DOS command or

program from MATLAB. When the command or program finishes, your
MATLAB variables are just as you left them. Use your editor to write
program lines just as you would type them in MATLAB's command mode.
Then save the file with extension .m in the directory where you will run
MATLAB. Such MATLAB programs are called m-files. You may run your
m-file by typing the file name (without the .m extension) at the MATLAB
command prompt (≫).

Script Files. There are two kinds of m-files, called script files and
functions. Running a script file is exactly like typing the commands it
contains at the ≫ prompt. Your m-file will automatically be a script file
unless you specify otherwise, as described later. Practice by entering,
saving, and running plotsin.m as listed next:
t = -6:.2:6;
y = sin(t);
plot(t,y)
title('SINE')
pause
grid
xlabel('t')
ylabel('sin(t)')

When the pause is executed, you will need to press a key to go on. If you
type whos after running plotsin, you will see that the variables t and y
remain in memory. Comments are important to a script file. They are
marked with the symbol %. Anything following this symbol on a line is
assumed to be a comment and is ignored by the MATLAB program
interpreter.

Functions. Functions differ from script files in that they have designated
input and output variables. Any other variables used within a function are
local variables, which do not remain after the function terminates and which
have no effect on variables outside the function. Many of the functions
supplied with MATLAB are actually m-files. A good example is triu.m:

≫ type triu

The word function at the beginning of the file makes it a function rather
than a script file. The function name in this line must match the file name.
The input variables of triu are x and k, meaning that the first input
argument will be referred to as x and the second as k within the function.
Likewise, the function line designates y as the output. There is nothing
special about the variable names x,k, and y when the function is used. It is
only that whatever inputs and output you use will be referred to as x,k, and
y inside the function. The variables m,n,j, and i are created temporarily
when triu runs and disappear when it terminates. They are local variables
and have no effect on variables with the same names outside the function.
In contrast, a script file has no local variables and does no substitution of
input and output variable names.
As an exercise, enter and save the function perp.m:

function y=perp(x)
% PERP(x) is a complex number perpendicular
to x.
j = sqrt(-1);
y = j*real(x)-imag(x);

Evaluate perp on various complex numbers. Replace the last line by y =

x*j;. Do you get the same result? Why?

Printing Files and Graphics. To display an m-file on the screen, use the
instruction ≫ type filename. To make a copy at a printer, use the
DOS command ≫ !print filename.m. Graphics hardcopy is
available through the commands meta and gpp. See the MATLAB manual
for more information.
An Introduction to MATLAB: Loops and Control

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

MATLAB has control statements like those in most computer languages.

We will only study the for loop here. See the MATLAB manual for details
on if and while statements.

What for loops do is allow a statement or a group of statements to be

repeated. For example,

for i = 1:n,x(i) = 0,end

assigns the value 0 to the first n elements of the array x. If n is less than 1,
the instruction is still valid but nothing will be done. If n is not defined,
then the following message will appear:

??? Undefined function or variable.

Symbol in question==>n

If n contains a real value, the integer part of n is used. If n is a complex

number, the integer part of the real part is taken. (This should, however, be
avoided.) If x is not declared earlier or is of smaller dimension than n, then
extra memory is allocated.

The command end must follow the command for. If it is not present,
MATLAB will wait for remaining statements until you type the command
end, and nothing will be executed in the meantime.
More than one statement may be included in the loop. The statement 1:n is
the way to specify all the integer values between 1 and n. A step different
than 1 may be specified. For example, the statement for
i=1:2:5,x(i)=1,end is equivalent to x(1)=1,x(5)=1. Negative
steps are also allowed, as in i=n:-1:1.

We may use a for loop to draw a circle of radius 1. Type

≫ j=sqrt(-1);
≫ n=360;
≫ for i=1:n,circle(1)=exp(2*j*i*pi/n);end;
≫plot(circle)

Note how easy it is to plot a curve. But also note how slowly the for loop is
executed. MATLAB is not very good at executing things one by one. It
prefers, by far, a vector-oriented statement. Using the range specification as
in the for loop, it is possible to speed up the process by replacing the
explicit for loop by an implicit for loop using the colon, like this:

≫ circle = exp((2*j*pi/n)*[1:n]);

Note how much faster this command is executed. In general, for loops
should be avoided as much as possible. For example, the first for loop you
typed could have been replaced by the command x=zeros(1,n);, which
is much more efficient. The function zeros fills the variable with 0's to the
specified size. Similarly, the function ones fills the variable with 1's. The
size can also be determined by the size of the argument. If A is a matrix of
size m,n, then the command B=ones(a) fills the matrix B with 1's and
forces the matrix B to have exactly the same size as the matrix A.

Avoiding for Loops. Since for loops are very inefficient in MATLAB
(they are computed sequentially, adding several more computations for
every loop), it is preferable to use the matrix capabilities of MATLAB to
replace for loops and speed up processing time.

i. Replace the for loop

≫ for i = 1:10, ≫ x(i) = i; ≫ end;
 with
≫ x = 1:10
to get
x = [1 2 3 4 5 6 7 8 9 10]
ii. Replace the for loop
≫ z = something; ≫ for i = 1:10, ≫ x(i) =
z*i; ≫ end;
with
≫ z = something ≫ x = z*(1:10);
to get
x = [z 2*z 3*z ... 10*z]
iii. Replace the for loop
≫ z = something; ≫ x(1) = z; ≫ for i = 2:10.
≫ x(i) = z*x(i-1); ≫ end;
with
≫ z = something; ≫ x = z.^(1:10);
to get
x = [z z**2 z**3 ... z**10]
iv. Replace the for loop
≫ for i = 0:2:100, ≫ x(i) =
1.0*exp(j*2*pi*i/100); ≫ end;
with
≫ x = 1.0*exp(j*2*pi*(0:2:100)/100);
to get
x = [exp(0) exp(j*4*pi/100) ...
exp(j*200*pi/100)]
The Edix Editor

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

On a PC, creating and editing m-files is done outside of MATLAB. The

Edix editor is one of the visual types of editors. It can be accessed either
from DOS by using > edix filename.m or from MATLAB by using
≫ !edix filename.m. The latter will return you back to MATLAB
after you quit the editor. MATLAB script files must end with a .m
extension. They are also called m-files. Once in the Edix editor, a help list
of the editor's commands can be displayed on the screen by typing “Alt-h.”
(Note: The “Alt” key must be held down while the "h" key is depressed.)
Pressing the “space” key will get you out of the help system.

A new file is created by calling the Edix editor with a file name that has not
been previously used. If the file name already exists on disc, then that file is
retrieved and displayed on the screen and is ready to be edited. Since Edix
is a visual editor, changes are made by moving the cursor to wherever
something needs to be typed in or deleted. Once changes have been made to
a file, then that file needs to be saved (written to disc) by typing “Alt-w.”
When all editing is finished, you may exit from the Edix editor by typing
“Alt-x.” Following are some useful Edix editor commands:

Alt- help-list of all the Edix commands;

h(F7)
Alt-
saves (writes) the file to disc;
w

Alt-x exits the Edix editor;

Alt-d deletes the current line;

the first time used, marks the beginning of a block of lines

Alt-
to be edited; the second time, marks the end of a block of
m
lines to be edited and highlights this block in white; the
(F8)
third time, "un-marks" the highlighted block;

makes a copy of the highlighted block at the current cursor

Alt-c
location (can be repeated);

Alt- (moves the highlighted block to the current cursor location

m (can be repeated);

(F5) put the cursor at the beginning of the line; and

(F6) puts the cursor at the end of the line.

Useful Mathematical Identities

Note:This module is part of the collection, A First Course in Electrical

and Computer Engineering. The LaTeX source files for this collection
were created using an optical character recognition technology, and
because of this process there may be more errors than usual. Please contact
us if you discover any errors.

e
jθ
, cos θ, and sin θ
n ∞ n
jθ θ 1
e =lim n→∞ (1 + j ) = ∑ (jθ) =cos θ + j sin θ
n n=0 n!
n n
∞ (−1) 2n ∞ (−1) 2n+1
cos θ = ∑ θ ; sin θ = ∑ θ
n=0 (2n)! n=0 (2n+1)!

∞ n ∞ n
(−1) (−1)
2n 2n+1
cos θ = θ ; sin θ = θ
(2n)! (2n + 1)!
n=0 n=0

Trigonometric Identities
Equation:
2 2
sin θ+ cos θ = 1

Equation:

sin (θ + φ) =sin θ cos φ+ cos θ sin φ

Equation:

cos (θ + φ) =cos θ cos φ− sin θ sin φ

Equation:

sin (θ − φ) =sin θ cos φ− cos θ sin φ

Equation:

cos (θ − φ) =cos θ cos φ+ sin θ sin φ

Euler's Equations
Equation:
jθ
e =cos θ + j sin θ

Equation:
jθ −jθ
e − e
sin θ =
2j

Equation:
jθ −jθ
e + e
cos θ =
2

De Moivre's Identity
Equation:
n
(cos θ + j sin θ) =cos nθ + j sin nθ

Binomial Expansion
Equation:
N
N
N N N!
n N −n
(x + y) = x y ; =
n n (N − n)!n!
n=0
Equation:
N
N
N
2 =
n
n=0

Geometric Sums
Equation:
∞

k
a
az = |z| < 1
1 − z
k=0

Equation:
N −1 N
a(1 − z )
k
az = z ≠ 1
1 − z
k=0

Taylor's Series
Equation:
∞ k
(x − a)
(k)
f (x) = f (a)
k!
k=0

Equation:

(Maclaurin's Series if a = 0)