0% found this document useful (0 votes)

49 views

Chapter

1. The document discusses the historical figures Jean-Baptiste Joseph Fourier and Heinrich Rudolf Hertz who made important contributions to signals and systems. 2. It provides an introduction to signals and systems, defining a signal as a function of time that represents a physical quantity, and a system as a combination of components that processes input signals to produce output signals. 3. The introduction classifies signals as either deterministic or random, and describes the time-domain and frequency-domain approaches for analyzing deterministic signals.

Uploaded by

Mostafa Rabie

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

49 views

Chapter

Uploaded by

Mostafa Rabie

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 67

2 Signals and systems

Knowing is not enough; we must apply. Willing is not enough, we must do.
johann wolfgang von goethe

HISTORICAL PROFILES

Jean-Baptiste Joseph Fourier (1768–1830), a French mathematician, ﬁrst

presented the series and transform (covered in this chapter) that bears his
name. Fourier’s results were not enthusiastically received by the scientific
world. He could not even get his work published as a paper.
Born in Auxerre, France, as the son of a tailor, Fourier was orphaned at age 8.
He attended a local military college run by Benedictine monks, where he
decided to train for the priesthood. He soon distinguished himself as a student
and made rapid progress, especially in mathematics. By the age of 14, he had completed a study
of the six volumes of Bézout’s Cours de mathématiques. Up until this time there had been a
conflict inside Fourier about whether he should follow a religious life or one of mathematical
research. However, in 1793 a third element was added to this conflict when he became involved
in politics and joined the local Revolutionary Committee.
Like most of his contemporaries, Fourier was swept into the politics of the French Revolution. Due
to his political involvement, he narrowly escaped death twice. In 1798 Fourier joined Napoleon’s
army in its invasion of Egypt as scientific adviser. While in Cairo, Fourier helped found the Cairo
Institute. Fourier returned to France in 1801 and resumed his post as Professor of Analysis at the
École Polytechnique. It was while there that he did his important mathematical work on the
propagation of heat. In 1822, he published his Théorie analytique de la chaleur. In 1826,
Fourier became a member of the French Academy, and in 1827 succeeded
Laplace as the president of the council of the École Polytechnique.
As a politician, Fourier achieved uncommon success, but his fame chiefly rests on
his strikingly original contributions to science and mathematics. His life was not
without problems, however, since his theory of heat still provoked controversy.

Heinrich Rudolf Hertz (1857–1894), a German experimental physicist, demon-

strated that electromagnetic waves obey the same fundamental laws as light.

Downloaded from https://www.cambridge.org/core. The Librarian-Seeley Historical Library, on 14 Dec 2019 at 05:16:59, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/9781316257364.003
16 Signals and systems

Hertz was born into a prosperous family in Hamburg, Germany. He attended the University of
Berlin, where he received his doctoral degree magna cum laude in 1880 under the prominent
physicist Hermann von Helmholtz. He became a professor of physics at Karlsruhe, where he began
his quest for electromagnetic waves. Their existence had been predicted in 1873 by the
mathematical equations of James Clerk Maxwell, a British scientist. Hertz successfully generated
and detected electromagnetic waves; he was the first to establish the fact that light is a form of
electromagnetic radiation. But Hertz saw no practical use for his discovery. Hertz also simplified
Maxwell’s theory to a mathematical formalism and that led to its widespread acceptance. In
1887, Hertz noted for the first time the photoelectric effect of electrons in a molecular structure.
Although Hertz died of blood poisoning at the age of 37, his discovery of electromagnetic waves
paved the way for the practical use of such waves in wireless telegraph, radio, television, radar,
and other communication systems. The unit of frequency, the hertz, bears his name.

2.1 INTRODUCTION

Communication theory involves applying mathematical models and techniques to communi-

cation systems. The purpose of the communication system is to convey a message (infor-
mation), which is usually in the form of signals. In other words, signals are means of
representing information, while systems are means of processing the signals. Although a signal
is not necessarily a function of time, in this book we deal exclusively with signals that are time-
dependent. In general,

A signal is a function of time that represents a physical quantity.

For our purposes, a signal is a waveform that describes or encodes information. In other words,
a signal is a function that carries information. Examples of signals include sine or cosine
signals, audio or radio signals, video or image signals, speech signals, seismic signals, and radar
signals. For example, the voice of my wife is a signal that makes me react in a certain manner.
A signal may be voltage waveform v(t), current waveform i(t), or any other variable.
Signals may be classiﬁed broadly as either deterministic or random. A deterministic (or
nonrandom) signal is one whose value is known precisely at all instants of time. Examples of
such signals include sine waves, square waves, and pulses. A random (or stochastic) signal is
one whose values at any instant of time can only be described statistically using the mean value,
variance, etc. Noise is a good example of a random signal. We will restrict ourselves to
deterministic signals in this chapter; random signals will be treated in Chapter 6.
There are two ways of describing a deterministic signal: time-domain and frequency-
domain techniques. The frequency-domain approach applies line spectral analysis based on
the Fourier series expansion of periodic signals and continuous spectral analysis based on the
Fourier transform of nonperiodic signals. Thus spectral analysis should be regarded as an

Input signal Output signal

Figure 2.1. Block diagram representation of a
x(t) System y(t) system.
(Excitation) (Response)

inestimable mathematical tool for studying communication systems. The spectral analysis can
be done in the laboratory using the spectrum analyzer.
The term system is broad. It is used in politics, education, economics, engineering, etc. For
our purposes, a system is a combination of several components to perform a given task. It may
also be regarded as a device that can manipulate, change, or transmit signals. In fact, any
complete set of mathematical relationships between input and output variables constitutes a
system.

A system is a mathematical model of a physical entity that processes signals (inputs) to

produce new signals (outputs).

The block diagram of a system is shown in Figure 2.1. Examples include a filter, a camera, an
automobile ignition system, an aircraft control system, an audio compact disk (CD) player,
radar, sonar, a computer (hardware), and a computer program (software).
We begin this chapter by understanding the concepts of signals and systems. Then we review
spectral analysis of the signals using Fourier series and Fourier transform, assuming that the
reader has some familiarity with them. We apply the concepts learned in this chapter to filters,
which are important components of communication systems. We finally use MATLAB to
perform some of the signal analysis covered in this chapter.

2.2 CLASSIFICATION OF SIGNALS

As mentioned earlier, the term signal is used for an electric quantity such as a voltage or current
(or even an electromagnetic wave) when it is used for conveying information. Engineers prefer
to call such variables signals rather than mathematical functions of time because of their
importance in communications and other disciplines. Table 2.1 provides a list of common
signals we will encounter.
There are several ways of looking at the same thing. Signals are no exceptions. In addition to
classifying signals as deterministic or random, signals may also be classiﬁed as follows:

• Continuous or discrete: if the independent variable t is continuous (deﬁned for every value
of t, i.e. over a continuum of values of t), the corresponding signal x(t) is said to be a
continuous-time signal. If the independent variable assumes only discrete values t = nT,
where T is ﬁxed and n is a set of integers (e.g. n = 0, 1, 2, 3, . . .), the corresponding
signal x(t) is a discrete-time signal. Examples of continuous and discrete signals are shown
in Figure 2.2.

Table 2.1. A list of common functions

Signals Description

Sinusoidal signal f(t) = A cos t

Unit step function 1, t 0
uðtÞ ¼
0, otherwise

Unit ramp function t, t 0
rðtÞ ¼
0, otherwise

Impulse (or delta) function δðtÞ ¼ 0, t 6¼ 0

at
Exponential function f(t) = e
t
Rectangular pulse 1, τ=2 < t < τ=2
Π ¼
τ 0, otherwise
(
Triangular pulse t jt j
Δ ¼ 1 , τ<t<τ
τ τ
0, otherwise

Sinc function sin t

sincðtÞ ¼
t

Signum function 1, t > 0
sgnðt Þ ¼
1, t < 0

Gaussian function t2 =2σ 2

g ðt Þ ¼ e

(a) (b)
x(t)
y(t)

t t

Figure 2.2. (a) Continuous-time signal; (b) discrete-time signal.

• Analog or digital: while terms continuous and discrete describe the nature of a signal along
the time (horizontal axis), the terms analog and digital describe the signal amplitude (vertical
axis). A digital signal can take a ﬁnite number of values, e.g. 0 and 1. An analog signal has an
amplitude that assumes any value.
• Periodic or aperiodic: a signal x(t) is called periodic if for a constant positive T,

xðtÞ ¼ xðt þ T Þ, for all t (2.1)

The small value of T that satisﬁes Eq. (2.1) is known as the period of x(t). A signal which
cannot satisfy Eq. (2.1) for any T is said to be aperiodic or nonperiodic. Examples of
periodic signals are sinusoidal functions (sine and cosine functions), while examples
of aperiodic functions are exponential and singularity functions (unit step, impulse function,
etc.). Periodic functions are very useful in science and engineering, especially in communi-
cations. The average or mean value Xave of a periodic signal x(t) is given by

ðT
1
X ave ¼ xðt Þdt (2.2)
T
0

• Energy or power: let x(t) represent the voltage signal across a resistance R. The corresponding
current produced is i(t) = x(t)/R so that the instantaneous power dissipated is Ri2(t) = x2(t)/R.
Since we may not know whether x(t) is a voltage or current signal, it is customary to
normalize power calculations by assuming R = 1 ohm. Thus, we may express the instantan-
eous power associated with signal x(t) as x2(t). The total energy of the signal over a time
interval of 2T is
ðT ð∞
lim 2
E¼ jxðt Þj dt ¼ jxðtÞj2 dt (2.3)
T !∞
T ∞

where the magnitude square has been used in case x(t) is a complex-valued signal. Since power
is the time average of the energy, the average power of the signal is

ðT
lim 1
P¼ jxðt Þj2 dt (2.4)
T!∞ 2T
T

A signal is said to be an energy signal when the total energy E of the signal satisﬁes the
condition
0 <E<∞ (2.5)

Similarly, a signal is called a power signal when its average power satisﬁes the condition

0 <P<∞ (2.6)

Thus, an energy signal must have finite power, non-zero energy (P = 0, 0 < E < ∞) and a
power signal must have finite, non-zero power (0 < P < ∞, E = 0). Deterministic and aperiodic
signals are energy signals, whereas periodic and random signals are power signals. From Eqs.
(2.3) and (2.4), we note that an energy signal has zero average power, while a power signal has
infinite energy. In other words, a signal with finite energy has zero average power, whereas a
signal with finite power has infinite energy. Thus, energy signals and power signals are

mutually exclusive. If a signal is a power signal, it cannot be an energy signal and vice versa. Of
course, a signal may be neither an energy nor a power signal. Thus,

x(t) is an energy signal if it has ﬁnite energy, 0 < E < ∞, so that P = 0

x(t) is a power signal if it has ﬁnite power, 0 < P < ∞, so that E = 0
x(t) is neither an energy nor a power signal if neither property is satisﬁed.

EXAMPLE 2.1
Show that the signal

10, t0
xðt Þ ¼
0, t<0
is a power signal.
Solution
From Eq. (2.4),

ðT ðT
lim 1 2 lim 1 lim 1
P¼ x ðt Þdt ¼ 100 dt ¼ ð100T Þ ¼ 50
T!∞ 2T T !∞ 2T T !∞ 2T
T 0

Similarly, from Eq. (2.3), E = ∞. Since P is ﬁnite while the energy is inﬁnite, we conclude that
x(t) is a power signal.
PRACTICE PROBLEM 2.1
Show that the signal y(t) = 2 cos(πt) is a power signal.
Answer: Proof.

2.3 OPERATIONS ON SIGNALS

Having examined different ways a signal can be classiﬁed, we will now consider some
important basic operations on signals. The operations include time shifting, time reﬂecting,
and time scaling.

• Shifting operation: the signal x(t a) denotes a time-shifted version of x(t). If a > 0, the signal
is delayed (or shifted right) by a seconds, as shown in Figure 2.3(a). If a is negative, the
signal is advanced (or shifted left) by a seconds, as illustrated in Figure 2.3(b).

(a)
x(t) x(t – 2)

10 10

–1 0 1 t 0 1 2 3 t
(b)
y(t) y(t + 6)

5 5

–8 –6 –4 –2 0 t
–2 0 2 t

Figure 2.3. Time-shifting a signal: (a) x(t) is delayed by 2 s; (b) y(t) is advanced by 6 s.

(a) (b) (c)

x(t) x(t/3)
x(2t)

2 2 2

–1 0 1 t –1/2 0 1/2 t –3 0 3 t

Figure 2.4. Time-scaling a signal.

• Scaling operation: the signal x(at) is known as the time-scaled version of x(t). If |a| > 1, the
signal x(at) is compressed because it exists in a smaller time interval than x(t). If |a| < 1,
the signal x(at) is expanded since the signal exists in a larger time interval. For example,
given the signal in Figure 2.4(a), its compressed version x(2t) and its expanded version x(t/3)
are shown in Figure 2.4(b) and 2.4(c) respectively. If we regard x(t) as the signal from a tape
recorder, then x(2t) is the signal obtained when the recorder plays twice as fast, and x(t/3) is
the signal we get when the recorder plays one-third the speed.
• Inverting operation: the signal x( t) is the inverted or reﬂected version of x(t) about t = 0. In
other words, x( t) is the mirror image of x(t) about the vertical axis. For example, if x(t) is the
signal from a tape recorder when played forward, then x( t) is the signal when the recorder
plays backward. Note that time inversion is a special case of time scale with the scaling factor
a = 1. Thus, to time-invert x(t), we merely replace t with t, as illustrated in Figure 2.5.

(a) (b)
x(t) x(–t)

4 4

0 1 2 3 t –3 –2 –1 0 t

Figure 2.5. Time-inverting a signal.

EXAMPLE 2.2
Given the signal
8
< t, 0 t 1
xðt Þ ¼ 1, 1 t 2
:
t þ 3, 2 t 3
shown in Figure 2.6, obtain and sketch: (a) x( t + 4), (b) x(2t 1).
Solution
Each of the required signals can be obtained from x(t) in two ways.
(a) Method 1: (Graphical approach) x( t + 4) combines both time inverting and time-shifting.
Since x( t + 4) = x( [t 4]), we ﬁrst time-invert x(t) to get x( t) as shown in Figure 2.7(a)
and then shift x( t) to the right by 4 seconds to get x( t + 4) as shown in Figure 2.7(b).
Method 2: (Analytic approach) To obtain x( t + 4) from the given x(t), replace every t with
t + 4. So we get
8
< t þ 4, 0 t þ 4 1
xð t þ 4Þ ¼ 1, 1 t þ 4 2
:
ð t þ 4Þ þ 3, 2 t þ 4 3
Note that the equality 0 t + 4 ! t 4 and t + 4 1 ! 3 t so that 0 t + 4 1 !
3 t 4. By treating other inequalities the same way, we obtain
8 8
< t þ 4, 3 t 4 < t 1, 1 t 2
xð t þ 4Þ ¼ 1, 2 t 3 ¼ 1, 2 t 3
: :
t 1, 1 t 2 4 t 3t4
which is what we have in Figure 2.7(b).
(b) Method 1: (Graphical approach) x(2t 1) is both time-scaled (compressed) and time-shifted.
Since x(2t 1) = x(2[t 1/2]), we ﬁrst time-scale x(t) to get x(2t) as shown in Figure 2.8(a)
and then shift x(2t) to the right by 1/2 second to get x(2t 1) as shown in Figure 2.8(b).

x(t) Figure 2.6. For Example 2.2.

0 1 2 3 t

(a) Figure 2.7. For Example 2.2.

x(–t)

–3 –2 –1 0 t

(b)
x(–t + 4)

0 1 2 3 4 t

(a) Figure 2.8. For Example 2.2.

x(2t)

0 1 2 t
(b)
x(2t – 1)

0 1/2 1 3/2 2 t

y(t) Figure 2.9. For Practice problem 2.2.

0 1 2 3 t

(a) Figure 2.10. For Practice problem 2.2.

y(2 – t)

–1 0 1 2 t
(b)
y(1 + t/3)

–3 0 3 6 t

Method 2: (Analytic approach) Similarly, to obtain x(2t 1) from x(t), replace every t with
2t 1.
8 8
< 2t 1, 0 2t 1 1 < 2t 1, 1=2 t 1
xð2t 1Þ ¼ 1, 1 2t 1 2 ¼ 1, 1 t 3=2
: :
ð2t 1Þ þ 3, 2 2t 1 3 4 2t, 3=2 t 2
which is the same as the sketch in Figure 2.8(b).

PRACTICE PROBLEM 2.2

Let

2t, 0t1
yðt Þ ¼
3 t, 1t3
which is sketched in Figure 2.9. Determine and sketch: (a) y(2 t), (b) y(1 + t/3).

1 þ t, 1t1
Answers: (a) yð2 t Þ ¼
4 2t, 1t2

2 þ 2t=3, 3t0
(b) yð1 þ t=3Þ ¼
2 t=3, 0 t 6
They are both shown in Figure 2.10.

2.4 CLASSIFICATION OF SYSTEMS

As said earlier, a system relates an input signal x(t) (known as the excitation) to the output
signal y(t) (known as the response). We may write the input-output relationship as

yðtÞ ¼ f ½xðtÞ (2.7)

On the basis of this relationship, we may classify systems as follows:

• Linear or nonlinear: a system is linear when its output is linearly related to its input. It is non-
linear otherwise. A linear system has an advantage in that superposition principle applies.
If y1(t) is the response of x1(t) and y2(t) is the response of x2(t), then a system is linear if it
satisﬁes the following two conditions:
(1) the response of x1(t) + x2(t) is y1(t) + y2(t);
(2) the response of ax1(t) is ay1(t), where a is a constant.
An electric network consisting of linear elements such as resistors, capacitors, and
inductors is a linear system, where an electronics network consisting of nonlinear elements
such as diodes and transistors is a nonlinear system.

• Continuous or discrete: a system is continuous-time when the input and output signals are
time-continuous. It is a discrete-time system if input and output signals are discrete-time.
• Time-varying or time-invariant: a system is said to be time-invariant or ﬁxed if its input-output
relationship does not vary with time. Otherwise, it is said to be time-varying. In other words, if a
time shift in the input signal produces a corresponding time shift in the output signal so that
yðt τÞ ¼ f ½xðt τÞ (2.8)
the system is time-invariant.
A system is time-varying if it does not satisfy Eq. (2.8). For example, suppose the input-
output relationship of a system is described by
yðtÞ ¼ AxðtÞ þ B (2.9)
The system is time-invariant if A and B are time-independent; it is time-varying if A and B
vary with time.
• Causal or noncausal: a signal is causal if it has zero values when t < 0. A causal system
is one whose output signal (response) does not start before the input signal (excitation) is
applied. Causal systems are also referred to as physically realizable or nonanticipatory
systems. A noncausal system is one in which the response depends on the future values
of the input. Such a system is not physically realizable; it does not exist in real life
but is mathematically modeled using a time delay. For example, a system described by
y(t) = f [x(t 1)] is causal, whereas a system described by y(t) = f [x(t + 1)] is noncausal.
• Analog or digital: the input signal to a system determines whether the system is analog or
digital. An analog system is one whose input signal is analog. Examples of analog systems
include analog switches, analog ﬁlters, and analog cellular telephone systems. A digital

system is one whose input is in the form of a sequence of digits. Typical examples of
digital systems include the digital computer, digital ﬁlters, and digital audiotape (DAT)
systems.

2.5 TRIGONOMETRIC FOURIER SERIES

The Fourier techniques, named after the French physicist Jean-Baptiste Fourier (1768–1830)
who first investigated them, are important tools for scientists and engineers. Fourier representa-
tion (Fourier series and Fourier transform) of signals plays a major role in the analysis of
communication systems for at least two main reasons. First, they help characterize signals in
terms of frequency-domain parameters such as bandwidth which are of major concern to
communication engineers. Second, they provide direct physical interpretation. We will consider
Fourier series in this section and treat Fourier transform later in this chapter.
Fourier series allow us to represent a periodic function exactly in terms of sinusoids. Any
practical periodic function f(t) with period T can be expressed as an infinite sum of sine or
cosine functions that are integral multiples of ωo. Thus, f(t) can be expressed as
X∞
f ðt Þ ¼ a0 þ ðan cos nω0 þ bn sin nω0 Þ (2.10)
|{z} n¼1
dc |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
ac

where
2π
ω0 ¼ (2.11)
T
is called the fundamental frequency in radians per second. Equation (2.10) is known as
quadrature Fourier series. The sinusoid sin nω0 t or cos nω0 t is called the nth harmonic of
f(t); it is an odd harmonic if n is odd or an even harmonic if n is even. The constants an and bn
are the Fourier coefficients. The coefficient a0 is the dc component or the average value of f(t).
(Recall that sinusoids have zero average value.) The coefficients an and bn (for n 6¼ 0) are the
amplitudes of the sinusoids in the ac component. Thus,

The Fourier series of a periodic function f(t) is a representation that resolves f(t) into a dc
component and an ac component comprising an inﬁnite series of harmonic sinusoids.

A function that can be represented by a Fourier series as in Eq. (2.10) must meet certain
requirements because the infinite series in Eq. (2.10) may or may not converge. These condi-
tions on f(t) to yield a convergent Fourier series are as follows:
(1) f(t) is single-valued everywhere;
(2) f(t) has a finite number of finite discontinuities in any one period;

(3) f(t) has a ﬁnite number of maxima and minima in any one period;
þT
t oð

(4) The integral jf ðtÞjdt is ﬁnite for any to.

to
These conditions are called Dirichlet conditions. Although they are not necessary conditions,
they are sufficient conditions for a Fourier series to exist. Luckily, all practical periodic
functions obey these conditions.
A major task in Fourier series is the determination of the Fourier coefficients a0 , an , and bn .
The process of determining the coefficients, known as Fourier analysis, involves evaluating the
following integrals

ðT
1
a0 ¼ f ðtÞdt (2.12a)
T
0

ðT
2
an ¼ f ðtÞ cos nωo t dt (2.12b)
T
0

ðT
2
bn ¼ f ðt Þ sin nωo t dt (2.12c)
T
0

Note that the dc component in Eq. (2.12a) is the average value of the signal f(t). In Eq. (2.12), the
interval 0 < t < T is chosen for convenience. The integrals would be the same if we chose instead
the interval T/2 < t < T/2 or to T/2 < t < to + T/2, where to is a constant. Some trigonometric
integrals, provided in Appendix A, are very helpful in Fourier analysis. It can be shown that the
Fourier series of an even periodic function consists of the dc term and cosine terms only (bn = 0)
since cosine is an even function. Similarly, the Fourier series expansion of an odd function has only
sine terms (a0 = 0, an = 0). Table 2.2 provides the Fourier series of some common periodic signals.
An alternative form of Eq. (2.10) is the amplitude-phase form
∞
X
f ðt Þ ¼ A0 þ An cos ðnω0 þ φn Þ (2.13)
n¼1

where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bn
A0 ¼ ao , An ¼ a2n þ b2n , φn ¼ tan 1
(2.14)
an
The plot of the coefficients An versus nω0 is called the amplitude spectrum of f(t); while the plot
of the phase ϕn versus nω0 is the phase spectrum of f(t). Both the amplitude and phase spectra
form the frequency spectrum or line spectrum of f(t). Equation (2.13) is also known as polar
Fourier series. Thus, there are two forms of trigonometric Fourier series: the quadrature form
and the polar form.

Table 2.2. The Fourier series of common signals

Signal Fourier series

1. Square wave
∞
4AX 1
f ðtÞ ¼ sin ð2n 1Þωo t
π n¼1 2n 1

2. Rectangular pulse train

∞
Aτ 2AX 1 nπτ
f ðtÞ ¼ þ sin cos nωo t
T T n¼1 n T

3. Sawtooth wave
∞
A AX sin nωo t
f ðtÞ ¼
2 π n¼1 n

4. Triangular wave
∞
A 4AX cos ð2n 1Þωo t
f ðtÞ ¼
2 2
π n¼1 ð2n 1Þ2

5. Half-wave rectiﬁed sine

∞
A A 4A X cos 2nωo t
f ðtÞ ¼ þ sin ωo t
π 2 π n¼1 4n2 1

6. Full-wave rectiﬁed sine

∞
2A 4AX sin nωo t
f ðtÞ ¼
π π n¼1 4n2 1

The average power of signal f(t) is deﬁned as

ðT
1 2
P¼ f ðtÞdt (2.15)
T
0

If the Fourier series expansion of f(t) in Eq. (2.10) is substituted into Eq. (2.15), we can readily
show that

∞
1X
P ¼ a20 þ a2n þ b2n (2.16)
2 n¼1

indicating that the average power of a periodic signal is the sum of the squares of its
Fourier coefﬁcients. In other words, the total average power is the sum of the powers
in the dc component and the powers in the harmonic components. This conﬁrms the
fact that a periodic signal is a power signal and that every component of its Fourier
series expansion is also a power signal contributing its individual power to the total
power.
The same conclusion can be reached using the polar Fourier series. By substituting Eq. (2.13)
into Eq. (2.15) or substituting Eq. (2.14) into Eq. (2.16), we obtain

∞
1X
P ¼ A20 þ A2 (2.17)
2 n¼1 n

Thus, the signal power can be found either in the time domain using Eq. (2.15) or in the
frequency domain using Eq. (2.16) or (2.17). This is known as Parseval’s theorem:

ðT ∞ ∞
1 2 1X 1X
P¼ f ðt Þdt ¼ a20 þ a2n þ b2n ¼ A20 þ A2 (2.18)
T 2 n¼1 2 n¼1 n
0

EXAMPLE 2.3
Determine the Fourier series of the periodic impulse train shown in Figure 2.11. Obtain the
amplitude and phase spectra.
Solution
The periodic train can be written as

∞
X
f ðt Þ ¼ 10 δðt nT Þ
n¼ ∞

Using the sampling property of the impulse function,

f(t) Figure 2.11. For Example 2.3.

–2T –T 0 T 2T 3T t

An Figure 2.12. For Example 2.3; amplitude

spectrum of the impulse train in Figure 2.11.
20/T

10/T

0 ω0 2ω0 3ω0 4ω0 ω

f(t) Figure 2.13. For Practice problem 2.3.

–2 –1 0 1 2 3 t

–1

ðT
1 10
a0 ¼ 10δðt Þdt ¼
T T
0

ðT
2 20
an ¼ 10δðt Þ cos nω0 dt ¼
T T
0

ðT
2
bn ¼ 10δðt Þ sin nω0 dt ¼ 0
T
0

It is not surprising that bn = 0 since the impulse function is even. From Eq. (2.14),
10 20
A0 ¼ , An ¼ , φn ¼ 0
T T
And Eq. (2.13) becomes
" #
∞
10 X 2π
f ðt Þ ¼ 1þ2 cos nωo t , ωo ¼
T n¼1
T

Since φn is zero, the phase spectrum is zero everywhere. However, the amplitude spectrum is
the plot of An as shown in Figure 2.12.
PRACTICE PROBLEM 2.3
Find the Fourier series expansion of the square wave in Figure 2.13. Plot the amplitude and
phase spectra.

(a) Figure 2.14. For Practice problem 2.3; amplitude

An and phase spectra of the function shown in
Figure 2.13.
4/p

4/3p

4/5p

0 p 2p 3p 4p 5p w

(b)
fn

0 p 2p 3p 4p 5p w

–90°

∞
X 1
Answer: f ðt Þ ¼ π4 sin 2nπt, n ¼ 2k 1. See Figure 2.14 for the amplitude and phase
k¼1
n
spectra.

EXAMPLE 2.4
Obtain the Fourier series for the sawtooth waveform in Figure 2.15 and plot the amplitude and
phase spectra.
Solution
The function can be written as
2
g ðt Þ ¼ t, π<t<π
π
Since g( t) = g(t), g(t) has odd symmetry so that a0 = an = 0. The period of the function is
T = 2π so that ωo = 2π/T = 1.
Tð=2 ðπ ðπ
2 2 2 4
bn ¼ g ðt Þ sin nωo t dt ¼ t sin nt dt ¼ 2 t sin nt dt (2.4.1)
T 2π ππ π
T =2 0

where we have taken advantage of the symmetry of g(t) by integrating from 0 to π and
multiplying by 2. But from Appendix A,
ð
1 t
t sin at dt ¼ 2 sin at cos at (2.4.2)
a a
Applying this to Eq. (2.4.1), we get

g(t) Figure 2.15. For Example 2.4.

–2p –p 0 p 2p 3p t

–2

(a) Figure 2.16. For Example 2.4.

An
1.273

0.637

0.424
0.3183

0 1 2 3 4 w

(b)
fn
90°

0 1 2 3 4 5 w

–90°

ðπ
π
4 4 1 t 4 π 4
bn ¼ 2 t sin nt dt ¼ 2 2 sin nt cos nt ¼ 0 cos πn 0þ0 ¼ cos πn
π π n n 0 π2 n πn
0

Since cos nπ = ( 1)n, the Fourier series expansion of g(t) is

∞
4X ð 1Þnþ1
g ðt Þ ¼ sin nt
π n¼1 n

having only sine terms. Since An = bn and φn = 90 + α, where α is due ( 1)n+1. α = 0 when n
is odd and α = 180 when n is even. Hence,

90 , n ¼ odd
φn ¼
90 , n ¼ even
The amplitude and phase spectra are shown in Figure 2.16.

h(t) Figure 2.17. For Practice problem 2.4.

–2 –1 0 1 2 3 4 t

An Figure 2.18. Answer for Practice problem 2.4

(not to scale).
2
1.62

0.18
0.06

0 p 2p 3p 4p 5p w

PRACTICE PROBLEM 2.4

Determine the Fourier series of the periodic function in Figure 2.17. Sketch the amplitude and
phase spectra.
∞
16
X 1
Answer: hðtÞ ¼ 2 π2 2
cos nπt, n ¼ 2k 1, the phase spectrum is zero and the
k¼1
n
amplitude spectrum is shown in Figure 2.18.

2.6 EXPONENTIAL FOURIER SERIES

Since we know that a sinusoid of frequency nω0 can be expressed in terms of exponentials ejnω0 t
and e jnω0 t , we should intuitively expect that trigonometric Fourier series in Eqs. (2.10) and
(2.13) can also be expressed in exponential form. Although it is more convenient to deal with
the trigonometric Fourier series since no complex algebra is involved, the exponential form is
useful for two reasons. First, it is more compact and it readily leads to the Fourier transform to
be covered in the next section. Second, the trigonometric Fourier generates a one-sided
spectrum, while the exponential Fourier series produces a two-sided spectrum (as we shall
see), which is more convenient to use and is preferred most of the time.
A periodic signal f(t) of period T and frequency ω ¼ 2π=T may be represented as an
exponential Fourier series as

∞
X
f ðt Þ ¼ C n e jnω0 t (2.19)
n¼ ∞

where the complex coefﬁcient is given by

ðT
1 jnω0 t
Cn ¼ f ðtÞe dt ¼ jC n j∠ϕn (2.20)
T
0

where |Cn| and ϕn are the magnitude and phase of Cn. The plots of the magnitude and phase of
Cn versus nω0 are called the complex amplitude spectrum and complex phase spectrum or
simply the exponential spectra of f(t). We should note the following:
(1) C0 corresponds to the dc value of f(t) because Eq. (2.20) is identical to Eq. (2.12a) when
n = 0.
(2) The frequencies present are ω0 (known as the fundamental frequency) and all integer
multiples nω0 (known as the harmonics). The presence of the negative frequencies
contradicts our common notion of frequency being the number of repetitions per unit time.
The negative frequencies are due to the existence of the negative exponential term e jnω0 t .
Notice that each exponential term comes in pairs – positive and negative. The positive
portion on its own does not represent a real signal but when its negative complement is
added the two together form a real signal.
(3) The coefﬁcients of the exponential Fourier series are related to those of the trigonometric
Fourier series by
an jbn ¼ An ∠φn ¼ 2C n ¼ 2jC n j∠ϕn (2.21)
Thus, we can obtain the exponential Fourier series from the trigonometric series, and
vice versa.
As we did for trigonometric series, we can obtain the average power P of signal f(t) from the
exponential series expansion. Substituting Eq. (2.19) into Eq. (2.15) gives
ðT ðT "
∞
#
1 2 1 X
jnω0 t
P¼ f ðt Þdt ¼ f ðt Þ Cne dt
T T n¼ ∞
0 0
Interchanging the order of summation and integration yields
2 T 3
∞ ð ∞ ∞
X 1 X X
P¼ Cn 4 f ðt Þejnω0 t dt 5 ¼ C n C n ¼ jC n j2 (2.22)
n¼ ∞
T n¼ ∞ n¼ ∞
0

where C n is the complex conjugate of C n . Thus, we obtain another form of Parseval’s theorem as:
ðT ∞
1 2 X
P¼ f ðtÞdt ¼ jC n j2 (2.23)
T n¼ ∞
0

This is an alternative and more compact way of expressing Parseval’s theorem. The power
∞
X
spectrum of signal f(t) is jC n j2 . The power spectrum shows how the total power is
n¼ ∞
distributed among the dc and the harmonic components.

EXAMPLE 2.5
Determine the exponential Fourier series of the periodic impulse train of Example 2.3.
Solution
The periodic train can be written as
∞
X
f ðt Þ ¼ 10 δðt nT Þ
n¼ ∞
Using the sampling property of the impulse function,
ðT
1 jnω0 t 10
Cn ¼ 10δðtÞe dt ¼
T T
0
Hence the Fourier series is
∞
10 X 2π
f ðtÞ ¼ ejnωo t , ωo ¼
T n¼ ∞ T
PRACTICE PROBLEM 2.5
Expand the rectangular pulse train shown in Figure 2.19 in exponential Fourier series.
∞
X sin λ jλt
Answer: f ðt Þ ¼ 2 e , λ ¼ nπ 5
n¼ ∞
λ

EXAMPLE 2.6
Find the exponential Fourier series expansion of the periodic function

f ðtÞ ¼ et , 0 < t < 2π with f ðt þ 2πÞ ¼ f ðtÞ:

Solution
Since T = 2π, ω = 2π/T =1. Hence

f(t) Figure 2.19. Rectangular pulse trains for Practice

problem 2.5.
10

–10 –9 –1 0 1 9 11 t

ðT 2π
ð
2π
1 1 1 1 1 2π
Cn ¼ f ðtÞe jnωo t
dt ¼ et e jnt
dt ¼ eð1 jnÞt
¼ e e j2nπ
1
T 2π 2π 1 jn 0 2π ð1 jnÞ
0 0

Using Euler’s identity,

j2nπ
e ¼ cos 2nπ j sin 2nπ ¼ 1 j0 ¼ 1
we obtain
e2π 1 85
Cn ¼ ¼
2π ð1 jnÞ 1 jn
Thus, the complex Fourier series is
∞
X 85
f ðtÞ ¼
n¼ ∞
1 jn
To plot the exponential spectra, we obtain the magnitude and phase of Cn as
85
jC n j ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi , ϕn ¼ tan 1 n
1 þ n2
By inserting in negative and positive values of n, we obtain the amplitude and the phase plots of
Cn versus nωo ¼ n, as in Figure 2.20.

(a) Figure 2.20. For Example 2.6: (a) amplitude

|Cn| spectrum; (b) phase spectrum.
85

60.1

38
26.9
20.6
16.7
–5 –4 –3 –2 –1 0 1 2 3 4 5 nw 0

(b)
fn
90° 78.7°
71.6° 76°
63.4°
45°

–5 –4 –3 –2 –1 0 1 2 3 4 5 nw 0

–90°

(a) Figure 2.21. For Practice problem 2.6:

|Cn| (a) amplitude spectrum; (b) phase spectrum.

0.32

0.16
0.11
0.8

–4 –3 –2 –1 0 1 2 3 4 n
(b)
fn

90°

–4 –3 –2 –1 0 1 2 3 4 n

–90°

g(t) Figure 2.22. For Example 2.7

–5 – 4 –3 –2 –1 0 1 2 3 4 5 t

PRACTICE PROBLEM 2.6

Determine the exponential Fourier series expansion of the periodic function f(t) = t, 1 < t < 1 with
f(t + 2n) = f(t). Show the exponential spectra.
X∞
jð 1Þn jnπt
Answer: f ðtÞ ¼ nπ e : The exponential spectra are shown in Figure 2.21.
n¼ ∞
n6¼0

EXAMPLE 2.7
Determine the power of the periodic signal in Figure 2.22 and show the power spectrum.
Solution
Notice that T = 4 so that ω = 2π/T = π/2. Hence the average power is
T=2
ð ð1
1 2 1 36ð2Þ
P¼ g ðt Þdt ¼ 62 dt ¼ ¼ 18 W
T 4 4
T =2 1

Table 2.3. Exponential Fourier series coefﬁcients for Example 2.9

n 0 1 2 3 4 5

Cn 3 1.91 0 0.636 0 0.382

|Cn|2 Figure 2.23. For Example 2.7; power spectrum

of the signal in Figure 2.22.
9

3.65 3.65

0.405 0.405
0.146 0.146
–5 –3 –1 0 1 3 5 n

ðT ð1
1 h i
1 jnωo t 1 jnπt=2 6 2 jπnt=2 6 jnπ=2
Cn ¼ f ðtÞe dt ¼ 6e dt ¼ e ¼ e ejnπ=2
T 4 4 jπn 1 ð 2jπnÞ
0 1

But
ex e x
sin x ¼
2j
6
Cn ¼ sin ðnπ=2Þ ¼ 3 sinc ðnπ=2Þ
nπ
sin x
where the sinc function is deﬁned as sinc ðxÞ ¼ x . Table 2.3 shows some values of Cn for n =
0, 1, 2, 3, 4, 5. Using Parseval’s theorem,
∞
X ∞
X
P¼ jC n j2 ¼ C 20 þ 2 jC n j2 ¼ 32 þ 2 1:912 þ 02 þ ð 0:6366Þ2 þ 02 þ 0:3822 þ
n¼ ∞ n¼1
¼ 17:4 W
which is 3.35% less than the exact value of 18 W. Greater accuracy can be achieved by taking
more terms in the summation. The power spectrum is the plot of jC n j2 versus n, as in
Figure 2.23.

PRACTICE PROBLEM 2.7

Obtain the power of the periodic signal in Figure 2.24. Sketch the power spectrum.

f(t) Figure 2.24. For Practice problem 2.7.

–2p –p 0 p 2p 3p t

|Cn|2 Figure 2.25. For Practice problem 2.7.

0.25

0.2 0.2

0.022 0.022
0.008 0.008
–5 –3 –1 0 1 3 5 n

Answer: 0.5 W; see Figure 2.25 for the power spectrum.

2.7 FOURIER TRANSFORM

Fourier series analysis has limited applications because it is restricted to periodic power signals,
while many signals of practical interest are nonperiodic energy signals.
However, besides their importance in many applications, Fourier series provide a foundation
for Fourier transform. Fourier transform is a generalization of the complex Fourier series in the
limit, as we will soon see. Fourier transform is by far the most commonly used tool in
communications systems and virtually every field of science and engineering. It finds a wide
range of applications in different fields such as signal processing, linear systems, electromag-
netics, image analysis, filtering, spectroscopy, tomography, partial differential equations, quan-
tum mechanics, and optics.
As a transition from Fourier series to Fourier transform, we assume a periodic signal and let
its period approach infinity. Consider the exponential form of Fourier series in Eq. (2.19),
namely
∞
X
f ðt Þ ¼ cn e jnωo t (2.25)
n¼ ∞

where
T=2
ð
1 jnω0 t
cn ¼ f ðt Þe dt (2.26)
T
T=2

The spacing between adjacent harmonics is

2π
Δω ¼ ðn þ 1Þω0 nω0 ¼ ω0 ¼ (2.27)
T
We substitute Eq. (2.26) into Eq. (2.25) to get
2 3
Tð=2
∞
X 61
f ðt Þe jnω0 t dt 5ejnωo t
7
f ðt Þ ¼ 4
n¼ ∞
T
2 T =2 3
Tð=2
∞
X 6Δω
f ðt Þe jnω0 t dt 5ejnωo t
7
¼ 4 (2.28)
n¼ ∞
2π
2 T =2 3
Tð=2
∞
1 X
f ðt Þe jnω0 t dt 5Δωejnωo t
6 7
¼ 4
2π n¼ ∞
T =2

As we let T ! ∞, the lines in the spectrum frequency come closer and closer so that Δω
becomes the differential frequency increment dω, the summation becomes integration, and the
harmonic frequency nω0 takes the values of frequency ω. Thus, Eq. (2.28) becomes
2 3
ð∞ ð∞
1 4 f ðt Þe jωt dt5ejωt dω
f ðtÞ ¼ (2.29)
2π
∞ ¼∞

The inner integral is known as the Fourier transform of x(t) and is represented by F(ω), i.e.
ð∞
jωt
F ðωÞ ¼ F½ f ðt Þ ¼ f ðt Þe dt (2.30)
∞

where F is the Fourier transform operator.

The Fourier transform of a signal f(t) is the integration of the product of f(t) and e jωt over
the interval from ∞ to +∞.
F(ω) is an integral transformation of f(t) from the time domain to the frequency domain and is
generally a complex function. F(ω) is known as the spectrum of f(t). The plot of its magnitude
|F(ω)| versus ω is called the amplitude spectrum, while that of its phase ∠F ðωÞversus ω is
called the phase spectrum. If a signal is real-valued, its magnitude spectrum is even, i.e. |F(ω)| =
|F( ω)| and its phase spectrum is odd, i.e. ∠F ðωÞ ¼ ∠F ð ωÞ. Both the amplitude and the
phase spectra provide the physical interpretation of the Fourier transform.

We can write Eq. (2.30) in terms of F(ω) and we obtain the inverse Fourier transform as
ð∞
1 1
f ðtÞ ¼ F ½F ðωÞ ¼ F ðωÞejωt dω (2.31)
2π
∞

We say that the signal f(t) and its transform F(ω) form a Fourier transform pair and denote their
relationship by:
f ðt Þ , F ðωÞ (2.32)
We will denote signals by lowercase letters, and their transforms in uppercase letters.
Since the Fourier transform is developed from the Fourier series, it follows that the conditions
for its existence follow from those of the Fourier series, namely, Dirichlet conditions. Specific-
ally, the Fourier transform F(ω) exists when the Fourier integral in Eq. (2.30) converges (i.e.
exists). The Dirichlet conditions are:
1. f(t) is bounded;
2. f(t) has a finite number of maxima and minima;
3. f(t) has a finite number of discontinuities;
4. f(t) is integrable, i.e.
ð∞
jf ðt Þjdt < ∞ (2.33)
∞

These are the sufﬁcient conditions for f(t) so that its Fourier transform exists. For example,
functions such as tu(t) (ramp function) and f ðt Þ ¼ et , t 0 do not have Fourier transforms
because they do not satisfy the conditions above. However, any signal that is a power or energy
signal has a Fourier transform.
Since F(ω) is a complex-valued function, we avoid the complex algebra involved by
temporarily replacing jω with s and then replacing s with jω at the end.

EXAMPLE 2.8
Find the Fourier transform of the following functions: (a) δðtÞ, (b) ejω0 t , (c) sin ω0 t, (d) e at
uðtÞ.

Solution
(a) For the impulse function,

ð∞
jωt

jωt
F ðωÞ ¼ F½δðtÞ ¼ δðtÞe dt ¼ e t¼0 ¼ 1 (2.8.1)
∞

where the shifting property of the impulse function has been applied. Thus,
F½δðtÞ ¼ 1 (2.8.2)

This shows that the magnitude of the spectrum of the impulse function is constant; that is,
all frequencies are equally represented in the impulse function.
(b) From Eq. (2.8.2)

δðtÞ ¼ F 1 ½1
Using the inverse Fourier transform formula in Eq. (2.31),
ð∞
1 1
δðt Þ ¼ F ½1 ¼ 1ejωt dω
2π
∞

or
ð∞
ejωt dω ¼ 2πδðt Þ (2.8.3)
∞

Interchanging variables t and ω results in

ð∞
ejωt dt ¼ 2πδðωÞ (2.8.4)
∞

Using this result, the Fourier transform of the given function is

ð∞ ð∞

F ejωt ¼ e jωo t
e jωt
dt ¼ ejðωo ωÞt
dt ¼ 2πδðω0 ωÞ
∞ ∞

Since the impulse function is an even function, δðω0 ωÞ ¼ δðω ω0 Þ,

F ejω0 t ¼ 2πδðω ω0 Þ (2.8.5)
By simply changing the sign of ω0, we readily obtain

F e jω0 t ¼ 2πδðω þ ω0 Þ (2.8.6)
Also, by setting ω0 = 0,

F½1 ¼ 2πδðωÞ (2.8.7)

(c) By using the result in Eqs. (2.8.5) and (2.8.6), we get

jω0 t

e þ e jω0 t
F½ sin ω0 t ¼ F
2j
1 1 (2.8.8)
¼ F ejω0 t F e jω0 t
2j 2j
¼ jπ ½δðω þ ω0 Þ δðω ω0 Þ

at
at e , t>0
(d) Let xðt Þ ¼ e uðtÞ ¼
0, t<0
ð∞ ð∞ ð∞
jωt at jωt ðaþjωÞt
X ðωÞ ¼ xðt Þe dt ¼ e e dt ¼ e dt
∞ 0 0 (2.8.9)
1 ðaþjωÞt ∞ 1

L ½e at uðtÞ ¼ X ðωÞ ¼ e 0 ¼ a þ jω
a þ jω
PRACTICE PROBLEM 2.8
Determine the Fourier transform of the following functions: (a) the rectangular pulse Π ðt=τ Þ,
(b) δ(t + 3), (c) 2cosω0t.
2 ωτ j3ω
Answers: (a) ω sin 2 ¼ τ sin c ωτ
2 , (b) e , (c) 2π ½δðω þ ω0 Þ δðω ω0 Þ

EXAMPLE 2.9
Obtain the Fourier transform of the signal shown in Figure 2.26.
Solution
ð∞ ð0 ð1
jωt jωt jωt
X ðωÞ ¼ xðt Þe dt ¼ ð AÞe dt þ Ae dt
∞ 1 0
A 0
jωt A jωt 1

¼ e 1 e 0
jω jω
jA
¼ 1 ejω e jω þ 1
ω
j2A
¼ ð1 cos ωÞ
ω

PRACTICE PROBLEM 2.9

Derive the Fourier transform of the triangular pulse deﬁned as

x(t) Figure 2.26. For Example 2.9.

–1 0 1 t

–A

(
jt j
Λðt=τ Þ ¼ 1 , jt j τ
τ
0, jt j > τ
ωτ
Answer: τ sin c2 2

EXAMPLE 2.10
Find the Fourier transform of the two-sided exponential pulse shown in Figure 2.27. Sketch the
transform.
Solution
a jt j eat , t < 0
Let f ðt Þ ¼ e ¼
e at : t > 0

The Fourier transform is

ð∞ ð0 ð∞
jωt at jωt at jωt
F ðωÞ ¼ f ðt Þe dt ¼ e e dt þ e e dt
∞ ∞ 0
11
¼ þ
a jω a þ jω
2a
¼ 2
a þ ω2
F(ω) is real in this case and it is sketched in Figure 2.28.

PRACTICE PROBLEM 2.10

Determine the Fourier transform of the signum function in Figure 2.29, i.e.

1, t > 0
f ðt Þ ¼ sgnðt Þ ¼
1, t < 0
Sketch |F(ω)|.
Answer: See Figure 6.30.

x(t) Figure 2.27. For Example 2.10.

e–a|t|

0 t

X(w) Figure 2.28. Fourier transform of x(t) in

Figure 2.27; for Example 2.10.
2/a

1/a

–a a w

sgn(t) Figure 2.29. The signum function of Practice

problem 2.10.

0 t

–1

|X(w)| Figure 2.30. Fourier transform of the signal in

Figure 2.29.

0 w

2
Answer: jω . See Figure 2.30.

2.8 PROPERTIES OF THE FOURIER TRANSFORM

In this section, we will develop some of the important properties of the Fourier transform and
show how they are used in ﬁnding the transforms of complicated functions from the transforms
of simple functions. For each of the properties, we follow this same pattern: we ﬁrst state it,
derive it, and then illustrate it with an example.

2.8.1 Linearity
Since integrals are linear operators, the linearity property holds for the Fourier transform just as
it holds for Laplace transform. If F1(ω) and F2(ω) are the Fourier transforms of f1(t) and f2(t)
respectively, then

F½a1 f 1 ðt Þ þ a2 f 2 ðt Þ ¼ a1 F 1 ðωÞ þ a2 F 2 ðωÞ (2.34)

where a1 and a2 are constants. This property states that the Fourier transform of a linear
combination of functions is the same linear combination of the transform of the individual
functions. By deﬁnition,
ð∞
jωt
F½a1 f 1 ðt Þ þ a2 f 2 ðt Þ ¼ ½a1 f 1 ðt Þ þ a2 f 2 ðt Þe dt
∞
ð∞ ð∞ (2.35)
jωt jωt
¼ a1 f 1 ðtÞe dt þ a2 f 2 ðtÞe dt
∞ ∞
¼ a1 F 1 ðωÞ þ a2 F 2 ðωÞ
This can be extended to a linear combination of an arbitrary number of signals.
For example, cos ω0 t ¼ 12 ðejω0 t þ e jω0 t Þ. Using the linearity property,
1 jω0 t
F½ cos ω0 t ¼ F e þ F e jω0 t (2.36)
2
¼ π ½δðω ω0 Þ þ δðω þ ω0 Þ
where we have applied Eqs. (2.8.5) and (2.8.6) in Example 2.8.

2.8.2 Time scaling

Let F ðωÞ ¼ F½f ðt Þ and a be a real constant. Then

1 ω
F½ f ðat Þ ¼ F (2.37)
jaj a

resulting in a new frequency ω/a. Equation (2.37) implies that expansion in one domains leads
to compression in the other domain and vice versa. To establish the time-scaling property, we
note that by deﬁnition,

ð∞
jωt
F½ f ðat Þ ¼ f ðat Þe dt
∞

Letting λ = at, dλ = adt so that

ð∞
jωλ=a dλ 1 ω
F½ f ðat Þ ¼ f ðλÞe ¼ F (2.38)
a a a
∞

Notice that if a = 1, then Eq. (2.37) becomes

F½f ð t Þ ¼ F ð ωÞ ¼ F ðωÞ (2.39)
where the asterisk stands for the complex conjugate. This is known as time reversal.
For example, for the rectangular pulse p(t) = AΠ ðt=τ Þ in Practice problem 2.8,
ωτ
F½pðtÞ ¼ Aτ sin c (2.40)
2
Using Eq. (2.37),
Aτ ωτ
F½pð2t Þ ¼ sin c (2.41)
2 4
The frequency scaling property states
1
F½f ðt=aÞ ¼ F ðaωÞ (2.42)
jaj

2.8.3 Time shifting

If F ðωÞ ¼ F½ f ðt Þ and to is a constant, then
jωt o
F½ f ðt to Þ ¼ e F ðωÞ (2.43)

This implies that a delay or time shift in the time domain implies a phase shift in the frequency
domain. To ﬁnd the Fourier transform of a shifted signal, we multiply the Fourier transform of
the original signal by e jωto . Only the phase is affected by time shifting; the magnitude does not
change. To derive this property, we note that by deﬁnition
ð∞
F½ f ðt t o Þ ¼ f ðt t o Þe jωt dt (2.44)
∞
Let λ = t to, dλ = dt, and t = λ + to, so that
ð∞
jωðλþt o Þ
F½ f ðt to Þ ¼ f ðλÞe dλ
∞
ð∞ (2.45)
jωt o jωλ jωt o
¼e f ðλÞe dλ ¼ e X ðωÞ
∞

By following similar steps,

F½f ðt þ to Þ ¼ ejωto F ðωÞ

For example, from Example 2.8(d),

at 1
F½e uðt Þ ¼ (2.46)
a þ jω
aðt 3Þ
We obtain the transform of xðtÞ ¼ e uðt 3Þ as
h
aðt 3Þ
i e j3ω
F e uðt 3Þ ¼ (2.47)
a þ jω

2.8.4 Frequency shifting

This is the dual of the time-shifting property. This property states that if X ðωÞ ¼ F½xðtÞ and ω0
is constant, then

F½ f ðt Þe jω0 t ¼ F ðω ω0 Þ (2.48)

This means that a shifting in the frequency domain is equivalent to a phase shift in the time
domain. By deﬁnition,
ð∞
F½ f ðt Þe jω0 t
¼ f ðtÞejω0 t e jωt
dt
∞
ð∞ (2.49)
jðω ω0 Þt
¼ f ðt Þe dt ¼ F ðω ω0 Þ
∞

For example, since cos ω0 t ¼ 12 ðejω0 t þ e jω0 t

Þ. Applying the property in Eq. (2.48) yields
1 1
F½ f ðt Þ cos ω0 t ¼ F f ðtÞejω0 t þ F f ðt Þe jω0 t
2 2
(2.50)
1 1
¼ F ðω ω0 Þ þ F ðω þ ω0 Þ
2 2
This is an important result in modulation. For example, if the amplitude spectrum of f(t) is as
shown in Figure 2.31(a), then the corresponding amplitude spectrum of f(t) cos ω0t will be as
shown in Figure 2.31(b).

2.8.5 Time differentiation

If F ðωÞ ¼ F½f ðt Þ, then the time differential property gives

F½ f 0 ðtÞ ¼ jωF ðωÞ (2.51)

(a)
|F [x(t)]|

–B 0 B w
(b)
| F [x(t) cos w0t]|

½ X(w +w 0) A/2 ½ X(w –w0)

–w0 – B –w 0 –w0 + B 0 w0 – B w0 w0 + B w

Figure 2.31. Amplitude spectra of: (a) signal f(t); (b) modulated signal x(t) cos ω0t.

This states that the transform of the derivative of f(t) is obtained by multiplying its transform
F(ω) by jω. By deﬁnition,
ð∞
1 1
f ðtÞ ¼ F ½F ðωÞ ¼ F ðωÞejωt dω (2.52)
2π
∞

Taking the derivative of both sides with respect to t gives

ð∞
df ðtÞ jω
¼ F ðωÞejωt dω ¼ jωF 1 ½F ðωÞ
dt 2π
∞

F½f 0 ðtÞ ¼ jωF ðωÞ (2.53)

Repeated applications of Eq. (2.51) give the Fourier transform of the nth derivative of x(t) as
h i
F f ðnÞ ðt Þ ¼ ðjωÞn F ðωÞ (2.54)
at
For example, if f(t) = e u(t), then

f 0 ðt Þ ¼ ae at
uðt Þ þ δðt Þ ¼ af ðt Þ þ δðtÞ (2.55)
where the δ(t) accounts for the discontinuity at t = 0. Taking the Fourier transform of the ﬁrst
and the last terms, we obtain

1
jωF ðωÞ ¼ aF ðωÞ þ 1 ) F ðωÞ ¼ (2.56)
a þ jω
which agrees with what we got in Example 2.8(d).

2.8.6 Frequency differentiation

This property states that if F ðωÞ ¼ F½ f ðt Þ, then

dn
F½ð jtÞn f ðt Þ ¼ F ðωÞ (2.57)
dωn
This property is also called multiplication by a power of t. We establish this by using the basic
deﬁnition of the Fourier transform.
0∞ 1
ð ð∞
dn dn @ jωt A dn
X ðω Þ ¼ f ðt Þe dt ¼ f ðt Þ e jωt dt
dωn dωn dωn
∞ ∞
ð∞ ð∞ (2.58)
n jωt n jωt
¼ f ðt Þð jt Þ e dt ¼ ð jt Þ f ðt Þe dt
∞ ∞
¼ Fðð jtÞn f ðt ÞÞ
For example, from Practice problem 2.8,
ωτ
L ½Π ðt=τ Þ ¼ τ sin c (2.59)
2
Letting n = 1

d ωτ d sin ωτ=2
L ½ jtΠ ðt=τ Þ ¼ τ sin c ¼τ
dω 2 dω ωτ=2
ωτ=2½τ=2 cos ðωτ=2Þ τ=2 sin ðωτ=2Þ
¼τ (2.60)
ðωτ=2Þ2
ωτ=2 cos ðωτ=2Þ sin ðωτ=2Þ
¼
ω2 =2

2.8.7 Time integration

If F ðωÞ ¼ F½f ðt Þ, then
2 3
ðt
F ðωÞ
F4 f ðt Þdt 5 ¼ þ πF ð0ÞδðωÞ (2.61)
jω
∞

This states that the transform of the integral of x(t) is obtained by dividing the transform of x(t) by
jω plus the impulse term that reﬂects the dc component X(0). If we replace ω by 0 in Eq. (2.30),

ð∞
F ð0Þ ¼ f ðt Þdt (2.62)
∞

indicating that the dc component is zero when the integral of x(t) over all time vanishes. The
time integration property in Eq. (2.61) will be proved later when we consider the convolution
property.
For example, we know from Example 2.8(a) that F ½δðtÞ ¼ 1 and that integrating the impulse
function gives the unit step function u(t). By applying Eq. (2.61),
2 3
ðt
1
F½uðtÞ ¼ F4 δðtÞdt5 ¼ þ πδðωÞ (2.63)
jω
∞

2.8.8 Duality
The duality property states that if F(ω) is the Fourier transform of f(t), then the Fourier
transform of f(t) is 2πf( ω); i.e.
F½f ðtÞ ¼ F ðωÞ ) F½F ðt Þ ¼ 2πf ð ωÞ (2.64)
This expresses the fact that the Fourier transform pairs are symmetric. To derive the property,
we recall from Eq. (2.31) that
ð∞
1 1
f ðtÞ ¼ F ½F ðωÞ ¼ F ðωÞejωt dω (2.65)
2π
∞

or
ð∞
2πf ðt Þ ¼ F ðωÞejωt dω (2.66)
∞

Replacing t with t gives

ð∞
jωt
2πf ð t Þ ¼ F ðωÞe dω (2.67)
∞

If we interchange t and ω, we obtain

ð∞
jωt
2πf ð ωÞ ¼ F ðt Þe dt ¼ F½F ðt Þ (2.68)
∞

as expected.

(a) Figure 2.32. A typical illustration of the duality

x(t) X(w)
property of the Fourier transform: (a) transform
of impulse; (b) transform of unit dc level.
1 1

0 t 0 w
(b)
X(t) F [X(t)]

1 2px(w)

0 t 0 w

|t|
For example, if f(t) = e , then
2
F ðωÞ ¼ (2.69)
ω2 þ1
2
By the duality property, the Fourier transform of F ðt Þ ¼ t2 þ1 is
jωj
2πf ð ωÞ ¼ 2πf ðωÞ ¼ 2πe (2.70)
Figure 2.32 illustrates another example of the duality property. If f(t) = δ(t) so that F(ω) =1, as
in Figure 2.32(a), then the Fourier transform of f(t) = 1 is 2πf(ω) as in Figure 2.32(b).

2.8.9 Convolution
If x(t) and h(t) are two signals, their convolution y(t) is given by the convolution integral
ð∞
yðt Þ ¼ hðt Þ xðt Þ ¼ hðτ Þxðt τ Þdτ (2.71)
∞

If X(ω), H(ω), and Y(ω) are the Fourier transforms of x(t), h(t), and y(t) respectively, then

Y ðωÞ ¼ F½hðt Þ xðt Þ ¼ H ðωÞX ðωÞ (2.72)

which states that convolution of the time-domain signals corresponds to multiplying their
Fourier transforms.
To derive the convolution property, we take the Fourier transform of both sides of Eq. (2.71)
and obtain
2 3
ð∞ ð∞
Y ðωÞ ¼ 4 hðτ Þxðt τ Þdτ5e jωt dt (2.73)
∞ ∞

Table 2.4. Properties of the Fourier transform

Property f(t) F(ω)

Linearity a1 f 1 ðt Þ þ a2 f 2 ðtÞ a1 F 1 ðωÞ þ a2 F 2 ðωÞ

1
ω
Scaling f(at) jaj F a

jωa
Time shift f(t a)u(t a) e F ðωÞ

Frequency shift ejωo t f ðt Þ F ðω ωo Þ

Modulation cos ðωo tÞf ðtÞ 1

2 ½F ðω þ ωo Þ þ F ðω ωo Þ

Time differentiation df jωF ðωÞ

dn f ðjωÞn F ðωÞ
dtn

Time integration Ðt F ðωÞ

f ðt Þdt jω þ πF ð0ÞδðωÞ
∞

n
Frequency differentiation tn f ðtÞ ðjÞn dω
d
n F ðωÞ

Time reversal f( t) F ð ωÞ or F ðωÞ

Duality F(t) 2πf( ω)

Time convolution f 1 ðtÞ f 2 ðt Þ F 1 ðωÞF 2 ðωÞ

Frequency convolution f 1 ðtÞf 2 ðtÞ 1

2π F 1 ðωÞ F 2 ðωÞ

If we change the order of integration and factor out h(τ) since it does not depend on t, we get
2∞ 3
ð∞ ð
Y ðωÞ ¼ hðτ Þ4 xðt τ Þe jωt dt5dτ
∞ ∞

The inner integral can be simpliﬁed by letting λ = t τ so that t = λ + τ and dt = dλ.

2∞ 3
ð∞ ð
Y ðωÞ ¼ hðτ Þ4 xðλÞe jωðλþτÞ dλ5dτ
∞ ∞
ð∞ ð∞ (2.74)
jωτ jωλ
¼ hðτ Þe dτ xðλÞe dλ ¼ H ðωÞX ðωÞ
∞ ∞

as expected.
These properties of the Fourier transform are listed in Table 2.4. The transform pairs of some
common functions are presented in Table 2.5.

Table 2.5. Fourier transform pairs

f(t) F(ω) Power or energy signal

δ(t) 1 Power

1 2πδðωÞ Power

u(t) 1 Power
πδðωÞ þ jω

uðt þ τ Þ uðt τÞ 2 sinωωτ Energy

|t| 2 Neither
ω2

sgn(t) 2 Power
jω

at 1
e uðtÞ aþjω
Energy

eat uð t Þ 1
a jω
Energy

tn e at
uðtÞ n! Neither
ðaþjωÞnþ1

ajtj 2a Energy
e a2 þω2

ejωo t 2πδðω ωo Þ Power

sin ωo t jπ ½δðω þ ωo Þ δðω ωo Þ Power

cos ωo t π ½δðω þ ωo Þ þ δðω ωo Þ Power

at ωo
e sin ωo tu(t) ðaþjωÞ2 þω2o
Energy

at aþjω
e cos ωo tu(t) Energy
ðaþjωÞ2 þω2o

ωτ
t 1, jtj < τ=2 τ sinc Energy
Π ¼ 2
τ 0, jtj > τ=2
ωτ
1 jt j=τ, jtj < τ τ sinc2 Energy
Δ τt ¼ 2
0, jtj > τ

a2 t 2 ω2 =4α2 Energy
e e
∞
X ∞
X Power
f ðt nT Þ ωo F ðnωo Þðω nωo Þ, ωo ¼ 2π
T
n¼ ∞ n¼ ∞

∞
X ∞
X Power
δðt nT Þ ωo δðω nωo Þ, ωo ¼ 2π
T
n¼ ∞ n¼ ∞

EXAMPLE 2.11
A signal f(t) has a Fourier transform given by
5ð1 þ jωÞ
F ðωÞ ¼
8 ω2 þ 6jω
Without ﬁnding f(t), ﬁnd the Fourier transform of the following.
(a) f(t 3)
(b) f(4t)
(c) e j2t f ðt Þ
(d) f( 2t)
Solution
We apply the appropriate property for each case.

jω3 5ð1 þ jωÞe jω3

(a) F½ f ðt 3Þ ¼ e F ðωÞ ¼
8 ω2 þ j6ω
5
4 ð1 þ jω=4Þ 5ð4 þ jωÞ

(b) F½ f ð4t Þ ¼ 14 F ω4 ¼ ¼
8 ω2 =16 þ j6ω=4 128 ω2 þ j24ω
5½1 þ jðω þ 2Þ 5ð1 þ jω þ j2Þ
(c) F½e j2t f ðt Þ ¼ F ðω þ 2Þ ¼ 2
¼ 2
8 ðω þ 2Þ þ 6jðω þ 2Þ 4 ω 4ω þ 6jω þ j12
5
ð1 jω=2Þ 5ð2 jωÞ
(d) F½ f ð 2t Þ ¼ 12 F ω2 ¼ 2 ω2 6jω ¼
8 4 2
32 ω2 12jω

PRACTICE PROBLEM 2.11

A signal x(t) has the Fourier transform
9
F ðωÞ ¼
9 þ ω2
Find the Fourier transform of the following signals:
(a) y(t) = f(2t 1)
(b) z(t) = df (2t)/dt
Ðt
(c) hðtÞ ¼ f ðλÞdλ
∞
18e jω=2 18jω 9
Answers: ðaÞ 2
, ðbÞ 2
, ðcÞ πδðωÞ þ
36 þ ω 36 þ ω jωð9 þ ω2 Þ
EXAMPLE 2.12
Determine the Fourier transform of the signal in Figure 2.33.
Solution
Although the Fourier transform of f(t) can be found directly using Eq. (2.30), it is much easier to
ﬁnd it using the derivative property. Taking the ﬁrst derivative of g(t) produces the signal in
Figure 2.34(a). Taking the second derivative gives us the signal in Figure 2.34(b). From this,

x(t) Figure 2.33. For Example 2.12.

–3 –2 –1 0 1 2 3 t

–4

(a) Figure 2.34. For Example 2.12: (a) ﬁrst

x¢(t) derivative of x(t); (b) second derivative of x(t).

–3 –2 –1 0 1 2 3 t
–2

(b)
6d(t – 2)
x²(t)

4d(t + 3)

–3 –2 –1 0 1 2 3 t

–4d(t – 3)

–6d(t + 2)

f 00 ðtÞ ¼ 4δðt þ 3Þ 6δðt þ 2Þ þ 6δðt 2Þ 4δðt 3Þ

Taking the Fourier transform of each term, we obtain

ðjωÞ2 F ðωÞ ¼ 4ej3ω 6ej2ω þ 6e j2ω 4e j3ω

ω2 F ðωÞ ¼ 4ðej3ω e j3ω Þ þ 6ðej2ω e j2ω

Þ
¼ j8 sin 3ω j12 sin 2ω
j
F ðωÞ ¼ 2 ð12 sin 2ω 8 sin 3ωÞ
ω
PRACTICE PROBLEM 2.12
Determine the Fourier transform of the function in Figure 2.35.

y(t) Figure 2.35 For Practice problem 2.12.

–2 –1 0 1 2 t

–4

Answer: Y ðωÞ ¼ j10

ω2 ðsin 2ω 2 sin ωÞ

EXAMPLE 2.13
Find the inverse Fourier transform of:
10jω
(a) GðωÞ ¼
ð jω þ 2Þðjω þ 3Þ
δðωÞ
(b) Y ðωÞ ¼
ðjω þ 1Þðjω þ 2Þ
Solution
(a) To avoid complex algebra, let s = jω. Using partial fractions,
10s 10s A B
GðsÞ ¼ ¼ ¼ þ , s ¼ jω
ð2 sÞð3 þ sÞ ðs 2Þðs þ 3Þ s 2 s þ 3

10ð2Þ
A ¼ ðs 2ÞGðsÞ ¼ ¼ 4
s¼2 2þ3

10ð 3Þ
B ¼ ðs þ 3ÞGðsÞ ¼ ¼ 6
s¼ 3 3 2
4 6
GðωÞ ¼
jω 2 jω þ 3

Taking the inverse Fourier transform of each term,

g ðt Þ ¼ 4e2t uð tÞ 6e 3t uðt Þ
(b) Because of the delta function, we use Eq. (2.31) to ﬁnd the inverse.

1 ∞
ð
δðωÞejωt dω 1 ejωt 1 1 1
yðt Þ ¼ ¼ ¼ ¼
2π ∞ ð2 þ jωÞðjω þ 1Þ 2π ð2 þ jωÞðjω þ 1Þ ω ¼ 0 2π 2 4π

where the sifting property has been applied.

PRACTICE PROBLEM 2.13

Obtain the inverse Fourier transform of:
e j2ω
(a) FðωÞ ¼
1 þ jω
πδðωÞ
(b) GðωÞ ¼
ð5 þ jωÞð2 þ jωÞ
ðt 2 Þ
Answers: (a) f ðtÞ ¼ e uðt 2Þ, (b) g(t) = 0.05

2.9 APPLICATIONS – FILTERS

Fourier analysis has several applications in communications. Such applications include

filtering, sampling, and amplitude modulation. In this section, we limit ourselves to just
filtering. Filtering is the process of separating a desired signal from unwanted signal (noise).
It is a universal tool used in many engineering fields such as electronics, communication, and
signal processing as well as in physics, biology, astronomy, economics, and finance. Communi-
cation systems in particular require extensive applications of filters.
A device or system that is capable of performing filtering is called a filter.

A ﬁlter is a circuit or system that passes certain frequencies of the input signal but rejects or
attenuates other frequencies.

Filters have been used in practical applications for more than eight decades. Filter technology
feeds related areas such as equalizers, impedance matching networks, transformers, shaping
networks, power dividers, attenuators, and directional couplers, and continually provides
practicing engineers with opportunities to innovate and experiment.
As a linear system, a ﬁlter has an input x(t), an output y(t), and an input response h(t). The
three are related by the convolution integral, namely,
ð∞
yðt Þ ¼ hðt Þ xðt Þ ¼ xðλÞhðt λÞdλ (2.75)
0

or Y ðωÞ ¼ F½hðt Þ xðt Þ ¼ H ðωÞX ðωÞ (2.76)

The transfer function of the ﬁlter is

Y ðωÞ
H ðωÞ ¼ ¼ jH ðωÞj∠θ (2.77)
X ðωÞ
where |H(ω)| is the magnitude of H (also known as the amplitude response) and θ is the phase of
H since H is generally complex.

As a filtering device, a filter is characterized by its stopband and passband. The passband of a
filter is the frequency range that the filter passes with little or no attenuation, while the stopband
is the range of frequencies that the filter does not pass (attenuates or eliminates).
As shown in Figure 2.36, there are four types of filter:
1. A low-pass filter passes low frequencies and stops high frequencies, as shown ideally in
Figure 2.36(a), i.e

1, BωB
jH ðωÞj ¼ (2.78)
0, otherwise

2. A high-pass ﬁlter passes high frequencies and rejects low frequencies, as shown ideally in
Figure 2.36(b).

0, BωB
jH ðωÞj ¼ (2.79)
1, otherwise

3. A bandpass ﬁlter passes frequencies within a frequency band and blocks or attenuates
frequencies outside the band, as shown ideally in Figure 2.36(c)

1, B1 jωj B2
jH ðωÞj ¼ (2.80)
0, otherwise

4. A band-stop ﬁlter passes frequencies outside a frequency band and blocks or attenuates
frequencies within the band, as shown ideally in Figure 2.36(d).

0 B 1 jω j B 2
jH ðωÞj ¼ (2.81)
1, otherwise

A filter is said to be ideal if it has a perfectly flat response within the desired frequency range
and zero response outside that range. As is evident from Eqs. (2.78) to (2.81) as well as
Figure 2.38, the magnitude of the transfer function of ideal filters |H(ω)| = 1 in the passband and
|H(ω)| = 0 in the stopband. Unfortunately, ideal filters cannot be built with practical components
such as resistors, inductors, and capacitors. This can be proved by taking the inverse Fourier
transform of the ideal filters; we obtain impulse responses h(t) which are noncausal and
therefore not physically realizable. For this reason, the ideal filters are known as unrealizable
filters. Physically realizable filters have an amplitude response |H(ω)| that varies gradually
without abrupt transitions between passband and stopbands as in Figure 2.38. However,
realizable filters whose characteristics approach those of ideal filters do exist.
Although to cover all types of practically realizable filters would require a whole book, we
will attempt to discuss a standard filter type to gain some insight into how we can make
realizable filters approach the behavior of ideal filters. Standard classes of filters include

(a)
|H(w)|

–B 0 B w
(b)
|H(w)|

–B 0 B w
(c)
|H(w)|

–B2 –B1 0 B1 B2 w

(d)
|H(w)|

–B2 –B1 0 B1 B2 w

Figure 2.36. Frequency response of four types of ideal filter: (a) low-pass filter; (b) high-pass filter; (c)
bandpass filter; (d) band-stop filter.

Butterworth, Chebyshev, elliptic, and Bessel filters. We consider only Butterworth low-pass
filters, which belong to the simplest class of filters. We consider only low-pass filters because
we can construct high-pass, bandpass, and band-stop filters from any low-pass filters by
frequency transformation.

|H(w)|

0.707

n=1

n=2

n=¥ n=8

0 1 2 w/wc

Figure 2.37. Magnitude response of Butterworth ﬁlters.

(a) (b)
1Ω 0.707 H

+
+ + +
x(t) 2F y(t) x(t) 1.414 F y(t)
– – – –

+ +
x(t) 1F 1F y(t)

– –

Figure 2.38. Typical RLC circuit realizations of Butterworth ﬁlters with ωc = 1: (a) ﬁrst-order;
(b) second-order; (c) third-order.

Butterworth filters are commonly used to meet specific design specifications and their
characteristics are readily available and extensively tabulated. They are characterized by the
fact that the square of the magnitude of the frequency response H(ω) is of the form
1
jH ðωÞj2 ¼ 2n (2.82)
ω
1þ ωc

where n is the order of the ﬁlter or the order of the differential equation that will describe the
transfer function in Eq. (2.69); n also corresponds to the number of storage elements (inductors
and capacitors) required to implement the ﬁlter. The parameter ωc =2πfc is the cutoff frequency,

Table 2.6. Coefﬁcients of Butterworth polynomial

n a0 a1 a2 a3 a4 a5 a6

1 1

2 1 1.414 1

3 1 2 2 1

4 1 2.613 3.414 2.613 1

5 1 3.236 5.236 5.236 3.236 1

6 1 3.864 7.464 9.141 7.464 3.864 1

which is the frequency at which the magnitude of the frequency response is 1/√2 times its value
at dc, i.e. |H(ωc)| = |H(0)|/√2. In other words, ωc is the frequency |H(ω)| in dB which is down by
3 dB on the Bode plot. Figure 2.37 illustrates a plot of |H(ω)| for various values of n. It is
evident from the plots that the Butterworth characteristic approaches that of the ideal filter when
n ! ∞. Butterworth filters are known as having a maximally flat frequency response (the
flattest possible curve) because the first 2n 1 derivatives of |H(ω)| are zero at the origin (dc
or ω = 0) for any given n.
Although Eq. (2.69) specifies the response magnitude of Butterworth filters, it does not
provide how to construct or realize the filters. To do this requires a transfer function of the form
K
H ðsÞ ¼ (2.83)
ðs p0 Þðs p1 Þðs p2 Þ . . . ðs pn Þ
where s = jω and p0 to pn are the poles of the filter and K is a constant. Rather than having the
denominator of the Butterworth transfer function in factored form as in Eq. (2.70), we can
multiply the factors and obtain
K
H ðsÞ ¼ (2.84)
a0 sn þ a1 ωc sn 1 þ a2 ω2c sn 2 þ þ an 1 ωcn 1 s þ an ωnc
The coefficients a0 to an (known as the coefficients of Butterworth polynomials or the denomin-
ator of H(s)) are listed in Table 2.6 for n = 1 to n = 6. The constant K can be determined by
realizing that H(s = 0) = 1, i.e. the dc gain is unity.
Figure 2.38 shows typical realizations of the first-, second-, and third-order Butterworth
filters. Notice that the first-order Butterworth filter is identical to an RC low-pass filter and
would not achieve a “good enough” approximation to the ideal filter. The approximation
improves and frequency response becomes “flat enough” in the passband as the order n
increases by adding more storage elements.
To determine the order of a Butterworth filter, it is usually specified that the stopband begins
at ω ¼ ωs with a minimum attenuation of δ. From Eq. (2.82), we need at least

2n
2 1 ωs 1
δ ¼ 2n ! 1
1þ ωs ωc δ2
ωc

Taking the logarithm of both sides leads to

1
log10 δ2
1
n (2.85)
ωs
2 log10 ωc

For example, if δ = 0.001 so that the stopband is at least 20 log10 (0.001) = 60 dB in

magnitude and the stopband begins at ω ¼ ωs ¼ 3ωc , then
log10 ð999999Þ
n ¼ 6:288
2 log10 ð3Þ
Since n must be integer, n = 7 or higher should be selected.

EXAMPLE 2.14
A Butterworth ﬁlter is designed to have a gain of 40 dB at ω ¼ 3ωc . What must the order of
the ﬁlter be? Obtain its transfer function.
Solution
40=20
40 dB ¼ 20 log10 jHj ! jHj ¼ 10 ¼ 0:01:
From Eq. (2.82),
1 2
jH ðωÞj2 ¼ 2n ¼ ð0:01Þ ! 1 þ 32n ¼ 1000
ω
1þ ωc

or 32n ¼ 999
Taking the logarithm of both sides gives

2n log10 3 ¼ log10 999 ! n ¼ 3:1434

Since n must be an integer, a fourth-order filter is required. From Eq. (2.84), when n = 4
K
H ðsÞ ¼
a0 s4 þ a1 ωc s3 þ a2 ω2c s2 þ a3 ω3c s þ a4 ω4c
From Table 2.4, we obtain the coefficients of the Butterworth filter. Since a4 = 1,

H ð0Þ ¼ 1 ¼ K=ω4c ! K ¼ ω4c

Hence,

ω4c
H ðsÞ ¼
s4 þ 2:613ωc s3 þ 3:414ω2c s2 þ 2:613ω3c s þ ω4c

PRACTICE PROBLEM 2.14

If a third-order Butterworth ﬁlter is designed to have a gain of 20 dB at 50 MHz, determine
the cutoff frequency of the ﬁlter and the corresponding transfer function.
ω3
Answer: 10.81 MHz, H ðsÞ ¼ s3 þ2ωc s2 þ2ω
c
2 sþω3 ,
c c
ωc ¼ 6:791 107 rad=s

EXAMPLE 2.15
The circuit in Figure 2.39 is to be designed as a second-order Butterworth filter with a cutoff
frequency of 10 rad/s. Assuming R = 1 Ω, find L and C.
Solution
The given circuit in Figure 2.39 is second-order because two storage elements are involved.
Using current division,
1=sC I
Io ¼ I¼
1=sC þ R þ sL 1 þ sC ðR þ sLÞ
RI
V ¼ I oR ¼
1 þ sRC þ s2 LC
Hence the transfer function of the circuit is
V ðsÞ R R=LC
H ðsÞ ¼ ¼ 2
¼ 2 (2.15.1)
I ðsÞ 1 þ sRC þ s LC s þ sR=L þ 1=LC
From Eq. (2.83) and Table 2.6, the transfer function for the second-order Butterworth filter is
K
H ðsÞ ¼ (2.15.2)
s2 þ 1:414ωc s þ ω2c
Comparing Eqs. (2.15.1) and (2.15.2), we notice that

K ¼ R=LC, ω2c ¼ 1=LC, 1:414ωc ¼ R=L

Given that R = 1 Ω and ωc = 10 rad/s,
R 1
L¼ ¼ ¼ 70:72 mH
1:414ωc 14:14
1 14:14
C¼ 2 ¼ ¼ 141:4 mF
ωc L 100

Io L Figure 2.39. For Example 2.15.

I C R V

L Figure 2.40. For Practice problem 2.15.

+
+
V1 C R V2
–

PRACTICE PROBLEM 2.15

Given the RLC circuit in Figure 2.40 in which R = 1 Ω, ωc ¼ 100 rad/s, ﬁnd L and C such that
H = V2/V1 produces a Butterworth frequency response.
Answer: L = 1.414 H and C = 7.07 mF

2.10 COMPUTATION USING MATLAB

MATLAB is a software package that is used throughout this book. It is particularly useful for
signal analysis. A review of MATLAB is provided in Appendix B for a beginner. This section
shows how to use the software to numerically perform most of the operations we had in this
chapter. Those operations include plotting, Fourier analysis, and ﬁltering. MATLAB has the fft
command for the discrete fast Fourier transform (FFT).

2.10.1 Plotting a signal

The MATLAB command plot can be used to plot x(t). For example, to plot
t
xðt Þ ¼ 2e þ 4 cos ð3t π=6Þ, 1<t<2
we use the following MATLAB statements:
» t=1:0.001:2;
» x=2*exp( t) + 4*cos(3*t pi/6);
» plot(t,x)

where an increment or step size 0.001 is selected. In MATLAB, t and x are taken as vectors and
must be of the same size in order to plot them. MATLAB has no command for ﬁnding the
Fourier transform F(ω), but you can use the command plot to plot F(ω) once you get it.

2.10.2 Fourier series

MATLAB does not provide a command for getting the Fourier series expansion of a signal. But
once the Fourier series expansion is obtained, MATLAB can be used to plot it for a ﬁnite

f(t) Figure 2.41. Rectangular pulse train.

–2 –1 0 1 2 3 t

number of harmonics or check whether the partial sums of the Fourier series approach the exact
signal. To illustrate this, consider the rectangular pulse train in Figure 2.41. Comparing that
signal with the signal in the ﬁrst entry of Table 2.2 shows the signal in Figure 2.41 is only a
raised version of the signal in the table. It is raised by A, where A = 1 and T = 2 or
ω ¼ 2π=T ¼ π. Thus, the Fourier series expansion has a dc value of A and is
∞
4X 1
f ðt Þ ¼ 1 þ sin kπt, k ¼ 2n 1 (2.86)
π n¼1 k

Keep in mind that the Fourier series must be truncated for computational reasons and only a
partial sum is possible even with computers. Let the harmonics be summed from n = 1 to n = N,
where N = 5, and suppose we want to plot f(t) for 2 < t < 2, the MATLAB commands for
generating the partial sum (or truncated series)
N
4X 1
f N ðt Þ ¼ 1 þ sin kπt, k ¼ 2n 1 (2.87)
π n¼1 k

are given below:

N=5;
t=-2:0.001:2;
f0=1.0; % dc component
fN=f0*ones(size(t));
for n=1:N
k=2*n-1;
mag=4/pi;
arg=k*pi*t;
fN = fN + mag*sin(arg)/k;
end
plot(t,fN)

The plot is shown in Figure 2.42 for N = 5. If we increase the value of N to N = 20, the plot
becomes that shown in Figure 2.43. Notice the partial sum oscillates above and below the actual
value of f(t). At the neighborhood of the points of discontinuity (t = 0, 1, 2, . . .), there is
overshoot and damped oscillation. In fact, an overshoot of about 9 percent of the peak value is
always present, regardless of the number of terms used to approximate f(t). This is known as the
Gibbs phenomenon.

2.5

1.5

0.5

–0.5
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2

Figure 2.42. Plot of partial sum fN(t) when N = 5.

2.5

1.5

0.5

–0.5
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2

Figure 2.43. Plot of partial sum fN(t) when N = 20.

We can perform the computation of the partial sum for exponential Fourier series. If we consider
the same rectangular pulse train in Figure 2.41, it can readily be shown that the truncated series is
N
X j
f N ðt Þ ¼ 1 þ C n e jnωo t , ωo ¼ π, Cn ¼ e jnπ
1 (2.88)
n¼ N
nπ
n6¼0

To ease programming, this can be written as

N
X N
X
f N ðt Þ ¼ 1 þ C n ejnπt þ C ne jnπt
(2.89)
n¼1 n¼1

where C n is the complex conjugate of Cn. The commands for computing the partial sum in Eq.
(2.89) are given below. By making N = 5 and 20, we obtain similar results to those shown in
Figures 2.42 and 2.43.
N=20;
t=-2:0.001:2;
f0=1.0; % dc component
fN=f0*ones(size(t));
for n=1:N
mag1= j*(exp(-j*n*pi) -1)/(pi*n);
mag2= conj(mag1);
arg=n*pi*t;
fN = fN + mag1*exp(j*arg) + mag2*exp(-j*arg);
end
plot(t,fN)

2.10.3 Filtering
In section 2.9, we noticed that it is sometimes necessary to be able to find the poles of the
transfer function H(s) of a filter. MATLAB can be used to find the roots of a polynomial by
using the command roots. For example, if the transfer function of a system is
sþ4
H ðsÞ ¼ (2.90)
s3 þ 6s2 þ 11s þ 6
We can find the poles of H(s) or roots of s3 þ 6s2 þ 11s þ 6 = 0 by entering
» roots( [1 6 11 6 ])

or
» den = [1 6 11 6 ]; % denominator of H(s)
» roots(den)

Either way, the roots are provided as 1, 2, and 3. In factored form, H(s) becomes
sþ4
H ðsÞ ¼
ðs þ 1Þðs þ 2Þðs þ 3Þ

Magnitude of the frequency response

0.9

0.8

0.7

0.6
Magnitude

0.5

0.4

0.3

0.2

0.1

0
0 1 2 3 4 5 6 7 8 9 10
Omega

Figure 2.44. Magnitude Bode plot of the frequency response of the third-order Butterworth ﬁlter.

The command buttap can be used to find the zeros and poles of the nth-order Butterworth filter. For
example, the following sequence of MATLAB statements produces the magnitude frequency
response of the fourth-order Butterworth filter. The plot of the response is provided in Figure 2.44.
» [z,p,k ]=buttap(4); % returns the zeros, poles, and constant k of the 4th-order
Butterworth filter
» num=k*poly(z); % forms the numerator
» den=poly(p); % forms the denominator
» [mag,phase,w ]=bode(num,den); % returns magnitude, phase (in degrees), and fre-
quency vector w (automatically)
» plot(w,mag) % plots the magnitude verse w
» title('Magnitude of the frequency response')
» xlabel('Omega')
» ylabel('Magnitude')

Rather than using linear scale for the magnitude, we could use log scale (in dB) by replacing the
plot(w,mag) statement with
» semilogx(w,20*log10(abs(mag)))

Summary
1. A signal is a time-varying function representing messages or information. Signals may be
classiﬁed as continuous or discrete, analog or digital, periodic or aperiodic, energy
or power.
2. A system is a functional relationship between the input x(t) and output y(t). Systems may
be classiﬁed as linear or nonlinear, continuous or discrete, time-varying or time-invariant,
causal or noncausal, analog or digital.
3. Spectral analysis is an inestimable mathematical tool for studying communi-
cation systems. It deals with the description of signals in the frequency domain
using Fourier series for periodic signals and Fourier transform for nonperiodic
signals.
4. The Fourier series of a periodic function is a summation of harmonics of a fundamental
frequency. Any periodic function satisfying Dirichlet conditions can be expressed in terms
of Fourier series in any of these three forms:

X ∞
f ðt Þ ¼ a0 þ ðan cos nω0 þ bn sin nω0 Þ ðquadrature formÞ
|{z} n¼1
dc |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
ac

ðT ðT
1 1
a0 ¼ f ðt Þdt, an ¼ f ðtÞ cos nωo t dt
T T
0 0

ðT
1
bn ¼ f ðtÞ sin nωo t dt
T
0

∞
X
f ðt Þ ¼ A0 þ An cos ðnω0 þ φn Þ ðamplitude phase formÞ
n¼1

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bn
A0 ¼ ao , An ¼ a2n þ b2n , φn ¼ tan 1
an
∞
X
f ðtÞ ¼ C n ejnω0 t ðexponential formÞ
n¼ ∞

ðT
1 jnω0 t
Cn ¼ f ðtÞe dt ¼ jC n j∠ϕn
T
0

5. One form of Parseval’s theorem (for periodic signals) states that the total average power of
a signal is the sum of the average powers of its harmonic components, i.e.

ðT ∞
1 2 X
P¼ f ðt Þdt ¼ jcn j2
T n¼ ∞
0

6. The Fourier transform F(ω) is the frequency-domain representation of f(t),

ð∞
jωt
F ðωÞ ¼ f ðt Þe dt
∞

7. The inverse Fourier transform is

ð∞
1
f ðt Þ ¼ F ðωÞe jωt dω
2π
∞

8. Important Fourier transform properties and pairs are summarized in Tables 2.4 and 2.5.
9. Another form of Parseval’s theorem (for energy signals) states that
ð∞ ð∞
2 1
E¼ f ðt Þdt ¼ jF ðωÞj2 dω
2π
∞ ∞

10. Filters are devices used for removing unwanted frequency components from a signal.
They are classified according to their suppressed frequency bands as low-pass, high-pass,
bandpass, and band-stop.
11. Ideal filters pass all frequency components of the input within their passband and reject
completely all frequency components outside the passband.
12. Butterworth filters are standard or prototype filters that approximate some aspects of ideal
filters by compromising the others.
13. MATLAB is a powerful tool for signal analysis. It is used in this chapter to plot, find
partial sum of a Fourier series, and design filters.

Review questions
2.1 A signal can be both a power signal and an energy signal.
(a) True. (b) False.

1W 0.7654 H 1.848 H Figure 2.45. For Review question 2.10.

1.848 F 0.7654 F 1W

2.2 If the input x(t) and output y(t) of a system are related as y(t) = 10x(t 2), then the
system is:
(a) Time-varying. (b) Time-invariant. (c) Causal. (d) Noncausal.
2.3 Parseval’s theorem implies superposition of average power.
(a) True. (b) False.
2.4 Which of the following signals is NOT a power signal?
(a) 3. (b) u(t). (c) cos 5t. (d) e 2|t|.
2.5 Which of the following signals are energy signals?
(a) 10. (b) sin 4tu(t). (c) δ(t). (d) e 2tu(t).
2.6 If x(t) = 10 + 8 cos t + 4 cos 3t + 2 cos 5t + . . ., the frequency of the sixth harmonic is:
(a) 12. (b) 11. (c) 9. (d) 6.
2.7 Which of these functions does not have a Fourier transform?
(a) etu( t). (b) tetu(t). (c) 1/t. (d) |t|u(t).
2.8 The inverse Fourier transform of δ(ω) is
(a) δ(t). (b) u(t). (c) 1. (d) 1/2π.
2.9 What kind of Butterworth ﬁlters are discussed in this chapter?
(a) Low-pass. (b) High-pass. (c) Bandpass. (d) Band-stop.
2.10 What is the order of the Butterworth ﬁlter shown in Figure 2.45?
(a) 3. (b) 4. (c) 5. (d) 6.

Answers: 2.1 b, 2.2 b,c, 2.3 a, 2.4 d, 2.5 b,d, 2.6 d, 2.7 c, 2.8 d, 2.9 a, 2.10 b

Problems
Sections 2.2 and 2.3 Classiﬁcations and operations on signals
2.1 Deﬁne each of the following terms:
(a) an analog signal
(b) a digital signal
(c) a continuous-time signal
(d) a discrete-time signal

x(t) Figure 2.46. For Problem 2.5.

0 1 2 t

2.2 Give a brief description of each of the following terms:

(a) a periodic signal
(b) an aperiodic signal
(c) a power signal
(d) an energy signal
2.3 Sketch the following signals.
(a) x1 ðtÞ ¼ 3uðt 1Þ uðt 2Þ
(b) x2 ðt Þ ¼ 2Π ð10t Þ
(c) x3 ðt Þ ¼ 5Δðt=4Þ
2.4 Give a sketch of each of the following signals.
(a) y1 ðt Þ ¼ 2 sinc ðπt=3Þ
(b) y2 ðt Þ ¼ Δðt Þ Π ðt 2Þ
(c) y3 ðt Þ ¼ 4 sgnðt þ 2Þ
2.5 Given that x(t) is shown in Figure 2.46, sketch the following signals.
(a) x1 ðt Þ ¼ xð t Þ
(b) x2 ðt Þ ¼ xð2 þ tÞ
(c) x3 ðt Þ ¼ 2xðt Þ þ 1
(d) x4 ðt Þ ¼ xð2tÞ
(e) x5 ðt Þ ¼ xðt=4Þ
2.6 Determine whether the following signals are power or energy signals or neither.
(a) xðtÞ ¼ e t (exponential)
(b) yðt Þ ¼ rðt Þ ¼ tuðtÞ (ramp)
(c) zðt Þ ¼ Δðt Þ (triangular pulse)
2.7 Is it possible to generate a power signal in a lab? Explain.
2.8 If f(t) is a power signal with average power P, ﬁnd the average power of g(t) = af(bt + c),
where a, b, and c are constants.

Section 2.4 Classiﬁcations of systems

2.9 Give a brief description of the following terms.
(a) a linear system
(b) a nonlinear system
(c) a continuous-time system
(d) a discrete-time system

2.10 Deﬁne the following terms.

(a) a time-varying system
(b) a time-invariant system
(c) a causal system
(d) a non-causal system
(e) an analog system
(f) a digital system
2.11 For each of the systems described below, x(t) is the input signal and y(t) is the output.
Specify whether each system is linear or nonlinear.
(a) y(t) = 10 + 2x(t)
(b) y(t) = x(t) + 2x2(t)
(c) y(t) = 3tx(t)
2.12 Repeat Problem 2.11 for the following cases.
Ðt
(a) xð4λÞdλ
∞
(b) y(t) = ln [x(t)]
(c) y(t) = sin(t) x(t)

Section 2.5 Trigonometric Fourier series

2.13 Determine the Fourier series expansion of the signal in Figure 2.47.
2.14 Find the Fourier series of the half-wave rectiﬁed cosine function shown in Figure 2.48.
2.15 Obtain the Fourier series for the periodic signal in Figure 2.49.
2.16 For the signal in Figure 2.50, determine the trigonometric Fourier series. Evaluate f(t) at
t = 2 using the ﬁrst three non-zero harmonics.

f(t) Figure 2.47. For Problem 2.13.

10
5

–4 –3 –2 –1 0 1 2 3 4 5 6 t

f(t) Figure 2.48. For Problem 2.14.

–5 –4 –3 –2 –1 0 1 2 3 4 5 t

f(t) Figure 2.49. For Problem 2.15.

–4 –3 –2 –1 0 1 2 3 4 5 6 t

f(t) Figure 2.50. For Problem 2.16.

–5 –4 –3 –2 –1 0 1 2 3 4 5 t

2.17 A periodic signal is expressed as

f ðt Þ ¼ 4 þ 2 cos ðt þ 15 Þ 0:5 cos ð3t þ 20 Þ þ 0:25 sin ð5t þ 25 Þ
Sketch the magnitude and phase spectra.
2.18 Given that
∞
X 20 3
f ðt Þ ¼ 2π2
cos 2nt sin 2nt
n¼1
n nπ
n¼odd

plot the first five terms of the amplitude and phase spectra for the signal.
2.19 An amplitude modulated (AM) waveform is given by
f ðt Þ ¼ ½40 20 sin ð2πt þ π=6Þ cos 5πt
Show that f(t) can be represented as
f ðt Þ ¼ a1 cos ðω1 t þ θ1 Þ þ a1 cos ðω2 t þ θ2 Þ þ a3 cos ðω3 t þ θ3 Þ
and determine a1 , a2 , a3 , ω1 , ω2 , ω2 , θ1 , θ2 , and θ3 .
2.20 Determine the Fourier series expansion of the signal defined over its period as:
8
< 4t, 0 < t < 1
f ðtÞ ¼ 4, 1 < t < 3
:
8 4t, 3 < t < 4
2.21 A periodic signal f ðtÞ ¼ 2t=π, π=2 < t < π=2 with f(t π) = f(t).
(a) Find its Fourier series expansion.
(b) Calculate the fraction of its power which is contained in the first four harmonics.

f(t) Figure 2.51. For Problem 2.23.

–T –t T t T t

f(t) Figure 2.52. For Problem 2.24.

–4 –3 –2 –1 0 1 2 3 t

Section 2.6 Exponential Fourier series

2.22 Obtain the exponential Fourier series for f ðt Þ ¼ t2 , π < t < π, f(t + 2πn) ¼ f(t).
2.23 For the triangular pulse in Figure 2.51, ﬁnd the exponential Fourier series.
2.24 Find the exponential Fourier series for the clipped sawtooth waveform in Figure 2.52.
2.25 The periodic signal f(t) is represented by
∞
X
f ðt Þ ¼ C n ejnωo t
n¼∞

Determine the coefﬁcients C 0n of each of the following signals.

(a) g(t) ¼ f(t 2)
(b) hðt Þ ¼ 2 dfdtðtÞ
2
(c) yðt Þ ¼ ddt2f dfdt

Sections 2.7 and 2.8 Fourier transform and its properties

2.26 Determine the Fourier transform of the following signals.
(a) xðt Þ ¼ e t sin πt
(b) yðtÞ ¼ 13 ½δðt þ 1=3Þ þ δðt 1=3Þ
(c) zðt Þ ¼ e t sgnðt Þ
2.27 Determine the Fourier transform of the signals in Figure 2.53.
2.28 Find the Fourier transform of the signals in Figure 2.54.
2.29 Determine the Fourier transform of the sawtooth pulse shown in Figure 2.55.

x(t) y(t) z(t)

4
2 2

–1 0 1 t 0 1 2 t –1 0 1 t

Figure 2.53. For Problem 2.27.

(a) (b) Figure 2.54. For Problem 2.28.

f(t) g(t)

3 2

–2 –1 0 1 2 t 0 1 2 3 4 t

–2

h(t) Figure 2.55. For Problem 2.29.

0 t t

2.30 Obtain the Fourier transform of these signals.

(a) f ðt Þ ¼ ð1 þ m cos αtÞ cos βt, ∞<t<∞
sin t, 0 < t < π
(b) g ðt Þ ¼
0, otherwise

2.31 Find the Fourier transform of the pulse

cos πt=τ, jt j < τ=2
pðtÞ ¼
0, otherwise
2.32 The raised cosine pulse is given by
rðt Þ ¼ 10ð1 þ cos πt ÞΠ ðt=2Þ
(a) sketch r(t),
(b) ﬁnd the Fourier transform of r(t).
2.33 Find the inverse Fourier transform of the following:
(a) F 1 ðωÞ ¼ cos ðπω=4Þ
e jω2
(b) F 2 ðωÞ ¼ 4þjω
ω2 þ2
(c) F 3 ðωÞ ¼ ω4 þ3ω 2 þ2

2.34 Determine the inverse Fourier transform of the spectrum in Figure 2.56.
2.35 Obtain f(t) corresponding to the F(ω) shown in Figure 2.57.
2.36 Determine the inverse Fourier transform of the signal whose spectrum is shown in
Figure 2.58.
F(w) Figure 2.56. For Problem 2.34.
2 sin pw

–1 0 1 w

–2

F(w) Figure 2.57. For Problem 2.35.

A/2

–wo – t –wo –wo + t –t 0 t wo – t wo wo + t w

F(w) Figure 2.58. For Problem 2.36.

–1 0 1 w

2.37 Given that the Fourier transform of g(t) is

20
GðωÞ ¼
ð1 þ jωÞ
ﬁnd the Fourier transform of the following:
(a) g( 2t)
(b) (1 + t)g(1 + t)u(1 + t)
(c) t dgdtðtÞ
(d) g(t) cos πt
2.38 The Fourier transform of x(t) is
4 þ jω
X ðωÞ ¼
ω2 þ j2ω þ 3
Find the Fourier transform of these signals
(a) xðt Þe j2t
(b) xðt Þ sin π ðt 1Þ
(c) x(t)*δ(t 2)
ðt
(d) xðτ Þdτ
∞
2.39 Use ðParseval’s theorem to evaluate the following integrals.
∞
sin2 t
(a) 2
dt
ð ∞∞ t
dt
(b) 4
dt
2
∞ ðt þ 4Þ
2.40 Use Parseval’s theorem to determine the energy of the following signals.
(a) xðt Þ ¼ e 3t uðt Þ
(b) yðt Þ ¼ Π ðt=4Þ
2.41 If f(t) = 5Π(t/2) and g(t) = f(t + 2) + f(t 2), determine G(ω).

Section 2.9 Applications – ﬁlters

2.42 Determine the frequency response V o ðωÞ=V i ðωÞ of the circuit in Figure 2.59. What type
of ideal ﬁlter does the circuit represent?

C L Figure 2.59. For Problem 2.42.

+ +

vi(t) R vo(t)

– –

Rs = 1 W L Figure 2.60. For Problem 2.46.

+
+
Vs C RL= 1 W Vo
– –

2.43 The transfer function of a low-pass ﬁlter has

1, jωj B
H ðωÞ ¼
0, otherwise

If the signal xðtÞ ¼ 10e 100πt uðtÞ is applied to the ﬁlter, determine the value of B in rad/s
that causes only one-third of the energy of x(t) to be passed.
2.44 Find the impulse response h(t) of a third-order Butterworth ﬁlter with ωc = 1 rad/s.
2.45 Determine H(s) such that
1
jH ðωÞj2 ¼
1 þ ω6

2.46 Given the RLC circuit in Figure 2.60 in which R = 1 Ω and ωc ¼ 10 rad/s, ﬁnd L and C
such that H = Vo/Vs produces a Butterworth frequency response.

Section 2.10 Computation using MATLAB

2.47 Use MATLAB to plot the following signals.
(a) xðt Þ ¼ t 2 2t þ 3, 2<t<2

(b) yðtÞ ¼ 4 cos ð2πt 12 Þ þ 3 sin 2πt, π<t<π
(c) zðt Þ ¼ 10ð1 e 2t Þ, 0<t<5
2.48 Use MATLAB to plot:
1 ffi
(a) jF ðωÞj ¼ pffiffiffiffiffiffiffiffi
4þω2
, 5<ω<5
2
(b) GðωÞ ¼ 10 sinc ð5ωÞ, 10 < ω < 10
2.49 Use MATLAB to plot the magnitude and phase of

jπω sin πω
F ðωÞ ¼ 10je
1 ω2
for 5 < ω < 5.
2.50 Write a MATLAB script to plot the partial sum of the trigonometric Fourier series of the
signal in Problem 2.13. Take N = 15.
2.51 Develop a MATLAB program to plot the partial sum of the trigonometric Fourier series
in Problem 2.14. Take N = 25.

2.52 Plot a 31-term partial sum of f(t) given in Problem 2.20 using MATLAB.
2.53 The response of the fifth-order Butterworth filter is
1
H ðsÞ ¼
s5 þ 3:236s4 þ 5:236s3 þ 5:236s2 þ 3:236s þ 1
Use MATLAB to find the poles of H(s).
2.54 Use MATLAB to plot the magnitude frequency response of the sixth-order Butterworth
filter. Use a semilog scale.

Signals & Systems
No ratings yet
Signals & Systems
63 pages
Lathi - Signals and Systems Lecture 1 2018
No ratings yet
Lathi - Signals and Systems Lecture 1 2018
28 pages
chapter_1_en_2024_2025chapter_2_en_2024_2025_sam_merou
No ratings yet
chapter_1_en_2024_2025chapter_2_en_2024_2025_sam_merou
10 pages
Signals and Systems Set I
No ratings yet
Signals and Systems Set I
25 pages
SS#1
No ratings yet
SS#1
63 pages
Keywords: Deterministic and Random Signals, Even and Odd Signals
No ratings yet
Keywords: Deterministic and Random Signals, Even and Odd Signals
14 pages
Analogue Communication 1
No ratings yet
Analogue Communication 1
69 pages
Chapter 1 - Introduction To Signals and Systems
No ratings yet
Chapter 1 - Introduction To Signals and Systems
51 pages
Lecture Notes For Signals and Systems
No ratings yet
Lecture Notes For Signals and Systems
12 pages
Principles of Communication
No ratings yet
Principles of Communication
24 pages
Signals and Systems: Lecture Notes
No ratings yet
Signals and Systems: Lecture Notes
107 pages
Signals and Systems
No ratings yet
Signals and Systems
67 pages
Signals and Systems-I PDF
No ratings yet
Signals and Systems-I PDF
94 pages
ECE 334 A Assignment 1
No ratings yet
ECE 334 A Assignment 1
7 pages
Types of Signals
No ratings yet
Types of Signals
22 pages
Signals and Systems: Dr. Mohamed Bingabr University of Central Oklahoma
No ratings yet
Signals and Systems: Dr. Mohamed Bingabr University of Central Oklahoma
44 pages
DEE2363 Signals and Systems: by Mr. Koay Fong Thai
No ratings yet
DEE2363 Signals and Systems: by Mr. Koay Fong Thai
100 pages
Lecture - 1
No ratings yet
Lecture - 1
46 pages
Signal MRCET
No ratings yet
Signal MRCET
63 pages
1 Introduction To Signals
No ratings yet
1 Introduction To Signals
88 pages
Chapter 1 Fundamental Background
No ratings yet
Chapter 1 Fundamental Background
22 pages
Week 1
No ratings yet
Week 1
30 pages
BEE2143 - Chap 1
No ratings yet
BEE2143 - Chap 1
72 pages
Analog Communications
No ratings yet
Analog Communications
69 pages
Signal
No ratings yet
Signal
47 pages
Signal and Production of Signal: Digital Signal (Figure 1b) Refers To More
No ratings yet
Signal and Production of Signal: Digital Signal (Figure 1b) Refers To More
4 pages
Unit 1 Introduction To Fourier Transform
No ratings yet
Unit 1 Introduction To Fourier Transform
91 pages
Lec 2 CH 2
No ratings yet
Lec 2 CH 2
37 pages
Analog Communication: Unit - I
No ratings yet
Analog Communication: Unit - I
43 pages
Unit 1
No ratings yet
Unit 1
30 pages
RKS Signals and Systems Lecture 01 03
No ratings yet
RKS Signals and Systems Lecture 01 03
18 pages
B. P. Lathi - 5th Edition Modern Digital and Analog Communication Systems Chapter 2
No ratings yet
B. P. Lathi - 5th Edition Modern Digital and Analog Communication Systems Chapter 2
33 pages
module 1
No ratings yet
module 1
9 pages
Signals & Systems Lecture Notes
No ratings yet
Signals & Systems Lecture Notes
185 pages
Signals & Systems: Department of Electrical Engineering University College of Engineering & Technology
No ratings yet
Signals & Systems: Department of Electrical Engineering University College of Engineering & Technology
37 pages
Keywords: Introduction To Signals and Systems, Signal Classification, Continuous and
No ratings yet
Keywords: Introduction To Signals and Systems, Signal Classification, Continuous and
12 pages
Unit - 1
No ratings yet
Unit - 1
47 pages
1st Class EEE 2297
No ratings yet
1st Class EEE 2297
17 pages
1.signals and Systems Introduction
No ratings yet
1.signals and Systems Introduction
55 pages
Refresher SignalsStatistics
No ratings yet
Refresher SignalsStatistics
25 pages
Wa0003
No ratings yet
Wa0003
64 pages
Matdid 744681
No ratings yet
Matdid 744681
109 pages
Chapter - 1 - Signals and Systems
No ratings yet
Chapter - 1 - Signals and Systems
70 pages
CAT1 Communications
No ratings yet
CAT1 Communications
37 pages
01 Introduction
No ratings yet
01 Introduction
92 pages
COM336 Lecture 1
No ratings yet
COM336 Lecture 1
65 pages
Communication
No ratings yet
Communication
18 pages
1-dsp SBT PDF
No ratings yet
1-dsp SBT PDF
68 pages
Communication Signals and Systems Design: Ii/Iv - B.Tech, Odd Semester
No ratings yet
Communication Signals and Systems Design: Ii/Iv - B.Tech, Odd Semester
31 pages
ADC Notes For B.tech
No ratings yet
ADC Notes For B.tech
30 pages
Lec1
No ratings yet
Lec1
60 pages
EC241 Part
No ratings yet
EC241 Part
119 pages
Lecture 1 Updated
No ratings yet
Lecture 1 Updated
43 pages
EC241 Part
No ratings yet
EC241 Part
123 pages
Signals & Systems
No ratings yet
Signals & Systems
49 pages
Unit-I Communication Com (Cum (Latin) "With" or "Together With") + Union (Latin) Union With or Union Together With. Introduction To Communication
No ratings yet
Unit-I Communication Com (Cum (Latin) "With" or "Together With") + Union (Latin) Union With or Union Together With. Introduction To Communication
25 pages
Brief Notes On Signals and Systems: C. Sidney Burrus
No ratings yet
Brief Notes On Signals and Systems: C. Sidney Burrus
75 pages
Review of Signals and Systems: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
No ratings yet
Review of Signals and Systems: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
32 pages
Review of Signals and Systems: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
No ratings yet
Review of Signals and Systems: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
32 pages
The Aim and Structure of Physical Theory
From Everand
The Aim and Structure of Physical Theory
Pierre Maurice Marie Duhem
4/5 (6)

Chapter

Uploaded by

Chapter

Uploaded by

2 Signals and systems

Jean-Baptiste Joseph Fourier (1768–1830), a French mathematician, ﬁrst

Heinrich Rudolf Hertz (1857–1894), a German experimental physicist, demon-

Communication theory involves applying mathematical models and techniques to communi-

A signal is a function of time that represents a physical quantity.

Input signal Output signal

A system is a mathematical model of a physical entity that processes signals (inputs) to

2.2 CLASSIFICATION OF SIGNALS

Table 2.1. A list of common functions

Sinusoidal signal f(t) = A cos t

Impulse (or delta) function δðtÞ ¼ 0, t 6¼ 0

Sinc function sin t

Gaussian function t2 =2σ 2

Figure 2.2. (a) Continuous-time signal; (b) discrete-time signal.

xðtÞ ¼ xðt þ T Þ, for all t (2.1)

x(t) is an energy signal if it has ﬁnite energy, 0 < E < ∞, so that P = 0

2.3 OPERATIONS ON SIGNALS

(a) (b) (c)

Figure 2.4. Time-scaling a signal.

Figure 2.5. Time-inverting a signal.

x(t) Figure 2.6. For Example 2.2.

(a) Figure 2.7. For Example 2.2.

(a) Figure 2.8. For Example 2.2.

y(t) Figure 2.9. For Practice problem 2.2.

(a) Figure 2.10. For Practice problem 2.2.

PRACTICE PROBLEM 2.2

2.4 CLASSIFICATION OF SYSTEMS

yðtÞ ¼ f ½xðtÞ (2.7)

2.5 TRIGONOMETRIC FOURIER SERIES

(4) The integral jf ðtÞjdt is ﬁnite for any to.

Table 2.2. The Fourier series of common signals

Signal Fourier series

2. Rectangular pulse train

5. Half-wave rectiﬁed sine

6. Full-wave rectiﬁed sine

The average power of signal f(t) is deﬁned as

Using the sampling property of the impulse function,

f(t) Figure 2.11. For Example 2.3.

An Figure 2.12. For Example 2.3; amplitude

0 ω0 2ω0 3ω0 4ω0 ω

f(t) Figure 2.13. For Practice problem 2.3.

(a) Figure 2.14. For Practice problem 2.3; amplitude

g(t) Figure 2.15. For Example 2.4.

(a) Figure 2.16. For Example 2.4.

Since cos nπ = ( 1)n, the Fourier series expansion of g(t) is

h(t) Figure 2.17. For Practice problem 2.4.

An Figure 2.18. Answer for Practice problem 2.4

PRACTICE PROBLEM 2.4

2.6 EXPONENTIAL FOURIER SERIES

where the complex coefﬁcient is given by

f ðtÞ ¼ et , 0 < t < 2π with f ðt þ 2πÞ ¼ f ðtÞ:

f(t) Figure 2.19. Rectangular pulse trains for Practice

Using Euler’s identity,

(a) Figure 2.20. For Example 2.6: (a) amplitude

(a) Figure 2.21. For Practice problem 2.6:

g(t) Figure 2.22. For Example 2.7

PRACTICE PROBLEM 2.6

Table 2.3. Exponential Fourier series coefﬁcients for Example 2.9

Cn 3 1.91 0 0.636 0 0.382

|Cn|2 Figure 2.23. For Example 2.7; power spectrum

PRACTICE PROBLEM 2.7

f(t) Figure 2.24. For Practice problem 2.7.

|Cn|2 Figure 2.25. For Practice problem 2.7.

Answer: 0.5 W; see Figure 2.25 for the power spectrum.

2.7 FOURIER TRANSFORM

The spacing between adjacent harmonics is

where F is the Fourier transform operator.

Interchanging variables t and ω results in

Using this result, the Fourier transform of the given function is

Since the impulse function is an even function, δðω0 ωÞ ¼ δðω ω0 Þ,

F½1 ¼ 2πδðωÞ (2.8.7)

(c) By using the result in Eqs. (2.8.5) and (2.8.6), we get

PRACTICE PROBLEM 2.9

x(t) Figure 2.26. For Example 2.9.

The Fourier transform is

PRACTICE PROBLEM 2.10

yðtÞ ¼ f ½xðtÞ (2.7)

F½1 ¼ 2πδðωÞ (2.8.7)

F½a1 f 1 ðt Þ þ a2 f 2 ðt Þ ¼ a1 F 1 ðωÞ þ a2 F 2 ðωÞ (2.34)

F½f ðt þ to Þ ¼ ejωto F ðωÞ

F½ f 0 ðtÞ ¼ jωF ðωÞ (2.51)

F½f 0 ðtÞ ¼ jωF ðωÞ (2.53)

Y ðωÞ ¼ F½hðt Þ xðt Þ ¼ H ðωÞX ðωÞ (2.72)

sin ωo t jπ ½δðω þ ωo Þ δðω ωo Þ Power

cos ωo t π ½δðω þ ωo Þ þ δðω ωo Þ Power

or Y ðωÞ ¼ F½hðt Þ xðt Þ ¼ H ðωÞX ðωÞ (2.76)