POKHARA UNIVERSITY

NEPAL ENGINEERING COLLEGE

A Course Manual

on

Signals & Systems

By

Asst. Prof. Bijaya Shrestha

Department of Electronics & Communication Engineering

July, 2015
Contents

1 Introduction 1
1.1 Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Classification of Signals . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2.1 Continuous-Time and Discrete-Time Signals . . . . . . . . . . 1
1.2.2 Even and Odd Signals . . . . . . . . . . . . . . . . . . . . . . 2
1.2.3 Periodic and Nonperiodic Signals . . . . . . . . . . . . . . . . 4
1.2.4 Deterministic and Random Signals . . . . . . . . . . . . . . . 5
1.2.5 Real and Complex Signals . . . . . . . . . . . . . . . . . . . . 5
1.2.6 Energy and Power Signals . . . . . . . . . . . . . . . . . . . . 5
1.2.7 Analog and Digital Signals . . . . . . . . . . . . . . . . . . . . 7
1.2.8 Causal, Anti-Causal, and Non-Causal Signals . . . . . . . . . 7
1.3 Transformation of the Independent Variable . . . . . . . . . . . . . . 8
1.3.1 Time Shifting . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Time Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.3 Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.4 Precedence Rule for Time Shifting and Time Scaling . . . . . 10
1.4 Elementary Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.1 Sinusoidal Signals . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.2 Exponential Signals . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.3 Unit Impulse Function . . . . . . . . . . . . . . . . . . . . . . 16
1.4.4 Unit Step Function . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4.5 Ramp Function . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4.6 Relationship Between Unit Impulse Function, Unit Step Func-
tion, and Ramp Function . . . . . . . . . . . . . . . . . . . . . 20
1.5 Some Other Useful Signals . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5.1 Signum Function . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5.2 Rectangular Pulse or Gate Function . . . . . . . . . . . . . . . 22
1.5.3 Sinc Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6 Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

i
CONTENTS CONTENTS

1.7 Interconnections of Systems . . . . . . . . . . . . . . . . . . . . . . . 24

1.8 Properties of Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.8.1 Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.8.2 Invertibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.8.3 Causality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.8.4 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.8.5 Time Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.8.6 Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2 Linear Time-Invariant Systems 32

2.1 The Convolution Sum . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 The Convolution Integral . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.3 Properties of LTI Systems . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.1 The Commutative Property . . . . . . . . . . . . . . . . . . . 41
2.3.2 The Distributive Property . . . . . . . . . . . . . . . . . . . . 41
2.3.3 The Associative Property . . . . . . . . . . . . . . . . . . . . 42
2.3.4 LTI Systems with and without Memory . . . . . . . . . . . . . 43
2.3.5 Invertibility of LTI Systems . . . . . . . . . . . . . . . . . . . 44
2.3.6 Causality for LTI Systems . . . . . . . . . . . . . . . . . . . . 44
2.3.7 Stability for LTI Systems . . . . . . . . . . . . . . . . . . . . . 46
2.3.8 The Unit Step Response of an LTI System . . . . . . . . . . . 46
2.4 Causal LTI Systems Described by Differential and Difference Equations 47
2.4.1 Linear Constant-Coefficient Differential Equations . . . . . . . 48
2.4.2 Linear Constant-Coefficient Difference Equations . . . . . . . 50
2.5 Block Diagram Representations of First-Order Systems Described by
Differential and Difference Equations . . . . . . . . . . . . . . . . . . 52
2.6 Convolution of a Rectangular Pulse Passed Through an RC Filter . . 53

3 Fourier Analysis for Continuous-Time and Discrete-Time Signals 57

3.1 Continuous-Time Fourier Series (CTFS) . . . . . . . . . . . . . . . . 57
3.1.1 Complex Exponential Fourier Series . . . . . . . . . . . . . . . 57
3.1.2 Trigonometric Fourier Series . . . . . . . . . . . . . . . . . . . 63
3.1.3 Compact Trigonometric Fourier Series . . . . . . . . . . . . . 67
3.1.4 Conversion of Trigonometric Fourier Series into Complex Ex-
ponential Fourier Series . . . . . . . . . . . . . . . . . . . . . 68
3.1.5 Conversion of Complex Exponential Fourier Series into Trigono-
metric Fourier Series . . . . . . . . . . . . . . . . . . . . . . . 69
3.2 Properties of Continuous-Time Fourier Series . . . . . . . . . . . . . . 70
3.3 Periodicity of Continuous-Time and Discrete-Time Complex Expo-
nential Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Signals & Systems ii Asst. Prof. Bijaya Shrestha, nec

CONTENTS CONTENTS

3.3.1 Continuous-Time Complex Exponential Signal, x(t) = ejω0 t . . 78

3.3.2 Discrete-Time Complex Exponential Signal, x[n] = ejω0 n . . . 79
3.4 Discrete-Time Fourier Series (DTFS) . . . . . . . . . . . . . . . . . . 79
3.5 Properties of Discrete-Time Fourier Series . . . . . . . . . . . . . . . 86
3.6 Continuous-Time Fourier Transform (CTFT) . . . . . . . . . . . . . . 89
3.7 Properties of Continuous-Time Fourier Transform . . . . . . . . . . . 98
3.8 Discrete-Time Fourier Transform (DTFT) . . . . . . . . . . . . . . . 107
3.9 Properties of Discrete-Time Fourier Transform (DTFT) . . . . . . . . 112

4 Discrete Fourier Transform (DFT) 119

4.1 Frequency-Domain Sampling: Discrete Fourier Transform (DFT) . . . 119
4.1.1 Frequency-Domain Sampling and Reconstruction of Discrete-
Time Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.1.2 Discrete Fourier Transfrom (DFT) Pairs . . . . . . . . . . . . 121
4.1.3 DFT as a Linear Transformation . . . . . . . . . . . . . . . . 123
4.1.4 Relationship of DFT to Other Transforms . . . . . . . . . . . 126
4.2 Properties of DFT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.2.1 Periodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.2.2 Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.2.3 Circular Symmetries of a Sequence . . . . . . . . . . . . . . . 129
4.2.4 Circular Time Shift of a Sequence . . . . . . . . . . . . . . . . 132
4.2.5 Circular Frequency Shift . . . . . . . . . . . . . . . . . . . . . 132
4.2.6 Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.2.7 Conjugation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.8 Multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.9 Parseval’s Relation . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.10 Circular Correlation . . . . . . . . . . . . . . . . . . . . . . . 134
4.3 Fast Fourier Transform (FFT) . . . . . . . . . . . . . . . . . . . . . . 135
4.3.1 Radix-2 FFT Algorithms . . . . . . . . . . . . . . . . . . . . . 135

5 Energy and Power 143

5.1 Parseval’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.1.1 Parseval’s Theorem for Power Signals . . . . . . . . . . . . . . 143
5.1.2 Parseval’s Theorem for Finite Energy Signals . . . . . . . . . 143
5.2 Power Spectral Density (PSD) . . . . . . . . . . . . . . . . . . . . . . 143
5.3 Energy Spectral Density (ESD) . . . . . . . . . . . . . . . . . . . . . 145
5.4 Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.4.1 Auto-Correlation . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.5 Cross-Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

Signals & Systems iii Asst. Prof. Bijaya Shrestha, nec

CONTENTS CONTENTS

6 Transmission of Signals 150

6.1 Transfer Function or Frequency Response . . . . . . . . . . . . . . . . 150
6.2 Distortionless Transmission . . . . . . . . . . . . . . . . . . . . . . . 151
6.3 Ideal Low Pass Filter (LPF) . . . . . . . . . . . . . . . . . . . . . . . 153
6.3.1 Frequency Response of Ideal LPF . . . . . . . . . . . . . . . . 153
6.3.2 Impulse Response of Ideal LPF . . . . . . . . . . . . . . . . . 154
6.4 Frequency Response and Impulse Response of RC Circuit . . . . . . . 155
6.5 Frequency Response and Impulse Response of RL Circuit . . . . . . . 156
6.6 Introduction to Communication System . . . . . . . . . . . . . . . . . 157

7 Transmission of Signals in Discrete-Time Systems 160

7.1 Transfer Function or Frequency Response of Discrete-Time Systems . 160
7.2 Types of Discrete-Time Systems or Digital Filters . . . . . . . . . . . 161
7.2.1 Finite-Duration Impulse Response (FIR) Systems . . . . . . . 161
7.2.2 Infinite-Duration Impulse Response (IIR) Systems . . . . . . . 163
7.3 Implementation of FIR Systems . . . . . . . . . . . . . . . . . . . . . 165
7.4 Implementation of IIR Systems . . . . . . . . . . . . . . . . . . . . . 165
7.4.1 Direct Form Structures . . . . . . . . . . . . . . . . . . . . . . 165
7.4.2 Cascade-Form Structures . . . . . . . . . . . . . . . . . . . . . 167
7.4.3 Parallel-Form Structures . . . . . . . . . . . . . . . . . . . . . 169
7.5 Solution of Linear Constant-Coefficient Difference (LCCD) Equations 172

Signals & Systems iv Asst. Prof. Bijaya Shrestha, nec

Chapter 1

Introduction

1.1 Signals
A signal is defined as a function of one or more independent variables that con-
tains information about the nature of some phenomenon. Voltages and currents as
a function of time in an electrical circuit are examples of signals. When the function
depends on a single variable, the signal is said to be one dimensional. A speech
signal is an example of a one-dimensional signal whose amplitude varies with time.
When the function depends on two or more variables, the signal is said to be multi-
dimensional. An image is an example of a two-dimensional signal whose brightness
depends on x- and y-coordinates. Other examples of signals are music, video, texts,
etc. We will study the signals with time t as an independent variable throughout
this course.

1.2 Classification of Signals

1.2.1 Continuous-Time and Discrete-Time Signals
A signal x(t) is said to be a continuous-time signal if it is defined for all time t. A
speech signal as a function of time is an example of a continuous-time signal. Fig-
ure 1.1 (a) represents an example of a continuous-time signal whose ampltude varies
continuously with time.

In contrast, a discrete-time signal is defined only at discrete times. The weekly

stock market index is an example of a discrete-time signal. Figure 1.1 (b) is an
example of a discrete-time signal.

1
 CHAPTER 1. INTRODUCTION

Figure 1.1: Example of (a) continuous-time signal and (b) discrete-time signal.

To distinguish between continuous-time and discrete-time signals, we will use the

symbol t as the independent variable for continuous-time signals and n for discrete-
time signals. Here, n is an integer value. Moreover, the independent variable is
enclosed in parentheses (.) for continuous-time signals and in brackets [.] for discrete-
time signals. The discrete-time signal in Figure 1.1 (b) can be expressed as a sequence
of numbers given by

x[n] = {0, 3, 2, 1, 2, 0, 0.}

↑

If the sequence does not have an arrow, the first element corresponds to the value
at n = 0.

1.2.2 Even and Odd Signals

A signal x(t) or x[n] is defined as an even signal if it is identical to its time-reversed
counterpart. That is,

x(−t) = x(t)
x[−n] = x[n].

A signal x(t) or x[n] is said to be an odd signal if

x(−t) = −x(t)
x[−n] = −x[n].

Any signal can be decomposed into even and odd parts. That is,

x(t) = xe (t) + xo (t). (1.1)

Signals & Systems 2 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Figure 1.2: Continuous-time and discrete-time even signal.

Figure 1.3: Continuous-time and discrete-time odd signal.

Substituting t by −t, we get

x(−t) = xe (−t) + xo (−t)

or,
x(−t) = xe (t) − xo (t) (1.2)
Solving Equations (1.1) and (1.2), we get
1
xe (t) = [x(t) + x(−t)] . (1.3)
2
and,
1
xo (t) = [x(t) − x(−t)] . (1.4)
2
Similarly for discrete-time signal x[n], we have
1
xe [n] = [x[n] + x[−n]] . (1.5)
2
and,
1
[x[n] − x[−n]] .
xo [n] = (1.6)
2
Note that the product of two even signals or of two odd signals is an even signal
and the product of an even signal and an odd signal is an odd signal.

Signals & Systems 3 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

1.2.3 Periodic and Nonperiodic Signals

A continuous-time signal x(t) is said to be periodic with period T if there is a positive
nonzero value of T for which

x(t + T ) = x(t)

for all values of t. Also,

x(t + mT ) = x(t),
where m is an integer. Figure 1.4 is an example of a continuous-time periodic signal.

Figure 1.4: Continuous-time periodic signal.

The fundamental period T0 of x(t) is the smallest positive value of T for which
x(t + T ) = x(t) holds. A signal which is not periodic or does not hold the above
condition is called the nonperiodic or aperiodic signal.

For discrete time, a signal or sequence x[n] is periodic with period N if there is
a positive integer N for which

x[n + N ] = x[n]

for all n. Also,

x[n + mN ] = x[n],
where m is an integer. Figure 1.5 shows an example of discrete-time periodic signal.
The fundamental period N0 of x[n] is the smallest positive integer N for which
x[n + N ] = x[n] holds.
Note that a sequence obtained by uniform sampling of a continuous-time periodic
signal may not be periodic. Also, the sum of two continuous-time periodic signals
may not be periodic but the sum of two periodic sequences is always periodic.

Signals & Systems 4 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Figure 1.5: Discrete-time periodic signal.

1.2.4 Deterministic and Random Signals

The signals whose values are completely specified for any given time are deterministic
signals. A deterministic signal is therefore a known function of time. For example,
sin(ωt) is the deterministic signal. Random signals are those signals that take random
values at any given time and must be characterized statistically. Thermal noise
generated in an eletronic device is the random signal.

1.2.5 Real and Complex Signals

A signal x(t) is a real signal if its value is a real number, and a signal x(t) is a
complex signal if its value is a complex number. The complex signal has the form
x(t) = x1 (t) + jx2 (t).

1.2.6 Energy and Power Signals

In electrical engineering, a signal may represent a voltage or a current. If voltage
v(t) is developed across a resistor with resistance R producing current i(t), then the
instantaneus power dissipated in this resistor is defined by

v 2 (t)
p(t) = v(t)i(t) = i2 (t)R = . (1.7)
R
If R = 1 Ω, then the instantaneous power is

p(t) = i2 (t) = v 2 (t). (1.8)

In signal analysis, power is defined normally on per-ohm basis, so that, regardless
of whether a given signal x(t) represents a voltage or a current, the instantaneous
power of the signal x(t) can be expressed as

p(t) = x2 (t). (1.9)

Now the total energy of the signal x(t) can be defined as

Signals & Systems 5 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

Z T /2
E = lim x2 (t)dt (1.10)
T →∞ −T /2
Z ∞
= x2 (t)dt
−∞

In general, the following equation is used to calculate the total energy of both the
real and complex signals.

Z ∞
E= |x(t)|2 dt. (1.11)
−∞

And, the average power is defined as

Z T /2
1
P = lim |x(t)|2 dt. (1.12)
T →∞ T −T /2

We can determine the average power of a periodic signal x(t) of fundamental period
T by using the equation given by
Z T /2
1
P = |x(t)|2 dt. (1.13)
T −T /2

For a discrete-time signal x[n], the total energy is defined by

∞
X
E= |x[n]|2 (1.14)
n=−∞

and its average power is defined by

N
1 X
P = lim |x[n]|2 . (1.15)
N →∞ 2N + 1
n=−N

As in continuous-time case, we can also determine the average power of a discrete-

time periodic signal x[n] with fundamental period N by using the expression given
as
N −1
1 X
P = |x[n]|2 . (1.16)
N n=0
A signal is referred to as an energy signal if and only if the total energy of the
signal satisfies the following condition.

Signals & Systems 6 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

0 < E < ∞.
The signal is referred to as a power signal if and only if the average power of the
signal satisfies the following condition.

0 < P < ∞.
The signal not satisfying any of the above conditions is referred to as the neither
energy nor power signal.

Note that an energy signal has zero average power and a power signal has infinite
energy. Periodic signals are usually viewed as power signals, whereas nonperiodic
signals are usually viewed as energy signals.

1.2.7 Analog and Digital Signals

The signals whose amplitude is continuously varying with respect to time and cannot
be defined by finite number of amplitude levels are called analog signals, whereas the
signals having only finite number of amplitude levels are called digital signals.

Figure 1.6: (a) Analog signal. (b) Digital signal.

Note that the discrete-time signals resulted after sampling analog signals may not be
digital. The digital signals are obtained after quantizing the resulted discrete-time
signals to a finite number of amplitude levels and then sending them to analog to
digital converter.

1.2.8 Causal, Anti-Causal, and Non-Causal Signals

The signals which exist only in the positive time instants are called as causal signals.
The signals that exist only in the negative time instants are called as anti-causal

Signals & Systems 7 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

signals. And, the signals that appear in both the positive and negative time instants
are called as non-causal signals. Figures 1.7(a), 1.7(b), and 1.8 show the examples
of a causal signal, anti-causal signal, and non-causal signal.

Figure 1.7: (a) Causal signal. (b)Anti-causal signal.

Figure 1.8: Non-causal signal.

1.3 Transformation of the Independent Variable

The independent variable in our case is time t for continuous-time case and n for
discrete-time case. The transformations of the independent variable involve: time
shifting, time scaling, and time reversal.

1.3.1 Time Shifting

A signal x(t) or x[n] may be shifted in time. The signal x(t − t0 ) is regarded as the
time-shifted version of the signal x(t) by t0 . A signal may be either advanced or

Signals & Systems 8 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

delayed in the time-axis depending on the sign of t0 . If t0 is positive, the shift is to

the right and the signal is delayed. On the other hand, if t0 is negative, the shift is
to the left and hence the signal is advanced. Let us consider a signal as shown in
Figure 1.9 (a) which is shifted to the right by 2 time units as shown in Figure 1.9
(b) and to the right by 2 time units as shown in Figure 1.9 (c).

Figure 1.9: (a) Original signal x(t). (b) Time-shifted signal x(t − t0 ) for t0 = 2. (c)
Time-shifted signal x(t − t0 ) for t0 = −2.

1.3.2 Time Scaling

The expansion or compression of a signal in time is called as time-scaling. A signal
x(αt) is regarded as the time-scaled version of the signal x(t). If |α| > 1, then the
signal x(t) is compressed and if |α| < 1, then the signal x(t) is expanded or stretched
in time. Let us consider an example of time-scaling as shown in Figure 1.10. Figure
1.10 (a) is the original signal which is compressed by 2 as shown in Figure 1.10 (b)
and expanded by 2 as shown in Figure 1.10 (c).

Signals & Systems 9 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

Let us consider x(t) to be an audio tape recording, then x(2t) is the recording
played at twice the speed and x(t/2) is the recording played at half-speed.

Figure 1.10: (a) Original signal x(t). (b) Time-scaled signal x(αt) for α = 2. (c)
Time-scaled signal x(αt) for α = 1/2.
.

1.3.3 Time Reversal

The signal x(−t) is referred to as the time-reversed or reflected version of the signal
x(t). We can obtain the signal x(−t) by reflecting or flipping the signal x(t) in time
about t = 0 as shown in Figure 1.11.

Considering the signal x(t) to be an audio tape recording, x(−t) represents the
same tape recording played backward.

1.3.4 Precedence Rule for Time Shifting and Time Scaling

Let the signal y(t) be derived from another signal x(t) such that

Signals & Systems 10 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Figure 1.11: (a) Original signal x(t). (b) Time-reversed version x(−t).
.

y(t) = x(αt − t0 )
then, to obtain y(t) from x(t), the time-shifting and time-scaling operations must be
performed in the correct order. We know that the scaling operation replaces t by
αt and time-shifting operation replaces t by t − t0 . So, the time-shifting operation
is performed first on x(t), resulting in an intermediate signal x(t − t0 ) and then
the time-scaling operation is performed on the intermediate signal resulting in the
desired output

y(t) = x(αt − t0 ).

1.4 Elementary Signals

Elementary signals are the basic building blocks for constructing more complex sig-
nals. These signals are extensively used in the study of signals and systems. Elemen-
tary signals include sinusoial and exponential signals, the step function, the impulse
function, and the ramp function.

1.4.1 Sinusoidal Signals

A sinusoidal signal can be defined by

x(t) = A cos(ωt + φ),

Signals & Systems 11 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

where A is the amplitude, w is the angular frequency in radians per second, and φ
is the phase angle in radians. A sinusoidal signal is an example of a periodic signal
with period
2π
T = .
ω

Figure 1.12: A sinusoidal signal.

A discrete-time sinusoidal signal is defined by

x[n] = A cos(ωn + φ).

If n is considered to be dimensionless, then both w and φ are measured in ra-
dians.This discrete-time signal may or may not be periodic. We know that for this
signal to be periodic with period N , we must have

x[n + N ] = x[n]
That is,

A cos(ωn + ωN + φ) = A cos(ωn + φ). (1.17)

Also,

A cos(ωn + φ) = A cos(ωn + 2mπ + φ), (1.18)

where m is an integer. Comparing Equations (1.17) and (1.18), we get

ωN = 2πm.
or,
2πm
ω= .
N

Signals & Systems 12 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

That means for a discrete-time sinusoidal signal to be periodic, the angular frequency
must be a rational multiple of 2π.

1.4.2 Exponential Signals

Real Exponential Signals

A continuous-time real exponential signal is expressed by

x(t) = Ceat ,

where c and a are real numbers. Depending on the sign of a, we have two types of
real exponential signals: exponentially growing signal for a > 0 and exponentially
decaying signal for a < 0.

Figure 1.13: (a) Exponentially growing signal. (b) Exponentially decaying signal.

A discrete-time real exponential signal is expressed by

x[n] = Cαn .

where C and α are real numbers. If |α| > 1 the magnitude of the signal grows
exponentially with n, whereas the signal decays exponentially for |α| < 1. Moreover,
if α is positive, all the values of x[n] are of the same sign whereas if α is negative
then the sign of x[n] alternates.

Signals & Systems 13 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

Periodic Complex Exponential Signals

A continuous-time periodic complex exponential signal is expressed by

x(t) = ejω0 t ,
where ω0 is the fundamental frequency and the fundamental period is
2π
T0 = .
ω0
By using Euler’s relation, we have

ejω0 t = cos(ω0 t) + j sin(ω0 t).

So, both the real and imaginary parts of the complex exponential signal are sinusoidal
signals.
A discrete-time complex exponential signal is expressed by

x[n] = ejω0 n .
And, the Euler’s relation gives

ejω0 n = cos(ω0 n) + j sin(ω0 n).

General Complex Exponential Signals

A general form of complex exponential signal is expressed by

x(t) = Ceat , (1.19)

where

C = |C|ejθ
and

a = r + jω0 .
So, the general complex exponential signal has the form

x(t) = |C|ejθ e(r+jω0 )t = |C|ert ej(w0 t+θ) . (1.20)

Using Euler’s relation, we have

x(t) = |C|ert cos(ω0 t + θ) + j|C|ert sin(ω0 t + θ). (1.21)

Signals & Systems 14 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Depending upon the values of r in Equation (1.21) we will have the following three
conditions:

For r = 0, the real and imaginary parts of a complex exponential are sinusoidal.
For r > 0, both the real and imaginary parts are growing sinusoidals.
For r < 0, both the real and imaginary parts are decaying sinusoidals.

Growing sinusoidal and decaying sinusoidal are shown in Figures 1.14 (a) and 1.14
(b) respectively.

Figure 1.14: (a) Growing sinusoidal signal. (b) Decaying sinusoidal signal.

Similarly, a general complex exponential signal in discrete time can be expressed

by

x[n] = Cαn , (1.22)

where

C = |C|ejθ
and

α = |α|ejω0 .
So,

x[n] = |C|ejθ |α|n ejω0 n = |C||α|n ej(ω0 n+θ) = |C||α|n cos(ω0 n+θ)+j|C||α|n sin(ω0 n+θ).
(1.23)
Therefore, from Equation (1.23), we see the following conditions.

Signals & Systems 15 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

For |α| = 1, the real and imaginary parts of a complex exponential signal are sinu-
soidal.
For |α| > 1, the real and imaginary parts are growing sinusoidal sequences.
For |α| < 1, the real and imaginary parts are decaying sinusoidal sequences.

1.4.3 Unit Impulse Function

Discrete-Time Unit Impulse Function
A discrete-time unit impulse (or unit sample) function is defined as
(
1, n = 0
δ[n] = (1.24)
0, n 6= 0
which is shown in Figure 1.15.

Figure 1.15: Discrete-time unit impulse function.

The discrete-time unit impulse function can be used to sample the value of a
signal at n = 0. That is,

x[n]δ[n] = x[0]δ[n].
Moreover, if we consider a unit impulse δ[n − n0 ] at n = n0 , then

x[n]δ[n − n0 ] = x[n0 ]δ[n − n0 ].

Continuous-Time Unit Impulse Function

A continuous-time unit impulse function is defined by

δ(t) = 0 for t 6= 0. (1.25)

and

Signals & Systems 16 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Z ∞
δ(t)dt = 1. (1.26)
−∞
From these equations, we can say that the impulse δ(t) is zero everywhere except
at the origin and the total area under the unit impulse is unity. The impulse δ(t) is
also referred to as the Dirac delta function. The graphical representation is shown
in the Figure 1.16(a).

Figure 1.16: (a) Continuous-time unit impulse function. (b) Rectangular pulse of
unit area.
The continuous-time unit impulse δ(t) is viewed as the limiting form of a rectan-
gular pulse of unit area as shown in Figure 1.16(b). As the duration of the pulse is
decreased, its amplitude is increased such that the area of the pulse remains unity.
So, for infinitesimally small duration, the rectangular pulse approximates the impulse
more closely. That is,

δ(t) = lim x∆ (t),

∆→0
where x∆ (t) is a rectangular pulse with duration ∆ and amplitude 1/∆.

We have some properties regarding to the continuous-time unit impulse function:

1. The unit impulse δ(t) is an even function of time. That is,

δ(−t) = δ(t).

2. Let us consider a continuous-time signal x(t) and the time-shifted impulse

function δ(t − t0 ), then we have
Z ∞
x(t)δ(t − t0 )dt = x(t0 ).
−∞
This is referred to as the sifting property.

Signals & Systems 17 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

3. The impulse function δ(t) has the time-scaling property as defined by:

1
δ(at) = δ(t), a > 0.
a

1.4.4 Unit Step Function

Discrete-Time Unit Step Function
A discrete-time unit step funciton is defined by
(
1, n ≥ 0
u[n] =
0, n < 0
which is shown in Figure 1.17.

Figure 1.17: Discrete-time unit step function.

Continuous-Time Unit Step Function

A continuous-time unit step function is defined by
(
1, t > 0
u(t) =
0, t < 0
which is shown in Figure 1.18.

Figure 1.18: Continuous time unit step function

Signals & Systems 18 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

1.4.5 Ramp Function

A continuous-time ramp function is defined by

(
t, t ≥ 0
r(t) = .
0, t < 0

Similarly, a discrete-time ramp function is defined by

(
n, n ≥ 0
r[n] = .
0, n < 0

The continuous-time and discrete-time ramp functions of unit slope are shown in
Figures 1.19(a) and 1.19(b) respectively.

Figure 1.19: (a) Continuous-time ramp funciton. (b) Discrete-time ramp function.

Signals & Systems 19 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

1.4.6 Relationship Between Unit Impulse Function, Unit Step

Function, and Ramp Function
Relationship Between Unit Impulse Function and Unit Step Function
In discrete-time, the unit step function is the running sum of the unit impulse func-
tion. That is,
n
X
u[n] = δ[m]. (1.27)
m=−∞

This is illustrated graphically as shown in Figure 1.20.

Figure 1.20: Running sum of Equation (1.27): (a) n < 0; (b) n > 0 .

On the other hand, the discrete-time unit impulse is the first difference of the discrete-
time unit step. That is

δ[n] = u[n] − u[n − 1].

In continuous-time, the unit step function is the running integral of the unit
impulse function. That is,
Z t
u(t) = δ(t)dτ. (1.28)
−∞

This is illustrated graphically as shown in Figure 1.21.

Figure 1.21: Running integral of Equation (1.28): (a) t < 0; (b) t > 0 .

Signals & Systems 20 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

On the other hand, the continuous-time unit impulse is the first derivative of the
continuous-time unit step. That is,

du(t)
δ(t) = .
dt

Relationship Between Unit Step Function and Ramp Function

The integral of the unit step function u(t) is a ramp funciton of unit slope. That is,
Z t
r(t) = u(τ )dτ.
−∞

Conversely, the continuous-time unit step funtion is the first derivative of the continuous-
time unit ramp function. That is,

dr(t)
u(t) = .
dt

1.5 Some Other Useful Signals

1.5.1 Signum Function
A discrete-time signum function is defined by

−1, n < 0

sgn[n] = 0, n=0,

1, n>0


and is shown in Figure 1.22 (a).

And, a continuous-time signum function is defined by


−1, t < 0

sgn(t) = 0, t=0,

1, t>0


and is shown in Figure 1.22 (b).

Signals & Systems 21 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

Figure 1.22: (a) A discrete-time signum function. (b) A continuous-time signum

function.

1.5.2 Rectangular Pulse or Gate Function

In continuous time, a rectangular pulse is expressed by

  ( T
t 1, |t| < 2
rect = T
,
T 0, |t| > 2

and is shown in Figure 1.23(a).

And, a discrete-time rectanguarl pulse is defined by

(
1, −N1 ≤ n ≤ N1
rect[n] = ,
0, otherwise

and is shown in Figure1.23(b).

Figure 1.23: (a) Continuous-time rectangular pulse. (b) Discrete-time rectangular

pulse.

Signals & Systems 22 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

1.5.3 Sinc Function

A sinc function in continuous time is expressed as
(
0, t = ±n
sinc(t) = sin(πt) . (1.29)
πt
, elsewhere

L’ Hospital rule is applied to compute the value of sinc(t) at t = 0 which is equal to

1. The sinc funcion is shown in Figure 1.24.

Figure 1.24: Sinc function.

1.6 Systems
A system can be defined as an interconnection of components, devices, or subsys-
tems that processes particular input signals to produce the desired outputs. E.g.,
amplifier, filter, etc.

A system is said to be a continuous-time system if continuous-time signals are

applied and result in continuous-time output signals. If the input and output signals
are in discrete time, then the system is referred to as the discrete-time system. The
continuous-time and discrete-time systems are represented graphically in Figures 1.25
(a) and (b) respectively.
Also, the continuous-time and discrete-time systems are represented symbolically
by Equations (1.30) and (1.31) respectively.

x(t) → y(t). (1.30)

Signals & Systems 23 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

Figure 1.25: (a) Continuous-time system. (b) Discrete-time system.

x[n] → y[n]. (1.31)

1.7 Interconnections of Systems

Many practical systems are built by interconnecting several subsystems. The inter-
connections can be of the following types.

1. A series or cascade interconnection of two systems is illustrated in Figure 1.26.

Figure 1.26: Series interconnection of two systems.

An example is a radio receiver followed by an amplifier.

2. A parallel interconnection of two systems is illustrated in Figure 1.27.

Figure 1.27: Parallel interconnection of two systems.

An example is an audio system with two microphones feeding into a single

amplifier.

Signals & Systems 24 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

3. We can have a series-parallel interconnection that is the combination of series

and parallel interconnections. An example is shown in the Figure 1.28.

Figure 1.28: Series-parallel interconnection.

4. A feedback interconnection system is shown in Figure 1.29.

Figure 1.29: Feedback interconnection.

1.8 Properties of Systems

1.8.1 Memory
A system is said to be memoryless if its output signal depends only on the present
value of the input signal. A resistor is a memoryless system because the voltage
(taken as output y(t)) dropped across it at any time depends on the current (taken
as input x(t)) at that same time. That is,

y(t) = Rx(t),
where R is the resistance. Another example is the identity system; i.e., y[n] = x[n].

Signals & Systems 25 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

On the other hand, a system is said to be with memory if its output signal
depends on the past or future values of the input signal. A capacitor is an example
of a continuous-time system with memory, since the voltage (taken as output y(t))
across it at any time depends on the past values of the current (taken as input x(t)).
That is,

1 t
Z
y(t) = x(τ )dτ,
C −∞
where C is the capacitance. An example of a discrete-time system with memory is
an accumulator or summer
n
X
y[n] = x[k].
k=−∞

1.8.2 Invertibility
A system is said to be invertible if the input of the system can be recovered from
the output. For an invertible system, there exists an inverse system when cascaded
with the original system produces the signal that is identical to the input signal to
the original system as illustrated in the Figure 1.30.

Figure 1.30: Invertible system.

An example of an invertible continuous-time system is

y(t) = 3x(t) (1.32)

for which the inverse system is
1
w(t) = y(t). (1.33)
3
This example is illustrated in Figure 1.31.
The following systems are invertible:

1. y[n] = 0: This sytem always produces zero output for any input signal.
2. y(t) = x2 (t): The output signal is same whether input signal has positive or
negative sign.

Signals & Systems 26 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Figure 1.31: An example of invertible system.

1.8.3 Causality
A system is said to be causal if the present value of the output signal depends only
on the present and/or past values of the input signal. If the present value of the
output signal depends also on the future value of the input signal, then the sytem is
referred to as the non-causal system.

All the real-time systems and memoryless systems are causal. E.g., resistor,
capacitor, inductor, delay, accumulator, etc. The moving-average system described
by

1
y[n] = (x[n + 1] + x[n] + x[n − 1])
3
is non-causal, since the output signal y[n] depends also on the future value of the
input signal, i.e., x[n + 1].

1.8.4 Stability
A system is said to be bounded input-bounded output (BIBO) stable if the output
is bounded for bounded input. That means the output of such a system does not
diverge or does not grow unreasonably larger if the input does not diverge.

In other words, a system is BIBO stable if the output signal is

|y(t)| ≤ My < ∞ for all t

for the input signal

|x(t)| ≤ Mx < ∞ for all t,

where Mx and My are finite positive numbers. The system

y(t) = ex(t)
is BIBO stable, whereas

Signals & Systems 27 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

y(t) = tx(t)
is BIBO unstable.

1.8.5 Time Invariance

A system is said to be time invariant if its behavior does not change over time. In
other words, a system is time invariant if a time shift of the input signal leads to an
identical time shift in the output signal. That is, if

x(t) → y(t),
then

x(t − t0 ) → y(t − t0 ).

Example:
Determine whether the system y(t) = sin[x(t)] is time variant or time invariant.

Solution:
Let x1 (t) be an input signal to the given system such that output signal is

y1 (t) = sin[x1 (t)]. (1.34)

Let’s delay the signal x1 (t) in time by t0 , i.e.,

x2 (t) = x1 (t − t0 ).
Then, the system output becomes

y2 (t) = sin[x2 (t)] = sin[x1 (t − t0 )]. (1.35)

Let’s delay the whole system by t0 , then from Equation (1.34), we have

y1 (t − t0 ) = sin[x1 (t − t0 )]. (1.36)

Comparing Equations (1.35) and (1.36), we get

y2 (t) = y1 (t − t0 ).
Hence the given system is time invariant.

Signals & Systems 28 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Example:
Determine whether the system y(t) = tx(t) is time variant or time invariant.

Solution:
Let x1 (t) be an input signal to the given system such that output signal is

y1 (t) = tx1 (t). (1.37)

Let’s delay the signal x1 (t) in time by t0 , i.e.,

x2 (t) = x1 (t − t0 ).
Then, the system output becomes

y2 (t) = tx2 (t) = tx1 (t − t0 ). (1.38)

Let’s delay the whole system by t0 , then from Equation (1.37), we have

y1 (t − t0 ) = (t − t0 )x1 (t − t0 ). (1.39)
Comparing Equations (1.38) and (1.39), we get

y2 (t) 6= y1 (t − t0 ).
Hence the given system is time variant.

1.8.6 Linearity
A system is said to be linear if it satisfies the superposition theorem. That means
if an input consists of the weighted sum of several signals, then the output is the
weighted sum of the responses of the system to each of those signals. Let x1 (t) and
x2 (t) be the input signals and y1 (t) and y2 (t) be the corresponding output signals,
i.e.,

x1 (t) → y1 (t)
x2 (t) → y2 (t)

then the system is linear if

x1 (t) + x2 (t) → y1 (t) + y2 (t) (additive property)

Signals & Systems 29 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 1. INTRODUCTION

and

ax1 (t) → ay1 (t) (homogeneity property)

The above two properties can be combined into a single statement:

ax1 (t) + bx2 (t) → ay1 (t) + by2 (t)

where a and b are constants.

Example:
Is the system y(t) = tx(t) linear?

Solution:
Let x1 (t) and x2 (t) be the input signals and y1 (t) and y2 (t) be the corresponding
output signals. Then,

x1 (t) → y1 (t) = tx1 (t)

x2 (t) → y2 (t) = tx2 (t)

Let x3 (t) be a linear combination of of x1 (t) and x2 (t). That is,

x3 (t) = ax1 (t) + bx2 (t)

where a and b are constants. If x3 (t) is the input to the system, then the output is

y3 (t) = tx3 (t)

= t(ax1 (t) + bx2 (t))
= atx1 (t) + btx2 (t)
= ay1 (t) + by2 (t)

That is,

ax1 (t) + by2 (t) → ay1 (t) + by2 (t).

Hence the given system is linear.

Example:
Is the system y(t) = x2 (t) linear?

Signals & Systems 30 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 1. INTRODUCTION

Solution:
Let x1 (t) and x2 (t) be the input signals and y1 (t) and y2 (t) be the corresponding
output signals. Then,

x1 (t) → y1 (t) = x21 (t)

x2 (t) → y2 (t) = x22 (t)

Let x3 (t) be a linear combination of of x1 (t) and x2 (t). That is,

x3 (t) = ax1 (t) + bx2 (t)

where a and b are constants. If x3 (t) is the input to the system, then the output is

y3 (t) = x23 (t)

= (ax1 (t) + bx2 (t))2
= a2 x21 (t) + b2 x22 (t) + 2abx1 (t)x2 (t)
= a2 y1 (t) + b2 y2 (t) + 2abx1 (t)x2 (t)

Since,

y3 (t) 6= ay1 (t) + by2 (t)

the given system is not linear.
.

Signals & Systems 31 Asst. Prof. Bijaya Shrestha, nec

Chapter 2

Linear Time-Invariant Systems

Many practical systems possess the properties of linearity and time invariance. The
systems with both of these properties are known as linear time-invariant (LTI) sys-
tems. An LTI system can be completely characterized in terms of its impulse re-
sponse. Impulse response is the output of a system when its input is the impulse
function.

Any arbitrary signal can be represented as a linear combination of time-shifted

impulses. If the impulse response of an LTI system is known, then by using the prop-
erties of linearity and time invariance, the output for any input signal can be deter-
mined as a linear combination of time-shifted impulse responses. The term ”convo-
lution sum” for discrete-time LTI systems and ”convolution integral” for continuous-
time LTI systems is used to describe the procedure to determine the output from the
input and the impulse response.

2.1 The Convolution Sum

Let us consider a discrete-time signal x[n] which is represented as a linear combination
of time-shifted impulses as illustrated in the Figure 2.1. The weights are the values
of the signal x[n] at the corresponding time shifts. Besides for −2 ≤ n ≤ 2, if there
are additional terms, then we can write

x[n] = · · ·+x[−2]δ[n+2]+x[−1]δ[n+1]+x[0]δ[n]+x[1]δ[n−1]+x[2]δ[n−2]+· · · (2.1)

In more compact form, we can express Equation (2.1) as

32
CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.1: Representation of a signal x[n] as a linear combination of time-shifted

impulses.

Signals & Systems 33 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

+∞
X
x[n] = x[k]δ[n − k]. (2.2)
k=−∞

This corresponds to the representation of an arbitrary signal x[n] as a linear com-

bination of time-shifted unit impulses δ[n − k], where the weights in this linear
combination are x[k].

If the system is linear and

δ[n − k] → hk [n]

then, from the superposition property, we can express the output as

+∞
X
y[n] = x[k]hk [n]. (2.3)
k=−∞

If the system is also time invariant, then

hk [n] = h0 [n − k]

where h0 [n − k] is the time-shifted version of unit impulse response h0 [n]. For sim-
plicity, let

h0 [n] = h[n].

Hence the output of a discrete-time LTI sytem for input x[n] becomes

+∞
X
y[n] = x[k]h[n − k]. (2.4)
k=−∞

Therefore, the output of an LTI system is given as a linear combination of time-shifted

impulse responses. The sum in Equation (2.4) is termed the convolution sum. The
operation on the right-hand side of Equation (2.4) is known as the convolution of
the sequences x[n] and h[n]. Symbolically, we can write

y[n] = x[n] ∗ h[n].

Signals & Systems 34 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Example:
Determine the output y[n] of a discrete-time LTI system for the following input x[n]
and impulse response h[n].

x[n] = {2, 1, 1} and h[n] = {1, 1}.

Solution:
We have,
+∞
X
y[n] = x[k]h[n − k]. (2.5)
k=−∞

1. Express x[n] and h[n] as functions of k. That is,

x[k] = {2, 1, 1} and h[k] = {1, 1}.

The functions x[k] and h[k] are shown in Figure 2.2 (a).

2. Flip h[k] over time as shown in Figure 2.2 (b).

For n = 0, using Equation (2.5), the output is
+∞
X
y[0] = x[k]h[−k].
k=−∞

The nonzero overlapping between x[k] and h[−k] is only at k = 0. And,

y[0] = 0 + 2 × 1 + 0 = 2.

3. For n = 1, h[1 − k] is the time-shifted version of h[−k] to the right by 1 time

unit as shown in Figure 2.2 (c). So,
+∞
X
y[1] = x[k]h[1 − k] = 0 + 2 × 1 + 1 × 1 + 0 = 3.
k=−∞

4. For n = 2, h[2 − k] is the time-shifted version of h[−k] to the right by 2 time

units as shown in Figure 2.2 (d). So,
+∞
X
y[2] = x[k]h[2 − k] = 0 + 1 × 1 + 1 × 1 + 0 = 2.
k=−∞

Signals & Systems 35 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

5. For n = 3, h[3 − k] is the time-shifted version of h[−k] to the right by 3 time

units as shown in Figure 2.2 (e). So,

+∞
X
y[3] = x[k]h[3 − k] = 0 + 1 × 1 + 0 = 1.
k=−∞

6. For n = 4, h[4 − k] is the time-shifted version of h[−k] to the right by 4 time

units as shown in Figure 2.2 (f). There are no nonzero overlapping between
x[k] and h[4 − k]. So,

+∞
X
y[4] = x[k]h[4 − k] = 0.
k=−∞

7. Similarly, for n > 4, there will be no nonzero overlapping between x[k] and
h[n − k]. So,
y[n] = 0.

8. For n = −1, h[−1 − k] is the time-shifted version of h[−k] to the left by 1 time
unit as shown in Figure 2.2 (g). There are no nonzero overlapping between
x[k] and h[−1 − k]. So,

+∞
X
y[−1] = x[k]h[−1 − k] = 0.
k=−∞

9. Similarly, for n < −1, there will be no nonzero overlapping between x[k] and
h[n − k]. So,
y[n] = 0.

10. Hence, the overall output of the given LTI system for the given input sequence
x[n] is
y[n] = {0, 2, 3, 2, 1, 0}.

Signals & Systems 36 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.2: Figures for determining the output of the given discrete-time LTI system
for the given input.

Signals & Systems 37 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

2.2 The Convolution Integral

A continuous-time LTI system is also completely characterized by its impulse re-
sponse; the output can be determined for any arbitrary continuous-time input signal
from the knowledge of its impulse response. The steps to determine the relation-
ship among input signal, output signal, and impulse response are analogous to the
discrete-time case. Recall that, a discrete-time input signal can be represented as a
weighted sum of time-shifted impulses as given by
+∞
X
x[n] = x[k]δ[n − k].
k=−∞

In continuous-time case, an input signal x(t) can be represented as a weighted integral

of time-shifted impulses as given by
Z +∞
x(t) = x(τ )δ(t − τ )dτ.
−∞

Let h(t) be defined as the impulse response of the sytem for an impulse input
δ(t). That is,

δ(t) → h(t).
Then using the time invariance property, we have

δ(t − τ ) → h(t − τ ).
Also, using the linearity property, the output of an LTI system is given as a linear
combination of time-shifted impulse responses given by
Z +∞
y(t) = x(τ )h(t − τ )dτ. (2.6)
−∞

The Equation (2.6) is termed as the convolution integral. Symbolically, we can write

y(t) = x(t) ∗ h(t).

Example:
Determine the output of an LTI system with unit impulse response

h(t) = u(t),
for the input signal

Signals & Systems 38 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

x(t) = e−at u(t), a > 0.

Solution:
We have,
Z +∞
y(t) = x(τ )h(t − τ )dτ.
−∞

1. Expressing x(t) and h(t) as functions of τ , we have

h(τ ) = u(τ ),

and
x(τ ) = e−aτ u(τ ), a > 0.
The functions x(τ ) and h(τ ) are shown in Figure 2.3(a).

2. Flip h(τ ) over time as shown in Figure 2.3(b).

3. For t < 0, h(t − τ ) is the time-shifted version of h(−τ ) to the left by t time
unit as shown in Figure 2.3(c). We see that for t < 0, the product of x(τ ) and
h(t − τ ) is zero. Therefore, output y(t) is zero.

4. For t > 0, h(t − τ ) is the time-shifted version of h(−τ ) to the right by t time
unit as shown in Figure 2.3(d). The output for t > 0 is then determined by
Z +∞ Z t t
1 1
e−aτ dτ = − e−aτ 1 − e−at .


y(t) = x(τ )h(t − τ )dτ = =
−∞ 0 a 0 a

5. For all t, the output is

1
1 − e−at u(t),


y(t) =
a

which is shown in Figure 2.3(e).

Signals & Systems 39 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.3: Figures for determining the output of the given LTI system.

Signals & Systems 40 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

2.3 Properties of LTI Systems

2.3.1 The Commutative Property
The convolution is the commutative operation. That is, for discrete time,

x[n] ∗ h[n] = h[n] ∗ x[n],

or,
+∞
X +∞
X
x[k]h[n − k] = h[k]x[n − k],
k=−∞ k=−∞

and for continuous time,

x(t) ∗ h(t) = h(t) ∗ x(t),

or, Z +∞ Z +∞
x(τ )h(t − τ )dτ = h(τ )x(t − τ )dτ.
−∞ −∞

This property tells us that the output of an LTI system with input x[n] and unit
impulse response h[n] is identical to the output of an LTI system with input h[n]
and unit impulse response x[n].

2.3.2 The Distributive Property

The convolution distributes over addition. The distributive property of convolution
is described by

x[n] ∗ (h1 [n] + h2 [n]) = x[n] ∗ h1 [n] + x[n] ∗ h2 [n] (2.7)

in discrete time and

x(t) ∗ (h1 (t) + h2 (t)) = x(t) ∗ h1 (t) + x(t) ∗ h2 (t),

in continuous time. This property is interpreted as below.
Let us consider two LTI systems with impulse responses h1 [n] and h2 [n] as shown
in Figure 2.4 (a). Here,

y1 [n] = x[n] ∗ h1 [n]

y2 [n] = x[n] ∗ h2 [n]

Signals & Systems 41 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.4: Interpretation of distributive property.

and

y[n] = x[n] ∗ h1 [n] + x[n] ∗ h2 [n] (2.8)

This corresponds to the right-hand side of Equation (2.7). From Figure 2.4 (b),

y[n] = x[n] ∗ (h1 [n] + h2 [n]) (2.9)

This corresponds to the left-hand side of Equation (2.7). Therefore, the distributive
property tells us that a parallel combination of LTI systems can be replaced by a
single LTI system whose unit impulse response is the sum of the individual unit
impulse responses in the parallel combination.

2.3.3 The Associative Property

The associative property of convolution is described by

x[n] ∗ (h1 [n] ∗ h2 [n]) = (x[n] ∗ h1 [n]) ∗ h2 [n] (2.10)

in discrete time and

x(t) ∗ [h1 (t) ∗ h2 (t)] = [x(t) ∗ h1 (t)] ∗ h2 (t)

in continuous time. We can interpret this porperty from Figures 2.5 (a) and (b). In
Figure 2.5 (a),

Signals & Systems 42 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.5: Interpretation of associative property.

y[n] = w[n] ∗ h[n] = (x[n] ∗ h1 [n]) ∗ h2 [n],

This corresponds to the right-hand side of Equation (2.10). In Figure 2.5 (b),

y[n] = x[n] ∗ h[n] = x[n] ∗ (h1 [n] ∗ h2 [n]).

This corresponds to the left-hand side of Equation (2.10).
Therefore, the associative property tells us that the series interconnection of two
LTI systems in Figure 2.5 (a) is equivalent to the single system in Figure 2.5 (b).

2.3.4 LTI Systems with and without Memory

A discrete-time LTI system is said to be memoryless if

h[n] = 0, for n 6= 0.
Here, the impulse response has the form

h[n] = Kδ[n],
where K = h[0] is a constant and the output of the system is

y[n] = Kx[n].
Similarly, a continuous-time LTI system is memoryless if

h(t) = 0, for t 6= 0.
The impulse response has the form

Signals & Systems 43 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

h(t) = Kδ(t),
and the system output is

y(t) = Kx(t).
If K = 1, then these systems become identity systems, with output equal to the
input. Also, the unit impulse response is equal to the unit impulse. Then, we have

x[n] = x[n] ∗ δ[n]

and

x(t) = x(t) ∗ δ(t).

2.3.5 Invertibility of LTI Systems

A continuous-time LTI system with impulse response h(t) is invertible only if there
exists an inverse system with impulse response h1 (t), when connected in series with
the original system, produces an output equal to the input to the first system as
illustrated in Figure 2.6 (a). Therefore, for invertibility, the series interconnection
in Figure 2.6 (a) must be identical to the identity system in Figure 2.6 (b). That
means,

h(t) ∗ h1 (t) = δ(t).

Similarly, a discrete-time LTI system with impulse response h[n] is invertible only
if there exists an inverse system with impulse response h1 [n] such that

h[n] ∗ h1 [n] = δ[n].

2.3.6 Causality for LTI Systems

For a discrete-time LTI system, we have
+∞
X
y[n] = x[k]h[n − k].
k=−∞

For a discrete-time LTI system to be causal, y[n] must not depend on x[k] for k > n.
This requires

Signals & Systems 44 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.6: A continuous-time invertible LTI system.

h[n − k] = 0 for k > n.

This implies that

h[n] = 0 for n < 0. (2.11)

According to Equation (2.11), the impulse response of a causal LTI system must be
zero before the impulse occurs. Therefore, the output of a causal discrete-time LTI
system becomes

n
X
y[n] = x[k]h[n − k].
k=−∞

Alternatively,

∞
X
y[n] = h[k]x[n − k].
k=0

Similarly, for a continuous-time LTI system to be causal,

h(t) = 0 for t < 0,

and the system output is given by

Z t Z ∞
y(t) = x(τ )h(t − τ )dτ = h(τ )x(t − τ )dτ.
−∞ 0

Signals & Systems 45 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

2.3.7 Stability for LTI Systems

We know that a system is stable if every bounded input produces a bounded output.
Let us consider an input signal x[n] that is bounded in magnitude. That is,

|x[n]| < Mx for all n.

Then the magnitude of the output can be determined by
+∞
X
|y[n]| = h[k]x[n − k] .
k=−∞

or,
+∞
X
|y[n]| ≤ |h[k]||x[n − k]|.
k=−∞

Since |x[n − k]| < Mx , then

+∞
X
|y[n]| ≤ B |h[k]| for all n.
k=−∞

This equation implies that if the impulse response is absolutely summable, that is, if
+∞
X
|h[k]| < ∞,
k=−∞

then the output y[n] is bounded in magnitude, and hence the system is stable.
Similarly, a continuous-time LTI system is stable only if its impulse response is
absolutely integrable, that is,
Z +∞
|h(τ )|dτ < ∞.
−∞

2.3.8 The Unit Step Response of an LTI System

We studied that an LTI system can be characterized by its unit impulse response. In
addition, we can use unit step response to describe the behavior of LTI systems. Unit
step response is the output of a system when input is unit step. The step response of
a discrete-time LTI system is obtained by performing the convolution sum between
the unit step and the impulse response; that is,

Signals & Systems 46 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

+∞
X
s[n] = u[n] ∗ h[n] = h[n] ∗ u[n] = h[k]u[n − k]. (2.12)
k=−∞

Since u[n − k] = 0 for n − k < 0 and u[n − k] = 1 for n − k ≥ 0, from Equation

(2.12),
n
X
s[n] = h[k]
k=−∞

Therefore, the unit step response of a discrete-time LTI system is the running sum
of its impulse response. Again,
n−1
X
s[n] = h[k] + h[n] = s[n − 1] + h[n].
k=−∞
or,

h[n] = s[n] − s[n − 1].

Therefore, the impulse response of a discrete-time LTI system is the first difference
of its step response.

Similarly, for a continuous-time LTI system, the unit step response is the run-
ning integral of its unit impulse response and the unit impulse response is the first
derivative of its unit step response. That is,
Z t
s(t) = h(τ )dτ.
−∞
and,

ds(t)
h(t) = .
dt
Hence, the unit step response can also be used to characterize an LTI system,
since the unit impulse response can be determined from it.

2.4 Causal LTI Systems Described by Differential

and Difference Equations
The input and output can be related also through a linear constant-coefficient dif-
ferential equation for a continuous-time system and a linear constant-coefficient dif-

Signals & Systems 47 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

ference equation for a discrete-time system.

2.4.1 Linear Constant-Coefficient Differential Equations

Let us consider an RC circuit as shown in Figure 2.7. Using Kirchhoff’s voltage law,
we get

Figure 2.7: An RC circuit.

dy(t)
x(t) = i(t)R + y(t) = RC + y(t)
dt
or,

dy(t)
y(t) + RC = x(t). (2.13)
dt
So, the input and output of an RC circuit can be related through the first order
differential equation.

To determine the output, this differential equation must be solved which requires
the initial condition of the system. The complete solution, y(t), consists of a homo-
geneous solution, yh (t), and a particular solution, yp (t), i.e.,

y(t) = yh (t) + yp (t). (2.14)

The homogeneous solution is obtained from the homogeneous differential equation

dy(t)
y(t) + RC = 0. (2.15)
dt
Let the homogeneous solution is of the form

yh (t) = c1 er1 t , (2.16)

where r1 is the root of the characteristic equation

1 + RCr1 = 0. (2.17)

Signals & Systems 48 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

1
So, r1 = − RC , and the homogeneous solution becomes
t
yh (t) = c1 e− RC . (2.18)
Consider the input signal

x(t) = Ke2t u(t), (2.19)

where K is a real number. Then, the particular solution is assumed to be

yp (t) = Y e2t u(t), (2.20)

where Y is to be determined. Substituting Equations (2.19) and (2.20) in (2.13) for
t > 0, we get

Y e2t + 2RCY e2t = Ke2t . (2.21)

Solving the Equation (2.21), we get

Y + 2RCY = K.
or,
K
Y = .
1 + 2RC
Then the particular solution becomes,
K
yp (t) = e2t , t > 0. (2.22)
1 + 2RC
Now, for t > 0, the complete solution becomes,
t K
y(t) = c1 e− RC + e2t . (2.23)
1 + 2RC
To determine the constant c1 , we need an initial condition. For a system to be
causal and LTI, it must have the condition of initial rest. This means, for a causal
LTI system, if x(t) = 0 for t < t0 , then y(t) must also equal to 0 for t < t0 . Therefore
y(t) = 0 for t < 0 and using y(0) = 0 in Equation (2.23), we get
K
0 = c1 +
1 + 2RC
or,
K
c1 = −
1 + 2RC

Signals & Systems 49 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Thus, for t > 0,

K t K
y(t) = − e− RC + e2t .
1 + 2RC 1 + 2RC
And, for all t,

K h
2t t
− RC
i
y(t) = e −e u(t).
1 + 2RC
This is the required solution of the given problem.

A general N th -order linear constant-coefficient differential equation is given by

N M
X dk y(t) X dk x(t)
ak = bk . (2.24)
k=0
dtk k=0
dtk

The response of a causal LTI system described by N th -order differential equation

(2.24) for t > t0 can be calculated with the initial conditions of rest, i.e.,

dy(t0 ) dN −1 y(t0 )
y(t0 ) = = ··· = = 0. (2.25)
dt dtN −1

2.4.2 Linear Constant-Coefficient Difference Equations

An N th -order linear constant-coefficient difference equation is given by
N
X M
X
ak y[n − k] = bk x[n − k]. (2.26)
k=0 k=0

Equations of this type can be solved exactly in a way we do for differential equations.
The complete solution is the sum of a homogeneous solution and a particular solution.
The homogeneous solution is the solution of the homogenous equation
N
X
ak y[n − k] = 0.
k=0

For a causal LTI system, we must have the condition of initial rest. That is, if
x[n] = 0 for n < n0 , then y[n] = 0 for n < n0 .

We have an alternative approach to solve the N th -order difference equation. The

Equation (2.26) can be rearranged in the form

Signals & Systems 50 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

(M N
)
1 X X
y[n] = bk x[n − k] − ak y[n − k] . (2.27)
a0 k=0 k=1

In order to calculate y[n], we need to know the auxiliary conditions y[n − 1], y[n −
2] · · · , y[n − N ].An equation of the form Equation (2.26) is called the recursive equa-
tion since we can calculate the output in terms of the input and the previous outputs.

Example:
Solve the following difference equation
1
y[n] − y[n − 1] = x[n] (2.28)
2
for

x[n] = Kδ[n].

Solution:
Equation (2.28) can be expressed as
1
y[n] = x[n] + y[n − 1]
2
Using the condition of initial rest, since x[n] = 0 for n < 0 , then y[n] = 0 for n < 0
as well. Now, for n ≥ 0

y[0] = x[0] + 21 y[−1] = K,

y[1] = x[1] + 12 y[0] = 12 K,

2
y[2] = x[2] + 12 y[1] = 12 K,
..
.
n
y[n] = x[n] + 21 y[n − 1] = 12 K.
Therefore, for all n,
 n
1
y[n] = Ku[n].
2
For K = 1, we get the impulse response of the system to be
 n
1
h[n] = u[n].
2

Signals & Systems 51 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

2.5 Block Diagram Representations of First-Order

Systems Described by Differential and Differ-
ence Equations
The systems described by linear constant-coefficient difference and differential equa-
tions can be represented by block diagrams. Block diagram representations make
us easy to understand the behavior and properties of the systems and also to simu-
late and implement them. Let us consider a discrete-time system described by the
following first-order difference equation

y[n] + ay[n − 1] = bx[n]. (2.29)

We can rewrite this equation in recursive form as

y[n] = bx[n] − ay[n − 1].

To represent this system in a block diagram, we need three operations: addition,
multiplication by a coefficient, and delay as shown in Figure 2.8.

Figure 2.8: Block diagram representation for the discrete-time system described by
Equation (2.29).

Let us consider a continuous-time system described by a first-order differential

equation given by

dy(t)
+ ay(t) = bx(t). (2.30)
dt
Let’s rearrange the Equation (2.30) in the form

1 dy(t) b
y(t) = − + x(t). (2.31)
a dt a
According to Equation (2.31), the output y(t) of the system can be determined by
performing three operations: addition, multiplication by a coefficient, and differen-
tiation as shown by the block diagram in Figure 2.9

Signals & Systems 52 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.9: Block diagram representation of a continuous-time system described by

Equation (2.30).

Since differentiators are normally difficult to implement, an alternative approach

is to use integrators. We can rewrite Equation (2.30) in the form

dy(t)
= bx(t) − ay(t) (2.32)
dt
and then integrating from −∞ to t, we get
Z t
y(t) = [bx(τ ) − ay(τ )]dτ. (2.33)
−∞

So, we can implement this system by as shown in Figure 2.10.

Figure 2.10: Block diagram representation of a continuous-time system described by

Equation (2.33).

2.6 Convolution of a Rectangular Pulse Passed

Through an RC Filter
Let us consider a rectangular pulse and an RC filter as shown in Figures 2.11 (a)
and (b) respectively. The impulse response of an RC filter is given by

Signals & Systems 53 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

1 −t/RC
h(t) = e u(t). (2.34)
RC

Figure 2.11: (a) Rectangular pulse. (b) RC filter.

We have,
Z +∞
y(t) = x(τ )h(t − τ )dτ. (2.35)
−∞

1. As a function of τ , we have

1 −τ /RC
h(τ ) = e u(τ ).
RC

The functions x(τ ) and h(τ ) are shown in Figures 2.12 (a).

2. Flip h(τ ) over time as shown in Figure 2.12 (b).

3. For t < 0, h(t − τ ) is the time-shifted version of h(−τ ) to the left by t time
unit as shown in Figure 2.12 (c). We see that for t < 0, the product of x(τ )
and h(t − τ ) is zero. Therefore, output y(t) is zero.

4. For 0 < t < 2, from Figure 2.12 (d), the output is

R t 1 −(t−τ )/RC
y(t) = 0 RC e dτ.
1 −t/RC t τ /RC
R
= RC e e
0 
dτ
1 −t/RC τ /RC t

= RC e RC e 0
= e−t/RC (et/RC − e0 )
y(t) = 1 − e−t/RC .

Signals & Systems 54 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

5. For t > 2, the overlapping always occurs from τ = 0 to 2. From Figure 2.12
(e), the output is obtained as
R 2 1 −(t−τ )/RC
y(t) = 0 RC e dτ.
1 −t/RC t τ /RC
R
= RC e e
0 
dτ
1 −t/RC τ /RC 2

= RC e RC e 0
y(t) = e−t/RC (e2/RC − 1).

6. Hence, the output can be summarized as


0,
 t<0
y(t) = 1 − e−t/RC
, 0<t<2

 −t/RC 2/RC
e (e − 1), t > 2.

Signals & Systems 55 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 2. LINEAR TIME-INVARIANT SYSTEMS

Figure 2.12: Figures for determining the output of an RC filter for a rectangular
pulse.

Signals & Systems 56 Asst. Prof. Bijaya Shrestha, nec

Chapter 3

Fourier Analysis for

Continuous-Time and
Discrete-Time Signals

3.1 Continuous-Time Fourier Series (CTFS)

3.1.1 Complex Exponential Fourier Series
A continuous-time periodic complex exponential signal with fundamental period T
and fundamental frequency w0 is defined by

x(t) = ejw0 t
where

2π
w0 = .
T
A set of harmonically related complex exponentials can be represented as

φk (t) = ejkw0 t , k = 0, ±1, ±2, · · ·

Each of these signals has a fundamental frequency that is an integer multiple of w0
and each is periodic with common period T .

A continuous-time periodic signal x(t) with fundamental period T can be repre-

sented as a linear combination of harmonically related complex exponentials. That
is,

57
 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

+∞ +∞
2π
ak ejk( T )t .
X X
jkw0 t
x(t) = ak e = (3.1)
k=−∞ k=−∞

This representation is known as the Fourier series (or exponential Fourier series)
representation and the set of coefficients {ak } are called the Fourier series coeffi-
cients.The term for k = 0 is a constant. The terms for k = −1 and k = +1 both
have fundamental frequency equal to w0 and are collectively called as the funda-
mental components or the first harmonic components. The terms for k = −2 and
k = +2 are referred to as the second harmonic components. In general, the terms
for k = −N and k = +N are collectively referred to as the Nth harmonic components.

To determine the coefficients {ak }, we multiply both sides of Equation (3.5) by

−jnw0 t
e and integrate from 0 to T to result
Z T Z T +∞
X
−jnw0 t
x(t)e dt = ak ejkw0 t e−jnw0 t dt. (3.2)
0 0 k=−∞

Interchanging the order of integration and summation, we obtain

Z T +∞
X Z T 
−jnw0 t j(k−n)w0 t
x(t)e dt = ak e dt . (3.3)
0 k=−∞ 0

Using Euler’s identity,

Z T Z T Z T
j(k−n)w0 t
e dt = cos((k − n)w0 t)dt + j sin((k − n)w0 t)dt. (3.4)
0 0 0

For, k 6= n, cos((k −n)w0 t) and sin((k −n)w0 t) are periodic signals with fundamental
T T
period |k−n| . Since, T is an integer multiple of |k−n| , then, for k 6= n, both of the
integrals on the right-hand side of Equation (3.4) are zero. For k = n, the integral
on the left-hand side of Equation (3.4) results T . Therefore,
Z T (
T. k = n
ej(k−n)w0 t dt =
0 0, k 6= n
And, the right-hand side of Equation (3.3) reduces to T an . That is,
Z T
x(t)e−jnw0 t dt = T an .
0

Signals & Systems 58 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

So,
Z T
1
an = x(t)e−jnw0 t dt
T 0
or,
Z T
1
ak = x(t)e−jkw0 t dt
T 0
or,
Z
1
ak = x(t)e−jkw0 t dt.
T T

The equation
+∞ +∞
2π
ak ejk( T )t
X X
x(t) = ak ejkw0 t = (3.5)
k=−∞ k=−∞

is referred to as the synthesis equation and the equation

Z Z
1 1 2π
ak = x(t)e −jkw0 t
dt = x(t)e−jk( T )t dt (3.6)
T T T T
is referred to as the analysis equation. For k = 0, the coefficient
Z
1
a0 = x(t)dt. (3.7)
T T
is the dc or constant component of x(t).

Example
Determine the Fourier series coefficients of the following periodic sawtooth wave.
Also, plot the magnitude and phase spectrum.

Solution
2π
Here, fundamental period, T = 1 sec., and hence, fundamental frequency, w0 = T
=
2π rad./sec..

For a single period from t = 0 to 1, using two point formula,

−1 − 1
x(t) − 1 = (t − 0)
1−0

Signals & Systems 59 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.1: A periodic sawtooth wave.

we get,

x(t) = 1 − 2t.
The Fourier series coefficients can be determined using

Z
1
ak = x(t)e−jkw0 t dt
T T
Z 1
= (1 − 2t)e−jk2πt dt
Z0 1 Z 1
−jk2πt
= e dt − 2 te−jk2πt dt. (3.8)
0 0

Using integral by parts,

Z 1 Z Z Z  
−jk2πt −jk2πt −jk2πtdt
te dt = t e dt − e dt dt
0 dt
1  −jk2πt 1 1 1  −jk2πt 1
= te 0
− e 0
−jk2π −jk2π −jk2π
−jk2π −jk2π
e e 1
=− + −
jk2π (k2π)2 (k2π)2
1 1 1
=− + 2
−
jk2π (k2π) (k2π)2
1
=− (3.9)
jk2π

Also,

Signals & Systems 60 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Z 1
1  −jk2πt 1
e−jk2πt dt = e 0
0 −k2π
1
=− (e−jk2π − 1)
k2π
1
=− (1 − 1)
k2π
= 0. (3.10)

Using Equations (3.9) and (3.10) in (3.8), we get

1 j
ak = =− .
jkπ kπ

For k = 0, we have
Z Z 1
a0 = x(t)dt = (1 − 2t)dt = 0.
T 0

Hence, the Fourier series coefficients are

(
0, k=0
ak = j
− kπ , k =
6 0.

Now, the magnitude spectrum is given by

(
0, k=0
|ak | = 1
|k|π
, k 6= 0.

and the phase spectrum is given by


0,
 k=0
π
∠ak = − 2 , k > 0

π
2
, k < 0.

The magnitude and phase spectrum of the given sawtooth wave are shown in Figure
3.2 (a) and (b) respectively.

Signals & Systems 61 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.2: (a) Magnitude spectrum. (b) Phase spectrum.

Convergence of the Fourier Series

Let us approximate a periodic signal x(t) by a linear combination of a finite number
of harmonically related complex exponentials, that is,
+N
X
xN (t) = ak ejkω0 t .
k=−N

Let eN (t) denote the approxmation error; that is,

+N
X
eN (t) = x(t) − xN (t) = x(t) − ak ejkω0 t .
k=−N

The energy in the error over one period can be calculated as

Z
EN = |eN (t)|2 dt.
T

As N increases, new terms are added and EN decreases. For N → ∞, the EN = 0.

In some cases, the integral in the equation

Z
1
ak = e−jkω0 t dt
T T
may diverge; that is, the value for some ak may be infinite. Moreover, even if all
of coefficients ak are finite, when these coefficients are substituted into the synthesis
equation, the infinite series may not converge to the original signal x(t).

Signals & Systems 62 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

The following Dirichlet conditions are required for the convergence of Fourier
series.

1. Condition 1: Over any period, x(t) must be absolutely integrable; that is,
Z
|x(t)|dt < ∞.
T

A periodic signal that violates the first condition is

1
x(t) = , 0 < t ≤ 1.
t
Here, x(t) is periodic with period 1.

2. Condition 2: There must not be more than a finite number of maxima and
minima during any single period of the signal. A signal that violates this
condition is  
2π
x(t) = sin , 0 < t ≤ 1.
t
3. Condition 3: There must be only a finite number of discontinuities over any
single period.

3.1.2 Trigonometric Fourier Series

A periodic signal x(t) can be represented by trigonometric Fourier series as
+∞
X
x(t) = b0 + [bk cos(kω0 t) + ck sin(kω0 t)] . (3.11)
k=1
2π
Here, fundamental period, T = ω0
and bk and ck are Fourier series coefficients.

Evaluation of b0 :
Integrating both sides of Equation (3.11) over one period, we get

Z Z +∞
Z X
x(t)dt = b0 dt + [bk cos(kω0 t) + ck sin(kω0 t)] dt
T T T k=1
+∞
X Z +∞
X Z
= b0 T + bk cos(kω0 t)dt + ck sin(kω0 t)dt
k=1 T k=1 T

Signals & Systems 63 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Since
Z
cos(kω0 t)dt = 0
T

and
Z
sin(kω0 t)dt = 0
T

Therefore,
Z
x(t)dt = b0 T
T

and
Z
1
b0 = x(t)dt.
T T

This is the dc component of the signal.

Evaluation of bk :
Multiplying both sides of Equation (3.11) by cos(nω0 t) and integrating over a single
period, we get

Z Z +∞
Z X
x(t) cos(nω0 t)dt = b0 cos(nω0 t)dt + [bk cos(kω0 t) + ck sin(kω0 t)] cos(nω0 t)dt
T T T k=1
Z +∞
X Z
= b0 cos(nω0 t)dt + bk cos(kω0 t) cos(nω0 t)dt
T k=1 T
+∞
X Z
+ ck sin(kω0 t) cos(nω0 t)dt
k=1 T
Z +∞ Z
1X
= b0 cos(nω0 t)dt + bk cos((k + n)ω0 t)dt
T 2 k=1 T
+∞ Z +∞ Z
1X 1X
+ bk cos((k − n)ω0 t)dt + ck sin((k + n)ω0 t)dt
2 k=1 T 2 k=1 T
+∞ Z
1X
+ ck sin((k − n)ω0 t)dt
2 k=1 T

Signals & Systems 64 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Since,
Z
cos(nω0 t)dt = 0
T

Z
cos((k + n)ω0 t)dt = 0, for all k, n
T

(
0, k 6= n
Z
cos((k − n)ω0 t)dt =
T T, k = n

Z
sin((k + n)ω0 t)dt = 0, for all k, n
T

Z
sin((k − n)ω0 t)dt = 0, for all k, n
T

Therefore,
Z
1
x(t) cos(nω0 t)dt = bn T
T 2

or,
Z
2
bn = x(t) cos(nω0 t)dt
T T

or,
Z
2
bk = x(t) cos(kω0 t)dt
T T

Evaluation of ck :

Multiplying both sides of Equation (3.11) by sin(nω0 t) and integrating over a single
period, we get

Signals & Systems 65 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Z Z +∞
Z X
x(t) sin(nω0 t)dt = b0 sin(nω0 t)dt + [bk cos(kω0 t) + ck sin(kω0 t)] sin(nω0 t)dt
T T T k=1
Z +∞
X Z
= b0 sin(nω0 t)dt + bk cos(kω0 t) sin(nω0 t)dt
T k=1 T
+∞
X Z
+ ck sin(kω0 t) sin(nω0 t)dt
k=1 T
Z +∞ Z
1X
= b0 sin(nω0 t)dt + bk sin((n − k)ω0 t)dt
T 2 k=1 T
+∞ Z +∞ Z
1X 1X
+ bk sin((n + k)ω0 t)dt + ck cos((k − n)ω0 t)dt
2 k=1 T 2 k=1 T
+∞ Z
1X
− ck cos((k + n)ω0 t)dt
2 k=1 T

Since,
Z
sin(nω0 t)dt = 0
T
Z
sin((n − k)ω0 t)dt = 0, for all k, n
T
Z
sin((n + k)ω0 t)dt = 0, for all k, n
T
(
0, k 6= n
Z
cos((k − n)ω0 t)dt =
T T, k = n
Z
cos((k + n)ω0 t)dt = 0, for all k, n
T

Therefore,
Z
1
x(t) sin(nω0 t)dt = cn T
T 2
or,

Signals & Systems 66 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Z
2
cn = x(t) sin(nω0 t)dt
T T
or,
Z
2
ck = x(t) sin(kω0 t)dt
T T

Symmetry Conditions
1. If x(t) is even, then

2 T /2
Z
b0 = x(t)dt
T 0
4 T /2
Z
bk = x(t) cos(kω0 t)dt
T 0
ck = 0

2. If x(t) is odd, then

b0 = 0
bk = 0
Z T /2
4
ck = x(t) sin(kω0 t)dt
T 0

3.1.3 Compact Trigonometric Fourier Series

We have, the trigonometric Fourier series representation as,
+∞
X
x(t) = b0 + [bk cos(kω0 t) + ck sin(kω0 t)]
k=1
p  
Let, d0 = b0 , dk = b2k + c2k , and θ = tan−1 cbkk . Then,
+∞
X
x(t) = d0 + [(dk cos θ) cos(kω0 t) + (dk sin θ) sin(kω0 t)]
k=1
or,

Signals & Systems 67 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

+∞
X
x(t) = d0 + dk cos(kω0 t − θ)
k=1

This is the compact trigonometric Fourier series representation. Here, the relation-
ship between the coefficients of trigonometric Fourier series and compact trigono-
metric Fourier series are:
b0 = d0 , bk = dk cos θ, and ck = dk sin θ.

3.1.4 Conversion of Trigonometric Fourier Series into Com-

plex Exponential Fourier Series

+∞
X
x(t) = b0 + [bk cos(kω0 t) + ck sin(kω0 t)]
k=1
+∞ 
ejkω0 t + e−jkω0 t ejkω0 t − e−jkω0 t
X    
= b0 + bk + ck
k=1
2 2j
+∞  
X bk − jck jkω0 t bk + jck −jkω0 t
= b0 + e + e
k=1
2 2

bk −jck bk +jck
Let, a0 = b0 , ak = 2
, and a−k = 2
. Then,

+∞
X
+ a−k e−jkω0 t
 jkω0 t 
x(t) = a0 + ak e
k=1
+∞
X +∞
X
= a0 + ak e jkω0 t
+ a−k e−jkω0 t
k=1 k=1
+∞
X −1
X
= a0 + ak ejkω0 t + ak ejkω0 t
k=1 k=−∞
+∞
X
= ak ejkω0 t
k=−∞

This is the complex exponential Fourier series representation.

Signals & Systems 68 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

3.1.5 Conversion of Complex Exponential Fourier Series into

Trigonometric Fourier Series
We can convert the complex exponential Fourier series into trigonometric Fourier
series by using

b0 = a0 , bk = ak + a−k , and ck = j(ak − a−k ).

Example
Determine the trigonometric Fourier series representation of the following periodic
sawtooth wave.

Figure 3.3: A periodic sawtooth wave.

Solution
A periodic signal x(t) can be represented by trigonometric Fourier series as,
+∞
X
x(t) = b0 + [bk cos(kω0 t) + ck sin(kω0 t)] .
k=1

The given signal has fundamental period, T = 1 sec. and fundamental frequency,
ω0 = 2π rad./sec..The given signal x(t) is odd. Therefore, b0 = 0 and bk = 0. And,

4 T /2
Z
ck = x(t) sin(kω0 t)dt
T 0
Z 1/2
=4 x(t) sin(kω0 t)dt
0

For 0 < t < 12 , using two-point formula,

Signals & Systems 69 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

0−1
x(t) − 1 = (t − 0)
1/2 − 0

∴ x(t) = 1 − 2t
Now,

Z 1/2
ck = 4 (1 − 2t) sin(kω0 t)dt
0
Z 1/2 Z 1/2
=4 sin(kω0 t)dt − 8 t sin(kω0 t)dt
0 0
Z 1/2 Z 1/2
=4 sin(k2πt)dt − 8 t sin(k2πt)dt
0 0
1/2 Z Z  Z  
4 dt
=− cos(k2πt) − 8 t sin(k2πt)dt − sin(k2πt)dt dt
k2π 0 dt
!
1/2 Z 1/2
4 t 1
=− (cos(kπ) − 1) − 8 − cos(k2πt) + cos(k2πt)dt
k2π k2π 0 k2π 0
!
1/2
4 1 1
(−1)k − 1 − 8 −


=− cos(kπ) + sin(k2πt)
k2π k4π (k2π)2 0
2 2 2
=− (−1)k + + (−1)k
kπ kπ kπ
2
=
kπ
Hence, the trigonometric Fourier series represenation of the given periodic signal is,
+∞  
X 2
x(t) = sin(kω0 t).
k=1
kπ

3.2 Properties of Continuous-Time Fourier Series

Let x(t) be a periodic signal with fundamental period T and fundamental frequency
ω0 = 2π/T . Let ak be its Fourier series coefficients, then we can use a shorthand
notation to indicate the relationship between x(t) and its Fourier series coefficients
as
FS
x(t) ←→ ak .

Signals & Systems 70 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

1. Linearity
Let x(t) and y(t) be two periodic signals with period T and
FS
x(t) ←→ ak
FS
y(t) ←→ bk
Then,
FS
z(t) = Ax(t) + By(t) ←→ ck = Aak + Bbk .
Proof
Z
1
ck = z(t)e−jkω0 t dt
T T
Z
1
= (Ax(t) + By(t))e−jkω0 t dt
T T
Z Z
1 −jkω0 t 1
=A x(t)e dt + B y(t)e−jkω0 t dt
T T T T
= Aak + Bbk .

2. Time Shifting
If
FS
x(t) ←→ ak
Then,
FS
y(t) = x(t − t0 ) ←→ bk = e−jkω0 t0 ak

Proof
Z Z
1 −jkω0 t 1
bk = y(t)e dt = x(t − t0 )e−jkω0 t dt
T T T T
Let t − t0 = τ , then
Z
1
bk = x(τ )e−jkω0 (τ +t0 ) dτ
T T Z
−jkω0 t0 1
=e x(τ )e−jkω0 τ dτ
T T
= e−jkω0 t0 ak

Signals & Systems 71 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

When a periodic signal is shifted in time, the magnitudes of its Fourier series
coefficients remain same.

3. Frequency Shifting
If
FS
x(t) ←→ ak
Then,
FS
y(t) = ejM ω0 t x(t) ←→ bk = ak−M

Proof

Z
1
bk = y(t)e−jkω0 t dt
T T
Z
1
= ejM ω0 t x(t)e−jkω0 t dt
T T
Z
1
= x(t)e−j(k−M )ω0 t dt
T T
= ak−M

4. Time Reversal
If
FS
x(t) ←→ ak
Then,
FS
y(t) = x(−t) ←→ bk = a−k

Proof
+∞
X
x(t) = ak ejkω0 t
k=−∞

Signals & Systems 72 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

So,
+∞
X
y(t) = x(−t) = ak e−jkω0 t
k=−∞

Put k = −m, then

+∞
X +∞
X
y(t) = a−m ejmω0 t = bm ejmω0 t
m=−∞ m=−∞

Hence,
bk = a−k
Note that if x(t) is even, then a−k = ak and if x(t) is odd, then a−k = −ak .

5. Time Scaling
If
FS
x(t) ←→ ak
Then
FS
x(αt) ←→ ak
We can prove this in a straightforward manner. Since,
+∞
X
x(t) = ak ejkω0 t
k=−∞

Then,
+∞
X
x(αt) = ak ejk(αω0 )t
k=−∞

Hence, the time-scaled version x(αt) has the same Fourier series coefficients ak .
However, it has the fundamental period T /α and fundamental frequency αω0 .

6. Conjugation
If
FS
x(t) ←→ ak
Then
FS
x∗ (t) ←→ a∗−k

Signals & Systems 73 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Proof

We have,
+∞
X
x(t) = ak ejkω0 t
k=−∞
Then, " #∗
+∞
X +∞
X
∗
x (t) = ak e jkω0 t
= a∗k e−jkω0 t
k=−∞ k=−∞
Put k = −m, then
+∞
X
∗
x (t) = a∗−m ejmω0 t
m=−∞
Hence,
FS
x∗ (t) ←→ a∗−k
Note that
(a) If x(t) is real, then
ak = a∗−k ⇒ a∗k = a−k

(b) If x(t) is real and even, then

a∗k = a−k = ak
That is, the Fourier series coefficients are also real and even.
(c) If x(t) is real and odd, then
a∗−k = −a−k = ak
So, the Fourier series coefficients are imaginary and odd.
7. Multiplication
Let x(t) and y(t) be two periodic signals with period T and
FS
x(t) ←→ ak
FS
y(t) ←→ bk
Then,
+∞
FS
X
z(t) = x(t)y(t) ←→ ck = al bk−l
l=−∞
The right-hand side is the convolution sum between ak and bk .

Signals & Systems 74 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Proof

We have,
+∞
X
x(t) = ak ejkω0 t
k=−∞
+∞
X
y(t) = bk ejkω0 t
k=−∞

Then,
+∞
X +∞
X
jlω0 t
z(t) = x(t)y(t) = al e bm ejmω0 t
l=−∞ m=−∞
+∞
X +∞
X
= al bm ej(l+m)ω0 t
l=−∞ m=−∞

Let l + m = k, then
+∞
X +∞
X
z(t) = al bk−l ejkω0 t
l=−∞ k=−∞
+∞ +∞
!
X X
= al bk−l ejkω0 t
k=−∞ l=−∞

Hence,
+∞
X
ck = al bk−l .
l=−∞

8. Periodic Convolution
Let x(t) and y(t) be two periodic signals with period T and
FS
x(t) ←→ ak
FS
y(t) ←→ bk

Then, Z
FS
z(t) = x(τ )y(t − τ )dτ ←→ ck = T ak bk
T
The left-hand side is referred to as the periodic convolution between x(t) and
y(t).

Signals & Systems 75 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Proof

Z
1
ck = z(t)e−jkω0 t dt
T T
Z Z
1
= x(τ )y(t − τ )dτ e−jkω0 t dt
T T T

Let t − τ = t0 , then

Z Z
1 0
ck = x(τ )y(t0 )dτ e−jkω0 (t +τ ) dt0
T
Z T T Z 
1 0 −jkω0 t0 0
= x(τ ) y(t )e dt e−jkω0 τ dτ
T T T
Z 
1 −jkω0 τ
= T bk x(τ )e dτ
T T
= T ak b k

9. Parseval’s Relation for Continuous-Time Periodic Signals

If

FS
x(t) ←→ ak

Then, the Parseval’s relation for continuous-time periodic signals is defined by

Z +∞
1 2
X
P = |x(t)| dt = |ak |2
T T k=−∞

Signals & Systems 76 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Proof

The average power of a periodic signal is

Z
1
P = |x(t)|2 dt
T T
Z
1
= x(t)x∗ (t)dt
T T
Z " +∞ #
1 X
= x(t) a∗k e−jkω0 t dt
T T k=−∞
+∞  Z 
∗ 1
X
−jkω0 t
= ak x(t)e dt
k=−∞
T T
+∞
X
= a∗k ak
k=−∞
+∞
X
= |ak |2
k=−∞

This relation states that the total average power in a periodic signal equals the
sum of the average powers in all of its harmonic components.
10. Differentiation
If
FS
x(t) ←→ ak
Then,
dx(t) F S
y(t) = ←→ bk = jkω0 ak
dt

Proof

We have,
+∞
X
x(t) = ak ejkω0 t
k=−∞

Differentiating both sides w.r.t. t, we get

+∞
dx(t) X
= jkω0 ak ejkω0 t
dt k=−∞

Signals & Systems 77 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Hence, we get
bk = jkω0 ak

11. Integration
If
FS
x(t) ←→ ak
Then,
Z t
FS 1
y(t) = x(t)dt ←→ bk = ak
−∞ jkω0

Proof

We have,
+∞
X
x(t) = ak ejkω0 t
k=−∞

Integrating both sides w.r.t. t, we get

Z t +∞
X 1
x(t)dt = ak ejkω0 t
−∞ k=−∞
jkω0

Hence, we get
1
bk = ak
jkω0

3.3 Periodicity of Continuous-Time and Discrete-

Time Complex Exponential Signals
3.3.1 Continuous-Time Complex Exponential Signal, x(t) =
ejω0 t
For the signal ejω0 t to be periodic with period T , the following condition must be
satisfied:
ejω0 (t+T ) = ejω0 t
That is,
ejω0 T = 1

Signals & Systems 78 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

or,
ω0 T = 2π
or,
2π
ω0 =
T
Here, ω0 is the fundamental frequency and T is the fundamental period. Note that
the signals ejω0 t are distinct for different values of ω0 . If ω0 is increased, the rate of
oscillation also increases.

3.3.2 Discrete-Time Complex Exponential Signal, x[n] = ejω0 n

For the signal ejω0 n to be periodic with period N , we must have the following condi-
tion:
ejω0 (n+N ) = ejω0 n
That is,
ω0 N = 2π
or,
2π
ω0 =
N
where N is the fundamental period and ω0 is the fundamental frequency. Hence, the
signal ejω0 n is periodic only if ω2π0 is the rational number (i.e., 1/N ) which is not the
case for continuous-time complex exponential signal ejω0 t .

Another difference can be made based on the following fact:

ej(ω0 +2π)n = ejω0 n

This means that the signal at frequency ω0 + 2π is the same as that at frequency ω0 .
So, the signals ejω0 n are not distinct for different values of ω0 .

3.4 Discrete-Time Fourier Series (DTFS)

A discrete-time periodic complex exponential signal with fundamental period N and
fundamental frequency w0 is defined by

x[n] = ejω0 n
where

Signals & Systems 79 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

2π
ω0 =
N
A set of harmonically related complex exponentials can be represented as

2π
φk [n] = ejkω0 n = ejk N n , k = 0, ±1, ±2, · · ·

Unlike in continous-time case, there are only N distinct signals in the set φk [n] since,

2π
φk+N [n] = ej(k+N ) N n
2π 2π
= ejk N n ejN N n
2π
= ejk N n
= φk [n]

In general,

φk+rN [n] = φk [n]

The only N distinct signals are:

2π 2π 2π
φ0 [n] = 1, φ1 [n] = ej N n , φ2 [n] = ej2 N n , · · · , φN −1 [n] = ej(N −1) N n

and any other φk [n] is identical to one of these signals. Therefore, the sequences
φk [n] are distinct only over a range of N successive values of k.

A discrete-time periodic signal x[n] can be represented as a linear combination

of harmonically related discrete-time complex exponential signals. That is,

2π
X X X
x[n] = ak φk [n] = ak ejkω0 n = ak ejk N n (3.12)
k=<N > k=<N > k=<N >

This is referred to as the discrete-time Fourier series representation and ak as the

Fourier series coefficients. We can range k for N successive values.
2π
To determine ak , we multiply both sides of Equation (3.12) by e−jr N n and sum
over N terms to obtain

Signals & Systems 80 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

2π 2π 2π
X X X
x[n]e−jr N n = ak ejk N n e−jr N n
n=<N > n=<N > k=<N >
2π
X X
= ak ej(k−r) N n
n=<N > k=<N >
2π
X X
= ak ej(k−r) N n
k=<N > n=<N >

Since,
(
X
j(k−r) 2π n 0, k 6= r
e N =
n=<N >
N, k = r
Therefore,
2π
X
x[n]e−jr N n = ar N
n=<N >
or,
1 X 2π
ar = x[n]e−jr N n
N n=<N >
or,
1 X 2π
ak = x[n]e−jk N n (3.13)
N n=<N >

The Equation (3.12) is referred to as the synthesis equation and (3.13) as the analy-
sis equation. The Fourier series coefficients are also known as the spectral coefficients.

Referring to Equation (3.12), let’s take k in the range from 0 to N − 1 to obtain

x[n] = a0 φ0 [n] + a1 φ1 [n] + · · · + aN −1 φN −1 [n]. (3.14)

Again, let’s take k in the range from 1 to N to obtain

x[n] = a1 φ1 [n] + a2 φ2 [n] + · · · + aN φN [n]. (3.15)

Since, φ0 [n] = φN [n], comparing Equations (3.14) and (3.15), we find that a0 = aN .
Similarly, for k ranging over any set of N successive integers, we have

ak = ak+N .
And, we conclude that the values ak repeat periodically with period N .

Signals & Systems 81 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Example
Determine and plot the magnitude and phase spectra of the periodic signal x[n] =
sin(ω0 n).

Solution
Here,
2π
ω0 =
N
is the fundamental frequency and N is the fundamental period of the given signal.
Let’s take the value of N to be 5.

Using Euler’s relation,

1 jω0 n 1
x[n] =e − e−jω0 n
2j 2j
Comparing this equation with the synthesis equation, we see that

1 1 π
a1 = ⇒ |a1 | = , ∠a1 = −
2j 2 2
1 1 π
a−1 = − ⇒ |a−1 | = , ∠a−1 =
2j 2 2

and the remaining coefficients over the interval of summation are zero. The magni-
tude and phase spectra are shown in Figures 3.4 (a) and (b) respectively.

Figure 3.4: (a) Magnitude spectrum. (b) Phase spectrum.

Signals & Systems 82 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Example
Determine and plot the magnitude and phase spectra of the following periodic signal
   
4π 4π π
x[n] = 2 + sin n + 2 cos n+ .
N N 2

Solution
Expanding the given signal x[n] using Euler’s identity, we get
1 j 4π n 1 1 4π π 1 4π π
e N − e−j N n + 2 ej ( N n+ 2 ) + 2 e−j ( N n+ 2 )
4π
x[n] = 2 +
2j 2j 2 2
1 j 4π n 1 −j 4π n 4π 4π
= 2 − j e N + j e N + jej N n − je−j N n
2 2
1 j 4π n 1 4π
= 2 + j e N − j e−j N n
2 2
Comparing this equation with the synthesis equation, we get
a0 = 2 ⇒ |a0 | = 2, ∠a0 = 0,
1 1 π
a2 = j ⇒ |a2 | = , ∠a2 = ,
2 2 2
1 1 π
a−2 = −j ⇒ |a−2 | = , ∠a−2 = − ,
2 2 2
and other coefficients ak are zero over the interval of summation in the synthesis
equation. For N = 5, the magnitude and phase spectra are shown in Figures 3.5 (a)
and (b) respectively.

Example
Determine and plot the spectrum of the following discrete-time periodic square wave.

Solution
We have,
1 X 2π
ak = x[n]e−jk N n
N n=<N >

For the given signal x[n], we can express the above equation as
N1
1 X 2π
ak = e−jk N n
N n=−N
1

Signals & Systems 83 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.5: (a) Magnitude spectrum. (b) Phase spectrum.

Figure 3.6: Discrete-time periodic square wave.

Let m = n + N1 , then
2N
1 X1 −jk 2π (m−N1 )
ak = e N
N m=0
2N
1 jk 2π N1 X1 −jk 2π m
= e N e N
N m=0
2π
!
1 2π 1 − e−jk N (2N1 +1)
= ejk N N1 2π
N 1 − e−jk N
π π π
1 2π ejk N ejk N (2N1 +1) − e−jk N (2N1 +1)
= ejk N N1 jk π (2N1 +1) π π
N e N  ejk N − e−jk N
π
1 ejk N (2N1 +1) sin k Nπ (2N1 + 1)
= π
N ejk N (2N1 +1) sin( kπ )
 π  N
1 sin k N (2N1 + 1)
= ,
N sin( kπ
N
)

and
2N1 + 1
a0 =
N

Signals & Systems 84 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

In our example, N = 10 and N1 = 2. So,

 π 
1 sin k 10 (2 × 2 + 1)
ak =
10 sin( kπ )
 π  10
1 sin k 2 )
= ,
10 sin( kπ
10
)

and

2×2+1 1
a0 = = .
10 2

The Fourier series coefficients are plotted as shown in Figure 3.7.

Figure 3.7: Fourier series coefficients.

Signals & Systems 85 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

3.5 Properties of Discrete-Time Fourier Series

Most of the properties of discrete-time Fourier series are similar to those of continuous-
time Fourier series. Let x[n] and y[n] be the periodic signals with period N and fun-
damental frequency ω0 = 2π N
. We use the following short hand notation to indicate
the relationship between a periodic signal x[n] and its Fourier series coefficients ak .
FS
x[n] ←→ ak

1. Linearity
If
FS
x[n] ←→ ak
FS
y[n] ←→ bk
Then,
FS
z[n] = Ax[n] + By[n] ←→ ck = Aak + Bbk .

2. Time Shifting
If
FS
x[n] ←→ ak
Then,
FS
y[n] = x[n − n0 ] ←→ bk = e−jkω0 n0 ak

3. Frequency Shifting
If
FS
x[n] ←→ ak
Then,
FS
y[n] = ejM ω0 n x[n] ←→ bk = ak−M

4. Time Reversal
If
FS
x[n] ←→ ak
Then,
FS
y[n] = x[−n] ←→ bk = a−k

Signals & Systems 86 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

5. Time Scaling
If
FS
x[n] ←→ ak
Then,
FS
x[αn] ←→ αak , α > 0.
The scaling operation changes the fundamental frequency to be αω0 and the
fundamental period to be N/α. Also, the Fourier series coefficients are scaled
by α.

6. Conjugation
If
FS
x[n] ←→ ak
Then
FS
x∗ [n] ←→ a∗−k

7. Multiplication
If
FS
x[n] ←→ ak
FS
y[n] ←→ bk

Then,
FS
X
z[n] = x[n]y[n] ←→ ck = al bk−l
l=<N >

The right-hand side is the periodic convolution between ak and bk .

8. Periodic Convolution
If
FS
x[n] ←→ ak
FS
y[n] ←→ bk

Then,
FS
X
z[n] = x[r]y[n − r] ←→ ck = N ak bk
r=<N >

The left-hand side is referred to as the periodic convolution between x[n] and
y[n].

Signals & Systems 87 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

9. Parseval’s Relation for Discrete-Time Periodic Signals

If
FS
x[n] ←→ ak

Then, the Parseval’s relation for discrete-time periodic signals is defined by

1 X X
P = |x[n]|2 = |ak |2
N n=<N > k=<N >

Proof

The average power of a discrete-time periodic signal is

1 X
P = |x[n]|2
N n=<N >
1 X
= x[n]x∗ [n]
N n=<N >
1 X X 2π
= x[n] a∗k e−jk N n
N n=<N > k=<N >
" #
X 1 X
−jk 2π
= a∗k x[n]e N
n

k=<N >
N n=<N >
X
= a∗k ak
k=<N >
X
= |ak |2
k=<N >

This relation states that the total average power in a periodic signal equals the
sum of the average powers in all of its harmonic components.

10. First Difference

If
FS
x[n] ←→ ak

Then,
 
FS 2π
y[n] = x[n] − x[n − 1] ←→ bk = 1 − e−jk N ak

Signals & Systems 88 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Proof

According to time-shifting property, we have

FS 2π
x[n − 1] ←→ e−jk N ak

and using linearity property,

 
FS −jk 2π
x[n] − x[n − 1] ←→ 1 − e N ak

3.6 Continuous-Time Fourier Transform (CTFT)

We know that periodic signals can be represented as linear combinations of har-
monically related complex exponentials; the representation being referred to as the
Fourier series representation. Aperiodic signals can also be represented as linear
combinations of complex exponentials. But, for aperiodic signals, the complex expo-
nentials are infinitesimally close in frequency and the representation takes the form
of an integral rather than a sum. The resulting spectrum of coefficients is called
the Fourier transform. And, the synthesis equation that uses these coefficients to
represent the signal is called the inverse Fourier transform.

The major concept behind the development of Fourier transform from the Fourier
series representation is that an aperiodic signal can be viewed as a periodic signal
with an infinite period. In the Fourier series representation of a periodic signal, we
have ω0 = 2π/T . As the period increases the fundamental frequency decreases and
the components become closer in frequency. If the period becomes infinite, then the
components form a continuum and the Fourier series sum becomes an integral.

Development of Continuos-Time Fourier Transform

Let us consider an aperiodic signal x(t) as shown in Figure 3.8 (a). We can make
a periodic signal x
e(t) with fundamental period T from this signal x(t) as shown in
Figure 3.8 (b). The signal x
e(t) is similar to the signal x(t) for one period extending
from −T /2 to T /2.

For the periodic sigal x

e(t), we have Fourier series representation as
+∞
X
x
e(t) = ak ejkω0 t (3.16)
k=−∞

Signals & Systems 89 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.8: (a) Aperiodic signal x(t). (b) Periodic signal x

e(t).

and Fourier series coefficients as

Z T /2
1
ak = e(t)e−jkω0 t dt
x (3.17)
T −T /2

e(t) = x(t) for −T /2 < t < T /2 and x(t) = 0 outside this interval, Equation
Since x
(3.17) can be rewritten as

Z T /2 Z +∞
1 −jkω0 t 1
ak = x(t)e dt = x(t)e−jkω0 t dt
T −T /2 T −∞

Let us define
Z +∞
X(jkω0 ) = T ak = x(t)e−jkω0 t dt (3.18)
−∞

then

1
ak = X(jkω0 ). (3.19)
T

Signals & Systems 90 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Using Equation (3.19), Equation (3.16) can be rewritten as

+∞
X 1
x
e(t) = X(jkω0 )ejkω0 t
k=−∞
T
+∞
1 X
= X(jkω0 )ejkω0 t
T k=−∞
+∞
ω0 X
= X(jkω0 )ejkω0 t
2π k=−∞
+∞
1 X
∴x
e(t) = X(jkω0 )ejkω0 t ω0 (3.20)
2π k=−∞

As T → ∞, x e(t) approaches x(t) and consequently ω0 → 0. Then, the summation

in the right-hand side of Equation (3.20) becomes an integral as
Z +∞
1
x(t) = X(jw)ejwt dw (3.21)
2π −∞

and Equation (3.18) becomes

Z +∞
X(jw) = x(t)e−jwt dt (3.22)
−∞

Here, x(t) and X(jw) are a Fourier transform pair and we can write a short-hand
notation as
FT
x(t) ←→ X(jw).
Equation (3.22) is termed the Fourier transform of x(t) and equation (3.21) the
inverse Fourier transform of X(jw). Equations (3.21) and (3.22) are also referred to
as the synthesis equation and analysis equation respectively.

Convergence of Fourier Transforms

The following Dirichlet conditions are sufficient to ensure for the convergence of
Fourier transforms.

1. x(t) should be absolutely integrable; that is

Z +∞
|x(t)|dt < ∞.
−∞

Signals & Systems 91 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

2. x(t) should have only a finite number of maxima and minima within any finite
interval.
3. x(t) should have only a finite number of discontinuities within any finite inter-
val. And, each of these discontinuities must be finite.

Example
Determine the Fourier transform of the following signal and plot the spectrum.
x(t) = e−at u(t) a > 0.

Solution
We have,
Z +∞
X(jω) = x(t)e−jωt dt
−∞
Z+∞
= e−at e−jωt dt
0
Z +∞
= e−(a+jω)t dt
0
∞
1
=− e−(a+jω)t
a + jω 0
1
∴ X(jω) =
a + jω
For magnitude spectrum and phase spectrum,

1 a − jω
X(jω) =
a + jω a − jω
a ω
= 2 2
−j 2
a +ω a + ω2
Therefore, the magnitude spectrum is
1
|X(jw)| = √
a2 + ω2
and the phase spectrum is
ω 
∠X(jw) = − tan−1 .
a
The magnitude spectrum and the phase spectrum of the given signal are plotted in

Signals & Systems 92 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

the Figure 3.9.

Figure 3.9: (a) Magnitude spectrum. (b) Phase spectrum.

Example

Determine and plot the Fourier transform of a rectangular pulse given as

(
1, |t| < T1
x(t) = .
0, |t| > T1

Signals & Systems 93 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Solution
The given signal is

Figure 3.10: A rectangular pulse.

The Fourier transform of x(t) can be evaluated as

Z T1
X(jω) = e−jωt dt
−T1
T
1 −jωt 1
=− e
jω −T1
1  −jωT1
− ejωT1

=− e
jω
2 ejωT1 − e−jωT1
 
=
ω 2j
sin(ωT1 )
∴ X(jω) = 2T1 .
ωT1
The spectrum X(jω) is as shown in Figure 3.11.

Figure 3.11: Fourier transform of a given rectangular pulse.

Signals & Systems 94 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Example
Determine the inverse Fourier transform of
(
1, |ω| < W
X(jω) = .
0, |ω| > W

Solution
The given spectrum is

Figure 3.12: A rectangular pulse in frequency domain.

The inverse Fourier transform of the given spectrum is determined as

Z W
1
x(t) = ejωt dω
2π −W
W
1 1 jωt
= e
2π jt
 jW t −W −jW t 
1 e −e
=
πt 2j
W sin(W t)
∴ x(t) = .
π Wt
This is shown in Figure 3.13.

Fourier Transform for Periodic Signals

Let us consider a signal x(t) which has the following Fourier transform
X(jw) = 2πδ(ω − ω0 ). (3.23)
Here, X(jω) is an impulse function of area 2π that exists at ω = ω0 . Taking inverse
Fourier transform of Equation (3.23), we get
Z ∞
1
x(t) = 2πδ(ω − ω0 )ejωt dω
2π −∞
= ejω0 t (3.24)

Signals & Systems 95 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.13: The inverse Fourier transform of the given spectrum.

In general, if
∞
X
X(jω) = 2πak δ(ω − kω0 ), (3.25)
k=−∞

then
∞
X
x(t) = ak ejkω0 t . (3.26)
k=−∞

The Equation (3.26) is actually the Fourier series representation of a periodic signal
with Fourier series coefficients {ak }. Therefore, the Fourier transform of a periodic
signal can be determined as a train of impulses occurring at the harmonically related
frequencies; the area of an impulse occurring at ω = kω0 is 2π times the Fourier
series coefficient ak .

Example

Find the Fourier transform of the following periodic square wave for T = 4T1 .

Figure 3.14: Periodic square wave.

Signals & Systems 96 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Solution
We know that Fourier transform of a continuous-time periodic signal x(t) can be
evaluated as ∞
X
X(jω) = 2πak δ(ω − kω0 ),
k=−∞

where {ak } are Fourier series coefficients and can be determined as

Z
1
ak = x(t)e−jkω0 t dt
T T
1 T1
Z
= x(t)e−jkω0 t dt
T −T1
1 T
1 1 −jkω0 t
= e
T −jkω0 −T1
1
e−jkω0 T1 − ejkω0 T1


=−
jk2π
1 ejkω0 T1 − e−jkω0 T1
 
=
kπ 2j
sin(kω0 T1 )
∴ ak = ,
kπ
and Z T1
1 2T1
a0 = dt = .
T −T1 T
For T = 4T1 , we have
sin k 2π T
sin kπ



T 4 2
ak = = ,
kπ kπ
and
2T1 1
a0 = = .
4T1 2
Therefore, the Fourier transform of the given signal x(t) is
∞
" #
X sin( kπ
2
)
X(jω) = 2π δ(ω − kω0 )
k=−∞
kπ
∞
X sin( kπ
2
)
= 2 δ(ω − kω0 ).
k=−∞
k

The frequency spectrum is shown in Figure 3.15.

Signals & Systems 97 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.15: Fourier transform of a periodic square wave.

3.7 Properties of Continuous-Time Fourier Trans-

form
1. Linearity
If
FT
x(t) ←→ X(jω)
and
FT
y(t) ←→ Y (jω),
then
FT
z(t) = Ax(t) + By(t) ←→ Z(jω) = AX(jω) + BY (jω).

Proof

Z ∞
Z(jω) = z(t)e−jωt dt
Z−∞
∞
= [Ax(t) + By(t)] e−jωt dt
−∞
Z ∞ Z ∞
−jωt
=A x(t)e dt + B y(t)e−jωt dt
−∞ −∞
∴ Z(jω) = AX(jω) + BY (jω)

2. Time Shifting
If
FT
x(t) ←→ X(jω),

Signals & Systems 98 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

then
FT
y(t) = x(t − t0 ) ←→ Y (jw) = e−jωt0 X(jω).

Proof
Z ∞
Y (jω) = y(t)e−jωt dt
Z−∞
∞
= x(t − t0 )e−jωt dt
−∞

Put t − t0 = τ , then
Z ∞
Y (jω) = x(τ )e−jω(τ +t0 ) dτ
−∞
Z ∞
−jωt0
=e x(τ )e−jωτ dτ
−∞
−jωt0
∴ Y (jω) = e X(jω)

3. Frequency Shifting
If
FT
x(t) ←→ X(jω),
then
FT
y(t) = ejω0 t x(t) ←→ X(j(ω − ω0 )).

Proof
Z ∞
Y (jω) = y(t)e−jωt dt
Z−∞
∞
= ejω0 t x(t)e−jωt dt
Z−∞
∞
= x(t)e−j(ω−ω0 )t dt
−∞
∴ Y (jω) = X(j(ω − ω0 )).

4. Time Reversal
If
FT
x(t) ←→ X(jω),
then
FT
y(t) = x(−t) ←→ Y (jω) = X(−jω).

Signals & Systems 99 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Proof

Z ∞
Y (jω) = y(t)e−jωt dt
Z−∞
∞
= x(−t)e−jωt dt
−∞

Put t = −t, then

Z ∞
Y (jω) = x(t)ejωt dt
∞
∴ Y (jω) = X(−jω)

5. Time Scaling
If
FT
x(t) ←→ X(jω),
then  
FT 1 jω
y(t) = x(at) ←→ Y (jω) = X ,
|a| a
where a is a real constant.

Proof

Z ∞
Y (jω) = y(t)e−jωt dt
Z−∞
∞
= x(at)e−jωt dt
−∞

Put at = τ , then
( R∞ ω
1
a −∞
x(τ )e−j a τ dτ, a > 0
Y (jω) = R∞ ω
− a1 −∞ x(τ )ej a τ dτ, a < 0
 
1 jω
∴ Y (jω) = X
|a| a
Therefore, time-scaling a signal x(t) by a factor of 0 a0 corresponds to the
frequency-scaling by 0 1/a0 and amplitude scaling by 0 1/|a|0 of its frequency
spectrum.

Signals & Systems 100 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

6. Conjugation
If
FT
x(t) ←→ X(jω),
then
FT
x∗ (t) ←→ X ∗ (−jω).

Proof

We have, Z ∞
X(jω) = x(t)e−jωt dt
−∞

Taking complex conjugate on both sides, we get

Z ∞ ∗
∗ −jωt
X (jω) = x(t)e dt
−∞
Z ∞
= x∗ (t)ejωt dt
−∞

Replacing ω by −ω, we get

Z ∞
∗
X (−jω) = x∗ (t)e−jωt dt.
−∞

Hence,
FT
x∗ (t) ←→ X ∗ (−jω).
Note

(a) If x(t) is real, then

X ∗ (−jω) = X(jω)
or
X ∗ (jω) = X(−jω).
Expressing X(jω) in rectangular form,

X(jω) = Re{X(jω)} + jIm{X(jω)}

and in polar form,

X(jω) = |X(jω)|ej∠X(jω)

Signals & Systems 101 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

we have,

Re{X(jω)} = Re{X(−jω)}
Im{X(jω)} = −jIm{X(−jω)}
|X(jω)| = |X(−jω)|
∠X(jω) = −∠X(−jω).

(b) If x(t) is real and even, then X(jω) is also real and even.
(c) If x(t) is real and odd, then X(jω) is purely imaginary and odd.
(d) Decomposing x(t) and X(jω) into even and odd parts; that is,

x(t) = xe (t) + xo (t)

X(jω) = Xe (jω) + Xo (jω),

then we can have

FT
xe (t) ←→ Re{X(jω)}
FT
xo (t) ←→ jIm{X(jω)}.

7. Multiplication
If
FT
x(t) ←→ X(jω),
and
FT
y(t) ←→ Y (jω),
then
FT 1
z(t) = x(t)y(t) ←→ Z(jω) = [X(jω) ∗ Y (jω)].
2π
Proof

Z ∞
Z(jω) = z(t)e−jωt dt
Z−∞
∞
= x(t)y(t)e−jωt dt
Z−∞∞  Z ∞ 
1 0 jω 0 t
= x(t) Y (jω )e dω e−jωt dt
0
−∞ 2π
Z ∞ Z−∞
∞ 
1 0 −j(ω−ω 0 )t 0
= x(t) Y (jω )e dω dt
2π −∞ −∞

Signals & Systems 102 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Put ω − ω 0 = σ, then
Z ∞ Z ∞ 
1 jσt
Z(jω) = x(t) Y (j(ω − σ))e dσ dt
2π −∞ −∞
Z ∞ Z ∞ 
1 jσt
= Y (j(ω − σ)) x(t)e dt dσ
2π −∞ −∞
Z ∞
1
= Y (j(ω − σ))X(jσ)dσ
2π −∞
Z ∞
1
= X(jσ)Y (j(ω − σ))dσ
2π −∞
1
∴ Z(jω) = [X(jω) ∗ Y (jω)]
2π
The multiplication property tells us that multiplication in the time domain cor-
responds to convolution in the frequency domain. The multiplication property,
sometimes, is also referred to as the modulation property.

8. Convolution
If
FT
x(t) ←→ X(jω),
and
FT
h(t) ←→ H(jω),
then
FT
y(t) = x(t) ∗ h(t) ←→ Y (jω) = X(jω)H(jω).
Proof
We have

y(t) = x(t) ∗ h(t)

Z ∞
= x(τ )h(t − τ )dτ.
−∞

Taking Fourier transform on both sides, we get

Z ∞ Z ∞ 
Y (jω) = x(τ )h(t − τ )dτ e−jωt dt.
−∞ −∞

Interchanging the order of integration, we have

Z ∞ Z ∞ 
−jωt
Y (jω) = x(τ ) h(t − τ )e dt dτ.
−∞ −∞

Signals & Systems 103 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

By using the time-shifting property, we have

Z ∞
h(t − τ )e−jωt dt = e−jωτ H(jω).
−∞

Therefore,
Z ∞
Y (jω) = x(τ )e−jωτ H(jω)dτ
−∞
Z ∞
= H(jω) x(τ )e−jωτ dτ
−∞
Y (jω) = H(jω)X(jω).

9. Parseval’s Relation for Continuous-Time Aperiodic Signals

If
FT
x(t) ←→ X(jω),
then Parseval’s relation for continuous-time aperiodic signals is defined by
Z ∞ Z ∞
2 1
E= |x(t)| dt = |X(jω)|2 dω.
−∞ 2π −∞

Proof
The total energy of a continuous-time aperiodic signal is determined as
Z ∞
E= |x(t)|2 dt
Z−∞
∞
= x(t)x∗ (t)dt
Z−∞
∞  Z ∞ 
1 ∗ −jωt
= x(t) X (jω)e dω dt
−∞ 2π −∞

Interchanging the order of integration, we have

Z ∞ Z ∞ 
1 ∗ −jωt
E= X (jω) x(t)e dt dω
2π −∞ −∞
Z ∞
1
= X ∗ (jω)X(jω)dω
2π −∞
Z ∞
1
= |X(jω)|2 dω.
2π −∞

Signals & Systems 104 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Therefore, Z ∞ Z ∞
2 1
|x(t)| dt = |X(jω)|2 dω.
−∞ 2π −∞
The Parseval’s relation tells us that the total energy may be determined either
by computing the energy per unit time (|x(t)|2 ) and integrating over all time
or by computing the energy per unit frequency (|X(jω)|2 /2π) and integrating
over all frequencies. For this reason, (|X(jω)|2 ) is often referred to as the
energy-density spectrum or the energy spectral density of the signal x(t).
10. Differentiation
If
FT
x(t) ←→ X(jω),
then
dx(t) F T
←→ jωX(jω).
dt
Proof
We have Z ∞
1
x(t) = X(jω)ejωt dω.
2π −∞
Differentiating both sides with respect to t, we get
Z ∞
dx(t) 1
= jωX(jω)ejωt dω.
dt 2π −∞
Therefore
dx(t) F T
←→ jωX(jω).
dt
11. Integration
If
FT
x(t) ←→ X(jω),
then Z t
FT 1
x(τ )dτ ←→ X(jω) + πX(0)δ(ω).
−∞ jω
Proof
We have Z ∞
1
x(t) = X(jω)ejωt dω.
2π −∞
Integrating both sides with respect to t, we get
Z t Z ∞
1 1
x(t)dt = X(jω)ejωt dω.
−∞ 2π −∞ jω

Signals & Systems 105 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Therefore Z t
1
FT
x(τ )dτ ←→ X(jω) + πX(0)δ(ω).
−∞ jω
The impulse term on the right-hand side is the dc constant that can result from
integration.
12. Duality
We have Fourier transform pairs
Z ∞
1
x(t) = X(jω)ejωt dω, (3.27)
2π −∞
Z ∞
X(jω) = x(t)e−jωt dt. (3.28)
−∞

There is a symmetry between Fourier transform and inverse Fourier transform

equations except the factor 2π and the sign change in the complex exponen-
tial. This symmetry leads us to interchange the roles of time and frequency;
the interchangeability property is termed duality. Figure 3.16 shows an exam-
ple of duality property in which a continuous rectangular pulse in either time
or frequency corresponds to a sinc function in either frequency or time.

Mathematically, duality property can be stated as below.

If
FT
x(t) ←→ X(jω),
then
FT
X(jt) ←→ 2πx(−ω).
Proof
Interchanging t and ω in Equation (3.27), we get
Z ∞
1
x(ω) = X(jt)ejtω dt. (3.29)
2π −∞
Replacing ω by −ω in Equation (3.29), we get
Z ∞
1
x(−ω) = X(jt)e−jtω dt.
2π −∞
Or Z ∞
2πx(−ω) = X(jt)e−jtω dt. (3.30)
−∞
Comparing Equation (3.30) with (3.27), we see that
FT
X(jt) ←→ 2πx(−ω).

Signals & Systems 106 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.16: Duality of rectangular pulses and sinc functions.

3.8 Discrete-Time Fourier Transform (DTFT)

We apply the similar approach to determine the Fourier transform of discrete-time
aperiodic signals as we did for the continuous-time case. Let us consider an aperiodic
signal x[n] as shown in Figure 3.17 (a). From this signal we can construct a periodic
signal x
e[n] as shown in Figure 3.17 (b) for which x[n] is one period. If the period N is
e[n] is identical to x[n] over the larger interval, and as N → ∞,
chosen to be larger, x
x
e[n] = x[n] for any value of n.

For x
e[n], we have the Fourier series representation as
2π
X
x
e[n] = ak ejk N n (3.31)
k=<N >

and Fourier series coefficients as

1 X 2π
ak = e[n]e−jk N n .
x (3.32)
N n=<N >

e[n] over a period extending from −N/2 to N/2 − 1 and x[n] is zero
Since x[n] = x

Signals & Systems 107 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.17: (a) An aperiodic signal x[n]. (b) A periodic signal x

e[n].

outside the interval −N1 ≤ n ≤ N2 , Equation (3.32) can be rewritten as

N2 ∞
1 X −jk 2π 1 X 2π
ak = x[n]e N = n
x[n]e−jk N n .
N n=−N N n=−∞
1

Let ∞
2π
X
X(ejkω0 ) = N ak = x[n]e−jk N n , (3.33)
n=−∞

then
1
ak = X(ejkω0 ). (3.34)
N
Using Equation (3.34) in (3.31), we get
X 1
x
e[n] = X(ejkω0 )ejkω0 n . (3.35)
k=<N >
N

Since ω0 = 2π/N ⇒ 1/N = ω0 /2π, Equation (3.35) can be rewritten as

1 X
x
e[n] = X(ejkω0 )ejkω0 n ω0 . (3.36)
2π k=<N >

As N → ∞, x e[n] approaches x[n] and consequently ω0 → 0. Then the summation in

the right-hand side becomes an integral. So, Equation (3.36) becomes
Z
1
x[n] = X(ejω )ejωn dω, (3.37)
2π 2π

Signals & Systems 108 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

and Equation (3.33) becomes

∞
X
jω
X(e ) = x[n]e−jωn . (3.38)
n=−∞

Here, the Equation (3.37) is the inverse discrete-time Fourier transform (I-DTFT)
and is called the synthesis equation whereas the Equation (3.38) is the discrete-time
Fourier transform (DTFT) and is called the analysis equation.

Example

Consider the signal x[n] = an u[n], 0 < a < 1. Find the Fourier transform of x[n].

Solution

Given signal x[n] is shown in Figure 3.18. The Fourier transform of x[n] can be

Figure 3.18: A signal x[n] = an u[n], 0 < a < 1.

determined as
∞
X
jω
X(e ) = x[n]e−jωn
n=−∞
X∞
n −jωn
= a e
n=0
∞
X
n
= ae−jω
n=0
1
∴ X(ejω ) = .
1 − ae−jω

Signals & Systems 109 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

For magnitude and phase spectra,

1
X(ejω ) =
1 − ae−jω
1
=
1 − a cos(ω) + ja sin(ω)
1 1 − a cos(ω) − ja sin(ω)
=
1 − a cos(ω) + ja sin(ω) 1 − a cos(ω) − ja sin(ω)
1 − a cos(ω) − ja sin(ω)
=
(1 − a cos(ω))2 + (a sin(ω))2
1 − a cos(ω) a sin(ω)
= 2 2
−j .
(1 − a cos(ω)) + (a sin(ω)) (1 − a cos(ω))2 + (a sin(ω))2

Therefore
1
|X(ejω )| = p
(1 − a cos(ω))2 + (a sin(ω))2
1
=p
1 − 2a cos(ω) + a cos2 (ω) + a2 sin2 (ω)
2

1
=p .
1 − 2a cos(ω) + a2

And  
jω −1 a sin(ω)
∠X(e ) = − tan .
1 − a cos(ω)
The magnitude spectrum and the phase spectrum are plotted in Figure 3.19 (a) and
(b) respectively.

Convergence of Discrete-Time Fourier Transform

The discrete-time Fourier transform X(ejω ) converges if x[n] is absolutely summable.
That is,
X∞
|x[n]| < ∞.
n=−∞

There is no convergence issue associated with the synthesis equation since the inte-
gration is of finite interval.

Signals & Systems 110 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Figure 3.19: (a) Magnitude spectrum. (b) Phase spectrum.

Discrete-Time Fourier Transform (DTFT) for Periodic Signals

In continuous time, we know that Fourier transform of ejω0 t is an impulse at ω = ω0
with area 2π. Since, discrete-time Fourier transform is periodic with period 2π, the
Fourier transform of ejω0 n has a train of impulses at ω0 , ω0 ± 2π, ω0 ± 4π, and so on
and each impulse has area 2π. That is, for x[n] = ejω0 n , the Fourier transform is
∞
X
X(ejω ) = 2πδ(ω − ω0 − 2πl).
l=−∞

Let us consider a periodic signal x[n] with period N and Fourier series represen-
tation X
x[n] = ak ejkω0 n ,
k=<n>

then, the Fourier transform of x[n] can determined as

∞
X
jω
X(e ) = 2πak δ(ω − ω0 k).
k=−∞

Signals & Systems 111 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Example
Determine the Fourier transform of the following periodic signal.

x[n] = cos(ω0 n).

Solution
We know
1 1
x[n] = ejω0 n + e−jω0 n
2 2
Since ∞
jω0 n F T
X
e ←→ 2πδ(ω − ω0 − 2πl),
l=−∞

therefore
∞
X ∞
X
X(ejω ) = πδ(ω − ω0 − 2πl) + πδ(ω + ω0 − 2πl)
l=−∞ l=−∞

That is,
X(ejω ) = πδ(ω − ω0 ) + πδ(ω + ω0 ), −π ≤ ω < π,
and X(ejω ) repeats periodically with period 2π. The spectrum is shown in Figure
3.20.

Figure 3.20: DTFT of cos(ω0 n).

3.9 Properties of Discrete-Time Fourier Transform

(DTFT)
We use the following short-hand notation to represent Fourier transform pair,
FT
x[n] ←→ X(ejω ).

Signals & Systems 112 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

1. Periodicity
The discrete-time Fourier transform is periodic in ω with period 2π. That is,

X ej(ω+2π) = X(ejω ).

2. Linearity
If
FT
x[n] ←→ X(ejω )
and
FT
y[n] ←→ Y (ejω ),
then
FT
Ax[n] + By[n] ←→ AX(ejω ) + BY (ejω ).

3. Time Shifting
If
FT
x[n] ←→ X(ejω ),
then
FT
x[n − n0 ] ←→ e−jωn0 X(ejω )

4. Frequency Shifting
If
FT
x[n] ←→ X(ejω ),
then
FT
ejω0 n x[n] ←→ X ej(ω−ω0 ) .

5. Time Reversal
If
FT
x[n] ←→ X(ejω ),
then
FT
x[−n] ←→ X(e−jω ).

6. Time Scaling

(a) Time Compression

y[n] = x[pn], |p| > 1.

Here, P is an integer value. The time scaling operation discards informa-
tion since y[n] retains only every pth value of x[n]. So, we define x[n] such

Signals & Systems 113 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

that it is zero unless n/p is integer.

If
FT
x[n] ←→ X(ejω ),
then ω
FT
y[n] = x[pn] ←→ Y (ejω ) = X(ej p ).
Proof

∞
X
jω
Y (e ) = y[n]e−jωn
n=−∞
X∞
= x[pn]e−jωn
n=−∞

Put pn = m, then
∞
X ω
jω
Y (e ) = x[m]e−j p m
m=−∞
jω
∴ Y (ejω ) = X(e ). p

(b) Time Expansion

If p is integer and |p| > 1, then
(
x[n/p], if n is a multiple of p
y[n] =
0, if n is not a multiple of p
is the expanded version of x[n].

If
FT
x[n] ←→ X(ejω ),
then
FT
y[n] = x[n/p] ←→ Y (ejω ) = X(ejpω ).
Proof

∞
X
Y (ejω ) = y[n]e−jωn
n=−∞
X∞
= x[n/p]e−jωn
n=−∞

Signals & Systems 114 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Put n/p = m, then

∞
X
jω
Y (e ) = x[m]e−jpm
m=−∞
jω jpm
∴ Y (e ) = X(e ).

7. Conjugation
If
FT
x[n] ←→ X(ejω ),
then
FT
x∗ [n] ←→ X ∗ (e−jω ).

(a) If x[n] is real valued, X(ejω ) is conjugate symmetric. That is,

X ∗ (e−jω ) = X(ejω ).

Note:

i.
Re{X(ejω )} = Re{X(e−jω )}
ii.
Im{X(ejω )} = −Im{X(e−jω )}
iii.
|X(ejω )| = |X(e−jω )|
iv.
∠X(ejω ) = −∠X(e−jω )
v.
FT
xe [n] ←→ Re{X(ejω )}
vi.
FT
xo [n] ←→ jIm{X(ejω )}
(b) If x[n] is real and even, then X(ejω ) is also real and even.
(c) If x[n] is real and odd, then X(ejω ) is odd and purely imaginary.

8. Multiplication
If
FT
x[n] ←→ X(ejω )

Signals & Systems 115 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

and
FT
y[n] ←→ Y (ejω ),
then Z
FT 1 jω
z[n] = x[n]y[n] ←→ Z(e ) = X(ejθ )Y (ej(ω−θ) )dθ.
2π 2π
Proof

∞
X
jω
Z(e ) = z[n]e−jωn
n=−∞
X∞
= x[n]y[n]e−jωn
n=−∞
∞ Z  
X 1
= y[n] X(e )e dθ e−jωn
jθ jθn

n=−∞
2π 2π
Z " ∞ #
1 X
= X(ejθ ) y[n]e−j(ω−θ)n dθ
2π 2π n=−∞
Z
1
∴ Z(ejω ) = X(ejθ )Y (ej(ω−θ) )dθ.
2π 2π

9. Convolution
If
FT
x[n] ←→ X(ejω )
and
FT
h[n] ←→ H(ejω ),
then
FT
y[n] = x[n] ∗ h[n] ←→ Y (ejω ) = X(ejω )H(ejω ).
Proof

∞
X
jω
Y (e ) = y[n]e−jωn
n=−∞
∞
" ∞
#
X X
= x[k]h[n − k] e−jωn
n=−∞ k=−∞
∞
" ∞
#
X X
= x[k] h[n − k]e−jωn
k=−∞ n=−∞

Signals & Systems 116 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

Using time-shifting property,

∞
X
h[n − k]e−jωn = e−jωk H(ejω ).
n=−∞

So,
∞
X
jω
Y (e ) = x[k]e−jωk H(ejω )
k=−∞
∞
X
= H(e ) jω
x[k]e−jωk
k=−∞
jω jω
= H(e )X(e )
∴ Y (ejω ) = X(ejω )H(ejω ).

10. Parseval’s Relation for Discrete-Time Aperiodic Signals

If
FT
x[n] ←→ X(ejω ),
then ∞ Z
X 1 2
E= |x[n]| = |X(ejω )|2 dω.
n=−∞
2π 2π

Proof

∞
X
E= |x[n]|2
n=−∞
X∞
= x[n]x∗ [n]
n=−∞
∞ Z 
X 1 ∗ jω −jωn
= x[n] X (e )e dω
n=−∞
2π 2π
Z " ∞ #
1 X
= X ∗ (ejω ) x[n]e−jωn dω
2π 2π n=−∞
Z
1
= X ∗ (ejω )X(ejω )dω
2π 2π
Z
1
∴E= |X(ejω )|2 dω
2π 2π

Signals & Systems 117 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 3. FOURIER ANALYSIS FOR CONTINUOUS-TIME AND
DISCRETE-TIME SIGNALS

According to Parseval’s relation, the total energy of a discrete-time aperiodic

signal x[n] can also be determined by integrating the energy per unit frequency,
|X(ejω )|2 /2π, over 2π interval. Here, |X(ejω )|2 is known as the energy-density
spectrum of the signal x[n].

11. First Difference

If
FT
x[n] ←→ X(ejω ),
then
FT
x[n] − x[n − 1] ←→ (1 − e−jω )X(ejω ).

12. Accumulation
If
FT
x[n] ←→ X(ejω ),
then n ∞
X FT 1 jω j0
X
x[m] ←→ X(e ) + πX(e ) δ(ω − 2πk).
m=−∞
1 − e−jω k=−∞

13. Differentiation in Frequency

If
FT
x[n] ←→ X(ejω ),
then
FT dX(ejω )
nx[n] ←→ j .
dω
Proof
∞
X
X(ejω ) = x[n]e−jωn .
n=−∞

Differentiating both sides with respect to ω, we get

∞
dX(ejω ) F T X
←→ −jnx[n]e−jωn .
dω n=−∞

FT dX(ejω )
∴ nx[n] ←→ j .
dω

Signals & Systems 118 Asst. Prof. Bijaya Shrestha, nec

Chapter 4

Discrete Fourier Transform (DFT)

Frequency-domain analysis of discrete-time signals is usually performed on a digital

signal processor. For performing the frequency-domain analysis, the signal in time
domain is converted into frequency domain. The frequency-domain representation is
given by the Fourier transform of the signal. Since, Fourier transform is a continuous
function of frequency, it can not be computed in a digital processor. To overcome this
problem, we represent the signal by samples of its spectrum. This kind of frequency-
domain representation leads to the discrete Fourier transform (DFT). The DFT is a
powerful computational tool to perform the frequency-domain analysis of discrete-
time signals. Again, the tool to reconstruct the signal from its frequency-domain
representation is referred to as the inverse discrete Fourier transform (IDFT). These
computational tools are very important in many digital signal processing applica-
tions, such as frequency analysis of signals, power spectrum estimation, and linear
filtering. The use of DFT and IDFT is also due to the fact that there are computa-
tionally efficient algorithms, known as fast Fourier transform (FFT) algorithms, to
compute the DFT and IDFT.

4.1 Frequency-Domain Sampling: Discrete Fourier

Transform (DFT)
4.1.1 Frequency-Domain Sampling and Reconstruction of Discrete-
Time Signals
Let us consider a discrete-time aperiodic signal x[n] with Fourier transform
∞
X
jω
X(e ) = x[n]e−jωn . (4.1)
n=−∞

119
 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Let us sample X(ejω ) in frequency at a spacing of δω between successive samples.

Since X(ejω ) is periodic with period 2π, only samples in the interval of 2π are
sufficient. Let us take N equidistant samples in the interval 0 ≤ ω ≤ 2π with
spacing δω = 2π
N
as shown in Figure 4.1.

Figure 4.1: Frequency-domain sampling of DTFT.

Evaluating Equation (4.1) at w = kδω = k 2π

N
, we obtain
∞
2π 2π
X
X(ejk N ) = x[n]e−jk N n k = 0, 1, 2, · · · , N − 1. (4.2)
n=−∞

The summation in Equation (4.2) can be subdivided into an infinite number of

summations, where each summation contains N terms. That is,
−1 N −1 2N −1
jk 2π −jk 2π −jk 2π 2π
X X X
X(e N ) = ··· + x[n]e N
n
+ x[n]e N
n
+ x[n]e−jk N n + · · ·
n=−N n=0 n=N
∞ +N −1
lNX
2π
X
= x[n]e−jk N n .
l=−∞ n=lN

In the inner summation, let n = n − lN , then

∞ N −1
jk 2π 2π
X X
X(e N )= x[n − lN ]e−jk N n .
l=−∞ n=0

Interchanging the order of the summation, we obtain

−1
N
" ∞ #
2π 2π
X X
X(ejk N ) = x[n − lN ] e−jk N n
n=0 l=−∞
N −1
2π
X
= xp [n]e−jk N n k = 0, 1, 2, · · · , N − 1, (4.3)
n=0

Signals & Systems 120 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

where ∞
X
xp [n] = x[n − lN ] (4.4)
l=−∞

is obtained by the periodic repetition of x[n] every N samples and hence is periodic
with fundamental period N . The xp [n] can be represented by Fourier series as
N −1
2π
X
xp [n] = ak ejk N n n = 0, 1, 2, · · · , N − 1, (4.5)
k=0

where the Fourier series coefficients are given by

N −1
1 X 2π
ak = xp [n]e−jk N n k = 0, 1, 2, · · · , N − 1. (4.6)
N n=0
Comparing Equations (4.3) and (4.6), we get
1 2π
ak = X(ejk N ) k = 0, 1, 2, · · · , N − 1. (4.7)
N
Therefore, Equation (4.5) becomes
N −1
1 X 2π 2π
xp [n] = X(ejk N )ejk N n n = 0, 1, 2, · · · , N − 1. (4.8)
N k=0
2π
The Equation (4.8) reconstructs the periodic signal xp [n] from the samples X(ejk N )of
the spectrum X(ejωn ). However, x[n] can be recovered from xp [n] only if there is
no aliasing in time domain. Figure 4.2 illustrates the condition for recovery of x[n]
from xp [n]. We have considered a finite-duration signal x[n], which is nonzero in the
interval 0 ≤ n ≤ L − 1. From Figure 4.2 (b), we observe that when N ≥ L,
x[n] = xp [n] 0 ≤ n ≤ N − 1.
So, x[n] can be recovered from xp [n]. But, when N < L as shown in Figure 4.2 (c),
there is time-domain aliasing and hence x[n] can not be recovered from xp [n].

4.1.2 Discrete Fourier Transfrom (DFT) Pairs

2π
We can not reconstruct the original sequence x[n] from its frequency samples X(ejk N n ), k =
0, 1, 2, · · · , N − 1, if x[n] is of infinite duration due to time aliasing of xp [n]. How-
ever, for a finite-duration sequence x[n] of length L ≤ N , xp [n] is a periodic repetition
of x[n], where xp [n] over a single period is given as
(
x[n], 0 ≤ n ≤ L − 1
xp [n] = . (4.9)
0, L≤n≤N −1

Signals & Systems 121 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.2: (a) Aperiodic signal of length L = 4. (b) Periodic version xp [n] of x[n]
for N ≥ L; here N = 6. (c) Periodic version xp [n] of x[n] for N < L; here N = 3.

And, x[n] can be reconstructed from its frequency samples by using Equation (4.8).
In Equation (4.9), we see that zeros have been padded for L ≤ n ≤ N − 1. It is to
be noted that zero-padding does not add any information about the spectrum of the
sequence.

For a finite-duration sequence x[n] of length L (i.e., x[n] = 0 for n < 0 and n ≥ L),
the Fourier transform can be determined as
L−1
X
X(e ) =jω
x[n]e−jωn 0 ≤ ω ≤ 2π. (4.10)
n=0

Sampling X(ejω ) at equally spaced frequencies kδω = k 2π

N
, k = 0, 1, 2, · · · , N − 1,
where N ≥ L, the frequency samples will be
L−1
jk 2π 2π
X
X[k] ≡ X(e N )= x[n]e−jk N n
n=0
N −1
2π
X
X[k] = x[n]e−jk N n k = 0, 1, 2, · · · , N − 1. (4.11)
n=0

Signals & Systems 122 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

The upper index in the sum has been increased from L − 1 to N − 1 since x[n] = 0
for N ≥ L. The relation in Equation (4.11) is called the discrete Fourier transform
(DFT) of x[n]. Equation (4.8) can be rewritten as
N −1
1 X 2π
x[n] = X[k]ejk N n n = 0, 1, 2, · · · , N − 1. (4.12)
N k=0

The relation in Equation (4.12) is called the inverse DFT (IDFT). Hence, we have
DFT pair equations as

N −1
2π
X
DFT: X[k] = x[n]e−jk N n k = 0, 1, 2, · · · , N − 1 (4.13)
n=0
N −1
1 X 2π
IDFT: x[n] = X[k]ejk N n n = 0, 1, 2, · · · , N − 1. (4.14)
N k=0

4.1.3 DFT as a Linear Transformation

Let us define a new term WN as
2π
WN = e−j N
which is the N th root of unity and is called the twiddle factor. Using twiddle factor,
Equations (4.13) and (4.14) may be expressed as
N
X −1
X[k] = x[n]WNkn k = 0, 1, 2, · · · , N − 1 (4.15)
n=0

N −1
1 X
x[n] = X[k]WN−kn n = 0, 1, 2, · · · , N − 1. (4.16)
N k=0

We can view DFT and IDFT as linear transformations on sequences {x[n]} and
{X[k]}, respectively. Let us define an N −point vector xN of signal x[n], an N −point
vector XN of frequency samples X[k] and, an N × N matrix WN as
   
x[0] X[0]
 x[1]   X[1] 
   
 x[2]   X[2] 
xN =   , XN =  ,
 ..   .. 
 .   . 
x[N − 1] X[N − 1]

Signals & Systems 123 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

···
 
1 1 1 1
1 WN1 WN2 ··· WNN −1 
2(N −1)
 
WN = 1
 WN2 WN4 ··· WN .

 .. .. .. .. 
. . . . 
2(N −1) (N −1)(N −1)
1 WNN −1 WN · · · WN

Now, the N −point DFT may be expressed in matrix form as

XN = WN x N

where WN is the matrix of linear transformation. And, IDFT may be expressed in

matrix form as
1 ∗
xN = W XN
N N

∗
where the matrix WN is the complex conjugate of the matrix WN .

Example

Compute the 4−point DFT of the sequence x[n] = {0, 1, 2, 3} with and without
linear transformation method.

Solution

Without linear transformation

N −1
2π
X
X[k] = x[n]e−jk N n k = 0, 1, 2, · · · , N − 1
n=0
3
2π
X
= x[n]e−jk N n
n=0

Signals & Systems 124 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

For k = 0 : X[0] = x[0] + x[1] + x[2] + x[3]

=0+1+2+3
=6
2π 2π 2π
For k = 1 : X[1] = x[0] + x[1]e−j1 4 1 + x[2]e−j1 4 2 + x[3]e−j1 4 3
= 0 + (1)(−j) + (2)(−1) + (3)(j)
= −2 + j2
2π 2π 2π
For k = 2 : X[2] = x[0] + x[1]e−j2 4 1 + x[2]e−j2 4 2 + x[3]e−j2 4 3
= 0 + (1)(−1) + (2)(1) + (3)(−1)
= −2
2π 2π 2π
For k = 3 : X[3] = x[0] + x[1]e−j3 4 1 + x[2]e−j3 4 2 + x[3]e−j3 4 3
= 0 + (1)(j) + (2)(−1) + (3)(−j)
= −2 − j2
Hence, X[k] = {6, −2 + j2, −2, −2 − j2}.

Using linear transformation

We have,
XN = WN x N
    
X[0] 1 1 1 1 x[0]
X[1] 1 W41 W42 W43  x[1]
X[2] = 1 W42 W44 W46  x[2]
    

X[3] 1 W43 W46 W49 x[3]

where
2π
W4 = e−j 4 = −j, W41 = −j, W42 = −1, W43 = j, W44 = 1, W46 = −1, W49 = −j.
Now,
    
X[0] 1 1 1 1 0
X[1] 1 −j −1 j  1
X[2] = 1 −1 1 −1 2
    

X[3] 1 j −1 −j 3
 
1×0+1×1+1×2+1×3
1 × 0 − j × 1 − 1 × 2 + j × 3
= 
1 × 0 − 1 × 1 + 1 × 2 − 1 × 3
1×0+j×1−1×2−j×3

Signals & Systems 125 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

 
6
−2 + j2
=
 −2 


−2 − j2

Hence, X[k] = {6, −2 + j2, −2, −2 − j2}.

4.1.4 Relationship of DFT to Other Transforms

Relationship of DFT to Fourier Series Coefficients of Continuous-Time
Periodic Signal

We know that a continuous-time periodic signal x(t) with fundamental period T =

2π/ω0 can be expressed in Fourier series as

∞
X
x(t) = ak ejkω0 t
k=−∞

where {ak } are the Fourier series coefficients. If x(t) is sampled at a uniform rate
with t = nTs = n/Fs , where Ts is the sampling interval and Fs is the sampling
frequency, we obtain the discrete-time sequence

∞
X
x[n] = x[nT s] = ak ejkω0 nT s
k=−∞
∞
X f0
= ak ejk2π Fs n
k=−∞
X∞
= ak ejk2πf n
k=−∞

where f = f0 /Fs = 1/N is called the relative frequency. So,

∞
2π
X
x[n] = ak ejk N n (4.17)
k=−∞

Signals & Systems 126 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Decomposing the summation in Equation (4.17) into an infinite number of summa-

tions, where each summation contains N terms, we obtain
−1 N −1 2N −1
2π 2π 2π
X X X
x[n] = · · · + ak ejk N n + ak ejk N n + ak ejk N n + · · ·
k=−N k=0 k=N
∞ lNX
+N −1
2π
X
= ak ejk N n .
l=−∞ k=lN

In the inner summation, let k = k − lN , then

∞ N −1
2π
X X
x[n] = ak−lN ejk N n .
l=−∞ k=0

Interchanging the order of summation, we obtain

−1
N
" ∞ #
2π
X X
x[n] = ak−lN ejk N n
k=0 l=−∞
N −1
2π
X
= ak−lN ejk N n
e (4.18)
k=0

where ∞
X
ak−lN =
e ak−lN
l=−∞

is an aliased version of {ak }. Comparing Equation (4.18) with the formula of IDFT
in Equation (4.14), we get
X[k] = Neak .

Relationship of DFT to Fourier Series Coefficients of Discrete-Time Peri-

odic Signal
A discrete-time periodic signal xp [n] can be represented by Fourier series as
N −1
2π
X
xp [n] = ak ejk N n (4.19)
k=0

where the Fourier series coefficients are given by

N −1
1 X 2π
ak = xp [n]e−jk N n k = 0, 1, 2, · · · , N − 1. (4.20)
N n=0

Signals & Systems 127 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

If we define a signal x[n] = xp [n], 0 ≤ n ≤ N − 1, then comparing Equation(4.20)

with the formula of DFT in Equation (4.13), we get

X[k] = N ak .

Also, the Equation (4.19) is comparable to the equation of IDFT.

Relationship of DFT to Fourier Transform of Discrete-Time Aperiodic

Signal
We have already discussed about this topic while making derivation of DFT. However,
we will discuss briefly here once again. For a discrete-time aperiodic signal x[n], the
Fourier transform is given by
∞
X
jω
X(e ) = x[n]e−jωn .
n=−∞

If we sample X(ejω ) at N equally spaced frequencies kδω = k 2π N

, k = 0, 1, 2, · · · , N −
1, we get
∞
2π
X
X[k] = x[n]e−jk N n , k = 0, 1, 2, · · · , N − 1.
n=−∞

The spectral components X[k] are the DFT values of the periodic sequence xp [n] of
period N , given by
∞
X
xp [n] = x[n − lN ].
l=−∞

The xp [n] is the periodic repetition of x[n]. For an infinite-duration sequence x[n],
xp [n] is the time-aliased version of x[n]. But, for a finite-duration sequence x[n], if
(
xp [n], 0 ≤ n ≤ N − 1
x[n] = ,
0, otherwise

the original sequence x[n] can be reconstructed from the DFT coefficients {X[k]}.

4.2 Properties of DFT

We use the following short-hand notation to denote the N −point DFT pair x[n] and
X[k].
DF T
x[n] ←−→ X[k]
N

Signals & Systems 128 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

4.2.1 Periodicity
If
DF T
x[n] ←−→ X[k],
N

then

x[n + N ] = x[n] for all n

X[k + N ] = X[k] for all k.

4.2.2 Linearity
If
DF T
x1 [n] ←−→ X1 [k]
N

and
DF T
x2 [n] ←−→ X2 [k],
N

then
DF T
Ax1 [n] + Bx2 [n] ←−→ AX1 [k] + BX2 [k].
N

4.2.3 Circular Symmetries of a Sequence

Recall that N −point DFT of a finite-duration signal x[n] of length L ≤ N is equal
to the N −point DFT of a periodic signal xp [n] with period N , which is the periodic
version of x[n] given as
∞
X
xp [n] = x[n − lN ].
l=−∞

If we shift the periodic sequence xp [n] by n0 units to the right, we get another periodic
sequence as
∞
X
x0p [n] = xp [n − k] = x[n − k − lN ]
l=−∞

The finite-duration sequence

(
x0p [n], 0 ≤ n ≤ N − 1
x0 [n] =
0, otherwise

is related to the original sequence x[n] by a circular shift. Here x0 [n] is the sequence
resulted by circular shifting of x[n] by n0 time units. We take counterclockwise direc-
tion as the positive direction. Therefore, the circular shift of an N −point sequence

Signals & Systems 129 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

is equivalent to the linear shift of its periodic extension, and vice versa. This is il-
lustrated in Figure 4.3 for 4−point sequence x[n]. The circular shift of the sequence
can be represented as the index modulo N . That is,

x0 [n] = x[(n − n0 )]N = x[N + n − n0 ]

An N −point sequence is circularly even if it is symmetric about point zero on

the circle. That is,
x[N − n] = x[n] 1 ≤ n ≤ N − 1.
An N −point sequence is called circularly odd if it is antisymmetric about the point
zero on the circle. That is,

x[N − n] = −x[n] 1 ≤ n ≤ N − 1.

The time reversal of an N −point sequence is attained by reversing its samples

about the point zero on the circle. That is,

x[(−N )]N = x(N − n) 0 ≤ n ≤ N − 1.

The time reversal is achieved actually by plotting x[n] in a clockwise direction on a

circle.

Circular Convolution
Let x1 [n] and x2 [n] be the two sequences of length N . And,
if
DF T
x1 [n] ←−→ X1 [k]
N

and
DF T
x2 [n] ←−→ X2 [k],
N

then
DF T
x3 [m] = x1 [n] N x2 [n] ←−→ X3 [k] = X1 [k]X2 [k].
N

Here, x3 [m] is called the circular convolution between x1 [n] and x2 [n] and is expressed
as
NX−1
x3 [m] = x1 [n]x2 [(m − n)]N m = 0, 1, 2, · · · , N − 1.
n=0

This property tells us that circular convolution of two sequences in time domain
results multiplication of their corresponding DFTs in frequency domain.

Signals & Systems 130 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.3: Circular shift of a sequence x[n].

Example

Perform the circular convolution between x1 [n] = {1, 2, 3, 4} and x2 [n] = {2, 3, 4, 5}.

Signals & Systems 131 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Solution
The circular convolution between x1 [n] and x2 [n] can be computed by using the
expression
N
X −1
x3 [m] = x1 [n]x2 [(m − n)]N m = 0, 1, 2, · · · , N − 1
n=0
X3
= x1 [n]x2 [(m − n)]4 m = 0, 1, 2, 3.
n=0

This expression can be implemented graphically as shown in Figure 4.4.

Hence, the result is

x3 [m] = {36, 38, 36, 30}.

Note
In circular convolution, if the length of the sequences x1 [n] and x2 [n] is not equal,
zeros are padded to the smaller sequence to make its length equal to that of the larger
one. Furthermore, we can compute linear convolution of two finite-length sequences
by using circular convolution. For that, the length of both the sequences must be
made equal to l1 + l2 − 1 by padding zeros, where l1 is the length of sequence x1 [n]
and l2 is the length of sequence x2 [n].

4.2.4 Circular Time Shift of a Sequence

If
DF T
x[n] ←−→ X[k],
N

then
DF T 2π
x[(n − n0 )]N ←−→ e−jk N n0 X[k].
N

4.2.5 Circular Frequency Shift

If
DF T
x[n] ←−→ X[k],
N

then
2π DF T
ejl N n x[n] ←−→ X[(k − l)]N .
N

Signals & Systems 132 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.4: Circular convolution of given sequences x1 [n] and x2 [n].

4.2.6 Time Reversal

If
DF T
x[n] ←−→ X[k],
N

then
DF T
x[(−n)]N ←−→ X[(−k)]N .
N

Signals & Systems 133 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

4.2.7 Conjugation
If
DF T
x[n] ←−→ X[k],
N

then
DF T
x∗ [n] ←−→ X ∗ [(−k)]N .
N

4.2.8 Multiplication
If
DF T
x1 [n] ←−→ X1 [k]
N

and
DF T
x2 [n] ←−→ X2 [k],
N

then
DF T 1
x1 [n]x2 [n] ←−→ X1 [k] N X2 [k].
N N

4.2.9 Parseval’s Relation

If
DF T
x[n] ←−→ X[k],
N

then
N
X −1 N
X −1
E= |x[n]|2 = |X[k]|2 .
n=0 k=0

4.2.10 Circular Correlation

If
DF T
x[n] ←−→ X[k]
N

and
DF T
y[n] ←−→ Y [k],
N

then
N −1
DF T e
X
rexy [l] = x[n]y ∗ [(n − l)]N ←−→ R ∗
XY [k] = X[k]Y [k].
N
n=0

Here, rexy [l] is the circular crosscorrelation sequence of x[n] and y[n].

Signals & Systems 134 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

4.3 Fast Fourier Transform (FFT)

We already explained that DFT has an important role in many applications of digital
signal processing; the major reason being the existence of efficient algorithms for its
computation. Fast Fourier transform (FFT) is an algorithm to compute the DFT
faster. FFT algorithm reduces the number of arithmetic operations such as multipli-
cations and additions so as to make the computation of DFT faster. The basic idea
of FFT algorithms is to compute DFT by decomposing it into smaller DFTs.

Let us recall the formulas of DFT and IDFT as

N
X −1
X[k] = x[n]WNkn k = 0, 1, 2, · · · , N − 1
n=0

and
N −1
1 X
x[n] = X[k]WN−kn n = 0, 1, 2, · · · , N − 1.
N k=0
Since the formulas for DFT and IDFT involve similar operations, the efficient algo-
rithms for DFT apply also for IDFT. From the formula of DFT, we can see that for
each value of k, direct computation of the DFT requires N complex multiplications
and N − 1 complex additions. Therefore, to compute N DFT values, N 2 complex
multiplications and N (N − 1) complex additions are required. The FFT algorithms
are used to reduce the number of arithemetic operations for faster computation of
DFT.

4.3.1 Radix-2 FFT Algorithms

FFT algorithms are applicable when N can be factored as

N = r1 r2 r3 · · · rv

where {rj } are prime numbers. If r1 = r2 = r3 = · · · = rv ≡ r, then N = rv .

In this case, the DFTs are of size r. Here, the number r is called the radix of the
FFT algorithm. If r = 2, then N = 2v and the algorithm is called the radix-2 FFT
algorithm. This algorithm is the most widely used FFT algorithm.

Decimation-In-Time FFT (DIT-FFT) Algorithm

In this algorithm, we split the N −point data sequence x[n] into two N/2−point
data sequences corresponding to the even-numbered and odd-numbered samples of

Signals & Systems 135 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

x[n]. Since the two N/2−point sequences are obtained by decimating x[n], the
algorithm is referred to as the decimation-in-time FFT (DIT-FFT) algorithm. The
DFT computations are done separately for these two sequences and added to get the
overall N −point DFT. That is,
N
X −1
X[k] = x[n]WNkn k = 0, 1, 2, · · · , N − 1
n=0
X X
= x[n]WNkn + x[n]WNkn
n even n odd
N/2−1 N/2−1
X X k(2m+1)
= x[2m]WN2km + x[2m + 1]WN
m=0 m=0
N/2−1 N/2−1
X X
= x[2m]WN2km + WNk x[2m + 1]WN2km .
m=0 m=0

 2π 2 2π
We know that WN2 = e−j N = e−j N/2 = WN/2 . Therefore,

N/2−1 N/2−1
X X
km
X[k] = x[2m]WN/2 + WNk km
x[2m + 1]WN/2
m=0 m=0
= G[k] + WNk H[k] k = 0, 1, 2, · · · , N − 1 (4.21)

where
N/2−1
X
km
G[k] = x[2m]WN/2
m=0

is the N/2−point DFT of even-numbered samples of x[n] and

N/2−1
X
km
H[k] = x[2m + 1]WN/2
m=0

is the N/2−point DFT of odd-numbered samples of x[n].

Since X[k] is periodic with period N , G[k] and H[k] will also be periodic with
k+N/2
period N/2. Also, WN = −WNk . Then, Equation (4.21) can be expressed as

X[k] = G[k] + WNk H[k] k = 0, 1, 2, · · · , N/2 − 1 (4.22)

N
X[k + ] = G[k] − WNk H[k] k = 0, 1, 2, · · · , N/2 − 1. (4.23)
2

Signals & Systems 136 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Further G[k] and H[k] can be computed as follows:

X X
km km
G[k] = g[m]WN/2 + g[m]WN/2
m even m odd
N/4−1 N/4−1
X X
2kb k 2kb
= g[2b]WN/2 + WN/2 g[2b + 1]WN/2
b=0 b=0
N/4−1 N/4−1
X X
kb k kb
= g[2b]WN/4 + WN/2 g[2b + 1]WN/4
b=0 b=0
k
= I[k] + WN/2 J[k] k = 0, 1, 2, · · · , N/2 − 1 (4.24)

where
N/4−1
X
kb
I[k] = g[2b]WN/4
b=0

is the N/4−point DFT of the even-numbered samples of g[m] and

N/4−1
X
kb
J[k] = g[2b + 1]WN/4
b=0

is the N/4−point DFT of the odd-numbered samples of g[m]. I[k] and J[k] are
k+N/4 k
periodic with period N/4 and WN/2 = −WN/2 . Then Equation (4.24) can be
expressed as
k
G[k] = I[k] + WN/2 J[k] k = 0, 1, 2, · · · , N/4 − 1 (4.25)
k
G[k + N/4] = I[k] − WN/2 J[k] k = 0, 1, 2, · · · , N/4 − 1. (4.26)

Similarly,
N/4−1 N/4−1
X X
kb k kb
H[k] = h[2b]WN/4 + WN/2 h[2b + 1]WN/4 k = 0, 1, 2, · · · , N/2 − 1.
b=0 b=0

In the radix-2 DIT-FFT algorithm, the total number of complex multiplications is

reduced to (N/2) log2 N and the total number of complex additions is reduced to
N log2 N . We can illustrate the radix-2 DIT-FFT algorithm for the computation of
8-point DFT with the help of block diagram as shown in Figure 4.5. According to
the figure, we need to perform three stages. We begin by computing four 2−point
DFTs, then two 4−point DFTs, and finally, one 8−point DFT. The 8−point DFT
can be computed with the help of signal flow graph as shown in Figure 4.6. This
signal flow graph is also called the butterfly structure by its look.

Signals & Systems 137 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.5: Block digram of radix-2 DIT-FFT algorithm for computing 8-point DFT.

Figure 4.6: Signal flow graph of radix-2 DIT-FFT algorithm for computing 8-point
DFT.

Signals & Systems 138 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Bit Reversal Technique

The shuffling of the data sequence can be done by the bit reversal technique. If we
express the index n in binary form, the order of the decimated sequence is obtained
by reading the binary representation of n in reverse order. The implementation of
this technique for 8−point data is shown in Table 4.1.

Table 4.1: Bit reversal technique for 8−point sequence.

n n2 n1 n0 n0 n1 n2 n
0 000 000 0
1 001 100 4
2 010 010 2
3 011 110 6
4 100 001 1
5 101 101 5
6 110 011 3
7 111 111 7

For 4−point DFT, radix-2 DIT-FFT algorithm can be described by the block
diagram and signal flow graph as shown in Figures 4.7 and 4.8 respectively.

Figure 4.7: Block diagram of radix-2 DIT-FFT algorithm for computing 4−point
DFT.

Decimation-In-Frequency FFT (DIF-FFT) Algorithm

We have the DFT formula as
N
X −1
X[k] = x[n]WNkn k = 0, 1, 2, · · · , N − 1.
n=0

Splitting the DFT formula into two summations, one of which involves the sum over
the first N/2 data points and the second sum involves the last N/2 data points, we

Signals & Systems 139 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.8: Signal flow graph of radix-2 DIT-FFT algorithm for computing 4−point
DFT.

get
N/2−1 N −1
X X
X[k] = x[n]WNkn + x[n]WNkn
n=0 n=−N/2
N/2−1 N/2−1
X kN/2
X
= x[n]WNkn + WN x[n + N/2]WNkn .
n=0 n=0
kN/2
Since WN = (−1)k ,
N/2−1
X 
x[n] + (−1)k x[n + N/2 WNkn .

X[k] =
n=0

Let us split X[k] into even and odd-numbered samples. Thus, we obtain
N/2−1
X
kn
X[2k] = [x[n] + x[n + N/2]] WN/2 k = 0, 1, 2, · · · , N/2 − 1
n=0

and
N/2−1
X
X[2k + 1] = {[x[n] − x[n + N/2]] WNn } WN/2
kn
k = 0, 1, 2, · · · , N/2 − 1.
n=0

Let
N N
g[n] = x[n] + x[n + ] n = 0, 1, 2, · · · , − 1
2 2
and
N N
h[n] = x[n] − x[n + ]W n n = 0, 1, 2, · · · , − 1,
2 N 2
then,

Signals & Systems 140 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.9: First stage of DIF-FFT algorithm for computing 8-point DFT.

N
2
−1
X
kn N
X[2k] = g[n]WN/2 k = 0, 1, 2, · · · , −1
n=0
2

and
N
2
−1
X
kn N
X[2k + 1] = h[n]WN/2 k = 0, 1, 2, · · · , − 1.
n=0
2

This computational procedure can be repeated through decimation of the N/2−point

DFTs, X[2k] and X[2k + 1]. For illustrative purpose, the first stage of 8−point DIF-
FFT algorithm is shown in Figure 4.9. As in DIT-FFT algorithm, this also requires
(N/2) log2 N complex multiplications and N log2 N complex additions. The signal
flow graph of radix-2 DIF-FFT algorithm for computing 8−point DFT is shown in
Figure 4.10.

Signals & Systems 141 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 4. DISCRETE FOURIER TRANSFORM (DFT)

Figure 4.10: Signal flow graph of radix-2 DIF-FFT algorithm for computing 8−point
DFT.

Applications of FFT
1. Filtering

2. Correlation

3. Spectrum analysis

4. Power spectrum estimation

Signals & Systems 142 Asst. Prof. Bijaya Shrestha, nec

Chapter 5

Energy and Power

5.1 Parseval’s Theorem

5.1.1 Parseval’s Theorem for Power Signals
Parseval’s theorem for a power signal states that the average power of a signal x(t)
can be determined from its Fourier series coefficients as
∞
X
P = |ak |2 .
k=−∞

For proof, go through the properties of continuous-time Fourier series of chapter 3.

5.1.2 Parseval’s Theorem for Finite Energy Signals

Parseval’s theorem for a finite energy signal states that the total energy of a signal
x(t) can be determined from its Fourier transform as
Z ∞
1
E= |X(jω)|2 dω.
2π −∞
For proof, go through the properties of continuous-time Fourier transform of chapter
3.

5.2 Power Spectral Density (PSD)

Power spectral density (PSD) of a power signal x(t) is defined as the distribution of
power per unit bandwidth as a function of frequency.

143
 CHAPTER 5. ENERGY AND POWER

Figure 5.1: A power signal.

Let us consider a power signal x(t) as shown in Figure 5.1. We can assume the
power signal as a limiting case of an energy signal. Let us consider the truncated
version xT (t) of the signal x(t) as shown in Figure 5.2. We know that the average

Figure 5.2: A truncated version of x(t).

power of x(t) is given by

Z T /2
1
P = lim |x(t)|2 dt.
T →∞ T −T /2

As long as T is finite, xT (t) has finite energy. The average power in terms of xT (t) is
1 ∞
Z
P = lim |xT (t)|2 dt.
T →∞ T −∞

Let XT (jω) be the Fourier transform of xT (t), then using Parseval’s theorem,
Z ∞ Z ∞ Z ∞
2 1 2
|xT (t)| dt = |XT (jω)| dω = |XT (f )|2 df.
−∞ 2π −∞ −∞

Therefore,
1 ∞
Z
P = lim |XT (f )|2 df
T →∞ T −∞
Z ∞
|XT (f )|2
= lim df
−∞ T →∞ T
Z ∞
= S(f )df
−∞

Signals & Systems 144 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 5. ENERGY AND POWER

where
|XT (f )|2
S(f ) = lim
T →∞ T
is called the power spectral density (PSD) or power density spectrum of the power
signal x(t). Therefore, the total average power of a power signal x(t) is the sum of
average powers contributed by all the spectral components.

Properties of PSD
1. PSD of a power signal x(t) is a non-negative real-valued function of frequency.
That is,
S(f ) ≥ 0.
2. PSD of a real-valued power signal x(t) is an even function of frequency. That
is,
S(f ) = S(−f ).
3. The total area under the curve of PSD of a power signal x(t) is equal to the
average power of the signal. That is,
Z ∞
P = S(f )df.
−∞

4. For a power signal, auto-correlation function and PSD function form a Fourier
transform pair. That is,
FT
R(τ ) ←→ S(f ).

5.3 Energy Spectral Density (ESD)

Energy spectral density (ESD) of an energy signal x(t) is defined as the distribution
of energy per unit bandwidth as a function of frequency.

Let us consider an energy signal x(t) with Fourier transform X(jω). Then, ac-
cording to Parseval’s theorem,
Z ∞ Z ∞ Z ∞
1 2 2
E= |X(jω)| dω = |X(f )| df = ψ(f )df
2π −∞ −∞ −∞
where
ψ(f ) = |X(f )|2
is called the energy spectral density (ESD) of x(t). Therefore, The total energy of a
signal x(t) is the sum of energies contributed by all the spectral components of the
signal x(t).

Signals & Systems 145 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 5. ENERGY AND POWER

Properties of ESD
1. ESD of an energy signal x(t) is a non-negative real-valued function of frequency.
That is,
ψ(f ) ≥ 0.

2. ESD of a real-valued signal x(t) is an even function of frequency. That is,

ψ(f ) = ψ(−f ).

3. The total area under the curve of ESD of an energy signal x(t) is equal to the
total energy of the signal.

4. For an energy signal x(t), the auto-correlation function and ESD function form
a Fourier transform pair. That is,
FT
R(τ ) ←→ ψ(f ).

5.4 Correlation
Correlation is defined as the measure of similarity between two signals. Correlation
is of two types:

1. Auto-correlation

2. Cross-correaltion.

5.4.1 Auto-Correlation
Auto-correlation is defined as the measure of similarity between a signal x(t) and its
delayed version x(t − τ ).

Auto-Correlation Function for Energy Signals

The auto-correlation function for an energy signal x(t) is defined by
Z ∞
R(τ ) = x(t)x∗ (t − τ )dt.
−∞

Also, Z ∞
R(τ ) = x(t + τ )x∗ (t)dt.
−∞

Signals & Systems 146 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 5. ENERGY AND POWER

Properties of Auto-Correlation Function for Energy Signals

1. Auto-correlation function exhibits conjugate symmetry. That is,
R(τ ) = R∗ (−τ ).

Proof
Z ∞
R(τ ) = x(t)x∗ (t − τ )dt
−∞
Taking complex conjugate on both sides, we get
Z ∞
∗
R (τ ) = x(t)x∗ (t − τ )dt.
−∞

Let τ = −τ , then
Z ∞
∗
R (−τ ) = x∗ (t)x(t + τ )dt
Z−∞
∞
= x(t + τ )x∗ (t)dt
−∞
∗
∴ R (−τ ) = R(τ ).

2. The value of auto-correlation function for τ = 0 (at origin) is equal to the total
energy of the signal. That is,
Z ∞
R(0) = |x(t)|2 dt = E.
−∞

Proof

For τ = 0,
Z ∞
R(0) = x(t)x∗ (t)dt
Z−∞
∞
= |x(t)|2 dt.
−∞
= E.

3. Auto-correlation function has maximum value at origin. That is,

R(τ ) ≤ R(0) for all τ.

Signals & Systems 147 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 5. ENERGY AND POWER

4. Auto-correlation function and ESD form a Fourier transform pair. That is,
FT
R(τ ) ←→ ψ(f ).

Proof
Z ∞
R(τ ) = x(t)x∗ (t − τ )dt
−∞

We have, Z ∞
∗
x (t − τ ) = X ∗ (f )e−j2πf (t−τ ) df.
−∞

So, Z ∞ Z ∞ 
∗ −j2πf (t−τ )
R(τ ) = x(t) X (f )e df dt.
−∞ −∞

Changing the order of integration, we get

Z ∞ Z ∞ 
∗ j2πf τ −j2πf t
R(τ ) = X (f )e x(t)e dt df
−∞ −∞
Z ∞
= X ∗ (f )X(f )ej2πf τ df
Z−∞
∞
= |X(f )|2 ej2πf τ df
Z−∞
∞
∴ R(τ ) = ψ(f )ej2πf τ df.
−∞

Hence,
FT
R(τ ) ←→ ψ(f ).

Auto-Correlation Function for Power Signals

The auto-correlation function for a power signal x(t) is defined by
Z T /2
1
R(τ ) = lim x(t)x∗ (t − τ )dt.
T →∞ T −T /2

Also,
Z T /2
1
R(τ ) = lim x(t + τ )x∗ (t)dt.
T →∞ T −T /2

Signals & Systems 148 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 5. ENERGY AND POWER

Properties of Auto-Correlation Function for Power Signals

1. Auto-correlation function exhibits conjugate symmetry. That is,

R(τ ) = R∗ (−τ ).

2. The value of auto-correlation function for τ = 0 (at origin) is equal to the

average power of the signal. That is,

R(0) = P.

3. Auto-correlation function has maximum value at origin. That is,

R(τ ) ≤ R(0) for all τ.

4. Auto-correlation function and PSD form a Fourier transform pair. That is,
FT
R(τ ) ←→ ψ(f ).

This is known as Wiener-Khintchine theorem.

5.5 Cross-Correlation
Cross-correlation is defined as the measure of similarity between a signal and the
delayed version of another signal. The cross-correlation function between two energy
signals x1 (t) and x2 (t) is defined by
Z ∞
R12 (τ ) = x1 (t)x∗2 (t − τ )dt.
−∞

And, the cross-correlation function between two power signals x1 (t) and x2 (t) is
defined by
Z T /2
R12 (τ ) = lim x1 (t)x∗2 (t − τ )dt.
T →∞ −T /2

Signals & Systems 149 Asst. Prof. Bijaya Shrestha, nec

Chapter 6

Transmission of Signals

6.1 Transfer Function or Frequency Response

Let us consider an LTI system with impulse response h(t). For any arbitrary input
signal x(t), the output signal y(t) can be determined by performing convolution
between input signal and impulse response as,

y(t) = x(t) ∗ h(t). (6.1)

Let
FT
x(t) ←→ X(jω),
FT
h(t) ←→ H(jω),
and
FT
y(t) ←→ Y (jω),
then, using the convolution property of Fourier transform for Equation (6.1), we get

Y (jω) = X(jω)H(jω).

Now,
Y (jω)
H(jω) =
X(jω)
is the transfer function or frequency response of the system. So, the frequency re-
sponse of a system is defined as the ratio of Fourier transform of output signal to
the Fourier transform of input signal.

Let
Y (jω) = |Y (jω)|ej∠Y (jω) ,

150
CHAPTER 6. TRANSMISSION OF SIGNALS

X(jω) = |X(jω)|ej∠X(jω) ,
and
H(jω) = |H(jω)|ej∠H(jω) ,
then
|Y (jω)|ej∠Y (jω) |Y (jω)| j(∠Y (jω)−∠X(jω))
|H(jω)|ej∠H(jω) = j∠X(jω)
= e .
|X(jω)|e |X(jω)|
That is, the magnitude response of the system is given by
|Y (jω)|
|H(jω)| = ,
|X(jω)|
and the phase response is given by

∠H(jω) = ∠Y (jω) − ∠X(jω).

6.2 Distortionless Transmission

Any undesired change in the wave-shape of an electric signal passing through a phys-
ical system is referred to as distortion.

The transmission of a signal through an LTI system is said to be distortionless if

the output signal is an exact replica of the input signal. However, a constant change
in magnitude and a constant change in time delay in the output replica is not treated
as distortion. So, the signal x(t) is said to be transmitted without distortion if the
output signal becomes
y(t) = Kx(t − td ). (6.2)
Taking Fourier transform on both sides of Equation (6.2), we get

Y (jω) = KX(jω)e−jωtd .

We know,
Y (jω)
H(jω) = = Ke−jωtd .
X(jω)
Here, the magnitude response is

|H(jω)| = K, (6.3)

and the phase response is

∠H(jω) = −ωtd . (6.4)
Therefore, for distortionless transmission,

Signals & Systems 151 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 6. TRANSMISSION OF SIGNALS

1. the magnitude response |H(jω)| must be constant as shown in Figure 6.1 (a).
This means that all the frequency components must be equally attenuated or
amplified and the system bandwidth must be infinite, and

2. the phase response ∠H(jω) must be directly proportional to the frequency as

shown in Figure 6.1 (b).

Figure 6.1: Distortionless transmission. (a) Magnitude response. (b) Phase response.

Thus, for distortionless transmission, the system bandwidth must be infinite. But,
no practical system has infinite bandwidth and therefore distortionless transmission
can not be achieved practically.

The following distortions (linear distortions) occur during transmission of signals

through LTI systems.

A. Amplitude Distortion

Amplitude distortion occurs when the magnitude response is not constant within the
frequency band of interest and the frequency components present in the input signal
are transmitted with different gain or attenuation.

B. Phase Distortion

Phase distortion occurs when the phase response is not linearly changing with fre-
quency and the different frequency components in the input signal are subjected to
different time delays during transmission.

Signals & Systems 152 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 6. TRANSMISSION OF SIGNALS

6.3 Ideal Low Pass Filter (LPF)

6.3.1 Frequency Response of Ideal LPF
Frequency response of an ideal low pass filter (LPF) is given by
(
Ke−jωtd , |ω| < ωc
H(jω) = . (6.5)
0, |ω| > ωc

Here, the magnitude response is

(
K, |ω| < ωc
|H(jω)| = ,
0, |ω| > ωc

and the phase response is

(
−ωtd , |ω| < ωc
∠H(jω) = ,
0, |ω| > ωc

where td is the transmission delay and ωc is the cut-off frequency of the low pass
filter. The magnitude response and the phase response of the ideal low pass filter
are shown in Figures 6.2 (a) and (b) respectively.

We see that the ideal low pass filter allows the frequency components from −ωc
to ωc to pass without distortion and rejects the components above |ωc |.

Figure 6.2: Ideal low pass filter. (a) Magnitude response. (b) Phase response.

Signals & Systems 153 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 6. TRANSMISSION OF SIGNALS

6.3.2 Impulse Response of Ideal LPF

We know that the inverse Fourier transform of frequency response is the impulse
response of the LTI system. Taking inverse Fourier transform of Equation (6.5), we
get

Z ωc
1
h(t) = Ke−jωtd ejωt dω
2π −ωc
K ωc jω(t−td )
Z
= e dω
2π −ωc
ωc
K
= ejω(t−td )
2πj(t − td )
 jω (t−t ) −ωc−jω (t−t ) 
K e c d −e c d
=
π(t − td ) 2j
Kωc sin(ωc (t − td ))
∴ h(t) = .
π wc (t − td )

The impulse response of an ideal LPF is the sinc function delayed by td as shown

Figure 6.3: Impulse response of ideal LPF.

in Figure 6.3. Since the impulse response is non causal, the ideal LPF can not be
realized practically.

Signals & Systems 154 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 6. TRANSMISSION OF SIGNALS

6.4 Frequency Response and Impulse Response of

RC Circuit
Let us consider an RC circuit as shown in Figure 6.4. Applying Kirchhoff’s voltage
law, we get

Figure 6.4: An RC circuit.

vin (t) = i(t)R + vo (t)

dvo (t)
or, vin (t) = RC + vo (t)
dt
dvo (t)
or, vo (t) + RC = vin (t) (6.6)
dt

Taking Fourier transform on both sides of Equation (6.6), we get

Vo (jω) + RCjωVo (jω) = Vin (jω)

or, Vo (jω)(1 + jωRC) = Vin (jω).

Therefore, the frequency response of the given RC circuit is

1
Vo (jω) 1 RC
H(jω) = = = 1 . (6.7)
Vin (jω) 1 + jωRC jω + RC

We have,
FT 1
e−at u(t) ←→ .
jω + a
Therefore, the inverse Fourier transform of H(jω) is

1 − t
h(t) = e RC u(t),
RC
which is the impulse response of the given RC circuit.

Signals & Systems 155 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 6. TRANSMISSION OF SIGNALS

6.5 Frequency Response and Impulse Response of

RL Circuit
Consider an RL circuit as shown in Figure (6.5). Applying Kirchhoff’s voltage law,
we get

Figure 6.5: An RL circuit.

di(t)
vin (t) = L + vo (t)
 dt
d voR(t)
or, vin (t) = L + vo (t)
dt
L dvo (t)
or, vin (t) = + vo (t)
R dt
L dvo (t)
or, vo (t) + = vin (t). (6.8)
R dt
Taking Fourier transform on both sides of Equation (6.8), we get
L
Vo (jω) + jωVo (jω) = Vin (jω)
R 
L
or, Vo (jω) 1 + jω = Vin (jω).
R
Therefore, the frequency response of the given RL circuit is
R
Vo (jω) 1 L
H(jω) = = L
= R
. (6.9)
Vin (jω) 1 + jω R jω + L
Since
FT 1
e−at u(t) ←→ ,
jω + a
the inverse Fourier transform of Equation (6.9) gives
R −Rt
h(t) = e L u(t),
L
which is the impulse response of the given RL circuit.

Signals & Systems 156 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 6. TRANSMISSION OF SIGNALS

6.6 Introduction to Communication System

Communication is the process of conveying an information-bearing (message) signal
from one point (source) to another point (destination). For this, different electronic
devices, such as, transducer, amplifier, modulator, demodulator, filter, antenna, etc.,
are interconnected to each other to form a communication system.

The essential components of a communication system are information source,

input transducer, transmitter, communication channel, receiver, output transducer,
and destination. The block diagram of a general communication system is shown in
Figure 6.6.

Figure 6.6: Block diagram of a communication system.

Information Source
The Information source originates a message, such as a human voice, television pic-
ture, or data. The message can be analog or digital in nature. Analog communication
sources produce continuously varying message signals. Microphone is a good exam-
ple of an analog source. On the other hand, digital communication sources produce
a finite set of possible messages. A digital computer is a good example of a digital
source. There are a finite number of characters (or messages) that can be emitted
by such source.

Input Transducer
The function of transducer is to change one form of energy into another form. The
message produced by the information source may or may not be in electrical form,
such as human voice and television picture. If the message is not in electrical form, an
input transducer is used to convert it into an electrical waveform referred to as base-
band signal or message signal. Microphone is an example of input transducer which
converts sound waves (e.g. human voice) into corresponding electrical waveform.

Signals & Systems 157 Asst. Prof. Bijaya Shrestha, nec

 CHAPTER 6. TRANSMISSION OF SIGNALS

Transmitter

Transmitter converts the message signal into a suitable form for reliable communica-
tion over a channel. Different processes are performed in the transmitter depending
upon the type of communication systems. Modulation is the most important process
performed in the transmitter. In this process, the message signal is superimposed
upon the high-frequency carrier signal for making the message signal suitable for
transmission. Amplification and filtration may also be required before transmission.

Communication Channel

The communication channel is the medium through which signal travels from the
transmitter to the receiver. During transmission, the signal is distorted due to several
factors: the channel imperfections, addition of noise and interfering signals (origi-
nating from other sources). As a consequence, the received signal is the corrupted
version of the transmitted signal.

The communication channels are either guided (wireline) or unguided (wireless).

The examples of guided channels are twisted-pair cable, coaxial cable, optical fiber,
and waveguide. And, the examples of unguided channels are air, vacuum, and sea-
water.

Receiver

The receiver receives the signal from the channel and recovers the original message
signal in electrical form by the process known as demodulation. Due to distortions
in the received signal, the demodulated signal may be degraded to some extent.
Equalizers are the devices that can be used to reduce the effect of distortions. Other
operations performed in the receiver are filtering and amplification.

Output Transducer

The output transducer converts the electrical message signal into its original form.
For example, speaker is the output transducer which converts the electrical signal
into corresponding sound wave.

Destination

The destination is the point where the message is finally delivered.

Signals & Systems 158 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 6. TRANSMISSION OF SIGNALS

Equalization
To compensate for linear distortion, we may use a network known as equalizer con-
nected in cascade with the system as shown in Figure 6.7. The equalizer is designed
in such a way that inside the frequency band of interest the overall magnitude re-
sponse and phase response of this cascade connection approximates the conditions for
distortionless transmission within prescribed limits. The overall frequency response

Figure 6.7: Use of equalizer in communication system.

of the cascade is Hc (jω)Heq (jω). For distortionless transmission,

Hc (jω)Heq (jω) = Ke−jωtd , (6.10)

where td is a constant time delay and K is some constant. Therefore, the frequency
response of the equalizer is inversely related to that of the channel as,

Ke−jωtd
Heq (jω) = , (6.11)
Hc (jω)

where Hc (jω) 6= 0.

In practice, the equalizer is designed such that its frequency response approxi-
mates the ideal value of Equation (6.11) close enough for the linear distortion to be
reduced to a satisfactory level.

Signals & Systems 159 Asst. Prof. Bijaya Shrestha, nec

Chapter 7

Transmission of Signals in
Discrete-Time Systems

7.1 Transfer Function or Frequency Response of

Discrete-Time Systems
For a discrete-time LTI system with impulse response h[n], the output y[n] for any
arbitrary input x[n] can be determined by performing the convlution sum between
x[n] and h[n] as
y[n] = x[n] ∗ h[n]. (7.1)
Let
FT
x[n] ←→ X(ejω ),
FT
h[n] ←→ H(ejω ),

and,
FT
y[n] ←→ Y (ejω ),
then, applying the convolution property of Fourier transform for Equation (7.1), we
get
Y (ejω ) = X(ejω )H(ejω ).
Therefore,
Y (ejω )
H(ejω ) =
X(ejω )
is the frequency response of a discrete-time LTI system.

160
CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Also, applying Z-transform on Equation (7.1), we get

Y (z) = X(z)H(z).

Therefore,
Y (z)
H(z) =
X(z)
is the frequency response of a discrete-time LTI system in terms of Z-transform.

7.2 Types of Discrete-Time Systems or Digital Fil-

ters
7.2.1 Finite-Duration Impulse Response (FIR) Systems
If the impulse response is of finite duration, then the system is referred to as the FIR
system. For a discrete-time LTI system, We have

∞
X
y[n] = h[k]x[n − k].
k=−∞

For FIR system, the impulse response must be of finite length. Let M be the length
of the impulse response. Then, causal FIR system is defined by

M
X −1
y[n] = h[k]x[n − k]
k=0
or, y[n] = h[0]x[n] + h[1]x[n − 1] + h[2]x[n − 2] + · · · + h[M − 1]x[n − M + 1].
(7.2)

This difference equation tells us that the output of a causal FIR system at any
instant depends on the present and past inputs. Thus, a causal FIR system has a
finite memory of length M − 1, where M is the length of the impulse response.

An example of FIR System

The moving average system is an example of the FIR system.

Signals & Systems 161 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Proof
A moving average system is expressed as
M −1
1 X
y[n] = x[n − k].
M k=0

Let M = 4, then, for x[n] = δ[n], we will have

M −1
1X
h[n] = δ[n − k].
4 k=0

Now,
3
1X 1
h[0] = δ[−k] = ,
4 k=0 4
3
1X 1
h[1] = δ[1 − k] = ,
4 k=0 4
3
1X 1
h[2] = δ[2 − k] = ,
4 k=0 4
3
1X 1
h[3] = δ[3 − k] = .
4 k=0 4

And, h[n] = 0 for n < 0 and n > 3. That is,

(
1
, 0≤n≤3
h[n] = 4 .
0, otherwise

Since, the impulse response is of finite duration, then the moving average system is
the FIR system.

Frequency Response of FIR System

For a causal FIR system, we have
M
X −1
y[n] = h[k]x[n − k].
k=0

Signals & Systems 162 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Taking Z-transform on both sides, we get

M
X −1
Y (z) = h[k]X(z)z −k .
k=0

Therefore,
M −1
Y (z) X
H(z) = = h[k]z −k
X(z) k=0

is the frequency response of the causal FIR system.

7.2.2 Infinite-Duration Impulse Response (IIR) Systems

If the impulse response is of infinite duration, then the system is known as the IIR
system. A causal IIR system can be expressed as
∞
X
y[n] = h[k]x[n − k]
k=0
= h[0]x[n] + h[1]x[n − 1] + h[2]x[n − 2] + · · · + h[∞]x[n − ∞].

Here, the impulse response is of infinite duration and the IIR system requires infinite
memory length.

An Example of IIR System

The cumulative average system is an example of the IIR system.

Proof
The cumulative average system can be expressed as
n
1 X
y[n] = x[k].
n + 1 k=0

Then, the impulse response of this system is

n
1 X
h[n] = δ[k].
n + 1 k=0

Signals & Systems 163 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Now,
h[0] = 1
1
h[1] =
2
1
h[2] =
3
..
.
1
h[n] = .
n+1
Since the impulse response is of infinite duration, the cumulative average system is
the IIR system.

Frequency Response of IIR System

An IIR system can be described by an N th order linear constant-coefficient difference
(LCCD) equation given as
N
X M
X
ak y[n − k] = bk x[n − k]. (7.3)
k=0 k=0

Or,
N
X M
X
a0 y[n] + ak y[n − k] = bk x[n − k].
k=1 k=0
Normalizing both sides by a0 and rearranging the terms, we get the recursive formula
as
M
X N
X
y[n] = bk x[n − k] − ak y[n − k]. (7.4)
k=0 k=1
Taking Z-transform on both sides of Equation (7.4), we get
M
X N
X
−k
Y (z) = bk X(z)z − ak Y (z)z −k
k=0 k=1
PM −k
Y (z) k=0 bk z
or, = .
X(z) 1+ N
P −k
k=1 ak z
Therefore, PM −k
k=0 bk z
H(z) = PN
1+ −k
k=1 ak z
is the frequency response of an IIR system.

Signals & Systems 164 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

7.3 Implementation of FIR Systems

Recalling Equation (7.2), we have

y[n] = h[0]x[n] + h[1]x[n − 1] + h[2]x[n − 2] + · · · + h[M − 1]x[n − M + 1].

We can implement this causal FIR system in the direct form as shown in Figure 7.1.

Figure 7.1: Direct-form structure of FIR system.

7.4 Implementation of IIR Systems

For an IIR system, we have the difference equation as

M
X N
X
y[n] = bk x[n − k] − ak y[n − k] (7.5)
k=0 k=1

and frequency response as

PM −k
k=0 bk z
H(z) = PN . (7.6)
1+ a z −k
k=1 k

We can implement the IIR systems in different structures.

7.4.1 Direct Form Structures

Direct Form I Structure
We can realize the Equation (7.5) or (7.6) by direct form I structure as shown in
Figure 7.2.

Signals & Systems 165 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Figure 7.2: Direct-form I structure.

Direct Form II Structure

The system described by direct form I structure can be considered as a cascade of
two LTI systems; one with input x[n] and output w[n] and the other with input w[n]
and output y[n]. We can interchange the order of these systems without changing
the input-output behavior of the cascade. This is shown in Figure 7.3 for M = N .

Since the direct form II structure minimizes the number of memory locations
(i.e., maximum(M, N )), it is also known as the canonic structure. The direct form
II realization is better than direct form I in terms of memory size requirement.

Example
Realize the following difference equation in direct form I and direct form II structures.

2y[n] − y[n − 1] + 0.5y[n − 2] = 2x[n] + 3x[n − 1] + x[n − 3]. (7.7)

Solution
Rearranging the given difference equation in recursive form, we get

y[n] = x[n] + 1.5x[n − 1] + 0.5x[n − 3] + 0.5y[n − 1] − 0.25y[n − 2]. (7.8)

Signals & Systems 166 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Figure 7.3: Direct-form II structure.

Taking Z-transform on both sides, we get

Y (z) = X(z) + 1.5z −1 X(z) + 0.5z −3 X(z) + 0.5z −1 Y (z) − 0.25z −2 Y (z)
or, Y (z)(1 − 0.5z −1 + 0.25z −2 ) = X(z)(1 + 1.5z −1 + 0.5z −3 )
Y (z) 1 + 1.5z −1 + 0.5z −3
or, H(z) = = . (7.9)
X(z) 1 − 0.5z −1 + 0.25z −2
We can implement the Equation (7.8) or (7.9) in direct-form I and direct-form II
structures as shown in Figures 7.4 and 7.5 respectively.

7.4.2 Cascade-Form Structures

In cascade-form realization, the transfer function is split such that
H(z) = H1 (z)H2 (z) · · · HK (z).

Example
Realize the following IIR system in cascade-form structure.
y[n] − 2y[n − 1] − 3y[n − 2] = x[n] + 4x[n − 1].

Signals & Systems 167 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Figure 7.4: Direct-form I structure of the given IIR system.

Figure 7.5: Direct-form II structure of the given IIR system.

Figure 7.6: Cascade-form realization.

Solution

Taking z-transform on both sides, we get

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

or, Y (z) 1 − 2z −1 − 3z −2 = X(z) 1 + 4z −1

Signals & Systems 168 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Y (z) 1 + 4z −1
or, H(z) = =
X(z) 1 − 2z −1 − 3z −2
1 + 4z −1
=
1 − 3z −1 + z −1 − 3z −2
1 + 4z −1
=
(1 + z −1 )(1 − 3z −1 )
1 + 4z −1
  
1
=
1 + z −1 1 − 3z −1
= H1 (z)H2 (z)

where
1 + 4z −1
H1 (z) =
1 + z −1
and
1
H2 (z) = .
1 − 3z −1
Now, we can realize the cascade-form structure as shown in Figure 7.7.

Figure 7.7: Cascade-form realization of the given IIR system.

7.4.3 Parallel-Form Structures

We can realize an IIR system in parallel-form structure by performing a partial-
fraction expansion of H(z). For N ≥ M , we can have

H(z) = C + H1 (z) + H2 (z) + · · · + HK (z).

Example
Realize the following IIR system in parallel-form structure.

y[n] − 2y[n − 1] − 3y[n − 2] = x[n] + 2x[n − 1] + 3x[n − 2].

Signals & Systems 169 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Figure 7.8: Parallel-form realization.

Solution
Taking z-transform on both sides, we get

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

or, Y (z) 1 − 2z −1 − 3z −2 = X(z) 1 + 2z −1 + 3z −2

Y (z) 1 + 2z −1 + 3z −2
or, =
X(z) 1 − 2z −1 − 3z −2

1 + 2z −1 + 3z −2
∴ H(z) =
1 − 2z −1 − 3z −2
z 2 + 2z + 3
=
z 2 − 2z − 3
z 2 + 2z + 3
=
z 2 − 3z + z − 3
z 2 + 2z + 3
=
(z + 1)(z − 3)
H(z) z 2 + 2z + 3
or, = .
z z(z + 1)(z − 3)
Performing partial-fraction expansion of Equation, we have
z 2 + 2z + 3 A B C
= + +
z(z + 1)(z − 3) z z+1 z−3

or, z 2 + 2z + 3 = A(z + 1)(z − 3) + Bz(z − 3) + Cz(z + 1).

Signals & Systems 170 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Put z = 0:
3 = A(1)(−3) ⇒ A = −1.

Put z = −1:
1 − 2 + 3 = B(−1)(−4) ⇒ B = 1/2.

Put z = 3:
9 + 6 + 3 = C(3)(4) ⇒ C = 3/2.

Therefore,

H(z) −1 1/2 3/2

= + +
z z z+1 z−3
(1/2)z (3/2)z
or, H(z) = −1 + +
z+1 z−3
1/2 3/2
∴ H(z) = −1 + −1
+ .
1+z 1 − 3z −1

Now, we can implement the given IIR system in parallel-form structure as shown in
Figure 7.9.

Figure 7.9: Parallel-form realization of the given IIR system.

Signals & Systems 171 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

7.5 Solution of Linear Constant-Coefficient Dif-

ference (LCCD) Equations
An N th order linear constant-coefficient difference (LCCD) equation is given as
N
X M
X
ak y[n − k] = bk x[n − k]. (7.10)
k=0 k=0

The complete solution y[n] of a difference equation is the sum of homogeneous solu-
tion or complementary solution yh [n] and particular solution yp [n]. That is,

y[n] = yh [n] + yp [n]. (7.11)

The complete solution y[n] can be determined also as the sum of zero-input
response yzi [n] and zero-state response yzs [n]. That is,

y[n] = yzi [n] + yzs [n]. (7.12)

Example 1.5.1
If a discrete-time LTI system is given by

y[n] − 0.5y[n − 1] = x[n] (7.13)

where x[n] = 0.25n u[n], then determine y[n].

Solution
Homogeneous Solution
Homogeneous solution of the given difference equation is the solution of the homoge-
nous equation:
y[n] − 0.5y[n − 1] = 0. (7.14)
Let yh [n] = λn , then

λn − 0.5λn−1 = 0
or, λn−1 (λ − 0.5) = 0
∴ λ = 0.5.

Therefore, we can let homogeneous solution to be

yh [n] = A0.5n . (7.15)

Signals & Systems 172 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Particular Solution
Let yp [n] = B(0.25)n u[n], then Equation (7.13) becomes

B(0.25)n u[n] − (0.5)B(0.25)n−1 u[n − 1] = 0.25n u[n]. (7.16)

For n = 1 (choose the value of n such that none of the terms of Equation (7.16)
vanishes), Equation (7.16) becomes

B(0.25)1 − (0.5)B(0.25)0 = 0.251

or, 0.25B − 0.5B = 0.25
∴ B = −1.

Therefore,
yp [n] = −0.25n u[n]. (7.17)
Complete Solution
Now, the complete solution is

y[n] = A0.5n − 0.25n u[n]. (7.18)

We know that an LTI system is initially at rest. In this example, y[n] = 0 for n < 0.
Put n = 0 in Equation (7.13), then we get

y[0] = 0.5y[−1] + x[0] = 0.250 = 1.

Again, put n = 0 in Equation (7.18), we get

y[0] = A0.50 − 0.250 = 1 ⇒ A = 2.

Therefore, for n ≥ 0,
y[n] = 2(0.5)n − 0.25n
and for all n,
y[n] = [2(0.5)n − 0.25n ]u[n].

Example 1.5.2
If a discrete-time LTI system is given by

y[n] − 0.5y[n − 1] = x[n] (7.19)

where x[n] = 0.25n u[n], then determine y[n] as the sum of zero-input response yzi [n]
and zero-state response yzs [n].

Signals & Systems 173 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Solution
Zero-Input Response or Natural Response
The zero-input response is the response of a difference equation which is determined
for x[n] = 0. In this example,the zero-input response yzi [n] is the solution of

y[n] − 0.5y[n − 1] = 0.

Let yzi [n] = λn , then

λn − 0.5λn−1 = 0 ⇒ λ = 0.5.
Therefore,
yzi [n] = A0.5n . (7.20)
For a given LTI system, y[n] = 0 for n < 0. Using y[−1] = 0 in Equation (7.20), we
get
0 = A0.5−1 ⇒ A = 0.
∴ yzi [n] = 0.
Note that the zero-input response of every LTI system is zero.
Zero-State Response or Forced Response
The zero-state response is the response of a difference equation for a particular input
signal x[n]. This response is actually equal to the complete solution with zero initial
condition. That means,
yzs [n] = yh [n] + yp [n] (7.21)
with zero initial condition.

For determining homogeneous solution, let yh [n] = λn . Then,

λn − 0.5λn−1 = 0
or, λn−1 (λ − 0.5) = 0
∴ λ = 0.5.

Therefore,
yh [n] = B0.5n . (7.22)

For determining particular solution, let yp [n] = C0.25n u[n], then Equation (7.19)
becomes
C0.25n u[n] − (0.5)C0.25n−1 u[n − 1] = 0.25n u[n].

Signals & Systems 174 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

For n = 1,
C0.251 − (0.5)C0.250 = 0.251
or, (C)(0.25) − (0.5)(C) = 0.25 ⇒ C = −1.
Therefore,
yp [n] = −0.25n u[n].
Considering the zero initial condition (i.e.,y[n] = 0 for n < 0), we have
yzs [n] = B0.5n − 0.25n u[n]. (7.23)
Put n = 0 in Equation (7.19) and consider initial condition y[−1] = 0, then we get
y[0] − 0.5y[−1] = 0.250
or, y[0] = 1.
Put n = 0 in Equation (7.23), we get
yzs [0] = B0.50 − 0.250 = 1 ⇒ B = 2.
Therefore, for n ≥ 0,
yzs [n] = 2(0.5)n − 0.25n
and for all n,
yzs [n] = [2(0.5)n − 0.25n ]u[n].
Complete Solution
Therefore, the complete solution is
y[n] = [2(0.5)n − 0.25n ]u[n].

Example 1.5.3
Determine the impulse response of a discrete-time LTI system described by
y[n] − 2y[n − 1] − 3y[n − 2] = x[n] + 2x[n − 1]. (7.24)
To determine homogeneous solution yh [n], let yh [n] = λn , then
λn − 2λn−1 − 3λn−2 = 0
or, λn−2 λ2 − 2λ − 3 = 0

or, λ2 − 2λ − 3 = 0
or, λ2 − 3λ + λ − 3 = 0
or, (λ + 1)(λ − 3) = 0
∴ λ = −1, 3.

Signals & Systems 175 Asst. Prof. Bijaya Shrestha, nec

CHAPTER 7. TRANSMISSION OF SIGNALS IN DISCRETE-TIME SYSTEMS

Therefore,
yh [n] = C1 (−1)n + C2 (3)n .
For x[n] = δ[n], the particular solution yp [n] is zero. Also, for this LTI system,
y[n] = 0 for n < 0. Therefore, the complete solution is

y[n] = C1 (−1)n + C2 (3)n n ≥ 0. (7.25)

Put n = 0 in Equation (7.24), then we get

y[0] − 2y[−1] − 3y[−2] = x[0] + 2x[−1]

or, y[0] = 1.

Put n = 1 in Equation (7.24), then we get

y[1] − 2y[0] − 3y[−1] = x[1] + 2x[0]

or, y[1] − (2)(1) = (2)(1)
or, y[1] = 4.

Put n = 0 in Equation (7.25), then we get

y[0] = C1 (−1)0 + C2 (3)0

or, C1 + C2 = 1. (7.26)

Put n = 1 in Equation (7.25), then we get

y[1] = C1 (−1)1 + C2 (3)1

or, − C1 + 3C2 = 4. (7.27)

Solving Equations (7.26) and (7.27), we get

1 5
C1 = − and C2 = .
4 4
Therefore, from Equation (7.25), the impulse response of the given system is
 
1 n 5 n
h[n] = − (−1) + (3) u[n].
4 4

Signals & Systems 176 Asst. Prof. Bijaya Shrestha, nec

Bibliography

[1] A. V. Oppenheim, A. S. Willsky and S. H. Nawab, Signals & Systems, 3rd ed.,
PHI Learning, New Delhi, 2012.

[2] S. Haykin and B. V. Veen, Signals & Systems, 2nd ed., Wiley-India, New Delhi,
2011.

[3] J. G. Proakis and D. G. Manolakis, Digital Signal Processing, 3rd ed., Pearson
Education, Delhi, 2004.

[4] S. Haykin, An Introduction to Analog and Digital Communications, Wiley-

India, New Delhi, 2010.

[5] S. Sharma, Signals & Systems, 8th ed., S.K. Kataria and Sons, New Delhi, 2015.