‫‪Lectures on‬‬

‫‪Digital Signal Processing‬‬

‫محاضرات في‬

‫
ا رة ا
‬

‫أ‪.‬د‪ .‬زاھر محسن المالكي‬

‫جامعة الكوفة ‪2014 -‬‬

‫‪1 / 124‬‬
 ‫‪Prof. Zahir M. Hussain‬‬ ‫الپروفسور زاھر محسن حسين المالكي‪:‬‬
‫ولد في مدينة النجف األشرف وتعلم فيھا ثم أكمل دراسته الجامعية األولية )ھندسة كھربائية عامة( والعليا )ماجستير ھندسة‬ ‫‪.1‬‬
‫إلكترون واتصاالت( في قسم الھندسة الكھربائية – جامعة بغداد والدكتوراه ھندسة الحاسوب )معالجة األشارة( في جامعة كوينزالند‬
‫التكنولوجية ‪ -‬أستراليا ‪.2002‬‬
‫وقد شغل مناصب تعادل عالميا مدرس مساعد )‪ 2001-1999‬في جامعة كوينزالند التكنولوجية( ثم مدرس ‪2001‬؛ استاذ مساعد‬ ‫‪.2‬‬
‫‪2002‬؛ استاذ ‪ 2005‬في ھندسة الحاسوب‪ -‬معالجة اإلشارة على المالك الدائم بجامعة ملبورن الملكية التكنولوجية ‪ RMIT‬وقد‬
‫شغل بعد عودته للوطن عام ‪ 2009‬منصب بروفسور في جامعة ملبورن الملكية إضافة إلى منصب االستاذية )ھندسة حاسبات( في‬
‫جامعة الكوفة لغاية ‪2013‬؛ وھو االن بروفسور بجامعة إيدث كوان األسترالية‪.‬‬
‫المشرف األول على ‪ 15‬طالب دكتوراه وطالب ماجستير واحد في جامعة ملبورن الملكية )قسم الھندسة الكھربائية والحاسوب(‬ ‫‪.3‬‬
‫تخرجوا جميعا )أخرھم تخرج ‪ .(2011‬االن ھو المشرف االول على طالبي دكتوراه و طالب ماجستير في جامعة الكوفة‪.‬‬
‫كما أشرف على ‪ 7‬طالب دكتوراه آخرين كمشرف ثان في جامعة ملبورن الملكية‪.‬‬ ‫‪.4‬‬
‫منح جائزة أفضل باحث في جامعة ملبورن الملكية ثالث مرات‪ :‬عام ‪ 2005‬وعام ‪ 2006‬وباألشتراك مع البروفسور رجارد ھارس‬ ‫‪.5‬‬
‫‪ Prof Richard Harris‬عام ‪ .2004‬كما استلم جائزة أحسن بحث في مؤتمرين عالميين‪.‬‬
‫منح جائزة أفضل مشرف للدراسات العليا في جامعة ملبورن الملكية عام ‪.2007‬‬ ‫‪.6‬‬
‫حصل على منحة األستكشاف من مجلس البحث األسترالي ‪2005‬؛ وأدت إلى ظھور علم المعالجة بالكلمات القصيرة‪.‬‬ ‫‪.7‬‬
‫كان رئيسا ً لقسم ھندسة األتصاالت في جامعة ملبورن الملكية عام ‪.2006‬‬ ‫‪.8‬‬
‫عمل في لجان جامعة ملبورن الملكية العليا كلجنة االكاديميين المتميزين ولجنة تمييز القرارات الجامعية‪.‬‬ ‫‪.9‬‬
‫له أكثر من ‪ 200‬بحث عالمي منھا ‪ 55‬بحثا في المجالت المتقدمة و ‪ 151‬بحثا في المؤتمرات العالمية وستة كتب وفصلين في‬ ‫‪.10‬‬
‫كتابين‪ .‬أكثر ھذه البحوث يمكن تنزيلھا من مواقع ‪ .IEEE, Elsevier, Springer‬أبرز اإلنجازات‪ :‬تصميم منظومة رقمية لمتابعة‬
‫التردد؛ حل جزئي لمشكلة الموجات متعددة المركبات؛ تصميم صنف متطور من تحويالت الزمان‪-‬التردد؛ تصميم خوارزمية لتخمين‬
‫التردد اآلني؛ إعادة التعريف الرياضي للتردد اآلني؛ تصميم نموذج رياضي النتقال الموجات الراديوية )مقارب للحسابات التجريبية‬
‫العالمية(؛ إدخال الدوال الموسعة في تحليل الموجات الالمستقرة؛ ابتكار علم المعالجة بالكلمات القصيرة؛ تصميم دالة ترابط لمعالجة‬
‫الصور؛ تبسيط المنظومات المعقدة مقابل الالخطية؛ ودراسات واسعة في التضمين المتعامد للجيل الرابع لإلتصاالت‪.‬‬
‫كان مسؤول البرامج والعالقات الدولية في جامعة ملبورن الملكية عامي ‪ 2004‬و ‪.2005‬‬ ‫‪.11‬‬
‫وكان رئيس مجموعة ھندسة الحاسوب‪ -‬معالجة االشارة في جامعة ملبورن الملكية ‪.2009-2005‬‬ ‫‪.12‬‬
‫والموجه العلمي لبرامج ھندسة االتصاالت في جامعة ملبورن الملكية ‪.2008 -2005‬‬ ‫‪.13‬‬
‫والموجه العلمي لمشروع اتصاالت الجيل الثالث بين جامعة ملبورن الملكية وشركة أن‪-‬إي – سي اليابانية ‪.2001‬‬ ‫‪.14‬‬
‫حضر أكثر من ‪ 50‬مؤتمراً عالميا ً وكان في اللجان العلمية وترأس عدة جلسات متخصصة في عدة مؤتمرات متقدمة‪.‬‬ ‫‪.15‬‬
‫كان محقق ونائب رئيس مؤتمر تنكون ‪ IEEE TENCON‬والذي استقبل اكثر من الف باحث عالمي عام ‪.2005‬‬ ‫‪.16‬‬
‫وھو عضو متقدم في نقابات المھندسين العالمية مثل ‪ IEEE‬ونقابة الحاسوب اإلسترالية ‪.ACS‬‬ ‫‪.17‬‬
‫ويعمل مقيما لكبريات المجالت العلمية العالمية‪ .‬استلم رسالة تقدير من ‪ IEEE‬كمقيم رصين للبحوث عام ‪.2006‬‬ ‫‪.18‬‬
‫أحرزت برامج الدكتوراه التي أشرف عليھا جوائز ‪ IEEE‬األولى ثالث مرات‪.‬‬ ‫‪.19‬‬
‫جائزة "المقيم الرصين" من ‪ IEEE‬في ‪.2006‬‬ ‫‪.20‬‬
‫جائزة أحسن بحث في مؤتمرين عالميين‪.TENCON 2005, DSPCDC 2006 :‬‬ ‫‪.21‬‬
‫كان ممتحنا خارجيا ألكثر من ‪ 50‬شھادة دكتوراه عالمية‪.‬‬ ‫‪.22‬‬
‫القى عدة ندوات علمية بناء على دعوات من جامعات عالمية‪.‬‬ ‫‪.23‬‬
‫دعي لمنصب بروفسور زائر في عدة جامعات منھا مركز ألكاتيل ‪ - Alcatel‬باريس‪.‬‬ ‫‪.24‬‬
‫منح لقب بروفسور من قبل جامعة باك‪-‬ھا الفيتنامية عام ‪.2008‬‬ ‫‪.25‬‬
‫منح جائزة االساتذة المتميزين علميا والمعروفين عالميا من قبل وزارة التعليم العالي العراقية في ‪ 15‬نيسان ‪.2012‬‬ ‫‪.26‬‬
‫التأثير العالمي لبحوث د‪ .‬زاھر المالكي‪ :‬نشرت مجلة معھد فرانكلين االمريكية ‪ 2011‬بحثـا ً للعالم الھولندي مارتن باستيانز )وھو‬ ‫‪.27‬‬
‫أحد الرموز العلمية منذ الثمانينات( طور فيه صنف التحويالت الزمنية – الترددية المعروفة باسم تحويالت تي )والمصممة من قبل‬
‫الدكتور زاھر في نفس المجلة عام ‪ (2006‬إلى تحويالت تي‪-‬الموسعة القادرة على تحليل الموجات عالية التردد‪ .‬وھناك المئات من‬
‫االقتباسات لھذه البحوث في المجالت العالمية الرائدة )مفصلة على موقع گوگل العلمي؛ منھا أقتباس نشرته جامعة كورنيل االمريكية‬
‫‪ 2013‬لعلماء من جامعة دوك االمريكية ومعھد الفيزياء النظرية بجامعة برلين؛ واقتباس نشرته جامعة أوكسفورد ‪2013‬؛ وأكثر من‬
‫‪ 20‬رسالة دكتوراه في الجامعات العالمية الرائدة‪ ،‬مثال معھد جورجيا للتكنولوجيا‪-‬الواليات المتحدة‪ ،‬وجامعة گرينوبيل الفرنسية(‪ .‬إن‬
‫اقتباس وتطوير أي بحث من قبل رموز العلم )إضافة إلى مكانة الناشر للبحث والناشر لإلقتباس( ھو المقياس العالمي المتبع حاليا ً‬
‫لمعرفة تأثير البحث ومكانة الباحث في االوساط العلمية العالمية‪.‬‬
‫موقع گوگل للباحث العلمي‪http://scholar.google.com/citations?user=DfNL-LwAAAAJ&hl=en :‬‬ ‫‪.28‬‬
‫البريد‪zahir.hussain@uokufa.edu.iq:‬‬ ‫‪.29‬‬

‫‪2 / 124‬‬
 Introduction
Signal Processing (SP) is a subject of central importance in engineering and applied
sciences. Signals are information-bearing functions, and SP includes the analysis and
processing of signals (by dedicated systems) to extract or modify information. Signal
processing is necessary since signals normally contain information that is not readily usable
or understandable; also might be disturbed by unwanted sources like noise. Although many
signals are non-electrical, it is common to convert them into electrical signals for
processing.
Most of natural signals (like the biomedical signals) or signals that are to be understood by
natural means (like music) are continuous functions of time, referred to as analog signals.
Analog Signal Processing (ASP) and analog systems were the only tools to deal with
analog signals in the past. Although ASP and analog systems are still widely used, Digital
Signal Processing (DSP) and digital systems are attracting more attention for the significant
advantages of digital systems over their analog counterparts. These advantages include
superiority in performance, speed, reliability, storage, and reduction in size and cost. In
addition, DSP can solve problems that cannot be solved using ASP, like the spectral
analysis of multicomponent signals, adaptive filtering, and operations at very low
frequencies. Following recent development in engineering in the 80’s and 90’s, it is clear
that DSP is one of the fastest growing industries with far-reaching effects on other
industries that deal with information like economics, meteorology, seismology,
bioengineering, oceanology, communications, astronomy, radar engineering, control
engineering, … etc.
We will cover the representation of analog and digital signals and systems in the time
domain and in the frequency domain. Core topics are convolution, transforms (Fourier,
Laplace, ZT, DTFT, DFT), filters, and random signal analysis. We also consider some
important applications of DSP like signal detection in noise, radar range estimation of
airplanes, binary communication systems, banking and financial applications, and
simulation of audio effects. Design and implementation of digital systems like integrator,
differentiator, resonator, and oscillator will also be considered.

Prerequisites: Basic knowledge in calculus and programming is essential.

Objectives:

The course aims at developing knowledge of analysing signals, natural or synthetic, and
processing them, using the appropriate systems, to reveal the information they convey in
their time and frequency structure. The course provides applications and topics in signal
processing, with MATLAB experiments to know how to implement analog and digital signal
processing systems practically.

References
1. Zahir M. Hussain et al, Digital Signal Processing, Springer, 2011.
2. G. Carlson, Signal and Linear System Analysis, John Wiley & Sons, Inc., 1998.
3. A. Ambardar, Analog and Digital Signal Processing, Brooks/ Cole, 1999.
4. A. V. Oppenheim and R. Schafer, Discrete-Time Signal Processing, Prentice-Hall,
1989.

3 / 124
 Chapter 1: Analog Signals and Systems
1.1 Definitions, Classifications, and Overview

1.1.1 Definitions

A signal is a piece of information (natural or synthetic) explained as a function of time (and

perhaps other variables like the dimensions x, y, etc.). This information can be represented
by the variations in the signal amplitude, phase, or frequency.

A system is a physical or mathematical (i.e., hardware or software) entity that performs

operations on signals to extract or modify information. For example, a low-pass filter is a
system that removes high frequency content from the signal.

1.1.2 Representation of Signals and Systems

It is the mathematical formulation of the signal and the function of the processing system.
This representation is the basis for hardware or software implementation (realization) of
signal processing systems.

1.1.3 Examples of Signals

1. The location of an oscillating pendulum (simple harmonic motion) [Fig.(1.1.1)] .

2. Profit as a function of time & market variables.
3. Trend of a stock price as a function of time.
4. Brain (EEG), heart (ECG), or eye (EOG) signals.
5. Speech, music, or sounds of natural phenomena [Fig.(1.1.2)].
6. Video signals.
7. Atmospheric pressure as a function of time.
8. Unemployment ratio as a function of social, political, and economic factors.
9. Photos sent from a satellite or space station.

Oscillating Pendulum Distance from equilibirium point

Distance

Equilibrium

Equilibrium Time, sec

Fig.(1.1.1): The location of an oscillating pendulum.

4 / 124
 A bird sound A whale sound

Norm. Signal Amplitude

Norm. Signal Amplitude

Time, sec Time, sec

Fig.(1.1.2): Sounds of animals.

1.1.4 Classification of Signals

1. Analog, discrete, and digital:
Analog: defined as continuous-time representation.
Discrete: defined only at discrete time instants.
Digital: discrete, quantized (to specific levels), and constant between adjacent
discrete instants [Fig.(1.1.3)].

Analog signals can be processed only by physical systems

(hardware), while digital signals can be processed by hardware
or software systems. Note that some modern communication
systems are totally based on software, as in SDR (Software-
Defined Radio).

Analog Discrete Digital

Fig.(1.1.3): Analog, discrete, and digital version of a sinusoid.
jω0t
2. Real & complex: e.g. x(t ) =sin(ω0t ) ; x(t ) =e . Complex representation of
signals and systems is a useful mathematical tool to find or analyze information
regarding delay (timing) and phase relationships.

Note that modern transmission of signals over mobiles, internet, and satellite
stations is in complex format.

5 / 124
3. Periodic & non-periodic: For example, the signal x(t ) =cos(ω0t ) +sin(3ω0t ) is
2π −t
periodic with period T = , while x (t ) = e is non-periodic. Electric supply at
0 ω0
home is periodic.

4. Deterministic & random: For example, x(t ) = sin(ω t ) is deterministic, i.e., its
0
exact value is known at any time, while noise n(t ) is random (cannot be
determined exactly as a function of time).

5. Power & energy: Power is defined as the time average of energy: P = E / T . A

Power signal is a signal with finite power over the time interval (−∞, ∞) , i.e.,
1 T 2
P = lim ∫−T x (t ) dt < ∞
T →∞ T
T 2
(Hence, the energy E = ∫−T x (t ) dt should be infinite for power signals).
An Energy signal has finite energy over the time interval (−∞, ∞) , i.e.,
T 2
E = lim ∫−T x (t ) dt < ∞ (Hence, P = 0 for energy signals)
T →∞
An example of power signals: x (t ) = sin(ω 0t ) , with P = 0.5 Watt ( E = ∞) .
−t
An example of energy signals: x (t ) = e , with E = 1 Joule ( P = 0) .

6. Mono-component & Multi-component: This depends on how many distinct

frequencies exist in the signal (when the number of frequency components is finite),
e.g., x (t ) = sin(ω0t ) + cos(5ω0t ) is a 2-component signal; its spectrum (Fourier

transform) is shown in Fig.(1.1.4) for f o = 1 Hz. The two components are

represented by two spikes at the relevant frequencies.
Spectrum, |X ( f )|

0.5

0.4

0.3

0.2

0.1

0
0 1 2 3 4 5 6 7 8 9 10
Frequency, Hz
Fig.(1.1.4): Spectrum of a two-component signal.

6 / 124
1.1.5 Analog and Digital Signal Processing (ASP & DSP)
Most signals in nature or in communication systems are analog. To process those signals
using digital systems, analog-to-digital (A/D) conversion is necessary. After processing,
digital-to-analog (D/A) conversion is applied to obtain the modified analog signal. The
general signal processing system is shown in Fig.(1.1.5).

Digital Digital
Signal Signal
x(t), y(t),
r(n) p(n)
Analog Analog
A/D DSP D/A
Input Output
Signal Signal

Fig.(1.1.5): Block diagram of a generic signal processing system.

1.1.6 DSP versus ASP

1) DSP is less susceptible to noise and power supply disturbances than ASP.
2) DSP is more accurate, esp. in reading the results of signal processing.
3) Storage of digital signals is easier.
4) DSP is more flexible & versatile, esp. in changing the system parameters to handle
changing environments (e.g., in adaptive filtering).

1.1.7 System Modeling

A system can be represented in the time domain as a mathematical operator
(transformation) T on the input signal, as shown in Fig.(1.1.6).

x ( t ), y ( t ),

Input Output
System, T
Signal Signal

y(t)= T[x(t )]

Fig.(1.1.6): Signal processing system as an operator.

1.1.8 Classification of Systems

1. Analog (i.e. continuous-time), discrete, and digital systems (as in the case of
signals).
2. Time-varying (non-stationary) and time-invariant (stationary): if the input x(t ) ,
which gives an output y (t ) , is shifted in time by to , i.e., the new input is x(t − to ) ,
then a time-invariant system will give an output which is the same as y (t ) but time-
shifted by the same amount to , i.e., y (t − t o ) , or:
y(t − t o ) = T [ x(t − to )] , where to is a constant, positive time-shift.

7 / 124
 3. Causal and non-causal systems: the output of a causal system is not dependent on
future values of the input signal x(t ) , i.e., y (t ) is not a function of x(t + to ) .

4. Static (memoryless) and dynamic (with memory) systems: a system whose output
does not depend on a previous value of the input signal x(t ) is called memoryless,
i.e., y (t ) is not a function of x (t − t o ) , where to is a positive time constant. For

example, a squarer is a memoryless system, where y (t ) = [ x(t )]2 .

5. Stable and unstable systems: if a bounded input signal (i.e., | x(t )|< ∞ ) produces a
bounded output signal y (t ) , the system is called bounded-input, bounded-output
(BIBO) stable.

6. Linear and non-linear systems: a system T is called homogeneous if it satisfies

the scaling property:
T [c ⋅ x(t )] = cT [ x(t )] , where c is a constant,
and is called additive if it satisfies the additivity condition:
T [ x (t ) + x (t )] = T [ x (t )] + T [ x (t )] .
1 2 1 2
A linear system satisfies the superposition property, which is the combination of
scaling (homogeneity) and additivity:
T [a ⋅ x (t ) + b ⋅ x (t )] = aT [ x (t )] + bT [ x (t )] , where a and b are constants.
1 2 1 2
Example-1: the system y (t ) = x(t ) + 2 is not linear. This system can be represented
by the operator T such that T [ x(t )] = x(t ) + 2 . Assume that
x(t ) = a ⋅ x (t ) + b ⋅ x (t ) , where a and b are constants. Now we have:
1 2
T [a ⋅ x (t ) + b ⋅ x (t )] = T [ x(t )] = x(t ) + 2 = a ⋅ x (t ) + b ⋅ x (t ) + 2
1 2 1 2
≠ aT [ x (t )] + bT [ x (t )] = a[ x (t ) + 2] + b[ x (t ) + 2] = ax (t ) + bx (t ) + 2a + 2b .
1 2 1 2 1 2
Hence, a system with independent internal sources is not linear.
Example-2: The system y (t ) = ln[ x(t )] is non-linear since ln[c ⋅ x(t )] ≠ c ln[ x(t )] .
Example-3: The system y (t ) = dx(t ) / dt is linear.

1.1.9 Linear Time-Invariant (LTI) Systems

This kind of systems is of fundamental importance in practical applications. It combines

linearity and time-invariance properties described above. It can be characterized by
differential equations with constant coefficients. LTI systems can be represented in the time
domain by using one of the following approaches:
1. Differential equations
2. State-space techniques
3. Impulse response.
In this course we will consider the impulse response approach only as a time-domain
representation for LTI systems.

8 / 124
1.2 Time-Domain/ Frequency-Domain Representation of Analog
Signals and LTI Systems
There are two approaches to analyze signals and systems: the time domain approach and
the frequency domain approach. The two domains are equivalent, connected by a
suitable transformation (like Fourier or Laplace transforms).

First we consider analog signals and systems. There are similar representations and
relationships for discrete and digital signals and systems that will be handled later.

1.2.1 Basic Functions and Relations

1.2.1.1 The Convolution Integral
The convolution between two functions h(t ) and x(t ) , denoted by h(t ) * x(t ) , is given by:
∞
y (t ) = h (t ) * x (t ) = ∫−∞ h(λ ) x (t − λ )dλ .
It includes integration of the product of the first function with a shifted and reflected
version of the second function. We will see later that the I/O function of any LTI system
can be described as a convolution.

Properties of the Convolution Integral:

1) Commutative: h(t ) * x (t ) = x (t ) * h (t )
2) Associative: h(t ) * [ x (t ) * v (t )] = [ h(t ) * x (t )] * v (t )
3) Distributive: h(t ) * [ x (t ) + v (t )] = h(t ) * x (t ) + h(t ) * v (t ) .

The above properties are important in finding the behavior of serial and/or parallel
combinations of LTI systems.

1.2.1.2 The Delta Function

The delta function, denoted in the time domain by δ (t ) , is of fundamental importance in

signal analysis. It is a generalized function, not an ordinary mathematical function, and
rigorous study of this function is complicated. It can be defined in conjunction with a
continuous function x(t ) by the integral:
∞
∫−∞ x (t )δ (t − t 0 ) dt = x (t 0 ) , to is constant.
Roughly speaking in practical language, δ (t ) is an even, tall, and narrow spike of
infinite height with zero width concentrated at t = 0 . Hence, δ (t − t ) is concentrated at
0
t = t , i.e., a loose engineering definition can be given by:
0
 0, t ≠ to
δ (t − t o ) = 
∞, t = to .
In applications we normally represent δ (t ) by an arrow of unit length, while a scaled
delta function aδ (t ) is represented by an arrow of height = a [see Fig.(1.2.1)].

9 / 124
 δ (t+3) δ (t)
1.0
0.5 δ (t-3)
0.5

Time, t
-3 0 1 2 3

- δ (t-2)

Fig.(1.2.1): Representation of the delta function.

Properties of the Delta Function: The delta function has the following properties:
∞
P1: ∫−∞ δ (t ) dt = 1 (unit area), or generally,
b 1 a < to < b
∫ δ (t − t o ) dt =
a 0 otherwise
P2: δ (t ) = δ ( −t ) (even).
P3: x (t ) * δ (t ) = x (t ) , or, generally, x (t ) * δ (t − t o ) = x (t − t o ) , where to is a constant.

Alternative Representations of the Delta Function

The Dirac delta function can also be defined as the limit of several even functions that
can satisfy the above properties in the limit. These definitions include:

1. Limit of the weighted rectangular pulse (box), Π 2 a (t ) [see Fig.(1.2.2) (left)]:

1 1 1 | t |≤ a 
δ (t ) = lim Π 2 a (t ) = lim ( a→0)  
a→0 2a 2 a 0 | t |> 0 
2. Limit of the weighted absolutely-decaying exponential:
1 −|x|/ a
δ ( x ) = lim ( a→0) e
2a
3. Limit of the weighted triangular pulse Λ (t ) [see Fig.(1.2.2) (right)]:
2a
1 
|t|
1 1 − , | t |≤ a 
δ (t ) = lim ( a→0) Λ 2a (t ) = lim ( a→0)  a 
a a | t |> 0 
 0 

10 / 124
 [1/(2a)] Π2a ( t ) (1/a) Λ 2a ( t )

1/(2a) 1/a

-a a t -a a t

Fig.(1.2.2): Weighted rectangular and triangular pulses.

1.2.1.3 The Unit Step Function and its Relationship to the Impulse Function
The unit step function u (t ) is defined as:
1 t > 0
u (t ) = 
0 t < 0
The unit step function has a discontinuity at t = 0 , where its value is undefined. If u (0)
is chosen as 0.5, u (t ) is called the Heaviside unit step. The above definition is equivalent
to the following integral relation:
t
u (t ) = ∫ δ (t ) dt .
−∞
Hence, we have the relation δ (t ) = du (t ) / dt .

1.2.2 Time-Domain Representation of Analog Signals and LTI Systems

In this section we study analog signals and systems in the time domain.

1.2.1.1 Mathematical Representation of Signals and Systems in the Time Domain

An analog signal is represented in the time domain as a function of time, while an LTI
analog system is fully identified (represented) in the time domain by its impulse
response (i.e., its output when the input is the Dirac delta function, δ (t ) ). Normally the
impulse response is denoted by h(t ) , as shown in Fig.(1.2.3).

LTI Analog System

x ( t ), y ( t ),
Input Output
h(t)
Signal Signal

y(t)= h(t)*x(t)

Fig.(1.2.3): Time-domain representation of an LTI system.

11 / 124
It can be shown that the system function or the input/ output (I/O) relationship is
described by the convolution of the impulse response h(t ) and the input signal x (t ) as
follows:
∞
y (t ) = h (t ) * x (t ) = ∫−∞ h(λ ) x (t − λ )dλ
Note: for causal systems h(t ) = 0 for t < 0 , otherwise instantaneous anticipating
quantities like h( −τ ) x (t + τ ) , τ > 0 , will appear in the above integral when λ = −τ < 0 ,
which implies that the system needs future values of the input.

1.2.2.2 Stability of Analog LTI Systems in the Time domain

An analog system is BIBO stable if its impulse response is absolutely summable, i.e.,

∞
∫−∞ h (t ) dt < ∞ .

This time domain approach to study the stability of systems is normally difficult for
complicated systems. We will see that the frequency domain approach using Laplace
transform is more practical.

Example 1: Consider the system described by the impulse response

e −at t ≥ 0
h (t ) = 
 0 t<0
where a is a positive constant. This system is causal (since h(t ) = 0 for t < 0 ) and stable
∞
since ∫−∞ h(t ) dt = 1 / a < ∞ .

This system can represent a capacitor-resitor circuit.

Q. Find the output of the above system when the input is x(t ) = cos(t ).

−|t |
Example 2: The system h(t ) = e is non-causal since h(t ) ≠ 0 for t < 0 , but it is
∞
stable since ∫−∞ h (t ) dt = 2 < ∞ .

1.2.3 Frequency-Domain Representation of Analog Signals and Systems

In this section we study analog signals and systems in the frequency domain, which is
equivalent to the time domain by help of suitable transformations. The frequency domain
can reveal further characteristics of signals and systems and provide better tools to analyze
and design signal processing systems. For example, it is difficult to study the system
stability in the time domain, especially for complicated systems.

Transformations should be unitary, i.e., it is invertible and preserves the four arithmetic
operations (in the general sense).

12 / 124
The most important transformations in applied sciences are:
1. Fourier Transform: which is a transformation from the time domain to the frequency
domain.
2. Laplace Transform: which is a transformation from the time domain to the
generalized (complex) frequency domain.

Both Fourier and Laplace transforms are unitary, as they are invertible and preserve
operations in both domains.

Fourier transform of a signal is normally referred to as the spectrum of the signal.

Fourier series is a specific transform that is used only for periodic signals.

The transforms above can be used to analyze all signals and systems under some basic
conditions that are met in practical applications.

1.2.3.1 Fourier Series Representation of Periodic Signals

A periodic time signal x (t ) with period To repeats its values every To seconds [see
Fig.(1.2.4)]. Fourier series (FS) is the decomposition of a periodic time signal into a linear
combination of sine and cosine functions. The frequencies of these functions are multiples
of the fundamental frequency of the signal, f o = 1 / To . This is called trigonometric
Fourier series.

Using Euler’s formula: e

± jθ
= cosθ ± j sin θ , we can obtain the “exponential” or
“complex” Fourier series from the trigonometric series. (Euler’s formula can be proved
using Taylor series expansion of e
jθ
, cos(θ ) , and sin(θ ) around θ = 0 . It represents a
connection between algebra and geometry).

Note that although Fourier series reveals the frequency content of the signal, it is not
exactly a frequency transform as the representation is still in the time domain.
s (t)
To = 1 / fo
Amplitude

Time
t

Fig.(1.2.4): A periodic signal s(t).

13 / 124
The Trigonometric Fourier Series
If x(t ) is a periodic signal with period To , then we can expand it as follows:
x (t ) = a0 + a1 cos(ω0t ) + a2 cos(2ω0t ) + ⋯
+ b1 sin(ω0t ) + b2 sin(2ω0t ) + ⋯
∞
[
= a0 + ∑ a n cos(nω0t ) + bn sin(nω0t )
n=1
]
2π
where ω0 = 2πf 0 = , and:
Τ0
1 Τ
a =
0 Τ ∫0 0
x(t )dt (the constant term).
0
2 Τ
an = 0 x(t ) cos(nω0t )dt ,
Τ ∫0
0
2 Τ
bn = 0 x(t ) sin(nω0t )dt .
Τ ∫0
0
Special Cases:
1. If x (t ) is odd, then x (t ) cos(nω t ) is odd, hence ao = a n = 0 , and the Fourier
0
series would be a series of sines without a constant term.
2. If x (t ) is even, then x (t ) sin( nω0t ) is odd, hence bn = 0 , and the Fourier series
would be a series of cosines.

Example: Consider the signals x (t ) and s (t ) in Fig.(1.2.5). The signal x (t ) can be odd if
it is shifted down by ½, hence we can use results related to odd functions. The signal s(t )
is even. The fundamental period of both signals is To = 2 . Fourier series of these two
signals are:

x(t) s(t)

1
1

-1 0 1 2 3 4
t 0 1.5 3 4.5
t
Fig.(1.2.5): Two square waves.

1 2 1 1
x (t ) = [sin(ω0t ) + sin(3ω0t ) + sin(5ω0t ) + ...] and
+
2 π 3 5
1 2 1 1 2π
s(t ) = + [cos(ω t ) − cos(3ω t ) + cos(5ω t ) + ...] , where ω0 = =π.
2 π 0 3 0 5 0 Τ0

14 / 124
The Exponential (Complex) Fourier Series

Using Euler formula, we can write the trigonometric Fourier series in the exponential form:
∞ + jnω0t ∞ + jn 2πf 0t
x (t ) = ∑ X ne = ∑ X ne
n=−∞ n=−∞
1 Τ − jnω t 1 Τ − jn2πf t
Xn = 0 = 0 dt , n = 0,1,2,⋯
Τ ∫0 Τ ∫0
0 x (t ) e dt 0 x (t ) e
0 0
This pair is similar in structure to the Fourier transform (FT) pair that will be discussed
1
later. In fact, it is easy to see that X n = X1 p ( f ) f =nf , where X1 p ( f ) is the FT of
To o
one period of x(t ) , hence, X1 p ( f ) / To is the envelope of FS coefficients.

Relationship Between Complex and Trigonometric Series Coefficients:

1 1
X o = ao , X n = 2 ( an − jbn ), and X −n = 2 ( a n + jbn ).

Q. Verify the above relations using Euler’s formula.

1.2.3.2 The Fourier Transform

Fourier Series (FS) applies only to periodic signals, with period To and fundamental
frequency f 0 = 1 / To . As To → ∞ , the signal becomes non-periodic and its FS will tend
to the Fourier transform (FT), which is normally defined as Fourier transform pair since the
time signal can be obtained by the inverse transformation F −1 :
∞ − j 2πft
X ( f ) = F { x (t )} = ∫∞ x (t )e dt …………. (1)
∞ X ( f )e + j 2πft df
x(t ) = F −1{ X ( f )} = ∫∞ …………. (2)

The Fourier transform (FT) reveals the frequency content of the signal, known as the
spectrum, and the frequency behavior of the system, known as the transfer function or
frequency response. The Fourier transform X ( f ) of the real time signal x(t ) is generally
complex. It is normally plotted as magnitude X ( f ) vs. frequency f (magnitude
spectrum) and phase ∠ X ( f ) vs. frequency f (phase spectrum). For systems, these

quantities are called the magnitude response and the phase response.

By help of Dirac delta function δ (t ) , Fourier transform can be applied to periodic signals
as well to obtain their frequency spectra. Note that Fourier series is not exactly a
transformation into the frequency domain.

15 / 124
 −2t
Example 1: Consider the signal x (t ) = e u (t ) [a decaying exponential]. Its FT is:
∞ −2t − j 2πft ∞ −( 2+ j 2πf ) 1
X ( f ) = ∫0 e .e dt = ∫0 e dt =
2 + j 2πf
1
Amplitude spectrum = X ( f ) = ,
2 2
4 + 4π f
−1 2πf
phase spectrum = ∠ X ( f ) = − tan ( ).
2
On MATLAB, we can use the FFT algorithm to compute the FT approximately, and
“phase” or “angle” statement to find the phase spectrum. See Fig.(1.2.6).

2
1 x(t) |X(f)|

Phase (rad)
Magnitude

1
Amplitude

0.4
0
0.5
0.2
-1

0 0 -2
-5 0 5 -5 0 5 -10 -5 0 5 10
Time, sec Frequency, Hz Frequency, Hz

Fig.(1.2.6): A decaying exponential signal with its magnitude and phase spectra.

Example 2: Consider the rectangular time pulse x(t ) = Π (t ) . Its FT is obtained as:
Τ
Τ / 2 −2πft
X(f ) = ∫ e
−Τ / 2
dt =
1
− j 2πf
e
− j 2πft Τ / 2 =
−Τ / 2
1
− j 2πf
[e
− jπfΤ
−e ]jπfΤ
[ ]
jπfΤ − jπfΤ
1 e −e sin(πfT ) sin(πfT )
= = (Euler’s) = T = Tsinc( fT ) .
πf 2j πf πfT
The signal with its FT and magnitude spectrum is shown in Fig.(1.2.7).
x(t) X(f) |X(f)|
1 1 1
1/T

2/T

0 0 0
t f f
1/T 3/T
-T/2 T/2
2/T

Fig.(1.2.7): A rectangular time pulse with its spectra.

16 / 124
Example 3: If x (t ) = δ (t ) , then its spectrum is given by:
∞ − j 2πft
X ( f ) = ∫−∞ δ (t )e dt = g (0) = 1 [From definition of the delta function] [Fig.(1.2.8)].
g (t )
If the time duration of a signal is narrow, its frequency duration is wide, and vise
versa.
x(t) = δ ( t ) X(f) = 1
← FT →
1

t, sec f, Hz
0 0

Fig.(1.2.8): The time-domain delta function and its spectrum.

Some Properties of FT: See Tables. Students should prove these properties.
1) Duality of FT:
F
if x (t ) ←→ X ( f ) is a FT pair, then:
F
X (t ) ←→ x ( − f ) is a FT pair.
Note that time and frequency variable are exchanged. If x (t ) is even, then:
F
X (t ) ←→ x ( f )
That is, time and frequency functions can be swapped.

F
Example 1: We proved Π Τ (t ) ←→ Tsinc( fT )
F
By duality we have: Bsinc( Bt ) ←→ Π B ( f ) , i.e., a time-domain sinc function
will be transformed into a frequency rectangular pulse.

Example 2: We proved F →1 . Then, by duality we have: 1←

δ (t ) ← F →δ ( f ) .

2) Time Shift: x(t − to ) ←→ X ( f

F
)e − j 2πft o
3) Frequency Shift:
F X ( f ) , then j 2πf t F
If x(t ) ←→
 0 ←→
 X( f − f ).
x(t )e
0
F → δ ( f ) [as shown above], then e j 2πf 0t ←
Example: Since 1← F →δ ( f − f ) ,
0
i.e., FT {complex exponential} is a frequency-shifted delta function.

17 / 124
Fourier Transforms of Sinusoids

jθ
From Euler’s formula: e = cos θ + j sin θ ............(1)
− jθ
= cos θ − j sin θ ............( 2)
e
1 jθ − jθ 1 jθ − jθ
∴ cos θ = (e +e ) ; sin θ = (e −e ).
2 2j
1 j 2πf 0t − j 2πf 0t
∴ cos(ω0t ) = cos(2πf 0t ) = [e +e ]
2
1 1
∴ FT {cos(2πf 0t )} = δ ( f − f 0 ) + δ ( f + f 0 ) , using the previous Example.
2 2
1
Similarly, FT {sin( 2πf t )} = [δ ( f − f ) − δ ( f + f )] . Hence, the magnitude
0 0 0
2j
spectra of sin(ωot ) and cos(ωo t ) are identical, as shown in Fig.(1.2.9).

Time signals Magnitude Spectra

1
1
sin(ω t), cos (ω t)

0.5 δ (f + fo) 0.5 δ (f - fo)

o

0.5
0.5

0 t, sec 0 f, Hz
-2 f =2
o
o

-0.5
-0.5

-1
-1
-1 -0.5 0 0.5 1 -5 0 5

Fig.(1.2.9): Sine and cosine (with same fo) have identical magnitude spectra.

Fourier Transform of Periodic Signals:

A periodic signal x(t ) can be represented by Fourier Series:

∞ + jn2πf t
x(t ) = ∑ Xke 0 ,
k =−∞
where { X } are the FS coefficients. Taking Fourier transform of both sides we have:
k
 ∞ + jk 2πf t  ∞
X ( f ) = F  ∑ X e 0  = ∑ X F e + jk 2πf 0t . = ∞ ∑ X k δ ( f − kf 0 ).
k k
k =−∞ 
 k = −∞   k =−∞
Hence, FT of a periodic signal x(t ) with period To is a sum of frequency impulses at
integer multiples of the fundamental frequency f o , that is, at f = kf o , weighted by FS
coefficients X .
k

18 / 124
Example: The complex Fourier series and Fourier transform of the square wave shown in
Fig.(1.2.10) are given by (Prove!):
1
∞ 1 k j πkt k
x(t ) = ∑ [ sinc( )]e 2 and X ( f ) = ∑ X k δ ( f − ) . Here T = 4
2 2 4 o
k = −∞ 
Xk
sec, f o = ¼ Hz. The envelope of X in the frequency domain is obtained by substituting
k
f for kf = k / 4 , hence (1/2) sinc(k / 2) becomes (1 / 2)sinc(2 f ) .
o
X(f)
x(t) 0.5 δ (f)

0.3183 δ (f+1/4) 0.3183 δ (f-1/4)

-1 0 1 2 5 t , sec f , Hz
-0.1061 δ (f-3/4)

Fig.(1.2.10): A square wave and its Fourier transform.

The above signal is useful in many applications. Its general form is:
∞
x (t ) = ∑ AΠT (t − nT ) , its Fourier coefficients are X = ( AT / T ) sinc(kf T ) ,
n = −∞ o k o o
where T is the duration of the “ON” state. The envelope of these coefficients is given by
E ( f ) = ( AT / T ) sinc( fT ) = X ( f ) / To , where X1 p ( f ) = FT {Π T (t )}= FT of one
o 1p
period. This signal can be simulated on MATLAB as follows: r=T/To*100; fo=1/To;
wo=2*pi*fo; x=A*0.5*(1+square(wo*(t+T/2),r));

1.2.3.3 The Laplace Transform

It is a generalization of Fourier transform that takes into account the amplitude variations
of the time signal in addition to the frequency content which is expressed in FT as a linear
jω t
combination of complex exponentials e . Hence, the Laplace transform (LT) gives
more insight for stability of systems. In addition, LT can deal with signals that have no FT,
n
like t u (t ) . Hence LT is the main tool in representing analog (continuous-time) feedback
systems where stability is of extreme importance. There are two definitions of LT as
explained below.

19 / 124