Signal Processing

Chapter 2 – Discrete-Time Signals and Systems

Adriano Vilela Barbosa

UFMG – Universidade Federal de Minas Gerais

DELT – Department of Electronics Engineering
CEFALA – Center for Research on Speech,
Acoustics, Language and Music

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 1 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

Contents

1 Discrete-Time Signals: Sequences

2 Discrete-Time Systems
3 Linear Time-Invariant Systems
4 Properties of Linear Time-Invariant Systems
5 Linear Constant-Coefficient Difference Equations
6 Frequency-Domain Representation of Discrete-Time Signals and
Systems
7 Representation of Sequences by Fourier Transform
8 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 2 / 90

 Introduction

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 3 / 90

Introduction

Signal – something that conveys information.

Signals are represented mathematically as functions of one or
more independent variables (e.g., speech is a function of time;
images are functions of two spatial variables).
The independent variable (usually time) can be either continuous
or discrete.
The dependent variable (amplitude) can also be either continuous
or discrete.
Digital signals are those for which both time and amplitude are
discrete.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 4 / 90

Introduction

Signal – something that conveys information.

Signals are represented mathematically as functions of one or
more independent variables (e.g., speech is a function of time;
images are functions of two spatial variables).
The independent variable (usually time) can be either continuous
or discrete.
The dependent variable (amplitude) can also be either continuous
or discrete.
Digital signals are those for which both time and amplitude are
discrete.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 4 / 90

Introduction

Signal – something that conveys information.

Signals are represented mathematically as functions of one or
more independent variables (e.g., speech is a function of time;
images are functions of two spatial variables).
The independent variable (usually time) can be either continuous
or discrete.
The dependent variable (amplitude) can also be either continuous
or discrete.
Digital signals are those for which both time and amplitude are
discrete.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 4 / 90

Introduction

Signal – something that conveys information.

Signals are represented mathematically as functions of one or
more independent variables (e.g., speech is a function of time;
images are functions of two spatial variables).
The independent variable (usually time) can be either continuous
or discrete.
The dependent variable (amplitude) can also be either continuous
or discrete.
Digital signals are those for which both time and amplitude are
discrete.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 4 / 90

Introduction

Signal – something that conveys information.

Signals are represented mathematically as functions of one or
more independent variables (e.g., speech is a function of time;
images are functions of two spatial variables).
The independent variable (usually time) can be either continuous
or discrete.
The dependent variable (amplitude) can also be either continuous
or discrete.
Digital signals are those for which both time and amplitude are
discrete.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 4 / 90

 Discrete-Time Signals: Sequences

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 5 / 90

Discrete-Time Signals: Sequences

Tempo
Contı́nuo Discreto
Amp.
Contı́nua Analógico Discreto
Discreta Quantizado Digital

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 6 / 90

Discrete-Time Signals: Sequences

Discrete-time signals are represented mathematically as sequences

of numbers.
Formally written as x = {x[n]}, −∞ < n < ∞, n integer.
In practice, such sequences usually arise from periodic sampling
of analog signals. In this case
x[n] = xa (nT), −∞ < n < ∞,
where xa (t) is the original analog signal. The quantity T is called
the sampling period and its reciprocal 1/T is the sampling fre-
quency.
Strictly speaking, x[n] denotes the nth sample of the sequence.
However, for convenience, we usually use x[n] to refer to the en-
tire sequence.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 7 / 90

Discrete-Time Signals: Sequences

Discrete-time signals are represented mathematically as sequences

of numbers.
Formally written as x = {x[n]}, −∞ < n < ∞, n integer.
In practice, such sequences usually arise from periodic sampling
of analog signals. In this case
x[n] = xa (nT), −∞ < n < ∞,
where xa (t) is the original analog signal. The quantity T is called
the sampling period and its reciprocal 1/T is the sampling fre-
quency.
Strictly speaking, x[n] denotes the nth sample of the sequence.
However, for convenience, we usually use x[n] to refer to the en-
tire sequence.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 7 / 90

Discrete-Time Signals: Sequences

Discrete-time signals are represented mathematically as sequences

of numbers.
Formally written as x = {x[n]}, −∞ < n < ∞, n integer.
In practice, such sequences usually arise from periodic sampling
of analog signals. In this case
x[n] = xa (nT), −∞ < n < ∞,
where xa (t) is the original analog signal. The quantity T is called
the sampling period and its reciprocal 1/T is the sampling fre-
quency.
Strictly speaking, x[n] denotes the nth sample of the sequence.
However, for convenience, we usually use x[n] to refer to the en-
tire sequence.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 7 / 90

Discrete-Time Signals: Sequences

Discrete-time signals are represented mathematically as sequences

of numbers.
Formally written as x = {x[n]}, −∞ < n < ∞, n integer.
In practice, such sequences usually arise from periodic sampling
of analog signals. In this case
x[n] = xa (nT), −∞ < n < ∞,
where xa (t) is the original analog signal. The quantity T is called
the sampling period and its reciprocal 1/T is the sampling fre-
quency.
Strictly speaking, x[n] denotes the nth sample of the sequence.
However, for convenience, we usually use x[n] to refer to the en-
tire sequence.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 7 / 90

Discrete-Time Signals: Graphical Representation

x[–1] x[0]
x[–2] x[1] x[n]
x [2]

7 8 9 10 11
–9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 n

Figure 2.1 Graphical representation of a discrete-time signal.

The sequence x[n] is only defined for integer values of n.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 8 / 90

Discrete-Time Signals: Graphical Representation
x[n] = xa (nT) T : sampling period

32 ms
(a)

256 samples
(b)
Figure 2.2 (a) Segment of a continous-time speech signal. (b) Sequence
of samples obtained from part (a) with T = 125µs.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 9 / 90

Basic Sequences

0, n 6= 0 1
δ[n] = Unit sample
... 1, n=0 ...

0 n
(a)

 Unit step
0, n<0 1
u[n] =
1, n≥0 ...
...

0 n
(b)
Figure 2.3 Some basic sequences.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 10 / 90

Basic Sequences

Real exponential
x[n] = Aαn ...

...
0 n
(c)

Sinusoidal
x[n] = A cos(ω0 n + φ)
...

0 n
...

(d)
Figure 2.3 (cont.) Some basic sequences.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 11 / 90
Basic Relations

Any sequence can be expressed as

∞
X
x[n] = x[k]δ[n − k]
k=−∞

p[n] = a−3 δ[n + 3] + a1 δ[n − 1] + a2 δ[n − 2] + a7 δ[n − 7]

a1
a–3
p[n]
2 7
–4 –2 0 1 3 4 5 6 8 n
a2 a7
Figure 2.4 Example of a sequence represented as a sum of
scaled, delayed impulses.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 12 / 90

Basic Relations

n
X
u[n] = δ[k]
k=−∞

δ[n] = u[n] − u[n − 1]

∞
X
u[n] = δ[n − k]
k=0

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 13 / 90

Complex Exponential

x[n] = Aαn , where A = |A| ejφ , α = |α| ejω0

x[n] = |A|ejφ |α|n ejω0 n

= |A||α|n ej(ω0 n+φ)

= |A||α|n cos(ω0 n + φ) + j|A||α|n sin(ω0 n + φ)
|α| > 1: oscillates with exponentially growing envelope
|α| < 1: oscillates with exponentially decaying envelope
|α| = 1: oscillates with constant amplitude → complex exponen-
tial sequence. In this case, by analogy with the continuous-time
case, ω0 is called the frequency of the complex exponential (in ra-
dians per sample), and φ is called the phase (in radians).

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 14 / 90

Complex Exponential: Important Differences

Discrete-time complex exponentials have two important properties

that distinguish them from their continous-time counterparts
Property 1: Discrete-time complex exponentials are periodic in
frequency

x[n] = Aej(ω0 +2π)n

= Aejω0 n ej2πn = Aejω0 n

The complex exponentials at frequencies ω0 and ω0 + 2πr, where
r is an integer, are identical
For these sequences, we only need to consider frequencies in an
interval of length 2π , such as −π < ω0 ≤ π or 0 ≤ ω0 < 2π

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 15 / 90

Complex Exponential: Important Differences

Discrete-time complex exponentials have two important properties

that distinguish them from their continous-time counterparts
Property 1: Discrete-time complex exponentials are periodic in
frequency

x[n] = Aej(ω0 +2π)n

= Aejω0 n ej2πn = Aejω0 n

The complex exponentials at frequencies ω0 and ω0 + 2πr, where
r is an integer, are identical
For these sequences, we only need to consider frequencies in an
interval of length 2π , such as −π < ω0 ≤ π or 0 ≤ ω0 < 2π

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 15 / 90

Complex Exponential: Important Differences

Discrete-time complex exponentials have two important properties

that distinguish them from their continous-time counterparts
Property 1: Discrete-time complex exponentials are periodic in
frequency

x[n] = Aej(ω0 +2π)n

= Aejω0 n ej2πn = Aejω0 n

The complex exponentials at frequencies ω0 and ω0 + 2πr, where
r is an integer, are identical
For these sequences, we only need to consider frequencies in an
interval of length 2π , such as −π < ω0 ≤ π or 0 ≤ ω0 < 2π

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 15 / 90

Complex Exponential: Important Differences

Discrete-time complex exponentials have two important properties

that distinguish them from their continous-time counterparts
Property 1: Discrete-time complex exponentials are periodic in
frequency

x[n] = Aej(ω0 +2π)n

= Aejω0 n ej2πn = Aejω0 n

The complex exponentials at frequencies ω0 and ω0 + 2πr, where
r is an integer, are identical
For these sequences, we only need to consider frequencies in an
interval of length 2π , such as −π < ω0 ≤ π or 0 ≤ ω0 < 2π

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 15 / 90

Complex Exponential
Property 2: Discrete-time complex exponentials are not necces-
sarily periodic in time
Condition for periodicity in time

x[n] = x[n + N], for all n, and where N an integer

As an example, consider the cosine function

A cos(ω0 n + φ) = A cos(ω0 n + ω0 N + φ)
Condition for periodicity: ω0 N = 2πk, where k is an integer

2π N
=
ω0 k
Thus, complex exponentials are periodic in time only if 2π/ω0 is a
ratio of two integers
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 16 / 90
Complex Exponential
Property 2: Discrete-time complex exponentials are not necces-
sarily periodic in time
Condition for periodicity in time

x[n] = x[n + N], for all n, and where N an integer

As an example, consider the cosine function

A cos(ω0 n + φ) = A cos(ω0 n + ω0 N + φ)
Condition for periodicity: ω0 N = 2πk, where k is an integer

2π N
=
ω0 k
Thus, complex exponentials are periodic in time only if 2π/ω0 is a
ratio of two integers
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 16 / 90
Complex Exponential
Property 2: Discrete-time complex exponentials are not necces-
sarily periodic in time
Condition for periodicity in time

x[n] = x[n + N], for all n, and where N an integer

As an example, consider the cosine function

A cos(ω0 n + φ) = A cos(ω0 n + ω0 N + φ)
Condition for periodicity: ω0 N = 2πk, where k is an integer

2π N
=
ω0 k
Thus, complex exponentials are periodic in time only if 2π/ω0 is a
ratio of two integers
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 16 / 90
Complex Exponential
Property 2: Discrete-time complex exponentials are not necces-
sarily periodic in time
Condition for periodicity in time

x[n] = x[n + N], for all n, and where N an integer

As an example, consider the cosine function

A cos(ω0 n + φ) = A cos(ω0 n + ω0 N + φ)
Condition for periodicity: ω0 N = 2πk, where k is an integer

2π N
=
ω0 k
Thus, complex exponentials are periodic in time only if 2π/ω0 is a
ratio of two integers
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 16 / 90
Complex Exponential
Property 2: Discrete-time complex exponentials are not necces-
sarily periodic in time
Condition for periodicity in time

x[n] = x[n + N], for all n, and where N an integer

As an example, consider the cosine function

A cos(ω0 n + φ) = A cos(ω0 n + ω0 N + φ)
Condition for periodicity: ω0 N = 2πk, where k is an integer

2π N
=
ω0 k
Thus, complex exponentials are periodic in time only if 2π/ω0 is a
ratio of two integers
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 16 / 90
Complex Exponential
By combining properties 1 and 2, it becomes clear that there are
N distinguishable frequencies for which the corresponding se-
quences are periodic with period N . One set of frequencies is
ωk = 2πk/N, k = 0, 1, . . . , N − 1.
High and low frequencies: rate of oscillation does not increase
continuously with ω0 . Values around ω0 = 2πk are referred to as
low frequencies (slow oscillations), whereas values around ω0 =
(π + 2πk) are referred to as high frequencies (rapid oscillations).

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 17 / 90

Complex Exponential

v 0 = 0 or v 0 = 2p

... ...

0 n
(a)

v 0 = p/8 or v 0 = 15p/8

... ...

0 n

(b)
Figure 2.5 cos(ω0 n) for different values of ω0 . Oscillations get faster
as ω0 increases from 0 to π (parts a–d) and slower as ω0 increases
from π to 2π (parts d–a).

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 18 / 90

Complex Exponential
v 0 = p/4 or v 0 = 7p/4

... ...

0 n

(c)

v0 = p
... ...

0 n
... ...

(d)
Figure 2.5 (cont.) cos(ω0 n) for different values of ω0 . Oscillations
get faster as ω0 increases from 0 to π (parts a–d) and slower as ω0
increases from π to 2π (parts d–a).
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 19 / 90
 Discrete-Time Systems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 20 / 90

Discrete-Time Systems

A discrete-time system is a transformation or operator that maps

an input sequence with values x[n] into an output sequence with
values y[n].
y[n] = T {x[n]}

T { •}
x [n] y[n]

Figure 2.6 Representation of a discrete-time system.

Example 2.3 – The Ideal Delay System

y[n] = x[n − nd ], −∞ < n < ∞,
where nd is a fixed positive integer. Question: What about nega-
tive nd ?

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 21 / 90

Example 2.4 – Moving Average

M2
1 X
y[n] = x[n − k]
M1 + M2 + 1 k=−M
1

x[k]

n–5
0 n k

Figure 2.7 Sequence values involved in computing a causal moving average

(n = 7, M1 = 0, M2 = 5).

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 22 / 90

Memoryless Systems

A system is referred to as memoryless if the output y[n] at every

value of n depends only on the input x[n] at the same value of n.
2
y[n] = (x[n]) , for each value of n
Ideal delay system: memoryless if and only if nd = 0.
Moving average system: memoryless if and only if M1 = M2 = 0.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 23 / 90

Memoryless Systems

A system is referred to as memoryless if the output y[n] at every

value of n depends only on the input x[n] at the same value of n.
2
y[n] = (x[n]) , for each value of n
Ideal delay system: memoryless if and only if nd = 0.
Moving average system: memoryless if and only if M1 = M2 = 0.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 23 / 90

Memoryless Systems

A system is referred to as memoryless if the output y[n] at every

value of n depends only on the input x[n] at the same value of n.
2
y[n] = (x[n]) , for each value of n
Ideal delay system: memoryless if and only if nd = 0.
Moving average system: memoryless if and only if M1 = M2 = 0.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 23 / 90

Linear Systems
Principle of superposition:

x1 [n] → y1 [n], x2 [n] → y2 [n]

T{ax1 [n] + bx2 [n]} = aT{x1 [n]} + bT{x2 [n]}

= ay1 [n] + by2 [n]

This can be generalized to the superposition of many inputs.

Specifically, if X
x[n] = ak xk [n] ,
k

then the output of a linear system will be

X
y[n] = ak yk [n] ,
k

where yk [n] is the system’s response to the input xk [n].

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 24 / 90
Linear Systems
Principle of superposition:

x1 [n] → y1 [n], x2 [n] → y2 [n]

T{ax1 [n] + bx2 [n]} = aT{x1 [n]} + bT{x2 [n]}

= ay1 [n] + by2 [n]

This can be generalized to the superposition of many inputs.

Specifically, if X
x[n] = ak xk [n] ,
k

then the output of a linear system will be

X
y[n] = ak yk [n] ,
k

where yk [n] is the system’s response to the input xk [n].

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 24 / 90
Example 2.6 – The Accumulator System

The Accumulator System is defined by

n
X
y[n] = x[k]
k=−∞

The accumulator system is a linear system. In order to prove this,

we must show that it satisfies the superposition principle for all
inputs. Let us define two arbitrary inputs:
n
X
y1 [n] = x1 [k] ,
k=−∞

n
X
y2 [n] = x2 [k] .
k=−∞

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 25 / 90

Example 2.6 – The Accumulator System

The Accumulator System is defined by

n
X
y[n] = x[k]
k=−∞

The accumulator system is a linear system. In order to prove this,

we must show that it satisfies the superposition principle for all
inputs. Let us define two arbitrary inputs:
n
X
y1 [n] = x1 [k] ,
k=−∞

n
X
y2 [n] = x2 [k] .
k=−∞

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 25 / 90

Example 2.6 – The Accumulator System

When the input is x3 [n] = ax1 [n] + bx2 [n], the superposition princi-
ple requires that the output be y3 [n] = ay1 [n] + by2 [n].
n
X
y3 [n] = x3 [k]
k=−∞
n
X
= (ax1 [k] + bx2 [k])
k=−∞
n n
X X
=a x1 [k] + b x2 [k]
k=−∞ k=−∞

= ay1 [n] + by2 [n]

Thus, the accumulator system satisfies the superposition principle

for all inputs and is therefore linear.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 26 / 90

Time-Invariant Systems

A time-invariant (also known as shift-invariant) system is a system

for which a time shift or delay of the input sequence causes a cor-
responding shift in the output sequence.
A system is said to be time-invariant if, for all integer n0 , the input
sequence
x1 [n] = x[n − n0 ]
produces the output sequence

y1 [n] = y[n − n0 ]

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 27 / 90

Time-Invariant Systems

A time-invariant (also known as shift-invariant) system is a system

for which a time shift or delay of the input sequence causes a cor-
responding shift in the output sequence.
A system is said to be time-invariant if, for all integer n0 , the input
sequence
x1 [n] = x[n − n0 ]
produces the output sequence

y1 [n] = y[n − n0 ]

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 27 / 90

Example 2.8 – The accumulator as a Time-Invariant
System
We define x1 [n] = x[n − n0 ]. To check for time invariance, we solve
for both y[n − n0 ] and y1 [n] and compare them to see whether they
are equal. First,
n−n0
X
y[n − n0 ] = x[k] .
k=−∞

Next, we find
n n
X X
y1 [n] = x1 [k] = x[k − n0 ]
k=−∞ k=−∞

Using the change of variables k1 = k − n0 results in

n−n0
X
y1 [n] = x[k1 ] = y[n − n0 ] .
k1 =−∞

Thus, the system is time-invariant.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 28 / 90
Example 2.8 – The accumulator as a Time-Invariant
System
We define x1 [n] = x[n − n0 ]. To check for time invariance, we solve
for both y[n − n0 ] and y1 [n] and compare them to see whether they
are equal. First,
n−n0
X
y[n − n0 ] = x[k] .
k=−∞

Next, we find
n n
X X
y1 [n] = x1 [k] = x[k − n0 ]
k=−∞ k=−∞

Using the change of variables k1 = k − n0 results in

n−n0
X
y1 [n] = x[k1 ] = y[n − n0 ] .
k1 =−∞

Thus, the system is time-invariant.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 28 / 90
Example 2.8 – The accumulator as a Time-Invariant
System
We define x1 [n] = x[n − n0 ]. To check for time invariance, we solve
for both y[n − n0 ] and y1 [n] and compare them to see whether they
are equal. First,
n−n0
X
y[n − n0 ] = x[k] .
k=−∞

Next, we find
n n
X X
y1 [n] = x1 [k] = x[k − n0 ]
k=−∞ k=−∞

Using the change of variables k1 = k − n0 results in

n−n0
X
y1 [n] = x[k1 ] = y[n − n0 ] .
k1 =−∞

Thus, the system is time-invariant.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 28 / 90
Example 2.9 – The Compressor System

The compressor system is defined by the equation

y[n] = x[Mn] , −∞ < n < ∞

with M a positive integer. The system discards (M − 1) out of
every M samples. The output y1 [n] that results from the input
x1 [n] = x[n − n0 ] is given by

y1 [n] = x1 [Mn] = x[Mn − n0 ]

Delaying the output y[n] by n0 samples yields

y[n − n0 ] = x[M(n − n0 )]
Thus, we see that y[n − n0 ] is not equal to y1 [n] for all M and n0
and, therefore, the system is not time invariant.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 29 / 90

Causality

A system is causal if, for every n0 , the output sequence value at

n = n0 depends only on the input sequence values for n ≤ n0 .
This implies that if x1 [n] = x2 [n] for n ≤ n0 , then y1 [n] = y2 [n] for
n ≤ n0 . In other words, the system is non-anticipative.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 30 / 90

Causality

A system is causal if, for every n0 , the output sequence value at

n = n0 depends only on the input sequence values for n ≤ n0 .
This implies that if x1 [n] = x2 [n] for n ≤ n0 , then y1 [n] = y2 [n] for
n ≤ n0 . In other words, the system is non-anticipative.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 30 / 90

Example 2.10 – The Forward and Backward
Difference Systems

The forward difference system is defined by:

y[n] = x[n + 1] − x[n] .

This system is not causal, since the current value of the output
depends on a future value of the input.
The backward difference system is defined by:

y[n] = x[n] − x[n − 1] .

This system is causal, since its output depends only on present
and past values of the input.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 31 / 90

Example 2.10 – The Forward and Backward
Difference Systems

The forward difference system is defined by:

y[n] = x[n + 1] − x[n] .

This system is not causal, since the current value of the output
depends on a future value of the input.
The backward difference system is defined by:

y[n] = x[n] − x[n − 1] .

This system is causal, since its output depends only on present
and past values of the input.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 31 / 90

Stability

A system is stable in the bounded-input, bounded-output (BIBO) sense

if, and only if, every bounded input sequence

|x[n]| ≤ Bx < ∞, for all n

produces a bounded output sequence

|y[n]| ≤ By < ∞, for all n

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 32 / 90

 Linear Time-Invariant Systems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 33 / 90

Linear Time-Invariant Systems
( +∞
)
X
y[n] = T {x[n]} = T x[k]δ[n − k]
k=−∞

From the principle of superposition:

+∞ +∞
X X
y[n] = x[k]T {δ[n − k]} = x[k]hk [n]
k=−∞ k=−∞

If the system is time invariant:

T {δ[n]} = h[n] → T {δ[n − k]} = h[n − k]

+∞
X
y[n] = x[k]h[n − k] = x[n] ∗ h[n]
k=−∞

This equation is known as the convolution sum.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 34 / 90
Computing the convolution sum: method 1
1
x[n] h[n]

3
–2 0 n 0 n

x–2 [n] = x[–2]d[n + 2] y–2 [n] = x[–2]h[n + 2]

–2 0 n 0 n

Figure 2.8 Representation of

x0 [n] = x[0]d[n] y0 [n] = x[0]h[n]

the output of a LTI system as the

0 n 0 n
superposition of responses to
individual samples of the input.
x3 [n] = x[3]d[n – 3] y3 [n] = x[3]h[n – 3]

3 3
0 n 0 n

x[n] = x–2 [n] + x0 [n] + x3[n] y[n] = y–2 [n] + y0 [n] + y3 [n]

3
–2 0 n 0 n

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 35 / 90

Computing the convolution sum: method 2

Here, the output sequence y[n] is seen, for any fixed value of n, as
the sum of the product of the sequences x[k] and h[n − k], where
both sequences are expressed as functions of k.
This requires finding the sequence h[n − k], −∞ < k < ∞, for all
values of n that are of interest.
Noting that h[n − k] = h[−(k − n)] allows it to be computed in two
simpler steps:
I Define h1 [k] = h[−k]
I Define h2 [k] to be h1 [k] delayed by n samples, i.e., h2 [k] = h1 [k − n]
This leads to h2 [k] = h1 [k − n] = h[−(k − n)] = h[n − k]
In general, the sequence h[n − k], −∞ < k < ∞, is obtained by
I Reflecting h[k] about the origin to obtain h[−k],
I Shifting the origin of the reflected sequence to k = n.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 36 / 90

Computing the convolution sum: method 2

Here, the output sequence y[n] is seen, for any fixed value of n, as
the sum of the product of the sequences x[k] and h[n − k], where
both sequences are expressed as functions of k.
This requires finding the sequence h[n − k], −∞ < k < ∞, for all
values of n that are of interest.
Noting that h[n − k] = h[−(k − n)] allows it to be computed in two
simpler steps:
I Define h1 [k] = h[−k]
I Define h2 [k] to be h1 [k] delayed by n samples, i.e., h2 [k] = h1 [k − n]
This leads to h2 [k] = h1 [k − n] = h[−(k − n)] = h[n − k]
In general, the sequence h[n − k], −∞ < k < ∞, is obtained by
I Reflecting h[k] about the origin to obtain h[−k],
I Shifting the origin of the reflected sequence to k = n.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 36 / 90

Computing the convolution sum: method 2

Here, the output sequence y[n] is seen, for any fixed value of n, as
the sum of the product of the sequences x[k] and h[n − k], where
both sequences are expressed as functions of k.
This requires finding the sequence h[n − k], −∞ < k < ∞, for all
values of n that are of interest.
Noting that h[n − k] = h[−(k − n)] allows it to be computed in two
simpler steps:
I Define h1 [k] = h[−k]
I Define h2 [k] to be h1 [k] delayed by n samples, i.e., h2 [k] = h1 [k − n]
This leads to h2 [k] = h1 [k − n] = h[−(k − n)] = h[n − k]
In general, the sequence h[n − k], −∞ < k < ∞, is obtained by
I Reflecting h[k] about the origin to obtain h[−k],
I Shifting the origin of the reflected sequence to k = n.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 36 / 90

Computing the convolution sum: method 2

Here, the output sequence y[n] is seen, for any fixed value of n, as
the sum of the product of the sequences x[k] and h[n − k], where
both sequences are expressed as functions of k.
This requires finding the sequence h[n − k], −∞ < k < ∞, for all
values of n that are of interest.
Noting that h[n − k] = h[−(k − n)] allows it to be computed in two
simpler steps:
I Define h1 [k] = h[−k]
I Define h2 [k] to be h1 [k] delayed by n samples, i.e., h2 [k] = h1 [k − n]
This leads to h2 [k] = h1 [k − n] = h[−(k − n)] = h[n − k]
In general, the sequence h[n − k], −∞ < k < ∞, is obtained by
I Reflecting h[k] about the origin to obtain h[−k],
I Shifting the origin of the reflected sequence to k = n.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 36 / 90

Computing the convolution sum: method 2

Here, the output sequence y[n] is seen, for any fixed value of n, as
the sum of the product of the sequences x[k] and h[n − k], where
both sequences are expressed as functions of k.
This requires finding the sequence h[n − k], −∞ < k < ∞, for all
values of n that are of interest.
Noting that h[n − k] = h[−(k − n)] allows it to be computed in two
simpler steps:
I Define h1 [k] = h[−k]
I Define h2 [k] to be h1 [k] delayed by n samples, i.e., h2 [k] = h1 [k − n]
This leads to h2 [k] = h1 [k − n] = h[−(k − n)] = h[n − k]
In general, the sequence h[n − k], −∞ < k < ∞, is obtained by
I Reflecting h[k] about the origin to obtain h[−k],
I Shifting the origin of the reflected sequence to k = n.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 36 / 90

Computation of the convolution sum

h [k]

–3 0 6 k
(a)

h[–k] = h[0 – k]

–6 0 3 k
(b)

h[n – k] = h[–(k – n)]

n–6 0 n n+3 k
(c)
Figure 2.9 Forming the sequence h[n − k]. (a) The sequence h[k]
as a function of k. The sequence h[−k] as a function of k. (c) The
sequence h[n − k] = h[−(k − n)] as a function of k for n = 4.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 37 / 90

Analytical Evaluation of the Convolution Sum
h [n – k]
x [k]

Example 2.13 – Compute the

convolution between the two n 0 k
n – (N – 1) (a)
sequences:
h[n] = u[n] − u[n − N]


x[n] = αn u[n] n – (N – 1)
0 n k
(b)

Use the formula:

0 n k
n – (N – 1)
N2 (c)
X αN1 − αN2 +1
αk = , N2 ≥ N1 y [n]

k=N1
1−α
0 k
N–1
(d)

Figure 2.10 Sequence involved in

computing a discrete convolution.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 38 / 90

 Properties of Linear Time-Invariant Systems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 39 / 90

Convolution: commutative property

x[n] ∗ h[n] = h[n] ∗ x[n]

Figure 2.11 Three linear time-invariant systems

with identical impulse responses.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 40 / 90

Convolution: distributive property

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

h1 [n]

+
x[n] y [n]
h2 [n]

(a)

h1 [n] + h2 [n]
x[n] y[n]
(b)
Figure 2.12 (a) Parallel combination of linear time-
invariant systems. (b) An equivalent system.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 41 / 90

Stability and causality of LTI systems

An LTI system is stable if, and only if, its impulse response is ab-
solutely summable:
∞
X
S= |h[k]| < ∞
k=−∞

An LTI system is causal if

h[n] = 0, n<0

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 42 / 90

Stability and causality of LTI systems

An LTI system is stable if, and only if, its impulse response is ab-
solutely summable:
∞
X
S= |h[k]| < ∞
k=−∞

An LTI system is causal if

h[n] = 0, n<0

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 42 / 90

Finite- and infinite-duration impulse response systems

A finite-duration impulse response (FIR) system is a system whose

impulse response is of finite duration. In other words, its impulse
response has only a finite number of non-zero samples.
I Important: FIR systems are always stable, as long as each of the
impulse response values is finite in magnitude.
Systems whose impulse response are of infinite duration (infinite
number of non-zero samples) are known as infinite-duration im-
pulse response (IIR) systems.
I IIR systems may or may not be stable.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 43 / 90

Finite- and infinite-duration impulse response systems

A finite-duration impulse response (FIR) system is a system whose

impulse response is of finite duration. In other words, its impulse
response has only a finite number of non-zero samples.
I Important: FIR systems are always stable, as long as each of the
impulse response values is finite in magnitude.
Systems whose impulse response are of infinite duration (infinite
number of non-zero samples) are known as infinite-duration im-
pulse response (IIR) systems.
I IIR systems may or may not be stable.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 43 / 90

Examples

The ideal delay system (example 2.3)

y[n] = x[n − nd ], −∞ < n < ∞

h[n] = δ[n − nd ]
I Stable, causal (nd ≥ 0), FIR

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 44 / 90

Examples

The ideal delay system (example 2.3)

y[n] = x[n − nd ], −∞ < n < ∞

h[n] = δ[n − nd ]
I Stable, causal (nd ≥ 0), FIR

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 44 / 90

Examples

The moving average system (example 2.4)

M2
1 X
y[n] = x[n − k]
M1 + M2 + 1 k=−M
1


1
 , −M1 ≤ n ≤ M2
h[n] = M1 + M2 + 1
0, otherwise


I Stable, causal (M1 ≤ 0, M2 ≥ 0), FIR

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 45 / 90

Examples

The moving average system (example 2.4)

M2
1 X
y[n] = x[n − k]
M1 + M2 + 1 k=−M
1


1
 , −M1 ≤ n ≤ M2
h[n] = M1 + M2 + 1
0, otherwise


I Stable, causal (M1 ≤ 0, M2 ≥ 0), FIR

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 45 / 90

Examples

The accumulator system (example 2.6)

n
X
y[n] = x[k]
k=−∞

h[n] = u[n]
I Unstable, causal, IIR

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 46 / 90

Examples

The accumulator system (example 2.6)

n
X
y[n] = x[k]
k=−∞

h[n] = u[n]
I Unstable, causal, IIR

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 46 / 90

Examples

The forward difference system (example 2.10)

y[n] = x[n + 1] − x[n]

h[n] = δ[n + 1] − δ[n]
I Stable, non-causal, FIR (finite impulse response)
The backward difference system (example 2.10)

y[n] = x[n] − x[n − 1]

h[n] = δ[n] − δ[n − 1]
I Stable, causal, FIR (finite impulse response)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 47 / 90

Examples

The forward difference system (example 2.10)

y[n] = x[n + 1] − x[n]

h[n] = δ[n + 1] − δ[n]
I Stable, non-causal, FIR (finite impulse response)
The backward difference system (example 2.10)

y[n] = x[n] − x[n − 1]

h[n] = δ[n] − δ[n − 1]
I Stable, causal, FIR (finite impulse response)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 47 / 90

Examples

The forward difference system (example 2.10)

y[n] = x[n + 1] − x[n]

h[n] = δ[n + 1] − δ[n]
I Stable, non-causal, FIR (finite impulse response)
The backward difference system (example 2.10)

y[n] = x[n] − x[n − 1]

h[n] = δ[n] − δ[n − 1]
I Stable, causal, FIR (finite impulse response)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 47 / 90

Examples

The forward difference system (example 2.10)

y[n] = x[n + 1] − x[n]

h[n] = δ[n + 1] − δ[n]
I Stable, non-causal, FIR (finite impulse response)
The backward difference system (example 2.10)

y[n] = x[n] − x[n − 1]

h[n] = δ[n] − δ[n − 1]
I Stable, causal, FIR (finite impulse response)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 47 / 90

Example: commutative property

Forward One-sample
x [n] difference delay y[n]

(a)

One-sample Forward
x [n] delay difference y [n]

(b)

Backward h[n] = (δ[n + 1] − δ[n]) ∗ δ[n − 1]

x [n] difference y [n]
= δ[n − 1] ∗ (δ[n + 1] − δ[n])
(c) = δ[n] − δ[n − 1]
Figure 2.13 Equivalent systems found
by using the commutative property of
convolution.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 48 / 90

Example: inverse system

h[n] = u[n] ∗ (δ[n] − δ[n − 1])

= u[n] − u[n − 1]
= δ[n]

Backward-
Accumulator
difference
x [n] system y [n] x [n]
system

Figure 2.14 An accumulator in cascade with a backward difference.

The cascade combination is equivalent to the identity system.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 49 / 90

 Linear Constant-Coefficient Difference Equations

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 50 / 90

Solution to a difference equation

General form:
N M
X X
ak y[n − k] = bm x[n − m]
k=0 m=0

General solution:

y[n] = yp [n] + yh [n]

yp [n]: particular solution
yh [n]: solution to the homogeneous equation

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 51 / 90

Solution to a difference equation

General form:
N M
X X
ak y[n − k] = bm x[n − m]
k=0 m=0

General solution:

y[n] = yp [n] + yh [n]

yp [n]: particular solution
yh [n]: solution to the homogeneous equation

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 51 / 90

Solution to the homogeneous equation

Homogeneous equation:
N
X
ak y[n − k] = 0
k=0

Solution to the homogeneous equation:

N
X
yh [n] = Am znm
m=1

The complex numbers zm are the solutions to the characteristic

equation:
N
X
ak z−k = 0
k=0

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 52 / 90

Solution to the homogeneous equation

Homogeneous equation:
N
X
ak y[n − k] = 0
k=0

Solution to the homogeneous equation:

N
X
yh [n] = Am znm
m=1

The complex numbers zm are the solutions to the characteristic

equation:
N
X
ak z−k = 0
k=0

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 52 / 90

Solution to the homogeneous equation

Homogeneous equation:
N
X
ak y[n − k] = 0
k=0

Solution to the homogeneous equation:

N
X
yh [n] = Am znm
m=1

The complex numbers zm are the solutions to the characteristic

equation:
N
X
ak z−k = 0
k=0

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 52 / 90

Example
Difference equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 2x[n − 1]

Homogeneous equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 0

Characteristic equation:

1 − 5z−1 + 6z−2 = 0
z2 − 5z + 6 = 0

Roots: z1 = 2 and z2 = 3

yh [n] = A1 (2)n + A2 (3)n

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 53 / 90
Example
Difference equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 2x[n − 1]

Homogeneous equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 0

Characteristic equation:

1 − 5z−1 + 6z−2 = 0
z2 − 5z + 6 = 0

Roots: z1 = 2 and z2 = 3

yh [n] = A1 (2)n + A2 (3)n

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 53 / 90
Example
Difference equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 2x[n − 1]

Homogeneous equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 0

Characteristic equation:

1 − 5z−1 + 6z−2 = 0
z2 − 5z + 6 = 0

Roots: z1 = 2 and z2 = 3

yh [n] = A1 (2)n + A2 (3)n

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 53 / 90
Example
Difference equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 2x[n − 1]

Homogeneous equation:

y[n] − 5y[n − 1] + 6y[n − 2] = 0

Characteristic equation:

1 − 5z−1 + 6z−2 = 0
z2 − 5z + 6 = 0

Roots: z1 = 2 and z2 = 3

yh [n] = A1 (2)n + A2 (3)n

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 53 / 90
General solution

The homogeneous equation has N coefficients to determine and

therefore a set of N boundary conditions are necessary to define
a unique solution for a given input x[n].
The boundary conditions can be output values for specific values
of n: y[−1], y[−2], . . . , y[−N]
Linearity, time invariance and causality of the system will depend
on the auxiliary conditions. For a difference equation with con-
stant coefficients, initial rest guarantees these properties.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 54 / 90

General solution

The homogeneous equation has N coefficients to determine and

therefore a set of N boundary conditions are necessary to define
a unique solution for a given input x[n].
The boundary conditions can be output values for specific values
of n: y[−1], y[−2], . . . , y[−N]
Linearity, time invariance and causality of the system will depend
on the auxiliary conditions. For a difference equation with con-
stant coefficients, initial rest guarantees these properties.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 54 / 90

General solution

The homogeneous equation has N coefficients to determine and

therefore a set of N boundary conditions are necessary to define
a unique solution for a given input x[n].
The boundary conditions can be output values for specific values
of n: y[−1], y[−2], . . . , y[−N]
Linearity, time invariance and causality of the system will depend
on the auxiliary conditions. For a difference equation with con-
stant coefficients, initial rest guarantees these properties.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 54 / 90

Non-recursive difference equation
If N = 0, no recursion is required to compute the output and no
auxiliary conditions are required. That is
M  
X bk
y[n] = x[n − k]
k=0
a0

M  
X bk
h[n] = δ[n − k]
k=0
a0
 
 bn , 0 ≤ n ≤ M
h[n] = a
 0
0, otherwise
The impulse response is finite in duration (FIR).
Indeed, the output of any FIR system can be computed nonrecur-
sively using the first equation above.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 55 / 90
 Frequency Domain Representation of
Discrete-Time Signals and Systems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 56 / 90

Frequency response
Consider an LTI system whose input is x[n] = ejωn :

T { •}
x [n] y[n]

The output is given by:

∞ ∞
!
X X
y[n] = h[k]ejω(n−k) = ejωn h[k]e−jωk = H(ejω )ejωn
k=−∞ k=−∞

∞
X
H(ejω ) = h[k]e−jωk
k=−∞

Thus, ejωn is an eigenfunction of the system, and H(ejω ) is the

associated eigenvalue. H(ejω ) is called the frequency response
of the system.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 57 / 90
Properties of the frequency response

In general, the frequency response H(ejω ) is complex:

jω )
H(ejω ) = HR (ejω ) + jHI (ejω ) = |H(ejω )|ej^H(e

The frequency response H(ejω ) is periodic with period 2π .

∞ ∞
X X
H(ej(ω+2π) ) = h[k]e−j(ω+2π)k = h[k]e−jωk e−j2πk
k=−∞ k=−∞
∞
X
= h[k]e−jωk = H(ejω )
k=−∞

I This is a direct consequence of the fact that the sequences ejωn and
ej(ω+2π)n , −∞ < n < ∞, are indistinguishable from each other.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 58 / 90

Properties of the frequency response

In general, the frequency response H(ejω ) is complex:

jω )
H(ejω ) = HR (ejω ) + jHI (ejω ) = |H(ejω )|ej^H(e

The frequency response H(ejω ) is periodic with period 2π .

∞ ∞
X X
H(ej(ω+2π) ) = h[k]e−j(ω+2π)k = h[k]e−jωk e−j2πk
k=−∞ k=−∞
∞
X
= h[k]e−jωk = H(ejω )
k=−∞

I This is a direct consequence of the fact that the sequences ejωn and
ej(ω+2π)n , −∞ < n < ∞, are indistinguishable from each other.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 58 / 90

Frequency response of the ideal delay

The ideal delay system is

y[n] = x[n − nd ]
To compute its frequency response, we simply make x[n] = ejωn in
the equation above

y[n] = ejω(n−nd ) = e−jωnd ejωn

Thus

H(ejω ) = e−jωnd

The frequency response of the system can also be found by com-

puting the Fourier transform of its impulse response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 59 / 90

Frequency response of the ideal delay

The ideal delay system is

y[n] = x[n − nd ]
To compute its frequency response, we simply make x[n] = ejωn in
the equation above

y[n] = ejω(n−nd ) = e−jωnd ejωn

Thus

H(ejω ) = e−jωnd

The frequency response of the system can also be found by com-

puting the Fourier transform of its impulse response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 59 / 90

Frequency response of the ideal delay

The ideal delay system is

y[n] = x[n − nd ]
To compute its frequency response, we simply make x[n] = ejωn in
the equation above

y[n] = ejω(n−nd ) = e−jωnd ejωn

Thus

H(ejω ) = e−jωnd

The frequency response of the system can also be found by com-

puting the Fourier transform of its impulse response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 59 / 90

Frequency response of the ideal delay

The ideal delay system is

y[n] = x[n − nd ]
To compute its frequency response, we simply make x[n] = ejωn in
the equation above

y[n] = ejω(n−nd ) = e−jωnd ejωn

Thus

H(ejω ) = e−jωnd

The frequency response of the system can also be found by com-

puting the Fourier transform of its impulse response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 59 / 90

Determining the output of a system from its frequency
response
A broad class of signals can be represented as a linear combina-
tion of complex exponentials in the form
X
x[n] = αk ejωk n
k

From the principle of superposition, the corresponding output of a

linear time-invariant system is
X
y[n] = αk H(ejωk )ejωk n
k

Thus, if we can find a representation of x[n] as a superposition

of complex exponential sequences and if we know the frequency
response of the system, then we can find the output of the system
using the equation above.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 60 / 90
Determining the output of a system from its frequency
response
A broad class of signals can be represented as a linear combina-
tion of complex exponentials in the form
X
x[n] = αk ejωk n
k

From the principle of superposition, the corresponding output of a

linear time-invariant system is
X
y[n] = αk H(ejωk )ejωk n
k

Thus, if we can find a representation of x[n] as a superposition

of complex exponential sequences and if we know the frequency
response of the system, then we can find the output of the system
using the equation above.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 60 / 90
Determining the output of a system from its frequency
response
A broad class of signals can be represented as a linear combina-
tion of complex exponentials in the form
X
x[n] = αk ejωk n
k

From the principle of superposition, the corresponding output of a

linear time-invariant system is
X
y[n] = αk H(ejωk )ejωk n
k

Thus, if we can find a representation of x[n] as a superposition

of complex exponential sequences and if we know the frequency
response of the system, then we can find the output of the system
using the equation above.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 60 / 90
Example 2.18 – Sinusoidal response of LTI systems

Consider a sinusoidal input to an LTI system

A jφ jω0 n A −jφ −jω0 n

x[n] = A cos(ω0 n + φ) =
e e + e e
2 2
The responses to the input components x1 [n] and x2 [n] are

x1 [n] = (A/2)ejφ ejω0 n → y1 [n] = H(ejω0 )(A/2)ejφ ejω0 n

x2 [n] = (A/2)e−jφ e−jω0 n → y2 [n] = H(e−jω0 )(A/2)e−jφ e−jω0 n
Thus, the total response is

A
H(ejω0 )ejφ ejω0 n + H(e−jω0 )e−jφ e−jω0 n

y[n] =
2

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 61 / 90

Example 2.18 – Sinusoidal response of LTI systems

Consider a sinusoidal input to an LTI system

A jφ jω0 n A −jφ −jω0 n

x[n] = A cos(ω0 n + φ) =
e e + e e
2 2
The responses to the input components x1 [n] and x2 [n] are

x1 [n] = (A/2)ejφ ejω0 n → y1 [n] = H(ejω0 )(A/2)ejφ ejω0 n

x2 [n] = (A/2)e−jφ e−jω0 n → y2 [n] = H(e−jω0 )(A/2)e−jφ e−jω0 n
Thus, the total response is

A
H(ejω0 )ejφ ejω0 n + H(e−jω0 )e−jφ e−jω0 n

y[n] =
2

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 61 / 90

Example 2.18 – Sinusoidal response of LTI systems

Consider a sinusoidal input to an LTI system

A jφ jω0 n A −jφ −jω0 n

x[n] = A cos(ω0 n + φ) =
e e + e e
2 2
The responses to the input components x1 [n] and x2 [n] are

x1 [n] = (A/2)ejφ ejω0 n → y1 [n] = H(ejω0 )(A/2)ejφ ejω0 n

x2 [n] = (A/2)e−jφ e−jω0 n → y2 [n] = H(e−jω0 )(A/2)e−jφ e−jω0 n
Thus, the total response is

A
H(ejω0 )ejφ ejω0 n + H(e−jω0 )e−jφ e−jω0 n

y[n] =
2

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 61 / 90

Example 2.18 – Sinusoidal response of LTI systems

If h[n] is real, it can be shown that H(e−jω0 ) = H ∗ (ejω0 ). Thus

y[n] = A|H(ejω0 )| cos(ω0 n + φ + θ) ,

where θ = ^H(ejω0 ) is the phase of the system function at fre-
quency ω0 .
In the case of the ideal delay, |H(ejω0 )| = 1 and θ = −ω0 nd .
Therefore

y[n] = A cos(ω0 n + φ − ω0 nd )
= A cos[ω0 (n − nd ) + φ]

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 62 / 90

Example 2.18 – Sinusoidal response of LTI systems

If h[n] is real, it can be shown that H(e−jω0 ) = H ∗ (ejω0 ). Thus

y[n] = A|H(ejω0 )| cos(ω0 n + φ + θ) ,

where θ = ^H(ejω0 ) is the phase of the system function at fre-
quency ω0 .
In the case of the ideal delay, |H(ejω0 )| = 1 and θ = −ω0 nd .
Therefore

y[n] = A cos(ω0 n + φ − ω0 nd )
= A cos[ω0 (n − nd ) + φ]

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 62 / 90

Ideal frequency-selective filters
These are systems for which the frequency response is unity over
a certain range of frequencies and zero elsewhere.
Hlp(e jv)

–2p –2p + vc –p –vc vc p 2p – vc 2p v

(a)

Hlp(e jv)

–p –vc vc p v
(b)

Figure 2.17 Ideal low-pass filter showing (a) periodicity of the frequency
response and (b) one period of the periodic frequency response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 63 / 90

Ideal frequency-selective filters
Hhp(e jv)

–p – vc 0 vc p v

(a)

Hbs(e jv)
1
Figure 2.18 Ideal frequency se-
lective filters. (a) High-pass filter.
(b) Band-stop filter. (c) Band-pass
–p – vb –va 0 va vb p v
(b) filter.

Hbp(e jv)

–p – vb –va 0 va vb p v
(c)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 64 / 90

Frequency response of the moving-average system


1
 , −M1 ≤ n ≤ M2
h[n] = M1 + M2 + 1
0, otherwise


M2
1 X
jω
H(e ) = e−jωn
M1 + M2 + 1 n=−M
1

1 sin [ω(M1 + M2 + 1)/2] −jω(M2 −M1 )/2

= e
M1 + M2 + 1 sin(ω/2)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 65 / 90

Frequency response of the moving-average system
|H(e jv)|
1

–2p –p –2p 2p p 2p v
5 5

\H(e jv)

–2p –p p 2p v

–p

Figure 2.19 (a) Magnitude and (b) phase of the frequency response
of the moving-average system for the case M1 = 0 and M2 = 4.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 66 / 90

 Representation of Sequences by Fourier
Transforms

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 67 / 90

Fourier Transform

∞
X
jω
X(e ) = x[n]e−jωn (analysis equation)
n=−∞

Z +π
1
x[n] = X(ejω )ejωn dω (synthesis equation)
2π −π

In general, the Fourier transform is a complex-valued function

of ω :
jω )
X(ejω ) = XR (ejω ) + jXI (ejω ) = |X(ejω )|ej^X(e

The quantities |X(ejω )| and ^X(ejω ) are the magnitude and phase
of the Fourier transform.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 68 / 90

Fourier Transform

∞
X
jω
X(e ) = x[n]e−jωn (analysis equation)
n=−∞

Z +π
1
x[n] = X(ejω )ejωn dω (synthesis equation)
2π −π

In general, the Fourier transform is a complex-valued function

of ω :
jω )
X(ejω ) = XR (ejω ) + jXI (ejω ) = |X(ejω )|ej^X(e

The quantities |X(ejω )| and ^X(ejω ) are the magnitude and phase
of the Fourier transform.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 68 / 90

Fourier Transform

The Fourier Transform X(ejω ) exists if x[n] is absolutely summable.

∞
X
|x[n]| < ∞
n=−∞

In this case, the series converges uniformly to a continuous func-

tion of ω .
Since a stable sequence is, by definition, absolutely summable, all
stable sequences have Fourier transforms.
It also follows that any stable system has a finite and continuous
frequency response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 69 / 90

Fourier Transform

The Fourier Transform X(ejω ) exists if x[n] is absolutely summable.

∞
X
|x[n]| < ∞
n=−∞

In this case, the series converges uniformly to a continuous func-

tion of ω .
Since a stable sequence is, by definition, absolutely summable, all
stable sequences have Fourier transforms.
It also follows that any stable system has a finite and continuous
frequency response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 69 / 90

Fourier Transform

The Fourier Transform X(ejω ) exists if x[n] is absolutely summable.

∞
X
|x[n]| < ∞
n=−∞

In this case, the series converges uniformly to a continuous func-

tion of ω .
Since a stable sequence is, by definition, absolutely summable, all
stable sequences have Fourier transforms.
It also follows that any stable system has a finite and continuous
frequency response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 69 / 90

Fourier Transform

The Fourier Transform X(ejω ) exists if x[n] is absolutely summable.

∞
X
|x[n]| < ∞
n=−∞

In this case, the series converges uniformly to a continuous func-

tion of ω .
Since a stable sequence is, by definition, absolutely summable, all
stable sequences have Fourier transforms.
It also follows that any stable system has a finite and continuous
frequency response.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 69 / 90

Example 2.21 – Absolute Summability for a
Suddenly-Applied Exponential

Compute the Fourier transform of x[n] = an u[n]:

∞ ∞
X
n −jωn
X 1
jω
X(e ) = ae = (ae−jω )n =
n=0 n=0
1 − ae−jω
Condition for convergence:

|ae−jω | < 1 or |a| < 1

F 1
an u[n] ←−
−→ X(ejω ) =
1 − ae−jω

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 70 / 90

Example 2.21 – Absolute Summability for a
Suddenly-Applied Exponential

Compute the Fourier transform of x[n] = an u[n]:

∞ ∞
X
n −jωn
X 1
jω
X(e ) = ae = (ae−jω )n =
n=0 n=0
1 − ae−jω
Condition for convergence:

|ae−jω | < 1 or |a| < 1

F 1
an u[n] ←−
−→ X(ejω ) =
1 − ae−jω

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 70 / 90

Fourier Transform – Square and Absolute Summability

Absolute summability is a sufficient condition for the existence of

the Fourier transform. It also guarantees uniform convergence.
Some sequences are not absolutely summable, but are square
summable. Square summable sequences have a Fourier trans-
form but without uniform convergence.
∞
X
|x[n]|2 < ∞
n=−∞

Square summable sequences converge in the mean-square sense.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 71 / 90

Fourier Transform – Square and Absolute Summability

Absolute summability is a sufficient condition for the existence of

the Fourier transform. It also guarantees uniform convergence.
Some sequences are not absolutely summable, but are square
summable. Square summable sequences have a Fourier trans-
form but without uniform convergence.
∞
X
|x[n]|2 < ∞
n=−∞

Square summable sequences converge in the mean-square sense.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 71 / 90

Fourier Transform – Square and Absolute Summability

Absolute summability is a sufficient condition for the existence of

the Fourier transform. It also guarantees uniform convergence.
Some sequences are not absolutely summable, but are square
summable. Square summable sequences have a Fourier trans-
form but without uniform convergence.
∞
X
|x[n]|2 < ∞
n=−∞

Square summable sequences converge in the mean-square sense.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 71 / 90

Example 2.22 – Square Summability for the Ideal
Lowpass Filter
Compute the inverse Fourier transform of the ideal lowpass filter:

|ω| < ωc

jω
1,
Hlp (e ) =
0, ωc < |ω| < π

Z ωc
1 1 ωc
hlp [n] = ejωn dω = [ejωn ]−ωc
2π −ωc 2πjn
1 sin(ωc n)
= (ejωc n − e−jωc n ) =
2πjn πn
hlp [n] is nonzero for n < 0 → system is noncausal.
hlp [n] is not absolutely summable (because of the discontinuity at
ω = ωc ).
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 72 / 90
Square Summability for the Ideal Lowpass filter
HM (e jv), M = 1 HM (e jv), M = 3

–p –vc 0 vc p v –p –vc 0 vc p v
(a) (b)

HM (e jv), M = 7 HM (e jv), M = 19

–p –vc 0 vc p v –p – vc 0 vc p v
(c) (d)

Figure 2.21 Convergence of the Fourier transform. The oscillatory

behavior at ω = ωc is often called the Gibbs phenomenon.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 73 / 90

Example 2.23 – Fourier Transform of a Constant

The sequence x[n] = 1, ∀n, is neither absolutely summable nor

square summable. However, it is possible (and useful) to define
the Fourier transform of the sequence x[n] to be the periodic im-
pulse train
∞
X
jω
X(e ) = 2πδ(ω + 2πr)
r=−∞

where δ(ω) is a function of a continuous variable.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 74 / 90

Example 2.24 – Fourier Transform of Complex
Exponential Sequences
Consider a sequence x[n] whose Fourier transform is the periodic
impulse train
∞
X
jω
X(e ) = 2πδ(ω − ω0 + 2πr)
r=−∞

We can determine x[n] by substituting X(ejω ) into the inverse Fourier

transform integral:

Z π
1
x[n] = 2πδ(ω − ω0 )ejωn dω
2π −π

= ejω0 n , ∀n
For ω0 = 0, this reduces to the sequence of Example 2.23.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 75 / 90
 Symmetry Properties of the Fourier
Transform

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 76 / 90

Definitions

Conjugate-symmetric sequence:

xe [n] = xe∗ [−n]

Re [n] + jIe [n] = Re [−n] − jIe [−n]

Conjugate-antisymmetric sequence

xo [n] = −xo∗ [−n]

Ro [n] + jIo [n] = −Ro [−n] + jIo [−n]

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 77 / 90

Definitions

Conjugate-symmetric sequence:

xe [n] = xe∗ [−n]

Re [n] + jIe [n] = Re [−n] − jIe [−n]

Conjugate-antisymmetric sequence

xo [n] = −xo∗ [−n]

Ro [n] + jIo [n] = −Ro [−n] + jIo [−n]

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 77 / 90

Decomposition

Any sequence x[n] can be expressed as a sum of a conjugate-

symmetric and a conjugate-antisymmetric sequence.

x[n] = xe [n] + xo [n]

1
xe [n] = (x[n] + x∗ [−n])
2
1
xo [n] = (x[n] − x∗ [−n])
2

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 78 / 90

Properties

Sequence x[n] Fourier transform X(ejω )

x∗ [n] X ∗ (e−jω )
x∗ [−n] X ∗ (ejω )
Re(x[n]) Xe (ejω )
j Im(x[n]) Xo (ejω )
xe [n] Re{X(ejω )}
xo [n] j Im{X(ejω )}

where:
xe [n]: conjugate-symmetric part of x[n]
xo [n]: conjugate-antisymmetric part of x[n]
Xe (ejω ): conjugate-symmetric part of X(ejω )
Xo (ejω ): conjugate-antisymmetric part of X(ejω )

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 79 / 90

Properties of real sequences

If x[n] is real:

X(ejω ) = X ∗ (e−jω ) (Fourier transform is conjugate-symmetric)

XR (ejω ) = XR (e−jω ) (real part is even)
XI (ejω ) = −XI (e−jω ) (imaginary part is odd)
|X(ejω )| = |X(e−jω )| (magnitude is even)
^X(ejω ) = −^X(e−jω ) (phase is odd)

F
If x[n] is odd: xo [n] ←−
−→ XI (ejω ) = −XI (e−jω ) (real and odd)
F
If x[n] is even: xe [n] ←−
−→ XR (ejω ) = XR (e−jω ) (real and even)

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 80 / 90

Example 2.25 – Illustration of Symmetry Properties
F 1
x[n] = an u[n] ←−
−→ X(ejω ) =
1 − ae−jω
5 5
4 4

Amplitude
Amplitude

3 3
2 2
1 1
0 0
–p –p 0 p p –p –p 0 p p
2 2 2 2
Radian frequency (v) Radian frequency (v)
(a) (c)
2 1.0

Phase (radians)
Amplitude

1 0.5
0 0
–1 – 0.5
–2 –1.0
–p –p 0 p p –p –p 0 p p
2 2 2 2
Radian frequency (v) Radian frequency (v)
(b) (d)
Figure 2.22 Frequency response for a system with impulse response
h[n] = an u[n]. (a) Real part. (b) Imaginary part. (c) Magnitude. (d) Phase.
Solid curves are used for a = 0.9 and dashed curves for a = 0.5.
Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 81 / 90
 Fourier Transform Theorems

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 82 / 90

Linearity

F
x1 [n] ←−
−→ X1 (ejω )

F
x2 [n] ←−
−→ X2 (ejω )

F
ax1 [n] + bx2 [n] ←−
−→ aX1 (ejω ) + bX2 (ejω )

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 83 / 90

Time shifting and frequency shifting

F
x[n] ←−
−→ X(ejω )

F
x[n − nd ] ←−
−→ e−jωnd X(ejω )

F
ejω0 n x[n] ←−
−→ X(ej(ω−ω0 ) )

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 84 / 90

Time reversal

F
x[n] ←−
−→ X(ejω )

F
x[−n] ←−
−→ X(e−jω )
If x[n] is real:
F
x[−n] ←−
−→ X ∗ (ejω )

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 85 / 90

Differentiation in frequency

F
x[n] ←−
−→ X(ejω )

F dX(ejω )
nx[n] ←−
−→ j
dω

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 86 / 90

Parseval’s Theorem

F
x[n] ←−
−→ X(ejω )

∞ Z +π
X 12
E= |x[n]| = |X(ejω )|2 dω
n=−∞
2π −π

The function |X(ejω )|2 is called the energy density spectrum.

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 87 / 90

The Convolution Theorem

F
x[n] ←−
−→ X(ejω )

F
h[n] ←−
−→ H(ejω )
If:
∞
X
y[n] = x[k]h[n − k] = x[n] ∗ h[n]
k=−∞

Then:

Y(ejω ) = X(ejω )H(ejω )

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 88 / 90

The Modulation or Windowing Theorem

F
x[n] ←−
−→ X(ejω )

F
w[n] ←−
−→ W(ejω )
If:

y[n] = x[n]w[n]

Then:
Z π
jω 1
Y(e ) = X(ejθ )W(ej(ω−θ) )dθ
2π −π

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 89 / 90

Example 2.26

Compute the Fourier transform of x[n] = an u[n − 5]

F 1
an u[n] ←−
−→
1 − ae−jω

x[n] = a5 an−5 u[n − 5]

a5 e−j5ω
X(ejω ) =
1 − ae−jω

Adriano Vilela Barbosa (UFMG) Signal Processing – Chapter 2 90 / 90