CHAPTER 2

Discrete-time
Signals and
Systems

We begin with the concepts of signals and systems in discrete time. A

number of important types of signals and their operations are introduced.
Linear and shift-invariant systems are discussed mostly because they are
easier to analyze and implement. The convolution and the difference equa-
tion representations are given special attention because of their impor-
tance in digital signal processing and in MATLAB. The emphasis in this
chapter is on the representations and implementation of signals and sys-
tems using MATLAB.

2.1 DISCRETE-TIME SIGNALS

Signals are broadly classified into analog and discrete signals. An analog
signal will be denoted by xa (t), in which the variable t can represent any
physical quantity, but we will assume that it represents time in seconds. A
discrete signal will be denoted by x(n), in which the variable n is integer-
valued and represents discrete instances in time. Therefore it is also called
a discrete-time signal, which is a number sequence and will be denoted by
one of the following notations:
x(n) = {x(n)} = {. . . , x(−1), x(0), x(1), . . .}
↑
where the up-arrow indicates the sample at n = 0.

22
Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 23

In MATLAB we can represent a finite-duration sequence by a row

vector of appropriate values. However, such a vector does not have any
information about sample position n. Therefore a correct representation
of x(n) would require two vectors, one each for x and n. For example, a
sequence x(n) = {2, 1, −1, 0, 1, 4, 3, 7} can be represented in MATLAB by
↑

>> n=[-3,-2,-1,0,1,2,3,4]; x=[2,1,-1,0,1,4,3,7];

Generally, we will use the x-vector representation alone when the sample
position information is not required or when such information is trivial
(e.g. when the sequence begins at n = 0). An arbitrary infinite-duration
sequence cannot be represented in MATLAB due to the finite memory
limitations.

2.1.1 TYPES OF SEQUENCES

We use several elementary sequences in digital signal processing for anal-
ysis purposes. Their definitions and MATLAB representations follow.
1. Unit sample sequence:
! " #
1, n = 0
δ(n) = = . . . , 0, 0, 1, 0, 0, . . .
0, n =
" 0 ↑

In MATLAB the function zeros(1,N) generates a row vector of N

zeros, which can be used to implement δ(n) over a finite interval. How-
ever, the logical relation n==0 is an elegant way of implementing δ(n).
For example, to implement
!
1, n = n0
δ(n − n0 ) =
0, n "= n0
over the n1 ≤ n0 ≤ n2 interval, we will use the following MATLAB
function.

function [x,n] = impseq(n0,n1,n2)

% Generates x(n) = delta(n-n0); n1 <= n <= n2
% ----------------------------------------------
% [x,n] = impseq(n0,n1,n2)
%
n = [n1:n2]; x = [(n-n0) == 0];

2. Unit step sequence:

!
1, n ≥ 0
u(n) = = {. . . , 0, 0, 1, 1, 1, . . .}
0, n < 0 ↑

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 24 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

In MATLAB the function ones(1,N) generates a row vector of N ones.

It can be used to generate u(n) over a finite interval. Once again an
elegant approach is to use the logical relation n>=0. To implement
!
1, n ≥ n0
u(n − n0 ) =
0, n < n0

over the n1 ≤ n0 ≤ n2 interval, we will use the following MATLAB

function.

function [x,n] = stepseq(n0,n1,n2)

% Generates x(n) = u(n-n0); n1 <= n <= n2
% ------------------------------------------
% [x,n] = stepseq(n0,n1,n2)
%
n = [n1:n2]; x = [(n-n0) >= 0];

3. Real-valued exponential sequence:

x(n) = an , ∀n; a ∈ R

In MATLAB an array operator “.^” is required to implement a real

exponential sequence. For example, to generate x(n) = (0.9)n , 0 ≤
n ≤ 10, we will need the following MATLAB script:

>> n = [0:10]; x = (0.9).^n;

4. Complex-valued exponential sequence:

x(n) = e(σ+jω0 )n , ∀n

where σ produces an attenuation (if <0) or amplification (if >0)

and ω0 is the frequency in radians. A MATLAB function exp is
used to generate exponential sequences. For example, to generate
x(n) = exp[(2 + j3)n], 0 ≤ n ≤ 10, we will need the following MATLAB
script:

>> n = [0:10]; x = exp((2+3j)*n);

5. Sinusoidal sequence:

x(n) = A cos(ω0 n + θ0 ), ∀n

where A is an amplitude and θ0 is the phase in radians. A MAT-

LAB function cos (or sin) is used to generate sinusoidal sequences.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 25

For example, to generate x(n) = 3 cos(0.1πn + π/3) + 2 sin(0.5πn),

0 ≤ n ≤ 10, we will need the following MATLAB script:

>> n = [0:10]; x = 3*cos(0.1*pi*n+pi/3) + 2*sin(0.5*pi*n);

6. Random sequences: Many practical sequences cannot be described

by mathematical expressions like those above. These sequences are
called random (or stochastic) sequences and are characterized by pa-
rameters of the associated probability density functions. In MATLAB
two types of (pseudo-) random sequences are available. The rand(1,N)
generates a length N random sequence whose elements are uniformly
distributed between [0, 1]. The randn(1,N) generates a length N Gaus-
sian random sequence with mean 0 and variance 1. Other random se-
quences can be generated using transformations of the above functions.
7. Periodic sequence: A sequence x(n) is periodic if x(n) = x(n + N ),
∀n. The smallest integer N that satisfies this relation is called the
fundamental period. We will use x̃(n) to denote a periodic sequence.
To generate P periods of x̃(n) from one period {x(n), 0 ≤ n ≤ N −1},
we can copy x(n) P times:

>> xtilde = [x,x,...,x];

But an elegant approach is to use MATLAB’s powerful indexing capa-

bilities. First we generate a matrix containing P rows of x(n) values.
Then we can concatenate P rows into a long row vector using the
construct (:). However, this construct works only on columns. Hence
we will have to use the matrix transposition operator ’ to provide the
same effect on rows.

>> xtilde = x’ * ones(1,P); % P columns of x; x is a row vector

>> xtilde = xtilde(:); % long column vector
>> xtilde = xtilde’; % long row vector

Note that the last two lines can be combined into one for compact
coding. This is shown in Example 2.1.

2.1.2 OPERATIONS ON SEQUENCES

Here we briefly describe basic sequence operations and their MATLAB
equivalents.
1. Signal addition: This is a sample-by-sample addition given by
{x1 (n)} + {x2 (n)} = {x1 (n) + x2 (n)}

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 26 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

It is implemented in MATLAB by the arithmetic operator “+”. How-

ever, the lengths of x1 (n) and x2 (n) must be the same. If sequences are
of unequal lengths, or if the sample positions are different for equal-
length sequences, then we cannot directly use the operator +. We have
to first augment x1 (n) and x2 (n) so that they have the same position
vector n (and hence the same length). This requires careful attention
to MATLAB’s indexing operations. In particular, logical operation of
intersection “&”, relational operations like “<=” and “==”, and the
find function are required to make x1 (n) and x2 (n) of equal length.
The following function, called the sigadd function, demonstrates these
operations.

function [y,n] = sigadd(x1,n1,x2,n2)

% implements y(n) = x1(n)+x2(n)
% -----------------------------
% [y,n] = sigadd(x1,n1,x2,n2)
% y = sum sequence over n, which includes n1 and n2
% x1 = first sequence over n1
% x2 = second sequence over n2 (n2 can be different from n1)
%
n = min(min(n1),min(n2)):max(max(n1),max(n2)); % duration of y(n)
y1 = zeros(1,length(n)); y2 = y1; % initialization
y1(find((n>=min(n1))&(n<=max(n1))==1))=x1; % x1 with duration of y
y2(find((n>=min(n2))&(n<=max(n2))==1))=x2; % x2 with duration of y
y = y1+y2; % sequence addition

Its use is illustrated in Example 2.2.

2. Signal multiplication: This is a sample-by-sample (or “dot”) multi-
plication) given by
{x1 (n)} · {x2 (n)} = {x1 (n)x2 (n)}
It is implemented in MATLAB by the array operator .*. Once again,
the similar restrictions apply for the .* operator as for the + operator.
Therefore we have developed the sigmult function, which is similar to
the sigadd function.

function [y,n] = sigmult(x1,n1,x2,n2)

% implements y(n) = x1(n)*x2(n)
% -----------------------------
% [y,n] = sigmult(x1,n1,x2,n2)
% y = product sequence over n, which includes n1 and n2
% x1 = first sequence over n1
% x2 = second sequence over n2 (n2 can be different from n1)
%

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 27

n = min(min(n1),min(n2)):max(max(n1),max(n2)); % duration of y(n)

y1 = zeros(1,length(n)); y2 = y1; %
y1(find((n>=min(n1))&(n<=max(n1))==1))=x1; % x1 with duration of y
y2(find((n>=min(n2))&(n<=max(n2))==1))=x2; % x2 with duration of y
y = y1 .* y2; % sequence multiplication

Its use is also given in Example 2.2.

3. Scaling: In this operation each sample is multiplied by a scalar α.
α {x(n)} = {αx(n)}
An arithmetic operator (*) is used to implement the scaling operation
in MATLAB.
4. Shifting: In this operation, each sample of x(n) is shifted by an
amount k to obtain a shifted sequence y(n).
y(n) = {x(n − k)}
If we let m = n−k, then n = m+k and the above operation is given by
y(m + k) = {x (m)}
Hence this operation has no effect on the vector x, but the vector n is
changed by adding k to each element. This is shown in the function
sigshift.

function [y,n] = sigshift(x,m,k)

% implements y(n) = x(n-k)
% -------------------------
% [y,n] = sigshift(x,m,k)
%
n = m+k; y = x;

Its use is given in Example 2.2.

5. Folding: In this operation each sample of x(n) is flipped around n = 0
to obtain a folded sequence y(n).
y(n) = {x(−n)}
In MATLAB this operation is implemented by fliplr(x)function for
sample values and by -fliplr(n) function for sample positions as
shown in the sigfold function.

function [y,n] = sigfold(x,n)

% implements y(n) = x(-n)
% -----------------------
% [y,n] = sigfold(x,n)
%
y = fliplr(x); n = -fliplr(n);

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 28 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

6. Sample summation: This operation differs from signal addition

operation. It adds all sample values of x(n) between n1 and n2 .
n2
$
x(n) = x(n1 ) + · · · + x(n2 )
n=n1

It is implemented by the sum(x(n1:n2)) function.

7. Sample products: This operation also differs from signal multi-
plication operation. It multiplies all sample values of x(n) between
n1 and n2 .
n2
%
x(n) = x(n1 ) × · · · × x(n2 )
n1

It is implemented by the prod(x(n1:n2)) function.

8. Signal energy: The energy of a sequence x(n) is given by
∞
$ ∞
$ 2
Ex = x(n)x∗ (n) = |x(n)|
−∞ −∞

where superscript ∗ denotes the operation of complex conjugation.1

The energy of a finite-duration sequence x(n) can be computed in
MATLAB using

>> Ex = sum(x .* conj(x)); % one approach

>> Ex = sum(abs(x) .^ 2); % another approach

9. Signal power: The average power of a periodic sequence x̃(n) with

fundamental period N is given by
N −1
1 $ 2
Px = |x̃(n)|
N 0

! EXAMPLE 2.1 Generate and plot each of the following sequences over the indicated interval.
a. x(n) = 2δ(n + 2) − δ(n − 4), −5 ≤ n ≤ 5.
b. x(n) = n[u(n)−u(n−10)]+10e−0.3(n−10) [u(n−10)−u(n−20)], 0 ≤ n ≤ 20.
c. x(n) = cos(0.04πn) + 0.2w(n), 0 ≤ n ≤ 50, where w(n) is a Gaussian
random sequence with zero mean and unit variance.
d. x̃(n) = {..., 5, 4, 3, 2, 1, 5, 4, 3, 2, 1, 5, 4, 3, 2, 1, ...}; −10 ≤ n ≤ 9.
↑

1 The symbol * denotes many operations in digital signal processing. Its font (roman

or computer) and its position (normal or superscript) will distinguish each operation.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 29

Solution a. x(n) = 2δ(n + 2) − δ(n − 4), −5 ≤ n ≤ 5.

>> n = [-5:5];
>> x = 2*impseq(-2,-5,5) - impseq(4,-5,5);
>> stem(n,x); title(’Sequence in Problem 2.1a’)
>> xlabel(’n’); ylabel(’x(n)’);

The plot of the sequence is shown in Figure 2.1a.

b. x(n) = n [u(n) − u(n − 10)] + 10e−0.3(n−10) [u(n − 10) − u(n − 20)], 0 ≤ n ≤
20.

>> n = [0:20]; x1 = n.*(stepseq(0,0,20)-stepseq(10,0,20));

>> x2 = 10*exp(-0.3*(n-10)).*(stepseq(10,0,20)-stepseq(20,0,20));
>> x = x1+x2;
>> subplot(2,2,3); stem(n,x); title(’Sequence in Problem 2.1b’)
>> xlabel(’n’); ylabel(’x(n)’);

The plot of the sequence is shown in Figure 2.1b.

Sequence in Example 2.1a Sequence in Example 2.1b

3
10
2
8
1 6
x(n)

x(n)
0 4

2
−1
0
−2
−5 0 5 0 5 10 15 20
n n
Sequence in Example 2.1c Sequence in Example 2.1d
6
1

0.5 4
xtilde(n)
x(n)

0
2
−0.5

−1 0

0 10 20 30 40 −10 −5 0 5
n n

FIGURE 2.1 Sequences in Example 2.1

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 30 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

c. x(n) = cos(0.04πn) + 0.2w(n), 0 ≤ n ≤ 50.

>> n = [0:50]; x = cos(0.04*pi*n)+0.2*randn(size(n));

>> subplot(2,2,2); stem(n,x); title(’Sequence in Problem 2.1c’)
>> xlabel(’n’); ylabel(’x(n)’);

The plot of the sequence is shown in Figure 2.1c.

d. x̃(n) = {..., 5, 4, 3, 2, 1, 5, 4, 3, 2, 1, 5, 4, 3, 2, 1, ...}; −10 ≤ n ≤ 9.
↑
Note that over the given interval, the sequence x̃ (n) has four periods.

>> n = [-10:9]; x = [5,4,3,2,1];

>> xtilde = x’ * ones(1,4); xtilde = (xtilde(:))’;
>> subplot(2,2,4); stem(n,xtilde); title(’Sequence in Problem 2.1d’)
>> xlabel(’n’); ylabel(’xtilde(n)’);

The plot of the sequence is shown in Figure 2.1d. !

! EXAMPLE 2.2 Let x(n) = {1, 2, 3, 4, 5, 6, 7, 6, 5, 4, 3, 2, 1}. Determine and plot the following
↑
sequences.
a. x1 (n) = 2x(n − 5) − 3x(n + 4)
b. x2 (n) = x(3 − n) + x(n) x(n − 2)

Solution The sequence x(n) is nonzero over −2 ≤ n ≤ 10. Hence

>> n = -2:10; x = [1:7,6:-1:1];

will generate x(n).

a. x1 (n) = 2x(n − 5) − 3x(n + 4).
The first part is obtained by shifting x(n) by 5 and the second part by
shifting x(n) by −4. This shifting and the addition can be easily done using
the sigshift and the sigadd functions.

>> [x11,n11] = sigshift(x,n,5); [x12,n12] = sigshift(x,n,-4);

>> [x1,n1] = sigadd(2*x11,n11,-3*x12,n12);
>> subplot(2,1,1); stem(n1,x1); title(’Sequence in Example 2.2a’)
>> xlabel(’n’); ylabel(’x1(n)’);

The plot of x1 (n) is shown in Figure 2.2a.

b. x2 (n) = x(3 − n) + x(n) x(n − 2).
The first term can be written as x(−(n − 3)). Hence it is obtained by first
folding x(n) and then shifting the result by 3. The second part is a multipli-
cation of x(n) and x(n − 2), both of which have the same length but different

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 31

Sequence in Example 2.2a

10
5
0

x1(n)
−5
−10
−15
−20
−6 0 15
n
Sequence in Example 2.2b
40

x2(n) 30

20

10

0
−7 0 12
n

FIGURE 2.2 Sequences in Example 2.2

support (or sample positions). These operations can be easily done using the
sigfold and the sigmult functions.

>> [x21,n21] = sigfold(x,n); [x21,n21] = sigshift(x21,n21,3);

>> [x22,n22] = sigshift(x,n,2); [x22,n22] = sigmult(x,n,x22,n22);
>> [x2,n2] = sigadd(x21,n21,x22,n22);
>> subplot(2,1,2); stem(n2,x2); title(’Sequence in Example 2.2b’)
>> xlabel(’n’); ylabel(’x2(n)’);

The plot of x2 (n) is shown in Figure 2.2b. !

Example 2.2 shows that the four sig* functions developed in this
section provide a convenient approach for sequence manipulations.

! EXAMPLE 2.3 Generate the complex-valued signal

x(n) = e(−0.1+j0.3)n , −10 ≤ n ≤ 10

and plot its magnitude, phase, the real part, and the imaginary part in four
separate subplots.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 32 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

Real Part Imaginary Part

2 1

1
0
0

−1
−1
−2

−3 −2
−10 −5 0 5 10 −10 −5 0 5 10
n n
Magnitude Part Phase Part
3 200

100
2

1
−100

0 −200
−10 −5 0 5 10 −10 −5 0 5 10
n n

FIGURE 2.3 Complex-valued sequence plots in Example 2.3

Solution MATLAB script:

>> n = [-10:1:10]; alpha = -0.1+0.3j;

>> x = exp(alpha*n);
>> subplot(2,2,1); stem(n,real(x));title(’real part’);xlabel(’n’)
>> subplot(2,2,2); stem(n,imag(x));title(’imaginary part’);xlabel(’n’)
>> subplot(2,2,3); stem(n,abs(x));title(’magnitude part’);xlabel(’n’)
>> subplot(2,2,4); stem(n,(180/pi)*angle(x));title(’phase part’);xlabel(’n’)

The plot of the sequence is shown in Figure 2.3. !

2.1.3 DISCRETE-TIME SINUSOIDS

In the last section we introduced the discrete-time sinusoidal sequence
x(n) = A cos(ω0 n + θ0 ), for all n as one of the basic signals. This signal
is very important in signal theory as a basis for Fourier transform and
in system theory as a basis for steady-state analysis. It can be conve-
niently related to the continuous-time sinusoid xa (t) = A cos(Ω0 t + θ0 )
using an operation called sampling (Chapter 3), in which continuous-time
sinusoidal values at equally spaced points t = nTs are assigned to x(n).

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 33

The quantity Ts is called the sampling interval, and Ω0 = ω0 /Ts is called

the analog frequency, measured in radians per second.
The fact that n is a discrete variable, whereas t is a continuous
variable, leads to some important differences between discrete-time and
continuous-time sinusoidal signals.

Periodicity in time From our definition of periodicity, the sinusoidal

sequence is periodic if
x[n + N ] = A cos(ω0 n + ω0 N + θ) = A cos(ω0 n + θ0 ) = x[n] (2.1)
This is possible if and only if ω0 N = 2πk, where k is an integer. This
leads to the following important result (see Problem P2.5):
%
The sequence x(n) = A cos(ω0 n + θ0 ) is periodic if and only if f0 =
ω0 /2π = k/N , that is, f0 is a rational number. If k and N are a
pair of prime numbers, then N is the fundamental period of x(n) and
k represents an integer number of periods kTs of the corresponding
continuous-time sinusoid.

Periodicity in frequency From the definition of the discrete-time si-

nusoid, we can easily see that
A cos[(ω0 + k2π)n + θ0 ] = A cos(ω0 n + kn2π + θ0 )
= A cos(ω0 n + θ0 )
since (kn)2π is always an integer multiple of 2π. Therefore, we have the
following property:
The sequence x(n) = A cos(ω0 n + θ) is periodic in ω0 with fundamen-
tal period 2π and periodic in f0 with fundamental period one.
This property has a number of very important implications:
1. Sinusoidal sequences with radian frequencies separated by integer mul-
tiples of 2π are identical.
2. All distinct sinusoidal sequences have frequencies within an interval of
2π radians. We shall use the so-called fundamental frequency ranges
−π < ω ≤ π or 0 ≤ ω < 2π (2.2)
Therefore, if 0 ≤ ω0 < 2π, the frequencies ω0 and ω0 + m2π are
indistinguishable from the observation of the corresponding sequences.
3. Since A cos[ω0 (n + n0 ) + θ] = A cos[ω0 n + (ω0 n0 + θ)], a time shift is
equivalent to a phase change.
4. The rate of oscillation of a discrete-time sinusoid increases as ω0 in-
creases from ω0 = 0 to ω0 = π. However, as ω0 increases from ω0 = π
to ω0 = 2π, the oscillations become slower. Therefore, low frequencies
(slow oscillations) are at the vicinity of ω0 = k2π, and high frequencies
(rapid oscillations) are at the vicinity of ω0 = π + k2π.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 34 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

2.1.4 SOME USEFUL RESULTS

There are several important results in discrete-time signal theory. We will
discuss some that are useful in digital signal processing.

Unit sample synthesis Any arbitrary sequence x(n) can be synthe-

sized as a weighted sum of delayed and scaled unit sample sequences, such
as ∞
$
x(n) = x(k)δ(n − k) (2.3)
k=−∞
We will use this result in the next section.

Even and odd synthesis A real-valued sequence xe (n) is called even

(symmetric) if
xe (−n) = xe (n)
Similarly, a real-valued sequence xo (n) is called odd (antisymmetric) if
xo (−n) = −xo (n)
Then any arbitrary real-valued sequence x(n) can be decomposed into its
even and odd components
x(n) = xe (n) + xo (n) (2.4)
where the even and odd parts are given by
1 1
xe (n) = [x(n) + x(−n)] and xo (n) = [x(n) − x(−n)] (2.5)
2 2
respectively. We will use this decomposition in studying properties of the
Fourier transform. Therefore it is a good exercise to develop a simple
MATLAB function to decompose a given sequence into its even and odd
components. Using MATLAB operations discussed so far, we can obtain
the following evenodd function.

function [xe, xo, m] = evenodd(x,n)

% Real signal decomposition into even and odd parts
% -------------------------------------------------
% [xe, xo, m] = evenodd(x,n)
%
if any(imag(x) ~= 0)
error(’x is not a real sequence’)
end
m = -fliplr(n);
m1 = min([m,n]); m2 = max([m,n]); m = m1:m2;
nm = n(1)-m(1); n1 = 1:length(n);
x1 = zeros(1,length(m)); x1(n1+nm) = x; x = x1;
xe = 0.5*(x + fliplr(x)); xo = 0.5*(x - fliplr(x));

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete-time Signals 35

Rectangular Pulse Even Part

1 1

0.8 0.8

xe(n)
x(n)
0.6 0.6

0.4 0.4

0.2 0.2

0 0
−10 −5 0 5 10 −10 −5 0 5 10
n n
Odd Part
0.6

0.4

0.2

xe(n)
0

−0.2

−0.4

−10 −5 0 5 10
n

FIGURE 2.4 Even-odd decomposition in Example 2.4

The sequence and its support are supplied in x and n arrays, respectively.
It first checks if the given sequence is real and determines the support
of the even and odd components in m array. It then implements (2.5)
with special attention to the MATLAB indexing operation. The resulting
components are stored in xe and xo arrays.

! EXAMPLE 2.4 Let x(n) = u(n) − u(n − 10). Decompose x(n) into even and odd components.

Solution The sequence x(n), which is nonzero over 0 ≤ n ≤ 9, is called a rectangular

pulse. We will use MATLAB to determine and plot its even and odd parts.

>> n = [0:10]; x = stepseq(0,0,10)-stepseq(10,0,10);

>> [xe,xo,m] = evenodd(x,n);
>> subplot(2,2,1); stem(n,x); title(’Rectangular pulse’)
>> xlabel(’n’); ylabel(’x(n)’); axis([-10,10,0,1.2])
>> subplot(2,2,2); stem(m,xe); title(’Even Part’)
>> xlabel(’n’); ylabel(’xe(n)’); axis([-10,10,0,1.2])
>> subplot(2,2,4); stem(m,xo); title(’Odd Part’)
>> xlabel(’n’); ylabel(’xe(n)’); axis([-10,10,-0.6,0.6])

The plots shown in Figure 2.4 clearly demonstrate the decomposition. !

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 36 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

A similar decomposition for complex-valued sequences is explored in

Problem P2.5.
The geometric series A one-sided exponential sequence of the form
{αn , n ≥ 0}, where α is an arbitrary constant, is called a geometric
series. In digital signal processing, the convergence and expression for the
sum of this series are used in many applications. The series converges for
|α| < 1, while the sum of its components converges to
$∞
1
αn −→ , for |α| < 1 (2.6)
n=0
1 − α
We will also need an expression for the sum of any finite number of terms
of the series given by
N
$ −1
1 − αN
αn = , ∀α (2.7)
n=0
1−α
These two results will be used throughout this book.
Correlations of sequences Correlation is an operation used in many
applications in digital signal processing. It is a measure of the degree to
which two sequences are similar. Given two real-valued sequences x(n) and
y(n) of finite energy, the crosscorrelation of x(n) and y(n) is a sequence
rxy (') defined as
∞
$
rx,y (') = x(n)y(n − ') (2.8)
n=−∞
The index ' is called the shift or lag parameter. The special case of (2.8)
when y(n) = x(n) is called autocorrelation and is defined by
$∞
rxx (') = x(n)x(n − ') (2.9)
n=−∞
It provides a measure of self-similarity between different alignments of the
sequence. MATLAB functions to compute auto- and crosscorrelations are
discussed later in the chapter.

2.2 DISCRETE SYSTEMS

Mathematically, a discrete-time system (or discrete system for short) is

described as an operator T [·] that takes a sequence x(n) (called excitation)
and transforms it into another sequence y(n) (called response). That is,
y(n) = T [x(n)]

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete Systems 37

In DSP we will say that the system processes an input signal into an output
signal. Discrete systems are broadly classified into linear and nonlinear
systems. We will deal mostly with linear systems.

2.2.1 LINEAR SYSTEMS

A discrete system T [·] is a linear operator L[·] if and only if L[·] satisfies
the principle of superposition, namely,

L[a1 x1 (n) + a2 x2 (n)] = a1 L[x1 (n)] + a2 L[x2 (n)], ∀a1 , a2 , x1 (n), x2 (n)
(2.10)

Using (2.3) and (2.10), the output y(n) of a linear system to an arbitrary
input x(n) is given by
& ∞ ' ∞
$ $
y(n) = L[x(n)] = L x(k) δ(n − k) = x(k)L[δ(n − k)]
n=−∞ n=−∞

The response L[δ(n − k)] can be interpreted as the response of a linear

system at time n due to a unit sample (a well-known sequence) at time k.
It is called an impulse response and is denoted by h(n, k). The output
then is given by the superposition summation
∞
$
y(n) = x(k)h(n, k) (2.11)
n=−∞

The computation of (2.11) requires the time-varying impulse response

h(n, k), which in practice is not very convenient. Therefore time-invariant
systems are widely used in DSP.

! EXAMPLE 2.5 Determine whether the following systems are linear:

1. y(n) = T [x(n)] = 3x2 (n)
2. y(n) = 2x(n − 2) + 5
3. y(n) = x(n + 1) − x(n − 1)
( ) ( )
Solution Let y1 (n) = T x1 (n) and y2 (n) = T x2 (n) . We will determine the
response of each system to the linear combination a1 x1 (n) + a2 x2 (n) and
check whether it is equal to the linear combination a1 x1 (n) + a2 x2 (n)
where a1 and a2 are arbitrary constants.
1. y(n) = T [x(n)] = 3x2 (n): Consider
( ) 2
T a1 x1 (n) + a2 x2 (n) = 3 [a1 x1 (n) + a2 x2 (n)]

= 3a21 x21 (n) + 3a22 x22 (n) + 6a1 a2 x1 (n)x2 (n)

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 38 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

which is not equal to

a1 y1 (n) + a2 y2 (n) = 3a21 x21 (n) + 3a22 x22 (n)

Hence the given system is nonlinear.

2. y(n) = 2x(n − 2) + 5: Consider
( )
T a1 x1 (n) + a2 x2 (n) = 2 [a1 x1 (n − 2) + a2 x2 (n − 2)] + 5
= a1 y1 (n) + a2 y2 (n) − 5

Clearly, the given system is nonlinear even though the input-output

relation is a straight-line function.
3. y(n) = x(n + 1) − x(1 − n): Consider

T [a1 x1 (n) + a2 x2 (n)] = a1 x1 (n + 1) + a2 x2 (n + 1) + a1 x1 (1 − n)

+ a2 x2 (1 − n)
= a1 [x1 (n + 1) − x1 (1 − n)]
+ a2 [x2 (n + 1) − x2 (1 − n)]
= a1 y1 (n) + a2 y2 (n)

Hence the given system is linear. !

Linear time-invariant (LTI) system A linear system in which an

input-output pair, x(n) and y(n), is invariant to a shift k in time is called
a linear time-invariant system i.e.,

y(n) = L[x(n)] ⇒ L[x(n − k)] = y(n − k) (2.12)

For an LTI system the L[·] and the shifting operators are reversible as
shown here.

x(n) −→ L [·] −→ y(n) −→ Shift by k −→ y(n − k)

x(n) −→ Shift by k −→ x(n − k) −→ L [·] −→ y(n − k)

! EXAMPLE 2.6 Determine whether the following linear systems are time-invariant.
1. y(n) = L[x(n)] = 10 sin(0.1πn)x(n)
2. y(n) = L[x(n)] = x(n + 1) − x(1 − n)
3. y(n) = L[x(n)] = 14 x(n) + 12 x(n − 1) + 14 x(n − 2)

%
Solution First we will compute the response yk (n) = L[x(n − k)] to the shifted
input sequence. This is obtained by subtracting k from the arguments of

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Discrete Systems 39

every input sequence term on the right-hand side of the linear transforma-
tion. To determine time-invariance, we will then compare it to the shifted
output sequence y(n − k), obtained after replacing every n by (n − k) on
the right-hand side of the linear transformation.
1. y(n) = L[x(n)] = 10 sin(0.1πn)x(n): The response due to shifted
input is

yk (n) = L[x(n − k)] = 10 sin(0.1πn)x(n − k)

while the shifted output is

y(n − k) = 10 sin[0.1π(n − k)]x(n − k) "= yk (n).

Hence the given system is not time-invariant.

2. y(n) = L[x(n)] = x(n + 1) − x(1 − n): The response due to shifted
input is

yk (n) = L[x(n − k)] = x(n − k) − x(1 − n − k)

while the shifted output is

y(n − k) = x(n − k) − x(1 − [n − k]) = x(n − k) − x(1 − n + k) "= yk (n).

Hence the given system is not time-invariant.

3. y(n) = L[x(n)] = 14 x(n) + 12 x(n − 1) + 14 x(n − 2): The response due
to shifted input is

yk (n) = L[x(n − k)] = 14 x(n − k) + 12 x(n − 1 − k) + 14 x(n − 2 − k)

while the shifted output is

y(n − k) = 14 x(n − k) + 12 x(n − k − 1) + 14 x(n − k − 2) = yk (n)

Hence the given system is time-invariant. !

We will denote an LTI system by the operator LT I [·]. Let x(n) and
y(n) be the input-output pair of an LTI system. Then the time-varying
function h(n, k) becomes a time-invariant function h(n − k), and the out-
put from (2.11) is given by
∞
$
y(n) = LT I [x(n)] = x(k)h(n − k) (2.13)
k=−∞

The impulse response of an LTI system is given by h(n). The mathemati-

cal operation in (2.13) is called a linear convolution sum and is denoted by
%
y(n) = x(n) ∗ h(n) (2.14)

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 40 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

Hence an LTI system is completely characterized in the time domain by

the impulse response h(n).

x(n) −→ h(n) −→ y(n) = x(n) ∗ h(n)

We will explore several properties of the convolution in Problem P2.14.

Stability This is a very important concept in linear system theory. The
primary reason for considering stability is to avoid building harmful sys-
tems or to avoid burnout or saturation in the system operation. A system
is said to be bounded-input bounded-output (BIBO) stable if every bounded
input produces a bounded output.
|x(n)| < ∞ ⇒ |y(n)| < ∞, ∀x, y
An LTI system is BIBO stable if and only if its impulse response is abso-
lutely summable.
∞
$
BIBO Stability ⇐⇒ |h(n)| < ∞ (2.15)
−∞

Causality This important concept is necessary to make sure that sys-

tems can be built. A system is said to be causal if the output at index n0
depends only on the input up to and including the index n0 ; that is, the
output does not depend on the future values of the input. An LTI system
is causal if and only if the impulse response
h(n) = 0, n<0 (2.16)
Such a sequence is termed a causal sequence. In signal processing, unless
otherwise stated, we will always assume that the system is causal.

2.3 CONVOLUTION

We introduced the convolution operation (2.14) to describe the response

of an LTI system. In DSP it is an important operation and has many other
uses that we will see throughout this book. Convolution can be evaluated
in many different ways. If the sequences are mathematical functions (of
finite or infinite duration), then we can analytically evaluate (2.14) for all
n to obtain a functional form of y(n).

! EXAMPLE 2.7 Let the rectangular pulse x(n) = u(n) − u(n − 10) of Example 2.4 be an input
to an LTI system with impulse response
h(n) = (0.9)n u(n)
Determine the output y(n).

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Convolution 41

Input Sequence
2

1.5

x(n)
1

0.5

0
−5 0 5 10 15 20 25 30 35 40 45
n
Impulse Response
2

h(n) 1.5

0.5

0
−5 0 5 10 15 20 25 30 35 40 45
n

FIGURE 2.5 The input sequence and the impulse response in Example 2.7

Solution The input x(n) and the impulse response h(n) are shown in Figure 2.5. From
(2.14)

$
9
$
9

y(n) = (1) (0.9)(n−k) u(n − k) = (0.9)n (0.9)−k u(n − k) (2.17)

k=0 k=0

The sum in (2.17) is almost a geometric series sum except that the term u(n−k)
takes different values depending on n and k. There are three possible conditions
under which u(n − k) can be evaluated.

CASE i n < 0: Then u(n − k) = 0, 0 ≤ k ≤ 9. Hence from (2.17)

y(n) = 0 (2.18)
CASE ii In this case the nonzero values of x(n) and h(n) do not overlap.
0 ≤ n < 9: Then u(n − k) = 1, 0 ≤ k ≤ n. Hence from (2.17)

$
n
$
n

y(n) = (0.9)n (0.9)−k = (0.9)n [(0.9)−1 ]k

k=0 k=0

1 − (0.9)−(n+1)
= (0.9)n = 10[1 − (0.9)n+1 ], 0≤n<9 (2.19)
1 − (0.9)−1

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 42 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

Output Sequence
8

y(n)
4

0
−5 0 5 10 15 20 25 30 35 40 45
n

FIGURE 2.6 The output sequence in Example 2.7

CASE iii In this case the impulse response h(n) partially overlaps the input x(n).
n ≥ 9: Then u(n − k) = 1, 0 ≤ k ≤ 9 and from (2.17)
$
9

y(n) = (0.9)n (0.9)−k

k=0
− (0.9)−10 n1 (2.20)
= (0.9) = 10(0.9)n−9 [1 − (0.9)10 ], n≥9
1 − (0.9)−1
In this last case h(n) completely overlaps x(n).

The complete response is given by (2.18), (2.19), and (2.20). It is shown in

Figure 2.6 which depicts the distortion of the input pulse. !

This example can also be done using a method called graphical convo-
lution, in which (2.14) is given a graphical interpretation. In this method,
h(n − k) is interpreted as a folded-and-shifted version of h(k). The output
y(n) is obtained as a sample sum under the overlap of x(k) and h(n − k).
We use an example to illustrate this.

! EXAMPLE 2.8 Given the following two sequences

x(n) = [3, 11, 7, 0, −1, 4, 2], −3 ≤ n ≤ 3; h(n) = [2, 3, 0, −5, 2, 1], −1 ≤ n ≤ 4

↑ ↑

determine the convolution y(n) = x(n) ∗ h(n).

Solution In Figure 2.7 we show four plots. The top-left plot shows x(k) and h(k), the
original sequences. The top-right plot shows x(k) and h(−k), the folded version
of h(k). The bottom-left plot shows x(k) and h(−1 − k), the folded-and-shifted-
by- −1 version of h(k). Then
$
x(k)h(−1 − k) = 3 × (−5) + 11 × 0 + 7 × 3 + 0 × 2 = 6 = y(−1)
k

The bottom-right plot shows x(k) and h(2 − k), the folded-and-shifted-by-2
version of h(k), which gives
$
x(k)h(2 − k) = 11 × 1 + 7 × 2 + 0 × (−5) + (−1) × 0 + 4 × 3 + 2 × 2 = 41 = y(2)
k

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Convolution 43

x(k) and h(k) x(k) and h(−k)

Solid: x Dashed: h Solid: x Dashed: h
10 10

5 5

0 0
n=0

−5 −5
−5 0 5 −5 0 5
k k
x(k) and h(−1−k) x(k) and h(2−k)
Solid: x Dashed: h Solid: x Dashed: h
10 10

5 5

0 0
n=−1 n=2

−5 −5
−5 0 5 −5 0 5
k k

FIGURE 2.7 Graphical convolution in Example 2.8

Thus we have obtained two values of y(n). Similar graphical calculations can
be done for other remaining values of y(n). Note that the beginning point (first
nonzero sample) of y(n) is given by n = −3 + (−1) = −4, while the end point
(the last nonzero sample) is given by n = 3 + 4 = 7. The complete output is
given by

y(n) = {6, 31, 47, 6, −51, −5, 41, 18, −22, −3, 8, 2}
↑

Students are strongly encouraged to verify the above result. Note that the re-
sulting sequence y(n) has a longer length than both the x(n) and h(n) sequences.
!

2.3.1 MATLAB IMPLEMENTATION

If arbitrary sequences are of infinite duration, then MATLAB cannot
be used directly to compute the convolution. MATLAB does provide a
built-in function called conv that computes the convolution between two
finite-duration sequences. The conv function assumes that the two se-
quences begin at n = 0 and is invoked by

>> y = conv(x,h);

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 44 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

For example, to do the convolution in Example 2.7, we could use

>> x = [3, 11, 7, 0, -1, 4, 2]; h = [2, 3, 0, -5, 2, 1];

>> y = conv(x, h)
y =
6 31 47 6 -51 -5 41 18 -22 -3 8 2

to obtain the correct y(n) values. However, the conv function neither
provides nor accepts any timing information if the sequences have arbi-
trary support. What is needed is a beginning point and an end point of
y(n). Given finite duration x(n) and h(n), it is easy to determine these
points. Let
{x(n); nxb ≤ n ≤ nxe } and {h(n); nhb ≤ n ≤ nhe }
be two finite-duration sequences. Then referring to Example 2.8 we ob-
serve that the beginning and end points of y(n) are
nyb = nxb + nhb and nye = nxe + nhe
respectively. A simple modification of the conv function, called conv m,
which performs the convolution of arbitrary support sequences can now
be designed.

function [y,ny] = conv_m(x,nx,h,nh)

% Modified convolution routine for signal processing
% --------------------------------------------------
% [y,ny] = conv_m(x,nx,h,nh)
% [y,ny] = convolution result
% [x,nx] = first signal
% [h,nh] = second signal
%
nyb = nx(1)+nh(1); nye = nx(length(x)) + nh(length(h));
ny = [nyb:nye]; y = conv(x,h);

! EXAMPLE 2.9 Perform the convolution in Example 2.8 using the conv m function.

Solution MATLAB script:

>> x = [3, 11, 7, 0, -1, 4, 2]; nx = [-3:3];

>> h = [2, 3, 0, -5, 2, 1]; ny = [-1:4];

>> [y,ny] = conv_m(x,nx,h,nh)

y =
6 31 47 6 -51 -5 41 18 -22 -3 8 2
ny =
-4 -3 -2 -1 0 1 2 3 4 5 6 7

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Convolution 45

Hence

y(n) = {6, 31, 47, 6, −51, −5, 41, 18, −22, −3, 8, 2}
↑

as in Example 2.8. !

An alternate method in MATLAB can be used to perform the convo-

lution. This method uses a matrix-vector multiplication approach, which
we will explore in Problem P2.17.

2.3.2 SEQUENCE CORRELATIONS REVISITED

If we compare the convolution operation (2.14) with that of the crosscor-
relation of two sequences defined in (2.8), we observe a close resemblance.
The crosscorrelation ryx (') can be put in the form

ryx (') = y(') ∗ x(−')

with the autocorrelation rxx (') in the form

rxx (') = x(') ∗ x(−')

Therefore these correlations can be computed using the conv m function

if sequences are of finite duration.
! EXAMPLE 2.10 In this example we will demonstrate one application of the crosscorrelation
sequence. Let

x(n) = [3, 11, 7, 0, −1, 4, 2]

↑

be a prototype sequence, and let y(n) be its noise-corrupted-and-shifted version

y(n) = x(n − 2) + w(n)

where w(n) is Gaussian sequence with mean 0 and variance 1. Compute the
crosscorrelation between y(n) and x(n).

Solution From the construction of y(n) it follows that y(n) is “similar” to x(n − 2) and
hence their crosscorrelation would show the strongest similarity at # = 2. To
test this out using MATLAB, let us compute the crosscorrelation using two
different noise sequences.

% noise sequence 1
>> x = [3, 11, 7, 0, -1, 4, 2]; nx=[-3:3]; % given signal x(n)
>> [y,ny] = sigshift(x,nx,2); % obtain x(n-2)
>> w = randn(1,length(y)); nw = ny; % generate w(n)
>> [y,ny] = sigadd(y,ny,w,nw); % obtain y(n) = x(n-2) + w(n)
>> [x,nx] = sigfold(x,nx); % obtain x(-n)

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 46 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

Crosscorrelation: Noise Sequence 1

250

200 Maximum

150

rxy
100

50

−50
−4 −2 0 2 4 6 8
Lag Variable l
Crosscorrelation: Noise Sequence 2

200 Maximum
150
rxy

100

50

−50
−4 −2 0 2 4 6 8
Lag Variable l

FIGURE 2.8 Crosscorrelation sequence with two different noise realizations

>> [rxy,nrxy] = conv_m(y,ny,x,nx); % crosscorrelation

>> subplot(1,1,1), subplot(2,1,1);stem(nrxy,rxy)
>> axis([-5,10,-50,250]);xlabel(’lag variable l’)
>> ylabel(’rxy’);title(’Crosscorrelation: noise sequence 1’)
%
% noise sequence 2
>> x = [3, 11, 7, 0, -1, 4, 2]; nx=[-3:3]; % given signal x(n)
>> [y,ny] = sigshift(x,nx,2); % obtain x(n-2)
>> w = randn(1,length(y)); nw = ny; % generate w(n)
>> [y,ny] = sigadd(y,ny,w,nw); % obtain y(n) = x(n-2) + w(n)
>> [x,nx] = sigfold(x,nx); % obtain x(-n)
>> [rxy,nrxy] = conv_m(y,ny,x,nx); % crosscorrelation
>> subplot(2,1,2);stem(nrxy,rxy)
>> axis([-5,10,-50,250]);xlabel(’lag variable l’)
>> ylabel(’rxy’);title(’Crosscorrelation: noise sequence 2’)

From Figure 2.8 we observe that the crosscorrelation indeed peaks at # = 2,

which implies that y(n) is similar to x(n) shifted by 2. This approach can be
used in applications like radar signal processing in identifying and localizing
targets. !

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Difference Equations 47

Note that the signal-processing toolbox in MATLAB also provides a

function called xcorr for sequence correlation computations. In its sim-
plest form

>> xcorr(x,y)

computes the crosscorrelation between vectors x and y, while

>> xcorr(x)

computes the autocorrelation of vector x. It generates results that are

identical to the one obtained from the proper use of the conv m function.
However, the xcorr function cannot provide the timing (or lag) informa-
tion (as done by the conv m function), which then must be obtained by
some other means.

2.4 DIFFERENCE EQUATIONS

An LTI discrete system can also be described by a linear constant coeffi-

cient difference equation of the form
N
$ M
$
ak y(n − k) = bm x(n − m), ∀n (2.21)
k=0 m=0

If aN "= 0, then the difference equation is of order N . This equation de-

scribes a recursive approach for computing the current output, given the
input values and previously computed output values. In practice this equa-
tion is computed forward in time, from n = −∞ to n = ∞. Therefore
another form of this equation is
M
$ N
$
y(n) = bm x(n − m) − ak y(n − k) (2.22)
m=0 k=1

A solution to this equation can be obtained in the form

y(n) = yH (n) + yP (n)
The homogeneous part of the solution, yH (n), is given by
N
$
yH (n) = ck zkn
k=1

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 48 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

where zk , k = 1, . . . , N are N roots (also called natural frequencies) of the

characteristic equation
$N
ak z k = 0
0
This characteristic equation is important in determining the stability of
systems. If the roots zk satisfy the condition
|zk | < 1, k = 1, . . . , N (2.23)
then a causal system described by (2.22) is stable. The particular part
of the solution, yP (n), is determined from the right-hand side of (2.21).
In Chapter 4 we will discuss the analytical approach of solving difference
equations using the z-transform.

2.4.1 MATLAB IMPLEMENTATION

A function called filter is available to solve difference equations nu-
merically, given the input and the difference equation coefficients. In its
simplest form this function is invoked by

y = filter(b,a,x)

where

b = [b0, b1, ..., bM]; a = [a0, a1, ..., aN];

are the coefficient arrays from the equation given in (2.21), and x is the
input sequence array. The output y has the same length as input x. One
must ensure that the coefficient a0 not be zero.
To compute and plot impulse response, MATLAB provides the func-
tion impz. When invoked by

h = impz(b,a,n);

it computes samples of the impulse response of the filter at the sample

indices given in n with numerator coefficients in b and denominator co-
efficients in a. When no output arguments are given, the impz function
plots the response in the current figure window using the stem function.
We will illustrate the use of these functions in the following example.

! EXAMPLE 2.11 Given the following difference equation

y(n) − y(n − 1) + 0.9y(n − 2) = x(n); ∀n

a. Calculate and plot the impulse response h(n) at n = −20, . . . , 100.

b. Calculate and plot the unit step response s(n) at n = −20, . . . , 100.
c. Is the system specified by h(n) stable?

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Difference Equations 49

Solution From the given difference equation the coefficient arrays are

b = [1]; a=[1, -1, 0.9];

a. MATLAB script:

>> b = [1]; a = [1, -1, 0.9]; n = [-20:120];

>> h = impz(b,a,n);
>> subplot(2,1,1); stem(n,h);
>> title(’Impulse Response’); xlabel(’n’); ylabel(’h(n)’)

The plot of the impulse response is shown in Figure 2.9.

b. MATLAB script:

>> x = stepseq(0,-20,120); s = filter(b,a,x);

>> subplot(2,1,2); stem(n,s)
>> title(’Step Response’); xlabel(’n’); ylabel(’s(n)’)

The plot of the unit step response is shown in Figure 2.9.

c. To determine the stability of the system, we have to determine h(n) for all n.
Although we have not described a method to solve the difference equation,

Impulse Response
1

0.5
h(n)

−0.5

−1
−20 0 20 40 60 80 100 120
n
Step Response

1.5
s(n)

0.5

−0.5
−20 0 20 40 60 80 100 120
n

FIGURE 2.9 Impulse response and step response plots in Example 2.11

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 50 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

we can use the plot of the impulse*response to observe that h(n) is practically
zero for n > 120. Hence the sum |h(n)| can be determined from MATLAB
using

>> sum(abs(h))
ans = 14.8785

which implies that the system is stable. An alternate approach is to use the
stability condition (2.23) using MATLAB’s roots function.

>>z = roots(a); magz = abs(z)

magz = 0.9487
0.9487

Since the magnitudes of both roots are less than one, the system is stable.
!
In the previous section we noted that if one or both sequences in
the convolution are of infinite length, then the conv function cannot be
used. If one of the sequences is of infinite length, then it is possible to use
MATLAB for numerical evaluation of the convolution. This is done using
the filter function as we will see in the following example.

! EXAMPLE 2.12 Let us consider the convolution given in Example 2.7. The input sequence is of
finite duration
x(n) = u(n) − u(n − 10)
while the impulse response is of infinite duration
h(n) = (0.9)n u(n)
Determine y(n) = x(n) ∗ h(n).

Solution If the LTI system, given by the impulse response h(n), can be described by a
difference equation, then y(n) can be obtained from the filter function. From
the h(n) expression
(0.9) h(n − 1) = (0.9) (0.9)n−1 u(n − 1) = (0.9)n u(n − 1)
or
h(n) − (0.9) h(n − 1) = (0.9)n u(n) − (0.9)n u(n − 1)
= (0.9)n [u(n) − u(n − 1)] = (0.9)n δ(n)
= δ(n)
The last step follows from the fact that δ(n) is nonzero only at n = 0. By
definition h(n) is the output of an LTI system when the input is δ(n). Hence
substituting x(n) for δ(n) and y(n) for h(n), the difference equation is
y(n) − 0.9y(n − 1) = x(n)

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Difference Equations 51

Output Sequence
8

y(n)
4

0
−5 0 5 10 15 20 25 30 35 40 45
n

FIGURE 2.10 Output sequence in Example 2.12

Now MATLAB’s filter function can be used to compute the convolution in-
directly.

>> b = [1]; a = [1,-0.9];

>> n = -5:50; x = stepseq(0,-5,50) - stepseq(10,-5,50);
>> y = filter(b,a,x);
>> subplot(2,1,2); stem(n,y); title(’Output sequence’)
>> xlabel(’n’); ylabel(’y(n)’); axis([-5,50,-0.5,8])

The plot of the output is shown in Figure 2.10, which is exactly the same as
that in Figure 2.6. !

In Example 2.12 the impulse response was a one-sided exponential se-

quence for which we could determine a difference equation representation.
This means that not all infinite-length impulse responses can be converted
into difference equations. The above analysis, however, can be extended to
a linear combination of one-sided exponential sequences, which results in
higher-order difference equations. We will discuss this topic of conversion
from one representation to another one in Chapter 4.

2.4.2 ZERO-INPUT AND ZERO-STATE RESPONSES

In digital signal processing the difference equation is generally solved for-
ward in time from n = 0. Therefore initial conditions on x(n) and y(n)
are necessary to determine the output for n ≥ 0. The difference equation
is then given by

M
$ N
$
y(n) = bm x(n − m) − ak y(n − k); n ≥ 0 (2.24)
m=0 k=1

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 52 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

subject to the initial conditions:

{y(n); −N ≤ n ≤ −1} and {x(n); −M ≤ n ≤ −1}

A solution to (2.24) can be obtained in the form

y(n) = yZI (n) + yZS (n)

where yZI (n) is called the zero-input solution, which is a solution due
to the initial conditions alone (assuming they exist), while the zero-state
solution, yZS (n), is a solution due to input x(n) alone (or assuming that
the initial conditions are zero). In MATLAB another form of the function
filter can be used to solve for the difference equation, given its initial
conditions. We will illustrate the use of this form in Chapter 4.

2.4.3 DIGITAL FILTERS

Filter is a generic name that means a linear time-invariant system designed
for a specific job of frequency selection or frequency discrimination. Hence
discrete-time LTI systems are also called digital filters. There are two
types of digital filters.

FIR filter If the unit impulse response of an LTI system is of finite

duration, then the system is called a finite-duration impulse response (or
FIR) filter. Hence for an FIR filter h(n) = 0 for n < n1 and for n > n2 .
The following part of the difference equation (2.21) describes a causal FIR
filter:
M
$
y(n) = bm x(n − m) (2.25)
m=0

Furthermore, h(0) = b0 , h(1) = b1 , . . . , h(M ) = bM , while all other h(n)’s

are 0. FIR filters are also called nonrecursive or moving average (MA)
filters. In MATLAB FIR filters are represented either as impulse response
values {h(n)} or as difference equation coefficients {bm } and {a0 = 1}.
Therefore to implement FIR filters, we can use either the conv(x,h)
function (and its modification that we discussed) or the filter(b,1,x)
function. There is a difference in the outputs of these two implementations
that should be noted. The output sequence from the conv(x,h) function
has a longer length than both the x(n) and h(n) sequences. On the other
hand, the output sequence from the filter(b,1,x) function has exactly
the same length as the input x(n) sequence. In practice (and especially
for processing signals) the use of the filter function is encouraged.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Problems 53

IIR filter If the impulse response of an LTI system is of infinite dura-

tion, then the system is called an infinite-duration impulse response (or
IIR) filter. The following part of the difference equation (2.21):
N
$
ak y(n − k) = x(n) (2.26)
k=0

describes a recursive filter in which the output y(n) is recursively com-

puted from its previously computed values and is called an autoregressive
(AR) filter. The impulse response of such filter is of infinite duration and
hence it represents an IIR filter. The general equation (2.21) also describes
an IIR filter. It has two parts: an AR part and an MA part. Such an IIR
filter is called an autoregressive moving average, or an ARMA, filter. In
MATLAB, IIR filters are described by the difference equation coefficients
{bm } and {ak } and are implemented by the filter(b,a,x) function.

2.5 PROBLEMS

P2.1 Generate the following sequences using the basic MATLAB signal functions and the basic
MATLAB signal operations discussed in this chapter. Plot signal samples using the stem
function.
1. x1 (n) = 3δ(n + 2) + 2δ(n) − δ(n − 3) + 5δ(n − 7), −5 ≤ n ≤ 15.
*5
2. x2 (n) = k=−5
e−|k| δ(n − 2k), −10 ≤ n ≤ 10.
3. x3 (n) = 10u(n) − 5u(n − 5) − 10u(n − 10) + 5u(n − 15).
4. x4 (n) = e0.1n [u(n + 20) − u(n − 10)].
5. x5 (n) = 5[cos(0.49πn) + cos(0.51πn)], −200 ≤ n ≤ 200. Comment on the waveform
shape.
6. x6 (n) = 2 sin(0.01πn) cos(0.5πn), −200 ≤ n ≤ 200. Comment on the waveform shape.
7. x7 (n) = e−0.05n sin(0.1πn + π/3), 0 ≤ n ≤ 100. Comment on the waveform shape.
8. x8 (n) = e0.01n sin(0.1πn), 0 ≤ n ≤ 100. Comment on the waveform shape.

P2.2 Generate the following random sequences and obtain their histogram using the hist
function with 100 bins. Use the bar function to plot each histogram.
1. x1 (n) is a random sequence whose samples are independent and uniformly distributed
over [0, 2] interval. Generate 100,000 samples.
2. x2 (n) is a Gaussian random sequence whose samples are independent with mean 10 and
variance 10. Generate 10,000 samples.
3. x3 (n) = x1 (n) + x1 (n − 1) where x1 (n) is the random sequence given in part 1 above.
Comment *on the shape of this histogram and explain the shape.
4
4. x4 (n) = k=1 yk (n) where each random sequence yk (n) is independent of others with
samples uniformly distributed over [−0.5, 0.5]. Comment on the shape of this histogram.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 54 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

P2.3 Generate the following periodic sequences and plot their samples (using the stem function)
over the indicated number of periods.
1. x̃1 (n) = {. . . , −2, −1, 0, 1, 2, . . .}periodic . Plot 5 periods.
↑
2. x̃2 (n) = e0.1n [u(n) − u(n − 20]periodic . Plot 3 periods.
3. x̃3 (n) = sin(0.1πn)[u(n) − u(n − 10)]. Plot 4 periods.
4. x̃4 (n) = {. . . , 1, 2, 3, . . .}periodic + {. . . , 1, 2, 3, 4, . . .}periodic , 0 ≤ n ≤ 24. What is the
↑ ↑
period of x̃4 (n)?

P2.4 Let x(n) = {2, 4, −3, 1, −5, 4, 7}. Generate and plot the samples (use the stem function) of
↑
the following sequences.
1. x1 (n) = 2x(n − 3) + 3x(n + 4) − x(n)
2. x2 (n) = 4x(4 + n) + 5x(n + 5) + 2x(n)
3. x3 (n) = x(n + 3)x(n − 2) + x(1 − n)x(n + 1)
4. x4 (n) = 2e0.5n x(n) + cos (0.1πn) x (n + 2) , −10 ≤ n ≤ 10

P2.5 The complex exponential sequence ejω0 n or the sinusoidal sequence cos (ω0 n) are periodic if
# ω0 K
the normalized frequency f0 = is a rational number; that is, f0 = , where K and N
2π N
are integers.
1. Prove the above result.
2. Generate exp(0.1πn), −100 ≤ n ≤ 100. Plot its real and imaginary parts using the stem
function. Is this sequence periodic? If it is, what is its fundamental period? From the
examination of the plot what interpretation can you give to the integers K and N above?
3. Generate and plot cos(0.1n), −20 ≤ n ≤ 20. Is this sequence periodic? What do you
conclude from the plot? If necessary examine the values of the sequence in MATLAB to
arrive at your answer.
P2.6 Using the evenodd function, decompose the following sequences into their even and odd
components. Plot these components using the stem function.
1. x1 (n) = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}.
↑
2. x2 (n) = e0.1n [u(n + 5) − u(n − 10)].
3. x3 (n) = cos(0.2πn + π/4), −20 ≤ n ≤ 20.
4. x4 (n) = e−0.05n sin(0.1πn + π/3), 0 ≤ n ≤ 100.
P2.7 A complex-valued sequence xe (n) is called conjugate-symmetric if xe (n) = x∗e (−n) and a
complex-valued sequence xo (n) is called conjugate-antisymmetric if xo (n) = −x∗o (−n).
Then, any arbitrary complex-valued sequence x(n) can be decomposed into
x(n) = xe (n) + xo (n) where xe (n) and xo (n) are given by
1 1
xe (n) = [x(n) + x∗ (−n)] and xo (n) = [x(n) − x∗ (−n)] (2.27)
2 2
respectively.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Problems 55

1. Modify the evenodd function discussed in the text so that it accepts an arbitrary
sequence and decomposes it into its conjugate-symmetric and conjugate-antisymmetric
components by implementing (2.27).
2. Decompose the following sequence:

x(n) = 10 exp([−0.1 + 0.2π]n), 0 ≤ n ≤ 10

into its conjugate-symmetric and conjugate-antisymmetric components. Plot their real

and imaginary parts to verify the decomposition. (Use the subplot function.)
P2.8 The operation of signal dilation (or decimation or down-sampling) is defined by

y(n) = x(nM )

in which the sequence x(n) is down-sampled by an integer factor M . For example, if

x(n) = {. . . , −2, 4, 3, −6, 5, −1, 8, . . .}

↑

then the down-sampled sequences by a factor 2 are given by

y(n) = {. . . , −2, 3, 5, 8, . . .}
↑

1. Develop a MATLAB function dnsample that has the form

function [y,m] = dnsample(x,n,M)

% Downsample sequence x(n) by a factor M to obtain y(m)

to implement the above operation. Use the indexing mechanism of MATLAB with
careful attention to the origin of the time axis n = 0.
2. Generate x(n) = sin(0.125πn), − 50 ≤ n ≤ 50. Decimate x(n) by a factor of 4 to
generate y(n). Plot both x(n) and y(n) using subplot and comment on the results.
3. Repeat the above using x(n) = sin(0.5πn), − 50 ≤ n ≤ 50. Qualitatively discuss the
effect of down-sampling on signals.
P2.9 Using the conv_m function, determine the autocorrelation sequence rxx (#) and the
crosscorrelation sequence rxy (#) for the following sequences.

x(n) = (0.9)n , 0 ≤ n ≤ 20; y(n) = (0.8)−n , − 20 ≤ n ≤ 0

Describe your observations of these results.

P2.10 In a certain concert hall, echoes of the original audio signal x(n) are generated due to the
reflections at the walls and ceiling. The audio signal experienced by the listener y(n) is a
combination of x(n) and its echoes. Let

y(n) = x(n) + αx(n − k)

where k is the amount of delay in samples and α is its relative strength. We want to
estimate the delay using the correlation analysis.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 56 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

1. Determine analytically the crosscorrelation ryx (#) in terms of the autocorrelation rxx (#).
2. Let x(n) = cos(0.2πn) + 0.5 cos(0.6πn), α = 0.1, and k = 50. Generate 200 samples of
y(n) and determine its crosscorrelation. Can you obtain α and k by observing ryx (#)?

P2.11 Consider the following discrete-time systems:

T1 [x(n)] = x(n)u(n) T2 [x(n)] = x(n) + n x(n + 1)

1 1 *n+5
T3 [x(n)] = x(n) + x(n − 2) − x(n − 3)x(2n) T4 [x(n)] = k=−∞
2x(k)
2 3
T5 [x(n)] = x(2n) T6 [x(n)] = round[x(n)]

where round[·] denotes rounding to the nearest integer.

1. Use (2.10) to determine analytically whether these systems are linear.
2. Let x1 (n) be a uniformly distributed random sequence between [0, 1] over 0 ≤ n ≤ 100,
and let x2 (n) be a Gaussian random sequence with mean 0 and variance 10 over
0 ≤ n ≤ 100. Using these sequences, verify the linearity of these systems. Choose any
values for constants a1 and a2 in (2.10). You should use several realizations of the above
sequences to arrive at your answers.

P2.12 Consider the discrete-time systems given in Problem P2.11.

1. Use (2.12) to determine analytically whether these systems are time-invariant.
2. Let x(n) be a Gaussian random sequence with mean 0 and variance 10 over 0 ≤ n ≤ 100.
Using this sequence, verify the time invariance of the above systems. Choose any values
for sample shift k in (2.12). You should use several realizations of the above sequence to
arrive at your answers.
P2.13 For the systems given in Problem P2.11, determine analytically their stability and causality.
P2.14 The linear convolution defined in (2.14) has several properties:

x1 (n) ∗ x2 (n) = x1 (n) ∗ x2 (n) : Commutation

[x1 (n) ∗ x2 (n)] ∗ x3 (n) = x1 (n) ∗ [x2 (n) ∗ x3 (n)] : Association
(2.28)
x1 (n) ∗ [x2 (n) + x3 (n)] = x1 (n) ∗ x2 (n) + x1 (n) ∗ x3 (n) : Distribution
x(n) ∗ δ(n − n0 ) = x(n − n0 ) : Identity

1. Analytically prove these properties.

2. Using the following three sequences, verify the above properties.

x1 (n) = cos(πn/4)[u(n + 5) − u(n − 25)]

x2 (n) = (0.9)−n [u(n) − u(n − 20)]
x3 (n) = round[5w(n)], −10 ≤ n ≤ 10; where w(n) is uniform over [−1, 1]

Use the conv m function.

P2.15 Determine analytically the convolution y(n) = x(n) ∗ h(n) of the following sequences, and
verify your answers using the conv_m function.
1. x(n) = {2, −4, 5, 3, −1, −2, 6}, h(n) = {1, −1, 1, −1, 1}
↑ ↑

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 Problems 57

2. x(n) = {1, 1, 0, 1, 1}, h(n) = {1, −2, −3, 4}

↑ ↑
3. x(n) = (1/4)−n [u(n + 1) − u(n − 4)], h(n) = u(n) − u(n − 5)
4. x(n) = n/4[u(n) − u(n − 6)], h(n) = 2[u(n + 2) − u(n − 3)]
P2.16 Let x(n) = (0.8)n u(n), h(n) = (−0.9)n u(n), and y(n) = h(n) ∗ x(n). Use 3 columns and 1
row of subplots for the following parts.
1. Determine y(n) analytically. Plot first 51 samples of y(n) using the stem function.
2. Truncate x(n) and h(n) to 26 samples. Use conv function to compute y(n). Plot y(n)
using the stem function. Compare your results with those of part 1.
3. Using the filter function, determine the first 51 samples of x(n) ∗ h(n). Plot y(n) using
the stem function. Compare your results with those of parts 1 and 2.
P2.17 When the sequences x(n) and h(n) are of finite duration Nx and Nh , respectively, then
their linear convolution (2.13) can also be implemented using matrix-vector multiplication.
If elements of y(n) and x(n) are arranged in column vectors x and y respectively, then from
(2.13) we obtain

y = Hx

where linear shifts in h(n − k) for n = 0, . . . , Nh − 1 are arranged as rows in the matrix H.
This matrix has an interesting structure and is called a Toeplitz matrix. To investigate this
matrix, consider the sequences

x(n) = {1, 2, 3, 4, 5} and h(n) = {6, 7, 8, 9}

↑ ↑

1. Determine the linear convolution y(n) = h(n) ∗ x(n).

2. Express x(n) as a 5 × 1 column vector x and y(n) as a 8 × 1 column vector y. Now
determine the 8 × 5 matrix H so that y = Hx.
3. Characterize the matrix H. From this characterization can you give a definition of a
Toeplitz matrix? How does this definition compare with that of time invariance?
4. What can you say about the first column and the first row of H?
P2.18 MATLAB provides a function called toeplitz to generate a Toeplitz matrix, given the first
row and the first column.
1. Using this function and your answer to Problem P2.17, part 4, develop another MATLAB
function to implement linear convolution. The format of the function should be

function [y,H]=conv_tp(h,x)
% Linear Convolution using Toeplitz Matrix
% ----------------------------------------
% [y,H] = conv_tp(h,x)
% y = output sequence in column vector form
% H = Toeplitz matrix corresponding to sequence h so that y = Hx
% h = Impulse response sequence in column vector form
% x = input sequence in column vector form

2. Verify your function on the sequences given in Problem P2.17.

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require
 58 Chapter 2 DISCRETE-TIME SIGNALS AND SYSTEMS

P2.19 A linear and time-invariant system is described by the difference equation

y(n) − 0.5y(n − 1) + 0.25y(n − 2) = x(n) + 2x(n − 1) + x(n − 3)

1. Using the filter function, compute and plot the impulse response of the system over
0 ≤ n ≤ 100.
2. Determine the stability of the system from this impulse response.
3. If the input to this system is x(n) = [5 + 3 cos(0.2πn) + 4 sin(0.6πn)] u(n), determine the
response y(n) over 0 ≤ n ≤ 200 using the filter function.
P2.20 A “simple” digital differentiator is given by

y(n) = x(n) − x(n − 1)

which computes a backward first-order difference of the input sequence. Implement this
differentiator on the following sequences, and plot the results. Comment on the
appropriateness of this simple differentiator.
1. x(n) = 5 [u(n) − u(n − 20)]: a rectangular pulse
2. x(n) = n [u(n) − u(n − 10)] + (20 − n) [u(n − 10) − u(n − 20)]: a triangular pulse
+ ,
πn
3. x(n) = sin [u(n) − u(n − 100)]: a sinusoidal pulse
25

Copyright 2010 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require