2.

61 Digital Signal Processing

2.9.5 Methods of Performing Linear Convolution

Method 1: Graphical Method

Let x1(n) and x2(n) be the input sequences and x3(n) be the output sequence.

1. Change the index "n" of input sequences to "m" to get x1(m) and x2(m).
2. Sketch the graphical representation of the input sequences x1(m) and x2(m).
3. Let us fold x 2 (m) to get x 2(–m). Sketch the graphical representation of the folded
sequence x2(–m).
4. Shift the folded sequence x2(–m) to the left graphically so that the product of x1(m) and
shifted x2(–m) gives only one nonzero sample. Now multiply x1(m) and shifted x2(–m) to get
a product sequence, and then sum up the samples of product sequence, which is the first
sample of output sequence.
5. To get the next sample of output sequence, shift x2(–m) of previous step to one position right
and multiply the shifted sequence with x1(m) to get a product sequence. Now the sum of the
samples of product sequence gives the second sample of output sequence.
2. To get subsequent samples of output sequence, the step 5 is repeated until we get a nonzero
product sequence.
Method 2: Tabular Method
The tabular method is same as that of graphical method, except that the tabular representation of the
sequences are employed instead of graphical representation. In tabular method, every input sequence, folded
and shifted sequence is represented by a row in a table.
Method 3: Matrix Method
Let x1(n) and x2(n) be the input sequences and x3(n) be the output sequence. In matrix method one of
the sequences is represented as a row and the other as a column as shown below.
Multiply each column element with row elements and fill up the matrix array.
Now the sum of the diagonal elements gives the samples of output sequence x3(n). (The sum of the
diagonal elements are shown below for reference).
x 2 (0) x 2 (1) x 2 (2) x 2 (3)

x 1 (0) x 1 (0) x 2 (0) x 1 (0) x 2 (1) x 1 (0) x 2 (2) x 1 (0) x 2 (3)

x 1 (1) x 1 (1) x 2 (0) x 1 (1) x 2 (1) x 1 (1) x 2 (2) x 1 (1) x 2 (3)

x 1 (2) x 1 (2) x 2 (0) x 1 (2) x 2 (1) x 1 (2) x 2 (2) x 1 (2) x 2 (3)

x 1 (3) x 1 (3) x 2 (0) x 1 (3) x 2 (1) x 1 (3) x 2 (2) x 1 (3) x 2 (3)

Chapter 2 - Discrete Time Signals and Systems 2. 62

......

x3(0) = ..... + x1(0) x2(0) + .....

x3(1) = ..... + x1(1) x2(0) + x1(0 ) x2(1) + .....
x3(2) = ..... + x1(2) x2(0) + x1(1) x2(1) + x1(0) x2(2) + .....
x3(3) = ..... + x1(3) x2(0) + x1(2) x2(1) + x1(1) x2(2) + x1(0) x2(3) + .....
......

\ x3(n) = {..... x3(0), x3(1), x3(2), x3(3), .....}

Example 2.22
Determine the response of the LTI system whose input x(n) and impulse response h(n) are given by,
x(n) = {1, 2, 0.5, 1} and h(n) = {1, 2, 1, –1}
­ - -

Solution
The response y(n) of the system is given by convolution of x(n) and h(n).
+∞
y(n) = x(n) ∗ h(n) = ∑
m = −∞
x(m) h(n − m)

In this example the convolution operation is performed by three methods.

The Input sequence starts at n = 0 and the impulse response sequence starts at n = –1. Therefore the
output sequence starts at n = 0 + (–1) = –1.
The input and impulse response consists of 4 samples, so the output consists of 4 + 4 – 1 = 7 samples.
Method 1 : Graphical Method
The graphical representation of x(n) and h(n) after replacing n by m are shown below. The sequence h(m)
is folded with respect to m = 0 to obtain h(–m).

x (m ) h (m ) h ( −m )

2 2 2

1 1 1 1 1 1
0.5

0 1 2 3 m −1 0 1 2 m −2 −1 0 1 m
−1 −1
F ig 1 : In p u t sequ e n ce. F ig 2 : Im p u lse resp o n se . F ig 3 : F o ld ed im p ulse respo nse.

The samples of y(n) are computed using the convolution formula,

+∞ +∞
y(n) = ∑
m = −∞
x(m) h(n − m) = ∑
m = −∞
x(m) hn (m) ; where hn (m) = h(n − m)

The computation of each sample using the above equation are graphically shown in fig 4 to fig 10. The
graphical representation of output sequence is shown in fig 11.
2. 63 Digital Signal Processing
+∞ +∞ +∞
When n = −1 ; y(−1) = ∑
m = −∞
x(m) h( −1 − m) = ∑
m = −∞
x(m) h−1(m) = ∑
m = −∞
v −1(m)

h −1 (m ) x (m ) v −1 (m )

2
X 2 ⇒
1 1 1 1 1
0.5

−3 −2 −1 0 m 0 1 2 3 m −3 −2 −1 0 1 2 3 m
T he s um of pro du c t s e qu en c e v −1(m )
−1 F ig 4 : C o m p u ta tio n of y ( −1 ).
g iv e s y ( −1 ). ∴ y ( −1) = 1
+∞ +∞ +∞
When n = 0 ; y(0) = ∑
m = −∞
x(m) h(0 − m) = ∑
m = −∞
x(m) h0 (m) = ∑
m = −∞
v 0 (m)

h 0 (m ) x (m ) v 0 (m )

2 2 2
X 2 ⇒
1 1 1 1
0.5

−2 −1 0 1 m 0 1 2 3 m −2 −1 0 1 2 3 m
T h e s u m o f p rod uc t s eq ue nc e v 0 ( m )
−1 F ig 5 : C o m p u ta tion of y (0 ).
giv es y(0 ) . ∴ y (0) = 2 + 2 = 4
+∞ +∞ +∞
When n = 1 ; y(1) = ∑
m = −∞
x(m) h(1 − m) = ∑
m = −∞
x(m) h1(m) = ∑
m = −∞
v1(m)

h 1 (m ) x (m ) v 1 (m )
4

X ⇒
2 2

1 1 1 1 1
0.5 0.5
−1 0 1 2 m 0 1 2 3 m −1 0 1 2 3 m
−1 T he s um of pro du c t s e qu en c e v 1 (m )
F ig 6 : C o m p u ta tio n of y (1 ).
g iv e s y (1). ∴ y(1 ) = 1 + 4 + 0.5 = 5 .5
+∞ +∞ +∞
When n = 2 ; y(2) = ∑
m = −∞
x(m) h(2 − m) = ∑
m = −∞
x(m) h2 (m) = ∑
m = −∞
v 2 (m)

h 2 (m ) x (m ) v 2 (m )

X ⇒
2 2
2
1 1 1 1 1 1
0.5
0 1 2 3 m 0 1 2 3 m 0 1 2 3 m
−1
−1 T he s um of pro du c t s e qu en ce v 2 (m )
F ig 7 : C o m p u ta tio n of y (2 ). g iv e s y (2). ∴ y (2 ) = −1 + 2 + 1 + 1 = 3
Chapter 2 - Discrete Time Signals and Systems 2. 64
+∞ +∞ +∞
When n = 3 ; y(3) = ∑
m = −∞
x(m) h(3 − m) = ∑
m = −∞
x(m) h3 (m) = ∑
m = −∞
v 3 (m)

h 3 (m ) x (m ) v 3 (m )

X ⇒
2 2
2
1 1 1 1
0.5 0.5
1 2 3 4 m 2 m
0 0 1 3 0 1 2 3 4 m
−1
−2
F ig 8 : C o m p u ta tio n of y (3 ). T he su m o f pro du ct s e qu en ce v 3 (m )
giv es y (3). ∴ y (3) = −2 + 0.5 + 2 = 0 .5
+∞ +∞ +∞
When n = 4 ; y(4) = ∑
m = −∞
x(m) h(4 − m) = ∑
m = −∞
x(m) h4 (m) = ∑
m = −∞
v 4 (m)

h 4 (m ) x (m ) v 4 (m )

2 X ⇒
2
1 1
1 1 1
0.5
0 1 2 3 4 5 m
−1
0 1 2 3 m 0 1 2 3 4 5 m
−0.5
T h e s u m o f p rod uc t s eq ue nc e v 4 (m )
F ig 9 : C o m p u ta tio n of y (4 ).
give s y(4 ). ∴ y (4 ) = −0 .5 + 1 = 0 .5

+∞ +∞ +∞
When n = 5 ; y(5) = ∑
m = −∞
x(m) h(5 − m) = ∑
m = −∞
x(m ) h5 (m) = ∑
m = −∞
v 5 (m)

h 5 (m ) x (m ) v 5 (m )

2 X 2
⇒
1 1 1 1
0.5
0 1 2 3 4 5 6 m 0 1 2 3 m 0 1 2 3 4 5 6 m
−1
−1
F ig 1 0 : C o m p u ta tio n of y (5 ). T h e s um of p rod uc t s eq ue nc e v 5 (m )
g ive s y(5 ). ∴ y (5 ) = −1

y (n )
The output sequence, y(n) = {1, 4, 5.5, 3, 0.5, 0.5, − 1
A
} 5.5

1
0.5 0.5

−2 −1 0 1 2 3 4 5 6 7 n
−1
F ig 11 : G ra ph ica l rep resen ta tio n of y (n ).
2. 65 Digital Signal Processing
Method 2 : Tabular Method
The given sequences and the shifted sequences can be represented in the tabular array as shown below.

Note : The unfilled boxes in the table are considered as zeros.

m –3 –2 –1 0 1 2 3 4 5 6
x(m) 1 2 0.5 1
h(m) 1 2 1 –1
h(–m) –1 1 2 1
h(–1 – m) = h–1(m) –1 1 2 1
h(0 – m) = h0(m) –1 1 2 1
h(1 – m) = h1(m) –1 1 2 1
h(2 – m) = h2(m) –1 1 2 1
h(3 – m) = h3(m) –1 1 2 1
h(4 – m) = h4(m) –1 1 2 1
h(5 – m) = h5(m) –1 1 2 1
Each sample of y(n) is computed using the convolution formula,
+∞ +∞
y(n) = ∑ x(m) h(n − m)
m = −∞
= ∑ x(m) h (m),
m = −∞
n where hn (m) = h(n − m)

To determine a sample of y(n) at n = q, multiply the sequence x(m) and hq(m) to get a product sequence
(i.e., multiply the corresponding elements of the row x(m) and hq(m)). The sum of all the samples of the product
sequence gives y(q).
3
When n = −1 ; y( −1) = ∑
m = −3
x(m) h −1(m) Q The product is valid only for m = −3 to + 3.

= x(–3) h–1(–3) + x(–2)h–1(–2) + x(–1)h–1(–1) + x(0) h–1(0) + x(1) h–1(1)

+ x(2) h–1(2) + x(3) h–1(3)
=0+0+0+1+0+0+0=1
The samples of y(n) for other values of n are calculated as shown for n = –1.
3
When n = 0 ; y(0) = ∑ x(m) h0 (m) = 0 + 0 + 2 + 2 + 0 + 0 = 4
m = −2
3
When n = 1 ; y(1) = ∑ x(m) h1(m) = 0 + 1+ 4 + 0.5 + 0 = 5.5
m = −1
3
When n = 2 ; y(2) = ∑ x(m) h2(m) = −1+ 2 + 1+ 1= 3
m = 0
4
When n = 3 ; y(3) = ∑ x(m) h3 (m) = 0 − 2 + 0.5 + 2 + 0 = 0.5
m = 0
5
When n = 4 ; y(4) = ∑ x(m) h4 (m) = 0 + 0 − 0.5 + 1+ 0 + 0 = 0.5
m = 0
6
When n = 5 ; y(5) = ∑ x(m) h5 (m) = 0 + 0 + 0 − 1+ 0 + 0 + 0 = −1
m = 0

The output sequence, y(n) = l 1, 4, 5.5, 3, 0.5, 0.5, − 1q

A
Chapter 2 - Discrete Time Signals and Systems 2. 66
Method 3 : Matrix Method
The input sequence x(n) is arranged as a column and the impulse response is arranged as a row as shown
below. The elements of the two-dimensional array are obtained by multiplying the corresponding row element
with the column element. The sum of the diagonal elements gives the samples of y(n).

h(n) h(n)
x(n) 1 2 1 −1 x(n) 1 2 1 −1

1 1 2 1 −1
1 1 ×1 1 ×2 1 ×1 1 × (−1)
2 2 4 2 −2
2 2 ×1 2×2 2 ×1 2 ×(−1) ⇒
0.5 0.5 1 0.5 −0.5
0.5 0.5 × 1 0.5 × 2 0.5 × 1 0.5 × (−1)
1 1 2 1 −1
1 1 ×1 1×2 1 ×1 1 × (−1)

y(–1) = 1 y(3) = 2 + 0.5 + (–2) = 0.5

y(0) = 2 + 2 = 4 y(4) = 1 + (–0.5) = 0.5 \ y(n) = {1, 4, 5.5, 3, 0.5, 0.5, –1}
-
y(1) = 0.5 + 4 + 1 = 5.5 y(5) = –1
y(2) = 1 + 1 + 2 + (–1) = 3

Example 2.23
Determine the output y(n) of a relaxed LTI system with impulse response,
h(n) = an u(n) ; where |a| < 1 and
When input is a unit step sequence, i.e., x(n) = u(n).

Solution
The graphical representation of x(n) and h(n) after replacing n by m are shown below. Also the sequence
x(m) is folded to get x(–m).

h (m ) x (m ) x ( −m )
1 1 1
a
2
a
3
a

0 1 2 3 m 0 1 2 3 m −3 −2 −1 0 m
F ig 1 : Im p u lse resp o n se. F ig 2 : Im p u lse sequ en ce. F ig 3 : F o ld ed inp u t sequ en ce.

Here both h(m) and x(m) are infinite duration sequences starting at n = 0. Hence the output sequence y(n)
will also be an infinite duration sequence starting at n = 0.
By convolution formula,
∞ ∞
y(n) = ∑ h(m) x(n − m) = ∑ h(m) x (m) ;
m = −∞ m =0
n where xn (m) = x(n − m)

The computation of some samples of y(n) using the above equation are graphically shown below.
2. 67 Digital Signal Processing
∞ ∞ ∞
When n = 0 ; y(0) = ∑
m = 0
h(m) x(0 − m) = ∑
m = 0
h(m) x 0 (m) = ∑
m = 0
v 0 (m)

h (m ) x 0 (m ) v 0 (m )

1 1 1
a
a
2
X ⇒
3
a

0 1 2 3 m −3 −2 −1 m 0 1 2
0 1 m
F ig 4 : C o m p u ta tio n of y (0 ). y (0) = 1

∞ ∞ ∞
When n = 1 ; y(1) = ∑
m = 0
h(m) x(1 − m) = ∑
m = 0
h(m) x 1(m) = ∑
m = 0
v1(m)

h (m ) x 1 (m ) v 1 (m )

1 1 1
a a
a
2 X ⇒
3
a

0 1 2 3 m −2 −1 m
0 1 −1 0 1 2 m
y (1) = 1 + a
F ig 5 : C o m p u ta tio n of y (1 ).
∞ ∞ ∞
When n = 2 ; y(2) = ∑
m = 0
h(m) x(2 − m) = ∑
m = 0
h(m) x 2 (m) = ∑
m = 0
v 2 (m)

h (m ) x 2 (m ) v 2 (m )
1 1 1
a a
2
a ⇒ a
2

a
3 X

0 1 2 3 m −1 0 1 2 m 0 1 2 3 m
2
y(2) = 1 + a +a
F ig 6 : C o m p u ta tio n of y (2 ).
Solving similarly for other values of n, we can write y(n) for any value of n as shown below.
n
y(n) = 1 + a + a 2 +......+ an = ∑a
p=0
p
; for n ≥ 0

y (n )
3
1+a+a +a
2
1 + a + a2
1+a

0 1 2 3 m
F ig 7 : G ra ph ica l rep rese n ta tio n o f y (n).
Chapter 2 - Discrete Time Signals and Systems 2. 68

2.10 Circular Convolution

2.10.1 Circular Representation and Circular Shift of Discrete Time Signal
Consider a finite duration sequence x(n) and its periodic extension xp(n). The periodic extension of x(n)
can be expressed as xp(n) = x(n + N), where N is the periodicity. Let N = 4. The sequence x(n) and its periodic
extension are shown in fig 2.24.

Let, x(n) = 1 ; n=0

=2; n=1
=3; n=2
=4; n=3
x p (n )
x (n )
4 4 4 4

3 3 3 3

2 2 2 2

1 1 1 1

0 1 2 3 n −4 −3 −2 −1 0 1 2 3 4 5 6 7 n
F ig 2.2 4 a : F in ite d u ra tio n seq u en c e x(n ). F ig 2.2 4 b : P erio d ic e xten sio n o f x (n ).
F ig 2.2 4 : A fin ite d u ra tion seq ue n ce a n d its p erio d ic e xten sio n .

Let us delay the periodic sequence xp(n) by two units of time as shown in fig 2.25(a). (For delay the
sequence is shifted right). Let us denote one period of this delayed sequence by x1(n). One period of the
delayed sequence is shown in fig 2.25(b).
x p (n −2 )
x 1 (n ) x1 (n) = xp ((n − 2))4
4 4 4 4

3 3 3 3

2 2 2 2
1 1 1 1

−2 −1 0 1 2 3 4 5 6 7 8 9 n 0 1 2 3 n
F ig 2 .2 5 a: x p (n ) d e la y ed b y tw o un its o f tim e. F ig 2 .2 5 b: O ne period of x p (n −2 ).
F ig 2 .2 5 : D elay ed versio n o f x p (n).
The sequence x1(n) can be represented by xp(n – 2, (mod 4)), or xp((n – 2))4, where mod 4 indicates that
the sequence repeats after 4 samples. The relation between the original sequence x(n) and one period of the
delayed sequence x1(n) are shown below.

x1(n) = xp(n – 2, (mod 4)) = xp((n – 2))4

\ When n = 0; x1(0) = xp((0 – 2))4 = xp((– 2))4 = x(2) = 3

When n = 1; x1(1) = xp((1 – 2))4 = xp((– 1))4 = x(3) = 4

When n = 2; x1(2) = xp((2 – 2))4 = xp((0))4 = x(0) = 1

When n = 3; x1(3) = xp((3 – 2))4 = xp((1))4 = x(1) = 2

2. 69 Digital Signal Processing
The periodic sequences xp(n) and x1(n) can be represented as points on a circle as shown in fig 2.26.
From fig 2.26 we can say that, x1(n) is simply xp(n) shifted circularly by two units in time, where the counter
clockwise (anticlockwise) direction has been arbitrarily selected for right shift or delay.
2 1 4

x (1) = 2 x 1 (1) = 4
3 1 ⇒2 4 ⇒1 3

x (n) 4 3 2
x (2) = 3 x (0) = 1 x 1 (2) = 1 x 1 (n) x 1 (0) = 3
R otate x p(n) antic loc kw ise tw o tim es to get x 1(n)
= x p ((n − 2)) 4

x (3) = 4 x 1 (3) = 2
F ig 2.2 6 a: C ircu la r rep rese n ta tio n o f x (n). F ig 2.2 6 b: C ircu la r rep rese n ta tio n o f x 1 (n ).
F ig 2.2 6 : C ircu la r rep rese n ta tio n o f a sig n a l a nd its dela yed version .
Let us advance the periodic sequence xp(n) by three units of time as shown in fig 2.27(a). Let us denote
one period of this advanced sequence by x2(n). One period of the advanced sequence is shown in fig 2.27(b).

x p (n + 3 ) x 2 (n ) x 2 (n) = x p ((n + 3)) 4

4 4 4 4

3 3 3 3

2 2 2 2
1 1 1 1

−3 −2 −1 0 1 2 3 4 5 6 7 8 n 0 1 2 3 n
F ig 2 .2 7 a: x p (n ) a d v an c ed by th ree u n its o f tim e. F ig 2 .2 7 b: O ne p e rio d of x p (n + 3 ).
F ig 2 .2 7 : A d va n c ed v ersion of x p (n ).
The sequence x2(n) can be represented by xp(n + 3, (mod 4)) or xp((n + 3))4, where mod 4 indicates that the
sequence repeats after 4 samples. The relation between the original sequence x(n) and one period of the
advanced sequence x2(n) are shown below.
x2(n) = xp(n + 3, (mod 4)) = xp((n + 3))4
\ When n = 0; x2(0) = xp((0 + 3))4 = xp((3))4 = x(3) = 4
When n = 1; x2(1) = xp((1 + 3))4 = xp((4))4 = x(0) = 1
When n = 2; x2(2) = xp((2 + 3))4 = xp((5))4 = x(1) = 2
When n = 3; x2(3) = xp((3 + 3))4 = xp((6))4 = x(2) = 3
The periodic sequences xp(n) and x2(n) can be represented as points on a circle as shown in fig 2.28.
From fig 2.28 we can say that x2(n) is simply xp(n) shifted circularly by three units in time where clockwise
direction has been selected for left shift or advance.
2 3 4 1
x (1) = 2 x 2 (1) = 1
3 1 4 2 ⇒1 3 ⇒2 4

x p (n) 4 1 2 3 x 2 (n)
x (2) = 3 x (0) = 1 x 2 (2) = 2 x 2 (0) = 4
R otate x p(n) cloc k w is e three tim es to get x 2(n)

x (3) = 4 x 2 (3) = 3
F ig 2.2 8 a: C ircu la r rep rese nta tio n o f x (n). F ig 2.2 8 b: C ircu la r rep rese nta tio n o f x 2 (n ).
F ig 2.2 8 : C ircu la r rep rese nta tio n o f a sig n a l a n d its ad v a nced versio n.
Chapter 2 - Discrete Time Signals and Systems 2. 70
Thus we conclude that a circular shift of an N-point sequence is equivalent to a linear shift of its
periodic extension and viceversa. If a nonperiodic N-point sequence is represented on the circumference of
a circle then it becomes a periodic sequence of periodicity N. When the sequence is shifted circularly, the
samples repeat after N shifts. This is similar to modulo-N operation. Hence, in general, the circular shift may
be represented by the index mod-N. Let x(n) be an N-point sequence represented on a circle and x¢(n) be its
circularly shifted sequence by m units of time.
Now, x¢(n) = x(n – m, mod N) º x((n – m))N ..... (2.53)
When m is positive, the equation (2.53) represents delayed sequence and when m is negative, the
equation (2.53) represents advanced sequence.

2.10.2 Circular Symmetries of Discrete Time Signal

The circular representation of a sequence and the resulting periodicity gives rise to new definitions for
even symmetry, odd symmetry and the time reversal of the sequence.
An N-point sequence is called even if it is symmetric about the point zero on the circle. This implies
that,
x(N - n) = x(n) ; for 0 £ n £ N - 1 ...... (2.54)
An N-point sequence is called odd if it is antisymmetric about the point zero on the circle.
This implies that,
x(N - n) = - x(n) ; for 0 £ n £ N - 1 ...... (2.55)
The time reversal of a N-point sequence is obtained by reversing its sample about the point zero on the
circle. Thus the sequence x(–n, (mod N) ) is simply written as,
x (-n, (mod N)) = x(N - n) ; for 0 £ n £ N - 1 ...... (2.56)
This time reversal is equivalent to plotting x(n) in a clockwise direction on a circle, as shown in
fig 2.29.
x (2) x (6)

x (1) x (5) x (7)

x (3)

x (n) x (0) x ( −n)

x (4) x (4) x (0)

x (5) x (7) x (3) x (1)

x (6) x (2)
F ig 2.2 9 : C ircu la r rep rese n ta tio n o f a n 8 -p o in t seq ue n ce a n d its fo ld ed seq u en ce.
2.10.3 Definition of Circular Convolution
The circular convolution of two periodic discrete time sequences x1(n) and x2(n) with periodicity of
N samples is defined as,
N −1 N −1
.....(2.57)
x3 ( n) = ∑ x1( m) x2 (( n − m)) N or x 3 ( n) = ∑ x2 ( m) x1(( n − m)) N
m=0 m= 0

where, x3(n) is the sequence obtained by circular convolution,

x1((n – m))N represents circular shift of x1(n)
x2((n – m))N represents circular shift of x2(n)
m is a dummy variable.
2. 71 Digital Signal Processing
The output sequence x3(n) obtained by circular convolution is also a periodic sequence with periodicity
of N samples. Hence this convolution is also called periodic convolution.
The convolution relation of equation (2.57) can be symbolically expressed as
x3(n) = x1(n) * x2(n) = x2(n) * x1(n) ..... (2.58)
where, the symbol * indicates circular convolution operation.
The circular convolution is defined for periodic sequences. But circular convolution can be performed
with nonperiodic sequences by periodically extending them.The circular convolution of two sequences
requires that, at least one of the sequences should be periodic. Hence it is sufficient if one of the sequences
is periodically extended in order to perform circular convolution.
The circular convolution of finite duration sequences can be performed only if both the sequences
consist of the same number of samples. If the sequences have different number of samples, then convert the
smaller size sequence to the length of larger size sequence by appending zeros.
Circular convolution basically involves the same four steps as that for linear convolution, namely,
folding one sequence, shifting the folded sequence, multiplying the two sequences and finally summing the
values of the product sequence. Like linear convolution, any one of the sequence is folded and rotated in
circular convolution.
The difference between the two is that in circular convolution the folding and shifting (rotating)
operations are performed in a circular fashion by computing the index of one of the sequences by modulo-N
operation. In linear convolution there is no modulo-N operation.
2.10.4 Procedure for Evaluating Circular Convolution
Let, x1(n) and x2(n) be periodic discrete time sequences with periodicity of N-samples. If x1(n) and
x2(n) are non-periodic then convert the sequences to N-sample sequences and periodically extend the sequence
x2(n) with periodicity of N-samples.
Now the circular convolution of x1(n) and x2(n) will produce a periodic sequence x3(n) with periodicity
of N-samples. The samples of one period of x3(n) can be computed using the equation (2.57). The value of
x3(n) at n = q is obtained by replacing n by q, in equation (2.57).
N −1
.....(2.59)
∴ x3 (q ) = ∑ x1( m) x2 ((q − m)) N
m=0
The evaluation of equation (2.59) to determine the value of x3(n) at n = q involves the following five
steps.
1. Change of index : Change the index n in the sequences x1(n) and x2(n), in order to get the
sequences x1(m) and x2(m). Represent the samples of one period of the
sequences on circles.
2. Folding : Fold x2(m) about m = 0, to obtain x2(-m).
3. Rotation : Rotate x2(-m) by q times in anti-clockwise if q is positive, rotate x2(-m) by
q times in clockwise if q is negative to obtain x2((q – m))N.
4. Multiplication : Multiply x1(m) by x2((q – m))N to get a product sequence. Let the product
sequence be vq(m). Now, vq(m) = x1(m) × x2((q – m))N.
5. Summation : Sum up the samples of one period of the product sequence vq(m) to
obtain the value of x3(n) at n = q. [i.e., x3(q)].
The above procedure will give the value of x3(n) at a single time instant say n = q. In general we are
interested in evaluating the values of the sequence x3(n) in the range 0 < n < N -1. Hence the steps 3 , 4 and
5 given above must be repeated, for all possible time shifts in the range 0 < n < N - 1.
Chapter 2 - Discrete Time Signals and Systems 2. 72

2.10.5 Linear Convolution via Circular Convolution

When two numbers of N-point sequences are circularly convolved, it produces another N-point
sequence. For circular convolution, one of the sequence should be periodically extended. Also the resultant
sequence is periodic with period N.
The linear convolution of two sequences of length N1 and N2 produces an output sequence of length
N1 + N2 -1. To perform linear convolution via circular convolution both the sequences should be converted
to N 1 + N 2 - 1 point sequences by padding with zeros. Then perform circular convolution of
N1 + N2 -1 point sequences. The resultant sequence will be same as that of linear convolution of N1 and N2
point sequences.

2.10.6 Methods of Computing Circular Convolution

Method 1 : Graphical Method
In graphical method, the given sequences are converted to same size and represented on circles. In
case of periodic sequences, the samples of one period are represented on circles. One of the sequence is
folded and shifted circularly. Let x1(n) and x2(n) be the given sequences. Let x3(n) be the sequence obtained by
circular convolution of x1(n) and x2(n). The following procedure can be used to get a sample of x3(n) at n = q.
1. Change the index n in the sequences x1(n) and x2(n) to get x1(m) and x2(m) and then represent the
sequences on circles.
2. Fold one of the sequence. Let us fold x2(m) to get x2(–m).
3. Rotate (or shift) the sequence x2(–m), q times to get the sequence x2((q – m))N. If q is positive then
rotate (or shift) the sequence in anticlockwise direction and if q is negative then rotate (or shift) the
sequence in clockwise direction.
4. The sample of x3(q) at n = q is given by,
N −1 N −1
x3 ( q ) = ∑ x1 ( m) x2 (( q − m)) N = ∑ x1( m) x2,q ( m)
m=0 m= 0

where, x2, q(m) = x2((q – m))N

Determine the product sequence x1 ( m) x 2,q ( m) for one period.
5. The sum of all the samples of the product sequence gives the sample x3(q) [i.e., x3(n) at n = q].
The above procedure is repeated for all possible values of n to get the sequence x3(n).
Method 2 : Tabular Method
Let x1(n) and x2(n) be the given N-point sequences. Let x3(n) be the N-point sequence obtained by
circular convolution of x1(n) and x2(n). The following procedure can be used to obtain one sample of x3(n)
at n = q.
1. Change the index n in the sequences x1(n) and x2(n) to get x1(m) and x2(m) and then represent the
sequences as two rows of tabular array.
2. Fold one of the sequence. Let us fold x2(m) to get x2(–m).
3. Periodically extend x2(–m). Here the periodicity is N, where N is the length of the given sequences.
4. Shift the sequence x2(–m), q times to get the sequence x2((q – m))N. If q is positive then shift the
sequence to the right and if q is negative then shift the sequence to the left.
2. 73 Digital Signal Processing
N −1 N −1
5. The sample of x3(q) at n = q is given by, x3 (q) = ∑ x1 (m) x 2 ((q − m)) N = ∑ x1(m) x2,q (m)
m=0 m=0
where x2, q(m) = x2((q – m))N
Determine the product sequence x1 ( m) x 2,q ( m) for one period.
6. The sum of the samples of the product sequence gives the sample x3(q) [i.e., x3(n) at n = q].
The above procedure is repeated for all possible values of n to get the sequence x3(n).
Method 3: Matrix Method
Let x1(n) and x2(n) be the given N-point sequences. The circular convolution of x1(n) and x2(n) yields
another N-point sequence x3(n).
In this method an (N ´ N) matrix is formed using one of the sequence as shown below. Another
sequence is arranged as a column vector (column matrix) of order (N ´ 1). The product of the two matrices
gives the resultant sequence x3(n).
LMx (0) x ( N − 1)
2 2 x2 ( N − 2) ..... x 2 ( 2) x2 (1) OP LM xx ((10)) OP LM xx ((10)) OP
1 3

MMx (1) x (0)

2 2 x2 ( N − 1) ..... x 2 ( 3) x 2 ( 2) PP MM x (2) PP MM x (2) PP
1

1
3

3
MMx M(2) x (1M)
2 2 x 2 ( 0)
M
..... x 2 ( 4)
M
x (3)
2
M
PP × MM M PP = MM M PP
MMx (N − 2) x ( N − 3) x2 ( N − 4) ..... x 2 ( 0) x ( N − 1) P
MM M PP MM M PP
2 2 2
MNx (N − 1) x ( N − 2) x 2 ( N − 3) ..... x 2 (1) x (0)
PP Mx (N − 2)P M x ( N − 2)P
1 3
2 2 2 Q MNx (N − 1) PQ MN x ( N − 1) PQ
1 3

Example 2.24
Perform circular convolution of the two sequences, x1(n) = {2, 1, 2, –1} and x2(n)= {1, 2, 3, 4}
- -
Solution
Method 1:Graphical Method of Computing Circular Convolution
Let x3(n) be the sequence obtained by circular convolution of x1(n) and x2(n).
The circular convolution of x1(n) and x2(n) is given by,
N − 1 N − 1
x3(n) = ∑
m = 0
x1(m) x 2((n − m))N = ∑ x (m) x
m = 0
1 2,n (m)

where x 2,n (m) = x 2((n − m))N and m is the dummy variable used for convolution.
The index n in the given sequences are changed to m and each sequence is represented as points on a
circle as shown below. The folded sequence x2(–m) and circularly shifted sequences x2(n– m) are also represented
on the circle.
x 1 (1) = 1 x 2 (1) = 2 x 2 (3) = 4

x 1 (2) = 2 x 1 (m ) x 1 (0) = 2 x (2) = 3 x 2 (m ) x 2 (0) = 1 x 2 (2) = 3 x 2 ( −m ) x 2 (0) = 1

2

x 1 (3) = −1 x 2 (3) = 4 x 2 (1) = 2

F ig 1 . F ig 2 . F ig 3 .
4 1 2 3

x 2((0 − m )) 4 1 ⇒ x 2((1 − m )) 4 x 2((2 − m )) 4 x 2 ((3 − m )) 4

3 4 2 ⇒ 1 3 ⇒2 4
= x 2,0 (m ) = x 2,1 (m ) = x 2,2(m ) = x 2,3 (m )

2 3 4 1
F ig 4 : C ircula rly sh ifted seq u en c es x 2 ( −m ) fo r n = 0 , 1 , 2 , 3 .
Chapter 2 - Discrete Time Signals and Systems 2. 74
The given sequences are 4-point sequences . \ N = 4.
Each sample of x3(n) is given by sum of the samples of product sequence defined by the equation,
3 3
x3 (n) = ∑
m = 0
x1(m) x 2,n (m) = ∑
m = 0
vn (m) ; where vn (m) = x1(m) x 2,n (m) .....(1)

Using the above equation (1), graphical method of computing each sample of x3(n) are shown in fig 5 to fig 8.
3 3 3
When n = 0 ; x 3 (0) = ∑
m = 0
x1(m) x 2 ((0 − m))4 = ∑
m = 0
x1(m) x 2,0 (m) = ∑
m = 0
v 0 (m)

1 4 1 ×4 = 4

2 x 1 (m ) 2 X 3 x 2, 0 (m ) 1 ⇒2 × 3 = 6 v 0 (m ) 2 ×1 = 2

−1 2 −1 × 2 = −2
T he su m o f sa m p les of v 0 (m ) giv es x 3 (0)
F ig 5: C o m p u ta tio n of x 3 (0 ). ∴ x (0 ) = 2 + 4 + 6 − 2 = 1 0
3

3 3 3
When n = 1 ; x 3 (1) = ∑
m = 0
x1(m) x 2 ((1 − m))4 = ∑
m = 0
x1(m) x 2,1(m) = ∑
m = 0
v1(m)

1 1 1 ×1 = 1

x 1 (m ) x 2, 1 (m ) ⇒ v 1 (m )
2 2 X 4 2 2 ×4 = 8 2 ×2 = 4

−1 3 −1 × 3 = −3
T he su m o f sa m ples of v 1 (m ) giv es x 3 (1)
F ig 6: C o m p u ta tio n of x 3 (1 ). ∴ x (1 ) = 4 + 1 + 8 − 3 = 1 0
3

3 3 3
When n = 2 ; x 3 (2) = ∑
m = 0
x1(m) x 2 ((2 − m))4 = ∑
m = 0
x1(m) x 2,2 (m) = ∑
m = 0
v 2 (m)

1 2 1 ×2 = 2

2 x 1 (m ) 2 X 1 x 2 , 2 (m ) 3 ⇒ 2 ×1 = 2 v 2 (m ) 2 ×3 = 6

−1 4 −1 × 4 = −4
T h e s u m o f s a m p le s of v 2 (m ) giv es x 3 (2)
F ig 7 : C o m p u ta tio n of x 3 (2 ). ∴ x 3 (2) = 6 + 2 + 2 − 4 = 6
3 3 3
When n = 3 ; x 3 (3) = ∑
m = 0
x1(m) x 2 ((3 − m))4 = ∑
m = 0
x1(m) x 2,3 (m) = ∑
m = 0
v 3 (m)

1 3 1 ×3 = 3

2 x 1 (m ) 2 X 2 x 2, 3 (m ) 4 ⇒ 2 ×2 = 4 v 3 (m ) 2 ×4 = 8

−1 1 −1 × 1 = −1

F ig 8 : C o m p u ta tio n of x 3 (3 ). T h e s u m o f s a m p le s of v 3 (m ) giv es x 3 (3)

∴ x 3 (3) = 8 + 3 + 4 − 1 = 1 4
\ x3(n) = {10, 10, 6, 14}
-
2. 75 Digital Signal Processing
Method 2 : Circular Convolution Using Tabular Array

The index n in the given sequences are changed to m and then, the given sequences can be represented
in the tabular array as shown below. Here the shifted sequences x2, n(m) are periodically extended with a
periodicity of N = 4. Let x3(n) be the sequence obtained by convolution of x1(n) and x2(n). Each sample of x3(n) is
given by the equation,

N − 1 N − 1
x 3 (n) = ∑
m = 0
x1(m) x 2((n − m))N = ∑
m = 0
x1(m) x 2,n (m), where x 2,n (m) = x 2((n − m))N

Note : The boldfaced numbers are samples obtained by periodic extension.

m –3 –2 –1 0 1 2 3

x1(m) 2 1 2 –1

x2(m) 1 2 3 4

x2((–m))4 = x2,0(m) 4 3 2 1 4 3 2

x2((1– m))4 = x2,1(m) 4 3 2 1 4 3

x2((2 – m))4 = x2,2(m) 4 3 2 1 4

x2((3 – m))4 = x2,3(m) 4 3 2 1

To determine a sample of x3(n) at n = q, multiply the sequence, x1(m) and x 2,q (m), to get a product
sequence x1(m) x 2,q (m). [i.e., multiply the corresponding elements of the row x1(m) and x2, q(m)]. The sum of all
the samples of the product sequence gives x3(q).
3
When n = 0 ; x 3(0) = ∑
m =0
x1(m) x 2,0 (m)

= x1(0) x 2,0 (0) + x1(1) x 2,0 (1) + x1(2) x 2,0 (2) + x1(3) x 2,0 (3)
= 2 × 1 + 1 × 4 + 2 × 3 + (−1) × 2 = 2 + 4 + 6 − 2 = 10
The samples of x3(n) for other values of n are calculated as shown for n = 0.
3
When n = 1; x3(1) = ∑
m =0
x1(m) x 2,1(m) = 4 + 1 + 8 − 3 = 10

3
When n = 2; x 3(2) = ∑
m =0
x1(m) x 2,2(m) = 6 + 2 + 2 − 4 = 6

3
When n = 3; x3 (3) = ∑
m =0
x1(m) x 2,3 (m) = 8 + 3 + 4 − 1 = 14

l
∴ x3 (n) = 10, 10, 6, 14 q
A
Method 3 : Circular Convolution Using Matrices
The sequence x1(n) can be arranged as a column vector of order N ´ 1 and using the samples of x2(n) the
N ´ N matrix is formed as shown below. The product of the two matrices gives the sequence x3(n).

LMx (0)
2 x 2(3) x 2(2) x 2(1)OP LMx (0)OP
1 LMx (0)OP
3

MMx (1)
2 x 2 (0) x 2(3) x (2) P
2 MMx (1) PP
1
= MMx (1) PP
3
x (3) P
MMxx ((32))
2 x 2(1)
x 2(2)
x 2 (0)
x 2(1)
2
x (0)PQ
P MMxx ((32))PP
1
MMxx ((32)) PP
3

N2 2 N Q
1 N Q
3
Chapter 2 - Discrete Time Signals and Systems 2. 76

LM1 4 3 2 OP LM 2OP LM1× 2 + 4 × 1+ 3 × 2 + 2 × −1OP LM10OP

MM2 1 4 3 P MM 1PP = MM2 × 2 + 1 × 1+ 4 × 2 + 3 × −1PP = MM10PP
4P
MM34 2
3
1
2 1 PQ
P MM−21PP MMN34 ×× 22 ++ 23 ×× 1+ 1 × 2 + 4 × −1
P
1+ 2 × 2 + 1 × −1PQ
MM146PP
N N Q N Q
\ x3(n) = {10, 10, 6, 14}
- ­

Example 2.25
Perform the circular convolution of the two sequences x1(n) and x2(n), where,

l
x1(n) = 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6 q
- ­
l
x 2(n) = 0.1, 0.3, 0.5, 0.7, 0.9, 1.1, 1.3, 1.5 q
- ­
Solution
Let x3(n) be the result of the circular convolution of x1(n) and x2(n). The given sequences consists of eight
samples. Then x3(n) will also have 8 samples.

The sequences are represented in the tabular array as shown below after replacing n by m. The sequence
x2(m) is folded and shifted.

The shifted sequences x2,n(m) are periodically extended with a periodicity of N = 8.

Note : The boldfaced numbers are samples obtained by periodic extension

m –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7

x1(m) 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

x2(m) 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5

x2((–m))8 = x2,0(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5 1.3 1.1 0.9 0.7 0.5 0.3

x2((1 – m))8 = x2,1(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5 1.3 1.1 0.9 0.7 0.5

x2((2 – m))8 = x2,2(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5 1.3 1.1 0.9 0.7

x2((3 – m))8 = x2,3(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5 1.3 1.1 0.9

x2((4 – m))8 = x2,4(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5 1.3 1.1

x2((5 – m))8 = x2,5(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5 1.3

x2((6 – m))8 = x2,6(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1 1.5

x2((7 – m))8 = x2,7(m) 1.5 1.3 1.1 0.9 0.7 0.5 0.3 0.1

Each sample of x3(n) is given by the equation,

7 7
x 3 (n) =
m =0
∑ x1(m) x 2 ((n − m))8 = ∑
m =0
x1(m) x 2,n (m) ; where x 2,n (m) = x 2 ((n − m))8

The samples of x3(0) are calculated as shown below.

2. 77 Digital Signal Processing
7 7
When n = 0 ; x 3 (n) = ∑ x (m) x
m= 0
1 2 ((0 − m))8 = ∑ x (m) x
m= 0
1 2, 0 (m)

= x1(0) x 2, 0 (0) + x1(1) x 2, 0 (1) + x1(2) x 2, 0 (2) + x1(3) x 2, 0 (3)

+ x1(4) x 2, 0 (4) + x1(5) x 2, 0 (5) + x1(6) x 2, 0 (6) + x1(7) x 2, 0 (7)
= 0.02 + 0.6 + 0.78 + 0.88 + 0.9 + 0.84 + 0.7 + 0.48 = 5.20
The samples of x3(n) for other values of n are calculated as shown for n = 0.
7 7
When n = 1; x3(1) = ∑
m=0
x1(m) x 2((1 − m))8 = ∑
m=0
x1(m) x 2,1(m) = 6.00

7 7
When n = 2; x3 (2) = ∑ x (m) x ((2 − m))
m=0
1 2 8 = ∑ x (m) x
m =0
1 2,2 (m) = 6.48

7 7
When n = 3; x3(3) = ∑ x (m) x ((3 − m))
m =0
1 2 8 = ∑ x (m) x
m =0
1 2,3 (m) = 6.64

7 7
When n = 4; x3 (4) = ∑ x (m) x ((4 − m))
m =0
1 2 8 = ∑
m=0
x1(m) x 2,4 (m) = 6.48

7 7
When n = 5; x3 (5) = ∑
m=0
x1(m) x 2((5 − m))8 = ∑
m=0
x1(m) x 2,5(m) = 6.00

7 7
When n = 6; x3(6) = ∑
m=0
x1(m) x 2 ((6 − m))8 = ∑ x (m) x
m =0
1 2,6 (m) = 5.20

7 7
When n = 7; x3(7) = ∑
m=0
x1(m) x 2 ((7 − m))8 = ∑ x (m) x
m =0
1 2,7 (m) = 4.08

lA
∴ x3(n) = 5.20, 6.00, 6.48, 6.64, 6.48, 6.00, 5.20, 4.08 q
Example 2.26
Find the linear and circular convolution of the sequences, x(n) = 1, 0.5 l q l q
and h(n) = 0.5, 1 .
­ ­ A A
Solution
Linear Convolution by Tabular Array
∞
Let , y(n) = x(n) * h(n) = ∑
m = −∞
x(m) h(n − m) ; where m is a dummy variable for convolution.

Since both x(n) and h(n) starts at n = 0, the output sequence y(n) will also start at n = 0.
Since the length of x(n) and h(n) is 2, the length of y(n) is 2 + 2 – 1 = 3.
Let us change the index n to m in x(n) and h(n). The sequences x(m) and h(m) are represented in the
tabular array as shown below.

Note : The unfilled boxes in the table are considered as zeros.

m –1 0 1 2
x(m) 1 0.5
h(m) 0.5 1
h(–m) = h0(m) 1 0.5
h(1 – m) = h1(m) 1 0.5
h(2 – m) = h2(m) 1 0.5
Chapter 2 - Discrete Time Signals and Systems 2. 78
Each sample of y(n) is given by the relation,
∞ ∞
y(n) = ∑
m = −∞
x(m) h(n − m) = ∑
m = −∞
x(m) hn (m) ; where hn (m) = h(n − m)

∞ 1
When n = 0 ; y(0) = ∑
m = −∞
x(m) h( −m) = ∑ x(m) h (m) = x(−1) h (−1) + x(0) h (0) + x(1) h (1)
m = −1
0 0 0 0

= 0 × 1 + 1 × 0.5 + 0.5 × 0 = 0 + 0.5 + 0 = 0.5

∞ 1
When n = 1 ; y(1) = ∑
m = −∞
x(m) h(1 − m) = ∑
m = 0
x(m) h1(m) = 1 + 0.25 = 125
.

∞ 2
When n = 2 ; y(2) = ∑
m = −∞
x(m) h(2 − m) = ∑
m= 0
x(m) h2(m) = 0 + 0.5 + 0 = 0.5

l
∴ y(n) = 0.5, 1.25, 0.5 q
A
Circular Convolution by Tabular Array
N− 1
Let, y(n) = x(n) ∗ h(n) =
m=0
∑ x(m) h((n − m)) N ; where m is a dummy variable for convolution.

The index n in the sequences are changed to m and the sequences are represented in the tabular array as
shown below. The shifted sequence hn(m) is periodically extended with periodicity N = 2.
Note : The boldfaced number is the sample obtained by periodic extension.

m –1 0 1
x(m) 1 0.5
h(m) 0.5 1
h((–m))2 = h0(m) 1 0.5 1
h((1 – m))2 = h1(m) 1 0.5

Each sample of y(n) is given by the equation,

N − 1 N − 1
y(n) = ∑
m = 0
x(m) h((n − m))N = ∑
m = 0
x(m) hn (m); where hn (m) = h((n − m))N

N − 1 1
When n = 0 ; y(0) = ∑
m= 0
x(m) h((0 − m))2 = ∑ x(m) h (m)
m = 0
0

= x(0) h0 (0) + x(1) h0 (1) = 1 × 0.5 + 0.5 × 1 = 0.5 + 0.5 = 10

.
N − 1 1
When n = 1 ; y(1) = ∑ x(m) h((1 − m))2 = ∑ x(m) h1(m)
m= 0 m = 0
= x(0) h1(0) + x(1) h1(1) = 1 × 1 + 0.5 × 0.5 = 1 + 0.25 = 125
.
∴ y(n) = 10l
. , 1.25 q
A
Example 2.27
The input x(n) and impulse response h(n) of a LTI system are given by,
x(n) = {–1, 1, 2, –2) ; h(n) = {0.5, 1, –1, 2, 0.75}
A A ­
Determine the response of the system a) using linear convolution and b) using circular convolution.
2. 79 Digital Signal Processing
Solution
a) Response of LTI system using linear convolution
Let y(n) be the response of LTI system. By convolution sum formula,
+∞
y(n) = x(n) ∗ h(n) = ∑
m = −∞
x(m) h(n − m) ; where m is a dummy variable used for convolution.

The sequence x(n) starts at n = 0 and h(n) starts at n = –1. Hence y(n) will start at n = 0 + (–1) = –1.
The length of x(n) is 4 and the length of h(n) is 5. Hence the length of y(n) is (4 + 5 – 1) = 8. Also y(n) ends at
n = 0 + (–1) + (4 + 5 –2) = 6.
Let us change the index n to m in x(n) and h(n). The sequences x(m) and h(m) are represented on the
tabular array as shown below. Let us fold h(m) to get h(–m) and shift h(–m) to perform convolution operation.
Note : The unfilled boxes in the table are considered as zeros.
m –4 –3 –2 –1 0 1 2 3 4 5 6 7
x(m) –1 1 2 –2
h(m) 0.5 1 –1 2 0.75
h(–m) 0.75 2 –1 1 0.5
h(–1 – m) = h–1(m) 0.75 2 –1 1 0.5
h(0 – m) = h0(m) 0.75 2 –1 1 0.5
h(1 – m) = h1(m) 0.75 2 –1 1 0.5
h(2 – m) = h2(m) 0.75 2 –1 1 0.5
h(3 – m) = h3(m) 0.75 2 –1 1 0.5
h(4 – m) = h4(m) 0.75 2 –1 1 0.5
h(5 – m) = h5(m) 0.75 2 –1 1 0.5
h(6 – m) = h6(m) 0.75 2 –1 1 0.5

Each sample of y(n) is given by summation of the product sequence, x(m) h(n – m). To determine a
sample of y(n) at n = q, multiply the sequence x(m) and hq(m) to get a product sequence [i.e., multiply the
corresponding elements of the row x(m) and hq(m)]. The sum of all the samples of the product sequence gives
y(q).
+∞ +∞
i. e. , y(n) = ∑ x(m) h(n − m) = ∑ x(m) h (m)
m = −∞ m = −∞
n

3
When n = −1 ; y(−1) = ∑
m = −4
x(m) h−1(m)

= x(–4) h–1(–4) + x(–3) h–1(–3) + x(–2) h–1(–2) + x(–1) h–1(–1) + x(0) h–1(0)
+ x(1) h–1(1) + x(2) h–1(2) + x(3) h–1(3)
= 0 + 0 + 0 + 0 + (–0.5) + 0 + 0 + 0 = –0.5

The samples of y(n) for other values of n are calculated as shown for n = –1.
3
When n = 0 ; y(0) = ∑
m = −3
x(m) h0 (m) = 0 + 0 + 0 + (−1) + 0.5 + 0 + 0 = −0.5

3
When n = 1 ; y(1) = ∑
m = −2
x(m) h1(m) = 0 + 0 + 1+ 1+ 1+ 0 = 3

3
When n = 2 ; y(2) = ∑
m = −1
x(m) h2 (m) = 0 + (−2) + ( −1) + 2 + ( −1) = −2
Chapter 2 - Discrete Time Signals and Systems 2. 80
4
When n = 3 ; y(3) = ∑
m= 0
x(m) h3(m) = −0.75 + 2 + (−2) + (−2) + 0 = −2.75

5
When n = 4 ; y(4) = ∑
m= 0
x(m) h4 (m) = 0 + 0.75 + 4 + 2 + 0 + 0 = 6.75

6
When n = 5 ; y(5) = ∑
m= 0
x(m) h5(m) = 0 + 0 + 1.5 + (−4) + 0 + 0 + 0 = −2. 5

7
When n = 6 ; y(6) = ∑
m= 0
x(m) h6 (m) = 0 + 0 + 0 + (−1.5) + 0 + 0 + 0 + 0 = −1.5

The response of LTI system y(n) is,

y(n) = {–0.5, –0.5, 3, –2, –2.75, 6.75, –2.5, –1.5}
- ­
b) Response of LTI System Using Circular Convolution
The response of LTI system is given by linear convolution of x(n) and h(n). Let y(n) be the response
sequence of LTI system. To get the result of linear convolution from circular convolution, both the sequences
should be converted to the size of y(n) and perform circular convolution of the converted sequences. Also the
converted sequences should start and end at the same value of n as that of y(n).
The length of x(n) is 4 and the length of h(n) is 5. Hence the length of y(n) is (4 + 5 – 1) = 8. Therefore both
the sequences should be converted to 8-point sequences.
The x(n) starts at n = 0 and h(n) starts at n = –1. Hence y(n) will start at n = 0 + (–1) = –1. The y(n) will
end at n = [0 +(–1)] + (4 + 5 – 2) = 6. Therefore the converted sequences should start at n = –1 and end at n = 6.
\ x(n) = {0, –1, 1, 2, –2, 0, 0, 0} and h(n) = {0.5, 1, –1, 2, 0.75, 0, 0, 0}
- -
The converted sequences x(n) and h(n) are represented on the tabular array after replacing the index n by
m as shown below. The sequence h(m) is folded and shifted.
The shifted sequences hn(m) are periodically extended with a periodicity of N = 8.

Note : The boldfaced numbers are samples obtained by periodic extension of the sequences.
m –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7
x(m) 0 –1 1 2 –2 0 0 0
h(m) 0.5 1 –1 2 0.75 0 0 0
h(–m) 0 0 0 0.75 2 –1 1 0.5
h((–1 – m))8 = h–1(m) 0 0 0 0.75 2 –1 1 0.5 0 0 0 0.75 2 –1 1
h((0 – m))8 = h0(m) 0 0 0 0.75 2 –1 1 0.5 0 0 0 0.75 2 –1
h((1 – m))8 = h1(m) 0 0 0 0.75 2 –1 1 0.5 0 0 0 0.75 2
h((2 – m))8 = h2(m) 0 0 0 0.75 2 –1 1 0.5 0 0 0 0.75
h((3 – m))8 = h3(m) 0 0 0 0.75 2 –1 1 0.5 0 0 0
h((4 – m))8 = h4(m) 0 0 0 0.75 2 –1 1 0.5 0 0
h((5 – m))8 = h5(m) 0 0 0 0.75 2 –1 1 0.5 0
h((6 – m))8 = h6(m) 0 0 0.75 2 –1 1 0.5 0 0 0 0.75 2 –1 1 0.5

Let y(n) be the sequence obtained by circular convolution of x(n) and h(n).
Now, each sample of y(n) is given by,
6 6
y(n) = ∑
m = −1
x(m) h((n − m))8 = ∑
m = −1
x(m) hn (m) ; where hn (m) = h((n − m))8
2. 81 Digital Signal Processing
To determine a sample of y(n) at n = q, multiply the sequence x(m) and hq(m) to get a product sequence
x(m) hq(m), [i.e., multiply the corresponding elements of the row x(m) and hq(m)]. The sum of all the samples of the
product sequence gives y(q).
6
When n = −1 ; y(−1) = ∑
m = −1
x(m) h−1(n) = x( −1) h−1( −1) + x(0) h−1(0) + x(1) h−1(1) + x(2) h−1(2)

+ x(3) h−1(3) + x(4) h−1(4) + x(5) h−1(5) + x(6) h−1(6)

= 0 + (−0.5) + 0 + 0 + 0 + 0 + 0 + 0 = −0.5

The samples of y(n) for other values of n are calculated as shown for n = –1.
6
When n = 0 ; y(0) = ∑ x(m) h0m = 0 + (−1) + 0.5 + 0 + 0 + 0 + 0 + 0 = −0.5
m = −1
6
When n = 1 ; y(1) = ∑ x(m) h1m = 0 + 1+ 1+ 1+ 0 + 0 + 0 + 0 = 3
m = −1
6
When n = 2 ; y(2) = ∑ x(m) h2m = 0 + (−2) + (−1) + 2 + (−1) + 0 + 0 + 0 = −2
m = −1
6
When n = 3 ; y(3) = ∑ x(m) h3m = 0 + (−0.75) + 2 + (−2) + (−2) + 0 + 0 + 0 = −2.75
m = −1
6
When n = 4 ; y(4) = ∑ x(m) h4m = 0 + 0 + 0.75 + 4 + 2 + 0 + 0 + 0 = 6.75
m = −1
6
When n = 5 ; y(5) = ∑ x(m) h5m = 0 + 0 + 0 + 1.5 + (−4) + 0 + 0 + 0 = −2.5
m = −1
6
When n = 6 ; y(6) = ∑ x(m) h6m = 0 + 0 + 0 + 0 + (−1. 5) + 0 + 0 + 0 = −1.5
m = −1

The response of LTI system y(n) is,

y(n) = {–0.5, –0.5, 3, –2, –2.75, 6.75, –2.5, –1.5}
-

Note : 1. Since circular convolution is periodic, the convolution is performed for any one period.
2. It can be observed that the results of both the methods are same.

2.11 Sectioned Convolution

The response of an LTI system for any arbitrary input is given by linear convolution of the input and
the impulse response of the system. If one of the sequences (either the input sequence or impulse response
sequence) is very much larger than the other, then it is very difficult to compute the linear convolution for the
following reasons.

1. The entire sequence should be available before convolution can be carried out. This makes long
delay in getting the output.

2. Large amounts of memory is required to store the sequences.

The above problems can be overcome in the sectioned convolutions. In this technique the larger
sequence is sectioned (or splitted) into the size of smaller sequence. Then the linear convolution of each
section of longer sequence and the smaller sequence is performed. The output sequences obtained from the
convolutions of all the sections are combined to get the overall output sequence. There are two methods of
sectioned convolutions. They are overlap add method and overlap save method.
Chapter 2 - Discrete Time Signals and Systems 2. 82
2.11.1 Overlap Add Method
In the overlap add method, the longer sequence is divided into smaller sequences. Then linear
convolution of each section of longer sequence and smaller sequence is performed. The overall output
sequence is obtained by combining the output of the sectioned convolution.
Let, N1 = Length of longer sequence
N2 = Length of smaller sequence
Let the longer sequence be divided into sections of size N3 samples.
Note : Normally the longer sequence is divided into sections of size same as that of smaller sequence.

N3 + N2 −1

N3 N2 −1
N3 + N2 −1

N2 −1 N 3 − ( N 2 − 1) N2 −1
N3 + N2 − 1

N2 − 1 N 3 − ( N 2 − 1) N2 −1
O v erlapped region

Note : Samples in the shaded region are added. O v erlapped region

F ig 2 .3 0 : O ve rla p p in g of o u tp ut seq u en c e o f sec tio n ed c o n vo lution b y o v erla p a d d m eth o d .

The linear convolution of each section with smaller sequence will produce an output sequence of size
N3 + N2 –1 samples. In this method the last N2 –1 samples of each output sequence overlaps with the first
N2 –1 samples of the next section. [i.e., there will be a region of N2 –1 samples over which the output sequence
of qth convolution overlaps the output sequence of (q +1)th convolution]. While combining the output
sequences of the various sectioned convolutions, the corresponding samples of overlapped regions are
added and the samples of non-overlapped regions are retained as such.

2.11.2 Overlap Save Method

In the overlap save method, the results of linear convolution of the various sections are obtained
using circular convolution. In this method, the longer sequence is divided into smaller sequences. Each
section of the longer sequence and the smaller sequence are converted to the size of the output sequence of
sectioned convolution. The circular convolution of each section of the longer sequence and the smaller
sequence is performed. The overall output sequence is obtained by combining the outputs of the sectioned
convolution.
Let, N1 = Length of longer sequence
N2 = Length of smaller sequence
Let the longer sequence be divided into sections of size N3 samples.
Note : Normally the longer sequence is divided into sections of size same as that of smaller sequence.
In the overlap save method, the results of linear convolution are obtained by circular convolution.
Hence each section of longer sequence and the smaller sequence are converted to the size of output sequence
of size N3 + N2 – 1 samples.The smaller sequence is converted to size of N3 + N2 –1 samples, by appending with
zeros.The convertion of each section of longer sequence to the size N3 + N2 –1 samples can be performed in
two different methods.
2. 83 Digital Signal Processing
Method-1
In this method, the first N2 –1 samples of a section is appended as last N2 –1 samples of the previous
section (i.e., the overlapping samples are placed at the beginning of the section). The circular convolution of
each section will produce an output sequence of size N3 + N2 –1 samples. In this output the first N2 –1 samples
are discarded and the remaining samples of the output of sectioned convolutions are saved as the overall
output sequence.
N3 + N2 −1
N3 N2 −1

N3 + N2 −1

N2 −1 N 3 − ( N 2 − 1) N2 −1

N3 + N2 −1

N2 −1 N 3 − ( N 2 − 1)
A ppended
w ith zero
F ig 2 .3 1 : A p p en d ing o f sec tio n s o f in pu t seq u en c e
in m eth o d 1 o f o ve rlap sa ve m e th o d .
N3 + N2 −1

N2 −1 N 3 − ( N 2 − 1) N2 −1

N3 + N2 −1

N2 −1 N 3 − ( N 2 − 1) N2 −1

N3 + N2 − 1
N2 −1 N3
O v erlapped region

Note : Samples in the shaded region are discarded.

O v erlapped region

F ig 2 .3 2 : O ve rla p p in g of o u tp ut seq u en c e of sec tio n ed c o n vo lution

b y m eth o d 1 o f o ve rla p sa ve m ethod .
Method-2
In this method, the last N2–1 samples of a section is appended as last N2 –1 samples of the next section
(i.e, the overlapping samples are placed at the end of the sections). The circular convolution of each section
will produce an output sequence of size N3 + N2 –1 samples. In this output the last N2 –1 samples are discarded
and the remaining samples of the output of sectioned convolutions are saved as the overall output sequence.

N3 + N2 −1
N 3 −( N 2 − 1) N2 −1 N2 −1
A ppended
w ith zero
N3 + N2 −1

N 3 −( N 2 − 1) N2 −1 N2 −1

N3 + N2 −1

N3 N2 −1

F ig 2 .3 3 : A p p en d ing o f sec tio n s o f in pu t seq u en c e

in m eth o d 2 o f o ve rlap sa ve m eth o d .
Chapter 2 - Discrete Time Signals and Systems 2. 84

N3 + N2 −1

N3 N2 −1
N3 + N2 −1

N2 −1 N 3 − ( N 2 − 1) N2 −1

N3 + N2 −1
N2 − 1 N 3 −( N 2 − 1 ) N2 −1
O v erlapped region

Note : Samples in the shaded region are discarded. O v erlapped region

F ig 2 .3 4 : O ve rla p p ing of o utp ut seq u en c e o f sec tio n ed co n vo lu tio n

b y m eth o d 2 o f o verla p sa v e m eth od .

Example 2.28
Perform the linear convolution of the following sequences by a) Overlap add method, and b) Overlap
save method.

x(n) = {1, –1, 2, –2, 3, –3, 4, –4} ; h(n) = {–1, 1}

Solution
a) Overlap Add Method

In this method the longer sequence is sectioned into sequences of size equal to smaller sequence. Here
x(n) is a longer sequence when compared to h(n). Hence x(n) is sectioned into sequences of size equal to h(n).

Given that, x(n) = {1, –1, 2, –2, 3, –3, 4, –4}

Let x(n) can be sectioned into four sequences, each consisting of two samples of x(n) as shown below.
x1(n) = 1 ; n = 0 x2(n) = 2 ; n = 2 x3(n) = 3 ; n=4 x4(n) = 4 ; n=6
= –1 ; n = 1 = –2 ; n = 3 = –3 ; n=5 = –4 ; n = 7
Let y1(n), y2(n), y3(n) and y4(n) be the output of linear convolution of x1(n), x2(n), x3(n) and x4(n) with h(n)
respectively.
Here h(n) starts at n = nh = 0
x1(n) starts at n = n1 = 0, \ y1(n) will start at n = n1 + nh = 0 + 0 = 0
x2(n) starts at n = n2 = 2, \ y2(n) will start at n = n2 + nh = 2 + 0 = 2
x3(n) starts at n = n3 = 4, \ y3(n) will start at n = n3 + nh = 4 + 0 = 4
x4(n) starts at n = n4 = 6, \ y4(n) will start at n = n4 + nh = 6 + 0 = 6

Here linear convolution of each section is performed between two sequences each consisting of 2
samples. Hence each convolution output will consists of 2 + 2 – 1 = 3 samples. The convolution of each section
is performed by tabular method as shown below.

Note :
1. Here N1 = 8, N2 = 2, N3 = 2. \ (N2 – 1) = 2 – 1 = 1 and (N2 + N3 – 1) = 2 + 2 – 1 = 3
2. The unfilled boxes in the tables are considered as zero.
3. For convenience of convolution operation the index n is replaced by m in x1(n), x2(n), x3(n), x4(n) and h(n).
2. 85 Digital Signal Processing
Convolution of Section 1
+∞

m –1 0 1 2
y1(n) = x1(n) ∗ h(n) = ∑ x (m) h(n − m)
m = −∞
1

+∞
x1(m) 1 –1
= ∑ x (m) h (m) ;
1 n n = 0, 1, 2
h(m) –1 1 m = −∞
where hn (m) = h(n − m)
h(–m) = ho(m) 1 –1
h(1 – m) = h1(m) 1 –1
When n = 0 ; y1(0) = ∑ x1(m ) h0 (m) = 0 − 1+ 0 = − 1

h(2 – m) = h2(m) 1 –1
When n = 1 ; y1(1) = ∑ x (m) h (m) = 1+ 1 = 2 1 1

When n = 2 ; y (2) = ∑ x (m) h (m) = 0 − 1+ 0 = −1

1 1 2
Convolution of Section 2
+∞

m –1 0 1 2 3 4 y 2 (n) = x 2(n) ∗ h(n) = ∑

m = −∞
x 2(m) h(n − m)

x2(m) 2 –2 +∞

h(m) –1 1
= ∑ x (m) h (m)
m = −∞
2 n ; n = 2, 3, 4

where hn (m) = h(n – m)

h(–m) 1 –1
h(2 – m) = h2(m) 1 –1 When n = 2 ; y 2(2) =∑ x (m) h (m) = 0 − 2 + 0 = −2
2 2

h(3 – m) = h3(m) 1 –1 When n = 3 ; y (3) = ∑ x (m) h (m) = 2 + 2

2 = 4
2 3

h(4 – m) = h4(m) 1 –1 When n = 4 ; y (4) = ∑ x (m) h (m) = 0 − 2 + 0 = −2

2 2 4

Convolution of Section 3

m –1 0 1 2 3 4 5 6
x3(m) 3 –3
h(m) –1 1
h(–m) 1 –1
h(4 – m) = h4(m) 1 –1
h(5 – m) = h5(m) 1 –1
h(6 – m) = h6(m) 1 –1
+∞ +∞
y3(n) = x3(n) ∗ h(n) = ∑ x (m) h(n − m) = ∑ x (m) h (m)
m = −∞
3
m = −∞
3 n ; n = 4, 5, 6

where hn (m) = h(n – m)

When n = 4 ; y3(4) = ∑ x (m) h (m) = 0 − 3 + 0 = −3
3 4

When n = 5 ; y (5) = ∑ x (m) h (m) = 3 + 3

3 3 = 6 5

When n = 6 ; y (6) = ∑ x (m) h (m) = 0 − 3 + 0 = −3

3 3 6

Convolution of Section 4
m –1 0 1 2 3 4 5 6 7 8
x4(m) 4 –4
h(m) –1 1
h(–m) 1 –1
h(6 – m) = h6(m) 1 –1
h(7 – m) = h7(m) 1 –1
h(8 – m) = h8(m) 1 –1
Chapter 2 - Discrete Time Signals and Systems 2. 86
+∞ +∞
y 4 (n) = x 4 (n) ∗ h(n) = ∑x
m = −∞
4 (m) h(n − m) = ∑ x (m) h (m)
m = −∞
4 n ; n = 6, 7, 8

where hn (m) = h(n − m)

∑ x (m) h (m) = 0 − 4 + 0 = −4
When n = 6 ; y 4 (6) = 4 6

When n = 7 ; y (7) = ∑ x (m) h (m) = 4 + 4

4 4 7 = 8
When n = 8 ; y (8) = ∑ x (m) h (m) = 0 − 4 + 0 = −4
4 4 8

To Combine the Output of Convolution of Each Section

It can be observed that the last sample in an output sequence overlaps with the first sample of next output
sequence. In this method the overall output is obtained by combining the outputs of the convolution of all
sections. The overlapped portions (or samples) are added while combining the output. The output of all sections
can be represented in a table as shown below. Then the samples corresponding to same value of n are added to
get the overall output.
n 0 1 2 3 4 5 6 7 8
y1(n) –1 2 –1
y2(n) –2 4 –2
y3(n) –3 6 –3
y4(n) –4 8 –4
y(n) –1 2 –3 4 –5 6 –7 8 –4

\ y(n) = x(n) * h(n) = {–1, 2, –3, 4, –5, 6, –7, 8, –4}

b) Overlap Save Method

In this method, the longer sequence is sectioned into sequences of size equal to smaller sequence. The
number of samples that will be obtained in the output of linear convolution of each section is determined. Then
each section of longer sequence is converted to the size of output sequence using the samples of original longer
sequence. The smaller sequence is also converted to the size of output sequence by appending with zeros. Then
the circular convolution of each section is performed.
Here x(n) is a longer sequence when compared to h(n). Hence x(n) is sectioned into sequences of size
equal to h(n). Given that, x(n) = {1, –1, 2, –2, 3, –3, 4, –4}
Let x(n) be sectioned into four sequences, each consisting of two samples of x(n) as shown below.
x1(n) = 1 ; n = 0 x2(n) = 2 ; n = 2 x3(n) = 3 ; n=4 x4(n) = 4 ; n=6
= –1 ; n = 1 = –2 ; n = 3 = –3 ; n=5 = –4 ; n = 7
Let y1(n), y2(n), y3(n) and y4(n) be the output of linear convolution of x1(n), x2(n), x3(n) and x4(n) with h(n)
respectively. Here linear convolution of each section will result in an output sequence consisting of
2 + 2 – 1 = 3 samples.
The sequence h(n) is converted to 3-sample sequence by appending with zero. \ h(n) = {–1, 1, 0}
Method - 1
In method 1, the overlapping samples are placed at the beginning of the sections. Each section of longer
sequence is converted to 3-sample sequences, using the samples of original longer sequence as shown below.
It can be observed that the first sample of x2(n) is placed as overlapping sample at the end of x1(n). The first sample
of x3(n) is placed as overlapping sample at the end of x2(n). The first sample of x4(n) is placed as overlapping
sample at the end of x3(n). Since there is no fifth section, the overlapping sample of x4(n) is taken as zero.
x1(n) = 1 ; n = 0 x2(n) = 2 ; n = 2 x3(n) = 3 ; n = 4 x4(n) = 4 ; n = 6
= –1 ; n = 1 = –2 ; n = 3 = –3 ; n = 5 = –4 ; n = 7
= 2; n=2 = 3 ; n=4 = 4; n=6 = 0 ;n=8
2. 87 Digital Signal Processing
Now perform circular convolution of each section with h(n). The output sequence obtained from circular
convolution will have three samples. The circular convolution of each section is performed by tabular method
as shown below.
Here h(n) starts at n = nh = 0
x1(n) starts at n = n1 = 0, \ y1(n) will start at n = n1 + nh = 0 + 0 = 0
x2(n) starts at n = n2 = 2, \ y2(n) will start at n = n2 + nh = 2 + 0 = 2
x3(n) starts at n = n3 = 4, \ y3(n) will start at n = n3 + nh = 4 + 0 = 4
x4(n) starts at n = n4 = 6, \ y4(n) will start at n = n4 + nh = 6 + 0 = 6

Note : 1. Here N1 = 8, N2 = 2, N3 = 2. \ (N2 – 1) = 2 – 1 = 1 and (N2 + N3 – 1) = 2 + 2 – 1 = 3

2. The boldfaced numbers in the tables are obtained by periodic extension.
3. For convenience of convolution operation, the index n in x1(n), x2(n), x3(n), x4(n) and h(n)
are replaced by m.
mf
Convolution of Section 1 y1(n) = x1(n) ∗ h(n) = ∑
m = mi
x1(m) h((n − m))N
m –2 –1 0 1 2 2

x1(m) 1 –1 2 = ∑ x (m) h (m) ;

m = 0
1 n n = 0, 1, 2,

h(m) –1 1 0 where hn (m) = h((n − m))N

h((–m))3 = h0(m) 0 1 –1 0 1 When n = 0 ; y1(0) = ∑ x (m) h (m) = −1 + 0 + 2 = 1
1 0

h((1 – m))3 = h1(m) 0 1 –1 0 When n = 1 ; y (1) = ∑ x (m) h (m) = 1+ 1+ 0 = 2

1 1 1

h((2 – m))3 = h2(m) 0 1 –1 When n = 2 ; y (2) = ∑ x (m) h (m) = 0 − 1 − 2 = −3

1 1 2

Convolution of Section 2

m –2 –1 0 1 2 3 4
x2(m) 2 –2 3
h(m) –1 1 0
h(–m) 0 1 –1
h((2 – m))3 = h2(m) 0 1 –1 0 1
h((3 – m))3 = h3(m) 0 1 –1 0
h((4 – m))3 = h4(m) 0 1 –1

mf 4
y 2(n) = x 2(n) ∗ h(n) = ∑
m = mi
x 2(m) h((n − m))N = ∑ x (m) h (m) ;
m = 2
2 n n = 2, 3, 4

where hn (m) = h((n − m))N

When n = 2 ; y 2(2) =∑ x (m) h (m) = −2 + 0 + 3 = 1
2 2

When n = 3 ; y (3) = ∑ x (m) h (m) = 2 + 2 + 0 = 4

2 2 3

When n = 4 ; y (4) = ∑ x (m) h (m) = 0 + −2 − 3 = −5

2 2 4

Convolution of Section 3

mf 6
y 3(n) = x 3 (n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
3 N = ∑ x (m) h (m) ;
m = 4
3 n n = 4, 5, 6

where hn (m) = h((n − m))N

Chapter 2 - Discrete Time Signals and Systems 2. 88
m –2 –1 0 1 2 3 4 5 6
x3(m) 3 –3 4
h(m) –1 1 0
h(–m) 0 1 –1
h((4 – m))3 = h4(m) 0 1 –1 0 1
h((5 – m))3 = h5(m) 0 1 –1 0
h((6 – m))3 = h6(m) 0 1 –1

When n = 4 ; y3(4) = ∑ x (m) h (m) = −3 + 0 + 4 = 1

3 4

When n = 5 ; y (5) = ∑ x (m) h (m) = 3 + 3 + 0 = 6

3 3 5

When n = 6 ; y (6) = ∑ x (m) h (m) = 0 − 3 − 4 = −7

3 3 6

Convolution of section 4

m –2 –1 0 1 2 3 4 5 6 7 8
x4(m) 4 –4 0
h(m) –1 1 0
h(–m) 0 1 –1
h((6 – m))3 = h6(m) 0 1 –1 0 1
h((7 – m))3 = h7(m) 0 1 –1 0
h((8 – m))3 = h8(m) 0 1 –1

mf 8
y 4 (n) = x 4 (n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
4 N = ∑ x (m) h (m) ; n = 6, 7, 8
m = 6
4 n

where hn (m) = h((n – m))N

∑ x (m) h (m) = −4 + 0 + 0 = −4
When n = 6 ; y 4 (6) = 4 6

When n = 7 ; y (7) = ∑ x (m) h (m) = 4 + 4 + 0 = 8

4 4 7

When n = 8 ; y (8) = ∑ x (m) h (m) = 0 − 4 + 0 = −4

4 4 8

To Combine the Output of the Convolution of Each Section

It can be observed that the last sample in an output sequence overlaps with the first sample of next output
sequence. In overlap save method the overall output is obtained by combining the outputs of the convolution
of all sections. While combining the outputs, the overlapped first sample of every output sequence is discarded
and the remaining samples are simply saved as samples of y(n) as shown in the following table.
n 0 1 2 3 4 5 6 7 8
y1(n) 1 2 –3
y2(n) 1 4 –5
y3(n) 1 6 –7
y4(n) –4 8 –4
y(n) * 2 –3 4 –5 6 –7 8 –4

y(n) = x(n) * h(n) = {*, 2, –3, –4, –5, 6, –7, 8, –4}

Note : Here y(n) is linear convolution of x(n) and h(n). It can be observed that the results of both the methods
are same, except the first N2 – 1 samples.
2. 89 Digital Signal Processing
Method 2
In method 2, the overlapping samples are placed at the end of the section.Each section of longer
sequence is converted to 3-sample sequence, using the samples of original longer sequence as shown below.
It can be observed that the last sample of x1(n) is placed as overlapping sample at the end of x2(n). The last
sample of x2(n) is placed as overlapping sample at the end of x3(n). The last sample of x3(n) is placed as
overlapping sample at the end of x4(n). Since there is no previous section for x1(n), the overlapping sample of
x1(n) is taken as zero.

x1(n) = 1 ; n = 0 x2(n) = 2 ; n = 2 x3(n) = 3 ; n=4 x4(n) = 4 ; n = 6

= –1 ; n = 1 = –2 ; n = 3 = –3 ; n=5 = –4 ; n = 7

= 0; n=2 = –1 ; n = 4 = –2 ; n = 6 = –3 ; n = 8

Now perform circular convolution of each section with h(n). The output sequence obtained from circular
convolution will have three samples. The circular convolution of each section is performed by tabular method as
shown below.

Here h(n) starts at n = nh = 0

x1(n) starts at n = n1 = 0, \ y1(n) will start at n = n1 + nh = 0 + 0 = 0

x2(n) starts at n = n2 = 2, \ y2(n) will start at n = n2 + nh = 2 + 0 = 2

x3(n) starts at n = n3 = 4, \ y3(n) will start at n = n3 + nh = 4 + 0 = 4

x4(n) starts at n = n4 = 6, \ y4(n) will start at n = n4 + nh = 6 + 0 = 6

Note : 1. Here N1 = 8, N2 = 2, N3 = 2. \ (N2 – 1) = 2 – 1 = 1 and (N2 + N3 – 1) = 2 + 2 – 1 = 3

2. The boldfaced numbers in the tables are obtained by periodic extension.
3. For convenience of convolution the index n is replaced by m in x1(n), x2(n), x3 (n), x4 (n) and h(n).
mf
Convolution of Section 1
y1(n) = x1(n) ∗ h(n) = ∑
m = mi
x1(m) h((n − m))N
m –2 –1 0 1 2
2
x1(m) 1 –1 0 = ∑ x (m) h (m) ;
m = 0
1 n n = 0, 1, 2
h(m) –1 1 0
where hn (m) = h((n − m))N
h((–m))3 = h0(m) 0 1 –1 0 1
When n = 0 ; y1(0) = ∑ x (m) h (m) = −1 + 0 + 0 = –1
1 0
h((1 – m))3 = h1(m) 0 1 –1 0
When n = 1 ; y (1) = ∑ x (m) h (m) = 1+ 1 + 0 = 2
1 1 1
h((2 – m))3 = h2(m) 0 1 –1
When n = 2 ; y (2) = ∑ x (m) h (m) = 0 − 1 + 0 = −1
1 1 2
Convolution of Section 2

m –2 –1 0 1 2 3 4
x2(m) 2 –2 –1
h(m) –1 1 0
h(–m) 0 1 –1
h((2 – m))3 = h2(m) 0 1 –1 0 1
h((3 – m))3 = h3(m) 0 1 –1 0
h((4 – m))3 = h4(m) 0 1 –1
Chapter 2 - Discrete Time Signals and Systems 2. 90
mf 4
y 2 (n) = x 2(n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
2 N = ∑ x (m) h (m); n = 2,
m = 2
2 n 3, 4,

where hn (m) = h((n − m))N

When n = 2 ; y 2(2) =∑ x 2 (m) h2(m) = −2 + 0 − 1 = – 3
When n = 3 ; y (3) = ∑ x (m) h (m) =
2 2 3 2+ 2+ 0 = 4
When n = 4 ; y (4) = ∑ x (m) h (m) =
2 2 4 0 − 2 + 1 = −1

Convolution of Section 3

m –2 –1 0 1 2 3 4 5 6
x3(m) 3 –3 –2
h(m) –1 1 0
h(–m) 0 1 –1
h((4 – m))3 = h4(m) 0 1 –1 0 1
h((5 – m))3 = h5(m) 0 1 –1 0
h((6 – m))3 = h6(m) 0 1 –1

mf 6
y3(n) = x3 (n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
3 N = ∑ x (m) h (m) ;
m = 4
3 n n = 4, 5, 6

where hn (m) = h((n − m))N

When n = 4 ; y3(4) = ∑ x3(m) h4 (m) = −3 + 0 − 2 = –5
When n = 5 ; y (5) = ∑ x (m) h (m) =
3 3 5 3 +3+ 0 = 6
When n = 6 ; y (6) = ∑ x (m) h (m) =
3 3 6 0 − 3 + 2 = −1

Convolution of Section 4

m –2 –1 0 1 2 3 4 5 6 7 8
x4(m) 4 –4 –3
h(m) –1 1 0
h(–m) 0 1 –1
h((6 – m))3 = h6(m) 0 1 –1 0 1
h((7 – m))3 = h7(m) 0 1 –1 0
h((8 – m))3 = h8(m) 0 1 –1

mf 8
y 4 (n) = x 4 (n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
4 N = ∑ x (m) h (m) ; n = 6, 7, 8
m = 6
4 n

where hn (m) = h((n – m))N

∑
When n = 6 ; y 4 (6) = x 4 (m) h6 (m) = −4 + 0 − 3 = −7
When n = 7 ; y (7) = ∑ x (m) h (m) =
4 4 7 4+4+0 = 8
When n = 8 ; y (8) = ∑ x (m) h (m) =
4 4 8 0 − 4 + 3 = −1
2. 91 Digital Signal Processing
To Combine the Output of the Convolution of Each Section
It can be observed that the last sample in an output sequence overlaps with the first sample of next output
sequence. In overlap save method the overall output is obtained by combining the outputs of the convolution of
all sections. While combining the outputs, the overlapped last sample of every output sequence is discarded and
the remaining samples are simply saved as samples of y(n) as shown in the following table.

n 0 1 2 3 4 5 6 7 8
Note :
y1(n) –1 2 –1
Here y(n) is linear convolution
y2(n) –3 4 –1 of x(n) and h(n). It can be
y3(n) –5 6 –1 observed that the results of both
the methods are same except the
y4(n) –7 8 –1 last N2–1 samples.
y(n) –1 2 –3 4 –5 6 –7 8 *

\ y(n) = x(n) * h(n) = {–1, 2, –3, 4, –5, 6, –7, 8, *}

Example 2.29
Perform the linear convolution of the following sequences by a) Overlap add method and b) Overlap
save method.

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

Solution
a) Overlap Add Method

In this method the longer sequence is sectioned into sequences of size equal to smaller sequence. Here
x(n) is a longer sequence when compared to h(n). Hence x(n) is sectioned into sequences of size equal to h(n).

Given that x(n) = {1, 2, 3, –1, –2, –3, 4, 5, 6}. Let x(n) can be sectioned into three sequences, each
consisting of three samples of x(n) as shown below.

x1(n) = 1 ; n = 0 x2(n) = –1 ; n = 3 x3(n) = 4 ; n = 6

=2;n=1 = –2 ; n = 4 =5;n=7
=3;n=2 = –3 ; n = 5 =6;n=8

Let y1(n), y2(n) and y3(n) be the output of linear convolution of x1(n), x2(n) and x3(n) with h(n) respectively.

Here h(n) starts at n = nh = 0

x1(n) starts at n = n1 = 0, \ y1(n) will start at n = n1 + nh = 0 + 0 = 0

x2(n) starts at n = n2 = 3, \ y2(n) will start at n = n2 + nh = 3 + 0 = 3
x3(n) starts at n = n3 = 6, \ y3(n) will start at n = n3 + nh = 6 + 0 = 6
Here linear convolution of each section is performed between two sequences each consisting of three
samples. Hence each convolution output will consists of 3 + 3 – 1 = 5 samples. The convolution of each section
is performed by tabular method as shown below.

Note : 1.Here N1 = 9, N2 = 3, N3 = 3, \ (N2 – 1) = 3 – 1 = 2 and (N2 + N3 – 1) = 3 + 3 – 1 = 5.

2.The unfilled boxes in the table are considered as zero.
3.For convenience of convolution operation, the index n is replaced by m in x1(n), x2(n), x3(n)
and h(n).
Chapter 2 - Discrete Time Signals and Systems 2. 92
Convolution of Section 1

m –2 –1 0 1 2 3 4
+∞

x1(m) 1 2 3 y1(n) = x1(n) ∗ h(n) = ∑ x (m) h(n − m)

m = −∞
1

h(m) 2 1 –1 +∞

h(–m) = h0(m) –1 1 2
= ∑ x (m) h (m)
m = −∞
1 n

h(1 – m) = h1(m) –1 1 2 for n = 0, 1, 2, 3, 4

where hn (m) = h(n − m)
h(2 – m) = h2(m) –1 1 2
h(3 – m) = h3(m) –1 1 2
h(4 – m) = h4(m) –1 1 2

When n = 0 ; y1(0) = å x (m) h (m) = 0 + 0 + 2 + 0 + 0 = 2

1 o

When n = 1 ; y (1) = å x (m) h (m) = 0 + 1 + 4 + 0

1 1 1
= 5
When n = 2 ; y (2) = å x (m) h (m) = –1 + 2 + 6
1 1 2
= 7
When n = 3 ; y (3) = å x (m) h (m) = 0 – 2 + 3 + 0
1 1 3
= 1
When n = 4 ; y (4) = å x (m) h (m) = 0 + 0 – 3 + 0 + 0 = –3
1 1 4

Convolution of Section 2

m –2 –1 0 1 2 3 4 5 6 7
x2(m) –1 –2 –3
h(m) 2 1 –1
h(–m) = h0(m) –1 1 2
h(3 – m) = h3(m) –1 1 2
h(4 – m) = h4(m) –1 1 2
h(5 – m) = h5(m) –1 1 2
h(6 – m) = h6(m) –1 1 2
h(7 – m) = h7(m) –1 1 2

∞ ∞
y 2(n) = x 2(n) ∗ h(n) = ∑ x (m) h(n − m) = ∑ x (m) h (m); n = 3, 4, 5, 6, 7
m = −∞
2
m = −∞
2 n

where hn (m) = h(n − m)

∑ x (m) h (m) = 0 + 0 − 2 + 0 + 0 = –2
When n = 3 ; y 2 (3) = 2 3

When n = 4 ; y (4) = ∑ x (m) h (m) = 0 − 1 − 4 + 0

2 2 4= −5
When n = 5 ; y (5) = ∑ x (m) h (m) = 1 − 2 − 6
2 2 5= −7
When n = 6 ; y (6) = ∑ x (m) h (m) = 0 + 2 − 3 + 0
2 2 6= −1
When n = 7 ; y (7) = ∑ x (m) h (m) = 0 + 0 + 3 + 0 + 0 = 3
2 2 7
2. 93 Digital Signal Processing
Convolution of Section 3
m –2 –1 0 1 2 3 4 5 6 7 8 9 10
x3(m) 4 5 6
h(m) 2 1 –1
h(–m) = h0(m) –1 1 2
h(6 – m) = h6(m) –1 1 2
h(7 – m) = h7(m) –1 1 2
h(8 – m) = h8(m) –1 1 2
h(9 – m) = h9(m) –1 1 2
h(10 – m) = h10(m) –1 1 2
∞ ∞
y3(n) = x3(n) ∗ h(n) = ∑ x (m) h(n − m) = ∑ x (m) h (m) ;
m = –∞
3
m = –∞
3 n n = 6, 7, 8, 9, 10

where hn (m) = h(n − m)

When n = 6 ; y3 (6) =∑ x (m) h (m) = 0 + 0 + 8 + 0 + 0 = 8
3 6

When n = 7 ; y (7) = ∑ x (m) h (m) = 0 + 4 + 10 + 0

3 3 7 = 14
When n = 8 ; y (8) = ∑ x (m) h (m) = –4 + 5 + 12
3 3 8 = 13
When n = 9 ; y (9) = ∑ x (m) h (m) = 0 – 5 + 6 + 0
3 3 9 = 1
When n = 10 ; y (10) = ∑ x (m) h (m) = 0 + 0 – 6 + 0 + 0 = – 6
3 3 10

To Combine the Output of the Convolution of Each Section

It can be observed that the last N2 – 1 sample in an output sequence overlaps with the first N2 – 1 sample
of next output sequence. In this method, the overall output is obtained by combining the outputs of the convolution
of all sections. The overlapped portions (or samples) are added while combining the output.
The output of all sections can be represented in a table as shown below. Then the samples corresponding
to same value of n are added to get the overall output.
n 0 1 2 3 4 5 6 7 8 9 10
y1(n) 2 5 7 1 –3
y2(n) –2 –5 –7 –1 3
y3(n) 8 14 13 1 –6
y(n) 2 5 7 –1 –8 –7 7 17 13 1 –6

\ y(n) = x(n) * h(n) = {2, 5, 7, –1, –8, –7, 7, 17, 13, 1, – 6}

b) Overlap Save Method
In this method the longer sequence is sectioned into sequences of size equal to smaller sequence. The
number of samples that will be obtained in the output of linear convolution of each section is determined. Then
each section of longer sequence is converted to the size of output sequence using the samples of original longer
sequences. The smaller sequence is also converted to the size of output sequence by appending with zeros.
Then the circular convolution of each section is performed.
Here x(n) is a longer sequence when compared to h(n). Hence x(n) is sectioned into sequences of size
equal to h(n). Given that x(n) = {1, 2, 3, –1, –2, –3, 4, 5, 6}.
Let x(n) be sectioned into three sequences each consisting of three samples as shown below.
Let, N1 = Length of longer sequence
N2 = Length of smaller sequence
N3 = N2 = Length of each section of longer sequence.
Chapter 2 - Discrete Time Signals and Systems 2. 94
x1(n) = 1 ; n = 0 x2(n) = –1 ; n = 3 x3(n) = 4 ; n = 6
=2;n=1 = –2 ; n = 4 =5;n=7
=3;n=2 = –3 ; n = 5 =6;n=8
Let y1(n), y2(n) and y3(n) be the output of linear convolution of x1(n), x2(n) and x3(n) with h(n) respectively.
Here linear convolution of each section will result in an output sequence consisting of 3 + 3 – 1 = 5 samples.
Hence each section of longer sequence is converted to five sample sequence, using the samples of
original longer sequence as shown below. It can be observed that the first N2 – 1 samples of x2(n) is placed as
overlapping sample at the end of x1(n). The first N2 – 1 samples of x3(n) is placed as overlapping sample at the end
of x2(n). Since there is no fourth section, the overlapping samples of x3(n) are considered as zeros.
x1(n) = 1 ; n = 0 x2(n) = –1 ; n = 3 x3(n) = 4 ; n = 6
= 2;n=1 = –2 ; n = 4 =5;n=7
= 3;n=2 = –3 ; n = 5 =6;n=8
= –1 ; n = 3 = 4; n=6 =0;n=9
= –2 ; n = 4 = 5; n=7 = 0 ; n = 10
The sequence h(n) is also converted to five sample sequence by appending with zeros.
\ h(n) = {2, 1, –1, 0, 0}
Now perform circular convolution of each section with h(n). The output sequence obtained from circular
convolution will have five samples. The circular convolution of each section is performed by tabular method as
shown below.
Here h(n) starts at n = nh = 0
x1(n) starts at n = n1 = 0, \ y1(n) will start at n = n1 + nh = 0 + 0 = 0
x2(n) starts at n = n2 = 3, \ y2(n) will start at n = n2 + nh = 3 + 0 = 3
x3(n) starts at n = n3 = 6, \ y3(n) will start at n = n3 + nh = 6 + 0 = 6

Note : 1. Here N1 = 9, N2 = 3, N3 = 3 \ (N2 – 1) = 3 – 1 = 2 and [N2 + N3 – 1] = 3 + 3 – 1 = 5 samples.

2. The boldfaced numbers in the table are obtained by periodic extension.
3. For convenience of convolution operation the index n is replaced by m in x1(n), x2(n), x3(n) and h(n).
Convolution of Section 1

m –4 –3 –2 –1 0 1 2 3 4
x1(m) 1 2 3 –1 –2
h(m) 2 1 –1 0 0
h((–m))5 = h0(m) 0 0 –1 1 2 0 0 –1 1
h((1 – m))5 = h1(m) 0 0 –1 1 2 0 0 –1
h((2 – m))5 = h2(m) 0 0 –1 1 2 0 0
h((3 – m))5 = h3(m) 0 0 –1 1 2 0
h((4 – m))5 = h4(m) 0 0 –1 1 2

mf 4
y1(n) = x1(n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
1 N = ∑ x (m) h (m); n = 0, 1, 2, 3, 4
m = 0
1 n

where hn (m) = h((n − m))N

2. 95 Digital Signal Processing

When n = 0 ; y1(0) = å x (m)h (m) = 2 + 0 + 0 + 1 – 2 = 1

1 o

When n = 1 ; y (0) = å x (m)h (m) = 1 + 4 + 0 + 0 + 2 = 7

1 1 1

When n = 2 ; y (2) = å x (m)h (m) = –1 + 2 + 6 + 0 + 0 = 7

1 1 2

When n = 3 ; y (3) = å x (m)h (m) = 0 – 2 + 3 – 2 + 0 = – 1

1 1 3

When n = 4 ; y (4) = å x (m)h (m) = 0 + 0 – 3 – 1 – 4 = – 8

1 1 4

Convolution of Section 2

m –4 –3 –2 –1 0 1 2 3 4 5 6 7
x2(m) –1 –2 –3 4 5
h(m) 2 1 –1 0 0
h(–m) = h0(m) 0 0 –1 1 2
h((3 – m))5 = h3(m) 0 0 –1 1 2 0 0 –1 1
h((4 – m))5 = h4(m) 0 0 –1 1 2 0 0 –1
h((5 – m))5 = h5(m) 0 0 –1 1 2 0 0
h((6 – m))5 = h6(m) 0 0 –1 1 2 0
h((7 – m))5 = h7(m) 0 0 –1 1 2
mf 7
y 2 (n) = x 2(n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
2 N = ∑ x (m) h (m); n = 3, 4, 5, 6, 7
m = 3
2 n

where hn (m) = h((n − m))N

∑
When n = 3 ; y 2 (3) = x 2(m) h3 (m) = −2 + 0 + 0 – 4 + 5 = – 1
When n = 4 ; y 2(4) =∑ x (m) h (m) = −1 − 4 + 0 + 0 – 5 = −10
2 4

When n = 5 ; y (5) = ∑ x (m) h (m) = 1 − 2 − 6 + 0 + 0 = − 7

2 2 5

When n = 6 ; y (6) = ∑ x (m) h (m) = 0 + 2 − 3 + 8 + 0 = 7

2 2 6

When n = 7 ; y (7) = ∑ x (m) h (m) = 0 + 0 + 3 + 4 + 10 = 17

2 2 7

Convolution of Section 3

m –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 9 10
x3(m) 4 5 6 0 0
h(m) 2 1 –1 0 0
h(–m) = h0(m) 0 0 –1 1 2
h((6 – m))5 = h6(m) 0 0 –1 1 2 0 0 –1 1
h((7 – m))5 = h7(m) 0 0 –1 1 2 0 0 –1
h((8 – m))5 = h8(m) 0 0 –1 1 2 0 0
h((9 – m))5 = h9(m) 0 0 –1 1 2 0
h((10 – m))5 = h10(m) 0 0 –1 1 2

mf 10
y3 (n) = x 3(n) ∗ h(n) = ∑ x (m) h((n − m))
m = mi
3 N = ∑ x (m) h (m)
m = 6
3 n ; n = 6, 7, 8, 9, 10

where hn (m) = h((n − m))N

Chapter 2 - Discrete Time Signals and Systems 2. 96

When n = 6 ; y3 (6) = ∑ x (m) h (m) = 8 + 0 + 0 + 0 + 0 = 8

3 6

When n = 7 ; y (7) = ∑ x (m) h (m) = 4 + 10 + 0 + 0 + 0 = 14

3 3 7

When n = 8 ; y (8) = ∑ x (m) h (m) = – 4 + 5 + 12 + 0 + 0 = 13

3 3 8

When n = 9 ; y (9) = ∑ x (m) h (m) = 0 – 5 + 6 + 0 + 0 = 1

3 3 9

When n = 10 ; y (10) = ∑ x (m) h (m) = 0 + 0 – 6 + 0 + 0

3 3 10 = –6
To Combine the Output of Convolution of Each Section
It can be observed that the last N2–1 samples in an output sequence overlaps with the first N2–1 samples
of next output sequence. In overlap save method the overall output is obtained by combining the outputs of the
convolution of all sections. While combining the outputs, the overlapped first N2–1 samples of every output
sequence is discarded and the remaining samples are simply saved as samples of y(n) as shown in the following
table.
n 0 1 2 3 4 5 6 7 8 9 10
y1(n) 1 7 7 –1 –8
y2(n) –1 10 –7 7 17
y3(n) 8 14 13 1 –6
y(n) * * 7 –1 –8 –7 7 17 13 1 –6

\ y(n) = x(n) * h(n) = {*, *, 7, –1, –8, –7, 7, 17, 13, 1, –6}
Note : Here y(n) is linear convolution of x(n) and h(n). It can be observed that the results of both the
methods are same except the first N2 – 1 samples.

2.12 Inverse System and Deconvolution

2.12.1 Inverse System
The inverse system is used to recover the input from the response of a system. For a given system, the
inverse system exists, if distinct inputs to a system leads to distinct outputs. The inverse systems exists for
all LTI systems.
The inverse system is denoted by H–1. If x(n) is input and y(n) is the output of a system, then y(n) is
the input and x(n) is the output of its inverse system.
x (n) y (n) y (n) w (n) = x(n)
H H −1

F ig 2.3 5 a : System . F ig 2.3 5 b : In verse system .

F ig 2.3 5 : A system and its inverse system .

Let h(n) be the impulse response of a system and h¢(n) be the impulse response of inverse system. Let
us connect the system and its inverse in cascade as shown in fig 2.36.
Identity sy s tem

H H
-1

y (n)
x(n) h(n) h ’(n) w (n) = x(n)

F ig 2.3 6 : C a sca d e co n n ectio n o f a system a n d its in v erse.

2. 97 Digital Signal Processing
Now it can be proved that,
h(n) * h¢(n) = d(n) .....(2.60)
Therefore the cascade of a system and its inverse is identity system.
Proof :
With reference to fig 2.36 we can write,
y(n) = x(n) * h(n) .....(2.61)
w(n) = y(n) * h¢(n) .....(2.62)
On substituting for y(n) from equation (2.61) in equation (2.62) we get,
w(n) = x(n) * h(n) * h¢(n) .....(2.63)
In equation (2.63),
if, h(n) * h¢(n) = d(n), then, x(n) * d(n) = x(n)
In a inverse system, w(n) = x(n), and so,

h(n) * h¢(n) = d(n). Hence proved.

2.12.2 Deconvolution
In an LTI system the response y(n) is given by convolution of input x(n) and impulse response h(n).
i.e., y(n) = x(n) * h(n)
The process of recovering the input from the response of a system is called deconvolution. (or the
process of recovering x(n) from x(n) * h(n) is called deconvolution).
When the response y(n) and impulse response h(n) are available, then the input x(n) can be computed
using the equation (2.64).

x(n) =
1 LM n −1
y(n) − ∑ x(m) h(n − m)
OP .....(2.64)
h(0) MN m= 0 PQ
Proof :
Let x(n) and h(n) be finite duration sequences starting from n = 0. Consider the matrix method
of convolution of x(n) and h(n) shown below.

h(0) h(1) h(2) h(3)

x(0) x(0)h(0) x(0)h(1) x(0)h(2) x(0)h(3)

x(1) x(1)h(0) x(1)h(1) x(1)h(2) x(1)h(3)

x(2) x(2)h(0) x(2)h(1) x(2)h(2) x(2)h(3)

x(3) x(3)h(0) x(3)h(1) x(3)h(2) x(3)h(3)

Chapter 2 - Discrete Time Signals and Systems 2. 98
From the above two-dimensional array we can write,

y(0)
y( n) = x(0) h(0) ⇒ x(0) =
h(0)
y(1) − x(0) h(1)
y(1) = x(1) h(0)+ x(0) h(1) ⇒ x(1) =
h(0)
y(2) − x(0) h(2) − x(1) h(1)
y(2) = x(2) h(0)+ x(1) h(1)+ x(0) h(2) ⇒ x(2) =
h(0)
y(3) − x(0) h(3) − x(1) h(2) − x(2) h(1)
y(3) = x(3) h(0)+ x(2) h(1)+ x(1) h(2)+ x(0) h(3) ⇒ x(3) =
h(0)
and so on.
From the above analysis, in general for any value of n, the x(n) is given by,

y(n) − x(0) h(n) − x(1) h(n − 1) − ...... − x(n − 1) h(1)

x( n) =
h(0)
1 LM n−1 O
∴ x(n) =
h(0)
y(n) −
MN ∑ x(m) h(n − m)PP
m =0 Q
Example 2.30
n
A discrete time system is defined by the equation, y(n) = ∑
m =0
x(m) ; for n ≥ 0. Find the inverse system.

Solution
n
Given that, y(n) = ∑ x(m)
m=0

0
When n = 0; y(0) = ∑ x(m) = x(0)
m =0
1
When n = 1; y(1) =
m=0
∑ x(m) = x(0) + x(1) = y(0) + x(1)

2
When n = 2; y(2) = ∑
m =0
x(m) = x(0) + x(1) + x(2) = y(1) + x(2)

3
When n = 3; y(3) = ∑
m=0
x(m) = x(0) + x(1) + x(2) + x(3) = y(2) + x(3)

and so on,

From the above analysis we can write,

x(0) = y(0) ; x(1) = y(1) – y(0) ; x(2) = y(2) – y(1) ; x(3) = y(3) – y(2) and so on,

In general for any value of n, the signal x(n) can be written as,

x(n) = y(n) – y(n –1)

Therefore the inverse system is defined by the equation,

x(n) = y(n) – y(n –1)

­
2. 99 Digital Signal Processing

Example 2.31
When a discrete time system is excited by an input x(n), the response is y(n) = { 2, 5, 11, 17, 13, 12 }
­--
If the impulse response of the system is h(n) = { 2, 1, 3 }, then what will be the input to the system?
-
Solution
Let N1 be number of samples in x(n) and N2 be number of samples in h(n), then the number of samples N3
in y(n) is given by,
N3 = N1 + N2 – 1
\ N1 = N3 – N2 + 1 = 6 – 3 + 1 = 4 samples
Therefore x(n) is 4 sample sequence.
Each sample of x(n) is given by,

1 LMy(n) − x(m) h(n − m)OP

n − 1
x(n) =
h(0) MN ∑m=0 PQ
y(0) 2
When n = 0 ; x(0) = = =1
h(0) 2
1 LMy(1) − x(m) OP
When n = 1 ; x(1) =
h(0) MN ∑
m=0
h(1 − m)
PQ
1 1
= y(1) − x(0) h(1) = 5 − 1× 1 = 2
h(0) 2
1 LMy(2) − x(m) h(2 − m) OP
1
When n = 2 ; x(2) =
h(0) MN ∑
m =0 PQ
1 1
= y(2) − x(0) h(2) − x(1) h(1) = 11 − 1 × 3 − 2 × 1 = 3
h(0) 2
1 LMy(3) − x(m) h(3 − m) OP
2
When n = 3 ; x(3) =
h(0) MN ∑
m=0 PQ
1 1
= y(3) − x(0) h(3) − x(1) h(2) − x(2) h(1) = 17 − 1 × 0 − 2 × 3 − 3 × 1 = 4
h(0) 2
∴ x(n) = {x(0), x(1), x(2), x(3)} = {1, 2, 3, 4}
A

2.13 Correlation, Crosscorrelation and Autocorrelation

The correlation of two discrete time sequences x(n) and y(n) is defined as,
+∞
.....(2.65)
rxy (m) = ∑ x(n) y(n − m)
n = −∞
where rxy(m) is the correlation sequence obtained by correlation of x(n) and y(n) and m is the variable used
for time shift. The correlation of two different sequences is called crosscorrelation and the correlation of a
sequence with itself is called autocorrelation. Hence autocorrelation of a discrete time sequence is defined as,
+∞
rxx (m) = ∑ x(n) x(n − m) .....(2.66)
n = −∞
If the sequence x(n) has N1 samples and sequence y(n) has N2 samples then the crosscorrelation
sequence rxy(m) will be a finite duration sequence consisting of N1 + N2 – 1 samples.If the sequence x(n) has
N samples, then and the autocorrelation sequence rxx(m) will be a finite duration sequence consisting of
2N – 1 samples.
Chapter 2 - Discrete Time Signals and Systems 2. 100
In the equation (2.65), the sequence x(n) is unshifted and the sequence y(n) is shifted by m units of
time for correlation operation. The same results can be obtained if the sequence y(n) is unshifted and the
sequence x(n) is shifted opposite to that of earlier case by m units of time, hence the crosscorrelation
operation can also be expressed as,
+∞
.....(2.67)
rxy (m) = ∑ x(n + m) y(n)
n = −∞

2.13.1 Procedure for Evaluating Correlation

Let, x(n) = Discrete time sequence with N1 samples
y(n) = Discrete time sequence with N2 samples
Now the correlation of x(n) and y(n) will produce a sequence rxy(m) consisting of N1+N2–1 samples.
Each sample of rxy(m) can be computed using the equation (2.65). The value of rxy(m) at m = q is obtained by
replacing m by q, in equation (2.65).
+∞
.....(2.68)
∴ rxy (q) = ∑ x(n) y(n − q)
n = −∞

The evaluation of equation (2.68) to determine the value of rxy(m) at m = q involves the following three
steps.
1. Shifting : Shift y(n) by q times to the right if q is positive, shift y(n) by q times to the
left if q is negative to obtain y(n - q).
2. Multiplication : Multiply x(n) by y(n - q) to get a product sequence. Let the product
sequence be vq(n). Now, vq(n) = x(n) × y(n - q).
3. Summation : Sum all the values of the product sequence vq(n) to obtain the value of
rxy(m) at m = q. [i.e., rxy(q)].
The above procedure will give the value rxy(m) at a single time instant say m = q. In general we are
interested in evaluating the values of the sequence rxy(m) over all the time instants in the range -¥ < m < ¥ .
Hence the steps 1, 2 and 3 given above must be repeated, for all possible time shifts in the range -¥ < m < ¥ .
In the correlation of finite duration sequences it is possible to predict the start and end of the resultant
sequence. If x(n) is N-point sequence and starts at n = n1 and if y(n) is N2-point sequence and starts at n = n2 then,
the initial value of m = mi for rxy(m) is mi = n1 – (n2 + N2 – 1). The value of x(n) for n < n1 and the value of y(n) for
n < n2 are then assumed to be zero.The final value of m = mf for rxy(m) is mf = mi + (N1+N2– 2).
The correlation operation involves all the steps in convolution operation except the folding.
Hence it can be proved that the convolution of x(n) and folded sequence y(-n) will generate the crosscorrelation
sequence rxy(m).
i.e., r (m) = x(n) * y(-n) .....(2.69)
xy

The procedure given above can be used for computing autocorrelation of x(n). For computing
autocorrelation using equation (2.68) replace y(n – q) by x(n – q). Similarly when equation (2.69) is used,
replace y(–n) by x(–n).
The autocorrelation of N-point sequence x(n) will give 2N –1 point autocorrelation sequence.
If x(n) starts at n = nx then initial value of m = mi for rxx(m) is mi = – (N –1). The final value of m = mf for rxx(m)
is mf = mi + (2N–2).
2. 101 Digital Signal Processing
Properties of Correlation
1. The crosscorrelation sequence rxy(m) is simply a folded version of ryx(m),
i.e., rxy(m) = ryx(-m)
Similarly for autocorrelation sequence,
rxx(m) = rxx(-m)
Hence autocorrelation is an even function.
2. The crosscorrelation sequence satisfies the condition,

rxy (m) ≤ rxx ( 0) ryy ( 0) = Ex Ey

where, Ex and Ey are energy of x(n) and y(n) respectively.

On applying the above condition to autocorrelation sequence we get,
rxx (m) ≤ rxx (0) = E x

From the above equations we infer that the crosscorrelation sequence and autocorrelation
sequences attain their respective maximum values at zero shift/lag.
3. Using the maximum value of crosscorrelation sequence, the normalized crosscorrelation sequence
is defined as,
rxy (m)
ρxy (m) ≤
rxx (0) ryy (0)
Using the maximum value of autocorrelation sequence, the normalized autocorrelation sequence
is defined as,
rxx (m)
ρxx (m) ≤
rxx ( 0)

Methods of Computing Correlation

Method 1: Graphical Method
Let x(n) and y(n) be the input sequences and rxy(m) be the output sequence.
1. Sketch the graphical representation of the input sequences x(n) and y(n).
2. Shift the sequence y(n) to the left graphically so that the product of x(n) and shifted y(n) gives
only one nonzero sample. Now multiply x(n) and shifted y(n) to get a product sequence, and then
sum up the samples of product sequence, which is the first sample of output sequence.
3. To get the next sample of output sequence, shift y(n) of previous step to one position right
and multiply the shifted sequence with x(n) to get a product sequence. Now the sum of the
samples of product sequence gives the second sample of output sequence.
4. To get subsequent samples of output sequence, the step 3 is repeated until we get a nonzero
product sequence.
Method 2: Tabular Method
The tabular method is same as that of graphical method, except that the tabular representation of the
sequences are employed instead of graphical representation. In tabular method, every input sequence and
shifted sequence is represented on a row in a table.
Chapter 2 - Discrete Time Signals and Systems 2. 102
Method 3: Matrix Method
Let x(n) and y(n) be the input sequences and rxy(m) be the output sequence. We know that the
convolution of x(n) and folded sequence y(-n) will generate the crosscorrelation sequence rxy(m). Hence fold
y(n) to get y(-n), and compute convolution of x(n) and y(-n) by matrix method.
In matrix method one of the sequence is represented as a row and the other as a column as shown
below.
y (0) y (−1 ) y (−2 ) y (−3 )

x (0) x (0)y(0) x (0)y( −1) x (0)y( −2) x (0)y( −3)

x (1) x (1)y(0) x (1)y( −1) x (1)y( −2) x (1)y( −3)

x (2) x (2)y(0) x (2)y( −1) x (2)y( −2) x (2)y( −3)

x (3) x (3)y(0) x (3)y( −1) x (3)y( −2) x (3)y( −3)

Multiply each column element with row elements and fill up the matrix array.
Now the sum of the diagonal elements gives the samples of output sequence rxy(m). (The sum of the
diagonal elements are shown below for reference).
:
:
rxy(0) = ..... + x(0) y(0) + .....
rxy(1) = ..... + x(1) y(0) + x(0 ) y(-1) + .....
rxy(2) = ..... + x(2) y(0) + x(1) y(-1) + x(0) y(-2) + .....
rxy(3) = ..... + x(3) y(0) + x(2) y(-1) + x(1) y(-2) + x(0) y(-3) + .....
:
:

Example 2.32
Perform crosscorrelation of the sequences, x(n) = {1, 1, 2, 2} and y(n) = {1, 0.5, 1}.

Solution
Let rxy(m) be the crosscorrelation sequence obtained by crosscorrelation of x(n) and y(n).
The crosscorrelation sequence rxy(m) is given by,
+∞
rxy = ∑ x(n) y(n − m)
n = −∞

The x(n) starts at n = 0 and has 4 samples.

2. 103 Digital Signal Processing
\ n1 = 0, N1 = 4
The y(n) starts at n = 0 and has 3 samples.
\ n2 = 0, N2 = 3
Now, rxy(m) will have N1 + N2 – 1 = 4 + 3 – 1 = 6 samples.
The initial value of m = mi = n1 – (n2 + N2 –1)
= 0 – (0 + 3 – 1) = – 2
The final value of m = mf = mi + (N1 + N2 – 2)
= –2 + (4 + 3 – 2) = 3
In this example the correlation operation is performed by three methods.
Method 1 : Graphical Method
The graphical representation of x(n) and y(n) are shown below.

x (n ) y (n )
2 2
1 1 1 1
0.5
0 1 2 3 n 0 1 2 n
F ig 1. F ig 2.
The 6 samples of rxy(m) are computed using the equation,
+∞ +∞
rxy (m) = ∑
n = −∞
x(n) y(n − m) = ∑
n = −∞
x(n) y m (n) ; where y m (n) = y(n − m)

The computation of each sample of rxy(n) using the above equation are graphically shown in fig 3 to fig 8.
The graphical representation of output sequence is shown in fig 9.
+∞ +∞ +∞
When m = −2 ; rxy ( −2) = ∑
n = −∞
x(n) y(n − ( −2)) = ∑
n = −∞
x(n) y −2 (n) = ∑
n = −∞
v − 2 (n)

x (n ) y −2 (n) v −2 (n)

2 2
X ⇒
1 1 1 1 1
0.5
0 1 2 3 n −2 −1 0 n −2 −1 0 1 2 3 n
T he su m o f pro du ct s eq ue n ce
v −2 (n ) giv es rxy ( −2)
F ig 3 : C o m p uta tio n o f r x y ( −2).
∴ rxy ( −2) = 0 + 0 + 1 + 0 + 0 + 0 = 1

+∞ +∞ +∞
When m = −1 ; rxy (−1) = ∑
n = −∞
x(n) y(n − ( −1)) = ∑
n = −∞
x(n) y −1(n) = ∑
n = −∞
v − 1(n)

x (n ) y −1 (n) v −1 (n )

2 2
X ⇒ 1
1 1 1 1
0.5 0.5
0 1 2 3 n −1 0 1 n −1 0 1 2 3 n
T he s um of pro du c t s e qu en c e
v −1(n ) g iv e s rx y ( −1 )
F ig 4 : C om p u ta tio n of rx y ( −1 ).
∴ rxy ( −1) = 0 + 0 .5 + 1 + 0 + 0 = 1 .5
Chapter 2 - Discrete Time Signals and Systems 2. 104
+∞ +∞ +∞
When m = 0 ; rxy (0) = ∑ x(n) y(n) = ∑ x(n) y 0 (n) = ∑ v 0 (n)
n = −∞ n = −∞ n = −∞

x (n ) y 0 (n ) v 0 (n )

2 2 2
X ⇒ 1
1 1 1 1
0.5 0.5
0 1 2 3 n 0 1 2 n 0 1 2 3 n
T he su m o f pro du ct s eq u en ce

F ig 5 : C o m p u ta tio n o f rxy (0 ). v 0 (n ) g ive s rx y (0)

∴ rx y (0) = 1 + 0 .5 + 2 + 0 = 3.5
+∞ +∞ +∞
When m = 1 ; rxy (1) = ∑
n = −∞
x(n) y(n − 1)) = ∑
n = −∞
x(n) y1(n) = ∑
n = −∞
v1(n)

x (n ) y 1 (n ) v 1 (n )

2 2 2
X ⇒
1 1 1 1 1 1
0.5
0 1 2 3 n 0 1 2 3 n 0 1 2 3 n
T he su m o f pro du ct s eq u en ce
F ig 6 : C o m p u ta tio n of rxy (1 ). v 1 (n) giv es rxy (1 )
∴ rx y (1 ) = 0 + 1 + 1 + 2 = 4

+∞ +∞ +∞
When m = 2 ; rxy (2) = ∑
n = −∞
x(n) y(n − 2) = ∑
n = −∞
x(n) y 2 (n) = ∑
n = −∞
v 2 (n)

x (n ) y 2 (n ) v 2 (n )

2 2 2
X ⇒ 1
1 1 1 1
0.5
0 1 2 3 n 0 1 2 3 4 n 0 1 2 3 4 n
T he su m o f pro du ct s eq u en ce
F ig 7 : C o m pu ta tio n o f rx y (2). v 2 (n ) giv es rx y (2 )
∴ rx y (2 ) = 0 + 0 + 2 + 1 + 0 = 3

+∞ +∞ +∞
When m = 3 ; rxy (3) = ∑
n = −∞
x(n) y(n − 3) = ∑
n = −∞
x(n) y 3(n) = ∑
n = −∞
v 3(n)

x (n ) y 3 (n ) v 3 (n )

2 2
X ⇒ 2
1 1 1 1
0.5
0 1 2 3 n 0 1 2 3 4 5 n 0 1 2 3 4 5 n
T he su m o f pro du ct s eq u en ce
F ig 8 : C o m p uta tio n o f rxy (3 ). v 3 (n ) giv es rx y (3 )
∴ rx y (3) = 0 + 0 + 0 + 2 + 0 + 0 = 2
2. 105 Digital Signal Processing
The crosscorrelation sequence, rxy(m) = {1, 1.5, 3.5, 4, 3, 2}
-
r x y (m )

4 ­
3.5
3

2
1.5
1

−2 −1 0 1 2 3 m
F ig 9 : G ra p h ic a l rep resenta tio n o f rxy (m ).
Method 2: Tabular Method
The given sequences and the shifted sequences can be represented in the tabular array as shown below.

n –2 –1 0 1 2 3 4 5
x(n) 1 1 2 2
y(n) 1 0.5 1
y(n –(–2)) = y–2(n) 1 0.5 1
y(n –(–1)) = y–1(n) 1 0.5 1
y(n) = y0(n) 1 0.5 1
y(n – 1) = y1(n) 1 0.5 1
y(n – 2) = y2(n) 1 0.5 1
y(n – 3) = y3(n) 1 0.5 1

Note: The unfilled boxes in the table are considered as zeros.

Each sample of rxy(m) is given by,
+∞ +∞
rxy (m) = ∑
n = −∞
x(n) y(n − m) = ∑
n = −∞
x(n) y m (n) ; where ym (n) = y(n − m)

To determine a sample of rxy(m) at m = q, multiply the sequence x(n) and yq(n) to get a product sequence
[i.e., multiply the corresponding elements of the row x(n) and yq(n)]. The sum of all the samples of the product
sequence gives rxy(q).
3
When m = −2 ; rxy (−2) = ∑
n = −2
x(n) y −2(n) = 0 + 0 + 1+ 0 + 0 + 0 = 1

3
When m = −1 ; rxy (−1) = ∑ x(n) y
n = −1
−1(n) = 0 + 0.5 + 1+ 0 + 0 = 1.5

3
When m = 0 ; rxy (0) = ∑ x(n) y
n =0
0 (n) = 1+ 0.5 + 2 + 0 = 3.5

3
When m = 1 ; rxy (1) = ∑
n=0
x(n) y1(n) = 0 + 1+ 1+ 2 =4

4
When m = 2 ; rxy (2) = ∑ x(n) y (n)
n =0
2 = 0 + 0 + 2 + 1+ 0 =3

5
When m = 3 ; rxy (3) = ∑ x(n) y (n)
n=0
3 = 0 +0 + 0 + 2+ 0 +0 = 2

∴ Crosscorrelation sequence, rxy (m) = {1, 1.5, 3.5, 4, 3, 2}

A
Chapter 2 - Discrete Time Signals and Systems 2. 106
Method 3: Matrix Method
Given that, x(n) = {1, 1, 2, 2} ; y(n) = {1, 0.5, 1} ; \ y(–n) = {1, 0.5, 1}
- - -
The sequence x(n) is arranged as a column and the folded sequence y(–n) is arranged as a row as shown
below. The elements of the two-dimensional array are obtained by multiplying the corresponding row element
with column element. The sum of the diagonal elements gives the samples of the crosscorrelation sequence,
rxy(m).

y ( −n ) y ( −n )
x (n) 1 0.5 1 x (n) 1 0.5 1

1 1 ×1 1 × 0.5 1 ×1 1 1 0.5 1

1 1 ×1 1 × 0.5 1 ×1 ⇒ 1 1 0.5 1

2 2 ×1 2 × 0.5 2 ×1 2 2 1 2

2 2 ×1 2 × 0.5 2 ×1 2 2 1 2

rxy (–2) = 1 ; rxy (–1) = 1 + 0.5 = 1.5 ; rxy(0) = 2 + 0.5 + 1 = 3.5

rxy(1) = 2 + 1 + 1 = 4 ; rxy(2) = 1 + 2 = 3 ; rxy(3) = 2
\ rxy(m) = {1, 1.5, 3.5, 4, 3, 2}
-

Example 2.33
Determine the autocorrelation sequence for x(n) = {1, 2, 3, 4}.
Solution
Let, rxx(m) be the autocorrelation sequence.
The autocorrelation sequence rxx(m) is given by,
+∞
rxx (m) = ∑
n = −∞
x(n) x(n − m)

The x(n) starts at n = 0 and has 4 samples.

\ nx = 0 and N = 4
Now, rxx(m) will have, 2N – 1 = 2 ´ 4 – 1 = 7 samples.
The initial value of m = mi = – (N – 1) = – (4 – 1) = –3
The final value of m = mf = mi + (2N – 2) = –3 + (2 ´ 4 – 2) = 3
The autocorrelation is computed by tabular method. Hence the sequence x(n) and the shifted sequences
of x(n) are tabulated in the following table.

n –3 –2 –1 0 1 2 3 4 5 6
x(n) 1 2 3 4
x(n –(–3)) = x–3(n) 1 2 3 4
x(n –(–2)) = x–2(n) 1 2 3 4
x(n –(–1)) = x–1(n) 1 2 3 4
x(n) = x0(n) 1 2 3 4
x(n – 1) = x1(n) 1 2 3 4
x(n – 2) = x2(n) 1 2 3 4
x(n – 3) = x3(n) 1 2 3 4
2. 107 Digital Signal Processing
Each sample of rxx(m) is given by,
+∞ +∞
rxx (m) = ∑
n = −∞
x(n) x(n − m) = ∑
n = −∞
x(n) xm (n) ; where xm (n) = x(n − m)

To determine a sample of rxx(m) at m = q, multiply the sequence x(n) and xq(n) to get a product sequence
[i.e., multiply the corresponding elements of the row x(n) and xq(n)]. The sum of all the samples of the product
sequence gives rxx(q).
3
W hen m = − 3 ; rxx ( − 3) = ∑
n = −3
x(n) x − 3 (n ) = 0 + 0 + 0 + 4 + 0 + 0 + 0 = 4

3
W hen m = − 2 ; rxx ( − 2) = ∑
n = −2
x(n) x − 2 (n ) = 0 + 0 + 3 + 8 + 0 + 0 = 11

3
W hen m = − 1 ; rxx ( − 1 ) = ∑
n = −1
x(n) x − 1(n ) = 0 + 2 + 6 + 12 + 0 = 20

3
W hen m = 0 ; rxx ( 0 ) = ∑
n = 0
x(n) x 0 (n ) = 1 + 4 + 9 + 16 = 30

4
W hen m = 1 ; rxx (1 ) = ∑
n = 0
x(n) x 1(n ) = 0 + 2 + 6 + 12 + 0 = 20

5
W hen m = 2 ; rxx ( 2) = ∑
n = 0
x(n) x 2 (n ) = 0 + 0 + 3 + 8 + 0 + 0 = 11

6
W hen m = 3 ; rxx (3) = ∑
n = 0
x(n) x 3 (n ) = 0 + 0 + 0 + 4 + 0 + 0 + 0 = 4

∴ Autocorrelation sequence, rxx (m ) = {4, 11, 20, 30, 20, 11, 4}

A
2.14 Circular Correlation
The circular correlation of two periodic discrete time sequences x(n) and y(n) with periodicity of N
samples is defined as,
N−1
rxy (m) = ∑ x(n) y* ((n − m))
n=0
N
.....(2.70)

where, rxy ( m) is the sequence obtained by circular correlation

y*((n – m))N represents circular shift of y*(n)
m is a variable used for circular time shift
The circular correlation of two different sequences is called circular crosscorrelation and the circular
correlation of a sequence with itself is called circular autocorrelation. Hence circular autocorrelation of a
discrete time sequence is defined as,
N−1
rxx ( m) = ∑ x( n) x* ((n − m))
n=0
N
.....(2.71)

The output sequence obtained by circular correlation is also periodic sequence with periodicity of N
samples. Hence this correlation is also called periodic correlation. The circular correlation is defined for
periodic sequences. But circular correlation can be performed with non-periodic sequences by periodically
extending them.The circular correlation of two sequences requires that, at least one of the sequences should
be periodic. Hence it is sufficient if one of the sequences is periodically extended in order to perform circular
correlation.
Chapter 2 - Discrete Time Signals and Systems 2. 108
The circular correlation of finite duration sequences can be performed only if both the sequences
consists of same number of samples. If the sequences have different number of samples, then convert the
smaller size sequence to the size of larger size sequence by appending zeros.

In the equation (2.70), the sequence x(n) is unshifted and the sequence y*(n) is circularly shifted by
m units of time for correlation operation. The same results can be obtained if the sequence y*(n) is unshifted
and the sequence x(n) is circularly shifted opposite to that of earlier case by m units of time, hence the circular
correlation operation can also be expressed as,
N−1
rxy ( m) = ∑ x(( n + m))
n=0
N y* (n) .....(2.72)

Circular correlation basically involves the same three steps as that for correlation, namely shifting one
of the sequence, multiplying the two sequences and finally summing the values of product sequence. The
difference between the two is that in circular correlation the shifting (rotating) operations are performed in a
circular fashion by computing the index of one of the sequences by modulo-N operation. In correlation, there
is no modulo-N operation.
2.14.1 Procedure for Evaluating Circular Correlation
Let, x(n) and y(n) be periodic discrete time sequences with periodicity of N-samples. If x(n) and y(n)
are non-periodic then convert the sequences to N-sample sequence and periodically extend the sequence
y(n) with periodicity of N-samples.
Now the circular correlation of x(n) and y(n) will produce a periodic sequence rxy ( m) with periodicity
of N-samples. The samples of one period of rxy ( m) can be computed using the equation (2.70).

The value of rxy ( m) at m = q is obtained by replacing m by q, in equation (2.70), as shown below.

N−1
rxy (q ) = ∑ x( n) y* ((n − q))
n=0
N
.....(2.73)

The evaluation of equation (2.73) to determine the value of rxy ( m) at m = q involves the following
four steps.
1. Conjugation : Take conjugate of y(n) to get y*(n). If y(n) is a real sequence then y*(n)
will be same as y(n). Represent the samples of one period of the sequences
x(n) and y*(n) on circles.
2. Rotation : Rotate y*(n) by q times in anticlockwise if q is positive, rotate y*(n) by
q times in clockwise if q is negative to obtain y*((n – q))N.
3. Multiplication : Multiply x(n) by y*((n – q))N to get a product sequence. Let the product
sequence be vq(m). Now, vq(m) = x(n) × y*((n – q))N.
4. Summation : Sum up the samples of one period of the product sequence vq(m) to
obtain the value of rxy ( m) at m = q. [i.e., rxy (q ) ].

The above procedure will give the value of rxy ( m) at a single time instant say m = q. In general, we are
interested in evaluating the values of the sequence rxy ( m) in the range 0 < m < N - 1. Hence the steps 2 ,
3 and 4 given above must be repeated, for all possible time shifts in the range 0 < m < N - 1.
2. 109 Digital Signal Processing

2.14.2 Methods of Computing Circular Correlation

Method 1 : Graphical Method
In graphical method the given sequences are converted to same size and represented on circles. In
case of periodic sequences, the samples of one period are represented on circles. Let x(n) and y(n) be the
given real sequences. Let rxy ( m) be the sequence obtained by circular correlation of x(n) and y(n). The
following procedure can be used to get a sample of rxy ( m) at m = q.
1. Represent the sequences x(n) and y(n) on circles.
2. Rotate (or shift) the sequence y(n), q times to get the sequence y((n – q))N. If q is positive then
rotate (or shift) the sequence in anticlockwise direction and if q is negative then rotate (or shift)
the sequence in clockwise direction.

3. The sample of rxy (q ) at m = q is given by,

N−1 N−1
rxy (q ) = ∑ x(n) y((n − q )) N = ∑ x(n) yq ( n)
n= 0 n=0
where, yq ( n) = y(( n − q )) N

Determine the product sequence x(n)yq(n) for one period.

4. The sum of all the samples of the product sequence gives the sample rxy (q ) [i.e., rxy ( m) at m = q].

The above procedure is repeated for all possible values of m to get the sequence rxy ( m).
Method 2 : Using Tabular Array
Let x(n) and y(n) be the given real sequences. Let rxy ( m) be the sequence obtained by circular
correlation of x(n) and y(n). The following procedure can be used to get a sample of rxy ( m) at m = q.
1. Represent the sequences x(n) and y(n) as two rows of tabular array.
2. Periodically extend y(n). Here the periodicity is N, where N is the length of the given sequences.
3. Shift the sequence y(n), q times to get the sequence y((n – q))N. If q is positive then shift the
sequence to the right and if q is negative then shift the sequence to the left.
4. The sample of rxy (q ) at m = q is given by,
N−1 N−1
rxy (q) = ∑ x(n) y((n − q)) N = ∑ x(n) yq ( n)
n=0 n=0
where, yq (n) = y((n − q)) N
Determine the product sequence x(n)yq(n) for one period.
5. The sum of all the samples of the product sequence gives the sample rxy (q ) [i.e., rxy ( m)
at m = q].
The above procedure is repeated for all possible values of m to get the sequence rxy ( m).
Method 3: Using Matrices
Let x(n) and y(n) be the given N-point sequences. The circular correlation of x(n) and y(n) yields
another N-point sequence rxy ( m).
Chapter 2 - Discrete Time Signals and Systems 2. 110
In this method an N ´ N matrix is formed using the sequence y(n) as shown below. The sequence x(n)
is arranged as a column vector (column matrix) of order N ´ 1. The product of the two matrices gives the
resultant sequence rxy ( m).
L r (0) O
OP LM xx((10)) OP MM r (1) PP
xy
LMy(0) y(1) y ( 2) ..... y( N − 1) y( N )
xy
MMy(N) y( 0) y (1) ..... y ( N − 2) y( N − 1) P MM P M r ( 2) P
x( 2) P
MMy(MN − 1) y( N ) y(0) ..... y ( N − 3) y( N − 2) P
P × M
M M PP = MM M PP xy

MMy(2)
M M M M PP M M P MM M PP
y (3) y (4) ..... y( 0) y(1)
P MMx(N − 2)PP M r (N − 2)P
MNy(1) y( 2) y( 3) ..... y( N ) y( 0) PQ MNx(N − 1) PQ MM r (N − 1) PP xy

N Q xy

Example 2.34
Perform circular correlation of the two sequences, x(n) = {1, 1, 2, 1} and y(n)= {2, 3, 1, 1}
- -
Solution
Method 1:Graphical Method of Computing Circular Correlation
The given sequences are represented as points on circles as shown in fig 1 and 2.
x (1) = 1 y (1) = 3

x (2) = 2 x (n) x (0) = 1 y(2) = 1 y (n) y (0) = 2

x (3) = 1 y (3) = 1
F ig 1. F ig 2.
3 2 1 1

y ((n −0))4 y ((n −1) ) 4 y ((n −2) ) 4 y ((n −3) ) 4

1 2 ⇒ 3 1 ⇒ 2
= y 2 (n)
1 ⇒1 3
= y 0 (n) = y 1 (n) = y 3 (n)

1 1 3 2
F ig 3 : C ircu la rly sh ifted seq u e nces y (n -m ), for m = 0 , 1 , 2, 3 .
Let rxy (m) be the sequence obtained by circular correlation of x(n) and y(n). The given sequences are 4
sample sequences and so N = 4. Each sample of rxy (m) is given by the equation,
N − 1 N − 1
rxy (m) = ∑
n = 0
x(n) y((n − m))N = ∑
n = 0
x(n) y m (n), where y m (n) = y((n − m))N

Using the above equation, graphical method of computing each sample of rxy (m) are shown in fig 4 to fig 7.
3 3 3
When m = 0 ; rxy (0) = ∑
n = 0
x(n) y((n − 0))4 = ∑
n = 0
x(n) y 0 (n) = ∑ v (n)
n = 0
0

1 3 1 ×3 = 3

y 0 ( n)
2 x (n) 1 X 1 2 ⇒ 2 ×1 = 2 v 0 (n) 1 ×2 = 2

1 1 1 ×1 = 1

T h e s um of s am ple s o f v 0 (n) giv es rxy (0 )

F ig 4 : C om p u tation o f rxy (0 ).
∴ rx y (0 ) = 2 + 3 + 2 + 1 = 8
2. 111 Digital Signal Processing
3 3 3
When m = 1 ; rxy (1) = ∑ x(n) y((n − 1))
n = 0
4 = ∑ x(n) y (n) = ∑ v (n)
n = 0
1
n = 0
1

1 2 1 ×2 = 2

2 x (n) 1 X 3 y 1( n) 1 ⇒ 2 ×3 = 6 v 1( n) 1 ×1 = 1

1 1 1 ×1 = 1

T h e s um of s a m p le s o f v 1(n ) g ive s rx y (1)

F ig 5 : C o m p u ta tio n of rxy (1 ).
∴ rxy (1) = 1 + 2 + 6 + 1 = 10

3 3 3
When m = 2 ; rxy (2) = ∑ x(n) y((n − 2))
n = 0
4 = ∑ x(n) y (n) = ∑ v (n)
n = 0
2
n = 0
2

1 1 1×1 = 1

2 x (n) 1 X 2 y 2 (n) 1 ⇒ 2 ×2 = 4 v 2 ( n) 1 ×1 = 1

1 3 1 ×3 = 3
T he su m o f sa m ples of v 2 (n) g iv e s rxy (2)
F ig 6 : C o m pu ta tio n o f rxy (2 ).
∴ rxy (2) = 1 + 1 + 4 + 3 = 9

3 3 3
When m = 3 ; rxy (3) = ∑ x(n) y((n − 3))
n = 0
4 = ∑ x(n) y (n) = ∑ v (n)
n = 0
3
n = 0
3

1 1 1 ×1 = 1

2 x (n) 1 X 1 y 3 (n) 3 ⇒ 2 ×1 = 2 v 3 (n) 1 ×3 = 3

1 2 1 ×2 = 2

T h e s um of s am ple s o f v 3 (n) g iv e s rxy (3)

F ig 7 : C o m p uta tio n o f rxy (3 ).
∴ rx y (3) = 3 + 1 + 2 + 2 = 8

\ rxy (m) = {8, 10, 9, 8}

-­

Method 2 : Circular Correlation Using Tabular Array

The given sequences are represented in the tabular array as shown below. Here the shifted sequences
ym(n) are periodically extended with a periodicity of N = 4. Let rxy (m) be the sequence obtained by circular
correlation of x(n) and y(n). Each sample of rxy (m) is given by the equation,

N − 1 N − 1
rxy (m) = ∑
n = 0
x(n) y((n − m))N = ∑
n = 0
x(n) y m (n), where y m (n) = y((n − m))N

Note : The boldfaced numbers are samples obtained by periodic extension.

Chapter 2 - Discrete Time Signals and Systems 2. 112

n 0 1 2 3 4 5 6

x(n) 1 1 2 1

y(n) 2 3 1 1

y0((n – 0))4 = y0(n) 2 3 1 1 2 3 1

y1((n – 1))4 = y1(n) 1 2 3 1 1 2 3

y2((n – 2))4 = y2(n) 1 1 2 3 1 1 2

y3((n – 3))4 = y3(n) 3 1 1 2 3 1 1

To determine a sample of rxy (m) at m = q, multiply the sequence, x(n) and y q (n), to get a product sequence
x(n) xq(n) [i.e., multiply the corresponding elements of the row x(n) and yq(n)]. The sum of all the samples of the
product sequence gives rxy (m).
3
When m = 0 ; rxy (0) = ∑ x(n) y
n=0
0 (n)

= x(0) y 0 (0) + x(1) y 0 (1) + x(2) y 0 (2) + x(3) y 0 (3)

= 1× 2 + 1× 3 + 2 × 1+ 1× 1 = 2 + 3 + 2 + 1 = 8

The samples of rxy (m) for other values of m are calculated as shown for m = 0.
3
When m = 1; rxy (1) = ∑
n=0
x(n) y1(n) = 1 + 2 + 6 + 1 = 10

3
When m = 2; rxy (2) = ∑ x(n) y 2(n) = 1 + 1 + 4 + 3 = 9
n =0
3
When m = 3; rxy (3) = ∑
n =0
x(n) y3(n) = 3 + 1 + 2 + 2 = 8

l
∴ rxy (m) = 8, 10, 9, 8 q
A
Method 3 : Circular Correlation Using Matrices

The sequence rxy (m) can be arranged as a column vector of order N ´ 1 and using the samples of y(n) the
N ´ N matrix is formed as shown below. The product of the two matrices gives the sequence rxy (m).

LMy(0) y(1) y(2) y(3) OP LMx(0)OP LMr xy (0) OP

MMy(3) y(0) y(1) y(2)P MMx(1) PP =
MMrxy (1) P
y(1) P (2) PP
MMy(y(12)) y(3)
y(2)
y(0)
y(3) y(0)PQ
P MMx(x(32))PP MMr
xy

(3) PQ
N N Q Nr
xy

LM2 3 1 1OP LM1OP LM8 OP

MM1 2 3 1 PP MM1PP = MM10PP
MMN13 11 2
1
3
2
PPQ MMN12PPQ MMN98 PPQ
\ rxy (m) = {8, 10, 9, 8}
-
2. 113 Digital Signal Processing

2.15. Summary of Important Concepts

1. The discrete signal is a function of a discrete independent variable.
2. In a discrete time signal, the value of discrete time signal and the independent variable time are discrete.
3. The digital signal is same as discrete signal except that the magnitude of the signal is quantized.
4. A discrete time sinusoid is periodic only if its frequency is a rational number.
5. Discrete time sinusoids whose frequencies are separated by an integer multiple of 2p are identical.
6. The sampling is the process of conversion of continuous time signal into discrete time signal.
7. The time interval between successive samples is called sampling time or sampling period.
8. The inverse of sampling period is called sampling frequency.
9. The phenomenon of high frequency component getting the identity of low-frequency component during
sampling is called aliasing.
10. For analog signal with maximum frequency Fmax, the sampling frequency should be greater than 2Fmax.
11. When sampling frequency Fs is equal to 2Fmax, the sampling rate is called Nyquist rate.
12. The signals that can be completely specified by mathematical equations are called deterministic signals.
13. The signals whose characteristics are random in nature are called nondeterministic signals.
14. A signal x(n) is periodic with periodicity of N samples if x(n + N) = x(n).
15. When a signal exhibits symmetry with respect to n = 0 then it is called an even signal.
16. When a signal exhibits antisymmetry with respect to n = 0, then it is called an odd signal.
17. When the energy E of a signal is finite and nonzero, the signal is called energy signal.
18. When the power P of a signal is finite and nonzero, the signal is called power signal.
19. For energy signals, the energy will be finite and average power will be zero.
20. For power signals the average power is finite and energy will be infinite.
21. A signal is said to be causal, if it is defined for n ³ 0.
22. A signal is said to be noncausal, if it is defined for both n ≤ 0 and n > 0.
23. A signal is said to be anticausal, if it is defined for n ≤ 0.
24. A discrete time system is a device or algorithm that operates on a discrete time signal.
25. When a system satisfies the properties of linearity and time invariance, it is called an LTI system.
26. When the input to a discrete time system is unit impulse d(n), the output is called impulse response, h(n).
27. In a static or memoryless system, the output at any instant n depends on input at the same time.
28. A system is said to be time invariant if its input-output characteristics do not change with time.
29. A linear system is one that satisfies the superposition principle.
30. A system is said to be causal if the output does not depends on future inputs/outputs.
31. When a system output at any time n depends on future inputs/outputs, it is called a noncausal system.
32. System is said to be BIBO stable if and only if every bounded input produces a bounded output.
33. When a system output at any time n depends on past outputs, it is called a recursive system.
34. A system whose output does not depends on past outputs is called a nonrecursive system.
35. The convolution of N1 and N2 sample sequences produce a sequence consisting of N1+N2–1 samples.
36. In an LTI system, response for an arbitrary input is given by convolution of input with impulse response.
37. The output sequence of circular convolution is also periodic sequence with periodicity of N samples.
38. The inverse system is used to recover the input from the response of a system.
39. The process of recovering the input from the response of a system is called deconvolution.
40. The correlation of two different sequences is called crosscorrelation.
41. The correlation of a sequence with itself is called autocorrelation.
Chapter 2 - Discrete Time Signals and Systems 2. 114

2.16. Short Questions and Answers

Q2.1 Perform addition of the discrete time signals, x1(n) = {2, 2, 1, 2} and x2(n) = {–2, –1, 3, 2}.
Solution

In addition operation, the samples corresponding to same value of n are added.

When n = 0, x1(0) + x2(0) = 2 + (–2) = 0 When n = 2, x1(2) + x2(2) = 1 + 3 = 4
When n = 1, x1(1) + x2(1) = 2 + (–1) = 1 When n = 3, x1(3) + x2(3) = 2 + 2 = 4
\ x1(n) + x2(n) = {0, 1, 4, 4}

Q2.2 Perform multiplication of discrete time signals, x1(n) = {2, 2, 1, 2} and x2(n) = {–2, –1, 3, 2}.
Solution

In multiplication operation, the samples corresponding to same value of n are multiplied.

When n = 0, x1(0) × x2(0) = 2 ×(–2) = –4 When n = 2, x1(2) × x2(2) = 1 × 3 = 3
When n = 1, x1(1) × x2(1) = 2 ×(–1) = –2 When n = 3, x1(3) × x2(3) = 2 × 2 = 4
\ x1(n) × x2(n) = {–4, –2, 3, 4}

Q2.3 Express the discrete time signal x(n) as a summation of impulses.

If we multiply a signal x(n) by a delayed unit impulse d(n – m), then the product is x(m), where x(m)
is the signal sample at n = m [because d(n – m) is 1 only at n = m and zero for other values of n].
Therefore, if we repeat this multiplication over all possible delays in the range –¥ < m < ¥ and sum
all the product sequences, then the result will be a sequence that is equal to the sequence x(n).
\ x(n) = ... x(–2) d(n + 2) + x(–1) d(n + 1) + x(0) d(n) + x(1) d(n – 1) + x(2) d(n – 2) + ...
+∞
= ∑ x(m) δ(n − m)
m =−∞

Q2.4 What are the basic elements used to construct the block diagram of discrete time system?
The basic elements used to construct the block diagram of discrete time system are adder, constant
multiplier and unit delay element.
x 1 (n) x 1(n ) + x 2 (n ) x 1 (n) ax 1 (n) x (n) x (n − 1)
−1
+ a z

x 2 (n)

F ig b : C o n sta nt m ultip lier. F ig c : U nit d ela y elem en t.

F ig a : A d d er.
2 4
Q2.5 Let, x(n) = {1, 2, 3, 4}, be one period of a periodic
sequence. What is x(n – 2, mod4)?
The x(n) can be represented on the circle as shown 3 x (n) 1 1 x (n − 2 , m od4) 3
in fig Q2.5a. The x(n – 2, mod4) is circularly shifted
sequence of x(n) by two units of time as shown in
fig Q2.5b.(Here, mod 4 stands for periodicity of 4). 4 2

\ x(n – 2, mod4) = {3, 4, 1, 2} F ig Q 2.5 a . F ig Q 2.5 b .

2. 115 Digital Signal Processing
Q2.6 Why linear convolution is important in digital signal processing ?
The response or output of an LTI discrete time system for any input x(n) is given by linear
convolution of the input x(n) and the impulse response h(n) of the system. (This means that if
the impulse response of a system is known, then the response of the system for any input can
be determined by convolution operation.)
Q2.7 In y(n) = x(n)* h(n), how will you determine the start and end point of y(n)? What will be
the length of y(n)?
Let, length of x(n) be N1 and starts at n = nx. Let, length of h(n) be N2 and starts at n = nh.
Now, y(n) will start at n = nx + nh
y(n) will end at n = (nx + nh) + (N1 + N2 – 2)
The length of y(n) is N1 + N2 – 1.
Q2.8 What is zero padding? Why is it needed?
Appending zeros to a sequence in order to increase the size or length of the sequence is called
zero padding.
In circular convolution, when the two input sequences are of different size, then they are
converted to equal size by zero padding.
Q2.9 List the differences between linear convolution and circular convolution.
Linear convolution Circular convolution
1. The length of the input sequence 1. The length of the input sequences
can be different should be same.
2. Zero padding is not required. 2. If the length of the input sequences are
different, then zero padding is required.
3. The input sequences need not be 3. Atleast one of the input sequence
periodic. should be periodic or should be
periodically extended.
4. The output sequence is nonperiodic. 4. The output sequence is periodic.
The periodicity is same as that of input
sequence.
5. The length of output sequence will 5. The length of the input and output
be greater than the length of input sequences are same.
sequences.

Q2.10 Perform the circular convolution of the two sequences x1(n) = {1, 2, 3} and x2(n) = {4, 5, 6}.
Solution
Let x3(n) be the sequence obtained from circular convolution of x1(n) and x2(n). The sequence
x1(n) can be arranged as a column vector of order 3 ´1 and using the samples of x2(n) a 3 ´ 3 matrix
is formed as shown below. The product of two matrices gives the sequence x3(n).

LMx (0)
2 x2 (2) x 2 (1) OP LMx (0)OP LMx (0)OP
1 3 LM4 6 5 OP LM1OP LM31OP
⇒
MMNxx (1)
2
2 (2)
x2 (0) x 2 (2)
x2 (1)
P Mx (1)P = MMNxx (1)
1
x (0) PQ MN x (2) PQ
2 1
3
P
(2) PQ
3
MMN56 4 6
5
P M2P = MMN2831PPQ
4PQ MN 3PQ

\ x3(n) = x1(n) * x2(n) = {31, 31, 28}.

Chapter 2 - Discrete Time Signals and Systems 2. 116
Q2.11 Perform the linear convolution of the two sequences x1(n) = {1, 2} and x2(n) = {3, 4} via circular
convolution.
Solution
Let x3(n) be the sequence obtained from linear convolution of x1(n) and x2(n). The length of
x3(n) will be 2 + 2 – 1 = 3. Let us convert x1(n) and x2(n) into three sample sequences by padding
with zeros as shown below.
x1(n) = {1, 2, 0} and x2(n) = {3, 4, 0}
Now the circular convolution of x1(n) and x2(n) will give x3(n). The sequence x1(n) is arranged as
a column vector and using the sequence x2(n), a 3 ´3 matrix is formed as shown below. The
product of the two matrices gives the sequence x3(n).
LMx (0)
2 x 2 (2) x 2 (1) OP LMx (0)OP LMx (0)OP
1 3 LM3 0 4 OP LM1OP LM 3 OP
⇒
MMNxx (1)
2
2 (2)
x 2 (0) x 2 (2)
x 2 (1)
P Mx (1)P = MMNxx (1)
1
x (0) PQ MN x (2) PQ
2 1
3
P
(2) PQ
3
MMN40 3 0
4
P M2P = MMN108PPQ
3PQ MN0PQ

\ x3(n) = x1(n) * x2(n) = {3, 10, 8}

Q2.12 Compare the overlap add and overlap save method of sectioned convolutions.
Overlap add method Overlap save method
1. Linear convolution of each section of 1. Circular convolution of each section of
longer sequence with smaller sequence longer sequence with smaller sequence
is performed. is performed. (after converting them to
the size of output sequence).
2. Zero padding is not required. 2. Zero padding is required to convert
the input sequences to the size of
output sequence.
3. Overlapping of samples of input 3. The N2–1 samples of an input section of
sections are not required. longer sequence is overlapped with next
input section.
4. The overlapped samples in the output 4. Depending on method of overlapping
of sectioned convolutions are added input samples, either last N2–1 samples
to get the overall output. or first N2–1 samples of output sequence
of each sectioned convolution are
discarded.

Q2.13 In what way zero padding is implemented in overlap save method?

In overlap save method, the zero padding is employed to convert the smaller input sequence to
the size of the output sequence of each sectioned convolution. The zero padding is also employed
to convert either the last section or the first section of the longer input sequence to the size of the
output sequence of each sectioned convolution. (This depends on the method of overlapping
input samples).
Q2.14 List the similarities and differences in convolution and correlation of two sequences.
Similarities
1. Both convolution and correlation operation involves shifting, multiplication and summation of
product sequence.
2. Both convolution and correlation operation produce same size of output sequence.
2. 117 Digital Signal Processing
Differences
1. Correlation operation does not involve change of index and folding of one of the input
sequence.
2. The convolution operation is commutative, [i.e., x(n) * y(n) = y(n) * x(n)], whereas in correlation
operation in order to satisfy commutative property, while performing correlation of y(n) and
x(n), the shifting has to performed in opposite direction to that of performing correlation of x(n)
and y(n).
Q2.15 Let rxy(m) be the correlation sequence obtained by correlation of x(n) and y(n), how will you
determine the start and end point of rxy(m)? What will be the length of rxy(m) ?
Let, length of x(n) be N1 and starts at n = n1. Let length of y(n) be N2 and starts at n = n2.
Now, rxy(m) will start at mi = n1 – (n2 + N2 – 1)
rxy(m) will end at mf = mi + (N1 + N2 – 2)
The length of rxy(m) is N1 + N2 – 1.
Q2.16 What are the differences between crosscorrelation and autocorrelation?
1. Crosscorrelation operation is correlation of two different sequences, whereas autocorrelation
is correlation of a sequence with itself.
2. Autocorrelation operation is an even function, whereas crosscorrelation is not an even function.
Q2.17 Perform the correlation of the two sequences, x(n) = {1, 2, 3} and y(n) = {2, 4, 1}.
Solution
Given that, x(n) = {1, 2, 3 } and y(n) = {2, 4, 1}. \ y(–n) = {1, 4, 2 }
- - -
The sequence x(n) is arranged as a column and the folded sequence y(–n) is arranged as a row as
shown below. The elements of the two dimensional array are obtained by multiplying
the corresponding row element with column element. The sum of the diagonal elements gives the
samples of the crosscorrelation sequence, rxy(m).
y(−n) y(−n)
x(n) 1 4 2 x(n) 1 4 2

1 1 ×1 1 ×4 1 ×2 1 1 4 2
⇒
2 2 ×1 2 ×4 2 ×2 2 2 8 4

3 3×1 3×4 3×2 3 3 12 6

rxy (–2) = 1 ; rxy (–1) = 2 + 4 = 6 ; rxy(0) = 3 + 8 + 2 = 13 ; rxy(1) = 12 +4 = 16 ; rxy(2) = 6 ;

\ rxy(m) = {1, 6, 13, 16, 6}
-
Q2.18 Perform the circular correlation of the two sequences, x(n) = {1, 2, 3} and y(n) = {2, 4, 1}.
Solution
Let rxy (m) be the sequence obtained from circular correlation of x(n) and y(n). The sequence x(n)
can be arranged as a column vector of order 3 ´1 and using the samples of y(n) a 3 ´3 matrix is
formed as shown below. The product of two matrices gives the sequence rxy (m).

LMy(0) y(1) y(2) OP LMx(0)OP LMrxy (0)

OP LM2 4 1 OP LM1OP LM13OP
⇒
MMNy(2)
y(1)
y(0) y(1)
y(2)
P Mx(1)P = MMrr
y(0) PQ MN x(2) PQ
xy (1)
(2) PQ
P MMN41 2 4
1
P M2P = MMN1712PPQ
2PQ MN 3PQ
N xy

\ rxy (m) = {13, 17, 12}

Chapter 2 - Discrete Time Signals and Systems 2.118
Q2.19 Perform circular autocorrelation of the sequence, x(n) = {1, 2, 3, 4}.

Solution

Let rxx ( m) be the sequence obtained from circular autocorrelation of x(n). The sequence x(n) can
be arranged as a column vector of order 4´1 and again by using the samples of x(n) a 4´4 matrix
is formed as shown below. The product of two matrices gives the sequence rxx ( m) .
LMx(0) x(1) x(2) x(3) OP LMx(0)OP LMr xx (0) OP LM1 2 3 OP LM1OP LM 30OP
4

MMx(3)
x(2)
x(0)
x(3)
x(1)
x(0)
x(2)
x(1)
PP MMx(1) P = MMrr
x(2) P
xx (1)
xx (2) P
P ⇒ MM43 1
4
2
1
3
2
PP MM23PP = MM 2422PP
MNx(1) x(2) x(3) x(0) PQ MNx(3)PQ MNr xx (3) PQ MN2 3 4 1 PQ MN4PQ MN 24PQ
\ rxx ( m) = {30, 24, 22, 24 }

Q2.20 What is the difference between circular crosscorrelation and circular autocorrelation?
Circular crosscorrelation operation is circular correlation of two different sequences, whereas
circular autocorrelation is circular correlation of a sequence with itself.

2.17 MATLAB Programs

Program 2.1
Write a MATLAB program to generate the standard discrete time signals unit
impulse, unit step and unit ramp signals.

%******************* program to plot some standard signals

n=-20 : 1 : 20; %specify the range of n

%******************* unit impulse signal

x1=1;
x2=0;
x=x1.*(n==0)+x2.*(n~=0); %generate unit impulse signal
subplot(3,1,1);stem(n,x); %plot the generated unit impulse signal
xlabel(‘n’);ylabel(‘x(n)’);title(‘unit impulse signal’);

%******************* unit step signal

x1=1;
x2=0;
x=x1.*(n>=0)+x2.*(n<0); %generate unit step signal
subplot(3,1,2);stem(n,x); %plot the generated unit step signal
xlabel(‘n’);ylabel(‘x(n)’);title(‘unit step signal’);

%******************* unit ramp signal

x1=n;
x2=0;
x=x1.*(n>=0)+x2.*(n<0); %generate unit ramp signal
subplot(3,1,3);stem(n,x); %plot the generated unit ramp signal
xlabel(‘n’);ylabel(‘x(n)’);title(‘unit ramp signal’);
OUTPUT
The output waveforms of program 2.1 are shown in fig P2.1.
2. 119 Digital Signal Processing

F ig P 2 .1 : O u tp u t w av efo rm s o f pro g ra m 2 .1. F ig P 2 .2 : O u tp u t w av efo rm s o f pro g ra m 2 .2.

Program 2.2
Write a MATLAB program to generate the standard discrete time signals exponential
and sinusoidal signals.

%******************* program to plot some standard signals

n=-20 : 1 : 20; %specify the range of n

%******************* exponential signal

A=0.95;
x=A.^n; %generate exponential signal
subplot(2,1,1);stem(n,x); %plot the generated exponential signal
xlabel(‘n’);ylabel(‘x(n)’);title(‘exponential signal’);

%******************* sinusoidal signal

N=20; %declare periodicity
f=1/20; %compute frequency
x=sin(2*pi*f*n); %generate sinusoidal signal
subplot(2,1,2);stem(n,x); %plot the generated sinusoidal signal
xlabel(‘n’);ylabel(‘x(n)’);title(‘sinusoidal signal’);

OUTPUT
The output waveforms of program 2.2 are shown in fig P2.2.

Program 2.3
Write a MATLAB program to find the even and odd parts of the signal x(n)=0.8n.

%To find the even and odd parts of the signal, x(n)= 0.8^n

n= -5 :1 :5; %specify the range of n

A=0.8;
x1=A.^n; %generate the given signal
x2=A.^(-n); %generate the folded signal
Chapter 2 - Discrete Time Signals and Systems 2.120
if(x2==x1)
disp(‘“The given signal is even signal”’);
else if (x2==(-x1))
disp(‘“The given signal is odd signal”’);
else
disp(‘“The given signal is neither even nor odd signal”’);
end
end

xe=(x1+x2)/2; %compute even part

xo=(x1-x2)/2; %compute odd part

subplot(2,2,1);stem(n,x1);
xlabel(‘n’);ylabel(‘x1(n)’);title(‘signal x(n)’);

subplot(2,2,2);stem(n,x2);
xlabel(‘n’);ylabel(‘x2(n)’);title(‘signal x(-n)’);

subplot(2,2,3);stem(n,xe);
xlabel(‘n’);ylabel(‘xe(n)’);title(‘even part of x(n)’);

subplot(2,2,4);stem(n,xo);
xlabel(‘n’);ylabel(‘xo(n)’);title(‘odd part of x(n)’);

F ig P 2 .3 : O u tp u t w av efo rm s o f pro g ra m 2 .3. F ig P 2 .4 : O u tp u t w av efo rm s o f pro g ra m 2 .4.

OUTPUT

“The given signal is neither even nor odd signal”

The output waveforms of program 2.3 are shown in fig P2.3.

Program 2.4
Write a MATLAB program to perform amplitude scaling and time shift on the
signal x(n) = 1+n; for n = 0 to 2.

Program to declare the given signal as function y(n)

% declare the given signal as function y(n)

function x = y(n)
x=(1.0 + n).*(n>=0 & n<=2);
2. 121 Digital Signal Processing
Note: The above program should be stored as a separate file in the current
working directory

Program to perform amplitude scaling and time shift on y(n)

%To Perform Amplitude scaling and Time shift on signal x(n)=1+n;

%for n= 0 to 2
%include y.m file in current work directory which declare given signal as
%function y(n)

n=-5:1:5; %specify range of n

y0 =y(n); %assign the given signal as y0

y1 =1.5*y(n); %compute the amplified version of x(n)
y2 =0.5*y(n); %compute the attenuated version of x(n)
y3 =y(n-2); %compute the delayed version of x(n)
y4 =y(n+2); %compute the advanced version of x(n)

%plot the given signal and amplitude scaled signal

subplot(2,3,1);stem(n,y0);
xlabel(‘n’);ylabel(‘x(n)’);title(‘Signal x(n)’);
subplot(2,3,2);stem(n,y1);
xlabel(‘n’);ylabel(‘x1(n)’);title(‘Amplified signal 1.5x(n)’);
subplot(2,3,3);stem(n,y2);
xlabel(‘n’);ylabel(‘x2(n)’);title(‘Attenuated signal 0.5x(n)’);

%plot the given signal and time shifted signal

subplot(2,3,4);stem(n,y0);
xlabel(‘n’);ylabel(‘x(n)’);title(‘Signal x(n)’);
subplot(2,3,5);stem(n,y3);
xlabel(‘n’);ylabel(‘x3(n)’);title(‘Delayed signal x(n-2)’);
subplot(2,3,6);stem(n,y4);
xlabel(‘n’);ylabel(‘x4(n)’);title(‘Advanced signal x(n+2)’);

OUTPUT
The input and output waveforms of program 2.4 are shown in fig P2.4.

Program 2.5
Write a MATLAB program to perform convolution of the following two discrete
time signals.
x1(n)=1; 1<n<10 x2(n)=1; 2<n<10
%******************Program to perform convolution of two signals
%******************x1(N)=1; n= 1 to 10 and x2(n)=1; n= 2 to 10

n = 0 : 1 : 15; %specify range of n

x1=1.*(n>=1 & n<=10); %generate signal x1(n)

x2=1.*(n>=2 & n<=10); %generate signal x2(n)
N1=length(x1);
N2=length(x2);
x3=conv(x1,x2); %perform convolution of signals x1(n) and x2(n)
n1=0 : 1 : N1+N2-2; %specify range of n for x3(n)
Chapter 2 - Discrete Time Signals and Systems 2.122
subplot(3,1,1);stem(n,x1);
xlabel(‘n’);ylabel(‘x1(n)’);
title(‘signal x1(n)’);

subplot(3,1,2);stem(n,x2);
xlabel(‘n’);ylabel(‘x2(n)’);
title(‘signal x2(n)’);

subplot(3,1,3);stem(n1,x3);
xlabel(‘n’);ylabel(‘x3(n)’);
title(‘signal, x3(n) =
x1(n)*x2(n)’);

OUTPUT
F ig P 2 .5 : O u tp u t w av efo rm s o f pro g ra m 2 .5.
The input and output waveforms of
program 2.5 are shown in fig P2.5.

2.18 Exercises
I. Fill in the blanks with appropriate words
1. A signal x(n) may be shifted in time by m units by replacing the independent variable n by _______.
2. The _______ of a signal x(n) is performed by changing the sign of the time base n.
3. If the average power of a signal is finite then it is called _______.
4. The smallest value of N for which x(n + N) = x(n) is true is called _______.
5. In a discrete time signal x(n), if x(n) = x(–n) then it is called _______ signal.
6. In a discrete time signal x(n), if x(–n) = –x(n) then it is called _______ signal.
7. The output of the system with zero input is called _______.
8. A discrete time system is _______ if it obeys the principle of superposition.
9. A discrete time system is _______ if its input-output relationship do not change with time.
10. The response of an LTI system is given by _______ of input and impulse response.
11. If the output of a system depends only on present input then it is called _______.
12. A system is said to be _______ if the output does not depends on future inputs and outputs.
13. An LTI system is causal if and only if its impulse response is _______ for negative values of n.
14. When a system output at any time n depends on past output values, it is called _______ system.
15. An N-point sequence is called _______ if it is symmetric about point zero on the circle.
16. An N-point sequence is called _______ if it is antisymmetric about point zero on the circle.
17. The _______ is called aperiodic convolution.
18. The _______ is called periodic convolution.
19. Appending zeros to a sequence in order to increase its length is called _______.
20. The two methods of sectioned convolutions are _______ and _______method.
2. 123 Digital Signal Processing
21. In _______ method of sectioned convolution, overlapped samples of output sequences are _______.
22. In _______ method, the overlapped samples in one of the output sequences are discarded.
23. The correlation of two different discrete time sequences is called _______ .
24. The cascade of a system and its inverse is _______.
25. The process of recovering the input from the response of a system is called ______ .

Answers
1. n – m 8. linear 15. even 22. overlap save
2. folding 9. time invariant 16. odd 23. cross correlation
3. power signal 10. convolution 17. linear convolution 24. identity system
4. fundamental period 11. memoryless or static 18. circular convolution 25. deconvolution
5. symmetric 12. causal 19. zero padding
6. antisymmetric 13. zero 20. overlap add, overlap save
7. natural response 14. recursive 21. overlap add, added

II. State whether the following statements are True/False

1. The discrete signals are continuous function of an independent variable.
2. In digital signal the magnitudes of the signal are unquantized.
3. A discrete time signal x(n) is defined for noninteger values of n.
4. An impulse signal has a nonzero sample only for one value of n.
5. When we multiply a discrete time signal by unit step signal, the signal is converted to one-sided signal.
6. Shifting a signal to left is called delay and shifting to right is called advance.
7. Any discrete time signal can be expressed as a summation of impulses.
8. Periodic signals are power signals.
9. When the energy of a signal is infinite, it is called energy signal.
10. The output of a system for impulse input is called impulse response.
11. A system can be realized in real time only if it is noncausal and stable.
12. Dynamic systems does not require memory but static systems require memory.
13. A system is time invariant if the response to a shifted version of the input is identical to a shifted version
of the response based on the unshifted input.
14. An LTI system is unstable if the impulse response is absolutely summable.
15. A system whose output depends only on the present and past input is called a recursive system.
16. The circular shift of an N-point sequence is equivalent to a linear shift of its periodic extension.
17. For an N-point sequence represented on a circle, the time reversal is obtained by reversing its sample
about the point zero on the circle.
18. When a nonperiodic N-point sequence is represented on a circle then it becomes periodic with
periodicity N.
19. In linear convolution the length of the input sequences should be same.
20. In circular convolution the length of the input sequences need not be same.
21. In circular correlation the length of the input and output sequences are same.
Chapter 2 - Discrete Time Signals and Systems 2. 124
∞ ∞
22. The correlation operation, rxy (m) = ∑ x(n) y(n − m) is not same as rxy (m) = ∑ x(n + m) y(n) .
n =−∞ n =−∞
23. The cross correlation sequence rxy(m) is folded version of ryx(m).
24. The inverse systems exist for all LTI systems.
25. The final value of m in autocorrelation sequence of N-point sequence is, mf = mi + (2N – 1).
Answers
1. False 6. False 11. False 16. True 21. True
2. False 7. True 12. False 17. True 22. False
3. False 8. True 13. True 18. True 23. True
4. True 9. False 14. False 19. False 24. True
5. True 10. True 15. False 20. False 25. False

III. Choose the right answer for the following questions

x(n − 1)
1. x(n) = with initial condition x(0) = –1, gives the sequence,
4

a) x(n) =
FG 1 IJ n
b) x(n) = −
FG 1 IJ n
c) x(n) =
FG 1 IJ −n
d) x(n) =
FG −1IJ −n

H 4K H 4K H 4K H4K
2. The process of conversion of continuous time signal into discrete time signal is known as,
a) aliasing b) sampling c) convolution d) none of the above
3. If Fs is sampling frequency then the relation between analog frequency F and digital frequency f is,
F F F 2F
a) f = b) f = s c) f = d) f =
2Fs F Fs Fs

4. If Fs is sampling frequency then the highest analog frequency that can be uniquely represented in its
sampled version of discrete time signal is,
Fs 1
a) b) 2Fs c) Fs d)
2 Fs

5. The sampling frequency of the following analog signal, x(t) = 4 sin150pt + 2 cos50pt should be,
a) greater than 75 Hz b) greater than 150 Hz c) less than 150 Hz d) greater than 50 Hz
6. Which of the following signal is the example for deterministic signal?
a) step b) ramp c) exponential d) all of the above
7. For energy signals, the energy will be finite and the average power will be,
a) infinite b) finite c) zero d) cannot be defined
n
8. In a signal x(n), if 'n' is replaced by , then it is called,
3
a) upsampling b) folded version c) downsampling d) shifted version
9. The unit step signal u(n) delayed by 3 units of time is denoted as,
a) u(n + 3) = 1; n ≥ 3 b) u(3 − n) = 1; n ≥ 3 c) u(n − 3) = 1; n ≥ 3 d) u(3n) = 1; n > 3
= 0; n < 3 = 0; n < 3 = 0; n < 3 = 0; n < 3
2. 125 Digital Signal Processing
10. The zero input response (or) natural response is mainly due to,
a) Initial stored energy in the system b) Initial conditions in the system
c) Specific input signal d) both a and b
11. If x(n) = an u(n) is the input signal, then the particular solution yp(n) will be,
a) Kn an u(n) b) K an u(n)
c) K1 an u(n) + K2 an u(n) d) K a–n u(n)
12. The discrete time system, y(n) = x(n–3) – 4x(n–10) is a,
a) dynamic system b) memoryless system c) time varying system d) none of the above
13. An LTI discrete time system is causal if and only if,
a) h(n) ¹ 0 for n < 0 b) h(n) = 0 for n < 0 c) h(n) ¹ ¥ for n < 0 d) h(n) ¹ 0 for n > 0
14. Which of the following system is causal?

a) h(n) = n
FG 1 IJ n
u(n + 1) b) y(n) = x2(n) – x(n+1) c) y(n) = x(–n) +x(2n–1) d) h(n) = n
FG 1 IJ n
u(n)
H 2K H 2K
15. An LTI system is stable, if the impulse response is,
∞ ∞ ∞
a) ∑ h(n) = 0 b) ∑ h(n) < ∞ c) ∑ h(n) ≠ 0 d) either a or b
n= −∞ n= −∞ n= −∞

16. The system y(n) = sin[x(n)] is,

a) stable b) BIBO stable c) unstable d) none of the above
17. Two parallel connected discrete time systems with impulse responses h1(n) and h2(n) can be replaced
by a single equivalent discrete time system with impulse response,
a) h1(n) * h2(n) b) h1(n) + h2(n) c) h1(n) – h2(n) d) h1(n) * [h1(n) + h2(n)]
18. Sectioned convolution is performed if one of the sequence is very much larger than the other in order
to overcome,
a) long delay in getting output b) larger memory space requirment
c) both a and b d) none of the above
19. In overlap save method, the convolution of various sections are performed by,
a) zero padding b) linear convolution c) circular convolution d) both b and c
20. If x(n) is N1-point sequence,if y(n) is N 2-point sequence, if r xy(m) is the correlation sequence
starts at m = mi , then the value of m corresponding to last sample of rxy(m) is,
a) mf = mi + (N1 + N2 – 2) b) mf = mi + (2N – 2) c) mf = mi + (N1 + N2 – 1) d) mf = mi + (2N + 1)
21. For a system, y(n) = nx(n), the inverse system will be,

a) y 1
nej b) 1 y n
n
bg c) ny(n) d) n –1y(n)

22. For a system y(n) = x(n–3) the impulse response of the system and the inverse system will be ––––––
and –––––– respectively.
a) h(n) = d(n + 3), x(n) = y(n – 3) b) h( n) = δ(3n), x(n) = y n ej
3
c) h(n) = d(n – 3), x(n) = y(n + 3) d) h(n) = d(n + 3), x(n) = y(3n)
Chapter 2 - Discrete Time Signals and Systems 2. 126
23. The circular correlation rx 1 x 2 (q) of the sequence x1(n) and x2(n) of length 'N' can be defined by the
equation,
∞ N −1
a) ∑ x1(n) x2 (n − q) b) ∑ x1(n) x∗2 (n − q)
n= −∞ n=0
N −1 ∞
c) ∑ x1(n) x∗2 b(n − q)gN d) ∑ x1(n) x∗2 b(n − q)gN
n=0 n=−∞

24. The evaluation of correlation involves,

a) shifting, rotating and summation b) shifting, multiplication and summation
c) change of index, folding and summation d) change of index, folding, shifting & multiplication
25. The circular correlation of N-point sequences is evaluated in the range,
a) – N < m < N b) –N<m<0 c) 0<m<N d) 0 < m < N– 1

Answers
1. b 6. d 11. b 16. a 21. b
2. b 7. c 12. a 17. b 22. c
3. c 8. a 13. b 18. c 23. c
4. a 9. c 14. d 19. c 24. b
5. b 10. a 15. d 20. a 25. d

IV. Answer the following questions

1. Define discrete and digital signal.
2. Explain briefly, the various methods of representing discrete time signal with examples.
3. Define sampling and aliasing.
4. What is Nyquist rate?
5. State sampling theorem.
6. Define the impulse and unit step signal.
7. Express the discrete time signal x(n) as a summation of impulses.
8. How will you classify the discrete time signals?
9. What are energy and power signals?
10. When a discrete time signal is called periodic?
11. What is discrete time system?
12. What is impulse response? Explain its significance.
13. Write the difference equation governing the Nth order LTI system.
14. Write the expression for discrete convolution.
15. List the various methods of classifying discrete time systems.
16. Define time invariant system.
17. What is linear and nonlinear systems?
18. What is the importance of causality?
19. What is BIBO stability? What is the condition to be satisfied for stability?
20. What are FIR and IIR systems?
21. Write the convolution sum formula for FIR and IIR systems.
22. What are recursive and nonrecursive systems? Give examples.
2. 127 Digital Signal Processing
23. Write the properties of linear convolution.
24. Prove the distributive property of linear convolution.
25. What are the two ways of interconnecting LTI systems?
26. Define circular convolution.
27. What is the importance of linear and circular convolution in signals and systems?
28. How will you perform linear convolution via circular convolution?
29. What is sectioned convolution? Why is it performed?
30. What are the two methods of sectioned convolution?
31. What is inverse system? What is its importance?
32. Define deconvolution.
33. Define cross correlation and autocorrelation?
34. What are the properties of correlation?
35. What is circular correlation?
V. Solve the following problems
E2.1 Determine whether the following signals are periodic or not. If periodic, find the fundamental
period.

a) x(n) = sin
5π
n+6
FG IJ
b) x(n) = sin
7n
+π
FG IJ
c) x(n) = cos
4πn FG IJ
8 H K 3 H K 12 H K
d) x(n) = cos
π 2
n
FG IJ
e) x(n) = e j9 n f) x(n) = 4 sin
3πn
+ 5 cos
3πn
32 H K 2 4
E2.2 Determine the even and odd parts of the signals.
π
1 −j n
a) x(n) = 2n b) x(n) = 8e 6 l
c) x(n) = 6, 4, 2, 2 q
a
A
E2.3 a) Consider the analog signal x(t) = 2 sin80pt. If the sampling frequency is 60 Hz, find the sampled
version of discrete time signal x(n). Also find an alias frequency corresponding to Fs = 60 Hz.
b) Consider the analog signals, x1 (t) = 4 cos2π (30t) and x2 (t) = 4 cos 2π (5t). Find a sampling
frequency so that 30 Hz signal is an alias of 5 Hz signal.
c) Consider the analog signal, x(t) = 3 sin40π t − sin100π t + 2cos 50π t. Determine the minimum
sampling frequency and the sampled version of analog signal at this frequency. Sketch the
waveform and show the sampling points. Comment on the result.
E2.4 Determine whether the following signals are energy or power signals.

a) x(n) =
FG 5 IJ n
u( n)
FG
b) x(n) = cos
3π IJ
n
H 9K H 4 K c) x(n) = u(2n) d) x(n) = 2 u(3 – n)

E2.5 Construct the block diagram and signal flow graph of the discrete time systems whose input-
output relations are described by the following difference equations.
a) y(n) = 2y(n – 1) + 2.1 x(n – 1) + 0.5 x(n – 2)
b) y(n) = 1.6 x(n –2) + 0.7 x(n) + 3y(n – 1) + 0.3y(n – 2)
E2.6 Determine the response of the discrete time systems governed by the following difference equations.
a) y(n) = 0.1y(n – 1) + x(n – 1) + 0.7x(n) ; x(n) = 2–n u(n) ; y(–1) = –1
b) y(n) + 2.1y(n – 1) + 0.2y(n – 2) = x(n) + 0.56x(n –1) ; x(n) = u(n) ; y(–2) = 1; y(–1) = –3
Chapter 2 - Discrete Time Signals and Systems 2. 128
E2.7 Test the following systems for time invariance.
a) y(n) = x(n + 1) + x(n + 2) b) y(n) = nax(n) c) y(n) = x2(n + 2) + C d) y(n) = (n –1) x2(n) + C
E2.8 Test the following systems for linearity.
a) y(n) = x2(n) + x3(n – 1) b) y(n) = bx(n + 2) + nex(n) c) y(n) = a x(n) + b x( n)
1 N M
d) y(n) = x(n) + e) y(n) = ∑b m x( n + m) + ∑ c m y( n + m)
x(n) m= −1 m=0

E2.9 Test the causality of the following systems.

a) y(n) = x(n) – x(–n – 2) + x(n –1) b) y(n) = a x(2n) + x(n2)
n n 4
c) y(n) = ∑ x( m) + ∑ x(2 m)
m= −1 m= −∞
d) y(n) = (0.3)n u(n + 2) e) y(n) = ∑ x( n − k )
k= −4

E2.10 Test the stability of the following discrete time systems.

a) y(n) = x2(n) + x(n + 1) b) y(n) = nx(n –1) c) h(n) = (0.4)n u(n + 3)
d) h(n) = (8)n u(4 – n) e) y(n) = x(n – 3)
E2.11 Determine the range of values of 'a' and 'b' for the stability of an LTI system with impulse response,

h(n) =
R|S( −4a) n
; n≥0
|T 2b −n
; n<0
E2.12 a) Determine the impulse response for the cascade of two LTI systems having impulse responses,

h1 (n) =
FG 1 IJ n
u(n) and h2 (n) = d(n − 3) h2 (n)
H7 K x (n )
b) Determine the overall impulse response + y (n )
of the interconnected discrete time
system shown in fig E2.12.
h1(n) + h 3 (n)

F ig E 2 .1 2.
Take, h1 (n) =
FG 1 IJ n
u(n) ; h2 (n) =
FG 1 IJ n
u(n) ; h3 (n) =
FG 1 IJ n
u(n)
H 3K H6K H 9K
E2.13 Determine the response of an LTI system whose impulse response h(n) and input x(n) are given by,
l
a) h( n) = 1, 4, 1, −2, 1 q , l
x( n) = 1, 3, 5, −1, −2 q
A A
b) h( n) =
RS1 ; 0≤n≤2
, x( n) = a n u( n); a <1
T0 ; n≥3
E2.14 Perform circular convolution of the two sequences,
l
a) x1 ( n) = 1, 2, −1, 1 ; q l
x2 ( n) = 2, 4, 6, 8 q
b) x ( n) = l0,
1 0.6, −1, 15
., 2 ; q x ( n) = l−2,
2 3, 0.2, 0.7, 0.8 q
E2.15 The input x(n) and impulse response h(n) of an LTI system are given by,
l
x(n) = −1, 1, −1, 1, −1, 1 ; q l
h(n) = −0.5, 0.5, −1, 0.5, −1, −2 q
A A
Find the response of the system using a) Linear convolution, b) Circular convolution.
E2.16 Perform linear convolution of the following sequences by,
a) Overlap add method b) Overlap save method
l
x( n) = 1, −1, 2, 1, −1, 2, +1, −1, +2 q ; l
h( n) = 2, 3, −1 q
2. 129 Digital Signal Processing
E2.17 Perform crosscorrelation of the sequences,
l
x( n) = −1, 2 3, −4, ; q l
h( n) = 2, −1, −3, q
A A
E2.18 Determine the autocorrelation sequence for x(n) = 1, 4, 3, −5, 2 . l q
A
E2.19 Find the inverse system for the following discrete time system,
n
y(n) = ∑ c p x(p − 2) ; for n ≥ 0
p =0

E2.20 A discrete time system is excited by an input x(n), and the response is, y(n) = 4, 3, 6, 7.5, 3, 30, − 8 . m r
A
l
If the impulse response of the system is h(n) = 2, 4, −2 , then what will be the input to the system? q
A
E2.21 Perform circular correlation of the sequence, x(n) = −1, 1, 2, 6 and y(n) = 4, −2, −1, 2 . l q l q
Answers
E2.1 a) periodic; N=16 b) nonperiodic c) periodic; N=6 d) periodic; N=32 e) nonperiodic. f) periodic; N=8

π
E2.2 a) xe ( n) =
1 −2n
a + a2n b) xe ( n) = 8 cos n l q
c) x e ( n) = 1, 1, 2, 6, 2, 1, 1
2 6 A
x (n) = l−1, − 1, − 2, 0, 2, 1, 1q
1 π o
xo (n) = a −2n − a 2n
2
xo (n) = − j8 sin n
6
A
4πn
E2.3 a) x( n) = 2 sin ; Alias frequency = 100Hz b) Fs = 25Hz
3
2 πn πn
c) Fs,min = 100 Hz ; x( nT) = 3sin + 2 cos (sin πn = 0, for integer n)
5 2
The component sin100pt will give always zero samples when
sampled at 100Hz for any value of n (Refer fig E2.3c).
E2.4 a) E = 1.435J ; P = 0 ; Energy signal.
b) E = ¥ ; P = 0.5W ; Power signal.
c) E = ¥ ; P = 0.25W ; Power signal.
d) E = ¥ ; P=2W ; Power signal.

E2.5 a)
x (n ) 2.1x(n −1) y (n ) F ig E 2 .3 c : S a m p lin g p o ints.
−1 2.1
z + +
)
)

−1
−2

(n
x (n

2y

F ig E 2 .5 a.1 : B lo ck −1
0.5

−1
z z
0.5 2
d ia g ra m . x (n ) −1
1 z 2.1 1 1 1 y (n )

−1 −1
z 0.5 2 z
F ig E 2 .5 a.2 : S ig n al flow grap h .
Chapter 2 - Discrete Time Signals and Systems 2. 130
E2.5 b)
x (n ) y (n ) x (n ) 1 0.7 1 1 y (n )
0.7 + +
−1 3 −1
z z
−1 −1
z + 3 z

0.
6

3
1.
−1
−1 z
z
−1
z 1.6 0.3 z
−1

F ig E 2 .5 b.2 : S ig n al flow grap h .

F ig E 2 .5 b.1 : B lo c k d ia g ra m .

E2.6
LM
a) y(n) = −2.775(0.1) n + 3.375
FG 1 IJ OP u(n)
n
b) y(n) = 0.47 − 0.02 ( −0.1) n + 6.65 ( −2) n u(n)
MN H 2 K PQ
E2.7 a) c) Time invariant b) d) Time variant
E2.8 a) e) Linear b) c) d) Nonlinear
E2.9 a) b) c) d) e) Noncausal
E2.10 a) c) d) e) Stable system b) Unstable system
1 1
E2.11 For stability, 0 < a < and 0 < b <
4 2
FG 1 IJ ( n − 3) LM F 1 I n
FG IJ + FG 3 IJ FG 1IJ OP u(n)
3 1
n n
E2.12 a) h( n) =
H 7K u(n − 3) b) h( n) = 4
MN GH 6 JK −
H K H 2 K H 3K PQ
2 9
n
E2.13 a) y( n) = l1, 7, 18, 20, −6, −16, 5, 3, −2 q b) y( n) = ∑ ak ; for n = 0, 1, 2
A k =0
n
= ∑ a k ; for n > 2
k = n−2

E2.14 a) x3 ( n) = 8, −6, 4, 14 l q l
b) x3 ( n) = 6.08, −0.55, 6.4, −4.28, 0.72 q
A A
E2.15 l
y( n) = 0.5, −1, 2, −2.5, 3.5, −1.5, 1, −0.5, −0.5, 1, −2 q
A
l
E2.16 a) Overlap add method : y( n) = 2, 1, 0, 9, −1, 0, 9, −1, 0, 7, −2 q
b) Overlap save method : y( n) = l*, *, 0, 9, −1, 0, 9, −1, 0, 7, −2q
E2.17 r ( m) = l3, −5, −13, 13, 10, −8q
xy
A
1
E2.18 r ( m) = l2, 3, − 11, − 9, 55, − 9, − 11, 3, 2q
xx E2.19 x( n) = [ y( n + 2) − y( n + 1)] ; for n ≥ −1
cn + 2
A with initial condition x( −2) = y(0)

E2.20 l
x( n) = 2, −2.5, 10, −18.75, 49 q E2.21 rxy (m) = 4, −8, −1, 29 l q
A
Solution for Exercise Problems E2. 1

Digital Signal Processing - A. Nagoor Kani Chapter 2 - Discrete Time Signals and Systems

Solution for Exercise Problems

E2.1. Determine whether the following signals are periodic or not. If periodic, find the fundamental period.

a) x(n) = sin
FG 5π n + 6IJ
H8 K
Solution

b g
Now, x n + N = sin
FG 5π bn + Ng + 6IJ = sin FG 5πn + 6 + 5πN IJ
H8 K H8 8 K

5πN
Since sin (q + 2pM) = sin q, for periodicity, should be integral multiple of 2p.
8

5πN
Let, = M × 2π , M and N are integers.
8
8 16
∴ N = M × 2π × = M
5π 5
16
N = M , if M = 5, 10, 15, 20 ..... N will be a integer.
5

When M = 5 , N = 16.

b
x n + N = sing FG 5π n + 6 + 5π × 16IJ = sin FG 5π n + 6 + 10πIJ = sinFG 5π n + 6IJ = x(n).
H8 8 K H8 K H8 K
\ x(n) is periodic.

Fundamental period is 16 samples.

b) x(n) = sin
FG 7n + π IJ
H3 K
Solution

b g
x n + N = sin
FG 7(n + N) + πIJ = cos
FG 7n + π + 7NIJ
H 3 K H3 3 K

7N
Since cos (q + 2pM) = cos q, for periodicity, should be equal to integral multiple of 2p.
3
7N 2π × 3 6π
Let, = M × 2π ⇒ N = M = M
3 7 7
Here, N cannot be an integer for any integer value of M, and so, x(n) will not be periodic.

c ) x(n) = cos
FG 4π n IJ
H 12 K
Solution

bg FG π nIJ
x n = cos
H3 K
F π I F nπ + Nπ IJ
xbn + Ng = cos G (n + N)J = cos G
H3 K H 3 3 K
Nπ
= 2πM ⇒ N = 6M
3
For M = 1, 2, 3, ..... N will be integer.

For M = 1, N = 6.

\ x(n) is periodic.

Fundamental period is 6 samples.

E2. 2 DSP, Chapter 2 -Discrete Time Signals and Systems

d) x(n) = cos
FG π n IJ
2
H 32 K
Solution

Given that, x(n) = cos

FG π n IJ
2
H 32 K
π
∴ x(n + N) = cos (n + N)2
32
π 2
= cos (n + N2 + 2nN)
32
Fπn 2 πN2 πN I
= cos GH 32 +
32
+
16
n JK
πN2 πN
Let, = 2πM1 Let, = 2πM2
32 16

∴ N = 8 M1 \ N = 32 M2

Now, N is integer for M1 = 12, 22, 32, 42 ..... Now, N is integer for M2 = 1, 2, 3, 4 .....

When M1 = 42 and M2 = 1, we get a common value for N as, N = 32.

F π n + π32 + π32 nI 2
When N = 32 ; x(n + N) = cos GH 32 32 16 JK 2

FF π I I
= cosG G n + 2πnJ + 16 × 2πJ
2
H H 32 K K
Fπ
= cosG n + 2πnJ
I 2
H 32 K For integer M,
π 2 cos(q + 2pM) = cosq
= cos n = x(n)
32
\ x(n) is periodic with fundamental period, N = 32 samples.

e) x(n) = e j9n

Solution

b g
x n + N = e j9(n+N) = e j9n . e j9N

Since, e j2 πM = 1
2π
Let, b9Ng = M × 2π ⇒ N=
9
M

For any integer value of M, N will not be an integer.

Hence x(n) is non periodic.

3πn 3πn
f) Given that, x(n) = 4 sin + 5 cos
2 4
Solution
3πn 3πn
Let, x1(n) = 4 sin Let, x2 (n) = 5 cos
2 4

b g
∴ x1 n + N1 = 4 sin
b
3 π n + N1 g b g
∴ x 2 n + N2 = 5 cos
b
3 π n + N2 g
2 4

= 4 sin
FG 3πn + 3πN IJ 1 .....(1) = 5 cos
FG 3πn + 3πN IJ 2 .....(2)
H2 2 K H4 4 K
3πN1 4 3πN2 8
Let, = 2πM1 ⇒ N1 = M1 Let, = 2πM2 ⇒ N2 = M2
2 3 4 3
Let, M1 = 3 ; \ N1 = 4 Let, M2 = 3 ; \ N2 = 8
Solution for Exercise Problems E2. 3
substitute N1 = 4 in equation (1), substitute N2 = 8 in equation (2),

b
∴ x1 n + N1 = 4 sin g FG 3πn + 3π × 4IJ FG 3πn + 3π × 8IJ
H2 2 K b g
∴ x 2 n + N2 = 5 cos
H4 4 K
F 3πn + 3 × 2πIJ
= 4 sin G F 3πn + 3 × 2πIJ
H2 K = 5 cos G
For integer M, For integer M, H4 K
sin(q + 2pM) = sinq 3πn cos(q + 2pM) = cosq 3πn
= 4 sin = x1(n) = 5 cos = x 2 (n)
2 4
\ x1(n) is periodic with fundamental period, N1 = 4 samples. \ x2(n) is periodic with fundamental period, N2 = 8 samples.

Here, x(n) = x1(n) + x2(n), and x(n) is periodic with period N1 = 4, and x2(n) is periodic with period N2 = 8.

Therefore, x(n) is periodic with period N, where N is LCM of N1 and N2.

The LCM of 4 and 8 is 8.

\ x(n) is periodic with fundamental period, N = 8.

E2.2. Determine the even and odd parts of the signals.

1
a) x(n) = ⇒ x(n) = a −2n
a 2n

Solution

1
x(−n) = ⇒ x(−n) = a 2n
a −2n
Even part of the signal,

1 1 −2n
xe (n) = x(n) + x(−n) = a + a2n
2 2
Odd part of the signal is,

1 1
x0 (n) = x(n) − x(−n) = a −2n − a 2n
2 2

π
−j n
b) x(n) = 8 e 6

Solution

x(n) = 8 e
π
−j n
6
= 8 cos
LM π π
n − j sin n
OP
N 6 6 Q
x( −n) = 8 e
π
− j ( − n)
6 = 8 cos
LM π π
n + j sin n
OP
N 6 6 Q
xe (n) =
LM1 π π π OP π π
× 8 cos n − j sin n + cos n + j sin n = 8 cos n
N2 6 6 6 Q 6 6

1 L π π π π O π
x (n) = × 8 Mcos n − j sin n − cos n − j sin nP = − j8 sin n
0
2 N 6 6 6 6 Q 6

l
c) x(n) = 6, 4, 2, 2 q
A
Solution

l
Given that, x(n) = 6, 4, 2, 2 q
A
x(0) = 6, x(1) = 4, x(2) = 2, x(3) = 2
l
x( −n) = 2, 2, 4, 6 q
A
x(0) = 6; x( −1) = 4; x(−2) = 2; x( −3) = 2
E2. 4 DSP, Chapter 2 -Discrete Time Signals and Systems
1 1
Even part, x e (n) =
2
b
x(n) + x( −n) g Odd part, x 0 (n) =
2
b
x(n) − x( −n) g
at n = –3 ; x(n) + x(–n) = 0 + 2 = 2 n = –3 ; x(n) – x(–n) = 0 – 2 = –2
n = –2 ; 0+2 = 2 n = –2 ; = 0 – 2 = –2
n = –1 ; 0+4 = 4 n = –1 ; = 0 – 4 = –4
n= 0; 6 + 6 = 12 n= 0; = 6–6 = 0
n= 1; 4+0 = 4 n= 1; = 4–0 = 4
n= 2; 2+0 = 2 n= 2; = 2–0 = 2
n= 3; 2+0 = 2 n= 3; = 2–0 = 2

1 1
xe (n) = x(n) + x( −n) x0 (n) = x(n) − x( −n)
2 2
l
xe (n) = 1, 1, 2, 6, 2, 1, 1 q l
x0 (n) = −1, − 1, − 2 , 0, 2, 1, 1 q
A A
E2.3. a) Consider the analog signal x(t) = 2sin80pt. If the sampling frequency is 60 Hz, find the sampled version of
discrete time signal x(n). Also find an alias frequency corresponding to Fs = 60 Hz.

Solution

x(n) = x( t) n
t = nT =
Fs

∴ x(n) = 2 sin 80 πt = 2 sin 80 π ×

n
= 2 sin
4 πn FG IJ
t=
n
60
60 3 H K
Now, 2 sin
FG 4π n + 2πnIJ = 2 sinFG 10πn IJ
H3 K H 3 K
F 10πn IJ is, f = 5
The frequency of 2sin G
H 3 K 3

F 5
Also, f = ⇒ F = fFs = × 60 = 100
Fs 3

Hence for Fs = 60 Hz, F = 100 Hz is an alias frequency.

b) Consider the analog signals x1(t) = 4 cos 2p (30t), x2(t) = 4 cos 2p (5t). Find a sampling frequency so that
30 Hz signal is an alias of 5 Hz signal.

Solution

Let the sampling frequency be, Fs = 30 – 5 = 25 Hz

∴ x1(n) = x1(t) = 4 cos 2π 30 ×

FG n IJ
t = nT =
n
Fs
H 25 K
∴ x1(n) = 4 cos
12π
n = 4 cos 2πn +
2 πnFG
= 4 cos
2π
n
IJ
5 5 H 5 K
x 2 (n) = x 2 (t) = 4 cos 2π 5 ×
FG n IJ = 4 cos
2π
n
t = nT =
n
Fs
H 25 K 5

c) Consider the analog signal, x(t) = 3sin40pt – sin100pt + 2cos50pt

Determine the minimum sampling frequency and the sampled version of analog signal at this frequency. Sketch the
waveform and show the sampling points.Comment on the result.

Solution

x(t) = 3 sin 40πt − sin100 πt + 2 cos 50πt ≡ x(t) = 3 sin 2πF1t − sin 2πF2 t + 2cos 2πF3 t
40 100 50
∴ F1 = = 20 Hz ; F2 = = 50 Hz ; F3 = = 25 Hz
2 2 2
Solution for Exercise Problems E2. 5
The maximum analog frequency in the signal is 50 Hz.

The minimum sampling frequency should be twice that of this maximum analog frequency.

Fs ≥ 2 Fmax ⇒ Fs ≥ 2 × 50

Let, Fs = 100Hz

∴ x(nT) = x(t) n
t=nT=
Fs

n n n 2πn π
x(nT) = 3 sin 40π × − sin 100π + 2cos 50π × = 3 sin − sin πn + 2 cos n
100 100 100 5 2
sin πn = 0, for integer values of n.
2πn πn
∴ x(nT) = 3 sin + 2cos
5 2

3 sin 40πt
1
⇒ F1 = 20 Hz, T1 = = 0.05 sec
20

sin 100πt

⇒ F2 = 50 Hz, T2 = 0.02 sec

2cos 50πt

⇒ F3 = 25 Hz, T3 = 0.04

Fs = 100 Hz

Ts = 0.01sec

In the analog signal x(nT), the component sin 100pt will give always zero samples when sampled at 100Hz for any value of n.
This is the drawback in sampling at nyquist rate, which is Fs = 2 Fmax.

E2.4. Determine whether the following signals are energy or power signals.

a) x(n) =
FG 5 IJ n

u(n)
H 9K
Solution

x(n) =
FG 5 IJ u(n)
n
for all n.
H 9K
∴ x(n) = (0.55)n ; n ≥ 0
+∞ ∞ ∞ ∞
2
2 n
∑ d(0.55) i = ∑ b0. 302g
2 n
Energy, E = ∑
n = −∞
x(n) = ∑
n=0
(0.55)n =
n= 0 n= 0

∞
1
∴ E= ∑ (0.302)
n= 0
n
=
1 − 0.302
= 1. 43 Joules

N
1 2
Power, P = Lt
N→∞ 2N + 1
∑ x(n)
n = −N
N N
1 n 1
= Lt
N→∞ 2N + 1
∑ d(0.55) i 2
= Lt
N→∞
∑
2N + 1 n = 0
(0.302)n
n= 0

1 (0.302)N+1 − 1 1 0−1
= Lt = × =0
N→ ∞ 2N + 1 0.302 − 1 ∞ −0.698
P is zero and E is finite.
So x(n) is energy signal.
E2. 6 DSP, Chapter 2 -Discrete Time Signals and Systems
3π 1 + cos 2θ
b) x(n) = cos n cos2 θ =
4 2
Solution
F 1+ cos 2 × 3π n I
∑ GG 4 J
+∞ +∞ 2 +∞
2 3π
Energy, E = ∑ x(n) = ∑ cos JJ n =
n = −∞ GH
n= −∞ 2 4
K n = −∞

F 1I F 3π I
+∞
1 F 3π I 1 +∞ +∞
= G J ∑ G 1 + cos
H 2K H nJ =
2 K 2 GH
∑ 1 + ∑ cos 2 nJK = 2 b∞ + 0g = ∞ n

n = −∞ n = −∞ n = −∞

Power, P = Lt
1 1 N

∑ x(n) = Lt 2N + 1 ∑ cos GH 4 nJK

F 3π I
2
N
2
2N + 1N→ ∞
n = −N
N→ ∞
n = −N

FG1+ cos 2 × 3π nIJ

1 H 4 K
N
1 1 L 3π O +N +N
= Lt
2N + 1
∑
N→ ∞ 2
= Lt
n = −N 2N + 1 2 NM
M ∑ 1 + ∑ cos nP
2 PQ N→ ∞
n = −N
n

n = −N

1 1 L O
= Lt M11+414+2144
2N + 1 2 MN
N→ ∞
.....31 + 1 + 11+44
1 + 12.....
44+ 31 + 0P
N termsPQ N terms

1 1 1
= Lt × 2N + 1 = = 0.5
N→ ∞ 2N + 1 2 2
Since P is finite and E is infinite, x(n) is power signal.

3π
It can be shown that cos n is periodic and sum of samples of one period of periodic cosine signal is zero.
2

cos
3π
b
n + N = cos
3πn 3πN
g+
FG IJ 3π 3π
2 2 2 H K n = 0 ; cos
2
n=1 n = 4 ; cos
2
n=1

3 πN 3π 3π
Let, = 2 πM n = 1 ; cos n=0 n = 5 ; cos n=0
2 2 2
4M 3π 3π
∴ N= n = 2 ; cos n = −1 n = 6 ; cos n = −1
3 2 2
Let, M = 3, Now, N = 4 3π 3π
n = 3 ; cos n=0 n = 7 ; cos n=0
2 2
3π
∴ cos n is periodic with
2
period 4 samples.

c) x(n) = u(2n)
Solution
+∞ ∞
2
E = ∑ x(n) = ∑ u(2n) 2
= ∑ u(n) = 1+ 1+ 1+ 1 ..... ∞ = ∞
n = −∞ n= 0 n = even

N N
1 2 1 1
P =
N→∞
Lt
2N + 1 ∑
n = −N
x(n) = Lt
N→∞ 2N + 1 ∑ u(2n)
n= 0
2
= Lt
N→∞
= 1+ 1+......+1
2N + 1 144244 3
N
1+ terms
2

N
FG 1 + 1IJ 1 1
+
1
= Lt
1
1+
N
=
FG IJ Lt
H N 2K = 2 ∞ = 2 =
1
.
N→∞ (2N + 1) 2 H K N→∞ F 1I
NG 2 + J 2+
1 2 4
H NK ∞
Since P is finite, E is infinite, x(n) is power signal.
d ) x(n) = 2 u(3 − n)
Solution
+∞ −3 −3

∑ b2 u(3 − n)g
2 2
E = ∑ x(n) = = ∑ 4 .......1 + 1 + 1 = ∞
14 4244 3 u (3 −n )
n = −∞ n = −∞ n = −∞ inf inite terms

+N −3
1 2 1
P =
N→∞
Lt
(2N + 1) ∑
n = −N
x(n) = Lt
N→∞ (2N + 1) ∑ 4 u(3 − n)
n=N

1 4
= Lt 4 1+ 1+ 1......+1 = Lt N− 2 n −3 −2 −1 0
N→∞ (2N + 1) 1442443 N→∞ (2N + 1)
N − 2 terms
Solution for Exercise Problems E2. 7

N × 4 1−
2 FG IJ 4 1−
FG 2 IJ
∴ P = Lt
N H K =
H ∞ 4
= =2
K
N→∞
N 2+
1 FG IJ 2+
1 2
N H K ∞

Since P is finite, E is infinite x(n) is power signal.

E2.5. Construct the block diagram and signal flow graph of the discrete time systems whose input-output relations are
described by the following difference equations.
a) y(n) = 2y(n − 1) + 2.1 x(n − 1) + 0.5 x(n − 2).

Solution

B loc k D ia gra m S ign a l F lo w G ra p h

x (n )
x (n ) z
−1
z
−1
2.1 x (n −1)

x (n − 1) 2.1 x (n −1)
−1
z 2.1

x (n − 1 )
x (n − 1 )
0.5 x (n − 2)

−1
z z
−1
0.5

x (n − 2) 0.5 0.5 x (n − 2) x (n − 2)

y (n )
y (n )
2 y(n −1 )
−1
z z
−1

y (n − 1)
2
2 y(n −1 )

B loc k D ia gra m S ign a l F lo w G ra p h

−1
x (n ) 2.1x(n −1) y (n ) x (n ) 1 z 2.1 1 1 1 y (n )
−1 2.1
z + +
)
)

−1
−2

−1 −1
(n

z
x(n

z 2
2y

0.5
0 .5

−1 −1
z 0.5 2 z

b) y(n) = 1.6x(n − 2) + 0.7 x(n) + 3y(n − 1) + 0.3 y(n − 2).

Solution

B lo c k D ia gra m S ig n a l F lo w G ra p h

0.7
x (n ) 0.7 0.7x(n) x (n ) 0.7x(n)

x (n )
x (n )

−1 −1 1.6 x (n −2)
z z

1.6
−1
−1
z
z
x (n −2 )
x (n − 2 ) 1.6 1.6 x (n − 2)
E2. 8 DSP, Chapter 2 -Discrete Time Signals and Systems
B loc k D ia gra m S ign a l F lo w G ra p h

y (n ) 3 y(n −1 ) y (n )

−1
z
−1 3
z

3 y(n − 1 ) y (n −1 )
y (n − 1 )
3

y (n − 1 ) 0.3 y (n −2) y (n −1)

−1
z
−1 z
0.3

0.3 y (n − 2 )
y (n −2 )
0.3

B loc k D ia gra m S ign a l F lo w G ra p h

x (n ) y (n ) y (n )
x (n ) 1 0.7 1 1
0.7 + +
−1 3 −1
−1
z z
−1
z + 3 z

0 .3
6
1.
−1
−1 z
z
−1 −1
z 1.6 0.3 z

E2.6. Determine the response of the discrete time systems governed by the following difference equations.

a) y(n) = 0.1 y(n − 1) + x(n − 1) + 0.7 x(n) ;

x(n) = 2 − n u(n) ; y( −1) = − 1

Solution

y(n) = 0.1 y(n – 1) + x(n – 1) + 0.7 x(n)

\ y(n) – 0.1 y(n –1) = 0.7 x(n) + x(n –1) .....(1)

Homogeneous solution

When the input is zero the equation (1) can be written as,

y(n) – 0.1 y(n –1) = 0 .....(2)

On substituting y(n) = ln in equation (2) we get,

ln – 0.1 l(n – 1) = 0

\ l(n – 1) (l – 0.1) = 0 Þ l = 0.1

The homogeneous solution yh(n) is given by,

yh(n) = Cln = C(0.1)n for n ³ 0 = C(0.1)n u(n) .....(3)

Particular solution

Given that, x(n) =

FG 1IJ u(n)
n
; ∴ y(n) = K
FG 1IJ u(n)
n

H 2K H 2K
Using the above values for x(n) and y(n) in equation(1) we get,

K
FG 1IJ u(n) − 0.1KFG 1IJ
n (n −1)
u(n − 1) = 0.7 ×
FG 1IJ u(n) + FG 1IJ
n n −1
u(n − 1) .....(4)
H 2K H 2K H 2K H 2K
To determine the value of ‘K’ evaluate equation(4) for n = 1.

K
FG 1IJ u(1) − 0.1KFG 1IJ u(0) = 0.7FG 1IJ u(1) + FG 1IJ u(0)
1 0 1 0

H 2K H 2K H 2K H 2K
1.35
0.5K − 0.1K = 0.35 + 1 ⇒ 0.4K = 1.35 ⇒ K= = 3.375
0.4
Solution for Exercise Problems E2. 9
\ The particular solution yp(n) is given by,

y p (n) = K
FG 1IJ u(n) = 3.375 × FG 1IJ u(n)
n n

.....(5)
H 2K H 2K
Total response

∴ Response, y(n) = yh (n) + y p (n)

LM
y(n) = C(0.1)n + 3.375 ×
FG 1IJ OP u(n)
n
(or) y(n) = C(0.1)n + 3.375
FG 1IJ n
; for n ≥ 0 .....(6)
MN H 2 K PQ H 2K
At n = 0, from equation (1) we get,
y(0) − 0.1y(−1) = 0.7 x(0) + x( −1) .....(7)

Given that : y(−1) = − 1 and x(n) = 2 −n u(n)

∴ x(0) = 1 and x(−1) = 0

On substituting the above values in equation (7) we get,

y(0) + 0.1 = 0.7

y(0) = 0.7 – 0.1

\ y(0) = 0.6

Put n = 0 and y(0) = 0.6 in equation (6).

y(0) = C(0.1)0 + 3 . 375 ×

FG 1IJ 0
= C + 3 . 375
H 2K
0.6 = C + 3 . 375
C = 0.6 − 3 . 375
C = −2 . 775
\ The total response is given by,
LM
y(n) = −2 . 775 (0.1)n + 3.375
FG 1IJ OP u(n)
n

MN H 2 K PQ
b) y(n) + 2.1 y(n – 1) + 0.2 y(n –2) = x(n) + 0.56 x(n – 1) ; x(n) = u(n) ; y(–2) = 1 ; y(–1) = –3.
Solution
y(n) + 2.1 y(n – 1) + 0.2 y(n – 2) = x(n) + 0.56 x(n – 1) .....(1)

Homogeneous Solution

When the input is zero, the equation(1) can be written as,

y(n) + 2.1 y(n –1) + 0.2 y(n –2) = 0 .....(2)

n
substitute y(n) = l in equation (2)

\ ln + 2.1 ln – 1 + 0.2 ln – 2 = 0

l(n –2) [l2 + 2.1 l + 0.2] = 0

The characteristic equation is,

λ2 + 2 .1 λ + 0.2 = 0 ⇒ bλ + 0.1g bλ + 2g = 0
\ The roots are, l1 = –0.1, l2= –2.

The homogenous solution yh(n) is given by,

yh (n) = C1 λn1 + C2 λn2

b g + C b−2g
yh (n) = C1 −0.1
n
2
n
for n ≥ 0

= C b −0.1g + C b −2g
n n .....(3)
1 2 u(n)
E2. 10 DSP, Chapter 2 -Discrete Time Signals and Systems
Particular Solution

Given that , x(n) = u(n) ; \ y(n) = K u(n). .....(4)

Using the above values for x(n) and y(n) in equation(1) we get,

K u(n) + 2.1 K u(n –1) + 0.2 K u(n –2) = u(n) + 0.56 u(n –1) .....(5)

To find ‘K’ evaluate equation (5), for n = 2.

\ K u(2) + 2.1 K u(1) + 0.2 K u(0) = u(2) + 0.56 u(1)

K + 2.1 K + 0.2 K = 1 + 0.56

1.56
3.3K = 1. 56 ⇒ K= = 0.47
3.3
\ yp(n) = 0.47 u(n)
\ Total response,
y(n) = yh(n) + yp(n)
y(n) = [C1(–0.1)n + C2(–2)n + 0.47] u(n)
y(n) = C1(–0.1)n + C2(–2)n + 0.47 for n ³ 0. .....(6)
At n = 0 from equation (1) we get,
y(0) + 2.1 y(–1) +0.2 y(–2) = x(0) + 0.56 x(–1) .....(7)
Given, y(–1) = –3, Also, x(n) = u(n)
y(–2) = 1 \ x(0) = 1 and x(–1) = 0.
On substituting the above values in equation (7),
y(0) + 2.1 (–3) + 0.2 (1) = 1 + 0.
y(0) – 6.3 + 0.2 = 1 Þ y(0) – 6.1 = 1
\ y(0) = 1 + 6.1 = 7.1
At n = 1 from equation (1) we get,
y(1) + 2.1 y(0) + 0.2 y(–1) = x(1) + 0.56 x(0)
We know that, y(0) = 7.1 , x(0) = 1
y(–1) = –3 , x(1) = 1
\ y(1) + 2.1 (7.1) + 0.2 (–3) = 1 + 0.56 Þ y(1) + 14.31 = 1.56
\ y(1) = 1.56 – 14.31 = –12.75
Put n = 0 and y(0) = 7.1 in eqaution(6).
y(0) = C1(–0.1)0 + C2(–2)0 + 0.47
7.1 = C1 + C2 + 0.47
\ C1 + C2 = 7.1 – 0.47
\ C1 + C2 = 6.63 .....(8)
Put n= 1 and y(1) = –12.75 in equation(6).
y(1) = C1(–0.1)1 + C2(–2)1 + 0.47
–12.75 = –0.1C1 – 2 C2 + 0.47
0.1C1 + 2C2 = 12.75 + 0.47
\ 0.1C1 + 2 C2 = 13.22 .....(9)

Equation (8) ´ 2 Þ 2 C1 + 2 C2 = 13.26

Equation (9) Þ 0.1 C1 + 2 C2 = 13.22

(–) (–) (–)

1.9 C1 = –0.04
−0.04
C1 = = − 0.02
1. 9
∴ C2 = 6.63 − C1 = 6.63 + 0.02 = 6.65
\ y(n) = –0.02 (–0.1)n + 6.65 (–2)n + 0.47, for n ³ 0.
= [0.47 – 0.02 (–0.1)n + 6.65 (–2)n] u(n).
Solution for Exercise Problems E2. 11
E2.7. Test the following systems for time invariance.
a) y(n) = x(n + 1) + x(n + 2)
Solution
Given that, y(n) = H {x(n)} = x(n + 1) + x(n + 2)
Response for Delayed Input
y(n –m) = H {x(n – m)} = x(n –m + 1) + x(n – m + 2)
Response for Unshifted Input
y(n) = H {x(n)} = x(n + 1) + x(n + 2)
Delayed Response

l q
y d (n) = z −m H x(n) = z −m x(n + 1) + x(n + 2) = z −m x(n + 1) + z −m x(n + 2)
= x(n − m + 1) + x(n − m + 2)

Here, y(n − m) = y d (n)

Hence system is time invariant.
b) y(n) = n ax(n)
Solution
Given that, y(n) = H{x(n)} = y(n) = n ax(n)
Response for Delayed Input
y(n – m) = H {x(n – m)} = (n–m) ax(n – m)
Delayed Response

l q
y d (n) = z −m H x(n) = z −m n a x(n) = n a x(n − m)

Here, y(n − m) ≠ y d (n)

Hence the system is time variant.
c) y(n) = x2(n + 2) + C
Solution
Given that, y(n) = H{x(n)} = x2(n + 2) + C
Response for Delayed Input
y(n –m) = x2(n – m + 2) + C
Delayed Response

l q
y d (n) = z −m H x(n) = z −m x 2 (n + 2) + C = z −m x 2 (n + 2) + C = x 2 (n − m + 2) + C

Here, y(n − m) = y d (n)

Hence the systems is time invariant.
d) y(n) = (n – 1) x2(n) + C
Solution
Given that, y(n) = H{x(n)} = (n - 1)x2(n) + C
Response for Delayed Input
y(n –m) = (n – m – 1)x2(n – m) + C
Delayed Response

l q
y d (n) = z −m H x(n) = z −m (n − 1) x 2 (n) + C = (n − 1) x 2 (n − m) + C.

Here, y(n − m) ≠ y d (n)

Hence the systen is time variant.

E2.8. Test the following systems for linearity.

a) y(n) = x2(n) + x3(n – 1)
Solution
Let, ‘H ’ be the system,
\ y(n) = H{x(n)} = x2(n) + x3(n – 1)
Consider two signals, x1(n) and x2(n).
E2. 12 DSP, Chapter 2 -Discrete Time Signals and Systems
Let y1(n) and y2(n) be responses of system ‘H’ for inputs x1(n) and x2(n).
\ y1(n) = H {x1(n)} = x12(n) + x13(n –1)
y2(n) = H {x2(n)} = x22(n) + x23(n –1)
\ a1 y1(n) + a2 y2(n) = a1 [x12(n) + x13(n –1)] + a2 [x22(n) + x23(n –1)] .....(1)
Consider a linear combination of inputs.
a1 x1(n) + a2 x2(n) = x3(n)
\ y3(n) = H {x3(n)} = x32(n) + x3(n – 1)
= [a1 x1(n) + a2 x2(n)]2 + [a1 x1(n – 1) + a2 x2(n – 1)]3 .....(2)
From equations (1) and (2) we can say that,
y3(n) ¹ a1 y1(n) + a2 y2(n)
Hence the system is non-linear.
b) y(n) = bx(n + 2) + n e x(n)
Solution
Let ‘H ’ be the system.
\ y(n) = H{x(n)} = b x(n + 2) + n ex(n)
Consider two signals x1(n) and x2(n).
Let y1(n) and y2(n) be their respective outputs.
l q
∴ y1(n) = H x1(n) = b x1(n + 2) + ne x1(n)

y (n) = H lx (n)q = b x (n + 2) + ne
2 2 2
x 2 ( n)

∴ a y (n) + a y (n) = a eb x (n + 2) + ne .....(1)

1 1 2 2 1j + a eb x (n + 2) + ne j
1
x1 ( n )
2 2
x 2 (n)

Consider a linear combination of inputs.

∴ a1 x1(n) + a 2 x 2 (n) = x 3 (n)

m r
∴ y 3 (n) = H x 3 (n) = b x 3 (n + 2) + ne x 3 (n)

= b a1 x1 (n + 2) + a 2 x 2 (n + 2) + ne[ a1 x1(n) + a 2 x 2 (n)]

= a1 b x1(n + 2) + a 2 b x 2 (n + 2) + nea1x1(n) ea 2 x 2 (n) .....(2)

From equations (1) and (2) we get,

y 3 (n) ≠ a1y1(n) + a 2y 2 (n)

Hence the system is non-linear system.

c) y(n) = a x(n) + b x(n)

Solution
Let ‘H ’ be the system.
l q
∴ y(n) = H x(n) = a x(n) + b x(n)

Consider two signals x1(n) and x2(n).

Let y1(n) and y2(n) be their respective outputs.

l q
∴ y1(n) = H x1(n) = a x1(n) + b x1(n)

y (n) = H lx (n)q = a
2 2 x 2 (n) + b x 2 (n)

∴ a1 y1(n) + a2 y2 (n) = a1 a x1(n) + b x1(n) + a2 a x2 (n) + b x2 (n) .....(1)

Consider a linear combination of inputs.

∴ a1 x1(n) + a2 x 2 (n) = x3 (n)

m r
∴ y 3 (n) = H x 3 (n) = a x 3 (n) + b x 3 (n)

= a a1 x1(n) + a 2 x 2 (n) + b a1 x1(n) + a 2 x 2 (n) .....(2)

From equations (1) and (2) we can say that,

y3(n) ¹ a1 y1(n) + a2 y2(n)
Hence the system is non-linear system.
Solution for Exercise Problems E2. 13
1
d) y(n) = x(n) +
x(n)

Solution
Let ‘H ’ be the system.

1
∴ y(n) = H x(n) = x(n) +l q x(n)

Consider two signals x1(n) and x2(n).

Let y1(n) and y2(n) be their respective outputs.

1 1
l q
∴ y1(n) = H x1(n) = x1(n) +
x1(n)
l
; y 2 (n) = H x 2 (n) = x 2 (n) + q x 2 (n)

a1 a2
∴ a1 y1(n) + a 2 y 2 (n) = a1 x1(n) + + a 2 x 2 (n) + .....(1)
x1(n) x 2 (n)

Consider linear combination of inputs.

∴ a1x1(n) + a2 x2 (n) = x3 (n)

1 1
m
∴ y 3 (n) = H x 3 (n) = x 3 (n) + r x 3 (n)
= a1x1(n) + a 2 x 2 (n) +
a1x1(n) + a 2 x 2 (n)
.....(2)

From equations (1) and (2) we get,

y3(n) ¹ a1 y1(n) + a2 y2(n)

Hence the system is non-linear system.

N M
e) y(n) = ∑b
m = −1
m x(n + m) + ∑c
m =0
m y(n + m)

Solution

Let, H be the system represented by the given equation.

N M
∴ y(n) = H x(n) = l q ∑b m x(n + m) + ∑c m y(n + m)
m = −1 m=0

Consider two signals, x1(n) and x 2 (n). Let y1(n) and y 2 (n) be the respective outputs.

N M
l q ∑ b x (n + m) + ∑ c y (n + m)
y1(n) = H x1(n) = m 1 m 1
m= −1 m=0

N M
l q ∑ b x (n + m) + ∑ c y (n + m)
y 2 (n) = H x 2 (n) = m 2 m 2
m= −1 m=0

L N O L M O
a y (n) + a y (n) = a M ∑ b x (n + m) + ∑ c y (n + m)P + a M ∑ b x (n + m) + ∑ c y (n + m)P
N M

.....(1)
1 1 2
MN
2 1
m= −1 PQ MN
m 1
m=0 PQ m 1 2
m= −1
m 2
m=0
m 2

Now consider linear combination of inputs

a1x1(n) + a 2 x 2 (n) = x 3 (n)

N M
∴ y 3 (n) = H x 3 (n) = m r ∑b m = −1
m x 3 (n + m) + ∑c
m=0
m y 3 (n + m)

N M
= ∑b
m = −1
m a1x1(n + m) + a 2 x 2 (n + m) + ∑c
m= 0
m y 3 (n + m)

N M M
.....(2)
= a1 ∑b
m = −1
m x1(n + m) + a 2 ∑b
m = −1
m x 2 (n − m) + ∑c
m= 0
m y 3 (n + m)

By time invariant property,

If y3(n) = H {a1 x1(n) + a2 x2(n)} ; then, y3(n + m) = H {a1 x1(n + m) + a2 x2(n + m)}

If y2(n) = H {x2(n)}, then y2(n + m) = H {x2(n + m)}

E2. 14 DSP, Chapter 2 -Discrete Time Signals and Systems
If y1(n) = H {x1(n)}, then y1(n + m) = H {x1(n + m)}

\ y3(n + m) = H {a1 x1(n + m) + a2 x2(n + m)} = a1 H {x1(n + m)} + a2 H {x2(n + m)}

= a1 y1(n + m) + a2 y2(n + m) .....(3)

Using equation (3), the equation(2) can be written as,
N N M
y 3 (n) = a1 ∑b
m = −1
m x1(n + m) + a 2 ∑b
m = −1
m x 2 (n + m) + ∑c
m=0
m a1y1(n + m) + a 2 y 2 (n + m)

N N M M
= a1 ∑b
m = −1
m x1(n + m) + a 2 ∑b
m = −1
m x 2 (n + m) + a1 ∑c
m=0
m y1(n + m) + a 2 ∑c
m=0
m y 2 (n + m)

F b N M I
= a1 GH ∑
m = −1
m x1(n + m) + ∑c
m= 0
m y1(n + m) JK
F b N M I
+ a2 GH ∑
m = −1
m x 2 (n − m) + ∑c
m= 0
m JK
y 2 (n + m) .....(4)

From equations (1) and (4) we can say that,

y 3 (n) = a1 y1(n) + a 2 y2 (n)

Hence the system is linear.
E2.9. Test the causality of the following systems.
a) y(n) = x(n) – x(–n – 2) + x(n –1)
Solution
When, n = –2, y(–2) = x(–2) – x(0) + x(–3)
n = –1, y(–1) = x(–1) – x(–1) + x(–2)
n = 0, y(0) = x(0) –x(–2) + x(–1)
n = 1, y(1) = x(1) –x(–3) + x(0)
n=2, y(2) = x(2) – x(–4) + x(1)
For n £ –2, the system response depends on future input.
Hence the system is noncausal.

b) y(n) = a x(2n) + x(n2)

Solution
When, n = –1, y(–1) = a x(–2) + x(1) ; Response depends on future input
n = 0, y(0) = a x(0) + x(0)
n = 1, y(0) = a x(2) + x(1) ; Response depends on future input.
Except n = 0 for all other values of n, the response depends on future input.
Hence the system is noncausal.
n n
c) y(n) = ∑ x(m) + ∑ x(2m)
m = −1 m = −∞

Solution
0 0
n = 0, y(0) = ∑ x(m) + ∑ x(2m) ⇒ y(0) = x( −1) + x(0) + ..... + x(−4) + x(−2) + x(0).....
m = −1 m = −∞

∴ y(0) depends on present and past inputs.

1 1
n = 1, y(1) = ∑ x(m) + ∑ x(2m) ⇒ y(1) = x( −1) + x(0) + x(1) + ..... + x( −4) + x( −2) + x(0) + x(2)
m = −1 m = −∞

y(1) depends on future input x(2).

The system response depends on future input for n > 0.

Hence it is noncausal system.
Solution for Exercise Problems E2. 15
d) y(n) = (0.3)n u(n + 2) Note : For causality y(n) = 0 ; for n < 0.

Solution
1
( 0 .3 ) 3
1 1
(0 .3)
n
( 0 .3 ) 2 ( 0 .3 ) 2
1 y (n )
1
0 .3 u (n + 2 ) 0 .3
1 1
X ⇒
0.3
2
( 0 .3 )
−3 −2 −1 0 1 2 −3 −2 −1 0 1 2 −3 −2 −1 0

Here y(n) ¹ 0 for n < 0. Therefore the system is noncausal.

4
e) y(n) = ∑ x(n − k)
k = −4

Solution
4
y(n) = ∑ x(n − k) = x(n + 4)
k = −4
+ x(n + 3) + x(n + 2) + x(n + 1) + x(n) + x(n − 1) + x(n − 2) + x(n − 3) + x(n − 4)

For any value of n the system response depends on future inputs.

Hence, the system is noncausal.

E2.10. Test the stability of the following discrete time systems.

a) y(n) = x2(n) + x(n + 1)
Solution
The given system involves squaring operation and so it is nonlinear.
The operations performed by system is squaring and shifting.
A bounded input signal will remain bounded even after squaring and shifting.
Hence the system is BIBO stable.
b) y(n) = n x(n – 1)
Solution
The given system involves multiplication by n and so it is time variant system.
If x(n) doesnot tend to “0” as n tends to infinity then the system is unstable.
c) h(n) = (0.4)n u(n + 3)
Solution
Here, h(n) = 0.4n u(n + 3) = 0.4n ; For n = –3 to + ¥
+∞ ∞ −1 ∞
1
∑ h(n) = ∑ (0.4) = ∑ (0.4) n n
+ ∑ (0.4) n
= (0.4)−3 + (0.4)−2 + (0.4)−1 +
1 − 0.4
= 26.04 = Constant
n = −∞ n = −3 n = −3 n= 0

Hence it is stable system.

d) h(n) = (8)n u(4 – n)

Solution
Here, h(n) = 8n u(4 – n) = 8n ; for n = – ¥ to +4
∞
For stability, ∑ |h(n)|
n = −∞
< ∞

+∞ 4 0 4
∴ ∑ h(n)
n = −∞
= ∑ (8)
n = −∞
n
= ∑ (8) + ∑ (8)
n = −∞
n

n =1
n

∞
= ∑ (8)
n= 0
−n
+ 81 + 8 2 + 8 3 + 84

=
∞
F 1I
∑ GH 8 JK
n
+ 4680 =
∞

∑ (0.125) n
+ 4680 =
1
+ 4680
n= 0 n= 0 1 − 0.125
= 4681.14 = Constant
Hence it is stable system.
E2. 16 DSP, Chapter 2 -Discrete Time Signals and Systems
e) y(n) = x(n – 3)
If x(n) = d(n), then y(n) = h(n)
d(n – 3) = 1, only
\ h(n) = d(n – 3) when n = 3, and
∞ +∞ zero for all other
∴ ∑
n = −∞
h(n) = ∑
n = −∞
δ(n − 3) = 1
values of n

Hence the system is stable.

E2.11. Determine the range of values of ‘a’ and ‘b’ for the stability of LTI system with impulse response,

h(n) =
|RS ( −4a) n
; n ≥ 0
T|(2b) −n
; n < 0

Solution
The condition to be satisfied for the stability of the system is,
+∞ −1 ∞

∑ h(n) = ∑ (2b)−n + ∑ (−4a)n

n = −∞ n= −∞ n= 0
−1 ∞
= ∑ |2b|−n + ∑ |4a|n
n = −∞ n= 0

∞ ∞
= ∑ |2b| n
+ ∑ |4a| n

n =1 n= 0

∞ ∞
= ∑ |2b| n
− |2b|0 + ∑ |4a| n

n= 0 n= 0
∞
1
If 0 <|2b| < 1, then ∑|2b| n
=
1 − |2b|
n= 0

∞
1
If 0 <|4a| < 1, then ∑|4a| n
=
1 − |4a|
n= 0

+∞
1 1
∴ ∑ h(n) = 1 − |2b| − 1+ 1 − |4a| = Constant
n = −∞

∴ Condition for stability is,

1
0 < |2b| < 1 ⇒ 0 < |b| <
2
1
0 < |4a| < 1 ⇒ 0 < |a| <
4

E2.12. a) Determine the impulse response for the cascade of two LTI systems having impulse responses,

h1 (n) =
FG 1 IJ n

u(n) and h2 (n) = δ (n − 3)

H7 K
Solution
The impulse response of the cascade system is given by,
h(n) = h1(n) ∗ h2 (n) = h2 (n) ∗ h1(n)
∞
= ∑ h (m) h (n − m) ;
m= −∞
2 1 'm' is dummy variable

∴ h(n) = ∑
∞
h2 (m) h1(n − m) =
∞

∑ δ(m − 3)
FG 1IJ n −m
=
∞

∑ δ(m − 3)
FG 1IJ FG 1IJ
n −m

m= 0 m=0
H 7K m=0
H 7K H 7K
FG 1IJ ∑ δ(m − 3) FG 1IJ
=
n ∞ −m

H 7K H 7K m=0

F 1I will be nonzero, only when m = 3.

The product of d(m – 3) and GH JK
−m

7
F 1I F 1I for n ≥ 3
∴ h(n) = G J G J
n −3

H 7K H 7K
F 1I
h(n) = G J u(n − 3) for all n.
n− 3

H 7K
Solution for Exercise Problems E2. 17
b) Determine the overall impulse response of the interconnected discrete time system shown in fig E2.12.

Take, h1 (n) =
FG 1 IJ n

u(n) ; h2 (n) =
FG 1 IJ n

u(n) ; h3 (n) =
FG 1 IJ n

u(n)
H 3K H6K H 9K
h 2 ( n)

x (n )
+ y (n )

h 1 ( n) + h 3 ( n)

Solution F ig E 2 .1 2
The given system can be redrawn as,

h 2 (n)

x (n ) y (n ) x (n ) y (n )
h 2 (n) + h 2 (n) + [(h 2 (n) + h1(n)) ∗ h3 (n)]

+ h 3 (n)

h1(n) x (n ) h(n) y (n )

b g
h(n) = h2 (n) + h1(n) + h2 (n) ∗ h3 (n) = h2 (n) + h1(n) ∗ h3 (n) + h2 (n) ∗ h3 (n) b g b g
Evaluation of h1(n) * h3(n)
∞
h1(n) ∗ h3 (n) = ∑ h (m) h (n − m)
m= −∞
1 3

=
n

∑ h (m) h (n − m) =
n

∑ GH 3 JK
F 1I FG 1IJ
m n −m
=
FG 1IJ ∑ FG 1IJ
n n m
9m =
FG 1IJ ∑ FG 9 IJ
n n m

m= 0
1 3
m= 0
H 9K H 9K H 3K m= 0
H 9K H 3K m= 0

FG 9 IJ − 1 n +1

F 1I H 3 K
n
= G J
H 9K 9 − 1
3
FG 9 IJ 9 − 1 n
FG 9 IJ 9 − 1 n

F 1I H 3 K 3
= G J
n
= G J
F 1I H 3 K 3 F 1I L 1 F 9 I 9 − 1 OP
= G J M G J
n n n

H 9K 9
−1
H 9K 2 H 9 K MN 2 H 3 K 3 2 PQ
3

G J G J − 21 FGH 91IJK = 32 FGH 31IJK − 21 FGH 91IJK for n ≥ 0.

3 F 1I F 9 I
n n n n n
=
2 H 9K H 3K

3 F 1I
G J u(n) − 21 FGH 91IJK u(n) for all 'n'
n n
=
2 H 3K
Evaluation of h2(n) * h3(n)

h2 (n) ∗ h3 (n) =
+∞

∑ h (m) h (n − m) = ∑ h (m) h (n − m)
F 1I FG 1IJ = FG 1IJ ∑ FG 1IJ FG 1IJ
n
=
n

∑ GH 6 JK
m n −m n n m −m

m= −∞
2 3
m= 0
H 9K H 9K H 6K H 9K
2 3
m= 0 m= 0

FG 3 IJ − 1 LM FG 3 IJ − 1OP n+1 n+1

F 1 I F 1I n
F 1I F 3 I = FG 1IJ H 2 K
n
= G J ∑ G J 9 = G J ∑ G J
m n
F 1I M H 2 K P
n m n n

H 9 K H 2 K H 9 K 3 − 1 = GH 9 JK MM 1 PP
m
H 9K H 6K m=0 m=0
2 MN 2 PQ
F 1I L F 3 I 3 − 2OP = FG 1IJ LMFG 3 IJ 3 − 2OP = FG 1IJ FG 3 IJ 3 − 2 FG 1IJ
= G J M2G J
n n n n n n n

H 9 K MN H 2 K 2 PQ H 9 K MNH 2 K PQ H 9 K H 2 K H 9K
F 3I F 1In
F 1 I F 1I F 1 I n
= G J 3 − 2 G J = 3 G J − 2 G J = 3 G J u(n) − G J u(n) for all n.
F 1I n n n n

H 18 K H 9K H 6K H 9K H 6K H 9K
Overall Impulse Response

Now the overall impulse response h(n) is given by,

b
h(n) = h2 (n) + h1(n) ∗ h3 (n) + h2 (n) ∗ h3 (n) g b g
E2. 18 DSP, Chapter 2 -Discrete Time Signals and Systems

h(n) =
FG 1IJ u(n) + LMFG 3 IJ FG 1IJ u(n) − FG 1IJ FG 1IJ u(n)OP + LM3 FG 1IJ u(n) − FG 1IJ u(n)OP
n n n n n

H 6K MNH 2 K H 3 K H 2 K H 9 K PQ MN H 6 K H 9 K PQ
FG 1IJ u(n) − 3 FG 1IJ u(n) + 3 FG 1IJ u(n) =
n n n
LM4 F 1I n
FG IJ
3 1
n
3 FG 1IJ OP u(n)
n
⇒ h(n) = 4
H 6K 2 H 9K 2 H 3K MN GH 6 JK −
H K
2 9
+
2 H 3 K PQ

E2.13. Determine the response of an LTI system whose and impulse response h(n) and input x(n) are given by,

a) l
h(n) = 1, 4, 1, − 2, 1 q
A
l
x(n) = 1, 3, 5, − 1, − 2 q
A
Solution

The response y(n) of the system is given by convolution of x(n) and h(n).
+∞
y(n) = x(n) ∗ h(n) = ∑ x(m) h(n − m)
m= −∞

Input sequence starts at n = –1

Impulse response starts at n = –2

Therefore the output sequence start at, n = –1 + (–2) = –3

The output consists of 5 + 5 –1 = 9 samples.

The 9 samples of output sequence are computed by table method as shown below.

m –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7

x(m) 1 3 5 –1 –2

h(m) 1 4 1 –2 1

h(–m) 1 –2 1 4 1

h(–3–m) 1 –2 1 4 1

h(–2–m) 1 –2 1 4 1

h(–1–m) 1 –2 1 4 1

h(0 – m) 1 –2 1 4 1

h(1 – m) 1 –2 1 4 1

h(2 – m) 1 –2 1 4 1

h(3 – m) 1 –2 1 4 1

h(4 – m) 1 –2 1 4 1

h(5 – m) 1 –2 1 4 1

3
When n = −3 ; y(−3) = ∑ x(m) h(−3 − m) = 0 + 0 + 0 + 0 + 1+ 0 + 0 + 0 + 0 = 1
m = −5

3
When n = −2 ; y(−2) = ∑ x(m) h(−2 − m) = 0 + 0 + 0 + 4 + 3 + 0 + 0 + 0
m = −4
=7

3
When n = −1 ; y(−1) = ∑ x(m) h(−1 − m) = 0 + 0 + 1+ 12 + 5 + 0 + 0 = 18
m = −3

3
When n = 0 ; y(0) = ∑ x(m) h(0 − m)
m = −2
= 0 − 2 + 3 + 20 − 1+ 0 = 20

3
When n = 1 ; y(1) = ∑ x(m) h(1 − m)
m = −1
= 1 − 6 + 5 − 4 − 2 = −6
Solution for Exercise Problems E2. 19
4
When n = 2 ; y(2) = ∑ x(m) h(2 − m) = 0 + 3 − 10 − 1− 8 + 0 = −16
m = −1
5
When n = 3 ; y(3) = ∑ x(m) h(3 − m) = 0 + 0 + 5 + 2 − 2 + 0 + 0 = 5
m = −1
6
When n = 4 ; y(4) = ∑ x(m) h(4 − m) = 0 + 0 + 0 − 1+ 4 + 0 + 0 + 0 = 3
m = −1
7
When n = 5 ; y(5) = ∑ x(m) h(5 − m) = 0 + 0 + 0 + 0 − 2 + 0 + 0 + 0 + 0 = −2
m = −1

l
∴ y(n) = 1, 7, 18, 20, − 6, − 16, 5, 3, − 2 q
A

E2.13. b) h(n) =
|RS1 ; 0 ≤n≤ 2

T|0 ; n ≥ 3
n
x(n) = a u(n) ; |a| < 1
Solution
x (m ) h (m ) h ( −m )
0
a =1 1 1 1
a
2
a
3
a

0 1 2 3 m 0 1 2 3 6 m −3 −2 −1 0
4 5 −6 −5 −4 m
By convolution formula,
∞
y(n) = ∑ x(m) h(n − m)
m = −∞

∞ ∞
When n = 0 ; y(0) = ∑ x(m) h(0 − m) = ∑ v0 (m)
m=0 m=0

x (m ) h( −m ) v 0 (m )
0 0
a =1 1 1=a
a
a
2
x ⇒
3
0
a ∴ y (0) = a

0 1 2 3 m −3 −2 −1 0 1 m 0 1 2 m
∞ ∞
When n = 1 ; y(1) = ∑ x(m) h(1 − m) = ∑ v1(m)
m=0 m=0

x (m ) h (1 − m ) v 1 (m )
0
0
a =1 1=a
1 1
a 1
a
a
2
⇒
3
x y (1) = a
0
+ a
1
a

0 1 2 3 m −2 −1 0 1 m −1 0 1 2 m
Similarly,
When, n = 2 ; y(2) = a0 + a1 + a2
When, n = 3 ; y(3) = a1 + a2 + a3
When, n = 4 ; y(4) = a2 + a3 + a4
When, n = 5 ; y(5) = a3 + a4 + a5
n
∴ y(n) = ∑a
k =0
k
; for n = 0, 1, 2

n
= ∑a
k =n − 2
k
; for n > 2
E2. 20 DSP, Chapter 2 -Discrete Time Signals and Systems
E2.14. Perform circular convolution of the two sequences,
l
a) x1 (n) = 1, 2, − 1 −1 q and x2 (n) = 2, 4, 6, 8 l q
A A
Solution
N −1
Let x 3 (n) = x1(n) ∗ x 2 (n) = ∑ x (m) x
m=0
1 2,n (m) ; x 2,n (m) = x 2 ((n − m))N ; N = 4

x 1 (1) = 2 x 2 (1) = 4 x2 (3) = 8

x 1 ( 2 ) = −1 x 1( m ) x 1( 0 ) = 1 x2 (2) = 6 x 2 (m ) x 2 (0 ) = 2 x 2 (2 ) = 6 x 2 ( −m ) x 2 (0) = 2

x 1 ( 3 ) = −1 x 2 (3 ) = 8 x 2 (1) = 4

3 3
When n = 0 ; x 3 (0) = ∑ x (m) x
m=0
1 2,0 (m) = ∑ v (m) = 2 + 16 − 6 − 4 = 8
m=0
0

2 8 16

−1 x 1 (m ) 1 x 6 x 2 ,0 ( m ) 2 ⇒ −6 v 0 (m ) 2

−1 4 −4

3 3
When n = 1 ; x 3 (1) = ∑ x (m) x
m=0
1 2,1(m) = ∑ v (m) = 4 + 4 − 8 − 6 = −6
m=0
1

2 2 4

−1 x 1(m) 1 x 8 x 2, 1(m ) 4 ⇒ −8 v 1(m) 4

−1 6 −6

3 3
When n = 2; x 3 (2) = ∑ x (m) x
m=0
1 2,2 (m) = ∑ v (m) = 6 + 8 − 2 − 8 = 4
m=0
2

2 4 8

−1 x1(m) 1 x 2 x 2 , 2 (m) 6 ⇒ −2 v 2 (m ) 6

−1 8 −8

3 3
When n = 3 ; x 3 (3) = ∑ x (m) x
m=0
1 2,3 (m) = ∑ v (m) = 8 + 12 − 4 − 2 = 14
m=0
3

2 6 12

−1 x 1(m) 1 x 4 x 2, 3 (m ) 8 ⇒ −4 v 3 (m) 8

−1 2 −2

l
x 3 (n) = 8, − 6, 4, 14 q
A
b) Perform the circular convolution of the two sequences,
l
x1 (n) = 0, 0.6, − 1, 1.5, 2 ; x2 (n) = −2, 3, 0.2, 0.7, 0.8 q l q
A A
Solution
The response x3(n) of the system is given by convolution of x1(n) and x2(n).
Solution for Exercise Problems E2. 21

N −1 4
x 3 (n) = x1(n) ∗ x 2 (n) = ∑ x (m) x ((n − m))
m=0
1 2 N = ∑ x (m) x ((n − m))
m=0
1 2 5

4
= ∑ x (m) x
m=0
1 2,n (m)

m –4 –3 –2 –1 0 1 2 3 4

x1(m) 0 0.6 –1 1.5 2

x2(m) –2 3 0.2 0.7 0.8

x2((–m))5 = x2,0(m) 0.8 0.7 0.2 3 –2 0.8 0.7 0.2 3

x2((1–m))5 = x2,1(m) 0.8 0.7 0.2 3 –2 0.8 0.7 0.2

x2((2–m))5 = x2,2(m) 0.8 0.7 0.2 3 –2 0.8 0.7

x2((3–m))5 = x2,3(m) 0.8 0.7 0.2 3 –2 0.8

x2((4–m))5 = x2,4(m) 0.8 0.7 0.2 3 –2

When n = 0 ;
4
x 3 (0) = ∑ x (m) x
m=0
1 2,0 (m) = x1(0) x 2,0 (0) + x1(1) x 2,0 (1) + x1(2) x 2,0 (2) + x1(3) x 2,0 (3) + x1(4) x 2,0 (4)

b g b
= (0 × −2) + 0.6 × 0.8 + −1 × 0.7 + 1.5 × 0.2 + 2 × 3 = 6.08 g b g b g
Similarly

g b g b b g b g
When n = 1 ; x 3 (1) = (0 × 3) + 0.6 × −2 + −1 × 0.8 + 1.5 × 0.7 + 2 × 0.2 = −0.55

When n = 2 ; x (2) = (0 × 0.2) + b0.6 × 3g + b −1 × −2g + b1.5 × 0.8g + b2 × 0.7g = 6.4

3

When n = 3 ; x (3) = (0 × 0.7) + b0.6 × 0.2g + b −1 × 3g + b1.5 × −2g + b2 × 0.8g = − 4.28

3

When n = 4 ; x (4) = (0 × 0.8) + b0.6 × 0.7g + b −1 × 0.2g + b1.5 × 3g + b2 × −2g = 0.72

3

∴ x (n) = l6.08, − 0.55, 6.4, − 4 . 28, 0.72q

3

A
E2.15. The input x(n) and impulse response h(n) of an LTI system are given by,

x(n) = l − 1, 1, − 1, 1, − 1, 1 q and l
h(n) = −0.5, 0.5, − 1, 0.5, − 1, − 2 q
A A
Find the response of the system using,
a) Linear Convolution
b) Circular Convolution

Solution

a) Response of LTI System Using Linear Convolution

+∞
Let, y(n) = x(n) ∗ h(n) = ∑ x(m) h(n − m)
m= −∞
+∞
= ∑ x(m) h (m) ; where h (m) = h(n − m)
n n
m= −∞

x(n) starts at n = −1 and h(n) starts at n = 0.

∴ y(n) will start at n = 0 + ( −1) = − 1

Length of x(n) is 6 and h(n) is 6.

Hence length of y(n) is 6 + 6 –1 = 11. Also y(n) ends at n = 0 + (–1) + (6 + 6 –2) = 9

E2. 22 DSP, Chapter 2 -Discrete Time Signals and Systems

The 11 samples of y(n) computed by table method as shown below.

m –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 9

x(m) –1 1 –1 1 –1 1

h(m) –0.5 0.5 –1 0.5 –1 –2

h(–m) –2 –1 0.5 –1 0.5 –0.5

h–1(–m) –2 –1 0.5 –1 0.5 –0.5

h0(m) –2 –1 0.5 –1 0.5 –0.5

h1(m) –2 –1 0.5 –1 0.5 –0.5

h2(m) –2 –1 0.5 –1 0.5 –0.5

h3(m) –2 –1 0.5 –1 0.5 –0.5

h4(m) –2 –1 0.5 –1 0.5 –0.5

h5(m) –2 –1 0.5 –1 0.5 –0.5

h6(m) –2 –1 0.5 –1 0.5 –0.5

h7(m) –2 –1 0.5 –1 0.5 –0.5

h8(m) –2 –1 0.5 –1 0.5 –0.5

h9(m) –2 –1 0.5 –1 0.5 –0.5

4
When n = −1, y( −1) = ∑ x(m) h
m = −6
−1(m)

= x( −6) h−1(−6) + x( −5) h−1( −5) + x( −4)h−1(−4) + x( −3) h−1(−3) + x( −2) h−1(−2) + x( −1)h−1(−1)
+ x(0) h−1(0) + x(1) h−1(1) + x(2)h−1(2) + x(3) h−1(3) + x(4) h−1(4)
= 0 + 0 + 0 + 0 + 0 + (−1 × − 0.5 ) + 0 + 0 + 0 + 0 + 0 = 0.5

4
When n = 0 ; y(0) = ∑ x(m) h (m) 0 b g b
= 0 + 0 + 0 + 0 + −1 × 0.5 + 1 × −0.5 + 0 + 0 + 0 + 0 = −1 g
m= −5

4
When n = 1 ; y(1) = ∑ x(m) h (m) 1 b g b g b
= 0 + 0 + 0 + −1 × −1 + 1 × 0.5 + −1 × −0.5 + 0 + 0 + 0 = 2 g
m= −4

4
When n = 2 ; y(2) = ∑ x(m) h (m) 2 b g b g b
= 0 + 0 + −1 × 0.5 + 1 × −1 + −1 × 0.5 + 1 × −0.5 + 0 + 0 = −2.5 g b g
m= −3

4
When n = 3 ; y(3) = ∑ x(m) h (m) 3 b g b g b g b
= 0 + −1 × −1 + 1 × 0.5 + −1 × −1 + 1 × 0.5 + −1 × −0.5 + 0 = 3.5 g b g
m= −2
4
When n = 4 ; y(4) = ∑ x(m) h (m) 4 b g b g b g b
= −1 × −2 + 1 × −1 + −1 × 0.5 + 1 × −1 + −1 × 0.5 + 1 × −0.5 = −1.5 g b g b g
m= −1
5
When n = 5 ; y(5) = ∑ x(m) h (m) 5 b g b g b g b
= 0 + 1 × −2 + −1 × −1 + 1 × 0.5 + −1 × −1 + 1 × 0.5 + 0 = 1 g b g
m= −1
6
When n = 6 ; y(6) = ∑ x(m) h (m) 6 b g b g b
= 0 + 0 + −1 × −2 + 1 × −1 + −1 × 0.5 + 1 × −1 + 0 + 0 = −0.5 g b g
m= −1
7
When n = 7 ; y(7) = ∑ x(m) h (m) 7 b g b g b
= 0 + 0 + 0 + 1 × −2 + −1 × −1 + 1 × 0.5 + 0 + 0 + 0 = −0.5 g
m= −1
8
When n = 8 ; y(8) = ∑ x(m) h (m) 8 b g b
= 0 + 0 + 0 + 0 + −1 × −2 + 1 × −1 + 0 + 0 + 0 + 0 = 1 g
m= −1
9
When n = 9 ; y(9) = ∑ x(m) h (m) 9 b g
= 0 + 0 + 0 + 0 + 0 + 1 × −2 + 0 + 0 + 0 + 0 + 0 = − 2
m= −1

The response of LTI system y(n) is,

l
y(n) = 0.5, − 1, 2, − 2.5, 3.5, − 1. 5, 1, − 0.5, − 0.5, 1, − 2 q
A
Solution for Exercise Problems E2. 23
b) Response of LTI system using circular convolution
The response y(n) is 11-point sequence. The y(n) start at n = –1 and end of n = 9. Hence both x(n) and h(n)
should be converted to 11-point sequence such that they start at n = –1 and end at n = 9 by appending zeros for
missing samples.

l
∴ x(n) = −1, 1, − 1, 1, − 1, 1, 0, 0, 0, 0, 0 q
A
l
h(n) = 0, − 0.5, 0.5, − 1, 0.5, − 1, − 2, 0, 0, 0, 0 q
A
9 9
Now, y(n) = x(n) ∗ h(n) = ∑ x(m) h((n − m))
m= −1
11 = ∑ x(m) h (m) ;
m= −1
n where hn (m) = h((n − m))11

The 11 samples of y(n) are computed by table method as shown below.

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

x(m) –1 1 –1 1 –1 1 0 0 0 0 0

h(m) 0 –0.5 0.5 –1 0.5 –1 –2 0 0 0 0

h(–m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0

h–1(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0 –2 –1 0.5 –1 0.5

h0(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0 –2 –1 0.5 –1

h1(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0 –2 –1 0.5

h2(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0 –2 –1

h3(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0 –2

h4(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0

h5(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0

h6(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0

h7(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0

h8(m) 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0

h9(m) 0 0 0 –2 –1 0.5 –1 0.5 –0.5 0 0 0 0 0 –2 –1 0.5 –1 0.5 –0.5

9
When n = −1 ; y( −1) = ∑ x(m) h −1(m) b g
= −1 × −0.5 + 0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 = 0.5
m= −1

9
When n = 0 ; y(0) = ∑ x(m) h (m) 0 b g b g
= −1 × 0.5 + 1 × −0.5 + 0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 = −1
m= −1
9
When n = 1 ; y(1) = ∑ x(m) h (m) 1 b g b g b g
= −1 × −1 + 1 × 0.5 + −1 × −0.5 + 0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 = 2
m= −1
9
When n = 2 ; y(2) = ∑ x(m) h (m) 2 b g b g b g b g
= −1 × 0.5 + 1 × −1 + −1 × 0.5 + 1 × −0.5 + 0 + 0 + 0 + 0 + 0 + 0 + 0 = −2.5
m= −1
9
When n = 3 ; y(3) = ∑ x(m) h (m) 3 b g b g b g b g b
= −1 × −1 + 1 × 0.5 + −1 × −1 + 1 × 0.5 + −1 × −0.5 + 0 + 0 + 0 + 0 + 0 + 0 = 3.5 g
m= −1
9
When n = 4 ; y(4) = ∑ x(m) h (m) 4 b g b g b g b g b
= −1 × −2 + 1 × −1 + −1 × 0.5 + 1 × −1 + −1 × 0.5 + 1 × −0.5 + 0 + 0 + 0 + 0 + 0 = −1.5 g b g
m= −1
9
When n = 5 ; y(5) = ∑ x(m) h (m) 5 b g b g b g b
= 0 + 1 × −2 + −1 × −1 + 1 × 0.5 + −1 × −1 + 1 × 0.5 + 0 + 0 + 0 + 0 + 0 = 1 g b g
m= −1
9
When n = 6 ; y(6) = ∑ x(m) h (m) 6 b g b g b g b
= 0 + 0 + −1 × −2 + 1 × −1 + −1 × 0.5 + 1 × −1 + 0 + 0 + 0 + 0 + 0 = −0.5 g
m= −1
9
When n = 7 ; y(7) = ∑ x(m) h (m) = 0 + 0 + 0 + b1× −2g + b−1 × −1g + b1× 0.5g + 0 + 0 + 0 + 0 + 0 = −0.5
m= −1
7
E2. 24 DSP, Chapter 2 -Discrete Time Signals and Systems
9
When n = 8 ; y(8) = ∑ x(m) h (m) 8 b g b g
= 0 + 0 + 0 + 0 + −1 × −2 + 1 × −1 + 0 + 0 + 0 + 0 + 0 = 1
m= −1
9
When n = 9 ; y(9) = ∑ x(m) h (m) 9 b g
= 0 + 0 + 0 + 0 + 0 + 1 × −2 + 0 + 0 + 0 + 0 + 0 = −2
m= −1

The response of LTI system y(n) is,

l
y(n) = 0.5, − 1, 2, − 2.5, 3.5, − 1. 5, 1, − 0.5, − 0.5, 1, − 2 q
A
E2.16. Perform linear convolution of the following sequences by,
i) Overlap add method
ii) Overlap save method
x(n) = 1, l − 1, 2 1, − 1, 2, 1, − 1, 2 q
l
h(n) = 2, 3, − 1 q
Solution
Overlap Add Method
l
x(n) = 1, − 1, 2, 1, − 1, 2, 1, − 1, 2 q
x1(n) = 1, n = 0 x 2 (n) = 1, n = 3 x 3 (n) = 1, n = 6
= −1, n = 1 = −1, n = 4 = −1, n = 7
= 2, n = 2 = 2, n = 5 = 2, n = 8
l
h(n) = 2, 3, − 1 q
Let, y1(n), y2(n), y3(n) be output of linear convolution of x1(n), x2(n), x3(n) with h(n) respectively.
Here, h(n) starts at nh = 0.
x1(n) starts at, n = n1 = 0 \ y1(n) starts at, n = 0 + 0 = 0
x2(n) starts at, n = n2 = 3 \ y2(n) starts at, n = 3 + 0 = 3
x3(n) starts at, n = n3 = 6 \ y3(n) starts at, n = 6 + 0 = 6
Here, N1 = 9, N2 = 3, N3 = 3
N2 – 1 = 2
N2 + N3 – 1 = 5
Convolution output of each section will consists of 3 + 3 – 1 = 5 samples.
Convolution of Section - 1

m –2 –1 0 1 2 3 4

x(m) 1 –1 2

h(m) 2 3 –1

h(–m)=h0(m) –1 3 2

h1(m) –1 3 2

h2(m) –1 3 2

h3(m) –1 3 2

h4(m) –1 3 2
+∞
y1(n) = x1(n) ∗ h(n) = ∑ x (m) h(n − m)
m= −∞
1

+∞
= ∑ x (m) h (m) ;
m= −∞
1 n n = 0, 1, 2, 3, 4

where hn (m) = h(n − m)

When n = 0 ; y1(0) = ∑ x (m) h (m) = 0 + 0 + 2 + 0 + 0 = 2
1 0

When n = 1 ; y (1) = ∑ x (m) h (m) = 0 + 3 − 2 + 0 = 1

1 1 1

When n = 2 ; y (2) = ∑ x (m) h (m) = − 1 − 3 + 4 = 0

1 1 2

When n = 3 ; y (3) = ∑ x (m) h (m) = 0 + 1+ 6 + 0 = 7

1 1 3

When n = 4 ; y (4) = ∑ x (m) h (m) = 0 + 0 − 2 + 0 + 0 = −2

1 1 4
Solution for Exercise Problems E2. 25
∴ y1(n) = l 2, 1, 0, 7, − 2 q
An=0

Convolution of sections 2 and 3

The convolution of section –2 and 3 are identical to that of section -1 except the starting value of n.

∴ y 2 (n) = l 2, 1, 0, 7, − 2 q
An= 3

∴ y 3 (n) = l 2, 1, 0, 7, − 2 q
An= 6

Overall Output

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

y1(n) 2 1 0 7 –2

y2(n) 2 1 0 7 –2

y3(n) 2 1 0 7 –2

y(n) 2 1 0 9 –1 0 9 –1 0 7 –2

l
y(n) = 2, 1, 0, 9, − 1, 0, 9, − 1, 0, 7, − 2 q
Overlap save Method

l
x(n) = 1, − 1, 2, 1, − 1, 2, 1, − 1, 2 q
h(n) = l2, 3, − 1 q
N1 = 9, N2 = 3, Let N3 = 3
x1(n) = 1, n = 0 x 2 (n) = 1 , n = 3 x 3 (n) = 1, n = 6
= −1, n = 1 = −1, n = 4 = −1, n = 7
= 2, n = 2 = 2, n = 5 = 2, n = 8

Let, y1(n), y2(n) and y3(n) be output of linear convolution of x1(n), x2(n) and x3(n) with h(n) respectively.
Now each output will consists of 3 + 3 – 1 = 5 samples. Hence convert x1(n), x2(n), x3(n) and h(n) to 5 sample sequence as shown
below.
x1(n) = 1, n = 0 x 2 (n) = 1, n = 3 x 3 (n) = 1, n = 6
= −1, n = 1 = −1, n = 4 = −1, n = 7
= 2, n = 2 = 2, n = 5 = 2, n = 8
= 1, n = 3 = 1, n = 6 = 0, n = 9
= −1, n = 4 = −1, n = 7 = 0, n = 10

h(n) = {2, 3, –1, 0, 0}

Now perform circular convolution of each section with h(n).
x1(n) starts at, n = 0 ; \ y1(n) starts at, n = 0.
x2(n) starts at, n = 3 ; \ y2 (n) starts at, n = 3.
x3(n) starts at, n = 6 ; \ y3(n) starts at, n = 6.

Convolution of Section 1

m –4 –3 –2 –1 0 1 2 3 4

x1(m) 1 –1 2 1 –1

h(m) 2 3 –1 0 0

h(–m)= h0(m) 0 0 –1 3 2 0 0 –1 3

h1(m) 0 0 –1 3 2 0 0 –1

h2(m) 0 0 –1 3 2 0 0

h3(m) 0 0 –1 3 2 0

h4(m) 0 0 –1 3 2
E2. 26 DSP, Chapter 2 -Discrete Time Signals and Systems
N −1 4
y1(n) = x1(n) ∗ h(n) = ∑ x (m) h((n − m)) = ∑ x (m) h (m) ;
m=0
1 n
m=0
1 n where hn (m) = h((n − m))5

4
When n = 0 ; y1(0) = ∑ x (m) h (m) =
m=0
1 0 2 + 0 + 0 − 1 − 3 = −2

4
When n = 1 ; y1(1) = ∑ x (m) h (m) =
m=0
1 1 3 − 2 + 0 + 0 + 1= 2

4
When n = 2 ; y1(2) = ∑ x (m) h (m) =
1 2 − 1− 3 + 4 + 0 + 0 = 0
m=0

4
When n = 3 ; y1(3) = ∑ x (m) h (m) =
m=0
1 3 0 + 1+ 6 + 2 + 0 = 9

4
When n = 4 ; y1(4) = ∑ x (m) h (m) =
m=0
1 4 0 + 0 − 2 + 3 − 2 = −1

∴ y1(n) = l − 2, 2, 0, 9, − 1 q
A
n=0

Convolution of Section 2

The output of convolution of section -2 will be identical to that of section-1 except the starting value of n.

∴ y 2 (n) = l − 2, 2, 0, 9, − 1 q
A
n= 3

Convolution of Section 3

m –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 9 10

x3(m) 1 –1 2 0 0

h(m) 2 3 –1 0 0

h0(m) 0 0 –1 3 2

h6(m) 0 0 –1 3 2 0 0 –1 3

h7(m) 0 0 –1 3 2 0 0 –1

h8(m) 0 0 –1 3 2 0 0

h9(m) 0 0 –1 3 2 0

h10(m) 0 0 –1 3 2

10 10
y 3 (n) = x 3 (n) ∗ h(n) = ∑ x (m) h((n − m)) = ∑ x (m) h (m) ;
m=6
3 5
m=6
3 n where hn (m) = h((n − m))5

10
When n = 6 ; y 3 (6) = ∑ x (m) h (m) =
m=6
3 6 2+0+0+0+0 = 2

10
When n = 7 ; y 3 (7) = ∑ x (m) h (m) =
3 7 3−2+0+0+0=1
m=6

10
When n = 8 ; y 3 (8) = ∑ x (m) h (m) =
m=6
3 8 − 1− 3 + 4 + 0 + 0 = 0

10
When n = 9 ; y 3 (9) = ∑ x (m) h (m) =
3 9 0 + 1+ 6 + 0 + 0 = 7
m=6

10
When n = 10 ; y 3 (10) = ∑ x (m) h
m=6
3 10 (m) = 0 + 0 − 2 + 0 + 0 = −2

∴ y 3 (n) = l 2, 1, 0, 7, − 2 q
A n= 6
Solution for Exercise Problems E2. 27
Overall Output
n 0 1 2 3 4 5 6 7 8 9 10

y1(n) –2 2 0 9 –1

y2(n) –2 2 0 9 –1

y3(n) 2 1 0 7 –2

y(n) * * 0 9 –1 0 9 –1 0 7 –2

y(n) = {*, *, 0, 9, –1, 0, 9, –1, 0, 7, –2}

Hence both the results are same except the first (N2 – 1) samples.
E2.17. Perform crosscorrelation of the sequences,
l
x(n) = −1, 2, 3 − 4 q and l
y(n) = 2, − 1, − 3 q
A A
Solution
The crosscorrelation sequence rxy(m) is given by,
∞
rxy (m) = ∑ x(n) y(n − m)
n= −∞

The x(n) starts at n = 0, and has 4 samples.

\ n1 = 0, N1 = 4
The y(n) start at, n = –1 and has 3 samples.
\ n2 = –1 , N2 = 3

Now rxy(m) will have, N1 + N2 – 1 = 4 + 3 –1 = 6 samples.

The initial value of m = mi = n1 – (n2 + N2 – 1) = 0 – (–1 + 3 – 1) = –1

The final value of m = mf = mi + (N1 + N2 – 2) = –1 + (4 + 3 –2) = 4

The 6 samples of crosscorrelation sequence are computed using table method as shown below.

n –2 –1 0 1 2 3 4 5

x(n) –1 2 3 –4

y(n) 2 –1 –3

y(n–(–1))= y–1(n) 2 –1 –3

y(n–0)= y0(n) 2 –1 –3

y(n–1)= y1(n) 2 –1 –3

y(n–2)= y2(n) 2 –1 –3

y(n–3)= y3(n) 2 –1 –3

y(n–4)= y4(n) 2 –1 –3

Each sample of rxy(m) is given by,

+∞ +∞
rxy (m) = ∑ x(n) y(n − m) = ∑ x(n) y
n= −∞ n= −∞
m (n) ; where ym (n) = y(n − m)

3
When m = −1 ; rxy ( −1) = ∑ x(n) y
n= −2
−1(n) = 0+0+3+0+0+0 = 3

3
When m = 0 ; rxy (0) = ∑ x(n) y (n) =
n= −1
0 0 + 1 − 6 + 0 + 0 = −5

3
When m = 1 ; rxy (1) = ∑ x(n) y (n) =
n=0
1 − 2 − 2 − 9 + 0 = −13

3
When m = 2 ; rxy (2) = ∑ x(n) y (n) =
n=0
2 0 + 4 − 3 + 12 = 13
E2. 28 DSP, Chapter 2 -Discrete Time Signals and Systems
4
When m = 3 ; rxy (3) = ∑ x(n) y (n) =
n=0
3 0 + 0 + 6 + 4 + 0 = 10

5
When m = 4 ; rxy (4) = ∑ x(n) y
n=0
4 (n) = 0 + 0 + 0 − 8 + 0 + 0 = −8

∴ rxy (m) =
RS 3, −5, − 13, 13, 10, − 8
UV
T A W
E2.18. Determine the autocorrelation sequence for x(n) = {1, 4, 3, –5, 2}
-
Solution
The autocorrelation sequence rxx(m) is given by,
+∞
rxx (m) = ∑ x(n) x(n − m)
n= −∞

The x(n) starts at n = –1, and has 5 samples.

\ nx = –1, and N = 5
The rxx(m) will have, 2N –1 = 10 – 1 = 9 samples
The initial value of m = mi = – (N – 1) = – (5 – 1) = –4
The final value of m = mf = mi + (2 N – 2) = –4 + (10 – 2) = 4

n –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7

x(n) 1 4 3 –5 2

x–4(n) 1 4 3 –5 2

x–3(n) 1 4 3 –5 2

x–2(n) 1 4 3 –5 2

x–1(n) 1 4 3 –5 2

x0(n) 1 4 3 –5 2

x1(n) 1 4 3 –5 2

x2(n) 1 4 3 –5 2

x3(n) 1 4 3 –5 2

x4(n) 1 4 3 –5 2

Each sample of autocorrelation sequence rxx(m) is given by,

+∞ +∞
rxx (m) = ∑ x(n) x(n − m) = ∑ x(n) x
n= −∞ n= −∞
m (n) ; where xm (n) = x(n − m)

3
When m = −4 ; rxx ( −4) = ∑ x(n) x
n= −5
−4 (n) = 0+0+0+0+2+0+0+0+0 = 2

3
When m = −3 ; rxx ( −3) = ∑ x(n) x
n= −4
−3 (n) = 0+0+0− 5+8+0+0+0 = 3

3
When m = −2 ; rxx ( −2) = ∑ x(n) x
n= −3
−2 (n) = 0 + 0 + 3 − 20 + 6 + 0 + 0 = −11

3
When m = −1 ; rxx ( −1) = ∑ x(n) x
n= −2
−1(n) = 0 + 4 + 12 − 15 − 10 + 0 = −9

3
When m = 0 ; rxx (0) = ∑ x(n) x (n) =
n= −1
0 1+ 16 + 9 + 25 + 4 = 55

4
When m = 1 ; rxx (1) = ∑ x(n) x (n) =
n= −1
1 0 + 4 + 12 − 15 − 10 + 0 = −9

5
When m = 2 ; rxx (2) = ∑ x(n) x (n) =
n= −1
2 0 + 0 + 3 − 20 + 6 + 0 + 0 = −11
Solution for Exercise Problems E2. 29
6
When m = 3 ; rxx (3) = ∑ x(n) x (n) = 3 0+0+0− 5 + 8+0+0+0= 3
n= −1
7
When m = 4 ; rxx (3) = ∑ x(n) x (n) = 4 0+0+0+0+2+ 0+0+0+0= 2
n= −1

∴ rxx (m) = l 2, 3, − 11, − 9, 55, − 9, − 11, 3, 2 q

A
E2.19. Find the inverse system for the following discrete time system
n
y(n) = ∑ c p x(p − 2) ; for n ≥ 0.
p=0

Solution n
Given that, y(n) = ∑ c x(p − 2)
p=0
p
; for n ≥ 0

0
When n = 0 ; y(0) = ∑ c x(p − 2) = c x(−2) = x(−2)
p=0
p 0
⇒ x(−2) = y(0)

1
When n = 1 ; y(1) = ∑ c x(p − 2) = c x(−2) + c x(−1)
p=0
p 0 1

= x( −2) + c x( −1)
1
= y(0) + c x(−1) ⇒ x( −1) = y(1) − y(0)
c
2
When n = 2 ; y(2) = ∑ c x(p − 2) = c x(−2) + c x(−1) + c x(0)
p=0
p 0 1 2

= x(−2) + cx(−1) + c 2 x(0)

= y(0) + y(1) − y(0) + c 2 x(0)

1
= y(1) + c 2 x(0) ⇒ x(0) = y(2) − y(1)
c2
1
Therefore, in general, x(n) = n+2
y(n + 2) − y(n + 1) ; for n ≥ −1 with initial condition x( −2) = y(0).
c

E2.20. A discrete time system is excited by an input x(n), and the response is, y(n) = 4, 3, 6, 7.5, 3, 30, − 8 . If the l q
A
l q
impulse response of the system is h(n) = 2, 4, − 2 , then what will be the input to the system?
A
Solution
Let, N1 = Number of samples in x(n)

N2 = Number of samples in h(n)

N3 = Number of samples in y(n)

Now, N3 = N1 + N2 – 1 Þ N1 = N3 – N2 + 1 = 7 – 3 + 1 = 5 samples

Each sample of x(n) is given by,

1 LM n −1 O
x(n) = y(n) −
MN ∑ x(m) h(n − m)PP
h(0) m=0 Q
y(0) 4
When n = 0 ; x(0) = = =2
h(0) 2
LM
1 OP 1 1
When n = 1 ; x(1) =
MN ∑
h(0)
y(1) −
m= 0
x(m) h(1 − m) =
PQ h(0)
y(1) − x(0) h(1) =
2
3 − (2 × 4) = −2.5

1 L O 1 1
1
When n = 2 ; x(2) = M
h(0) MN
y(2) − ∑ x(m) h(2 − m)P =
PQ h(0) y(2) − x(0) h(2) − x(1) h(1) =
2
b g
6 − (2 × −2) − −2.5 × 4 = 10
m=0
E2. 30 DSP, Chapter 2 -Discrete Time Signals and Systems

1 LM 2 O= 1
When n = 3 ; x(3) = y(3) −
MN ∑ x(m) h(3 − m)PP y(3) − x(0) h(3) − x(1) h(2) − x(2) h(1)
h(0) m=0 Q h(0)
1
=
2
b g
7.5 − (2 × 0) − −2.5 × −2 − (10 × 4) = −18 . 75

1 LM 2 O= 1
When n = 4 ; x(4) = y(4) −
MN ∑ x(m) h(4 − m)PP y(4) − x(0) h(4) − x(1) h(3) − x(2) h(2) − x(3) h(1)
h(0) m=0 Q h(0)
1
=
2
b g
3 − (2 × 0) − −2.5 × 0 − (10 × −2) − (−18.75 × 4) = 49

l q l
∴ x(n) = x(0), x(1), x(2), x(3), x(4) = 2, − 2.5, 10, − 18.75, 49 q
A
E2.21. Perform circular correlation of the sequences, x(n) = −1, 1, 2, 6 l q and l q
y(n) = 4, − 2, − 1, 2

Solution

Let rxy (m) be the sequence obtained by circular correlation of x(n) and y(n).

The circular correlation is given by,

N −1 N −1
rxy (m) = ∑ x(n) y((n − m)) = ∑ x(n) y
n= 0
N
n= 0
m (n), where y m (m) = y((n − m))N

Here, N = 4, The circular correlation is performed by table method as shown below.

q 0 1 2 3 4 5 6 7

x(n) –1 1 2 6

y(n) 4 –2 –1 2

y0(n) 4 –2 –1 2 4 –2 –1 2

y1(n) 2 4 –2 –1 2 4 –2 –1

y2(n) –1 2 4 –2 –1 2 4 –2

y3(n) –2 –1 2 4 –2 –1 2 4

3
When m = 0 ; rxy (0) = ∑ x(n) y (n) = −4 − 2 − 2 + 12 = 4
0
n= 0

3
When m = 1 ; rxy (1) = ∑ x(n) y (n) = −2 + 4 − 4 − 6 = −8
1
n=0

3
When m = 2 ; rxy (2) = ∑ x(n) y (n) = 1+ 2 + 8 − 12 = −1
2
n= 0

3
When m = 3 ; rxy (3) = ∑ x(n) y (n) = 2 − 1+ 4 + 24 = 29
3
n= 0

l
∴ rxy (m) = 4, − 8, − 1, 29 q