7

Convolution

In Fourier transform, Laplace transform and z-transform we

have already introduced convolution theorems as a part of
their properties. The actual significance of convolution
theorem as well as its use in digital signal processing have not
been introduced. The aim of this chapter is to introduce
convolution in detail.

7.1 CONVOLUTION THEOREM FOR CONTINUOUS SYSTEM

If the impulse response of an analog or continuous time

system be h(t), it means that for an input δ (t) its output
is h (t). Hence for linear system its output for a shifted
impulse δ (t –T) is h (t –T). The following are the
mathematical representation of these operations:

δ(t) h(t)
δ (t −T) h (t – T)
x(t) δ (t − T) x(T)h(t − T)

Obviously,
From last two statements, we can say that for an
input x(T)δ (t–T) the output is x(T)h(t–T) and for an input of
continuous summation or integration of x(T) δ (t–T), the
output is also continuous summation of x(T) h(t–T).
Therefore, we can write by the linearity of the system

The right-hand side of Eq. (7.1) is defined as convolution

of x(t) with h(t) and hence we can say that the output y(t) is
the convolution of h(t) and x(t). y(t) = Convolution of h(t)
and x(t)

The symbol ‘*’ in Eq. (7.2) stands for convolution.

7.2 PROPERTIES OF LINEAR CONVOLUTION

The linear convolution is the convolution process for the continuous time
systems. The convolution of continuous time (CT) signals possesses the
following properties:

Commutative Property: x1(t)*x2(t) = x2(t)*x1(t)

Associative Property : x1(t)*[x2(t) *x3(t)] = [x1(t)*x2(t)] *x3(t)

Distributive Property : x1(t)*[x2(t) + x3(t)] = x1(t)*x2(t) + x1(t)*x3(t)

If any function is convolved with impulse function d (t), it results the
function itself. Again, convolution of any function with δ (t – t0), it results
the shifted function. If the original function is impulse, this is also true.
Mathematically, we can write

x(t)* δ (t) = x(t)

x(t)* δ (t – t0) = x(t – t0)
δ (t)* δ (t) = δ (t)
δ (t)* δ (t – t0) = δ (t – t0)
δ (t – t1)* δ (t – t2) = δ (t – t1– t2)

7.3 GRAPHICAL CONVOLUTION

In this article graphical convolution is explained here. Let us

consider two sided signal. Let us consider a signal shown
in Fig. 7.1. let x (t) and x (t) are identical to x(t). Therefore,
1 2

Fig. 7.1 Signal x(t)

for a given t.
The meaning of the words ‘graphically process’ is to
find x (λ), x (t – λ) and to multiply these two. The
1 2

multiplication of these two gives the area under the function

for a given ‘t’. From Fig. 7.2, x (λ) is nothing but x(t) with the
1

name of x-axis changed from t to λ. In Fig. 7.3, x (λ) is the

2

function x(t) and x (–λ) is the mirror image of x (λ) shown

2 2

in Fig. 7.4. In Fig. 7.5(b) x (–λ) is advanced to the right by an

2

amount ‘t’ giving x (t– λ) for positive ‘t’ and Fig. 7.5(c) shows
2

the plot of multiplication of x (λ) and x (t– λ).

1 2

Fig. 7.2 Plot of x1(l)

Fig. 7.3 Plot of x (λ)

2
Fig. 7.4 Plot of x (λ)
2

Fig. 7.5 Product of x (λ) and x (t – λ)

1 2
The area under x (λ) and x (t – λ) is given by
1 2

This will continue upto t = 2T.

For t > 2T:
The area under x (λ) and x (t– λ) will be zero because x (t– λ)
1 2 2

will go out of the range of x (λ).1

Therefore, for positive ‘t’, the convolution is given by

y(t) = x (t)*x (t) = A (2T – t)

1 2
2 (7.4)

For negative t, this process gives the same result except

that x (–λ) in Fig. 7.4 will be shifted to the left only.
2

y(t) =A (2T + t)
2 (7.5)

where t has been changed by –t in Eq. (7.4).

Therefore from Eqs. (7.4) and (7.5) we can write that

Compactly, we can write that

Fig. 7.6 gives the plot of y(t) = x (t)*x (t) = x(t)*x(t)

1 2

Fig. 7.6 Plot of y(t) = x (t) * x (t)

1 2

Example 7.1 Plots of x (t) and x (t) are shown below.

1 2

Convolve x (t) and x (t).

1 2
Fig. E7.1
Solution
Figs. E7.2(a) and E7.2(b) show plots of x (l) and x (–l)
1 2

respectively.

Fig. E7.2

x (–λ) will be shifted towards right for positive values of ‘t’.

2

For 0 < t < 2:

Here t lies between 0 and 2 seconds and the product x (λ)x (t –
1 2

λ) gives a rectangle of height 1.5 and width 1 second.

Therefore, the area under multiplication is 3 and hence
convolution. These have been shown in Fig.
E7.3(a), (b) and (c).

Fig. E7.3

We can write y(t) = 1.5 (1)

For 2 < t < 3:
Here t lies between 2 and 3 seconds. For t beyond 2 seconds, a
part of x (t – λ) will go out of range of x (λ). The part which
2 1

remains under x (λ) has the width of [2– (–1 + t)] = (3 – t).
1

For t = 3, no part of x (t – λ) will remain under x (λ).

2 1

We can write y(t) = 1.5(3 – t) (2)

In Figs. E7.1, E7.2, E7.3, E7.4, and E7.5, in the proper sealing
of x (t) and x (t). The heights of x (t) and x (t) should match in
1 2 1 2

all laser.
Fig. E7.4

Fig. E7.5
For t > 3 second:
x (t – λ) will go outside the range of x (λ). Therefore, the
2 1

convolution becomes zero.

We can write y(t) = 0 (3)

Therefore, we can write

Fig. E7.6 shows the plot for positive values of t.
Fig. E7.6

Repeating the process for negative values of t for –1 < t < 0,

For –2 < t < –1 and for–∞ < t < –2. All these have been shown
in Fig. E7.7.
Fig. E7.7
Therefore, we can write
Fig. E7.8 shows the plot for negative values of t.

Fig. E7.8

Fig. E7.9 shows the plot of y(t) for whole range of t which is
asymmetrical.

Fig. E7.9
 7.4 DISCRETE CONVOLUTION

If the input x(n) is given to discrete time system having an

impulse {h(n)}, the output is given by convolving the {h(n)}
with x(n).

y(n) = h(n)* x(n) = x(n)* h(n) (7.8)

where the symbol ‘*’ represents the convolution of x(n)

with h(n) and this process is also called convolution sum. It is
defined by

The limits will reduce to 0 to n for one-sided x(n) and h(n).

Proof:
Fig. 7.7 shows an input sequence {x(n)}.
Fig. 7.7 Input sequence {x(n)}

Let us express x(n) as a sum of delta sequences. We can write

We know that the output becomes h(n) if input is δ(n).

Mathematically, we can express this as follows:

Therefore, we can write that

For two-sided sequences h(n) and x(n) we can write that

Hence we can obtain any system response by simply
performing multiplications and additions without using z-
transform.

7.4.1 Properties of Discrete Convolution

In discrete convolution h(n) and x(n) are interchangeable, we

can write

Discrete convolution is commutative, distributive and

associative.
Commutative property: x (n)*x (n) = x (n)*x (n)
1 2 2 1

Associative property : x (n)*[x (n) *x (n)] = [x (n)*x (n)] *x (n)

1 2 3 1 2 3

Distributive property : x (n)*[x (n) + x (n)] = x (n)*x (n)

1 2 3 1 2

+ x (n)*x (n)
1 3

7.5 IMPORTANT PROPERTIES OF SYSTEMS

Figure 7.8 shows the two LTI systems having common

input x(n). Their outputs are y (n) and y (n) respectively
1 2

which are added to get the net output y(n). The systems are in
parallel mode.
Fig. 7.8 Replacement of parallel systems

y(n) = x(n)*h (n) + x(n) *h (n) = x(n)*[h (n) + h (n)]

1 2 1 2

Therefore, the parallel system can be replaced by a single

system with impulse response [h (n) + h (n)].
1 2

Figure 7.9 shows the two systems which are in cascade i.e., in
series.

Fig. 7.9 Replacement of cascade systems

From Fig. 7.9, we get y(n) = x(n)*h (n) *h (n). Since the
1 2

convolution is commutative, it is possible to replace the

systems in cascade by a single system having impulse
response given by h (n)*h (n).
1 2

The following cases arise for convolution of discrete sequence:

 The sequences x(n) and y(n) both are one-sided and of finite
length.
For example

 The sequences x(n) and y(n) are of finite length but one or both
sequences may be one-sided and two-sided.
For example

 One sequence is of finite length whereas other is of infinite length

(one-sided or two-sided). For example
  Both the sequences are of finite length (one-sided or two-sided).
For example,

It is possible to find the convolution in such diverse cases.

7.6 CONVOLUTION THEOREM IN Z-TRANSFORM

If and Then

We can prove that

Proof:
Let us consider one-sided sequence x (n) having zw–
1

transform X (z) i.e.,

1

Similarly,
Therefore,

Let us collect the information of x (n) for getting the inverse

3

transform of X (z) X (z).

1 2

and so on.
This gives the general standard form

We get the similar result for two-sided sequences.

Therefore, IZT of

7.7 CIRCULAR OR PERIODIC CONVOLUTION

If we convolve two finite sequences x(n) with y(n), we can determine the
convolution sum. If x(n) and y(n) are periodic, the sequence x(n) and y(n–
k) is of infinite length and hence the sum does not converge. Therefore, we
get the infinite convolution which is absolutely meaningless. It is possible
in such cases that for over one period. the convolution result gives the
similar result of normal convolution. Figure 7.10 shows the sequences x(n)
and y(n).

Fig. 7.10 x(n) and y(n)

We can write for one period of y(n)

Therefore, the value of y(0) which slips out the period enters at the left and
hence the number of terms remains the same during this period. Therefore,
we can say that over one period of N terms the convolution can be obtained
as a circular one if and only if the sequences are of same length. This
convolution is denoted by N. This sum is useful for periodic sequences and
results a periodic sequences.

Example 7.2 Two periodic signals with period is given by

Determine the convolution between them using graphical convolution and

matrix convolution.

Solution

Graphical Method: x1(n) is placed on the four points of a circle in anti-

clockwise direction to represent x1(k) with x1(0) at the bottom.
Similarly, x2(n) is arranged on four points of a circle clockwise direction to
represent x2(–k).

The corresponding numbers on two circles are multiplied and the sum of
the products gives x3(n) for n= 0.
Fig. E7.10

Next x2(–k) on the circle is rotated by ‘1’ unit in positive direction to

represent x2(1 – k). The numbers on x2(1 – k) is multiplied with
corresponding numbers of x1(k) and the sum of the products gives x3(n)
for n = 1. The above said process is repeated for n = N – 1 = 4.

For n = 0: x3(0) = 1 × 3 + 2 × 3 + 1 × 3 + 1 × 4 = 16

For n = 1: x3(1) = 1 × 1 + 2 × 3 + 3 × 3 + 4 × 1 = 20

For n = 2: x3(2) = 1 × 1 + 2 × 1 + 3 × 3 + 4 × 3 = 24

For n = 3: x3(2) = 1 × 3 + 2 × 1 + 3 × 1 + 4 × 3 = 20

Matrix Method: Let us first arrange x1(–k) in the first row of a matrix
starting with x1(0) followed by x1(3), x1(2) and x1(1). To form the next row,
the first row is shifted to the right circularly so that the term which drops
off to the right reappears on the left.

x(k) is arranged in vertical column matrix with x2(0) at the top and x2(N – 1)
at the bottom. The product of the two matrices gives x3(n) for n = 0, 1, 2, 3.
Example 7.3 Perform the linear and circular convolution of the two
sequences x1(n) and x2(n) when

Solution

Tabular Method: For linear convolution the tabular method is used.

y(0) = 3

y(1) = 1 + 6 = 7

y(2) = 4 + 2 + 9 = 15

y(3) = 2 + 8 + 3 + 12 = 25

y(4) = 4 + 12 + 4 = 20

y(5) = 6 + 16 = 22

y(6) = 8

y(7) = 0

y(8) = 0 etc.
Fig. E7.11

Matrix Method: For circular convolution we use the matrix method: 14

323 21 23 25

ADDITIONAL SOLVED EXAMPLES

Example 7.4 Convolve and

Solution
Let
which gives
z(0) = x(0)y(0)
z(1) = x(0)y(1) + x(1)y(0)
z(2) = x(0)y(2) + x(1)y(1) + x(2)y(0)
z(3) = x(0)y(3) + x(1)y(2) + x(2)y(1) + x (3)y(0)
Here
x(0) = 1, x(1) = 2, x (2) = 3 and x (3) = 4 and
y(0) = 1, y (1) = –2, y (2) = 0. y (3) = 0
z(0) = x (0)y(0) = 1 × 1 = 1
z(1) = x (0)y(1) + x (1)y(0) = 1 × (–2) + 2 × (1) = 0
z(2) = x (0)y(2) + x (1)y(1) + x (2)y(0) = 1 × 0 + 2 × (–2) + 3 ×
1 = –1
z(3) = x (0)y(3) + x (1)y(2) + x (2) y (1) + x(3)y (0) = 1 × 0 + 2
× 0 + 3 × (–2) + 4 × 1 = –2
z(4) = x (0)y (4) + x (1)y(3) + x(2)y(2)+ x(3)y(1) + x(4)y (0)
= 1 × 0 + 2 × 0 + 3 × 0 + 4 × (–2) + 0 × 1 = –8
z(5) = x(0)y (5) + x(1)y (4) + x(2)y (3)+ x(3)y (2) + x(4) y (1)
+ x(5) y(0)
= 1 × 0 + 2 × 0 + 3 × 0 + 4 × 0 + 0 × (–2) + 0 × 1 = 0
Therefore, we can conclude that if one sequence has m terms
and other has n terms, the convolved sequence will have
(m + n – 1) terms.

Example 7.5 Convolve and ,using

graphical convolution.
Solution
Step 1: Plot x (k) by replacing n with k (changing the name of
variable n to k) referred to Fig. E7.12(a).
Fig. E7.12

Step 2: Plot y (k) by replacing n with k (changing the name of

variable n to k) referred to Fig. E7.12(b).
Step 3: Plot y(– k) which is mirror image y (k) referred to Fig.
E7.12 (c).
Step 4: Plot y (1–k) by advancing y (k) to the right by one step
referred to Fig. E7.12(d). Similarly, for y(2 – k) and so on.
Step 5:
For n = 0. Let us multiply elements x(k) with y(–k) and the
sum gives z(0).
z(0) = 1 × 4 = 4
For n = 1, let us multiply elements x(k) with y(1 – k) and the
sum gives z(1).
z(1) = 1 × 1 + 4 × 3 + 2 × 0 = 13
For n = 2, let us multiply elements x(k) with y(2 – k) and the
sum gives
z(2). z(2) = 1 × 2 + 3 × 1 + 4 × 2 = 13
For n = 3, let us multiply elements x(k) with y (3 – k) and the
sum gives z(3).
z(3) = 3 × 2 + 2 × 1 = 8
For n = 4, let us multiply elements x(k) with y (4 – k) and the
sum gives z(4).
z(4) = 1 × 0 + 3 × 0 + 2 × 2 = 4
For n = 5, it will go out of range x (k) and hence z(5) = 0.

Tabular Method:
Let us prepare a table putting the elements of {x(n)}
horizontally and that of {y (n)} vertically shown in Fig. E7.13.
The elements of each box is the multiplication of the elements
of {x(n)} and {y(n)}.

Fig. E7.13
Fig. E7.14

Example 7.6 Find the convolution of

and using graphical convolution
Solution
Fig. E7.15

SIGNIFICANT POINTS

The convolution of x(t) with h(t) is defined by

Properties of linear convolution:
Commutative Property: x (t)*x (t) = x (t)*x (t)
1 2 2 1

Associative Property: x (t)*[x (t) *x (t)] = [x (t)*x (t)] *x (t)

1 2 3 1 2 3

Distributive Property: x (t)*[x (t) + x (t)] = x (t)*x (t) + x (t)

1 2 3 1 2 1

*x (t)
3

Discrete convolution:
If the input x(n) is given to discrete time system having an
impulse {h(n)}, the output is given by convolving the {h(n)}
with x(n).

y(n) = h(n)*x(n) = x(n)*h(n)

where the symbol ‘*’ represents the convolution of x(n)

with h(n) and this process is also called convolution sum. It is
defined by

The limits will reduce to 0 to n for one sided x(n) and h(n).
Properties of discrete convolution:
Commutative property: x (n)*x (n) = x (n)*x (n)
1 2 2 1
Associative property: x (n)*[x (n) *x (n)] = [x (n)*x (n)] *x (n)
1 2 3 1 2 3

Distributive property: x (n)*[x (n) + x (n)] = x (n)*x (n)

1 2 3 1 2

+ x (n)*x (n)
1 3

SHORT QUESTIONS AND ANSWERS

Q1. Define the convolution of x(t) with h(t).

Ans.: The convolution of x(t) with h(t) is defined by

Hence the output y(t) is the convolution of h(t) and x(t).

∴ (y) = Convolution of h(t)

and x(t) =x(t) h(t) =

*

The symbol ‘*’ in Eq. (7.2) stands for convolution.

Q2. State the properties of linear convolution.

Ans.: The linear convolution possesses the following
properties:
Commutative property: x (T)*x (t) = x (t)*x (t)
1 2 2 1

Associative property: x (t)*[x (t) *x (t)] = [x (t)*x (t)] *x (t)

1 2 3 1 2 3
Distributive property: x (t)*[x (t) + x (t)] = x (t)*x (t)
1 2 3 1 2

+ x (t)*x (t)
1 3

Q3. What do you mean by discrete convolution?

Ans.: If the input x(n) is given to discrete time system having
an impulse {h(n)}, the output is given by convolving the
{h(n)} with x(n).

y(n) = h(n)*x(n) = x(n)*h(n)

where the symbol ‘*’ represents the convolution of x(n)

with h(n) and this process is also called convolution sum. It is
defined by

The limits will reduce to 0 to n for one sided x(n) and h(n).

Q4. State the properties of discrete convolution.

Ans.: The linear convolution possesses the following
properties:
Commutative property: x (n)*x (n) = x (n)*x (n)
1 2 2 1

Associative property: x (n)*[x (n) *x (n)] = [x (n)*x (n)]

1 2 3 1 2

*x (n)
3

Distributive property: x (n)*[x (n) + x (n)] = x (n)*x (n)

1 2 3 1 2

+ x (n)*x (n)
1 3

EXERCISES

Q1. The plots of x (t) and x (t) are shown in Fig. 1 and Fig.
1 2

2 respectively. Convolve x (t) and x (t).

1 2

Fig. 1
Fig. 2

[Ans. ]

Q2. Find the convolution between x (n) and h(n) = y(n)

where x(n) is given below:

1.
2.
[Ans. (i) y(n) = u(n) + u(n – 1) + 2u (n – 2) + 5u(n – 3)
(ii) y (n) = … + x(–3) + u(n + 3) + x(–2) u(n + 2) + x(–
1) u(n + 1) + x(0) u(n) + x (1) u(n – 1) + x(2) u(n – 2) +
………]

Q3. Two periodic signals x (n) and x (n) with period four given
1 2

below:
Determine the circular convolution between them i.e., x (n) 3

= x (n)*x (n) using (i) graphical method and ii) matrix

1 2

method.

[Ans.: ]

Q4. Find the linear convolution of the following two

sequences:

[Ans.: ]

MULTIPLE CHOICE QUESTIONS

1. If y(t) = x1(t)*x2(t), find the correct expression from the following:

1.
2.

3.

4.
2. Find the correct expression:

0. x(t)* δ (t– t0) = x(t– t0)

1. x(t)* δ (t– t0) = 1
2. x(t)* δ (t– t0) = x(t0)
3. x(t)* δ (t– t0) = x(t)

3. Find the correct expression:

0. δ (t– t1) * δ (t– t2) = δ (t– t– t2)

 1. δ (t– t1) * δ (t– t2) = 0
2. δ (t– t1) * δ (t– t2) = δ (t)
3. δ (t– t1) * δ (t– t2) = δ (t– t2)

4. If the sequences h(n) and x(n) are two-sided, then

0.

1.

2.

3.
5. Find the correct representation for the following:

0. x1 (n)*x2(n) = x2(n)*x1(n)
1. x1 (n)*[x2(n) * x3(n)] = [x1(n)*x2(n)]*x3(n)
2. x1 (n)*[x2(n) + x3(n)] = x1(n)*x2(n) + x1(n)] + x3(n)
3. all are valid
4. only (a) and (c) are valid

Answers

1. (a) 2. (a) 3. (a) 4. (a) 5. (d)