DISCRETE-TIME

CONVOLUTION
Dr. Muhammad
Zaigham Abbas Shah
Response of LTI systems
Review: Discrete-time signals
 A discrete-time signal is a set of
numbers, one can
x=[2 0 -1 3]
Review: Resolution of a DT Signal into pulses
x = [2 0 -1 3]

Impulses at n = 0, 1, 2, and 3 with

amplitudes
x[0] = 2, x[1] = 0, x[2] = -1, x[3] = 3

This can be written as,

xn 2 n   [n  2]  3 [n  3]
xn  x[0] [n  0]  x[1] [n  1]  x[2] [n  2]  x[3] [n  3]
K1
xn  x[k ] [n  k ] K is the length of
k 0 x

xn   x[k ] [n  k ] For infinite pulses
k 
Example 1: Resolve the following discrete-time signals into impulses
x[n] 2 4 0 3

Impulses occur at n = -1, 0, 1, 2 with amplitudes x[-1] = 2,

x[0] = 4, x[1] = 0, x[2] = 3

2
xn   x[k ] [n  k ]
k  1

 x[ 1] [n  ( 1)]  x[0] [n  0]  x[1] [n  1]  x[2] [n  2]

x[n] 2 [n  1]  4 [n]  3 [n  2]

Follow the same procedure for r[n]

Weighted sum of time-shifted impulses
 Characterization of LTI systems
 LTI systems can be characterized in two ways

 Using Difference equations

 Relationship between discrete-time inputs and
discrete-time outputs
 Also called Input-Output equations
y[n]  x[n]  34 x[n  1]  2 x[n  5]

y[n]  x[n]  34 x[n  1]  2 x[n  5]  1

11 y[n  1]  75 y[n  4]
Characterization of LTI systems
 Impulse response
 System’s response to an impulse
 Decompose the input signal vector into weighted-time-
shifted impulses
 Find the output of the system as the sum of its impulse
response

x[n] [a0 a1 a2 ] 

 x[k ] h[ n  k ]
a0 [n  0]  a1 [n  1]  a2 [n  2]   Syste k 

m
H(.)
Convolution
 Convolution is the process by which an input
interacts with an LTI system to produce an output

 Convolution between of an input signal x[n] with a

system having impulse response h[n] is given as,


y[ n]  x[ n] * h[ n]   x[ k ] h[ n  k ]
k 

where * denotes the convolution

Convolution
We have already established that we can resolve the
discrete-time input as weighted, time-shifted impulses

Let’s generalize this 

x[ n]   x[ k ] [ n  k ]
k 

Now, we apply this signal to an LTI system ‘H’ to get an

output ‘y’   
y[ n] H 
x[ k ]  [ n  k ] 
where h[n] is the response of
k   the system H to each impulse

y[ n]   x[ k ] h[ n  k ]
k  Flip, Shift, Multiply, Add


y[ n]  x[ n] * h[ n]   x[ k ] h[ n  k ]
k 
Ways to compute D.T. Convolution
 Finite State Machine Algorithm (FSMA)/Table
method
 Flip, Shift, Multiply, Add

 Analytical method
FSMA/Table method
Steps to follow:
Step 1 List the index ‘k’ covering a sufficient range

Step 2 List the input x[k]

Step 3 Obtain the reversed sequence h[-k], and align the

rightmost element of h[n-k] to the leftmost element
of x[k]
Step 4 Cross-multiply and sum the nonzero overlap terms
to produce y[n]
Step 5 Slide h[n-k] to the right by one position

Step 6 Repeat step 4; stop if all the output values are zero
or if required.
Example 2: Find the convolution of the two sequences x[n] and h[n]
given by,
x[k] = [3 1 2] h[k] = [3 2 1]

k: -2 -1 0 1 2 3 4 5

x[k]: 3 1 2

h[0-k]: 1 2 3

h[1-k]: 1 2 3

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

Hint: The value of k starts from (– length of h + 1) and continues till (length of h + length of x –
1)
k: -2 -1 0 1 2 3 4 5

x[k]: 3 1 2

h[0-k]: 1 2 3

h[1-k]: 1 2 3

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

y[0] = 3 x 3
=9
k: -2 -1 0 1 2 3 4 5

x[k]: 3 1 2

h[0-k]: 1 2 3

h[1-k]: 1 2 3

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

y[0] = 3 x 3
=9
y[1] = 3 x 2 + 1 x 3
=9
k: -2 -1 0 1 2 3 4 5

x[k]: 3 1 2

h[0-k]: 1 2 3

h[1-k]: 1 2 3

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

y[0] = 3 x 3
=9
y[1] = 3 x 2 + 3 x 1
=9
y[2] = 3 x 1 + 1 x 2 + 2 x
3= 11
k: -2 -1 0 1 2 3 4 5

x[k]: 3 1 2

h[0-k]: 1 2 3

h[1-k]: 1 2 3

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

y[0] = 3 x 3 y[3] = 1 x 1 + 2 x 2 = 5
=9
y[1] = 3 x 2 + 3 x 1
=9
y[2] = 3 x 1 + 1 x 2 + 2 x
3= 11
k: -2 -1 0 1 2 3 4 5

x[k]: 3 1 2

h[0-k]: 1 2 3

h[1-k]: 1 2 3

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

y[0] = 3 x 3 y[3] = 1 x 1 + 2 x 2 = 5
=9
y[1] = 3 x 2 + 3 x 1 y[4] = 2 x 1 = 2
=9
y[2] = 3 x 1 + 1 x 2 + 2 x
3= 11
k: -2 -1 0 1 2 3 4 5

x[k]: 0 0 3 1 2

h[-k]: 1 2 3 0 0

h[1-k]: 0 1 2 3 0

h[2-k]: 1 2 3

h[3-k]: 1 2 3

h[4-k]: 1 2 3

h[5-k]: 1 2 3

y[0] = 3 x 3 y[3] = 1 x 1 + 2 x 2 = 5
=9
y[1] = 3 x 2 + 3 x 1 y[4] = 2 x 1 = 2
=9
y[2] = 3 x 1 + 1 x 2 + 2 x y[5] = 0 (no overlap)
3= 11
y[n] {9 9 11 5 2 0}
Example 3: Find the convolution of the two
sequences, x[n] and h[n], represented by,

x[n] = [1 2 3] h[n] = [10 20 30 40

50]
Example 4: Find the convolution of the two
sequences x[n] and h[n] represented by,

x[n] = {2 1 -2 3 -4} h[n] = [3 1 2 1

4]
Properties of Convolution

Commutative…
x1[n] * x2 [n]  x2 [n] * x1[n]

Associative…
{x1[n] * x2 [n]} * x3[n]  x1[n] *{x2 [n] * x3 [n]}

Distributive…

{x1[n]  x2 [n]} * x3 [n]  x1[n] * x3 [n]  x2 [n] * x3 [n]

And……we’re done!!