Digital Signal Processing

Lecture 4-5

-- Impulse Response of an LTI system

-- Convolution Sum
-- Properties of Convolution

1
 Analysis of Discrete Time Linear Time
Invariant Systems
There are two basic methods for analyzing
the behavior or response of a linear system
to a given input signal.

• Method based on direct input-output

equation of a system.

• Decomposition of the input signal into a

sum of elementary signals (usually
impulses). 2
 Impulse Response
• The impulse response is the output of an LTI system due to
an impulse input applied at time t = 0 or n = 0.
• The impulse response completely characterizes the behavior
of any LTI system.
• If the input to a linear system is expressed as a weighted
superposition of time-shifted impulses, then the output is
the weighted superposition of the response to each time
shifted impulse.
• If the system is also time invariant, then the system response
to a time shifted impulse is a time shifted version of the
system response to an impulse.
• Thus the output of an LTI system is a weighted
superposition of time shifted impulse responses. This
weighted superposition is termed as the Convolution Sum
for discrete time systems and Convolution integral for
3
continuous time systems.
 The Convolution Sum
Consider the product of a signal x[n] and the impulse
sequence [n], written as
x[n][n] = x[0][n], x[n][n-1] = x[1][n-1]
Generalize this relationship to the product of x[n] and
a time shifted impulse sequence to obtain
x[n][n-k] = x[k][n-k]
This property allows us to express x[n] as the
following weighted sum of time shifted impulses:
x[n] = ….. + x[-2][n+2] +x[-1][n+1]+x[0][n]+
x[1][n-1]+x[2][n-2]+….

x[n ]   x[k ]n  k  (1)
k  
4
 The Convolution Sum (cont.)
 Resolution of a DT signal into pulses

5
 The Convolution Sum (cont.)
Let the operator H denote the system to which the
input x[n] is applied. Thus using (1) to represent
the input x[n] to the system results in the output
  
y[n]  H   x[k ][n  k ]
k   

Using the linearity property, we write


y[n]   x[k ]Hn  k 
k  
(2)
or

y[n]   x[k ]h[n  k ]
k  
(3)

6
where h[n-k] = H{[n-k]}.
The sum in (3) is termed the Convolution Sum and
is denoted by the symbol *, that is,

(4)

7
 Steps to compute Convolution Sum
1- Folding: fold h[k] about k=0, to obtain h[-k].

2- Shifting: shift h[-k] by n samples to obtain h[n-k].

3-Multiplication: multiply x[k] by h[n-k] to obtain

product sequence x[k]h[n-k].

4- Summation: sum all the values of the product sequence

to obtain value of output at time n.
• The formula that gives the response y(n) of LTI system
as a function of the input signal x[n] and the unit
sample response h[n] is called convolution sum.
This sum of products is in fact a function of n that
represents the overlap between x[n] and the time-reversed
and shifted version of h[n]. 8
 Convolution Sum
• Example 1:
The impulse response of a linear time invariant
system is h[n] = {1, 2, 1, -1}

Determine the response of the system to the input

signal x[n] = {1, 2, 3, 1}

Solution: The first step in the computation of

convolution is to fold h[k]. Complete solution is
given in the next slide.

9
 -1 1 2 1
-1 1 2 1
h[-k] -1 1 2 1
x[n] 0 1 2 3 1
-1 1 2 1
-1 1 2 1
-1 1 2 1
-1 1 2 -1
1

x[n]*h[n] = [0, 1, 4, 8, 8, 3, -2, -1] 10

Example 2: Assume that an LTI system has impulse
response
h[n] = {1 2 1}

Determine the output of this system in response to the input

x[n] = {2 3 -2}
x[n]*h[n] = {0, 2, 7, 6, -1, -2}
Solution:
1 2 1
1 2 1
h[-k] 1 2 1
x[k] 0 2 3 -2
1 2 1
1 2 1
1 2 1 11
Another way to get convolution sum.

12
13
Example 3: Determine the output y[n] of a linear
time invariant system with impulse response h[n] =
anu[n], |a| < 1, when the input is a unit step
sequence, that is x[n] = u[n].
Solution:
 

 x[k ]hn  k    a
nk
x[n] * h[n]  u[ n  k ]
k   k 0

 
a n
 (a)
k 0
k
 a n  (1 / a ) k
k 0

n n 1
n 2
 a [1  1 / a  1 / a  ] 
a a

1 1/ a a 1
14
Example 4:
Determine the impulse response for the cascade
of two linear time-invariant systems having
impulse responses h1[n]=(1/2)nu[n] and
h2[n] = (1/4)nu[n].
Solution:
k n k

1 n
1
   
h1 n * h 2 n   h1 [k ]h 2 [n  k ]      
k 0 k 0  2   4 

n n n
1
 
4

k 0
2 k

1
4

  1  2  2 2  ...... 
1
n
 2 n 1  1   1  n   1 
n
    2    
4  2  1   2    2  
   15
Tutorial 5:
Find the convolution of the two sequences x[n] and
h[n] given by,
1- x[n] = [3 1 2] h[n] = [3 2 1]

2- x[n] = [1 2 4] h[n] = [1 1 1 1 1]

3- x[n] = {0 1 -2 3 -4} h[n] = [0.5 1 2 1 0.5]

4- x[n] = [1, 1, 0, 1, 1], h[n] = [1, -2, -3, 4]

5- h[n] = (1/2)nu[n], x[n] = (1/4)nu[n]

16
 Properties of Convolution
• Commutative Law
x[n]*h[n] = h[n]*x[n]
• Associative Law
[x[n]*h1[n]]*h2[n] = x[n]*[h1[n]*h2[n]]
• Distributive Law
x[n]*[h1[n] + h2[n]] = x[n]*h1[n] + x[n]*h2[n]

17
 Infinite Geometric Series
Formulas
• a/(1-r) when r<1; k(sum limit) starts from 0

• arm/(1-r) when sum does not start from 0, i-e

k=m to infinity.

• a(1-rn)/(1-r) when r>1. or a(rn+1-1)/(r-1)

18