KING SAUD UNIVERSITY

ELECTRICAL ENGINEERING DEPARTMENT.

EE 301: Signals and Systems

Chapter 3:
LTI Systems and Convolution

Dr A. I. Sulyman
Assistant Professor
EE Department
King Saud University
Office: College of Eng. 2C 12/1 (Tel: x70586)
***********
********
Course Instructor: Dr A. I. Sulyman

Outline

1. Introduction

2. DT convolution, examples

3. The convolution integral, examples

4. Properties

2
Course Instructor: Dr A. I. Sulyman

1 - Introduction
x n 

n
 2 1 0 1 2 

3.1 Representation of DT Signals in terms of Impulse Sequence

Consider the DT signal x[n] shown. The components of x[n] at each DT time instant “k”

can be written as

 x(2) n  2
x(n) (n  2)  x(2) (n  2)  
0 n  2

 x(1) n  1
x(n) (n  1)  x(1) (n  1)  
0 n  1

 x(0) n  0
x(n) (n)  x(0) (n)  
0 n0

 x(1) n  1
x(n) (n  1)  x(1) (n  1)  
0 n 1

 x(2) n  2
x(n) (n  2)  x(2) (n  2)  
0 n2

Thus, x[n] can be represented in terms of impulse sequence as

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

 In general, we have:

x ( n )  x (  ) ( n   )  ...  x ( 3) ( n  3)
 x ( 2) ( n  2)  x ( 1) ( n  1)  x (0)( n )  x (1) ( n  1)  x ( 2) ( n  2)
 x (3) ( n  3)  ....  x ( ) ( n   )

  x(k ) (n  k )
K  

3
Course Instructor: Dr A. I. Sulyman


 Similarly for CT signals: x(t )   x( ) (t   )d


 If x(n) is only a +ve time signal, then we have


x(n)   x(k ) (n  k )
k 0

3.2 LTI Systems & Discrete-time Convolution

 Summary of LTI property:

Input LTI System Output

x1 ( n )  y1 ( n )
x2 (n )  y 2 (n )
Linearity:
(i) x3 ( n )  x1 ( n )  x2 ( n )  y 3 ( n )  y1 ( n )  y 2 ( n ) [Additive]
(ii) x4 ( n )  x1 ( n )  y 4 ( n )  y1 ( n ) [Homogeneity]

Time Invariance (T.I.):

x5 ( n )  x1 ( n  n0 )  y5 ( n )  y1 ( n  n0 ) [T. I.]

 Response of LTI system to impulse sequence δ (n):

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

1 n  0
 If x ( n )   ( n )   , then the output y(n)=h(n) is the characteristics of the
0 otherwise
system.

 Since h(n) is obtained by passing an impulse into the system, it is called the impulse
response of the system.

LTI
 Let  (n )  h (n )

LTI
 Then  (n  k )  h(n  k ) (T.I.)

4
Course Instructor: Dr A. I. Sulyman

LTI
 Also, x[k ] ( n  k )  x (k )h(n  k ) (homogeneity,   x (k ) )

 
 Finally  x(k ) (n  k )
k 0
LTI
  x(k )h(n  k ) (additivity)
k 0


 Recall from above that: x ( n )   x ( k ) ( n  k )
k 0


 y ( n )   x ( k ) h( n  k )
LTI
 Therefore x (n ) 
k 0

 This is DT convolution expression for calculating the output of LTI systems given any
input signal x (n ) .

 Notation: we write the conv. operation in short hand as y ( n )  x ( n ) * h( n ) .

 Causality:
 n
From above, y ( n )   x ( k )h( n  k )   x ( k )h( n  k )  x ( n  1)h( n  n  1)  ...
k 0 k 0
Therefore for causality of the system, we need h(n) = 0 for n < 0 (so that future inputs
x ( n  1) , x ( n  2) , etc., are not required in order to calculate present output y (n ) ).
n
 Therefore for causal LTI systems: y ( n )   x(k ) h(n  k )
k 0

 Also consider the transformation: i  n  k

0 n
 y (n )   x ( n  i )h (i )   h ( k ) x ( n  k )
in k 0
n n
Thus y ( n )  x ( n ) * h ( n )  h ( n ) * x ( n )   x ( k )h ( n  k )   h ( k ) x ( n  k )
k 0 k 0

 This is called the commutative property of DT convolution.

Example 1 (Ex 2.2 in text)

Calculate the output of an LTI system with impulse response h(n) and an input x(n), as
illustrated below:
h n 
1
n
0 1 2

x n  2

0 .5
n
0 1 2

5
Course Instructor: Dr A. I. Sulyman

Solution:
n
y ( n )   x ( k )h ( n  k )
k 0
2
x k 
0 .5
k
0 1 2

Case n < 0
x ( k )  h( n  k )  0 , thus y ( n )  0
h n  k , n  0
k
3 2 1 0

Case n = 0
0
y (0)   x ( k )  h(0  k )  0.5
k 0
h n  k , n  0
k
3 2 1 0

Case n = 1
1
y (1)   x ( k )  h(1  k )  0.5  2  2.5
k 0
h n  k , n  1
k
2 1 0 1
2

Case n = 2
2
y ( 2)   x ( k )  h( 2  k )  0.5  2  0  2.5
k 0

h n  k , n  2
k
0 1 2 3

Case n = 3
3
y (3)   x ( k )  h(3  k )  0  2  0  2
k 0
h n  k , n  3
k
0 1 2 3 4

Case n > 3
x ( k )  h( n  k )  0 , thus y ( n )  0

Sketch:

6
Course Instructor: Dr A. I. Sulyman

y n  2.5

0.5 2

n
0 1 2 3

Exercises:
(i) Calculate the output of a DT LTI system with impulse response h(n) = u(n) – u(n-3),
1
given the input sequence x n    2 n .u( n  1).u( n )
2

 [Hint: verify by sketching, that h(n) and x(n) are same as in Example 1 above].

(ii) Given an LTI system with input-output pairs below, determine the response of the
system to the input signal: x ( n )  cos n
x1 ( n )   ( n )   ( n  1)  y1 ( n )  0.5 ( n )  0.5 ( n  1)  0.5 ( n  2)  0.5 ( n  3)
x2 ( n )   ( n )   ( n  1)  y 2 ( n )  0.5 ( n )  0.5 ( n  1)  0.5 ( n  2)  0.5 ( n  3)

7
Course Instructor: Dr A. I. Sulyman

Example 2 (Ex 2.3 in text):

Calculate the output of an LTI system with input x(n) and a unit impulse response h(n)
given by
n
1
x(n )    u(n )
2
h( n )  u n 

Solution:

 The convolution operation can be described as “sliding” the sequence h(n-k) past x(k),
as shown below.

1
x k  1
2
1
4 k
0 1 2

h n  k , n   1
k
3 2 1 0

h n  k , n  0 in general
k
n 0

For general case n < 0: x (k )  h( n  k ) = 0, thus y ( n )  0 .

h n  k , n  0
k
2 1 0 1 2

h n  k , n  0 in general
k
n

 1 k
( ) 0k n
For general case n ≥ 0: x ( k )h(n  k )   2 .
 0 otherwise

1
n
1 k 1  ( ) n 1
Therefore y ( n )   ( )  2 , n  0.
k 0 2
1
1 ( )
2

8
Course Instructor: Dr A. I. Sulyman

 1 n 1 
1  ( ) 
 Thus for all n, y ( n )   2 u ( n ) , n
 1  (1) 
 
 2 

Example 3: (requires the understanding of the implication of LTI property)

(a) Find the impulse response of a discrete-time (DT) LTI system with input-output
pairs shown below.
(b) Calculate the output of the system for the input signal: x ( n )  cos n

x1 ( n )   ( n  1)   ( n  2 )   ( n  3 )  y1 ( n )  5 ( n  1)  5 ( n  2 )  5 ( n  3 )  3 ( n  4 )

x 2 ( n )  3 ( n  3 )  3 ( n  2 )  y 2 ( n )   15  ( n  2 )  15  ( n  3 )  9  ( n  4 )

Solution: see class notes

9
Course Instructor: Dr A. I. Sulyman

Example 4 (Ex 2.5 in text):

Consider an LTI system with input x[n] and unit impulse response h[n] specified as
follows:

x(n) = 2n u(-n)
h(n) = u(n)

Calculate the response of the system using DT convolution.

Solution:

1
1 x k 
1
2
4 k
2 1 0 1

h n  k 
k
n2 n 1 n

 When n  0 , x ( k )h( n  k ) has non-zero samples in interval k  0 .

0 0
 Therefore y ( n )   x ( k )h ( n  k ) 
k  
2
k  
k

 Making a change of variable: r = -k, and using the infinite sum formula:

1

k 0
k 
1
, 0< |  | <1, will give the result
0
1
 y (n )   2k   2, n  0.
k   1  (1 / 2)

 When n  0 , x ( k )h( n  k )  0 for k  n , zero otherwise.

n n
 Therefore y(n)   x(k )h(n  k ) 
k 
2
k 
k
 2n  2 [after substituting l=-k, then m=l+n]

y n  2

1 1 1
4 2 n
3 2 1 0 1 2

10
Course Instructor: Dr A. I. Sulyman

3.3 The Convolution Integral

LTI system

Input x (t) h (t) Output y (t)

 Given any input x(t) to an LTI system, what is the system response y(t)?

 If x (t )   (t ) , then y (t )  h(t ) (the impulse response of the system).

Representation of CT signals interms of impulses:

1
 Let   (t ) denote the impulse function of width  , & height


 Let x(t) be an arbitrary input, we can approx. x(t) as x(t )  x 0 (t )  x1 (t )  x 2 (t )  .....

x (t ) x0 (t) x1 ( t ) x2 (t)

   
t
  2 2 3

where
x0 (t )    (t ).x(0). 
x1 (t )    (t  ) x(). 
x 2 (t )    (t  2 ) x(2 ). 



 x (t )     (t  i ) x (i ). 
i 0

Now,   (t ) LTI
 h (t ) (Impulse Response)
  (t  i ) LTI
 h (t  i ) (Time Invariance)
LTI
  (t  i ). x (i )   h (t  i ). x (i )  (Homogeneity)
 
Finally, 
i 0
 (t  i ) x (i )   
i 0
h (t  i ) x (i )  (Additivity)

11
Course Instructor: Dr A. I. Sulyman

 Take limit of both sides as  

 0

 Therefore, the output of a C.T. system can be calculated as

  
y ( t )  lim   h ( t  i  ) x ( i  ) i  1    i  
 0
 i0 
Substitute   i 

y (t )  lim  h (t   ) x ( ) 
 0
 0
Taking the limit; (1) lim h (t )  h(t )
0 
 
(2) lim   
0
0 0

(3)   d


 Therefore, y(t )  h(t   ) x( ) d
0
[ convolution integral ]

 For causality, h(t )  0 for t < 0

 h (t   )  0 ,  t
 Hence,

t
y(t )   h(t   ) x( ) d
0

 Substitute  '  t   ,   t   ' & d  d :   0  '  t  0  t

 t  't t 0
0
y (t )   h( ' ) x(t   ' ).  d '
;set    
t
t
 y (t )   h( ) x(t   )d
0
t t
i.e., y (t )  h(t ) * x(t )  x(t ) * h(t )   h( ) x(t   )d   ht   x d
0 0

12
Course Instructor: Dr A. I. Sulyman

Examples:

1. Given an input signal x(t), calculate the output of an LTI system which delays it’s
input by two units of time.

x (t) 2 units delay y (t)

 (t )  h (t )   (t  2)
t
y (t )  x (t ) * h (t )   x ( )h (t   )d
0
t
  x( ) t    2d  x(t  2) (where we have used the sifting property of  ( ) )
0
This result is expected since the system delays its input by two time units!.

2. Given that the impulse response of an LTI system is

h(t )   (t  1)   (t  2)   (t  3) , compute the response of the system to an input
1  t 0  t 1

signal x (t ) given by x (t )   1  t - 1  t  0
0 otherwise


h (t) x (t)
1
1
t t
0 1 2 3 1 0 1

y (t )  x (t ) * h (t )
 x (t ) *  (t  1)   (t  2)   (t  3)
 x (t  1)  x (t  2)  x (t  3)

y (t)
1
t
0 1 2 3 4

13
Course Instructor: Dr A. I. Sulyman

Exercise 3: (Ex 2.6 in text)

 Let x(t) be the input to an LTI systems with unit impulse response h(t), where

x (t )  e  at u(t ) a0
h (t )  u (t )

Solution:

y (t )  x (t ) * h (t )
t
  x ( )h(t   )d
0
x ( )  e  a 
1

0

1
h (t   ) , t  0

t 0
1 h (t   ) , t  0


0 t
 For t  0 , x ( )h(t   )  0 . Thus y (t )  0 for t  0 .

e  a , 0    t
1

 e  a 0    t 
 For t  0 , x ( )h(t   )   0 t
0 otherwise

 1  e  at  , t >0.
t t
1  a t 1
 Therefore, y (t )   x ( )h(t   )d   e  a d   e
0 0
a 0 a

 For all t, y ( t ) 
1
a
1  e  at  u ( t )
1
a
t
0

14
Course Instructor: Dr A. I. Sulyman

3.4 Properties of Convolution

Summary of general response of LTI systems, for general input signal x(t), x(n):

y(n)   x ( k )h ( n  k )  x ( n ) * h ( n )
K  
convolution sum

y (t )   x( )h(t   )d  x(t ) * h(t )

convolution integral

For +ve-time signal we have: k ,  0,  .

For causal system, we have:   0, t  , k  0, n 

These general expressions will be used to discuss the following properties of convolution.

1. The Commutative Property:

 
x n  * h n   h n  * x ( n ) , i.e. y n    h(k ) xn  k    xk hn  k 
k   k  

 
xt  * ht   ht  * xt  , i.e. y t    h  x t   d    x  h ( t   ) d  .
 

Proof: (see section 3.3)

2. The Distributive Property:

x n  * h1 n   h2 n   x n  * h1 n   x n  * h2 n 
x t  * h1 t   h2 t   x t  * h1 (t )  x t  * h2 t 

 Application: two systems with impulse responses h1 (t ) , h2 (t ) operating in parallel

can be implemented as one system with impulse response h(t )  h1 (t )  h2 (t ) .
t
Proof: For causal system: y t    xt ht   d
0
t t
Let h  h1  h2 : y t    x  h1 t   d   x  h2 t   d  x (t ) * h1 (t )  x (t ) * h2 (t )
0 0

3. Associative Property:

x n  * h1 n  * h 2 n   x n  * h1 n * h 2 n 
x t  * h 1 t  * h 2 t   x t  * h 1 t * h 2 t 

15
Course Instructor: Dr A. I. Sulyman

 Application: two systems with impulse responses h1 (t ) , h2 (t ) operating in serires can

be implemented as one system with impulse response h (t )  h1 (t ) * h2 (t ) .
Proof hint: For causal LTI system,

 
 

y t    y 1   h 2 t   d     x ( ) h1 ( t   ) d  h 2 ( t   ) d   

4. LTI Systems with and without memory

y n    h k x n  k 
K

y t    h  x t   d


Let h n    n  and h t    t  ;   arbitrary constant .


y n      k x n  k   x n 
K

y t       x t   d  x t 

- Thus, LTI system is memoryless iff:

hn    n   y n   xn  i.e. output at time “n” depends only on the input at “n”

ht    t   y t   xt  i.e. output at time “t” depends only on the input at “t”

5. Invertibility of LTI systems : obtaining an LTI inverse of a system

 If B is known to be an inverse of A, test linearity and T.I.

(a) Linearity:
 1 x1 t    2 x 2 t  
A
 1 y1 t    2 y 2 t 
 1 y1 t    2 y 2 t    1 x 1 t    2 x 2 t 
B

B satisfies Homogeneity + Additivity  B is a linear inverse of A

(b) Time Invariance (TI):

x t    
A
y t   
y t     x t   
B

then B satisfies TI test  B is Time-Invariant inverse of A

 Note: if an LTI system is invertible, we can always obtain an LTI inverse of the
system.

16
Course Instructor: Dr A. I. Sulyman

6. Stability for LTI Systems

y n    h k x n  k  (output signal of LTI systems)
K

Assume that xn   B for all n

y n    hk xn  k 
K

  h k  x n  k 
K

 B  h k 
K
Therefore, the system is stable if the impulse response is absolutely summable. That is, if


 hk   
k  

 Similarly, for CT LTI system, the system is stable if



 h   d 

 

Example 1: Check the stability of LTI system with h(n) & h(t) given as: h n    n  n0  ,
h t    t  t0  .

h n    n  n 0    h n     n  n 0   1  Stable
n n

h t    t  t 0    h  d      t 0 d  1  Stable

 

Exercise 2: Check the stability of an LTI with input-output relation: y t   x   d 

t
 

Solution:
t
If x (t )   (t ), then y t   h t  . Therefore h t      d  u t 

 
Stability Check:  h   d 

  u   d 

   unstable

17