EEE4221 Digital Signal Processing

Lecture 4 : Linear Time-Invariant Systems

Instructor: Dr D. Banda
Email: daliso.banda@unza.zm

March 2022
 References
Our main reference text books in this course are
[1] John G. Proakis, Dimitris Manolakis, “Digital Signal Processing:
Principles, Algorithms and Applications”, 4th Edition, (2006), Prentice
Hall, ISBN-10: 0131873741, ISBN-13: 978-0131873742.
[2] Alan V. Oppenheim, Ronald W. Schafer, John R. Buck, Discrete-Time
Signal Processing, 3rd Edition, (2010), Prentice-Hall, ISBN: 0137549202 /
0-13-754920-2.
[3] DIMITRIS G. MANOLAKIS, VINAY K. INGLE, Applied Digital Signal
Processing, 2nd Edition, (2011) Cambridge University Press, ISBN 978-0-
521-11002-0.

However, feel free to use some additional text which you might find relevant
to our course.

Based on notes produced by Jerry MUWAMBA , DEPT. of EEE, School of Engineering, UNZA

2
Linear-Time Invariant System

Can be decomposed into a sequence of delayed impulses

scaled by x[k] as follows:

Remember this result for the next slides

3
Linear-Time Invariant System

For a linear system:

Time variant:

Time invariant:

4
 Linear-Time Invariant System
 From linearity and time invariance we arrive at the convolution sum

y  n   x k  h n  k   x n  h n
k 

 Special importance for their mathematical tractability

 Most signal processing applications involve LTI systems
 LTI system can be completely characterized by its impulse response

 The response of an LTI system to any input signal x[n] can be

determined from it’s impulse response h[n] using the convolution sum.

5
 Evaluating the Convolution the Sum:

1. By Superposition
2. Analytically
3. Numerically using Octave/Matlab
4. Convolution as a “scanning” operation

6
LTI System Example by Superposition
2 2

1 1

0 0
-5 0 5 -5 0 5

2 2

1 1

0 0
-5 0 5 -5 0 5
4

2
2
1

0 0
-5 0 5
-5 0 5

7
LTI System Example by Superposition

8
LTI System Example by Superposition

9
 Example 1 Analytical Evaluation of the Convolution sum

 For h[n] = [1 1 1] and x[n] = [ 1 2 3 4] determine the system response

y[n].

10
 Example 1 Analytical Evaluation of the Convolution sum

 For h[n] = {1 1 1} and x[n] = {1 2 3 4} determine the system response

y[n].

 ANS: y[n] = {1 3 6 9 7 4} (confirm numerically using octave)

11
 Example 4.1 Analytical Evaluation of the Convolution sum

 Consider a system with impulse response

1, 0  n  N  1,
h  n  u  n  u n  N   
0, otherwise.
 The input is x  n   a nu  n 
 The output y  n  is determined from the sums over all k of the
product x  k  h  n  k  for different values of n.
 Figure Example 4.1(a) shows sequences x  k  and h  n  k  , plotted
for n a negative. Vividly, nonzero portions of the sequences do not
overlap, thus y  n  0 , n  0.
 Figure Example 4.1(b) illustrates the two sequences when 0  n  N  1
. Clearly for this range, x  k  h  n  k   a ,
k

n
 It follows that y  n   a k , for 0  n  N  1
k 0

12
Example 4.1 Analytical Evaluation of the Convolution sum
x k 
h n  k  Vividly, nonzero portions of
the sequences do not overlap,
 
thus y n  0 , n  0.

x k  h n  k   ak

Figure Example 4.1

13
 Example 4.1 Final steps

 Since y  n  is the sum of n+1 terms of a geometric series, we obtain

1  a n 1
y  n  , 0  n  N  1,
1 a
 Finally, Figure Example 4.1(c) shows the two sequences when N  1  n
or 0  n  N  1 . As before, x  k  h  n  k   a , n  N 1  k  n ,
k

 but now the lower limit on the sum is n  N  1 , thus

n
y  n   ak , for N  1  n.
k  n  N 1

 Therefore,

14
 LTI System Convolution as a “scanning” operation

 The samples of the sequence x[k] are in natural order whereas sequence
h[k] samples are in reverse order (flipping or time reversal h[-k ]).
 To determine the value of y[n] for n = n0, the flipped impulse response
sequence is shifted so that the sample h[0] is aligned to sample x[n0] of
the input. Eg. consider:

15
 FIR versus IIR systems
• If the impulse response has a finite number of nonzero samples
we have a finite (duration) impulse response (FIR) system.
• Otherwise, we have a system with infinite (duration) impulse
response (IIR).
• Here is a block diagram representation of an FIR system.

• See board for example IIR system

• These filters can also be 2D filters for image processing
(Discuss lab).
16
 Properties of LTI Systems
 Convolution is commutative
 
x  n  h  n   x k  h n  k    h k  x n  k   h n  x n
k  k 

 Convolution is distributive
x  n   h1  n  h2  n  x  n   h1  n   x  n   h2  n 

17
 Properties of LTI Systems
 Cascade connection of LTI systems

18
 Stable and Causal LTI Systems
 An LTI system is (BIBO) stable if and only if
 Impulse response is absolutely summable, i.e., if


 h k   
k 

 Let’s write the output of the system as

 
y  n   h k  x n  k    h k  x n  k 
k  k 
 If the input is bounded, i.e.,
x[n]  Bx
 Then the output is bounded by

y  n  Bx  h k 
k 
 The output is bounded if the absolute sum is finite
 An LTI system is causal if and only if h  k   0 for k  0

19
 Linear Constant-Coefficient Difference Equations
 An important subclass of LTI systems are of the form
N M

 a y  n  k    b x n  m
k 0
k
m0
m

 The output is not uniquely specified for a given input

 The initial conditions are required
 Linearity, time invariance, and causality depend on the initial conditions
 If initial conditions are assumed to be zero system is linear, time invariant,
and causal
 Example
 Moving Average
y[n]  x[n]  x[n  1]  x[n  2]  x[n  3]
 Difference Equation Representation
0 3

 a y  n  k    b x  n  m
k 0
k
m0
m where ak  bm  1

20
 Eigenfunctions of LTI Systems

 Eigenfunction of an LTI systems is the signal which when fed into the system,
the response is that same signal multiplied by a complex constant
(eigenvalue).

21
 Eigenfunctions of LTI Systems
 Complex exponentials are eigenfunctions of LTI systems:
x  n   e j n

 Let’s see what happens if we feed x[n] into an LTI system:

 
y  n   h k x n  k    h  k e j ( n  k )

k  k 
eigenfunction
 
 j k  j n
y  n     h  k e  e  H  e j
 e j n

 k  
eigenvalue
 The eigenvalue is called the frequency response of the system

H e j
   h k e  j k

k 

 H(ej) is a complex function of frequency

 Specifies amplitude and phase change of the complex exp. input.
22
 Eigenfunctions of LTI Systems
 Example:
Calculate The system response for a system with h[n]={0.5 0.5}
Determine y[n] if x[n] = cos(πn/2)

23
24
 End of Lecture 4

Thank you for your attention!

25