EE 831 Advanced Digital Signal Processing

Lec 2

Fall 2020
Dr Adil Masood Siddiqui
[email protected]
 Conv Sum by Analytical Method

Example:
x[n]  u[n]
h[n]  a nu[n]
y[0]  a 0  1
y[1]  1  a
y[2]  1  a  a 2
y[n]  1  a  a 2  .....  a n
n
1   n 1

r 0
 
r

1
1  a n 1
y[n] 
1 a
 Conv Sum by Analytical Method

Example:
x[n]  u[n]  u[n  N ]
h[n]  a nu[n]
y[0]  a 0  1
y[1]  1  a
y[2]  1  a  a 2
y[n]  1  a  a 2  .....  a n

 0 n0

 1  a n 1
y[n]   0  n  N 1
 1 a
a n  N 1  1  a 
N

 1 a  N 1  n

  
2D Convolution
Example of 2D Convolution
Low Pass Filtering/Image Smoothing
Noise Removal Using Median Filtering
High Pass Filtering/Edge Detection
Properties of Convolution
Stability
Stability
Causality
Example
 Linear Constant Coefficient Difference
Equation
• For all computationally realizable LTI systems, the input and the output
satisfy a difference equation of the form

• This leads to the recurrence formula

• Which can be used to compute the present output from the present and
M past values of the input and past N values of output
 LCCDE: Linear Constant Co-efficient
Difference Equation
• This addition and multiplication are infinite, if
the h[n] is infinite.
• E.g. in Accumulator
n
y[ n ]   x[ k ]
k  
n 1
y[ n ]  x[ n ]   x[ k ]
k  
n 1
y[ n  1]   x[ k ]
k  

y[ n ]  x[ n ]  y[ n  1]
y[ n ]  y[ n  1]  x[ n ]
 LCCDE
• Means, we can represent input and output in
terms of difference equation
y[n]  y[n  1]  x[n]
N M

 a y[n  k ]   b
k 0
k
b 0
m x[n  m]

More Generalized form is

N
 ak  M
 bk 
y[n]     y[n  k ]     x[n  k ]
k 1  a 0  b 0  a 0 

Auto-Regressive Part Moving Average Part

First Order Difference Equation
LCCDE Implementation
N M
y[n]   ak y[n  k ]   bk x[n  k ]
k 1 b 0
LTI Systems- Frequency Domain Analysis
 Frequency Representation

y[n]   h[k ]x[n  k ]
n 

for x[n]  e jn


y[n]  
n 
h[k ]e j ( n  k )

   j k 

j n
y[n]  e h[ k ]e 
 n  
Interpretation of impulse and frequency response

LTI System h(n)

jwn jwn
e jwn e H (e )
LTI System
02/06/2021
 Frequency Representation

02/06/2021
 Example
h[n]  [1 1]
j
n
x[n]  e 4

1
y[n]   h[k ]x[n  k ]
k 0
j j
( n 0 ) ( n 1)
 h0 e 4
 h1e 4

j j
n 
e 4
(1  e 4
)
j
n
e 4
(1  (cos( / 4)  j sin( / 4)))
j
n
e 4
(a  jb)
j
n
e 4
re j

02/06/2021
02/06/2021
02/06/2021
 Periodicity of Freq Response

H (e j (  2 ) )  
n 
h[n]e  j (  2 ) n

  h[n]e
n 
 j n  j 2 n
e
 j 2 n
e  cos(2 n)  j sin(2 n)  1 for n integer
j (  2 ) j
H (e )  H (e )

02/06/2021
 Convergence of Frequency Response

02/06/2021
 Ideal Delay:
y[n]  x[n  nd ]
x[n]  e jn
x[n  nd ]  e j ( n  nd )  e j n e  jnd
j  j nd
H (e )  e

In ideal delay system, input remains same,

and only phase changes

02/06/2021 28
 Frequency Response: Example
y[n]  x[n  1]  x[n  1]
h[n]    n  1    n  1
H (e jw )  Frequency Re sponse
y[n]  e j ( n 1)  e j ( n 1)
j  j
e  e
 2 e j n ( )
2
 2e j cos 
H (e jw )  2 cos 
Freq Response tells us, how the input
frequency is mapped at output
02/06/2021
02/06/2021
02/06/2021
02/06/2021
Continuous Time Fourier Transform
Example
Discrete Time Fourier Transform
Existence of DTFT
Comparison between Fourier Series & DTFT
 Example:
x[n]  a nu[n]

X (e ) 
jw
 x[n]e
n  
 jwn


  a u[n]e
n  
n  jwn


  a n e  jwn
n 0

  (ae  jw ) n
n 0

1
  k 
k 0 1
1
a nu[n]  | a | 1
1  ae  jw
 Ideal Delay:

y[n]  x[n  nd ]
x[n]  e jwn

jw ( n  nd ) jwn  jwnd
x[n  nd ]  e e e

In ideal delay system, input remains same,

and only phase changes
 Example:
a nu[ n  5]  ?
1
a u[ n] 
n
 jw
1  ae
a 5 a 5 a nu[n  5]
5 5 jw
n 5 ae
5
aa u[n  5]   jw
1  ae
Example:
 Ideal Delay
Or

H (e ) 
jw
 h[ n
n  
]e  jwn


H (e jw )    [
n  
n  nd ]e  jwn


H (e jw )  e  jwnd    (t   ) f (t )dt  f ( )


H R  cos wnd
H I   sin wnd
Magnitude and Phase of Fourier
DTFT Properties

46
DTFT Properties
Symmetric Properties: Rebuild
if x[n]  X (e jw )
then x[ n]  X (e  jw )
x[n]*  X * (e  jw )
x[n]  x[ n]  X (e jw )  X (e  jw )
even even
x[n]   x[n]  X (e jw )   X (e  jw )
odd odd
x[n]  x*[n]  X (e jw )  X * (e  jw )
real Hermition Symmetric
Consequence of Hermitian symmetry

If X (e jw )  X * (e  jw )
then Re[ X (e jw )] is even
jw
Im[ X (e )] is odd
| X (e jw ) | is even
X (e jw ) is odd
If x[n] is real and even, X(ejw) will be real and even
If x[n] is real and odd, X(ejw) will be imaginary and odd
Even Function
Odd Function
Example:
Example:
 Example:
Let we have LTI system with response
h[n]   (n  2)  2 (n  1) 
3 (n)  2 (n  1)   (n  2)
H ( z )  e j 2 w  2e jw  3  2e  jw  e  j 2 w
H ( z )  2 cos(2 w)  4 cos( w)  3
Comments: Frequency response is real,
so system has “zero” phase shift, this is
to be expected since unit sample
response is real and even
 Delayed Even Function
h[n]   (n)  2 (n  1) 
3 (n  2)  2 (n  3)   (n  4)
H ( z )  1  2e  jw  3e  j 2 w  2e  j 3 w  e  j 4 w
 e  j 2 w (e j 2 w  2e jw  3  2e  jw  e  2 jw )
H ( z )  e  j 2 w (2 cos(2 w)  4 cos( w)  3)

All even functions have real

Fourier Transform, If we make
it causal, it will be a complex
DTFT and LTI System
DTFT:
Example:
Moving Average
Inverse Fourier Transform

1 a b n
 jw  jw
 a u[n] 
n
b u[n]
(1  ae )(1  be ) a b ba
 Freq Response of LCCDE
1 1
y[n]  y[n  1]  x[n]  x[n  1]
2 4
1 n 1 1 n 1
( ) u[n]  ( ) u[n  1]
2 4 2

Use freqz(B,A,X) for frequency response of the equation