Scribd
Upload
271 views2 pages

Linear Convolution Using DFT

The document discusses using the discrete Fourier transform (DFT) to perform linear convolution. It explains that linear convolution is needed for real-time filtering but circular convolution from the DFT property is not suitable. Zero padding is used to increase the length of the input sequences to the required length for the DFT. The procedure involves taking the DFT of the input and impulse response, multiplying them together in the frequency domain, and taking the inverse DFT to obtain the output of the linear convolution. An example is provided to demonstrate computing the linear convolution of two sequences using this DFT method.

Uploaded by

samhith
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
271 views2 pages

Linear Convolution Using DFT

The document discusses using the discrete Fourier transform (DFT) to perform linear convolution. It explains that linear convolution is needed for real-time filtering but circular convolution from the DFT property is not suitable. Zero padding is used to increase the length of the input sequences to the required length for the DFT. The procedure involves taking the DFT of the input and impulse response, multiplying them together in the frequency domain, and taking the inverse DFT to obtain the output of the linear convolution. An example is provided to demonstrate computing the linear convolution of two sequences using this DFT method.

Uploaded by

samhith
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Linear Convolution using DFT

The circular convolution property states that the product of two DFTs is equivalent to the circular
convolution of the corresponding time-domain sequence. But to determine the output of a real time
(linear) filter, the circular convolution is not suitable. Therefore in this section, the linear convolution
through circular convolution using DFT is explained.

Let x(n) be the finite duration sequence with the length of L and h(n id the impulse response of a
digital filter with length of M. The resultant output sequence will have a length of N=L+M-1 samples.
Therefore, the DFT of size N is required to represent y(n) in the frequency domain. The number of
samples in x(n) and x(n) has been increased to meet this condition using zero padding technique.
Adding zero at the end does not affect the frequency response of the system. Thus with zero padding,
the DFT can be used to perform linear filtering.

The procedure is illustrated below.

1. Using zero padding increase the length of x(n) and y(n) to N from L and M
respectively.
2. Compute N-point DFT for x(n) and h(n) and represent as X(k) and H(k)
respectively.
3. Compute Y(k)=X(k)H(k)
4. Compute Inverse DFT to obtain the y(n). i.e. y(n)=IDFT{Y(k)}.

Example
Perform the linear convolution of the two sequences given below using the DFT relations.
n, 0  n  3 1, n  1,1
h( n)   and x(n)  
0, otherwise 0, n0
Solution

L  4, M  3, therefore N  6
Computation of H(k)
 2 
3  j  nk
H ( k )   h ( n )e  N 

n 0
 2 
3  j  nk
  h ( n )e  6 
( N  6)
n 0
   2 
 j k  j k
e  2e
 3
 3e j ( k )
 3 

For k=0,1,2…5 H (0)  6


H (1)  e j ( /3)  2e j (2 /3)  3e j ( )
H (2)  e j (2 /3)  2e j (4 /3)  3e j (2 )
H (3)  e j  e j (2 )  3e j (3 )
H (4)  e j (4 /3)  2e j (8 /3)  3e j (4 )
H (5)  e j (5 /3)  2e j (10 /3)  3e j (5 )
Computation of X(k)
 2 
1  j  nk
X (k )   x(n)e
n 1
 N 
 2   2 
j k  j k
e  6 
0e  6 

 2   2 
j k  j k
X (k )  e e
 6   6 

For k=0,1,2…5 X (0)  1  1  2


X (1)  e j (2 /6)  e j (2 /6)
X (2)  e j (2 /3)  e j (2 /3)
X (3)  e j ( )  e j ( )
X (4)  e j (4 /3)  e j (4 /3)
X (5)  e j (5 /3)  e j (5 /3)
Computation of Y(k)

Y(k)=H(k)X(k)

For k=0,1,2,…5 Y  0   H  0  X  0   12
Y 1  H 1 X 1  e j ( /3)  2e j (2 /3)  3e  j ( )  e j (2 /6)  e j (2 /6) 
 1  6e j (2 /3)  2e j ( )  3e j (4 /3)
Y  2   H  2  .X  2   e j (2 /3)  2e j (4 /3)  3e j (2 )  e j (2 /3)  e  j (2 /3) 
 8 
 j 
 j (4 /3)  j (2 /3)  j (2 )
 1  4e  2e  2e  3e  3 

Y  3  H  3 X  3  e  j ( )
 2e  j (2 )
 3e  j (3 )  e j ( )  2e  j (2 )  e  j ( ) 
 1  2e j  4e j 2  2e j 3  3e j 4
Y  4   H  4  X  4   e j (4 /3)  2e j (8 /3)  3e j (4 )  e j (4 /3)  e  j (4 /3) 
 8   4   16   12 
 j   j   j   j 
 1  4e  2e  3   2e  3   2e  3 
 3 

  j  5   10 
 j  
Y  5  H  5 X  5  e  3   2e  3   3e j (5 )  e j (5 /3)  e j (5 /3) 
 
 10   5   15   20 
 j   j   j   j 
 1  4e  3 
 2e  3 
e  3 
 3e  3 

Computation of y(n)
The Inverse Discrete Fourier Transform (IDFT) is given by
1 N 1

y ( n) 
N i 0
Y (k )e j (2 K / N ) , 0  n  N  1

1 5
Here N  6, y (n)   Y (k )e j (2 nK /6)
6 k 0
The different values of y (n) can be evaluated using the above expression and the outputs obtained are

y(n)={0,1,2,4,2,3}

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy