Skip to content Skip to navigation

Connexions

You are here: Home » Content » Fast Convolution Using the FFT

Navigation

Content Actions

  • Download module PDF
  • Add to ...
    Add the module to:
    • My Favorites
    • A lens
    • An external social bookmarking service
    • My Favorites (What is 'My Favorites'?)
      'My Favorites' is a special kind of lens which you can use to bookmark modules and collections directly in Connexions. 'My Favorites' can only be seen by you, and collections saved in 'My Favorites' can remember the last module you were on. You need a Connexions account to use 'My Favorites'.
    • A lens (What is a lens?)

      Definition of a lens

      Lenses

      A lens is a custom view of Connexions content. You can think of it as a fancy kind of list that will let you see Connexions through the eyes of organizations and people you trust.

      What is in a lens?

      Lens makers point to Connexions materials (modules and collections), creating a guide that includes their own comments and descriptive tags about the content.

      Who can create a lens?

      Any individual Connexions member, a community, or a respected organization.

    • External bookmarks
  • E-mail the author

Recently Viewed

This feature requires Javascript to be enabled.

Fast Convolution Using the FFT

Module by: Robert Nowak

Summary: This module describes FFT, convolution, filtering, LTI systems, digital filters and circular convolution.

Important Application of the FFT

Exercise 1

How many complex multiplies and adds are required to convolve two NN-pt sequences? yn=m=0N-1xmhn-m y n m 0 N 1 x m h n m

Solution 1

There are 2N-1 2 N 1 non-zero output points and each will be computed using NN complex mults and N-1 N 1 complex adds. Therefore, Total Cost=2N-1N+N-1ON2 Total Cost 2 N 1 N N 1 O N 2

  1. Zero-pad these two sequences to length 2N-1 2 N 1 , take DFTs using the FFT algorithm xnXk x n X k hnHk h n H k The cost is O2N-1log2N-1=ONlogN O 2 N 1 2 N 1 O N N
  2. Multiply DFTs XkHk X k H k The cost is O2N-1=ON O 2 N 1 O N
  3. Inverse DFT using FFT XkHkyn X k H k y n The cost is O2N-1log2N-1=ONlogN O 2 N 1 2 N 1 O N N

So the total cost for direct convolution of two NN-point sequences is ON2 O N 2 . Total cost for convolution using FFT algorithm is ONlogN O N N . That is a huge savings (Figure 1).

Figure 1
Figure 1 (figure1.png)

Summary of DFT

  • xn x n is an NN-point signal (Figure 2).
Figure 2
Figure 2 (figure2.png)
  • Xk=n=0N-1xn-2πNkn=n=0N-1xnWNkn X k n 0 N 1 x n 2 N k n n 0 N 1 x n WN k n where WN=-2πN WN 2 N is a "twiddle factor" and the first part is the basic DFT.

What is the DFT

Xk=XF=kN=n=0N-1xn-2πFn X k X F k N n 0 N 1 x n 2 F n where XF=kN X F k N is the DTFT of xn x n and n=0N-1xn-2πFn n 0 N 1 x n 2 F n is the DTFT of xn x n at digital frequency FF. This is a sample of the DTFT. We can do frequency domain analysis on a computer!

Inverse DFT (IDFT)

xn=1Nn=0N-1Xk2πNkn x n 1 N n 0 N 1 X k 2 N k n

  • Build xn x n using Simple complex sinusoidal building block signals
  • Amplitude of each complex sinusoidal building block in xn x n is 1NXk 1 N X k

Circular Convolution

DFT

xnhnXkHk x n h n X k H k (1)

Regular Convolution from Circular Convolution

  • Zero pad xn x n and hn h n to length=lengthx+lengthh-1 length length x length h 1
  • Zero padding increases frequency resolution in DFT domain (Figure 3)

Figure 3
Subfigure 3.1: 8-pt DFT of 8-pt signalSubfigure 3.2: 16-pt DFT of same signal padded with 8 additional zeros
Subfigure 3.1 (figure3-1.png)Subfigure 3.2 (figure3-2.png)

The Fast Fourier Transform (FFT)

  • Efficient computational algorithm for calculating the DFT
  • "Divide and conquer"
  • Break signal into even and odd samples keep taking shorter and shorter DFTs, then build NN-pt DFT by cleverly combining shorter DFTs
  • NN-pt DFT: ON2ONlog2N O N 2 O N 2 N

Fast Convolution

  • Use FFT's to compute circular convolution of zero-padded signals
  • Much faster than regular convolution if signal lengths are long
  • ON2ONlog2N O N 2 O N 2 N

See Figure 4.

Figure 4
Figure 4 (figure4.png)

Comments, questions, feedback, criticisms?

Send feedback