You are on page 1of 2

Lecture 15: Fast Fourier Transform (FFT)

 Introduction: Fast Fourier Transforms (FFTs) are computational efficient algorithms for
evaluating DFT. Using FFT algorithms the number of computations required for computing
DFT are significantly reduced compared to the direct computation of DFT.

Direct computation of DFT is inefficient primarily because it does not exploit the symmetry
 j 2
and periodicity properties of the phase factor or twiddle factor N  e N . The term  N is
 j 2  j 2
2
known as phase factor where | N |  |e N |  1 always and N  (e N )
N
always. Therefore, the phase factor is a vector on a unit circle. In particular the phase factor
has two important properties:
 N
K 
(i) Symmetry property: N  2
 N K
(ii) Periodicity property: N 
K N 
 N K

The direct computation of DFT is given by the equation:


N 1  j 2 N 1

  x[n] 
k .n
X [k ]  x[n] e N
= N
k n
where 0  k  N  1
n0 n0

Using the above equation, direct calculation of DFT for one point or value of ‘k’ requires
‘N’ complex multiplications and ‘N-1’ complex additions. Consequently, to compute all ‘N’
values of the DFT requires N x N = N2 complex multiplications and N(N-1) complex
additions. However using FFT algorithm the number of complex multiplications required is
N
log 2 N and complex additions is N log 2 N .
2

Number of Number of For Direct Computation of For FFT algorithm


stages ‘M’ points ‘N’ DFT

Multiplication Addition Multiplication Addition


(N2) N(N-1) N N log 2 N
log 2 N
2
2 4 16 12 4 8
3 8 64 56 12 24
4 16 256 240 32 64
5 32 1024 992 80 160

Lecture Notes on Digital Signal Processing (EE1706): By Roshan Pradhan Page 1


The FFT algorithms follow divide and conquer approach for computation of DFT. This
approach is based on the decomposition of N-point DFT into successively smaller DFTs.

 Radix-2 DIT (Decimation in time) FFT algorithm: Radix-2 means that the number of
DFT points ‘N’ can be expressed as a power of 2, i.e. N = 2M where ‘M’ is an integer. In this
method, a ‘N’ point DFT can be obtained by successively adding or combining N-number of
2 point DFTs, i.e. a ‘N’ point DFT is successively divided into smaller groups (groups of 2
in this case) till a 2-point DFT is obtained and then these 2-point DFTs are successively
combined to obtain a ‘N’ point DFT. Hence, this method is known as divide and conquer
approach.

Lecture Notes on Digital Signal Processing (EE1706): By Roshan Pradhan Page 2

You might also like