Efficient Computation of the DFT:

The FFT Algorithm

Hafiz Malik
Dept. of Electrical & Computer Engineering
The University of Michigan-Dearborn
[email protected]
http://www-perosnal-engin.umd.umich.edu/~hafiz

10/20/20 DFT - Hafiz Malik 1

 Acknowledgement
• These lecture slides are based on
– Chapter 5 of
• Digital Signal Processing: Principles, Algorithms and
Applications 3rd e
• John G. Proakis and Dimitris K Manolakis
– Lecture slides of Prof. Richard M. Stern
• Department of Electrical and Computer Engineering,
Carnegie Mellon University

10/20/20 DFT - Hafiz Malik 2

The Cooley-Tukey decimation-in-time algorithm

• Consider the DFT algorithm for an integer power of 2, N  2

N 1 N 1
X[k]   x[n]WN nk   x[n]e j2nk / N ; WN  e j2 / N
n0 n0
• Create separate sums for even and odd values of n:
X[k]   x[n]WN nk   x[n]WN nk
n even n odd
• Letting n  2r for n even and n  2r  1 for n odd, we obtain
 N / 21  N / 2 1
X[k]   x[2r]WN 2rk   x[2r  1]WN  2r1 k
r 0 r 0

10/20/20 DFT - Hafiz Malik 3

 The Cooley-Tukey DIT algorithm
• Splitting indices in time, we have obtained
 N / 21  N / 2 1
X[k]   x[2r]WN 2rk   x[2r  1]WN 2r1 k
r 0 r 0

• But W 2  e  j 2 2 / N  e  j2 /( N / 2)  W and W 2rk W k  W k W rk

N N/2 N N N N /2
So …
(N/ 2)1 ( N/ 2)1
X[k]   x[2r]WNrk/ 2  WNk  x[2r  1]WNrk/ 2
n0 n0

N/2-point DFT of x[2r] N/2-point DFT of x[2r+1]

10/20/20 DFT - Hafiz Malik 4

 Computational Gain so far …

• We have split the DFT computation into two halves:

N 1
X[k]   x[n]WN nk
k 0
( N/ 2)1 ( N/ 2)1
  x[2r]WNrk/ 2  WNk  x[2r  1]WNrk/ 2
n0 n0

• Have we gained anything? Consider the nominal number of

multiplications for N  8
– Original form produces 2
8  64 multiplications
– New form produces multiplications
2(4 2 )  8  40
– So we’re already ahead ….. Let’s keep going!!
10/20/20 DFT - Hafiz Malik 5
 Signal flowgraph notation
• In generalizing this formulation, it is most convenient to adopt a graphic
approach …
• Signal flowgraph notation describes the three basic DSP operations:
– Addition

x[n]
– Multiplication by a constant x[n]+y[n]
y[n]
– Delay
a
x[n] ax[n]

z-1
x[n] x[n-1]

10/20/20 DFT - Hafiz Malik 6

Signal flowgraph representation of 8-point DFT
k
• Recall that the DFT is now of the form X[k]  G[k]  WN H[k]
• The DFT in (partial) flowgraph notation:

10/20/20 DFT - Hafiz Malik 7

 Continuing with the decomposition …
• So why not break up into additional DFTs? Let’s take the
upper 4-point DFT and break it up into two 2-point DFTs:

10/20/20 DFT - Hafiz Malik 8

The complete decomposition into 2-point DFTs

10/20/20 DFT - Hafiz Malik 9

Now let’s take a closer look at the 2-point DFT

• The expression for the 2-point DFT is:

1 1
X[k]   x[n]W2nk   x[n]e  j 2nk / 2
n0 n0

• Evaluating for k  0,1 we obtain

X[0]  x[0]  x[1]
X[1]  x[0]  e  j 21 / 2 x[1]  x[0]  x[1]

which in signal flowgraph notation looks like ...

This topology is referred to as the
basic butterfly

10/20/20 DFT - Hafiz Malik 10

The complete 8-point DIT FFT Structure

10/20/20 DFT - Hafiz Malik 11

Computational Complexity of N-point FFTs

• Let
N  2 where   log 2 (N)
• (log2(N) columns)(N/2 butterflys/column)(2
mults/butterfly)
N log 2 (N)
or ~ multiplys
10/20/20 DFT - Hafiz Malik 12
Additional timesavers: reducing multiplications
in the basic butterfly
• As we derived it, the basic butterfly is of the form

WNr

WNr N / 2
• Since W N / 2  1 we can reducing computation by 2 by
N
premultiplying byWNr

WNr 1
10/20/20 DFT - Hafiz Malik 13
 Alternate FFT structures

• We developed the basic decimation-in-time

(DIT) FFT structure, but other forms are
possible simply by rearranging the branches of
the signal flowgraph
• Consider the rearranged signal flow diagrams
on the following panels …..

10/20/20 DFT - Hafiz Malik 16

 Alternate DIT FFT structures
• DIT structure with input bit-reversed, output natural

10/20/20 DFT - Hafiz Malik 17

 Alternate DIT FFT structures
• DIT structure with input natural, output bit-reversed:

10/20/20 DFT - Hafiz Malik 18

 Alternate DIT FFT structures
• DIT structure with both input and output natural:

10/20/20 DFT - Hafiz Malik 19

 Alternate DIT FFT structures
• DIT structure with same structure for each stage:

10/20/20 DFT - Hafiz Malik 20

Comments on alternate FFT structures

• A method to avoid bit-reversal in filtering

operations is:
– Compute forward transform using natural input, bit-
reversed output (as in OSB 9.10)
– Multiply DFT coefficients of input and filter response
(both in bit-reversed order)
– Compute inverse transform of product using bit-
reversed input and natural output (as in OSB 9/14)
• Latter two topologies (as in OSB 9.15 and 9.16)
are now rarely used

10/20/20 DFT - Hafiz Malik 21

 The decimation-in-frequency (DIF)
FFT algorithm

• Introduction: Decimation in frequency is an

alternate way of developing the FFT algorithm
• It is different from decimation in time in its
development, although it leads to a very
similar structure

10/20/20 DFT - Hafiz Malik 22

 FFT: Decimation in Frequency
• Consider the original DFT equation ….
N 1
X (k )   x(n)WNnk
n 0
• Separate the first half and the second half of time samples:
( N / 2 ) 1 N 1
X (k )  n 0
x (n)W nk
N  
n N / 2
x(n)WNnk
( N / 2 ) 1 ( N / 2 ) 1
 
n 0
x(n)W nk
N W ( N / 2) k
N n 0
x( n  N / 2)WNnk

  x(n)  (1) 
( N / 2 ) 1
 k
x( n  N / 2) WNnk
n 0

• Note that these are not N/2-point DFTs

10/20/20 DFT - Hafiz Malik 23
 FFT: Decimation in Frequency
  x(n)  (1) 
( N / 2 ) 1
X (k )  k
x ( n  ( N / 2)) WNnk
n 0

• For k even, let k  2r

  x(n)  (1) 
( N / 2 ) 1
X (k )  2r
x(n  ( N / 2)) WNn 2 r
n 0
( N / 2 ) 1
   x ( n )  x ( n  ( N / 2)) W nr
N /2
n 0

• For k odd, let k  2r 1

  x(n)  (1) 
( N / 2 ) 1
X (k )  2r
( 1) x(n  ( N / 2)) WNn ( 2 r 1)
n 0
( N / 2 ) 1
   x(n)  x(n  ( N / 2))W W
n 0
n
N
nr
N /2

• These expressions are the N/2-point DFTs of

10/20/20
x ( n )  x ( n  ( N / 2)) and [ x ( n )
DFT - Hafiz Malik
 x ( n  ( N / 2))]WN
n
24
 These equations describe the
following structure:

10/20/20 DFT - Hafiz Malik 25

 Continuing by decomposing the odd
and even output points results …

10/20/20 DFT - Hafiz Malik 26

… and replacing the N/4-point DFTs by
butterflys we obtain

10/20/20 DFT - Hafiz Malik 27

The DIF FFT is the transpose of the DIT FFT
• To obtain flowgraph transposes:
– Reverse direction of flowgraph arrows
– Interchange input(s) and output(s)
• DIT butterfly: DIF butterfly:

• Comment:
– We will revisit transposed forms again in our discussion of
filter implementation

10/20/20 DFT - Hafiz Malik 28

The DIF FFT is the transpose of the DIT FFT
• Comparing DIT and DIF structures:
DIT FFT structure: DIF FFT structure:

• Alternate forms for DIF FFTs are similar to those of

DIT FFTs

10/20/20 DFT - Hafiz Malik 29

 Alternate DIF FFT structures
• DIF structure with input natural, output bit-reversed:

10/20/20 DFT - Hafiz Malik 30

 Alternate DIF FFT structures
• DIF structure with input bit-reversed, output natural:

10/20/20 DFT - Hafiz Malik 31

 Alternate DIF FFT structures
• DIF structure with both input and output natural:

10/20/20 DFT - Hafiz Malik 32

 Alternate DIF FFT structures
• DIF structure with same structure for each stage:

10/20/20 DFT - Hafiz Malik 33

 FFT structures for other DFT sizes
• Can we do anything when the DFT size N is not
an integer power of 2 (the non-radix 2 case)?
• Yes! Consider a value of N that is not a power
of 2, but that still is highly factorable …

• Let N  p1 p2 p3 p4 ...p ; q1  N / p1, q2  N / p1 p2 , etc.

Then let

N 1
X[k]   x[n]WNnk
n0
q1 1 q1 1 q1 1
p1rk ( p1r1)k ( p r 2)k
  x[ p1r]WN   x[ p1r  1]WN   x[ p1r  2]WN 1  ...
r0 r0 r0
10/20/20 DFT - Hafiz Malik 34
 Non-radix 2 FFTs (continued)
• An arbitrary term of the sum on the previous
panel is

q1 1
 x[ p1r  l]WN( p1rl)k
r0
q1 1 q1 1
p1rk lk
  x[ p1r  l]WN WN  WN  x[ p1r  l]Wqrk
lk
1
r0 r0

• This is, of course, a DFT of size of points

spaced by q1
p1

10/20/20 DFT - Hafiz Malik 35

 Non-radix 2 FFTs (continued)
• In general, for the first decomposition we use

p1 1 q1 1
X[k]   WNlk  x[ p1r  l]Wqrk
1
l 0 r0
• Comments:
– This procedure can be repeated for subsequent factors of
N
– The amount of computational savings depends on the
extent to which N is “composite”, able to be factored into
small integers
– Generally the smallest factors possible used, with the
exception of some use of radix-4 and radix-8 FFTs

10/20/20 DFT - Hafiz Malik 36

 Direct DFT & FFT Comparison
• Time (1 multiplication per microsec)

Direct DFT
N FFT
26 64 .02 sec .002 sec
29 512 1 .02 sec
212 4096 67 .2
215 32768 1 hr 11 mins 2
218 262144 3 days 4 hrs 19

10/20/20 DFT - Hafiz Malik 37