Lab.8.

Discrete Fourier Transform (DFT)

and Fast Fourier Transform (FFT)
(Theory and Implementation)
 Learning Objectives
 DFT algorithm.
 Conversion of DFT to FFT algorithm.
 Implementation of the FFT algorithm.
 DFT Algorithm
 The Fourier transform of an analogue signal
x(t) is given by:

X     x t e  jt
dt


 The Discrete Fourier Transform (DFT) of

a discrete-time signal x(nT) is given by:
N 1 2
j nk
X k    xn e N

n0

 Where: k  0,1, N  1
xnT   xn
 DFT Algorithm
j
2 N 1
 If we let: e N
 WN then: X k    xn WNnk
n 0

Sampled signal
2

1
Amplitude

-1

-2
0 20 40 60 80 100 120
Sample
Frequency Domain
1

0.8
Magnitude

0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5
Normalised Frequency
 DFT Algorithm
N 1
x[n] = input
X k    xn WNnk X[k] = frequency bins
n 0 W = twiddle factors

X(0) = x[0]WN0 + x[1]WN0*1 +…+ x[N-1]WN0*(N-1)

X(1) = x[0]WN0 + x[1]WN1*1 +…+ x[N-1]WN1*(N-1)
:
X(k) = x[0]WN0 + x[1]WNk*1 +…+ x[N-1]WNk*(N-1)
:
X(N-1) = x[0]WN0 + x[1]WN (N-1)*1 +…+ x[N-1]WN (N-1)(N-1)

Note: For N samples of x we have N frequencies

representing the signal.
 Performance of the DFT Algorithm
 The DFT requires N2 (NxN) complex
multiplications:
 Each X(k) requires N complex multiplications.
 Therefore to evaluate all the values of the DFT
( X(0) to X(N-1) ) N2 multiplications are
required.
 The DFT also requires (N-1)*N complex
additions:
 Each X(k) requires N-1 additions.
 Therefore to evaluate all the values of the DFT
(N-1)*N additions are required.
 Performance of the DFT Algorithm

120 120
Number of Multiplications

Number of Additions
100 100

80 80

60 60
40 40
20 20
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Number of Samples Number of Samples

 Can the number of computations required

be reduced?
 DFT  FFT
 A large amount of work has been devoted
to reducing the computation time of a DFT.
 This has led to efficient algorithms which
are known as the Fast Fourier Transform
(FFT) algorithms.
 DFT  FFT
N 1
X k    xn WNnk ; 0  k  N  1 [1]
n 0

x[n] = x[0], x[1], …, x[N-1]

 Lets divide the sequence x[n] into even

and odd sequences:
 x[2n] = x[0], x[2], …, x[N-2]
 x[2n+1] = x[1], x[3], …, x[N-1]
 DFT  FFT
2
 Equation 1 can be rewritten as: e
j
N
 WN
N N
1 1
2 2
X k    x2n  WN2 nk   x2n  1 WN2 n 1k [2]
n0 n0

 Since:
2 2
j 2 nk j nk
WN2 nk  e N
e N 2
WN2 n 1k  W Nk  W Nnk
 W Nnk 2
2

 Then: N
2
1
N
2
1

X k    x2n W Nnk  WNk  x2n  1 W Nnk

n0 2 n 0 2

 Y k   WNk Z k 
 DFT  FFT
 The result is that an N-point DFT can be
divided into two N/2 point DFT’s:
N 1
X k    xn WNnk ; 0  k  N  1 N-point DFT
n0

 Where Y(k) and Z(k) are the two N/2

point DFTs operating on even and odd
samples respectively:
N N
1 1
2 2
X k    x1 n W Nnk  WNk 
 2 N
x n W nk
Two N/2-
n0 2 n 0 2 point DFTs
 Y k   WNk Z k 
 DFT  FFT
 Periodicity and symmetry of W can be
exploited to simplify the DFT further:
N N
1 1
2 2
X k    x1 n W Nnk  WNk  x2 n W Nnk [3]
n 0 2 n 0 2


N N
1  N 1  N
2 N 2
 N n k   n k  
X  k     x1 n W  x n W
 2 k  2
N  WN 2
2 N
 2  n 0 2 n 0 2
N 2 2 N 2 2
Or: WN
k
2
e
j
N
k
e
j
N 2
e
j
N
k
e  j
 e
j
N
k
 W Nk : Symmetry

N 2 2 N 2
k j k j j k
And: WN 2
e N 2
e N 2 2
e N 2
 W Nk : Periodicity
2 2
 DFT  FFT
 Symmetry and periodicity:

W86
W85 W87
WNk+N/2 = -WNk
WN/2k+N/2 = WN/2k
W84 W80 = W88
W8k+4 = -W8k
W8k+8 = W8k
W83 W81 = W89
W82
 DFT  FFT
 Finally by exploiting the symmetry and
periodicity, Equation 3 can be written as:
N N
1 1
2 2
 N
X k     x1 nW Nnk  WNk  x2 nW Nnk
 2 n 0 2 n0 2
[4]
 Y k   W Nk Z k 
 DFT  FFT
N 
X k   Y k   WNk Z k ;
k  0,    1
2 
 N N 
X  k    Y k   W N Z k ; k  0,   1
k

 2 2 

 Y(k) and WNk Z(k) only need to be

calculated once and used for both
equations.
 Note: the calculation is reduced from 0 to
N-1 to 0 to (N/2 - 1).
 DFT  FFT
N 
X k   Y k   WNk Z k ;
k  0,    1
2 
 N N 
X  k    Y k   W N Z k ; k  0,   1
k

 2 2 

 Y(k) and Z(k) can also be divided into N/4

point DFTs using the same process shown
above:
Y k   U k   W Nk V k  Z k   Pk   W Nk Qk 
2 2

 N  N
Y  k    U k   W Nk V k  Z  k    Pk   W Nk Q k 
 4 2  4 2

 The process continues until we reach 2

point DFTs.