APPENDIX 2: Fast Fourier Transform

Matrix Factorizations

Recall the Fourier matrix as an N N matrix:

1
1
1
...

1
w
w2
...

1
w2
w4
...
1
FN =

N .
..
..
..
.
.

wN1

w2(N1)

...

1
wN2
w2(N2)
..
.
w(N1)(N2)

wN1

2(N1)

..

w(N1)

We are attempting to represent it as the product of matrices with 2 non-zero entries in

each row - at least in the case N = 2M . Just how everything changes for different values
of N , say N prime, could be a matter for investigation also. As FN is symmetric, we
could replace w by w if we wanted. This would make the notation and calculations simpler
perhaps.
(a) n = 4 case. w = e2i/4 = i =

F4

1.

1 1
1 1 i
=

2 1 1
1 i

1
1
1
1

1
i

1
i

Idea of FFT is to represent F as a product of matrix operations in which the rows of the
matrices contain only 2 non-zero elements, thereby reducing number of calculations.

F4

1 1
1 0 0
=
2 1 1
0 0

0 0
1
1 0
1 i

0 0
2 1
1 i
0

0 1
1 0
0 1
1 0

0
1

0
1

But whats the meaning of this decomposition? The usual explanation is in terms of
butterflies, but lets compute what the individual matrices in this product actually do to
an element of 2 (Z4 ):

a+c
a
1 0 1
0
1 b
b+d
0 1 0

ac
c
1 0 1 0
bd
d
0 1 0 1
1

(b) n = 8 case: w = e2i/8 = ei/4 so that

w2 = i,
Thus

w3 = iw,

w4 = 1,

1
1
1
w
i
1

1 i 1
1
i
1 iw
F8 =

3
1
1
1
( 2)

1 w i

1
i
1
i
1 iw

1 1 0
0
0 0
w
0
0
0
0
1

0
1 i
0 0 0
1
0
0 0
0 0 0
=
1
1
0
0
0 0

2
0 0 1 w 0 0

0 0 0
0
1 i
0 0 0
0
0 0

F8

1
0

0
1
0

1
2

0
0

0 1
1 0
0 0
0 0
0 1
1 0
0 0
0 0

w5 = iw,
1
iw
i
w
1
iw
i
w

0
1
0
0
0
1
0
0

1
0

0
1
0

1
2

0
0

0
0
0
1
0
0
0
1
0
0
1
0
0
0
1
0
0
1
0
0
0
1
0
0

1
1
1
1
1
1
1
1

0
0

iw

0
iw

1
w
i
iw
1
w
i
iw

0 0
0 0
0 i
1 0
0 0
0 0
0 i
1 0
0
0
1
0
0
0
1
0

0
0
0
1
0
0
0
1

w6 = i,
1
i
1
i
1
i
1
i

w7 = iw.

1
iw

iw

i
w

0
0

0
i

1
0
0
0
1
0
0
0

0
1
0
0
0
1
0
0

0
0
0
0

1
0

0
1
.
0
0

0
0

1 0
0 1