Topic 5

Fourier Transform
Introduction
Transform: a mathematical conversion from one way of
thinking to another to make a problem easier to solve

problem in solution in
original original
way of solution inverse way of
transform in transform
thinking transform thinking
way of thinking
Introduction

problem in solution inverse solution

time Fourier in time
in Fourier
domain transform domain
Frequency transform
domain

• Other transforms
• Laplace
• z-transform
• wavelet transform
Introduction
• Fourier transform (FT) is a mathematical transform that
decomposes a function (or a signal) into its
constituent frequencies, such as the expression of a musical chord
in terms of the volumes and frequencies of its constituent notes.

• The term Fourier transform refers to both the frequency

domain representation and the mathematical operation that
associates the frequency domain representation to a function of
time.

• FT of a function of time is a complex-valued function of frequency

o Magnitude (absolute value) represents the amount of that frequency present
in the original function
o Argument (Phase) is the phase offset of the basic sinusoid in that frequency
Introduction
• Fourier transform is not limited to functions of time, but
the domain of the original function is commonly referred to as
the time domain.

• There is also an inverse Fourier transform (IFT) that

mathematically synthesizes the original function from its
frequency domain representation

• Linear operations performed in one domain (time or frequency)

have corresponding operations in the other domain, which are
sometimes easier to perform.
o The operation of differentiation in the time domain corresponds to
multiplication by the frequency, so some differential equations are easier to
analyze in the frequency domain.
o Convolution in the time domain corresponds to ordinary multiplication in the
frequency domain.
Introduction
• After performing the desired operations, transformation of the
result can be made back to the time domain. Harmonic analysis is
the systematic study of the relationship between the frequency and
time domains, including the kinds of functions or operations that
are "simpler" in one or the other, and has deep connections to
many areas of modern mathematics.

• Functions that are localized in the time domain have Fourier

transforms that are spread out across the frequency domain and
vice versa, a phenomenon known as the uncertainty principle.
Definition
From Fourier series (FS), we can find the frequency content of a
signal.
FS has 2 limitations
•It is valid only for periodic signal
•The frequency content is only discrete

If we have general class of signals, either periodic, aperiodic or of

finite duration whose periodicity can’t be determined, the question
arises about the way to determine its frequency content. The answer
lies in Fourier Transform (FT) by which we convert it to frequency
domain and hence know frequency content of any signal, of any
duration, either periodic or not.
 Definition
Let us take a finite duration signal x(t) as shown below.
Let’s make it periodic ~
x (t ) with a period T by repeating x(t)
Here, ~x (t )  x(t ) as T   .
Fourier series for ~x (t ) is

~
x (t )   C n e jn0t
n 
(1)
T
2
1

 jn0t
Cn  x (t ) e dt
T T (2)

2

Now as T   ,
So we can write

2 1 1
x(t )e dt d 
 j t
0   0 n0   Cn  X ( )d where
T 2 
2
1 0 d 
 
T 2 2 X ( )  x(t )e  jt dt is the Fourier transform of x(t)


Putting the value of C in (1), we find

n

 
1 1
x(t )   X ( )de jt  X ( ) e jt
d
n  2 2 
Definition
FT pairs:

X ( )  x(t )e  jt dt FT


1

jt
x(t )  X ( ) e d IFT
2 

x(t )  X ( ) or x(t )  X (j )

Examples
Rectangular pulse
t
x(t ) rect  
 

 /2
X ( )  1e  jt
dt 
j

1  jt
e 
 /2
 /2 
j
e 
1  j / 2
 e j / 2 
 /2

  
sin  
2   2   2    
 sin     sin c 
  2   2   2 
2
Examples
Exponential function
 at 1
x(t ) e u (t ) X ( ) 
a  j
 jat 1
x(t ) e u (t ) X ( ) 
j (  a )
Properties
Application: Filter Design
Let us consider a low-pass RC filter
.
dy (t ) dy (t )
x(t ) iR  y (t ) RC  y (t )  i C
dt dt
Taking FT on both sides
X ( ) RCjY ( )  Y ( ) (1  j RC) Y( )
Y ( ) 1
H ( )  
X ( ) 1  j RC
1
H ( ) 
1   RC 
2

The cut-off frequency of the filter occurs at

half-power / 3-dB point where 1
H ( ) 
2
1
Putting this value, we get  c 
RC
Similarly for HPF
jRC
H ( ) 
1  jRC
 DTFT
DTFS is applicable for periodic signal. If periodicity is absent or not known, as in
the case of finite-duration signal, how to do Fourier analysis? Let x(n) is aperiodic.
We can construct periodic xN0(n) by repeating x(n).

lim x ( n )  x n 
N 0   N0

x N 0 ( n)   r
D e
r  N0
jr0 n


1
Dr 
N0
 x
n 
( n ) e  jr0 n

1  

Dr  Lim
N0   N 0
 x ( n )e
n 
 jrn
Lim 
  0 n 
x(n)e  jrn
2

1 
1 X ( )   x(n)e  jn

2
 x ( n )e
n 
 j n
d  X ( )d
2 n 
DTFT

X ( )   x(n)e  jn DTFT
n 
x(n)  X ( )
1
x ( n) 
2 2
X ( ) e jn d Fourier Integral and IDTFT
DTFT of Elementary signals
n 1 Put z=ejω in ZT to get FT
a u ( n) 
1  ae  j
1 
u ( n)  X ( )   x(n)e  jn
1  e  j n 

j n 1
e u ( n) 
1  e j (   )
1  cos(   )
cos n u (n) 
21  cos   
sin(   )
sin n u (n) 
21  cos   
DTFT Properties
Periodic with a period of 2π:
X(ω+2π) = X(ω)
X ( )  X (  ) X ( )  X (  )
Magnitudes are even symmetric
while phases are odd

Time inversion results frequency x (  n)  X (   )

inversion

Time shifting results multiplication: x(n  k )  e  jk X ( )

x1 (n)  x 2 (n)  X 1 ( ) X 2 ( )
Convolution in time and frequency 1
domains: x1 (n) x 2 (n)  X 1 ( )  X 2 ( )
2
DTFT
Example: Find the response of the DT system with x(n) = (0.8)nu(n)
and h(n) = (0.5)nu(n)

n 1
a u ( n) 
1  ae  j

1 e j 1 e j
X ( )   j
 j H ( )   j
 j
1  0.8e e  0.8 1  0.5e e  0.5
5 j 8 j
j 2  e e
e 3 3
Y ( )  X ( ) H ( )   j  j
( e j  0.8)(e j  0.5) (e  0.5) (e  0.8)

 5 n 8 n
 y (n)   (0.5)  (0.8)  u (n)
 3 3 
DFT

X ( )   x(n)e  jn
n 
X(ω) is periodic over 2π. Let X(ω) is sampled every δω and for 0≤ ω ≤ 2π we have N
samples. Then δω = 2π/N. So, for ω = 2πk/N
 2  
X k    x(n)e  j 2nk / N k = 0, 1, 2, ….., N-1
 N  n 
1 N1 2N  1
.....   x ( n )e
n  N
 j 2nk / N
  x ( n )e
n 0
 j 2nk / N
  x
n N
( n ) e  j 2nk / N
 ...
 lN  N  1
  x ( n ) e  j 2nk / N

l  n lN
Changing the limit of inner sum n=0 to n=N-1 and
interchanging the order of summation 
x p (n)   x(n  lN )
 2  N  1 
X k    x(n  lN )e  j 2nk / N l 

 N  n 0 l  periodic repetition of x(n) with period N

when N ≥ L (data length)
DFT
N1
X (k )  x(n)e  j 2nk / N N point DFT
n 0
N1
1
x ( n) 
N
 X
k 0
( k ) e j 2nk / N
IDFT

Example: x(n) = {1 0 0 1} 3
X (k )  x(n)e  j 2nk / 4
n 0
Ans: X(k) = {2 1+j1 0 1-j1}

Draw magnitude and phase spectra. Against point N/2, magnitudes

are even symmetry while phases are of odd symmetry.
Putting the values of k as 0, 1, 2, 3

Check! X(0)= Sum of

data
DFT Properties
Periodicity: If x(n+N) = x(n), then X(k+N) = X(k) for all k

Linearity: a1x1(n) + a2x2(n)  a1X1(k) + a2X2(k)

Real valued sequence: If x(n) is real, X(N-k) + X*(k) = X(-k)

N1
2nk
Real and even sequence: X (k )  x(n) cos  DCT
(x(n) is real and even) n 0 N
Real and odd sequence: N1
2nk
(x(n) is real and odd) X (k )  j  x(n) sin
n 0 N
DFT Multiplication
N1 N1
X 1 (k )  x1 (n)e  j 2nk / N
X 2 (k )  x 2 (n)e  j 2nk / N X 3 (k )  X 1 (k ) X 2 (k )
n 0
n 0
1 N1 j 2 mk / N
1 N1 k = 0, 1, 2, . . . . ., N-1
x3 (m) 
N
X
k 0
3 ( k )e 
N
 X (k )X
k 0
1 2 ( k )e j 2 mk / N

N1 N1 N1 j 2 mk / N

1   j 2 nk / N    j 2 lk / N 
  1   2
1 N1 N1
 x ( n ) e x (l ) e e  N  1 j 2k ( m  n  l ) / N 
N    x1 (n) x2 (l )   e 
k 0  n 0   l 0  N n 0 l 0  k 0 

When m-n-l = pN , the inner sum is N and l=((m-n))N

Otherwise the inner sum is 0. So,
N1
x3 (m)  x1 (n) x 2 ((m  n)) N m = 0, 1, 2, . . ., N-1
n 0

Circular Convolution
Multiplication of 2 DFT sequences is equivalent to circular convolution
of the 2 sequences in time domain
Circular Convolution
Example: Using circular convolution, determine the response of an
LTI system with x(n) = { 2, 1, 2, 1) and h(n) = {1, 2, 3, 4}.
Check your results using DFT.

N1
y(m)  x(n) h((m  n)) N
n 0

Using circular convolution, we get y(n) = {14,16,14,16}

Using DFT,
X(k) = {6, 0, 2, 0} and H(k) = {10, -2+j2, -2, -2-j2}

Y(k) =X(k)H(k)= {60, 0, -4, 0}

Using IDFT
y(n) = {14,16, 14, 16}
DFT Application

Filtering
Filtering of long data sequence:
overlap-save method and overlap-add method
Frequency analysis of signals
……
Thank you