ECE

EE 8443
3512 – PatternContinuous
– Signals: Recognition
and Discrete

LECTURE 10: FOURIER TRANSFORM PROPERTIES

• Objectives:
Linearity
Time Shift and Time Reversal
Multiplication
Integration
Convolution
Parseval’s Theorem
Duality

• Resources:
BEvans: Fourier Transform Properties
MIT 6.003: Lecture 8
DSPGuide: Fourier Transform Properties
Wiki: Audio Timescale ModificationISIP: Spectrum Analysis

URL:
The Fourier Transform

EE 3512: Lecture 10, Slide 2

Example: Cosine Function
See “The Fourier Transform”

EE 3512: Lecture 10, Slide 3

Example: Periodic Pulse Train

Note: Since this is a periodic signal, we

use a Fourier series to compose the
signal as a sum of complex exponentials.
Then we take the Fourier transform of
each complex exponential.

EE 3512: Lecture 10, Slide 4

Linearity
• Recall our expressions for the Fourier Transform and its inverse:

1

jt
x(t )  X ( j  ) e d (synthesis)
2 

1
X ( j )  X ( )   x(t ) e  jt dt (analysis)
T 
• The property of linearity:
F ax(t )  by (t )  aX ( j )  bY ( j )  ax(t )  by (t )  aX ( j )  bY ( j )
Proof:

1
F ax(t )  by (t )   ax (t )  by (t ) e  jt dt
T 
 
1 1
  ax (t ) e  jt dt   by (t ) e  jt dt
T  T 
1   1  
 a   x(t ) e dt   b   y (t ) e  jt dt 
 j t

T   T  
 aX ( j )  bY ( j )

EE 3512: Lecture 10, Slide 5

Time Shift
• Time Shift:
x(t  t 0 )  X ( j )e  jc0
Proof: 
1
Fx(t  t 0 )   x(t  t 0 ) e  jt dt
T 
make a change of variables : λ  t-t 0 , which implies t  λ  t 0

1
Fx(t  t 0 )   x( ) e  j (   t0 ) d
T 

1   jt0
   x ( ) e  j
d e
T   
 X ( j  ) e  j t 0
• Note that this means time delay is equivalent to a linear phase shift in the
frequency domain (the phase shift is proportional to frequency).
• We refer to a system as an all-pass filter if:
X ( j )  1 X ( j )  0
• Phase shift is an important concept in the development of surround sound.

EE 3512: Lecture 10, Slide 6

Time Scaling
• Time Scaling:
1 j
x(at )  X ( )
a a
Proof: 1

Fx(at )   x (at ) e  jt dt
T 
assume a  0, make a change of variables : λ  at , which implies t   / a, and dt  (1 / a )d
 
1  j ( ) 1
Fax(t )   x ( ) e a ( )d
T  a
1 1 

 ( )   x ( ) e  j ( / a ) 
d 
a  T  
1 j
 ( )X ( )
a a
• Generalization for a < 0 , the negative value is offset by the change in the
limits of integration.
• What is the implication of a < 1 on the time-domain waveform? On the
frequency response? What about a > 1?
• Any real-world applications of this property? Hint: sampled signals.78
EE 3512: Lecture 10, Slide 7
Time Reversal
• Time Reversal:
x(t )  X ( j )

Proof:
1 j
Fx(t )  X( )  X (  j )
a a a  1
We can also note that for real-valued signals:
X ( j )  X ( j ) X ( j )
 X ( j ) X ( j )  X * ( j ) (complex conjugate)

• Time reversal is equivalent to conjugation in the frequency domain.

• Can we time reverse a signal? If not, why is this property useful?

EE 3512: Lecture 10, Slide 8

Multiplication by a Power of t
• Multiplication by a power of t:
n
n n d
t x(t )  ( j ) X ( j )
d n

1
Proof: X ( j )   x (t ) e  jt dt
T 
differenti ate with respect to  :

dX ( j ) 1
  ( jt ) x(t ) e  jt dt
d T 
multiply by j :

dX ( j ) 1
j  ( j )  ( jt ) x(t ) e  jt dt
d T 

1
  (t ) x (t ) e  jt dt  Ft x(t )
T 

• We can repeat the process for higher powers of t.

EE 3512: Lecture 10, Slide 9

Multiplication by a Complex Exponential (Modulation)
• Multiplication by a complex exponential:
x(t )e jt  X ( j (   0 )) for any real number  0
Proof:

1

F x(t)e jω0t
   x (t )e jω0t e  jt dt
T 

1
  x (t )e  j ( ω0 )t dt
T 
 X ( j (  ω0 ))
• Why is this property useful?
• First, another property:
e j0t  e  j0t 1
x(t ) cos( 0 t )  x(t )[ ]  [ X ( j (   0 ))  X ( j (   0 ))]
2 2

• This produces a translation in the frequency

domain. How might this be useful in a
communication system?

EE 3512: Lecture 10, Slide 10

Differentiation / Integration
• Differentiation in the Time Domain:
dn n
n
x (t )  ( j  ) X ( j )
dt
• Integration in the Time Domain:
t
1
 x (  ) d 

j
X ( j )  X (0) ( )

• What are the implications of time-domain differentiation in the frequency

domain?
• Why might this be a problem? Hint: additive noise.
• How can we apply these properties? Hint: unit impulse, unit step, …

EE 3512: Lecture 10, Slide 11

Convolution in the Time Domain
• Convolution in the time domain:

x(t )  h(t )  X ( j ) H ( j )
• Proof: 
x(t )  h(t )   x( )h(t   )d


 
  j t
Fx(t )  h(t )     x( )h(t   )d e dt
    

 
  x( )   h(t   )e dt  d
 j t

   
change of variables :   t    d  dt

 
  x( )   h( )e  j (   )
d  d
   
   

   x ( )e  j
d    h( )e  j
d 
   
 X ( j ) H ( j )

EE 3512: Lecture 10, Slide 12

Other Important Properties
• Multiplication in the time domain:

1 1
x(t )  y (t ) 
2
[ X ( j )  Y ( j )] 
2  X ( )Y (   )d


• Parseval’s Theorem:
 
1 2
 
2
x (t ) dt  X ( j ) d

2 

• Duality:
X (t )  2 x( )

• Note: please read the textbook carefully for the derivations and interpretation
of these results.

EE 3512: Lecture 10, Slide 13

Summary

EE 3512: Lecture 10, Slide 14