Signals and Systems Fall 2021

Continuous-Time Fourier
Transform
Prof. Adnan Kavak
Computer Engineering
Kocaeli University

Textbook: Linear Systems and Signals, B.P. Lathi

This slides were used in EE313 course at UT Austin, which are adopted and used in
this course with Courtesy of Prof. Brian L. Evans
 Fourier Integral
 
G  f    g t  e  j 2 f t
dt X     xt  e  j  t dt
 
 1 
g t    G  f  e j 2 ft
df xt    X   e j t
d
 2  
Communicat ion Systems Signal Processing

• Conditions for Fourier transform of x(t) to exist

x(t) is single-valued with finite maxima and minima in
any finite time interval
x(t) is piecewise continuous; i.e., it has a finite number of
discontinuities in any finite time interval
x(t) is absolutely integrable 
 g t  dt  
• Conditions not obeyed for cos(t), sin(t) and u(t)


We’ll find ways to define Fourier transforms for them

Continuous-Time Fourier Transform
• Forward direction

Analysis: extracts spectrum information from the signal x(t)

• Inverse direction

Synthesis: creates time-domain signal x(t) from its spectral info

 Fourier Transform Pairs
• From the sifting property of the Dirac delta,

F  t     t  e  j  t dt  e  j 0  1

• Consider a Dirac delta in the Fourier domain
1  1 j 0t 1
F    
1
    e j  t d  e 
2  2 2
Using linearity property, F{ 1 } = 2p d(w)

x(t) = 1 X(w) = 2 p d(w)

1 F (2p)

t w
0 0
12 - 4
(2p) means that the area under the Dirac delta is (2p)
 Fourier Transform Pairs

  t   j t  /2
F     rect  e dt   e  jt dt

   / 2

     
sin   sin   

1  j / 2
e  e j / 2   2 
2 
  2    sinc   
 
j      2 
 
 2 

f(t) F(w) t
1 F
w
t
-t/2 0 t/2 -6p -4p -2 p 0 2p 4p 6p
t t t t t t
 Fourier Transform Pairs
1  1 j0t
F    0  
1
    0  e d 
jt
e
2   2
1 j 0 t
e     0  or e j0t  2    0 
2
1 j0t

Since cos0t   e  e  j0t
2

cos0t       0      0 

f(t) F(w)
(p) (p)
F
t w
0 -w0 0 w0
 Causal Exponential Signal
• Formula: w = -8 : 0.01 : 8;
H = 1 ./ (1 + j*w);
Hmag = abs(H);
Hphase = phase(H);
figure;
plot(w, Hmag);
title('Magnitude Response');
ylim( [-0.0 1.1] );
figure;
plot(w, Hphase);
title('Phase Response');

Decays if Oscillates
Re{a} > 0
Magnitude-Phase form

a=1
 Fourier Transform Properties

• What system properties does it possess?

 Memoryless (in fact requires infinite memory)
 Causal
 Linear
 Time-invariant (doesn’t apply)
• What does it tell you about a signal?
Answer: Measures frequency content
• What doesn’t it tell you about a signal?
Answer: When those frequencies occurred in time
 Duality
• Forward/inverse transforms are similar
 1 
F     f t  e  j t
dt f t    F   e j t
d
 2
f t   F   F t   2 f   
• Example: rect(t/t)  t sinc(w t / 2)
Apply duality t sinc(t t/2)  2 p rect(-w/t)
rect(·) is even t sinc(t t /2)  2 p rect(w/t)
f(t) t
F(w)
1

w
t
-t/2 0 t/2 -6p -4p -2 p 0 2p 4p 6p
t t t t t t
 Scaling

• Given f t   F   and that a  0

1  
f at   F 
a a
|a| > 1: compress time axis, expand frequency axis
|a| < 1: expand time axis, compress frequency axis
• Extent in time domain is inversely proportional
to extent in frequency domain (a.k.a bandwidth)
f(t) is wider  spectrum is narrower
f(t) is narrower  spectrum is wider
 Shifting in Time
• Shift in time f t  t0   e  j  t0 F  
Does not change magnitude of the Fourier transform
Shifts phase of Fourier transform by -wt0
(so t0 is the slope of the linear phase)
• Derivation 
F  f (t  t0 )   f (t  t0 )e  j  t dt


Let u = t – t0, so du = dt and integration limits stay same

 
F f (t  t0 )   f (u )e  j  ( u t0 )
du  e  j  t0
 f (u )e  j  u du  e  j  t0 F ( )
 
 Sinusoidal Amplitude Modulation
y t   f t  cos 0t 
Multiplica tion in the time domain is
convolutio n in the frequency domain :
1
Y    F       0       0 
2
Recall that

x t   t      x t   d  x t 


x t   t  t0       t0 x t   d  x t  t0 


So,
1 1
Y    F    0   F   0 
2 2
 Sinusoidal Amplitude Modulation
F(w)
• Example: y(t) = f(t) cos(w0 t)
1
f(t) is an ideal lowpass signal
Assume w1 << w0
w
1/2 F(w+w )
Y(w) 1/2 F(w-w )
-w1 0 w1
0 0
1/2

w
-w0 - w1 -w0 + w1 0 w 0 - w1 w0 + w1
-w0 w0

• Demodulation (i.e. recovery of f(t) from y(t)) is

modulation followed by lowpass filtering
• Similar derivation for modulation with sin(w0 t)
Frequency-shifting Property

e j 0 t f t   2 F   0 
e  j 0 t f t   2 F   0 
1
cos0t  f t   F   0   F   0 
2
1 1
cos0t  f t   F   0   F   0 
2 2
1
sin 0t  f t    jF   0   jF   0 
2
j j
sin 0t  f t   F   0   F   0 
2 2