Introduction to

Telecommunication

The Fourier Transform

Part I - Definitions

Asst. Prof. Dr. Ertuğrul SAATÇI

Overview
 Today we will start to discuss the Fourier Transform by
examining its relationship to the Fourier Series.

 The Fourier Transform introduces the frequency domain which

is an equivalent but insightful means of representing signals.

 Reading
 2.1, 2.3

1
Lecture Objectives
 In this lecture we will
● define the Fourier Transform.
● relate it to the Fourier Series.
● discuss its characteristics.
● examine several examples.

Fourier Theory
 Two basic types of signals
● Periodic
 Power signals
● Aperiodic
 Energy signals

 Fourier Domain Options

● Fourier Series
 Representation valid for all time if signal is periodic.
 Representation is valid only over a certain interval for
aperiodic signals.
● Fourier Transform
 Applies directly to aperiodic signals.
 Requires introduction of the impulse for application to
periodic signals.

2
Limitations of the Fourier Series
 The continuous time Fourier Series (CTFS) is a useful analytical
tool but has limitations:
● It can represent periodic signals for all time and can
represent aperiodic signals for a finite time, but cannot
represent an aperiodic signal for all time.
● It inherently depends on the fundamental frequency (i.e.,
the observation interval) chosen – signals with different
fundamental periods must be converted to a common
observation interval.
 The Fourier Transform will overcome these limitations by
allowing us to represent periodic and aperiodic signals without
depending on the observation interval.

Illustrative Example
 To help us understand the relationship between the Continuous
Time Fourier Series and the Continuous Time Fourier Transform
consider the following signal with Tw=To/2.

To

-TW /2 TW /2
rect(t)

 t  nT0  
 t  2nTW 
1 x(t )  A  rect    A  rect  
n   TW  n   TW 
t
-1/2 1/2

3
 To

Illustrative Example -TW /2 TW /2

 We know that the Continuous Time Fourier Series is

 t0 T0
1 1 1
x(t )  
n 
cn e jn 2 f0t , cn 
T0  x(t )e  jn 2 f0t dt , f0  
T0 2TW
t0
TW / 2
A  e  jn 2 f0t 
TW /2
1

 jn 2 f 0t
cn  Ae dt   
T0 TW /2
T0   jn 2 f 0  T /2
W

 T 

A
jn 2 f 0T0

e jn 2 f0TW / 2  e  jn 2 f0TW /2 
A
jn2

2 j sin  n2 f 0 W 
 2 
 n
sin   
A  n A  2   A sinc  n 
 sin      
n  2 2  n 2 2
2

Illustrative Example
Reducing the Duty Cycle
 Now let us reduce the duty cycle such that
● Tw remains constant and To=10Tw
● The average power is kept constant (i.e., we increase the
amplitude by To to maintain constant average power).

 t  10nTW  A  n
x(t )  5 A  rect   cn  sinc  
n   TW  2  10 
 If we further reduce the duty cycle such that
● Tw remains constant and To=1000Tw
● The average power is kept constant (i.e., we increase the
amplitude by To).

 t  1000nTW  A  n 
x(t )  500 A  rect   cn  sinc  
n   TW  2  1000 

4
Illustrative Example
The Magnitude Response
 Plotting for To=2Tw and To=10Tw

To=2Tw cn , A  1 To=10Tw cn , A  1

A n A  n
cn  sinc   cn  sinc  
2 2 2  10 
Decreasing duty cycle slows the CTFS in frequency and
shows more detail of the underlying sinc function.

Illustrative Example
The Magnitude Response – cont.
 If we normalize the x-axis by the
To=2Tw fundamental period, we see that increasing
the period increases the sampling rate of
the underlying function.
 If we let the period go to infinity (i.e., we
let the square pulse train approach a single
pulse), the Fourier Series representation
approaches a continuous function – this is
the Fourier Transform.

To=10Tw To=1000Tw

5
The Frequency Domain
 The original signal x(t) is said to be in the time domain since its
argument t represents time.
 The Fourier Transform X(f) representation is said to be in the
frequency domain since its argument f represents frequency.
 Notes:
● Frequency is the reciprocal of time.
● The Fourier Transform is referred to as an analysis of the
signal x(t) since it extracts
 the frequency components of x(t) at each value of f.
● The Inverse Fourier Transform is referred to as synthesis
since it recombines
 the components X(f) to obtain the original signal x(t).
● The physical meaning of X(f) depends on the meaning of
x(t). If x(t) has units of volts, X(f) has units volts/Hz.
 Thus it represents how much of the overall voltage
signal is present at each frequency.

The Frequency Domain

 We can think of the Fourier Transform and the Inverse Fourier
Transform as means for moving between the time and
frequency domains.
 Note that no information is lost in the transformation and
 both are equivalent representations of a signal.

6
The Frequency Domain
Further Notes
 The function X(f) is also sometimes referred to as the
amplitude spectral density or the spectrum of x(t).

 We often represent Fourier Transform pairs using the notation

x(t )  X ( f )
and we refer to x(t) and X(f) as a Fourier Transform pair.

 Sometimes the Fourier Transform is defined in terms of radian

frequency (angular frequency):
 
1
 x(t )e  jt dt  X ( )e
jt
X ( )  x(t )  d

2 

Example 2.1
 Consider the rectangular pulse of duration T and amplitude A
as shown in the figure below.
 1 1
1,  t 
rect  t   
2 2
0, t   1 or t  1
 2 2
t 
g  t   A rect  
T 
Find the Continuous Time Fourier Transform.
 T /2

 g (t )e 
 j 2 ft
G( f )  dt  Ae  j 2 ft dt
 T /2

7
Example 2.1 – cont.
T /2
T /2
 e  j 2 ft  A
G( f )  A  e  j 2 ft
dt  A    e j 2 fT / 2  e  j 2 fT /2 
T /2   j 2 f  T /2 j 2 f
A A
 2 j sin  2 fT / 2   sin  fT 
j 2 f f
sin  fT 
 AT  ATsinc  fT 
 fT

 t  
 ATsinc  fT 

FT
A rect   
 T

Fourier Transform Pair

Example 2.1 – cont.

t 
A rect   ATsinc  fT 
T 

8
Example 2.1 – cont.

t Tsinc  fT 
rect  
T 

Example 2.1 – cont.

Time vs. Frequency
g(t) Expanding a G(f)
T = 0.01 signal in time
makes it narrow
in frequency.

Faster changes
require higher
frequency
components.
t (sec) f (Hz)

T = 0.1

t (sec) f (Hz)

9
Example 2.1 – cont.
Time vs. Frequency
 As mentioned earlier, time and frequency are reciprocal.

 If a function speeds up in time, it slows down in frequency.

● If a signal changes rapidly it requires more high frequency
components.
● Signals which change rapidly in time are said to have a
large bandwidth (a measure of the frequency content).

 If a function slows down in time, it speeds up in frequency.

● If a signal changes slowly in time it requires less high
frequency components and more low frequency
components.
● Signals which change slowly in time are said to have a
small bandwidth.

Definitions of Bandwidth for Baseband

Signals
 Bandwidth is a term used to describe a positive frequency
range over which the signal has significant content. There are
various definitions for bandwidth including:
● Absolute Bandwidth (Babs)
 Defined as B where W ( f )  0, f  B
● 3-dB Bandwidth (half-power bandwidth - (B3dB))
 Defined as B where
2
W ( f ) max
W( f )  f B
2
,
● X-dB Bandwidth 2
 Defined as B where
20 log10  W ( f )    20 log10  W ( f ) max   X  , f B
● First Null Bandwidth (Bfirst null)
 For baseband systems this is equal to the frequency of
the first null in the spectrum.

10
Bandwidth
Baseband Signal

Bandwidth
Bandpass Signal

11
Example 2.2
 Find the 20dB
bandwidth,
the absolute
bandwidth and
the first-null
bandwidth.
 B20dB=2.75kHz
 Babs=infinite
 Bnull=1kHz

Example 2.3
 Determine the first-
null bandwidth,
absolute bandwidth
and 20dB
bandwidth of the
signal.
 B20dB=1.35kHz
 Babs=infinite
 Bnull=1.4kHz

12
Example 2.4
FT of Exponential
 A truncated decaying exponential is defined mathematically
using the unit step function u(t). Find the Continuous Time
Fourier Transform of this signal.

1, t  0
1

u (t )   , t  0
2
0, t  0
e  at t0
g (t )  e  at u (t )  
 0 t0

Example 2.4
FT of Exponential – cont.
  

 g (t )e e dt   e  ( a  j 2 f )t dt
 j 2 ft  at  j 2 ft
G( f )  dt  u (t )e
  0

 ( a  j 2 f ) t 
 e  1
   
 a  j 2 f  0 a  j 2 f
1  2 f 
G( f )  , arg  G ( f )    tan 1  
a 2   2 f   a 
2

 Fourier Transform pair for the decaying exponential pulse is

1
e  at u (t ) 
a  j 2 f

13
Example 2.4
FT of Exponential – cont.

As time waveform decreases more slowly, the

more low frequency content in the wave.

Example 2.4
FT of Exponential – cont.

1
G( f )   2 f 
arg  G ( f )    tan 1  
a 2   2 f 
2
 a 

14
Approach to Finding Fourier Transform
Pairs
 We could continue to find transform pairs according to the
definition, but this is inefficient.

 In general, we compile a table of known transform pairs.

 We also compile a table of simple rules for modifying transform

pairs (i.e., properties).

 Using the known pairs and transform properties we can find

most transforms needed.

Summary
 In this lecture we have discussed a vital tool in communication
system analysis termed the Fourier Transform.

 The Fourier Transform is useful for providing a frequency

domain representation of periodic and aperiodic signals that is
valid for all time.
 We will examine periodic signals next week.

 The Fourier Transform is an incredibly useful tool in many fields

of engineering.

 Understanding the relationship between time and frequency is

perhaps one of the most important concepts in this course.

15
Introduction to
Telecommunication

The Fourier Transform

Part II - Properties

Asst. Prof. Dr. Ertuğrul SAATÇI

Overview
 Today we continue our discussion of the Fourier Transform (FT)
by discussing properties of the FT.

 In EE550 most of these properties were presented and proven.

We will typically present them without proof.

 We will also go through examples using tables of pairs and

properties to find the Fourier Transform of arbitrary time
signals.

 Reading
 2.2

16
 Lecture Objective
 Today you will learn about properties of the Fourier Transform
(FT)
 that allows you to find FT of a large class of signals without
actually going through the FT integral.

Fourier Transform Properties

Conjugation x*  t   X *   f 
Linearity  x t    y t    X  f   Y  f 
Time  shifting x  t  t 0   e  j 2 f t X  f 
0

Frequency  shifting e j 2 f t x  t   X  f  f 0 
0

Time reversal x  t   X   f 

 x  t    j 2 f  X  f 
d
Time  differentiation
dt
t
1
Time  integration  x   d  j 2 f X  f  , If X (0)  0
1  f 
Time / freq  scaling x  at   X 
a a
Multiplication x t  y t   X  f  *Y  f 
Convolution x t  * y t   X  f  Y  f 

17
Linearity
 If
z t    x t    y t 
 Then
 
Zf   z  t  e  j 2 ft dt    x  t    y  t   e
 j 2 ft
dt
 
 
  x t  e  y t  e dt   X  f    Y  f 
 j 2 ft
dt    j 2 ft

 

 x t    y t    X  f   Y  f 

Time Shifting
 Let
z  t   x  t  t0 
 Then
 
Zf   z  t  e  j 2 ft dt   x t  t  e 0
 j 2 ft
dt ,   t  t0
 
 

 x   e  x   e d  e  j 2 f t0 X  f 
 j 2 f   t0   j 2  f t0
 d  e  j 2 f 

 

x  t  t 0   e  j 2  f t0 X  f 

18
Frequency Shifting
 Let
z  t   e j 2 f 0 t x  t 
 Then
 
Zf   z  t  e  j 2 ft dt   e 
x  t  e j 2 ft dt
j 2 f 0 t

 


 x t  e dt  X  f  f 0 
 j 2  f  f 0  t



e j 2 f 0 t x  t   X  f  f 0 

Time Scaling
 Let
z  t   x  at 
 Then the Fourier Transform is
 
Zf   z  t  e  j 2 ft dt   x  at  e
 j 2 ft
dt ,   at
 
  
1  j 2 f 1 f
 j 2  1 f 
  x   e a
d   x   e a
d  X  
a  a  a a

1  f 
x  at   X 
a a

19
Time Scaling
Interpretation
 Scaling a signal in time by α scales the Fourier transform (i.e.,
the signal in frequency) by 1/α.

 Does this make sense? Recall our previous discussion that time
and frequency are reciprocal.

 Let’s assume that α>1.

 Scaling a signal in time by α speeds the signal up in time.
• The resulting transform is scaled by 1/α which slows the
transform down in frequency – this means that more high
frequency values are present to accomplish faster changes.

 Scaling a signal in time by 1/α slows the signal down in time.

• The resulting transform is scaled by α which speeds the
transform up in frequency – this means that more low
frequency values are present to account for slower changes.

Convolution and Multiplication Properties

 Convolution property:

x t  * y t   X  f Y  f 
 Multiplication property:

x t  y t   X  f  *Y  f 

 Thus, convolution in the time domain results in multiplication in

the frequency domain while multiplication in the time domain
results in convolution in the frequency domain.

 These properties can greatly simplify some system analysis.

20
Convolution and Multiplication Properties
Convolution in Time
 Let
z t   x t  * y t 
 Then  
Zf   z  t  e  j 2 ft dt    x  t  * y  t  e
 j 2 ft
dt
 

   j 2 ft
   x   y  t    d  e dt , Changing the order of integration


  
 x 
 
   y  t    e  j 2 ft
dt  d   x   e  j 2 f  Y  f  d
 

F  y  t  

Convolution and Multiplication Properties

Convolution in Time – cont.

Zf   x   e
 j 2 f 
Y  f  d


 Y  f   x   e j 2 f  d  Y  f  X  f 


x  t  * y  t  
F
 X  f Y  f 

21
Convolution and Multiplication Properties
Multiplication in Time
 Let
z t   x t  y t 
 Then  
Zf   z  t  e  j 2 ft dt   x t  y t  e
 j 2 ft
dt
 

   j 2 ft
 x 
 
t   Y    e j 2 t
d  e dt , Changing the order of integration


  
  Y     x  t  e
 j 2  f    t
dt  d    Y   X  f    d
 

  

X  f  

Convolution and Multiplication Properties

Multiplication in Time – cont.


Zf   Y   X  f    d


 X  f  *Y  f 

x  t  y  t  
F
 X  f  *Y  f 

22
Time Differentiation
 Using the Fourier Transform representation of x(t) and taking
the derivative of x(t) with respect to t

x t    X  f e
j 2 ft
df


d   

d
 x  t   
dt  
X  f  e j 2 ft
df     j 2 f  X  f  e
j 2 ft
df
dt  
 F 1  j 2 f X  f 
 Thus,

d
dt
 x  t   j 2 f X  f 

Modulation
 A common operation in communication systems is modulation
or the multiplication of a signal by a high frequency sinusoid:

z  t   x  t  cos  2 f ct 
 We can find the Fourier Transform of z(t) using the
multiplication property.
Z  f   X  f  * F cos  2 f c t 
1 1 
 X  f  *    f  fc     f  fc 
2 2 
 Using the sifting property of the impulse
1 1
Zf  X  f  fc   X  f  fc 
2 2

23
 Example 2.5
Radio Frequency (RF) Pulse
 Consider the pulse signal g(t) shown in the figure below, which
consists of a sinusoidal wave of unit amplitude and frequency
fc, extending in duration t=-T/2 to t=T/2. This signal is
sometimes referred to as an RF pulse when the frequency fc
falls in the radio-frequency band. Find the Fourier Transform of
this signal.

t 
g  t   rect   cos  2 f ct 
T 

Example 2.5 – cont.

t 
 Find the Fourier Transform of the signal g  t   rect   cos  2 f c t  .
Recall the modulation property: T 
1 1 t 
Zf  X  f  f c   X  f  f c  ; also rect    Tsinc  f T 
2 2 T 
Thus, we can write directly

sinc   f  f c  T   sinc   f  f c  T 
T T
G f  
2 2

24
 Example 2.5 – cont.

x  t   rect  t  X  f   sinc  f 

Xf

Example 2.5 – cont.

x  t   rect  t  cos  200 t 

1
sinc  f  100 
2

1 1
X  f   sinc  f  100   sinc  f  100 
2 2

25
 Parseval’s Theorem
(=Rayleigh’s Energy Theorem)
 While the time domain signal x(t) and the frequency domain
signal X(f) appear quite different they do have the same
energy.

 That is
 

 x  t  dt   Xf
2 2
Ex  df
 

 In other words, it doesn’t matter whether we calculate the

energy of a signal in the time domain or in the frequency
domain, we get the same result.
 This should make sense since the two representations are
equivalent.

Example 2.6
 Find the energy in the pulse sinc(t).
The energy for any signal can be defined as


 x t 
2
E dt


Substituting for this signal gives

sin  t  sin 2  t 
  2 
sinc  t  dt 
Making a change of variables
  dt  
2
E dt ,
   t 
 
t 2
x  t
1 sin 2  x 

1

sin 2  x  2
 
  x 2
dx  2 
 0 x 2
dx 
 2
1

 /2

26
Example 2.6 – cont.
 Using our knowledge of Fourier Transforms and Parseval’s
Theorem we could have solved this very easily.
Parseval’s Theorem states
 

 x t   Xf
2 2
Ex  dt  df
 

Further, the Fourier Transform of sinc(t) is

sinc  t   rect  f 

The energy can easily be found to

be 1 from inspection of this plot.

Duality
 Due to the similar nature of the Fourier Transform and the
Inverse Fourier Transform, the CTFT exhibits the duality
property.

 The duality property says that if we have the Fourier Transform

pair

x t   X  f 

then we also have the Fourier Transform pair.

X t   x   f 

27
Example 2.7
 From our previous development we know the following FT pair:

rect  t  
F
 sinc  f 
The duality property says that

sinc  t  
F
 rect   f   rect  f 
 Check: Find the Inverse Fourier Transform for rect(f).

Example 2.7 – cont.

 Check: Find the Inverse Fourier Transform for rect(f).

 1/ 2
e j 2 f t
1/2
x t    X  f e
j 2 f t
df   e j 2 f t
df 
 1/2
j 2 t 1/2

e j t
e  j t
1 e j t
e  j t
sin  t 
     sinc  t 
j 2 t j 2 t  t 2j t

28
Example 2.7 – cont.
 More generally, consider the function X(f)=rect(f/B):

 B/2
e j 2 f t
B /2
x t    X  f e
j 2 f t
df   e j 2 f t
df 
  B /2
j 2 t  B /2

e j tB
e 1 e j tB j tB
e  j tB
sin  tB 
     Bsinc  tB 
j 2 t j 2 t  t 2j t

Example 2.8
 From the table we know the following FT pair:

1  f 
The duality property says that

 t   1
 Check: Find the Fourier Transform for δ(t).

 
Xf   x t  e
 j 2 f t
dt    t  e
 j 2 f t
dt  e  j 2 f t 1
t 0
 

29
Example 2.9
 Using Parseval’s Theorem, determine the energy of

x  t   cos  2 f 0t 
 From Parseval’s Theorem (using frequency domain integration):
  2
1 1
 X  f  df     f  f 0     f  f 0  df
2
Ex 
 
2 2

1 1 
   4   f  f   4   f  f  df 
2 2
0 0

 We can double check the result using time domain integration:
  

 x t   cos  2 f 0t  dt   cos  2 f t  dt  
2 2
Ex  dt  2
0
  

Example 2.9 – cont.

How do we evaluate δ2(t)?
 How do we evaluate δ2(t)?
  2
1  t 
   t  dt  lim 
2
 rect    dt
a 0
 
a  a 
a/2
1 1 1
 lim
a 0 a 2
a/2
 dt  lim
a 0 a 2
a  lim  
a 0 a

 Thus we have the following:



 g  t    t  t  dt  g  t 

0 0

 g  t    t  t  dt  
2
0


30
 Example 2.10
Triangular Pulse
 Use the integration property to
determine the Fourier Transform
of triangular pulse z(t)=tri(t).

 Taking the derivative of tri(t) twice we have

 1  1
d
tri  t   rect  t    rect  t  
dt  2  2

Example 2.10 – cont.

 Taking the derivative of tri(t) twice we have

 1  1
d
tri  t   rect  t    rect  t  
dt  2  2
d   1  1  d
rect  t    rect  t    
dt   2  2   dt

u  t  1  u  t    u  t   u  t  1  
   t  1    t     t     t  1 
   t  1  2  t     t  1
F   t  1  2  t     t  1  e j 2 f  2  e  j 2 f  2 cos  2 f   2

X f 

31
 Example 2.10 – cont.
 From the integration property we have

1 1
Zf  Xf  2 cos  2 f   2 
 j 2 f   j 2 f 
2 2

2 2 sin 2  f 
 1  cos  2 f    2sin 2  f  
 j 2 f   j 2 f   f 
2 2 2

sin  f  sin  f 
  sinc 2  f 
 f   f 
1
Note: X  0   0 and sin 2  x   1  cos  2 x  
2

Summary
 In this lecture we have examined several properties of the
Fourier Transform and examples which demonstrate their
usefulness.

 The properties are very similar to those for the Fourier Series.

 We will find these properties very useful in determining the

Fourier Transform of arbitrary signals.
 Using a simple table of Fourier Transforms and FT
properties, we can determine the FT of most signals of
interest.

32