You are on page 1of 287

# Fourier Representations of Signals and

## Linear Time-Invariant Systems

Chapter 3

Outline

Introduction
Complex sinusoids and frequency response of LTI
systems
Fourier representations for four classes of signals
Discrete-time periodic signals: Discrete-time Fourier
series
Continuous-time periodic signals: Fourier series
Discrete-time nonperiodic signals: Discrete-time
Fourier transform
Continuous-time nonperiodic signals: Fourier transform
Properties of Fourier representation
Linearity and symmetric properties
Convolution property

Research Laboratory

Outline

## Differentiation and Integration properties

Time- and Frequency-shift properties
Finding inverse Fourier transforms by using partialfraction expansions
Multiplication property
Scaling property
Parseval relationships
Time-bandwidth product
Duality
Exploring Concepts with MATLAB
Summary

Research Laboratory

Introduction

## Represents signals in terms of complex

sinusoids, we can

## Obtain an alternative expression for the I/P-O/P

behavior of an LTI system

## Represents an I/P signal as a weighted superposition of

complex sinusoids
The O/P of an LTI system is a weighted superposition of
the system response to each complex sinusoids

and systems

## The weight associated with a sinusoid of a given

frequency represents the contribution of that sinusoid to
the overall signal

Research Laboratory

Introduction

Fourier analysis

representations

## Discrete-time Fourier series (DTFS)

Fourier series (FS)
Discrete-time Fourier transform (DTFT)
Fourier transform (FT)

Research Laboratory

## Complex sinusoids and frequency

response of LTI systems

## Response of an LTI system to a sinusoidal I/P

The representation is obtained by using
convolution and a complex sinusoidal I/P signal

## Figure 3.1 (p. 196)

The output of a complex sinusoidal
input to an LTI system is a complex
sinusoid of the same frequency as
the input, multiplied by the
frequency response of the system.

Research Laboratory

## Complex sinusoids and frequency

response of LTI systems
Consider a discrete-time LTI system with impulse response h[n]
and unit amplitude complex sinusoidal I/P x[n] e jn , the O/P is
y[n]

h[k ]x[n k ]

h[k ]e j ( n k ) e jn

k
j

where H (e )

h[k ]e jk H (e j )e jn

## system. It is a function of frequency but not a function of time n

Research Laboratory

## Complex sinusoids and frequency

response of LTI systems
Consider a continuous-time LTI system with impulse response h(t )
and unit amplitude complex sinusoidal I/P x(t ) e jt , the O/P is

y (t ) h( ) x(t )d h( )e

j ( t )

d e

jt

h( )e j d

H ( j )e jt
where H ( j )

h( )e j d is the frequency-response of a

continuous-time system.
H ( j ) is a function of frequency but not a function of time t
Research Laboratory

## Complex sinusoids and frequency

response of LTI systems
Polar form:
b
c a jb c c e
, where c a b , arg{c} arctan{ }
a
H ( j ) H ( j ) e j arg{H ( j )} , where
j arg{c}

H ( j ) : magnitude response
arg{H ( j )}: phase response
y (t ) H ( j )e jt H ( j ) e j arg{H ( j )}e jt H ( j ) e j (t arg{ H ( j )})
the system modifies the amplitude of the I/P by H ( j )
and the phase by arg{H ( j )}
Research Laboratory

## Complex sinusoids and frequency

response of LTI systems
Ex. 3.1 RC circuit: Frequency response

t
RC

1
I/P: x(t ), O/P: y (t ), Impulse response: h(t )
e u (t )
RC
Find an expression for the frequency response and plot the
magnitude and phase response

## Figure 3.2 (p. 197)

RC circuit for Example 3.1.

Research Laboratory

10

## Complex sinusoids and frequency

response of LTI systems

)
( j
1
1
j
j
RC
RC
Sol. H ( j ) h( )e d [
e u ( )]e d
e
d

RC
0
RC
1
( j
)
1
1
1
1
1

RC

e
|0
(0 1)
1
1
RC ( j
RC ( j
( j RC 1)
)
)
RC
RC
1 j RC

( RC ) 2 1

1
RC
1
2
2
H ( j ) (
) (
)
,
2
2
2
( RC ) 1
( RC ) 1
( RC ) 1
arg{H ( j )} arctan{ RC}
Research Laboratory

11

## Complex sinusoids and frequency

response of LTI systems

## Figure 3.3 (p. 198)

Frequency response of the RC circuit in Fig. 3.2. (a) Magnitude response. (b) Phase response.

Research Laboratory

12

## Complex sinusoids and frequency

response of LTI systems
If ek is an eigenvector of a matrix A with eigenvalue k , then Aek k e k
We say that the complex sinusoid (t ) e jt is an eigenfunction of the
LTI system H associated with the eigenvalue H ( j ),
because satisfies H {e jt } H ( j )e jt H { (t )} (t )
i.e. H { .} and H ( . ) are different!!
H

## Figure 3.4 (p. 198)

Research Laboratory

13

[ n]

h[n]

x[k ] [n k ]

y[n]

## x[k ]h[n k ] x[n]* h[n]

x[n] ak e jk n

y[n] ak H (e jk )e jk n

Research Laboratory

14

(t )

h(t )

x(t ) x(t ) * (t )

x( ) (t )d

y (t )

## x( )h(t )d x(t )* h(t )

e jt

H ( j )e jt

x(t ) ak e jk t

y (t ) ak H ( jk )e jk t
k

## Q: What is the value of k , ak , and the range of k ?

Research Laboratory

15

## Complex sinusoids and frequency

response of LTI systems
Consider expressing the I/P to an LTI system as the weighted sum of
M

k 1

## If e jk t is an eigenfunction of the system with eigenvalue of H ( jk ),

then each term in the I/P,ak e jk t , produces an O/P term ak H ( jk )e jk t
Hence, we can express the O/P of the system as
M

y(t ) ak H ( jk )e jk t
k 1

## the operation of convolution, h(t ) * x(t ), becomes multiplication,

ak H ( jk ), because is expressed x(t ) as a sum of eigenfunctions
Research Laboratory

16

## Complex sinusoids and frequency

response of LTI systems

## This property is a powerful motivation for

representing signals as weighted superposition of
complex sinusoids
Rather than describing the signals behavior as a
function of time, the weights describe it as a function
of frequency

Research Laboratory

17

signals

## Discrete-time, nonperiodic signal

Research Laboratory

18

[ n]

h[n]

x[k ] [n k ]

y[n]

## x[k ]h[n k ] x[n]* h[n]

x[n] ak e jk n

y[n] ak H (e jk )e jk n

Research Laboratory

19

## Fourier representations for four classes

of signals

Research Laboratory

20

representations

## Consider representing a periodic signal as a

weighted superposition of complex sinusoids
Since the weighted superposition must have the
same period as the signal, each sinusoid in the
superposition must have the same period as the
signal
It implies that the frequency of each sinusoid must be
an integer multiple of the signals fundamental
frequency
0=2p/N, N=2p/0
0k0
NN/k

Research Laboratory

21

## Periodic signals: Fourier series

representations
A sinusoid whose frequency is an integer multiple of a
fundamental frequency is said to be a harmonic of the
sinusoid at the fundamental frequency.
Thus, e jk0t is the k -th harmonic of e j0t .

Research Laboratory

22

## Periodic signals: Fourier series

representations
If is x[n] a discrete-time signal with fundamental period N , then
we seek to represent x[n] by the DTFS,
x[ n] A[ k ]e jk 0 n ,

(3.4)

where
2p
0
is the fundamental frequency of x[n];
N
A[k ] is the weight applied to the k -th harmonic;
() denotes approximate value. (i.e. DTFS is not assumed to be existed yet)
The frequency of the k -th sinusoid in the superposition is k 0 .
Each of these sinusoids has a common period N .
Research Laboratory

23

## Periodic signals: Fourier series

representations
If is x(t ) a continuous-time signal with fundamental period T , we seek to
represents x(t ) by the FS,
x (t ) A[ k ]e jk0t ,

(3.5)

2p
is the fundamental frequency of x(t )
T
A[k ] is the weight applied to the k -th harmonic;
() denotes approximate value. (i.e. FS is not assumed to be existed yet)
where 0

## The frequency of the k -th sinusoid in the superposition is k0 .

Each of these sinusoids has a common period T
Research Laboratory

24

## f[n]= ejk0n(i.e., f[n]=f[n+N]),N=2p/0

f[k+N]=ej(k+N)0n=ejk0nejN0n=ejk0nej2pn=ejk0n=f[k]

## f(t)= ejk0t(i.e., f(t)=f(t+T)),T=2p/0

f[n+N]=ejk0(n+N)=ejk0nejk0N=ejk0nej2pk=ejk0n=f[n]

## f[x]=f[x+N], is a function of x with period N; x, N are integer

f(x)=f(x+T), is a function of x with period T

f(t+T)=ejk0(t+T)=ejk0tejk0T=ejk0tej2pk=ejk0t=f(t)

f[k]=ejk0t(i.e., f[k]!=f[k+N])

f[k+N]=ej(k+N)0t=ejk0tejN0t!=ejk0t=f[k], tejN0t!=1

Research Laboratory

25

## Periodic signals: Fourier series

representations
Consider the complex sinusoids
e j ( N k ) 0 n e jN 0 n e jk 0 n e j 2p n e jk 0 n e jk 0 n
e jk 0 n are N -periodic in the frequency index k , there are only N distinct
complex sinusoids of the form e jk 0 n
N 1

k 0

## In contrast, continuous-time complex sinusoids e jk0t with distinct

frequencies k0 are always distinct (+)0 = 0 0 0
Hence, Eq.(3.5) can be expressed as x (t )

Research Laboratory

26

A[k ]e jk0t

representations

## For periodic signal x[n]

k
2p
jk
n
2p
jk 0 n
N
0
, e
e
e
N

Fundamental period of

## ak e jk n with period N (i.e.,x[n] x[n N ])

2p
n
N
( )
k

jk 0 n
a
e
k
kZ

Let k k 0 , x[n]
e

jk 0 n

j ( k mN ) 0 n

ae

N
k
N
is least common multiple (LCM) of
(i.e., N )
k

jk 0 n

,
N 1

k 0

## A[k ] ... ak N ak ak N ...

Research Laboratory

27

k mN

## Periodic signals: Fourier series

representations
Goal : Find the A[k ] such that x[n] and x (t ) are good
approximations to x[n] and x(t ), respectively
Mean-square-error (MSE):
1
Discrete-time case: MSE
N

N 1

x[n] x[n]

n 0

2
1 T
Continuous-time case: MSE x(t ) x (t ) dt
T 0

## i.e. the DTFS and FS coefficients minimize the MSEs.

Research Laboratory

28

representations

## There is no restriction on the period of the sinusoids

used to represent a non-periodic signal. Hence, the
Fourier transform representations employ complex
sinusoids having a continuum of frequencies.
The signal is represented as a weighted integral of
complex sinusoids where the variable of integration is
the sinusoids frequency.

Research Laboratory

29

## f[n]= ejn(i.e., f[n]!=f[n+N])

f(T)=ej(+T)n=ejnejTn=ejn=f(), if T=2p

## f(t)= ejt(i.e., f(t)!=f(t+T))

f[n+N]=ej(n+N)=ejnejN!=ejn=f[n], , ejN!=1

## f[x]=f[x+N], is a function of x with period N; x Z, N is positive integer

f(x)=f(x+T), is a function of x with period T, x R, T is positive real

f(t+T)=ej(t+T)=ejtejT!=ejkt=f(t), , ejT!=1

f()=ejt(i.e., f()!=f(+T))

f(T)=ej(+T)t=ejtejTt!=ejt=f(), t, ejTt!=1

Research Laboratory

30

## Nonperiodic signals: Fourier-transform

representations
Continuous-time sinusoids are used to represent continuous signal in the FT.
(+) = , 2

## Continuous-time sinusoids with distinct frequencies are distinct, so the FT

involves frequencies from - to , as shown by the equation
1
jt
x (t )
X
(
j

)
e
d,

2p
X ( j )
Hence,
represents the "weight" or coefficient applied to a sinusoid
2p
of frequency in the FT representation.

Research Laboratory

31

## Nonperiodic signals: Fourier-transform

representations
Discrete-time sinusoids are used to represent discrete signal in the DTFT.
ej(+T)n=ejnejTn=ejn, if T=2p

## Discrete-time sinusoids are unique only over a 2p interval of frequencies,

since discrete-time sinusoids with frequencies separated by an integer
multiple of 2p are identical.
The DTFT involves frequencies from -p to p , as shown by the equation
1 p
j
j n
x[n]
X
(
e
)
e
d ,

p
2p
X (e j )
Hence,
represents the "weight" or coefficient applied to a sinusoid
2p
of frequency n in the DTFT representation.
Research Laboratory

32

[ n]

h[n]

x[k ] [n k ]

y[n]

## x[k ]h[n k ] x[n]* h[n]

x[n] ak e jk n

y[n] ak H (e jk )e jk n

Research Laboratory

33

DTFS, FS

## DTFS & FS()

k0k0()

f(k)=ejk0n(i.e., f(k)=f(k+N))DTFS
N
N 1

x[n] A[k ]e jk 0n
k 0

f(k)=ejk0tFS
(i.e., k )

x (t )

Research Laboratory

34

A[k ]e jk0t

DTFT,FT

DTFT & FT
()()

f()=ejn(i.e., f()=f(+2p))
DTFT2p

1
x[n]
2p

p
-

X (e j )e jn d

f()=ejtFT-

1
x (t )
2p
Research Laboratory

35

X ( j )e jt d,

## The DTFS representation of a periodic signal x[n]

with fundamental period N and fundamental
frequency 0=2p/N is given by
N 1

x[n] X [k ]e jk 0 n ,

(3.10)

k 0

1
X [k ]
N

N 1

jk 0 n
x
[
n
]
e
arethe DTFS coefficients of the signal x[n].

n 0

## We say that x[n] and X [k ] are a DTFS pair and is denoted by

x[n]

DTFS ; 0

Research Laboratory

X [k ]
36

## Either x[n] or X[k] provides complete description of

the signal.
The DTFS coefficients X[k] are termed a frequencydomain representation for x[n], because each
coefficient is associated with a complex sinusoid of
a different frequency.
k determines the frequency of the sinusoid
associated with X[k], so we may say that X[k] is a
function of frequency.
DTFS is the only Fourier representation that can be
numerically evaluated and manipulated in a
computer

Research Laboratory

37

[ n]
x[n]

h[n]

x[k ] [n k ]

y[n]

x[n] X [k ]e

N 1

y[n] X [k ]H (e jk 0 )e jk 0n

jk 0 n

k 0

k 0

1
X [k ]
N

## x[k ]h[n k ] x[n]* h[n]

N 1

N 1

x[n]e

N 1

Y [k ]e jk 0 n

jk 0 n

k 0

n 0

Y [k ] X [k ]H (e jk 0 )
N 1

y[n] Y [k ]e jk 0n
k 0

Research Laboratory

38

## Ex. 3.2 Determining DTFS coefficients

Find the frequency-domain representation of the
following signal

Prob. 3.2

Research Laboratory

## Figure 3.5 (p. 203)

Time-domain signal for Example 3.2.

39

## Discrete-time periodic signals: DTFS

Sol.
The signal has period N =5, so 0 2p / 5.
Also, the signal has odd symmetry. So we sum over n 2 to n 2
1 2
X [k ] x[n]e jk 2p n / 5
5 n 2
1
{x[ 2]e jk 4p / 5 x[ 1]e jk 2p / 5 x[0]e j 0 x[1]e jk 2p / 5 x[2]e jk 4p / 5 }
5
1
1
1
1
{1 e jk 2p / 5 e jk 2p / 5 } {1 j sin(2kp / 5)}
5
2
2
5
4p
1
X [2] j sin( ) / 5 0.232e j 0.531;
5
5
1
2p
1
X [1] j sin( ) / 5 0.276e j 0.760 ; X [0] 0.2e j 0 ;
5
5
5
Research Laboratory

40

## Discrete-time periodic signals: DTFS

1
2p
X [1] j sin( ) / 5 0.276e j 0.760 ;
5
5
1
4p
X [2] j sin( ) / 5 0.232e j 0.531.
5
5
Now, suppose we calculate X [k ] using n 0 to n 4
1
X [k ] {x[0]e j 0 x[1]e j 2 kp / 5 x[2]e j 4 kp / 5 x[3]e j 6 kp / 5 x[4]e j 8 kp / 5 }
5
1
1
1
{1 e j 2 kp / 5 0 e j 4 kp / 5 0 e j 6 kp / 5 e j 8 kp / 5 }
5
2
2
1
1 j 2 kp / 5 1 j 2 kp / 5
1
2 kp
{1 e
e
} {1 j sin(
)}
5
2
2
5
5
Note that e j 8 kp / 5 e j 2 kp e j 2 kp / 5 e j 2 kp / 5
Research Laboratory

41

## |X[k]| is the magnitude spectrum of x[n]

arg{X[k]} is the phase spectrum of x[n]

Research Laboratory

42

## Figure 3.6 (p. 204)

Research Laboratory

Magnitude and phase of the DTFS coefficients for the signal in Fig. 3.5.

43

Prob. 3.2

## Sol. The signal has period N =6, so 0 2p / 6 p / 3.

p
p
p
jk ( 1)
jk 0
jk 1
1 2
1
X [k ] x[n]e jk 0 n (2e 3
1e 3 2e 3 )
N n 3
6
1
kp
2
kp
1
(2 2 cos( ) 1) cos( )
6
3
3
3
6

1 2
x[n] X [k ] cos(kp / 3)
6 3
DTFS ;p /3

Research Laboratory

44

## Ex. 3.3 Computing DTFS coefficients by inspection

Find the DTFS coefficients of x[n]=cos(pn/3+f)

Prob. 3.3

Research Laboratory

45

## Discrete-time periodic signals: DTFS

Sol. The signal has period N =6, so 0 2p / 6 p / 3.

x[n+6]=cos(pn/3+2p+f)=x[n]

## We expend the cosine by Euler's Eq.

x[n] cos(
x[ n]

pn
3

f)

X [k ]e

j(

pn
3

f )

e
2

j(

pn
3

f )

1 jf j p3n 1 jf j p3n
e e
e e
2
2

jkp n / 3

k 2

## X [2]e j 2p n / 3 X [1]e jp n / 3 X [0] X [1]e jp n / 3 X [2]e j 2p n / 3 X [3]e jp n

1 jf

e , k 1

1 jf

X [k ]
e ,k 1
2

0, otherwise on - 2 k 3

Research Laboratory

46

## Prob. 3.3 (a) x[n] 1 sin(np /12 3p / 8)

Sol. The signal has period 0 p /12, so N =24.
x[n] 1
x[n]

j(

p
12

12

3p
)
8

X [k ]e

e
2j

jk 0 n

n 11

j(

p
12

3p
)
8

12

X [k ]e

jk

p
12

n 11
j

3p
8

7p
8

e
e
X [0] 1, X [1]
, X [1]
2j
2j

Research Laboratory

47

3p
8

e
2j

p
12

3p
8

e
2j

p
12

## Ex. 3.4 DTFS representation of an impulse train

Find the DTFS coefficients of the N-periodic impulse
train
x[n]

[n lN ]

## Figure 3.9 (p. 207)

A discrete-time impulse train with period N.

Research Laboratory

48

## Discrete-time periodic signals: DTFS

Sol.
1 N 1
1 N 1
1
jk 0 n
jkn 2p / N
X [k ] x[n]e
[n]e

N n 0
N n 0
N

Research Laboratory

49

## Ex. 3.5 The inverse DTFS

Find x[n] from the
DTFS coefficients
in Fig. 3.10

## Prob. 3.4, 3.5

Research Laboratory

50

## Figure 3.10 (p. 208)

Magnitude and phase of
DTFS coefficients for Example
3.5.

## Discrete-time periodic signals: DTFS

Sol.
2p
The DTFS has period 9, so 0
9
x[n]

X [k ]e

jk 2p n / 9

k 4

0 e j 8p n / 9 e j 2p / 3e j 6p n / 9 2e jp / 3e j 4p n / 9 0 e j 2p n / 9 1e jp e j 0p
0 e j 2p n / 9 2e jp / 3e j 4p n / 9 e j 2p / 3e j 6p n / 9 0 e j 8p n / 9
2 cos(6p n / 9 2p / 3) 4 cos(4p n / 9 p / 3) 1
Research Laboratory

51

## Ex. 3.6 DTFS representation of a square wave

Find the DTFS coefficient of x[n]
1, M n M
x[n]
, where N 2M 1
0, M n N M

## Figure 3.11 (p. 209)

Square wave for Example 3.6.

Research Laboratory

52

## Discrete-time periodic signals: DTFS

Sol.
2p
The period is N , so 0
N
1 N M 1
1 M jk 0 n
jk 0 n
X [k ]
x[n]e

e
, let m n M ,

N n M
N n M
1 2 M jk 0 ( m M ) 1 jk 0 M 2 M jk 0 m
e
e
e
,

N m 0
N
m 0
For k 0, N , 2 N ,
1
X [k ]
N

2M

m 0

Research Laboratory

, e jk 0 e jk 0 1,

2M 1
, k 0, N , 2 N ,
N

53

## Discrete-time periodic signals: DTFS

For k 0, N , 2 N ,

1 jk 0 M 1 e jk 0 (2 M 1) 1 e jk 0 (2 M 1) / 2 1 e jk 0 (2 M 1)
X [k ] e

jk 0
N
(1 e
)
N e jk 0 / 2 1 e jk 0
1 e jk 0 (2 M 1) / 2 e jk 0 (2 M 1) / 2 1 sin(k 0 (2 M 1) / 2)

,

jk 0 / 2
jk 0 / 2
N
e
e
sin(k 0 / 2)
N
Substituting 0

2p
, we have
N

2M 1
N , k 0, N , 2 N ,
X [k ]
1 sin(kp (2 M 1) / N ) , k 0, N , 2 N ,
sin(kp / N )
N
Research Laboratory

54

## Discrete-time periodic signals: DTFS

Using L'Hopital's rule by treating k as a real number, it is easy
to show that
1 sin(kp (2 M 1) / N ) 2 M 1
lim

k 0, N , 2 N , N
sin(kp / N )
N
Hence, X [k ] is commonly written as
1 sin(kp (2 M 1) / N )
X [k ]
N
sin(kp / N )

Research Laboratory

55

## Figure 3.12 (p. 211)

The DTFS coefficients for the square
wave shown in Fig. 3.11, assuming a
period N = 50: (a) M = 4. (b) M = 12.

Research Laboratory

56

## hence, the magnitude spectrum is the absolute value of X[k]

the phase spectrum is

## 0 for X[k] > 0

p for X[k] < 0 (-1=ejp)

Prob. 3.6

Research Laboratory

57

## Prob. 3.6(a) Find the DTFS

2p p l n k l 1
,
(a ) N 10, 0
1
10 5 n k
1
X [k ]
N

x[n]e

n 5

jk n
5

5 jk 5
1
5 jk 5 n 1

e
10 n 2 4
10 n 2 4

5 jk 5
5 jk 5
5 jk 5
1 e

e
e
2p
jk

4
4
4
8
1

5
e

p
p
125
10
5 jk 5
5 jk 5
1 e
1 e
4
4
Research Laboratory

58

-5

Prob. 3.6(b)
2p p l n k l 1
(b) N 10, 0
,
10 5 n k
1

4p
p
p
5p
jk
jk 0
jk
jk

jk n
jk n
jk n
1
1
1
e 5 e 5
e 5 e 5
5
5
5
X [k ] x[n]e
(1) e
1 e

(1)
p
p

jk
jk
N n 5
10 n 4
10
n 1

1 e 5
1 e 5

4p
4p
jk
jk

2
k
p
2 kp
4p
4p
10
10
jk
jk

e
sin
e
sin
4p
k
p
4
p
k
p

jk
j
jk
j
1
1 e 5
1 e 5 1
5
5
5
5
5
5

( e
)
(e
)
( e
)
(e
)

p
p
p
p

jk

jk

jk

jk
10
10
kp
kp
5
5
10
10
1

e
1

e
e
sin
e
sin

10
10

2 kp
2 kp
kp

sin
sin

j
sin(
)
5 kp
5 kp
1
5 (e j 10 e j 10 )
5
2

kp
10 sin kp
5

sin
10
10

Research Laboratory

59

## Examining Ex. 3.6, the DTFS coefficients

have even symmetry (i.e. X[k]=X[-k]) .
We may rewrite Eq. (3.10) as a series
involving harmonically related cosines.
Assume that N is even, then N/2 is an
integer, let k range from N/2+1 to N/2, we
have
N
2

x[n]

Research Laboratory

X [k ]e jk 0n

N
1
2

60

## Discrete-time periodic signals: DTFS

x[n]

N
2

X [k ]e jk 0 n

N
1
2

N
1
2

N
j 0 n
N
X [k ]e jk 0 n X [ ]e 2
2
N
k ( 1)
2

N n

N
1
2

N j 20
X [0] X [ ]e
( X [ m]e jm0 n X [ m]e jm0 n )
2
m 1
X [m] X [m], and N 0 2p
N
1
2

N jp n
e jm0 n e jm0 n
x[n] X [0] X [ ]e 2 X [m](
)
2
2
m 1
N
1
2

N
X [0] X [ ]cos(p n) 2 X [ m]cos( m0 n)
2
m 1
Research Laboratory

61

## Discrete-time periodic signals: DTFS

N
1
2

N 0
N
x[n] X [0] 2 X [ m]cos( m0 n) X [ ]cos(
n)
2
2
m 1
X [k ], k 0, N / 2
Define B[k ]
,
2 X [k ], k 1, 2, , N / 2 1
N
2

k 0

## A similar expression may be derived for odd N

Research Laboratory

62

(3.17)

## Ex. 3.7 Building a square wave from DTFS coefficients

The contribution of each term to the square wave may
be illustrated by defining the partial-sum approximation
to x[n] in Eq. (3.17) as
J

k 0

N
2

## This approximation contains the first 2J+1 terms

centered on k=0 in Eq. (3.10) (p.202). Assume a square
wave has period N=50 and M=12. Find the cases that
J=1,3,5,23, and 25.
Research Laboratory

63

## Figure 3.14a (p. 213)

Individual terms in the DTFS expansion of
a square wave (left panel) and the
corresponding partial-sum approximations
J[n] (right panel). The J = 0 term is 0[n]
= and is not shown. (a) J = 1. (b) J = 3.

Research Laboratory

64

Research Laboratory

65

## Figure 3.14b (p. 213)

(c) J = 5. (d) J = 23. (e) J = 25.

## The FS representation of a periodic signal x(t) with

fundamental period T and fundamental frequency
0=2p/T is given by
x(t )

X [k ]e

jk0t

(3.19)

1 T
X [k ] x(t )e jk0t dt
(3.20)
0
T
X [k ] are the FS coefficients of the signal x(t ).
We say that x(t ) and X [k ] are an FS pair and is denoted by
FS ;

x(t )

Research Laboratory

X [k ]
66

## Either x(t) or X[k] provides complete description of

the signal.

The FS coefficients X[k] are known as a frequencydomain representation for x(t) because each
coefficient is associated with a complex sinusoid of a
different frequency.

## k determines the frequency of the sinusoid

associated with X[k], so we may say that X[k] is a
function of frequency.

Research Laboratory

67

## The infinite series in Eq. (3.19) is not guaranteed to

converge for all possible signals

Define

x (t )

X [k ]e jk0t

## and choose X[k] according to Eq. (3.20)

1 T
2
If x(t ) dt , then the MSE between x(t ) and x (t ) is zero.
T 0
1 T
2
Or, mathematically, x(t ) x (t ) dt 0
T 0

Research Laboratory

68

## Note that, in contrast to discrete-time case, an MSE of

zero does not imply that x(t) and x (t ) are equal
pointwise, or x(t ) x (t ) at all value of t. It only implies
that there is zero power in their difference

## Pointwise convergence of x (t ) to x(t) is guaranteed at

all value of t except those corresponding to
discontinuities if the Dirichlet conditions are satisfied:

x(t) is bounded

## x(t) has a finite number of discontinuities in one period

Research Laboratory

69

## If x(t) satisfies the Dirichlet conditions and is not

continuous, then x (t ) converges to the midpoint of the
left and right limits of x(t) at each discontinuity.

## The following three examples illustrate how the FS

representation is determined.

Research Laboratory

70

## Ex. 3.9 Direct calculation of FS coefficients

Determine the FS coefficient of x(t) depicted in Fig.
3.16

## Figure 3.16 (p. 216)

Time-domain signal for Example 3.9.

Research Laboratory

71

## Continuous-time periodic signals: FS

Sol :
T 2, 0 2p / 2 p
1 T
X [k ]= x(t )e jk0t dt
T 0
1
1 2 (2 jkp )t
1 2 2t jkp t
e (2 jkp ) t |02
dt
dt e
e e
2(2 jkp )
2 0
2 0
1 e4
1
4 jk 2p
)
(1 e e

4 jk 2p
4 jk 2p

Research Laboratory

72

## Figure 3.17 (p. 217)

Magnitude and phase spectra for Example 3.9.

Research Laboratory

73

## |X[k]| is the magnitude spectrum of x(t)

arg{X[k]} is the phase spectrum of x(t)
Since x(t) is periodic, the interval of integration in Eq.
(3.20) may be chosen as any interval one period in
length.
Choosing the appropriate interval of integration often
simplifies the problem

Research Laboratory

74

## Ex. 3.10 FS coefficients for an impulse train

Determine the FS coefficient of x(t) for the signal

x(t )

(t 4l )

Research Laboratory

75

## Continuous-time periodic signals: FS

Sol :
T 4, 0 2p / 4 p / 2,
x(t ) is even symmetry it is easilier to evaluate X [k ] over - 2 t 2
p
jk t
1 T
1 2
1
jk0t
X [k ]= x(t )e
dt (t )e 2 dt
T 0
4 2
4
in this case, the magnitude spectrum is constant and the phase spectrum is zero.

Note that we cannot evaluate the infinite sum in Eq. (3.19) in this case and that
x(t) does not satisfy the Dirichlet condition. However, the FS expension of an
impulse train is useful in spite of convergence difficulties.

Research Laboratory

76

## Ex. 3.11 Calculation of FS coefficients by inspection

Determine the FS coefficient of x(t) for the signal

x(t ) 3cos( t / 2 / 4)

Prob. 3.7,3.8

Research Laboratory

77

## Continuous-time periodic signals: FS

Sol :
T 4, 0 2 / 4 / 2, x(t )

X [ k ]e jk t / 2

## Using Euler's formula to expend the cosine yields

e j ( t / 2 / 4) e j ( t / 2 / 4) 3 j / 4 j t / 2 3 j / 4 j t / 2
x(t ) 3cos( t / 2 / 4) 3
e e
e
e
2
2
2
3 j / 4
, k 1
2 e

3
X [k ] e j / 4 , k 1
2
0, otherwise

Research Laboratory

78

## Figure 3.18 (p. 219)

Magnitude and phase spectra for Example 3.11.

Research Laboratory

79

Prob. 3.7

## Find FS for x(t ) 2sin(2 t 3) sin(6 t )

T 1, 0 2
e j (2 t 3) e j (2 t 3) e j 6 t e j 6 t e 3 j j 2 t e3 j j 2 t e j 6 t e j 6 t
x(t ) 2

2j
2j
j
j
2j
2j
j
j
je 3 j e j 2 t je3 j e j 2 t e j 6 t e j 6 t
2
2
x(t )

X [k ]e

jk0t

X [k ]e

j
2 , k 3
3j
je , k 1

X [k ] je 3 j , k 1

j
,k 3
2
0, otherwise
Research Laboratory

80

j 2 kt

Prob. 3.8
T 1, 0 2
1 j t j t
(e e )
2

x(t ) cos( t )
x(t )

X [k ]e

jk0t

X [k ]e

j 2 k t

1
1 T
1 T2
jk0t
jk0t
X [k ] x(t )e
dt T x(t )e
dt 21 cos( t )e j 2 k t dt

T 0
T 2
2
1
2
1

e j t e j t j 2 k t
1 12 j (1 2 k ) t
e j (1 2 k ) t
j (1 2 k ) t
(
)e
dt 1 e
e
dt

2
2 2
2 j (1 2k )

j (1 2 k )

1
2

j (1 2 k )

e
2 j (1 2k )

1
2

Research Laboratory

j (1 2 k )

1
2

j (1 2 k )

e
2 j (1 2k )
81

1
2

1
2

1
2

e j (1 2 k ) t

2 j (1 2k )

1 2k
1 2k
sin(
) sin(
)
2
2

(1 2k )
(1 2k )

1
2

1
2

## Ex. 3.12 Inverse FS

Find the time-domain signal x(t) corresponding to the
FS coefficients
1 k jk / 20
X [k ] ( ) e
2

## Assume that the fundamental period T=2

Prob. 3.9

Research Laboratory

82

## Continuous-time periodic signals: FS

Sol : T 2, 0 2 / 2 ,
x(t )

X [k ]e

jk0t

1
1 k jk / 20 jk t
1 k jk / 20 jk t 1 k jk / 20 jk t
( ) e
e ( ) e
e ( ) e
e
k 2
k 2
k 0 2

1 l jl / 20 jl t 1 k jk / 20 jk t
1
1
= ( ) e
e
( ) e
e
1
1
1
l 1 2
k 0 2
1 e j ( k t / 20)
1 e j ( k t / 20)
2
2
1 j ( k t / 20) 1 j ( k t / 20) 1 j ( k t / 20)
1 j ( k t / 20)
e
e
1 e
1 e

1
2
2
2

1
1
1 j ( k t / 20) 1 j ( k t / 20)
1 e j ( k t / 20) 1 e j ( k t / 20)
1 e
1 e

2
2
2

1 j ( k t / 20) 1 1 j ( k t / 20)
3
e
1 e

3
2
4 2

1 1
5
1 e j ( k t / 20) e j ( k t / 20)
cos( t / 20) 5 4 cos( t / 20)
4 2
4
Research Laboratory

83

## Prob. 3.9 (a)

(a) X [k ] j [k 2] j [k 2] 2 [k 3] 2 [k 3], 0
x(t )

X [k ]e jk0t

( j [k 2] j [k 2] 2 [k 3] 2 [k 3])e jk t

je j 2 t je j 2 t 2e j 3 t 2e j 3 t
j (e j 2 t e j 2 t ) 2(e j 3 t e j 3 t )
j 2 j sin(2 t ) 2 2 cos(3 t ) 2sin(2 t ) 4 cos(3 t )
Research Laboratory

84

## Prob. 3.9 (b)

x (t )

X [k ]e

jk0t

j 2 ( t 1)
e
e

k 4

e j 2 ( t 1) (1 e

k
2

jk t
2

k 4

j 4 ( t 1)
2

1 e

j 9 ( t 1)
2

j ( t 1)
2

j ( t 1)
2

e j 2 ( t 1)e

9
( t 1)
4

1 e
9 (t 1)
sin(
)
4

(t 1)
sin(
)
4

Research Laboratory

j 5 ( t 1)
2

85

j ( t 1)
4

(e

(e

9
( t 1)
4

j ( t 1)
4

9
( t 1)
4

j ( t 1)
4

## Ex. 3.13 FS for a square wave

Determine the FS representation of the square wave
in Fig. 3.21

## Figure 3.21 (p. 221)

Square wave for Example 3.13.

Prob. 3.10

Research Laboratory

86

Sol :

0 2 / T ,

## x(t) is even symmetry integrating over the range

T
T
t
2
2

1 T2
1 T0 jk0t
1 jk0t T0
jk0t
For k 0, X [k ] T x(t )e
dt e
dt
e
|T0

0
T 2
T
Tjk0
2sin(k0T0 )
2 e jk0T0 e jk0T0

(
)
Tk0
2j
Tk0
2T
1 T2
1 T2
1 T0
j 00t
For k 0, X [0] T x(t )e
dt T x(t )dt dt 0
T 2
T 2
T T0
T
2sin(k0T0 ) 2T0

k 0
Tk0
T

By means of L'Hopital's
rule, it is straightforward to show that lim
Hence, we write X [k ]
Research Laboratory

2T

sinc(k 0 )
Tk0
k 2
T
T
87

## Figure 3.22a&b (p. 222)

The FS coefficients, X[k], 50 k
50, for three square waves. (see Fig.
3.21.) (a) Ts/T = 1/4 . (b) Ts/T =
1/16. (c) Ts/T = 1/64.

Research Laboratory

88

## Figure 3.22c (p. 222)

Research Laboratory

89

Sinc function
sin( u )
sinc(u )
u

analysis

i.e.

## The maximum of this function is unity at u=0

The zero crossing occurs at integer values of u
mainlobe & sidelobes

2sin(k 2 T0 / T ) 2sin(k 2 T0 / T )
2T sin( k 2T0 / T ) 2T0
2T

(T0 / T ) 0

sinc(k 0 )
k 2
k 2 (T0 / T )
T ( k 2T0 / T )
T
T

Research Laboratory

90

## Sinc function sinc(u) = sin(u)/(u)

Research Laboratory

91

## The form of FS described by Eqs. (3.19) and (3.20)

is termed the exponential FS

## The trigonometric FS is often useful for real-valued

signals and is expressed as

## x(t ) B[0] ( B[k ]cos(k0t ) A[k ]sin(k0t )),

(3.25)

k 1

1 T
B[0] x(t )dt
T 0
2 T
B[k ] x(t ) cos(k0t )dt
T 0
2 T
A[k ] x(t ) sin(k0t )dt
T 0
Research Laboratory

92

(3.26)

signal

## Using Eulers formula to expend cosine and

sine functions in Eq. (3.26), we have

B[k]=X[k]+X[-k]

A[k]=j(X[k]-X[-k])

Prob. 3.10

Research Laboratory

(3.27)

93

## The trigonometric FS coefficients of the

square wave in Ex. 3.13 are obtained from
(3.28)

B[0]=2T0/T

B[k]=2sin(k2T0/T)/(k)

## A[k]=0 (i.e. it is because that x(t) is even)

Research Laboratory

94

## Ex. 3.14 Square-wave partial sum approximation

Let the partial-sum approximation to the FS in (3.29)
be given by
J

J
xJ (t ) B[k ]cos(k0t ), xJ (t )
x(t )
k 0

## Consider a square wave with T=1 and T0=1/4. Depict

one period of the J-th term in this sum and find x J (t )
for J=1,3,7,29, and 99

Research Laboratory

95

## Continuous-time periodic signals: FS

Sol :
1
2 ,k 0

k 1
2
B[k ] (1) 2 , k is odd
k
0,k is even

Research Laboratory

96

## Figure 3.25a (p. 225)

Individual terms (left panel) in the FS expansion of a square wave and the corresponding partial-sum approximations
wave has period T = 1 and Ts/T = . The J = 0 term is 0(t) = and is not shown. (a) J = 1.

Research Laboratory

97

J(t)

## Figure 3.25b-3 (p. 226)

(b) J = 3. (c) J = 7. (d) J = 29.

Research Laboratory

98

## Figure 3.25e (p. 226)

(e) J = 99.

Research Laboratory

99

H
y(t ) h(t )* x(t )

x(t )

x(t )

X [k ]e

jk0t

y (t )

X [k ]H ( jk )e
0

Let Y [k ] X [k ]H ( jk0 )

1
X [k ] x(t )e jk0t dt
T 0

y (t )

Y [k ]e

100

jk0t

jk0t

## Ex. 3.15 RC Circuit: Calculation the O/P by means of

FS
Find the FS representation of the O/P y(t) of the RC
circuit depicted in Fig. 3.2 in response to the squarewave I/P in Fig. 3.21, assuming that T0/T=1/4, T=1,
and RC=0.1s

Research Laboratory

101

## Continuous-time periodic signals: FS

Sol :

x(t )

X [k ]e

jk0t

y (t ) H {x(t )}

y (t )

Y [k ]e

jk0t

X [k ]H {e

FS ;0
, y (t )
Y [k ] H ( jk0 ) X [k ]

## From Ex. 3.1 H ( j )

1/ RC
j 1/ RC

10
jk 2 10
2sin(k0T0 ) sin(k / 2)
From Ex. 3.13, X [k ]

Tk0
k

0 2 , RC 0.1s H ( jk0 )

Y [k ]

10
sin(k / 2)
jk 2 10
k

Research Laboratory

102

jk0t

H ( jk ) X [k ]e

jk0t

## Continuous-time periodic signals: FS

Y [k ]

10
sin(k / 2)
jk 2 10
k

k
Y [k ]
0, we can approximate y (t ) by

y (t )

100

Y [k ]e jk0t

k 100

Research Laboratory

103

## The FS coefficients Y[k], 25

k 25, for the RC circuit output
in response to a square-wave
input. (a) Magnitude spectrum.
(b) Phase spectrum. c) One
period of the input signal x(t)
dashed line) and output signal
y(t) (solid line). The output
signal y(t) is computed from
the partial-sum approximation
given in Eq. 3.30).

Research Laboratory

104

x[n] is given by
1
x[n]
2
j

X (e )

X (e j )e jn d

x[n]e

jn

(3.31)
(3.32)

## X (e j ) are the DTFT coefficients of the signal x[n].

We say that x[n] and X (e j ) are an DTFT pair and is denoted by
x[n]

DTFT

Research Laboratory

X (e j )

105

## X (e j ) is the frequency-domain representation of

x[n] as a function of a sinusoidal frequency

## Eq. (3.31) is usually termed the inverse DTFT, since

it maps the frequency-domain representation back
into time-domain

## DTFT is used primarily to analyze the action of

discrete-time system on discrete-time signals

## The infinite sum in Eq. (3.32) converges if x[n] has

finite duration and is finite values.

Research Laboratory

106

## If x[n] is of infinite duration, the the sum

converges only for certain classes of signals

If

## then the sum in Eq. (3.32) converges uniformly to

a continuous function of

n

## has finite energy), then the sum in Eq. (3.32) converges in a

mean-square error sense,but does not converge pointwise.
Research Laboratory

107

## Find the DTFT of the sequence x[n]=anu[n]

Prob. 3.11

Research Laboratory

108

## Discrete-time nonperiodic signals: DTFT

Sol :
Using Eq. (3.32), we have
j

X (e )

x[n]e

jn

u[n]e

jn

a n e jn
n 0

## The sum diverges for a 1, for a 1, we have

1
, for a 1
j
1ae
n 0
n 0
1
1
1 a cos ja sin
If a is real valued, X (e j )

## 1 a e j 1 a cos ja sin (1 a cos ) 2 (a sin ) 2

1
1
j
X (e )

1
1
2
2 2
2
(1 a cos ) (a sin )
(a 1 2a cos ) 2
X (e ) a e
j

n j n

(a e j ) n

a sin
arg{ X (e j )} arctan(
)
1 a cos
Research Laboratory

109

## The DTFT of an exponential signal x[n] =

(a)nu[n]. (a) Magnitude spectrum for a =
0.5. (b) Phase spectrum for a = 0.5. (c)
Magnitude spectrum for a = 0.9. (d) Phase
spectrum for a = 0.9.

Research Laboratory

110

## |X(ej)| is the magnitude spectrum of x(t)

arg{X(ej)} is the phase spectrum of x(t)

Research Laboratory

111

Prob. 3.11
X (e j )

## Find DTFT for x[n] 2(3)n u[n]

x[n]e jn

2(3) u[n]e
n

jn

2 (3e j ) n
n

1 j m
2

2 ( e )
1 j
m0 3
1 e
3
m n

Research Laboratory

112

1, n M
x[n]
0, n M

## Figure 3.30 (p. 233)

Example 3.18. (a) Rectangular
pulse in the time domain. (b)
DTFT in the frequency domain.

Research Laboratory

113

## Discrete-time nonperiodic signals: DTFT

Sol :
j

X (e )

x[n]e

jn

1 e jn , let m n M n m M

n M

j (2 M 1)

e
j

M
2M
2M
, 0, 2 , 4 ,
e
j
j ( m M )
jM
j
X (e ) e
e (e j ) m
1 e
m 0
m 0
2 M 1,
0, 2 , 4 ,

j (2 M 1)
j (2 M 1) / 2
(e j (2 M 1) / 2 e j (2 M 1) / 2 )
jM 1 e
jM e
e
e
j
1 e
e j / 2 (e j / 2 e j / 2 )

e j (2 M 1) / 2 e j (2 M 1) / 2 sin((2M 1) / 2)

j / 2
j / 2
e
e
sin( / 2)

Research Laboratory

114

## Discrete-time nonperiodic signals: DTFT

lim

0, 2 , 4 ,

X (e ) e

sin((2M 1) / 2)
2M 1
sin( / 2)
jM

Research Laboratory

1 e j (2 M 1) sin((2M 1) / 2)

j
1 e
sin( / 2)

115

## Ex. 3.19 Inverse DTFT of a rectangular spectrum

1, W
X (e )

0, W
j

Research Laboratory

116

## Figure 3.31 (p. 234)

Example 3.19. (a) Rectangular
pulse in the frequency domain. (b)
Inverse DTFT in the time domain.

## Discrete-time nonperiodic signals: DTFT

Sol :
W
1
1
jn
e

sin(Wn), n 0

W
1
1
2 nj
n
j
jn
j n
W
x[n]
X
(
e
)
e
d

e
d

2
2 W
W
, n 0
1
W
1
W
Wn
lim
sin(Wn) x[n]
sin(Wn) sinc( )
n 0 n

Research Laboratory

117

x[n] [n]

## Prob. 3.12, 3.13

Research Laboratory

118

Sol :
j

X (e )

x[n]e

jn

jn

[
n
]
e
1

DTFT
Hence, [n]
1

## Figure 3.32 (p. 235)

Example 3.20. (a) Unit impulse in the time domain.
(b) DTFT of unit impulse in the frequency domain.

Research Laboratory

119

X (e j ) (),

## Figure 3.33 (p. 236)

Example 3.21. (a) Unit impulse in
the frequency domain. (b) Inverse
DTFT in the time domain.

Research Laboratory

120

## Discrete-time nonperiodic signals: DTFT

Sol :
1
1
1
j
jn
jn
x[n]
X (e )e d
( )e d

2
2
2
1
DTFT
Hence,

(),
2
In this example, we have defined only one period of X (e j ).
alternatively, we can defind X (e j ) as
j

X (e )

( 2 k )

Research Laboratory

121

Prob. 3.12(d)
j

X (e )

x[n]e

jn

(1)e

n 7

jn

(1)e jn
n 1

e j 7 1 e j e j 8 1 e j 7 e j (1 e j 7 )

j
j
1 e
1 e
1 e j
j 7 / 2
j 7 / 2
j 7 / 2
j / 2
j / 2
j / 2
j 7
j

e
(
e

e
)
e
(
e

e
)
(1 e
)(1 e )

j
1 e
e j / 2 (e j / 2 e j / 2 )
e j 7 / 2

2 j sin( )2 cos( )
2
2

2 j sin( )
2

Research Laboratory

122

Prob. 3.13(a)

X (e ) 2cos(2)

1
1
j
jn
jn
x[n ]
X
(
e
)
e
d

cos(2

)
e
d

1 j 2
1 j ( n 2)
j 2
jn
j ( n 2)

(
e

e
)
e
d

(
e
d

e
d )

2
2
1
2 1, for n 2,
2
0, otherwise ( e jk - e jk 0)

Research Laboratory

123

Prob. 3.13(c)
1
j
jn
x[n]
X
(
e
)
e
d

j
1 0 j 2 jn

[ e e e d e e 2 e jn d ]
0
2

j
1 0 j 2 jn

[ e e e d e e 2 e jn d ]
0
2

j 0 (1 jn )
j 1 e e j n e e j n 1
(1 jn )

[ e
d e
d ]
[

0
2
2
1 jn
1 jn
e j n e j n cos( n )

1
1
j
2 jn
(1 e cos( n))[

]
(1 e cos( n))
2
1 jn 1 jn 2
1 n2
j

[1

e
cos( n)]
2
(1 n )
Research Laboratory

cos( n ) ( 1) n

124

n(1 e (1) n )
(1 n 2 )

response
1

y
[
n
]

## ( x[n] x[n 1])

1
2
Consider two systems:
y [n] 1 ( x[n] x[n 1])
2
2
1

h
[
n
]

( [n] [n 1])
1
2
the impulse responses of the two systems are
h [n] 1 ( [n] [n 1])
2
2
Find the frequency response and plot the magnitude response of each system

Research Laboratory

125

## Discrete-time nonperiodic signals: DTFT

Sol :
H1 (e j )
j

H 2 (e )

h1[n]e jn

h2 [n]e

jn

1 1
e j e
2 2

j
2

j
1 1 j
e
e je 2
2 2

e
2

e
2j

je

cos( )
2

sin( )
2

H1 (e j ) cos( ) , arg{H1 (e j )}
2
2

, 0

2 2
H 2 (e j ) sin( ) , arg{H 2 (e j )}
2
, 0
2 2
Research Laboratory

126

2 2

## Figure 3.36 (p. 239)

The magnitude responses of two simple discrete-time systems. (a) A
system that averages successive inputs tends to attenuate high
frequencies. (b) A system that forms the difference of successive
inputs tends to attenuate low frequencies.

Research Laboratory

127

## Ex. 3.23 Multipath communication channel:

Frequency response
y[n] x[n] ax[n 1]
In Ex. 2.12, we have h[n] [n] a [n 1]
and hinv [n] (a) n u[n], a 1
j

## Compare the magnitude response of both systems for a 0.5e 3 , a 0.9e

Research Laboratory

128

2
3

j

Sol : H (e )

h[n]e

jn

j

1
2 2

## H (e ) ((1 a cos( arg{a})) ( a sin( arg{a})) )

2

1
2

(1 a 2 a cos( arg{a})) , a 1
2

the frequency response of the inverse system is the inverse of the frequency
response of the origional system.
H inv (e j )

1
1
1

## H (e j ) 1 a e j ( arg{a}) 1 a cos( arg{a}) j a sin( arg{a})

H inv (e j )

j
H (e )

Research Laboratory

1
(1 a 2 a cos( arg{a}))
2

129

1
2

## Figure 3.37 (p. 241)

Magnitude response of the system in
Example 3.23 describing multipath
propagation. (a) Echo coefficient a =
0.5ej/3. (b) Echo coefficient a =
0.9ej2/3.

Research Laboratory

130

## Figure 3.38 (p. 241)

Magnitude response of the inverse
system for multipath propagation in
Example 3.23. (a) Echo coefficient a =
0.5ej/3. (b) Echo coefficient a = 0.9ej/3

Research Laboratory

131

## The FT representation of a nonperiodic signal x(t) is

given by
1
x(t )
2

X ( j )e jt d

X ( j ) x(t )e jt dt
-

(3.35)
(3.36)

## X ( j ) are the FT coefficients of the signal x(t ).

We say that x(t ) and X ( j ) are an FT pair and is denoted by
x(t )

FT

Research Laboratory

X ( j )
132

## X ( j ) is the frequency-domain representation of x(t)

as a function of a sinusoidal frequency

## Eq. (3.35) is usually termed the inverse FT, since it

maps the frequency-domain representation back
into time-domain

## FT is used primarily to analyze the action of

continuous-time system on continuous-time signals

## The integral in Eq. (3.35), (3.36) may not converge

for all x(t) and X ( j )

Research Laboratory

133

## Continuous-time nonperiodic signals: FT

1
jt
Define x (t )
X
(
j

)
e
d

2
It can be shown that the square error between x(t ) and x (t )
(i.e.error energy)

x(t ) x (t ) dt is zero if
2

x(t ) dt

## It implies that there is zero energy in the difference of terms.

Research Laboratory

134

## Continuous-time nonperiodic signals: FT

Pointwise convergence is guaranteed at all values of t
except to those corresponding to discontinuities if x(t )
satisfies the Dirichlet condition for nonperiodic signals:
x(t ) is absolutely integrable:

x(t ) dt

## x(t ) has finite number of maximum, minimum, and

discontinuities in any finite interval.
The size of each discontinuity is finite.

Research Laboratory

135

## Almost all physical signals encountered in

engineering practice satisfy the second and the third
conditions.

## Many idealized signals, such as the unit step, are

neither absolutely or square integrable.

## strict sense, the FT does not converge for such signals

Research Laboratory

136

## Ex. 3.24 FT of a real

decaying exponential
Find FT of x(t)=e-atu(t)
in Fig. 3.39(a)
Figure 3.39 (p. 243)
Example 3.24. (a) Real time-domain exponential signal.
(b) Magnitude spectrum.
(c) Phase spectrum.

Research Laboratory

137

## Continuous-time nonperiodic signals: FT

Sol :
the FT does not converge for a 0 since x(t) is not
absolutely integrable, that is

e u (t )dt e dt for a 0
at

at

For a 0,

X ( j ) e u (t )e
at

X ( j )

dt e
0

( a j ) t

1 ( a j ) t
dt
e
a j

1
1
2 2

(a 2 )
Research Laboratory

jt

, arg{ X ( j )} arctan( )
a
138

a j

1, T0 t T0
x(t )

0, t T0

## Figure 3.40 (p. 244)

Prob. 3.14

Research Laboratory

## Find the FT of x(t )

Example 3.25. (a) Rectangular
pulse in the time domain. (b) FT in
the frequency domain.

139

Sol :

X ( j ) x(t )e

jt

T0

dt e
T0

jt

1 jt
dt
e
j

T0
T0

sin(T0 ), 0

## For 0, X (0) 2T0

T0
lim sin(T0 ) 2T0 X ( j ) sin(T0 ) 2T0sinc(
)
0

0,
sin(T0 ) 0

sin(T0 )

X ( j ) 2
, arg{ X ( j )}

, 2 sin(T ) 0
0

Research Laboratory

140

## Prob. 3.14 (a)

X ( j ) x(t )e

(2 j ) t

jt

dt e 2t u( t )e jt dt

1
1
(2 j ) t 0
dt
e
|
2 j
j 2

Research Laboratory

141

## Ex. 3.26 Inverse FT of a rectangular spectrum

1, W W
X ( j )

0, W

Research Laboratory

142

Find x(t )

## Continuous-time nonperiodic signals: FT

Sol :
1
x(t )
2

1
X
(
j

)
e
d

2
W
For t 0, x(0)
jt

jt
1
W e d 2 j t e

jt

1
W
1
W
Wt
lim sin(Wt ) x(t ) sin(Wt ) sinc( )
t 0 t

Research Laboratory

143

W
W

1
sin(Wt ), t 0
t

## Figure 3.42 (p. 246)

Example 3.26. (a) Rectangular spectrum in the frequency domain.
(b) Inverse FT in the time domain.

Research Laboratory

144

## Ex. 3.27 FT of the unit impulse

Find the FT of x(t)=d(t)

Research Laboratory

145

## Continuous-time nonperiodic signals: FT

Sol :

X ( j ) x(t )e

jt

dt d (t )e jt dt 1

FT
d (t )
1

Research Laboratory

146

## Ex. 3.28 Inverse FT of an impulse spectrum

Find the inverse FT of X(j)=2d()

Research Laboratory

147

## Continuous-time nonperiodic signals: FT

Sol :
1
1
jt
x(t )
X ( j )e d

2
2
FT
1
2d ( )

Research Laboratory

148

2d ( )e jt d 1

Prob. 3.15(d)

2 j
0
j
1
1
jt
2 jt
2 jt
e d )
e

d
e
e
(

d
e
)

j
(
X
x (t )

0
2 2
2
2
0
1
1
2 ( j 2 1d ( j ) 0 1d ) 2 (2 j 2 j ) 0, t 0

cos(2t )
1
j 2t
j 2t

,t 0
((1 e ) ( e 1))

t
2 t

j, e

Research Laboratory

149

## Ex. 3.29 Characteristics of digital communication

signals

Ar , t T0 / 2
A rectangular pulse: xr (t )
0, t T0
2 t
Ac
(1

cos(
), t T0 / 2
2
T0
A raised-cosine pulse: xc (t )
0, t T
0

Choose constants Ar and Ac such that both BPSK signals have unit power.

## Figure 3.44 (p. 249)

Pulse shapes used in BPSK communications. (a)
Rectangular pulse. (b) Raised cosine pulse.

Research Laboratory

150

## Figure 3.45 (p. 249)

BPSK signals constructed by using
(a) rectangular pulse shapes and
(b) raised-cosine pulse shapes.

Research Laboratory

151

## Continuous-time nonperiodic signals: FT

Suppose that the frequency band assigned to each user is 20 kHz.
To prevent interference with adjacent channels, we assume that
the peak value of the magnitude spectrum of the transmitted signal
outside the 20kHz band is required to be -30dB below the peak
in-band magnitude spectrum.
Use the FT to determine the maximum number of bits per second
that can be transmitted when the rectangular and raise-cosine pulse
shapes are utilized.
Research Laboratory

152

## Continuous-time nonperiodic signals: FT

Sol :
Although BPSK signal is not periodic, its magnitude squared is T0 periodic
Thus, its respective power is calculated as
Pr

1
T0

1
Pc
T0

T0
2

Ar2 dt Ar2

T0
2

T0
2

Ac
2 t 2
1
(
(1

cos(
))
dt

T 2
T0
T0
0

2
c

A
4T0

2
T0
2

[1 2 cos(

T0
2

T0
2

Ac2
2 t 2
(1 cos(
)) dt
4
T0

2 t 1 1
4 t
3
) cos(
)]dt Ac2
T0
2 2
T0
8

Let Pr Pc 1, Ar 1, Ac
Research Laboratory

T0
2

153

8
3

## Spectrum of rectangular pulse in dB, normalized by T0.

Research Laboratory

154

## Continuous-time nonperiodic signals: FT

Using the result of Ex. 3.25, the FT of xr (t ) is given by
X r ( j )

2sin(T0 / 2)

= 2

X r' ( jf )

2 f
f

## The normalized spectrum of the signal in dB is given by 20log10{

X r' ( jf )
T0

From Fig. 3.46, it implies that we must choose T0 so that the 10th zero crossing is at 10kHz
The kth zero crossing occurs when f k / T0 . So we require 10000=10/T0 or T0 10 3
It implies a data transmission rate of 1000 bits/sec.

Research Laboratory

155

## Continuous-time nonperiodic signals: FT

The FT of xc (t ) is given by
1 8
X c ( j )
2 3
2

2
3

T0
2

2 t jt
T (1 cos( T0 ))e dt
0

T0
2

2
jt
T e dt 3
0

T0
2

T0
2

Research Laboratory

e jt dt

T0
2

T0
2

1 2
2 3

156

2 t

2 t

j
1 j T0
(e
e T0 )e jt dt
2
T0
2

T0
2

2
j ( ) t
T0

dt

1 2
2 3

T0
2

T0
2

j (

2
)t
T0

dt

## Continuous-time nonperiodic signals: FT

T0
2

e j t dt 2

sin( T0 / 2)

T0
2

X c ( j )

2
X c' ( jf )

2
3

T0
2

e jt dt

T0
2

1 2
2 3

T0
2

2
j ( ) t
T0

T0
2

dt

1 2
2 3

T0
2

j (

2
)t
T0

dt

T0
2

2 sin(T0 / 2)
2 sin(( 2 / T0 )T0 / 2)
2 sin(( 2 / T0 )T0 / 2)

3
2 / T0
3
2 / T0
2 sin( fT0 )
2 sin( ( f 1/ T0 )T0 )
2 sin( ( f 1/ T0 )T0 )
0.5
0.5
3
f
3
( f 1/ T0 )
3
( f 1/ T0 )

Research Laboratory

157

## Figure 3.47 (p. 252)

The spectrum of the raised-cosine
pulse consists of a sum of three
frequency-shifted sinc functions.

Research Laboratory

158

## Spectrum of the raised-cosine pulse in dB, normalized by T0.

Research Laboratory

159

## Continuous-time nonperiodic signals: FT

The normalized spectrum of the signal in dB is given by
20log10 {

X c' ( jf )
T0

## From Fig. 3.48, we require 10000=2/T0 or T0 2 104 s

It implies a data transmission rate of 5000 bits/sec.
the use of the raise-cosine pulse shape increases the data
transmission rate by a factor of five relative to the rectangular
pulse shape in this application.

Research Laboratory

160

## Properties of Fourier representations

Research Laboratory

161

Periodic signals

## A discrete set of frequencies ( e jk 0 n , e jk0t ) is involved in the

series
Frequency domain representation involves a discrete set of
weights or coefficients (X[k])

Non-periodic signals

## Signal is represented as a weighted integral of complex

sinusoids over a continuum of frequencies

Research Laboratory

162

## Discrete-time signals have periodic frequency-domain

representation because the discrete-time complex
sinusoids are 2p periodic

## Continuous-time signals have non-periodic frequencydomain representation because continuous-time

sinusoids with distinct frequencies are always distinct.

## Representations that are discrete (continuous) in one

domain are periodic (non-periodic) in the other domain

Research Laboratory

163

## Properties of Fourier representations

Research Laboratory

164

## Linearity and symmetry properties

Linearity property:

## Used to find Fourier representations of signals that are

constructed as sums of signals whose representations are

## In FS & DTFS, the signals are assumed to have the same

fundamental period

Research Laboratory

165

## Ex. 3.30 Linearity in the FS

3
1
x(t ) y (t )
2
2
Find the FS coefficient Z [k ]
z (t )

Prob. 3.16

Research Laboratory

166

z(t)=(3/2)x(t)+(1/2)y(t)

## Representation of the periodic signal z(t) as

a weighted sum of periodic square waves:
z(t) = (3/2)x(t) + 1/2y(t).
(a) z(t). (b) x(t). (c) y(t).

Research Laboratory

167

## Linearity and symmetry properties

Sol.
z (t )

3
1
x(t ) y (t )
2
2

1
kp
sin( )
kp
4
1
kp
FS ;2p
y (t )
Y [k ]
sin( )
kp
2
3
1
3
kp
1
kp
FS ;2p
z (t )
Z [k ] X [k ] Y [k ]
sin( )
sin( )
2
2
2 kp
4
2 kp
2

FS ;2p
x(t )
X [k ]

Research Laboratory

168

Prob. 3.16(b)
(b)

1
u[n], 1
,
j
1e
n

DTFT

1, W
1
DTFT
j
sin(Wn), 0 W p X (e )
pn
0, W p
1
pn
4
1
DTFT
4 u[n]
sin( )
1
pn
4
2
1 e j
2
n

Research Laboratory

169

1,

0, p p
4

## Symmetry properties: Real and imaginary signals

*

X ( j ) x(t )e dt x* (t )e jt dt (3.37)

jt

## X ( j ) x(t )e j ( )t dt X ( j ) X ( j ) is complex-conjugate symmetric

*

Re{X ( j )}=Re{X * ( j )}

X * ( j ) X ( j )

## Im{X ( j )}= Im{X * ( j )}

Research Laboratory

## Re{X ( j )} real part is even

X * ( j ) X ( j )

170

## The real part of the transform is an even function

of frequency
The imaginary part of the transform is an odd
function of frequency
Magnitude spectrum is even
Phase spectrum is odd

## X*[k]=X[N-k] because the DTFS coefficients are N

periodic and thus X[-k]=X[N-k]

Research Laboratory

171

## Magnitude spectrum is even

Phase spectrum is odd

## Suppose x(t ) is real, then x(t ) x* (t ),

X ( j ) x(t )e j ( ) t dt X ( j )
*

X ( j ) X ( j ) e j arg{ X ( j )} ;
X ( j ) X ( j ) e
*

j arg{ X ( j )} *

X ( j ) e j arg{ X ( j )}

X * ( j ) X ( j ),
X ( j ) X ( j ) magnitude spectrum is even
arg{ X ( j )} j arg{ X ( j )} phase spectrum is odd

Research Laboratory

## Symmetry properties: Real and imaginary signals

Research Laboratory

## The complex-conjugate symmetry property in FT leads

to a simple characterization of the O/P of an LTI system
A j (t ) A j (t )
e
e
2
2
y (t ) H ( j )e jt , H ( j ) H ( j ) e j arg{H ( j )}

x(t ) A cos(t )
Eq. (3.2): x(t ) e jt

A j (t arg{ H ( j )})
A j (t arg{ H ( j )})
y (t ) H ( j ) e
H ( j ) e
2
2
H ( j ) H ( j ) , arg{H ( j )} arg{H ( j )}
A j (t arg{ H ( j )}) j (t arg{ H ( j )})
(e
e
)
2
H ( j ) A cos(t arg{H ( j )})

y (t ) H ( j )

Research Laboratory

## The system modifies the amplitude of the I/P sinusoid by

|H(j)| and phase by arg{H(j)}

## the frequency response of a system with a real-valued impulse

response is easily measured using a sinusoid oscillator & an
oscillscope

## Figure 3.50 (p. 257)

Research Laboratory

A sinusoidal input to an LTI system results in a sinusoidal output of the same frequency,
with the amplitude and phase modified by the systems frequency response.

## Similarly, for discrete-time LTI system

x[n] A cos(n )

A j ( n ) A j ( n )
e
e
2
2

## x[n] e jn y[n] H (e j )e jn , and H (e j ) H (e j ) e j arg{H ( e

A j ( n arg{H ( e j )})
A j ( n arg{H ( e j )})
j
H (e ) e
y[n] H (e ) e
2
2
H (e j ) H (e j ) , arg{H (e j )} arg{H (e j )}
j

## A j ( n arg{H ( e j )}) j ( n arg{ H ( e j )})

)
e
y[t ] H (e ) (e
2
H (e j ) A cos(n arg{H (e j )})
j

Research Laboratory

)}

## Now, suppose that x(t) is purely imaginary

x* (t ) x(t )
From (3.37),
*

jt
*
jt

X ( j ) x(t )e dt x (t )e dt = x(t )e j ( ) t dt

= X ( j )
*

## Re{X ( j )}=Re{X * ( j )}= Re{X ( j )}

Im{X ( j )}= Im{X * ( j )}= Im{X ( j )}

## Real part of FT is odd, imaginary part is even

Research Laboratory

## x* (t ) x(t ), x(t ) x(t ) x* (t ) x(t )

X * ( j ) [ x(t )e jt dt ]* x* (t )e jt dt = x(t )e j ( t ) dt x( )e j d X ( j )
X * ( j ) X ( j ) Im{ X ( j )} 0

## Suppose that x(t) is real & odd

x* (t ) x(t ), x(t ) x(t ) x* (t ) x(t )

X * ( j ) [ x(t )e jt dt ]* x(t )e j ( t ) dt x( )e j d X ( j )
X * ( j ) X ( j ) Re{ X ( j )} 0

Research Laboratory

representations.

## Time signal is real and even,

the frequency-domain representation is real
Time signal is real and odd,
the frequency-domain representation is imaginary
Real and even time-domain signals have
real and even frequency-domain representation
Real and odd time-domain signals have
imaginary and odd frequency-domain representation

Prob. 3.17

Research Laboratory

Prob. 3.17

## Prob. 3.17 Determine the time-domain

signal correspond to the following
frequency domain representation are
real or complex valued and even or odd:
(a)

Research Laboratory

Discussion

## How to determine whether f(x) is real or complex?

A: real if f(x)=f*(x); imaginary if f(x)=-f*(x)
How to determine whether f(x) is even or odd?
A: even if f(x)=f(-x); odd if f(x)=-f(-x)

Research Laboratory

Prob. 3.17(a)

: arg{X ( j)}

X ( j ) X ( j )
arg{ X ( j )} arg{ X ( j )}

X ( j ) X ( j ) e j arg{ X ( j )} X ( j ) e j arg{ X ( j )} X * ( j )
p
j ( )

2
, 0 1 j X ( j ) , 0 1
X ( j ) e
X ( j )

p
X ( j ) e j 2 , 1 0 j X ( j ) , 1 0

p
j( )

2
X ( j ) e , 0 1 j X ( j ) , 0 1
X ( j )

X ( j )
p
X ( j ) e j 2 , 1 0 j X ( j ) , 1 0

Prob. 3.17(a)

X * ( j) X ( j), X ( j) X ( j)

1
x (t )
2p
*

1
X ( j )e d 2p

1
v

2p
1
x(t )
2p

jt

1
j ( t )
X
(
j

)
e
d

2p

X ( j )e
*

1
jvt
X ( jv)e d (v)
2p

2p
v

jt

1
d
2p

Research Laboratory

j ( ) t
(

X
(

))
e
d

## X ( jv)e jvt dv x(t ), x(t ) is odd

X ( j )e j ( )t d

Convolution property

## The most important property of Fourier

representation is the convolution property.

## With the convolution property, we may analyze the

I/P-O/P behavior of a linear system in the frequency
domain by multiplying transforms instead of
convolving time signals.

## The convolution property is a consequence of

complex sinusoids being eigenfunctions of LTI
systems.

Research Laboratory

## y (t ) h(t ) * x(t ) h( ) x(t )d , x(t )

1
2

X ( j )e j (t ) d

y (t ) h( ) x(t )d h( ) X ( j )e j (t ) d d

1
j
X ( j )e jt d

h
(

)
e
d

2
1
jt

H
(
j

)
X
(
j

)
e
d

2
FT
y (t ) h(t ) * x(t )
Y ( j ) H ( j ) X ( j )

Research Laboratory

## Ex. 3.31 Solving a convolution problem in the

frequency domain
sin( t )
sin(2 t )
, h(t )
t
t
Find y (t ) h(t )* x(t )

Let x(t )

Research Laboratory

## Convolution of nonperiodic signals

Sol.
1, W
1
FT
From Ex. 3.26,
sin(Wt ) X ( j )
t
0, W
1,
sin( t ) FT
x (t )
X ( j )
,
t
0,
1, 2
sin(2 t )
FT
h (t )
, H ( j )
,
t
0, 2
FT
Since y (t ) h(t ) * x (t )
Y ( j ) H ( j ) X ( j )

1,
Y ( j ) H ( j ) X ( j )
X ( j ),
0,
Research Laboratory

y (t ) x (t )

sin( t )
t

## Ex. 3.32 Finding inverse FTs by means of the

convolution property

x(t ) X ( j )

FT

## Prob. 3.18, 3.19, 3.20

Research Laboratory

2
sin
( )
2

## Convolution of nonperiodic signals

Sol.
We may write X ( j ) as the product Z ( j ) Z ( j ), where Z ( j )
The convolution property states that
FT
z (t ) * z (t )
Z ( j ) Z ( j ), so x(t ) z (t ) * z (t )

## Using the result of Ex. 3.25, we have

t 2, 2 t 0
1, t 1 FT

z (t )
Z ( j ), x(t ) z (t ) * z (t ) 2 t , 0 t 2
0, t 1
0, otherwise

## Figure 3.52 (p. 261)

Signals for Example 3.32. (a)
Rectangular pulse z(t). (b) Convolution
of z(t) with itself gives x(t).

Research Laboratory

sin( )

Prob. 3.18(a)

(a)

1
e u (t )
,
j a
at

FT

3
2
y (t ) x(t ) * h(t ) Y ( j ) X ( j ) H ( j )
j 1 j 2
FT

Research Laboratory

## Two nonperiodic discrete-time signals x & h

DTFT
DTFT
x[n]
X (e j ), h[n]
H (e j ),
DTFT
y[n] h[n]* x[n]
Y (e j ) H (e j ) X (e j )

## Proof of this result is left as an exercise

Research Laboratory

Filtering

The multiplication that occurs in the frequencydomain representation give rise to the notation of
filtering
A system performs filtering on the I/P signal by
presenting a different response to components of the
I/P that are at different frequencies.
Filtering implies that some frequency components of
the I/P are eliminated while others are passed by the
system unchanged.

## Low-pass, high-pass, band-pass

Passband, stopband, transition band

Research Laboratory

Filtering
Figure 3.53 (p. 263)
Frequency response of ideal continuous- (left panel)
and discrete-time (right panel) filters. (a) Low-pass
characteristic. (b) High-pass characteristic. (c) Bandpass characteristic.

Research Laboratory

Filtering

## The magnitude response of a filter is commonly

described in units of decibels (dB), defined as

20 log|H(j)| or 20 log|H(ej)|

## The edge of the passband is usually defined by the

frequencies for which the response is -3dB,
corresponding to a magnitude response of ( 1/ 2 )

## |Y(j)|2= |H(j)|2 |X(j)|2

The -3dB points are usually termed the cutoff frequencies of
the filter

Research Laboratory

Filtering

## From Ex. 1.21

1 t /( RC )
hC (t )
e
u (t )
RC
Since yR (t ) x(t ) yC (t ),
Figure 3.54 (p. 264)

RC
xt
y
y t
1 t /( RC )
hR (t ) (t )
e
u (t )
RC
Plot the magnitude responses of both systems on a linear scale and in dB
and characterize the filtering properties of the systems.
circuit with input ( ) and outputs

Research Laboratory

c(t)

and

R(

).

Filtering
Sol.
1 t /( RC )
1
FT
e
u (t )
H C ( j )
RC
j RC 1
1 t /( RC )
j RC
FT
hR (t ) (t )
e
u (t ) H R ( j ) 1 H C ( j )
RC
j RC 1

hC (t )

Research Laboratory

## RC circuit magnitude responses as a function of normalized frequency RC.

(a) Frequency response of the system corresponding to yC(t), linear scale.
(b) Frequency response of the system corresponding to yR(t), linear scale.
(c) Frequency response of the system corresponding to yC(t), dB scale.
(d) Frequency response of the system corresponding to yR(t), dB scale,
shown on the range from 0 dB to 25 dB.

Research Laboratory

## RC circuit magnitude responses as a function of normalized frequency RC.

(a) Frequency response of the system corresponding to yC(t), linear scale.
(b) Frequency response of the system corresponding to yR(t), linear scale.
(c) Frequency response of the system corresponding to yC(t), dB scale.
(d) Frequency response of the system corresponding to yR(t), dB scale,
shown on the range from 0 dB to 25 dB.

Research Laboratory

Filtering

## The convolution property implies that the frequency

response of a system may be expressed as the ratio
of the FT or DTFT of the O/P to that of the I/P.
Y ( j )
For continuous-time system: H ( j )
X ( j )
j
Y
(
e
)
j
For discrete-time system: H (e )
X (e j )

## If the I/P spectrum is nonzero at all frequencies, the

frequencies of a system may be determined from knowledge
of the I/P and O/P spectra.

Research Laboratory

Filtering

## Given x(t ) e2t u (t ), y (t ) e t u (t ).

Find the frequency response and the impulse response
of this syetm.

Research Laboratory

Filtering
Sol.
1
x(t ) e u (t ) X ( j )
j 2
1
FT
t
y (t ) e u (t ) Y ( j )
j 1
Y ( j ) j 2
1
H ( j )

1
X ( j ) j 1
j 1
2 t

FT

h(t ) (t ) e t u (t )

Research Laboratory

Filtering

## For continuous-time system: X ( j ) H inv ( j )Y ( j )

For discrete-time system: X (e j ) H inv (e j )Y (e j )

## An inverse system is also known as an equalizer

The process of recovering the I/P form O/P is know as
equalization

## In practice, causality restriction makes it difficult to

build an inverse system.

## Approximated inverse is used.

Research Laboratory

Filtering

## Ex. 3.35 Multipath communication channel:

Equalization
Given y[n] x[n] ax[n 1], a 1.
Use the convolution property to find the impulse response
of an inverse syetm.

Prob. 3.21

Research Laboratory

Filtering
Sol.
1, n 0

0, otherwise

## hinv [n]* h[n] [n], take the DTFT of both side,

H inv (e j ) H (e j ) 1 H inv (e j )

1
H (e j )

DTFT
h[n] [n] a [n 1]
H (e j ) 1 ae j

H inv (e j )

1
1

H (e j ) 1 ae j

## hinv [n] (a ) n u[n]

Research Laboratory

## (Ref. p.774, Appendix C.3)

Prob. 3.21(a)

Find x(t) if

(a)

FT
e at u (t )

1
,
j a

1
1
1
H ( j )
, Y ( j )

j 4
j 3 j 4
1
1

Y ( j )
1
j 3 j 4 j 4
X ( j )

1
1
H ( j )
j 3
j 3
j 4
x(t ) e 3t u (t )
Research Laboratory

## For reference only

Research Laboratory

## Differentiation and integration are operators

that applies to continuous functions

## Continuous-time signals: x(t)

Continuous-frequency: FT X(jw) and DTFT X(ejW)

Research Laboratory

207

Differentiation in time

## Consider a nonperiodic signal x(t)

1
jwt
x(t )
X
(
j
w
)
e
dw

2
d
d 1
1
jwt
x(t ) X ( jw )e d w
dt
dt 2
2
d
FT
x(t )
jw X ( jw )
dt

[ X ( jw ) jw ]e jwt d w

## Differentiation a signal in the time domain corresponds to

multiplying its FT by jw
Differentiation accentuates the high-frequency components
of the signal
Differentiation destroys an DC component of x(t)

Research Laboratory

208

Differentiation in time

## Ex. 3.37 Verifying the differentiation property

d at
jw
FT
(e u (t ))
dt
a jw

Prob. 3.22,3.23

Research Laboratory

209

Differentiation in time
Sol.
d at
(e u (t )) ae at u (t ) e at (t ) ae at u (t ) (t )
dt
d at
a
jw
FT
(e u (t ))
1
dt
a jw
a jw

Research Laboratory

210

Prob. 3.22(a)

Find FT if
(a)

1
d
FT
e u (t )
, x(t )
jw X ( jw )
a jw dt
at

z (t ) e

2 t

FT

e 2t u (t ) e 2( t )u (t )

1
1
4
Z ( jw )

Appendix C.4
2
2 jw 2 jw 4 w
d
4 jw
FT
x(t ) z (t ) X ( jw ) jw Z ( jw )
dt
4 w2

Research Laboratory

211

Differentiation in time

## The differentiation property may be used to find the

frequency response of a continuous-time system
described by the differential equation
M
N
M
dk
dk
k
k
a
y
(
t
)

b
x
(
t
)

a
(
j
w
)
Y
(
j
w
)

b
(
j
w
)
X ( jw )

k
k
k
k
k
k
dt
dt
k 0
k 0
k 0
k 0
N

Y ( jw )
H ( jw )

X ( jw )

k
b
(
j
w
)
k
k 0
N

k
a
(
j
w
)
k
k 0

Research Laboratory

212

Differentiation in time

## The frequency response of a system described by a

linear constant-coefficient differential equation is a
ratio of two polynomials in the jw
The frequency response is the systems steady-state
response to a sinusoid.

## It cannot represent initial conditions

It can only describe a system that is in a steady-state
condition

Research Laboratory

213

Differentiation in time

x(t )

X [k ]e

jkw0t

d
d
jkw0t
jkw0t
x(t ) X [k ]e

X
[
k
]
jk
w
e

dt
dt k
k
d
FS ;w0
x(t )
jkw0 X [k ]
dt

## Differentiation forces the time-averaged value of

the differentiated signal to be zero; hence, the FS
coefficient for k=0 is zero

Research Laboratory

214

Differentiation in time

## Ex. 3.39 Find the FS representation of the

triangular wave depicted in Fig. 3.59(a)

## Signals for Example 3.39. (a) Triangular wave y(t).

(b) The derivative of y(t) is the square wave z(t).

Research Laboratory

215

Differentiation in time
Sol.
T 1
d
y (t ), as shown in Fig. 3.59(b). z (t ) 4 x (t ) 2, x(t ) could be obtained from Ex. 3.13 with 0 =
dt
T 4
(p. 221)
0, k 0

FS
z (t ) 4 x(t ) 2
Z [k ] 4 X [k ] 2 [ k ] 4sin( k )
2 ,k 0

k
d
Z [k ]
FS ;w0
z (t ) y (t )
Z [k ] jkw0Y [k ] Y [k ]
,k 0
dt
jkw0

Define z (t )

Y [0] is the average value of y (t ) and is determined by inspection of Fig. 3.59(a). Y [0]
T
2 ,k 0
2
Z [k ]
Z [k ]
Z [k ]

w0
Y [k ]

T Y [k ]
k
2
T
sin(
)
T
jkw0 jk 2
j 2 k

2 ,k 0
T

2 2
jk

Research Laboratory

216

T
2

Differentiation in frequency

## Consider the FT X(jw)

X ( jw ) x(t )e jwt dt

d
d
jwt
jwt

X ( jw )
x
(
t
)
e
dt

x
(
t
)(

jt
)
e
dt

dw
dw
d
FT
jtx(t )
X ( jw )
dw

## Differentiation of a FT in the frequency domain

corresponds to multiplication of the signal by jt in
the time domain

Research Laboratory

217

Differentiation in Frequency

## Ex. 3.40 FT of a Gaussian pulse

Determine the FT of the Gaussian pulse, defined by
g (t )

1
e
2

t2
2

## , as depicted in Fig. 3.60

Prob. 3.25,3.26,3.27,3.28

Research Laboratory

218

## Figure 3.60 (p. 275)

Gaussian pulse g(t).

Differentiation in Frequency
d
d 1 t2
t t2
Sol.
g (t )
e
e tg (t )
(1)
dt
dt 2
2

d
FT
FT
g (t )
jwG ( jw ) tg (t )
jwG ( jw )
dt
d
1 d
FT
FT
jtg (t )
G ( jw ) tg (t )
G ( jw )
dw
j dw
1 d
d
jwG ( jw )
G ( jw ) wG ( jw )
G ( jw )
(2)
j dw
dw
2

c G ( j 0),

w2
2

## , c is determined by (see Appendix A-4)

t2
2

1
1
e dt 1 c 1,
e
2
2

Research Laboratory

219

t2
2

FT

w2
2

Differentiation in Frequency

## Consider the DTFT X(ejW)

jW

X (e )

x[n]e jWn

d
d

j Wn
j Wn
X (e j W )
x
[
n
]
e

x
[
n
](

jn
)
e

dW
d W n
n
d
DTFT
jnx[n]
X (e j W )
dW

Research Laboratory

220

x(t ) te at u (t ) tz (t ), z (t ) e at u (t )
d
d
FT
FT
jtz (t )
Z ( jw ), tz (t ) j
Z ( jw )
dw
dw
1
d 1
FT
Z ( jw )
, X ( jw ) j

jw a
dw jw a
j

Research Laboratory

221

1 j

jw a )

jw a )

d
d
3t
2 t
y (t ) {te u (t ) * e u (t )} {x(t ) * z (t )}
dt
dt
1
FT
3t
x(t ) te u (t ) X ( jw )
2
jw 3)
1
z (t ) e u (t ) Z ( jw )
jw 2
2 t

FT

Y ( jw ) jw ( X ( jw ) Z ( jw )}

Research Laboratory

222

jw

jw 3) jw 2
2

Integration

## Integration applies only to continuous dependent

variables.

Time: FT and FS
Frequency: FT and DTFT

## We limit our consideration here to integrating

nonperiodic signals with respect to time

Research Laboratory

223

Integration
d
1
Let y (t ) x( )d x(t ) y (t ) Y ( j )
X ( j ), 0

dt
j
The value at 0 is modified by adding a term c ( ), where c depends on
the average value of x(t ). Th correct result is obtained by setting c X ( j 0)
t
1
FT
x( )d j X ( j ) X ( j 0) ( )
t

Integration

## Integration may be viewed as an averaging operation,

It tends to smooth signals in time
Deemphasizing the high-frequency components of the signal

Prob. 3.29

Research Laboratory

224

1
z ( )d j Z ( j ) Z ( j0) ( )
1
1
1
X ( j )
( ), Z ( j )
, Z ( j 0) 1
j j 1
j 1
t

FT

x(t ) z ( )d e u ( )d e d

1 e t , if t 0

1 et u (t )
0, otherwise

Research Laboratory

225

## FT of the unit step

t

u (t ) ( )d ,

FT
(t )
1
FT
u (t )
U ( j )

Research Laboratory

1
( )
j

226

## Figure 3.61 (p. 279)

Representation of a step function as the
sum of a constant and a signum function.

Research Laboratory

227

1, t 0
1 1

## u (t ) sgn(t ), where sgn(t ) 0, t 0

2 2
1, t 0

1 FT
1
FT
( ), let sgn(t )
S ( j ) U ( j ) ( ) S ( j )
2
2
d
FT
sgn(t ) 2 (t ) j S ( j ) 2 ( (t )
1);
dt
2
, 0
S ( j 0) 0 because sgn(t) is odd S ( j ) j
0, 0

2
, with the understanding that S ( j 0) 0
j
1
1
U ( j ) ( ) S ( j ) ( )
2
j
It is common to write S ( j )

Research Laboratory

228

Integration

Research Laboratory

229

## Time- and Frequency-shift properties

Time-shift property
Let z (t ) x(t t0 ), find the FT of z (t )

Z ( j ) z (t )e

jt0

jt

dt x(t t0 )e

jt

dt x( )e j ( t0 ) d

x( )e j d e jt0 X ( j )

Z ( j ) X ( j ) , arg{Z ( j )} arg{ X ( j )} t0

e jt0

## A shift in time introduce the phase shift

Research Laboratory

230

## Time- and Frequency-shift properties

Research Laboratory

231

## Ex. 3.41 Finding an FT using the time-shift

property
Use the FT of the rectangular pulse x(t) depicted in
Fig. 3.62(a) to determine the FT of the time-shifted
rectangular pulse z(t) depicted in Fig. 3.62(b)

## Figure 3.62 (p. 281)

Application of the time-shift
property for Example 3.41.

Research Laboratory

232

Sol.
z (t ) x(t T1 ),
Z ( j ) e

jT1

X ( j ) e

jT1

sin(T0 )

## Prob. 3.30, 3.31, 3.32, 3.33

Research Laboratory

233

Prob. 3.31(a)

x(t ) e 2t u (t 3)
Let y (t ) x(t 3) e
Y ( j ) e

2( t 3)

6 2 t

u (t ) e e u (t )

1
j 2

x(t ) y (t 3), X ( j ) e

Research Laboratory

j 3

Y ( j ) e

234

j 3 6

1
j 2

## The time-shifting property may be used to find the

frequency response of a system described by the
difference equation
N

a y[n k ] b x[n k ]
k 0

k 0

z[n k ] e
DTFT

jk

Z (e ) a k (e
j

k 0

H (e j )

Y (e )

j
X (e )

b (e
k 0
N

a (e
k 0

Research Laboratory

j k

j k

235

) Y (e ) bk (e j ) k X (e j )

j k

k 0

j 2

1 2e
H (e )
j
j 3
3 2e 3e
a0 3, a1 2, a3 3, b0 1, b2 2
j

a y[n k ] b x[n k ]
k 0

k 0

## 3 y[n] 2 y[n 1] 3 y[n 3] x[n] 2 x[n 2]

Research Laboratory

236

## Time- and Frequency-shift properties

Frequency-shift property
FT
x(t )
X ( j )

1
1
jt
jt
z (t )
Z
(
j

)
e
d

X
(
j
(

))
e
d

2
2

1
j ( ) t
j t 1
j t
j t

X
(
j

)
e
d

e
X
(
j

)
e
d

e
x(t )
2
2

## A frequency shift corresponds to multiplication in

the time domain by a complex sinusoid whose
frequency is equal to the shift

Research Laboratory

237

## Time- and Frequency-shift properties

Research Laboratory

238

## Ex. 3.42 Finding an FT using the frequencyshift property

Determine the FT of z(t), where
j10 t

e
, t

z (t )

0, t

Prob. 3.34

Research Laboratory

239

## Time- and Frequency-shift properties

Sol.
1, t
x(t )
0, t
FT
x(t )
X ( j )

sin( )

FT
z (t ) e j10t x(t )
X ( j ( 10))
FT
z (t )
X ( j ( 10))

Research Laboratory

2
sin(( 10) )
10

240

Prob. 3.34(a)

Z (e )

1 e

j ( /4)

, 1,

1
x[n ] u[n ] X ( e )
j
1 e
DTFT
j n
j ( )
e x[n ] X (e
)
n

Z (e ) X (e
z[ n ] e

Research Laboratory

n
4

DTFT

j ( /4)

),

x[n ] e

n
4

nu[n ]

241

## Ex. 3.43 Using multiple properties to find an

FT
Determine the FT of x(t), where
d
x(t ) {(e3t u (t ))*(et u (t 2))}
dt

Prob. 3.35

Research Laboratory

242

## Time- and Frequency-shift properties

Sol.
d
d
{(e 3t u (t )) *(e t u (t 2))} {w(t ) * v(t )}
dt
dt
Applying the convolution and differentiation properties from Table 3.5,
x(t )

X ( j ) j{W ( j )V ( j )}
1
1
FT
3t
e u (t )
w(t ) e u (t ) W ( j )
a j
3 j
at

( t 2)

FT

u (t 2) e
FT

j 2

v(t ) e u (t 2) e {e

1
e j 2
1
FT
t

(i.e. e u (t )
)
1 j 1 j
1 j

( t 2)

e j 2
u (t 2)} e
1 j
FT

j e j 2
X ( j ) j{W ( j )V ( j )} e
(3 j )(1 j )
2

Research Laboratory

243

## Finding the inverse Fourier transforms by

using partial-fraction expansions

## The frequency response of a system described by a

linear constant-coefficient differential/difference
equation is given by a ratio of two polynomials in
j/ej.
The partial-fraction expansion can be used to find the
inverse transforms for ratios of polynomials.

Research Laboratory

244

## Inverse Fourier Transform

Suppose X ( j ) is expressed as a ratio of ploynomials in j
bM ( j ) M b1 ( j ) b0
B( j )
X ( j )

N
N 1
( j ) aN 1 ( j ) a1 ( j ) a0 A( j )
Assume that M N , If M N , we may express X ( j ) as
X ( j )

M N

k 0

f k ( j ) k

B( j )
A( j )

FT
(t )
1 and the differentiation property can be used to
M N

## find the inverse FT of

k 0

Research Laboratory

f k ( j ) k

245

## Inverse Fourier Transform

bM ( j ) M b1 ( j ) b0
B ( j )

X ( j )
( j ) N aN 1 ( j ) N 1 a1 ( j ) a0 A( j )
Let the roots of A( j ) be d k 1, 2,

## by determining the roots of v N aN 1v N 1

a1v a0 0.

X ( j )

b ( j )
k

k 1
N

( j d
k 1

FT
e dt u (t )

k
assuming d k are different N

k 1

Ck
j d k

1
, for d 0
j d
N

x(t ) Ck e u (t ) X ( j )
dk t

k 1

Research Laboratory

FT

k 1

Ck
j d k

246

Prob. 3.36

j
1
2
(a) X ( j )

2
( j ) 3 j 2 j 1 j 2
N

x(t ) Ck e dk t u (t ) e t u (t ) 2e 2t u (t )
k 1

5 j 12
2
3
(b) X ( j )

2
( j ) 5 j 6 j 2 j 3
N

x(t ) Ck e dk t u (t ) 2e 2t u (t ) 3e 3t u (t )
k 1

Research Laboratory

247

Multiplication property

Multiplication property
defines the Fourier representation of a product of timedomain signals y(t)=x(t)z(t), x(t) & z(t) are nonperiodic

1
y (t ) x(t ) z (t )
2

1
(2 ) 2

1
X
(
j

)
e
d

j t

X ( j ) Z ( j )e

j ( ) t

Z ( j )e jt d

1
d d
2

1
2

jt
X
(
j

)
Z
(
j
(

)
d

e d

1
1
jt
jt
X
(
j

)
*
Z
(
j

)
e
d

Y
(
j

)
e
d
2

2
1
FT
y (t ) x(t ) z (t )
Y ( j )
X ( j ) * Z ( j )
2
1

## Multiplication in time domain corresponds to convolution

their FT in frequency domain and multiply by 1/(2)

Research Laboratory

248

Multiplication property

Research Laboratory

249

Multiplication property

## The multiplication property enables us to study the effect of

truncating a time-domain signal on its frequency-domain
representation.

## The process of truncating a signal is also known as windowing

1
y (t ) x(t ) w(t ) Y ( j )
X ( j ) *W ( j ),
2
where
FT

1, t T0 FT
2
w(t )
W ( j ) sin(T0 )

0, t T0

## Prob. 3.39, 3.40(b)

Research Laboratory

250

## (b) Convolution of the signal and window FTs resulting

from truncation in time.

## The effect of windowing.

(a) Truncating a signal in time by using a window function w(t).

Research Laboratory

251

Prob. 3.39

x(t )

4
2
2
sin
(2
t
)

z
(t )
2 2
t

2, 2
2
FT
z (t ) sin(2t ) Z ( j )
t
0, otherwise
X ( j )

1
Z ( j ) * Z ( j )
2

Research Laboratory

1
2

252

Prob. 3.40(b)

FT
x (t ) 2 y (t ) z (t )
X ( j ) Y ( j ) * Z ( j )

1, t 1 FT 2sin( )
1, t 2 FT 2sin(2 )
w(t )

, w(t / 2)

0, t 1
0, t 2
FT
w(t t0 )
e jt0W ( j )
2sin( 2)
F 1
Let Y ( j )
, y (t ) e j 2 t [u(t 1) u(t 1)]
2
e j 2 sin(2 )
Let Z( j )

1
[ u ( t 2) u ( t 2)],t t 2
1
t

2
1
F
2
z (t ) w(
)
[u(t ) u(t 4)]
2
2
2
1

j 2t
x (t ) 2 y (t ) z (t ) 2 e [u(t 1) u(t 1)] [u(t ) u(t 4)]
2

## e j 2 t [u(t ) u(t 1)]

1

253

Scaling property
z (t ) x(at )
Z ( j )

z (t )e

jt

dt

x(at )e jt dt

j
1
x( )e a d , a 0

at a
1

1 x( )e j a d , a 0 a
a

FT
Z ( j )
z (t ) x(at )

x( )e

j
a

j
1
X( )
a
a

## Scaling in time-domain introduces inverse scaling in

frequency-domain and an amplitude change

Research Laboratory

254

## Figure 3.70 (p. 300)

Research Laboratory

## The FT scaling property. The

figure assumes that 0 < a < 1.

255

Scaling property

## Ex. 3.48 Scaling a rectangular pulse

1, t 1
1, t 2
Use the FT of x(t )
to find the FT of y(t )
.

0, t 1
0, t 2

Research Laboratory

256

Scaling property
Sol.
1, t 1
2
FT
x(t )
, x(t ) X ( j ) sin( )

0, t 1
2
2
FT
y (t ) x(t / 2)
Y ( j ) 2 X ( j 2 ) 2
sin(2 ) sin(2 )
2

## Figure 3.71 (p. 301)

Application of the FT scaling property in
Example 3.48. (a) Original time signal. (b)
Original FT. (c) Scaled time signal y(t) =
x(t/2). (d) Scaled FT Y(j) = 2X(j2).

Research Laboratory

257

Scaling property

## Ex. 3.49 Using multiple properties to find an

inverse FT
d
e j 2
Find x(t ) if X ( j ) j
{
}
d 1 j ( )
3

## Prob. 3.41, 3.42

Research Laboratory

258

Scaling property
Sol.
1
s (t ) e u (t ) S ( j )
1 j
t

FT

d
e j 2
d
1
d
j
X ( j ) j
{
} j
{e j 2
} j
{e j 2 S ( )}

d 1 j ( )
d
d
3
1 j( )
3
3
j
Let Y ( j ) S ( ) y (t ) 3s(3t ) 3e 3t u (3t ) 3e 3t u (t )
3
j
j 2
Define W ( j ) e S ( ) e j 2Y ( j ) w(t ) y (t 2) 3e 3(t 2)u (t 2)
3
d
Since X ( j ) j
W ( j ) x(t ) tw(t ) 3te 3( t 2)u (t 2)
d
Research Laboratory

259

Scaling property
Consider the periodic signal x(t ) with fundamental period T ,
then z (t ) x(at ) is also periodic with fundamental period T / a.
For convenience, assume a 0.
Z [k ]

T
a

T
a

a
a
jka0t
jka0t
dt
e
)
at
(
x

dt
e
)
t
(
z

T 0
T 0

FS ;a0
Z [k ] X [k ], a 0
z (t ) x(at )

## The FS coefficients of x(t ) and z (t ) are identical.

Scaling operation changes the harmonic spacing from 0 to a0 .
The scaling of discrete-time signals is further addressed in Prob. 3.80
Research Laboratory

260

FT
x(t )
X ( j ) e j e

1
z (t ) x(at ) Z ( j ) X ( j / a)
a
FT

y (t ) x(2t ) Y ( j ) X ( j ( )), a 2
2
2
FT

1
Y ( j ) e
2

Research Laboratory

j ( )
2

1 j 2
e
e
2
2

261

FS ;

x(t )
X [k ] e

jk / 2

ke

2 k

, 0

FS ;a0
z (t ) x(at )
Z [k ] X [k ], a 0

y (t ) x(3t ), Y [k ] X [k ] e

Research Laboratory

jk / 2

262

ke

2 k

, 0 3

Parseval relationships

## The Parsavel relationships state that the energy or

power in the time-domain representation of a signal is
equal to the energy or power in the frequency-domain
representation.
Energy and power are conserved in the Fourier
representation

Research Laboratory

263

Parseval relationships
The energy in a continuous-time nonperiodic signal is

1
*
*
jt
Wx x(t ) dt x(t ) x (t )dt x(t )
X ( j )e d dt

1
1
*
jt
*

X
(
j

)
x
(
t
)
e
dt
d

X
( j ) X ( j )d

2
2

X ( j ) d
2

Research Laboratory

264

Parseval relationships

## The energy in time-domain representation of the

signal is equal to the energy in the frequency-domain
representation, normalized by 2
|X(j)|2 is the energy spectrum of the signal.

## The power or energy spectrum of a signal is defined as the

square of the magnitude spectrum.

Research Laboratory

265

Parseval relationships

Research Laboratory

266

Parseval relationships

## Ex. 3.50 Calculating the energy in a signal

sin(Wn)
,
n
use the Parseval's theorem to evaluate

Let x[n]

n=-

sin 2 (Wn)
x[n]
2 n2
n=-
2

Prob. 3.43

Research Laboratory

267

Parseval relationships
Sol.
1, W
sin(Wn) DTFT
j
x[n]
X (e )
n
0, W

= x[n]
n=-

1
2

Research Laboratory

X (e j ) d

1
2

268

1d

Prob. 3.43(a)

1
d
2
2
j 2

4
j 2

1
2

d
2

X ( j ) d
2

2
X ( j )
x(t ) 2 e 2t u (t )
j 2
1
1
2

X ( j ) d

Research Laboratory

x 2 (t )dt 2 e 2t dt 4 e 4t dt
0

269

Time-bandwidth product

## Compressing a signal in time leads to expansion in

the frequency domain and vice versa
The bandwidth of a signal is the extent of the signals
significant frequency component.

## the width of the mainlobe/2

the width of the mainlobe

## Root-mean-square measures of the effective duration and

bandwidth

Research Laboratory

270

Time-bandwidth product

## Figure 3.72 (p. 305)

Rectangular pulse illustrating the
inverse relationship between the time
and frequency extent of a signal.

Research Laboratory

271

Time-bandwidth product
Assume x(t ) is centered about the origion and is low pass.
The effective duration is defined by
1
2

t 2 x(t ) 2 dt

Td
2

x
(
t
)
dt

## and the effective bandwidth is defined by

w 2 X ( jw ) 2 d w

Bw
2

X
(
j
w
)
d
w

Research Laboratory

1
2

272

Time-bandwidth product
It can be shown that the time-bandwidth product for any signal is
lower bounded according to the relationship
1
Td Bw
(3.65)
2
This bound indicates that we cannot simultaneously decrease the
duration and bandwidth of a signal.
Eq. (3.65) is also known as the uncertainty principle after its application
to modern physics, which exact position and exact momentum of an
electron cannot be determined simultaneously.
Gaussian pulses are the only signals that satisfy this relationship with
equality.
Research Laboratory

273

Time-bandwidth product

## Ex. 3.51 Bounding the bandwidth of a rectangular

pulse

1, t T0
x(t )
, find a lower bound of Bw

0, t T0

Research Laboratory

274

Time-bandwidth product
Sol.
1, t T0
x(t )
,
0, t T0
t 2 x(t ) dt

Td
2

x
(
t
)
dt

1
2

1
2

T0
2
t
dt

T
1
1
T

3
0
T00
t

3
2T0 3 T0
dt
T0

T0

1
1
3
Td Bw Bw

2
2Td 2T0

Research Laboratory

275

Duality

## Figure 3.73 (p. 307)

Duality of rectangular pulses and sinc functions.

Research Laboratory

276

Duality

We observed

## Consistent symmetry between the time- and frequencydomain representation of signals

Symmetries in Fourier representation properties

## Convolution in one domain corresponds to multiplication in the

other domain
Differentiation in in one domain corresponds to multiplication by
the independent variable in the other domain
Etc.

## The symmetries are a consequence of the symmetry

in the definition of time- and frequency-domain
representation.
We can interchange time and frequency. The
interchangeability property is termed duality

Research Laboratory

277

Duality

## The duality property of the FT

1

jwt
dw
e
)
w
j
(
X

)
t
(
x

2
(3.66)
d
e
)

(
z

(
y
let
,

2
X ( jw ) x(t )e jwt dt

1
FT
(3.67)
z (w )
z (w )e jwt dt y (t )
If t , w , y (t )

2
1
FT
jwt
2 y (w ) (3.68)

)
t
(
z

dt
e
)
t
(
z
If w , t , y (w )

2
(3.67) & (3.68) imply symmetry between time and frequency.
FT
F ( jw ),
Given the FT pair: f (t )
FT
2 f (w )
we can have the new FT pair: F ( jt )

Research Laboratory

278

Duality

## Figure 3.74 (p. 309)

Research Laboratory

279

Duality

## Ex. 3.52 Applying duality

Find the FT of x(t )

1
1 jt

## Prob. 3.44, 3.45,

Research Laboratory

280

Duality
Sol.
1
FT
f (t ) e u (t ) F ( jw )
& F ( jt )
2 f ( w )
1 jw
1
FT
F ( jt )

2 f (w ) 2 ew u ( w )
1 jt
t

Research Laboratory

FT

281

Duality

Research Laboratory

282

Duality

## Ex. 3.53 FS-DTFT duality

Determine the inverse DTFT of Fig. 3.75(a)

## Figure 3.75 (p. 311)

Example 3.53. (a) Triangular
spectrum. (b) Inverse DTFT.

Research Laboratory

285

Duality
Sol.
DTFT
FS
Duality property of FS-DTFT: x[n]
X (e j ), X (e jt )
x[k ]
FS
Let z (t ) X (e jt ), if z (t )
Z [k ] then x[n] Z [-n]

## Assuming T 2 , specifically, z (t ) y (t ).Using the time-sift property, we have

2
, k 0

jk / 2
Z [k ] e
Y [k ] = 4 j k -1 sin( k )
2 ,k 0

k
, n 0

## x[n] Z [n] 4 j ( n 1) sin( n ) 4( j ) n 1 sin( n )

2
2 ,k 0

( n) 2
n2

Research Laboratory

286

X ( jw ) u (w )
FT
x(t )
X ( jw )

1
F ( jt ) u (t ) f (w )
(w ) 2 x(w )
jw
1 1
1
j
x(t )
(t )
(t )
(t )
2 j (t )
2 jt
2 t
FT

Research Laboratory

283

## Exploring concepts with MATLAB

Skip

Research Laboratory

287

Summary

## We develop techniques for representing signals as

weighted superposition of complex sinusoids
The weights are a function of the complex sinusoidal
frequencies and provide a frequency-domain
description of the signal.
Four representations

## Fourier representation properties relate the effect of

an action on a signal in the time-domain to a
corresponding change in the frequency-domain
representation