Chapter 4:

Applications of Fourier
Representations
Chih-Wei Liu
Outline
 Introduction
 Fourier Transform of Periodic Signals
 Convolution/Multiplication with (Non-)Periodic Signals
 Fourier Transform of Discrete-Time Signals
 Sampling
 Reconstruction of Continuous-Time Signals
 Discrete-Time Processing of Continuous-Time Signals
 Fourier Series of Finite-Duration Nonperiodic Signals

2
Introduction
 Four classes of signals in Fourier representations
 Continuous- and discrete-time signals
 Periodic and nonperiodc signals
 In order to use Fourier methods to analyze a general system
involving a mixing of noperiodic (says, impulse response) and
periodic (input) signals, we must build bridges between Fourier
representations of different classes of signals
 FT/DTFT are most commonly used for analysis applications
 We must develop FT/DTFT representations of periodic signals
 DTFS is the primary representation used for computational
applications (the only one can be evaluated on a computer)
 Use DTFS to represent the FT, FS, and DTFT

3
FT Representations of Periodic Signals

Recall the FS representation of a periodic signal x(t): x(t )   X [k ]e 0
jk t

k 
 Note that 1 FT
2 ( )
 Using the frequency-shift property, we have e jk0t 
FT
2   k0 
 Let’s take the FT of x(t):
  jk0t 
 
  
FT  x (t )   X [k ]e    X [k ]FT e jk0t
 2  X [k ] (  k0 )
 k    k  k  

 
x (t )   X [k ]e
k 
jk0t

FT
X  j   2  X [k ]   k 
k 
0

◆ The FT of a periodic signal is a series of impulses spaced by the

fundamental frequency 0.
◆ The kth impulse has strength 2X[k]
◆ The shape of X(j) is identical to that of X[k]

4
FS and FT Representations of a
Periodic Continuous-Time Signal

0

X ( j )
6
4
2

0 0 0 0
 2

5
Example 4.1 FT of a Cosine
Find the FT representation of x(t) = cos(ot)
<Sol.> 1
 , k  1
 0   X  k    2
1. FS of x(t): cos  t FS ;0

 0, k 1

2. FT of x(t): cos 0t   X  j      0      0 

FT

cos( 0t ) X ( j )


2 2 0 0
 
0 0

6
Example 4.2 FT of a Unit Impulse Train

Find the FT of the impulse train pt    t  nT 
<Sol.>

n  

1. p(t) = periodic, fundamental period = T

2. FS of p(t):

Pk     t e  jk0t dt  1 / T
1 T /2
T T / 2

3. FT of p(t): P( j)
2 
2
P  j       k 
0 T
T k  


4 2 2 4
  
T T T T

 The FT of an impulse train is also an impulse train.

7
Relating the DTFT to the DTFS
 Recall the DTFS representation of a periodic DT signal x[n]:
N 1
x  n    X [k ]e jk 0n
k 0

 Note that the inverse DTFT of a frequency shifted impulse, i.e. (-k), is
a complex sinusoid. With one period of e jk0n , we have
1 jk0n DTFT
e   (  k0 )       ,    k0  
2
 Let’s construct an infinite series of shifted impulses separated by 2, i.e. to
obtain the following 2-periodic function

1 jk0n DTFT
e    (  k0  m  2 )
2 m  

 Let’s take the DTFT of x[n]:

N 1 N 1 
x[n ]   X [k ]e jk 0 n
 X ( e )  2  X [k ]   (  k0  m  2 )
DTFT j

k 0 k 0 m  
8 Since X[k] is N periodic and N0 = 2, we may combine the two sums
 Relating the DTFT to the DTFS
N 1 
x  n    X [k ]e jk 0n

DTFT
X e j
  2  X [k ]    k  0
k 0 k 

 The DTFT of a periodic signal is a series of impulses spaced by the

fundamental frequency 0

Weighting factor = 2
DTFS DTFT

4
4
2
 The DTFS X[k] and the corresponding
DTFT X(e j) have similar shape.
4 2 4
9 2
Example 4.3 DTFT of a Periodic Signal
Determine the inverse DTFT of the frequency-domain representation depicted in the
following figure, where Ω1 = π /N.

2

 3  2   2 3 4
2
<Sol.>
1. We express one period of X(e j) as

X  e j  
1 1
    1       1  ,      
2j 2j
from which we infer that 2. The inverse DTFT:
 1/  4 j  , k 1 N 1

X  k    1/  4 j  , k  1 x[n ]   X [k ]e jk0n
k 0
 0, otherwise on 1  k  N  2

 1 j1n  j1n 
x n 
1
2 2 j  e  e  
1
 2 sin  1n 
10
 
Outline
 Introduction
 Fourier Transform of Periodic Signals
 Convolution/Multiplication with (Non-)Periodic Signals
 Fourier Transform of Discrete-Time Signals
 Sampling
 Reconstruction of Continuous-Time Signals
 Discrete-Time Processing of Continuous-Time Signals
 Fourier Series of Finite-Duration Nonperiodic Signals

11
Convolution and Multiplication with
Mixtures of Periodic and Nonperiodic Signals
 Example: a periodic signal fed into a stable, nonperiodic impulse response
Periodic input x(t) y(t) = x(t) h(t)
Stable filter
Nonperiodic impulse response h(t)

 FT is applied to continuous-time (CT) case

 DTFT is applied to discrete-time (DT) case

 For CT signals: y  t   x  t   h  t  

FT
Y  j   X  j  H  j 
 If x(t) is CT periodic signal, then

x  t  
FT
X  j   2  X  k      k 
k 
0


y  t   x  t  * h  t   Y  j   2
FT
 X [k ]   k H  j 
k 
0

12
y  t   x  t  * h  t   Y  j   2
FT
 H  jk  X [k ]   k 
k 
0 0
Convolution with Mixtures of Periodic and
Nonperiodic Signals

y  t   x  t  * h  t   Y  j   2
FT
 H  jk  X [k ]   k 
k 
0 0

X ( j ) Magnitude
2  X [0]
2  X [ 2] Y ( j )
adjustment
 0 0
2  X [0] H ( j 0)
 4 0  2 0 2 0 4 0 2  X [2 ] H ( j 2  0 )
2  X [1]  0 0

H ( j )  4 0  2 0 2 0 4 0

2  X [1] H ( j 0 )


 4 0  2 0   0  0 2 0 4 0

13
 Example 4.4
A periodic square wave is applied to a system with impulse response h(t)=(1/(t))sin(t.
Use the convolution property to find the output of this system.

<Sol.>

1. The frequency response of the LTI system is by taking the FT of the impulse response h(t):
1,  
h  t  
FT
H  j    H ( j )

0,  
2. FT of the periodic square wave:   

2 sin k / 2   
X  j   
X ( j )
k  
k   k  2
3. The FT of the system output is Y(j) = H(j)X(j): 
     
Y  j   2          2      
2 2


 2  2
H(j) acts as a low-pass filter, passing harmonics 
at /2, 0, and /2, while suppressing all others.

4. Inverse
14 FT of Y(j): y t   1 / 2   2 /   cos t / 2  
 
2 2
Convolution with
Mixtures of Periodic and Nonperiodic Signals
Periodic input x[n] y[n] = x[n] h[n]
Stable filter
Nonperiodic impulse response h[n]

 For DT signals:
 If x[n] is DT periodic signal, then
N 1 
x  n    X [k ]e jk 0n

DTFT
X e j
  2  X [k ]    k  0
k 0 k 

y  n   x  n  * b  n   Y  e
DTFT j
  2  H  e  X [k ]    k .
jk 0
0
k 

 The form of Y(e j) indicates that y[n] is also periodic with the same period as x[n].

15
Multiplication of
Periodic and Nonperiodic Signals
 Continuous-time case:
 Original multiplication property:
1
y  t   g  t  x  t  
FT
Y  j   G  j   X ( j )
2
 If x(t) is periodic. The FT of x(t) is
 
x (t )   X [k ]e
k 
jk0t

FT
X  j   2  X [k ]   k 
k 
0


y  t   g  t  x  t   Y  j   G ( j )   X [k ] (  k0 )
FT

k 

y  t   g  t  x  t   Y  j  
FT
 X [k ]G  j(  k ) 
k 
0

Multiplication of g(t) with the periodic function x(t) gives an FT consisting of a

weighted sum of shifted versions of G(j)
16
 
y  t   g  t  x  t   Y  j  
FT
 X [k ]G  j(  k ) 
k 
0

G ( j )

Y ( j )
G ( j ) G ( j )

X ( j )  2 0 2 0

 0 0

 2 0 2 0

 0 0 y(t) becomes nonperiodic signal !



The product of periodic and

nonperiodic signals is nonperiodic

17
Example 4.4
Consider a system with output y(t) = g(t)x(t).
Let x(t) be the square wave and g(t) = cos(t /2), Find Y(j) in terms of G(j).
<Sol.>
 FS representation of the square wave:
sin  k / 2 
x  t  
FS ; / 2
 X k  
k
And, we have G  j      1 / 2      1 / 2 


sin k / 2
Y  j       1 / 2  k / 2     1 / 2  k / 2
k   k

18
 Example 4.6 AM Radio
A simplified AM transmitter and receiver are depicted in Fig. 4.13(a). The effect of propagation
and channel noise are ignored in this system. The signal at the receiver antenna, r(t), is
assumed equal to the transmitted signal. The passband of the low-pass filter in the receiver is
equal to the message bandwidth,  W < < W. Analyze this system in the frequency domain.

<Sol.>

1. The transmitted signal is expressed as

r  t   m  t  cos ct  
FT
R  j   1/ 2  M  j   c    1/ 2  M  j   c  
19
 G ( j )


 2C 2C

Y ( j )
R ( j )

 
C C

g  t   r  t  cos ct  

FT
G  j   1/ 2  R  j   c    1/ 2  R  j   c  

G  j   1/ 4  M  j   2c    1/ 2  M  j     1/ 4  M  j   2c  

The original message is recovered by low-pass filtering to remove the message

replicates centered at twice the carrier frequency.

20
Multiplication of
Periodic and Nonperiodic Signals
 Discrete-time case:
 Original multiplication property:
1
y[n ]  x[n ]z[n ]   Y ( e j ) 
DTFT
X ( e j )  Z ( e j )
2
 If x[n] is periodic. The DTFS of x[n] and its DTFT is
N 1 
x  n    X [k ]e jk 0n

DTFT
X e j
  2  X [k ]    k  0
k 0 k 

 
1
j
Y (e )  X ( e )  Z ( e )    X [k ] (  k0 ) Z ( e j (  ) )d
j j

2 
k  
In any 2 interval of , there are exactly N multiples of the form (  0),
since 0= 2/N.
N 1  N 1
Y ( e )   X [k ]  (  k0 ) Z ( e
j j (   )
)d   X [k ]Z ( e j ( k0  ) )

k 0 k 0

   X [k ]Z  e  
N 1
y  n   x  n  z  n   Y  e
DTFT j j  k 0 
21
k 0
 Example 4.7 Windowing Effect
Windowing (or truncating): one common data-processing applications, which access only
to a portion of a data record.
Consider the signal xn  cos
 7   9 
n   cos n 
 16   16 
Using only the 2M+1 values of x[n], n≦ M, evaluate the effect of computing the DTFT.
<Sol.>
N 1 
1. Since x  n    X [k ]e jk 0n
DTFT
X e j
  2  X [k ]    k  0
k 0 k 

Then X e j      9       7       7       9 
 
 16   16   16   16 
which consists of impulses at ± 7/16 and ± 9/16.
1, n M
2. Define a signal y[n] = x[n]w[n], where wn  
0, n M
1
y[n ]  x[n ]z[n ]   Y ( e j ) 
DTFT
X ( e j )  Z ( e j )
2
22 Y e   W e 
j 1
2
 
j  9 /16 
W e  
j  7 /16 

W e 
j  7 /16 
 
W e 
j  9 /16 
  
3. The windowing introduces replicas of W(e j ) centered at the frequencies 7/16 and
9/16, instead of the impulses that are present in X(e j ) .
4. We may view this state of affairs as a smearing of broadening of the original impulses
5. The energy in Y(e j ) is now smeared over a band centered on the frequencies of the
cosines

Effect of windowing a data record. Y(e j) for

different values of M, assuming that 1 = 7/16
and 2 = 9/16.
(a) M = 80, (b) M = 12, (c) M = 8.

 If the number of available data points is small

relative to the frequency separation, the DTFT
is unable to distinguish the presence of two
distinct
23 sinusoids (e.g. M=8)
Fourier Transform of Discrete-Time Signals
 A mixtures of discrete-time and continuous-time signals
 By incorporating discrete-time impulses into the description of the signal in
the appropriate manner.
Basic concepts:
Consider the following signals: x (t )  e jt and g [ n ]  e jn
Let’s force g[n] to be equal to the samples of x(t) with sample period Ts; i.e., g[n] = x(nTs).

e jn  e jTs n i.e.  = Ts


Now, consider the DTFT of a DT signal x[n]: X e     x  ne
j  jn

n 
Define the continuous time signal x(t) with the Fourier transform X(j):

X   j   X  e j
|
Ts   x  n e  jTs n
n 

Taking the inverse FT of X(j), using the FT pair   t  nTs   e  jTs n
FT

24
x  t    x n   t  nT  .
n 
s
Relating the FT to the DTFT
 
x  t   
n 
x  n   t  nTs  
FT
X   j   
n 
x  n  e  jTsn ,

◆ x(t)  a CT signal corresponds to x[n];

X(j)  Fourier transform corresponds to the CT Fourier transform X(e j)
The DTFT X(e j) is periodic in  while the FT X(j) is 2/Ts periodic in .

2 2

Ts

X d ( j )
2


2 2
25 Ts Ts
Example 4.8
Determine the FT pair associated with the DTFT pair

X  e j  
1
x  n   a n u  n  
DTFT

1  ae  j
This pair is derived in Example 3.17 assuming that |a| < 1 so the DTFT converges.
<Sol.> 
1. We first define the continuous-time signal x t     t  nTs 
n
a
n 0
2. Using =  Ts gives
1
x  t  
FT
X   j  
1  ae  jTs

26
Relating the FT to the DTFS
 Suppose that x[n] is an N-periodic signal, then
N 1 
x  n    X [k ]e jk 0n

DTFT
X e j
  2  X [k ]    k 
0
k 0 k 

where X[k] = DTFS coefficients

 Now, define x(t)  a CT signal corresponds to x[n], then the Fourier
transform X(j) of the continuous time signal x(t) is

X  ( j )  X ( e )
j
 Ts
 2  X [k ] (T
k  
s  k 0 )

0
 2 
k  
X [k ] (Ts (  k
Ts
))

Use the scaling property of the impulse, (a) = (1/a)(), to rewrite X(j) as

2  k 0 
X   j      T 
X k   
Ts k   s 

27
 
2  k 0 
x (t )   x[n ] (t  nTs )  X  ( j ) 
FT
 X [k ]    T 

n   Ts  s 

1. Recall that X[k] is N-periodic, which implies that X(j) is periodic with periodic
N0/Ts = 2/Ts
2. Recall that x[n] is N-periodic, which implies that x(t) is periodic with periodic NTs

0

X d ( j )

2 2

Ts 2 Ts 
NTs
X[k] and x[n] are N-periodic functions
x(t) and X(j) are periodic impulse trains with period NTs and 2/Ts.
28
Outline
 Introduction
 Fourier Transform of Periodic Signals
 Convolution/Multiplication with (Non-)Periodic Signals
 Fourier Transform of Discrete-Time Signals
 Sampling
 Reconstruction of Continuous-Time Signals
 Discrete-Time Processing of Continuous-Time Signals
 Fourier Series of Finite-Duration Nonperiodic Signals

29
Sampling
 We use FT representation of discrete-time signals to analyze the effects of
uniformly sampling a signal.
 Sampling the (continuous-time) signal is often performed in order to
manipulate the signal on a computer.

CT signals DT signals
Sampling
x[n ]  x (nTs )
 Let x(t) be a CT signal and x[n], a DT signal, is the “samples” of x(t) at
integer multiples of a sampling interval. 
p  t      t  nTs .
n 


x  t    xn  t  nT 
s

30
= n  
Sampling Process
 We note that
 
x (t )   x[n] (t  nT )   x(nT ) (t  nT )
n  
s
n  
s s

 
  x(t ) (t  nT )  x(t )   (t  nT )  x(t ) p(t )
n  
s
n  
s

 Then 1
x (t )  x (t ) p(t ) 
FT
X  ( j )  X ( j )  P( j )
2
Example 4.2
1 2 
X  ( j ) 
2
X ( j ) 
Ts
  (  k )
k  
s s = 2/Ts

1

Ts
 X ( j(  k ))
k  
s

An infinite sum of shifted versions of the original signal, X(j)

31
 Sampling Theorem
 
1
x (t )   x[n ] (t  nTs )  X  ( j ) 
FT
 X ( j(  k )) s
n   Ts k  

 The shifted version of X(j) may overlap with each other if s is not large
enough, compared with the bandwidth of X(j), i.e. W.

X ( j )
X d ( j)

 
 2s  s s 2s
sW < W
X d ( j ) sW > W X d ( j)
aliasing

 
 2 s s ss
 2 s 3s  2s  s s 2s 3s
s  W

As Ts increases or s decreases, the shifted replicas of X(j) moves closer together, finally
32
overlapping one another when s < 2W.
 Example 4.9
Consider the effect of sampling the sinusoidal signal x  t   cos  t 
Determine the FT of the sampled signal for Ts = 1 / 4.
 
1
<Sol.>
Recall that
x (t )  
n  
x[n ] (t  nTs ) 
FT
X  ( j ) 
T
 X ( j(  k )) s
s k  
1. First find that X ( j )   (   )   (   )
2. Then, we have X   j  
 

Ts
      k        k 
k  
s s

which consists of pairs of impulses separated by 2, centered on integer multiples of

s = 2/Ts

Ts = ¼.

33
 X d ( j )

2
Ts = 1

 4   2 2 4

for Ts = 3/2
X d ( j )
2 / 3


 3  2  
   2 3
3 3

34
Down-Sampling:
Sampling Discrete-Time Signals
 
1
 Recall x (t ) 
 
n  
x[ n ] ( t  nTs )  X FT
 ( j ) 
Ts
 X ( j(  k ))
k  
s

 Let y[n] = x[nq] be a subsampled version of x[n], q is positive integer.


y (t )   y[n] (t  nT ' )
n  
s
s '  s q
 
1 k
  x[nq] (t  nqTs )  Y ( j ) 
FT

qTs

k  
X ( j ( 
q
s ))
n  

k m
 Let  l  , 0  m  q  1 , then
q q
1 
k 1 q 1  1 
m 
Y ( j )   X ( j (  s ))   
q m 0  Ts
 X ( j (  ls  s ) 
qTs k   q l   q 
1 q 1 m
  X  ( j (  s ))
q m 0 q
35
Down-Sampling:
Sampling Discrete-Time Signals
 Since y[n ] 
DTFT
 
 Y e j  Y ( j )     Y ( j )    , therefore
Ts ' qTs

1 q 1 m 1 q 1  m
Y ( e )   X  ( j (  s ))
j
  X  ( j(  s ))
q m 0 q 
 q m 0 qTs q
qTs

1 q 1  m 2 1 q1 j  m
  X  ( j(  ))   X  ( (  2 ))
q m 0 qTs q Ts q m 0 Ts q q

Use X e  X
j
( j )    , we have 1 q 1 1
j (   2 m )
 
Ts Y ( e j
)  
q m 0
X (e q
)

 Summary
 
1 k
y (t )   x[nq] (t  nqTs )  Y ( j ) 
FT
 X ( j (  s ))
n   qTs k   q
1
1 q 1
y[n ]  x[ qn ]  Y e    X ( e q
j (   2 m )
DTFT j
)
36
q m 0
 Down-Sampling by q
X q (e j (( q1)2 ) )

2 2 2 (q  1) 2 2 (q  1)


q

4 2   2 4
2 q 2 q

X d (e j (2 ) ) 
q

2

2 (q 1) 2  Wq 2  Wq 2 (q 1) 4 3 2   2 3 4

Figure 4.29
Effect of subsampling on the DTFT. (a) Original signal spectrum. (b) m = 0 term in Eq. (4.27)
(c) m = 1 term in Eq. (4.27). (d) m = q – 1 term in Eq. (4.27). (e) Y(ej), assuming that W < /q.
(f) Y(ej), assuming that W > /q.

 If the highest frequency component of X (e j ), W, is less than /q, the

37
aliasing can be prevented.
Reconstruction of Sampled Signal
 Reconstruction?

 Not so easy !! The samples of a signal do not always uniquely determine the
corresponding continuous-time signal

 Nyquist Sampling Theorem

Let x(t ) 
FT
 X ( j ) be a band-limited signal, i.e. X(j)=0 for >m.
If s > 2 m, where s = 2/Ts is the sampling frequency, then x(t) is uniquely
determined by its samples x(nTs), n = 0,  1,  2, … .
38
Ideal/Perfect Reconstruction
 If the signal is not band-limited, an antialiasing filter is necessary
 Hereafter, we consider a band-limited signal
 Suppose that x(t ) 
FT
 X ( j ), then the FT of the sampled signal is

1
X  ( j ) 
Ts
 X ( j  jk )
k 
s

X  ( j )
X ( j )

 
 m m  s  m m s
(a)
Ideal low-pass filter
H r ( j )
X ( j )  X  ( j ) H r ( j )
 Ideal reconstruction:
Ts ,   s 2 
H r ( j )    s / 2 s / 2
 0 otherwise Non-causal !!
39 s: Nyquist frequency the perfect-reconstruction cannot be
implemented in any practical system
 X ( j )  X  ( j ) H r ( j ) x(t )  x (t )  hr (t )
H r ( j )

s t
Ts sin( ) t
 2
hr (t )   sin c( )
 s / 2 s / 2
t Ts
 H r ( j )
hr (t )  FT

  
x (t )  x (t )  hr (t )    x[n ] (t  nTs )   hr (t )
 n   
 
t  nTs
  x[n]hr (t  nTs ) 
n  

n  
x[n ] sin c(
Ts
)

X  (t )

Infinite length of hr(t) =

Non-causal filter
Perfect reconstruction

40
A Practical Reconstruction –
Zero-Order Hold Filter
 The zero-order hold filter, which holds the input value for Ts seconds.

a stair-step approximation to the original signal

 Impulse response (rectangular pulse):
1, 0  t  Ts sin(Ts / 2)
ho (t )   h0 (t ) 
FT
H 0 ( j )  2e  jTs / 2
0, t  0, t  Ts 

1. A linear phase shift corresponding to a time delay of Ts/2 seconds

2. A distortion of the portion of X(j) between  m and m. [The distortion is
produced by the curvature of the mainlobe of Ho(j).]
3. Distorted and attenuated versions of the images of X(j), centered at nonzero
multiples of s.
41
 X ( j ) H O ( j )


 m m 
 s s 2 s

X  ( j ) arg  H O ( j )



  s s 2 s
 s  m m s


X O ( j )


 s  m m s 2 s

 Ts seconds holding for x[n]  Time shift of Ts/2 in xo(t) for distortion-1
 Stair-step approximation  main reason for distortion-2 & 3
 1. Using Equalizer for non-flattened mag. in passband
2. Using anti-image filter
 Compensation filter
Both distortions 1 and 2 are reduced by increasing s or, equivalently, decreasing Ts,
42 e.g. oversampling !!
Compensation Filter
 The frequency response: Equalizer
H c ( j )

Anti-imaging

m  s  m m s  m

 Ts
 2sin(T / 2) ,   m
H c ( j )   s
 0,    s  m


43
 Example 4.13
In this example, we explore the benefits of oversampling in reconstructing a continuous-time
audio signal using an audio compact disc player. Assume that the maximum signal frequency is
fm = 20 kHz. Consider two cases:
(a) reconstruction using the standard digital audio rate of 1/Ts1 = 44.1 kHz, and
(b) reconstruction using eight-times oversampling, for an effective sampling rate of 1/Ts2 =
352.8 kHz.
In each case, determine the constraints on the magnitude response of an anti-imaging filter so
that the overall magnitude response of the zero-order hold reconstruction system is between
0.99 and 1.01 in the signal passband and the images of the original signal’s spectrum centered
at multiples of the sampling frequency are attenuated by a factor of 10 3 or more.
<Sol.>
0.99 1.01
 H 'c ( jf )  , 20 kHz  f  20 kHz
H 'o ( jf ) H 'o ( jf )

44
45
Outline
 Introduction
 Fourier Transform of Periodic Signals
 Convolution/Multiplication with (Non-)Periodic Signals
 Fourier Transform of Discrete-Time Signals
 Sampling
 Reconstruction of Continuous-Time Signals
 Discrete-Time Processing of Continuous-Time Signals
 Fourier Series of Finite-Duration Nonperiodic Signals

46
Discrete-Time Signal Processing
 Several advantages for DSP algorithm
 Signal manipulations are more easily performed by using arithmetic
operations of a computer than through the use of analog components
 DSP systems are easily modified in real-time
 Direct dependence of the dynamic range and signal-to-noise ratio on the
number of bits used to represent the discrete-time signal
 Easily implement the decimation and the interpolation

H a ( j ) H O ( j ) H c ( j )

G ( j )

47
System Response Analysis
 Assume that the discrete-time processing operation is represented by a
DT system with frequency response H(e j)  = Ts , Ts  sampling interval

H a ( j ) H O ( j ) H c ( j )

X a ( j )  H a ( j ) X ( j )


 X  j    k  
1
X  ( j )  a s
Ts k 


 H  j    k   X  j    k  
1
 a s s
Ts k 


Y ( j )  H  e   H a  j   ks  X  j   ks  
1 jTs

Ts k 


Y ( j )  H o ( j ) H c ( j ) H  e   H a  j   ks  X  j   ks  
1 jTs

48 Ts k 
The anti-imaging filter Hc(j) eliminates frequency components above s/2, hence
eliminating all the terms in the infinite sum except for the k = 0 term.
Therefore, we have

H o ( j ) H c ( j ) H  e jTs  H a ( j ) X ( j )
1
Y ( j ) 
Ts
The overall system is equivalent to a continuous-time LTI system having the response:

H o ( j ) H c ( j ) H  e jTs  H a ( j )
1
G ( j ) 
Ts

 If we choose the anti-aliasing and anti-imaging filters such that

(1/ Ts ) H o ( j ) H c ( j ) H a ( j )  1 G( j )  H (e jTs )
That is, we may implement a CT system in DT by choosing sampling parameters
appropriately and designing a corresponding DT system.

49
Oversampling A high sampling rate leads to high computation cost

 Oversampling can relax the requirements of the anti-aliasing filter as well as

the anti-image filter (a wide transition band)
 Anti-aliasing filter is used to prevent aliasing by limiting the bandwidth of
the signal prior to sampling.
Ws  stopband of filter,
Wt = Ws  W  width of transition band.

s-W s-W > Ws Wt = Ws  W

< s- 2W
50 
s-Ws s
Decimation
 Two sampled signals x1[n] and x2[n] with different sampling intervals Ts1 and
Ts2 X ( j )

Sampling !!

4 2 2 4

x2[n] g[n]
4 2 2 4

q-fold spread out

4 2 2 4
51
The maximum frequency component of X2(e j ) satisfies WTs2 < /q
Decimation Filter
 Decimation filter hd[n] : a low-pass filter prevents from aliasing problem
when downsampling by q.

 
j 1, Ts 
H d (e )   q
0, otherwise

52
Interpolation
 Interpolation increases the sampling rate and requires that we somehow
produce values between two samples of the signal

2   2

q-fold squeeze

2 
4

2

WTs1 WTs1 2 4 2
q q q q q q
53 Image occurs !!
Interpolation Filter
 Decimation filter hi[n] : a low-pass filter prevents from image problem when
upsampling by q.

Discrete-time
Upsampling xi[n]
x[n] by q
low pass
Hi(ej)

 
 1,  T 
H i ( e j )   s
q
0, otherwise

 2   2

q q

 2 2
54
Outline
 Introduction
 Fourier Transform of Periodic Signals
 Convolution/Multiplication with (Non-)Periodic Signals
 Fourier Transform of Discrete-Time Signals
 Sampling
 Reconstruction of Continuous-Time Signals
 Discrete-Time Processing of Continuous-Time Signals
 Fourier Series of Finite-Duration Nonperiodic Signals

55
 Why DTFS for Finite Nonperiodic Signals?
 The primary motivation is for the numerical computation of Fourier
transform
 DTFS is only Fourier representation that can be evaluated numerically !!
 Let x[n] be a finite-duration signal of length M
 Introduce a periodic DT signal x[n ] with period NM
zero padding

M 1
j
X (e )   x n e
n 0
 j n

56
2 2
Relating the DTFS to the DTFT
 Consider a periodic DT signal x[n ] with period NM, then
N 1 M 1
1 1
 
~ ~  jk0 n ~  jk 0n
X [k ]  x [ n ]e Apply x[n], we have X [k ]  x[ n ]e
N n 0  = 2/N
M 1 o
N n 0


j
Recall that X ( e )   x n e
n 0
 j n
, we conclude that

~ 1
X [k ]  X ( e j )
N   k 0

The DTFS coefficients of x[n ] are samples of the DTFT of x[n], divided by N and
evaluated at intervals of 2/N.

 The effect of sampling the DTFT of a finite-duration nonperiodic

signal is to periodically extend the signal in the time domain
 Dual to sampling in time domain
In order to prevent overlap in time domain, we requires NM, or the sampling
frequency 02/M
57
Example 4.14  3 
 
cos n  , 0  n  31
Consider the signal x  n     8 

 0, otherwise
Derive both the DTFT, X(e j ), and the DTFS, X[k], of x[n] , assuming a period N > 31.
Evaluate and plot |X(e j )| and N|X[k]| for N = 32, 60, and 120.
<Sol.>
1. We rewrite the signal x[n] as g[n]w[n], where g[n] = cos(3n/8) and a rectangular window
1, 0  n  31 DTFT sin 16 
w n   W  e j   e  j 31 / 2
0, otherwise sin( 2)
2. Then j 1
X (e )  G ( e j )  W ( e j )
2
e  j 31( 3 / 8) / 2 sin 16    3 8  e  j 31( 3 / 8) / 2 sin 16    3 8 
X e  
j

2 sin     3 8 2  2 sin     3 8 2 
3. Sample at =k0 and divide by N, we have

58
  e  j ( ko 3 / 8)16  e j ( ko 3 / 8)16  e  j ( ko 3 / 8)16
X k     j ( k 3 / 8)2  j ( k 3 / 8)2
  j ( k o  3 / 8)  e
1 o
e o

 2 Ne 2

 e  j ( ko 3 / 8)16  e j ( ko 3 / 8)16  e  j ( ko 3 / 8)16
  j ( k o 3 / 8) / 2  j ( k 0 3 / 8)2  j ( k o 3 / 8)2
 2 Ne  e  e
 e  j 31( ko 3 / 8) / 2  sin 16  k o  3 8 
 
 2N  sin   k o  3 8 2 
 e  j 31( ko 3 / 8) / 2  sin 16  k o  3 8 
 
 2N  sin   k o  3 8 2 

59
Relating the FS to the FT
 The relationship between the FS and the FT of a finite-duration non-
periodic continuous-time signal is analogous to that of discrete-time case
 Let x(t) have duration T0, so that x (t )  0, t  0 or t  T0

 Construct a periodic signal x (t )   x (t  mT ) with T  To
m 
 Consider the FS of x (t )
1 T 11 TTo0

jkot jkjko0tt
X [k ]   x (t )e
  dt   
x
x (
( tt))ee dt
dt
T 0 TT 0 0

 Recall that 
 X  j    x (t )e
T
x (t ) 
DT  jt
dt   x (t )e  jt dt
 0

1
X [k ]  X ( j )
T  ko

The FS coefficients are samples of the FT, normalized by T

60