Chapter

Discrete Time
Fourier Transform 8
8.1 Introduction
Discrete time Fourier Transform (DTFT) is used to perform frequency analysis of periodic and non-
periodic discrete time signals. We have already learned about the z-Transform and we know that discrete
frequency axis is always periodic with period 2. In this chapter we will study all properties of DTFT of discrete
time signals

8.2 DTFT representation of non periodic discreate-time signals

The periodic discrete time signals are represented using DTFS. For periodic signal with period N has
DTFS coefficients Xn which are discrete and periodic with period N and

x[n]   Xk e jonk
k  N 

We can assume a non periodic signal with N   i.e. Infinite period and o  0 thus the discrete
components will come close to each other and will form a continuous plot. Thus DTFT is a continuous plot
which is periodic with period 2. We can write
1
x[n]  X( e j )e jn d ...(1)
2 2

and X(e j)   x[n]e jn ...(2)
n

REMEMBER

• Thus DTFT convert a non periodic discrete signal into continuous periodic signal with period 2
x[n]  X(e j)
• From eq.(2) we can see that DTFT of a signal is z-transform at unit circle
DTFT = X(e j)  X ( z )|
z e j
Thus if a signal is absolutely summable or ROC of it’s z-transform contain unit circle then best method is to
find X(z) and replace z by e j

GATE MASTERS PUBLICATION

 370 Signals & Systems

8.3 Periodicity of DTFT

The DTFT is a periodic function in w with a period 2, i.e.,
X( e j(  2 ) ) = X(ej) …(3)
Proof We know that

X(e j )   x ( n )e  j n
n 
For any integer k, we have

X( e j(  2 k ) )   x( n)e j( 2 k )n
n

  x( n)e jne j 2kn
n

X( e j(  2 k ) )   x( n)e jn
n

X( e j(  2 k ) ) = X(e j)
where we have used the fact that e–j2kn = 1. Hence X(e j) is periodic with period 2. But this property
is just a consequence of the fact that frequency range for any discrete-time signal is unique over the frequency
interval of (–, ) or , equivalently, (0, 2), and any frequency outside this interval is equivalent to a frequency
within this interval.

8.4 Convergence of DTFT

There is one sufficient but not necessary condition for existence of DTFT of signal x[n]. The condition is
that signal should be absolutely summable that is

 x( n)   ...(4)
n
Thus if ROC of z-transform of x[n] has unit circle then we can directly find DTFT by replacing z by e j.

Study Note
Only one condition exist for existence of DTFT of x[n]. There is no condition on number of discontinuity in x[n]
or finite number of maxima and minima because x[n] is discrete.

Example 8.1

Find the Fourier transform of the unit sample squence x ( n ) = ( n )

Solution 8.1
The Fourier transform X(e j) of the sequence x(n) is given by

X(e j )   x( n)e jn
n


  ( n)e jn  e jn 1
n n0

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 371

(n)  1 …(5)

x (n ) X(ej)
1
1

–4 –3 –2 –1 1 2 3 4 n –  

(a) (b)
Fig (a) Unit impulse sequence and (b) its Fourier transform

Note: The unit sample sequence has a Fourier transform representation consisting of equal contributions at all
frequencies. The unit sample sequence and its Fourier transform are shown in Fig. (a) and (b), respectively.

Example 8.2

Find the Fourier transform of the causal sequence [shown in Fig. (a)]
x ( n ) = a n u (n ), | a | < 1.
x(n) x(n)

1 0a<1 1 a 0

1/a
2
1/a

1 3 5 7 9
0 1 2 3 4 n 0 2 4 6 8 n

–1/a

(a)
Fig 8.2 : (a) Signal x(n)

Solution 8.2
We can see that the signals are absolutely summable and the z-transform is
1
X (z ) ,
1  az1
pole of X(z) is at z = a and signal is right sided, thus ROC is |z| > |a|

1
 X (z )  , | z|  | a |
1  az1
Since |a| < 1 thus ROC of X(z) has unit circle and DTFT will be

1
X( z)|z  e j 
1  ae j
To find magnitude and phase response of X(e j) we will replace
e j  = (cos  + j sin )
1
Thus X( e j) 
1  a(cos   j sin )

GATE MASTERS PUBLICATION

372 Signals & Systems

1
Magnitude of X(e j) 
(1  a cos )2  ( a sin )2
 a sin  
and phase of X(e j) is  tan1  
 1  a cos  
The phase and magnitude plot both will be periodic with period 2.

Example 8.3

Find the Fourier transform of the sequence

x ( n ) = a n u( –n – 1), | a | > 1.
x (n)

… –4 –3 –2 –1 0 n

Solution 8.3
The z-transform of x[n] is

X (z )   a nz n
n
 n
z
  
n1  a 

 z/a 
X (z )   
1 z / a 
The pole of X(z) is z = a, since signal is left sided signal the ROC will be |z| < |a|.
Now |a| > 1 thus ROC contain unit circle
Thus X( e j)  X( z)|z  e j
z

a  z z e j 

e j

a  e j

 cos   j sin  
 X(e j )   
 a  cos   j sin  

1
The magnitude of X(e j) is
( a  cos )2  (sin )2

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 373

 sin    sin  
and phase of X(e j) is tan 1    tan 1  
 cos    a  cos  

Example 8.4

Find the fourier transform of signal

x ( n ) = a |n | , | a | < 1.
Solution 8.4
When x[n] = a|n| and |a| < 1 then signal x[n] is absolutely summable and
x[n] = anu[n] + a–nu[–n – 1]
ztransform 1
a n u[ n] ,ROC| z |  | a |
1  az1
z  transform az
and a nu [n  1]   ,ROC| z |  |1 / a |
1  az
1 az
 a n u[ n]  a  n u[ n  1]  1

1  az 1  az
and ROC of X(z) is |a| < |z| < |1/a|
Since |a| < 1 thus ROC contain unit circle
1 ae j
 X(e j)  X( z)|z  e j  
1  ae j 1  ae j
1  ae j  ae j  a 2

1  ae j  ae j  a 2
1  a2
 X(e j ) = 
1  2a cos   a 2

Example 8.5

Find the fourier transform of rectangular pulse

1, | n|  N1
x ( n ) = u[n + N1] – u[n – N1 – 1] = 
0, | n|> N1
Solution 8.5
Since x[n] is non zero only for finite number of n thus it is absolutely summable.
Since x[n] = u[n + N1] – u[n – N1 – 1]

 1   N1 ( N1  1)  zN1
 X (z )   1  
z  z   1
(1  z2 N1 1 )
 1  z  (1  z )

Thus the pole at z = 1 will be cancelled by zero at z = 1 because 1  z2N1 1 will give 2N1 + 1 zeros on cicle
of unit radius.
Thus ROC of X(z) is entire z plane except z =  and ROC contain unit circle.
(1  e(2N1 1) j )
 X( e j)  X( z)|z e j  e jN1
(1  e j )
Taking e j( N1 1/2) comman from numerator and e–j/2 from denominator..

GATE MASTERS PUBLICATION

374 Signals & Systems

sin( ( N1  1 / 2))
 X(e j) 
sin(  / 2)

Example 8.6

Find the inverse fourier transform of


X ( e j) =  2(   2m)
m=

Solution 8.6
The plot of X(e j) is
X (e j )

2
… …

–2 0 2 4 6 
The DTFT is periodic with period 2. The signal x[n] is

x[n]  1
 j
X (e )e  j nd 
2 
For one period between – and  X(e j) = 2()

1 jn
 x[n]   2()e d  1
2 
 1 2() (  to  )

or 1 2  (  2m)
m

REMEMBER

• We can see that the DTFT of 1 exist, z-transform of 1 do not exist


1  2(   2m )
m 

• Signal x[n] = 1 is not absolutely summable thus we cannot find result by applying the eq.(2)

Example 8.7

Find the DTFT of unit step function u [ n ]

Solution 8.7
Since signal x[n] = u[n] is not absolutely summable we cannot apply eq.(2) to find X(e j) and neither
X( z)|z e j because ROC of z-transform of u[n] do not contain unit circle. We will use another method to

find X(e j) of u[n].

1
Let f [n]  for   n  
2

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 375

1
 for n0
and g[n]   2
 1 for n0
 2
So g[n] – g[n – 1] = [n]
If g[ n]  G( e j )

g[ n  1]  G( e j )e j

[ n]1

1
Thus G(ej)=
1e j


and f [ n]   (   2m)
m

Since u[n] = f [n] + g[n]

j j
 u[ n]  F( e )  G( e )

1

1  e j
   (  2m)
m

Thus for – <  <  we get

1
u[ n]  ()
1  e j

Example 8.8

Find the inverse DTFT of the rectangular pulse spectrum defined only for –    

   1, |  |  c
X ( e j  )  rect  
 2c  0, c  |  |  

which is depicted in Fig. (a). Alternatively, we can define X ( e j) over all  by writing it as an
infinite sum of rectangle functions shifted by integer multiple of 2
    2 m 
X ( e j)   rect  
m  2 c 

Solution 8.8
To find x[n] we can apply equation

1
x[n]   X( e j )e jn d
2 
for – <  <  we have
  
X( e j)  rect  
 2c 

GATE MASTERS PUBLICATION

 376 Signals & Systems


1
x[n]  e jn d
2 

c

1
 ( e jcn  e jcn )
2jn

sin(c n)
x[n] 
n

sin( c n)   
  rect   ,     
n  2c 
The signal x(n) is depicted in Fig. (b)
x (n )
X(ej)
c

1

–2 – –c 0 c  2  0

(a) (b)
Fig : (a) Rectangular pulse spectrum and (b) its inverse Fourier transform

j   
REMEMBER • Thus a discrete low pass filter with cut off frequency c is H(e )  rect  2  ,      
 c
sin(c n)
and it’s time domain signal will be h[n] 
n
• Thus we have learned two method to find DTFT of a signal x[n]. If x[n] is absolutely summable
then X(e j)  X( z)|z e j . But if signal is power signal then we cannot apply eq.(2) to find

X(e j) and we have to apply some manipulations to find result

• We cannot find DTFT of signals which are neither energy and nor power signals.

8.5 Properties of DTFT

We assume that X(e j) is DTFT of x[n], that is
x[ n]  X( e j )

8.5.1 Linearity
If
x1( n)  X1( e j ) and x2 ( n)  X2 ( e j )
then
ax1( n)  bx2 ( n)  aX1( e j )  bX2 ( e j ) …(6)
Proof The Fourier transform of ax1(n) + bx2(n) is

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 377


 [ax1(n)+bx 2 (n)]   [ax1( n)  bx2 ( n)]e jn
n
 
 a  x 1( n )e  j n  b  x 2 ( n )e  j n

n    
n 
X 1 (e j  ) X 2 (e j  )

= aX1(e j) + bX2(e j)

8.5.2 Time shifting

If
x( n)  X( e j )
then
y(n)  x( n  n0 ) Y( e j )  X( e j )e jn0 …(7)
Proof The Fourier transform of x(n – n0) is

 [ x ( n  n0 )]   x( n  n0 )e jn
n
A change of variables is performed by letting m = (n – n0), which also yields n = (m + n0), m = – as
n = – and m =  as n = . Therefore,

 [ x ( n  n0 )]   x( m)e j( m n ) 0

n

 e jn0  x( m)e jm
m

 [ x ( n  n0 )]  X( e )e jn0
j

REMEMBER

• One consequence of this property is that, when a signal is shifted in time, the magnitude of its Fourier transform
remain unaltered. That is, if we express X(e j) in polar form as
j )
 [ x ( n )]  X( e j )  | X( e j )| e jX( e
then
j  )n )
j  jn0
 [ x ( n  n0 )]  X( e )e  | X( e j )| e j( X( e 0

Thus, the effect of a time shift on a signal is to introduce into its Fourier transform a phase shift, n0, which is a
linear function of .
• In time shifting the effect goes with some sign
If x[n]  X(e j)
then x[n – n0]  X( e j )e jn0

Example 8.9

Find DTFT of sequence

1 1 
x[n ]  sin c  ( n  2) 
4 4 

GATE MASTERS PUBLICATION

 378 Signals & Systems

Solution 8.9
It will be difficult to solve for X(e j) using eq.(2) because summation will be difficult

sin n
Since 4  rect    ,     
n  /2
 

sin ( n  2)
4     j 2
  rect  e ,     
( n  2) /2
1 1      j 2
 sin c  ( n  2)   rect  e ,     
4 4   /2

8.5.3 Frequency shifting

If
x( n)  X( e j )
then
y(n)  x( n)e j0 n  Y( e j )  X( e j( 0 ) ) …(8)
j0 n
Proof The Fourier transform of x( n)e is given by

 [ x ( n )e j 0 ]   [x( n)e j n ]e jn
0

n

  x( n)e j(  )n 0

n

j( 0 )
 [ x ( n )e j 0 ]  X( e )

Hence, a frequency shift corresponds to multiplication in time domain by a complex sinusoid whose
frequency is equal to the frequency shift.

Note: In case of frequency shifting the effect in seen with apposite sign, that is if
x[n]  X(e j)
then e+j0nx[n]  X(e j(– 0))

Example 8.10

If DTFT of x [ n ] is X ( e j ) then DTFT of (–1) n x [ n ] is __________.

Solution 8.10
Given that
x[n]  X(ej – )
then (–1)nx[n] = e jn x[n]  X(e j( – ))
Thus shifting X(e j) on  axis by  left side

8.5.4 Time Reversal

If x[n]  X(ej)
then
y(n) = x(–n)  Y(e j) = X(e–j) …(9)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 379

Proof The Fourier transform of y(n) is given by


Y( e j  )   y( n)e jn
n

  x(  n)e jn
n
Substituting m = –n into the above equation gives

Y( e j  )   x( m)e jm
n

  x( m)e j( )m
n

Y( e j  ) = X(e–j)
An interesting consequence of the time-reversal property is that if x(n) is even, then its Fourier transform
is also even, i.e.,
if x(–n) = x(n),
then X(e–j) = X(ej) …(10)
Similarly, if x(n) is odd, then so is its Fourier transform, i.e.,
if x(–n) = – x(n),
then X(e–j) = – X(ej) …(11)

Note: Thus even signal has even DTFT and odd signal has odd DTFT

8.5.5 Time Expansion

Suppose x[n] is a discrete time signal and X(e j) is DTFT of x[n], that is


X(e j)   x[n]e jn
n

if we generate a new signal y[n] = x[an] then DTFT of y[n] will be


Y( e j  )   x [an ]e  j n
n 

There will be no exact relation between Y(e j) and X(e j) because X(e j) has all values of x[n] and Y(e j)
has x[a], x[2a] … and not all value of x[n]. But if we expand signal x[n] that is
Let m be a positive integer, and define the signal

  n
x , if n is a multiple of m
x(m)(n)    m  …(12)
0, if n is not a multiple of m.

xm(n) can be obtained from x(n) by placing (m – 1) zeros between successive values of the original
signal. Intuitively, we can think of x(m)(n) as a slowed-down version of x(n).
Now, if
x(n)  X(ej)
then

GATE MASTERS PUBLICATION

 380 Signals & Systems

y(n) = x(m)(n)  Y(e j) = X(e jm) …(13)

Proof The Fourier transform of y(n) = x(m)(n) is given by

Y( e j)   y( n)e jn
n
 
 n   j n
  x( m)( n)e jn   x e
n n  m
n
A change of variables is performed by letting r  , which also yields r = – as n = – and r =  as
m
n = . Therefore,

Y( e j  )   x ( r )e  j mr
r 

Y(e j) = X(ejm)

Note that as the signal spread out and slowed down in time by taking m > 1, its Fourier transform is
compressed.

8.5.6 Differencing in Time Domain

If a signal y[n] = x[n] – x[n – 1]
and DTFT of x[n] is X(e j) then DTFT of y[n] will be
Y(e j) = X(e j) – X(ej)e–j
Y( e j) = (1 – e –j)X(e j) …(14)

8.5.7 Differentiation in Frequency Domain

If
x(n)  X(ej)
then
dX( e j )
–jnx(n)  …(15)
d
or
dX( e j )
nx(n)  j …(16)
d
Proof We have

X(e j )   x( n)e jn
n
By differentiating both sides w.r.t. , we obtain

dX( e j ) 
d
 [ jnx( n)]e jn
n

dX( e j )
=  [ jnx ( n )]
d
Therefore,
dX( e j )
–jnx(n) 
d

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 381

Multiplying both sides by j yields

dX( e j )
nx(n)  j
d
Example 8.11

Determine the Fourier transform of the sequence

y ( n ) = na n u ( n ), | a | < 1
Solution 8.11
Define a squence
x(n) = anu(n), |a| < 1
Its Fourier transform is give by
1
X(e j ) 
1  ae j
Now consider the given sequence
y(n) = nanu(n) = nx(n)
Now apply the differentiation in frequency domain property to get the Fourier transform of y(n):
dX( e j ) d  1 
Y( e j  )  j  j
d d  1  ae j 


ae j

(1  ae j )2

ae j
nanu(n)  , |a | 1
(1  ae j )2

Example 8.12

Determine the inverse DTFT of

1
X ( e j)  , |a| 1
(1  ae j  )2

Solution 8.12
We have

ae j
nanu(n)  , |a | 1
(1  ae j )2

Using the time-shifting property, we obtain

 ae j  j
(n + 1)an+1u(n + 1)    j 2 
e
 (1  ae ) 
a
(n + 1)an+1u(n + 1) 
(1  ae  j  )2

1
(n + 1)anu(n + 1) 
(1  ae  j  )2
Therefore,

GATE MASTERS PUBLICATION

 382 Signals & Systems

x(n) = (n + 1)anu(n + 1)
It is worth noting that although the RHS is multiplied by a step function that begins at n = –1, the sequence
(n + 1)anu(n + 1) is still zero prior to n = 0, since the factor n + 1 is zero at n = –1. Thus, we can
alternatively express x(n) as
x(n) = (n + 1)anu(n)
and hence,
1
(n + 1)anu(n) 
(1  ae  j  )2

8.5.8 Convolution Property

If
x1(n)  X1(ej) and x2(n)  X2(ej)
then
x 1( n )* x 2 ( n )  X1(e j)X2(ej) …(17)
The fourier transform maps the convolution of two signals into the product of their Fourier transforms.
Proof The Fourier transform of x1(n) * x2(n) is

  x 1( n )* x 2 ( n )   [x1( n)* x2 ( n)]e jn
n

   
   x1( m)x2 ( n  m)  e jn
n   m 
By interchangin the order of summation, we have

   
  x 1( n )* x 2 ( n )   x1( m)  x2 ( n  m)e jn 
m  n 
By applying the time-shifting property the bracketed term is X2(e j)e–jm. Substituting this into the
above equation yields

  x 1( n )* x 2 ( n )   x1( m)( X2 ( e j )e jm )
m


 X2 ( e j )  x1( m)e jm
m

  x 1( n )* x 2 ( n ) = X2(e j)X1(e j) = X1(e j)X2(e j)

Therefore,
x1(n) * x2(n)  X1(e j)X2(ej)
The convolution property states that convolution in the time domain corresponds to multiplication in
the frequency domain.

8.5.9 Accumulation Property

If
x[n]  X(ej)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 383

then

n
X (e j  ) j0

 x[k] 
1e  j
 X (e )  (  2m ) ...(18)
k  m 

This can be simply proved by convolution property

n  1  
Since  x[k]  x[n]* u[ n] X(e j )  1  e j  m

(   2m)

k 

X (e j  ) 
 x[n] * u[n] 
1 e j
   X (e j 0 )(  2m )
m 

8.5.10 Multiplication Property

If
x1(n)  X1(ej) and x2(n)  X2(ej)
then
1 1
x1(n)x2(n)  [X1(e j  )  X 1(e j  )]   X1(e j  )X 2(e j () )d  …(19)
2 2 2

where  denotes the periodic convolution. The Fourier transform maps the multiplication of two signals
into the convolution of their Fourier transforms.
Proof The Fourier transform of x1(n)x2(n) is given by

  x 1( n )x 2 ( n )   [x1( n)x2 ( n)]e jn
n
  1 
   2  X1( e j )e jn d  x2 ( n)e jn
n   2 

Interchanging the order of intergration and summation and noting that X1(ej) does not depend on n,
we obtain

1   
  x ( n  2 )   X1( e j )  [x2 ( n)e jn ]e jn  d
2 2   n 
By applying the frequency-shifting property the bracketed term is X2(e j( – )). Substituting this into the
above equation yields
1
  x 1( n )x 2 ( n )  X1( e j )X2 ( e j(   ) )d
2 2
1
 [X 1(e j  )  X 2 (e j  )]
2
Therefore,
1
x1(n)x2(n)  [ X 1(e j  )  X 2 (e j  )]
2
The convolution integral in Eq. (19) is known as the periodic convolution of X1(e j) and X2(ej) because
it is the convolution of two periodic functions having the same period. We note that the limits of integration

GATE MASTERS PUBLICATION

 384 Signals & Systems

extend over a single period.

8.5.11 Conjugation and conjugate symmetry

If
x(n)  X(ej)
then
x*(n)  X*(e–j) …(20)
Proof The Fourier transform of x*(n) is
 *
 j n   
  x *(n )   x *( n)e    x( n)e jn 
n  n 
*
  
   x( n)e j( )n 
 n 

= [X(e –j)]*
F[x*(n)] = X *(e–j)
1. When the signal is real
that is x[n] = x*[n]
If DTFT of x[n] is X(e j) then that of x*[n] is X *(e–j)
Thus for a real signal
X(e j ) = X *(e–j)
Thus DTFT will be conjugate symmetric

Note: Similarly if x[n] is imaginary then X(e j) that is DTFT will be conjugate anti symmetric

2. When the signal is real and even

Here x[n] = x*[n] and x[n] = x[–n]
Thus X(e j ) = X *(e–j) and X(ej) = X(e–j)
 X( e j ) = X *(ej) = X(e–j)
 X(e j) is real and even
Thus for real and even time domain signal DTFT is also real and even

3. When the signal is real and odd

That is x[n] = x*[n] and x[n] = –x[–n]
Thus X( e j ) = X *(e –j) and X(ej) = –X(e–j)
 X(e j ) = –X *(ej) = –X(e–j)
Thus X(e j) is imaginary and odd
for a real and odd signal DTFT is imaginary and odd

4. If signal x[n] is real then DTFT of even and odd part of x[n] is
Given that x[n]  X(ej)
and x[n] = x*[n]
 X( e j ) = X *(e–j)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 385

x[ n]  x[ n] X( e j )  X( e j )
Even part of x[n]  
2 2
j
X (e )  X *(e j  )

2
 xe[n]  Real {X(ej)}
Similarly
x[ n]  x[ n] X( e j )  X( e j )
odd part of x[n]  
2 2
X (e j  )  X *(e j  )

2
xo[n]  jImg{X(ej)}

Note: Thus for a real signal the real part of DTFT is due to even part of signal and imaginary part of DTFT is due
to odd part of signal.

8.5.12 Parseval’s Theorem

Let x(n) be an energy signal, and if
x(n)  X(ej)
then


1
Ex   | x( n)|2  2  | X( e j )|2 d ...(21)
n 2

Proof Consider the LHS of above equation, we have

 
Ex   | x( n)|2   x( n)x *( n)
n n
*
  1 
  x( n)  X( e j )e jn d 
 
n  2 2  
  1 
  x (n) 2  X *(e j  )e  j nd 
n   2 
  
1
  X *(e j  )  x (n )e  j n d 
2 2  n  
1
 X *(e j  )X (e j  )d 
2 2

1
 | x( n)|2  
2 2 
| X( e j )|2 d
n

Example 8.13

Find the energy of the sequence

 n
x ( n ) = sinc  c 
  

GATE MASTERS PUBLICATION

 386 Signals & Systems

assuming  c < .
Solution 8.13
We have

c  c n      2 m 

sinc 
  
  rect  
m  2c 

 c n       2m 
x(n) = sinc    X( e j )   rect  
   c m  2c 
Hence, from the Parseval’s relation we have

1
Ex   |x( n)|2  2  | X( e j )|2 d
n 2

 2
1 2   
Ex   2
rect   d
2  c  2c 

We know that

   1, | |  c
rect   
 2 c  0, c  |  |  

  
Substituting the value of rect   into Eq.(6.61) yields
 2 c 
c
1 2  
Ex 
2 2c  1d  22 2c  c
c c

8.5.13 Duality Theorem

In DTFT the time domain signal (x[n]) is discrete and nonperiodic but fourier domain signal (X(e j)) is
continuous and periodic with period 2.
DTFT
discrete non periodic signal   continuous periodic signal (period 2)

X(e j ) =  x[ n]e jn ...(22)
n

1
and x[n]  X( e j )e jn d ...(23)
2 2

If we think about apposite transform then continuous time fourier series (CTFS) convert continuous
periodic signal (x(t)) into discrete non periodic signal (Ck) and we if we take period of x(t) as 2 then
CTFS
continuous periodic signal with period (2)  discrete non periodic (Ck)
Since period is 2  0 = 1
 
 x(t )   Ck e j tk   Ck e jtk
0
…(24)
k  k 

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 387

1 1
Ck  x(t )e j0kt dt  x(t )e jkt
2 2 2 2
and …(25)

If we compare equation 22, 23, 24 and 25 we can see that variable  and t are continuous and k and n
are discrete
If we manipulate equation (22) and (23) by replacing n = –k and  = t we get

X(ejt)   x[k]e jtk
k 

1
and x[–k]  X( e jt )e jtk dt
2 2

 The time domain signal X(e jt) formed by replacing  by t in expression of X(ej) will be periodic with
period 2 and it’s continuous time fourier series (CTFS) coefficients will be x[–k]
Thus if
DTFT
x [n] X( ej )

CTFS
X(ejt) Ck = x (–k )

Thus we can say that duality exist between CTFS and DTFT and it can be used to find DTFT of power
signals for which we cannot apply equation (2) to find DTFT.

8.6 Insight in some equations

Let X(e j) is DTFT of x[n]

1. X(ejo)   x[n]
n

Since X(e j)   x[ n]e jn
n

 X(e jo)   x[n]
n

2. X ( e j )   (1)n x[n]
n

Since X( e j )   x[n]e jn
n

 X ( e j )   (1)n x[n]
n

1 j
3. x[0]   X( e )d
2 

1 j jn
Since x[n]   X( e )e d
2 

GATE MASTERS PUBLICATION

388 Signals & Systems


1 j
 x[0]   X( e )d
2 

The average value of X(e j) is x[0].


1  dX( e j ) 
4.   d  d = 0
2 
dX( e j )
Since nx[n]  j
d
n j
x[ n]  dX( e )
j d

n 1 dX( e j ) jn

j
x[ n]   d e d
2 

Putting n = 0 both side of the equation we get


1  dX (e j  )
 0    d  d 
2 
 2
dX( e j ) 
5.  d  2  | nx[ n]|2

d n

dX( e j )
Since nx[n]  j
d
Now applying parsevals theorem we get
 2
1 dX( e j ) 

 d
2 
d   | nx[n]|2
n

REMEMBER whenever we have to find angle of X(ej ) then the time domain signal x(n) which is already real
can be made even by shifting. Now the new signal(y(n)) obtained by shifting of x(n) is real and
even then it’s DTFT will also be real and even thus will have zero phase. Using the time shifting

property we can easily find the phase of X(ej ). The following example illustrate this procedure.

Example 8.14

L et X ( e j  ) denote the Fourier transform of the signal x ( n ) depicted in Fig. Perform the
following calculations without explicity evaluating X ( e j ):
( a ) Evaluate X ( e j 0 )
( b ) Find  X ( e j  )

j
( c ) Evaluate  X( e )d


( d ) Find X ( e j  )

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 389

( e ) Determine and sketch the signal whose Fourier transform is Real{ X ( e j  )}


j
( f ) Evaluate  | X( e )|2 d



dX( e j  )
( g ) Evaluate  d

d

Solution 8.14
From Fig. we can observe that
x(n)  {1 0 1 2 1 0 1 2 1 0 1}, 3  n  7


x (n)
2

–3 0 7
–2 –1 1 2 3 4 5 6 n

–1

Fig. : Signal x(n) for Example

(a) we have
 7
X( e j0 )   x( n)   x( n)  6
n n3
(b) We can observe that y(n) = x(n + 2) is real and even. Therefore (using the properties of the
DTFT) the DTFT Y(ej) =   y ( n ) is real and even. Hence, Y(ej) has zero phase. We also know
that
y(n) = x(n + 2)
  y ( n ) =   x ( n  2)

Y(e j) = X(ej)ej2

or, equivalently,
X(e j) = Y(e j)e –j
Since Y(ej) has zero phase, we have
X(e j) = –2
(c) we have

j
 X(e )d = 2x(0) = 4


(d) we have

X ( e j )   (1)n x( n)
n
7
  (1)n x( n)
n3

GATE MASTERS PUBLICATION

 390 Signals & Systems

= –x(–3) + x(–2) – x(–1) + x(0) – x(1) + x(2) – x(3) + x(4)

– x(5) + x(6) – x(7)
=1+0–1+2–1+0–1+2–1+0+1=2
(e) we have
xe(n)  Real{X(ej)}
Therefore, the desired signal is
x( n)  x(  n)
xe(n) 
2
The signal is as shown in Fig.
x e( n)
2

0 7
–4 –3 –2 –1 1 2 3 4 5 6
n

(f) We have
 
j
 | X(e )|2 d  2  | x( n)|2
 n

7
 2  | x( n)|2  28
n3

(g) we have
 2
dX( e j ) 

 d  2  | nx( n)|2

d n

 2  | n|2| x( n)|2
n
7
 2  | n|2| x( n)|2  316
n3

8.7 DTFT of Periodic Signals

If a signal x[n] is periodic then to find DTFT we cannot apply eq.(2) because x[n] is not absolutely
summable. To find DTFT of periodic signals we use DTFS. Any discrete time periodic signal can be written as

x[n]   Xk e j0 nk
k  N 

2
Here 0 
N
The DTFT of
1  2(), 0   < 2
Using frequency shifting property e j 0 n
 2( – 0), 0   < 2

e j0kn  2( – 0k), 0   < 2

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 391

Xk e j0 kn  2Xk  – 0k), 0   < 2

 x[n]   Xk e j0kn  2  Xk (  k0 ) ...(22)

k  N  k  N 

Since the DTFT is periodic with period 2, it follows that X(e j) consists a set of N impulses of strength
2Xk, k = 0, 1, 2, ..., N – 1, repeated at intervals of N0 = 2. Thus, X(e j) can be compactly written as

X(e j)  2  Xk (  k0 ) …(23)
k 
Thus, the Fourier transform of a periodic signal is simply an impulse train with impulses located at
 = k0, each of which has a strength 2Xk, and all impulses are separated from each other by 0.

Example 8.15

Find and sketch the Fourier transform of the discrete-time impulse train

x(n)   ( n  mN )
m
depicted in Fig.
x ( n)

–2N –N 0 N 2N n

Fig. : Discrete-time impulse train

Solution 8.15
2
The signal is periodic with period N and frequency 0  . The given signal for one period may be written
N
as
1, n0
x(n)  
0, 1  n  ( N  1)
= (n)
Therefore, the Fourier series coefficients are
N 1
1
Xk   x( n)e jk n 0
N n 0
N 1
1
  ( n)e jk(2  / N )n
N n 0

1 1
 e  jk (2  / N )n 
N n 0 N
The DTFT of a periodic signal is given by

X(e j)  2  Xk (  k0 )
k 
1
Substituting Xk  into the above equation gives
N

GATE MASTERS PUBLICATION

392 Signals & Systems


1
X(e j)  2  (  k0 )
k  N

2  2k 

N
     N 
k 

which is illustrated in Fig.

X(e j )

… …

0 
2
N

Fig. : DTFT of a discrete-time impulse train

Example 8.16

Find the Fourier transform of the periodic signal

x ( n ) = cos( 0 n )
2
with 0  .
5
Solution 8.16
Consider the given signal
1 j0 n 1  j0 n
x(n) = cos(0n)  e  e
2 2
We know that


e j0 n   2(  0  2m)

m
Therefore,
1 1
X(e j)  [e j 0n ]  [e  j 0n ]
2 2
1  1 
X( e j )   2 (   0  2 m)   2(  0  2m)
2 m 2 m

1   2  1   2 
X( e j )   2     2m    2     2m 
2 m  5  2 m  5 
The DTFT X(e j) is illustrated in fig.
X(e j )


… …
2
–2 –00 0 2 – 0 2+0 

Fig. : DTFT of x(n) = cos(0n)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 393

Example 8.17

Find the Fourier transform of the periodic signal

x ( n ) = sin( 0n )
2
with 0  .
5

Solution 8.17
Consider the given signal
1 j0 n 1  j0 n
x(n) = sin(0n)  e  e
2j 2j
We know that

e j0 n   2(  0  2m)
m
Therefore,
1 1
X( e j )   [e j 0 n ]   [e  j 0 n ]
2j 2j
 
1 1
X( e j )   2(  0  2m )   2(  0  2m )
2j m  2j m 

 
1  2  1  2 
X( e j )   2     2m    2     2m 
2j m   5  2j m   5 
The DTFT X(e j) us illustrated in Fig.
X(e j )


j
…
2 –0 –0 2
… –2 0  2+0 
0


–
j

Fig. : DTFT of x(n) = sin(0n)

8.8 Inverse Discrete Time Fourier Transform

Case 1:
If in a question an expression of X(e j) is given, and expression is rational function in ej with some
numerator and denominator for example

2  e j
X( e j ) 
e2 j  5e j  7
Then in such case replace e j by z and we will get a function X(z), Now simply apply inverse
z-transform with ROC of X(z) such that it contain unit circle as DTFT of signal exist it must be
absolutely summable.

GATE MASTERS PUBLICATION

394 Signals & Systems

Case 2:
In a question X(e j) is a simple function of , for example

2sin(  / 10)  j10 

X( e j )  e
10
or X(e j) = (u[ + 3] – u[ – 3])e j 

Then in such cases to find x[n] we will apply equation


1
x[n]   X( e j )e jn d
2 

j
Here X(e j) = X (e j  ) e j X (e )

Example 8.18

Find the DTFT time domain signal for these

1  e j 
(a) X ( e j  ) 
1  2 e j 
  
(b) X ( e j )  rect   ,     
 2 c 
(c) X ( e j  ) = 1
(d) X ( e j  ) = e – j 10
Solution 8.18

1  e j
(a) X (ej ) 
1  2e j

1  z 1 1 z 1
 X (z )  1
 1

1  2z 1  2z 1  2 z 1
The pole of X(z) is at –2, since ROC must contain unit circle  ROC is |z| < 2. Thus signal is left
sided signal
1
  –(–2)nu[–n – 1]
1  2 z1
 z 1
and  (–2)n–1 u[–n]
1  2 z 1
 x[n] = –(–2)nu[–n – 1] + (–2)n–1u[–n]
  
(b) X( e j )  rect   ,     
 2c 
   1, |  |  c
We know that rect   
 2 c  0, elsewhere

1 j  jn
 x[n]   X( e )e d
2 

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 395

1 c  jn
 e d
2 
c

1
 ( e jcn  e jcn )
2jn
sin( c n)
x[n] 
n
(c) X(e j ) = 1
We will not answer by applying eq.(1) but we know that
[n]  1
Thus x[n] = [n]
(d) X(e j ) = e –j10
Here also we will not get x[n] by applying eq.(1) but we know that
[n – 10]  e–j10
 x[n] = [n – 10]

8.9 Signal Transmission through LTI System

If x(n) and y(n) are the input and output of an LTI system with impulse response h(n), then
y(n) = x(n) * h(n)
Application of the time convolution property to the above equation yields [assuming that both x(n) and
h(n) are Fourier transformable]
 [ x (1  n )  x ( 1  n )] =  [ x ( n )]*  [h( n )]
Y( e j) = X(ej)H(ej) …(26)
or

Y( e j )
H(e j)  …(27)
X( e j )

x(n) h(n) y(n)

j
X (e j  ) H (e ) Y (e j  )  X (e j  )H (e j  )

Fig.8.1 : LTI system analysis using DTFT

The use of the convolution property for LTI systems is demonstrated in Fig.8.1 Equation (26) can be
expressed in the polar form as
j ) j )
| Y( e j )| e jY( e  | X (e j  )|e j X (e  |H (e j  )|e j H (e j  ) 
   
j ) j  )H( e j  )
| Y( e j )| e jY( e  | X( e j )|| H( e j )| e jX( e
Therefore,
|Y(e j)| = |X(e j)||H(e j)| …(28)
and
Y( e j) = X(e j) + H(e j) …(29)

GATE MASTERS PUBLICATION

 396 Signals & Systems

REMEMBER During transmission, the input signal magnitude spectrum | X ( e j )| is changed to
j j j
|X( e )||H( e )|. Similarly, the input signal phase spectrum  X ( e ) is changed to
X(e j) + H(e j). An input signal spectral component of frequency  is modified in amplitude
by a factor |H(e j)| and is shifted in phase by angle H(e j). Clearly |H(e j)| is magnitude
responce and H(ej) is the phase response of the system. H(ej) is called the frequency response
of the system.
When H(e j) exist then system is stable, since the system will be causal only when all poles give
right sided signal. Thus here system will be stable and causal and cannot be causal alone.

8.9.1 Response of LTI system to Complex Exponentials

We have seen in z-transform that

LTI System
e j o n e j o n H ( z ) z e j o
h[n] H(z)

Similarly if H(e j) is given in the question then output for e j0n will be H(ej0).ej0n.
If h[n] is real and it’s DTFT (H(e j0)) will be conjugate symmetric then

e j 0 n  e  j 0 n LTI System H( e j0 )e j0 n  H( e j0 )e j0 n

cos[0n] 
2 h( n )  H (e j  ) 2

Since h[n] is real thus H(e j) will be conjugate symmetric. Thus H(e j0) = H *(e –j0)
If H(e j0) = |H(ej0)| 
then H(e –j0) = |H(ej0)| –
Thus output will be

 e j0 n e j   e j0 n e j 
| H( e j0 )|   = cos( n + )|H(e j0)|
0
2

LTI System
cos[0n] H(ej0).cos[0n] = |H(ej0)|cos(0n + )
j
h( n )  H (e )

Similarly

LTI System
sin[0n] H(ej0).sin[0n] = |H(ej0)|sin(0n + )
j
h( n )  H (e )

REMEMBER In above analysis we have taken h[n] as real, generally h[n] is real and if not specified in
question then assume h[n] is real.

8.9.2 Solving Difference Equation of System

If in the question a difference equation of system in given then even if the terms are given in DTFT
format then change the entire equation into z-transform and then apply z-transform analysis to find result.

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 397

Example 8.19

Consider a discrete time LTI system with input x(n) and output y(n) defined by difference
equation
3 1
y(n) - y(n - 1) + y(n - 2) = 2x(n)
4 8
(a) Find the impulse response of the system if DTFT of transfer function exist.
n
1
(b) Find the output of the system for input x(n) =   u(n)
4

Solution 8.19
(a) We will use z-transform for analysis:
3 1
Y( z)  Y( z)z1  Y( z)z2 = 2X(z)
4 8

Y( z) 2
 H(z)  
X( z) 1  3 z1  1 z2
4 8
 16  16
 H(z)   2 

 8  6 z  z  (2  z )(4  z1 )
1 1

Now to find h[n] we will apply inverse z-transform such that ROC of H(z) has unit circle
1 1 1
Since poles of H(z) are at and , since ROC must contain unit circle so ROC is | z |  .
2 4 2
Thus both poles give right sided signal

8 8 4 2
 H(z)  1
 1
 
2z 4 z 1 1
1  z1 1  z1
2 4
n n
1 1
 h[n]  4   u[ n]  2   u[ n]
2
   4
n
1
(b) To find output for input x[n]    u[ n]
 4
Since Y(z) = X(z) H(z)
1  16 
  1 1 
 1 1   (2  z )(4  z ) 
1  4 z 
 
4 2 8
  
1 1  1 1 2 1 1
1 z 1 z
4 1  4 z  2
 
  1  n 1
n
1
n 
 y[n]  4   u[ n]  2( n  1)  u[ n]  8   u[ n]
  4   4 2 

GATE MASTERS PUBLICATION

398 Signals & Systems

Example 8.20

Determine the response of the system with impulse response

n
1
h ( n )    u( n )
2
when the input is the complex exponential sequence
x ( n ) = Ae j (/2)n , – < n < 
Solution 8.20
Given that
n
1
h(n)    u( n)
2
Taking its DTFT, we obtain

1
H ( e j ) 
1
1    e j
2

At   , the above equation yields
2
1
H(e j/2) 
1
1    e j  /2
2

1 2  j 26.6
  e
1 5
1 j  
2
The response of the system to the complex exponential is given by
y(n) = AH(ej0)ej0n = AH(e j/2)e j/2n
 2  j 26.6  j /2 n
 A e e
 5 
2
y(n)  A e j( n /2 26.6 )
5

Example 8.21

Determine the response of the system with impulse response

n
1
h ( n )    u( n )
2
to the input signal

 
x ( n )  5  5 sin  n   10 cos( n ),   n  
2 

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 399

Solution 8.21
Given that
n
1
h(n)    u( n)
2
Taking its DTFT, we obtain

1
H(e j  ) 
1
 
1    e j
2

The first term in the input signal x(n) is a constant signal corresponding to  = 0, thus

1
H( e j 0 )  2
1
1
2

The second term in x(n) has a frequency . At this frequency, the frequency response of the system is
2

1 1 2
H(e j/2)    e j 26.6
1 1 5
1  e j  /2 1 j
2 2
Finally, the third term in x(n) has a frequency  = . At this frequency, the frequency response of the system
is

1 1 1 2
H (e j  )    
1 1 3 3
1    e j 1   
2 2 2

Hence the response of the system to x(n) is

 
y(n)  5H( e j 0 )  H( e j  /2 )5sin  n   H( e j  )10cos( n)
2 
10   20
y(n)  10  sin  n  26.6   cos( n)
5 2  3

Example 8.22

If DTFT of a signal x [ n ] is X ( e j ) then find the phase of x [ n ] if X ( e j ) is

j
X(e )

A
… …

–2 0  2 
4

GATE MASTERS PUBLICATION

 400 Signals & Systems

Solution 8.22
The plot of X(e j) show that it is real and X(e j(+/4)) is real and even thus time domain signal corresponding
to it will be real and even. Thus X(e j(+/4)) is real and even
If x[n]  X(ej)

j n
 x[ n]e 4  X(e j(+/4))


j n 
Thus x[ n]e 4 will be real and even thus phase of x[n] will be n.
4

8.9.3 Distortionless System or Distortion less Transmission

Consider a discrete-time LTI system with impulse response h(n) and frequency response H(ej). Let a
signal x(n) with Fourier transform X(e j) be applied to the input of the system. Let a signal y(n) with fourier
transform Y(e j) denote the output of the system. In several applications, such as signal amplification or message
signal transmission over a communication channel, we require that the output waveform be a replica of the
input waveform, i.e., distortionless transmission. As in the continuous-time case, transmission is said to be
distortionless if the input and the output have identical waveshapes within a multiplicative constant and a
constant time delay. Thus, in distortionless transmission, the input x(n) and the output y(n) satisfy the condition
(Fig. 8.2)
Input signal LTI Output signal
x (n ) System y (n ) = Gx (n – n d)

Fig 8.2 : Distortionless transmission from system

y(n) = Gx(n – nd) …(30)
where the constant G accounts for a change in amplitude and the constant nd accounts for a delay in
transmission. Taking the Fourier transform of the above equation yields

Y(e j) = G X (e j  )e  j nd

Y( e j )
= H(e j) = G e  j nd
X( e j )
This is the frequency response required of a system for distortionless transmission. From this equation
it follows that
1. the magnitude response |H(e j)| must be a constant, i.e., we must have
|H(e j)| = G …(31)
for some constant G.
2. the phase response H(e j) must be a linear function of  with slope –nd and intercept zero, i.e.,
we must have
H(e j) = –nd …(32)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 401

|H(ej )| H(ej)

0  0 

Slope = nd

(a) (b)

Fig 8.3 : (a) Magnitude response and (b) phase response of a system for distortionless transmission

Note: For distortionless transmission we need magnitude response of LTI system to be constant and phase response
to be linearly related to 

8.9.4 Phase Delay and Group Delay

The two important parameters that characterize the form of the response y(n) of an LTI system excited
by an input signal x(n) composed of a weighted sum of sinusoidal signals are phase delay and group delay. These
two parameters are associated with the frequency response H(e j) of the system. For a linear phase system
(whose phase varies linearly with frequency), both the phase delay and the group delay are constant.

Phase Delay
The time delay experienced by a single-frequency signal (i.e. sinusoidal signal) when it passes through
a system is referred to as system phase delay. Suppose input to system is cos(0n + ) then output will be

LTI System
cos[0n + ] H(ej0) cos[0n + ]
h[n] H(ej )

 y[n] = |H(e j0)|cos[0n +  + H(ej0)]

  H( e j0 )  
 | H( e j0 )|cos  0  n    
  0  

Thus we can see that delay in time is

H (e j 0 )
p(0)   …(33)
0
Group Delay
When the input signal constains many sinusoidal components with different frequencies that are not
harmonically related, each component will go through different phase delays when processed by a frequency
selective LTI system, and the signal delay is determined using a different parameter called the group delay, as
defined below:

d H (e j  )
g()   …(34)
d 0

GATE MASTERS PUBLICATION

 402 Signals & Systems

The time delay g() is called the group delay or enveolope delay. Thus, the group delay at each frequency
equals the negative of the phase at that frequency. If g() is constant, all the components are delayed by the
same interval.

Example 8.23

Consider a discrete-time LTI system with frequency response H ( e j ) and real impulse response
h ( n ). The group-delay function for such a system is defined as

dH( e j  )
g(  )  
d
where  H ( e j ) has no discontinuities. Suppose that, for this system,

| H ( e j /2 ) | = 2,  H ( e j0 ) = 0,  g    2
2
Determine the output of the system for each of the following inputs:
   7 
( a ) cos  n  (b) sin  n 
2   2 4

Solution 8.23

   
(a) The input signal x(n)  cos   has a frequency 0  2 . Hence, the response of the system to x(n) is
 2n 

y(n)  | H( e j0 )|cos(0 ( n   g (0 )))

    
 | H (e j  /2 )|cos   n   g    
2  2 

 
y(n)  2cos  ( n  2) 
2 
 
y(n)  2cos  n   
2 
(b) Consider the given input signal
 7 
x(n)  sin  n 
 2 4

 
x(n)   sin  n  
2 4 

The input signal x(n) has a frequency 0  . Hence, the response of the system to x(n) is
2
y(n) = |H(ej0)|sin(0(n – g(0)) + )

 7      
 | H( e j  /2 )|sin   n   g     
 2   2  4 
 7 
y(n)  2sin  ( n  2)  
 2 4

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 403

 7 
y(n)  2sin  n  7  
 2 4

 7 
 2sin  n 
 2 4

 7 3 
y(n)  2sin  n 
 2 4 

Example 8.24

Consider a discrete-time low-pass filter whose impulse response h ( n ) is known to be real

and whose frequency response magnitude in the region –     is given as

 
1, | | 
| H ( e j) |  4
0, otherwise

Determine the real-valued impuse response h(n) for this filter when the corresponding
group-delay function is specified as
5 5
( a ) g () = 5, (b) g ()  , (c) g ()  
2 2

Solution 8.24
Let H0(e j) = |H(e j)|. Then the inverse DTFT of H0(e j) is given by

1
h0(n)   H0 ( e j )e jn d
2 

 /4  /4
1 j n 1 e j n
 e d 
2 /4 2 jn  /4

1 e j  /4n  e  j  /4n
 
n 2j

sin(  / 4n)
h0(n) 
n

d H (e j  )
If g()     (where  is a constant), then H(e j) =– + , where  is a constant. Given
d
that h(n) is real, therefore H(e j) is an odd function, and hence for H(e j) is an odd function, and hence
for H(e j) to be an odd function, the constant  = 0. Therefore,
j )
H(e j)  | H( e j )| e jH( e  H0 ( e j )e j
Taking the inverse DTFT, we obtain
h(n) = h0(n – )

GATE MASTERS PUBLICATION

 404 Signals & Systems

 
sin  ( n   )
 4 
h(n) 
( n   )

(a) Given that g() = 5 = . Substituting  = 5 in the expression of h(n) yields

 
sin  ( n  5)
 4 
h(n) 
( n  5)
5 5
(b) Given that g()   . Substituting   in the expression of h(n) yields
2 2
  5 
sin   n   
4
  2 
h(n) 
 5
n 
 2
5 5
(c) given that g()    . Substituting    in the expression of h(n) yields
2 2
  5 
sin   n   
h(n)  4  2 
 5 
 n 
 2

8.10 Discrete Domain filters

We have seen that discrete frequency axis is periodic with period 2, thus in discrete frequency we can
say that
0  DC
  + (highest frequency)
–  – (highest negative frequency)
Thus if we have any filter whose impulse response is h[n] and DTFT of h[n] is H(e j) then to find type of
filter we should follow these steps
Step 1 : Find the magnitude of H(e j)|=0 i.e. |H(e j0)|
let A = |H(e j0)|
Step 2 : Find the magnitude of H(e j)|= and –
Let B = |H(e j)| or |H(e –j)|
generally h[n] is real so |H(e j)| and |H(e –j)| will be same
Step 3 :
Case-1: If A > B and B is close to zero then we can say that filter passes low frequency component
of input and attenuates the high frequency component thus it is a low pass filter
Case 2. If B > A and A is close to zero then the system will attenuate low frequency component in
input and will pass high frequency component in input. Thus it act as a high pass filter.
Case-3: If A and B both are small and close to zero then system will alternate both low and high
frequency component in input but surely |H(ej)| will be greater than 1 at some frequency thus
system act as a band pass filter.

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 405

Case-4: If A and B both are greater than 0 then system may be band stop filter that is the filter will
attenuate some frequency of input or an all pass filter which pass all frequency of input without
alternation.
Following are the characteristics of these filters:
1. Low-pass filters are those characterized by a passband that extends from  = 0 to  = c, where c
is called the cutoff frequency of the filter [Fig. 8.4(a)]. A low-pass filter attenuates high-frequency
components of the input and passes the low-frequency components.

|H(e j )|

– –c 0 c  

(a)
Fig 8.4 : (a) Low-pass filter
2. High-pass filters are characterized by a stopband that extends from  = 0 to  = c and a passband
that extends from  = c to  [Fig. 8.4(b)]. A high-pass filter attenuates low frequencies and passes
the high frequencies.
|H(e j )|

– – c 0 c  

(b)
Fig 8.4 : (b) High-pass filter
3. Bandpass filters are characterized by a passband that extends from  = 1 to  = 2 and all other
frequencies are stopped [Fig. 8.4(c)]. A bandpass filter passes signals within a certian frequency
band and attenuates signals outside that band.
|H(e j )|

– –2 –0 –1 0 –2 –0 –1  

(c)
Fig 8.4 : (c) Bandpass filter
4. Bandstop filters stop frequencies extending from  = 1 to  = 2 and pass all other frequencies
[Fig.8.4 (d)]. A bandstop filter attenuates signals within a certain frequency band and passes signals
outside that band.

GATE MASTERS PUBLICATION

 406 Signals & Systems

|H(e j)|

– –2 –0 –1 0 –2 – 0 –1  

(d)

Fig 8.4 : (d) Bandstop filter

5. An all-pass filter is characterized by a magnitude response that is constant for all frequencies
[Fig.8.4(e)]. The characteristics of an all-pass filter are completely determined by its phase-shift
characteristics. An all-pass filter has nonconstant group delay and different frequencies in the
input are delayed by different amounts.

|H(e j )|

– 0  

(e)

Fig 8.4 : (e) All-pass filter

Note: If in a question H(z) is given and we have find the type of filter then simply replace z by e j and follow the
procedure

Properties of Ideal all pass filter :

General transfer function is
N
 bi z  i
i 0
H(z)  N
c i z i
i 0
For all pass filter we need magnitude response to be constant for all , here order of numerator will be
same as denominator. Thus for all pass filter (and generally c0 = 1).
N
 cN* i z  i
i 0
H(z)  N
 c i z 1
i 0

That is for first order N = 1

 c *  c * z 1  c *  z 1
H(z)   1 0 1   1 1
 c 0  c1z  1  c1z
1
The pole is at (–c1) and zero at  .
c1*

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 407

1
Tor all pass filter if pole is at pi then zero will be at zi  .
pi*

8.11 Compiling Properties of DTFT

Table 8.1 Properties of the Discrete Time Fourier Transform
Property Signal Fourier transform
x[ n] X( e j ) periodic with

y[ n] Y( e j )  period 2
Linearity ax[n] + by[n] aX(e j) + bY(e j)
Time Shifting x[n – n0] X (e j  )e  j n0

Frequency shifting x ( n )e j 0 n X (e j  0  )

Conjugation x*[ n ] X *(e –j)
Time Reversal x [– n ] X( e – j  )
  n
x , if n  multiple of k
Time Expansion x(k) [n]    k  X(e jk)
0, if n  multiple of k

Convolution x[n] * y[n] X(e j)Y(e j)
1
X( e j  )Y( e j(    ) )d 
2 2
Multiplication x[ n ] y [ n ]

Differencing in Time x[n] – x[n – 1] (1 – e –j)X(e j)

n
1
Accumulation  x{k} X( e j )
k   1  e j

X( e j 0 )  (   2k )
k 
dX( e j )
Differentiation in Frequency nx [n ] j
d

X (e j  )  X *(e  j  )

 e{X (e j  )}  e{X (e  j  )}
 j j
Conjugate Symmetry for Real x[n] real  m{X (e )}   m{X (e )}
Signals  j j
| X (e )|  | X (e )|
X (e j  )  X (e  j  )


Symmetry for Real, Even x[n] real an even X(e j) real and even
Signals
Symmetry for Real, odd x[n] real an odd X(e j) purely imaginary and
Signals odd
Even-odd Decomposition of xe[n] = v{x[n]} [x[n] real] e{X(e j)}
Real Signals xe[n] = Od{x[n]} [x[n] real] jm{X(e j)}
Parseval’s Relation for Aperiodic Signals

1
 | x[ n]|2  2  | X( e j )|2 d
n  2

GATE MASTERS PUBLICATION

 408 Signals & Systems

If x[n] is discrete time signal and it’s DTFT is X(e j) then for
The domain signal DT FT
real  Conugate symmetric
imaginary  Conjugate anti symmetric
odd  odd
even  even
real and even  real and even
real and odd  imaginary and odd
imaginary and even  imaginary and even
imaginary and odd  real and odd

REMEMBER In above expression we have double sided arrow that is if time domain signal is real then DTFT
is conjugate symmetric and if time domain signal is conjugate symmetric then DTFT will be real

Table 8.2 Sumary of Fourier Series and Transform Expressions

Continuous time Continuous time

Time domain Frequency domain Time domain Frequency domain

1 1
x(t )   ak e jk t
0
ak 
T0 T
x(t )e jk0t x[n]   ak e jk(2 / N )n ak   x[ n]e jk(2 /N)n
Fourier k 
0
k ( N ) N k ( N )
Series
continuous time discrete frequency discrete time duality discrete frequency
periodic in time aperiodic in frequency du periodic in time periodic in frequency
al
ity

  
1 1
j t  j t x[ n]  X( e j )e jnd X( e j )   x[n]e jn
2 2
x(t )   X( j)e d X( j)   x(t)e dt
Fourier 2   n

Transform

continuous time duality continuous frequency discrete time continuous frequency

aperiodic in time aperiodic in frequency aperiodic in time periodic in frequency

8.12 Energy Spectral Density

Parseval’s theorem relates the total signal energy in a signal x(n) to its Fourier transform through

  1/2
1 j 2 j 2
Ex   | x( n)|2   | X( e )| d   | X( e )| df
2 
n 1/2

Parseval’s theorem states that the total energy Ex may be determined either by computing the energy
| X( e j )|2
per unit time (|x(n)|2) and summing over all time or by intergrating the energy per unit frequency
2
over a full 2 interval of distinct discrete-time frequencies. For this reason |X(e j)|2 represents energy per unit
bandwidth and is often referred to as the energy spectral density or energy density spectrum of the signal x(n)
and is denoted by x(e j). Hence,
x (e j) = |X(e j)|2 …(35)
For discrete-time signals, the unit of ESD is simply the square of the signal unit (whatever that may be)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 409

8.12.1 Relationship between input and outpud ESDs of an LTI System

Consider an LTI system with frequency response H(e j), input x(n), and output y(n). If x(n) and y(n)
are energy signals, then their ESDs are x(e j) = |X(e j)|2 and y(e j) = |Y(e j)|2, respectively. Since, we
know that
Y(e j) = H(e j)X(e j)
it follows that
|Y(e j)|2 = |H(e j)X(e j)|2
|Y(e j)|2 = |H(e j)|2|X(e j)|2
y( e j ) = |H(e j)|2x(e j) …(36)

8.12.2 Relation of ESD to Autocorrelation

The autocorrelation function Rxx(m) of a real energy signal is defined as

Rxx(m)   x( n)x( n  m)
n

Rxx(m) = x(m) * x(–m)

taking the Fourier transform of the above equation, we have
 [R xx ( m )] = X(e j)X(e –j)
= X ( e j ) X * ( e j )
 [R xx ( m )] = |X(e j)|2

 [R xx ( m )] = x(e j)
Rxx(m)  x(e j) …(37)
Thus, the autocorrelation function Rxx(m) and ESD makes a Fourier transform pair.

8.13 Power Spectral Density (PSD)

PSD has the same relation to power signals as ESD has to energy signals. The expression of PSD may be
obtained by assuming the power signal as a limiting case of an energy signal.
Consider a power signal of infinite duration. Since the total energy of a power signal cannot be found, let
us first find the ESD of a truncated version xN(n) of the infinite duration signal x(n). The signal xN(n) may be
expressed as

 x( n), N  n  N
xN(n)   …(38)
0, otherwise

The truncated signal xN(n) is of finite duration, therefore, it is an energy signal. Now, if xN(n)  XN(e j),
then using the Parseval’s theorem, we have

 
1 j 2
ExN   | xN ( n)|2   | XN ( e )| d
2 
n

Substituting the value of xN(n) from Eq. (38) into the above equation gives

GATE MASTERS PUBLICATION

 410 Signals & Systems

N 
1
 | x( n)|2  j 2
 | XN ( e )| d
2 
n N

N 
1
lim  | x( n)|2  lim 1 1
| XN ( e j )|2 d
N 2 N  1
n N N 2N  1 2 


N
| XN ( e j )|2

1 1
lim  | x( n)|2   N 2N  1 d
lim
N 2N  1 2 
n N

The LHS of the above equation represents the average power Px of the signal x(n). Therefore,
N
1
Px  lim  | x( n)|2
N 2N  1
n N

| XN ( e j )|2

1
  N 2N  1 d
2 
lim


1 j
Px   Gx ( e )d
2 
…(39)

where
| XN ( e j )|2
Gx(e j)  lim …(40)
N  2N  1
is the PSD. The units of PSD depend on the units of the underlying signal x(n).

8.13.1 Relationship Between input and output PSDs of an LTI System

Consider an LTI system with frequency response H(e j), input xN(n), and output yN(n). If x(n) and y(n)
are power signals, then their PSDs are
| XN ( e j )|2
Gx(e j)  lim
N  2N  1

| YN ( e j )|2
and Gy(e j)  lim
N  2N  1
respectively. Since, we know that
YN(e j) = H(e j)XN(e j)
it follows that
|YN(e j)|2 = |H(e j)XN(e j)|2
|YN(e j)|2 = |H(e j)|2|XN(e j)|2
| YN ( e j )|2 | X ( e j )|2
lim  | H( e j )|2 lim N
N  2N  1 N  2N  1
Gy(e j) = |H(e j)|2Gx(e j) …(41)

8.13.2 Relation of PSD to Autocorrelation

The autocorrelation function Rxx(m) of a power signal is defined as
N
1
Rxx(m)  lim  x( n)x( n  m)
N 2N  1
n N

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 411


1
 lim
N 2 N  1
 x N ( n)xN ( n  m)
n

1
Rxx(m)  lim [ x N ( m)* x N (  m)]
N  2 N  1

taking the Fourier transform of the above equation, we obtain

1
 [R xx ( m )]  lim XN ( e j )XN ( e j )
N  2 N  1

1
 lim XN ( e j )XN* ( e j )
N  2 N  1

1
 [R xx ( m )]  lim | XN ( e j )|2
N  2 N  1

 [R xx ( m )] = Gx(e j)
Rxx(m)  Gx(e j) …(42)
Thus, the autocorrelation function Rxx(m) and PSD makes a Fourier transform pair.

8.14 Relationship between DTFT and DTFS

We know that DTFS give discrete periodic Xk and DTFT give continuous periodic X(e j). DTFS is
2
calculated for periodic signals with period N and 0  , and
N
x[n]   Xk e j0kn
k  N 

for DTFT
1
x[n]  X( e j )e jn d
2 2

from here we can see that Xk i.e. DTFS coefficients of periodic signal are samples of X(e j) taken at
integral multiple of 0.

1
 Xk  X( e j )| k0
2
In period of 0 to 2 we will get N samples only and Xk will also become periodic.

Example 8.25

Determine the Fourier transform for –   <  in the case of each of the following periodic
signals:
 
( a ) x 1 ( n )  sin  n  
 3 4
 
( b ) x 2 ( n )  2  cos  n  
6 8

Solution 8.25

GATE MASTERS PUBLICATION

412 Signals & Systems

 
(a) x1[n]  sin  n  
3 4
   
j n j j n j
e 3 e 4 e 3 e 4

2j

Since 1  2() – <  < 


j n  
 e 3  2      – <  < 
 3

j n  
and e 3  2      – <  < 
 3
 
j   j 
e 4 2     e 4 2   
  3  3 
 sin  n      
3 4 2j 2j

 3
j  j 4  
e 4    
 X1 (e j)  e      – <  < 
 3  3

(b) Here
 
x[n]  2  cos  n  
6 8  
   
j n j j j
e 6 e 8 e 6e 8
 2 
2 2
We know that
Since 2  4(), – <  < 

j n  
 e 6  2      , – <  < 
 6

j n  
and e 6  2      – <  < 
 6
 
 j  j
 x[n]  4()       e 8       e 8 – <  < 
 6  6

Example 8.26

Determine the inverse Fourier transform of


      
( a ) X1( e j  )    2 (   2 m)      2  2 m       2  2 m  
m     
2 j, 0
( b ) X2 ( e j  )  
 2 j ,     0

Solution 8.26

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 413

(a) The inverse DTFT x1(n) of X1(e j) is given by


1
x1(n )   X1( e j )e jn d
2 

1        jn
  2()      2       2  e d
2 

1 1 j n
 e j n  e j n  e
0 2  2 

 2
2
1 j  /2 n 1  j  /2 n
 e j0  e  e
2 2
 
x1(n)  1  cos  n 
2  
(b) The inverse DTFT x2(n) of X2(e j) is given by

1 j jn
x2(n )   X2 ( e )e d
2 
 
1 1
  2 je jn d   2 je jn d
2  2 0
0 
j e j n j e j n
 
 jn 
 jn 0

1 1
 [1  e jn ]  [e jn  1]
n n
1
 [e jn  1  1  e jn ]
n
1 jn
 [e  e jn  2]
n
1
 [2 cos( n)  2]
n
2
 [cos( n)  1]
n
4  n 
x2 ( n )  sin2  
n  2 

Example 8.27

Given that x ( n ) has Fourier transform X ( e j ), express the Fourier transform of the following
signals in terms of X ( e j ):
( a ) x1( n ) = x (1 – n ) + x (–1 – n )
x * ( n )  x( n )
( b ) x2 ( n) 
2
( c ) x 3 ( n ) = ( n – 1) 2 x ( n )

GATE MASTERS PUBLICATION

414 Signals & Systems

Solution 8.27
Given that
x(n)  X(e j)
(a) Using the time-shifting property, we have
x(n + 1)  X(e j)e j
and x(n – 1)  X(e j)e –j
Using the time-reversal property on this, we have
x(–n + 1) = x(1 – n)  X(e –j)e –j
and x(–n – 1) = x(–1 – n)  X(e –j)e j
Therefore,
 [ x 1( n )] =  [ x (1  n )  x ( 1  n )]
X1(e j) = X(e –j)e –j + X(e –j)e j
X1(e j) = 2X(e –j)cos()
(b) Using the conjugation property, we have
x*(n)  X *(e –j)
Using the time reversal-property on this, we have
x*(–n)  X *(e j)
Therefore,
 x *( n )  x ( n ) 
 [x2(n)]    
 2 
X *( e j )  X( e j )
X 2( e j  ) 
2
= {X(e j)} = X(e j)
X2 (e j)
= X(e j)
(c) Consider the given signal x3(n) = (n – 1)2 x(n) = n2x(n) – 2nx(n) + x(n).
Using the differentiation in frequency-domain property, we have

dX( e j )
nx(n)  j
d
Using the differentiation in frequency-domain property again, we have
d  dX( e j ) 
n[nx(n)]  j j 
d  d 
d2 X( e j )
n2x(n)  
d
Therefore,
 [x3(n)] =  [(n – 1)2x(n)]
=  [n2x(n) – 2nx(n) + x(n)]
d2 X( e j ) dX( e j )
X3( e j )   2j  X( e j )
d d
d2 X( e j ) dX( e j )
X 3( e j )   2j  X( e j )
d d

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 415

Example 8.28

Find the discrete-time Fourier transform of

( a ) x 1 ( n ) = cos( 0 n ) u ( n )
( b ) x 2 ( n ) = sin( 0 n ) u ( n )
Solution 8.28
(a) Consider the given signal
x1(n) = cos(0n)u(n)
Using the multiplication property, we have
X1(e j) =  [cos(0n)]*  [u(n)], – < 

 1 
= [( – 0) + ( + 0)] *   j
 ( )
 1  e 

 1 1  2
   j
    [(   0 )  (   0 )]
1  e 1  e j0 
X1(e j) =  + 2[( – 0) + ( + 0)]
(b) Consider the given signal
x2(n) = sin(0n)u(n)
Using the multiplication property, we have
X2(e j) =  [sin(0n)]*  [u(n)], – < 

  1 
 [(   0 )  (   0 )]*   j
 ( )
j 1  e 
2
 1 1  
   j0
   j [(   0 )  (  0 )]
j 1  e 1  e j0 
 sin(0 )
  2 [(  0 )  (  0 )]
1  cos( 0 )

   
2sin  0  cos  0 
 2   2   2 [(    )  (    )]
  0 0
 
2sin2  0 
 2 

 0  2
X2(e j)   cot     [(   0 )  (   0 )]
 2 

Example 8.29

A linear time-invariant system is described by the following difference equation:

y ( n ) = ay (n – 1) + bx( n), 0 < a < 1
( a ) Determine the magnitude response | H ( e j )| and phase response  H ( e j ) of the system
( b ) Choose the parameter b such that the maximum value of | H ( e j )| is unity.
Solution 8.29
(a) Consider the given difference equation

GATE MASTERS PUBLICATION

416 Signals & Systems

y(n) = ay(n – 1) + bx(n), 0 < a < 1

Taking the Fourier transform of the above equation we obtain
Y(e j) = aY(e j)e –j + bX(e j)
Y( e j ) b
 H( e j ) 
j
X( e ) 1  ae j
b
H( e j  ) 
1  a cos( )  ja sin( )
Therefore,
| b|
|H(e j)| 
(1  a cos( ))2  ( a sin( ))2
| b|
|H(e j)|  2
1  a  2a cos()

1 a sin()
and H(e j) = angle of b  tan
1  a cos( )
(b) Since the parameter a is positive, the denominator of |H(e j)| attains a minimum at  = 0. Therefore,
|H(e j)| attains its maximum value at  = 0. At this frequency, we have
| b|
|H(e j0)|  1
1 a
|b| = 1 – a
b = (1 – a)

Example 8.30

Consider a discrete-time LTI system with frequency response

   
 1  e j 2   4e j 4    j   4 
H ( e j)   e     
 1  j 2 
1 e
 2 
Determine the Fourier transform of the output if the input is
 n 
x ( n )  cos  
 2 
Solution 8.30
Consider the given input
 n 
x(n)  cos 
 2 
Taking the Fourier transform of the above equation, we obtain
     
X( e j)                   
  2  2 
Fourier transform of the output is given by
Y(e j) = H(e j)X(e j)

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 417

     
 H( e j )            
  2  2 
   
Y(e j)  H( e j  /2 )      H( e j  /2 )    
 2   2 
Given that

   
1  e j 2   4e j 4   j   4 
H( e j )   e
 1  1 e j 2  
 2 
Therefore,

 2 4 
1  e j 2  4e j 2   j     
H(e j/2)    e 2 4
2
 1  j 
 1 e 2 
 2 
 
1  e j   4e j 2    j 4
 e
 1  1 e j  
 2 
 
1  1  4   j 4
H(e j/2)   e
 1 1 
 2 
j /2
H(e ) = 8e –j /4

Similarly, we can obtain

H(e –j/2) = 8e j3/4
Substituting the values of H(e j/2) and H(e –j/2) in he expression of Y(e j), we obtain

  j  3
 j   
Y(e j)  8  e 4       e 4      
  2  2  

GATE MASTERS PUBLICATION

 418 Signals & Systems

where * denotes convolution and |c| . Determine

a stricter constraint on c which ensures that
2
  

1 Subjective Practice Problems  sin 4 n 

y[n]  
 n 


1 . Use the Fourier transform synthesis equation to 6 . Suppose we are given the following facts about an
determine the inverse Fourier transform of LTI system S with impulse response h [ n ] and
j frequency response H(e j):
X(e j)  | X( e j )| e jX( e )
, where
n
1
  1.  4  u[ n]  g[ n], where g[n] = 0 for n  2
1, 0  | | 
4 3
 
|X(e j)|   and X( e j )   . and n < 0
0,  2 2. H(e j/2) = 1.
 | |  
 4 3. H(e j) = H(e j(– )).
Use your answer to determine the values of n for Determin h[n].
which x[n] = 0.
7 . Let the inverse Fourier transform of Y(e j) be
2 . For each of the following Fourier transforms, use 2
Fourier transform properties to determine whether  sin c n 
y[n]    .
the corresponding time-domain signal is (i) real,  n 
imaginary, or neither and (ii) even, odd, or neither. where 0 < c < . Determine the value of c which
10
ensures that
 j
(a) X1(e j)  e  (sin k ) 1
k 1 Y(e j)  .
2
(b) X2(e j) = jsin() cos(5)
(c) X3(e j) = A() + e jB() where 8 . Given the fact that

   1  a2
a|n|  , | a |  1,
1, 0  | | 
8 3 1  2a cos   a 2
A()   and B()    .
0,  2 use duality to determine the Fourier series coefficients
 | |  
 8 of the following continuous time signal with period
3 . Determine x[n] when its Fourier transform is T = 1.
 3  1
1  sin 2   x(t) 
5  4cos(2t )
X(e j)     5(),     
1  e j  sin  
  9 . The following are the Fourier transforms of discrete-
 2 
time signals. Determine the signal corresponding to
4 . The following four facts are given about a real signal
each transform.
x[n] with Fourier transform X(e j):
1. x[n] = 0 for n > 0.   3
2. x[0] > 0 1, 4
 | | 
4
(a) X(e j)  
3. m{X(e j)} = sin – sin2. 0, 3  
 |  |  , 0  |  | 
  4 4
1 j 2
4.  | X( e )| d  3.
2 
(b) X(e j) = 1 + 3e–j + 2e–j2 – 4e–j3 + e–j10
(c) X(e j) = e–j/2 for –
Determine x[n] (d) X(e j) = cos2 + sin23

5 . Let 10. Determine which, if any, of the following signals have

2 Fourier transforms that satisfy each of the following
  
 sin 4 n   sin c n  conditions:
y[n]    * .
 n   n  1. e{X(e j)} = 0.

GATE MASTERS PUBLICATION

 Discrete Time Fourier Transform 419

2. m{X(e j)} = 0 11. Consider the signal

3. There exists an integer a such that e jX(e j) is  n   n 
real. x[n]  sin    2cos  
 8   4 

j
Suppose that this signal is the input to LTI systems
4.  X(e )d  0 with the following impulse responses. Determine the

output in each case.
5. X(e j) periodic.
sin( n / 6)
6. X(e j0) = 0. (i) h[n] 
n
(a) x[n] as in Figure (a) sin( n / 6) sin( n / 2)
(b) x[n] as in Figure (b) (ii) h[n]  
n n
n
1 sin( n / 6)sin( n / 3)
(c) x[n]    u[ n] (iii) h[n] 
2 n
|n|
1 sin( n / 6)sin( n / 3)
(d) x[n]    (iv) h[n] 
2 2 n2
(e) x[n] = [n – 1] + [n + 2]
12. Consider an LTI system with unit sample response
(f) x[n] = [n – 1] + [n + 3]
sin( n / 3)
(g) x[n] as in Figure (c) h[n] 
n
(h) x[n] as in Figure (d)
Determine the output for each of the following inputs:
(i) x[n] = [n – 1] – [n + 1]
(i) x[n] = the square wave(amplitude =1) depicted
x [n]
in figure
2
3 
2
1 1
(ii) x[n]   [ n  8k ]
k 
2
(iii) x[n] = (–1)n times the square wave depicted in
–1 0 1 2 3 4 5 6 n
Figure
(a) (iv) x[n] = [n + 1] + [n – 1]
x [ n] x [ n]
1
1
–8 0 8 16 n
–2 –1 0 2 n
–1
Fig.:
(b)
x [n] 13. Consider a system consisting of the cascade of two
LTI systems with frequency responses
2
2  e j
1 H1( e j ) 
1  1 / 2e j
0 n and
–1
1
(c) H2 ( e j ) 
1  j 1  j 2
1 e  e
x [ n] 2 4
(a) Find the difference equation describing the
2 overall system.
1 (b) Determine the impulse response of the overall
system
0 n
–1
14. A causal LTI system is described by the difference
(d) equation

GATE MASTERS PUBLICATION

 420 Signals & Systems

y[n] – ay[n – 1] = bx[n] + x[n – 1],  n 

where a is real and less than 1 in magnitude. (ii) p[ n]  cos  
 2 
Find a value of b such that the frequency response  n 
of the system satisfies (iii) p[ n]  sin  
 2 
|H(e j)| = 1, for all . 

This kind of system is called an all-pass system, as it (iv) p[ n]   [ n  2k ]

k 
does not attenuate the input e jn for any value of . 
(v) p[ n]   [ n  4k ]
15. Let k 

 n   n  X(e j )
x1[n]  cos    sin  
 3   2 
be a signal, and let X 1( e j) denote the Fourier 1
transform of x1[n]. Sketch x1[n], together with the
signals with the following Fourier transforms;
–2 – –  0   2 4 
(i) X2(e j) = X1(e j)e j, || <  2 2
(ii) X3(e j) = X1(e j)e –j3/2, || < 
19. X(e j) denotes the DTFT of a length-9 sequence x(n)
16. Determine whether each of the following statements given by
is true or false. Justify your answers. In each x( n)  {2, 3,  1, 0, 4, 3, 1, 2, 4},
statement, the Fourier transform of x[n] is denoted 

by X(e j). –2  n  6
Evaluate the following functions of X(ej) without
(a) If X(e j) = X(e j(– 1)), then x[n] = 0 for |n| > 0 computing the transform itself:
(b) If X(e j) = X(e j(– )), then x[n] = 0 for |n| > 0 (a) X(e j0)
(b) X(e j)
(c) If X(e j) = X(e j/2)), then x[n] = 0 for |n| > 0 
j
(d) If X(e j) = X(e j2), then x[n] = 0 for |n| > 0
(c)  X( e )d


17. A particular discrete-time system has input x[n] and j
(d)  | X( e )|2 d
output y[n] and output y[n]. The Fourier transforms 
of these signals are related by the equation  2
dX( e j )
(e)  d
j j  j j dX( e j ) 
d
Y( e )  2 X( e )  e X( e ) 
d 20. Consider the filter
(i) Is the system linear? Clearly justify your answer. y(n) = 0.9y(n – 1) + bx(n)
(a) Determine b so that |H(e j0)| = 1.
(ii) Is the system time invariant? Clearly justify your
(b) Determine the frequency at which |X(e j)|
answer. 1
 .
(iii) What is y[n] if x[n] = [n]? 2
(c) Is this filter low-pass, high-pass, or bandpass?
18. Let x [n] be a discrete-time signal with Fourier
transform X(e j), which is illustrated in Figure.

Sketch the Fourier transform of
w[n] = x[n]p[n]
for each of the following signals p[n]:
(i) p[n] = cos n

GATE MASTERS PUBLICATION