5

Discrete-time Fourier Transform

5.1 INTRODUCTION
A continuous-time signal can be represented in the frequency domain using Laplace
transform or continuous-time Fourier transform (CTFT). Similarly, a discrete-time signal can
be represented in the frequency domain using Z-transform or discrete-time Fourier transform.
The Fourier transform of a discrete-time signal is called discrete-time Fourier transform
(DTFT). DTFT is very popular for digital signal processing because of the fact that using
this the complicated operation of convolution of two sequences in the time domain can be
converted into a much simpler multiplicative operation in the frequency domain. In this
chapter, we discuss about DTFT, its properties and its use in the analysis of signals.

5.2 DISCRETE-TIME FOURIER TRANSFORM (DTFT)

The Fourier transform of discrete-time signals is called the discrete-time Fourier transform
(DTFT).
If x(n) is the given discrete-time sequence, then X(w) or X(e jw ) is the discrete-time
Fourier transform of x(n).
The DTFT of x(n) is defined as:
‡
F[x (n)] X (X ) Ç x (n) e  jX n
n ‡

The inverse DTFT of X(w) is defined as:

Q
1
F 1[X (X )] x ( n) Ô X (X ) e
jX n
dX
2Q Q
358
 Discrete-time Fourier Transform | 359

We also refer to x(n) and X(w) as a Fourier transform pair and this relation is expressed
as:
x (n) Ž  X (X )
FT

5.3 EXISTENCE OF DTFT

The Fourier transform exists for a discrete-time sequence x(n) if and only if the sequence is
absolutely summable, i.e. the sequence has to satisfy the condition:
‡
Ç x(n)  ‡
n ‡

The DTFT does not exist for the sequences that are growing exponentially (ex. an u(n),
a > 1) since they are not absolutely summable. Therefore, DTFT method of analyzing a
system can be applied for a limited class of signals. Moreover this method can be applied
only to asymptotically stable systems and it cannot be applied for unstable systems. That is,
DTFT can be used only for the systems whose system function H(z) has poles inside the unit
circle.
The Fourier transform X(w) of a signal x(n) represents the frequency content of x(n).
We can say that, by taking Fourier transform, the signal x(n) is decomposed into its
frequency components. Hence X(w) is called signal spectrum.
The difference between the Fourier transforms of a discrete-time signal and analog
signal are as follows:
1. The Fourier transform of analog signals consists of a spectrum with a frequency
range – ¥ to ¥. But the Fourier transform of discrete-time signals is unique in the
frequency range –p to p (or equivalently 0 to 2p ). Also Fourier transforms of
discrete-time signals are periodic with period 2p. Hence the frequency range for
any discrete-time signal is limited to –p to p (or 0 to 2p ) and any frequency
outside this interval has an equivalent frequency within this interval.
2. Since the analog signals are continuous, the Fourier transform of analog signals
involves integration, but the Fourier transform of discrete-time signals involves
summation because the signals are discrete.

5.4 RELATION BETWEEN Z-TRANSFORM AND FOURIER TRANSFORM

The Z-transform of a discrete sequence x(n) is defined as:
‡
X ( z ) Ç x ( n) z  n
n ‡
where z is a complex variable.
The Fourier transform of a discrete-time sequence x(n) is defined as:
‡
X (X ) Ç x (n) e  jX n
n ‡
360 | Digital Signal Processing

The X(z) can be viewed as a unique representation of the sequence x(n) in the complex
z-plane.
Let z = re jw
‡
\ X ( z ) Ç x (n) (re jX )  n
n ‡
‡
Ç [x (n) r  n ] e  jX n
n ‡

The RHS is the Fourier transform of x(n) r–n, i.e. the Z-transform of x(n) is the Fourier
transform of x(n) r–n.
When r = 1,
‡
X ( z ) Ç x (n) e  jX n X (X )
n ‡

The RHS is the Fourier transform of x(n). So we can conclude that the Fourier transform of
x(n) is same as the Z-transform of x(n) evaluated along the unit circle centred at the origin of
the z-plane.
‡ ‡
n
\ X (X ) = X ( z ) z e jX Ç x ( n) z z e jX Ç x(n) e jX n
n ‡ n ‡

For X(w ) to exist, the ROC must include the unit circle. Since ROC cannot contain any poles
of X(z) all the poles must lie inside the unit circle. Therefore, we can conclude that Fourier
transform can be obtained for any sequence x(n), from its Z-transform X(z) if the poles of
X(z) are inside the unit circle.

EXAMPLE 5.1 Find the DTFT of the following sequences:

(a) d (n) (b) u(n)
(c) d (n – m) (d) u(n – m)
(e) anu(n) (f) –anu(–n – 1)
(g) d (n + 3) – d (n – 3) (h) u(n + 3) – u(n – 3)
Solution:
(a) Given x(n) = d (n)
ÎÑ1 for n 0
E (n) Ï
ÑÐ0 for n › 0
‡
X (X ) F{E ( n)} Ç E (n) e  jX n n 0 1
n ‡

\ F{E ( n)} 1

FT
E (n ) Ž  1
 Discrete-time Fourier Transform | 361

(b) Given x(n) = u(n)

ÎÑ1 for n • 0
u(n) Ï
for n  0
ÑÐ0
‡
X (X ) F{u(n)} Ç u(n) e  jX n
n ‡
‡
Ç (1) e  jX n 1  e  jX
1
n 0
1
\ F{u( n)}
1  e  jX
FT 1
u(n) Ž 
1  e  jX
(c) Given x(n) = d (n – m)
ÎÑ1 for n m
E ( n  m) Ï
ÑÐ0 for n › m
‡
X (X ) F{E (n  m)} Ç E (n  m) e  jX n e  jX n n m e  jX m
n ‡
\ F{E ( n  m )} e  jX m

E (n  m) Ž  e  jX m
FT

(d) Given x(n) = u(n – m)

ÎÑ1 for n • m
u(n  m) Ï
ÑÐ0 for n  m
‡ ‡
X (X ) F{u(n  m)} Ç u(n  m) e  jX n Ç (1) e jX n
n ‡ n m

e  jX m
e  jX ( m  1)
 e  jX ( m  2)  "
e  jX m (1  e  jX  e  j 2X  ")
e  jX m
1  e  jX
e  jX m
\ F{u(n  m)}
1  e  jX

FT e  jX m
u( n  m ) Ž 
1  e  jX
362 | Digital Signal Processing

(e) Given x(n) = anu(n)

‡
X (X ) F{a n u(n)} Ç a n u( n) e  jX n
n ‡
‡
Ç (ae jX )n
1
n 0 1  ae  jX
1
\ F{a n u(n)}
1  ae  jX

FT 1
a n u(n) Ž 
1  ae  jX
(f) Given x(n) = –anu(–n – 1)
X (X ) F{ a n u(  n  1)}
‡
Ç  anu( n  1) e jX n
n ‡
1 ‡ ‡
Ç  an e jX n  Ç a n e jX n  Ç (a1e jX )n
n ‡ n 1 n 1

 ËÍ a 1e jX  ( a 1e jX )2  (a 1e jX )3  "ÛÝ

 a 1e jX ËÍ1  ( a 1e jX )1  (a 1e jX )2  "ÛÝ

 a 1e jX
1  a 1e jX
1
1  ae  jX
1
\ F{ a n u(  n  1)}
1  ae  jX
FT 1
 a n u(  n  1) Ž 
1  ae  jX

(g) Given x(n) = d (n + 3) – d (n – 3)

X (X ) F{E ( n  3)  E ( n  3)}
‡
Ç {E (n  3)  E (n  3)} e jX n
n ‡

e jX n  jX n
n 3  e n 3

e j 3X  e  j 3X 2 j sin 3X
 Discrete-time Fourier Transform | 363

(h) Given x(n) = u(n + 3) – u(n – 3)

X (X ) F{u(n  3)  u(n  3)}
‡ ‡ ‡
 jX n =  jX n 
Ç {u( n  3)  u ( n  3)} e Ç (1) e Ç (1) e jX n
n ‡ n 3 n 3

e j 3X
e j 2X jX
 e 1 e  jX
e  j 2X
"  e  j 3X
 e  j 4X  "
e j 3X  e j 2X  e jX  1  e  jX  e  j 2X

EXAMPLE 5.2 Find the DTFT of:

(a) x(n) = {1, –2, 2, 3} (b) x(n) = 3nu(n)
n
È 1Ø
(c) x(n) = (0.5)n u(n) + 2n u(–n – 1) (d) x (n) = É Ù u(n  1)
Ê 4Ú
ÎÑ n, 4 n 4 ÑÎ1, 0 n 3
(e) x ( n) Ï (f) x (n) Ï
ÑÐ0, otherwise ÑÐ0, otherwise
(g) x(n) = a |n|
Solution:
(a) Given x(n) = {1, –2, 2, 3}
‡
X (X ) F{x (n)} Ç x ( n) e  jX n
n ‡

x (0)  x (1) e  jX  x (2) e  j 2X  x (3) e  j 3X

1  2e  jX  2e  j 2X  3e  j 3X
(b) Given x(n) = 3nu(n). The given sequence is not absolutely summable. Therefore, its
DTFT does not exist.
(c) Given x(n) = (0.5)nu(n) + 2n u(–n – 1)
‡
X (X ) F{x (n)} Ç {(0.5) n u(n)  2 n u(  n  1)} e  jX n
n ‡
‡ ‡
 jX n 
Ç {(0.5) n
u ( n )} e Ç {2n u( n  1)} e  jX n
n ‡ n ‡
‡ 1
Ç (0.5)n e jX n  Ç 2n e jX n
n 0 n ‡
‡ ‡
Ç (0.5e  jX )n  Ç (21 e jX )n
n 0 n 1

1 2 1 e jX

1  0.5e  jX 1  2 1 e jX
1 1
 
1  0.5e jX
1  2e  jX
364 | Digital Signal Processing
n
È 1Ø
(d) Given x (n) = É Ù u( n  1)
Ê 4Ú

ÎÑÈ 1 Ø n ÞÑ ‡ È 1Øn
X (X ) F ÏÉ Ù u(n  1) ß Ç É Ù u( n  1) e  jX n
ÑÐÊ 4 Ú Ñà n ‡ Ê 4 Ú
‡ È 1Øn ‡ È1 n
 jX n - jX Ø
Ç ÉÊ ÙÚ e Ç ÉÊ e ÙÚ
n 1 4 n 1 4
1
È 1  jX Ø Ë ‡ 1  jX Û jX Ë 1 Û
ÉÊ e ÙÚ Ì Ç e Ü 4e Ì  jX Ü
4 ÍÌ n 0 4 ÝÜ Í 1  (1/4) e Ý
jX
4e
1  (1/4) e  jX
ÎÑ n, 4 n 4
(e) Given x ( n) Ï
ÑÐ0, otherwise
4
X (X ) F{x (n)} Ç ne  jX n
n 4
j 4X j 3X
 4e  3e  2e j 2X  e jX  e  jX  2e  j 2X  3e  j 3X  4e  4 jX
 2 j {4 sin 4X  3 sin 3X  2 sin 2X  sin X}
ÎÑ1, 0 n 3
(f) Given x (n) Ï
ÑÐ0, otherwise
By definition of Fourier transform,
‡ 3
X (X ) Ç x ( n) e  jX n Ç (1) e  jX n
n ‡ n 0

1e  j 4X
1  e  j 2X e  j 2X
1  e  jX 1  e(  jX /2) e(  jX /2)
e  j 2X {e j 2X  e  j 2X } Ë 2 j sin 2X Û  j 2X  jX /2
= (  jX /2) ( jX /2) (  jX /2) Ì Üe
e {e e } Í 2 j sin(X /2) Ý
sin 2X - j (3/2) X
= e
sin(X /2)
(g) Given x(n) = a |n|
‡
X (X ) F{a } Ç a e  jX n
n n

n ‡
1 ‡ ‡ ‡
Ç a  n e jX n  Ç an e jX n Ç (ae jX )n  Ç (ae jX )n
n ‡ n 0 n 1 n 0
jX
ae 1 1 a 2

jX
  jX
1  ae 1  ae 1  2 a cos X  a2
 Discrete-time Fourier Transform | 365

EXAMPLE 5.3 Find the DTFT of the following sequences:

È nQ Ø È nQ Ø
(a) sin É Ù u(n) (b) cos É Ù u(n)
Ê 2 Ú Ê 3 Ú
n n2
È 1Ø È nQ Ø È 1Ø
(c) ÉÊ 2 ÙÚ sin ÉÊ 4 ÙÚ u( n) (d) É Ù u(n  2)
Ê 2Ú
(e) cos (w0n) u(n) (f) sin(w0 n) u(n)
Solution:
È nQ Ø
(a) Given x(n) = sin É Ù u(n)
Ê 2 Ú

Î È nQ Ø Þ ‡ Î È nQ Ø Þ  jX n ‡ È nQ Ø
X (X ) F Ïsin É Ù u(n) ß
Ê Ú Ç Ï ÉÊ 2 ÙÚ ß
sin u ( n ) e = Ç sin ÉÊ 2 ÙÚ e jX n
Ð 2 à n ‡ Ð à n 0

‡ e j ( nQ /2)  e  j ( nQ /2)  jX n 1 Ë ‡ j[(Q /2)  X ]n ‡ Û

Ç e ÌÇ e  Ç e  j[(Q /2)  X ]n Ü
n 0 2j 2 j ÍÌ n 0 n 0 ÝÜ
1 Ë 1 1 Û
Ì Q  X   Q  X Ü
2 j Í1  e j [( /2) ]
1e j [( /2) ]
Ý
1 Ë 1  e  j (Q /2) e  jX  1  e j (Q /2) e  jX Û
Ì Ü
2 j ÌÍ 1  e  j 2X  e  jX [ e j (Q /2)  e  j (Q /2) ] ÜÝ

e  jX sin (Q /2) e  jX
1  e  j 2X 1  e  j 2X

È nQ Ø
(b) Given x(n) = cos É Ù u(n)
Ê 3 Ú

Î È nQ Ø Þ ‡ Î È nQ Ø Þ
X (X ) F Ïcos É Ù u(n) ß
Ê 3 Ú Ç Ïcos ÉÊ 3 ÙÚ u(n)ß e jX n
Ð à n ‡ Ð à
‡ Ë e j ( nQ /3)  e  j ( nQ /3) Û Ë ‡ j[(Q /3)  X ]n ‡ Û
u(n) Ü e  jX n  Ç e  j[(Q /3)  X ]n Ü
1
Ç Ì ÌÇ e
n ‡ Ì
Í 2 ÜÝ 2 ÌÍ n 0 n 0 ÜÝ
1Ë 1 1 Û
Ì j[(Q /3)  X ]
 – j[(Q /3)  X ] Ü
2 Í1  e 1e Ý
1 Ë1 e  j ( Q /3)
e  jX
1 e j ( Q /3) 
e jX Û
Ì Ü
2 ÍÌ 1  e  j 2X  e  jX [ e j (Q /3)  e  j (Q /3) ] ÝÜ

1Ë 2  e  jX Û
Ì Ü
2 ÍÌ 1  e  jX  e  j 2X ÝÜ
366 | Digital Signal Processing
n
È 1Ø È nQ Ø
(c) Given x(n) = É Ù sin É Ù u( n)
Ê 2Ú Ê 4 Ú

ÑÎÈ 1 Ø ÑÞ
n
È nQ Ø
X (X ) F ÏÉ Ù sin É Ù u(n) ß
Ê Ú Ê 4 Ú
ÐÑ 2 àÑ
‡ ÎÑÈ 1 Ø n È nQ Ø ÞÑ  jX n ‡ È 1Øn È nQ Ø
Ç ÏÉ Ù sin É Ù u(n) ß e Ç É Ù sin É Ù e  jX n
Ê Ú Ê 4 Ú Ê Ú Ê 4 Ú
Ð 2
n ‡ Ñ àÑ n 0 2

‡ È 1 Ø n ÎÑ e j ( nQ / 4)  e  j ( nQ /4) ÞÑ
 jX n
Ç ÉÊ 2 ÙÚ Ï ße
n 0 ÐÑ 2j àÑ
1 Ë ‡ ‡ È 1 Ø  j[(Q /4)  X ]n Û
n n
È 1 Ø j[(Q /4)  X ]n
ÌÇ ÉÊ 2 ÙÚ e  Ç ÉÊ 2 ÙÚ e Ü
2 j ÍÌ n 0 n 0 ÝÜ
1 Ë ‡ ‡ È 1  j[(Q /4)  X ] Ø Û
n n
È 1 j[(Q /4)  X ] Ø
ÌÇ ÉÊ e ÙÚ  Ç ÉÊ e ÙÚ Ü
2 j ÍÌ n 0 2 n 0 2 ÝÜ
1 Ë 1 1 Û
Ì  Ü
2 j Í 1  (1/2) e j[(Q / 4)  X ] 1  (1/2) e  j[(Q /4)  X ] Ý
(1/2) e  jX sin(Q /4)
1  (1/4) e  j 2X  e  jX cos(Q /4)
(1/2 2 ) e  jX
1  (1/ 2) e  jX  (1/4)e  j 2X
n –2
È 1Ø
(d) Given x(n) = É Ù u( n  2)
Ê 2Ú
‡ ËÈ 1 Ø n  2 Û
X (X ) F{x (n)} Ç ÌÉ Ù u(n  2) Ü e  jX n
n ‡ ÌÍ Ê 2 Ú ÜÝ
‡ È 1Øn2
Ç ÉÊ 2 ÙÚ e jX n
n 2
2
1  j 3X È 1 Ø  j 4X
e  j 2X  e É Ù e "
2 Ê 2Ú
Ë È 1Ø
2 Û
e  j 2X Ì1  e  jX  É Ù e  j 2X  "Ü
1
ÌÍ 2 Ê 2Ú ÜÝ
e  j 2X

1  (1/2) e  jX
 Discrete-time Fourier Transform | 367

(e) Given x(n) = cos(w0 n) u(n)

‡
X (X ) F{x (n)} Ç {cos(X 0 n) u( n)} e  jX n
n ‡

‡ Ë e jX 0 n  e  jX0 n Û
 jX n
ÇÌ Üe
n 0 Í Ì 2 Ü
Ý
1 ÎÑ ‡ j (X0 –X ) n ‡ ÞÑ
Ï Ç [e ]  Ç [e  j (X 0 X ) ]n ß
2 ÑÐ n 0 n 0 Ñà
1 Ë 1 1 Û
Ì j (X 0 X )
  j (X 0 X ) Ü
2 ÌÍ 1  e 1e ÜÝ
1 Ë 1  e  j (X 0 X )  1  e j (X 0 X ) Û
Ì Ü
2 ÍÌ 1  e  j 2X  e  jX (e jX 0  e  jX0 ) ÝÜ

1  e  jX cos X 0
1  2e  jX cos X 0  e  j 2X
(f) Given x(n) = sin(w0n) u(n)
‡
X (X ) F{x (n)} Ç {sin (X 0 n) u(n)} e  jX n
n ‡

‡ ÎÑ e jX 0 n  e  jX0 n ÞÑ ‡ ÎÑ e j (X 0 X ) n  e  j (X 0 X ) n ÞÑ
 jX n
ÇÏ ß e ÇÏ ß
n 0 ÐÑ 2j àÑ Ñ
n 0 Ð 2j àÑ
1 Ë 1 1 Û
Ì  Ü
2 j ÍÌ 1  e j (X 0 X ) 1  e  j (X 0 X ) ÝÜ

1 Ë 1  e  j (X 0 X )  1  e j (X 0 X ) Û
Ì Ü
2 j ÍÌ 1  e  j 2X  e  jX (e jX 0  e  jX 0 ) ÜÝ

e  jX sin X 0
1  2e  jX cos X 0  e  j 2X

EXAMPLE 5.4 Find the DTFT of the rectangular pulse sequence:

ÎÑ A, n N
x (n) Ï
ÑÐ0, n !N

ÎÑ A, n N
Solution: Given x (n) Ï
ÑÐ0, n !N
368 | Digital Signal Processing

N 1 N
X (X ) Ç Ae  jX n Ç Ae  jX n  Ç Ae jX n
n N n N n 0
N N N 1 N
Ç Ae jX n  Ç Ae  jX n Ae jX Ç e jX n  A Ç e  jX n
n 1 n 0 n 0 n 0

Ë1  ejX N Û Ë1  e  jX ( N  1) Û
Ae jX Ì jX Ü
 AÌ  jX Ü
ÍÌ 1  e ÝÜ ÍÌ 1  e ÝÜ
Ëe  ejX jX ( N 1) Û Ë1  e  jX ( N  1) Û
AÌ jX Ü  AÌ  jX Ü
ÌÍ 1  e ÜÝ ÌÍ 1  e ÜÝ
Ë e jX  1  e jX ( N 1)  e jX N  1  e jX  e  jX ( N 1)  e  jX N Û
AÌ Ü
ÍÌ 1  1  e jX  e  jX ÝÜ
Ë (e jX N
e  jX N
)  (e j X ( N  1)
e  jX ( N 1)
)Û
AÌ jX – jX Ü
ÍÌ 2  (e  e ) ÝÜ
Ë 2 cos X N  2 cos X ( N  1) Û
= AÌ Ü
Í 2  2 cos X Ý
Ë 2 sin X [ N  (1/2)] sin(X /2) Û A sin X [ N  (1/2)]
AÌ Ü
Í 2 sin 2 (X /2) Ý sin(X /2)

5.5 INVERSE DISCRETE-TIME FOURIER TRANSFORM

The process of finding the discrete-time sequence x(n) from its frequency response X(w ) is
called the inverse discrete-time Fourier transform.
Q
1
Ô X (X ) e dX
jX n
x (n)
2Q Q
The integral solution of the above equation for x(n), i.e, for the inverse Fourier transform is
useful for analytic purpose, but it is usually very difficult to evaluate for typical functional
forms of X(w). An alternate and more useful method of determining the values of x(n)
follows directly from the definition of the Fourier transform.
‡
X (X ) Ç x (n) e  jX n "  x ( 2) e j 2X  x ( 1) e jX  x (0)  x (1) e  jX  x (2) e  j 2X  "
n ‡

From the defining equation of X(w ) we can say that, if X(w ) can be expressed as a series of
complex exponentials as shown in the above equation for X(w ), then x(n) is simply the
coefficient of e–jw n. Inverse Fourier transform can be obtained by using the partial fraction
method or by using the convolution theorem.

EXAMPLE 5.5 Determine the signal x(n) for the given Fourier transforms:
(a) X(w) = e–jw for –p £ w £ p (b) X(w) = e–jw (1 + cos w)
 Discrete-time Fourier Transform | 369

Solution:
(a) Given X(w) = e–jw
Q
1
x ( n) F 1{X (X )} Ôe
 jX
e jX n dX
2Q Q

Q Q
1 1 Ë e jX ( n 1) Û
Ôe
jX ( n 1)
dX Ì j ( n 1) Ü
2Q Q
2Q ÌÍ ÜÝ Q

1 Ë e jQ ( n 1)  e  jQ ( n 1) Û 1 Ë e jQ ( n 1)  e  jQ ( n 1) Û

Ì Ü Ì Ü
2Q ÌÍ j(n  1) ÜÝ Q (n  1) ÌÍ 2j ÜÝ
sin Q (n  1)
Q (n  1)
(b) Given X (X ) e  jX (1  cos X )
È e jX  e  jX Ø
= e  jX É1  Ù
Ê 2 Ú

e  jX  0.5  0.5e  j 2X

x (0)  x (1) e  jX  x (2)e  j 2X

Therefore, x(0) = 0.5, x(1) = 1, x(2) = 0.5
x(n) = 0, otherwise
i.e. x(n) = {0.5, 1, 0.5}

EXAMPLE 5.6 Obtain the impulse response of the system described by

ÎÑ1, for X X0
H (X ) Ï
ÑÐ0, for X 0 X Q

ÎÑ1, for X X0
Solution: Given H (X ) Ï
ÑÐ0, for X 0 X Q
The impulse response h(n) is given by
Q
1
Ô H (X ) e dX
jX n
h( n)
2Q Q
X0 X
1 È e jX n Ø
0
1 jX n
2Q Ô (1) e dX
2Q ÉÊ jn ÙÚ
X 0 X 0

1 È e jX 0 n  e  jX 0 n Ø sin(X 0 n)
2Q ÉÊ jn Ù
Ú Qn
370 | Digital Signal Processing

EXAMPLE 5.7 Find the inverse Fourier transform of the following:

Î Q 2Q
ÑÑ1, X
X (X ) 3 3
Ï
Ñ0, 2Q
X Q
ÑÐ 3
Solution: The Fourier transform X(w ) is given. Then, the inverse Fourier transform of X(w )
is given by
Q
1
Ô X (X ) e dX
jX n
x ( n)
2Q Q

1 Ë Q /3 2Q /3 Û 1 Ë È e jX n Ø Q /3 È e jX n Ø 2Q /3 Û
Ì Ô 1 e jX n dX  Ô 1 e jX n dX Ü Ì  É jn Ù Ü
2Q ÌÍ 2Q /3 Q /3 ÜÝ 2Q Ì ÉÊ jn ÙÚ Ê Ú Ü
Í 2Q /3 Q /3 Ý

1 Ë e  jn(Q /3)  e  jn(2Q /3)  e jn(2Q /3)  e jn(Q /3) Û

Ì Ü
nQ ÍÌ 2j ÝÜ
1 Ë e jn(2Q /3)  e  jn(2Q /3) e jn(Q /3)  e  jn(Q /3) Û 1 Ë 2Q QÛ
Ì  Ü Ìsin n 3  sin n 3 Ü
nQ ÍÌ 2j 2j ÝÜ nQ Í Ý

EXAMPLE 5.8 Find the inverse Fourier transform of

X (X ) 2  e  jX  3e  j 3X  4e  j 4X

Solution: Given X (X ) 2  e  jX  3e  j 3X  4e  j 4X
‡
We know that X (X ) Ç x (n) e  jX n
n ‡
= " + x ( 2) e j 2X + x ( 1) e jX + x (0)  x (1) e  jX  x (2) e  j 2X
 x (3) e  j 3X  x (4) e  j 4X  "
Comparing the above two values of X(w), we get
i.e. x (0) 2, x (1) 1, x (2) 0, x (3) 3, x (4) 4
x (n) {2, 1, 0, 3, 4}

5.6 PROPERTIES OF DISCRETE-TIME FOURIER TRANSFORM

5.6.1 Linearity Property

The linearity property of DTFT states that
If F[x1(n)] = X1(w) and F[x2(n)] = X2(w)
Then F[ax1(n) + bx2(n)] = aX1(w) + bX2(w)
 Discrete-time Fourier Transform | 371
‡
Proof: F{ax1 (n)  bx2 (n)} Ç [ ax1 ( n)  bx2 ( n)] e  jX n
n ‡
‡ ‡
Ç ax1 (n) e  jX n  Ç bx2 (n) e  jX n aX1 (X )  bX2 (X )
n ‡ n ‡

5.6.2 Periodicity Property

The periodicity property of DTFT states that the DTFT X(w) is periodic in w with period 2p,
i.e.
X (X  2 nQ ) X (X )
Implication: We need only one period of X(w) {i.e., X ° (0, 2Q ) or ( Q , Q )} for analysis
and not the whole range –¥ < w < ¥.

5.6.3 Time Shifting Property

The time shifting property of DTFT states that
If F [x(n)] = X(w)
–jw m
Then F [x(n – m)] = e X(w) where m is an integer.
‡
Proof: F{x (n  m)} Ç x (n  m) e jX n
n ‡
Let n – m = p
\ n=p+m
‡
\ F{x (n  m)} Ç x ( p) e  jX ( p  m)
p ‡
‡
e  jX m x ( p) e  jX p
Ç e  jX m X (X )
p ‡

This result shows that the time shifting of a signal by m units does not change its amplitude
spectrum but the phase spectrum is changed by –w m.

5.6.4 Frequency Shifting Property

The frequency shifting property of DTFT states that
If F{x ( n)} X (X )
Then F{x (n) e jX0 n } = X (X  X 0 )
‡
Proof: F{x ( n) e jX 0 n } = Ç {x (n) e jX 0 n } e  jX n
n ‡
‡
Ç x (n) e j (X X0 ) n X (X  X 0 )
n ‡

This property is the dual of the time shifting property.

372 | Digital Signal Processing

5.6.5 Time Reversal Property

The time reversal property of DTFT states that
If F{x(n)} = X(w)
Then F{x(–n)} = X(ÿ-w)
‡
Proof: F{x (  n)} Ç x (  n) e  jX n
n ‡
‡
Ç x (n) e jX n
n –‡
‡
Ç x (n) e  j ( X ) n
n –‡
X ( X )

That is, folding in the time domain corresponds to the folding in the frequency domain.

5.6.6 Differentiation in the Frequency Domain Property

The differentiation in the frequency domain property of DTFT states that
If F{x(n)} = X(w)
d
Then F{n x (n)} j [ X (X )]
dX
‡
Proof: F{x (n)} X (X ) Ç x (n) e  jX n
n ‡

Differentiating both sides w.r.t. w, we get

d d ÑÎ ‡  jX n ÑÞ
{X (X )} Ï Ç x ( n) e ß
dX dX ÐÑ n – ‡ àÑ
‡
Ç x(n) dX e jX n
d
n ‡
‡
Ç x(n) ( jn) e jXn
n ‡
ÎÑ ‡ ÞÑ
 j Ï Ç n x (n) e jX n ß
ÐÑ n ‡ àÑ
‡
nx (n) e  jX n
d
\ Ç F{n x (n)} j [ X (X )]
n ‡ dX
 Discrete-time Fourier Transform | 373

5.6.7 Time Convolution Property

The time convolution property of DTFT states that
If F{x1(n)} = X1(w) and F{x2(n)} = X2(w)
Then F{x1(n) * x2(n)} = X1(w) X2(w)
‡
Proof: x1 (n) x2 (n) Ç x1 (k ) x2 (n  k )
k ‡
‡
\ F{x1 (n) x2 (n)} Ç [ x1 (n) x2 (n)] e  jX n
n ‡
‡ ÎÑ ‡ ÞÑ
Ç Ï Ç [ x1 (k ) x2 (n  k )]ß e jX n
n ‡ Ð Ñ k ‡ àÑ
Interchanging the order of summations, we get
‡ ‡
F{x1 (n) x2 (n)} Ç x1 ( k ) Ç x2 (n  k ) e  jX n
k ‡ n ‡

Put n – k = p in the second summation.

\ n =p+k
‡ ‡
F{x1 (n) x2 ( n)} Ç x1 ( k ) Ç x2 ( p) e  jX ( p  k )
k ‡ p ‡
‡ ‡
Ç x1 (k ) e  jX k Ç x2 ( p) e  jX p
k ‡ p ‡
X1 (X ) X2 (X )
That is, the convolution of the signals in the time domain is equal to multiplying their
spectra in the frequency domain.

5.6.8 Frequency Convolution Property

The frequency convolution property of DTFT states that
If F{x1(n)} = X1(w) and F{x2(n)} = X2(w)
Q
1
Then F{x1 (n) x2 ( n)} X1 (X ) X 2 (X ) Ô X1 (R ) X2 (X  R ) dR
2Q Q
‡
Proof: F[x1 (n) x2 (n)] Ç x1 (n) x2 (n) e  jX n
n ‡
‡ Ë 1 Q Û
Ç Ì 2Q Ô X1(R ) e jR n dR Ü e jX n x2 (n)
n ‡ ÌÍ Q ÜÝ
374 | Digital Signal Processing

Interchanging the order of summation and integration, we get

Q Ë ‡ Û
Ì Ç x 2 ( n ) e  j ( X  R ) n Ü dR
1
F{x1 (n) x2 (n)} Ô
2Q Q
X1 (R )
ÍÌ n ‡ ÝÜ
Q
1
2Q ÔQ
X1 (R ) X 2 (X  R ) dR

This operation is known as periodic convolution because it is the convolution of two periodic
functions X1(w) and X2(w).

5.6.9 The Correlation Theorem

The correlation theorem of DTFT states that
If F{x1(n)} = X1(w) and F{x2(n)} = X2(w)
Then F{Rx1 x2 (l)} X1 (X ) X 2 (  X ) = * x1 x2 (X )

The function * x1 x2 (X ) is called the cross energy spectrum of the signals x1(n) x2(n).

5.6.10 The Modulation Theorem

If F{x(n)} = X(w)
1
Then F{x ( n) cos X 0 n} {X (X  X 0 )  X (X  X 0 )}
2
‡ e jX 0 n  e  jX0 n  jX n
Proof: F{x (n) cos X 0 n} Ç x ( n) e
n ‡ 2
1 ÎÑ ‡  j (X  X 0 ) n 
‡ ÞÑ
Ï Ç x ( n) e Ç x (n) e  j (X  X 0 ) n ß
2 ÐÑ n ‡ n ‡ àÑ
1
{X (X  X 0 )  X (X  X 0 )}
2

5.6.11 Parseval’s Theorem

If F{x(n)} = X(w)
‡
Ç x ( n)
2
Then E
n ‡
Q
1
Ô X (X ) dX
2
2Q Q
 Discrete-time Fourier Transform | 375
‡ ‡
Ç Ç
2
Proof: E x (n) = x (n) x *(n)
n ‡ n ‡
‡ ÎÑ 1 Q ÞÑ
*

Ç x ( n) Ï Ô X (X ) e jX n dX ß
n ‡ ÐÑ 2Q Q àÑ
‡ ÎÑ 1 Q ÞÑ
Ç x(n) Ï 2Q Ô X * (X ) e jXn dX ß
n ‡ ÐÑ Q àÑ
Interchanging the order of summation and integration, we get

1
Q ÎÑ ‡  jX n ÞÑ dX
E Ô X *
(X ) Ï Ç x ( n) e ß
2Q Q ÐÑ n ‡ àÑ
Q
1
2Q ÔQ
X * (X ) X (X ) dX

Q
1
Ô
2
X (X ) dX
2Q Q

5.6.12 Symmetry Properties

The DTFT X(w) is a complex function of w and can be expressed as:
X (X ) X R (X )  jX I (X )

where XR(w) is real part and XI (w) is imaginary part of X(w) respectively. We have
‡
X (X ) Ç x (n) e  jX n
n ‡
‡ ‡
= Ç x (n) cos X n  j Ç x (n) sin X n
n ‡ n ‡
‡ ‡
i.e. X R (X )  jX I (X ) Ç x (n) cos X n  j Ç x (n) sin X n
n ‡ n ‡

Comparing LHS and RHS, we have

‡
X R (X ) Ç x (n) cos X n
n ‡
‡
X I (X )  Ç x (n) sin X n
n ‡
376 | Digital Signal Processing

Since cos(–w) n = cosw n, and sin(–w) n = – sinw n

‡ ‡
X R ( X ) Ç x ( n) cos( X ) n Ç x (n) cos X n X R (X )
n ‡ n ‡

i.e. X R ( X ) X R (X ) (Even symmetry)

‡ ‡
X I ( X )  Ç x (n) sin ( X ) n Ç x (n) sin X n  X I (X )
n ‡ n ‡

i.e. X I ( X )  X I (X ) (Odd symmetry)

Therefore, XR(w) is an even function of w and XI(w) is an odd function of w.
We can write XR(w) in the polar form as:
X (X ) X (X ) e jR (X )
where |X(w)| is the magnitude and q (w) is the phase of X(w).
Expanding X (X ) X (X ) {cos R (X )  j sin R (X )}
i.e. X R (X )  jX I (X ) X (X ) cos R (X )  j X (X ) sin R (X )
Comparing LHS and RHS, we get
X R (X ) X (X ) cos R (X )
X I (X ) X (X ) sin R (X )
X (X ) {X R (X )}2  {X I (X )}2
2

i.e. X (X ) {X R (X )}2  {X I (X )}2

X I (X )
and tan R (X )
X R (X )
X I (X )
or R (X ) = tan 1
X R (X )

Similarly, X ( X ) {X R ( X )}2  {X I ( X )}2

X R (X )  X I (X )
2 2

X (X )
Therefore, X(w) is an even function of w,
X I ( X )
R ( X ) tan 1
X R ( X )
 X I (X )
= tan 1
X R (X )
X I (X )
 tan 1 R (X )
X R (X )
 Discrete-time Fourier Transform | 377

\ R ( X )  R (X )
That is X ( X )  X (X )

Therefore X (X ) is an odd function of w.

TABLE 5.1 Properties of DTFT

Property Sequence DTFT

x(n) X(w)
x1(n) X1(w)
x2(n) X2(w)
Linearity ax1(n) + bx2(n) aX1(w) + bX2(w)
Time shifting x(n – m) e–jw m X(w)
Time reversal x(–n) X(–w0)
Frequency shifting x(n)e jw 0n X(w – w)
d
Differentiation in frequency domain nx(n) j [ X (X )]
dX
Convolution x1(n) * x2(n) X1(w) X2(w)
Multiplication x1(n) x2(n) X1(w) * X2(w)
Correlation Rx1x2 (l) X1(w) X2(–w)

Modulation theorem x(n) cos w0n 1

{X (X  X 0 )  X (X  X 0 )}
2
 Q
 1
X (X ) dX
2

2
Parseval’s theorem x ( n)
n   2Q
Q
Symmetry property x*(n) X(–w)
x*(–n) X*(w)
xR(n) Xe(w)
jxI (n) X0(w)
xe(n) XR(w)
x0(n) jXI (w)

EXAMPLE 5.9 Using properties of DTFT, find the DTFT of the following:
n 2 n 3
È 1Ø È 1Ø
(a) ÉÊ 4 ÙÚ (b) É Ù u (n  3)
Ê 3Ú
(c) E (n  2)  E ( n  2) (d) u(n  1)  u(n  2)
(e) n2 n u(n) (f) u(  n)
n
(g) n3 u(  n) (h) e3n u( n)
378 | Digital Signal Processing

Solution:
(a) Using the time shifting property, we have
ÎÑÈ 1 Ø n 2 ÞÑ ÎÑÈ 1 Ø n ÞÑ
F ÏÉ Ù ß e  j 2X F ÏÉ Ù ß
Ê 4Ú Ê 4Ú Ñ
ÐÑ àÑ ÐÑ à
ÎÑÈ 1 Ø n ÞÑ ‡ È 1Ø n
F ÏÉ Ù ß Ç ÉÊ 4 ÙÚ e jX n
ÑÐÊ 4 Ú Ñà n ‡

1 È 1 Ø  n ‡ È 1Øn
 jX n   jX n
Ç ÉÊ ÙÚ e
4
Ç ÉÊ ÙÚ e
4
n ‡ n 0

‡ È 1Øn ‡ È 1Øn
Ç ÉÊ 4 ÙÚe jX n  Ç É Ù e  jX n
n 1
Ê Ú
n 0 4

1 jX 1 1 15 /16
e 
4 1  (1/4) e jX 1  (1/4) e  jX (17/16)  (1/2) cos X

ÎÑÈ 1 Ø n  2 ÞÑ
e  j 2X
15 /16
\ F ÏÉ Ù ß
Ê 4Ú (17 /16)  (1 / 2) cos X
ÐÑ àÑ
(b) Using the time shifting property, we have
ÎÑÈ 1 Ø n 3 ÞÑ ÎÑÈ 1 Ø n ÞÑ ÎÑ ÞÑ
e  j 3X F ÏÉ Ù u(n) ß e  j 3X Ï
1
F ÏÉ Ù u( n  3) ß  jX ß
Ê 3Ú Ê Ú ÑÐ1  (1/3) e Ñà
ÐÑ àÑ ÐÑ 3 àÑ
(c) Using the time shifting property, we have
F{E (n  2)  E (n  2)} = F{E (n  2)}  F{E (n  2)}

= e  j 2X F{E (n)}  e j 2X F{E (n)}

e  j 2X  e j 2X  2 j sin 2X
(d) Using the time shifting property, we have
F{u(n  1)  u(n  2)} = F{u( n  1)}  F{u( n  2)}
= e jX F{u(n)}  e j 2X F{u( n)}
e jX e j 2X
= 
1  e  jX 1  e  jX
(e) Using differentiation in the frequency domain property, we have
ÎÑ È 1 Ø n ÞÑ d Ë ÎÑÈ 1 Ø ÞÑÛ
n
F Ï n É Ù u( n) ß j Ì F ÏÉ Ù u(n) ßÜ
ÑÐ Ê 2 Ú Ñà dX Ì ÑÐÊ 2 Ú ÑàÜÝ
Í
 Discrete-time Fourier Transform | 379

d Ë 1 Û
j Ì  jX Ü
dX Í 1  (1/2) e Ý
 jX
{ [  (1 / 2) e (  j )]} (1 / 2) e  jX
j
{1  (1 / 2) e  jX }2 {1  (1/ 2) e  jX }2
(f) Using the time reversal property, we have
ÑÎ 1 ÑÞ 1
F{u(  n)} = F{u(n)} X X = Ï  jX ß jX
ÑÐ1  e ÑàX X 1  e
(g) Using differentiation in frequency domain and time reversal properties, we have
d
F{n3– n u(  n)} = j [F{3 n u(  n)}]
dX
d dË 1 Û
j [F{3n u( n)}]X X j Ì  jX Ü
dX dXÍ 1  3e ÝX X
jX
d Ë 1 Û  [  3e ( j )] 3e jX
j Ì jX Ü
j
dX Í 1  3e Ý [1  3e jX ]2 {1  3e jX }2
(h) Using the frequency shifting property, we have
F{e3n u(n)} F{u(n)} X X  3

ÑÎ 1 ÑÞ 1
Ï  jX ß  j (X  3)
ÑÐ1  e ÑàX X  3 1 e

EXAMPLE 5.10 Find the inverse Fourier transform for the first order recursive filter
H(w) = (1 – ae–jw )–1

(1  ae  jX ) 1
1
Solution: Given H (X )
1  ae  jX
1  ae  jX  a2 e  j 2X  a3e  j 3X  "
Let h(n) be the inverse Fourier transform of H(w).
‡
\ H (X ) Ç h(n) e  jX n "  h( 2) e j 2X  h( 1) e jX  h(0)  h(1) e  jX + h(2) e  j 2X  "
n ‡

On comparing the two expressions for H(w), we can say that the samples of h(n) are the
coefficients of e–jw n.
\ h(n) {1, a, a2 , !, ak , !}

ÎÑ a n , n•0
i.e. h ( n) Ï or h(n) a n u( n)
ÑÐ0, n0
380 | Digital Signal Processing

EXAMPLE 5.11 Determine the output sequence from the output spectrum:
1 e j 2X  1  e  j 2X
Y (X )
4 1  ae  jX
1 e j 2X  1  e  j 2X
Solution: Given Y (X )
4 1  ae  jX
The output sequence y(n) is the inverse Fourier transform of Y(w).

ÎÑ 1 e j 2X  1  e  j 2X ÞÑ
\ y ( n) F 1 Ï ß
ÐÑ 4 1  ae  jX àÑ

1 Ë 1 ÎÑ e j 2X ÞÑ 1 Î
Ñ 1 ÞÑ Î e  j 2X ÑÞÛ
1 Ñ
ÌF Ï  jX ß  F Ï  jX ß  F Ï  jX ßÜ
4 ÌÍ ÑÐ 1  ae Ñà ÐÑ1  ae àÑ ÑÐ1  ae ÑàÜÝ

Using the time shifting property, we have

ÎÑ e j 2X ÞÑ Î
1 Ñ 1 ÑÞ
F 1 Ï  jX ß = F Ï  jX ß
= a n u( n) a n  2 u(n  2)
n n 2
ÐÑ1  ae àÑ ÐÑ1  ae àÑ n n 2

Also, we know that

ÎÑ 1 ÞÑ
F 1 Ï  jX ß
a n u( n)
ÐÑ1  ae àÑ
Using the time shifting property, we have

ÑÎ e  j 2X ÑÞ ÑÎ 1 ÑÞ
F 1 Ï  jX ß
F 1 Ï  jX ß
a n  2 u(n  2)
ÑÐ1  ae Ñà ÑÐ1  ae Ñà n n 2

1 n2
\ y ( n) {a u(n  2)  a n u(n)  a n  2 u(n  2)}
4
EXAMPLE 5.12 The impulse response of a LTI system is h(n) = {1, 2, 1, –2}. Find the
response of the system for the input x(n) = {1, 3, 2, 1}.
Solution: The response of the system y(n) for an input x(n) and impulse response h(n) is
given by
y(n) = x(n) * h(n)
Using the convolution property of Fourier transform, we get
Y (X ) X (X ) H (X )
\ y ( n) F 1{X (X ) H (X )}
Given x(n) = {1, 3, 2, 1}
 Discrete-time Fourier Transform | 381

\ X (X ) 1  3e  jX  2e  j 2X  e  j 3X
Given h(n) = {1, 2, 1, –2}

\ H (X ) 1  2e  jX  e  j 2X  2e  j 3X

Y(w) = X (X ) H (X ) (1  3e  jX  2e  j 2X  e  j 3X )(1  2e  jX  e  j 2X  2e  j 3X )
1  5e  jX  9e  j 2X  6e  j 3X  2e  j 4X  3e  j 5X  2e  j 6X
Taking inverse Fourier transform on both sides, we get
y(n) 1  5E (n  1)  9E ( n  2)  6E (n  3)  2E (n  4)  3E (n  5)  2E (n  6)
or y(n) {1, 5, 9, 6,  2,  3,  2}

EXAMPLE 5.13 Find the convolution of the signals given below using Fourier transform:
n
È 1Ø
x1 (n) ÉÊ 2 ÙÚ u(n)
n
È 1Ø
x 2 ( n) ÉÊ 3 ÙÚ u(n)

n
È 1Ø
Solution: Given x1 (n) ÉÊ 2 ÙÚ u(n)

1
\ X1 (X )
1  (1 / 2) e  jX
n
È 1Ø
x 2 ( n) ÉÊ 3 ÙÚ u(n)

1
\ X2 (X )
1  (1/ 3) e jX
Using the convolution property of Fourier transform, we get
F[ x1 (n) x2 (n)] X1 (X ) X 2 (X )
Ë 1 ÛË 1 Û
Ì  jX Ü Ì  jX Ü
Í 1  (1/ 2) e Ý Í 1  (1/ 3) e Ý
ÎÑ Ë 1 ÛË 1 Û ÞÑ
\ x1 ( n) x2 (n) F 1 Ï Ì  jX Ü Ì  jX Üß
ÐÑ Í 1  (1 / 2)e Ý Í 1  (1 / 3)e Ý àÑ

È 1 ØÈ 1 Ø Ë e jX ÛË e jX Û
Let X (X ) É Ù Ì jX Ü Ì jX Ü
Ê 1  (1 / 2)e  jX Ú Ê 1  (1/ 3)e  jX ÙÚ
É
ÍÌ e  (1 / 2) ÝÜ ÍÌ e  (1/ 3) ÝÜ
382 | Digital Signal Processing

X (X ) e jX
\
e jX [e jX  (1 / 2)] [ e jX  (1/ 3)]
A B 3 2
jX
 jX jX
 jX
e  (1 / 2) e  (1/ 3) e  (1/ 2) e  (1 / 3)

3e jX 2e jX
\ X (X ) 
e jX  (1/ 2) e jX  (1/ 3)
1 1
3  jX
2
1  (1 / 2)e 1  (1 / 3)e  jX
Taking inverse Fourier transform on both sides, we have
n n
È 1Ø È 1Ø
x ( n) x1 (n) x2 (n) 3 É Ù u(n)  2 É Ù u( n)
Ê 2Ú Ê 3Ú

EXAMPLE 5.14 Consider a discrete-time LTI system with impulse response

n
È 1Ø
h ( n) ÉÊ 2 ÙÚ u(n). Use Fourier transform to determine the response to the signal

n
È 3Ø
x (n) ÉÊ 4 ÙÚ u(n).

Solution: Given the impulse response h(n) and the input x(n) to the system, the response
y(n) is given by
y ( n) x ( n) h ( n)
Using the convolution property of Fourier transform, we have

y( n) x ( n) h ( n) F 1 [ X (X ) H (X )]
where X(w) and H(w) are the Fourier transforms of x(n) and h(n), respectively.
n
È 3Ø
x ( n) ÉÊ 4 ÙÚ u(n)

1
\ X (X )
1  (3 / 4) e  jX
n
È 1Ø
h ( n) ÉÊ 2 ÙÚ u( n)

1
\ H (X )
1  (1 / 2) e  jX
 Discrete-time Fourier Transform | 383

Ë 1 ÛË 1 Û
\ Y(w) = X (X ) H (X ) Ì  jX Ü Ì  jX Ü
Í 1  (3 / 4) e Ý Í 1  (1/ 2) e Ý
Ë e jX ÛË e jX Û
Ì jX Ü Ì jX Ü
ÍÌ e  (3 / 4) ÝÜ ÍÌ e  (1 / 2) ÝÜ

Y (X ) e jX
\
e jX [ e jX  (3 / 4)] [ e jX  (1/ 2)]
A B 3 2
jX
 jX jX
 jX
e  (3 / 4) e  (1/ 2) e  (3 / 4) e  (1/ 2)

3e jX 2e jX 3 2
\ Y (X ) jX
 jX  jX

e  (3 / 4) e  (1/ 2) 1  (3 / 4) e 1  (1/ 2) e  jX

Taking inverse Fourier transform on both sides, we get the response

n n
È 3Ø È 1Ø
y ( n) 3 É Ù u( n)  2 É Ù u( n)
Ê 4Ú Ê 2Ú

5.7 TRANSFER FUNCTION

If X(w) is the Fourier transform of the input signal x(n), and Y(w) is the Fourier transform of
the output signal y(n), then H(w), the transfer function of the system is given by
Y (X )
H (X )
X (X )
i.e. the transfer function of an LTI system is defined as the ratio of the Fourier transform of
the output to the Fourier transform of the input. It is also defined as the Fourier transform of
the impulse response h(n) of the system. It can be obtained as follows:
If the input to the system is of the form e jw n, then the output y(n) is given by
‡
y(n) h(n) x (n) Ç h( k ) e jX ( n  k ) e jX n H (X )
k ‡

i.e. y(n) = x(n) H(w)

Q
X (X ) e jX n dX
1
2Q ÔQ
We have x ( n)

Q Q
X (X ) e jX n dX H (X ) X (X ) H (X ) e jX n dX
1 1
2Q ÔQ 2Q ÔQ
y ( n)
384 | Digital Signal Processing

y(n) in terms of Y(w) is given by

Q
Y (X ) e jX n dX
1
2Q ÔQ
y ( n)

Comparing the above two expressions for y(n), we get

Y(w) = X(w) H(w)

Y (X )
\ H (X )
X (X )
where H(w) is known as the transfer function of the system.

5.8 FREQUENCY RESPONSE OF DISCRETE-TIME SYSTEMS

The frequency response of a linear time-invariant discrete-time system can be obtained by
applying a spectrum of input sinusoids to the system. The frequency response gives the gain
and phase response of the system to the input sinusoids at all frequencies.
Let h(n) be the impulse response of an LTI discrete system, and let the input x(n) to the
system be a complex exponential e jw n.
The output of the system y(n) can be obtained by using convolution sum.
‡
y(n) h(n) x ( n) Ç h(k ) x (n  k )
k ‡
For x(n) = e jw n,
‡
y ( n) h(k ) e jX ( n  k )
Ç
k ‡
‡
e jX n Ç h(k ) e  jX k
k ‡
jX n
eN H
N (X )
Input Frequency response

That is, if we force the system with a complex exponential e jw n, then the output is of the
form H(w)e jw n. Therefore, the output of the system is identical to the input modified in
amplitude and phase by H(w). The quantity H(w) is the frequency response of the system.
It is same as the transfer function H(w) in the frequency domain. The frequency
response H(w) is complex and can be expressed in the polar form as:

H (X ) H (X ) e j ²H (X )
where the magnitude of H(w), i.e. |H(w)| is called the magnitude response and phase angle of
H(w), i.e. ÐH(w) is called the phase response. The plot of |H(w)| versus w is called the
magnitude response plot and the plot of ÐH(w) versus w is called the phase response plot.
 Discrete-time Fourier Transform | 385

Properties of frequency response

Frequency response is a complex function that describes the magnitude and phase shift of a
filter over a range of frequencies. If h(n) is a real sequence, the frequency response H(w) has
the following properties:
1. H(w) takes on values for all w, i.e. on a continuum of w.
2. H(w) is periodic in w, with period of 2p.
3. The magnitude response |H(w)| is an even function of w and symmetrical about p.
4. The phase response ÐH(w) is an odd function of w and antisymmetrical about p.

EXAMPLE 5.15 Write a difference equation that characterizes a system whose frequency
response is:
1  e  jX  3e  j 2X
H (X )
1  (1/ 3) e  jX  (1/ 6) e  j 2X

Y (X ) 1  e  jX  3e  j 2X
Solution: Given H (X )
X (X ) 1  (1/ 3) e  jX  (1/ 6) e  j 2X
On cross multiplication, we get

Y (X )  e  jX Y (X )  e  j 2X Y (X ) X (X )  e  jX X (X )  3e  j 2X X (X )
1 1
3 6
Taking inverse Fourier transform on both sides, we get the difference equation:
1 1
y(n)  y(n  1)  y(n  2) x (n)  x (n  1)  3x (n  2)
3 6

EXAMPLE 5.16 Find the frequency response of the following causal systems:
3 1
(a) y(n)  y(n  1)  y(n  2) x (n)  x (n  1)
16 2
1 3
(b) y(n)  y( n  1)  y(n  2) x ( n)  x (n  1)
4 8
Solution: The frequency response of a system is given by H(w) = [Y(w)/X(w)], where Y(w)
and X(w) are the Fourier transforms of output and input signals respectively.
3 1
(a) Given y(n)  y(n  1)  y(n  2) x (n)  x (n  1)
16 2
Taking Fourier transform on both sides, we have

Y (X )  e  jX Y (X ) 
3  j 2X
Y (X ) X (X )  e  jX X (X )
1
e
16 2
È Ø È Ø
Y (X ) É 1  e  jX  e  j 2X Ù X (X ) É 1  e  jX Ù
3 1
i.e.
Ê 16 Ú Ê 2 Ú
Y (X ) 1  (1/ 2) e  jX e jX [e jX  (1 / 2)]
\ H (X )
X (X ) 1  e  jX  (3 /16) e  j 2X e j 2X  e jX  (3 /16)
386 | Digital Signal Processing

1 3
(b) Given y ( n) y( n  1)  y(n  2) x ( n)  x (n  1)
4 8
Taking Fourier transform on both sides, we have

Y (X )  e  jX Y (X )  e  j 2X Y (X ) X (X )  e  jX X (X )
1 3
4 8
È Ø
Y (X ) É1  e  jX  e  j 2X Ù X (X ) (1  e  jX )
1 3
i.e.
Ê 4 8 Ú
Y (X ) 1  e  jX e jX (e jX  1)
\ H (X )
X (X ) 1  (1/ 4) e  jX  (3 / 8) e  j 2X e j 2X  (1 / 4) e jX  (3 / 8)

EXAMPLE 5.17 A discrete system is given by the following difference equation:

y(n)  5 y(n  1) x (n)  4 x (n  1)
where x(n) is the input and y(n) is the output. Determine its magnitude and phase response.
Solution: Given y(n)  5 y(n  1) x (n)  4 x (n  1)
Taking Fourier transform on both sides, we have
Y (X )  5e  jX Y (X ) X (X )  4e  jX X (X )
i.e. Y (X ) {1  5e  jX } X (X ) {1  4e  jX }
The transfer function of the system, i.e. the frequency response of the system is:
Y (X ) 1  4 e  jX e jX  4
H (X )
X (X ) 1  5e  jX e jX  5
cos X  j sin X  4
cos X  j sin X  5
The magnitude response of the system is:

(4  cos X )2  (sin X )2 17  8 co s X
H (X )
(cos X  5)2  (sin X )2 26  10 cos X
The phase response of the system is:
Î sin X Þ 1 Î sin X Þ
H (X ) tan 1 Ï ß  tan Ï ß
Ð 4  cos X à Ð cos X  5 à

EXAMPLE 5.18 The output y(n) for a linear shift-invariant system, with input x(n) is
given by
y ( n) x (n)  2 x (n  1)  x (n  2)
Determine the magnitude and phase response of the system.
 Discrete-time Fourier Transform | 387

Solution: Given y ( n) x (n)  2 x (n  1)  x (n  2)

Taking Fourier transform on both sides, we have
Y (X ) X (X )  2e  jX X (X )  e  j 2X X (X )
The transfer function of the system, i.e. the frequency response of the system is:
Y (X )
H (X ) 1  2e  jX  e  j 2X
X (X )
1  2[cos X  j sin X ]  cos 2X  j sin 2X
[1  2 cos X  cos 2X ]  j[2sin X  sin 2X ]
The magnitude response of the system is:

H (X ) [1  2 cos X  cos 2X ]2  [2 sin X  sin 2X ]2

The phase response of the system is given as:
Î 2 sin X  sin 2X Þ
H (X ) tan 1 Ï ß
Ð1  2 cos X  cos 2X à

EXAMPLE 5.19 The impulse response of a system is:

ÎÑ1 for 0 n N 1
h(n) Ï
ÑÐ0 otherwise
Find the transfer function, frequency response, magnitude response and phase response.
Solution: The transfer function H(w) is obtained by taking the Fourier transform of h(n).
‡ N 1
\ H (X ) Ç h(n) e  jX n Ç (1) e  jX n
n ‡ n 0

1e  jX N

1  e  jX
The frequency response of the system is same as the transfer function, i.e.
1  e  jX N
H (X )
1  e  jX
The magnitude function is given as:
H (X ) {H (X ) H * (X )}1/2

ÎÑË 1  e  jX N Û Ë 1  e jX N
1/2
Û ÞÑ
ÏÌ  jX Ü Ì jX Üß
ÐÑÌÍ 1  e ÜÝ ÌÍ 1  e ÜÝ àÑ
388 | Digital Signal Processing

jX N
 e  jX N ÑÞ
1/2
ÑÎ1  1  e
Ï jX  jX ß
ÑÐ 1  1  e  e Ñà

ÎÑ 2  (e jX N  e  jX N ) ÞÑ Î1  cos X N Þ
1/2 1/2

Ï ß Ï ß
Ð 1  cos X à
jX  jX
ÑÐ 2  (e  e ) Ñà

In order to determine the phase function, the real and imaginary parts of H(w) have to be
separated.
1  e  jX N 1  e jX 1  e jX  e  jX N  e  jX ( N  1)
\ H (X ) –
1  e  jX 1  e jX 2  (e jX  e  jX )
1  (cos X  j sin X )  (cos X N  j sin X N )  [cos X ( N  1)  j sin X ( N  1)]
2  2 cos X
1  cos X  cos X N  cos X ( N  1)
Now, H R (X )
2  2 cos X
 sin X  sin X N  sin X ( N  1)
H I (X )
2  2 cos X
H I (X )
\ H (X ) tan 1
H R (X )
Î  sin X  sin X N  sin X ( N  1) Þ
tan 1 Ï ß
Ð1  cos X  cos X N  cos X ( N  1) à

EXAMPLE 5.20 The impulse response of a LTI system is given by h(n). Find the
frequency response, magnitude and phase response.
Solution: Given h(n) = 0.6n u(n)
The frequency response of the system is:
1
H (X ) = F{0.6 n u(n)}
1  0.6e  jX
Here H(w) is a complex function of frequency. To separate the real and imaginary parts of
H(w), multiply the numerator and denominator by the complex conjugate of the denominator.
Thus,
1 1  0.6e jX 1  0.6e jX
H (X ) –
1  0.6e  jX 1  0.6e jX 1  0.6e jX  0.6e  jX  0.36
1  0.6 cos X 0.6 sin X
j
1.36  1.2 cos X 1.36  1.2 cos X
1  0.6 cos X 0.6 sin X
\ H R (X ) and H I (X )
1.36  1.2 cos X 1.36  1.2 cos X
 Discrete-time Fourier Transform | 389

The magnitude function of H(w ) is:

H (X ) {H R (X )2  H I (X )2 }1/2
1/2
ÎÑ Ë 1  0.6 cos X Û2 Ë 0.6 sin X Û 2 ÞÑ
ÏÌ Ü Ì Ü ß
1.36  1.2 cos X Ý Í 1.36  1.2 cos X Ý Ñ
ÐÑ Í à
1
(1.36  1.2 cos X )1/2
The phase function of H(w ) is:
H I (X ) Î  0.6 sin X Þ
H (X ) tan 1 tan 1 Ï ß
H R (X ) Ð1  0.6 cos X à
The magnitude function |H(w )| can be obtained as follows:

H (X ) {H (X ) H * (X )}1/2 {H (X ) H ( X )}1/2
1/2 1/2
È 1 1 Ø Ë 1 Û
É  jX
– jX Ù Ì jX  jX Ü
Ê 1  0.6e 1  0.6e Ú Í 1  0.36  0.6 (e  e ) Ý
1
(1.36  1.2 cos X )1/2

EXAMPLE 5.21 A causal LTI system is described by the difference equation:

y(n)  ay (n  1) bx (n)  x ( n  1)
where a is real and less than 1 in magnitude. Find a value of b(b ¹ a) such that the
frequency response of the system satisfies |H(w)| = 1 for all w (an all pass system, the
magnitude of the frequency response is constant independent of frequency).
Solution: Given y(n)  ay (n  1) bx (n)  x ( n  1)
Taking Fourier transform on both sides, we have
Y (X ) [1  ae  jX ] X (X ) [b  e  jX ]

Y (X ) b  e  jX b  cos X  j sin X
\ H (X )
X (X ) 1  ae  jX 1  a cos X  ja sin X

(b  cos X )2  sin 2 X 1  b2  2b cos X

H (X )
2

(1  a cos X )2  a 2 sin 2 X 1  a2  2 a cos X

For magnitude to be independent of frequency,
d
H (X ) 2
0
dX
390 | Digital Signal Processing

d È 1  b2  2b cos X Ø
i.e. 0
dX ÉÊ 1  a2  2a cos X ÙÚ
Simplifying, we get b = 1/a.

EXAMPLE 5.22 A causal and stable LTI system has the property that
n n
È 4Ø È 4Ø
ÉÊ 5 ÙÚ u(n)  n É Ù u( n)
Ê 5Ú
(a) Determine the frequency response H(w) for the system.
(b) Determine a difference equation relating any input x(n) and the corresponding
output y(n).
n
È 4Ø
Solution: Given x ( n) ÉÊ 5 ÙÚ u(n)
1
\ X (X )
1  (4 / 5) e  jX
n
È 4Ø
y ( n) n É Ù u( n)
Ê 5Ú
d (4 / 5) e  jX
\ Y (X ) j [ X (X )]
dX [1  (4 / 5) e  jX ]2
(4 / 5) e  jX 1 (4 / 5) e  jX
\ Y (X ) X (X )
[1  (4 / 5) e  jX ] [1  (4 / 5) e  jX ] [1  (4 / 5) e  jX ]
Therefore, the frequency response is:
Y (X ) (4 / 5) e  jX
H (X )
X (X ) [1  (4 / 5) e  jX ]
Cross multiplying, we get
4  jX 4  jX
Y (X )  e Y (X ) e X (X )
5 5
Taking inverse Fourier transform, we get the difference equation:
4 4
y(n)  y(n  1) x (n  1)
5 5

EXAMPLE 5.23 Determine the impulse response of all the four types of ideal filters
shown in Figure 5.1.
Solution:
(a) For an ideal low-pass filter shown in Figure 5.1(a),
ÎÑ1, for 0 X Xc
H (X ) Ï
ÑÐ0, otherwise
 Discrete-time Fourier Transform | 391

Figure 5.1 Frequency response of ideal filters, (a) low-pass filter, (b) high-pass filter, (c) band pass
filter, and (d) band stop filter.

Q
H (X ) e jX n dX
1
2Q ÔQ
We have h( n)

X
1 c jX n e jX c n  e  jX c n sin X c n
2Q XÔ
e d X
Q n(2 j) Qn
c

sin X c n
i.e. h( n) ‡ n ‡
Qn
(b) For an ideal high-pass filter shown in Figure 5.1(b),
ÎÑ0, for 0 X Xc
H (X ) Ï
ÑÐ1, for X c X Q
Q
H (X ) e jX n dX
1
2Q ÔQ
We have h( n)

X Q
1 Ë c jX n Û
Ì Ô e dX  Ô e jX n dX Ü
2Q Ì Q Xc Ü
Í Ý
[ e  jX c n  e  jQ n  e jQ n  e jX c n ]
1
2Q jn

1 Ë È e jQ n  e  jQ n Ø È e jX c n  e  jX c n Ø Û
Ì
Q n ÌÍ ÉÊ Ù É ÙÜ
2j Ú Ê 2j Ú ÜÝ
392 | Digital Signal Processing

1
[sin Q n  sin X c n]
Qn
sin X c n

Qn
sin X c n
i.e. h(n)  ‡ n ‡
Qn
(c) For an ideal band pass filter shown in Figure 5.1(c),

ÎÑ1, for X c1 X X c2
H (X ) Ï
ÑÐ0, otherwise

Q
1
Ô H (X ) e dX
jX n
We have h( n)
2Q Q

1 È Xc1 Xc 2 Ø
É Ô e dX  Ô e dX Ù
jX n jX n
2Q ÉÊ X X c1 ÚÙ
c2

1
[e  jX c1n  e  jX c 2 n  e jX c 2 n  e jX c1n ]
2Q jn

1 Ë È e jXc 2 n  e  jX c 2 n Ø È e jXc1n  e  jX c1n Ø Û

Qn
ÌÉ Ù É ÙÜ
ÌÍ Ê 2j Ú Ê 2j Ú ÜÝ
1
[sin X c 2 n  sin X c1n]
Qn
sin X c2 n  sin X c1n
i.e. h(n) ‡ n ‡
Qn
(d) For an ideal band stop filter shown in Figure 5.1(d),

ÎÑ1, for X X c1 and X c 2 X Q

H (X ) Ï
ÑÐ0, for X c1 X X c 2

Q
H (X ) e jX n dX
1
2Q ÔQ
We have h(n)

È X c 2 X c1 Q Ø
É Ô e jX n dX  Ô e jX n dX  Ô e jX n dX Ù
1
2Q ÉÊ Q X c1 Xc2 ÙÚ
 Discrete-time Fourier Transform | 393

[e  jX c 2 n  e  jQ n  e jX c1n  e  jX c1n  e jQ n  e jX c 2 n ]
1
2Q jn
1 Ë È e jX c 2 n  e  jX c 2 n Ø È e jX c1n  e  jX c1n Ø È e jQ n  e  jQ n Ø Û
Ì É Ù É Ù É ÙÜ
Qn ÌÍ Ê 2j Ú Ê 2j Ú Ê 2j Ú ÜÝ
1
[sin X c1n  sin X c 2 n]
Qn
sin X c1n  sin X c 2 n
i.e. h( n) ‡ n ‡
Qn

SHORT QUESTIONS WITH ANSWERS

1. Define Fourier transform of a discrete-time signal.
Ans. The Fourier transform of a discrete-time signal x(n) is defined as:
‡
F{x (n)} X (X ) Ç x (n) e  jX n
n ‡
‡
The Fourier transform exists only if Ç x (n)  ‡, i.e. only if the sequence is
n ‡
absolutely summable.
2. Define inverse discrete-time Fourier transform.
Ans. The inverse discrete-time Fourier transform of X(w) is defined as:
Q
X (X ) e jX n dX
1
F 1{X (X )}
2Q ÔQ
x ( n)

3. Why the Fourier transform of a discrete-time signal is called signal spectrum?

Ans. By taking Fourier transform of a discrete-time signal x(n) it is decomposed
into its frequency components. Hence the Fourier transform is called signal spectrum.
4. List the differences between Fourier transform of discrete-time signal and analog signal.
Ans.
1. The Fourier transform of analog signal consists of a spectrum with frequency
range – ¥ to ¥, but the Fourier transform of a discrete-time signal is unique in
the range – p to p (or 0 to 2p), and also it is periodic with periodicity of 2p.
2. The Fourier transform of analog signal involves integration, but Fourier
transform of discrete-time signal involves summation.
5. Give some applications of discrete-time Fourier transform.
Ans. Some applications of discrete-time Fourier transform are:
1. The frequency response of LTI system is given by the Fourier transform of the
impulse response of the system.