F 1.

02,
0.56,
2.09,
0.69,
-0.95

Transform Domain
Representation of
Discrete Time Signals

The Discrete Fourier

Transform
(I)

Yogananda Isukapalli
1

1
 §F§1.02,
0.56,
2.09,
§ 0.69,
§ -0.95

1. Fourier Series Periodic waveforms

• Any periodic waveform, f(t), can be

represented as the sum of an infinite number of
sinusoidal and cosinusoidal terms, together
with a constant term:
¥ ¥
f (t ) = a 0 + åa
n =1
n cos(nw t ) + åb
n =1
n sin( nw t ) (1)

where
w = 2p / T p is the first harmonic or fundamental
angular frequency
Tp is the repetition period of the
waveform

nw discrete nth harmonics of w

2
 F 1.02,
0.56,
2.09,
0.69,
-0.95

Tp / 2
1
a0 =
Tp
-T p / 2
ò f (t )dt is a constant, like a DC
voltage level
Tp / 2
2
an =
Tp ò f (t ) cos(nw t )dt
-T p / 2
and

Tp / 2
2
bn =
Tp ò f (t ) sin(nw t )dt
-T p / 2

Eqn (1) can also be represented as

¥
f (t ) = å
n = -¥
d n e jnw t (2)

Tp / 2
1
where dn =
Tp ò f (t )e - jnw t dt
-T p / 2
(3)

is complex and d n has the units of volts 3

3
 F 1.02,
0.56,
2.09,
0.69,
-0.95

2. Fourier Transform Non-periodic

waveforms

• Consider a non-periodic waveform, obtained

by making the period T p of the periodic
waveform to be infinite, i.e. T p ® ¥
• As T p is increased, the spacing between the
harmonic components, 1 / T p = w / 2p ,
decreases to dw / 2p , eventually becoming
zero.
ß
discrete frequency variable nw changes to the
continuous variable w

Þ Amplitude and phase spectra become

continuous
Thus, d n ® d (w ) as T p ® ¥
4

4
 F 1.02,
0.56,
2.09,
0.69,
-0.95

With these changes, eqn (3) becomes

¥
dw
d (w ) =
2p ò
-¥
f (t )e - jw t dt (4)

Dividing by dw / 2p ,
¥
d (w )
dw / 2p
= F ( jw ) = ò
-¥
f (t )e - jw t dt (5)

F ( jw ) Fourier Transform

F ( jw ) Amplitude spectral density

(continuous with units volts per
hertz)
2 Energy spectral density
F ( jw )
(continuous with units joules
per hertz)
5

5
 F 1.02,
0.56,
2.09,
0.69,
-0.95

3. Discrete Fourier Transform

• In practice the Fourier components of data
are obtained by digital computation rather
than by analog processing.
• So, the analog waveforms are digitized using
a sample-and-hold circuit followed by an
analog-to-digital converter and under the
Nyquist criterion for sampling.
• Thus the data to be transformed is discrete
and probably non-periodic.
• It is not possible to apply Fourier transform
because it is for continuous data.
• Analog transform for use with discrete data
Discrete Fourier Transform (DFT)

6
 F 1.02,
0.56,
2.09,
0.69,
-0.95

Given: Finite (non-periodic) Duration

Discrete - Time Signal
x[nT]

0 1 2 3 4 5 6
x[nT] = 0 n<0, n>N-1
¹0 for 0£n£N-1
DFT of x[nT] X(kW) where W=2p/NT
N -1
X (kW) = FD [ x(nT )] = å x[nT ]e - jWnkT
n =0 DFT
N -1 2p
- j( ) nk
= å x[nT ]e
n =0
N

k = 0,...... N -1

!
N -1 2p
1 j( ) nk
x[nT ] =
N å X (kW)e
k =0
N
n = 0,...... N -1 IDFT
7

7
 F 1.02,
0.56,
2.09,
0.69,
-0.95

• Note: From now on we assume

X(k) represents X(kW), x[n] represents x[nT]
Relation between DFT and Fourier transform
• The DFT equation can be seen analogous to
the Fourier transform equation (5) by putting
x(nT) = f(t), kW = w , and nT = t.
• Making these substitutions in eqn (5), and
putting dt=T and replacing the integral with a
summation gives
N -1

å
n =0
x(nT )e - jkWnT T = F ( jw ) for 0 £ t £ ( N - 1)T (6)

• Now comparing eqn (6) with the DFT eqn

gives F ( jw ) = TX (k ) i.e.
The Fourier transform components may be
obtained by multiplying the DFT components by
the sampling interval.
8

8
 F 1.02,
0.56,
2.09,
0.69,
-0.95

Example:
x[n]={1,1,0,0} n=0,1,2,3
3 2p
- j( ) nk
X (k ) = å x[n]e 4

n =0 k = 0 ,1, 2 , 3

X(0)=1+1=2, X(1)=1+e-jp/2=1-j
X(2)=1+ e-jp =0, X(3)=1+ejp/2=1+j
Note X(4)=X(0), X(5)=X(1), X(6)=X(2),
and so on: X(k+4)=X(k) k=0,1,2,3
{1,1,0,0} <=>{2,1-j,0,1+j}

Example: Consider an analog signal x(t)

sampled with T=0.01 and the sampled
values are :
n 0 1 2 3 4 5
x[n] 5.0 -1.5 6.5 -3.0 6.5 -1.5
9

9
F 1.02,
0.56, 5 - j(
2p
) nk
X (k ) = å x[n]e
2.09,
0.69,
6
-0.95
n =0 k = 0 ,1, 2 , 3, 4 , 5

5
X (0) = å x[n] = 12
n =0
p
5 - j ( 2 )n
X (1) = å x[n]e 6 =0
n =0
p
5 - j ( 2 )2n
X (2) = å x[n]e 6 = -3
n =0
5 2p
- j(
X (3) = å x[n]e
) 3n
6 = -24
n =0
5 2p
- j(
X (4) = å x[n]e
)4n
6 = -3
n =0
5 2p
- j(
X (5) = å x[n]e
)5n
6 =0
n =0

Now taking the IDFT:

2p
1 5 j( ) nk
x[n] = å X (k )e 6
6 k =0 n = 0 ,1, 2 , 3, 4 , 5 10

10
F 1.02,
0.56,
2.09,
0.69,
-0.95 First Write :
2p 2p 2p
1 j( )2n j( )3n j( )4n
x[n] = [12 - 3e 6
+ 24e 6
- 3e 6
6
2p 2p
Since: e
j(
6
)4n
=e
- j(
6
)2n
Then:
2p 2p 2p
j( )2n - j( )2n j( )3n
x[n] = 2 - 0.5{e 6
+e 6
} + 4e 6

2pn
= 2 - cos( ) + 4 cos(pn)
3
x[0] = 2 - cos(0) + 4 cos(0) = 5
2p
x[1] = 2 - cos( ) + 4 cos(p ) = -1.5
3
4p
x[2] = 2 - cos( ) + 4 cos(2p ) = 6.5
3
6p
x[3] = 2 - cos( ) + 4 cos(3p ) = -3
3
8p
x[4] = 2 - cos( ) + 4 cos(4p ) = 6.5
3
10p
x[5] = 2 - cos( ) + 4 cos(5p ) = -1.5
3 11

11
 F 1.02,
0.56,
2.09,
0.69,
-0.95
DFT
{5.0, -1.5, 6.5, -3.0, 6.5, -1.5} Û {12, 0, -3, 24, -3, 0}
IDFT

• DFT: N-point transform is unique

N -1 2p
- j( ) nk
X (k ) = å x[n]e
n =0
N
(7)
k = 0,...... N -1
N -1 2p
1 j( ) nk
x[n] =
N å X (k )e N
n = 0,...... N -1
k =0 (8)
2p
- j( )
let..W n = e N

2p
j( )
W n-1 = e N ...Then

N -1
X (k ) = å x[n]W
n =0
kn
N
(9)
k = 0,...... N -1
N -1
1
x[n] =
N å
k =0
X (k )W N- kn n =0,...... N -1
(10)

12

12
 F 1.02,
0.56,
2.09,
0.69,
-0.95

Suppose we multiply (9) by WN-kp

where p=0,…..N-1 and sum from k=0 to N-1
N -1 N -1 N -1

å X (k )W
k =0
- kp
N = åå x[n]WNk ( n - p )
k =0 n =0

N -1
é N -1 k ( n - p ) ù
å
k =0
x[n]êå WN
ë n =0
ú
û p =0,1,...... N -1
N -1 N -1 2p
jk ( p - n )
= å WNk ( n - p ) = å e N

n =0 n =0
Orthogonal Property :
a.) p=n gives
N -1

å1 = N
k =0
b.) p¹n rewrite
N -1

åa
k =0
k

2p 13
j ( p -n)
a =e N
13
 F 1.02,
0.56,
2.09,
0.69,
-0.95

a¹1, p-n is between -(N-1) and +(N+1)

Follows : N -1
1-a N
å
k =0
a = k

1-a
.....a ¹ 1
2 PN
j ( p -n)
aN =e N
=1
N -1
Implies : å =0
a k

k =0

• Proved the orthogonality property of

distinct set of complex discrete
exponentials :
N -1

åW
k ( n- p ) N .. p = n
N = 0..otherwise
k =0

• We have now proved:

N -1 N -1

å X (k )W = å x[n]Nd (n - p )
- kp
N
k =0 k =0

= Nx( p ) 14

14
 F 1.02,
0.56,
2.09,
0.69,
-0.95

N -1
1
å X (k )WN
- kp
X ( p) =
N k =0 p = 0 ,1,..... N -1

or
N -1 2p
1 - j( )( kn )
x[n] =
N
å X ( k )e
k =0
N

• Conclusion: Given {x[n]}n=0, 1, ……..N-1

we obtain a unique set of values :
{X(k)}k=0, 1, ….N-1 as a result :
{x[n]} Û {X(k)}
for n=0,1…N-1 & k=0,1…N-1

15

15