ON-LINE DSP

CLASSES
LESSION -2 Date 21/4/2021
Topic :-- Fourier Analysis of Digital Signals

All signals, Analog or Digital, are expressed in two domains. Time Domain and Frequency
Domain. In other words, the function to describe the signal can be expressed as both
a. As a function of time ‘t’ (Time Domain)
b. As a function of frequency ‘f’ (Frequency Domain).

(a) When the amplitude variation of the signal w.r.t. time is considered, we express the signal
in Time Domain, such as

s (t) = 4 cos (100 π t) ….(1)

This is a signal with a cosinusoidal variation of amplitude, with a frequency of 50 Hz, and a
peak value of 4 V. This describes how the instantaneous amplitude of the signal varies w.r.t
time. We can also depict this in the form of a graph as follows. Both of these are Time
Domain Representations.

(b) This very signal can also be expressed as follows –

s (t) = 2 [ {e( j 314.28t ) }+ {e(− j 314.28 t )} ] ….(2)

This expression describes that the signal exists at two different locations having the
same frequency value with the amplitude divided into two equal parts. These two parts
are the mirror images of each other. This is described in a graphical form as follows—
In this graph, the independent variable is “frequency”. Thus this is the “Frequency Domain
Representation”.
Discrete Time Periodic Signals

When an Analog Signal is sampled, we get a Discrete Time Signal. A sampled signal is
expressed as a function of the Sample Number ‘n’.

A Discrete time cosine signal is as shown in the figure below. This consists of the samples of the
same Analog cosine signal in Eq.-1. The signal has been sampled 12 times per cycle.

Time Domain Expression of the Discrete Time Periodic Signal

A Discrete Time Function is a function of the Sample Number ‘n’. The Expression for the
Sampled Cosine Function shown above is

{ ( )}
s ( n )=4 cos 2 π
1
12
n ….(3)

From this expression the following General properties of a Discrete

Time Sinusoidal Function are observed --

i. Amplitude of the function is a real number ‘A’ (A= 4 Volt)

ii. Periodicity of the function is an integer ‘N’, with the unit “Samples per
Cycle” (N = 12)
iii. Frequency of the function is in the form of a “Rational Fraction” of the
general form” f = k/N” with the unit “Cycles per Sample”.
(AS USUAL Frequency is the Reciprocal of Periodicity, Hence f = 1/12).
iv. Any two Discrete Time Function differing by a frequency of 2 π are
IDENTICAL.
v. The HIGHEST possible oscillation in a Discrete Time Sinusoid can be 1/2.
Frequency Domain Expression of the Discrete Time Periodic Signal

(SPECTRUM OF DISCRETE TIME PERIODIC SIGNAL USING FOURIER SERIES )

To express any signal in Frequency Domain, we have to perform Fourier Analysis.

A Periodic Signal is expressed in a Fourier Series.
The Fourier Series expression of a Discrete Time Periodic Signal x (n) is defined as –

k=(N−1) 2 πkn

x(n) = ∑ Ck e N
(for k=0,1,2,..) …(4)
k=0

Where the term Ck is The Fourier Coefficient given by

n= ( N−1 ) −2 π k n
1
C k=
N
∑ x ( n) e N
(k=0,1,2,..(N-1)) …(5)
n=0

We can make the following observations from these two expressions--

a. In the eq- 4 the Fourier Series of the function x(n) is resolved into a FINITE series
of ‘N’ number of phasor terms at harmonics of the frequency k/N.
b. Each of these Fourier Coefficient Ck represent the amplitude and phase of this
sequence of phasors, each of which is at the harmonics of the frequency k/N.
This is given by the eq- (5).
c. Since the Fourier Coefficient Ck is a function of the Discrete Frequency term
“k/N” the Fourier Series representation of the Discrete Time Signal is in
Frequency Domain. Hence the Fourier Series represents “Spectrum” of the
signal.
d. Since the Fourier Coefficient Ck is a function of the given signal x (n), this can be
calculated and substituted in expression (4) in order to represent the Discrete
Time Signal in a Fourier Series.
e. The Fourier Coefficients Ck are at the same periodicity (N), as the periodicity of
the given function x (n). Thus, since ‘N’ consecutive samples of the function x (n)
completely describes the complete sequence, so also ‘N’ consecutive harmonic
 components of ‘Ck’ also can describe the complete Fourier Series (Complete
Spectrum.)

NUMERICALS
Example 1: -- Evaluate the Spectrum (Fourier Series Representation) of the signal
x (n) = cos (π/3) n

Solution: -- The Discrete Time cosine function is given in the form

x (n) = cos ω n where ω = π/3
ω=2πf
ω 1
∴f = =
2π 6
Comparing with the standard form of frequency of a Discrete Time Signal,
f = k/N
1
we see that f=
6
This represents a Discrete Cosine of periodicity N = 6
Using this in the Equation (5) we can evaluate the Fourier Coefficients in terms of
frequency “f” as
n =5
1
C k= ∑ x ( n ) e− j 2 π k n /6
6 n=0

This is to be evaluated for n = 0 , 1 , 2 , 3 , 4 , 5 (N-1) = 5

Or in terms of the Angular Frequency “ω” as

n=5
C k= 1 ∑ x ( n ) e− j ωn
6 n=0

Using these Fourier Coefficients, the Fourier Series of Discrete cosine function x (n) will
be
 k =5
x (n) = ∑ C k e j 2 π k n/ 6 …………(A)
k=0

This is to be evaluated within a period of “N = 6 samples” for sample numbers

from k = 0 , 1 , 2 , 3 , 4 , 5

Or
k =5

x (n) = cos ω n = ∑ C k e j ωn
k=0

A cosine function can also be represented as

x (n) = cos (2π n/6) = 1 e(

2
[{ j2π n
6 )}+ {e(− j26 π n )} ]

( { )}+ 1 {e(− j 26 π n )}
j2π n
x (n) = 1 e 6
……..(B)
2 2

{
∑ C k e j 2 π k n/ 6 = 12 e( )}+ 1 {e(− j 26 π n )}
n=5 j2π n
6

n=0 2

 Both the expressions (A) and (B) define the same function. In the expression (A), the
function is defined as a sequence of phasors of Amplitude Ck , each of which have a
phase angle 2 π k n/6. This phase angle is a function of the integer k.
 Equation (B) is the trigonometric representation of the same Cosine Function.
 Hence, equation (B) is also the Fourier Series of the signal x (n). Since a Fourier Series of
a Periodic signal (of periodicity “N”) is also periodic in the same periodicity as of the
given signal x (n), we can also define the sequence x (n) by using any 6 number of
consecutive samples.
 Using this logic, eq (B) indicates that the Fourier Series of x (n) which should consist of 6
consecutive terms, seem to have only two terms.
 In the first term of eq (B) we find that k = +1 , and in the second term k = - 1 . In these
terms C +1 = C -1 = 1/2. These two terms are associated with phasors at the frequency of f1
= + 1/6 and f2 = -1/6.
  This only means that the Fourier Series has been defined with 6 consecutive terms from
C -1 to C+4 , i.e. C-1 , C0 , C+1 , C+2 , C+3 , C+4 ,
 In this sequence the other terms C0 , C+2 , C+3 and C+4 have amplitude of 0.
 Now if we use the Angular Frequency, we have
ω1 = 2 π f1 = 2 π .(1/6) = + π/3
ω2 = 2 π f2 = 2 π .(-1/6) = - π/3

From the above, the graph of the Fourier Series (or in other word, The Spectrum of the signal
x (n) ) is a plot of the Fourier Coefficients as a function of the Angular Frequency as follows --

Example 2: -- Evaluate and sketch the spectrum of the Periodic signal x (n) whose one cycle is
given below

Solution: -- By the definition of the Fourier series we get

k=(N−1) k =3
x (n) = ∑ Ck e
j 2 π k n/ N
= ∑ C k e j 2 π k n/ 4
k=0 k=0

This is to be evaluated within a period of “N = 4 samples” for sample numbers

from k = 0 , 1 , 2 , 3
From the definition of Fourier Coefficient we get
n=(N−1)
1
C k=
N
∑ x ( n) e
−j2π k n/N

n=0

Since N = 4 we have
n=3 ¿ n=3 ¿
1
C k=
4
∑ ¿ x ( n ) e− j 2 π k n/ 4 = 14 ∑ ¿ x ( n ) e− j π k n/ 2
n=0 n=0
 n=3 ¿
1
C k=
4
∑ ¿ x ( n ) e− j π k n /2
n=0

This is to be evaluated using the sample values of the given function x (n).
for n = 0 , 1 , 2 , 3 ,
We have the sample values of the function given as
x (0) = 1 , x (1) = 1 , x (2) = 0 , x (3) = 0.
Substituting these in the expression for Ck we get

1
C k=
4
{ x (0) e
− j π k .0/2
+ x(1) e
− j π k .1/2
+0+ 0 }
1
C k = { 1+1 . e }
− j π k /2
4
This is to be evaluated for k = 0 , 1 , 2 , 3 . Substituting these values of k we get
C0 = 1/2 (at k = 0)
1
C1 = 4 { 1− j1 } ( Since at k = 1 , e− j π / 2 is the imaginary value = - j 1 )
C2 = 0 ( Since at k = 2 , e− j2 π /2 e− j π is the real value = -1 )
1
C3 = 4 { 1+ j 1 } ( Since at k = 3 , e− j3 π /2 is the imaginary value = + j 1 )
Evaluating the magnitudes of the Fourier Coefficients we get
√2 √2
| C0 |= 1/2 :: | C1 |= :: | C2 |=0 :: | C3 |=
4 4
These Fourier Coefficients are defined at the respective frequencies as
C0 is at f0 = 0 (ω0 =0 ) :: C0 is at f1 = 1/4 (ω1 = π/2) ::
C2 is at f2 = 2/4 (ω2 = π) :: C3 is at f3 = 3/4 (ω1 =3 π/2)

With the calculation as done above , we can sketch the spectrum as follows ---
Example 3 :-- Sampled Pulse Train ( Signal x (n) as shown in figure below )

Solution :--

 The Fourier coefficients are evaluated with the formula

n=(N−1)
1
C k=
N
∑ x ( n ) e− j 2 π k n / N
n=0

 The signal consists of a group of L samples of constant amplitude A (from n = 0 to

n = (L-1), followed by some samples of 0 amplitude , in a period of N.
 This will have a fundamental frequency f0 given by
L
f 0= (cy cles /sample)
N
 Hence the formula for Ck can be written as
n=(L−1)
A
C k=
N
∑ e− j 2 π k n/ N
n=0

Such expressions can be expressed in a general form as follows---

n=(M −1)

P= B ∑ (a)n ( Where “B” & “M” are constants ) …………. (A)

n=0
  These have as solution given by Geometric Sum Formula as

Expressing the formula for Fourier Coefficients in the General Form of (A), we have

n=(L−1)
A n
C k=
N
∑ {e− j 2 π k / N } ………….(B)
n=0

Where P = Ck , B = (A/N) , M = L & a = { e− j2 π k / N }

 Substituting these quantities and Solving with the Geometric Sum Formula we get
AL − j2 π k
C k= (Since, with k=0. e N = 1 )
N
− j2 π k
 When ‘k’ is in the range 0 < k < (L-1) we have e N
≠1

A {1−( e )}
− j 2 πk L
N

C =( )
{1−( e )}
k
N − j 2 πk
N

A {1−( e )}
− j 2 πkL
N
C =( )
{1−(e )}
k
N − j 2 πk
N

)) {e ( j 2NπkL )−e( − j 2NπkL )}

A (e
( − j 2 πkL
N
C =( )
)) {e( j 2Nπk )−e(− jN2 πk )}
k
N
( e(
− j 2 πk
N

[ ] ( N )
πkL
sin
( ( ))
− j 2 πk (L−1)

C k= ( )
A
N
e
N

sin ( )
πk
N

 An expression in the form of the last term of the equation above is called a “Sample Sinc
Function” denoted as
 [ ]
sin( πkLN ) = Sa L
sin ( )
πk
N
(N)

 Thus we get the Fourier Coefficients of the given function as

AL
C k= (when k=0)
N
&

( )(e( )) Sa
( NL )
− j 2 πk (L−1)
A N
When 0< k < (L-1) ….(C)
C k=
N

 Plotting of Ck w.r.t. Frequency gives us the Spectrum of the signal.

 The Shape of the “Sample Sinc Function” is similar to a Sinc Function. However the
expression (C) defines the Sample Sinc Function at discrete frequencies. These are
Harmonics of the Fundamental frequency f0 of the given signal .
 We had f0 defined as
L
f 0=
N
Thus the Spectrum of the signal will appear at these discrete values of the frequency. This is
shown in the figure below.