Discrete - Time Signals and Systems

Fourier Series
Analysis and Synthesis 1

Yogananda Isukapalli
 General definition of orthogonal signals

A set {fn (t )}, n = 0,±1,±2,..... ± M ,..... , is orthogonal ove

some interval [a, b] if;
b
ln m = n
òa fm (t )f (t )dt = 0 m ¹ n
*
n

‘*’ denotes conjugate for complex functions

In other words, each signal in the set is orthogonal
to every other signal in the set
Example
Family of Harmonically related complex
exponentials
fn (t ) = {e jnw t }
0
n = 0,±1,±2,..... ± M ,....
2p
w0 = ,
T0
T0 ...is the fundamental period
fn (t ) are periodic signals,
fn (t + T0 ) = fn (t )
Which means,
fn (t + T0 ) = e jnw 0 ( t +T0 )

( jnw0 t + jnw0T0 )
=e
= e jnw0t e jnw0T0
2p
jn T0
jnw0 t jnw0 t jn 2p
=e e T0
=e e
e jn 2p = 1, for all n , ' n' is an integer
\fn (t + T0 ) = e jnw0t = fn (t )
Continue example….
The set {fn (t )}is orthogonal over any period
(t1 , t1 + T0 )
Proof:
t1 +T0 t1 +T0

ò n (t )dt = ò e dt
jnw0t - jmw0t
f (t )fm
*
e
t1 t1
t1 +T0

ò e dt
jnw0t - jmw0t
= e
t1
t1 +T0

ò
j ( n - m )w0 t
= e dt
t1
 t1 +T0
2p
1 æ j ( n-m) t ö
= çe T0 ÷
j (n - m) çè ÷
ø t1
2p 2p
1 æ j ( n - m ) ( t + T ) j ( n - m ) t1 ö
çe ÷
1 0
= T0
-e T 0

j (n - m) çè ÷
ø
1
2p 2p 2p
1 æ j ( n - m ) t j ( n - m ) T j ( n - m ) t1 ö
çe ÷
1 0
= T0
e T0
-e T0

j (n - m) çè ÷
ø
2p 2p
1 æ j ( n - m ) t j ( n - m ) t1 ö
çe ÷
1
= T0
-e T 0

j (n - m) çè ÷
ø
= 0, if m ¹ n
 for m=n
t1 +T0 t1 +T0

ò n ò
jnw0 t - jnw0t
f (t )fm
*
(t )dt = e e dt
t1 t1
t1 +T0

ò e dt = (t )
t1 +T0
= 0
t1
t1

= t1 + T0 - t1 = T0 = ln
Thus we have proved for the family of harmonically related
f
complex exponentials n (t ) = {e jnw0t
} n = 0,±1,±2,..... ± M ,....
b
[ln = T0 ] m=n

òf (t )f n (t )dt =
*
m
a
0 m¹n
Fig. 4.1
The Fourier series and integral is a most beautiful and
fruitful development, which is central to the areas of
communications, signal processing and antennas. Taken
by the beauty of Fourier series , Maxwell called it a
great ‘mathematical poem’
It is Fourier’s investigation into the propagation of heat
in solid bodies that led to the powerful insight called
Fourier series and Fourier integral
The result, any signal can be expressed as a sum of
sinusoids, was announced in a paper on the theory of heat.
However, due to the lack of mathematical rigor and
generality, the paper was not published, the impact was
felt decades later
 Fourier Series: Definition
Approximation of periodic signals [ x(t ± T0 ) = x(t ) ],
by harmonically related periodic exponential
functions, {e jnw t } leads to Fourier series theory
0

¥
x(t ) = å k
a e
k = -¥
jw0 kt Eq.1

2p
Where, w0 =
T0
The frequency of k th complex exponential ,
f k = k T0 or wk = kw0 , All the frequencies are integer
multiples of fundamental frequency, f 0 = 1 T0 Hz
Calculation of Fourier coefficients
¥
x(t ) = å k
a e
k = -¥
jw0 kt

T0
- j ( 2p T0 )lt
ò
0
x (t ) e dt

T0
æ ¥ ö
(
= ò ç å ak e jw0 kt ÷. e - j (2p T0 )lt dt )
0 è k = -¥ ø
Integration and summation are inter changeble,
¥ T0

= å
k = -¥
ak ò e jw0 ( k -l ) t dt
0
 From orthogonality property,
T0
T0 k = l
òe
jw 0 ( k - l ) t
dt =
0
0 k ¹l
¥ T0

\ å ak ò e jw0 ( k -l ) t dt = alT0
k = -¥ 0

From above proof , Fourier analysis equation,

T0 æ 2p ö
- j çç ÷÷ kt Eq.1
1
ak =
T0 ò x(t )e
0
è T0 ø
dt

And the Fourier synthesis equation,

æ 2p ö
¥ j çç ÷÷ kt Eq.2
x(t ) = åa e
k = -¥
k
è T0 ø

Eq.1 and 2 play the fundamental role of signal analysis and synthesis
 General example
x(t), periodic signal, with period T0 x(t)

- 5T0 2 - 2T0 - 3T0 2 - T0 - T0 2 0 T0 2 T0 Time ‘t’

Fig. 4.2

ak : Fourier coefficients, contains spectral information, often complex

numbers, so absolute value is shown on the plot
ak
Spectrum

Fig. 4.3 ……….. - 3w0 - w0 0 w0 3w0 ….. Frequency, w = kw

0
 Example 1
x(t ) = cos w0t + sin 2 w0t
= cos w0t + 1 2(1 - cos 2w0t )
= 1 2 + cos w0t - 1 2 cos 2w0t
e jw 0 t + e - jw 0 t
using the Euler identity , cos w0t = ,
2
e jw 0 t + e - jw 0 t 1 æ e j 2 w 0 t + e - j 2 w 0 t ö
x(t ) = 1 2 + - çç ÷÷
2 2 2
è ø
1 e jw 0 t e - jw 0 t e j 2 w 0 t e - j 2 w 0 t
= + + - -
2 2 2 4 4
 ¥
x(t ) = å k
a e
k = -¥
jw0 kt

= ..... + a- 2 e - j 2w0t + a-1e - jw0t + a0 + a1e jw0t + a2 e j 2w0t + .....

Through comparison,
1
a-1 = a1 =
2
1
a- 2 = a2 = -
4
1
a0 =
2
ak = 0, k > 2
In cases as above, where one can reduce the function into complex
exponentials, there is no need to integrate to get the Fourier
coefficients
 Example 2
x(t ) = 8 cos 3w0t
e jw 0 t + e - jw 0 t
using the Euler identity , cos w0t = ,
2
æ e j 3w0t + e - j 3w0t ö
x(t ) = 8çç ÷÷
è 2 ø
= 4e j 3w0t + 4e - j 3w0t , comparing again with definition,
¥
x(t ) = å k
a e
k = -¥
jw0 kt

= ..... + a- 2 e - j 2w0t + a-1e - jw0t + a0 + a1e jw0t + a2 e j 2w0t + .....

a-3 = a3 = 4
ak = 0, otherwise
 Example 3: Periodic train of rectangular
pulse of width(t ) and period(T0)

A A A

3T0 - T0 - T0 2 - t 2 0 t 2 T0 2 T0 3T0 Time, t

Fig. 4.4 -
2 2

Calculation of Fourier coefficients, ak

T0 æ 2p ö
- j çç ÷÷ kt
1
ak =
T0 ò x(t )e
0
è T0 ø
dt
The limits of the period {- T0 2 , T0 2}, with amplitude ' A' ,
T0 2
A
ò
- jw0 kt
= x (t ) e dt
T0 -T0 2
The function is non - zero in the width 't ' ,
t 2
A
ò
- jw0 kt
= e dt
T0 -t 2
æA 1 öæ - jw0 kt t 2 ö
= çç . ÷÷ç e ÷
è T0 - jw0 k øè ø
-t 2

=-
A
jT0w0 k
(
e - jw 0 k t 2 - e jw 0 k t 2 )
A
=- (- 2 j sin(w0 k t 2) )
jT0w0 k
 2A æ æ 2p ö ö 2p
= sin çç çç ÷÷k t 2 ÷
÷ ! w0 =
æ 2p
ö è è T0 ø ø T0
T0 çç ÷÷k
è T0ø
A æ pkt ö
ak = sin çç ÷÷, k ¹0
pk è T0 ø
For k = 0,
T0 æ 2p ö
- j çç ÷÷ 0 t
1
a0 =
T0 ò x(t )e
0
è T0 ø
dt

t 2
A A
a0 = ò 1.dt = [t 2 - (-t 2)]
T0 -t 2
T0
At
a0 = , 'a 0 ' is also known as ' DC ' term
T0
 Example 4: The Square wave
A square wave is defined as,
ì T0 The period of square wave T0
ï 1 for 0 £ t £ 2
x(t ) = í
T
ï0 for 0 £ t £ T0
î 2

- 2T0 - 3T0 - T0 - T0 2 0 T0 2 T0 3T0 2T0 Time, t

2 2
Fig. 4.5

Duty cycle 50%

 T0 æ 2p ö
- j çç ÷÷ kt
1
ak =
T0 ò x(t )e
0
è T0 ø
dt

The limits of the period are {0, T0 }, with amplitude '1' ,

T0 æ 2p ö
- j çç ÷÷ kt
1
= òe è T0 ø
dt , The function is non - zero in {0, T0 2},
T0 0

T0 2 æ 2p ö
- j çç ÷÷ kt
1
=
T0 ò
0
e è T0 ø
dt

æ ö
ç ÷æ æ 2p ö T0 2
ö
ç1 1 ÷ç - j ççè T0 ÷÷ø kt ÷
=ç . ÷çe ÷
ç T0 - æ 2p ö ÷ç ÷
ç j çç ÷÷k è
÷ 0 ø
è T
è 0 ø ø
 æ - 1 öç æ æ 2p ö T0
- j çç ÷÷ k - j çç ÷÷ k 0 ö
æ 2p ö

-e è 0 ø ÷
T 2 T
= çç ÷÷ e è 0 ø
è j 2pk øçè ÷
ø
æ - 1 ö - jpk
= çç ÷÷ e -1( )
è j 2pk ø
!e - jpk
(
= e - jp
)
k
and e - jp = -1 e - jpk = (- 1) ,
k

1 - (- 1)
k
ak = , for k ¹ 0
j 2pk
T0 2 æ 2p ö
- j çç ÷÷ 0 t
1
For k = 0, a0 =
T0 ò
0
e è T0 ø
dt

T0 2
1 1 1
a0 =
T0 ò0 1.dt = T0 [T0 2] = 2
1
a0 = , a 0 is also known as ' DC ' term
2
 The final answer for the Fourier series coefficients of the square
wave can be summarized in three different cases,
ì 1
ï jpk for k = ±1, ± 3, ± 5,....
ï
ak = í0 for k = ±2 , ± 4 , ± 6 ,....
ï1
ï2 for k = 0
î Fig. 4.6

Spectrum of a square wave from Fourier series coefficients, fundamental

frequency of 25Hz, frequency range [-225 - 225] is shown, k=-9 to +9
 Synthesis of Square wave
Given the Fourier series coefficients, using the synthesis
Equation the signal can be reconstructed
æ 2p ö
¥ j çç ÷÷ kt
x(t ) = åa e
k = -¥
k
è T0 ø

The quality of signal synthesized improves with the no. of

Fourier coefficients used
The summation of terms in the above equation can
sometimes result in simple cosine or sine functions
Synthesis using the 1st three harmonics
æ 2p ö
¥ j çç ÷÷ kt
x(t ) = åa e
k = -¥
k
è T0 ø
, using the first 3 harmonics,
æ 2p ö
3 j çç ÷÷ kt
x(t ) = åa e
k = -3
k
è T0 ø

= a-3e - j 3w0t + a- 2 e - j 2w0t + a-1e - jw0t + a0 + a1e jw0t + a2 e j 2w0t + a3e j 3w0t
-1 1
! a-3 = , a3 =
j 3p j 3p
a- 2 = a2 = 0
-1 1
a-1 = , a1 =
jp jp
1
a0 =
2
x3 (t ) = a-3e - j 3w0t + a- 2 e - j 2w0t + a-1e - jw0t + a0 + a1e jw0t + a2 e j 2w0t + a3e j 3w0t
= a0 + a-1e - jw0t + a1e jw0t + a-3e - j 3w0t + a3e j 3w0t
1 - 1 - jw0t 1 jw0t - 1 - j 3w0t 1 j 3w0t
= + e + e + e + e
2 jp jp j 3p j 3p
1 j - jw0t - j jw0t j - j 3w0t - j j 3w0t
= + e + e + e + e
2 p p 3p 3p
! - j = e - jp 2 , and j = e jp 2

2 p
(
1 1 j p 2 - jw 0 t - j p 2 jw 0 t
= + e e +e e +
3p
)
e e (
1 jp 2 - j 3w0t - jp 2 j 3w0t
+e e )
1 2 2
= + cos(w0t - p 2) + cos(3w0t - p 2), or
2 p 3p
1 2 2
x3 (t ) = + sin(w0t ) + sin(3w0t )
2 p 3p
 Synthesis, fundamental frequency of 25 Hz , 3 harmonics
1 2 2
x3 (t ) = + cos(2p (25)t - 2 ) +
p
cos(2p (75)t - p2 )
2 p 3p
Fourier series coefficients Fig. 4.7
For 1st three harmonics

x(t), with a fundamental frequency of 25 Hz

x3(t), with a
fundamental
frequency
of 25 Hz

Fig. 4.8
 1 2 2 2 2
x7 (t ) = + cos(50p t - p2 ) + sin(150p t ) + sin(250p t ) + sin(350p t )
2 p 3p 5p 7p
Fourier series coefficients for 1st seven harmonics
Fig. 4.9

Synthesized

x7 (t )

x(t )
original Fig. 4.10
General formula for synthesized signal with ' N ' harmonics
1 2 2
x N (t ) = + sin(w0t ) + sin(3w0t ) + !
2 p 3p
1st three

1st seven

1st seventeen

Notice the Gibbs phenomenon at discontinuities Fig. 4.11

 Reference
James H. McClellan, Ronald W. Schafer
and Mark A. Yoder, “Signal Processing
First”, Prentice Hall, 2003