www.jntuworld.

com

Fourier Series &

Fourier Transform

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

JBJ Fourier (1768- 1830)

A product of the French Revolution,
(Jean Baptiste) Joseph Fourier held positions
within the Napoleonic Empire during his
early career. After Napoleons fall from
power, Fourier devoted his talents
exclusively to science. Although he won
the Institut de France prize in 1811 for
his work on heat diffusion, criticism of its
mathematical rigor and generality led him to
publish the classic book Thorie analytique de
la chaleur in 1823. Within this book he
introduced the world to the series that
bears his name.

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Trigonometric Fourier Series

Orthogonal Basis Set

cos n0t, sin n0t, 1;

n= 1,2,

Orthogonality Check
T

T
0 cos m0t cos n0t dt 2 for m n
0 for m n

cos m t sin n t dt 0 for all m,n

0

T
0 sin m0t sin n0t dt 2 for m n

= 0 for m n
1 is orthogonal to cos n0t and sin n0t

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

for all n

www.jntuworld.com

Trigonometric Series of Periodic Signals

Trigonometric Fourier series can exactly represent a continuous
waveform over an interval (t1,t2), with fundamental period T equal
to the specified interval. However, as the basis functions are
periodic with their periods having integer ratios, their sum is
periodic with period T. Thus trigonometric series are naturally
suited for representing periodic signals.

f (t ) a0 an cos n0t bn sin n0t

n 1

where f (t ) periodic with T

1
a0
T

2
an
T

f (t )dt average value of

f (t )

2
f (t ) cos n0t dt ; bn
T

f (t ) sin n t dt
0

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Trigonometric Series: Examples

square wave
1
-

-1
full wave rectified sine wave
1
-T

triangular wave
-/2

2T

3/2

sawtooth wave
A
-1

4
4
cos 0t
cos 20t
3
15
4

cos 30t .....

35

-1

4
4
sin 3t
sin 5t .....
3
5

3T

/2

sin t

1
0

sin t

sin 3t

8
sin 5t
2
25

9 2
8

sin 7t .....
49 2

A A
A
sin 2 t
sin 4 t
2
2
A

sin 6 t .....
3

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Trigonometric Fourier Series: Useful Observations

f(t) with even symmetry does not have sine components.
f(t) with odd symmetry has no cosine components and no constant
term.
f(t) with half-wave symmetry does not have even harmonics.
Constant (dc) term hides odd symmetry.
A shift in origin of f(t) changes coefficients of both sine and cosine
terms.
Enables view of a periodic signal in terms of its harmonic content,
which becomes equivalent description of the signal in frequency
domain. Frequency domain plots (of ans and bns) or spectra would be
discrete, with values specified only at multiples of fundamental
frequency.
Amplitude scaling of f(t) causes scaling of all components by same
factor.
Compression in time causes expansion in frequency, and vice versa.
Time shift of f(t) requires each harmonic to be phase-shifted by a
value proportional to its frequency i.e., time-delay of t0 in f(t) causes
phase of nth harmonic to lag by n0t0.

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Periodic Waveshape as Sum of Harmonics: Example

f1 (t ) sin(0t )

f 2 (t ) sin(20t )

f3 (t ) f1 (t ) f 2 (t ) sin(0t ) sin(20t )

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Harmonic Content of Audio Waveforms (1 KHz)

Factory Whistle

Soprano

Piano
Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Complex Exponential Fourier Series

Orthogonal Basis Set

e jn0t :
n 0, 1, 2,...
Orthogonality Check
T

jm0t

jn0t

T for m n
dt
0 for m n

Obtaining Exponential Series

f (t )

Fe

jn0t

1
, where Fn f (t )e jn0t dt
T 0

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

www.jntuworld.com

Complex Exponential Series: Examples

square wave
1

2 j 5t
2
2
e j e j 3t j e jt
5
3

2
2
2
j e jt j e j 3t j e j 5t ...

3
5

... j

2
-1
full wave rectified sine wave
1
-

t
-T

triangular wave
-/2

2T

3T

1
0

/2

3/2

sawtooth wave
A
0

4
4 j 3t
4 jt
j 5t
e

j
e
25 2
9 2
2
4
4
4
j 3t
j 2 e jt
e

e j 5t ...
2
2

9
25

...

-1

-1

2 j 30t
2 j 20t 2 j0t 2
e

35
15
3

2 j0t
2 j 20t
2 j 30t

e
e

e
...
3
15
35

...

jA j 4 t
jA j 2 t A
...
e

4
2
2
jA j 2 t
jA j 4 t

e
...
2
4

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

10

www.jntuworld.com

Some Notes on Complex Exponential Fourier Series

Negative frequency is only a mathematical concept and has no
physical meaning.

For periodic f(t), average power Fn

( Parseval ' s relation)

For real valued signal f(t),

Fn F n (conjugate symmetry)
Fn F n (magnitude spectrum even)
and Fn F n (phase spectrum odd)

Relation with Trigonometric Series

an
bn
an
bn
Fn j ; F n j
2
2
2
2
an Fn F n ; bn j Fn F n
Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

11

www.jntuworld.com

Extending Exponential Series of Periodic Signal fp (t)

to obtain Fourier Transform of Aperiodic Signal f(t)
Let f ( t ) lim f p ( t )

f p (t )

1
Fn
T

f p ( t )e

jn0t

dt

2
T

F ( n0 ) FnT

f p (t )e jn0t dt

2
0
T
n0 n , f p (t ) f (t )
as T , 0

F ( )

f (t )e

Fn e jn0t ; using Fn

jt

dt

(Analysis Equation)
F () is the spectral density function of f (t )

F (n0 )
T

F (n0 ) jn0t
e
T
n
n

=
n

= F (n0 )e jn0t
n

1
=
2

0
2

F (n )e jnt

as T , 0

2
0
T

n0 n , f p (t ) f (t ),
1
f (t )
2

F ( )e jt d

(Synthesis Equation)

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

12

www.jntuworld.com

Dirichlets Conditions
Sufficient conditions to be satisfied by a periodic signal for existence
and convergence of fourier series:
absolute integrable over one period (weak condition)
no more than finite number of maxima and minima in a period
(strong condition)
no more than finite number of discontinuities in a period, and each
discontinuity must be finite (strong condition)
Sufficient conditions to be satisfied by a signal for existence of fourier
transform:
absolute integrable (weak condition)
no more than finite number of maxima and minima in any finite
interval (strong condition)
no more than finite number of discontinuities in any finite interval,
and each discontinuity must be finite (strong condition)

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

13

www.jntuworld.com

Fourier Transforms of Some Functions

e

at

te

F
F
1
2a
a t
u (t )
a 0; e
2
a 0; (t ) 1
2
a j
a

at

u (t )

1
2
a j

a 0; f (t )
0

A Sa

2
2

F
1
2
A 2 A ; u (t )
; sgn(t )
j
j

j0t

2 0 ; cos 0t 0 0
F

sin 0t j 0 0
F

Periodic function f p t 2

n0

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

14

www.jntuworld.com

Properties of Fourier Transforms (1/2)

F

1. Linearity: a1 f1 ( t ) a2 f 2 ( t ) a1 F1 () a2 F2 ( )
F

2. Time shift: f ( t t0 ) F () e jt0

3. Frequency shift: f ( t) e

j0t

F ( 0 )
F

4. Time and frequency scaling: f ( at)

1

F
a
a

5. Time and frequency reversal: f ( t ) F ( )

d n f (t ) F
n
6. Time differentiation:

F ( )

n
dt
F d n F ( )
n
7. Frequency differentiation: ( jt ) f ( t )
d n
F

8. Time convolution: f1 ( t ) f 2 ( t ) F1 ( ) F2 ( )
F

9. Frequency convolution: f1 ( t ) f 2 ( t )

1
F1 ( ) F2 ( )
2

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

15

www.jntuworld.com

Properties of Fourier Transforms (2/2)

F

10. Conjugation: f ( t ) F ( )

conjugate symmetry

for real f ( t ), F () = F ( )

for real even f ( t), F ( ) is real; for real o dd f (t ), F ( ) is imaginary

t

11. Time integration:

f ( )d

F ( )
F (0) ( )
j

F
1
12. Frequency integration:
f ( t ) f (0) ( t ) F ( ) d
jt

F ( t ) F
13. Duality/symmetry: F (t ) 2 f ( );
f ( )
2

1
14. Area relations: F (0) f ( t ) dt; f (0)
F ( ) d

15. Parseval energy theorem:

1
f ( t ) dt
2
2

Generalized Parseval's relation:

F ( ) d
2

1
f ( t ) g ( t ) dt
2

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

F ( ) G ( ) d

16

www.jntuworld.com

Properties of Fourier Transforms: Hints for Proofs

1.
2.
3.
4.
5.
6.
7.
8.

Direct consequence of linearity of integral operations

Change of variable from t to (t- t0) in synthesis equation
Change of variable from to (- 0) in analysis equation
Change of variable in fourier transform of f(at)
Substitute a = -1 in property 4
Differentiate synthesis equation with respect to t
Differentiate analysis equation with respect to
Use convolution integral, replace either function from synthesis equation,
reverse order of integration. Then use analysis equation twice
9. Proceed in a similar manner as for property 8
10. Conjugate analysis equation, reverse sign of for main property. Apply for
real signals f(t) = f *(t), for real even f(t) = f(-t) and for real odd f(t) = -f(-t)
11. Note that LHS = f(t)*u(t). Use property 8 and FT of u(t) to obtain RHS
12. Obtain inverse transform of u() using property 13. Then proceed in similar
manner to that of property 11.
13. In synthesis equation change variable to and reverse its sign. Similar
operations in analysis equation for second part
14. Obtain f(0) from synthesis and F(0) from analysis equations
15. f (t ) 2 f (t ) f (t ), for f ( t ) use conjugate of synthesi s equation, then

reverse order of integration.

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

17

www.jntuworld.com

Some More Fourier Transform Pairs

f(t)

F()

|t|

at

0
2
a j 02

sin 0t u(t )
t

t
t

Sa

w
wt
Sa
2
2

t n 1
e at u(t )
n 1!
t

a j

2 e 2
t

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

18

www.jntuworld.com

Some Interesting Correspondences between

Operations on Fourier Transform Pairs
One Domain

Other Domain

Convolution

Multiplication

Compression

Expansion

Periodicity
(Repetition)

Discreteness
(Sampling)

Even-odd split

Real-imaginary split

Shift

Multiply by complex exponential

Differentiation

Multiply by domain variable and j

Integration

Divide by domain variable and j, + impulse at origin

Conjugation

Conjugation with sign reversed domain variable

Analytical form

Single-sidedness (causal/anti-causal)

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

19

www.jntuworld.com

Energy Spectral density Sf()

Based on Parseval's energy theorem, we can perceive energy distribution in frequency
domain and accordingly define a useful quantity called Energy Spectral Density:
F ( )
S f ( )
2
Properties of Energy Spectral Density

1) S f ( ) 0 ; esd is real and non-negative

2) S f ( ) S f ( ); for a real signal, esd is even function of

3) Energy E f

( )d

Cross Energy Spectral density

From generalized Parseval's theorem, we define cross-energy spectral density of
two signals f1 (t ) and f 2 (t )
F1 ( ) F2 ( )
S12 ( )
; (generally complex)
2
Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

20

www.jntuworld.com

Power Spectral density Pf()

Power signals in general do not have fourier transforms. However, they can be described in
frequency domain in terms of power spectral density defined as below:
1 FT ( )
Pf ( ) lim
, where FT ( ) is the FT of power signal f (t ) trucated to interval T
T T
2
2

1 FT ( )
is the mean energy spectral density of the truncated power signal
note :
T
2

Properties of power spectral density:

2

1. Pf ( ) real and 0

2. Pf () Pf ( ) for real f ( t)

3. Average Power of f (t ) =

P ( )d
f

Cross Power Spectral density

We define cross-power spectral density of two signals f1 (t ) and f 2 (t ) as
1 F1 ( ) F2 ( )
P12 ( ) lim
; (generally complex)
T T
2
Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

21

www.jntuworld.com

Concept of Signal Bandwidth

Frequency band over which a signal has significant spectral density can be
observed from its magnitude spectrum. If maximum spectral density is at zero
frequency, signal is said to be lowpass; and if at a non-zero frequency it is
bandpass.
Half-power (or 3 dB) bandwidth
for a low pass signal is |0-c| where F (c ) F (0) / 2
for a bandpass filter with maximum gain at c, it is |1 2| where
1 <c < 2 and F (1 ) F (2 ) F (c ) / 2

Pulse width Bandwidth Relation

Pulse width may be defined as the interval between the two half power
instants on either side of the peak (similar in concept to half power
bandwidth). Due to a property of Fourier Transform, an inverse relation
exists between pulse width and bandwidth of a signal. That is, narrow pulses
have large bandwidths and vice versa. We also state this relation
alternatively as pulse width x bandwidth = constant.
Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

22

www.jntuworld.com

Introduction to Hilbert Transform

1
1

H [ f (t )] f (t ) f (t )

f ( )
d
t

f (t ) f (t ); Hilbert transform does not change domain.

1 F
( ) j sgn( ) F ( )
j sgn( ), F
t
Positive frequencies are phase shifted by -900 and negative
frequencies by +900 . Magnitude spectrum is unchanged.
f (t ) and f (t ) have same energy and are mutually orthogonal.
(For proof of orthogonality, use generalised Parseval's relation)

Hilbert transform applied twice: H f (t ) = f (t ) f (t )

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

23

www.jntuworld.com

Some Hilbert Transform Pairs

f t

cos t
sin t
(t )
1

Sa(t )
1

sin t
t

t a
t a
constant

jt

a
;a 0
2
2
a

f (t )
sin t
cos t
1

t
(t )

1 cos t
t
1 ta
ln
t a
zero

j sgn( ) e jt

t
a2 2

Presented by APN Rao, Dept ECE, GRIET, Hyderabad. Aug 2011

24