DSP First, 2/e

Lecture 7
Fourier Series Analysis
READING ASSIGNMENTS

 This Lecture:
 Fourier Series in Ch 3, Sect. 3-5
 Also, periodic signals, Sect. 3-4

 Other Reading:
 Appendix C: More details on Fourier Series

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 3

LECTURE OBJECTIVES
 Work with the Fourier Series Integral
T0
 j ( 2 k / T0 ) t
ak  T
1
0 
0
x (t )e dt
 ANALYSIS via Fourier Series
 For PERIODIC signals: x(t+T0) = x(t)
 Draw spectrum from the Fourier Series coeffs

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 4

HISTORY

 Jean Baptiste Joseph Fourier

 1807 thesis (memoir)
 On the Propagation of Heat in Solid Bodies
 Heat !
 Napoleonic era

 [Link]

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 5

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 6
SPECTRUM DIAGRAM
 Recall Complex Amplitude vs. Freq
1 * 10 1
X k ak
X j / 3  j / 3 2
2 k 7e 7e
 j / 2 j / 2
4e j k
4e
X k Ake
–250 –100 0 100 250
f (in Hz)

 
N
*  j 2 f k t
x (t )  Xa0   { aXkk e 1
2
j 2 f k t
 1
2
aXkk e
k 1
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 7
 Harmonic Signal->Periodic

x ( t )   ak e j 2 k F0 t Sums of Harmonic
complex exponentials
are Periodic signals
k 

PERIOD/FREQUENCY of COMPLEX EXPONENTIAL:

2 1
2 F0  0  or T0 
T0 F0
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 8
Notation for Fundamental
Frequency in Fourier Series
 The k-th frequency is fk = kF0
 
x (t )   ak e j 2 k F0 t   ak e j 2 f k t
k  k 

 Thus, f0 = 0 is DC
 This is why we use upper case F0 for the
Fundamental Frequency

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 9

Harmonic Signal (3 Freqs)
a1
a3 a5

T = 0.1

5
x ( t )   ak e j 2 (10 k ) t

k  5

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 11

Periodic signals->Harmonic?


x ( t )   ak e j 2 k F0 t Can all periodic
signals be written as
harmonic signals?
k 

 Fourier’s contribution was to postulate the answer is yes

 Called Fourier Series
 For heat transfer it is easy to solve PDE for sinusoidal
sources, but difficult for general sources
 Made formal by Dirichlet and Riemann

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 12

STRATEGY: x(t)  ak
 ANALYSIS
 Get representation from the signal
 Works for PERIODIC Signals
 Measure similarity between signal & harmonic
 Fourier Series
 Answer is: an INTEGRAL over one period
T0
 j0 k t
ak  1
T0  0
x (t )e dt
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 13
CALCULUS for complex exp
d t t d j t t
e  e  e  j e
dt dt
b b
1 1
e dt  e
t

t


e b
 e a 
a a

b b
1 j t 1 j b

j t
e dt  e  e  e 
j a

a
j a
j
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 14
INTEGRAL Property of exp(j)
 INTEGRATE over ONE PERIOD
T0 T0
 j ( 2 / T0 ) mt T0  j ( 2 / T0 ) mt
e
0
dt 
 j 2 m
e
0

T0
 ( e  j 2 m  1)
 j 2 m
T0
2
e
 j ( 2 / T0 ) mt
dt 0 m 0 0 
0
T0
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 15
ORTHOGONALITY of exp(j)

1
T0 
0 k 
j ( 2 / T0 ) t  j ( 2 / T0 ) kt

T0 0
e e dt 
1 k 
j ( 2 / T )(  k ) t 
e 0
m   k
T0
1 j ( 2 / T0 )(  k ) t

T0 0
e dt

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 16

Fourier Series Integral
 Use orthogonality to determine ak from x(t)
T0 Fundamental Freq.
 j ( 2 / T0 ) k t
ak  1
T0 x(t )e
0
dt F0 1 / T0

*
a k ak when x (t ) is real
T0

a0  T10 x (t )dt (DC component)

0
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 19
Isolate One FS Coefficient

x (t )   ak e j ( 2 / T0 ) k t
k 
T0 T0 
 j ( 2 / T0 ) k t   j ( 2 / T0 ) t
1
T0  x (t )e  j ( 2 / T0 ) t
dt  T0   ak e
1
e dt
0 0  k  
T0 T0
  
T0 x ( t ) e
1  j ( 2 / T0 ) t
dt   ak  T0 e
1 j ( 2 / T0 ) k t  j ( 2 / T0 ) t
e dt  a
 
0 k   0 
T0
Integral is zero
except for k 
0 
 a  T1 x (t )e  j ( 2 / T0 ) t dt
0

Aug 2016
 is dummy variable, could be k
© 2003-2016, JH McClellan & RW Schafer 20
Fourier Series: x(t)  ak
 ANALYSIS
 Given a PERIODIC Signal
 Fourier Series coefficients are obtained via
an INTEGRAL over one period
T0
 j0 k t
ak  1
T0  0
x (t )e dt
 Next, consider a specific signal, the FWRS
 Full Wave Rectified Sine

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 21

Full-Wave Rectified Sine
x (t )  sin(2 t / T1 ) Period is T0  12 T1

Absolute value flips the

negative lobes of a sine wave

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 22

Full-Wave Rectified Sine {ak}
T0
1
ak  x (t )e  j ( 2 / T0 ) kt dt
T0 0 Full-Wave Rectified Sine
T0 x (t )  sin( 2 t / T1 )
ak  T10 sin( T0 t )e  j ( 2 / T0 ) kt dt Period : T0  12 T1
0
 x (t )  sin( t / T0 )
T0 j ( / T0 ) t  j ( / T0 ) t
e  e
 1
T0  e  j ( 2 / T0 ) kt dt
0
2j
T0 T0

 j 21T0 e  j ( / T0 )( 2 k  1) t dt  1
j 2 T0 e  j ( / T0 )( 2 k 1) t
dt
0 0

 j ( / T0 )( 2 k  1) t T0  j ( / T0 )( 2 k 1) t T0
e e
 j 2T (  j ( / T )( 2 k  1))  j 2 T0 (  j ( / T0 )( 2 k 1))
0 0
Aug 2016 0© 2003-2016, JH McClellan & RW Schafer 0 23
Full-Wave Rectified Sine {ak}
 j ( / T0 )( 2 k  1) t T0  j ( / T0 )( 2 k 1) t T0
e e
ak  j 2T (  j ( / T )( 2 k  1))  j 2T0 (  j ( / T0 )( 2 k 1))
0 0
0 0


 2 ( 21k  1) e  j ( / T0 )( 2 k  1)T0  1   1
2 ( 2 k 1)
e  j ( / T0 )( 2 k 1)T0
1

  ( 21k  1) e  j ( 2 k  1)  1 1
 ( 2 k 1) e  j ( 2 k 1)
 1

 
2 k 1 ( 2 k  1)
 ( 4 k 2  1)
 2k2
 ( 1)  1 (4k 2  1)
Aug 2016 © 2003-2016, JH McClellan & RW Schafer 24
Fourier Coefficients: ak

 ak is a function of k
 Complex Amplitude for k-th Harmonic
 2
ak 
 ( 4k  1)
2

 Does not depend on the period, T0

 DC value is a0 2 /  0.6336

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 25

Spectrum from Fourier Series
Plot ak for Full-Wave Rectified Sinusoid

 2 F0 1 / T0 and 0 2 F0
ak 
 (4k  1)
2

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 26

Reconstruct From Finite Number of
Harmonic Components

Full-Wave Rectified Sinusoid x (t )  sin( t / T0 )

T0 10 ms  2
ak 
 F0 100 Hz  (4k 2  1)
a0 2 /  0.6336

N
x N (t ) a0   a k e j 2 k F0 t
a e 
k
 j 2 k F0t

k 1

How close is x N (t ) to x (t )  sin( t / T0 ) ?

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 27

Reconstruct From Finite Number of
Spectrum Components

Full-Wave Rectified Sinusoid x (t )  sin( t / T0 )

T0 10 ms
 F0 100 Hz

a0 2 /  0.6336

 2
ak 

Aug 2016
( 4 k 2
 1) © 2003-2016, JH McClellan & RW Schafer 28
 Synthesis: up to 7th Harmonic
1 2 ) 2 2 2
y (t )   cos(50 t  2
sin(150 t )  sin( 250 t )  sin(350 t )
2  3 5 7

Aug 2016 © 2003-2016, JH McClellan & RW Schafer 34