Signals and Systems

Lecture 6 Fourier Series Representation of

Periodic Signals

Asst. Prof. Dr. Erturul SAATI

Figures and images used in these lecture notes are adopted from
Signals & Systems by Alan V. Oppenheim and Alan S. Willsky, 1997

A Historical Perspective
L. Eulers study on the motion of a
vibrating string in 1748
L Euler
L. E l showed
h d (in
(i 1748)
The configuration of a vibrating
string at some point in time is a
linear combination of these normal
modes
D. Bernoulli argued (in 1753)
All physical motions of a string could
be represented by linear
combinations of normal modes
But, he did not pursue this
mathematically
th ti ll
J.L. Lagrange strongly criticized (in 1759)
The use of trigonometric series in
examination of vibrating strings
Impossible to represent signals with
corners using trigonometric series

1
A Historical Perspective
In 1807, Jean Baptiste Joseph Fourier
Submitted a paper of using trigonometric
series to represent any periodic signal
It is examined by S.F. Lacroix, G. Monge,
P.S. de Laplace and J.L. Lagrange
But Lagrange rejected it!
In 1822, Fourier published a book Theorie
analytique de la chaleur
The Analytical Theory of Heat
Ship encountering the superposition of three wave trains,
trains each with a
different spatial period.
When these waves reinforce one another, a very large wave can
result.

A Historical Perspective
Fouriers main contributions
Studied vibration, heat diffusion, etc.

Found series of harmonically related sinusoids to be useful in

representing the temperature distribution through a body

Claimed that any periodic signal could be represented by such a series

(i.e., Fourier series discussed in Chap 3)

Obtained a representation for aperiodic signals

(
(i.e.,
, Fourier integral
g or transform discussed in Chap
p 4 & 5))

(Fourier did not actually contribute to the mathematical theory of Fourier

series)

2
A Historical Perspective
Impact of Fouriers work
Theory of integration, point-set topology, eigenfunction expansions, etc.

Motion of planets, periodic behavior of the earths climate, wave in the

ocean, radio & television stations

Harmonic time series in the 18th & 19th centuries

Gauss etc. on discrete-time signals and systems

Faster Fourier transform (FFT) in the mid-1960s

Cooley & Tukey in 1965 discovered independently
Can be found in Gausss notebooks

Fourier Series Representation of Periodic

Signals
The objective is to represent complicated signals as linear combination of
basic functions, i.e.,

x(t ) a (t )
k
k k

so that if the response of the LTI system to k(t) is known, then the
response to x(t) can be expressed as the weighted sum of these
responses.
Try to select basic functions k(t) such that the response to k(t) is
k k(t)
k(t) are the eigenfunctions of LTI systems
k are the eigenvalues of LTI systems

Hence, the output is:

y (t ) a (t )
k
k k k

3
Fourier Series Representation of Periodic
Signals
Why complex exponential is used as basic function k?

The most basic periodic signal with a well-defined frequency.

Complex exponentials are eigenfunctions of LTI systems.

Response of an LTI system to a complex exponential is the
same complex exponential with only change in amplitude.

CT: e st H ( s )e st eigenfunction

DT: z H ( z ) z
n n
eigenvalue

Response is scaled version of the input with H(s) or H(z).

H(s) and H(z) are complex amplitude factors as functions of
complex variables s and z.

The Response of LTI Systems to Complex

Exponentials
Consider a CT LTI system with impulse response h(t) and input x(t)=est:

h( ) x(t ) d h( )e
s ( t )
y (t ) d

h( )e
s
y (t ) e st d , the response to e st is of the form

y (t ) H ( s )e st H ( s ) x (t ), where,

h( )e
s
H (s) d is a complex constant

Hence, complex exponentials are eigenfunctions of CT LTI systems.

The constant H(s) for a specific value of s is the eigenvalue associated
with the eigenfunction est.

4
The Response of LTI Systems to Complex
Exponentials
Consider a DT LTI system with impulse response h[n] and input x[n]=zn:

y[n] h[k ]x[n k ] h[k ]z
k k
nk

y[n] z n h[k ]z
k
k
, the response to z n is of the form

y[n] H [ z ]z n H [ z ]x[n], where,

H [ z] h[k ]z
k
k
is a complex constant

Hence, complex exponentials are eigenfunctions of DT LTI systems.

The constant H[z] for a specific value of z is the eigenvalue associated
with the eigenfunction zn.

The Response of LTI Systems to Complex

Exponentials
Eigenfunctions and Superposition Properties
In order to show the usefulness of decomposing more general signals in
terms of eigenfunctions for the analysis of LTI systems:
Let
L t x(t)
(t) be
b a linear
li combination
bi ti off three
th complex
l exponentials,
ti l
x(t ) a1e s1t a2 e s2t a3e s3t
From the eigenfunction property, the response to each separately is

a1e s1t a1 H ( s1 )e s1t

a2 e s2t a2 H ( s2 )e s2t
a3e s3t a3 H ( s3 )e s3t
and from the superposition property the response to the sum is the
sum of the responses,
y (t ) a1 H ( s1 )e s1t a2 H ( s2 )e s2t a3 H ( s3 )e s3t

5
 The Response of LTI Systems to Complex
Exponentials
Eigenfunctions and Superposition Properties
More generally, for both CT and DT, if the input to an LTI system is
represented as a linear combination of complex exponentials, then
th output
the t t can also
l bbe represented
t d as a linear
li combination
bi ti off the
th
same complex exponential signals.
CT: x (t ) ak e sk t then the output will be y (t ) ak H ( sk )e sk t
k k

DT: x[n] ak zkn then the output will be y[n] ak H ( zk ) zkn

k k
In this representation each coefficient of the output is obtained as
the product of the corresponding coefficient ak of the input and the
systems eigenvalue H(sk) or H(zk) associated with the eigenfunctions
eskt or zkn, respectively.
respectively
In CT we focus on purely imaginary values of s (s=j), and thus, we
consider only complex exponentials of the form ejt.
In DT we restrict the range of values of z to those of unit magnitude
(z= ej) so that we focus on complex exponentials of the form ejn.

The Response of LTI Systems to Complex

Exponentials
Example 1
Consider an LTI system with the input x(t) and output y(t) which are
related by a time shift of 3, y(t)=x(t-3).
If the
th input
i t to
t th
the system
t is
i the
th complex
l exponential
ti l signal
i l x(t)=e
(t) j2t, then
th
x(t ) e j 2t y (t ) e j 2 ( t 3) e j 6 e j 2t H ( s ) x(t )
Since ej2t is an eigenfunction, then the associated eigenvalue is H(j2)=e-j6.
y (t ) x(t 3) The impulse response is h(t ) (t 3)

h( )e ( 3)e
s
H (s) d s
d e 3 s , H ( j 2) e 3 s e j 6

As a second example, consider the input signal x(t)=cos(4t)+cos(7t), then
x(t ) cos(4t ) cos(7t ) y (t ) cos(4(t 3)) cos(7(t 3))
1 1 1 1
x(t ) e j 4t e j 4t e j 7 t e j 7 t
2 2 2 2
1 1 1 1
y (t ) e j12 e j 4t e j12 e j 4t e j 21 e j 7 t e j 21 e j 7 t
2 2 2 2
x(t ) a1e s1t a2 e s2t a3e s3t a4 e s4t y (t ) H ( s1 )a1e s1t H ( s2 )a2 e s2t H ( s3 )a3e s3t H ( s4 )a4 e s4t

6
Fourier Series Representation of CT Periodic Signals
Linear Combinations of Harmonically Related Complex
Exponentials

Consider a periodic signal, x(t ) x(t T ) for all t

Then, the fundamental period To: minimum positive nonzero value of T

2
and the fundamental frequency: o
To
Two basic periodic signals:

sinusoid: x(t ) cos 0t

periodic complex exponential: x(t ) e j0t
2
Both periodic with fundamental period T0
0
and fundamental frequency of 0

Fourier Series Representation of CT Periodic Signals

Linear Combinations of Harmonically Related Complex
Exponentials

The set of harmonically related complex exponentials:

2
jk t
k (t ) e jk t e
0 T
; k 0, 1, 2, ...
Each of k has a fundamental frequency that is an integer multiple of 0,
and all have a common period T (for |k|2, the fundamental period of
k(t) is a fraction of T).

A linear combination of harmonically related complex exponentials of the

form (take sk=kj0, then eskt=ejk0t)
2
jk t
x(t ) a (t ) a e
k
k k
k
k
jk0t
ae
k
k
T

is the Fourier series representation of x(t).

Need to find k !!

7
Fourier Series Representation of CT Periodic Signals
The Fourier Series Representation
The Fourier series representation of a periodic signal x(t):
2
jk t
x(t ) ae
k
k
jk0t
ae
k
k
T

Periodic with period T

The term for k=0 DC or constant term

The terms for k=1 1st harmonic component (fundamental

component) periodic with fundamental period T

The terms for k=2 2nd harmonic component periodic with

fundamental period T/2

More generally, the terms for k=N Nth harmonic component

periodic with fundamental period T/N

Fourier Series Representation of CT Periodic Signals

Example 2
Consider a periodic signal x(t), with fundamental frequency 2,
3
x(t ) a e
k 3
k
jk 2 t

1 1 1
where a0 1, a1 a1 , a2 a2 , a3 a3
4 2 3
Rewriting x(t) and collecting each of the harmonic components which
have the same fundamental frequency gives
1 1 1
x(t ) 1 (e j 2 t e j 2 t ) (e j 4 t e j 4 t ) (e j 6 t e j 6 t )
4 2 3
1 2
x(t ) 1 cos 2 t cos 4 t cos 6 t
2 3

8
Fourier Series Representation of CT Periodic Signals
Example 2
Figure illustrates graphically
how the signal x(t) is built up
from its harmonic components

1
x(t ) 1 cos 2 t cos 4 t
2
2
cos 6 t
3
x(t ) x0 (t ) x1 (t ) x2 (t )
x3 (t )

Fourier Series Representation of CT Periodic Signals

Fourier Series of Real Periodic Signals
If periodic signal x(t) is real, then x*(t)=x(t),

x(t )
k
ak e jk0t x* (t ) ae
k
* jk0t
k

Replacing k by k in the summation, we have

x(t ) a
k
*
k e jk0t x(t ) ae
k
k
jk0t

which, by comparison with above equation, requires that

ak a* k , or equivalently, ak* a k , for real periodic signals
Note that this is the case in Example 2,
2 where aks
s are real and ak=a-k.

An alternative form for the Fourier series of real periodic signals can be
obtained by rearranging the summation as

x(t ) ae
k
k
jk0t
a0 [ak e jk0t a k e jk0t ]
k 1

9
Fourier Series Representation of CT Periodic Signals
Fourier Series of Real Periodic Signals
Substituting ak* for a-k,

x(t ) a0 [ak e jk0t ak*e jk0t ]
k 1
Since the two terms inside the summation are complex conjugates of
each other

x(t ) a0 2e ak e jk0t
k 1
If ak is expressed in polar form as ak Ak e jk

x(t ) a0 2e Ak e j ( k0t k ) x(t ) a0 2 Ak cos(k0t k )
k 1
k 1

Another form is obtained by writing ak in rectangular form as ak=Bk+jCk

x(t ) a0 2 [ Bk cos(k0t ) Ck sin(k0t )]
k 1

Fourier Series Representation of CT Periodic Signals

Determination of the Fourier Series Representation of a
CT Periodic Signal

Assume that a given periodic signal x(t) has a FS representation, then we

need a procedure to determine the FS coefficients ak of:

x(t ) ae
k
k
jk0 t

Multiplying both sides with e-jnot and integrating over one period (from 0
to T=2/0), we obtain
T T
T is the fundamental period
x(t )e
jn0t
dt ae k
jk0t jn0 t
e dt of x(t).
0 0 k
T T j ( k n )0t
x(t )e 0 dt
jn t
k e
a dt
0 k 0
T T T

e 0 dt cos((k n)0t )dt j sin((k n)0t )dt

j ( k n ) t

0 0 0

10
 Fourier Series Representation of CT Periodic Signals
Determination of the Fourier Series Representation of a
CT Periodic Signal

For kn, cos((k-n)0t) and sin((k-n)0t) are periodic sinusoids with

fundamental period T/|k-n|. Therefore integrating over an interval of
length T gives zero.
zero The same result can be obtained if we integrate over
any interval of length T.
T
T , k n T T
0 e 0 dt 0, k n
j ( k n ) t
x(t )e 0 dt
jn t
a e k
j ( k n )0 t
dt anT
0
k 0
Thus, the equation for determining the coefficients:
T T
1 1
an
T 0
x(t )e jn0t dt ak x(t )e jk0t dt
T 0
Integration can be done over any interval of length T:

T , k n 1
e , an x(t )e jn0t dt
j ( k n )0t
dt
T
0, k n T T

Fourier Series Representation of CT Periodic Signals

Summary
Fourier series representation of a periodic CT signal x(t) is given by a
pair of equations:
2
jk t
x(t )
k
ak e jk0t
k
ak e T synthesis equation

2
1 1 jk t
ak x(t )e jk0t dt x(t )e T dt analysis equation
T T T T
0 : fundamental frequency, T : fundamental period
The ak coefficients are often called the Fourier series coefficients or
the spectral coefficients of x(t).

The coefficient a0 (ak with k=0) is the DC or constant component of

x(t).
1
T T
ao x(t )dt , is the average value of x(t ) over one period.

11
Fourier Series Representation of CT Periodic Signals
Example 3
Consider the signal x(t)=sin(ot) whose fundamental frequency is o.
Determine the Fourier series coefficients.

For this simple case, it is easier to expand the sinusoidal signal as a

linear combination of complex exponentials and identify the FS
coefficients by inspection.

1 jot 1 jot
sin(0t )
2j
e e
2j
x(t ) sin(0t ) ae
k
k
jko t

1 1
a1 , a1 ,
2j 2j
ak 0, k 1 or 1

Fourier Series Representation of CT Periodic Signals

Example 4

Consider the signal x(t ) 1 sin(0t ) 2 cos(0t ) cos(20t )
4
which has fundamental frequency o. Determine the Fourier series
coefficients.
Again, x(t) can be expanded directly in terms of complex exponentials
and the FS coefficients can be identified by inspection.
1 jot jot 1
x(t ) 1 [e e ] [e jot e jot ] [e j (2ot / 4) e j (2ot / 4) ]
2j 2
1 1 1 1
x(t ) 1 (1 )e jot (1 )e jot ( e j ( / 4) )e j 2ot ( e j ( / 4) )e j 2ot
2j 2j 2 2
1 1 1 1
a0 1, a1 1 1 j , a1 1 1 j,
2j 2 2j 2
1 2 1 2
a2 e j ( / 4) (1 j ), a2 e j ( / 4) (1 j ), ak 0, k 2
2 4 2 4

12
Fourier Series Representation of CT Periodic Signals
Example 4
A bar graph of the magnitude and
phase of ak:

a0 1

a1 1 1 j
2
1.1180
1
a1 1 j
2
ak 0.5000
a2 2 (1 j )
4

a2 2 (1 j )
4
a 0, k 2
k 0.7854

-0.4636

Fourier Series Representation of CT Periodic Signals

Example 5
The periodic square wave in the figure is defined over one period as

1, t T1
x(t )
0, T1 t T / 2

Determine the Fourier series coefficients.

This signal is periodic with fundamental period T and fundamental

frequency o=2/T.
Because of the symmetry of x(t) about t=0,
t=0 it is convenient to perform
integration over the interval -T/2t<T/2. Using these limits of
integration and for k=0, 0 can be found as
T
1 1 1 2T
a0
T T
x (t ) dt
T T1
1dt 1
T
(DC, average value)

13
Fourier Series Representation of CT Periodic Signals
Example 5
For k0, k coefficients can be obtained as
T T1
1 1 1 1
ak x(t )e jk0t dt 1e jk0t dt e jk0t
T T T T1 jk0T T1

jk0T1 jk0T1
2 e e 2sin(k0T1 ) 2 sin(k0T1 )

k0T 2j k0T 2
k T
T
sin(k0T1 )
ak , k 0
k
For T=4T1, x(t) is a square wave that is unity for half the period and
zero for half the period. In this case, oT1=/2.

1 sin( k / 2)
T 4T1 a0 and ak , k0
2 k

Fourier Series Representation of CT Periodic Signals

Example 5
A bar graph of the FS coefficients
k for a fixed value of T1 and T=4T1
severall values
l off T.
T

sin( k / 2 )
T 4T1 ak , k0
k
T=8T1
sin( k / 4)
T 8T1 ak , k0
k
T=16T1
sin( k / 8)
T 16T1 ak , k0
k

14
Convergence of the Fourier Series
Fourier maintained that any periodic signal could be
represented
p by
y a Fourier series.

The truth is that Fourier series can be used to represent an

extremely large class of periodic signals but not all of them.

The question is that when a periodic signal x(t) does in fact have
a Fourier series representation?

2
jk t
x(t )
k
ak e jk0t
k
ak e T

2
1 1 jk t
ak
T T
x(t )e jk0t dt x(t )e T dt
T T

Convergence of the Fourier Series

Fourier series expansion is possible if x(t) is a periodic function
satisfying Dirichlet conditions.

These conditions guarantees that

x(t) equals its Fourier series representations,
except at isolated values of t for which x(t) is discontinuous.

For a periodic signal with a finite number of discontinuities in

each period, the FS representation equals the signal everywhere
except at isolated points of discontinuity.
At the
these
e di
discontinuities,
ontin itie the FS e expansion
p n ion converges
on e ge to the
average of the values on either side of the discontinuity.

For a periodic signal that has no discontinuities, the FS

representation converges and equals the original signal at every
value of t.

15
 Convergence of the Fourier Series
Dirichlet Conditions
Condition 1: Over any period, x(t) must be absolutely integrable, that is,

1 1

T
x(t ) dt ak
T T
x(t )e jk0t dt x(t ) dt
T T
1
T T
So if x(t ) dt then ak

A periodic signal that violates the first Dirichlet condition is

1
x(t ) , 0 t 1
t
x(t) is periodic with period 1.

Convergence of the Fourier Series

Dirichlet Conditions
Condition 2: x(t) has a finite number of maxima and minima within any
period.
An example of a function that meets condition 1 but not condition 2 is
2
x(t ) sin , 0 t 1
t 1
For this function, which is periodic with period T=1, x(t ) dt 1.
0
The function has, however, an infinite number of maxima and minima
in the interval.

16
 Convergence of the Fourier Series
Dirichlet Conditions
Condition 3: x(t) has a finite number of discontinuities within a period.
Furthermore, each of these discontinuities is finite .

A periodic signal that violates the third Dirichlet condition is shown in

the figure below.
The signal, of period T=8, is composed of infinite number of sections,
each of which is half the height and half the width of the previous
section. There are an infinite number of discontinuities in each period,
thereby violating Condition 3.

Convergence of the Fourier Series

How the Fourier Series Converges for a Periodic
Signal with Discontinuities
In 1898, Albert Michelson (an American physicist) used his harmonic
analyzer to compute xN(t) the truncated Fourier series (values of N up to
80) approximation for the square wave.
wave
N
xN (t ) ae
k N
k
jk0t

x1 (t ) a1e j10t a0 a1e j10t

x79 (t ) a79 e j 790t a1e j10t
a0 a1e j10t a79 e j 790t
Michelson is observed that
the partial sum in the vicinity of discontinuity exhibits ripples and
the peak amplitude of these ripples does not seem to decrease with
increasing N.

17
Convergence of the Fourier Series
How the Fourier Series Converges for a Periodic
Signal with Discontinuities

Michelson wrote to Josiah Gibbs (mathematical

physicist)
In 1899,
1899 Gibbs showed that
the partial sum near discontinuity exhibits
ripples and
the peak amplitude remains constant
with increasing N
This is known as the Gibbs phenomenon.
xN (T1 ) xN (T1 )
xN (T1 )
2

Convergence of the Fourier Series

Remarks

Complex exponentials are continuous for all t. Hence
is continuous as well.
x(t ) ae
k
k
jk0 t

What if x(t) has discontinuity at t=t0?

If Dirichlet conditions are satisfied, x(t) is equal to x(t ) ae k
jk0 t

at every continuity point of x(t). k

At discontinuities, x(t ) ae
k
k
jk0 t
equals to the average of the

values on either side of the discontinuity, i.e., x(t0 ), x(t0 )

x(t0 ) x(t0 )
If x(t ) is discontinuous at t t0 , x(t0 )
k
ak e jk0t
2

18