Motivation

We mentioned at the start of the last section that sine

waves have a special property in relation to linear systems.

Section 5

A sine wave at the input

leads to a (possibly different) sine wave at the output.

Fourier Series

The Fourier series is introduced using an analogy with

splitting vectors up into components.
The symmetry properties that enable us to predict that
certain coefficients are zero are presented.

It would therefore be useful to be able to express an

arbitrary signal in terms of a sum of sine waves.

original
signal

System

split into
sine waves
61

Sine wave with

amplitude or
System phase changed

Sine wave Linear

output

recombine

62

Motivation: Car Suspension

Splitting up Vectors

Supposing we know that our car suspension will start

to oscillate (bounce up and down uncomfortably) at
frequency f .

We want to express a signal f (t) in the range

t in terms of some basic signals, i.e. sine waves.
Lets look first at how we do a similar thing with vectors.
Consider how we express the arbitrary vector r in terms
of the basis vectors i and j.

r=ai+bj
We want to measure a variety of typical road profiles
and calculate how much of frequency f they each
contain (with the car travelling at a particular speed).
This will tell us which combinations of road profile and
speed are likely to be a problem.
The Fourier series enables us to represent the road
profile as the sum of a set of sinusoidal components
at different frequencies.
63

where
a=r.i
i. i

b=r.j
j.j

j
i
The basis vectors are orthogonal: i.j = 0.
64

Basis Functions
Just as we represent r using orthogonal basis vectors,
we want to represent f (t) in the range to using
orthogonal basis functions. We only need two vectors,
but we need an infinite number of functions.
1 (i.e. a constant term)
cos(t) cos(2t) cos(3t)
sin(t) sin(2t) sin(3t)

cos(4t)
sin(4t)

...
...

sin(nt) sin(mt) dt =
Z

bn =

d=

sin(nt)f (t) dt

sin(nt) sin(nt) dt

R
1f (t) dt

R
11 dt

1
sin(nt)f (t) dt
=

Z

1
=
f (t) dt
2
Z

The equivalent of our vector expression for r in terms

of i and j, (i.e. r = ai + bj) is an expression for f in

,n=m

1
cos(nt)f (t) dt
=
an = R

cos(nt) cos(nt) dt
cos(nt)f (t) dt

R
cos(nt) sin(mt) dt = 0
R
cos(nt) 1 dt = 0
R

sin(nt) 1 dt = 0
(
Z
0 , n 6= m

cos(nt) cos(mt) dt =

The equivalents of our vector dot product expressions

to calculate the component of r in each direction (eg.
a = (r.i)/(i.i)) are:
R

If n and m are positive integers greater than zero.

Fourier Series

0 , n 6= m
,n=m

terms of all the basis functions.

1 1 dt = 2
f (t) =

p(t)q(t) dt as our dot product for funcSo, using

tions, the basis functions are orthogonal.

65

n=1

an cos(nt) +

n=1

bn sin(nt) + d 1

66

Fourier Series Properties

Fourier Series Example 1

Represent the square wave
f (t) as a Fourier series.

1
2

f(t)
1

t
2

1
an =
cos(nt)f (t) dt = 0

Z
1
2 (1 (1)n )
bn =
sin(nt)f (t) dt =

n
Z
1
f (t) dt
=0
d=
2
Z

Thus, we can model the square wave function f (t)

using:

1. We can use any range of length 2 instead of

t in the Fourier formulae. For example,
0 t 2 is equally OK.

2. We are only modelling the function f (t) in the

specified range (eg. to , or 0 to 2). Outside
this range the model will just repeat with period 2.
This is fine if the function we wish to model is periodic itself, but if the function is not periodic the Fourier
model will probably only be useful over the range on
which it was built.
Original function

f (t) = d +
=
=

(an cos(nt) + bn sin(nt))

n=1

X 2 (1 (1)n )

n=1
"

Fourier series representation

based on range 0 to 2

sin(nt)

4
sin(3t)
sin(5t)
sin(t) +
+
+ ...

3
5

67

68

Fourier Series Example 2

Represent f (t) = et as a Fourier series between
and .
an =

Z
1

cos(nt)et dt =

bn =

1
sin(nt)et dt

d=

1 t
e dt
2

(1)n e e


2
1+n


(1)n e e n


=
1 + n2


e e
=
2

Fourier Model of Exponential

Fourier model of e t
built on range
to repeats
every 2

et

Thus, in the range < t < we can model the

function f (t) = et using:
f (t) = d +

(an cos(nt) + bn sin(nt))

n=1

e e

n
X (1)
1
+
[cos(nt) n sin(nt)]
2
2
1
+
n
n=1

3.68 3.68 cos(t) + 3.68 sin(t)

+1.47 cos(2t) 2.94 sin(2t) . . .
69

70

Avoiding Integration

Symmetric Signals
ODD function
f (t) = f (t)
EVEN function f (t) = f (t)

eg: sin(t)
eg: cos(t)

cos(t)

The a n terms
model the EVEN
component in
the function

If we can spot a symmetry in the function to be represented then we can avoid evaluating one or more of
the Fourier integrals.
No even component all an = 0
No odd component all bn = 0
Zero mean d = 0

sin(t)

The b n terms
model the ODD
component in
the function
t

The d term
models the
mean value of
the function

71

EVEN function
with nonzero
mean: b n = 0

Purely ODD
function with
zero mean:
a n = 0 and d = 0

Function with
zero mean:
d=0

72

Fourier Series Example 3

Fourier Series Example 4

Find the Fourier series representation for the function

f (t) below.

Find the Fourier series representation for the function

f (t) = cos(t + /4).

f(t)

EVEN function
with zero mean:
b n = 0 and d = 0

This function has a mean value of zero so d = 0.

Z
1
cos(nt) cos(t + /4) dt
an =

=
cos(nt + t + ) + cos(nt t ) dt
2
4
4
Z

We only have to calculate an

1
= , when n = 1 and 0 otherwise.
2

R
cos(nt)f (t) dt
an = 1

R0
= 1
cos(nt)(t /2) dt
R
1
+ 0 cos(nt)(t /2) dt

1
sin(nt) cos(t + /4) dt
bn =

Z

=
sin(nt + t + ) + sin(nt t ) dt
2
4
4

R
= 2 0 cos(nt)(t /2)
dt
(
0
= n22 ((1)n 1) =
4

, n even
, n odd
n2

1
= , when n = 1 and 0 otherwise.
2

so the Fourier series is:

so the Fourier series is:
1
1
4
cos(t) + cos(3t) +
cos(5t) + . . .
f (t) =

9
25


73

f (t) =

cos(t) sin(t)

2
74

Section 5: Summary
Section 6
Periodic functions, (so far only with period 2), can be
represented using the the Fourier series.

General Fourier Series

We can use symmetry properties of the function to
spot that certain Fourier coefficients will be zero, and
hence avoid performing the integral to evaluate them.
The Fourier series for arbitrary period is presented.
Functions with zero mean have d = 0.

We compare three techniques for calculating a general range Fourier series: direct integration, using a
related series of delta functions, and using the electrical data book.

Purely odd functions have an = 0.

Purely even functions have bn = 0.
Segments of non-periodic functions can be represented
using the Fourier series in the same way. The Fourier
series representation just repeats outside the range
on which it was built.

75

During the direct integration example, some symmetry

arguments for simplifying integrals are illustrated.

76