Notes On Fourier Series and Fourier Transforms: 1 Periodic Functions

Notes on Fourier Series and Fourier Transforms
S Chaturvedi
August 31, 2017
1 Periodic functions
A (real or complex) function f (t) of a real variable t is said to be a periodic
function if there is a smallest T such that
f (t) = f (t + T )
for all t and T is called its period.
If f1 (t) and f2 (t) are two periodic functions with periods T1 and T2 then
their linear combination
g(t) = af1 (t) + bf2 (t)
is periodic function if and only if T1 and T2 are commensurate i.e. T1 /T2 is
a rational number :
T1
= m/n, m, n integers
T2
or in other words T1 = m and T2 = n. The period T of g(t) will be the
smallest number divisible by both T1 and T2 i.e LCM(m, n).
A function f (t) defined over a finite interval say 0 t can be em-
bedded into a periodic function g(t) with period T in many different ways.
The function g(t) is called the periodic extension of f (t). For instance, two
possible periodic extensions of f (t) shown below
are:
1
2 Fourier Series
Any piecewise continuous periodic function f (t) of period T can be expanded
as
X 2
f (t) = cn eint , =
n=
T
in terms of the set of functions
fn (t) = {eint , 2/T ; n = 0, 1, 2, }.
This set of functions form an orthonormal basis with respect to the scalar
product:
1 T /2
Z
(f, g) = f (t)g(t)
T T /2
i.e Z T /2 Z T /2
1 1
(fn , fm ) = fn (t)fm (t) = ei(mn)t = nm
T T /2 T T /2
Using this orthogonality property, given an f (t), we can compute the cn s,

the Fourier coefficients, as follows:
1 T /2
Z
cn = (fn , f ) = dt eint f (t)
T T /2
The Fourier series thus digitizes a periodic signal f (t) in that it stores
the periodic signal f (t), 0 t T in terms of a denumerable set cn , n =
0, 1, 2, . Often one does not need all the cn s and the signal can be
fairly well approximated by a small number of cn s.
Using eint = cos nt + i sin nt, the Fourier series can also be expressed
in the sine-cosine form as
inf ty
a0 X
f (t) = + [an cos nt+bn sin nt], a0 2c0 , an (cn +cn ); an i(cn cn )
2 n=1
Given an f (t) the Fourier coefficients that appear in this form can be com-
puted using the relations
2 T /2 2 T /2
Z Z
an = dtf (t) cos nt, n = 0, 1, , bn = dtf (t) sin nt, n = 1, 2
T T /2 T T /2
2
3 Convergence
If f (t) is continuous at t = t0 then the Fourier series evaluated at t = t0
converges to f (t0 )
If f (t) is discontinuous at t = t0 the the Fourier series evaluated at t = t0
converges to the average value of f (t) at t = t0 i.e to [f (t0+ ) + f (t0 )]/2
4 Parseval Identity
Z T /2
1 2
X 1 1X
dt|f (t)| = |cn | = |a0 |2 +
2
[|an |2 + |bn |2 ]
T T /2 n=
4 2 n=0
5 T : Fourier Series to Fourier transform

A non periodic function may be viewed as a periodic function with T = .
We now consider this limit of the Fourier series. Substituting for the cn s in
Fourier series we have
Z T /2
int 1 0
X
f (t) = e dt0 f (t0 )eint
n=
T T /2
Z T /2
1 X in(tt0 )
= dt0 f (t0 ) e
T /2 T n=
Z T /2
1 X 2 in(tt0 )
= dt0 f (t0 ) e
T /2 2 n= T
Introducing a discrete variable x taking values {n, n = 0, 1, 2} with the

difference x between adjacent values being 2/T we find that the sum on
the RHS becomes an integral in the limit T leading to
Z Z
0 0 1 0
f (t) = dt f (t ) dxeix(tt )
2
which after interchanging the order of integration and replacing x by may
also be written as
Z Z
1 it 1 0
f (t) = de dt0 eit f (t0 )
2 2
3
The first of the two equations leads to the notion of the Dirac delta function
Z
0 1 0
(t t ) = dxeix(tt )
2
and the second to the notion of the Fourier transform

Z
1
f (t) = deit F (]
2
Z
1
F () F[f (t)] = dteit f (t)
2
|F ()|2 is called the power spectrum of f (t) : it gives the amount of the
frequency present in the signal f (t)
6 Delta function
The delta function, by definition, has the property:
Z
f (t)(t t0 ) = f (t0 )

It can be pictured as a function which is zero everywhere except at the point

where its argument vanishes where it has an infinite spike. As a result the
integral Z b
f (t)(t t0 )
a
equals f (t0 ) if t0 is contained in the interval (a, b) and equals 0 if not.
7 Parseval Identity
Z Z
2
dt|f (t)| = d|f ()|2

Functions f (t) for which

Z
dt|f (t)|2 <

4
are called square integrable functions. From the Parseval identity it follows
that the Fourier transforms of square integrable functions are also square
integrable. Further, it can be shown that linear combinations of square in-
tegrable functions are also square integrable : the set of square integrable
functions form a Hilbert space a vector space equipped with the scalar prod-
uct Z
(f, g) = dt f (t)g(t)

8 Discrete Fourier Transform

The Fourier transform F (j), j = 0, 1, , N 1 of a function f (k), k =
0, 1, , N 1 over a discrete set of N points labelled 0, 1, 2, , N 1 is
defined as
N 1
1 X jk
F (j) = f (k), e2i/N
N k=0
[ Note that here stands for the N th root of unity in contrast to the variable
which appears in the definition of Fourier transforms.] The inverse relation
expressing f in terms of its Fourier transform is
N 1
1 X jk
f (k) = F (j),
N j=0
which follows from the orthonality relation

N 1
1 X (j`)k
= j`
N k=0
[When j = `, the RHS is clearly equal to 1. When j 6= `, the RHS is evidently

the sum of the N roots of unity which we know equals 0.]
As before we have the Parseval identity
N
X 1 N
X 1
2
|F (j)| = |f (j)|2
j=0 j=0
The Discrete Fourier transform above in the special case when N = 2n is

variously referred to in the literature as the Fast Fourier transform as in this
5
case there are efficient algorthims for computing the Fourier transform. In
the Quantum Information Theory literature it is referred to as the Quantum
Fourier Transorm
The three Fourier transforms that we have considered respectively cor-
respond to functions defined on a circle, the real line, and a periodic lattice
of N points (N points on a circle )

Notes On Fourier Series and Fourier Transforms: 1 Periodic Functions

Uploaded by

Copyright:

Available Formats

Notes On Fourier Series and Fourier Transforms: 1 Periodic Functions

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Notes On Fourier Series and Fourier Transforms: 1 Periodic Functions

Uploaded by

Copyright:

Available Formats

Notes on Fourier Series and Fourier Transforms

fn (t) = {eint , 2/T ; n = 0, 1, 2, }.

Using this orthogonality property, given an f (t), we can compute the cn s,

5 T : Fourier Series to Fourier transform

Introducing a discrete variable x taking values {n, n = 0, 1, 2} with the

and the second to the notion of the Fourier transform

It can be pictured as a function which is zero everywhere except at the point

equals f (t0 ) if t0 is contained in the interval (a, b) and equals 0 if not.

Functions f (t) for which

8 Discrete Fourier Transform

which follows from the orthonality relation

[When j = `, the RHS is clearly equal to 1. When j 6= `, the RHS is evidently

The Discrete Fourier transform above in the special case when N = 2n is

You might also like