Appendix 1: Fourier Series

Appendix 1
Fourier Series
A1.1 Definitions
A periodic function of time, , obeys the equation
(1.1)
where is the period and
is the angular frequency. Any periodic function can be represented as a Fourier

series
(1.2)
where the are the Fourier coefficients. can have discontinuities but
must satisfy the condition
(1.3)
The cosine and sine functions are orthogonal, meaning
(1.4)
(1.5)
(1.6)
where is the Kronecker delta. They are also orthogonal to the function
[prove it]. Any reasonable function (i.e. one satisfying (1.3)) can be expanded in
the set of functions —the set is said to be complete.
To find the coefficients for any from equations (1.4)–(1.6) do the following:
228
A1.2 Examples 229
(i) Multiply by the corresponding function.

(ii) Integrate over the whole range
(iii) Use the orthogonality relations (1.4)–(1.6) to show that all integrals vanish
except the one containing the chosen function coefficient corresponding to
the relevant function
(iv) Rearrange the equation to give the coefficient explicitly
Thus
(1.7)
(1.8)
(1.9)
A1.2 Examples
(i) The square wave. This function may be defined as periodic with period
with
The Fourier series for has the form of (1.2) where the coefficients are
given by (1.8)–(1.9). Substituting for gives
and
and
230 Fourier Series
The series can then be written as
The cosine and constant terms vanish: this always happens for odd func-
tions, those with ; the sine terms always vanish for even
functions, which have . The Fourier series represents a de-
composition of the square wave into its component harmonics.
(ii) A parabola. It is left as an exercise to show, using formulae (1.2) and
(1.8)–(1.9) and (8), that the Fourier series of the function in the
interval is
[This requires some tedious integration by parts. It is possible to circumvent

some of the integration by noting the function is symmetrical about the
point , while all the sine functions are antisymmetrical].
(iii) The sawtooth function, defined by on the interval
and periodic repetition: [You need to sketch the function
over several periods, say to 3, to see why it is called a sawtooth.]
The Fourier series of has the form of (1.2) with . All integrals
containing cosines vanish. Doing the remaining integrals yields:
A1.3 Complex Fourier Series

The function may also be expanded in complex exponentials.
(1.10)
The exponentials are orthogonal:
(1.11)
A1.4 Some Properties of Fourier Series 231
Note that orthogonality for complex functions requires one of the functions be
complex conjugated (see Appendix 5. Multiplying both sides of (1.10) by
and integrating from to gives the formula for the complex coefficients :
(1.12)
The complex Fourier series (1.10) may be shown to be equivalent to the real
form (1.2). Using DeMoivre’s theorem:
we rearrange the real Fourier series (1.2) into the complex form by choosing the
coefficients
(1.13)
Note that if is a real function.
A1.4 Some Properties of Fourier Series

(i) Periodicity. The Fourier series is periodic with the defined period; in fact
the Fourier series is not completely specified until the period is given.
The Fourier series converges on the required function only in the given
interval, outside the interval it will be different [see the parabola (ii) above]
except in the special case when the function itself has the same period [e.g.
the square wave (iii) above].
(ii) Even and odd functions. Sines are odd functions ; the DC
term and cosines are even functions . Even functions can
be expanded in cosine terms plus a DC shift, odd functions can be expanded
in sine terms.
(iii) Half-period series. A function defined only on the interval can
be expressed in terms either sine or cosine terms. To see this, consider the
function extended to the interval as an even function by
. The Fourier series on the interval with period will contain
only cosine terms but will converge on the required function in the original
interval . Similarly, extending the function as an odd function with
will yield a Fourier series with only sine terms.
(iv) Least squares approximation. When a Fourier series expansion of a con-
tinuous function is truncated after terms it leaves a least squares
approximation to the original function: the area of the squared difference
232 Fourier Series
between the two functions is minimised and goes to zero as increases.

This property follows from orthogonality of sines and cosines: it is useful
because increasing does not change the existing coefficients, and only
the higher terms need be computed in order to improve the approximation.
If is expanded in functions that are not orthogonal, for example the
monomials , then raising the truncation point requires all the coeffi-
cients to be recomputed, not just the new ones. See Appendix 5 (ii).
(v) Gibbs’ phenomenon. The Fourier series still converges in the least squares
Fig. A1.1. Fourier series approximation to a square wave. The Fourier series has been
truncated at 1,2,4,8, and 16, terms. Note the overshoot and compression of the oscillations
around the discontinuity.
sense when is discontinous, but the truncated series oscillates near the
discontinuity with a constant overshoot of about 19% of the height of the
A1.4 Some Properties of Fourier Series 233
discontinuity. As increases the oscillations are compressed to a pro-

gressively smaller region about the discontinuity but the maximum error
remains the same. The convergence is said to be non-uniform because the
error depends not only on but also on . These oscillations are known as
Gibbs’ phenomenon. Figure A1.1 illustrates the effect for the square wave.
.

Appendix 1: Fourier Series

Uploaded by

Appendix 1: Fourier Series

Uploaded by

Appendix 1

is the angular frequency. Any periodic function can be represented as a Fourier

The cosine and sine functions are orthogonal, meaning

(i) Multiply by the corresponding function.

The series can then be written as

[This requires some tedious integration by parts. It is possible to circumvent

A1.3 Complex Fourier Series

The exponentials are orthogonal:

Note that if is a real function.

A1.4 Some Properties of Fourier Series

between the two functions is minimised and goes to zero as increases.

discontinuity. As increases the oscillations are compressed to a pro-

You might also like