Vous êtes sur la page 1sur 8

Fourier Analysis - introduction from a practical point of view

In calculus or mathematics, Fourier analysis is a topic which grew from the study
of Fourier series. This field began with the study of the way periodic or general functions
might be represented by summations of simpler trigonometric functions (sine or cosine
series).

Fourier analysis has many applications in science, physics, digital signal processing, imaging,
numerical analysis, acoustics... and in many other areas.

We are going to comment on some electronic applications, where there are two emphasized
domains: analysis in time domain, to find the voltages or currents at specific moments, and
the analysis in the frequency domain, to study (in order to possibly manipulate) the
frequencies present in a given signal.

Fourier series

This figure shows a sine wave with a peak value Vp and a period T.

When AC analysis is presented in circuit theory courses, they usually focus on the sine wave
because it helps understand many effects. Based on those effects or behaviors of electronic
components we can start the study of non-sinusoidal signals.

Periodic waves

The shapes below (a half-sine, a sawtooth and a square wave) complete their basic form over
a period T, and each cycle is an exact copy of the first period or cycle. This kind of waveform
with repetitive cycles is called a periodic wave with period T.
The addition of sine waves with appropriate amplitudes and phases can represent triangular
waves, or a different combination of sine series and waves can produce a sawtooth signal,
and another different combination may produce a half-wave signal. In other words, any
periodic wave is a superposition of sine waves.

The sine waves are harmonically related, which means that the frequencies are harmonics (or
multiples) of a fundamental (lowest) frequency. Given a periodic wave, you can find the
period T when measured with an oscilloscope. The reciprocal of T is equal to the fundamental
frequency.

In formulas:

The fundamental frequency f1 = 1 / T


The second harmonic has a frequency f2 = 2f1
The third harmonic has a frequency f3 = 3f1

In general, the nth harmonic has a frequency


fn = nf1

Formulas for the Fourier series

Expressed in words, the figure above means:


periodic wave = DC component +
first harmonic +
second harmonic +
... +
nth harmonic

In mathematical terms:

v = V0 + V1 sin(wt + Ф1) + V2 sin(2wt + Ф2) + ... + Vn sin(nwt + Фn)

This famous equation is called the Fourier series, and it means that a periodic wave is a
superposition of harmonically related sine waves.

Voltage v is the value of the periodic wave at any instant, this value can be calculated by
adding the DC component and the instant values of all the the harmonics. The first term in the
Fourier series is V0, which is a constant and represents the DC component. The
coefficients V1, V2... Vn are the peak values of the harmonics. The angular frequency w is
equal to 2πf, so you can see that each term in the Fourier series represents the next higher
harmonic.

Harmonic waves theoretically continue to infinity. However, five to ten terms are often
enough to represent a periodic signal within an accuracy of 5%. With the right combination
of magnitudes, angles and phases, it is possible to reproduce any periodic waveform.

Now, let’s do a brief experiment with our numerical software. You can see how to produce
a square wave, here.

We can implement this code:

% Define and plot one cycle of a square wave


x = 0 : .01 : 2*pi;
y = square(x);
plot(x,y)
axis([min(x) max(x) min(y)-.5 max(y)+.5])

% Plot f1 and f3 and add them up


figure
y1 = sin(x);
y2 = .2*sin(3*x);
plot(x,y1,'b', x,y2,'b:', x, y1+y2, 'g', 'linewidth', 2)
axis([min(x) max(x) min(y)-.5 max(y)+.5])

We get two figures:


The figure on the left is a plain square wave. We want to find an equivalent of that periodic
wave by utilizing only sines. The figure on the right represents our first attempt to do it. We
use the first and third harmonics of the original wave, we add them up and show the results.
We can clearly see that the green line is getting closer to a square wave. In order to produce a
better approximation, we’d need to adjust our amplitudes and include more termsin the
series.

Fourier Series - an understandable introduction...

In mathematics, the Fourier series is an infinite sequence of terms used to solve special types
of Problems.

The series consists of an infinite sum of sines and cosines that repeats over fixed intervals,
and so is very useful for analyzing periodic functions.

Fourier theorem is the key to the analysis in the frequency domain when talking about
electronic applications.

According to the Fourier theorem, a wave or signal is composed of a series of sinusoidal


components whose frequencies are those of the fundamental freq. and its harmonics, each
component having the proper amplitude and phase. The sequence of components that form
this wave is called its spectrum.

There are, at least, two ways to perform a circuit analysis. First, you can find what happens
with the periodic waveform at every instant of time; second, you can determine what happens
to each of its harmonics. Sometimes, the first analysis (time domain) is fast, but often the
second analysis (frequency domain) is better.

Spectral components
Suppose Vp-p represents the peak-to-peak value of a sawtooth wave. It can be demonstrated
that the amplitude of each harmonic is

Vn = Vp-p /(nπ)

This means that the peak value of the nth harmonic is equal to Vp-p divided by nπ.

For example, the top subplot in this figure shows a sawtooth wave with a peak-to-peak value
of 10,

the harmonics have then the following peak values:

V1 = 10/(1π) = 3.18
V2 = 10/(2π) = 1.59
V3 = 10/(3π) = 1.06

The second subplot in the figure above shows the first seven magnitudes of the respective
harmonics. The bottom subplot shows how the addition of different harmonics (red lines) can
produce something close to the original sawtooth (green line): the more considered terms the
more accuracy.

You can learn some ideas on how to produce periodic square or sawtooth waves, here.

The Matlab code to produce the first two subplots in the figure above, is:

x = 0 : .01 : 5*pi;
y = 5 * sawtooth(x);
subplot(311), plot(x, y)
axis([min(x) max(x) min(y)-.5 max(y)+.5])
xlabel('Sawtooth wave')

n = 1 : 7;
v = 10 ./(n*pi);
subplot(312), stem(n, v)
axis([min(n)-1 max(n)+1 0 max(v)+.5])
xlabel('First harmonics')

With an oscilloscope (time domain tool) you can see the periodic signal as a function of time.
The vertical axis represents voltage and the horizontal axis represents time, in fact, the
oscilloscope represents the instantaneous value Vof the periodic wave.

The spectrum analyzer is different because it’s an instrument of the frequency domain. The
horizontal axis represents the frequency, and the vertical axis represents the amplitude of the
harmonics. For example, you would use an oscilloscope to observe the top subplot above.
You’d use a spectrum analyzer to observe the middle subplot above (that's what the Fourier
series are for). This kind of information is called the spectrum, where each level of the
vertical lines represents the peak value of the harmonics, and the horizontal axis shows the
corresponding frequency.

Any periodic wave has a spectrum or a set of vertical lines representing the harmonics. The
spectrum naturally differs from one periodic signal to another.

Three basic spectra

The figures below show three periodic waves and their spectra. In each case, Vp-p is the peak-
to-peak value of the periodic wave and for convenience we show only the harmonics up
to n = 7. The formula of the harmonic peak appears for each case.
You can learn some ideas on how to produce square and sawtooth functions, here.

The code to produce the Fourier series of harmonics above is:

clear, clc, close all

x = 0 : .01 : 5*pi;
y = sawtooth(x);
subplot(211), plot(x, y)
axis([min(x) max(x) min(y)-.5 max(y)+.5])
xlabel('Sawtooth wave')

n = 1 : 7;
v = 2 ./(n*pi);
subplot(212), stem(n, v)
axis([min(n)-1 max(n)+1 0 max(v)+.5])
xlabel('First harmonics')

figure
y = square(x);
subplot(211), plot(x, y)
axis([min(x) max(x) min(y)-.5 max(y)+.5])
xlabel('Square wave')

n = 1 : 2 : 7;
v = 4 ./(n*pi);
subplot(212), stem(n, v)
axis([min(n)-1 max(n)+1 0 max(v)+.5])
xlabel('First harmonics')

figure
y = abs(sin(x));
subplot(211), plot(x, y)
axis([min(x) max(x) min(y)-.5 max(y)+.5])
xlabel('Full-wave rectified sine')

n = 1 : 7;
v = 4 ./(pi*(4*n.^2 - 1));
subplot(212), stem(n, v)
axis([min(n)-1 max(n)+1 0 max(v)+.5])
xlabel('First harmonics')

Vous aimerez peut-être aussi