Calculating a Fourier series with MATLAB manually problem

20 次查看（过去 30 天）
Owen
Owen 2024-3-30
编辑： Paul 2024-4-1
Hi there, I'm trying to transfer this academic terms learnings on the fourier series to MATLAB, and running into difficulties.
I want to be able to plot a fourier series approximation against the original function but when I try and calculate the fourier series to 5 steps of n in the final line I get a division by zero error.
I can't see where this division by zero occurs, the only division I am using is dividing by the period, T, which is just pi.
I understand this is probably a bad way to do things! And I don't think I understand the symsum enough, but any help gratefully recieved.
clear all
close all hidden
clc
syms t % time
syms n % number of terms to calculate the sum to
syms T % period
syms a_n(t, n, T) % fourier co-effecients
syms b_n(t, n, T)
syms a_0(t, n, T)
syms f(t)
f = sin(t)^2
tvec = -3 : 0.01 : 3;
xvec = sin(tvec).^2;
figure
plot(tvec, xvec)
title('Plot of f(t)')
T = sym(pi)
a_0 = (2/T)*int(f, t, 0, T)/2
a_n = (2/T)*int(f*cos(((2*pi)/T)*n*t), t, 0, T)
b_n = (2/T)*int(f*sin(((2*pi)/T)*n*t), t, 0, T)
exp = a_0 + symsum(a_n*cos(((2*pi)/T)*n*t) + b_n*sin(((2*pi)/T)*n*t), n, 1, 5)
% division by zero error
  10 个评论
显示 8更早的评论
Torsten
Torsten 2024-3-31
I'm interested in why for this specific function we get an a_n that is invalid for n=1.
But I alrady answered it - it's a MATLAB bug.
Owen
Owen 2024-3-31
Thank you, I had misread that it was specifically a bug that b_n is not also invalid, but I see you meant the other way round, thanks!

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Paul
Paul 2024-3-31
编辑：Paul 2024-4-1
Ran in:
I don't think anything is wrong with the code in the Question, at least based on visual inspection (though it's always best to use sym(pi) rather than pi in symbolic expressions). Rather, int often doesn't recognize special cases when evaluating parameterized, Fourier integrals. The user needs to identify those cases and manually process them.
syms t % time
syms n % number of terms to calculate the sum to
f = sin(t)^2;
T = sym(pi);
a_0 = 2/T*int(f,t,0,T)/2
a_0 = 
a_n = 2/T*int(f*cos(2*sym(pi)/T*n*t),t,0,T)
a_n = 
b_n = 2/T*int(f*sin(2*sym(pi)/T*n*t),t,0,T)
b_n = 
I don't know why int() captures the special cases of n for b_n but does not for a_n. So we have to handle the special case for a_n manually (keeping in mind that we only care about n >= 1)
a_n = piecewise(n==1,limit(a_n,n,1),a_n)
a_n = 
%b_n = piecewise(n==1,limit(b_n,n,1),b_n)
We can simplify things by placing appropriate assumptions on n. This step isn't really necessary, but adds some clarity IMO.
assume(n,'integer')
assumeAlso(n,'positive'); % n >= 1 for the sin/cos Fourier series
a_n = simplify(a_n)
a_n = 
b_n = simplify(b_n)
b_n = 
0
The reconstruction yields the expected result
fr = a_0 + symsum(a_n*cos(2*sym(pi)/T*n*t),n,1,5)
fr = 
However, an incorrect result for a_n ensues if evaluating the integrals with the assumptions on n.
a_n = 2/T*int(f*cos(2*sym(pi)/T*n*t),t,0,T)
a_n = 
0
b_n = 2/T*int(f*sin(2*sym(pi)/T*n*t),t,0,T)
b_n = 
0
That looks like a bug.
Alternatively, we can Hold the integrals and then release() them after evaluating them at specified values of n (the @doc functionality isn't finding symbolic release() ?)
a_n(n) = 2/T*int(f*cos(2*sym(pi)/T*n*t),t,0,T,'Hold',true)
a_n(n) = 
b_n(n) = 2/T*int(f*sin(2*sym(pi)/T*n*t),t,0,T,'Hold',true)
b_n(n) = 
a_n = release(a_n(1:5))
a_n = 
b_n = release(b_n(1:5))
b_n = 
fr = a_0 + sum(a_n.*cos(2*sym(pi)/T*(1:5)*t))
fr = 
  1 个评论
Owen
Owen 2024-3-31
Thanks Paul this has cleared everything up and thank you for the work-arounds, great to have some suggestions of new functrionality for me to learn and glad to know that my code was not the problem!

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Symbolic Math Toolbox 的更多信息

标签

产品


版本

R2023b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by