Can someone explain what is represented by the patterns traced out by the circles in the left part of this animation? If these circles can be thought of as lying in the complex plane they are represented by $e^{i\omega_0kt}$ for different $k$. The Fourier coefficient $\hat{f}(k)$ of the function being expanded should be a complex number specifying the magnitude and phase of these circles, in order for their sum to approach this function. So shouldn't these circles be drawing the function? Why is the function the projection on the imaginary axis? This is like ignoring all the cosine terms no? But why do that and why is the sum of the circles not $f$?

If the addition of the circles is the way to represent $f$ using its Fourier decomposition shouldn't the complex value that results from this addition at each point in time represent $f$ at that time? For another example, youtube.com/watch?v=8Q0NxFt-s7Y what is happening in this animation? Why not take only y here?

  • $\begingroup$ The connection between the circles and the complex plane is just a convention. You can let the vertical axis be real and the horizontal be imaginary or vice versa. In other words, to connect a circle to a complex number you have to choose an arbitrary orthogonal coordinate system. Your statment ".. the projection of the imaginary axis?" is one convention. It could just as easily be the real axis. $\endgroup$ – Somos Nov 1 '17 at 22:44
  • $\begingroup$ @Somos I know the question is why project. I used the imaginary as a convention too, to explain. $\endgroup$ – vounoo Nov 1 '17 at 22:47
  • $\begingroup$ For convenient practical reasons. If you trace a point on the smallest circle, with a separate time axis, it will trace out a 3 dimensional curve sort of like a helical spiral. Unless you have 3D projection system, you settle for a 2D projection of the 3D picture. $\endgroup$ – Somos Nov 1 '17 at 22:53
  • $\begingroup$ @Somos Why trace it with a different time axis in the first place? The Sum of fourier coefficients times the $e^{i\omega t}$ should be what these circles trace out, if this is not correct explain why please $\endgroup$ – vounoo Nov 1 '17 at 23:01

The Fourier series for a square wave that is $-1$ for $-\pi < t < 0$ and is $1$ for $0 < t < \pi$ is a sine series. This sine series in $t$ is the imaginary part of an exponential series, which is what they are plotting. So, they consider only the sum of the exponentials with positive imaginary argument $e^{int}$, and they look only at the vertical component. They add the exponentials as you would add vectors in $\mathbb{R}^2$, which is the same as complex addition. Then they plot the height, which is the imaginary component, as a function of $t$, and that is the graph on the right. But I guess they are plotting the graph on the right backwards in time, because the earlier values of time are to the right.

  • $\begingroup$ But if the addition of the circles is the way to represent $f$ using its fourier decomposition shouldn't the complex value that results from this addition at each point in time represent $f$ at that time? For another example, youtube.com/watch?v=8Q0NxFt-s7Y what is happening in this animation? Why not take only y here? $\endgroup$ – vounoo Nov 2 '17 at 15:45
  • $\begingroup$ @vounoo : You could add in the other components. Because this is a sin series, you'll have terms $\frac{1}{2i}(e^{int}-e^{-int})$. This results in cancellation of the horizontal components. Then dividing by $2i$ gives a real number. Remember, this is just a visualization to help you understand how the various components are used to build up the final shape. Wheels on wheels. $\endgroup$ – Disintegrating By Parts Nov 2 '17 at 16:27
  • $\begingroup$ @vounoo : The visualization in the second example is a full-scale complex Fourier series, where the graph is a picture in the complex plane. The reference to Ptolemy is Historical. I believe he originated the idea of using wheels on wheels to describe various things, long before the Fourier series. :) $\endgroup$ – Disintegrating By Parts Nov 2 '17 at 16:32
  • $\begingroup$ So you are basically saying that this animation with the square wave should in fact include some circles spinning in the opposite directions and had he done that, the net value of the expansion would oscillate on the imaginary axis thus giving the wave. The animation is simply using only positive frequency terms and this results in a real component that goes away when you start to include the negative frequencies? $\endgroup$ – vounoo Nov 3 '17 at 10:04
  • $\begingroup$ Adding the effect of all these circles and removing the real component is equivalent to adding the negative frequencies of the expansion? $\endgroup$ – vounoo Nov 3 '17 at 10:08

What is being illustrated is behavior in time of a system of epicycles. This is, the equation $x(t)+iy(t)=a_1e^{it}+a_2e^{2it}+\dots$ for some constants $a_1,a_2,\dots$ is a system. Its graph is a curve in a 3 dimensional $(x,y,t)$ space. The projection onto $(x,y)$ is the left side of the animation. The projection onto $(y,t)$ is the right side. We are seeing two projections of the "world line" of a moving system of circle epicycles.


The rotating circles at left are indeed giving you $\cos(x)$, $\cos(2x)$, $\cos(3x)$ and so on. The first circle rotates at a constant rate. The end of the clock hand has an $x$ coordinate of $\sin(t)$ (we don't use the $x$ coordinate) and a $y$ coordinate of $\cos(t)$. The second circle rotates at three times that rate, and has a $x$ coordinate of $\sin(3t)$ and a $y$ coordinate of $\cos(3t)$. The third has an $x$ coordinate of $sin(5t)$ and a $y$ coordinate of $\cos(5t)$. Stacking the circles as they are arranged has the effect of summing the $y$'s.

The vertical line to the right is sweeping at a constant rate, so $x=t$. Replace $t$ with $x$ in all the cosines.

The animation is doing a real-valued Fourier transform as a sum of cosines. Recall that $\cos(x) = \frac{1}{2}e^{ix} +\frac{1}{2}e^{-ix}$, so if you're not yet comfortable with the real Fourier transformation, you can think of it as a complex transform that's symmetric about the origin.

If you had an infinite number of circles, the sum would indeed be that square wave. The animation is showing that as you add cosine terms, the sum approaches the square wave in the limit.

Of course, this is showing the result of the transformation, not the process that obtained it. You're not seeing the computation of $F(n) = \int_0^\pi f(x)\cos n \pi x ~ dx$ that derives the coefficients, you're just seeing the summation of the series that results from it.

  • $\begingroup$ I dont think this is an appropriate answer for what I asked can you please elaborate? $\endgroup$ – vounoo Nov 1 '17 at 22:27
  • $\begingroup$ how are the circles "indeed giving you cos⁡(x), cos(2x)cos⁡(2x), cos(3x)cos⁡(3x) and so on??" $\endgroup$ – vounoo Nov 1 '17 at 22:32
  • $\begingroup$ I'm not sure what you're missing, but I've tried to expand the explanation. It really is showing the summation of a Fourier series - in this case, the result of a discrete cosine transformation. (It's not showing how the coefficients were obtained - I don't know whether that may have been the source of your confusion?) $\endgroup$ – Kevin Kenny Nov 2 '17 at 13:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.