I would just want to prove $\cos(t) + \cos(\pi t)$ is non periodic.

I don't know where to start it. Also I know that individually these signals ie $\cos(t)$ and $\cos(\pi t)$ are periodic with frequencies $1/(2 \pi)$ Hz and $1/2$ Hz respectively. But since the superpositon of these signals is not periodic should we represent them using Fourier transform only? What is the harm in drawing the frequency components at $1/(2 \pi)$ Hz and $1/2$ Hz using fourier series?

  • 1
    $\begingroup$ Fourier series are for signals that are periodic and your signal is not periodic, as Aaron's answer points out. I don't know what "drawing" the frequency components means, but it is perfectly fine to write the signal as $\cos(t) + \cos(\pi t)$ which is not a Fourier series, and to "draw" the signal as having impulses (or vertical lines) at $f = \pm (1/2\pi)$ and $f = \pm (1/2)$ if that is what you are looking for. $\endgroup$ Jan 1, 2014 at 7:13
  • $\begingroup$ The answer is as usual 42, sorry, no, 44. Try it out, use your favorite plotting program to plot $f(x)$ against $f(x+44)$ ;-) (This is an extension of the famous $\sin(22)$ test, resp. the pi approximation of Archimedes.) $\endgroup$ Jan 1, 2014 at 12:04
  • $\begingroup$ @LutzL : Are you speaking about periodicity. The signal is not periodic. $\endgroup$
    – dexterdev
    Jan 1, 2014 at 14:48
  • $\begingroup$ Of course not. But the difference $f(x+44)-f(x)$ has size $2\sin(22-7\pi)<0.018$, which is mostly invisible in function plots. Even smaller is the difference in $f(x+710)-f(x)$, bounded by $2\sin(355-113\pi)<0.0000603$. $\endgroup$ Jan 1, 2014 at 15:07
  • $\begingroup$ @LutzL : Actually I was trying to look the theoretical aspects and I trusted software called 'GRAPH' for that matter in plotting. Now that is lesson for me. $\endgroup$
    – dexterdev
    Jan 2, 2014 at 5:41

1 Answer 1


$cos(t)$ and $cos(\pi t)$ are both periodic with periods $2\pi$ and $2$ respectively. To find the period of the sum we need to find an integers $n,m$ such that $\pi n = 2 m$ or $\pi / 2 = m /n $, which is not possible since $\pi$ is irrational.

I think the Fourier series uses specific frequencies to form the signal. You can't necessarily put a frequency component at irrational frequencies.

Fourier series work fine with irrational frequencies too, but the frequencies are all harmonics, that is, integer multiples, of the fundamental frequency.

  • 1
    $\begingroup$ The last sentence is incorrect. Fourier series work just fine with irrational frequencies, but the frequencies are all harmonics, that is, integer multiples, of the fundamental frequency $f_0$ (or $\omega_0$ depending on which flavor of Fourier series one prefers). But notice that since $\omega_0 = 2\pi f_0$ and $\pi$ is irrational, if one set of folks are using a rational fundamental frequency, the other set of folks are using irrational fundamental frequency! $\endgroup$ Jan 1, 2014 at 7:08
  • $\begingroup$ @Aaron : I have edited your answer regarding the point which Dilip Sarwate mentioned. $\endgroup$
    – dexterdev
    Jan 1, 2014 at 8:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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