I have $x_1(t)$ here. To get $x_2(t)$, I need to differentiate $x_1(t)$. Express $x_2(t)$ as $2u(t+2)-4u(t)+2u(t-2)$.
From Fourier transform definition integral, I got $X_2(j\omega)=\frac{2e^{j\omega 2}}{j\omega}-\frac{4\pi \delta(\omega)}{j\omega}+\frac{2e^{-j\omega2}}{j\omega}$.
Is this correct? It seems weird and would be complicated to calculate the magnitude and phase spectrum. Thanks! enter image description here

  • $\begingroup$ Simplifying it, I got $X_2(j\omega) = \frac{4\cos(2\omega)}{j\omega} - \frac{4\pi \delta(\omega)}{j\omega}$ $\endgroup$
    – keanehui
    Apr 5, 2020 at 8:42
  • $\begingroup$ You are on the right path. Since $u(t)$ is not a stable function, it's fourier transform will have discontinuity. $\endgroup$
    – jithin
    Apr 5, 2020 at 9:49
  • 1
    $\begingroup$ @jithin that's not right: the rect function (not stable) has the Fourier transform sinc, and that is as continuous as any function will ever get! $\endgroup$ Apr 5, 2020 at 12:21
  • 1
    $\begingroup$ @MarcusMüller Pssssssst, not all functions have plugged discontinuities. $\endgroup$ Apr 5, 2020 at 12:27
  • $\begingroup$ @CedronDawg :) but really, the rect hint is all I'm willing to give here – I must have tried to derive the Fourier transform of the Heaviside function so many times that I forgot how to do it right, because because I don't get the (ugly) right result when I try to do it again. Keanehul, $u(t)$ is not the function you're looking for, honestly! $\endgroup$ Apr 5, 2020 at 12:33

1 Answer 1


It looks like the idea of that exercise is to compute the Fourier transform of $x_3(t)$, and then from that derive the Fourier transform of the original function $x_1(t)$.

Note that $x_3(t)$ is just a sum of scaled and shifted Dirac impulses, so computing its Fourier transform is trivial. The Fourier transform of $x_2(t)$ is then easily derived by multiplying $X_3(j\omega)$ by $\pi\delta(\omega)+1/j\omega$, and noting that the Dirac delta is cancelled because $X_3(0)=0$. So you just end up dividing by $j\omega$. In the same way, $X_1(j\omega)$ is obtained from $X_2(j\omega)$. Also note that none of the three Fourier transforms $X_1(j\omega)$, $X_2(j\omega)$, and $X_3(j\omega)$ have any Dirac impulses.

It's a good idea to cross-check your result by realizing that $x_1(t)$ can also be represented as the convolution of two rectangular functions, i.e., its Fourier transform must be a squared sinc-function.

  • $\begingroup$ Thanks! And is it possible to find $X_2(j\omega)$ from only $x_1(t)$ or $x_2(t)$ using the Fourier transform table, which is a hint from the exercise? $\endgroup$
    – keanehui
    Apr 6, 2020 at 3:49
  • $\begingroup$ @keanehui: Sure, you can find $X_2(j\omega)$ from $x_2(t)$ if you know the transform of $u(t)$ and if you know the frequency domain consequence of shifting in the time domain. $\endgroup$
    – Matt L.
    Apr 6, 2020 at 7:53

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.