0
$\begingroup$

If I have $$x[n] = u[n] - u[n-4]$$ where $u[n]$ is the unit step function, and $$h[n] = e^{-i\frac{\pi}{2}n}$$

does $x[n] * h[n] = 0$?

I tried doing the convolution sum and I got: $1 - i - 1 + i = 0$ and I also just tried plotting points and think I got 0 also, but I'm not sure if I did either of these correctly. Is there any easier or intuitive way to think about convolving a discrete complex exponential with a function, or (if it's easier the other way around), convolving a discrete box with another function?

$\endgroup$
2
$\begingroup$

The discrete-time Fourier transform (DTFT) of a length $N$ "discrete box" sequence starting at $n=0$ is

$$\sum_{n=0}^{N-1}e^{-jn\omega}=e^{-j(N-1)\omega/2}\frac{\sin\left(\frac{N\omega}{2}\right)}{\sin\left(\frac{\omega}{2}\right)}\tag{1}$$

The numerator of $(1)$ has zeros at frequencies

$$\omega_{0,k}=\frac{2k\pi}{N}\tag{2}$$

You can interpret the convolution as filtering the signal $x[n]=e^{-jn\pi /2}$ with a filter with its frequency response given by $(1)$. With $N=4$ you get zeros of the frequency response at $\omega_{0,k}=k\pi /2$, so the complex exponential with frequency $\omega=-\pi/2$ is completely suppressed because the filter's frequency response has a zero at exactly that frequency.

$\endgroup$
  • $\begingroup$ Thanks very much. Is the exponent in the second half of the first equation missing the frequency component? $\endgroup$ – Austin Sep 12 '16 at 15:01
  • $\begingroup$ @Jake: That's right, I corrected it. $\endgroup$ – Matt L. Sep 12 '16 at 15:33
2
$\begingroup$

You can use the convolution theorem to convert the convolution to the product of Fourier transforms of the given signals. $$h[n]*x[n]=\mathcal{F}^{-1}\{X(\omega)H(\omega)\}$$

$$H(\omega)=\mathcal{F}\{h[n]\}=\delta(\omega-\frac{\pi}{2})$$ $$X(\omega)=\mathcal{F}\{x[n]\}=e^{-j\frac{3 \omega}{2}}X_r(\omega)$$ where $$X_r(\omega)=\left\{\begin{matrix} \frac{\sin(2\omega)}{\sin(\omega/2)} &\omega\neq 0 \\ 4 & \omega=0 \end{matrix}\right.$$

However since $X(\omega)\delta(\omega-\frac{\pi}{2})=X(\frac{\pi}{2})\delta(\omega-\frac{\pi}{2})$ and $X(\frac{\pi}{2})=e^{-j\frac{3\pi}{4}}\frac{\sin(2\pi/2)}{\sin(\pi/4)}=0$, the inverse Fourier transform and therefore the covolution is zero.

You can also assume $x[n]$ is the impulse response of an LTI system and $h[n]$ is an input signal. Since the complex exponential is eigenfunction of the system, the output whould be an scaled version of the input i.e. $Ah[n]$ where $A=X(\frac{\pi}{2})$ and $X$ is the FT of $x[n]$.

$\endgroup$
  • $\begingroup$ The result of the convolution is indeed zero. The mistake is in the length of the discrete box. You used a length of $3$ samples, but it is actually $4$ samples. $\endgroup$ – Matt L. Sep 12 '16 at 7:24

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.