When finding the overall system impulse,

how to do arithmetic with $\delta$ functions?

E.g. \begin{align} h[n]&=h_1[n]\star h_2[n]\\ &=\sum_{k=-\infty}^{\infty} h_1[k]h_2[n-k]\\ &=\sum_{k=-\infty}^{\infty} \left(2\delta[k-2]-3\delta[k+1])(\delta[n-k-1]+2 \delta[n-k+2]\right)\\ &=\sum_{k=-\infty}^{\infty} (2\delta[k-2]\delta[n-k-1]+2\delta[k-2]2 \delta[n-k+2]-3\delta[k+1]\delta[n-k-1]-3\delta[k+1]2 \delta[n-k+2]) \end{align}

The two middle terms are convolutions. But what about the two other? What to do with them?

Or am I supposed to leave the indices there, rather than evaluating the $\delta$ functions?

  • $\begingroup$ you are sure they are multiplied and not convolved, right ? $\endgroup$ – Fat32 Mar 30 '17 at 14:50
  • $\begingroup$ Here it's multiplication, not convolution. But the above expression was found through convolution of the signals in $()()$. $\endgroup$ – mavavilj Mar 30 '17 at 14:54
  • $\begingroup$ You shall better formulate the convolution first, it seems there is an unconventional use of the variables n and k. $\endgroup$ – Fat32 Mar 30 '17 at 15:03
  • $\begingroup$ That seems to be the convolution between two sums of two deltas. What about doing the operation in the Fourier domain, and then coming back? $\endgroup$ – Tendero Mar 30 '17 at 15:05
  • 1
    $\begingroup$ @Tendero I haven't been taught Fourier transforms yet. $\endgroup$ – mavavilj Mar 30 '17 at 15:07

You make it unnecessarily complicated.

Given $$\begin{align} h_1[n]&=2\delta[n-2]-3\delta[n+1]\\ h_2[n]&=\delta[n-1]+2 \delta[n+2] \end{align}$$ to calculate $$h[n]=h_1[n]\star h_2[n]$$

the easiest way is to use the sampling property of delta, which is

$$f[n]\star\delta(n\pm n_0)=f[n\pm n_0]$$

So you can write as follows

$$\begin{align} h_1[n]\star h_2[n] &= h_1[n]\star \left(\delta[n-1]+2 \delta[n+2]\right)\\ &=\color{green}{h_1[n-1]}+2\color{blue}{h_1[n+2]}\\ &=\color{green}{2\delta[(n-1)-2]-3\delta[(n-1)+1]}+2\color{blue}{\left(2\delta[(n+2)-2]-3\delta[(n+2)+1]\right)}\\ &=2\delta[n-3]-3\delta[n]+4\delta[n]-6\delta[n+3] \end{align}$$


Using the sifting property, $$ x[n] \delta[n-m] = x[m]\delta[n-m] $$

specifically applied when $x[n]$ is $\delta[n-k]$ where $k$ is an integer, $$ \delta[n-k] \delta[n-m] = \delta[m-k]\delta[n-m] $$

which is equivalent to $\delta[n-m]$ when $k=m$ and otherwise identically $0$, when $m \neq k$ .

You can apply the result term by term. Give it a try and tell us your result.

Hint: (from your post) $$h_1[n] = 2\delta[n-2] - 3\delta[n+1]$$ $$h_2[n] = \delta[n-1] + 2\delta[n+2]$$ $$h[n]=(h_1[n]*h_2[n])$$

I suggest you at this point (for these very short impulse responses) to use the distribution property of convolution and the convolution property of the $\delta[n]$.

i.e. $$ (x_1[n] + x_2[n]) \star h[n] = x_1[n] \star h[n] + x_2[n] \star h[n] $$

  • 1
    $\begingroup$ I don't see how the distribution property applies here. Or does $\delta$ also have convolution property? $\endgroup$ – mavavilj Mar 30 '17 at 16:01
  • $\begingroup$ if you look at my last line, you can see the distrbution of convolution over addition there. Then you shall replace $x_1[n]$, $x_2[n]$ and $h[n]$ with appropriate signals. Then you will apply convolution property of impulse to obtain the final result. $\endgroup$ – Fat32 Mar 30 '17 at 21:12

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.