Consider a discrete time LTI causal system $S: y = S(u)$, with its impulse response $h:{Z} \rightarrow R:h(n)=3^{n+1}{H(n)}$ with $H$ the Heaviside function. We know the values of input:

$$u(0) = 2$$ $$u(1) = 1$$ $$u(2) = 1$$ $$u(3) = 1$$ $$u(4) = 2$$

Also the $y(2) = 66$. I'm trying to find the value of $y(3)$.

I may say, for a LTI, causal system the output is the convolution of the input with the impulse response:

$$y(n) = \sum^\infty_{k=0}3^{k+1}u(n-k)$$


$$y(3) = 3u(3)+9u(2)+27u(1)+81u(0) = 201$$

Which is the correct answer! But I don't understand, why should I stop at $u(0)$ or why the negative values of the input $u(n), n< 0$ are zero here ?

Alternative solution:

recursion gives us this form:

$$y(n) = 3y(n-1)+3u(n)$$

  • 1
    $\begingroup$ I think that usually for such questions, the input for negative time values is defined as zero. $\endgroup$ Dec 15, 2019 at 7:51

1 Answer 1


Unless explicitly stated, you shall not assume that the input is zero for $n<0$, hence the convolution sum should be

$$y(n) = \sum^\infty_{k=-\infty}3^{k+1}u(n-k)=\sum^\infty_{k=0}3^{k+1}u(n-k).$$

In this situation, without making any assumption on $u[n]$ for $n<0$ you can still compute $y[n]$ for $n \geq 0$ based on the recursive relation as you have outlined :

$$y(n) = 3y(n-1)+3u(n) \implies y[3] = 3y[2] + 3u[3] = 201$$

What can be said about $u[n]$ for $n<0$ ? Actually nothing but, given that $y[2] = 66$ we may try the folowing :

$$ \begin{align} y[2] &= \sum^2_{k=0}3^{k+1}u(2-k) + \sum^\infty_{k=3}3^{k+1}u(2-k) \\ &= 66 + \sum^\infty_{k=3}3^{k+1}u(2-k) \\ 66 &= 66 + \sum^\infty_{k=3}3^{k+1}u(2-k) \\ 0 &= \sum^\infty_{k=3}3^{k+1}u(2-k) \end{align} $$

That the sum in the parenthesis equals zero, has a trivial solution of $x[n] = 0$ for $ n <0$.

However, we may also ask whether another, nonzero, solution is also possible that the sum be zero? Consider, for example, the input $$u[n] = (-1)^{2-n} ~~3^{n-3}$$ for $n<0$ which yields for the sum

$$\sum^\infty_{k=3}3^{k+1}u(2-k) = 1-1+1-1+1-...$$

This is an alternating sequence and it does not converge. However, if it's possible to restrict the length of the input to be even for $n<0$, then the sum will be zero too...

Nevertheless, in order to uniquely specify $u[n]$ for $n<0$ a single sum is not enough. We must know values of output $y[n]$ for all $n<0$ for that. Hence you cannot conlcude that $x[n]=0$ for $n<0$ under the given situation.


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.