1
$\begingroup$

I am trying to understand how the output behaves when the input and the system function do not have a common region of convergence (ROC) for an LSI system.

Consider an LSI system with $x[n]$, $h[n]$, $y[n]$ as input, impulse response, output signal respectively: We know that \begin{gather} Y(z) = X(z)\cdot H(z)\\[5pt] \text{with } \ ROC(Y) \supseteq ROC(X) \ \cap \ ROC(H) \end{gather} So what if $ROC(X) \ \cap \ ROC(H)=\phi$, what would be the output? Is it even properly defined?

I tried with some examples considering $x[n]$ and $h[n]$ with disjoint ROCs, the output is not defined, here is one example: \begin{gather} x[n] = (1/4)^nu[-n-1] \\[5pt] h[n] = (1/2)^nu[n]\\[5pt] ROC(X): |z|<1/4\\[5pt] ROC(H): |z|>1/2\\[10pt] y[n] = \sum\limits_{k=-\infty}^{\infty} (1/4)^ku[-k-1](1/2)^{n-k}u[n-k]\\[5pt] y[n]= (1/2)^n\sum\limits_{k=-\infty}^{-1} (1/2)^ku[n-k]\\[5pt] y[n] = (1/2)^n\sum\limits_{k=1}^{\infty} 2^ku[n+k]\\[5pt] \text{Clearly } y[n] \text{ is not defined} \end{gather} My question is whether this is true in every case that the output would not be defined or there are some special $x[n]$ and $h[n]$ with disjoint ROCs where $y[n]$ is properly defined.

$\endgroup$
0
$\begingroup$

If the ROCs don't overlap, the corresponding convolution sum doesn't converge, hence the output is undefined. Note that in your question it is assumed that the $\mathcal{Z}$-transforms of the input as well as of the system's impulse response both exist.

You can construct examples for which neither the $\mathcal{Z}$-transform of the input nor of the output exist, yet the input and output signals are perfectly defined. E.g., the input sequence $x[n]=A\cos(\omega_0n)$ has no $\mathcal{Z}$-transform, neither has the corresponding output sequence $y[n]=B\cos(\omega_0n+\phi)$, where $B$ and $\phi$ depend on the system's frequency response. [In this example it is assumed that the ROC of the system's transfer function $H(z)$ includes the unit circle, i.e. the system is stable.]

$\endgroup$
1
  • $\begingroup$ Oh right, If Z-transform doesn't exist we might consider it as something with a null region of convergence. Existence of Z-transforms is assumed, Thanks! $\endgroup$ – Aang Oct 25 '20 at 13:56

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.