Let an IAR system be defined by the following difference equation:

$$y[n]-\frac{1}{4} y[n-2]=x[n]+3x[n-1]$$

and an input signal $x[n]=(-0.5)^n$.

The transfer function is defined as $H^z(z)=\frac{1+3z^{-1}}{1-0.25 z^{-2}}$. The value $z=-0.5$ is a pole of the system, hence, it is not in the ROC. We cannot use $y[n]=H^z(-0.5)x[n]$ to solve, as we would for any other eigen-signal.

Is there any other way to solve this? What can we say about the output? Is it infinity? is it undefined?


If the input is $x[n]=\left(-\frac12\right)^n$ for all $n$, then the output is infinite, because the convolution sum doesn't converge. Note that also the input grows without bounds for negative $n$.

If, on the other hand, the input starts at $n=0$, i.e., $x[n]=\left(-\frac12\right)^nu[n]$ then you can use the $\mathcal{Z}$-transform to compute the output:


With $$X(z)=\frac{1}{1+\frac12 z^{-1}}\tag{2}$$

you obtain

$$\begin{align}Y(z)&=\frac{1+3z^{-1}}{\left(1+\frac12 z^{-1}\right)^2\left(1-\frac12 z^{-1}\right)}\\&=\frac{a}{\left(1+\frac12 z^{-1}\right)^2}+\frac{b }{1+\frac12 z^{-1}}+\frac{c}{1-\frac12 z^{-1}}\tag{3}\end{align}$$

with some real-valued constants $a$, $b$, and $c$.

Inverse $\mathcal{Z}$-transform of $(3)$ gives


EDIT: First, to clarify, I use a basic example that shows that a double pole in the $\mathcal{Z}$-transform is no problem. Assume a system with impulse response $h[n]=(1/2)^nu[n]$ and an input signal $x[n]=(1/2)^nu[n]$. Computing the output in the time domain is straightforward:


The $\mathcal{Z}$-transform of the output sequence is simply given by the multiplication of the $\mathcal{Z}$-transforms $H(z)$ and $X(z)$:


The inverse $\mathcal{Z}$-transform of $(6)$ equals the result $(5)$, which is very easy to confirm.

The probable cause of the OP's confusion could be the fact that for an input signal $x[n]=z_0^n$, $-\infty<n<\infty$, the output is given by


if and only if $z_0$ is inside the ROC of $H(z)$, because otherwise the convolution sum doesn't converge and the expression $H(z_0)$ is meaningless. However, in the computation of the outputs $y[n]$ in the examples above we never evaluate $H(z)$ for any $z$ outside the ROC, so the situation is entirely different.

  • $\begingroup$ I think your solution for $x[n]=(-0.5)^nu[n]$ is incorrect. How are you using $H^z(z)$ for $z\notin ROC$? $\endgroup$
    – havakok
    Jan 7 '21 at 12:53
  • $\begingroup$ You did not evaluate it. You say "of course you can use it to multiply it with the Z-transform of the input". I disagree. Can you support your claim? $\endgroup$
    – havakok
    Jan 7 '21 at 13:56
  • $\begingroup$ Yes, that fact is well known in the general case and I want you to support that it can also be done in the specific case of an eigen-signal which is a pole of the transfer function. A book or a paper that supports this delicate point or mathematical reasoning. $\endgroup$
    – havakok
    Jan 7 '21 at 14:05

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.