Suppose I have the following equation, where $x[n]$ is the input and $y[n]$ is the output: $$x[n]-x[n-1]=-10y[n-1]+2y[n]+2y[n+1].$$

The question asks me to compute the impulse response from this equation.

Attempts: $x[n]=1$ when $n=0$; $0$ otherwise. I need to find $h[n]=y[n]$, but, from the equation, it always depends on the future value $y[n+1]$, which is unknown.


1 Answer 1


One way to solve such a difference equation is to use the $\mathcal{Z}$-transform. Transforming the given difference equation yields


which results in the following transfer function:


The poles of $H(z)$, i.e. the zeros of the denominator polynomial in $(1)$, are both real-valued:


with $a=\frac{-1+\sqrt{21}}{2}\approx 1.79$ and $b=\frac{-1-\sqrt{21}}{2}\approx -2.79$. Note that $1<|a|<|b|$ is satisfied, so both real-valued poles are outside the unit circle of the complex plane.

With $(2)$, $H(z)$ can be rewritten as


This transfer function can have three possible regions of convergence (ROCs), each of them corresponding to a different impulse response:

$$\begin{align}(1)\quad &|z|>|b|>|a|\\ (2)\quad & |b|>|z|>|a|\\ (3)\quad & |b|>|a|>|z|\end{align}$$

ROC $(1)$ corresponds to a right-sided impulse response, ROC $(2)$ corresponds to a two-sided impulse response, and ROC $(3)$ corresponds to a left-sided sequence. Only ROC $(3)$ includes the unit circle $|z|=1$, so only the impulse response corresponding to this ROC results in a stable, yet anti-causal, system. The three impulse responses corresponding to the three ROCs given above are

$$\begin{align}(1)\quad & h_1[n]=\frac{1}{2(b-a)}\left((1-a)a^{n-1}-(1-b)b^{n-1}\right)u[n-1]\\ (2)\quad & h_2[n]=\frac{1}{2(b-a)}\left((1-a)a^{n-1}u[n-1]+(1-b)b^{n-1}u[-n]\right)\\ (3)\quad & h_3[n]=\frac{1}{2(b-a)}\left(-(1-a)a^{n-1}+(1-b)b^{n-1}\right)u[-n]\end{align}$$

These impulse responses describe three different systems, all of them characterized by the same transfer function given by Eq. $(3)$. $h_1[n]$ corresponds to a causal unstable system, $h_2[n]$ to a non-causal unstable system, and $h_3[n]$ to an anti-causal stable system.

  • $\begingroup$ Actually the question does not state that this system is non-causal, it is my assumption based on the equation. The equation is derived from a 2nd order ODE in continous time, then mapping to DT using Euler approximation. I think your answer might apply in this case $\endgroup$
    – TonyJ
    Nov 11, 2015 at 11:58
  • 1
    $\begingroup$ @TonyJ: But the causal impulse response is not stable, so it's not useful in practice. $\endgroup$
    – Matt L.
    Nov 11, 2015 at 12:45

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.