I have two discrete-time LTI systems whose transfer functions satisfy $h_1[n] * h_2[n]= \delta[n]$. We also know that system 1 is causal and stable. Does $h_1[n] * h_2[n]= \delta[n]$ tell us anything about $h_2$?

Intuitively, I would say that $h_2[n]$ is stable, but I don't have any idea how to reason about it.


2 Answers 2


It tells us that the systems are inverses of each other. The DFT of

$$h_1[n]*h_2[n]= \delta[n]$$


$$H_1[k] \cdot H_2[k] = 1 $$

so we get

$$H_2[k] = \frac{1}{H_1[k]}, H_1[k] = \frac{1}{H_2[k]}$$

In order for $h_2[n]$ to be causal and stable, $h_1[n]$ has to be minimum phase. Causality and stability of $h_1$ are not sufficient to guarantee causality and stability of the inverse.

A good example of this is an all pass filter. It's perfectly causal and stable but its inverse is anti-causal, i.e. $h_{AP,inv}[n] = h_{AP}[-n]$


Well Hilmar is 100% correct but I want to give a practical application of 2 such systems.

If 2 systems are inverse of each other you can use them to make a noise removing system:

The block with $H_{1}(z)$ is usually a sensor which measures the main frequencies of the input signal sends them to a PID , then the PID using software sets the resonant frequency.


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.