Suppose you have an LTI system for which the (complex) frequency response $H(j\omega)$ has been measured in some frequency window $[\omega_1,\omega_2]$. Now imagine that you want to provide an input $x(t)$ to this system, such that the output is a specific, known waveform $y(t)$ that has negligible frequency content outside $[\omega_1,\omega_2]$.

My question is:

What are the necessary and sufficient conditions that our measured $H(j\omega)$ values in $[\omega_1,\omega_2]$ need to satisfy such that the required input $X(j\omega) = Y(j\omega)/H(j\omega)$ exists, and makes practical sense? For example, the system can clearly not have any zeros on the $j\omega$ axis in our frequency window.

I realize that a sufficient condition is for the system to be invertible. However, no such knowledge of the system is available, as we only have access to the frequency response in a specific frequency interval. What can we say about the possibility of backing out the frequency response only with the information at hand?

  • $\begingroup$ This sounds exactly like you're building an equalizer for a channel that you can't know before observing it. $\endgroup$ Dec 3, 2019 at 22:55
  • $\begingroup$ @MarcusMüller Even though having an equalizer would solve the problem, wouldn't it be overkill? Since my question is about equalizing a specific waveform. Cascading the system with an equalizer would compensate the effect of the transfer function for any input signal. I just want to find the input that results in one particular signal to appear at the output of the system (if it exists). $\endgroup$ Dec 4, 2019 at 2:50
  • $\begingroup$ To me, something that flattens the channel wherever necessary is an equalizer (that term imho doesn't require any all-frequencies behaviour, anyway, and that would be impossible to begin with – there can't be an infinitely wide equalizer), and it's a problem that is solved with equalizer in communications theory all the time: Divide observable spectrum into one or multiple bands of interest, and revert the channel on that/these individually, with only the goal of reverting the channel in all respects that are detrimential to your measurement/estimation problem. Exactly modern wideband comms! $\endgroup$ Dec 4, 2019 at 8:30


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.