Originally Posted by jwatte
Yes, I give you that. And my hat off to you for managing to actually do that in real life in one instance! It's not something I'd even try -- along with climbing really tall mountains, running a Marathon, etc. It CAN be done...
I realize now that I may be thoroughly behind the times in trying to do this correction manually, using frequency response flatness. I would try JohnPM's very excellent Room EQ Wizard for finding parametric filter settings. I invariably use it to collect impulse responses for further analysis, for which it does a superior job.
A minimum phase system has the wonderful property that any correction (valid inverse filter) applied in the frequency domain will also automatically correct the time domain -- and vice-versa. Parametric equalizers are composed of second order filter sections which are necessarily minimum phase.
After studying some of the literature on minimum vs. non-minimum phase responses of rooms, I have learned a lot. For one thing, there is a non-minimum phase portion of a room response, which has been found to be due to the introduction of zeros in the room transfer function. For non-geeks, zeros create resonant dips, whereas poles create resonant peaks.
So any attempt to smooth out resonant dips with a parametric equalizer will not address the non-
minimum phase room component, and is probably destined to failure for time domain correction. Of course, any boost in a parametric equalizer is not a good idea, but now I better understand some of the mathematical complications. You must ignore the dips even when finding the inverse filter for the peaks, because any such atttempt can lead you down the unsuccessfu non-minimum phase correction path.
I have since done a bit of my own research using a signal processing transformation called the "complex cepstrum," which has the amazing ability to tease out minimum phase from non-minimum phase signal components. I was able to confirm earlier published research about non-minimum phase room conditions.
Sorry about the technicality of this discussion, but it may provide some useful insights for those who understand it.