How do feedback eliminators work? Can they be time domain?

Question

Firstly, the reason I need a feedback destroyer is for playing open mics. I'm not in control of soundcheck, nor do I have time to fiddle with it.

While searching for an anti-feedback VST and found a forum link talking about something called the beringer shark. The way it functions is described like this:

For the uninitiated, Behringer's Feedback Destroyer is a series of 12 very, very narrow notch filters that scan the audio for huge peaks, lock onto them and notch them away. They are very dynamic so when filter 12 gets used, filter 1 starts scanning again. I usually never get more than 8 locked to get rid of all my feedback problems.

I have the means to program my own audio tools, but I'd like to get a better idea of how this works (or how a feedback eliminator could work).

I would have thought that it would be easiest to cut out feedback rings in frequency domain (just search for the out of place super loud ring and reduce those bins). But frequency domain is typically a bit slow and never(?) analog, so that might explain the "scan and notch" technique.

My guess on how the scan and notch technique works is that there's a series of notch-filters with parallel (but silent) band-pass filters. The notch and band pairs move around (how fast?) until the band peaks, at which point both stop. The notch is then mixed in to some level relative to how loud the band is.

This all seems really computationally intensive, so I suspect there's a better way.

Answer 1

It doesn’t scan, per se, it waits for a peak to form, as in a real time analyzer or even a chromatic tuner. Just as the tuner can tell what is the loudest frequency in a sound, so does the feedback eliminator detect the loudest frequency. Then it creates and applies a notch filter for that frequency.

Side note: I suggest comparing brands. Some brands are far less expensive and are also less effective. If you pay a bit more, you’ll get much better performance.