intelligent, automated feedback death?
I’m going to be playing and doing live processing in a large outdoor festival this summer, and there will be no sound check. (The Newport Folk Festival–hopefully Pete Seeger won’t come out with an axe and whack my computer (actually, that’s not really fair, because the legend is apparently false, and besides, it would actually be an honor to have Pete S. trash my computer).) I can’t afford to take any chances on something happening such as happened to me the other day, where too-hot monitor levels + overall house level (which happened to not be the same as the sound check) caused significant feedback problems. I’m also going to be actively playing instruments, so I won’t be able to diddle w. the computer if something starts going south (although I suppose I could control levels with a foot pedal). My question is this: given the above scenario, does anyone have suggestions about intelligent, automated feedback killing? (I’ve got ideas, but the proviso that they be intelligent is problematic.) Ideally the gain structure would be solid enough to prevent such problems before they occur, but I want to be prepared for unpredictable worst case scenarios. A fallback would be for me to use the aforementioned pedal to attenuate the gain on input signals, should I hear any problems. For any ideas which will help me do better than that, I will be eternally in your debt!
Also, and this is semi-related, I have dim memories of a kind of "I Am Sitting In A Room" patch tucked away in some folder of the distro (in days of yore), perhaps even w. some kind of feedback detection in place. Is this memory a byproduct of my copious crack intake at the time, or is it out there somewhere? Couldn’t find it.
I think omx.peaklim~ is a good place to start. Put this across your
input, then decrease the gain, do your processing, and peak-limit the
output. The most important things are controlling the input and output
gain. A multiband limiter may also be very helpful at the output.
minmax~ might also be helpful at scaling the output into a safe range.
Something I’ve found useful in (albeit internal) feedback situations
are custom compressors with a negative slope at high volumes (and a
slow release at these values). I’ve used these in a matrix~ 8 8 with
live input into effects where every inlet is connected to every outlet,
and it’s very safe. I had an older version using the threshold~ object
from the Jimmies, but you could probably reconstruct something similar.
Basically if you go over a certain threshold, you want to quickly ramp
down the gain, and only slowly increase it back up to it’s last level.
(maybe every time it’s over, you could have it subtract a few dB even)
I think a lot of this depends on the algorithms you’re using. If it’s
a feedback delay, then there are definitely steps that you can take to
help with this. I made a post a week or so ago about some things to
help with that. I would be very careful of feedback delays with high
feedback coefficients, filters with high Q’s, and long reverbs. It’s
important to have ways of scaling these coefficients on the fly if
necessary. (this is what bondo and * 1. are for!)
Test it the patch at home with your microphone positioned to pick up
strong interference from your speakers. If it stabilizes there, it’ll
probably work in the stadium. (caveat: NO DRY MIC SOUND from the
mixer when trying this…)