Sound installation feedback prevention

[sndpl]

Hello all,

I just got an axoloti recently for an installation project I was asked to help with on technical set up. Unfortunately I didn't have much time to get to learn the axoloti before having to do the installation, so it was a matter of just getting the essentials up and running and now the board is already installed there while I would still like to make some tweaks, specifically in preventing feedback.

It is a touchy installation with a large and fairly live room mic'ed up and playing back the room into the space through an array of various small speakers. The sound is going through the axoloti first on a delay line and granulation effect, going for somewhat of a Lucier type thing. While the speakers are pretty small and I have tuned volume levels as best I can to prevent feedback from happening it is of course still possible. While I have instructions for the gallery technicians if there is a problem of how to reset the audio if it does happen, with the installation running for 6 weeks I'd like to try and be able to automate some sort of feedback prevention, but I haven't had the time to learn the axoloti that well and was hoping some more experienced users here might have some tips...

My idea was to do a pretty crude implementation of just trying to shut the audio off either at the input or output automatically for 6-8 seconds (the time it takes for the feedback to dissipate when it occurs) every 3-5 minutes. It would be long enough and irregular enough that no visitors would notice and in case some major feedback did errupt somehow it wouldn't last more than a couple minutes before the cycle would automatically shut off and reset. Simple but effective and probably much easier than configuring some sort of reverse noise gate tuned for high frequencies or something like that. Going through all the objects I'm still a bit confused how to best do this...

Obviously I need some sort of VCA or gate, but to open and close it at the predetermined times I am finding tricky since I don't know all the objects and what they want signal wise yet... Unfortunately it seems that the LFOs don't actually run all that slow on the axoloti unless there is a way of making them slower than the knobs go. My idea was to basically just set a slow square LFO cycling at around 3-5 minutes like I need, set a very thin pulsewidth to equal around 6-8 seconds, invert the value and use that to control a VCA... but without a slow enough pulse width adjustable LFO that idea goes out the window.

Any recommendations or thoughts? Thanks!

[lokki]

while this is possible (i made a very slow attack decay module that you could use) i would opt for a different approach.

why don't you measure the volume of the incoming audio, and if it crosses a certain threshold you lower the input volume, so that less sound comes in. you could even mute the input completely for a couple of seconds. that way the output would still be continuous...

[sndpl]

part of the problem is that the room sound is very dynamic - depending on what is going on in the space, and how close someone might be to the microphones can have a big influence on the volume. Someone shuffling their feet close by for a few seconds could cause a large jump in volume but with no feedback while someone standing across the room shouting for a split second at a higher pitch could cause feedback. The volume control would have to be very dynamic- being based on both volume and frequency content as well as time since a short/fast spike in volume might be no problem, but high volume maintained for several seconds or longer might be nothing but feedback or might also just be temporary high sound activity in the room. Being able to do all this would be the ideal scenario but seems way too complicated for this particular instance, especially considering I can't sit at home experimenting with the board anymore.

[lokki]

i see. i will see if i can program you such a slow "lfo" that is high for only a couple of seconds every 5 minutes.

i would still try to cut some of the highs from the input (most likely the incoming "natural" sounds should be within a range of 20 to 8000 hz) and maybe also the outgoing sounds can be cut. then additionally you could shut the input off if very high frequencies get into the system, see attached image:

[Blindsmyth]

I think what feedback destroyer usually do is cut the input into lots of frequency bands, and if you one band raises a threshhold it will be drastically put down.

Propably you could use the vocoder objects by SSS for this:

[lokki]

here you go, a patch with included watch that does what you want. an envelope that goes low every 4.something minutes for 6-7 seconds.

downtime.axp (4.0 KB)

[sndpl]

thanks, this super helpful!