Sound installation feedback problem

[Sutchi]

I am setting up a sound installation in a room approximately 4m x 3m. The purpose of the installation is to pick up the voices of people in the space and play them back with either a distortion on the voice or a delay. I am of course running into feedback problems. Is there a way around this, I am very flexible on variations to the set up. Any help would be gratefully received.

[Wonks]

Your main issue will normally be in distorting the voice rather than simple delay.

Are you using an all -analogue system mics >mixer>insert effects/aux effects > power amp> speakers, or a digital mixer(and using its built-in effects), or an audio interface into a computer running processing software.

How is the sound being distorted? If it's through a guitar-type distortion pedal (or a software emulation of one) , then the signal is given a lot of gain, which is then being fed through a stage which clips the signal to provide the distortion. It's all that gain which will create the feedback. You need to keep the overall gain of the system down below 1/unity to prevent feedback.

Easiest for us if you try and describe the current set-up in as much detail as you can.

[Sam Spoons]

Also tell us where the Mics and speakers are relative yo the 'people in the room'.

[ef37a]

Now this is going to sound silly but, bear with...( big apps' Miranda)

You have a tape tech in a box and he opens a switch that allows a mic channel into a recorder and records 'a' voice. He then MUTES the mic and, quick as a flash, plays the recorded voice out through some speakers adding the while, delay or distortion to the signal. Feedback cannot occur because the record and replay paths are never open at the same time. This is akin to the system used between control rooms and studios to prevent talk back/monitor feedback.

All you need to do now is get someone to code a computer to do it! (I suspect MIDI might be useful at some point in the proceedings?)

Dave.

[James Perrett]

Are you doing anything to prevent the mics from picking up the speaker output? Modern phones have built in echo cancellation when you use them on the speaker phone setting so you probably need to find something similar for your setup. Delaying the audio would also help but you would probably still end up with some kind of feedback loop eventually.

[ef37a]

D**** you Sutchi!

You have got me thinking about this! A guitar looper pedal might be a good start as the recording device? Cheap enough that you can get in there and 'bend' it to your will.

Dave.

[Sutchi]

Thank you so much for all your responses. Currently this set up is not active, I have been trying out a simple version using a laptop and a mike. The mike will be probably placed hanging from the ceiling centrally in the room so as to pick up voices. The speaker will be somewhere around the edge of the room. I was hoping to have a laptop running to control the whole setup. It needs to be a fairly stable set up as it will be a switch on in the morning and run without too much intervention. If anyone has any more to add after reading this I would be very grateful.

[Sam Spoons]

I guess this is some kind of installation art? Do you intend to replay the distorted/delayed voices as the subject is speaking? If that is the case you are, I fear, doomed to failure. The mic is relatively distant from the subject who is, presumably, not expecting what is about to happen so will be speaking at a normal conversational level. The loudspeakers will be, maybe only twice the distance from the mic as the subject. The simple physics of this will limit the gain before feedback you can achieve and, at best, the playback will be at a very low level.

Delaying the playback will help a little but a brief experiment shows that a 200 ms delay simply makes the feedback build more slowly which is no help in your situation.

Perhaps if you explain a little more about the aims of the installation we can come up with something that might work?

[Wonks]

You can always put a reasonable single delay on any distorted voice, say 2 seconds, so that it isn't concurrent. This may or may not fit in with your intentions. However any fairly constant background noise will be picked up by the mic and unless the overall system gain is less than 1, the background noise replay will slowly get louder. Any louder spoken words will also create peaks of noise that will assist in this overall loudening of the playback.

So any playback needs to be quieter at the microphone position than the voice of anyone picked up by the microphone. By delaying the distorted voice, it will certainly make it more noticeable to the participants. However, it may encourage some people to treat the space as an echo chamber and shout louder to hear their voices being repeated.

You could use a more directional mic fixed to the ceiling, that is angled to face away from the speaker (a hyper-cardioid might work well here) so its least responsive sector is pointing towards the speaker. The downside is that the mic is picking up from a much smaller area of the space. If people move away from that area, then it could all go quite quiet (I have no idea how big the space is and have no idea how many people you expect and so how likely there is going to be someone speaking to drive the system).

There are other types of sound distortion that could be applied that don't have the major gain issues associated with standard gain-driven distortion methods. There is bit-depth reduction, so a 24-bit recording is converted into say a 6-bit recording, which will distort the original signal without adding gain. There's ring-modulation, to give a Dalek-type vocal sound. There's backwards reverb, which records a few seconds of sound, then plays it backwards with some reverb added whilst at the same time recording a new section of sound to play backwards. Even extremes of phasing, chorus and flanging will make a voice sound weird. Band-passing the sounds can make them sound thin, whilst some fairly extreme EQing can give nasal and 'telephone' style sounds.

A basic DAW program can be set to continuously record over a selection of bars, and various plug-ins are turned on and off through automation, so you get a constant changing of effects. You could record from one input to several tracks at the same time, then have volume automation fade in and out the different effects (if you didn't want one effect to suddenly change to another). There are several ways to get the same end result, but it would probably be best to keep it simple and understandable by using several tracks with one or two effects on each rather than trying to do it all on one track with multiple lanes of automation.

You'd have to set the DAW to monitor the tracks and playback the monitored sound, so that any latency in the plug-ins would add to and delay time added deliberately, but I don't think that would be an issue.

[Mike Stranks]

... another thing to consider is that as soon as most people hear their voice replayed to them delayed or distorted doesn't matter - they immediately shut up!

I'm with Sam on this... I can't see this ever 'flying'. Sorry.

[Sam Spoons]

Agreed Mike. Though Wonks suggestion of a 2 sec delay, (and gated to avoid background noise setting off a feedback loop?) seems like the only way it might do something useful? The long delay would give the punters time to get used to the idea. It would help if the OP would give us a better idea of his aims.

[Sutchi]

Just to recap, the space is 4m x 3M with a standard room ceiling height. I just want to create a curious interactive sound space with a visitors to the space. I would think that the average amount of people in the space is one or two together at any one time. The delay idea was just me trying to cut down on feedback and the voice distortion just a bit of fun. I understand that the gain on the mike to pick up normal conversation will conflict with the speaker at the edge of the room. A directional mike might help but would obviously cut down on the opportunity to relay voices. I am fighting the laws of physics, I know. Would anyone be able to suggest an alternative if this is doomed to failure. I am still open to other ideas though in keeping with this idea of echoing back a visitors voice. I appreciate everyones time in thinking about this conundrum. I look forward to any suggestions.

[Wonks]

is there anything else in the room, paintings, sculptures, installation art etc? If there are objects standing within the room, it might be possible to put a small mic on a few objects, and highlight the mics with say a red circle round them. People would have to speak near the mics to be picked up, which means that you need far less gain on the mic - so far less risk of feedback.

I've thought about using the mic to trigger random samples of pre-recorded effected vocal sounds, or open the gate so that pre-recorded effected vocals can suddenly be heard. However there is the problem that if the playback sounds are loud enough for people to hear, then they will simply keep triggering more playback or keep the gate open. You could probably get technical and set the input trigger up with a reverse polarity copy of the playback sounds so that they cancel out, but that's getting quite complicated to set up. And triggered sounds may not be what you want at all. Motion detectors would probably make better trigger inputs, and with multiple narrow beam detectors, depending where people were, different sounds could be played back as they moved around. This might work with a drum brain sending out MIDI messages that a DAW could process. You could probably just use a drum sampler for the noises (with non-drum samples), but playback would be limited to those samples, whereas I'm sure you could play back a lot more random samples if connected to a computer.

[Wonks]

My only other suggestion is to issue wireless headphones so that there is no speaker to worry about, so all the sounds come through the headphones with no feedback issues. You'd just have to ensure that people put the headphones back on their hooks (and didn't walk off with them) and that you had a sufficient number of charged headphones to replace the ones in use as they run down.

[Drew Stephenson]

I was wondering if boundary mics with a second or so delay might be interesting as it would pick up things like footsteps as well.
A delay might cause people to stop, as Mike says, but then restart when it goes quiet. Delays at different frequencies might be interesting as well as helping to tackle feedback.

[zenguitar]

Mount a large number of low power speakers in the ceiling so that you get good coverage of the space without using much gain.

The my thought would be to choose a much longer delay, like 15-45 seconds. Long enough to avoid feedback, but short enough for people to recognise what they were saying before they forgot the conversation/comment.

Give the space a number of focal points with something on display that invites comment and put an omni mic at each one but broadcast the delayed speech to the entire room. Then see how people interact with the delayed speech at random. You could set a different delay for each microphone, or a random delay time for each microphone reset every minute.

Sum the delayed signals to a mono mix to broadcast to the room, but also record it so that you could use the randomised conversations for a future work.

Call it 'The sound of social media'. I claim my BA in Art

Andy

[Mike Stranks]

Good idea for longer delays, Zen!

It's very disconcerting to people to hear their own voices just a second or two after the said something.

Back in the days of tape we sometimes played a 'jolly jape' ( ) on presenters on pre-recorded programmes. These were people who were used to wearing headphones and receiving ongoing production directions as they were 'on air' so they could listen and speak at the same time with no difficulty.

We'd take the 'off-tape' signal (slightly 'behind' the signal being recorded) and route that into the presenter headphones. Hearing their own slightly delayed voice would bring the most seasoned to a halt in a few seconds...

So Andy's (Zen) suggestion is good... the trick would be to ensure the delay is not so long that the speakers haven't moved on by the time the voice comes back through the loudspeakers!

[AlecSp]

I am fighting the laws of physics, I know.

Completely

Would anyone be able to suggest an alternative if this is doomed to failure. I am still open to other ideas though in keeping with this idea of echoing back a visitors voice.

Unfortunately, while I'd love to be more positive, I believe this is doomed to failure.

[Sutchi]

I am thinking of using a laptop to control this planned installation. With this in mind what would be the best way to achieve such a long delay in play back?

[Sam Spoons]

If using a laptop and needing minimum intervention I'd first try a long delay patch in a DAW and, if it's sensitive enough, a gate to remove background noise (which may help prevent a slow building feedback loop) it'll also mean not every bit of speech gets through only the loud bits which may add a useful element of randomness to the exercise.

I still think you'll have difficulty getting enough level but I'll have a little play in the studio later with the drum overhead an a vocal monitor to see if I can make anything work.

[Wonks]

Too long a delay and the people may have disappeared before anything has happened. I'd suggest using a delay just long enough to prevent feedback.

Just beware of two very chatty people in the room who just keep talking without listening; keeping the noise gate open and the system recording which could very well end in feedback if it went on for long enough.

[Hugh Robjohns]

In the situation described above, it might be worthwhile including a limiter across the output adjusted to hold the system gain just below the level needed to establish a howlround.

H

[Kwackman]

If you do ever get this to work (and I hope you do!) for fun I'd suggest to use a pitch changer to distort the delayed voices- an octave higher or lower, or both!
Visitors won't hear their own voice, but their words sounding either like Lee Marvin, or like Pinky & Perky.

[Sam Spoons]

I have done a couple of experiments using a distant mic and a speaker with a 500ms delay. All it seems to do is to delay the onset of the feedback (and distort it BTW). Adding a gate had no useful affect either (the idea being to remove the background noise). The wanted audio is picked up by the mic when it is replayed and generates a feedback loop. I think to make this work you'll have to get into some programming to interrupt the mic while the speaker is active. Not sure how though sorry......

[Sutchi]

Thanks to everyone for your input . This is my first time on this forum and I am very impressed with the suggestions and time spent trying to help me.