Phase Cancellatoin and Hello

Posted: Mon Nov 17, 2008 10:37 pm

Hi guys,

Not sure which section to post this in so have posted it more than once as it is important to me sorry.

Im new to the froum. A little about me first - Im studying Audio System Design at Southampton Solent University (UK) in my final year.

To the crux of the matter. For my final year project I have set myself the challenge of designing and making a unit to detect and supress acoustic feedback in an audio system - both digital and analogue. As feedback is a sinewave created by the delay in a system in correlation with a specific frequencey I have come up with a theory in order to supress it without using standard filters.

I am looking into using phase cancellation in order to do this. I was wondering if there was anyone who had expertise in this field or research or just any ideas really as I'm just getting off the ground with this.

Thanks in advance for any help.

Regards

Robbie de Jong

Posted: Tue Nov 18, 2008 1:21 am

The product forums don't get as much traffic. Perhaps a mod will move this to live sound.

So mixing in a sine wave that is out of phase of the feedback frequency. Interesting.

It sounds like a complicated control system problem. If you are off more than 90 degrees in phase, the feedback will not be reduced. If you add too much of the out of phase signal you will make the problem worse by switching the phase of the feedback as the room acoustics will naturally accentuate this frequency.

How do you tell the difference between feedback and a crescendo of a synth lead?

Good luck in your research, is sounds like a great idea.

Posted: Tue Nov 18, 2008 8:11 am

Let's see? I'm an analog kind of person. Well, I could take the recording feed from the PA mixer. I could then take that signal through another amplifier, invert the phase. This signal could then be applied to a noise gate/downward expander. Its level would be kept low but when the feedback rears its ugly head, the noise gate/downward expander will open up. Allowing signal to pass which would then be combined with the output of the PA mixer before the amplifier. So as feedback would open the noise gate above a set threshold. Combined out of phase with the output signal from the PA mixer should help to reduce feedback problems. So that is utilizing phase shifts technique to help tame feedback. This is on the assumption that the feedback will always be louder than the performers. Which it generally is. And the entire cancellation of the signal would be minimized by the action of the feedback. Of course it can also cause a certain amount of pumping to occur since this would be a broadband oriented device. It could be simplified to be high pass oriented since most feedback is high-frequency as opposed to low-frequency feedback.

See if your teacher like that idea?
Ms. Remy Ann David

Posted: Tue Nov 18, 2008 8:26 am

For our church PA I do that sort of thing, but just with a compressor. So the level of the lapel microphone is never allowed to get up to the level that would feed back. Same principal. The feedback is always louder than the person talking, so if you don't let it get that loud, then you don't have a problem. I like Remy's suggestion of the high pass filter. I could add one in the side chain.

Posted: Wed Nov 19, 2008 4:34 am

Adding a high-pass filter to a compressor's sidechain makes the compressor LESS sensitive to low-frequency information, which means that bad artifacts like plosives, pops, etc will not be limited when they should be.

Posted: Wed Nov 19, 2008 4:45 am

Do you want to stop feedback after or before it occurs?

Posted: Wed Nov 19, 2008 4:47 am

moonbaby wrote:

Adding a high-pass filter to a compressor's sidechain makes the compressor LESS sensitive to low-frequency information, which means that bad artifacts like plosives, pops, etc will not be limited when they should be.

I think you are missing the point: the idea is to reduce feedback. The compressor does not even kick in unless the volume goes above normal speaking level.

Also, this is a normal setup, and very useful. If it was engaged when at normal speaking volume this would be a common configuration for a de-esser.

Plosives and pops can more effectively be reused using an EQ rather than a compressor due to their transient nature.

You are right about the plosives mainly having low frequency content.

Feedback is normaly triggered by a louder than usual sound. So if you are running the microphone on the edge of feedback, a loud section of speech will push the system into feedback. A compressor on the insert path can help fight feedback. The threshold is raised just above normal speech. The raito at about 1:4. Normal attack and release. Because it is on the insert path whatever Joe is running sound, can ride the main fader to their hearts content. If they get into a ringing situation, don't hear it, and leave the board, the compressor will catch it before it feeds back.

Posted: Wed Nov 19, 2008 5:42 am

I understood your statement. First off, "compression" brings UP the gain on low-level sources, and reduces gain on higher level information. If you are setting the threshold control to a high enough level as to only respond to the PEAKS, that is referred to as "limiting". Compression will INDUCE feedback issues, by its' very nature of attempting to raise the gain of a lower-level source.
And FYI, de-essing works by accessing the detection circuit and BOOSTING the offensive band via a notch or bandpass filter design, with a gain BOOST at a very narrow bandwidth. A simple highpass filter will make the detection slam everything that's above the filter's hingepoint. This will yield a dull, muddy signal that will have all of it's life pounded out of it.

Posted: Wed Nov 19, 2008 7:54 am

RobbiedeJong,
Sorry for hijacking your thread. I was just trying to elaborate on what Remy was saying. I think the new JP22 thread is getting to me and moonbaby.

moonbaby wrote:

I understood your statement. First off, "compression" brings UP the gain on low-level sources, and reduces gain on higher level information.

No. Compression reduces the volume. Make up gain on the compressor increases it. Compression only effects the dynamic above the threshold.

Quote:

If you are setting the threshold control to a high enough level as to only respond to the PEAKS, that is referred to as "limiting". Compression will INDUCE feedback issues, by its' very nature of attempting to raise the gain of a lower-level source.

Turning the make up gain up would induce feedback just like raising the fader. Changing the ratio or threshold will not induce feedback unless the make up gain is set to auto. Limiting is a type compression. Specifically compression with a high ratio.

Quote:

And FYI, de-essing works by accessing the detection circuit and BOOSTING the offensive band via a notch or bandpass filter design, with a gain BOOST at a very narrow bandwidth. A simple highpass filter will make the detection slam everything that's above the filter's hingepoint. This will yield a dull, muddy signal that will have all of it's life pounded out of it.

Why are you yelling boost?
I agree that a better De-esser can be created using a pass band filter than using a high pass filter. However a high pass filter in the side chain of a compressor is a common implementation of a de-esser. (The one built into Pro Tools uses this method) If you use a spectrum analyzer to watch a vocal track you will see that the "ess" sound is about a 2 octave band of noise near 6kHz. It peaks up fairly high. At this high level there there is no non-ess activity above 6kHz, so a high pass filter will function adequately in place of a pass band. Don't get me wrong, there is harmonic content above 6k, but it is not loud enough to trigger a properly adjusted high pass filter.

Now if you have cymbals on your vocal track, then using the high pass filter would be a problem, but for just vocals, the high pass is fine.

Posted: Fri Nov 21, 2008 11:33 am

You are quoting somebody's tech literature without much real-world application /operation. A "common application" is to use an equalizer plugged into the sidechain and then boost the closest frequencies to the offending range (of, say, sibilance) to nail the problem.
BTW, I have another quote for you to ponder. It was recently stated by our own Remy Ann regarding compression and feedback issues:

--------------------------------------------------------------------------------
"Compression is great for recordings. Almost impossible to deal with if you are trying to amplify anything as all it does is cause feedback problems. When doing a lot of work, you might want limiting but not compression. That isn't to say you don't have to use the compression. And in your first assumptions, Yamaha is a great brand. No problems there. "

Limiting is NOT merely a higher ratio of compression. It is a higher THRESHOLD, set only to LIMIT peaks. You don't seem to get this basic
fact.

Posted: Tue Dec 09, 2008 1:32 am

Cheers for the replies guys. I am looknig into designing a DSP unit to detect and attenuate the feedback before it becomes strong enough to get through the programme material. Looking into other research done for hearing aids this should allow another 10dB (on average) extra level for the rest of the programme.

Posted: Tue Dec 09, 2008 2:00 am

RobbiedeJong wrote:

Looking into other research done for hearing aids this should allow another 10dB (on average) extra level for the rest of the programme.

Are you saying that the technology used for feed back elimination in hearing aids adds 10dB of dynamic range to the audio? Not sure that I follow you here. Where is that information from?

Posted: Tue Dec 09, 2008 2:02 am

Im saying that the dynamic range is improved before feedback becomes a problem again.

This topic sponsored by: