Audio feedback eliminator

Thread Starter

Earl Jefthas

Joined Aug 23, 2009
6
Hi fellow electronic geeks

I am required to build a small box that will fit onto a microphone, the circuit is required to elliminate feedback that is occuring.

First of all I had a simple idea how to do this, I will have to make use of a pre-amp to amplify the signal from the mic and the output of the mic will than go into a bandpass filter.

The bandpass filter will cut out the frequency of the feedback.

But what I found was that I will have to check the output of the and look whether it is unstable(feedback is occuring) or stable. If feedback is occuring I will have to lower the gain of the mic and by doing so the feedback will be elliminated.

The porblem that I have is that I can not think of an idea how to check the output of the mic and than to allow a voltage drop to enable a voltage control amplifier.

Can someone please offer some advice?

(I don't have sound engineering experience)
 

rjenkins

Joined Nov 6, 2005
1,013
I don't know how you would detect feedback without knowing in advance what the audio should sound like.. Any frequency detector would eqally detect a sustained vocal or instrument note.

One common way around the problem is a frequency shifter. This changes the pitch of the microphone audio by a tiny amount (around 5Hz if I remember right).

This prevents feedback building up, because any given spot frequency is shifted repeatedly, each pass around the feedback 'loop'.

The real mike audio is only ever shifted once, with no noticable detrement.
 

flat5

Joined Nov 13, 2008
403
5 hz shift. Cool. Never heard of that.

A parametric filter can be used to notch out one frequency or a narrow band of frequency, so to speak.
 

Audioguru

Joined Dec 20, 2007
11,248
A frequency shifter does not allow much more output level before the feedback makes a "frequency shift" sound.

If the sound system has a fairly flat frequency response then a notch filter does not allow much more output level before another frequency causes feedback howling.
 

Thread Starter

Earl Jefthas

Joined Aug 23, 2009
6
The feedback that needs to be considered is the feedback from loudspeaker. The frequency that i will have to eliminate is the high frequency noise which sometimes occur when the output of the loudspeaker is reflected back to the microphone.

This why I was think of designing a filter that will cut out that frequency, but I was informed by an audio expert that the only way to reduce that type of feedback noise is to lower the volume until that noise is not present anymore and than increase the volume again.

So I will have to use a voltage controlled amplifier, the output of the mic will still go through the filter but some how when the feedback noise frequency is present the VCA must be triggered to lower the gain of the mic until the noise is eliminated.
 

hgmjr

Joined Jan 28, 2005
9,027
The feedback that needs to be considered is the feedback from loudspeaker. The frequency that i will have to eliminate is the high frequency noise which sometimes occur when the output of the loudspeaker is reflected back to the microphone.

This why I was think of designing a filter that will cut out that frequency, but I was informed by an audio expert that the only way to reduce that type of feedback noise is to lower the volume until that noise is not present anymore and than increase the volume again.

So I will have to use a voltage controlled amplifier, the output of the mic will still go through the filter but some how when the feedback noise frequency is present the VCA must be triggered to lower the gain of the mic until the noise is eliminated.
If you take your idea a bit further, you could consider detecting the presence of the oscillator condition, measure its frequency and then dynamically adjust a notch filter to remove the unwanted frequency. In this way the correction would be able to accommodate variations in the distance between the mike and the speaker which is a strong determinant of the oscillatory frqeuency in the first place.

hgmjr
 

Ron H

Joined Apr 14, 2005
7,063
I used to work with Dave Hershberger, who is an absolute wizard. For several years, he made and sold a kit called DSP-3, which he had designed, to hams (radio amateurs). It adaptively filtered out repetitive noise such as ignition noise. I did a little searching and found DSP - An Intuitive Approach, which was published in QST magazine in February 1996. It might not work in an acoustic feedback loop. I'm also not sure the audio quality would be sufficient.
 

Audioguru

Joined Dec 20, 2007
11,248
The best way to eliminate acousical feedback howling is to record the sounds with the speakers turned off then play back the sounds with the mic turned off.

Turn down the volume then get closer to the mic.
Face a directional mic away from directional speakers.
Use a mic and speakers with a flat frequency response (most cheap speakers have a peak at about 5kHz) and some "vocals" mics also have a peak at about 5kHz).
 

Thread Starter

Earl Jefthas

Joined Aug 23, 2009
6
Thanks quys, I will combine my ideas with yours.

I will use a mic and speakers with a flat frequency response and I will have detection device that will detect the oscillation condition and than from there I will voltage controlled amplifier that will lower the gain of the microphone.
 

flat5

Joined Nov 13, 2008
403
The frequency response of the environment is part of the equation too. Having a good graphic equalizer and a few tunable notch filters can help.
Start with the sound as flat as you can get it and do what you have to do.
Like AG says, close mic, keep the speakers far away (isolated).
I would remove unneeded lows and highs.
 

rogs

Joined Aug 28, 2009
279
Have to agree with The RB about the Behringer DEQ2496 -very impressive unit. There is a cheaper version the 'Shark': http://www.behringer.com/EN/Products/DSP110.aspx which also works pretty well - and has all sorts of other fun tricks in the box!

I have to say that having tried all the other options mentioned in this thread at one time or another, none come anywhere near the effectiveness of these 'auto learning' DSP processors, in my experience.

What seems like a fairly easy problem in theory - locate the offending feedback frequency and notch it out - is actually quite difficult in to achieve in practice.
Any change in the set up can 'move' the dominant feedback frequency.
Moving the mic - introducing more people into the acoustic environment - moving a loudspeaker -anything can (and usually will!) change the dominant feedback frequency, and thus the 'notch' freqency of the required bandpass filter.

These DSP units will 'track' the acoustic environment, and automatically optimise the required notch filter set up.

Of course, nothing beats the best solution of all -- reduce the loop gain to less than unity at all frequencies, by turning the volume down!!
 
Top