audio feedback eliminator circuit needed

Thread Starter

k2man

Joined Feb 25, 2014
7
I am designing a device that is similar to a megaphone to amplify voice. I have a feedback problem due to the close proximity of the mic to the speaker. This is a hand held battery powered device. The microphone is in an enclosure, which almost prevents feedback, but doesn't always. I've read about notch filters and frequency shifting as possible solutions. I'm using a dynamic mic with an op amp preamp feeding into a class D 30W amplifier, all running at 24-28V. I need a circuit to reduce or eliminate feedback. Can anyone help point me to a circuit that would accomplish this while being compact and efficient? Absolute fidelity of the audio is not critical, understandability of the speech is what is needed.
 

AnalogKid

Joined Aug 1, 2013
10,971
If the microphone location is fixed with respect to the speaker, and the offending oscillation frequency is fairly repeatable, then a Twin-T notch filter should pick off the offending freq with minimal impact. Two flavors, a passive network in series with the audio, and an active circuit where the twin-T network is in an opamp feedback loop. The active circuit has higher Q; that is, the notch is narrower and deeper.

With six frequency-selecting components, this guy is not very tuneable. But if the distance between the mic and speaker is fixed, and a scope gives you a consistant, repeatable number for the frequency of oscilation, you should be able to get the attenuation you need with 1% Rs and Cs.

Here are a few (!) circuit options:

http://images.search.yahoo.com/sear...mage&va=twin+t+notch+filter+circuit#index=srp

ak
 

Thread Starter

k2man

Joined Feb 25, 2014
7
t06afre - thanks for the link on DSP. I'll consider digital if I need to - I was hoping to find a simple analog solution, just because I have no digital in the system yet. I'll do some investigating there.

AnalogKid -Yes, my microphone is fixed wrt the speaker always. I've heard in testing some fairly low frequency feedback as well as higher frequency feedback. I'll check out the twin T notch - that would fit in my analog circuitry easily. Is this good only for a single frequency though? Is there a simple way to constantly shift the frequency of everything a little - I understand this will kill feedback. I don't think that shifting everything a few hz would affect my application at all.
 

#12

Joined Nov 30, 2010
18,224
I'd like to mention that you can lower the Q of the twin-T to get a broader frequency response, but you probably already know that. There is also a possibility that you can add audio frequency absorbent material in the mic enclosure. Knew that too? Well, just trying to cover all the bases.

Some microphone containers have strategically placed holes to let sound cancel on the backside of the element.
 

AnalogKid

Joined Aug 1, 2013
10,971
A low density closed-cell foam around the mic is a good idea, especially if one of the feedback paths is mechanical conduction through the enclosure rather than acoustic through the air.

Varying the Q of the filter is a trade-off between selectivity and depth. A high Q fiter can attenuate over 60dB, but the range of affected freqs is very small. But you can have multiple twin T filter sections in series (isolated from each other with opamps). First, identify the freqs that are causing the problem. Then try to determine how much attenuation you need. For example, do you have to yell into the mic to start oscillations, or will anything above a whisper set it off?

ak
 

ifixit

Joined Nov 20, 2008
652
Forget using RC filters.

Change the mechanical design so that the Mic will only pick up the sound from your mouth from no more than 2 inches away. Experiment with putting a sound blocking material in front of, or around the Mic.

Good Luck,
Ifixit
 

#12

Joined Nov 30, 2010
18,224
Varying the Q of the filter is a trade-off between selectivity and depth. A high Q fiter can attenuate over 60dB
Exactly my point. If you've ever tuned a church or auditorium, you will know that 60 db of feedback doesn't happen in the large scale acoustic world. One would be hard pressed to design a room or a football field that has a 60 DB difference in response to some few audio frequencies.
 

Thread Starter

k2man

Joined Feb 25, 2014
7
Here are my amplifiers. It's all battery powered. The microphone is nearly soldered to the preamp board - less than 15mm of wire to the board. The power amp is less than 100mm of wire from the preamp. The speaker is less than 100mm of wire from the power amp. So it is all pretty tight. The mic is inside an enclosure - sort of like talking into a beer glass with a vent hole drilled into it. The mic enclosure can't be lined with foam, as the whole thing can get pretty wet. I think I'm going to try to redesign the mic enclosure to better block any sound transmission from the rear of the enclosure.

I was hoping there was some simple electronic circuit I could also add as "cheap insurance" to prevent feedback, as I'm running such high power into a small speaker, and I've burnt up several speaker coils in testing. Isn't there a way to shift frequency a small bit (I don't think it would be a problem with my application - I need clear, understandable speech) that will kill feedback - in a simple analog circuit?
 

Attachments

Veracohr

Joined Jan 3, 2011
772
I've burnt up several speaker coils in testing.
That sounds like your speaker doesn't have the power capability to handle the amp feeding it.

I don't know how megaphones are constructed, but I suppose it's possible that, if the whole thing is contained in one physical unit, the problem might be mechanical feedback rather than acoustic.
 

AnalogKid

Joined Aug 1, 2013
10,971
The only thhing that jumps out from your schematics is that you don't need output coupling capacitors on both the peramp output and the power amp input. I'd eliminate the two 10 uF caps from the preamp output and let the 1 uF caps on the power amp input do the work.

The big questions are the frequencies of oscillation as shown on a scope. The answer might be as simple as adding a couple of parts to the preamp if the feedback gods are smiling at you.

ak
 

Kermit2

Joined Feb 5, 2010
4,162
Troubleshooting is not complete.

You need to determine if the feedback is acoustic or mechanical.

Isolate the microphone physically far away from the amp and determine if your problem is cured. If not then decrease the gain on the microphone, and/or add directional isolation to the microphone to decrease capture of off axis acoustic energy. ( put it in a tube shaped structure). your face can also reflect the amplified sound back into the mic input. using a foam 'wind screen' might help if this is the problem.
 

Thread Starter

k2man

Joined Feb 25, 2014
7
The mechanical stuff I've pretty much already done, but I've got a few more mods to try. I'm pretty limited here though. I can decrease the gain, but then the product becomes useless. So I need the gain. So what I need is a way to electronically kill the feedback, or prevent it in the first place. As I'm dealing with voice only, and I really only need the voice to be intelligible, I can alter the pitch or vary the frequency some, or put some time delay in - any of that will be acceptable. I just need some help designing a circuit that I can stuff in a small space that doesn't eat a lot of electrons.
 

rogs

Joined Aug 28, 2009
279
Unfortunately, the only effective electronic solutions to this problem are not simple, and are not likely to fit into a small space... or be available as a simple circuit!

As already suggested by several folk on here, you really would be better off spending your efforts on dealing with the problem mechanically.

Very few (if any?) megaphones, for example, include electronic feedback control hardware.. it simply doesn't work well without quite complex circuitry.. which takes up space!

Megaphones make use of a couple of simple acoustic properties which help the situation... The microphone is normally placed facing backwards - away from the loudspeaker and towards the person speaking - and the loudspeaker uses a horn enclosure, to both provide a much more efficient use of the available audio power, and to direct the sound away from the microphone, thus helping with any feedback problems.

I don't mean to seem negative here.. just realistic!. Google 'Feedback Destroyers' and there's quite lot out there describing the various feedback control techniques.... notch filters... frequency shifters... DSP controlled 'learning' equalizers... delay lines... etc
None of them simple to do effectively in practice, I'm afraid...
 

#12

Joined Nov 30, 2010
18,224
I don't mean to seem negative here.. just realistic!.
You want to see, "negative"? Here it comes: You can buy an (8) D cell megaphone on Amazon.com for $25. That's why I refused to repair one last month. It was loaded with corrosion like it was left out in the rain for a year, no parts available, and I can replace the whole thing for less than 1 hour of labor charge.:(
 

Thread Starter

k2man

Joined Feb 25, 2014
7
Rogs, thanks for the advice. I've explored the available techniques - feedback destroyers, most of which use notch filters. I very much understand that this is not a simple problem that has been solved before, but I think it can be done electronically, and can fit in a small space. I am designing a product (not my first) that has not been built before (again, not the first time I've done such a thing). I am going to explore shifting the frequency (old technology) as notch filters that actively shift are more complex. I like simple and elegant where possible. I'm hoping that someone out there in AllAboutCircuits Land has done something similar. If not, I'll try to report back with my solution and how well it worked.
 

rogs

Joined Aug 28, 2009
279
The idea of frequency shifting to solve this problem goes back a long way ... in practice, you'll probably get about 6dB of improvement before it all collapses in a heap, and you suddenly get loud 'howl round' with very little warning!....

Here in the UK, the Surrey Electronics website ( http://www.surreyelectronics.com/ ) still lists a device for sale that resulted from an article in Wireless World magazine way back in 1973 -- so the old 'analogue' way of doing it still seems viable.

Have you looked at the very simplest option .. a noise cancelling microphone?..
You might like to experiment with using two mic capsules, - mounted fairly close together (but far enough apart to be able to speak into one, and not the other) - and wired in antiphase, to create your own noise cancelling mic.... electret capsules are pretty small and cheap....
 
Last edited:

Thread Starter

k2man

Joined Feb 25, 2014
7
Does a noise cancelling mic help any with feedback? I didn't think they did. I am limited in microphones for other reasons - but I could use a second one (although expense starts to mount here) for noise cancelling. I am using a dynamic mic for this application.

Thanks for the link to Surrey Electronics. Very interesting that this technology is still around, based on a 1973 design! I am wondering if I can shift the frequency up 5Hz, then down 5Hz, then back up... in a triangular modulated pattern. Wondering if this would give me any better improvement.

I have a Roland VT-3 Voice Transformer on order - a newer model of the Boss VT-1, which does have a slide control to vary the pitch of a voice by +- 1 octave. I am going to wire that into my amp and bench test it to see what improvement I can get from that. If that doesn't work, my next best option I think will be to introduce a delay in the output. For my application, I don't think this will be too bad - I'm not so sure though, as it may cause difficulty for the user, although my user's won't be able to hear the speaker very well. I may just have to try it out.
 

ronv

Joined Nov 12, 2008
3,770
I'm just going to think out loud.
If it is mechanical (and at the resonant point it probably is) a notch at that frequency might work.
If the response is pretty flat it will just move outside the notch for a different frequency.
One thing I have noticed about feedback is it seems to take a while to develop, so some kind of sweeping notch or band pass might work.
 
Top