3
\$\begingroup\$

First time here, so here I am with an electronic-related question. Please note that altough I have a scientific background, I'm quite new to electronics.

Setup

I'm trying to build a small audio amp using a LM386 chip. I found a lot of designs on different websites, and I ended up building this :

circuit
(source: free.fr)

The design is largely based on this one and this one. The cap between pins 1 and 8 sets a higher gain for the amp, and if I understood correctly, the feedback betweens pins 1 and 5 helps cutting higher frequencies and reducing hiss.

The input is a simple electret mic, and all this is powered by an Arduino board.

Problem

For my application of this amp, the microphone and the speaker are very close from each other, and I need a high gain. You might see the problem coming : I have to avoid the audio feedback (larsen) between them. This is why I added a simple RC low-pass filter at the end of the audio path (dotted box on the schematic). Its cutoff frequency is roughly around 1kHz, which should be ok because feedback yields frequencies higher than 4kHz.

This seems to be reducing the feedback a bit (I can put the mic and the speaker closer), but it isn't enough : I still get feedback at the same frequencies.

I tried to add a filter right after the mic output (before the amp), but then I can't hear any output signal in the speaker : it seems to be cutting all the input signal.

So, I'm looking for a filtering solution (either using extra components, or taking advantage of the LM386) that could help completely shutting off the highest frequencies of the signal.

I hope my setup and problem is clear. If not, do not hesitate to ask for more details !

Thank you !

\$\endgroup\$
9
  • \$\begingroup\$ Forget about how to do it, what are you trying to make that requires a microphone so close to a speaker??? High-level answer such as "washing machine" is what I'm looking for. Please don't say "I'm building an amplifier". \$\endgroup\$
    – Andy aka
    Commented Mar 20, 2015 at 14:03
  • \$\begingroup\$ In fact I'm trying to build what could perhaps be described as "controlled feedback instrument". The microphone and the speaker are at each opening of a PVC pipe. I want to amplify the natural resonance of the pipe, but I have to get rid of the "direct feedback" between mic and speaker, which is orders of magnitude stronger. (article explaining the concept) I've been able to make a proof of concept using my computer soundcard, but I'd like to make it a standalone device. \$\endgroup\$
    – Thoduv
    Commented Mar 20, 2015 at 14:15
  • \$\begingroup\$ The natural resonance of something is just a number - how can you amplify that number? Sorry I just don't understand what you are trying to achieve, \$\endgroup\$
    – Andy aka
    Commented Mar 20, 2015 at 16:01
  • \$\begingroup\$ Acoustic resonators act as (quite) linear filters on sound waves : they damp frequencies and amplify others. If a sound wave travels through a pipe, the outcoming signal measured by a microphone will have a different spectral content, namely the resonant frequency will be favored. At the next time step, if we have put a speaker in this pipe, and we feed it with the signal from the microphone, the resonant frequency will be amplified again by the pipe and so on. Quickly the feedback mechanism will saturate around a periodic signal whose frequency is the resonator fundamental frequency. \$\endgroup\$
    – Thoduv
    Commented Mar 20, 2015 at 16:38
  • \$\begingroup\$ OK so you are trying to find the resonant frequency of the pipe by pushing its zeros (ala pole zero information) to the jw line aka reducing losses/dampening. Then, I suspect you will want something that just about maintains oscillation without causing distortion i.e. you need a voltage controlled amplifier in the feedback loop to control the gain of the amplifier. Yes? No? Maybe you also need filters to eradicate certain frequencies that the pipe might resonate at due to integer multiples of wavelength? \$\endgroup\$
    – Andy aka
    Commented Mar 20, 2015 at 16:49

1 Answer 1

Reset to default
2
\$\begingroup\$

Running your signal through a low pass filter is probably not what you want. Normally when feedback is an issue, you want to apply a notch filter with a reasonably high Q at the feedback frequency. That frequency is not trivial to predict and will vary with position of the microphone, the room it is in, humidity, and a slew of other factors. It's highly possible your feedback frequency will be around 2kHz or lower, and by the time you low pass the signal enough to stop the feedback, it will sound like your hand is on your mouth while you're talking.

My first move would be adding an easily controllable gain stage ahead of the LM386, and take the extra gain away from the LM386. Turn it up until it squeals, and see if it's enough gain.

\$\endgroup\$
3
  • \$\begingroup\$ Thank you for your answer. I was going to try a notch filter, but I prefered to start with a very simple low-pass filter, so I could see if I was able to get rid of the feedback (signal quality is not a big trouble for my application). When you say "controllable gain stage", I assume you mean tuning the gain of the LM386 ? I think this could be done for instance by adding a potentiometer in the pins 1-8 track. I'm going to try this. \$\endgroup\$
    – Thoduv
    Commented Mar 20, 2015 at 13:04
  • \$\begingroup\$ Haven't read a 386 datasheet in several years, but I'm pretty sure you can't put a pot between pins 1 and 8. \$\endgroup\$
    – Matt Young
    Commented Mar 20, 2015 at 13:07
  • \$\begingroup\$ I tried to lower the gain, and it indeed suppressed the feedback, but it's also too low for my purposes. Then I tried to use a second-order RC to get a sharper filtering, but it doesn't seem to bother much the ~4kHz feedback. \$\endgroup\$
    – Thoduv
    Commented Mar 20, 2015 at 16:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.