0
\$\begingroup\$

enter image description here

opAmps are fed by +/- 9V.

I have two more filters like this one, in total three filters.

If I use a signal generator as input signal and observe output via an oscilloscope; all those filters give me satisfying results(not the same with simulated results, but close enough considering I am using %5 resistors and other possible non-idealities.)

However, in reality, the input signal comes from a microphone output. And mic takes sound from a 10 cm away speaker(it is being so loud that, microphone gives around 600 mVpp directly(***), without any amplification done.) At this point, the problems start to occur.

(***) - is this normal? I mean I always saw that microphones give output in the range of maybe tens of milivolts while I was doing research on the web. The microphone is an electret one, and I don't have any datasheet or so.

The filters which works fine before, starts to behave problematic. For example, even if I don't connect any input signal(the output of mic), the filters output a sinusoidal-ish waveforms around 400-500 mVpp at their(filter's) designed center frequency when the loud speaker is on. I even connected input to ground, but did not help. One thing I could not understand is why filters interact with sound when there is no physical(electrical) connection with the microphone?

Also, when connected with microphone, the filters still does not behave as before. I think it might be resulting from some kind of loading effect. There is a clear difference between mic out and signal generator out, which I could not figure out. Two things I suspect are,

  • The filter opAmps are NE5532 which has a very low input impedance of 30k ohms or so.

  • I think adding a voltage buffer to microphone output may help since it will(?) make output impedance more similar to the output impedance of a signal generator, but I was not sure and did not try.

I am also attaching the mic driver circuit, which is pretty simple.

enter image description here

It is the exact circuit in my implementation with +9V instead of +5V.

The microphone output itself is fine, it gives pretty similar waveforms to loudspeaker output.

I might look combining two questions in one, but I could not be sure if they are related whith each other or not.

\$\endgroup\$
22
  • 1
    \$\begingroup\$ "if I don't connect any input signal(the output of mic), the filters output a sinusoidal-ish waveforms around 400-500 mVpp" - what happens if you ground the input? Please show the complete schematic, including power supply and decoupling capacitors. \$\endgroup\$
    – Bruce Abbott
    Jan 5 '20 at 2:42
  • 1
    \$\begingroup\$ Your diagrams have left out some essentials: How are the opamps and the mic supplied with DC in reality and what's the GND. Having poorly or totally non-decoupled voltages or weak GND or both are two common ways to call troubles such as oscillation. \$\endgroup\$
    – user287001
    Jan 5 '20 at 11:57
  • 2
    \$\begingroup\$ Integrated circuits need stable supply voltages, they must not be fed through inductors. The decoupling caps are commonly inserted with the shortest possible wires between the supply voltage input pins of the IC and the capacitor. \$\endgroup\$
    – user287001
    Jan 5 '20 at 12:11
  • 2
    \$\begingroup\$ See manufacturer's layout guidelines in the datasheet: ti.com/lit/ds/symlink/ne5532.pdf \$\endgroup\$
    – user287001
    Jan 5 '20 at 13:00
  • 2
    \$\begingroup\$ The microphone you used has an FET inside to match the electret element to a more conventional 10K-ish impedance. It also amplifies the signal. That's why you get 600mV out instead of 10-20mV. It's also why it requires a power supply. \$\endgroup\$
    – gbarry
    Jan 5 '20 at 21:08
3
\$\begingroup\$

I simulated your filter in LTspice. In transient analysis with the input open circuit it broke into oscillation after 40ms. With a short circuit it took 200ms, but with 50uV of noise added it only took 20ms.

AC analysis of the first stage shows phase increasing with frequency when passing through the center frequency of the filter. This is a sign of potential instability.

enter image description here

Increasing R2 from 1kΩ to 1.05kΩ I got this:-

enter image description here

The phase now flips to negative but the amplitude peaks sharply at over +20dB, indicating that the filter is right on the edge of being unstable.

With 1.1kΩ I got:-

enter image description here

The response is now similar to the original design with 1kΩ, but the phase is going in the correct direction for stability. This is confirmed by transient analysis, which does not show any hint of oscillation.

Trying to get the sharpest filter response with 5% tolerance components is dangerous because a small variation could push it into instability. You should either use 1% tolerance resistors, or 'detune' the filter and accept a wider bandwidth. Alternatively you could make a passive LC filter using high Q inductors, which is guaranteed to be stable.

even if I don't connect any input signal(the output of mic), the filters output a sinusoidal-ish waveforms around 400-500 mVpp at their(filter's) designed center frequency when the loud speaker is on. I even connected input to ground, but did not help. One thing I could not understand is why filters interact with sound when there is no physical(electrical) connection with the microphone?

The speaker circuit might be inducing noise into the ground, power supply, or signal wires, which is 'shocking' the filters into oscillation, or you might have a 'ground loop' which is only present when the speaker is turned on.

Also, when connected with microphone, the filters still does not behave as before. I think it might be resulting from some kind of loading effect. There is a clear difference between mic out and signal generator out, which I could not figure out.

The microphone has a higher impedance than the generator. If the filters are close to instability then a small change (such as higher source impedance which effectively makes R1 larger) could push them over the edge.

Adding a buffer between the microphone and filters is a good idea because it isolates them from the source and reduces interaction between them, but as you also have oscillation with the input short-circuited a buffer may not stop it. Even if it did, you should increase the filter stability margin to avoid problems in the future.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks for such a detailed answer! The speaker circuit and filter part, indeed, use same power source. I will first seperate them using different power supply for each. Then buffer thing + making R2 1.1k ohms. After all these changes, I will let you know about the final situation. \$\endgroup\$
    – muyustan
    Jan 5 '20 at 16:14
  • \$\begingroup\$ btw, did you change both R2 or only the one on the first stage? \$\endgroup\$
    – muyustan
    Jan 5 '20 at 16:26
  • 1
    \$\begingroup\$ I changed both, but only simulated the first stage for the graphs shown here. The second stage performed similarly (which is expected since it is identical). \$\endgroup\$
    – Bruce Abbott
    Jan 5 '20 at 16:33
2
\$\begingroup\$

An audio amplifier that uses ceramic capacitors in the audio path produces an output when something touches or vibrates a capacitor because a ceramic capacitor is "microphonic". Film capacitors should be used.

Why don't your opamps show a filtered power supply? I assume it has positive and negative voltages. We do not know the voltages.

The electret microphone draws about 0.5mA then with a 10k resistor from 5V the mic has no voltage. The 10k resistor should be about 4.7k when powered from 5V and can be 10k when powered from 9V. The resistor that feeds the mic should be powered from an RC filter to avoid feedback from the power supply or battery.

\$\endgroup\$
8
  • \$\begingroup\$ thanks, which type of filter? I assume high-pass, but a more explanation would be nice. opAmps are powered by +/- 9 volts. The thing with ceramic capacitors can be vital in my case, because I have lots of them. Though, the ones in the filter are film, not ceramic. \$\endgroup\$
    – muyustan
    Jan 5 '20 at 19:37
  • \$\begingroup\$ ...........?... \$\endgroup\$
    – muyustan
    Jan 6 '20 at 12:10
  • \$\begingroup\$ All electret microphones have a Jfet impedance converter inside. The output level is 10mV when speaking at conversation level at a distance of 10cm from your mouth. For high levels like inside a drum or inside a piano the common source Jfet wiring can be modified (Linkwitz mod, look in Google) to be a source-follower. Your filter is called Multiple Feedback Bandpass Filter, look in Google) and is usually fed from a very low impedance. But yours is fed from the about 2.7k impedance of the mic and its powering resistor, so reduce the value of the filter input resistor by 2.7k. \$\endgroup\$
    – Audioguru
    Jan 6 '20 at 21:09
  • \$\begingroup\$ Hmm, makes sense now. I was also suspicious about that thing. Does unity gain oltage buffering the mic output with an opamp also solve the situation? \$\endgroup\$
    – muyustan
    Jan 6 '20 at 21:12
  • \$\begingroup\$ Wouldn't the output impedance of the source be 2.7k//10k ohms then? \$\endgroup\$
    – muyustan
    Jan 6 '20 at 22:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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