3
\$\begingroup\$

I'm building an electret microphone pre-amp following this application note, the output of the preamp is connected to an LM386 and then to some headphones.

So my pre-amp circuit looks like this, where VCC is 3V, all other components are the same:

enter image description here

and the LM386 part:

enter image description here

And it works, I hear the sound being amplified but, it's buzzing quite a lot.

So here there are some details about the implementation and things that I've done to try reduce/identify the noise:

  • The preamp stage was implemented using a different op amp, the one that I had available, an OPA313 which doesn't seem to be the reason because it's a relatively low noise op amp, and most of it's characteristics are very similar to the one used in the AN.

  • My first guess was the LM386 circuit, so I grounded the input to check if the noise was still there, and it wasn't, so it's quite clear that the noise is coming from the preamp stage. Which makes sense because that stage has the highest gain.

  • The board was initially powered using a USB (with an RC filter at the input 100 ohms + 330uF). So to remove the possible 'unfiltered' noise I powered the board with a 9V battery and a 7805 and the buzzing was still there.

  • The wire that connects the microphone to the pre-amp is as short as it could get, to eliminate any induced noise there, so I basically connected the C3 cap pin to the op amp pin.

  • Since I don't have the electret capsule datasheet I used the component values from the AN, except for R1, because the pre-amp stage is powered from a 3V regulator.

  • The noise it's not at 50 Hz, it's a higher frequency, around 3 KHz.

I have tried with different electret capsules but the noise is still there. Am I missing something? The capsules I found are the generic ones, anyway I don't think that the amount of noise I'm hearing is due to that. I've been struggling with this for many days now and looks like the solution is really simple but I can't see it.

\$\endgroup\$
7
  • \$\begingroup\$ Find the EMI injection point by suppression with a cap or shunt R or injecting more noise with finger. It could be conducted, differential unbalanced impedance from Common Mode supply noise and layout wires . Ground one hand and use other to inject or suppress noise to duplicate test to find injection point. Understanding this will make the solution easier. ie filter, shielding (STP cable or Coax, CM Choke etc) \$\endgroup\$
    – D.A.S.
    Commented Aug 5, 2020 at 22:40
  • 1
    \$\begingroup\$ The noise is external to your setup and is being picked up by the op-amp most likely. Check that all signal grounds are good. Tony's above comment should be tried out as well. \$\endgroup\$
    – user105652
    Commented Aug 5, 2020 at 22:45
  • 1
    \$\begingroup\$ Is "buzz" a harmonic of line frequency? \$\endgroup\$
    – D.A.S.
    Commented Aug 6, 2020 at 0:26
  • \$\begingroup\$ You need to help us out by showing component values such as R2, which sets the gain of the pre-amp. Also C3 would normally have a series resistor or gain could be infinite at certain frequencies. Provide part values and we will try to solve this problem. \$\endgroup\$
    – user105652
    Commented Aug 6, 2020 at 2:12
  • 1
    \$\begingroup\$ "The noise it's not at 50 Hz, it's a higher frequency, around 3 KHz" - I would not describe 3kHz as 'buzzing'. How do you know it's 3kHz? \$\endgroup\$
    – Bruce Abbott
    Commented Aug 6, 2020 at 7:10

2 Answers 2

Reset to default
3
\$\begingroup\$

I find the frequency characteristics of the LM386 very unstable. There is likely to be a high frequency oscillation in the system. This can make a special noise. The problem is caused by 100nF at the output.lm386

If this is really the problem, it can be fixed with a serial RC circuit: lm386-1

Update:

The circuit is theoretically stable, failures are likely due to the assembly or MLCCs. (The 3V power supply voltage (noise) is also interesting because the operational amplifier gains it as an inverting amplifier, 75/2,2 (34x).) micamp

Lastly I would suggest replacing the capacitor or the resistor to ca. 10 times bigger at the input of LM386 in order to have appropriate lower cutoff frequency and replacing possible MLCC capacitors in signal chain (C1,C4 in this schematic) to film/tantalum capacitors. The film/tantalum construction is free from piezoelectric noise. A disadvantage of the MLCC can be the piezoelectric nature of the ceramic material. MLCCs can exhibit problematic or disruptive noise (ringing or singing) due to oscillations within the audible range.

\$\endgroup\$
5
  • \$\begingroup\$ Good catch. There is supposed to be a low-ohm resistor in series with that capacitor. It is supposed to make the LM386 more stable.+1 \$\endgroup\$
    – user105652
    Commented Aug 6, 2020 at 0:07
  • \$\begingroup\$ csabahu, thanks for taking your time to simulate the circuit, actually the schematic I shared was missing the 10 ohms resistor in series with the capacitor, I have updated the question. \$\endgroup\$
    – fcoppede
    Commented Aug 6, 2020 at 1:04
  • \$\begingroup\$ It is not completely stable either. Instead of 100n you need a smaller capacitor (22 ... 33n). \$\endgroup\$
    – csabahu
    Commented Aug 6, 2020 at 7:10
  • \$\begingroup\$ Thanks! You are right, I changed that to a 33nF cap. I will update the schematic. (The noise is still there) \$\endgroup\$
    – fcoppede
    Commented Aug 6, 2020 at 14:00
  • \$\begingroup\$ Please do an experiment. Connect a 100u or larger capacitor in parallel with C3 (2.2u). Positive towards the microphone. Thanks. \$\endgroup\$
    – csabahu
    Commented Aug 6, 2020 at 16:46
2
\$\begingroup\$

If the answer by "csabahu" does not work, then lets clean up the VDD to the microphone.

In that VDD, insert a series R and a shunting C.

[NOTE: this is a LOW PASS FILTER; connect as external_VDD, series R, then the shunting C, then the existing R1.]

Make the new R_VDD be 1/10 of R1.

Make the capacitor be 1uF or 10uF or 100uF. Ground the cap to the bottom of the electret microphone.

ALSO --- the gain of the opamp circuit may be enormous, because there is no resistor in series with the input capacitor. With that much gain (R2 * FET_gm), that circuit may be oscillating. Insert 1Kohm or 10Kohm in series with the CAP.

================================================

Assuming the FET does indeed convert the Electret (electrostatic/piezo sensor) voltage output into current variations, the 1/gm (1/transconductance) of the FET will be the Rout of the entire "microphone" function.

================================================

If you have high gain, the "electret" ground node must be the same physical location (solder joint) as the ground of the VDD/2 bypass capacitor C6.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ My schematic was missing the 10 ohm resistor in series with the capacitor (just updated it). About the R_VDD, that's exactly what I did, it did not solve the problem. Regarding the gain of the op amp, it's actually working as a transimpedance amplifier, so the gain is not that big, see the application note page 5. \$\endgroup\$
    – fcoppede
    Commented Aug 6, 2020 at 1:12
  • \$\begingroup\$ What is the value of R1? It would be necessary for the simulation... \$\endgroup\$
    – csabahu
    Commented Aug 6, 2020 at 10:54
  • \$\begingroup\$ Hello csabahu, My R1 is 2K2 and vcc for that stage is 3V, so that way I have around 2V across the electret. To simulate the electret in that transimpedance configuration you will have to use a current source at the input. That current source will have a DC value that will cause the 2V drop using a resistor \$\endgroup\$
    – fcoppede
    Commented Aug 6, 2020 at 13:58
  • \$\begingroup\$ Thanks. I have a model of electret microphone with JFET, current is 500uA. The voltage at elektret will approx. 1,9V. \$\endgroup\$
    – csabahu
    Commented Aug 6, 2020 at 14:23

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.