Help: Noisy microphone design problem

[EdenPal]

Hi All,
I'm new here.
I designed a PCB with ESP32 (standalone chip) and I connect microphone to the ADC pin of the ESP.
Claude wrote code for this design that the ESP32 will stream the sound to my phone by the WiFi channel.

I hear the sound but it is very noisy. I tired to add a software LP filter but it is still very noisy.
In addition, I added a hardware LP filter (RC filter after the opamp) and it still noisy.

I added the project file (EasyEDA) to this message.

Anyone can help me to understand what no good with my design?

[MariuszD]

Desolder L5 and connect an independent battery to C63; this way we will find out if the noise is transmitted through the power supply.

LMV321 has high noise, but I don't think it is the main cause.

[marco.van.steen]

A low signal amplitude into the ADC can be the cause.
In addition to the previous suggestion, can you check for us the DC voltages, referenced to GND, on:
1) pin 1 of the microphone (must be >1.8V)
2) MIC_out (must be 1.65V)
3) Node "1.65Vref" (must be 1.65V)
4) Pin 3 of U12 (must also be 1.65V)

That way we could check if the mic + preamp biasing and supply levels are OK.

[rickjf]

Your post got me curious about how the mic and preamp works.

Found an excellent detailed reference design by TI of your topology (see attached file.)
Single-Supply, Electret Microphone Pre-Amplifier - TI Reference Design

Question - what is the nature of the noise:
  low frequency hum or buzz, high frequency tone, wideband static, clipping or other?

[EdenPal]

Thank you very much for your answers.

marco.van.steen, about your answer, I checked all the voltage nets before I continued to the code.

MariuszD, thank you for the idea. I supply the microphone circuit with battery, like you said and the noises disappear.
Do you think that if I will add capacitance to the buck converter it will help?
I try to think what to do for the next project with sensetive analog design with DC-DC cirucit that supply it (not LDO).

Thank you very much!

[MariuszD]

Try to make a better filter, use a large inductance for L5 and a higher capacitance in parallel with C57.

If you have an oscilloscope, check the ripple on the 3V3 and 3V3_f rails.

L5 can also be replaced or connected in series with a resistor of a few hundred ohms.