Microphone – Default Gain of Electret Microphone Amplifier 4466

First thing I notices is the time constant with your input capacitance and 1Mohm resistor - it's 47 seconds or equivalent to a high pass 3dB frequency of 0.0034 Hz.

This might take ages for your inputs to settle - make the input capacitor 47nF is my recommendation - this puts the high pass frequency at about 3.4Hz. Even a 10nF wouldn't be unreasonable (high pass of 16 Hz).

Your op-amp is a TL084 so I'm less worried about input current leakages causing offset voltage problems AND I'm happier with non-polarized caps at the inputs because if it is an electret mic and has a bias voltage on it, the electrolytic (have to be polarized correctly) input caps will become reverse biased and this may indeed be a possible reason why you see a problem.

Regarding your 50k pots, these will potentially give higher noise than say 5kpots and with 5k pots you can reduce R5, R10, R15 and R16 to something like 10ks also reducing noise. Noise is proportional to the square root of resistance so always best to go lower a bit if you can.

C7 and C12 are also problematic in the the signal they see may become reverse polarized. The time constants you are using around these circuits are very large given that those caps are 47uF - go for 1uF and non-polarized. U2 just appears to be a gain of ten stage so lose the centre biasing - 0V is centre and that is just fine assuming your Vcc and Vdd are largely equal and opposite.

You are using TL084 output amplifiers but this is really daft because you have an output pot that will likely introduce an impedance of about 25k (mid point). Just use pots and get rid of the extra 3 op-amps that feed the individual outputs. C9, C10 and C11 are not needed either or if you feel happier with them use non-polarized 1uF caps.

Your math is about right but what's missing is that

(a) 1Vrms is 2.8Vp-p for a pure sinewave and well over 5Vp-p for normal speech, so (even without headroom for speaking a little louder) 1Vrms is too much output from your mic amp into a 5V range ADC, you probably want some headroom - maybe6-12dB, so 0.5 or 0.25V rms for a 60dBa input.

(b) if you're measuring normal speech as 60dBspl, you're measuring it from some distance away - like, across a room. Or at least 1 metre.

So if you're building a bug for espionage purposes, a gain like 1000-4000 is probably what you need.

But for other purposes, speak into the mic and you'll need much lower gain, like the numbers you're seeing from other sources.

(c) Sound level control is HARD - for top sound quality you need an experienced guy continually monitoring and adjusting the gain. for other purposes, don't imagine the problem stops with the mic amp - often it's just the first gain stage (and optimised for good noise performance).

Later stages involve more gain, and frequently some way of adjusting the gain, like a compressor, limiter to keep the audio levels within acceptable limits.