I'm starting to wonder whether the 1k resistors are too small, as they're smaller than the 2.2k output impedance of the microphone.
Those are the output impedance of the microphone. If you look at the mic capsule's datasheet you'll see an equivalent circuit:
I don't know why manufacturers always show the FET as a triangle. This is how it's actually configured:
So this is really a common source amplifier:
The output impedance of a common source amplifier is just \$R_\text{D}\$, the drain resistor, so when the datasheet says "output impedance (Zout) 2.2 KΩ", they really mean "output impedance of our example circuit".
With \$R_\text{S}\ = 0\$, the voltage gain of the common source amplifier is proportional to \$R_\text{D}\$, since the FET acts like a current source, so the resulting voltage is determined by V = I(FET) * Rd.
What resistor should you choose? It depends. Generally you want high gain in the first stage so you can lower the gain of subsequent stages, which lowers noise. The distortion also decreases as gain increases. You can't increase \$R_\text{D}\$ forever, though, there's a point at which current is too low and distortion increases and gain drops suddenly. Also, if your microphone is expected to pick up high SPLs, you shouldn't increase the gain too much or it will clip.
I don't know how to optimize the gain based on the parameters in the datasheet, but I'd like to know. For mass production, the gm of the FETs will vary from unit to unit (and possibly the FET type will be changed from one capsule to the next even though they have the same part number), so optimizing for maximum gain for a specific FET is probably a bad idea.
Negative Feedback
It gives you:
- better linearity
- lower gain
- even lower output impedance
- opportunity for frequency-dependent gain
As shown, you get DC feedback via R1+R2. But R1/C1 forms a low pass filter for signals coming from the collector and going to the base. Less negative feedback means less reduction in gain for high frerquencies. Cancel out all those double-negatives and you're left with a high-pass filter.
The transistor below corresponds to Q5 in your drawing.
simulate this circuit – Schematic created using CircuitLab
Best Answer
Actually the circuit you have after a few tweaks would be pretty good, although I didn't look up the particular opamp to see if it is appropriate for this use.
The first change I'd make is to put a cap to ground on the + input line to the opamps. Since the impedance is 5 kΩ, 1 µF would give you a rolloff of 32 Hz, which is fine since we're only trying to keep power supply noise from feeding into the amplifier.
The 220 nF input capacitor should be a bit higher. As it is now, it will form a low pass filter with the 4.7 kΩ resistor in series with it with a rolloff of 155 Hz. That's a bit skimpy, although perhaps your mic can't support frequencies lower than that. Still, I'd make it at least twice what it is, but more likely use a 1 uF cap. However, make sure not to use barium titanate or related ceramics because those exhibit microphonics. Just putting two of the caps you already have in parallel could be good enough.
Your voltage gain is only 100. You probably want more to make 5 Vpp from a electret mic. With this gain, you need 50 mVpp from the mic, which sounds a bit high. Maybe you're fine as is if you will be using a high enough resolution A/D so that the signal can only use part of the range most of the time, but you want overhead in reserve for short but loud sounds.
The 1 uF output cap allows you to float Vout at any level you want. Since you want 2.5 V, float it at that. A simple way to do that is to connect it to a voltage divider from the 5 V supply. Two 100 kΩ resistors would do that nicely.