An update on this circuit. I finally had a clear day to prototype the left half of the circuit.
?
I realised that an AD8220 was just an AD8226 with an extra FET stage in the input. Seeing as I've switched to using FET electrets, this stage was no longer necessary, thus I could go with the cheaper AD8226.
?
The main finding is that it absolutely does seem to need a low impedance reference voltage. The output was minuscule if I used a simple potential divider even with the parallel 47?F cap for the reference voltage.? When I shorted the reference voltage line to ground, I suddenly saw proper amplitude output, albeit the negative side clipped.
?
The datasheet for AD8226 elaborates on the need for a low impedance reference a bit further, saying that less than 2 ohms impedance is ideal and advocates using a op-amp buffer driven by a potential divider as a half voltage source in single rail setups.
?
I bodged in an AD8001 (found one lying around) and it gave a nice buffered 2.5v, and still within 10mA at 5V.? This seems to have added some RF noise sensitivity, but it might simply be the AD8226 now operating in a happy state with too much gain.
Another finding was that AD8226s (and AD8220 too) need a current return path from the reference voltage pin to the inputs. I added a 1M resistor between vGND and each input.
?
So this is way overkill when the industry standard seems to be a cheap unidirectional FET electret capsule and a bias resistor!
?
