This seems like a market opportunity.
A clean, amplified electret mic with 0.6V bias and 1.2Vpp at its reasonable maximum.
Yes. My hope is Onehorse or Sparkfun would make such a product. Ideally, it should also have some provision for running from a battery, even if just a JST connector that mates with commonly available batteries and just connects it to VIN.
Later this year I'm going to be putting quite a lot more work into sound-reactive LED control (and non-blocking Neopixel library), so there's going to be more need for a good quality, ready-to-use microphone board. If nobody wants to pick this up, PJRC will probably end up making one. But I'd much rather focus on Teensy and the software side.
No, that low 0.6V bias is unusual. Typically they are at 50% of the A/D input range in order to maximize that range.
The range is 1.2V, so this is 50%.
In older 8 bit AVR and PIC chips, indeed 1.2V range is unusually low. But even then, some AVR offer both 2.56V and 1.1V.
In modern 32 bit chips, 1.2V internal reference is pretty much standard. This isn't just Freescale. The STM chips and NXP's LPC chips have 1.2V internal reference. Atmel's SAMD chips (used on most of Arduino's 32 bit boards) have 1.0V internal reference.
Paul has a very good rationale for that decision, I'm sure!
It's a simple matter of the chip's internal reference voltage being 1.2V. The reduction in range is worthwhile for rejection of power supply noise on the reference.
Ordinary line level audio is also approx 1Vp-p, so this matches pretty well to the commonly used consumer gear audio signals.
That microphone module is still signing with a wonky 1.2-1.5Khz signal superimposed on the output.
Adafruit's design leaves a lot to be desired. If I can get someone to create a board with 0.6V offset, hopefully they'll be willing to do a better design, and ideally go through a couple iterations if the first attempt isn't great.
(click for full size)
This schematic has 2 paths for power supply noise to couple to the signal where a quiet signal matters most, before the amplifier gain. The mic bias at least has some attempt to filter away noise, but the 0.1 uF capacitor isn't nearly large enough. Together with 1K impedance, this put the single-pole corner frequency around 1.6 kHz, so it's only effective at filtering away noise in the top few octaves of human hearing.
The 1M resistors are a lesser issue, only because their impedance is so high. Their noise contribution will be about 60 dB less, because the mic imedance is 1000 times lower. Still, using an extra resistor and capacitor on this path would add minimal cost, and a smaller capacitor would work fine because the resistors are higher.
The power supply filtering could be done much better. Isolating the grounds is probably worthless and might even be harmful. This is the sort of thing that should be tested on a first prototype and iterated. Excellent analog design isn't easy and this is the sort of thing I'm not able to predict in advance.
But it is pretty easy to predict these ferrite beads are really only effective at RF frequencies. At audio bandwidth, they're just wires. FB2 probably wants to be become something useful for audio, like 100 uH. For example, Digikey 587-2048-1-ND might be a good starting point. However, even this is on the low end, less than 1 ohm impedance in the middle of the audio band. Even something like 587-2509-1-ND isn't a lot. Adding a series resistor, like 10 ohms or even 100 ohms, would also be a good idea. At the lowest frequencies where no reasonable inductor does any good, a resistor in series will really help. To make this effective as a filter, a large decoupling capacitor would also be needed.
Adafruit didn't use any decoupling capacitor at all, even though Maxim says one is needed. A 0.1 uF ceramic probably wants to be used in parallel with something big like 100 uF or more. This might seem excessive, but consider the most important audio frequencies are very low, just a few hundred Hz. Really good audio design would probably do much, much more, perhaps even a LDO regulator. But I believe an inductor and resistor in series and a pair of decoupling capacitors can probably be good enough.
The other really important thing missing from Adafruit's design (and almost all others in the maker market) is low-pass filtering. Well, other than the natural response of the microphone itself. Ideally, there should be at least a 2-pole filter, which costs another opamp and more resistors and capacitors (good NP0/C0G ones). If it's just a normal Sallen-Key circuit, the corner frequency should be fairly low, perhaps 6 to 8 kHz. Yes, that attenuates some of the higher frequencies people can hear, but attenuating signal and noise above the Nyquist rate is far more important. Any higher frequencies (above Nyquist) that make it to the ADC pin become terrible aliasing noise in the audio band.
Yes, I know this all adds up to many more parts, and some of them physically larger. My hope is the usual thought process of "can a smaller/cheaper part work" or "are so many parts needed" can give way to thinking more like "audio quality shouldn't suck, even when used in a hostile environment like a many-LED project". Onehorse, if you're reading this, any thoughts?