Microphone array background

 Microphone array. Design background.


Digital Design

Use a digital microphone. Something with a microphone based on the LMV1024(which I can't seem to find!) or maybe an AKU2004 if I can find someone that will sell it to me.

These will spit out a PDM encoded data stream @ ~ 3MHz, so something may be needed to move the data from the microphone to the sampling board. I don't think it practical to run an unbuffered 3Mhz signal over twisted pair for 2 meters (is it?)

Possibly use a LVDM transceiver to drive it.

Analogue design


WM-61A is a ~62dB SNR, but sensitivity is -35dB. Conversation @ 1 meter is about 5 * 10^(-2) Pa. -35dB senstivity is ~20mV PP, which gives 1mV PP for conversation level sound at 1 meter distance. Trying to feed into a 24-bit ADC gives ~7pV per step!

Ok. That's just crazy. Using a higher gain microphone such as the SPM0208HE5 which has much higher sensitivity at -22dB. (it can be programmed for zero gain, or 20dB gain. No idea if that's -22dB is before or after said gain). It's response curve is not exactly flat either (~ 5dB higher @ 8KHz).

So a pre-amp is definately needed. Something with about 200x gain would be a reasonable idea (~ 43dB gain)

Maybe the SSM2167, microphone signal conditioner. It operates of 3V, provides variable gain up to 60dB gain and has both ADC overload protect, and noise suppression. But, dammit, the noise floor is about -55dBV (~ 1mV rms) which is our typical level. So it would need pre-amp in front anyway.



AD8656 looks like a good op-amp to use (significantly cheaper than the OPA2228 used in the nist mic array).


#2. Use analogue microphones (WM-61A is good for ~ 62dB SNR), use shielded cable, runs them analogue into a CS5340 stereo ADC. The CS5340 will deliver 24-bit samples @ 96Khz (digikey has them at $9 each!) The CS5340 also sucks because it needs an active filter on the input (each op-amps).

This person has a multi-channel sync'ed design based on the CS5333 (a previous generation device).

These will output I2S which the NWG100 can take in via the SSI units.


Alternative ADCs:


PCM1802 has digital anti-alias filter. About $5. Better SNR. 5th order sigma-delta modulator.

PCM1807/1803A (~$2.50) Has single ended inputs, doesn't seem to need analogue buffering. xxx7 is SPI controlled. Analogue anti-alias filter. Lower SNR.

PCM4201. Differential input. Internal PLL.

PCM4204  (quad-ADC, 24 bit, 200KHz sample rate, ~ $10). Seems to require a fair bit of analogue interface circuitry. 4 x OPA1632 (@ $4) plsu 2 x OPA227 (@ $5) plus passives = $37 in chips plus passives, plus board, plus sockets.  Also needs dual supply (5v + 3.3v)

AD1974. quad-input. single 3.3v supply. differential inputs. Internal PLL, can generate clock from crystal or LR clock. 48-LQFP.

AD1939 codec. quad-input.

AD1988 codec. 6 ADCs, programmable gain, mic bias, 10 DACs. 48-lqfp. HD-audio interface!? HD-audio interface is more complicated than NGW100 SSI can handle. alas.

UDA1361 dual-input, single supply.