Hi Guys,

Is it utopia to imagine we could take a PDM bit stream (carrying audio signal let's say up to 15KHz) and just put a second order LC low pass filter to get the average value out of the stream?

I'm looking at a NATIONAL microphone (LMV1022):

http://www.national.com/opf/LM/LMV1022.html

www.national.com/an/AN/AN-1784.pdf

And I'm thinking that by just using a discrete low pass filter (with a buffer to avoid loading the microphone output), we should be able to reconstruct somewhat the original audio? They are using an FPGA to do a low pass filter + decimation + I2S packaging. But after all, if the PDM goes direct to an analog load, decimation is not needed and an analog low pass should be enough. Am I far from reality?

I'm trying to see if I could use the LMV1022 to replace a conventional (analog) microphone in an existing design.

Any suggestions appreciated. Thanks beforehand,

 

The data says: maybe. See the Wikipedia article - http://en.wikipedia.org/wiki/Pulse-density_modulation.

I would think that the bit stream would have to be buffered. Applying the buffered stream to an integrator might do the reconstruction you want. At least it will be well filtered.

 

Thanks for your answer. Appreciated.

You think a single order integrator will suffice? I was more thinking using a low pass filter as an integrator, order maybe of 10 (Tchebychev) for 100dB attenuation between fcut=3Khz and fstop=8KHz. The bit stream frequency is much higher than my 3KHz fcut, so the low pass will act as an integrator.

It maybe overkill but I'm experimenting. After, it will be easy to reduce order of the filter with components from linear technology like LTC1068.


Thanks for your time.