Simple MicroProcessor PI Idea

[hdphilip]

I'm using a 10-bit ADC and 64 samples in my averaging, for an effective 13 bits

but in order to get the 13 bit sampling from a 10 bit ADC, you have to pay the price!
your sampling rate would also be divided by 64, small target's may also be averaged out.

just a thought

Philip

[Davor]

It would depend only upon the available/required dynamic range. If you can supply 96 dB of dynamic range and a bit more for noise - go for full 16 bit. Otherwise averaging does it for you. A whole point is that nothing spectacular happens between two successive samples, and if it does - it is just noise.

E.g. a signal that is sampled at 16 bit with 4 lsb of noise is in effect a 12 bit sample.

Lets see what it takes.
Say you need to sample at noise level at source impedance of 1kohm @ 100kHz bandwidth (~1,3mV) and pre-aplification of 100x ... timed 65536 for full 16bit resolution ... you reach ~8.5V
Fine. But you typically have more pre-amplification gain, and your ADC can cope with a bit less signal so the largest sensible resolution is 12 or 13 bit. Everything beyond that is just sampling below noise level.

[goldfinder]

16 bit marginal

A few years back I had the same Idea and came to the conclusion that 16 bits is barely enough. To get the noise out you need a couple more bits and then you still have to do some digital filtering. At that time 18 or 20 bit ADCs were too expensive. What I came up with is to do a little more processing with the analog side and then do some smart DSP in the remaining in the uP.
Goldfinder