PI for hammered coins suggestions?

[Carl-NC]

For several years i've been using some audio modules with WM8738GED, to make audio links.
...
I am just surprised that no one from the md world did not come up with similar design.

The Minelab X-Terra is direct sampling and uses the WM8731 which is a 24-bit CODEC that samples up to 96k. I believe the Equinox uses a newer chip that samples even higher.

What is the advantage or disadvantage of using a 24 bit ADC vs a 14 bit ADC with a 1000 gain amplifier? Maybe a 16 bit ADC and a 256 gain amplifier.

More bits is always better if you can manage the analog noise. Realistically most of the 24b converters are sigma-delta and have 19-20 bits ENOB at most, which is still better than an ideal 16b SAR. But you can overdo it, see the example below.

I suspect that noise is very much the limiting factor.
For example, most 16-bit ADCs will have an equivalent number of bits of around 14 due to their own internal noise and non-linearities etc.

Some of the new 16b SARs have SNRs approaching the quantization noise.

For a back-of-the-napkin calculation, let's assume we have a 10V power system (+/-5V) and the preamp has a G=1000 and a feedback tau of 1us and uses a 1nV/rtHz opamp. Ignoring all resistor noise, the output noise of the opamp is 1nV*1000*sqrt(PI/2*1/(2*PI*1us)) = 500uV rms.

If we want to exactly match this to ADC quantization noise then the vq = 500uV so vlsb = 500uV*sqrt(12) = 1.732mV. For a 10V system 10V/2^N = 1.732mV so N = 12.5 bits. In this case a really clean 13b ADC will barely do the job. But in a PI system we can also do a lot of averaging to pound down the thermal noise and more precision in the ADC will help. A 16b ADC give 8x the precision which is good but an 18b ADC has 32x the precision and will give cleaner averaging results.

[ivconic]

Yes, i forgot X-Terra.

[jeep6789]

Is there anyone is willing to design a schematic,