What the hell do those guys want to do with a 24bit-ADC? This discussion makes me sick!

May be that some ADC has 24bit resolution, but real world shows drift effetcs with are universes worse! Let's have some estimation:
Assume, that input signal of 24bit-ADC is amplified by some real amplifier. Further assume, that gain of amplification is defined by the relationship of two resistors. In first order approximation drift of amplifier gain will depend then on the sum of drifts of both individual resistors. We don't want to talk about tolerance of components, because this can be nulled by calibration. So, we only talk about drift.
A normal good metal film resistor, which is most widely used in 'precision' front end sensor electronics has tolerance of <+-1%, temperatur drift of <+-50ppm/°C and long term drift of about <+-0.03% /2years.
Let's only take temperature drift and assume that ambient temperature changes to about +-20°C. Then, temperature drift of one such resistor is 20 x +-50ppm = +-0.1%. In worst case situation, temp drift of both resistors has opposite sign, so, gain drift of this amplifier can be about +-0.2%.

But +-0.2% drift corresponds to a resolution of 9bit!!!

So, for most of all front end sensor electronics temperatur drift has the same effect as using an ADC with a resolution of 9bit.

Ok, now you will say, 24bit-ADC will need the use of 'real' precision resistors. Such a 'real' precision resistor has a tolerance of <+-0.1%, temperatur drift of <+-10ppm/°C and long term drift of about <+-0.02%/2years.
From same calculation as above, we can see, that gain of amplifier shows now a temperature drift of +-0.04%. This corresponds to a resolution of 11bit!
But long term drift can no longer be ignored. Additional 2 x +-0.02% gives finally +-0.08% for our 'real super precision resistor' solution. But this corresponds to a resolution of 10bit!

So, why the hell using 24bit-ADC, when with 'real super precision resistors' only a drift can be achieved, which corresponds to 10bit resolution?

Ok, let's make the resolution of ADC 2 bits higher than 10bit, then resolution of ADC is much better than drift of components and is negelectable. And let's have additional 2 bits because ADC has also some drift. Then we end up with a 14bit-ADC. This is really optimum when using those 'real super precision resistors'. But anything else is bullshit!

Kai