> Max132 is a 18 bits ADC from Maxim, it is somehow nice to use
> and quite stable, but as a rule, pcb layout and wide and low resistance
> ground conductors are very important to consider when dealing above
> 12 or 14 bits.

I made software for a board with MAX132 and AVR. My experience was that the
MAX132 might become flaky when breathing (!) on it (no I didn't eat old
pizza :-). Although zero calibration was used to compensate for temperature
before reading each any every sample, the values went up through the roof!
When breathing on it extensivly the values even crossed the 21 bit border,
and the MAX132 then did stop completly (did not signal EOC anymore).

After about 10 seconds the results settled back into the valid range, and
within 60-90 seconds they returned completly to their normal precision.

I didn't design the hardware for that project, but everything looked fine,
and we tried several boards/max132s, and checked for flux, moisture and
heat problems. With no success. It all appeared to be the MAX132 itself.

Maxim did say they do not know of that problem, but did not want to send
an evaluation board for us to try or reproduce the problem themselves
(at least they said they don't).

So until now it's not sure if it's really a MAX132 problem. Maybe
one of you PICLIST readers has a working MAX132 board on your desk
right now to try it and confirm/deny?

Until then I'll not use that particular device in my own designs (but
others from Maxim which I am very satisfied with).