I am using ADS1118 at G=256 in a 16-channel K-thermocouple isolated DAQ at a fixed 10 S/s. With a small correction for the OVP network it works absolutely great, with less than +/-0.5 degree variation across 32 sampled parts from room temperature to 1300 degrees C including CJC. Very impressive.
However I find the accuracy is not quite as good when the ADC input is near zero volts or negative. There is a very repeatable non-linearity on all parts causing a 40-50 uV step function between -2.0 mV and +2.0 mV ADC input (roughly +/-50 deg). Granted that is only ~1 degree, but the positive range is so good it would require me to double my accuracy spec: they all read a consistent +1.0 (+/-0.2) degrees at 0.0 input and -197 at -200 C input from Fluke calibrator. CJC is at room ambient 26 deg during all tests.
Changing Rpu/Ppd and bypassing the 1K OVP circuit make no difference. Is there anything you can suggest, or is this small disjoint characteristic of this chip? Happy to supply schematic, layout and measurements if helpful, it's the datasheet circuit.
Thanks much,
Gary Schneider, Engr. Mgr, Hi-Techniques Inc.