Sounds like a thermocouple. The other two are not polarity sensitive. Is there a calibration pot or some software calibration routine? I have used a lot of digital equipment and all of it was only as accurate as the last calibration.
When I do water sampling for the state, I am required to calibrate the machine before I start and run the calibration again after. It is amazing how many times the "after" test fails.


Greg Fretwell