Thermocouple Measurement Error

[cm_opi]

I'm working on a DAQ system that has 32 analog and 32 thermocouple channels.  The analog channels work fine, but the thermocouples are giving us a weird "stairstepping" effect.  As can be in the attached chart of a sample dataset, the thermocouple readings remain steady for a short time (with noise and sometimes a slight slope) and then jump to another value.  After the reading stairsteps to the new value, it will stay around that reading for a short time until it jumps again.  In working on this problem we’ve come up with the following:

The effect seems follow the trend of what the temperature is doing, it just doesn’t read the right temperature at all times.  Drawing a line between the stairstep points would probably give us what the temperature profile is actually like.

The effect only happens on the thermocouple data, and the time delay between stairsteps is much longer than our data rate.

Connecting a thermocouple to all 32 channels did not do anything to correct this effect.  We see the stairstepping no matter how many of the channels we leave floating.

If we assume that it’s not a software issue, what would the next step be in troubleshooting?  For your reference, we’re using a SCXI-1600 USB device connected to an 1101-C thermocouple box.  Any help you can offer would be appreciated.

[aNIta B]

HI cm_opi,

       I'm quite curious as to what you sampling rate is, and what your number of samples is- could you please post those numbers?   Another question I have is, what kinds of thermocouples are you using?  Have you tried using any different thermocouples.  Or, is it possible for you to try out a different terminal block with the thermocouples you have?  Please let  me know, thanks!  

[JamesBeams]

It looks like one of the components of the thermocouple measurement (the TC channel itself, the CJC channel, or Autozero if available/enabled) is being sampled only once per 50-100 samples of the other measurement(s).  If you can measure the raw values of each of these components separately and plot them it may also help narrow down the problem.  Since the steps are somewhat large I'd guess it's the TC channel, but it's only a guess and I have no idea why this would happen.