Two suggestions:
1... Put one of the thermocouples into an icebath: i.e. crushed ice and water. That should keep a steady temperature of 32 degF.
If your signal shows a cyclic variation, then it's not in the environment, it's in the signal handling.
2... Remove the averaging and look at your raw data. There are bound to be clues there.
I suspect grounding problems, as suggested earlier.
Another possibility: How are you timing the acquisition? When I first ran cDAQ, I wasn't aware that they had their own timing source - I was trying to manage the timing on the host, but the difference in timing between the host and cDAQ was small and steady and created some similar problems. Every few minutes there was a hiccup because the timing skipped.
HTH
