Hi j29388,
This sounds like an interesting application. I have a few questions to help clarify some things, and I have a few comments in reply to your questions.
QUESTIONS
1) Are thermistors or RTDs options for you to replace thermocouples?
The temperature range of 0-100 degrees C is fairly limited. In fact, it's limited enough that other transducers such as Thermistors and RTDs would survive. Are these options for you? If they are, I recommend pursuit of higher accuracy with those transducers. Although thermocouples can be accurate, they are inherently compound measurements consisting of a primary voltage measurement and a secondary absolute temperature measurement (usually in the form of a thermistor, although sometimes in the form of a fixed point reference cell). It is often easier to achieve higher absolute accuracy with thermistors than with TCs if the temperature range is limited enough.
2) Your calibrator has a 2 decimal display. Does that mean x.xx degrees Celsius?
3) What is the specified accuracy of the calibrator you are using?
Just because it displays extra digits doesn't necessarily mean it is actually accurate to that much precision.
4) What do you mean that you "set out to calibrate"? Do you mean verify? Or do you mean to adjust some constants in the processing of the data?
5) Are you looking for absolute accuracy or relative accuracy between your thermocouples?
ANSWERS
A) "The thermocouples still seem to have large deviations from one another..."
Although I don't know the scale of what you mean by "large" it is not surprising that the various channels vary from each other. This can be because of TC wires themselves not being matched to each other, and from channel specific errors in your thermocouple instrument (the NI 9213). My guess (without being able to see your system) is that the bulk of the mismatch comes from the 9213, especially from the Cold Junction Compensation (CJC). A portion of the CJC error comes from isothermal error, which is the difference in temperature from the point where the CJC senses and the point where the TC wire meets the instruments connectors. This would be a unique error on each channel because each channel would have slightly different temperature gradients between it and the CJC inside the instrument.
B) "I was hoping for 0.1C accuracy for the project"
This will not be trivial to achieve. This is certainly beyond what we specify in our manual. I'm willing to bet it's also well beyond what the TC wire manufacturer specifies. You will definitely need to do a custom calibration and use best practices to squeeze out every last drop of accuracy possible.
C) "What can I do to improve this?"
It sounds like you have a good head start by employing your thermocouple calibrator. You will need to use this calibrator to provide reference points so that you can adjust various constants in your post processing so that your channels match.
In addition, you should follow some best practices to get more stable performance out of the 9213 than what it is specified for. I wouldn't say that best practices are always required, but the accuracy you are seeking is well beyond specifications. If you are seeking relative accuracy between channels, then I would say that the number one most important thing you can do is to put the 9213 in some place cozy and protected from air drafts and sudden temperature changes. Number two would be to get the system fully running and acquiring data for a long time before you calibrate. These two recommendations will ensure stable thermal gradients inside your instrument and therefore allow you to calibrate out the isothermal errors.
D) "Am I calibrating the TC wire or the 9213?"
You are calibrating both. This is a system level calibration which is the most accurate overall method because it lets you eliminate all stable errors from the whole system.
E) "Isn't the TC just two lengths of wire?"
Well,... yes, it is just two lengths of wire. But very tiny shifts in the alloy properties of the materials can cause significant changes in the thermal to voltage characteristics of those wires. Those wires are probably from the same spool, and are probably similar to each other, but they *might* not be entirely homogenous and *might* have different properties from each other. Additionally, manufacturers differ greatly from each other and only need to meet NIST defined accuracy requirements in order to call it "T type". Standard allowable error for T-type is ~1C. If you want to use a wire for greater accuracy than the manufacturer specifies it for then you'll need to do your own custom calibration. You'll even need to repeat the calibration periodically because thermocouples experience various types of destabilizing decay (especially for base metal types like your T-types).
SUMMARY
Whew... that's a lot of information. I'll try to summarize the key points: 1) It will be difficult to achieve 0.1C overall accuracy with a thermocouple. 2) Since you have a calibrator, you can drastically improve uncertainty but you will need to perform some sort of software compensation. 3) You will need to follow some best practices to stabilize as many errors as possible.
All in all, you can certainly make improvements as long as you're willing to be careful and perform some math. As to the math... we'll need another post for that.
