How to do better

Am setting up a temperature measurement system to evaluate different thermal insulation materials to be used in a heating unit in an office environment.
Some parameters design limitations include:
Insulated side must be able to be placed on high quality wood flooring with no damage (either heat or mechanical ie sharp edges screws etc)
Insulated side temp no greater than 90 F
Warmed side no more than 120 F.

BUDGET is always a limitation ours right now is measured in terms of pennies.

Instrumentation used is
Fluke Hydra Data Bucket 2635
Thermocouple type K comprised of field made junctions using TIG torch. The connections are approx 15 years old.


PROBLEM / ISSUE
Six TCs have been compared using a plastic bracket to hold them close but apart in open air at approx 85 or 86 ambient temp. Readings between all six vary by a maximum variation of 2 degrees. I have not succeeded to get instrument screen display to go to finer resolution than 1 ie cannot get it to go to 0.1 or 0.01 F. That is not a huge concern because if I log the data I get 4 or 5 digit resoulution on values in a file or printout.

I know there is a graphite unit used to weld thermocouple in a nitrogen (was it?) inert shield - I would love this - but - the budget - sigh,,,,


QUESTIONS
1. Can I get better performance by redoing junctions? What would you recommend silver solder or TIG torch.
2. Can I do this better and if so how with equipment on hand?
3. Any other bright ideas are welcome.

Thanks

Dan Bentler

Argon is the inert gas. Make sure wire is thermocouple grade, not extension grade. Bare wire thermocouples are fairly cheap from Omega.
No to silver solder. You just want the 2 dissimilar metals - silver solder will add other elements.

Can you not zero out the error in some way?
How repeatable are the readings from each probe when you repeatedly put it in an ice bath, or boiling water, or some other known temperature?
How much do the absolute values matter?

I mean if you are trying to compare different materials, isn't it the change in temperatures from one material to another that matters most?

You didn't mention what you have used the K thermocouples for previously, but if they have been exposed to high heat (~1000F or higher), or if they have any green oxidation on them; I would redo the ends. Cut off anything that has been exposed to the high temperature and just twist the ends together. As long as you are twisting clean wire together, it will work just fine for your application. You can solder them (soldering doesn't hurt the joint). Most commercial probes are welded as you don't have the temp limit of the solder to worry about. For your application (low temp, short duration, non-corrosive, non-oxidizing), a simple twist would work fine IMHO. Keep in mind that the reading will be the average temperature sensed over the entire twisted area; any place the wires are touching, really.

Keep in mind that thermocouples aren't all that accurate anyway. The limit of error for standard wire is the greater of 2.2C or .75%. RTDs would be a better choice of sensor for this application. The cheapest ones I found on Omega were $1 a piece; the drawback being that it's a 100 pack.
http://www.omega.com/ppt/pptsc.asp?ref=F1500_F2000_F4000&Nav=temc13

Here's some other reading I found on a quick Google search that might be good background reading for you.

http://www.picotech.com/applications/temperature.html
^Discusses accuracy in temp measurement.

http://www.picotech.com/applications/thermocouple.html
^Similar to above with a focus on thermocouples.


Brian

Phil the ice bath is next am setting up a test tank. You are right the delta T is what I am measuring but am measuring dT at several locations on the material on both heated and cool sides so would like best resolution I can get.

Brian -- TCs were coated with clear fingernail polish to avoid oxidation problems and were used in "clean" applications ie air or water or in a sensor well. No visible deterioration oxidation etc. Only one was used above 200 - insulation burned off for six inches so it needs some repair.

Gotta do some more reading but it appears to me I am near the end point for resolution using TCs. For thes narrow temperature (dT) values I have looks like RTD is better. Fluke will do RTD also.

Thanks for help and referrals guys. Keep em coming.

Dan

Sounds like they should be usable as-is, then.
You are asking for quite a bit out of a K T/C. There are other T/C types that would suit your temperature range a bit better, but going to RTD would make more sense.

Brian

I agree with Brian123, change the wire. For the cost of special limit of error type T wire, I'd go buy it and use it.

1) The chemical composition of T/C wire directly effects the Seebeck effect EMF generation. The burned part of the wire could be chemically altered.

2a Type T has half the ANSI uncertainty error of K or J wire.

2b Special limit of error wire has half the error of standard limit of error wire.

The foods and pharmas use Type T's (when they don't use RTDs) because if fits their temperature range and the inherently better accuracy of the Type T.




3) For your near ambient testing, all you need to do is to strip the T/C wire and twist the ends with a pair of pliers. You do not need to weld the junction. Welding is done on commercial T/Cs to ensure long life at high temps and vibration, but you're at room temp or slightly elevated - nothing extreme.

4) Make sure that wherever the cold junction compensator on the Fluke is, prevent it from getting drafts on it. It's probably heat sinked internally, but drafts on the CJC sensor will cause as direct a drift/shift problem as drafts on the sensor end.

5) You can silver solder a T/C hot junction if the T/C junction has good electrical/thermal connection between the wires and not solely through the silver solder because the Seebeck effect is due to the thermal gradient along the wire, not solely at the junction.