Perhaps pedantic:

Well - every metal has a thermoelectric, aka Seebeck, effect. Heat one end relative to the other and there is a voltage difference. What a thermocouple does is use metals with different rates of voltage production, typically joined at, well, the junction, but they could be joined by some other piece of some other metal and it would work just the same, though less conveniently packaged and not reporting based on the temperature of a single point.

Thus, the entire wire is what detects the temperature gradient. Which is also why there is a need for the "cold junction compensation" if you want precise measurements. https://www.tegam.com/what-exactly-is-cold-junctio...

3DDave,

In other words, if one end of this wire is inside a box and the rest of it is outside the box, I am measuring temperatures inside and outside the box?

--
JHG

You are measuring the difference in the temperature with the wires between the different ends of the wires.

3DDave,

So just the end of the cable is active?

--
JHG

It's all "active," but the opposite end is what is supposed to exposed to the temperature being measured; what makes the temperature measurement correct is the proper temperature compensation of the cold junction, which is supposed to mimic the measurement thermocouple system in its entirety, with the exception that the measurement end is at a known and controlled reference temperature. Some thermometers "cheat" by simulating a cold junction thermocouple electronically.

Every connector/connection inserts two additional thermocouple errors into the mix that have to be corrected for

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm

So typically, you don't run thermocouples (TC) any longer than you have to because the wire is expensive.
You run the TC to a junction, if this is not the measuring instrument then the junction needs to be made using special alloy links that match the TC so that you don't introduce a larger offset in voltage.
From there you run special lead wire, this is similar to TC wire but lower precision.
Where it terminates at your instrument the device will measure those junction temperatures and apply the correct cold junction corrections.
Each step needs to be done correctly or you can get huge errors.
The TC itself is usually made by welding the TC wire, either electrical spot welding or torch welding (usually oxygen/hydrogen so that no C can be introduced).
When we were making our own TCs I would weld up 5 or 6 if I needed 3. Then bundle them in a heat sink block and put them in the lab furnace. I would record the readings (in mV) at various temperatures. If one TC was too high or low on readings, I would mark it and then destroy it. But I would have a set that matched fairly well, and a spare.
If they were for process control or certification I would send them out to a real lab for calibration data.
For example the Pt TCs in our heat treat furnaces all had serial numbers and periodic re-calibration.
It also matters what TC alloys you use. Today there is no reason to ever use "K", the drift on it can be awful. Use "N" instead.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

Quote:

So typically, you don't run thermocouples (TC) any longer than you have to because the wire is expensive.

Ideally, yes, but we actually tend to make our thermal test TCs pretty long, so that they can reach the UUT (unit under test) inside the thermal chamber and get to the readout unit outside the thermal chamber. In one case we had about 15 ft of TC wire to accommodate the UUT functional test at temperature that involved spinning the UUT a few times in each direction.

I don't know if you meant "destroy" in the literal sense; my experience is that you can usually re-form the junction

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm

By destroy I meant the junction. I would cut them back 1" and re-weld.
And yes we had some long ones also, like 120' for profiling furnaces.
But these are a real pain to work with if you need a real calibration.
Most outside services will not touch them.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

Our calibration police(y) allows for on-site calibration, so we do ice bath cals if needed.

And, so long as the cold junction compensation duplicates the effects of all the junctions in the TC path, everything is technically accurate, albeit possibly noisy.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm

In industry T/Cs can run hundreds of feet. Often the wire used in the section with the junction is higher purity than what you'd want running thru conduits in wads everywhere. To deal with this thermocouple extension wire is used which is not for "measuring" but is used only to carry the signal back to the controls input. It costs less but still prevents any shift in reading due to the junction with the measuring length of T/C wire.

Keith Cress
kcress - http://www.flaminsystems.com

Thermocouple instruments are relatively expensive compared to other methods. While the sensor is pretty cheap there has to be a reference junction to get a real temperature. The temperature of the reference junction must be known , and measured, by some other technology, such as an RTD or physical, real-life ice bath.

A lot of people think that thermocouples must be the greatest thing ever for measuring temperature because they have high-tech sounding name and you can get so many different types, with magic letter identifications.

The truth is that they are overall pretty crappy.

Not particularly accurate, even when made and installed properly.
Easy to make improperly.
Easy to install improperly.
Subject to electrical noise.
What Compositepro wrote.

That said, for many industrial applications where you really only need to know the temperature to "about" accuracy, they work just fine.

The tell is that accurate temperature references are either RTDs or PRTs, specifically.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm

If you are at low enough temps use another method.
The only thing harder that using TCs correctly is optical methods.
They are even easier to get wrong.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed