7
\$\begingroup\$

Thermocouples have stated tolerances [wikipedia]. However, the characteristic behaviour is defined by the materials in contact, so there aren't size or shape considerations as for RTDs. What are the main sources of the variation/tolerances in thermocouples?

A comment to this answer blames variation in metallurgical properties, which seems plausible, but this is unsourced.

\$\endgroup\$
  • \$\begingroup\$ What now, in thermocouples, RTDs or both? \$\endgroup\$ – PlasmaHH May 24 '16 at 10:33
  • \$\begingroup\$ @PlasmaHH Whoops, typo - thermocouples \$\endgroup\$ – Gremlin May 24 '16 at 10:44
4
\$\begingroup\$

In my experience the difference between thermocouples of the same type typically comes from the the manufacturing process. I have been making a number of K-type thermocouples using electrical arc welding, with "exactly" the same wire lengths, at the same temperature and with the same amplifier they display different temperatures in the range of ±0.5 Celsius from each other.

My only explanations for this is that during the fusing different amounts of impurities are introduced into the junction, which is mostly based on the quality of the conductors, and that the phenomenon is related to the temperature of the junction at the moment of fusing.

Unsurprisingly Wikipedia has some useful info on this as well.

Basically, I'd say the guy in the thread you linked is right.

| improve this answer | |
\$\endgroup\$
  • 1
    \$\begingroup\$ Impurities in the junction will not affect the temperature reading provided the region with impurities is isothermal. More likely it's variations within the wire. \$\endgroup\$ – Spehro Pefhany May 24 '16 at 15:19
2
\$\begingroup\$

Thermocouples are nothing but two (or more) metals or alloys in contact, so the inherent variations must be metallurgical in nature. Metallurgical effects include the alloys, how evenly distributed they are during the drawing of the wire, and effects such as annealing and cold-working.

Of course the construction of a thermocouple can lead to differing measurements depending on way the thermocouple is made or mounted and how heat flows from the object being measured down (or up) the leads.

Extension wire is not always made from the same type of metals, or made to the same specifications, so the entire thermocouple assembly may show errors if the transition to extension leadwire is far from room temperature. The main reason for using different materials is to save money and allow the use of robust conductors, especially with precious metal thermocouple types (most commonly types S, R, B but gold has been used as well).

| improve this answer | |
\$\endgroup\$
1
\$\begingroup\$

To understand thermocouples there are three things you have to keep in mind:

  1. Thermocouples measure their own temperature. They do not measure air temperature, water temperature or anything else; they measure the temperature of the wire.
  2. Thermocouples do not measure absolute temperature, they measure differences in temperature or relative temperature. If you have a thermocouple based temperature gauge that gives an absolute reading another temperature reference is used also. (This is called cold junction compensation.)
  3. Thermocouple junctions are important and insignificant. The thermoelectric effect is a wire effect, the junctions are just where you change wires. There are people who are trying to make the junction as small as possible because they think that the thermocouple measures the temperature at the junction. thermocouples measure the temperature across the wire. It is not the thermal mass of the junction, but the thermal mass of the wire that is important.

The most common source of variation in thermocouples is the wire (which results in changes of thermal mass and electrical resistance and in millivolt systems like thermocouples resistance is very significant). but many times significant source of differences in thermocouple readings is not the thermocouple, but the rest of the system. Common mode noise can be a problem as can rounding error. Wire length is almost never a source of error (assuming you are using thermocouple wire), but shielding, connectors and oxidation can all be problems, specifically hot junction oxidation (even on welded junctions) can lead to errors and is the most common reason to replace old thermocouples (Heat, dissimilar metals and current is a recipe for unwanted reactions.).

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ This doesn't answer the question, because I'm talking manufacturer-specified tolerances that we can safely assume are based on repeatable measurements on the same system. Also, resistance shouldn't be significant in thermocouples because there is no current flow. \$\endgroup\$ – Gremlin May 24 '16 at 13:15
  • \$\begingroup\$ It is only an approximation (one that can be off by quite a bit) that a thermocouple EMF represents a difference in temperature. \$\endgroup\$ – Spehro Pefhany May 24 '16 at 13:29
  • \$\begingroup\$ @SpehroPefhany, the thermoelectric effect is not directly related to EMF. (and I know that it is non linear). \$\endgroup\$ – hildred May 24 '16 at 14:04
  • \$\begingroup\$ @Eoin, there is always current flow in metal, that is what makes it metal (otherwise it would be a salt or gas), but more than that thermocouples work because heat introduced into a metal increases electron movement in a measurable way (and using dissimilar metals introduces different movement of electrons in each wire. The voltage and current produced by the thermocouple is small which makes resistance very important. remember v=ir. As to manufacturer tolerances, any good engineer will measure what his product can do, but will tell you it is not as good as it is so the only surprises are good \$\endgroup\$ – hildred May 24 '16 at 14:14
  • 1
    \$\begingroup\$ @Eoin, In millivolt systems the voltage is so low that even with high impedance instruments resistance is still a factor. It is true that in non millivolt systems that the instrument impedance will dominate to such an extent as to reduce its error to negligible, but in a millivolt system which might be able to shrug off a couple ohms of wire resistance a single bad connection with say 50 or 100 ohms will introduce error. Also just because you can have voltage without current flow, you cannot measure voltage without it. \$\endgroup\$ – hildred May 24 '16 at 15:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.