Another suprising source of noise and error is at the thermocouple connections. This can introduce an error voltage equiv. of almost 10 degrees under some conditions. I correct this by connecting a second thermocouple in revers and series at the connector block. This will cancle most of the thermo-noise and correct the error created by the 3 thermocouples (your actual one pluss the 2 created at the terminal blocks).

Mhm, so there are 4 thermocouple connections with your methode? May be I did not understand you correctly, but with standard method only 3 thermocouple connections are needed, the actual one plus the 2 created at the terminal block. Validity of this statement can be shown as follows.

If you connect the thermocouple standardly you get the following scheme:





The actual thermocouple is formed by the connection of metal '1' and metal '2'. This thermocouple connection shall be at temperature T1. At connector terminals of according amplifier both metals are connected to a third and identical metal, namely metal '3', which is mostly made of copper. The additional two 'parasitic' thermocouple connections are at temperatures T2 and T3.

In the schematic also shown are the densities n1, n2 and n3. What do they stand for?
If you bring two different metals close together, electrons at the junction tend to equalize the two different Fermi levels of the two different metals. As consequence there is a wandering of electrons from one metal into the other. Wandering stops if potential difference formed by these electrons equals the difference of Fermi levels. 'n1' is the density of electrons in metal '1' after the equalization has taken place. The same is for 'n2' and 'n3'.

The voltage which can be measured at R is the sum of all involved thermocouples. And assuming that in first order approximation Boltzmann statistics is valid, we get:



You see, a fourth thermocouple is not needed.

What, if R, which stands for the amplifier section coming in contact with the thermocouple cables, is not made totally of copper, but additional metal connections must be taken into consideration?
Well, then additional parasitic thermocouples are formed and additional terms are created and have to be added. But: These additional parasitic thermocouples cancel fortunately, IF ALL THESE PARASITIC THERMOCOUPLES ARE HELD ON SAME TEMPERATURE AS T2!!

That's, why a heavy thermal block is used in such designs, providing big heat capacity. Its task is to accomplish thermal equalization of all involved parasitic thermocouples.

In thermocouple amplifier designs it's often overlooked, that heat dissipation of standard operational amplifiers can heavily violate this thermal equalization. And it's no surprise, that AD595, a special integrated circuit for thermocouple amplification, provides a current consumption of only about 130µA! May be your design suffers from this mechanism?

Kai