The mysterious case of temperature readings

By Joseph San Filippo

We built a large thermal vacuum chamber at our facility, capable of high vacuum and elevated temperature. Heating was achieved by means of numerous flexible resistive heaters epoxied directly to the exterior surface of the steel chamber and uniformly distributed.  A vacuum-rated thermocouple feedthrough connector was installed to allow measurement of the temperatures of items inside the chamber. Since the measurement thermocouples were to be type  T, copper + Constantan, we of course installed a feedthrough connector with copper and Constantan pins to match the thermocouple materials. We connected around 10 thermocouple pairs to the feedthrough connector inside the chamber.

We checked out each thermocouple circuit by connecting thermocouple wires from the feed through connector outside the chamber to a thermocouple readout; inside the chamber we connected a thermocouple simulator in place of the measurement junction. With the chamber open and at ambient temperature, we entered the chamber and dialed a temperature into the simulator. Outside, the readout exactly agreed with the simulator setting on every circuit: a successful checkout.

We then replaced the simulator with a thermocouple probe on each circuit and planted the probes on a dummy test article inside the chamber, closed the chamber and activated the heaters. This was done at ambient pressure; no vacuum. For the first few minutes all temperature readings from inside the chamber were OK, indicating around ambient temperature. Since there was no active air circulation inside the chamber, we expected to see a very slow rise on all channels due to convection and radiative heating from the walls. What we observed instead was a fairly rapid rise shortly after activating the heaters, with the indicated temperatures having no predictable relation to the ambient temperature or to the temperature of the chamber walls.

We shut down, opened the chamber, allowed everything to cool to ambient, then repeated the entire process with exactly the same results. All channels read perfectly with the simulator at ambient temperature and erroneously with heat applied. We then remembered the fundamentals of thermocouple operation: Errors in a thermocouple circuit happen where there is a temperature gradient along a length of conductor that also has a compositional inhomogeneity.

The only obvious gradient in this case was where the circuit exited the chamber, but there should be no inhomogeneity because we had used a matching thermocouple connector…or had we?   Upon close examination we found that we had displaced every connection in the type T feedthrough connector by exactly one position; every copper wire was connected to a Constantan connector contact and every  Constantan wire was connected to a copper contact - major inhomogeneities. When the entire connector and the wires were isothermal (all at the same temperature; no gradients) the readings were correct.  When we created a major temperature gradient by heating the chamber walls we introduced errors. We fixed the wiring errors and all channels read correctly with heat applied.