CJC accuracy issue with a removable connector

This is our termination box for a few 16-channel thermocouple input or simulator units. There's a 4-wire platinum RTD on the bottom of the board as the cjs. The metal barrier forces the t/c leads to be a little longer inside the box. Seems to work pretty well. Fortunately, thermocouple wire is usually a terrible heat conductor.

Our electronics boxes also have a selectable RTD right at their d-subs, but that's not nearly as good as the remote box.

It's surprising how accurate standard t/c's are reasonably close to room temp, with good electronics.

snipped-for-privacy@highlandsniptechnology.com wrote

I can see this works, because you have nothing in the box :) Not even a microwatt is generated, so the whole box is at the same temp within a small fraction of a degC.

Right. The selectable on-board RTDs, on the instrument end, are subject to all sorts of nasty thermal gradients.

There's a tiny self-heating on the platinum RTDs, but they are scanned so the heating is low duty cycle.

The bright yellow ones use pins made of the same metals as the thermocouples.

Cheers

Phil Hobbs

That just moves the dissimilar-metal junction a couple tenths of an inch!

Well, it's probably easier to fix inside the box, e.g. with a routed-out PCB paddle with a bit of metal clamped top and bottom. The thermal conductance of the two wires on the probe side won't be a big issue, I don't think. (A bit of a trap for non-native speakers there.) ;)

Cheers

Phil Hobbs

No, because both the male and female connectors are the same as the thermocouple wires; the dissimilar-metal junction happens at the end of thermocouple extension wires from the back of the female connector, i.e. after a serpentine wire path goes to a thermal isolation region inside the measurement box.

The plug/socket pairs that use thermocouple materials are bulkhead or inline types, not printed wiring board mount. You can't solder thermocouple wire like tinned copper....

That sounds like a pain. I'd probably go straight to copper using the method I posted upthread. If you stay on one level, the PCB copper probably tracks better than the extension wire anyway.

Cheers

Phil Hobbs

It might be a pain, but a bulkhead connector is painless when you want to curl the thermocouple(s) up and stuff it in storage... and a junction AT a bulkhead isn't good for accuracy because there's hot inside-the-box air on one side of the bulkhead, and ambient air on the other, and where does the thermosensor go, exactly? If it's on the PCB, you want the junction there, too.

Of course. We're talking about the yellow-peril connectors, where the pins and sockets are made from the TC metals. For a lab thermometer, my preference would be for a nice linearized thermistor soldered to a copper pour, in intimate contact with the aforementioned metal bits clamped around the aforementioned routed-out PCB paddle.

But to design a defence, one has to decide on one's threat model. Just look at the stunning success of the Maginot line. ;)

Cheers

Phil Hobbs

Phil Hobbs wrote

This is easy enough if you have a terminal block which is mounted on the PCB.

You can mount the A-D converter (some have internal temp sensors, which is handy) right next to it.

But with a removable connector?

I am seeing a 2-3C delta between the CJ and the sensor, and the sensor is on the PCB and right next to the fixed part of the connector.