Measuring Conductive Stuff With Thermocouples

So you’ve got a thermocouple in your hand and you’re wondering, “what happens when I try to measure something electrically conductive?” You probably tried it, and it probably worked fine, so you left it at that. But let’s dig a little more.

The first time I thought about this was while doing an ice-bath calibration: mix ice and water in a slush, guaranteeing that the temperature of said slush is 0.0°C, and stick the thermocouple right in. But wait, electronics and water? BUT WAIT again – often we’d like to measure a heat sink or some metal part to make sure it’s not getting too hot. Will the measurement still be good?

I was testing out a new 4 channel thermocouple logger today and re-encountered the question. The experiment was to answer the questions 1) to what precision do the 4 channels agree when held at an identical temperature and 2) can I quickly put together a heat conduction experiment to validate that the data logging works and I know how to use it properly.

Here’s the setup, caught at a very opportune moment of perfect harmony:

Image showing test setup, with 4 thermocouples sandwiched in the same piece of copper foil tape, and all 4 channels reading 21.3 celsius on the logger screen.
Beautiful!

Each of the four K-type TCs is sandwiched in a row between two pieces of copper foil tape, so that we can touch one side and watch the delayed temperature rise depending on distance from the touch.

As to experiment (1), perfect – all 4 channels agree perfectly as long as the copper is left long enough to equilibrate, and they vary generally ±0.2c with air currents in the room, despite the copper.

But how about experiment 2? My first thought here is “what about the conductivity of the copper in contact with the thermocouples, and then later my own body?” To make sure this wasn’t relevant, I put a layer of thermally conductive electrically insulating double-stick tape (like you’d use for light-duty heat sinks) between the thermocouples and copper foil. I then captured this lovely trace, which shows exactly what we expect: the channels nearer to the touch point rise first, then they quickly equilibrate and decay after the finger is removed:

Image of 4 channel graphs overlaid showing what I just described.

But what if I omit the insulating tape? Things get a little crazier.

Graph (in celsius this time) showing channels 1 and 3 rising with the touch, but 2 and 4 immediately falling.

This graph shows a few interesting things. First, in the no-touch steady-state, all channels agree. This is due to the law of intermediate metals, which states that a thermoelectric junction is defined by the two outermost metals, and the intermediate junctions don’t matter as long as all junctions are at the same temperature. For instance, our K-type thermocouple is made of chromel and alumel leads. A chromel-aluminum-alumel or chromel-copper-alumel thermocouple would behave just the same, as long as the CA+AA or CC+CA junctions were both measuring the same physical temperature at the same time. In other words, it doesn’t matter any which way the copper is shorting out the thermocouple, as long as all of the microscopic contact points where it does so to form a junction are all at the same temperature.

The law of intermediate metals is actually where we get the concept of a “cold junction” – the chromel-copper and alumel-copper joints where the thermocouple leads come into our meter’s copper traces form two additional thermocouples that must be accounted for, which is really one additional K-type thermocouple as long as both of them are at the same temperature. We can compensate for this second “cold junction” as long as we accurately know its temperature (which, to be honest, multimeters with those thermocouple adapters are really bad at.) But I digress.

The final piece of the puzzle is that the thermocouple logger I have reads each TC sequentially in a separate slice of time so as to avoid needing 4 ACTUAL channels of analog to digital conversion. So it doesn’t matter how one TC is shorted to the adjacent ones, since the adjacent ones are electrically isolated when the first one is being read.

The second interesting thing about the graph above is that channel 1 and channel 3 behave just as they did before, where channels 2 and 4 DROP in temperature as soon as the copper is touched. More interesting still, when I let go of the copper, the channels return to bang-on accurate temperature read out, as shown by the fact that the order of temperatures while DECAYING (finger off) matches the pervious experiment, while the order disagrees when rising (finger on).

My theory here is that two of the thermocouples have a dominant short to copper on the chromel lead, while the other two have a dominant short on the alumel side. My body is either sinking or sourcing some tiny current into the joint through the copper, which in one case slews the measurement high and in the other case slews it low. It’s also possible that my body only electrically matters on the side connected to excitation voltage vs the side connected to meter ground, rather than contributing equal and opposite offset.

Of course, it drops by little enough that the reading still looks accurate by eye if you didn’t have the other channels to compare. I imagine this same sort of effect is commonplace in single-channel single-sample scenarios, but goes unnoticed because the “electrical noise” component of the measurement is small in comparison to the actual thermal component.

Anyway, a cool little one-hour experiment, and now I know to electrically isolate my probes while measuring active surfaces, just in case!

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *