5
\$\begingroup\$

I need to connect 5 thermocouples on a PCB. My question is about the connectors, can I use any type of connector? My doubts arise from the fact that I could create other junctions, the chip I'm using (MAX31856MUD+) should compensate for the cold junction, but I still have doubts.

On the evaluation board, to connect the thermocouples, there are terminal blocks, like these:

enter image description here

my intention is to use this connector MOLEX 436451000 (10 pins inline) to connect the 5 thermocouples. Could I have problems with these connector?

https://www.molex.com/molex/products/part-detail/crimp_housings/0436451000

\$\endgroup\$
3
  • 1
    \$\begingroup\$ What range of temperatures are your thermocouples sensing? What accuracy (absolute trueness) do you seek to achieve? What thermocouple type (T, K, J, etc) are you planning to use? If you can describe your application, that would be good too. \$\endgroup\$
    – Nick Alexeev
    Mar 23 at 20:01
  • \$\begingroup\$ I need to measure range from -40 to +120 deg, thermocouple is K type. Accuracy +- 1 deg is acceptable \$\endgroup\$
    – Federico Massimi
    Mar 24 at 8:48
  • \$\begingroup\$ Is this industrial or a hobby project? Are you building one, ten, one thousand, one million of these? What are the tradeoffs -- is it worth another $1, $10, $100, per unit to go from 2 deg C to 1 deg C accuracy? d \$\endgroup\$
    – Technophile
    Apr 5 at 0:43

3 Answers 3

Reset to default
8
\$\begingroup\$

Using a terminal block like that means that the actual cold junction temperature is the temperature of the terminal block/thermocouple wire junctions.

To the extent your cold junction temperature sensor does not measure that temperature accurately you will have additional errors. To the extent the different varies you will have noise/drift in the measurement.

So your goal, being stuck with an inconvenient semiconductor package, is to get the chip as close to the terminal block as practical and to make sure that all the items are as isothermal as necessary to meet your target specifications for accuracy and drift.

It's not unusual to get differences of a few °C here and there, and they may change significantly as your circuit warms up and heat is conducted in and out of the enclosure.

Your Molex connector will have similar problems. The female spring part will have less thermal coupling to the pin and PCB than a terminal block, so it will likely be more prone to errors due to heat flowing down the T/C wires. On top of that, it is a crimped type. With a very good crimping tool you may be able to get reliable gas-tight joints but it's certainly a possible source of issues. With a bad crimp problems over time are almost guaranteed. For very good metallurgical reasons, popular thermocouple materials are somewhat difficult to reliably connect to without something like welding.

\$\endgroup\$
2
  • \$\begingroup\$ In my PCB the connector is very close to the chip, less than 1 cm. And underneath the chip and the connector i have two solid copper ground planes, and in case I can even add a third copper ground plane on the bottom layer to have an even more uniform temperature situation. What absolute error in temperature reading are we talking about? \$\endgroup\$
    – Federico Massimi
    Mar 24 at 8:53
  • \$\begingroup\$ @FedericoMassimi you tell us. What's in your circuit? Is it a micropower sensor or controlling a large heater? How much heat is dissipated in the enclosure -- closer to 1 mW, 1 W, or 1 kW? \$\endgroup\$
    – Technophile
    Mar 31 at 17:44
5
\$\begingroup\$

By adding dissimilar materials between your thermocouple and your measurement device you create errors in your measurement, as indicated in the MAX31856MUD+ datasheet:

enter image description here

In theory, if the temperature of the IC pin is the same temperature as the point at which the wires plug into the connectors, all the errors cancel out. In practice, there is a temperature differential and thus an error. This is true for any connector you use, be it your Molex connector or a PCB mount thermocouple connector such as the Evo Sensors k1x-femx-con-fp-x-pccx:

enter image description here

The work around is to bring the thermocouple wires as close to the electronics as possible and keep the temperature the same from the thermocouple / board interface up to the IC. For example, from the Analog Devices AD8494 data sheet:

enter image description here

You will get a smaller error if you put thru holes in your PCB about 0.5 mm away from the IC pins and solder the thermocouple wires directly to the PCB. The solder captures the wires, but doesn't really wet them. Then pour an alumina filled glop top glob top thermocouple wire / board interface up to the IC pins. Alumina is not electrically conducting, but an OK thermal conductor. This helps keep the temperature uniform between the two.

\$\endgroup\$
5
  • \$\begingroup\$ Repeat my previous comment: In my PCB the connector is very close to the chip, less than 1 cm. And underneath the chip and the connector i have two solid copper ground planes, and in case I can even add a third copper ground plane on the bottom layer to have an even more uniform temperature situation. What absolute error in temperature reading are we talking about? \$\endgroup\$
    – Federico Massimi
    Mar 24 at 8:54
  • \$\begingroup\$ Should "glop top" be "glob top"? \$\endgroup\$
    – psmears
    Mar 24 at 14:59
  • 1
    \$\begingroup\$ @FedericoMassimi About six months ago we used a MAX31856MUD+ with an Evo Sensors k1x-femx-con-fp-x-pccx connector about 1.7 cm apart. The PCB was two layer with a pour below and around the IC and connector. There was no alumina (we ran out) and we measured 29 °C when the room was 25 °C. The connector is made for K-type thermocouples, but the pins to the PCB are copper, which we suspect was much of the source of our error. \$\endgroup\$
    – C. Dunn
    Mar 24 at 16:02
  • \$\begingroup\$ But is this error fixed? That is, can I measure this error with a reference thermocouple and then compensate for it in software? \$\endgroup\$
    – Federico Massimi
    Mar 26 at 11:35
  • 1
    \$\begingroup\$ @FedericoMassimi This IC was the temperature measurement for a heater controller we made. We did some testing and and found subtracting 2.5 °C off in software minimized the error over the temperature cycle. This was an acceptable work around for us since our testing was done with the electronics at room temperature, and the client would be using the electronics at the same temperature. I'd be worried about using such a scheme if the client was using the electronics at a different temperature than what we tested, say -10 °C. \$\endgroup\$
    – C. Dunn
    Mar 28 at 18:03
0
\$\begingroup\$

From your comment "accuracy +- 1 deg is acceptable" you have some design choices, including:

  1. Use the connector you like casually and accept reduced accuracy, or
  2. Find sources for connectors made from matching thermocouple materials (I have seen such DB style connectors), or
  3. Take whatever measures are required to thermally isolate and insulate the connection so that it remains within less than the stated +- 1 degree error budget
\$\endgroup\$
2
  • \$\begingroup\$ For the connector that I should use there are no specific contacts for thermocouples, there are contacts of various materials, but I don't know which one could be more suitable, for example these: molex.com/molex/search/… One thing came to my mind, in my system I have 5 thermocouples, but only 4 are really used, the fifth is for service. Could I use the fifth thermocouple to measure the connector temperature (connector joint temperature) and use that measurement to correct the others? \$\endgroup\$
    – Federico Massimi
    Apr 2 at 12:52
  • \$\begingroup\$ @FedericoMassimi that sounds like a distributed "cold junction". Thermocouples give a signal proportional to the difference in temperature between the ends (and it's not 100% linear). To get an absolute temperature, you need an absolute temperature for either the cold or the hot junction (the latter is usually for cooling applications). \$\endgroup\$
    – Technophile
    Apr 5 at 0:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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