For a test bed, we want to measure temperatures with NTC sensors (~10kOhm at RT) for the range 10°C to 70°C. Our test machine accepts auxiliary signal through CAN-bus, so we used a Raspberry with a "CAN-hat" to prototype the temperature measurement (three-point probing) and configure the test machine.

We now want to scale it up to 80 (phase 1) then 320 (phase 2) channels. We already have all the necessary NTC sensors (recovered from previous projects). So far we have 2 solutions which are not satisfying:

  • An all inclusive solution by a provider for a hefty price

  • Acquiring eight 10-channel Ohm meters + eight RS-232 to CAN. That's about $100 to $150 per channel. That would slash the price in 2 compared to solution 1.

In both cases, that's more than the cost of the RPi + the CAN-hat per channel. I would expect that scaling up a prototype would mean reducing the price per unit.

Is there any less resource-intensive solution(s) we are missing? Ideally measurement with 3 or 4-point method would be wished.

Thank you. edit: accuracy around 1K, resolution around 0.1K.

  • \$\begingroup\$ What is your required resolution and accuracy? How much time do you have to take these readings? \$\endgroup\$ Aug 13, 2018 at 16:27
  • \$\begingroup\$ Have you considered Analog Switch Mux? \$\endgroup\$ Aug 13, 2018 at 18:04
  • \$\begingroup\$ Thanks for your questions. Resolution should lie around 0.1K and accuracy around 1K. Acquisition frequency is typically 30s to few minutes. I have not heard of Analog Switch Mux. \$\endgroup\$
    – hyamanieu
    Aug 14, 2018 at 7:29

1 Answer 1


At the quantities you desire, it would most likely be best to design a custom PCB with a micro with CAN bus, and ADC and an OPamp could be had for the 50$-70$ range depending on your requirements. (0.1C 0.5C or 1C accuracy?)

The circuit looks like this (doesn't matter if it's a PTC or NTC): enter image description here Source: Here

An add on board with analog for an RPI would probably be in the same price range. If you did multiple channels on the same board the cost would go down even more. Depending on the accuracy and temperature resolution, maybe less. There would be an additional cost of time and software writing for the custom solution.

Another option that might work is the LTC2986 which has a demo board with an 'arudino' (actually linear's own flavor), if I remember right they have at least 4 channels (or 8) with 0.1C accuracy.

enter image description here Source: Digikey.com

With a one of these solutions the NTC's may also reqire calibration (depending on your accuracy and resolution requirements) with a bath or some other method which might be more complexity

  • \$\begingroup\$ Thanks for the great answer! Calibration is what scares me the most, for 320 channels it'll be very repetitive and long. But our other "all-in" solutions have the same requirements. Accuracy should lie around 1K, resolution 0.1K. \$\endgroup\$
    – hyamanieu
    Aug 14, 2018 at 7:51

Your Answer

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

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