6

I am trying to measure the temperature on a heatsink that is at Earth ground potential, while my circuit ground is at line potential.

Typically I would use something like the Microchip MCP9700 in a TO-92 package, but there is seemingly no way this could stand up to a 2500V hi-pot test.

I found one UL rated NTC thermistor, but it is very expensive ($10) and it is only rated to 600V. No dielectric breakdown data was listed in the datasheet.

What am I missing? Surely this is a very common thing to do.

Note: My MOSFETs are in a TO-220FP package, so they are bolted directly on to the heatsink.

Should I just ditch the FP MOSFETs, go to a TO-220 and just find a TO-220 NTC and use insulation pads for everything? Or is there a better approach?

This is not looking for product recommendations, rather examples of commonly implemented solutions to this problem.

Daniel
  • 6,168
  • 21
  • 33
  • Maybe just add insulation pad for sensor only. It doesn't affect temperature measurement too much, accuracy of most temperature sensors will be ten times worse than error caused by thermal resistance of pad. – Jakub Rakus Dec 23 '15 at 23:27
  • @lustful-rat True. That may be what I have to do. After continuing my search, I found this item: http://www.vishay.com/docs/29114/ntcalug3.pdf but it is still only 1kV breakdown. – Daniel Dec 24 '15 at 00:20
  • Actually this would be perfect, if you could actually buy them http://www.vishay.com/docs/29164/ntcalug01t.pdf – Daniel Dec 24 '15 at 03:57
  • I'm looking for sensors with even higher dielectric strength (4400Vac). Did you find any that could reach this value in your search? – Akhil Dec 04 '19 at 14:13
  • @Akhil I did not – Daniel Dec 05 '19 at 19:27

2 Answers2

2

Ideas:

A cheap temperature sensor and cheap voltage to frequency convertor driving a cheap opto isolator sounds like an option. The circuit at live potential has to reconvert frequency to voltage to obtain temperature.

There are temperature sensing devices that have a digital circuit ambedded that produce T to F. These can also be used to drive an opto. Look at Maxim's portfolio.

Make a pyrometer using an infra red photodiode pointed at the heatsink - use its output to determine heatsink temperature. You might need to look into material emissivity and black body radiation a tad but it's not impossible and could be very cheap if you have more than a soupçon of processing power up at live potential.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • I should have been more clear.. there's no requirement for the sensor to be electrically at a different potential. The issue is that it must be in contact with a surface of a very different potential. Cost is a very real consideration here. I like the pyro idea. Looked into it as well :) Just very expensive. – Daniel Dec 24 '15 at 00:17
  • 1
    Additionally, there's no power supply at the other potential to drive an opto, so that would require a lot of parts – Daniel Dec 24 '15 at 00:18
1

I deal with the same problem. I've used thermistors from US Sensor. http://www.ussensor.com/ But you're right, they're not cheap. We pay roughly $19 for USP14235, rated for 3 kV.

Stephen Collings
  • 17,373
  • 17
  • 92
  • 180