Measuring FET case temperature accurately

method measure fet temp

Does anyone know of a good thermal probe for measuring FET temperature in a live circuit? I need to monitor FET case temperature in my full-bridge SMPS. I'm using a Fluke 80TK- the kind that connects to a DMM, but it's giving me erroneous readings when the circuit is running. It needs to physically make contact with the tab on the FET, but since the tab is not at ground potential and is switching at 100kHz, the reading is not accurate. I need some type of proble that is either non-contact, or is not effected by switching noise when in contact. I think thermal imaging would be a terrific solution, but WAY too costly for the application.

Thanks for any help or suggestions.

If you're sure that the contact between the bead of the probe and the tab of the FET causes error, you may consider to use Thermally Conductive Insulator Pad ( http://www.chomerics.com/products/therm_chotherm.htm ) between the tab and the bead. Do you think possible?

Few years ago, I have the similar problem when measuring the winding hot-spot temperature of a running motor. The bead of the thermocouple was embedded into the winding, but it did not touch 'electrically' with the motor winding. We measured the temperature with TWO thermocouple meters, first one was cheaper, and the 2nd one was Fluke 51. The reading measured using the first meter was not stable or fluctuated. However, the reading of Fluke 51 was stable. So, I think the input CMRR of the first meter was not high enough to filter out the common pickup noise on both thermocouple wires.

This calls for what is commonly called an IR thermometer. This detects the IR radiation from the heat of the device. On source of error is the emissivity constant being not 1.000 for aluminum. Just put some black paint on the tab and the readings will be correct. This method requires hand holding the thermometer.

flatulent said:

This calls for what is commonly called an IR thermometer. This detects the IR radiation from the heat of the device. On source of error is the emissivity constant being not 1.000 for aluminum. Just put some black paint on the tab and the readings will be correct. This method requires hand holding the thermometer.

Click to expand...

Would you pls advise what is the best accuracy that could be achieved using IR thermometer? Thanks.

I would suspect that a quality version would be within 1 C of the proper temperature. Another alternative is if you can shut your system down for a second or two every once in a while to allow your thermocouple measurement method to work without interference.

Quite a lot of IR thermometers have renge of -50 to +>500degC with resolution of 0.1degC...

IanP said:

Quite a lot of IR thermometers have renge of -50 to +>500degC with resolution of 0.1degC...

Click to expand...

I have been involved indirectly in the design of one of these. The resolution is no indication of the accuracy. the 0.1 C digit is just to impress the customers. The uncertainty of the emission coefficient of the surface measured and inherent errors in the instrument make 1 C a more achievable accuracy.

You can calibrate your IR instrument by putting DC through a transistor on a heat sink and measure it with the IR and thermocouple method. Run the heat at the level you want to measure in a working circuit. This should make your 0.1 C IR reading the accuracy of the measurement.

Thanks for all the terrific suggestions!

I seem to be having success using a small flake of mica insulator between the tab and the thermocouple. This and some thermal grease between everything is giving fairly accurate results- within a few degrees I suspect.

I do have access to an IR thermometer, but it does not pin point a measurement area very well. I think it will read a 2cm spot at a distance of 15cm or so. Do you think it would "focus" on an even smaller area if I brought it even closer? I'm not sure how I'd be able to tell. I'd like to measure a spot of about 5mm.

Yes, place the IR unit 1 cm from the transistor and compare the reading with the thermocouple reading.