Measurement of temperature on/besides Mosfet

[eTobey]

Hi,

i am wondering where i can get the most accurate measurement for the temperature of the mosfet. Besides (B) or on the case (A)?

There are 4 layers of  2 oz copper and the vias go to the back side to a heatsink. There is much copper around, but not directly connected.

[PwrElectronics]

If the fet has an exposed case tab, what I have done in the past is solder a type "T" thermocouple to that tab.  T thermocouple has 1 copper wire so it will take solder.  Only do this if its not at a hazardous voltage.  I have done this with 50VDC things OK but some thermocouple meters do not like if if the fet is high frequency PWM.

If you go to the PCB, there is a temperature gradient through the solder and also the lateral heat flow.  Temperature there will be cooler than at the case tab.  But, the error is probably not a lot, depending...

[Coordonnée_chromatique]

Hello, this is perhaps stupid, but could you tell why it is impossible to use the internal diode of a mosfet to track the die temperature ?

[tom66]

You can't use the internal diode normally, as the diode doesn't conduct unless the MOSFET is reverse biased. Whilst it would be possible to design a circuit that allowed the diode to be biased in reverse to test the diode forward voltage and measure temperature it would involve numerous additional parts to do so, hardly worth it.

However, you can use the RDS(on) change with temperature, provided you know the current you can back-calculate for temperature against a lookup table.  This requires a current sensor or a load with a known constant load characteristic.

[eTobey]

Since i put some redundant current sensors in, this would be an interesting thing to achieve. But i would have to calibrate this.

I have looked at some thermal images. It would be then more accurate, to measure it on the case. But only if it does not have leads that are to big.

[PwrElectronics]

At work, we usually give a thin coat of flat white paint (I have used black but our tech lab said white is better) over the board to improve the emissivity when using a thermal camera.  Especially if there are reflective metal surfaces.

[thm_w]

At work, we usually give a thin coat of flat white paint (I have used black but our tech lab said white is better) over the board to improve the emissivity when using a thermal camera.  Especially if there are reflective metal surfaces.

Black paint has better emissivity (0.94 vs 0.84) but maybe they are worried about improving the thermals of the test board, or have a camera thats already calibrated to 0.84.
https://www.thermoworks.com/emissivity-table/

If OP had a thermal camera I suspected they'd already be using it.