Infra-red (IR) temperature sensor for power MOSFETs

[@adamp779]

Just a random thought from a hobbyist: To calibrate an uncalibrated MOSFET put it in a controlled oven set to the desired Maximum Junction Temperature. Monitor the internal diode forward drop at a constant current. Let it heat-soak till it is heat is saturated, and the diode's forward drop Voltage has stabilized. During testing for maximum current allowed through the system at different ambient and PCB temperatures. Periodically cut the MOSFET off from the circuit and test the internal diode forward drop, at the same constant current. This should give a very accurate indication of the devices internal temperature. I am assuming the topology of the MOSFET has the current junction and the internal diode equivalent in close proximity.

[@sambenyaakov]

Possible in principle but highly not practical

[@Savage-lx5yj]

Great information! Excellent presentation! Thank you sir!

[@sambenyaakov]

Thanks

[@webster0203]

Amazing! I saw on ElectrArc240 channel a video about EVs using the inverter for heating the passenger area. Supposedly they make the inverter less efficient on purpose by lowering gate voltage or whatever.

[@sambenyaakov]

Thanks for the info! It will get hot even with low gate resistance😊

[@MrRohailbootwala]

I am using a similar concept for below my table. I covered 3.5 sides with mylar sheet and there's a buck converter + heating eliment with fan. I have placed the buck directly behind the fan XD. My feet stay warmish. It's 3°C outside. (I plan to use a peltier next and transfer laptop heat down below)

[@sambenyaakov]

@@MrRohailbootwala Great! 3°C outside??wow

[@MrRohailbootwala]

@@sambenyaakov indeed indeed. I am thinking that I can do better by placing water cooled peltiers at the top of window frame. Then transfer the air heat via water pipes

[@MrRohailbootwala]

@@sambenyaakov Just to clarify, I am not using gas for my room. I use a heating jacket for the Torso

[@juliatruchsess1019]

TI makes TMP708/709 thermal switch devices in SOT-23, with 0.5C accuracy. Just put them on the board as close to the FET as possible, much simpler than an IR solution. They have open-collector outputs so can be wire-ORed. There are also many SMD NTC thermistors available.

[@jaro6985]

Its offset from the case temp by a fair bit, you measured and what differential did you see?

[@juliatruchsess1019]

@@jaro6985 My board has a lot of copper on 6 layers, so the temperature in the area around my MOSFETs is pretty uniform, within a degree or two C. Good enough for emergency cutoff.

[@sambenyaakov]

This is OK for protecting the PCB

[@gaozihan4]

This in fact would be much more useful for higher voltage modules and medium voltage transformers, e.g. 10 kV SiC modules and transformers for them. We ever tried to design temp sensors based on NTC or fiber temp sensor but the cost is high and the implementation is very complicated (using HV power supply or special fiber transducer). With this IR sensor, just put them on the gate drive or current sensor boards and share the power supply and communication, and it can be very impressive if the IR light can penetrate well through the silicone gel.

[@sambenyaakov]

Thanks. Yes, that could be an interesting application of this technology. Thanks for your thoughts

[@pieters286]

Very interesting, should one also add the thermal resistance between heatsink and ambient temperature in thermal model at start?

[@sambenyaakov]

Yes, that's also needed for the model. Thanks for comment.

[@Tuong11a]

Thank you very much, very interesting video. If i not wrong in the Tesla EDU they use it for sensing Mosfet's temperature.

[@sambenyaakov]

Yes, I am mentioning this in video. Not sure about the car model.

[@salec7592]

Professor, why don't power device producers integrate an isolated pn junction on the device substrate, as an "near observer" temperature sensor?

[@sambenyaakov]

Already do. Infineon just came out with one.

[@mayankamipara6766]

Professor, How can we find the thermal resistance of Junction to tab in datasheet? In mostly datasheet there is two thermal resistance they mentioned Tjc and Tja.

[@sambenyaakov]

Tjc is for the junction to metal tab

[@mayankamipara6766]

​@@sambenyaakov yes, but for measurement with an IR temp sensor on the plastic tab, theta of tab to the junction required? Right? In my flyback smps i had measured case temperature with thermocouple with hole in heatsink and tab temperature with ir sensor parallely, they are nearly same..... I have used cree mosfet 1700V, 4A with heatsink.

[@and1oncnn]

wow thanks! It seems like you can read my mind. I was thinking of over temperature protection for power MOSFETs and looking for a simple solution just recently. Is it also enough just to monitor over current condition using the MOSFET driver feature and deduce the over temp condition from the overcurrent?

[@sambenyaakov]

Just current monitoring leads sometime to over protection. E.g. not allowing a short bursts of high current which are still safe.

[@sergepetrov8598]

Quite a sophisticated method with multiple stages. Why not integrate a sensor, anyway modern power transistor is de-facto an integrated circuit.

[@sambenyaakov]

Too late 😊 see www.infineon.com/dgdl/Infineon-IPT60T022S7-DataSheet-v02_01-EN.pdf?fileId=8ac78c8c8afe5bd0018b4c5965fd7aeb

[@tamaseduard5145]

👍🙏❤️

[@sambenyaakov]

😊👍🙏

[@biswajit681]

Long time no see 😅

[@sambenyaakov]

Very busy but there is hope

[@ychannel3]

Isolation would be the only advantage of this sensor, but there other ways to get around that.

[@sambenyaakov]

Indeed

[@zazio5535]

This sensor could measure the realtime temperature on top side, without other delay factors. NTC or other contact thermal sensors will always have to sit with some distance to the case. I remember Vishay or Onsemi had a power mosfet with thermal diode grown on the same die, this could be the ultimum "accurate" sensing..