Power MOSFETs and Thermal Runaway - John Dunn, Consultant, Ambertec, P.E., P.C.

Switching power MOSFETs have a thermal runaway mechanism. A MOSFET's on-resistance rises with rising temperature and if the drain to source current doesn't materially change in response to that, a vicious cycle can occur. A rising Rdson leads to rising temperture which leads to rising Rdson which leads to rising temperature which leads to .......

You can test for this without having to burn anything by measuring MOSFET case temperature. Stasrt by measuring temperature at regular intervals as the devices get hot. Usually intervals of ten seconds are a good choice.

Once the MOSFETs have reached some high temperature, you turn off operating power AND you KEEP ON measuring tempertures as the MOSFETs cool down. Keep using the same time interval and keep on making the measurements until the MOSFET temperature has pretty much returned to the starting value.

Next, you examine the thermal time constants of the curve of the temperature rise and the curve of the temperature fall. The two time constants should be nearly equal to each other, but if the time constant of the temperature rise seems to be much longer than the time constant of the temperture fall, you DO have a thermal runaway condition just waiting to do lots of damage when the MOSFETs eventually burn out.

The reason why the apparent time constant of the temperature rise can look larger that way is that the end-point of that thermal rise curve is itself rising. This stretches out the shape of the rise curve and makes the rising time constant look longer than it really is. This is illusory AND this is bad news if you see it.

Here's a real-life case (no pun intended) where this really happened:

Yes, there was a lot of smoke when this power supply broke down, but changing to lower Rdson devices and improving the heat sinking eventually made everything okay.

Comments

[Omid Mostaghimi]

Hi
Thanks very much for the post
could you please put some references?
Many thanks
Omid

[george storm]

Well warned! A colleague had just this problem in the early days of high-Voltage power MOSFETs - and those devices blew their tops off. His simplified theory said it should not happen, so he told his assigned apprentice to fit new ones as the devices must have been faulty - the rest is social history

[Senthil]

Good One...Thanks