Dear Dr. Engleberry:
This might seem like an odd request, but I am trying to duplicate a field failure and I need to destroy a MOSFET in the harshest, gnarliest and most-nastiest possible way. What can I do to let the smoke out of a power MOSFET?
—Charlotte in Charleston
Whenever anyone mentions smoke I’m reminded of the humidor of Arturo Fuente Opus X cigars waiting for me at my luxury condominium. Ah, I can’t wait to light the gas fireplace and apply flame to one of those premium sticks. Sorry, I was daydreaming for a moment. What were we talking about? Oh, destroying MOSFETs. It’s funny that it would be appropriate for an eminent man like me, with a lab-certified IQ of 181, to consider such matters, but I do. Often.
Semiconductor reliability is a miracle of modern technology. It’s interesting to reflect on how our customers drive costs lower and lower every year and yet—there is no effect on manufacturing quality; regardless of whether you buy a transistor for a penny or a dollar, you expect the same good reliability.
The fact is, for devices that are treated properly, you can expect many tens of thousands of hours of reliable service. However, in cases of electrical overstress, all kinds of bad things can happen, from the device simply ceasing to function, to exothermic effects like charring and burning.
A typical failure mode for a MOSFET is to short out. In this case, the internal resistance can be very small and nothing limits the peak current except the source impedance of the power source. If the current is limited, then this situation can persist indefinitely, though a more common outcome is the die and metal melts or vaporizes and opens the circuit. There is variability in exactly what happens—so if you destroy a group of parts, you’ll see the various responses, from benign to catastrophic.
Using the MOSFET as a protection device and expecting it to fail in a safe condition (open circuit) is a very bad idea. You can use the MOSFET as a protection device, but you do it by turning the device OFF when a fault condition is sensed.
To address your question, there are many ways to destroy a MOSFET. I will list a few that immediately come to mind. As fair warning, you should wear protective eyewear, make sure a fire extinguisher is standing by and use appropriate caution when doing these kinds of tests:
- 1. Apply a DC source to the gate to source to make sure it’s fully enhanced (fully turned on), then slowly increase the current limit of a power supply connected from drain to source. For each MOSFET, you’ll reach a DC current guaranteed to destroy the MOSFET.
2. Similar to Method 1, fully enhance the MOSFET, then rapidly apply the power supply connected from drain to source with the current limit set as high as it can go.
3. Connect a reasonably rated power supply from drain to source, then slowly increase the power supply connected from gate to source. You’ll fully enhance the FET and it will conduct, but that’s not what we’re trying to do. Keep slowly increasing the VGS voltage until the gate oxide breaks down. For a gate rated at 12V, you’ll see FET destruction at about 15V or so. For a 20V gate rating, it will fail somewhere around 25V.
As I mentioned, if you abuse a hundred MOSFETs until they fail, you’ll see various responses. The failure modes can be difficult to predict, but I assure you—you’ll see some you really don’t want.
Now, I would like to excuse myself. There is a leather recliner and an aromatic humidor waiting for me at home.