In the course of my career, I’ve spent a lot of time psychoanalyzing MOSFETs—trying to uncover the sources of their stress breakdown. Did they have a dysfunctional childhood? Did they get a bad draw in the gene department? Like tiger moms, are we driving them too hard?
In general, we like the MOSFET characteristics with positive temperature coefficients (PTC) where they lead to stabilizing negative feedback. We like PTC in gate thresholds where hot cells have a higher threshold voltage and turn on later than cooler cells. That’s good. The drain-source avalanche break-over voltage has a PTC where a hot part breaks over at a higher applied voltage. That’s good.
One of the keys to the miracle of MOSFETs is the PTC related to on-resistance. Hot ones have a higher RDSON. That’s good because it helps parallel devices share current more evenly. This happens externally when the engineer connects multiple devices in parallel, but it also happens internally. That’s why, year-by-year, you see the technological long march to lower on-resistance. More and more transistor cells are crammed closer and closer together. You think you’re buying one device, but you’re really buying many (millions?) in parallel.
The front page of the datasheet gives you an on-resistance rating, but it’s measured at 25o C. It would be a rare customer design that runs at 25o C, so, generally, the on-resistance will be higher. How much higher? I use a rule-of-thumb factor of 1.6 for a normally operating system.
When is the 1.6 fudge-factor insufficient? When the circuit is not running normally.
In a closed loop system, suppose the load increases or a weakened MOSFET losses increase. As long as the loop is in compliance, the PWM controller will increase the duty cycle and the MOSFET stresses will increase. We don’t think of FETs as having a thermal runaway problem, but they sure do as long as the control loop says so. Like a debt collector, the linear operation of the loop will increase FET stresses with no regard for safety and wellbeing.
What saves the system? Not the linear, closed-loop operation. The savior comes from the nonlinear control functions like over-current or short circuit detection which we hope are well-defined and well-behaved.
Do you have stories to tell about the closed-loop operation of your power circuit? Tell us about your victories and sad tales of woe in the comment section.
Ken Coffman is a Bay Area, Senior, Principal Field Applications Engineer for Intersil Corporation.