Switching supply burning mosfets, what is happening?

Question

I have been working on a SEPIC converter for an electronic cigarette power supply. This application involves using two 3.7v INR batteries in parallel, rated for an appropriately high current discharge, to power the converter and drive a kanthal heating element with a DC resistance ranging from about 0.5 ohms to 2 ohms.

I am driving the SEPIC MOSFET with the PWM output of an atmega328p at ~5.0v, through a 100 ohm limiting resistor and with a 10K pulldown to ground. The SEPIC portion of the circuit is as follows:

The only difference being that my MOSFET is an STL260N3LLH6 and my diode is MBR2045EMFST1G - my LTSpice library doesn't have entries for the parts I am actually using. The inductors are not coupled, and are rated for 48A current saturation. All capacitors are ceramics.

At low duty cycles I am getting the expected behavior, although sometimes depending on inductor value and coupling capacitor value the output voltage is about 10% higher than it should be for a particular duty cycle.

Once I get to a duty cycle of about 55-60%, the voltage output no longer continues to increase and at about 65% the MOSFET gets extremely hot and will burn up if operated for too long.

Initially I thought I was not driving the MOSFET gate hard enough, as I tried to operate my MCU at 3.3v, but operating the MCU and MOSFET at 5.0v has not fixed the problem. Inductor values from 1.5uH to 6.5uH have been swapped in with minimal changes to behavior as well. I have tried coupling capacitors from 1uF all the way to 90uF, still no luck. Switching frequencies from 400Hz to 128kHz have also been tried.

Can anyone point me in the right direction as to why my MOSFET is burning up? For curiosity's sake, I built up a buck-boost converter using the same parts and got a similar result.

Answer 1

How did you calculate those inductance and/or timing values? Also, 100 ohms will make that gate switch at a glacial rate. 62nC (datasheet) is a lot of gate charge.

Anyway, if we assume you've got 50% duty cycle and allow current to increase (let's assume the best case, starting from zero) linearly from 0 for 1msec the current at the end of the time (through the 4.7uH inductor) will be 787A.

You need to drive that gate hard with a gate driver to at least 5V and certainly not with the ~150-200 ohms (counting the AVR output resistance) you are driving it with, and to increase the switching frequency and/or the inductor values.

You are getting the MOSFET spending all its time in the linear region at high frequencies and the inductors saturating at low frequencies and the MOSFET not really fully turned on ever (excepting that 5V test).

If you can beg or borrow (I wouldn't recommend steal) an oscilloscope and observe the voltage and current waveforms you will be able to make rapid progress. If not, look at the simulations including current waveforms and put actual values like that 100+ ohms in series with the drive.

Answer 2

I think you need to first define your design on a basic level. How much voltage do you want at the output? Continuous operation? How much current? Figure out your basic requirements and start sizing your components from there.

Driving a power MOSFET from a MCU pin is a bad idea unless you're driving small signal FETs (logic level) The MCU pin is only able to supply a few mA. This means that the FET will switch rather slowly. Every time the MOSFET turns ON or OFF, it dissipates heat. Higher the switching frequency, higher is the loss. So the gate of the FET has to be driven hard with a gate voltage of 10V or more (max, especially if it's expect to drive high currents. Also, the MOSFET has to withstand VIN + VOUT. I would add a safety factor of 30% for it to be robust.

On the other hand, a high switching frequency allows you to reduce the size of your inductors and capacitors (to an extent) It's all about trade-offs...

Consider using LM5022 as a controller. I successfully built a SEPIC using that and coupled inductors from Coilcraft. You could get away with a narrower VIN range. The tricky part is starting the chip at low voltages, but maybe you can do a simple charge pump (https://www.powerelectronictips.com/faq-what-is-a-charge-pump-and-why-is-it-useful-part-1-faq/) to start the chip and then supply it from the output (if the output voltage is high enough). Remember to use diodes to decouple the two paths!

If all you need is a short high current pulse, maybe you should rethink your design.