# H-Brdige MOSFETs getting hot despite no load

I'm building a 600v H-Bridge using IPZA60R060 MOSFETs together with isolated gate drivers UCC21520. I designed the PCB layout to be as tight as possible and thought I'd done a good job routing the gate drive signals. I've got a weird problem that I can't work out.

With NO load attached the MOSFETs are pulling a small current from the DC bus (about 20-30mA per half bridge). This only occurs when applying the 30KHz PWM signal from a STM32F407. With low bus voltages it doesn't matter much but I want this H-Bridge to operate at up to 450volts. At 75volts they're consuming about 2Watts and get hot. Weirdly, the current seems to stay 20-30mA regardless of the bus voltage.

I've tried:

1. Increasing the deadtime both via hardware (resistor from DT pin of UCC21520) and via software. Nothing
2. Removing the diode discharge resistor so the gate is charged and discharged via a 4.7R resistor - Nope
3. Increasing this gate resistor to 10R then to 22R - No benefit
4. Reducing the switching frequency from 42KHz to 20KHz did slightly reduce the current draw (30 -> 20mA)

I don't know what else to consider. The gate drive supply is provided by a separate 3S LiPo at 12v. Is it a layout problem or something to do with the separated kelvin gate/source pins? The circuit is no good to go to 450volts at the moment and I'm pulling my hair out!

• What current do you calculate the MOSFETs should take to charge and discharge their internal capacitances, at these switching frequencies? Alternatively, if charging capacitance C from 75V 42000 times a second consumes 30mA, what is C?
– user16324
Oct 5, 2020 at 11:45
• I'm using a separate 12v LiPo battery for the gate drive power supply. The MOSFETs are drawing power straight from the DC Bus. I calculate about 69nC*42KHz = 3mA needed to charge each MOSFETs gate at 42KHz but this shouldn't be making them get hot. Oct 5, 2020 at 11:57
• Good. That's the gate capacitance. I was asking about the output capacitance, though I' don't know if it's really the problem. You may also be driving significant current via the bootstrap C.
– user16324
Oct 5, 2020 at 12:12
• I think you've mixed up the a and b drivers on the gate drives, so this would mean if there's anything connected to the load it would have some conduction even at no pwm difference... Now you said no load connected... Could you isolate each side? Is the cross conduction only happening on side with b driver at the bottom? Oct 5, 2020 at 18:50
• Thanks for your ideas. There's no load connected (just a ~10cm bit of wire). Each half bridge is on it's own. I've tried sending PWM to each half-bridge separately and the problem is still there but just on the MOSFETs that are switching so half the current is drawn. With these gate drivers the A & B drivers are supposedly interchangeable so for layout reasons i did it the way you have noticed. Oct 6, 2020 at 8:32

2 things you need to modify on the board.

1. You have used 1 source pin of MOSFET as VSSA of MOSFET driver and others for connecting to lower MOSFET while you should short these two source pin of one MOSFET together externally.

2. use 22uf and 0.1uf capacitor in parallel as a bootstrap capacitor.

I have found a solution:

The circuit was behaving as if there was a 25nF output capacitance. I don't know where this was coming from. The losses were all occurring in the MOSFETs since no other components were getting warm. Despite what I said the current draw was proportional to DC Bus voltage and the switching frequency (sorry for misinformation).

I analysed one half bridge and grounded the other. I measured the weird output capacitance by 2 methods:

1. With an 82KHz switching frequency, 50mA was drawn at 20v. Therefore: C = I / (V x f) = 27nF
2. With a large deadtime set and a 22ohm resistor between phase A and B (B being set low) I noticed a RC decay during the deadtime window. Based on the decay time I calculated a capacitance of 24nF.

Absolutely no idea where this phantom capacitance was coming from...

• I only have a 5MHz probe but I couldn't identify any nasty ringing going on. I could confirm there was no cross conduction though.

• The bootstrap capacitors were maintaining their charge, the gate drive signals were appropriate, all looked good on the scope.

I swapped out the IPZA60R060P7 MOSFETs which I thought had so much promise. I put in some PSMN027-100PS MOSFETs I had lying about. And it's completely fixed the problem! No DC bus current is drawn even at 82KHz and it works as I would expect.

I now just need to find a 600v MOSFET or IGBT that will work. Thank you for all your suggestions.

• Good work. Mind your units. 'k' for kilo. 'K' for kelvin. 'V' for volt. Capitals matter! Oct 18, 2020 at 11:38