full bridge buck boost converter blows mosfets

[turboii]

ok trying to build a full bridge buck boost converter, please help me troubleshoot a blown mosfet problem, can you see obvious problem?  i was just testing it with no load, no inductor, 24V connected to both sides as shown in schematic, and it blew up all the low side fets, high side fets are fine.  PWM#2 was inverted from PMW#1, with about 5us dead time between the two. I scoped Vgs from both high and low sides, they peak at 18V down to -5V. i understand that is within acceptable +/-25V range

after it blew up, i removed the lower 4 mosfets, replaced one of them on the input side. tried it again with a slower PWM at 30Hz and seems to have no problems. the gate voltages are fully above threshold after only a couple hundred ns, i dont see how a 500Hz pwm would cause problems

[T3sl4co1l]

Why are there two MOSFETs in series?  And rated 200V for a 24V supply?

And the right hand side stuck at 100% duty and shorted out?  This looks like an H-bridge inverter, not a buck-boost converter.  The top supplies have to be split to do that.

Frequency is irrelevant: transistors die in microseconds.  Switching edges and fault currents are the most common killers.

What current is flowing through the transistors in question?

Tim

[turboii]

sorry i drew it wrong.. i uploaded fixed.  its actually an interter board im repurposing for dc-dc converter.  it has 8 mosfets, so 2 in parallel on each corner, each has their own gate resistor and diode, and share the 10k pull down. i took them out of the schematic for simplicity. i realize the circuit is pointless as drawn, i connected both sides to 24v so i could probe it to see if the fets were working with no load. so zero current should have flowed, if no shoot through

normally there would be an inductor in the middle of the bridge and 24v would not be connected to the right (output) side

[jmelson]

Do the FETs blow with no load?  If so, then the deadtime between transistors on one side turning off and the other turning on is wrong.  You may need to have several microseconds of dead time between turn-off and the other turn-on to avoid this.

if it only fails with a load on the output, then the issue is likely related to inductance of the load continuing to draw/push current when the  transistors switch.

I also note there is no current sensing method on this.
Oh, one other thing.  The high-side driver experiences a high dv/dt when the transistors switch.  Check the datasheet for any limits on how much dv/dt it can handle from the LED side to the driver side.  Ah, another thing:  note that the Miller effect only happens when there is voltage on the drain terminal.  The swinging of the Vds couples through the drain-gate capacitance to fight charging or discharging the gate, this vastly slowing the Vgs.  Make sure you measure the Vgs with power on the +24 V supply, although this makes the measurement more complicated.

Jon

[turboii]

thanks for the pro tips, i will add current sensing, the board already has it, just need to wire it into the uC. i think i must have accidentally shorted something with my test leads to cause it to blow up, because it has been working great after replacing the blown fets, even with inductor connected. and looking at gate signals, there is plenty of dead time with no weird stuff going on

[strawberry]

current protection must not be longer delay than ~1us 

[turboii]

it works! i discovered my code was causing shoot through near zero duty and during transition between buck and boost, probably what blew the fets up. also found out minimum dead time is about 200ns, have it set to 500ns at the moment