Hi all

I have an LLC resonant converter with a half bridge and two STD25NF10L MOSFETs, and a LM5109 gate driver:


(ignore the "Digital controller")



It runs, but I am having problems when increasing the input voltage V_in. I have "maxed" out the power with 60V @ resonance (6kHz), and it has not exploded yet, but that was only a short period. When testing a lower voltage (50V), but letting it run for a while, the MOSFETs started smoking. I replaced them, and it still "works", but I looked at V_gs on both MOSFETs before increasing the voltage:



This is with V_in = 15V and 450 ns dead time. Yellow is high-side MOSFET, and green is low-side.

Here's the part I don't understand:

After V_gs has dropped, it spikes up to roughly 6V before going down again, oscillating. The peak of this spike is dependent on the input voltage V_in. Increasing the dead time only increases the "oscillating" period after the spike. This could explain why the MOSFETs started smoking.

What's going on, and what to do about it? Change MOSFETs? Change gate driver? I have thought about this for some time today and can't seem to grasp it.

Thanks!

 

Noise looks like it is coming from you scope ground leads.

The top MOSFET does not know where ground is. It only knows Gate-Source voltage. Please show G-S voltage if you scope can do that or use two traces to show that. (Do not connect scope ground to top Source.)

The Turn on signal is slow. 120//120 ohms. Maybe too slow. If you are doing zero current turn on then slow is OK.
The Turn off signal is fast. Maybe too fast. I can't see.

The yellow 6V spike is at the same time as the green spike. This is probably when the output voltage hits 0V.

 

The problem is fairly obvious, which is that gate drive must be relative to the source. Otherwise you are using a "blooper" circuit which is not going to be at all reliable. A slow turn on and slow turn off leave lots of time in the linear range where lots of power is turned into lots of heat,AS THE TS commented was happening.

 

@ronsimpson
@MisterBill2
I have isolated channels, so I am not measuring G-GND on the top mosfet, but G-S (GATE_HI - HS). On the lower one, G-S is G-GND. The oscillations are stable, does this not mean that it is less probable that it is probe noise?

I soldered off R44 and R43 (to disconnect the diode) to slow down the discharge:


Yellow: V_gs top mosfet
Green: V_gs bottom mosfet

This is with 60V input, the voltage peak after the gate discharge now decreases slightly with higher input voltage. Mosfets are still running hot. It is noteworthy that I do not have heatsinks or fans to cool the mosfets. They are DPAK soldered to quite large copper shapes, but not complete copper layers with vias to conduct the heat to the bottom etc.

Removing the diode made the turn-on and turn-off quite similar, should I basically tune the gate resistances? Right now it's 120 x 3 (parallel) => 39 ohms. What should I aim for? Should the diode be reintroduced, but with a different resistor to limit the discharge rate?

 

What is the current in the MOSFETs? It is probably hard to measure, so measure the current in the transformer.
It is nice to know the voltage and current in one of the MOSFETs.
Do you have a current probe?

Nice scope. Isolated!

 

OK on the actual drive connections. REALLY, though, the portion with the mosfets should also be shown. Because it is the whole system that needs to be evaluated to do an accurate analysis.
AND STILL!!! It is the time in the linear mode of operation that would be the source of most of the heating. THAT is the reason for having FAST switching, to reduce heating by not spending time in the linear mode.