IGBT bridge drivers keep blowing up (again :/)

[TmaxElectronics]

A while ago i built a full bridge driver for a tesla coil using dedicated half bridge driver ICs, which kept blowing up, so I made a post here about those (https://www.eevblog.com/forum/projects/igbt-full-bridge-issues/msg2900336/#msg2900336).
Now I implemented the suggestions I got here into a new design (schematic and board layout is attached) and switched to a different driver (L6494), that supports dead time control, but the drivers still keep failing.

They run for about 5 minutes, and then fail, shorting the input to ground and not providing any output. I did not observe any ringing at the outputs of the driver, and put in the correct drive resistors (10ohm for IMax of 1.5A),  so i (again) don't have any clue why they keep failing.

The attached switching waveform is unfortunately the only one I took a screenshot of, before the drivers broke.

Is there anything that could make these drivers blow up, other than inadequate input decoupling leading to ringing or too low output drive impedance?
I am kind of getting desperate having thrown a lot of money away by blowing up drivers 

[ace1903]

Embedded developer here so no expert knowledge but will try to help.
1uF seems too much capacitance for the bootstrap capacitor.  Have seen lot of schematics and usually lower value is used.
The datasheet suggests an external diode to be used when using large capacitance. There was st appnote explaining measures to be taken this capacitor to be kept charged all the time.
To me looks like the capacitor is drained after a while IGBT is underdriven and blows up.
If there is no one to suggest something smarter I would try external diodes, maybe smaller capacitance and for test
pure resistive load (maybe some heater or an incandescent lamp) . 

[moffy]

Maybe a clamping zener at the gate of each of the IGBT's to prevent dv/dt induced spikes.

[T3sl4co1l]

There's uh, no bootstrap power, so it doesn't work at all?

If you've patched that, then the incredibly loose layout, and complete lack of ground plane (and lack of one Kelvin gate drive pin?), likely puts all sorts of nasty transient voltages around the circuit.  In particular, peak voltages above 600V and below -25V will kill it.

IGBTs also don't behave well without negative Vge(off).  Not sure how much this has improved over the last few years (if at all), but expecting more than 50kHz from them at anywhere near full power (which ought to be ~20kW continuous for SOT-227 sized parts?) is woefully optimistic, even with biased drive.

Tim

[TmaxElectronics]

The bootstrap diode is integrated into the IC, and i am actually using the kelvin emitters of the IGBTs, but the schematic didn't show them.

I am also sceptical about the bootstrap capacitor being too large, because (i think) that would really only be a concern at power up. After that, all the bootstrap diode has to supply is the charge that was put into the gate of the transistor, so maybe gate charge is abnormally high? I think I will try to patch in an external diode though, as the internal resistance seems quite high at 175 Ohm.

And I think i explained myself poorly: what blows up here is not the IGBTs, but rather the drivers themself, even when there is no load being driven at all (and no power connected to the bridge, only to the driver).

[T3sl4co1l]

Ugh.  I read that as using FAN7390.  Why did you open your paragraph with the thing you aren't using and literally put the thing you are using, in parentheses?

So let me get this straight.

You attach a schematic for us to look at.  But you later add, it's the wrong schematic.

You attach a layout for us to look at.  But you later add, it's the wrong layout.  (QE2 isn't Kelvin connected and QE3 is backwards.)

Tim

[TmaxElectronics]

Ok the first thing was a bit confusing, i admit. I changed it now

But the schematic and the layout is still as I wired it, only that i didn't draw in a kelvin emitter connection. The datasheet of the IGBTs notes that either emitter terminal can be used as a kelvin one, so i just used the one that was more convenient for the layout on QE3.
And even though i forgot the kelvin connection for QE2 (yes using a dedicated symbol with KE would have saved me from that), that wouldn't cause the drivers to blow up without any load.

But you later add, it's the wrong schematic.

I wouldn't call it wrong, as it is still the way i connected everything. And where did i say that the layout was wrong?