SSTC FETs dying for no reason (was: MOSFET failure modes) / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

SSTC FETs dying for no reason (was: MOSFET failure modes)

1 2 3 next

Move Thread

LAN_403

Dr. Dark Current

Sat Apr 10 2010, 07:45AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

I have blown tens of those little IRF740 MOSFETs in SSTC experiments and can't figure out what's going on. I have even blocked their internal diodes, included a simple overcurrent shutdown circuit, but no matter what I do they always seem to blow after a while of experimenting with the coil. The gate waveforms are clean of course.

To the failure mode: The G-S junction always fails in the same way, with ~40ohms resistance, D-S usually reads a few hundred mv drop in both directions (note that the input current was always limited and there is no electrolytic filtering cap in the circuit, just two dc block film caps).

Anyone recognizes this failure mode? I'm really lost now, the only thing I can think of is the PLL circuit I am using, but I'd hate to go for a fixed oscillator.

I'll also try using ultrafast IGBTs, maybe they'll be more durable

Goodchild

Sat Apr 10 2010, 05:40PM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

I think it's as simple as buy bigger high rated MOSFETs IRF740 are not exactly high performers try some IRFP260s or some with similar rateings.

ScotchTapeLord

Sat Apr 10 2010, 06:51PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

What kind of snubbers are you using?

Dr. Dark Current

Sat Apr 10 2010, 07:43PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Goodchild: here we have 230V mains so >400V FETs, the IRF740 is rated 10A continuous 40A pulse, and I have around 12A peak and less than 4A rms. The little heatsinks never got hot, just warm.

ScotchTapeLord: I don't use any snubbers, I have two 1.5uF DC block caps soldered directly to the MOSFET legs, and two 18V "anti-series" zener diodes on each gate.

I had another idea, I use a very low-leakage-L GDT and the gates are discharged directly through a diode (charged through a 10ohm resistor), and I was wondering if the very fast turn-off couldn't be causing some dv/dt problems. So if everything fails, I'm gonna try and remove those diodes.

cjk2

Sat Apr 10 2010, 07:50PM

Registered Member #51 Joined: Thu Feb 09 2006, 04:17AM
Location:
Posts: 263

Try using a good snubber, put zeners on the gates of your fets, put a TVS from Drain to Source on your fets. If you still kill them, then they are probably overheating.

Also, check for shoot through.

Arcstarter

Sat Apr 10 2010, 08:41PM

Registered Member #1225 Joined: Sat Jan 12 2008, 01:24AM
Location: Beaumont, Texas, USA
Posts: 2253

Yea, voltage spikes could be a killer. They would avalanche, but that could possibly kill the die from overcurrent. But the DC blocking caps should not allow that, unless you avalanched both mosfets at the same time and shoot-through killed it. Or maybe it avalanched for too long and the die got so hot it died, so fast that the heatsink did not because noticeably hotter.

Do what cjk2 said, you won't regret it! Snubbers also deal with ringing caused by inductance which is half the battle in protecting from over voltage.

Dr. Dark Current

Sat Apr 10 2010, 08:56PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

But I would put the snubbers directly across the caps, does that make any sense?

Maybe if the caps had non-negligible ESL, but I think it is very small.
The die can't possibly overheat from avalanching, because the coil runs from halfwave rectified mains and no interrupter, so the thermal impedance shouldn't be a problem.

Sulaiman

Sat Apr 10 2010, 09:20PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Do you have a suitable diode from each source-to-drain ?
Are the esl and esr of the dc bus low?

The answer to both should be YES,
in which case it can only be Isquared.Rds.t heating.

If you suspect high esl or esr of the dc bus then put a voltage clamp across the bus (TVSs, MOVs etc.)

How does the current limit work?
IF it is before a large bus cap then massive current pulses can be drawn from the capacitor(s), easily enough to destroy mosfets.
If it is after the bus capacitor(s) then the current limit circuit would allow the dc bus to 'spike' to very high voltages when the load current 'comes back' to the bridge.

So DC bus voltage clamping, a diode across each source-drain and a suitable fuse in series with EACH transistor should be 'bullet-proof'.

Dr. Dark Current

Sat Apr 10 2010, 10:10PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Sulaiman
I have blocked the internal diodes with 30A series schottky diodes and 15amp/600V ultrafast soft recovery antiparallel diodes.
I believe the esl and esr are low enough, it's still a "birds nest" on the table but I tried to keep the wires as short as possible (and as I said, dc block caps go directly to the semiconductors legs).

The transistors never fail hot, they are cold or just warm (the coil will run for a minute with the heatsinks getting just warm).

The overcurrent shutdown senses current through the primary winding with a current transformer, which is fullwave rectified with ultrafast diodes, shunted and through a 1k resistor goes to a gate of a small SCR which pulls down the enable pins of the gate drivers until power is restarted.

So couldn't it be a too high dv/dt during the switchnig transition? Ive seen this value in the absolute maximum ratings of some transistor datasheets, but usually it's not there.

Thanks for the replies

ScotchTapeLord

Sat Apr 10 2010, 10:10PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

I had almost the exact same problem... solved 100% by putting capacitors from the drains of the top of each leg to the sources of the bottom of each leg.

This is what I figure:

If there's any kind of dead time (which there will be, unless you have even the tiniest bit of shoot-through, since there is either a point when all transistors or off or on during transition [in reality]), and there probably is with such fast-switchers as the IRF740, then there is a spike across the entire bridge. Without any kind of filtering, this spike is the result of the inductance of your wiring and your supply's wiring (probably house wiring?), along with the sharp drop in current. An excellent means of 40-ohm resistor production, but not for running a solid state tesla coil.

So try the snubbers, just a few uF across the bridge, and see if anything changes.
THEN switch to IGBTs! :D You won't be impressed with the capabilities of the IRF740, most likely.

1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob