Quote from: picowatt on June 23, 2012, 11:40:23 PM
TK,

Yes, it is possible for a MOSFET to fail in an open circuit or mostly open circuit failure mode.  You will likely agree, however, that this is more rare than failing in a shorted condition.  However, it is quite possible to fuse open a die connect lead/die connection or fracture part of the die itself, so a failed Q1 is always presented as one possible explanation for Q1 not turning on in FIG3 and FIG 7. 

As to your 2922... ouch!

PW

Heh... around here, mosfets and other power transistors fail either "open", or "I can't find all the pieces and I worry about the neighbors calling the cops".
But yes, I'll happily concede that shorts are more popular than opens. It's just that when I manage to short a mosfet there is usually a lot of energy behind it and the short very spectacularly becomes an "open" because the damn thing explodes from the power arc inside its little body.
I usually use gate-source protection zeners and I think I might have had 30 v ones in there instead of the more conservative 15 v ones I usually use, but my Class E waveforms were lousy and so it was likely the switching losses that heated up the mosfet so badly. I was giving it less than 30 volts drain-source and only 4 volts gate bias, though. And it's on a good heatsink.
I've got a bit better tuning in Class E now and I can light up a neon bulb from the secondary, with only 5 volts D-S into another NTE mosfet, a 2394. I think that one cost 13 dollars, so I'm taking it easy with it, still matching coils and playing with coupling. Yes, I'm on a good big finned heatsink with white slime and everything, considering adding a fan once the coil is working right and making corona.

Quote from: TinselKoala on June 24, 2012, 12:26:17 AM
Heh... around here, mosfets and other power transistors fail either "open", or "I can't find all the pieces and I worry about the neighbors calling the cops".
But yes, I'll happily concede that shorts are more popular than opens. It's just that when I manage to short a mosfet there is usually a lot of energy behind it and the short very spectacularly becomes an "open" because the damn thing explodes from the power arc inside its little body.
I usually use gate-source protection zeners and I think I might have had 30 v ones in there instead of the more conservative 15 v ones I usually use, but my Class E waveforms were lousy and so it was likely the switching losses that heated up the mosfet so badly. I was giving it less than 30 volts drain-source and only 4 volts gate bias, though. And it's on a good heatsink.
I've got a bit better tuning in Class E now and I can light up a neon bulb from the secondary, with only 5 volts D-S into another NTE mosfet, a 2394. I think that one cost 13 dollars, so I'm taking it easy with it, still matching coils and playing with coupling. Yes, I'm on a good big finned heatsink with white slime and everything, considering adding a fan once the coil is working right and making corona.

TK,

I dispute nothing you are saying.  In the NERD circuit, however, Rload does act to limit current.  But, also, some of the waveforms I have seen made prior to the demo do look like they are stressing the heck out of Q1.  I will accept any of the possible explanations I have given as to why Q1 is not functioning in FIG 3 and FIG 7.  I just won't accept that the 'scope is being read incorrectly.

As for your issues, even with a low impedance gate drive circuit, and depending on your switching frequency, it is still possible to have the gate voltage exceed its max voltage even with the zeners.  Just the small inductance in the gate lead, and any interconnect lead length, can ring or overshoot and cause the gate max V to be exceeded.  Hence, usually a low value non-inductive resistor in series right at the gate is often used to to work in concert with Ciss to dampen any overshoot or ringing at high frequencies.  This is usually difficult to see with a 'scope, as even the small capacitance of the 'scope probe will many times dampen the ringing/overshoot that would otherwise be visible.  And remember, high frequency does not always mean fast cycle times.  When using square waves, fast rise/fall times equate to high frequency regardless of the cycle rate (rem Fourier).

What freq were you running your coil at?

PW

TK,

And when I do look at high risetime gate drive signals with a 'scope, I tend to overdamp the switching (increasing the gate R, etc) which, although not always the most efficient (a bit more heat to deal with), allows me to sleep a bit better.  I always figure it is better to overdamp than underdamp.  Of course you don't want to be so slow that in a brdge, for example, one set does not turn off before the opposing set turns on.  Hence, plenty of dead time.  I really don't like working with high power switchers all that much and do so only when I have to.  My roots are in analog.

Where can GL's circuit be found, is it availble online?

PW 

Quote from: picowatt on June 23, 2012, 11:32:14 PM
The following excerpt from an earlier post better provides some multiple choices from which to choose as to why Q1 is not functioning properly in FIG3 and FIG7.

Choose away:

"The facts are that the 'scope capture in FIG3 of the first paper (not the COP=17 article) demonstrates that during the portion of the FG duty cycle wherein the FG output is a positive voltage, approx. +12volts is indicated as being applied to the gate of Q1.  This is more than sufficient gate drive to turn Q1 fully on.  During this same portion of the FG cycle, however, the CSR trace does not depict the current flow one would expect to see when Q1 is turned on.  Possibly Q1 was damaged during these tests, although it is fairly rare for a MOSFET to fail open circuit.  Possibly, in the confusion of paralleling MOSFET's, Q1's gate and source lead were reversed, placing all five MOSFET's in parallel and in the common gate configuration.  Possibly Q1 was overheating, and as it was believed at that time to be in parallel with the other four MOSFET's, Q1 was innocently disconnected and the four Q2 MOSFETs only used during these tests.  Possibly a clip lead or connection to Q1 became loose and disconnected.  We cannot know what happened to Q1 at that time, but we do know that the 'scope captures indicate that Q1 was not connected, at the time of the related captures, as is indicated by the first paper's schematic.

As well, Fig 7 in the first paper also indicates sufficient gate drive to turn on Q1, and again, the CSR trace depicts that Q1 is not turning on."

Guys - picowatt knows perfectly well why Q1 doesn't conduct during this phase of the oscillation.  Thankfully this is CERTAINLY a test that is going to be replicated by our EXPERTS.  And they've also been sent our function generator.  So there can be NO chance of it not being 'repeated'.  Roll on the time when I'll be able to show you all the difference between experts and these sad and sorry pretenders to any expertise at all. 

The most ridiculous of these so called experts is TK.  LOL.  A full on university academic with multiple papers published on mathematical models.  And yet he has time to bore us to tears with his exemplary lack of professionalism while he engages in these HIGHLY UNPRODUCTIVE forums - obsessing about blown MOSFETs - when he's not indulging in obscene calumny.  And that's not taking into account the amount of time he spends experimenting on apparatus in especially dimmed light all the better to hide his rather untidy, utterly confused work.  Work that a student would be shy to own to.

God help us all as we're exposed to their rather sick agenda

Regards,
Rosemary

My dear picowatt

Quote from: picowatt on June 24, 2012, 02:17:13 AM
And when I do look at high risetime gate drive signals with a 'scope, I tend to overdamp the switching (increasing the gate R, etc) which, although not always the most efficient (a bit more heat to deal with), allows me to sleep a bit better.  I always figure it is better to overdamp than underdamp.  Of course you don't want to be so slow that in a brdge, for example, one set does not turn off before the opposing set turns on.  Hence, plenty of dead time.  I really don't like working with high power switchers all that much and do so only when I have to.  My roots are in analog.

Where can GL's circuit be found, is it availble online?

Spare us all the need to pretend that you ever get out of your armchair to test anything at all.  Unless of course you'd care to show us some shots of your apparatus.  And some data.    You fool no-one.   

Rosie Pose