You can't just flip them like that! You need to redraw them with right part/symbol. And use the right names. The original circuit you showed used NPN transistors. You have, for some unexplained reason, swapped them for PNP devices. That's what's confusing things.

But you have P-channel MOSFETs. The load for those must be on the drain side as I drew it, but you've turned them round and put the load on the source side. That's not going to work as you think because the gate must be more -ve than the source to turn it on and so as volts builds up in the coil the MOSFET is being switched off so can never get the full drive you expect.

I'm not sure you fully understand how BJTs and MOSFETs work...

 

any way to check if that's the problem/fix it?

That might be related but isn't your fundamental issue, which is how you're trying to use the MOSFETs.

 

You can't just flip them like that! You need to redraw them with right part/symbol. And use the right names. The original circuit you showed used NPN transistors. You have, for some unexplained reason, swapped them for PNP devices. That's what's confusing things.

But you have P-channel MOSFETs. The load for those must be on the drain side as I drew it, but you've turned them round and put the load on the source side. That's not going to work as you think because the gate must be more -ve than the source to turn it on and so as volts builds up in the coil the MOSFET is being switched off so can never get the full drive you expect.

I'm not sure you fully understand how BJTs and MOSFETs work...

the original pic had NPN transistors, I have only have PNP transistors, so I had to wire it up accordingly, the drawing had somewhat ambiguous symbols for a reason.

All right, why the heck did they bond the drain to the chassis ground? If the load needs to be attached to the drain for reliable operation, and the drain is bonded to the heatsink, that means the FETs either MUST be used with a dedicated heatsink, or they MUST be ganged together, why would they do that, why wouldn't they bond the source to the heatsink instead, that would make more sense. That's probably why the feedback winding needed to be so large for the BJTs as well, it's simply overpowering the turn off effect you are referring to via brute force, and since it's being used backwards that would generate more heat, and tend to let the smoke out, am I correct?. Also, do N channel FETs have the same problem? Kinda looks like they do.

 

When a FET fails where the gate, source, and drain are all bridged, but the FET isn't hot, is that the failure mode when the gate over-volts? if so, would adding a pair of opposing zenier diodes between the gate and source prevent this failure mode? I'm thinking 18V zeniers so the gate voltage will clamp at +-18.6V

 

Suggest you look at how MOSFETs and BJTs are constructed, and you'll see why drain/collector end up being the main heat removal path. Yes it can be a pain sometimes but that's why good engineering knowledge is needed not random guesswork.

As to MOSFET failure mode, given where you're coming from, it's impossible to say. Let's get a circuit that has a reasonable possibility of working, then we can analyse probable failure modes. Gate punch through is one of the least likely.

I'll endeavour to draw up something using P-channel MOSFETs and PNP BJTs that has a chance of working without destroying itself later today. What transformer are you using? How is it constructed, # of turns per winding, and any other info?

 

all right, I'm using these cores:
https://www.digikey.com/en/products/detail/epcos-tdk-electronics/B64290A0699X087/3913464

4.5 windings of 14AWG wire for each half of the primary, 9 winds total. 5 winds of smaller wire for the feedback winding, transformer is hand wound, all windings can be changed easily, there's a smaller core available that can be wound into an inductor, no information on it, but I have an inductance tester, the minimum inductance is ~20 micro Henry.

Anyway gate punch through is more likely than you think, I have a talent for using semiconductor devices wrong, they were running of 4V, couldn't get anywhere near enough current to fail other ways, and they didn't heat up, the gates were connected to the sense coil, and nothing else, the gate capacitance probably formed a resonant circuit with the sense coil, which pumped itself using the power side of the circuit until the voltage was high enough to nuke the gates.

 

all right, I'm using these cores:
https://www.digikey.com/en/products/detail/epcos-tdk-electronics/B64290A0699X087/3913464

4.5 windings of 14AWG wire for each half of the primary, 9 winds total. 5 winds of smaller wire for the feedback winding, transformer is hand wound, all windings can be changed easily, there's a smaller core available that can be wound into an inductor, no information on it, but I have an inductance tester, the minimum inductance is ~20 micro Henry.

Which material for the toroid?

 

N87 core material, suitable for push-pull transformers.

 

Running the BJTs in the correct direction seems to have helped with the heat problem, as did adding shottkey diodes to take the stress off the body diode, the shottkey diodes are dropping 0.11V, the body diodes drop 4.5V, the shottkey that isn't mounted on a heatsink is heating up pretty quickly, and it's dropping a quarter of what the body diodes do, i might have to ditch the FETs, they bridge out internally first chance they get when driving them in the correct direction, still can't figure out why, but the problem is repeatable.

 

Are you still running PNP BJTs or did you swap them? Its very hard to make a workable solution with P-channel MOSFETs and PNP BJTs. Can you post your latest schematic?

 

I'm basically out of P channel FETs, if I order more in they can be N channel

This is what works with the BJTs, this always works with the BJTs, even if they are wired up backwards, the feedback winding needs to be bigger if they are wired wrong, but the schematic doesn't really show that anyway. The two 500Ohm resistors are providing a current path for the BJT bases, and biases them on, whichever BJT is faster on the draw bootstraps the circuit, this works because they are current driven, and react to current immediately, this doesn't work for the FETs because the gate is a capacitor, and the cap has to charge to turn the transistor on, it has to discharge to turn off, this basically gives FETs inertia, and lag that prevents bootstrapping, both always end up on at the same time.



Your corrected version of my hand drawn schematic should work, it sort of worked when I was running all the transistors backwards, anyway, but as you pointed out, the BJTs and FETs operating on the same channel is causing problems, if I get N-FETs they should work with the P-BJTs on hand, probably should get some smaller BJTs as well since using a 20A power BJT as a gate driver seems excessive.

 

N-channel MOSFETs are preferable to P-channel as generally have a lower on resistance. To drive them from BJTs you ideally need PNP devices as they will actively source current to 'pull-up' the MOSFET gate. Yes, you're right, those power darlingtons are overkill, but you'll still need something capable doing an amp or so, something like a ZTX705.

Here's the N-channel/PNP version. You don't need two base resistors for Q1/Q2. R3 is sufficient to bias both Q1 & Q2 as the DC resistance of L3 is tiny.

 

N-channel MOSFETs are preferable to P-channel as generally have a lower on resistance. To drive them from BJTs you ideally need PNP devices as they will actively source current to 'pull-up' the MOSFET gate. Yes, you're right, those power darlingtons are overkill, but you'll still need something capable doing an amp or so, something like a ZTX705.

Here's the N-channel/PNP version. You don't need two base resistors for Q1/Q2. R3 is sufficient to bias both Q1 & Q2 as the DC resistance of L3 is tiny.

View attachment 240274

That looks beautiful, I'll get some parts ordered to see how it works. Thanks for the help so far, I would never have realized I was using all the transistors backwards otherwise.

These look good:
https://www.digikey.com/en/products/detail/IRFB4110PBF/IRFB4110PBF-ND/935978?itemSeq=366829129



With the BJT version, the BJTs are current driven, electrons go through the base into the load, they want to flow in a loop and have to come from somewhere, they either have to come from the base of the other BJT, or there has to be some other path back to ground, the two resistors provide a path back to ground, the BJT only circuit runs better with the two resistors, or at least the waveform is cleaner on the scope, also if there's only one resistor, and it goes to the non-dominant gate the waveform is also messier, yeah it's weird, but these self driving resonant circuits rely on weirdness to bootstrap themselves. I can swap them for 1K resistors, especially since the base current is going to be lower.

 

Those MOSFETs will do, though they may still need a heatsink due to switching losses even though conduction losses will only be a few watts. What's your expected switching frequency?

With darlington drive BJTs the bias current is small, it only needs to be a few mA or so. So 1k is easily enough, you could even try 2k2. Ideally that feedback winding would be centre tapped and a single resistor bias applied there.

 

Expected frequency is between 3 and 6KHz depending on drive voltage, I thought about center tapping the feedback winding, it would slightly improve the efficiency, and reduce the parts count of the circuit, but for this particular application efficiency is only desirable because it reduces the amount of heat the transistors, and diodes need to get rid of, and increase the lifespan of the device.

 

Parts came in, it seems to work, I'll have to get it set up for full power testing.

 

All right, I'm building a Royer oscillator, basic push/pull configuration, if I use BJTs as the drive transistors there are no problems, aside from BJTs being massively inefficent when used in 12V power electronics, so they tend to cook themselves when I try to add a load. 2 BJTs in parellel wasn't quite enough, so I would need to use at least 6 20A BJTs total to build the oscillator, and add more heat sink, obviously this is getting ridiculous, so I thought, why not use FETs instead?

Anyhow, so I found these FETs:
https://www.digikey.com/en/products/detail/vishay-siliconix/SUP70101EL-GE3/7622840

They refuse to run properly on their own when used in the same configuration as the BJTs (the BJTs always require a control winding that is a few turns larger than the primary, exactly how many extra can be kinda random), or in the configuration in this schematic:

they both turn ON at the same time whenever they are driven directly by the feedback winding when a 500 ohm bootstrap resistor is added between the gate and ground (PNP instead of NPN), the BJTs were perfectly fine with this configuration, 1Kohm has the same effect, basically they will not bootstrap properly, adding more winds to the feedback winding does not help. I stuck a BJT to the gate as an amp, the feedback winding can't shut off the gate because the BJT blocks it, so I added a 250ohm pull up resistor to the gate to shut the FETs off when not being driven by the BJTs, this finally got the FETs to behave themselves, sort of, they started heating up rapidly, even with no load attached, and the voltage is a lot lower than it should be, it should be ~24V across the primary, but it's only +-7V, less than half of what's expected, the frequency is a lot lower as well ~1.3KHz instead of the ~3.9KHz the BJTs were running at, like they are being driven in their linear region (which doesn't make any sense, the BJT is switching the gate directly to ground), or aren't shutting off all the way. The waves look square, if the gate wasn't being shut off fast enough by the pull up resistor I would expect to see it on the scope, same for a slow turn on. Reducing the input voltage to 8V only slightly reduces the voltage, and frequency, dropping to 3.7 is very odd, the frequency goes up instead of down, 2.6KHz, peak voltage is ~4V, waveform has the same spikes the circuit has when using BJTs only.

Obviously I'm missing something here.

Is'nt there supposed to be another resistor biasing the lower transistor?

 

There is another resistor in my design, it seems to run better, but it's not technically necessary, and it's not in other people's designs I found. The design I started with is on comment #31, but originally i was using the BJTs backwards.