Re: "zvs" flyback driver on mains?
Proud Mary, Thu Feb 21 2008, 05:07PM

Perhaps you should put some anti-parasitic chokes into the circuit, and make them sufficiently different from each other to avoid unwanted oscillatory coupling between them. Good power supply by-passing, short leads, and use of a common earthing point (rather than 'daisy-chaining') will all help with stability too.

Re: "zvs" flyback driver on mains?
Sulaiman, Thu Feb 21 2008, 07:08PM

I have considered this and so far I have not tried because;
1) The series/dc inductor will be quite large.
2) The primary transformer will be quite large
3) I haven't yet found a use for such power!

I am considering a 'zvs' type oscillator as a CW SSTC primary, just haven't got around to it.

1200 V is adequate, TO220 sounds a little small.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Thu Feb 21 2008, 10:19PM

1) Not too large but more turns of thinner wire will be needed
2) What do you mean by primary transformer? The main ferrite HV power transformer? The 1200W I stated is based on the core size of my xfmr.
3) I did :) Anything from arc drawing and burning things to Jacobs ladder to a Tesla coil (the output will be fullwave rectified and adjustable from 0 to max)

As to the TO220 devices, I'll probably 'pour' something on them, maybe parafin wax or hot glue (after I find it non-conductive) etc. to avoid flashovers between the pins.

Re: "zvs" flyback driver on mains?
Sulaiman, Sat Feb 23 2008, 05:54AM

I wish you well, and hope to see the working system.

If you like I will crunch numbers as a cross-check, given; operating frequency, core cross-sectional area.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sat Feb 23 2008, 01:55PM

The core is 3.7 cm^2, I don't know the ferrite material number but you can be sure it's for power transformers. The frequency is a variable to experiment with, I'll try different tank capacitor sizes/primary turns and see what works best. However the aim is ~40-60kHz.

I've found a possible problem, I'll be using gate drivers which need below 1.5V on the input to change the output to low, and the IGBTs have ~2V forward voltage drop so theoretically they can not turn the gate drivers low. Will this be a problem in reality? If so what could I do to fix this?


EDIT: What about "angle controller" with a SCR to bring the supply voltage down to say, 200VDC? It would cause bad power factor but I really don't care, and the driver would be quite simpler to build. Just don't think about what happens when the SCR fails short...

Re: "zvs" flyback driver on mains?
Marko, Sat Feb 23 2008, 02:59PM

I want to make the power stage run from rectified mains (325V) to avoid bulky mains transformer.
I plan to use gate drivers running from low voltage to drive the gates, I'm wondering if the faster switching wouldn't cause more parasitic oscillations and cause the circuit to be unstable.

Steve C. and W seem to have reported some problems, but I don't have clue what could cause them. If drain-capacitor and source-GND connections are kept short I don't see how could Q build high enough to undergo oscillation.

You need to take care though not to overload the tank and drop it's Q too low to oscillate.

You should consider some kind of overcurrent protection. A small CT and a comparator.
Even simple SCR saved me a lot of mosfets in halfbridge circuit.

Also I want to run it with 1200V IGBT's as MOSFETs of this rating are xpensive and hard to obtain. However the IGBT's I'm looking at are TO220 devices, I'm wondering if there wouldn't be any flashover problem between its legs? IIRC they are around 1mm apart, couldn't 1kV+ actually flash over this gap under certain circumstances? And is 1200V rating enough?

I'm pretty sure packages are tested for those voltages, and 1kV shouldn't be but a fraction of mm. Just make sure not to get solder, wires and other metal parts much closer than that.

Potting the device probably wouldn't help much as an arc, if it develops would track the package and go through the gap between the potting and package, as you can hardly make a good seal there.

I've found a possible problem, I'll be using gate drivers which need below 1.5V on the input to change the output to low, and the IGBTs have ~2V forward voltage drop so theoretically they can not turn the gate drivers low. Will this be a problem in reality? If so what could I do to fix this?

This was a big problem to me, especially when you consider that voltage drops add! I'm not sure but this may be one of reasons why people have trouble with IGBT's in this circuit.


My idea is to use a PNP transistor series with a pulldown resistor forming an inverter circuit, and put a inverting driver chip or some other circuitry after it.


PNP transistor would react even if base is pulled like a volt under the Ucc and give you large margin of feedback voltage.

For very high frequency operation this also allows using small diodes like 1N4148's in series since forward drop isn't so important anymore.

Haven't tried anything of this yet, just theoretical - sorry!

Hope you'll manage something with this.

EDIT: What about "angle controller" with a SCR to bring the supply voltage down to say, 200VDC? It would cause bad power factor but I really don't care, and the driver would be quite simpler to build. Just don't think about what happens when the SCR fails short...

I wouldn't recommend that especially if you were going to power a tesla coil, your output waveform would be modulated my the short fraction of waveform where SCR's are conducting.

and if you use a filter cap and choke your DC voltage would be greatly dependent on the load.

If I wanted a power supply for royer converter my choice would be PFC buck-boost converter.

It would allow regulation 0-100% and steady DC output. This would actually be my ultimate general purpose HV transformer driver, next step would simply be SLR converter.


Good luck sir, I might join in the project but not sooner than few months, as soon as I get out of the mess here.

Marko

Re: "zvs" flyback driver on mains?
thermite, Sat Feb 23 2008, 03:49PM

quick side question:
who is vladimir mazilli and where is his site or circuit originally published ?

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sat Feb 23 2008, 03:52PM

I've actually considered SLR too, it's pretty simple to get it right, but... There are problems with open circuit output ... and I killed a FET after playing with it a bit, at low power, I'm pretty sure I had everything right. I know this is no excuse, but I think it's not good for anything else than capacitor charging [Tesla coil].



I remember his circuit to firstly appear in a thread called "zvs flyback driver" on the old 4hv.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sun Feb 24 2008, 08:51AM

Today I tried "low voltage" (115VDC) version with 500V FETs and gate drivers. It starts to oscillate as I apply power but as soon as I make the first spark with the transformer, the circuit bursts into MHz oscillations and there is no output from the transformer.
I have asolutely no idea what to do with this, I have the tank capacitor soldered directly across the MOSFET terminals.
Schematic:

Re: "zvs" flyback driver on mains?
..., Sun Feb 24 2008, 09:01AM

I had very similar problems, the thing just doesn't work reliably off mains, and adding gate drivers just further complicates things. Really I think that for higher power levels you are better off dumping the self resonant design and try to get a PLL working, or better yet just go for a halfbridge.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sun Feb 24 2008, 09:31AM

PLL for what? I think we were talking about HV transformer driver, and I'm actually going away from half bridge and any hard switched designs, as the switching losses are too big and I exploded any halfbridge at the power level I want to run.
So I'm investigating all and any soft-switching drivers that could be used to drive the ferrite transformer and tolerate both open load and shorted load conditions.


EDIT: what about running it from fixed frequency? (probably not)

Re: "zvs" flyback driver on mains?
uzzors2k, Sun Feb 24 2008, 10:05AM

Given the problems everyone seems to have with their royer oscillator deviations why not use a PWM chip to keep the frequency under control? You could still achieve ZVS by using the same output stage with a resonant capacitor, and just tune it manually or use PLL feedback like Peter suggested.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sun Feb 24 2008, 10:12AM

Yes but the problem is that the tank resonant frequency is diferent with open and shorted loads.

Re: "zvs" flyback driver on mains?
Sulaiman, Sun Feb 24 2008, 11:01AM

The problem is that the 'zvs' inverter is 'supposed' to operate in near-sine mode, with some primary "Q"
From memory the MINIMUM "Q" is theoretically 2.3, try to use Q>=5 for reasonable zvs operation,
this means that the zvs inverter should never work into a short-circuit.
In this case Q = (reactive power in resonant circuit)/(output power)
If you don't maintain the Q then the invertor will not really be operating in zvs mode.

Re: "zvs" flyback driver on mains?
Marko, Tue Feb 26 2008, 12:06PM

Today I tried "low voltage" (115VDC) version with 500V FETs and gate drivers. It starts to oscillate as I apply power but as soon as I make the first spark with the transformer, the circuit bursts into MHz oscillations and there is no output from the transformer.

So the circuit remains in 'Mhz oscillations' indefinitely after that?

I really have no clue what could be going on, not enough data, knowledge, intelligence on my side.

The problem is that the 'zvs' inverter is 'supposed' to operate in near-sine mode, with some primary "Q"
From memory the MINIMUM "Q" is theoretically 2.3, try to use Q>=5 for reasonable zvs operation,
this means that the zvs inverter should never work into a short-circuit.
In this case Q = (reactive power in resonant circuit)/(output power)
If you don't maintain the Q then the invertor will not really be operating in zvs mode.

I don't know what else to recommend...

Jan, what airgap do you have on your transformer? T think the core should be significantly gapped to give the circuit enough Q in worst case.

But, conner also seemed to have same thing happen on induction heater so I don't think that is the problem.


My biggest question here is, what could anywhere in the circuit cause it to oscillate in Mhz range? I really don't see any parasitic tank circuits that could do that.

Could the circuit just be jumping to Fres of transformer's secondary? But why it stays there?

I'm not sure why would faster gate drive provoke it either. I've made a mosfet converter of this type with active pullups and it worked well at low power levels.

Don't know,
Marko

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Tue Feb 26 2008, 04:13PM

Marko, I really believe that they are oscillations between the transistors' internal inductance and the tank caps (and their ESL). The fact that the oscillations are so fast that I can't really capture them on my 5MHz scope only supports that, the power supply for the gate drivers actually drops to ~1/2 voltage drawing about 0.5-1A.

This also explains why faster gate drive makes it worse, as the transitions are faster and provoke more ringing which eventually has enough amplitude to ring below the turn-on threshold of the gate drivers, and bad thing happens.

Re: "zvs" flyback driver on mains?
uzzors2k, Tue Feb 26 2008, 04:15PM


ZVS is most beneficial when switching some current, so tune it for ZVS with an arc. I'm not sure what happens without a load then, but I don't imagine the losses will be any worse than when running the flyback open circuit while hard switching. Is there really that much to gain from constant ZVS?

Re: "zvs" flyback driver on mains?
GeordieBoy, Tue Feb 26 2008, 05:17PM

If you short-circuit the output of a Royer oscillator this short-circuit is reflected back to the primary side of the transformer and it kills off most of the inductance in the parallel resonant tank circuit. The only inductance remaining will be the "un-coupled" leakage inductance of the transformer referred to the primary side which will resonate with the primary cap, so I would expect the operating frequency to go right up.

If the transformer has really tight coupling then a short-circuit on the output might even decrease the Q of the circuit so much that is stops oscillating compeletely. See Sulaiman's post above. This is more likely to happen at high power levels than low, because a thicker high-current arc will present a lower load impedance to the output of the royer oscillator.

Re: "zvs" flyback driver on mains?
Marko, Tue Feb 26 2008, 07:43PM

Marko, I really believe that they are oscillations between the transistors' internal inductance and the tank caps (and their ESL). The fact that the oscillations are so fast that I can't really capture them on my 5MHz scope only supports that, the power supply for the gate drivers actually drops to ~1/2 voltage drawing about 0.5-1A.

That inductance is absolutely tiny... I really don't see how could it have any significant Q on that much capacitance on it, and oscillate at 5Mhz. I don't see how is feedback loop necessary for continuous oscillation established.

Maybe you meant for gate circuit oscillation? Gate voltage going under threshold like that is very implausible unless your driver placement was totally crazy.

Can you clarify how exactly do you think this oscillation is forming?

Anyone else can give some insight?

ZVS is most beneficial when switching some current, so tune it for ZVS with an arc. I'm not sure what happens without a load then, but I don't imagine the losses will be any worse than when running the flyback open circuit while hard switching. Is there really that much to gain from constant ZVS?

Your resonating inductance will drop proportionally to load; you can think of it as, that with un-loaded transformer you have Lleakage series with Lmagnetizing, and with shorted transformer only Lleakage.

I'm not sure if this equation is correct:

XLp = wL1 - (w^3*M^2/Rload^2*w^2*L2^2),

M -mutual inductance,
L1, L2 - magnetizing inductances,
w ang. frequency;;

bleh,

If your core is not gapped enough your leakage inductance will be very small and circuit will probably crash if you short it's output or oscillate at very high frequency.

Maybe you could also use capacitive ballast in series with output like they do in those little royer CCFL inverters - I don't know what good does that do vs. simply gapping the core a bit more?



Hi Richie,

So you believe it's all nothing but simply a matter of Q getting too low?

Can any of those ''parasitic oscillations'' guys keep mentioning happen at all?

It is interesting that Steve reported same problem on a royer used as an induction heater, which had no core at all! I don't believe it would be too easy to get the Q too low... there, with so low coupling?

Just curious

Marko

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Tue Feb 26 2008, 08:08PM

yeah, at least I can try to.. BTW the 5MHz figure is the "rating" of my scope, I believe the frequency was way higher.

I think it goes like this: The drain voltage of one FET starts rising and once it crosses the other gate driver's threshold, it's supposed to keep the other FET on. However there is a very high frequency ringdown of unknown cause, superimposed on the rising drain voltage. If the amplitude of this ringing noise is bigger than the input hysteresis of the gate drivers (in my case probably below 100mV because of the PNP transistor amplifying the signal), it causes the other gate driver and FET to turn off again, initiating the very high frequency oscillation.

One thing to note is that in the case of the VHF oscillation, the power circuit does not draw any "fault" current (just slightly higher than with open transformer) and there is no output from the xfmr; the FETs survive it most of the time.

Re: "zvs" flyback driver on mains?
Sulaiman, Tue Feb 26 2008, 11:12PM

If you make a zvs inverter with separate supplies for gates and power, like this



and connect just V2, the system will oscillate !
(this is why I say you should connect the gate supply before the main supply)
Not at full power - but it oscillates

So not only do we have the parasitic oscillations when Drain1-source1-source2-drain2 inductance resonates with the capacitor, there is a second more subtle 'parasitic' oscillation mode.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sat Mar 08 2008, 12:17PM

So has anyone any idea how to get rid of the parasitic oscillation? Right now I've made one additional winding on the core and I'm using it for voltage feedback, to an opamp (switch at zero cross). I think the additional-winding-voltage-feedback is the way to go - but the only problem is that I can't get it to oscillate well.
Any ideas?

Re: "zvs" flyback driver on mains?
Marko, Sat Mar 08 2008, 07:18PM

and connect just V2, the system will oscillate !
(this is why I say you should connect the gate supply before the main supply)
Not at full power - but it oscillates

So not only do we have the parasitic oscillations when Drain1-source1-source2-drain2 inductance resonates with the capacitor, there is a second more subtle 'parasitic' oscillation mode.

Yes, that is normal, and circuit works same as it does usually except it is powered only through gate resistors.
Oscillation happens at frequency of the LC tank, it will only jump in amplitude once drain power is applied.



According to Richie, the parasitic oscillation forms when circuit is overloaded and primary inductance drops too low, making mosfets effectively parallel.

Also, too low inductance may simply reduce Q of the circuit too low for it to oscillate.


A proper solution seems to be capacitive ballast on the output, as it is done with CCFL inverters.

Low value HV cap (few tens to few hundred pF, depending on frequency -> wanted impedance) should be used in series with output, and relying on leakage inductance to the transformer to limit the current is apparently a bad way -

One royer CCFL inverter I had here kept blowing up because it's ballast capacitor was shorted, despite it lit the tube well even in that state.

Marko

Re: "zvs" flyback driver on mains?
Proud Mary, Fri Mar 28 2008, 12:51PM


You could try putting some anti-parasitic chokes on the gates - chokes which have low Z at the desired frequency, and a very much higher Z in the parasitic range. Make sure the two chokes are of slightly different inductances/self-resonance so they do not set up an oscillatory circuit all of their own!

Re: "zvs" flyback driver on mains?
Steve Conner, Fri Mar 28 2008, 01:01PM

I can't see any reason to use anything other than a SLR converter at this high power level. We already know the SLR works- Steve Ward has got some great arcs from SLR-driven ferrite transformers- and just needs a fixed frequency oscillator with 50% duty. A fullbridge of TO-247 IGBTs should handle >3kW in the SLR, even those crappy BUP314Ds that kick around Eastern Europe

I believe the parasitics happen in a single-ended mode. The resonant circuit is between gate and source, and it's powered by feedback through the Miller capacitance. (The same feedback mechanism as Sulaiman observed, just a different resonant mode.)

The transconductance of vertical MOSFETs is extremely high to start with, and increases even more with drain current, so as you crank your variac up, you get more gain, more feedback, until the loop gain exceeds 1 and it bursts into song.

I'd bet the cure would involve ferrite beads on gate leads, gate stopper resistors, RC snubbers from drain to ground, etc. IGBTs would also suffer less than MOSFETs because they're slower. The original Mazzilli oscillator did in fact use BUP314s (they are 1200V) running off rectified 220V mains, and powered a SSTC with good results.

But that doesn't solve the fundamental problem of running a ZVS off mains: that the choice of 1200V devices is poor, and they're slower and more expensive than the 600V ones which are driven by development for consumer SMPS. It was much easier for me to go to a PLL driven halfbridge for my induction heater than try to debug the ZVS.

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Fri Mar 28 2008, 01:37PM

I actually got this circuit going right, the problem was in the feedback path. Now i use a voltage sense winding on the transformer and use this to switch the transistors on zero crossings. It works well and no "parasitic" oscillations.

The only thing I'd want to improve is that the delay in the fb path causes that the transistors switch a bit too late. Maybe PLL wil help here? I have no experience with it so I have no idea how would I set it up.

As to the 1200V devices - well you can use a current fed halfbridge (that operates with the same principle) and 600V devices can be used at mains 325V voltage.

Re: "zvs" flyback driver on mains?
jpsmith123, Sat Mar 29 2008, 02:03AM

Plasmaddict, I'm curious as to how you built your transformer? Did you wind the coil yourself? Did you encapsulate it with anything?

(BTW somewhere I saw a schematic from "Information Unlimited" where they drove what is supposedly a 10 kv 60 ma "flyback" transformer from the AC mains, selectable for either 120 or 240 vac, with a half-bridge using IRFP450s and an IR2153 driver).

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sat Mar 29 2008, 07:39AM

I just used two quite large cores from old flyback transformers and two identical HV coils which I had no idea where they come from, in series. I was pushing around 600Watts of real power, and the little IRF740's got just warm.

However I noticed one odd thing: If I use larger resonant cap, the arcs are thick but shorter and with smaller caps they are longer and thinner! How can this be? The primary voltage is still the same so how can higher frequency draw bigger arcs?? I don't get it...

Re: "zvs" flyback driver on mains?
Arcstarter, Sat Mar 29 2008, 06:02PM

I experience the same thing with my 555 driving irfp460.I get 2 inch sparks at audible frequencies but 1 inch powerful arcs at above audible range. I haven't as much experience as you do but i figure it is because there are so many sparks at a high frequency that they fuse together. And i base this upon the theory of absolutely nothin'.:)

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sat Mar 29 2008, 06:30PM

Well, this resonant technology is good for very high powers where you don't want to use expensive high current devices. This is only true when driving highly inductive loads. I was pulling 560 Watts of real power and getting ~8" arcs, and the small TO220 IRF740 transistors were just slightly warm.

Re: "zvs" flyback driver on mains?
Marko, Sun Mar 30 2008, 04:27AM

Hi guys,

Been looking at this now for a while, good job Jan!
Any pix of the arcs for me?

I actually got this circuit going right, the problem was in the feedback path. Now i use a voltage sense winding on the transformer and use this to switch the transistors on zero crossings. It works well and no "parasitic" oscillations.

The only thing I'd want to improve is that the delay in the fb path causes that the transistors switch a bit too late. Maybe PLL wil help here? I have no experience with it so I have no idea how would I set it up.

Glad to see.. how did you use the feedback transformer? Clamped and directly fed into gate driver inputs? or there is something in between?

How do you start the circuit up?


I've been reading over the articles on parasitic oscillations of paralleled mosfets, but am still pretty puzzled why the problem happens in royer and why is it so bad.

Richie B. pointed out, that in shorted output state, where resonating inductance is low, mosfets in push-pull configuration act very much like they are parallel, and experience parasitic oscillations in same way.

But, if I'm using low impedance gate drivers on each mosfet, how can ever the gate be affected by miller capacitance, yet with zero voltage transitions?
Or I'm not getting you correctly?

Am I right, that this parasitic oscillation is happening through the circuit's feedback path itself? The 'feedback diodes' are directly connecting drain of one mosfet to other's gate.

Slow gate pullup resistors would definitely dampen out any fast oscillations, so that explains why they happen with gate drivers and not with the pullup resistors?

If that's true, could one solve the problem by using some sort of ''debounce'' circuit in series with feedback?


I don't understand what you mean about the 'mode' observed by sulaiman, could you guys clarify that a bit?

When I connect just the gate supply without the drain supply, the tank circuit gets powered through pullups and diodes and oscillates at it's resonant frequency, which is expected.
I never noticed anything wrong about that?

I'd bet the cure would involve ferrite beads on gate leads, gate stopper resistors, RC snubbers from drain to ground, etc. IGBTs would also suffer less than MOSFETs because they're slower. The original Mazzilli oscillator did in fact use BUP314s (they are 1200V) running off rectified 220V mains, and powered a SSTC with good results.

I don't really understand yet, what exactly do the ferrite beads do?

Is their point to saturate with high gate charging current, but provide high inductance during on/off times?



I'm not sure how would PLL work there - I was thinking, if it fails to lock and VCO is far off, wouldn't the tank circuit represent low impedance and blow the mosfets up?

And even if it could it wouldn't really do anything about the phase shift, as far as I know.

My understanding of PLL is still quite poor.


I hope you 'll design the elusive bridge CF-inverter at some point Jan, have good luck

Marko

Re: "zvs" flyback driver on mains?
Dr. Dark Current, Sun Mar 30 2008, 07:54AM

Hi marko, sorry no pics as the arcs were not that spectacular, they were similar to the ones in the "moster arcs from flyback" video which I've posted in "Flyback arc contest" thread. However they were a bit more "flappy" and noisy.

I used a winding directly on one of the main power xfmrs. This goes into a fast invertor and then to TC4422 gate drivers. Start up is realised by a start-up oscillator.

This is what it think, that this "diode feedback" design is a little flawed. I can't say how did I come to this conclusion as it is very hard to imagine (and simulating this circuit in your mind cand be a mind-bending experience), but it has something to do with the gate threshold voltage, transconductance of the FETs, gate capacitance, and of course is greatly affected by the pull up resistors.

Heh, I've already "designed" it (by this I mean I could draw a working schematic) but there needs to be the little thing solved in the feedback, and that's the delay. Of course it would work as is but the delay introduces more loss on the transistors, especially at high frequencies.



Does anyone have an idea why bigger resonant caps result in thicker but shorter arcs? Why are they shorter if f is lower?

Re: "zvs" flyback driver on mains?
Steve Conner, Sun Mar 30 2008, 10:32AM

I guess the arcs get longer and thinner with increasing frequency because you're getting closer to the self-resonant frequency of the secondary coils. So on one hand, you get resonant rise like a Tesla coil, boosting the open-circuit output voltage, which makes the arc longer.

On the other hand, as the frequency increases, the inductive reactance of the secondary coils limits the output current more once the arc is struck, making it thinner.

I've also noticed the same effect when playing with a ferrite transformer from an electronic NST, that used to power a neon beer sign. When I tried driving it actually at the secondary's resonant frequency, it flashed over and caught fire within 10 seconds.

Re: "zvs" flyback driver on mains?
Proud Mary, Sun Mar 30 2008, 11:23AM


When you change the capacitor in the tuned circuit, you are changing the frequency of oscillation. At this new frequency, the impedance of the supply line choke, the Q of the tuned circuit, and the efficiency of the flyback, will all be different.

LOPTs from PAL 625-line TV sets will have been optimized for operation at 15.625kHz so can be expected to do best around this frequency.

Like you, I tried a range of capacitors in the tuned circuit from the 0.68uF specified in the original Mazzilli circuit, through to 4uF.

I measured the LOPT peak output voltage by charging a capacitor through a diode, and observed a range of outputs from ~3kV to ~4kV depending on capacitor size (and therefore fo.) when running the oscillator from a low impedance 15.5V supply.

I used a 275V polycarbonate for the 0.68uF, and 375V motor start capacitors for the 1, 2, 3, and 4uF tried in my experiment.

I used cheap IRL640 MOSFETS in my version, with 300uH for the choke, a value which I increased beyond the original Mazzilli design values because of the lower frequency of my circuit.

Best results (3.95kV pp) were obtained with the 2uF, when the circuit oscillated at 18.6kHz, close to the original LOPT 15.625kHz design frequency. (The highest voltage - 4.1kV - was obtained with a 1uF tank capacitor, but when the output was loaded with a 4-stage CW, it sagged below the level that could be sustained at the lower frequency provided by the 2uF)

I have not tried computer monitor LOPTs, which are optimized to work at much higher frequencies.