Dulal ZVS - dual flyback driver doesn't work at high frequency (150 KHz)

[marko73]

Hi, everybody,
I recently built a dual ZVS flyback driver (4 mosfets: IRFP260N) and used it to power 2 identical flybacks (with voltage doubler inside) separated by driver transformer. I understand the basic principle of working of zvs, and how to calculate frequency, but i noticed strange things: when I drive it with capacitor of 1.5uF i have nice arcs which can melt a metal (f= 20kHz unloaded), but with capacitor 0.5 (f= 80kHz unloaded, 150kHz loaded) there is much larger arcs (i assume: higher the frequency = higher output voltage??), and there is the problem: in the moment when large arc is formed all mosfets are stuck on, they do not oscillate (this doesn't happen on lower frequency). I read something about making air gap on primary, and this flybacks are from TV so there is no enough space for it (currently only 3mm space on flyback where gap is 1mm), so can I make air gap on primary of driver core, and what dimensions it should be?
Power supply is transformer rated 25V 30A, full bridge rectifier 1000V 50A, capacitor 4700uF.
Important: I use 2 chokes each 70uH, but they may be slightly different, i couldnt find two same, so i wound one to have similar inductance to first one. Ferite cores have different permeability, one is green-blue, second yellow. 

[davekni]

Nice project!  A few questions:

Those look like normal TV flyback transformers.  I've never seen a voltage-doubler incorporated into a flyback.  Normal DC-output flyback transformers have effectively a single output diode, implemented as a set of series diodes, one between each section of secondary winding and the next section.  This reduces secondary AC voltage and associated winding capacitance loading.

Still, it appears that you are getting some effect of secondary capacitance (or some other artifact).  If unloaded frequency with 1.5uF is 20kHz, then 0.5uF should be sqrt(1.5/0.5) higher or 34.6kHz.  Operation at 80kHz is a different mode, perhaps resonating with leakage inductance rather than normal parallel inductance.  If these are normal NTSC or PAL standard TV flyback transformers, they are designed for roughly 15.7kHz.  It's hard to know what's happening at 80kHz.  Even with the segmented secondary windings, you are likely hitting some resonance of secondary winding capacitance.

Hopefully your chokes are powdered iron (or iron alloy), not ferrite.  If ferrite, they are almost certainly saturating, contributing to strange behavior.  Even if powdered iron, you may be saturating the choke cores.  It's hard to know without measurements or specifications for those parts.  If saturation isn't an issue for either choke, then a small difference in inductance shouldn't be a problem.

Oscillation dropping out and all FETs conducting is the normal behavior for ZVS oscillators when resonant Q gets too low.  (Also occurs at power-up if supply ramp-rate is slow and available current isn't high enough.)  At lower frequencies where secondary winding capacitance isn't significant, core gapping can keep leakage-inductance high enough to keep Q above the drop-out threshold.  Flyback transformers are made with an internal gap, which is likely sufficient.  At 80kHz, it's hard to know what is resonating and what is controlling Q.

[marko73]

Now i just measured each of capacitors and one of them is apperently dead, so there is 1u or 0.26u.
I just scoped primary for 1u, and it is 50kHz. There are spikes at origin, are they dangerous?
Any lower capacity than 1u result in all mosfets on.
When zvs is off, there is still voltage on output, so there must be one capacitor and diode inside flyback.
Is output voltage somehow related with operating frequency? At 80kHz there is corona (not virus ) discharges on cores on flybacks.
If i add more 2 flybacks, which are different than previous, will they get out of phase?

[davekni]

50kHz is still well above operating frequency for normal TV flyback transformers.  It's not surprising that high frequency hits a resonance of the secondary windings.  (There are higher-frequency flybacks for CRT computer monitors to support their higher resolution.)

Concerning "spikes at origin", can you post scope traces - even a camera snapshot of the scope screen?  When you say "origin", are you referring to when power is first switched on?  It is common for ZVS oscillators to have transient higher voltage at turn-on.  That is often problematic.  Depends on whether the voltage exceeds ratings of any parts, especially the FETs, but could also fry secondary diodes.

What output voltage to you see with the ZVS off, and for how long?  Most flyback transformers have only parasitic capacitance.  For typical designs, the CRT provided enough capacitance to filter the flyback output, so no additional capacitors are included inside the transformer.

Corona will be stronger at higher frequency even if voltage isn't.  Corona current is through stray capacitance that includes air (where one sees corona).  Capacitive impedance goes as 1/frequency, so corona current is higher at high frequency.  Voltage might be higher too - depends on details of secondary resonances.  The more scope plots you share, the more can be determined about what your circuit is doing.

If you add more flybacks, I'd connect secondaries in parallel.  Even two in series is stressing the transformers, as evidenced by the corona.  Flyback insulation isn't designed for high-voltage between the primary/core and the low-side of the secondary.  The corona will likely lead to eventual insulation failure with continued operation.  Parallel adds current rather than voltage, but that leads to longer arcs after initiated.  Arcs can be stretched out further because higher current ionizes more air, making the arc path lower resistance.

[marko73]

I don't have HV probe, but there is just little arc when connect ouput. I have one flyback from monitor, and it works fine on this high frequency. When there is only one flyback online everything is fine, but with two, sometimes FETs stuck on. I will try reducing frequency further. And connect them in parallel. So for higher voltage on output I need higher voltage on input (as long as it is lower than 50V, Irfp260 are rated for max 200V). This is supposed to be HV power supply for my 50cm secondary ARSGTC, later it will be dual MOT.
Scopes are for one flyback online, no load. By origin I meant when sine wave hit 0 there is spike, see picture.
When it is loaded there is no much change in frequency, just amplitude on primary drops to 80V.

[davekni]

The "spikes" at the voltage zero-crossings are actually short bursts of much-higher-frequency oscillation.  I get that too unless the resonant capacitor(s) are very close to the FETs with short leads.  Sometimes that can be fixed with smaller (10-100nF) caps from source-to-drain on the FETs.  The high-frequency oscillation doesn't seem to cause much trouble, unless you are listening to a radio station near that frequency.

The gates are taking about half of their on-time to charge to 12V.  That may be contributing to problems at higher frequency.  The simple fix is to reduce resistance of the gate pull-up resistors.  That does require higher-power resistors and zener diodes.  Other options require slightly more complex circuitry.  A couple such options are in my Jacob's ladder thread:
https://highvoltageforum.net/index.php?topic=831.msg5491#msg5491

It looks like you are connecting the "ground" lead of the scope probe to active circuit nodes.  Even with the scope or power supply isolated, the ground lead often has enough capacitance to alter circuit operation.  To be sure, I suggest always scoping with the ground clip on a DC voltage point, usually the negative supply (FET sources).  Then you are scoping individual gate or drain waveforms.

Parallel flyback secondaries will be best for feeding your ARSGTC.  TV flybacks are generally rated for somewhere between 20kV-30kV, at least for color TVs.  That's with asymmetric flyback drive.  With symmetric sine-wave drive, output voltage is less (for given diode reverse voltage capability), but even 15kV is more than you will get from two MOTs.  (The MOTs will provide much more current.  That's what my SRSGTC uses - two MOTs in series, 9kV using resonant MMC charging.)  BTW, I use a single flyback with sine-wave drive to get 21kV for my coin shrinker.

[marko73]

I reduced frequency to 30kHz and everything works fine, 4 flybacks, secondaries are in parallel. But zvs is capacitor killer, it is probably bad caps, what are best caps for zvs: mkp, mkt...?
Power supply I am using is obviously weak, it is already at limit, but I have one bigger transformer. Just before something bad happens, will this caps survive 50V without explosions, wire connecting them is melting.
Currently it is only 0.6kW and it can melt small screw washer.

[davekni]

You need polypropylene (PP) dielectric caps, either MKP or MFP.  Even within those polypropylene types, some are rated for higher AC current than others.  Depends on internal construction.  Most polypropylene caps will work.  Almost any other type will overheat and fail with such high AC current, such as MKT.

0.6kW into four flybacks may be close to their current capability.  At some point the internal diodes and/or secondary windings will overheat and fail.  I'd suggest enjoying the power you have and not pushing it much further, unless you have spare parts available.

When you use this for your ARSGTC, make sure you have some inductance or resistance (Terry Filter) between the flybacks and spark gap.  There's another thread here somewhere about feeding an SGTC with a flyback and ZVS oscillator.  Flyback transformers kept frying, probably due to the sudden discharges.  Internal voltage distribution within flybacks is uneven during sudden changes, which can lead to local internal breakdown.