High Voltage Forum

Tesla coils => Spark Gap Tesla Coils (SGTC) => Topic started by: jturnerkc on November 06, 2019, 05:19:13 PM

Title: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 06, 2019, 05:19:13 PM
This is my first, "lets get serious" SGTC build.
This will also be my first post here at the HV forum!
Technically I made a small SGTC first, but it was simple enough, so we're going larger, but still budget-minded scale now.
I'm not jumping to the NSTs, MOTs, etc. just yet - but I'm very happy with what I've been able to get out of this build.
I certainly have a few kinks to work out, and I'm still a relative novice in the world of high voltage, so please feel free to join me on this adventure.   

Watch it GO at 120 FPS!

https://drive.google.com/open?id=1CkyjDH6enRkY398ecstW0iXVxJOWfRqc

*UPDATE*
See ongoing discussion for current status.

*ORIGINAL* SPECS below:

Power Supply:
    - 0-40vDC 10A Switching Power Supply (Line Filter added)
    - ZVS Driver with center-tapped Primary (5+5)
    - CRT Flyback Transformer/Ignition Coil (rated output unknown; estimated 15-20kV, 5-6mA)
Input Power:
    - 28-36vDC
Secondary Coil:
    - 2.4" Diameter
    - 16" Winding Height
    - 28 AWG
    - Ratio: ~6.7:1
Primary Coil:
    - .25" copper tubing (20 ft.)
    - Minor Diameter: 6.4"
    - Major Diameter: 10.46" (at tap)
    - Turns at Tap: ~4.7
MMC:
    - 20x 0.33uF, 1200vDC (series) = 16.5pF, 24kVdc
Spark Gap:
    - Two Electrodes, ~.5" diameter (small, drawer knobs), Spaced ~.2"
Topload:
    - 1.5"x10" ring with top and bottom plate, covered with Aluminum Tape
    - Effective Capacitance: ~9pF
    - Breakout added
Arc Length:
    - ~12-15" (great looking arcs!)

I'm considering encasing the flyback transformer in oil or other medium to help with heat dissipation at higher input voltages.

I certainly welcome, and would love, any feedback, constructive criticisms, or ideas anyone may have to make this build even better!
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: davekni on November 07, 2019, 05:31:43 AM
Using a ZVS oscillator into a flyback makes me think you have enough electronic experience to transition to IGBT drive (DRSSTC) whenever you decide to.

Do you have any spec's on the flyback transformer, or know what input voltage and frequency it was designed for?  There will be a limit to getting more power by cooling.  As the flyback ferrite core gets close to saturation, it's losses will go up faster than the output power (efficiency drops).  If you can get multiple identical flyback transformers, paralleling them is an option.

Is your MMC made from the Chinese induction-cooker capacitors?  The 1200V 0.33uF rating makes that seem likely.  I'm having good success so far with those.
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: jturnerkc on November 07, 2019, 05:33:13 PM
Using a ZVS oscillator into a flyback makes me think you have enough electronic experience to transition to IGBT drive (DRSSTC) whenever you decide to.

Do you have any spec's on the flyback transformer, or know what input voltage and frequency it was designed for?  There will be a limit to getting more power by cooling.  As the flyback ferrite core gets close to saturation, it's losses will go up faster than the output power (efficiency drops).  If you can get multiple identical flyback transformers, paralleling them is an option.

Is your MMC made from the Chinese induction-cooker capacitors?  The 1200V 0.33uF rating makes that seem likely.  I'm having good success so far with those.

Thank you for your interest and support, Dave!

The DRSSTC certainly seems like a daunting task compared to the relatively basic construction of the SGTC. I do think solid state is next on the list, but I think I'll need to study up a bit more on the solid state implementation before i make that jump.

Finding specs on the FBT was a bit more difficult than I initially imagined, considering it's an old, discontinued, chinese model, but I believe I've determined it was designed for a horizontal frequency of 15.625kHz, which would imply it was used in a TV monitor.

The flyback cooling may or may not be necessary. It was a consideration based on an observation made after a short run. I had a bit of hot glue just securing the flyback from freely moving around - a temporary measure that quickly proved useless once the flyback warmed up ::) - and would prefer a 'cleaner' means of mounting and to use some decent terminals for tidier connection between the other components. I considered a small housing, simply because there doesn't seem to be a good mounting position for the flyback. I imagined filling the housing with oil and suspending the flyback could keep it cooler, if only by a bit.
If the heating gets any more significant it will be worth looking into further.

You're spot on with the MMC - it is comprised of the induction-cooker type capacitors and I'm also having good success with them. They are certainly a bargain considering what they can put up with.
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: davekni on November 08, 2019, 04:11:42 AM
15.625kHz is the analog PAL-standard TV horizontal frequency, used in Europe.  That's 64us, minus the 12us PAL blanking time, so 52us active time.  Presuming the 110-130V is the DC voltage of the flyback driver circuit, then the transformer input can handle 130V * 52us = 6.76mVs from peak negative to peak positive current.  (Half that time-voltage area if starting from 0 current.)  If you scope the transformer input in your ZVS circuit, hopefully the voltage-time area under one half cycle is less than 6.76mVs.

At 36Vdc in to your ZVS, peak voltage should be PI times that, or 113.1V.  For a sine wave, the area under a half cycle will be 2/PI times the peak voltage times the half-cycle time, so 72V * half_cycle_time, or 36V * period or 36V / frequency.  So, all you really need to measure is the frequency of your ZVS oscillator.  (I'm assuming your ZVS circuit is a standard Royer oscillator topology.)

Concerning transformer temperature, many power ferrite materials are designed for minimum loss at 100C.  I don't know if such materials are common in old flyback transformers.  I'd not be too concerned about it running "hot".

One final note on using TV flyback transformers with symmetric waveforms such as sine waves from a ZVS.  For a given output voltage, the internal diodes will see almost twice the reverse voltage.  In normal flyback use, the transformer output has high positive pulses and comparatively low negative pulses, reducing the diode's peak reverse-bias voltage requirements.  So, keep the output voltage (spark-gap breakdown voltage) not too far above half of the flyback transformer's output rating.
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: jturnerkc on November 08, 2019, 09:07:05 PM
15.625kHz is the analog PAL-standard TV horizontal frequency, used in Europe.  That's 64us, minus the 12us PAL blanking time, so 52us active time.  Presuming the 110-130V is the DC voltage of the flyback driver circuit, then the transformer input can handle 130V * 52us = 6.76mVs from peak negative to peak positive current.  (Half that time-voltage area if starting from 0 current.)  If you scope the transformer input in your ZVS circuit, hopefully the voltage-time area under one half cycle is less than 6.76mVs.

At 36Vdc in to your ZVS, peak voltage should be PI times that, or 113.1V.  For a sine wave, the area under a half cycle will be 2/PI times the peak voltage times the half-cycle time, so 72V * half_cycle_time, or 36V * period or 36V / frequency.  So, all you really need to measure is the frequency of your ZVS oscillator.  (I'm assuming your ZVS circuit is a standard Royer oscillator topology.)

Concerning transformer temperature, many power ferrite materials are designed for minimum loss at 100C.  I don't know if such materials are common in old flyback transformers.  I'd not be too concerned about it running "hot".

One final note on using TV flyback transformers with symmetric waveforms such as sine waves from a ZVS.  For a given output voltage, the internal diodes will see almost twice the reverse voltage.  In normal flyback use, the transformer output has high positive pulses and comparatively low negative pulses, reducing the diode's peak reverse-bias voltage requirements.  So, keep the output voltage (spark-gap breakdown voltage) not too far above half of the flyback transformer's output rating.

That is some great info and some specifics that will certainly help me out! I think I'll definitely look into grabbing another flyback to run in parallel, as you suggested earlier.
Sadly, during a run last night, i started smelling the undeniable scent of burning electronics. I noticed the high voltage pin on the flyback was actually arcing to the neighboring pin. This probably explains why it was getting so hot! I'm guessing it was doing that last time and I just didn't notice.  I never cut the other pins off, as I didn't think it would turn out to be an issue at the time. Everything is a learning experience, I suppose.
Of note: Feedback. Enough to receive a small shock from the power supply chassis. 
I was hoping that there was no internal damage to the FBT, but before I could run some further tests...my power supply seems to have stopped working. I'm getting zero DC voltage. I disconnected the supply and I can't see any burnt components or bloated caps, nothing. I'm pretty sure i was nowhere near 360W on the DC side when it was running. I'm going to look a bit closer tonight, remove the board, and see if i can figure out what's happened, but it seems my project is on hold until i can get the PS working again. I do have some batteries laying around that I could at least test the flyback with in the meantime. Hopefully the ZVS wasn't affected by whatever happened.

I'll update as I uncover more.

Thanks again, David!
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: davekni on November 09, 2019, 04:32:38 AM
Most TV flyback transformers have the primary high-voltage output (positive rectified output of ~20-30kV) on a read wire lead, not on any of the "pins".  One of the pins is the return (negative) side of the primary HV output.  I'd expect the red HV lead and the return pin to be wired to your spark-gap.

TV flyback transformers are a bit more complex that just the above.  They usually have some resistive dividers from intermediate taps to make adjustable 1-3kV outputs for other CRT electrodes (focus if I recall correctly).  The resistive dividers are intended to be returned to ground (to the primary HV output return pin).  I've had issues with those resistive-divider return pins arcing to other pins if left unconnected.  Usually doesn't cause internal damage, but does causing charring.

Drawing sparks to your DC supply is likely related to it frying.  If your Tesla coil secondary return (bottom end) isn't grounded, but instead has some electrical path back to the DC power, the return current will arc across components within the supply, likely causing failure.

Your system has three separate sections.  I'd recommend that each has a ground connection, to your power line safety ground.  The three are:
1) 24-36Vdc and circuitry powered by it (ZVS and the input side of the flyback transformer).  Negative DC lead is typically grounded.
2) Flyback output, MMC, spark gap, and Tesla primary coil.  Again, negative lead (flyback return pin) is typically grounded.
3) Tesla secondary coil.  This is the most important point to ground, as top-load capacitance is mostly to ground.
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: jturnerkc on November 09, 2019, 07:55:36 PM
Most TV flyback transformers have the primary high-voltage output (positive rectified output of ~20-30kV) on a read wire lead, not on any of the "pins".  One of the pins is the return (negative) side of the primary HV output.  I'd expect the red HV lead and the return pin to be wired to your spark-gap.

TV flyback transformers are a bit more complex that just the above.  They usually have some resistive dividers from intermediate taps to make adjustable 1-3kV outputs for other CRT electrodes (focus if I recall correctly).  The resistive dividers are intended to be returned to ground (to the primary HV output return pin).  I've had issues with those resistive-divider return pins arcing to other pins if left unconnected.  Usually doesn't cause internal damage, but does causing charring.

Drawing sparks to your DC supply is likely related to it frying.  If your Tesla coil secondary return (bottom end) isn't grounded, but instead has some electrical path back to the DC power, the return current will arc across components within the supply, likely causing failure.

Your system has three separate sections.  I'd recommend that each has a ground connection, to your power line safety ground.  The three are:
1) 24-36Vdc and circuitry powered by it (ZVS and the input side of the flyback transformer).  Negative DC lead is typically grounded.
2) Flyback output, MMC, spark gap, and Tesla primary coil.  Again, negative lead (flyback return pin) is typically grounded.
3) Tesla secondary coil.  This is the most important point to ground, as top-load capacitance is mostly to ground.

I misspoke and did intend to refer to the return "pin".
I do have the secondary coil of the TC grounded with a makeshift counterpoise at the moment, which I'm now speculating is not sufficient. I did not have the return pin of the Flyback grounded, it simply went straight into the spark gap. I can certainly understand why this circuit should be grounded. The grounding of this circuit is a source of confusion for me. Should it be isolated from the both of the other ground points?
When you refer to the negative DC lead being grounded, would the power supply not serve that purpose via the mains ground or are we talking about physically grounding the negative DC lead separate from the mains?
When you recommend each section have a "ground connection to [my] power line safety ground", are you suggesting that these can all share a common ground? I did not think this was the case.

After looking into the power supply further, I've determined that the transistors are definitely pooched, as well as the base resistors and associated diodes. I have a couple other random resistor failures. I've yet to test the IC, but everything around it was fine, with the exception of two resistors, so i do suspect it is something i should check as well. I may save the power supply repair for another time and just grab a cheap one for now, so I can get back to this project.
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: davekni on November 10, 2019, 06:18:16 AM
A counterpoise is helpful, especially for high frequencies caused by an arc.  I recommended connecting the counterpoise/secondary bottom to earth ground as well, via power wiring or water pipes.  The counterpoise alone may have several times the capacitance of the topload, but that can leave it with several percent of the top-load voltage, plenty enough to cause problems.

The flyback secondary return pin (negative side, since the red output wire is the positive side) could theoretically remain floating (along with the spark gap, MMC, and Tesla primary coil).  However, I recommend grounding it for two reasons.  Key one is the flyback's internal resistive divider and taps for focus etc. electrodes.  Other reason is due to stray capacitance from MMC/Tesla primary needing a return path.

Concerning the DC supply, it may or may not have a grounded output internally.  Adjustable lab supplies rarely have internal ground connection to the DC output.  This allows the user to ground the + or - or neither side (for say wiring two supplies in series).  Fixed supplies such as for laptop computers are more likely to be internally grounded, typically on the negative output lead.  Some aren't grounded, however, even when the power cord has a ground wire.  The power ground is used just for internal shielding between the supplies internal input and output side.

So, yes, I'm suggesting grounding all three sections to the power-line ground, which might as well be at the same point from a single power cord.  There's no short-circuit caused by this.  All three sections are electrically-isolated, coupled only magnetically.  First-to-second coupling is magnetic through the flyback.  Second to third is the Tesla primary-to-secondary magnetic coupling.  Grounding all three provides the single electrical connection to return any current due to stray capacitance.  A short-circuit would be created only if you ground two different nodes within one section, such as both the + and - outputs of the DC supply.

Good luck with getting it going again!
Title: Re: SGTC MK1 - Proud of my Accomplishment
Post by: jturnerkc on November 10, 2019, 10:58:28 PM
A counterpoise is helpful, especially for high frequencies caused by an arc.  I recommended connecting the counterpoise/secondary bottom to earth ground as well, via power wiring or water pipes.  The counterpoise alone may have several times the capacitance of the topload, but that can leave it with several percent of the top-load voltage, plenty enough to cause problems.

The flyback secondary return pin (negative side, since the red output wire is the positive side) could theoretically remain floating (along with the spark gap, MMC, and Tesla primary coil).  However, I recommend grounding it for two reasons.  Key one is the flyback's internal resistive divider and taps for focus etc. electrodes.  Other reason is due to stray capacitance from MMC/Tesla primary needing a return path.

Concerning the DC supply, it may or may not have a grounded output internally.  Adjustable lab supplies rarely have internal ground connection to the DC output.  This allows the user to ground the + or - or neither side (for say wiring two supplies in series).  Fixed supplies such as for laptop computers are more likely to be internally grounded, typically on the negative output lead.  Some aren't grounded, however, even when the power cord has a ground wire.  The power ground is used just for internal shielding between the supplies internal input and output side.

So, yes, I'm suggesting grounding all three sections to the power-line ground, which might as well be at the same point from a single power cord.  There's no short-circuit caused by this.  All three sections are electrically-isolated, coupled only magnetically.  First-to-second coupling is magnetic through the flyback.  Second to third is the Tesla primary-to-secondary magnetic coupling.  Grounding all three provides the single electrical connection to return any current due to stray capacitance.  A short-circuit would be created only if you ground two different nodes within one section, such as both the + and - outputs of the DC supply.

Good luck with getting it going again!

I was under the impression that grounding back to mains was a last resort, if anything, with Tesla Coils. No doubt, you know much more than I do about the subject. I had read, from a number of TC information sources, that there's the risk of affecting other electronics in the home, etc. and that high frequency through the water pipes, etc. can be undesirable as well. My understanding was the the best ground was an actual, physical earth ground, separate from the mains ground, at least for the Tesla Secondary.
Drawing a simple schematic, it sounds like you're saying I can just ground everything back to the mains outlet that I'm using to power the supply. Am I misunderstanding something? I realize that the sections are only magnetically coupled, but if a common ground is used, would that not then create a case where they are no longer electrically isolated?
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 11, 2019, 01:24:16 AM
Personally, I've always grounded to power mains, besides using aluminum window-screen on the floor (or on grass when outside).  However, my first Tesla coil was only 6 years ago.  Others here have a much longer history.  Perhaps I'm just lucky to have success with line grounding.  (I've never used line-grounding alone.  Always had some aluminum screen/foil/sheets on the floor too.  Perhaps using the line alone would be problematic.)

Yes, grounding all three sections makes them no-longer electrically isolated.  But it doesn't create any short-circuits.  Only one node of each section is grounded.  What it does provide is a return-path for currents caused by stray capacitance out to the universe.  Otherwise the return currents can arc across the transformers that separate the sections, or through other control circuitry within the power supply.

You could try line-grounding only the first two sections, and keeping your counterpoise separate for Tesla secondary return.  The risk there is if you do get a strike to your Tesla primary, the current spike to the line ground will be higher than if your counterpoise were included in the line grounding to absorb some of the current spike.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 11, 2019, 04:51:57 PM
Personally, I've always grounded to power mains, besides using aluminum window-screen on the floor (or on grass when outside).  However, my first Tesla coil was only 6 years ago.  Others here have a much longer history.  Perhaps I'm just lucky to have success with line grounding.  (I've never used line-grounding alone.  Always had some aluminum screen/foil/sheets on the floor too.  Perhaps using the line alone would be problematic.)

Yes, grounding all three sections makes them no-longer electrically isolated.  But it doesn't create any short-circuits.  Only one node of each section is grounded.  What it does provide is a return-path for currents caused by stray capacitance out to the universe.  Otherwise the return currents can arc across the transformers that separate the sections, or through other control circuitry within the power supply.

You could try line-grounding only the first two sections, and keeping your counterpoise separate for Tesla secondary return.  The risk there is if you do get a strike to your Tesla primary, the current spike to the line ground will be higher than if your counterpoise were included in the line grounding to absorb some of the current spike.

I have a new power supply getting dropped off today, so I plan to test the ZVS and flyback out. Hopefully the ZVS is ok, and hopefully I'll be able to go for another light tonight.
Just so I understand correctly - in your setup, you were grounding all sections back to mains - would that imply that i can simply ground the DC negative and the flyback return, straight back to the AC mains ground terminal on the power supply? Would it be easier to ground everything to the counterpoise, and just ground the counterpoise to mains?
I'm not particularly worried about a strike to the primary, at least on this build, but I am still questioning whether or not to ground the counterpoise of the secondary separately, to a nearby pipe or something, just so it's not feeding directly back into the same outlet, and may eventually put a proper earth ground in somewhere.
I'll defer to your experience on this one.  ;)
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 12, 2019, 05:25:20 AM
Either option should work.  Grounding the counterpoise separately to a pipe or whatever should be fine, as would tying in with the other connections to the power supply's ground terminal.

BTW, if the power supply has a separate ground terminal on the front, that's a good indication that the DC output is not grounded internally.  So, the DC output does need to be grounded, the negative output to ground by convention.  The fllyback output return (negative output pin) could be connected locally to the DC negative or with a separate wire back to the supply ground.  Locally is probably easier.  Then ground the counterpoise either to a pipe or back to the supply ground or both.

I usually have foil or screen against the floor or ground that extends from my coil back to the power supply, where it ties to the power line ground.  Is your counterpoise something like that?  I usually have more foil/screen extending past the coil and out to either side.  My scope input cables (coax) route against the foil, either just above or just below it.

Have fun!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 12, 2019, 02:49:05 PM
Either option should work.  Grounding the counterpoise separately to a pipe or whatever should be fine, as would tying in with the other connections to the power supply's ground terminal.

BTW, if the power supply has a separate ground terminal on the front, that's a good indication that the DC output is not grounded internally.  So, the DC output does need to be grounded, the negative output to ground by convention.  The flyback output return (negative output pin) could be connected locally to the DC negative or with a separate wire back to the supply ground.  Locally is probably easier.  Then ground the counterpoise either to a pipe or back to the supply ground or both.

I usually have foil or screen against the floor or ground that extends from my coil back to the power supply, where it ties to the power line ground.  Is your counterpoise something like that?  I usually have more foil/screen extending past the coil and out to either side.  My scope input cables (coax) route against the foil, either just above or just below it.

Have fun!

Wired up the new power supply last night. Started at 24v, but was not able to pull an arc off the HV return pin. I can hear the telltale hum of the flyback, but barely a spark when trying to pull an arc. I checked the ZVS and the FETs seem to meter fine, no shorts, and all resistors and diodes test fine.
After a closer inspection, I noticed a very slight bulge in the flyback housing. Sure enough, there's a few small cracks that formed as a result and I can only assume this flyback is done for. It must have been subjected to a beating long enough to experience an internal failure.
Luckily, I actually came across a pair of matching flybacks and should have those in a couple days.

For the counterpoise, i was using 4 12x12" sheets of 3/32" steel plates, attached together in a large square with aluminum tape (so it can be folded up), and placed on the garage floor underneath the secondary.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: profdc9 on November 12, 2019, 08:24:38 PM
You seem to have solved many problems I had when trying to do this.

One problem I had when trying to use a TV flyback transformer is that the peak voltage can be very high and damage the capacitors.  The open circuit voltage of some flyback transformers can be 30-40 kV!  The other problem is that I destroyed many transformers because when the arc is quenched, it causes a brief burst of high voltage, MHz-frequency RF in the transformer that over the course of a few minutes arced through and destroyed the internal insulation of the flyback transformer.   Then I tried to construct my own beefier  transformer on some huge U-shaped ferrite pieces, and those worked better, but even those were destroyed after a short run. 
I also had problems and could not use a static spark gap, because even with a strong fan blowing on the spark gap it was too hot.  So eventually I had to make a rotary spark gap out of an angle grinder, which is very difficult because it spins very fast and has to be balanced perfectly or will shake itself and everything else to pieces.

Eventually I went to dual MOTs and used a Terry filter which removes the RF spikes, and it was reliable.  This is one reason why DRSSTC can be so much easier, because there is no spark gap to cause voltage spikes, and especially good rotary spark gaps are hard to reconstruct.

Anyways that is a great build and very impressive you got it to work that well.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 12, 2019, 09:20:34 PM
You seem to have solved many problems I had when trying to do this.

One problem I had when trying to use a TV flyback transformer is that the peak voltage can be very high and damage the capacitors.  The open circuit voltage of some flyback transformers can be 30-40 kV!  The other problem is that I destroyed many transformers because when the arc is quenched, it causes a brief burst of high voltage, MHz-frequency RF in the transformer that over the course of a few minutes arced through and destroyed the internal insulation of the flyback transformer.   Then I tried to construct my own beefier  transformer on some huge U-shaped ferrite pieces, and those worked better, but even those were destroyed after a short run. 
I also had problems and could not use a static spark gap, because even with a strong fan blowing on the spark gap it was too hot.  So eventually I had to make a rotary spark gap out of an angle grinder, which is very difficult because it spins very fast and has to be balanced perfectly or will shake itself and everything else to pieces.

Eventually I went to dual MOTs and used a Terry filter which removes the RF spikes, and it was reliable.  This is one reason why DRSSTC can be so much easier, because there is no spark gap to cause voltage spikes, and especially good rotary spark gaps are hard to reconstruct.

Anyways that is a great build and very impressive you got it to work that well.

Thank you!
Did you happen to have your tank circuit grounded as David suggested above?
Could a low-pass filter be designed and added to snub the higher frequency slapping the flyback?
I struggled with this and transients during another unrelated project, but the same solution wouldn't work here.
I'm guessing the old vintage flybacks would hold up better since they don't have any other components like the "newer" flybacks. Might be worth a shot on top of the rest, but I am curious to see what kind of performance I'd get out of simply grounding the HV return pin on the flyback. Hmm...
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 13, 2019, 06:01:47 AM
I'm presuming you mean low-pass-filter (not high) between the flyback and spark gap.  Low-pass allows the "slow" MMC charging, but blocks the very-fast voltage drop of the spark-gap firing. 

profdc9 does bring up a good point that I'd missed.  The very-sudden voltage drop of the spark-gap firing may be what fried your flyback.  A high-voltage-capable inductor in series with the flyback output should help.  (That's probably the most critical part of any low-pass filter for this use.  A low-pass-filter with capacitors too will have resonances, which will ring to below ground after the spark.)

Inductors that can handle the sudden voltage of the spark-gap firing aren't trivial to find or build.  I learned that the hard way with my inductor-coupled Marx generator.  The charging inductors get a sudden 48kV spike when the Marx spark gaps fire.  My initial home-wound coils arc'ed across winding sections.  Changed to automobile spark coil secondaries, but they lasted only a few minutes.  In both cases, I think that uneven stray capacitance caused the voltage to distribute unevenly across the coil, causing insulation failure.  I finally ended up with strings of tiny commercial 1mH inductors in series, 133 inductors for one large inductor.  (Needed 24 total, so lots of tedious construction.)

A high-voltage-capable resistor in series with the flyback output would also help, but efficiency will suffer.

Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 13, 2019, 03:57:53 PM
I'm presuming you mean low-pass-filter (not high) between the flyback and spark gap.  Low-pass allows the "slow" MMC charging, but blocks the very-fast voltage drop of the spark-gap firing. 

profdc9 does bring up a good point that I'd missed.  The very-sudden voltage drop of the spark-gap firing may be what fried your flyback.  A high-voltage-capable inductor in series with the flyback output should help.  (That's probably the most critical part of any low-pass filter for this use.  A low-pass-filter with capacitors too will have resonances, which will ring to below ground after the spark.)

Inductors that can handle the sudden voltage of the spark-gap firing aren't trivial to find or build.  I learned that the hard way with my inductor-coupled Marx generator.  The charging inductors get a sudden 48kV spike when the Marx spark gaps fire.  My initial home-wound coils arc'ed across winding sections.  Changed to automobile spark coil secondaries, but they lasted only a few minutes.  In both cases, I think that uneven stray capacitance caused the voltage to distribute unevenly across the coil, causing insulation failure.  I finally ended up with strings of tiny commercial 1mH inductors in series, 133 inductors for one large inductor.  (Needed 24 total, so lots of tedious construction.)

A high-voltage-capable resistor in series with the flyback output would also help, but efficiency will suffer.

Correct. I meant low-pass, to attenuate the high frequency. Bit of dyslexia there, I guess.
I definitely noticed, even with the available higher voltage inductors I've found, that I'd probably be looking at a couple hundred bucks just to put something like that together (just as a rough guess).
What calculations did you use to determine the inductor specifications you required? I'm not sure I have the patience to wire up 3000 little baby inductors but perhaps I'll come across something in the future that would fit the bill.
I was about to ask about using a flyback secondary, but if the auto ignition coils failed, I'm betting a flyback wouldn't fair any better, and I certainly would prefer to not destroy another one just yet.

Would grounding the the return pin serve any purpose in reducing these spikes?
What about a simple grounded safety gap in parallel with the flyback and main spark gap, similar to those used with NSTs? I did come across a design where the individual was actually just using a simple horn gap, like a mini jacob's ladder as some sort of safety, but seems like it was more-so intended to protect from the effects of a primary strike, and not the voltage cause by the spark gap collapse. I'm not sure how that could be sufficient enough, but wouldn't it be great if the fix were that easy?
Perhaps I can look into hv resistors, for now, and sacrifice some efficiency to save my flybacks.

Needless to say, certain aspects of this project are pushing the limits of my electrical understanding, but that was technically the intent in the first place.
I really appreciate the information, detailed explanations, and patience!
There's really not any in depth information to be found on SGTC's using ZVS/Flyback combo anywhere, that I've found. Considering the preliminary results I was able to obtain with my current build, I'd love to hammer this out and be able to provide a proper, thorough, reference.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 14, 2019, 05:25:57 AM
Grounding the flyback HV return pin might help slightly.  At least it should prevent any internal arcing from the flyback secondary back to primary.

A string of small inductors isn't all that hard or expensive.  I needed 3000+ because I needed 24 inductors each capable of 48kV.  You need only one inductor at perhaps 30kV or whatever your spark-gap firing voltage is.  100 inductors is probably fine.  Each inductor sees only 300V, which shouldn't be enough to break down even the thin magnet wire insulation.  200 inductors would give lots of margin.  Here's links to a couple Digikey pages for inductor parts that should work, roughly $20 and $30 for 100 parts:
https://www.digikey.com/product-detail/en/taiyo-yuden/LHL08TB153J/587-5891-1-ND/7675011
https://www.digikey.com/product-detail/en/bourns-inc/RLB1014-104KL/RLB1014-104KL-ND/2561370
Checking more distributors might turn up a lower price for these or other similar parts.  I just searched for the lowest cost and highest inductance that could handle 40mA.  Do you know what your flyback output current is?  I was just taking a guess that it wouldn't be above 40mA.

The key detail to making the inductor string is to have uniformly distributed stray capacitance.  That keeps the voltage evenly distributed.  I made a soldering fixture of a string of 6.35mm magnets.  Cut the inductor leads to ~6mm, then lined up a row on each side of the magnet string, staggered so the leads touched in a series configuration.  Then it was easy to run down the string bonding the touching lead pairs with solder.  I made 19-long inductor strings this way, 10 on one side and 9 on the other side.  Each string went into 1" heat-shrink tubing.  The resulting insulated strings were layered (stacked).  I'd share pictures, but my fixture is buried in my storage shed at the moment.  If you want to go this route, I'll get it out and add further description. 
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 15, 2019, 03:59:48 PM
Grounding the flyback HV return pin might help slightly.  At least it should prevent any internal arcing from the flyback secondary back to primary.

A string of small inductors isn't all that hard or expensive.  I needed 3000+ because I needed 24 inductors each capable of 48kV.  You need only one inductor at perhaps 30kV or whatever your spark-gap firing voltage is.  100 inductors is probably fine.  Each inductor sees only 300V, which shouldn't be enough to break down even the thin magnet wire insulation.  200 inductors would give lots of margin.  Here's links to a couple Digikey pages for inductor parts that should work, roughly $20 and $30 for 100 parts:
https://www.digikey.com/product-detail/en/taiyo-yuden/LHL08TB153J/587-5891-1-ND/7675011
https://www.digikey.com/product-detail/en/bourns-inc/RLB1014-104KL/RLB1014-104KL-ND/2561370
Checking more distributors might turn up a lower price for these or other similar parts.  I just searched for the lowest cost and highest inductance that could handle 40mA.  Do you know what your flyback output current is?  I was just taking a guess that it wouldn't be above 40mA.

The key detail to making the inductor string is to have uniformly distributed stray capacitance.  That keeps the voltage evenly distributed.  I made a soldering fixture of a string of 6.35mm magnets.  Cut the inductor leads to ~6mm, then lined up a row on each side of the magnet string, staggered so the leads touched in a series configuration.  Then it was easy to run down the string bonding the touching lead pairs with solder.  I made 19-long inductor strings this way, 10 on one side and 9 on the other side.  Each string went into 1" heat-shrink tubing.  The resulting insulated strings were layered (stacked).  I'd share pictures, but my fixture is buried in my storage shed at the moment.  If you want to go this route, I'll get it out and add further description.

That does sound worth a try.
Does the inductors' own frequency matter here? Should DC resistance be lower to reduce energy storage and minimize any heat buildup, or higher to dissipate more power?
I feel like I already know the answer to this question, but is there any reason a perf board couldn't be used, lining inductors up in rows and soldering in series, instead of forming a long string? I can imagine a scenario where the proximity of all those magnetic fields would be undesirable, but would it particularly matter in this application?

Received a couple flybacks today, but eh... apparently wasn't paying close enough attention. They're a bit smaller than expected (about half the size of my original flyback! I had to laugh at myself a bit...). Is there a practical way to run the primaries in parallel and secondaries in series? I'll try to spec these out as much as I can, but I'm thinking these might be a bit too small.
I'd sure love to find some datasheets on these things... They are identified as MFB-10/332A or E144238. The equivalent is the HR 2287 T17, but I'm struggling to find any detailed info on either.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 16, 2019, 04:13:35 AM
Placing two flyback transformer secondaries in series will almost-certainly fry one of them.  Flybacks are generally designed to have the HV negative (return) terminal somewhere near the same potential as the primary.  In series, one of the secondary negative pins will be many kV away from it's primary.  Internal insulation on the negative side isn't designed to handle that voltage.

Paralleling both the primaries and secondaries should work, and is probably closer to what you need.  The output voltage is likely enough from one flyback.  Using two in parallel will double the output current, so charge your MMC in half the time.

Yes, perf-board will work fine.  The string method is faster for the large quantities I needed.  Yes, the inductors do interact magnetically with adjacent ones.  Alternating the orientation (180 degree rotation) for radial-lead inductors as I used increases inductance (makes magnetic loops).  That's what I wanted.  It does lower saturation current, however.  Making a string of radial-lead inductors all in the same orientation lowers inductance, but increases saturation current.  I'd recommend making the entire string either the same or alternating, rather than mixing the two options.  If the inductors are spaced out a few mm, then direction won't matter much.  The inductors I used have a small white dot mark on top to show orientation.

For the inductors I suggested previously, resistance isn't enough to make much difference.  For example, the 150-ohm ones, 100 in series will have 15k ohms.  If charging at 30mA from the flyback(s), that's 450V drop, not too much compared to ~30kV.

To avoid excess voltage across any given inductor during the spark discharge, stray capacitance within the inductor string should be reasonably uniform.  I'd suggest a physical layout on your proto-board matching this schematic:
 [ Invalid Attachment ]

In other words, don't wire a zig-zag.  Make series-connected rows of inductors, wiring the right edge of each row to the left edge of the next row.  That way the electric field will be roughly-uniform from top-to-bottom.  Leave a few mm between rows, whether or not you decide to space out inductors within each row.  Mount the proto-board inductor string in a plastic case or otherwise spaced away from metal.

Good luck with your fun project!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 18, 2019, 12:07:38 AM
Placing two flyback transformer secondaries in series will almost-certainly fry one of them.  Flybacks are generally designed to have the HV negative (return) terminal somewhere near the same potential as the primary.  In series, one of the secondary negative pins will be many kV away from it's primary.  Internal insulation on the negative side isn't designed to handle that voltage.

Paralleling both the primaries and secondaries should work, and is probably closer to what you need.  The output voltage is likely enough from one flyback.  Using two in parallel will double the output current, so charge your MMC in half the time.

Yes, perf-board will work fine.  The string method is faster for the large quantities I needed.  Yes, the inductors do interact magnetically with adjacent ones.  Alternating the orientation (180 degree rotation) for radial-lead inductors as I used increases inductance (makes magnetic loops).  That's what I wanted.  It does lower saturation current, however.  Making a string of radial-lead inductors all in the same orientation lowers inductance, but increases saturation current.  I'd recommend making the entire string either the same or alternating, rather than mixing the two options.  If the inductors are spaced out a few mm, then direction won't matter much.  The inductors I used have a small white dot mark on top to show orientation.

For the inductors I suggested previously, resistance isn't enough to make much difference.  For example, the 150-ohm ones, 100 in series will have 15k ohms.  If charging at 30mA from the flyback(s), that's 450V drop, not too much compared to ~30kV.

To avoid excess voltage across any given inductor during the spark discharge, stray capacitance within the inductor string should be reasonably uniform.  I'd suggest a physical layout on your proto-board matching this schematic:

In other words, don't wire a zig-zag.  Make series-connected rows of inductors, wiring the right edge of each row to the left edge of the next row.  That way the electric field will be roughly-uniform from top-to-bottom.  Leave a few mm between rows, whether or not you decide to space out inductors within each row.  Mount the proto-board inductor string in a plastic case or otherwise spaced away from metal.

Good luck with your fun project!

Right. I definitely understand running two flybacks in series would normally be an issue. Trying to run the two primaries in parallel, and then the secondaries in series would almost certainly result in a failure.
I was considering a possible circuit that I thought might work. Let me know what you think. Essentially there's a bare (extracted flyback) core transformer driven by the ZVS, that then oscillates the two flybacks, instead of running parallel primaries.

These flybacks I just picked up are angry little guys. Series might be way overkill - so I'll stick with running a single flyback for now. I will need to keep an eye on temperature though, considering the size of these FBTs.
I think I'll try out your inductor suggestion and see how that goes as well. Thanks for the clarification on that!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 18, 2019, 01:59:24 AM
Your proposed series connection could work, but it requires the upper secondary winding of your custom transformer to handle 15~30kV relative to the lower secondary and primaries, which are essentially at ground potential (compared to 30kV).  15~30kV is whatever voltage a single flyback output can generate.  If you can construct a custom transformer with such HV insulation, I'd suggest making your own flyback large enough to drive your project directly with one stage.  If you do try your series design, I'd add a connection from the negative (return) secondary pin of the upper flyback to it's primary.  This provides a path for any leakage current from you custom input transformer, preventing it from damaging the upper flyback.

Starting with a single flyback sounds like a good idea.  If you get enough voltage to trigger your spark gap, then series isn't needed.  Add a second in parallel if you want to double the firing frequency.  BTW, do you know roughly what voltage your spark gap is set for?

Small fans are quite cheap these days.  I'd suggest placing one blowing directly on the flyback(s).  Of course, there's still higher temperature inside than on the surface, but a fan will significantly improve the power capability.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 18, 2019, 03:30:02 AM
Your proposed series connection could work, but it requires the upper secondary winding of your custom transformer to handle 15~30kV relative to the lower secondary and primaries, which are essentially at ground potential (compared to 30kV).  15~30kV is whatever voltage a single flyback output can generate.  If you can construct a custom transformer with such HV insulation, I'd suggest making your own flyback large enough to drive your project directly with one stage.  If you do try your series design, I'd add a connection from the negative (return) secondary pin of the upper flyback to it's primary.  This provides a path for any leakage current from you custom input transformer, preventing it from damaging the upper flyback.

Starting with a single flyback sounds like a good idea.  If you get enough voltage to trigger your spark gap, then series isn't needed.  Add a second in parallel if you want to double the firing frequency.  BTW, do you know roughly what voltage your spark gap is set for?

Small fans are quite cheap these days.  I'd suggest placing one blowing directly on the flyback(s).  Of course, there's still higher temperature inside than on the surface, but a fan will significantly improve the power capability.

With the smaller flyback, since i have not yet established its actual voltage output, I have the spark gap set to fire at ~9kV (rudimentary measurement).
I did a short test run and achieved a respectable output. I'm running on a bit lower power (about 25V to the ZVS), until I'm satisfied it can handle a bit more. I hesitate to push this baby flyback too hard on its own before adding some additional protection.
On another note, I've finally finished up the new topload. I've also had an additional secondary I wrapped a while back and have been waiting to run (different dimensions, but near the same resonance) so will be swapping those out.

I'll have some new specs up shortly.
I suppose this one will have to officially be...
(https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcRsDsqsiNsi2-nE2er0QKge42dck_9nLZbIJ5EZD6Vjw48A8LkwUw&s)
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: Mads Barnkob on November 18, 2019, 12:21:53 PM
Your proposed series connection could work, but it requires the upper secondary winding of your custom transformer to handle 15~30kV relative to the lower secondary and primaries, which are essentially at ground potential (compared to 30kV).  15~30kV is whatever voltage a single flyback output can generate.  If you can construct a custom transformer with such HV insulation, I'd suggest making your own flyback large enough to drive your project directly with one stage.  If you do try your series design, I'd add a connection from the negative (return) secondary pin of the upper flyback to it's primary.  This provides a path for any leakage current from you custom input transformer, preventing it from damaging the upper flyback.

Reversing the polarity on one of the primaries and turn one flyback transformer around, mid-point ground them and you will not surpass the voltage rating of the single transformer. But if it makes it necessary to redo the whole grounding scheme.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 19, 2019, 12:49:52 AM
Your proposed series connection could work, but it requires the upper secondary winding of your custom transformer to handle 15~30kV relative to the lower secondary and primaries, which are essentially at ground potential (compared to 30kV).  15~30kV is whatever voltage a single flyback output can generate.  If you can construct a custom transformer with such HV insulation, I'd suggest making your own flyback large enough to drive your project directly with one stage.  If you do try your series design, I'd add a connection from the negative (return) secondary pin of the upper flyback to it's primary.  This provides a path for any leakage current from you custom input transformer, preventing it from damaging the upper flyback.

Reversing the polarity on one of the primaries and turn one flyback transformer around, mid-point ground them and you will not surpass the voltage rating of the single transformer. But if it makes it necessary to redo the whole grounding scheme.

Ok, that's genius and clearly works, at least in simulation. David does mention a point I had not considered in my original sim, however...the internal diodes. With a couple AC flybacks it looks like you'd get double the voltage across the spark gap, while each flyback is only putting out half of the voltage the spark gap sees! I'll definitely keep this in mind for the future.

I've swapped out the secondary of the SGTC and added my newly constructed topload. I should have some updates and specs on Mk. II in a few days!

Thank you both for all your insight and suggestions so far!!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 19, 2019, 05:47:34 AM
Are you using standard TV flyback transformers as I've been assuming?  Those almost always have internal diodes, so output DC, positive on the HV lead and negative on the return pin.  If you were using bare AC flyback transformers, then you would have HV diode(s) between the secondaries and spark gap.

The normal DC flyback transforms wouldn't work back-to-back in series.  Both ends would be positive, so no voltage between them.  Your series schematic doesn't show diodes either internal or external.  Is that what you are simulating - no diodes?  If the simulation values (inductances, capacitances, etc.) are reasonable, you won't be getting much voltage on the MMC, since it can charge for only one half-cycle of the flyback frequency.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 19, 2019, 06:25:49 AM
Are you using standard TV flyback transformers as I've been assuming?  Those almost always have internal diodes, so output DC, positive on the HV lead and negative on the return pin.  If you were using bare AC flyback transformers, then you would have HV diode(s) between the secondaries and spark gap.

The normal DC flyback transforms wouldn't work back-to-back in series.  Both ends would be positive, so no voltage between them.  Your series schematic doesn't show diodes either internal or external.  Is that what you are simulating - no diodes?  If the simulation values (inductances, capacitances, etc.) are reasonable, you won't be getting much voltage on the MMC, since it can charge for only one half-cycle of the flyback frequency.

That's a good point, and a good observation. I did not use diodes in the sim.
Maybe Mads can chime in.
Although I don't have the exact specs on these small flybacks i acquired, i think it's safe to assume there are diodes employed. I've attached some shots. Like I said, they are pretty small. I've not found much data on these in my searches. I certainly don't plan on pushing them much or in series. I don't think these little babies could handle much of anything like that. I'm only using one for the time being.
I may consider winding my own. It actually sounds like a fun challenge. A relatively easy build that I'm sure would be pretty satisfying after completion. 
Right now, I'm working on finishing up the new TC secondary and grounding scheme, as well as a fancy new stand.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 20, 2019, 05:30:25 AM
Thank you for the images.  I'd be quite surprised if they don't have internal diodes.  I can't see how your SGTC would work at all if they didn't include diodes.

Should have been obvious to me, but I hadn't realized that you were winding your own primary.  Flyback transformers don't typically have center-tapped primary windings, and you mentioned center-tapped primary in your initial post.  Custom primary winding has a likely-key advantage besides center-tap:  The coupling factor will be lower than with the built-in primary winding.  As I discuss some in my ZVS Jacob's ladder post:
    https://highvoltageforum.net/index.php?topic=831.0
Having a coupling factor below 0.86 allows the ZVS oscillator to run over the full range of output loading from short to open.  That's close to what you have as the MMC charges from 0V to the spark-gap trigger voltage:  shorted load to some lower load current (higher voltage).  It would be interesting to measure the coupling coefficient of your flyback.  However, that's not necessary unless you have trouble with the oscillations dropping out, which would cause run-away input current to the ZVS oscillator.

What frequency is the ZVS running with your new small flyback transformer?  What is the cross-sectional area of the flyback's ferrite core?  Knowing those two values would allow estimating the ZVS input voltage permitted before saturating the core.  I'll use 50kHz and 50mm^2 area for an example, and a presumed saturation flux density of 0.4T.  Period = 20us, divided by 2PI = 3.183us/radian.  50mm^2 * 0.4T = 20uVs/turn, for 200uVs for  your 10-turn primary.  200uVs / 3.183us = 62.83V peak.  ZVS peak voltage is ideally (no losses) PI * DC_input_voltage.  So, input voltage would be 62.83V / PI = 20Vdc.  If the core could handle 0.5T (the highest I've seen for any fferrite material), then Vdc could be up to 25V.  Of course, this is just for my example guess for frequency and core cross-sectional area.

If you want to optimize power from the small flyback, it can help to separately measure temperatures of the primary winding, ferrite core, and secondary winding.  That indicates where to improve.  Hot primary might be improved with litz-wire.  With your single-layer primary, litz is of less advantage than for multi-layer windings.  So, I'd guess that won't be the hottest.  If the core is hottest, it could be due to flux saturation and/or high frequency losses.  A hot secondary is likely due to the internal diodes, either just total current or switching losses if the frequency is high.

Looking forward to hearing how your fun project progresses!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: Mads Barnkob on November 20, 2019, 07:16:24 PM
They would have to be put in series, positive|negative - positive|negative for doubling the output voltage, whether or not you will destroy one from imbalance or not... its not exactly practical to add balancing resistors :)

Here is one that did that, on a single ZVS driver
/>
From my own experience, you have to push them real hard before they start to have corona on the outside enclosure and flash-over is not far after that, at the end of this video (input voltage was around 140VDC)

Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 21, 2019, 01:11:49 AM

They would have to be put in series, positive|negative - positive|negative for doubling the output voltage, whether or not you will destroy one from imbalance or not... its not exactly practical to add balancing resistors :)

Here is one that did that, on a single ZVS driver
[video]

From my own experience, you have to push them real hard before they start to have corona on the outside enclosure and flash-over is not far after that, at the end of this video (input voltage was around 140VDC)
[video]

Right. That first video was effectively what my sim was set up to recreate. As David pointed out, however, I did not take internal diodes into account.

I'll skip the series idea for now. Even though these are small flybacks, they seem to be packing quite a punch. However I’m quite confident one or both would be destroyed running in series.


Thank you for the images.  I'd be quite surprised if they don't have internal diodes.  I can't see how your SGTC would work at all if they didn't include diodes.

Flyback transformers don't typically have center-tapped primary windings, and you mentioned center-tapped primary in your initial post.  Custom primary winding has a likely-key advantage besides center-tap:  The coupling factor will be lower than with the built-in primary winding.

Having a coupling factor below 0.86 allows the ZVS oscillator to run over the full range of output loading from short to open.  That's close to what you have as the MMC charges from 0V to the spark-gap trigger voltage:  shorted load to some lower load current (higher voltage).  It would be interesting to measure the coupling coefficient of your flyback.  However, that's not necessary unless you have trouble with the oscillations dropping out, which would cause run-away input current to the ZVS oscillator.

What frequency is the ZVS running with your new small flyback transformer?  What is the cross-sectional area of the flyback's ferrite core?  Knowing those two values would allow estimating the ZVS input voltage permitted before saturating the core.  I'll use 50kHz and 50mm^2 area for an example, and a presumed saturation flux density of 0.4T.  Period = 20us, divided by 2PI = 3.183us/radian.  50mm^2 * 0.4T = 20uVs/turn, for 200uVs for  your 10-turn primary.  200uVs / 3.183us = 62.83V peak.  ZVS peak voltage is ideally (no losses) PI * DC_input_voltage.  So, input voltage would be 62.83V / PI = 20Vdc.  If the core could handle 0.5T (the highest I've seen for any ferrite material), then Vdc could be up to 25V.  Of course, this is just for my example guess for frequency and core cross-sectional area.

If you want to optimize power from the small flyback, it can help to separately measure temperatures of the primary winding, ferrite core, and secondary winding.  That indicates where to improve.  Hot primary might be improved with litz-wire.  With your single-layer primary, litz is of less advantage than for multi-layer windings.  So, I'd guess that won't be the hottest.  If the core is hottest, it could be due to flux saturation and/or high frequency losses.  A hot secondary is likely due to the internal diodes, either just total current or switching losses if the frequency is high.

I will take some temperature readings at different run times during the next light.
Is there a relatively simple method to measure the DC bias?

Could bias be removed or reduced using a tertiary winding?

Could modifying the ZVS frequency/duty cycle of the primary be used to minimize bias and risk of saturation?
Just wondering this makes me want to modify the ZVS for variable frequency, but to make that possible on a resonant circuit i'd have to find the perfect variable inductors and/or capacitors.

I ran a little Jacob's ladder to grab some frequency readings. Seems my ZVS is operating around 70kHz.
After a few test runs, I can't tell what's hotter - the primary, the core, or the secondary. They're all quite warm to the touch after about 1 minute, give or take. My ZVS mosfets and inductor were getting quite warm as well. I was running this setup with both the DC negative and HV return pin connected to mains ground. I'm not sure if it's related, but after grounding both, my amp draw seemed to  decrease.
I will try to get some of the other measurements you suggested, asap. I'm curious to get to the bottom of all this.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 21, 2019, 05:44:40 AM
I hadn't intended to bring up DC bias, but that is a good point.  The half-wave rectification of the internal flyback diodes will create some DC magnetic field in the core.  Core saturation will show up as distortion to the sine-wave shape of the ZVS output (flyback primary) leads.  Saturation will show up as steeper edges and flatter top compared to a true sine wave.  If one ZVS output has steeper edges than the other, that's the result of the DC field component caused by the output diodes.  Compensation could be done by moving the "center" tap a bit off of center.  Then the DC from the ZVS supply input inductor will flow through more turns one direction than the other, so add/subtract DC bias.  DC bias compensation may not be necessary depending on how close you happen to be to saturation.  I have a ZVS simulation with full-wave rectification on the output.  I'll change it to half-wave to get some feeling for the bias issue.  (Saturation isn't simulated, so it won't be perfect.)

Frequency is determined by the flyback primary inductance and your ZVS resonant capacitor value (and the secondary winding capacitance adds a bit too).  Changing frequency involves just changing ZVS capacitor value.  What value do you have now for your 70kHz result?  Did you get a chance to measure core cross-section (ferrite thickness and width inside your primary winding) - at least a guess?

Do you have any information on the average DC output current?  For example, if you know the spark repeat frequency and the spark-gap voltage (9kV last you mentioned), then the average current can be found by 16.5nF * 9kV * spark_frequency, where 16.5nF is your MMC capacitance.

You may not be far off from optimum already.  If the flyback is staying below 100C, it may be fine as is.  (I'd still suggest adding the output inductor string to protect from nano-second fall times.)  In other words, analysis and optimization is fun, but you may prefer to go with what is working and not drag out the details.  This is a fun project, but don't let me take it too far down the analytical path if that's not your interest.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 22, 2019, 01:38:57 AM
I hadn't intended to bring up DC bias, but that is a good point.  The half-wave rectification of the internal flyback diodes will create some DC magnetic field in the core.  Core saturation will show up as distortion to the sine-wave shape of the ZVS output (flyback primary) leads.  Saturation will show up as steeper edges and flatter top compared to a true sine wave.  If one ZVS output has steeper edges than the other, that's the result of the DC field component caused by the output diodes.  Compensation could be done by moving the "center" tap a bit off of center.  Then the DC from the ZVS supply input inductor will flow through more turns one direction than the other, so add/subtract DC bias.  DC bias compensation may not be necessary depending on how close you happen to be to saturation.  I have a ZVS simulation with full-wave rectification on the output.  I'll change it to half-wave to get some feeling for the bias issue.  (Saturation isn't simulated, so it won't be perfect.)

Frequency is determined by the flyback primary inductance and your ZVS resonant capacitor value (and the secondary winding capacitance adds a bit too).  Changing frequency involves just changing ZVS capacitor value.  What value do you have now for your 70kHz result?  Did you get a chance to measure core cross-section (ferrite thickness and width inside your primary winding) - at least a guess?

Do you have any information on the average DC output current?  For example, if you know the spark repeat frequency and the spark-gap voltage (9kV last you mentioned), then the average current can be found by 16.5nF * 9kV * spark_frequency, where 16.5nF is your MMC capacitance.

You may not be far off from optimum already.  If the flyback is staying below 100C, it may be fine as is.  (I'd still suggest adding the output inductor string to protect from nano-second fall times.)  In other words, analysis and optimization is fun, but you may prefer to go with what is working and not drag out the details.  This is a fun project, but don't let me take it too far down the analytical path if that's not your interest.

Can you share the schematic you're using for your sim? I'd like to replicate, in mine, what you're looking at.

For the ZVS, I actually just used the same capacitors that I used in my MMC. There are two .33uF 1200vdc; one between each FET drain and ground.
The small flyback cores measure ~9.7mm in diameter, ~49.4mm in height and ~35mm wide. Cute little thing, isn't it?

I do not yet know, for sure, what the output current is on the secondary. I can estimate the spark gap voltage at ~9kV, maybe 10, but I can’t confirm that with any certainty. I'd like to put a high voltage probe together and at least see what the voltage output is. At least I'd have some info to work with... Same with spark frequency. I can guess and throw some numbers into a calculator, but they may or may not be close to accurate. If I do get any more flybacks, I’m definitely going to make sure I can find the datasheet first!

I ran the same makeshift jacob's ladder with the flybacks in parallel tonight and it was running around 90kHz.
I did not take temperature readings with them in parallel, but with a single flyback running (zvs ~70kHz) I took readings after at least 1 minute of run time and got these readings:
Secondary/Housing: ~44°C
Primary: ~33°C
Core: ~46°C
I should have done a few different runs at different time intervals, but ran out of time tonight.

I certainly expect those temperatures to change quite a bit when actually running the tesla coil. I should get that done over the weekend.

I really do appreciate you taking the time to walk me through these things, especially those I had not considered, and help me improve this circuit. I couldn't ask for better insight and I do enjoy the analytical aspect and trying to optimize performance. What was working also resulted in destruction of my original flyback and I would prefer a more reliable circuit that I can run more than 1 minute with relative certainty that I’m not going to fry another one, or something else, during a demonstration. I'm even more determined now that I've destroyed a flyback and rendered a power supply useless (for the time being).
The complexity and number of components involved in building an SSTC or DRSSTC don't seem to be very practical for me at this time, so I am determined to optimize this current setup (or something like it). I've seen some assembled SSTC and DRSSTC drivers on eBay. I'd certainly be curious to try solid state as it seems much more stable, but hesitate to buy anything pre-assembled from overseas. It would be fun to "play with", but I'd, of course, prefer to build my own.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 22, 2019, 06:21:22 AM
Those temperatures seem quite reasonable.  The only issue would be if the secondary power dissipation was concentrated mostly in the diodes, which could make their internal die temperature much hotter than secondary surface temperature.  That's why I was asking about secondary current.  A wild guess for that flyback is ~5mA average (DC) secondary rated current.  You could probably push it a bit higher, especially since the sine-wave drive has the diodes conducting with a higher duty cycle than a normal flyback waveform.

Here's the ZVS I'm using to simulate something like your setup.  The voltages and currents are different, as this is derived from my Jacob's ladder ZVS.  The input ramps from 0 to 160V (most of what it sees from rectified 120V line power).  The secondary is 1:1 to the primary.  That's just for convenience in comparing input and output waveforms on the same simulation plot.  It's easy to scale the output manually.  It also uses a separate high-coupling-factor center-tapped coil to feed DC to the oscillator.  That's left over from some of my old ZVS induction circuits - works roughly the same as center-tapping the working coil (L3 in this case).  You can modify it to be like your circuit.
 [ Invalid Attachment ]

I wound a flyback from my stash (11 x 13.5mm rectangular core area) with 10T and made some measurements.  Coupling factor was 81%, and the secondary has somewhere around 1600 turns.  Your flyback has 74mm^2 core area, so can handle a bit more volt*seconds than my example calculation using 50mm^2.  I think you've built a very nice close-to-optimum system for these flybacks.

A resistor in series with the negative ground-return side of the secondary would allow measuring current.  1000 ohms would have 5V drop at 5mA.  The waveform will be higher pulses with zero between, which you could measure with a scope.  If using a meter to measure the voltage-drop, I'd suggest adding a 0.1~1uF capacitor across the resistor so the high-frequency pulses don't confuse the meter.  If using a scope, you can also see the spark frequency, as the current will drop as the MMC charges, then suddenly rise again after each spark.

In simulation I played with feeding the ZVS from only one side (removed L1 and L2, made L4 200uH and wired it to either VP or VN).  Moving the DC feed all the way to one side was about right to compensate for the DC component of output current.  It compensated well at the medium-to-low current conditions (higher output voltage).  At low output voltage and high output current, even that wasn't fully compensating for output current.  I'd initially guessed the "center"-tap would need to be moved only 1 or 2 turns from center.  Turns out that all the way to one end is good.  Feeding to VP should be compensating, adding current to the end of L3 that matches the positive end of L5 where current is leaving.  I'm 99% sure that's what worked in simulation, but was too distracted today to write it down.  Feeding to the wrong side doubles the total DC current bias to the core.  To check for yourself, wire it each way and plot the sum of L3 and L5 current to see the net DC component.  (Another reason it's easier to simulate with 1:1 turns ratio.)

Yes, if/when you get to DRSSTC, I'd build your own.  Much more fun and more educational too.  You're off to a great start!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 23, 2019, 05:01:46 AM
Thinking a bit more about filtering between flyback output and spark-gap:  A simple R/C filter is probably sufficient, avoiding the cost/work of soldering 100 inductors.  In this case, the R/C filter is removing high-frequencies generated by the spark gap, so would have the R on the spark-gap side:
 [ Invalid Attachment ]

Something like the above filter will waste a few % of input power, but should protect the flyback from fast spark-gap falling edges as well or better than the inductor string.  The resistor is constructed from 10 in series to avoid the cost/trouble of finding a single 20-30kV capable resistor.  Even with 10 resistors, each will see 2-3kV.  Resistors rated for a few kV aren't too hard to find, but any 1-2W rated resistor will likely be fine.  The voltage is there for a very short time.  I used 2W 10k resistors, six per string, for my small Marx generator (9kV per string).  That's only 1.5kV/resistor, but I did some testing to much higher voltage on those 2W parts.

Values for the above low-pass filter aren't critical.  Take some care in finding a capacitor, however.  Many cheep ceramic 20-30kV capacitors have bad capacitance vs. voltage curves, dropping down by 80-90% at rated voltage (only 10-20% of capacitance remaining).

One other caution, which I think you're already doing.  Make sure there's some load (MMC or arc gap) on the secondary when the ZVS starts up.  ZVS oscillators often produce a startup-burst that is ~2x the normal run voltage.  Energy builds up in the input inductor until oscillation starts.  Than that energy transfers to output oscillation until used up.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: MRMILSTAR on November 23, 2019, 04:15:31 PM
Thinking a bit more about filtering between flyback output and spark-gap:  A simple R/C filter is probably sufficient, avoiding the cost/work of soldering 100 inductors.  In this case, the R/C filter is removing high-frequencies generated by the spark gap, so would have the R on the spark-gap side:


Something like the above filter will waste a few % of input power, but should protect the flyback from fast spark-gap falling edges as well or better than the inductor string.  The resistor is constructed from 10 in series to avoid the cost/trouble of finding a single 20-30kV capable resistor.  Even with 10 resistors, each will see 2-3kV.  Resistors rated for a few kV aren't too hard to find, but any 1-2W rated resistor will likely be fine.  The voltage is there for a very short time.  I used 2W 10k resistors, six per string, for my small Marx generator (9kV per string).  That's only 1.5kV/resistor, but I did some testing to much higher voltage on those 2W parts.

Values for the above low-pass filter aren't critical.  Take some care in finding a capacitor, however.  Many cheep ceramic 20-30kV capacitors have bad capacitance vs. voltage curves, dropping down by 80-90% at rated voltage (only 10-20% of capacitance remaining).

One other caution, which I think you're already doing.  Make sure there's some load (MMC or arc gap) on the secondary when the ZVS starts up.  ZVS oscillators often produce a startup-burst that is ~2x the normal run voltage.  Energy builds up in the input inductor until oscillation starts.  Than that energy transfers to output oscillation until used up.

Looks like the well-known "Terry filter" used by many spark gap coilers for their NST-based tesla coils.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 23, 2019, 06:34:56 PM
Oh!  I'd seen the phrase "Terry filter" on this forum, but had no idea what it was.  Thank you for the insight.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 25, 2019, 09:20:41 PM
Thinking a bit more about filtering between flyback output and spark-gap:  A simple R/C filter is probably sufficient, avoiding the cost/work of soldering 100 inductors.  In this case, the R/C filter is removing high-frequencies generated by the spark gap, so would have the R on the spark-gap side:

Something like the above filter will waste a few % of input power, but should protect the flyback from fast spark-gap falling edges as well or better than the inductor string.  The resistor is constructed from 10 in series to avoid the cost/trouble of finding a single 20-30kV capable resistor.  Even with 10 resistors, each will see 2-3kV.  Resistors rated for a few kV aren't too hard to find, but any 1-2W rated resistor will likely be fine.  The voltage is there for a very short time.  I used 2W 10k resistors, six per string, for my small Marx generator (9kV per string).  That's only 1.5kV/resistor, but I did some testing to much higher voltage on those 2W parts.

Values for the above low-pass filter aren't critical.  Take some care in finding a capacitor, however.  Many cheep ceramic 20-30kV capacitors have bad capacitance vs. voltage curves, dropping down by 80-90% at rated voltage (only 10-20% of capacitance remaining).

One other caution, which I think you're already doing.  Make sure there's some load (MMC or arc gap) on the secondary when the ZVS starts up.  ZVS oscillators often produce a startup-burst that is ~2x the normal run voltage.  Energy builds up in the input inductor until oscillation starts.  Than that energy transfers to output oscillation until used up.

I think I'll go ahead and try out the inductor string first since I recently acquired 160 inductors, which I’ll probably try to get wired up tonight. I'll keep this in mind, though.

Meanwhile, during some testing with the small flybacks in parallel, once again I spotted the return pin of one of them arching to a neighboring pin (even after I had already cut all other pins, added extra insulation around the return pin, and sealed everything up with hot glue. I think I'll use epoxy or something next time.) Needless to say I don't particularly feel like digging glue out at the moment. The other flyback does not have the same issue.  Unfortunately, one doesn't seem powerful enough to charge my current capacitor configuration (or rather, my capacitor bank is too large for the flyback) and is not firing consistently.
I’m convinced these little turds aren't going to produce the results I'm looking for. They’re great for a Jacob’s ladder though.

I’ve already sourced two larger ones locally, but will take me a couple days before I can extract them. We’ll try this again with a properly sized transformer like I started with. I miss my thick, angry, streamers!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 26, 2019, 05:54:16 AM
If it's not firing consistently, then it sounds like insufficient voltage rather than insufficient current.  The latter would cause it to be slow (low repeat frequency), but not inconsistency.  At risk of frying the little flybacks, a bit higher ZVS input voltage would help.

The pin arcing may be caused by a feedback resistor within the flyback.  Several of my flyback transformers include a high-value resistor from the positive HV output lead back to a pin adjacent the negative HV return pin.  For example, the one I've been experimenting with recently has 264meg ohms from output lead to pin 7.  Pin 6 is the HV return.  I don't know if the 264meg resistor was for feedback to regulate voltage, or part of a voltage-divider to generate focus voltage (~2kV).  Either way, if your flyback has such an internal resistor, it's better to tie the associated pin to HV return.  (I have 20meg from pin 7 to pin 6 to monitor HV voltage.)

If that HV feedback pin is left floating, it may arc to the HV return or not.  The problem is that it might arc inside the flyback rather than at the pin.  Internal arcing will damage insulation and perhaps start a more extensive internal failure.  That's why I'd suggest tying it to the HV return, directly or through a resistor.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 26, 2019, 09:33:32 PM
If it's not firing consistently, then it sounds like insufficient voltage rather than insufficient current.  The latter would cause it to be slow (low repeat frequency), but not inconsistency.  At risk of frying the little flybacks, a bit higher ZVS input voltage would help.

The pin arcing may be caused by a feedback resistor within the flyback.  Several of my flyback transformers include a high-value resistor from the positive HV output lead back to a pin adjacent the negative HV return pin.  For example, the one I've been experimenting with recently has 264meg ohms from output lead to pin 7.  Pin 6 is the HV return.  I don't know if the 264meg resistor was for feedback to regulate voltage, or part of a voltage-divider to generate focus voltage (~2kV).  Either way, if your flyback has such an internal resistor, it's better to tie the associated pin to HV return.  (I have 20meg from pin 7 to pin 6 to monitor HV voltage.)

If that HV feedback pin is left floating, it may arc to the HV return or not.  The problem is that it might arc inside the flyback rather than at the pin.  Internal arcing will damage insulation and perhaps start a more extensive internal failure.  That's why I'd suggest tying it to the HV return, directly or through a resistor.

That's an interesting observation. I checked all the pins when I received the units and didn't have any resistance readings that stuck out like that.

When using the larger flyback, I was able to get consistent charge and firing at only 12V, and eventually kicked it up to about 30V which resulted in those really nice streamers pictured in my original post. I ran this small one at 24V, but didn't want to push it much further at the time.

When first receiving the flybacks and simply pulling an arc off the HV return, the small one seems more interested in making long, fire-y, rising, arcs instead of the shorter, more concentrated, arcs output from the larger flyback. It's an interesting contrast. Would the difference in arc behavior be attributed to voltage or current? Frequency?
Perhaps I can record a short video to illustrate more clearly what I'm speaking about and post all the specs I can come up with.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 27, 2019, 05:58:09 AM
My guess about insufficient voltage is that the MMC is charging to barely the spark-firing voltage.  When the voltage is marginal, a spark gap can fire or not depending on just how a corona streamer forms - what dust particles happen to be in the air, UV photons that happen to ionize some air, etc.  If the issue was limited current, the MMC would reach spark-gap firing voltage, just more slowly.

Another possibility occurred to me:  The flyback secondary winding has enough internal capacitance to resonate, and that is coupled to your primary ZVS resonant circuit.  That makes two resonant frequencies, one where the two winding voltages are in-phase and one where they are 180 degrees out-of-phase.  (The two resonant frequencies are discussed in some of the Tesla coil discussions, as the Tesla coil primary and secondary are two coupled resonant circuits.)  Perhaps the ZVS is occasionally locking into the higher-frequency out-of-phase mode.  I saw that occasionally a couple days ago in a ZVS-driven flyback experiment of my own - when the output was loaded more heavily.  The flyback is inefficient in that mode.

Concerning the nature of arcs, I'm just learning with my recent Jacob's ladder project.  My only previous experience was with spark-gap sudden discharges, not with continuous arcs.  Perhaps others here can assist.  If I had to guess, I'd say the fire-y arcs are higher current.  I doubt frequency matters as long as it's in the 10+kHz range.  The ionized air path definitely decays significantly in 1-2ms, but I don't think it decays much in <100us.

Many meter resistance ranges top-out at 20meg, so a 264meg resistor may have shown up as open.  An easy way to look for high resistances is with a DC voltage source and a volt meter.  Meters often have 10meg or 1meg input resistance on voltage ranges.  Apply a DC voltage to the HV output wire, then measure voltage on the pins.  If the DC voltage is negative and above ~30V, then the HV return pin can be found that way.  The internal HV diodes often have 20-30V forward drop, so at least that much voltage is required to see continuity from HV wire (positive output, which is the diode cathode) to the HV return pin.

If you want to head down the analytical path, I'd suggest measuring the flyback output turns.   There are a few ways to do so depending on what tools you have around.  Scope?  Probes good for a few hundred volts?  AC signal generator (some source of low voltage in the kHz range)?
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: Uspring on November 27, 2019, 02:03:02 PM
Quote
I doubt frequency matters as long as it's in the 10+kHz range.  The ionized air path definitely decays significantly in 1-2ms, but I don't think it decays much in <100us.

I tend to agree with this. Plasma is caused by heat and that takes some time to cool off. But even during longer pauses hot air remains, which is a lot thinner than at room temperature. That reduces breakdown voltages along the hot path, which causes an easy reignition there. One can make nicely working Jacobs ladders even at line frequency.

Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 28, 2019, 12:44:10 AM
My guess about insufficient voltage is that the MMC is charging to barely the spark-firing voltage.  When the voltage is marginal, a spark gap can fire or not depending on just how a corona streamer forms - what dust particles happen to be in the air, UV photons that happen to ionize some air, etc.  If the issue was limited current, the MMC would reach spark-gap firing voltage, just more slowly.

Another possibility occurred to me:  The flyback secondary winding has enough internal capacitance to resonate, and that is coupled to your primary ZVS resonant circuit.  That makes two resonant frequencies, one where the two winding voltages are in-phase and one where they are 180 degrees out-of-phase.  (The two resonant frequencies are discussed in some of the Tesla coil discussions, as the Tesla coil primary and secondary are two coupled resonant circuits.)  Perhaps the ZVS is occasionally locking into the higher-frequency out-of-phase mode.  I saw that occasionally a couple days ago in a ZVS-driven flyback experiment of my own - when the output was loaded more heavily.  The flyback is inefficient in that mode.

Concerning the nature of arcs, I'm just learning with my recent Jacob's ladder project.  My only previous experience was with spark-gap sudden discharges, not with continuous arcs.  Perhaps others here can assist.  If I had to guess, I'd say the fire-y arcs are higher current.  I doubt frequency matters as long as it's in the 10+kHz range.  The ionized air path definitely decays significantly in 1-2ms, but I don't think it decays much in <100us.

Many meter resistance ranges top-out at 20meg, so a 264meg resistor may have shown up as open.  An easy way to look for high resistances is with a DC voltage source and a volt meter.  Meters often have 10meg or 1meg input resistance on voltage ranges.  Apply a DC voltage to the HV output wire, then measure voltage on the pins.  If the DC voltage is negative and above ~30V, then the HV return pin can be found that way.  The internal HV diodes often have 20-30V forward drop, so at least that much voltage is required to see continuity from HV wire (positive output, which is the diode cathode) to the HV return pin.

If you want to head down the analytical path, I'd suggest measuring the flyback output turns.   There are a few ways to do so depending on what tools you have around.  Scope?  Probes good for a few hundred volts?  AC signal generator (some source of low voltage in the kHz range)?

That does make sense and I would agree that the issue would then appear to be voltage. I assumed 24V would be plenty to produce the energy needed. Without being able to measure the output, I couldn't be sure and didn't want to risk frying the little flyback.

It's very interesting that you mention two resonant frequencies. I'm not sure it's of note, but with a multimeter on either ZVS output, I was reading a relatively steady 70kHz. During a quick test the other night with a basic oscilloscope, I noticed the frequency would occasionally drop as low as 50kHz, but only for a very short time. Through my narrow perspective though, the waveforms appeared to be relatively as expected considering the crude Jacob's ladder I was using as a load, so I didn't think much of it. This may or may not be of any significance. We're getting into new territory for me now, and I'm learning more on every exchange. I need to widen my field of view when looking at these things and consider all variables.

I found it very curious that the two different flybacks had such drastically different arc formation. I might even speculate that my spark gap woes could be contributed to this. The arc formation of a typical Jacob’s ladder is exactly the kind of arcs this little flyback makes on its own and sustains them a considerable distance. It's quite impressive.

I experimented briefly with current limiting and was able to obtain a steady, controlled, arc across the spark gap (no capacitors connected). When I get back to it, I’ll try increasing voltage a bit while limiting available current and see what kind of results I get.

I did get ahold of another, larger, flyback today. Fortunately, I actually have the schematic for this one. (attached) I will do some comparisons and report the results.

Meter resistance range is also a good point. The one I'm using was actually a recent purchase and I can't recall the range off-hand. I'll have to double check. I hadn’t considered checking the voltage drop with an actual DC input, but that's a great idea.

I did put together a small signal generator that can output triangle, sine, and square wave up to 1MHz and can be finely adjusted. It can use any DC input. I usually just use a 9V battery.
I technically have a scope. It’s just a cheap, handheld, single-channel, bare-bones, oscilloscope that I purchased as a kit some time ago for very basic purposes. It gets the job done..mostly.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: MRMILSTAR on November 28, 2019, 05:40:21 AM
I have a Marx generator which is powered by an average-size flyback transformer driven by a ZVS driver. I use a 24 volt DC supply to power the ZVS driver. The flyback and ZVS driver have no problem with the 24 volts. Perhaps this will allay your concerns about the 24 volt power supply.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on November 28, 2019, 06:20:37 AM
Your small flyback transformers may be designed for a bit lower voltage and higher current.  As Steve suggests, it's probably OK to run the ZVS input voltage a bit higher.  The biggest issue may be the unconnected focus pin, if it decides to arc inside the potted windings.  Those small flybacks look new enough that they are likely from a color TV or monitor, which suggests they'd be good for at least 20kV output.

Measuring the turns ratio would help decide what input voltage is safe.  One way is to feed your signal generator sine wave to the flyback input through a resistor, and measure the flyback HV output with your meter.  Hopefully your signal generator can product enough power to get at least 100-200V on the HV output with only the meter load.  Set the signal generator for something in the 20-50kHz range, lower if you can get enough output voltage.  Measure the voltage across the flyback primary with the scope, to see what actual voltage is achieved from the signal generator and resistor.  Adjust the generator and/or resistor to get 100Vdc on the meter.  Record the flyback primary peak voltage from the scope (or P-P and divide by 2).  Adjust the resistor or generator for 200V, and record that input voltage.  The turns ratio is then (200V - 100V) / (peak_input_for_200V - peak_input_for_100V).  The output diode forward drop cancels in the calculation by using deltas (changes in voltages).  (This presumes a 1.0 coupling factor.  The real turns-ratio will be higher by 1/K.)

An alternative method:  Charge a capacitor (0.1 to 5uF) to 100V.  Then connect the capacitor to the flyback secondary, negative to the HV output wire and positive to the HV return pin.  Measure the flyback primary waveform with your scope.  Repeat with the capacitor charged to 200V.  For the primary peak voltage, use the initial fast edge voltage, not the following ring-down.  Calculation is the same as above, except that the actual turns-ratio will be lower by a factor of K.  (It's worth repeating each capacitor discharge multiple times, taking the scope reading from the trace with the cleanest waveform - initial fast step followed by ring-down without subsequent fast steps.  Mechanical touching of wires often makes mechanical bounce.  The good traces will avoid that noise, or at least have it well past the initial step.  Or, use a TRIAC, such as BTA8 or BTA16 or BTB16 or ... as the switch.)

K (coupling factor) is probably between 0.8 and 0.85 based on the couple flybacks I've measured.  If you want to measure K, perhaps the easiest is to run your ZVS at as low a voltage as it can handle.  Measure the frequency with the flyback secondary open (not arcing), then again with the secondary shorted.  If it's like what I see, the frequency change will be ~2.5:1.

Your new flyback has focus taken from an intermediate stage instead of the HV output.  My larger flybacks are that way too.  I'd still suggest grounding that pin (pin 7) to be safe.  Does the specification list anything about output voltage or current?  How about the design input Vcc voltage?

For comparing the flyback arc characteristics, measuring HV return current would be informative.  A series resistor, perhaps 100 ohms to ground.  Measure the voltage with your scope, or add a parallel capacitor and measure DC value with your meter.

Based on how well you are understanding all the information, it's hard to picture that you are just starting.  Impressive progress!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on November 30, 2019, 09:41:17 PM
I have a Marx generator which is powered by an average-size flyback transformer driven by a ZVS driver. I use a 24 volt DC supply to power the ZVS driver. The flyback and ZVS driver have no problem with the 24 volts. Perhaps this will allay your concerns about the 24 volt power supply.

Thanks MRMILSTAR. The capacitors are rated at 630 VAC or 1200 Volts DC. The voltage measured across the capacitors when the flyback is operating and producing a spark is approximately 50-60 Volts, but a safe drain to source rating for the FETs would be 200 Volts, so I'm not particularly concerned whether my ZVS can handle over 24V, I was questioning myself whether I should push the tiny flybacks any harder, especially with the amount of current being drawn (over 10A peaks when not being limited).
The larger flyback I recently got ahold of is only pulling ~5A at peak (when not limited), with the same 24V input.

Your small flyback transformers may be designed for a bit lower voltage and higher current. As Steve suggests, it's probably OK to run the ZVS input voltage a bit higher. The biggest issue may be the unconnected focus pin, if it decides to arc inside the potted windings.  Those small flybacks look new enough that they are likely from a color TV or monitor, which suggests they'd be good for at least 20kV output.

Your new flyback has focus taken from an intermediate stage instead of the HV output.  My larger flybacks are that way too.  I'd still suggest grounding that pin (pin 7) to be safe.  Does the specification list anything about output voltage or current?  How about the design input Vcc voltage?

For comparing the flyback arc characteristics, measuring HV return current would be informative.  A series resistor, perhaps 100 ohms to ground.  Measure the voltage with your scope, or add a parallel capacitor and measure DC value with your meter.

Based on how well you are understanding all the information, it's hard to picture that you are just starting.  Impressive progress!

Another interesting note about the larger flyback: My ZVS driver uses two 0.33 uF capacitors. Each capacitor is connected from a FET drain to ground. With a primary inductance of 22 uH resonating with 0.167 uF, the resulting frequency is approximately 80 kHz (if I'm doing the math right). Observed operational frequency is closer to 70kHz (which is pretty much spot on for what I was seeing with my small flybacks). However, with the larger flyback, this drops to an average of ~40kHz. Unfortunately, I can't compare this to the first large flyback because I didn't take frequency readings on it.

I ran it on the crude jacob's ladder I use just as a load. After about 10 minutes, maybe, temp reading on the secondary/housing was 60C. The primary and core were noticably cooler. ZVS fets were cool as well. This is quite different from the results when running the smaller flybacks, where EVERYTHING was hot.
Grounding pin 7 on the larger flyback, probably contributed to this difference, I can imagine.
Speaking of pin 7 - were you recommending to ground pin 7 and leave the HV return pin un-grounded, or tying the two pins together and then grounding?
From what I was able to gather, I believe this flyback is rated for 24.2kV, but I don't have an exact number on the ouput current. I came across some information in a repair manual that seems to contradict that though, and implies a 30.5kV "limited EHT rating". An uneducated guess would tell me I could safely assume around 150W?

It may or not be of note, but figured I'd mention that I'm using a bifilar winding for the primary this time (still 5+5). I did the same with my first flyback, but not the smaller ones. The difference may be negligible, but I thought I'd give it a try again, thinking it may help, to some extent, reduce the resonance imbalance cause by leakage inductance, and perhaps help stabilize frequency fluctuations I was seeing in the smaller flyback tests. Waveforms appear to be effectively the same, both when drawing and arc and not. Although the frequency is lower, I'm not seeing the fluctuation I observed with the small flyback, but all-in-all it appears more likely that the alternative winding method is not making any noticeable difference.

As far as measuring turns in the secondary, my signal generator definitely can't put out that kind of voltage (i put it together to be pocket size and was really just using it as a trigger), so I would have to use the capacitor method. I will definitely make a note of that info and perhaps try it in the future, but I think I'll set that aside for the time being. I do think you're right on with the .8-.85 coupling coefficient though.

I appreciate the support and encouragement. I'm not super new to electronics, but still consider myself a novice. This is, however, my first foray into high voltage. Before even attempting any of this, I spent months reading and simulating, etc. Even with all that, I think I'm picking up things here, with you, that I never found mentioned in my research. I'm a quick study, so I always appreciate a challenge, so I'm loving this. I can't thank you enough for your time and sharing your experience!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 01, 2019, 02:56:39 AM
Bifilar winding shouldn't matter much.  For the ZVS, the leakage inductance just adds to the inductance from positive DC input, so is a slight advantage.

I'd missed your original note about the two 0.33uF caps being from drain-to-ground.  Most ZVS circuits I've seen have a capacitor between the two drains, but not to ground.  Having capacitors to ground causes the FETs to carry the resonant current in addition to the DC input current, so increases their power dissipation.  Having capacitors to ground as you do may have advantages, though.  I was thinking it might help keep the oscillation to the lower resonant frequency of the two for the case of coupled resonant circuits.  Attempted simulation of that topology a few days ago, but LTSpice didn't converge well, and I didn't take time to play with the algorithm parameters (charge tolerance, ...).

That's likely why you're measuring 70kHz instead of 80kHz - the HV winding capacitance scales by turns^2 and adds to the primary capacitance in the lower-frequency mode.  (The flybacks I've been playing with drop even more due to secondary capacitance.)

BTW, due to my not reading carefully about caps to ground, I'd been presuming 0.66uF primary rather than 0.167uF.  So, that may have messed up some of my previous-post calculations.  However, with 0.33uF from each drain to ground, the effective resonant capacitance is actually 0.33uF, not 0.167uF.  That's because at any given point in time, one of the two 0.33uF caps is shorted by a FET.  Which cap toggles, but only one is active at a time.  The higher FET current is due to that cap shorting, carrying the resonant current from the other cap.  You can see this easily in simulation - measuring frequencies and FET currents in both topologies.

How did you determine 22uH?  Was that something I calculated (from bogus assumptions)?  The best way I know is to measure frequency with a much larger capacitor (ie 3-30uF) so the secondary capacitance is less significant.  (For best accuracy, measure with two different large capacitors and do the algebra to calculate inductance and parasitic capacitance - two measurements and two variables.  That's what I just did for a flyback last week - one I intentionally fried the internal diodes to get AC output.)  Ring-down is the best method with large capacitors.  Running the ZVS will result in low frequency and high current and therefore core saturation, which lowers inductance.

A larger flyback is likely to have a bit more inductance (for 10-turns), but probably not enough to explain 70kHz to 40kHz (about 3x inductance).  Inductance could actually be 3x higher, or perhaps the resonant mode is different.

The HV return pin definitely needs to remain grounded, along with pin 7.  Otherwise the HV return current passes through the high-value resistors of the pin 7 internal network.

150W seems reasonable, 5-6mA at 24-30kV.   With ZVS sine-wave input, the current can probably be a bit higher, and the voltage a bit lower.  (Lower voltage because sine-waves are symmetric, while flyback waveforms have much less reverse voltage.  Higher current because the diode forward-conduction will have higher duty cycle with sine-wave drive, so lower peak current.)

For measuring turns ratio, the signal generator doesn't need to make more than 1-2V peak.  The secondary turn count is likely at least 1000, so at least 100x your 10-turn primary.  However, it needs to get 1-2V peak into a fairly low impedance, which it may not be capable of doing.

For your new flyback for which you attached a schematic, pins 2 to 6 is shown as 90Vpp.  That would be a great place to measure with your scope.  Ground pin 2 and measure pin 6 or visa-versa.  With no HV load, measure pin 2-6 Vpp relative to ZVS input voltage.  That will determine what input DC voltage is safe to run.

Fun project!  Thank you for the detailed information.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 01, 2019, 07:32:44 PM
Bifilar winding shouldn't matter much.  For the ZVS, the leakage inductance just adds to the inductance from positive DC input, so is a slight advantage.

I'd missed your original note about the two 0.33uF caps being from drain-to-ground.  Most ZVS circuits I've seen have a capacitor between the two drains, but not to ground.  Having capacitors to ground causes the FETs to carry the resonant current in addition to the DC input current, so increases their power dissipation.  Having capacitors to ground as you do may have advantages, though.  I was thinking it might help keep the oscillation to the lower resonant frequency of the two for the case of coupled resonant circuits.  Attempted simulation of that topology a few days ago, but LTSpice didn't converge well, and I didn't take time to play with the algorithm parameters (charge tolerance, ...).

That's likely why you're measuring 70kHz instead of 80kHz - the HV winding capacitance scales by turns^2 and adds to the primary capacitance in the lower-frequency mode.  (The flybacks I've been playing with drop even more due to secondary capacitance.)

BTW, due to my not reading carefully about caps to ground, I'd been presuming 0.66uF primary rather than 0.167uF.  So, that may have messed up some of my previous-post calculations.  However, with 0.33uF from each drain to ground, the effective resonant capacitance is actually 0.33uF, not 0.167uF.  That's because at any given point in time, one of the two 0.33uF caps is shorted by a FET.  Which cap toggles, but only one is active at a time.  The higher FET current is due to that cap shorting, carrying the resonant current from the other cap.  You can see this easily in simulation - measuring frequencies and FET currents in both topologies.

Thank you for pointing out the power dissipation! That wasn't something I even looked at when simulating. The difference seemed negligible, operation-wise, but I hadn't made considerations for power dissipation. I don't particularly recall the reason for doing that other than some proposed method I may have read and probably didn't understand. I can't exactly remember my thought process there.
Your note regarding resonant capacitance - I should have known better on that one. That's certainly quite clear to see in simulation.
I think i did originally misspeak regarding my ZVS configuration because I was not using the same configuration for my sim. That's my fault for the confusion there and now I understand that may have contributed to some of my other confusions.
Quote
How did you determine 22uH?  Was that something I calculated (from bogus assumptions)?  The best way I know is to measure frequency with a much larger capacitor (ie 3-30uF) so the secondary capacitance is less significant.  (For best accuracy, measure with two different large capacitors and do the algebra to calculate inductance and parasitic capacitance - two measurements and two variables.  That's what I just did for a flyback last week - one I intentionally fried the internal diodes to get AC output.)  Ring-down is the best method with large capacitors.  Running the ZVS will result in low frequency and high current and therefore core saturation, which lowers inductance.

The 22uH was based on some rough guesses entered into an inductance calculator and also seemed to match up closely to results another individual observed using effectively the same ZVS/flyback configuration. I made some other assumptions that were incorrect, as you pointed out, so this may or may not be near accurate. It made sense at the time as the results were close to my observed output when using the smaller flybacks.

Quote
A larger flyback is likely to have a bit more inductance (for 10-turns), but probably not enough to explain 70kHz to 40kHz (about 3x inductance).  Inductance could actually be 3x higher, or perhaps the resonant mode is different.

Right, I was not expecting that significant of a drop! I have that second large flyback, now, that I'll throw on there today and see what kind of frequency and other readings I get in comparison.

Quote
The HV return pin definitely needs to remain grounded, along with pin 7.  Otherwise the HV return current passes through the high-value resistors of the pin 7 internal network.

150W seems reasonable, 5-6mA at 24-30kV. With ZVS sine-wave input, the current can probably be a bit higher, and the voltage a bit lower.  (Lower voltage because sine-waves are symmetric, while flyback waveforms have much less reverse voltage.  Higher current because the diode forward-conduction will have higher duty cycle with sine-wave drive, so lower peak current.)

For your new flyback for which you attached a schematic, pins 2 to 6 is shown as 90Vpp.  That would be a great place to measure with your scope.  Ground pin 2 and measure pin 6 or visa-versa.  With no HV load, measure pin 2-6 Vpp relative to ZVS input voltage.  That will determine what input DC voltage is safe to run.

Excellent tip! Thank you for that! That will certainly help to know that value.
I should have time today to continue working, so will update with new results asap.
Thank you, again!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 02, 2019, 04:09:37 AM
When measuring pin 2 to 6 Vpp, there's no need to derate the 90 Vpp rating.  The diode reverse voltage is the output peak-to-peak voltage.  (The output peak voltage will be lower because of sine-wave drive, with as much reverse voltage as forward voltage.)

The inductance for a flyback I wound with 10 turns is ~60uH.  It's probably physically a bit larger than your small ones.

Can you provide any more details about your comment: "I found it interesting that the voltage across the caps was quite low when tied to ground"?  The peak voltage across each cap should be the same grounded or drain-to-drain.  Peak-to-peak will be half, because each cap sees a one-sided waveform, not going more than a diode-drop below ground.

Having some of the resonant capacitance to ground could have an advantage when using IGBT parts for a ZVS oscillator.  It will make the IGBT current negative just before switching off.  I haven't explored the effects yet, but it might be useful for IGBTs with slow turn-off.

Looking forward to your new update!  Hopefully it's a successful day of experimenting.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 02, 2019, 05:49:05 AM
When measuring pin 2 to 6 Vpp, there's no need to derate the 90 Vpp rating.  The diode reverse voltage is the output peak-to-peak voltage.  (The output peak voltage will be lower because of sine-wave drive, with as much reverse voltage as forward voltage.)

The inductance for a flyback I wound with 10 turns is ~60uH.  It's probably physically a bit larger than your small ones.

Can you provide any more details about your comment: "I found it interesting that the voltage across the caps was quite low when tied to ground"?  The peak voltage across each cap should be the same grounded or drain-to-drain.  Peak-to-peak will be half, because each cap sees a one-sided waveform, not going more than a diode-drop below ground.

Having some of the resonant capacitance to ground could have an advantage when using IGBT parts for a ZVS oscillator.  It will make the IGBT current negative just before switching off.  I haven't explored the effects yet, but it might be useful for IGBTs with slow turn-off.

Looking forward to your new update!  Hopefully it's a successful day of experimenting.

I tried to go back and correct my voltage statement, but you got to it before I did.
I was only reading half of the waveform and out of ignorance was not figuring peak-to-peak for both caps and confused myself a bit. As you pointed out, one is always shorted, so I was only reading the Vpp of that particular cap.

I'm not well versed in LtSpice. It's a little too deep for me at this point, but I've managed to make good use of the Falstad simulator. I've attached the circuits for the two different ZVS topologies. I haven't entered my 'real-world' values as this was just a base model for operational comparison purposes. I also still don't have accurate values on my flyback primary/secondary. These can be a pain to simulate, I've noticed. A rough guess has yet to get me any further in mirroring my actual circuit, at least as closely as possible, especially when trying to including the rest of the TC circuit.
*Note: the 1u resistors were used just to prevent capacitor loop error in Falstad. I modeled it this way simply because I wanted to show 2 capacitors. The values I used for the transformer are: 83mH primary, Ratio is 1:2, and K is .99 (again, just a working model). For clarification on the screenshot - the top two waveforms are each of the capacitors, the third waveform down is one FET, and the 4th waveform is the same FET's power consumption. The left column is the typical ZVS topology, and the right column is how I currently have mine configured.)

I noticed the frequency increase as well as a slight increase in power consumption, as you mentioned. The FET waveform changes and the current drops slightly negative, as you pointed out, as well as the voltage. I'm curious about that brief positive spike just before the drop and subsequent ramp up. FET voltage remains relatively unchanged (~1V), but current increases quite a bit. The rms voltage of the capacitors effectively match the FETs, which makes sense (i think). Frequency and current seem to be the most noteworthy.
Looking back, I believe this topology is proposed to be used in higher frequency applications than is needed here...which I suppose makes sense, as no components need to be added/removed to achieve a higher frequency. Regardless of its intended application, it would appear, to my eyes anyways, that I should go the drain-to-drain route with the capacitors.

I think i did originally misspeak, and may have gone back and corrected my claim, regarding my ZVS configuration because I was not using the same configuration for my sim at the time. That's my fault for the mix-up there and now I understand that may have contributed to following confusions.

After flipping through some old notes, the drain-to-ground implementation seems likely to have spawned from notes I made on a conversation regarding FETs burning out in an H-Bridge motor driver application a while back.

I spent most of the time I had set aside today practicing my soldering skills putting together 120, 100mH, inductors, so I don't have much to update on just yet, other than that I am glad it was not more! I used some small boards I had laying around and put together a little two-tiered stack of 60 inductors each. I didn't have time to wire it up in the circuit, but it's ready to go.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 07, 2019, 09:48:21 PM

The small positive current spike is likely because the FET turns on slightly before its drain voltage reaches 0, so relatively-suddenly discharges its drain cap the last few volts.  BTW, the relative drain current between the two topologies depends on the relative resonant current to load current.  If you change the 1k load in simulation, it changes the load current, but not the resonant current.  A low-impedance resonant circuit would make the extra FET loss of the drain-to-ground topology worse.

Yes, the drain-to-ground topology smooths out the drain voltage during the transition points.  That may have value with slower switch devices (IGBTs).  In my bit of simulation, drain-to-ground caps made it more difficult to get the ZVS to oscillate at the lower frequency parallel mode of the input and output windings.  I'm not sure why.  I'll be experimenting with that for a DIY plasma ball eventually, but December is busy for me.

Coupling of 0.99 is high for your simulation transformer, but won't matter much until you add an output capacitor to simulate the flyback secondary's internal stray capacitance (wire and diodes).  With a realistic output capacitance and coupling (80-85%), you can experiment with the two resonant frequencies.


The parameters I used for the transformer are not representative of my actual application. I entered a higher inductance and just left the coupling parameter at default. In simulation, it made it easier to observe the waveforms, considering I can only increase the division so much. I found that just increasing the inductance would reduce the frequency and make those observations easier. This seemed like a simple enough solution at the time. Although it was not completely representative, it simply helped me visualize what is generally happening at different locations within the circuit, during different conditions.

The ZVS circuit can definitely be fickle in simulation.

A few updates:

The larger secondary and new topload that I was going to be using will be set aside for the time being, as I use the previous secondary and topload (since I already had that tuned for the most part). The calculated resonant frequency of the larger secondary (backed up by observations when testing), would indicate that I either need to increase the # of turns in my primary or double the capacitance of the MMC to reduce the primary resonant frequency. Simply adding another 20 caps in parallel would lower the frequency enough to come within ~2% of resonance (using JavaTC) I'm going to go ahead and wire up another 20 capacitors and set them aside until I decide what to do. That would be cheaper and less time consuming than purchasing more copper and re-'winding' the primary.
Decisions. Decisions.

In the meantime, I reduced the inductor string to 6H. Even after taking resistance readings, voltage drop, etc. the day before - the next day went to check everything again before connecting, and nothing. I ended up finding a single inductor reading as an open circuit. Odd, but I just need to swap that one out apparently. Each level of my 'inductor bank' had 60 inductors, so I removed that level and left myself with 60, so 6H.



During some simulation, however, I found that it may be beneficial to actually use 2 separate strings. When an inductor string is placed on both the HV output and return. It was important to ground the HV return and then attach the inductor between the ground point and the spark gap. Another observation I made, when simulating this configuration was that placing the MMC in parallel and the spark gap in series (swapping the components around) appeared to positively affect the circuit. I judge this from a certainly novice standpoint, but I see less noise and a more uniform operation when simulating in this way. The drain to ground implementaion also seems to aid in the scheme of things.

Here is how I have it laid out in simulation. (I know the flyback representation is probably not ideal, but I wanted to try and simulate the effect of grounding the focus pin with the HV return. I'm using three diodes due to 3 diodes being represented on the flyback schematic.



Testing last night was not entirely successful. I believe the issue is actually with my spark gap and crude adjustment method. I'm going to be working on improving that today and hopefully give it another run tonight with more thorough results. At this point, I'm just trying to get back to my previous results, before losing the first flyback, with these added stability improvements.

One note: the DC circuitry does not appreciate being tied to the same ground as is being used for the HV return (and pin 7). The first bang in the spark gap would immediately reset my regulator, and/or disrupt DC output of the Power Supply. I'm not exactly sure which, but I would lean towards the latter being the reason the regulator reset and shut the ZVS off. Then again, my oscilloscope reset itself as well and it was only connected to one of the zvs outputs. Either way, I didn't like it, so, for now, I'll be keeping the rest of the TC circuit isolated from DC and mains. Something odd like this is kind of what I expected and the reason I hesitated to use a shared ground, especially back to the mains ground.
The TC secondary is grounded to only the counterpoise, so I grounded the HV return (& pin 7) to the counterpoise for the time being and that did remedy the power disruption I was experiencing. I am considering running a line from the counterpoise to a nearby water pipe just for an added measure. In addition, I'm definitely noticing a difference in behavior depending on what I do with pin 7, but I can't determine what the best measure is. Performance seems better when it's simply left disconnected or grounded to the counterpoise, instead of grounding both the HV return and pin 7. Tying the HV return and pin 7 together seems to adversely affect performance.

Speaking of the counterpoise, I'm now using a 3'x3' piece of aluminum with a perforated pattern. I'm curious if it will work well.

 
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 09, 2019, 12:53:57 AM
Great detail.  Hopefully I can make a coherent response with all the thoughts it triggers.

For the inductor sets, I'd recommend mounting them side by side or end to end, which ever way extends the rows or columns in the same sequence they are wired within each board.  In other words, input to the inductor string on one edge and output on the opposite edge.  With the inductor boards stacked, the capacitance between boards bypasses inductors for the highest-frequencies of the spark gap.  (This may be less important if driving the MMC directly instead of the spark gap - see farther below.)  My other possible concern is the etched copper on the breadboard used for inductor support.  Is the copper just at the fingers and plating through holes?  Or, does it have any larger traces or planes (intended for distributing power or ground in a typical circuit)?  Such traces could bypass the start-to-end of each inductor array with close-enough gaps to potentially arc over.  Even if not arcing, they would add capacitance that bypass inductors much as the board stack.  6H vs. 12H isn't an issue.  It's that with 6H, the voltage is split among only 60 inductors, so higher voltage per inductor, so higher risk of breaking down the inductor's internal magnet wire enamel.

For the larger lower-frequency secondary, the trade-off is of energy-per-spark vs. spark repeat frequency.  Twice the capacitance is twice the energy, which will take twice as long to recharge for given ZVS input power.  I don't have any particular recommendation.  Others here are more likely to have experience with which would produce a more impressive system.

For simulation, note that the flyback schematic had diodes between secondary winding sections, and the focus tap came after the first diode.  Focus is a DC tap, not AC as in your simulation.  Unrelated thought on focus:  since this flyback has focus coming from a tap rather than the full HV output, there's a better chance of keeping it from arcing when open.  It will be at only 1/3rd of your output voltage.  I'd still recommend tying it to the HV return pin, but perhaps not critical.  (I suspect your small flybacks take focus from the full HV output as do the smallest ones I have.)

For grounding, I'd suggest a ground plane under the entire "low" voltage side, the front half in your image.  Given the relative proximity to the higher-voltage spark-gap components, I'd recommend that the ground plane include a vertical part between the two halves.  Ie. a piece of aluminum bent into an L, with horizontal leg under all the LV circuitry and vertical leg separating the two halves, with some cutout or pass-through for the flyback primary leads.  I'd ground this sheet-metal piece to the line ground at the input to the power supply (at the point the line cord enters your system), and to the negative supply output terminal.  (For added protection against supply confusion, run the supply output wires (both + and -) for a few turns through a large ferrite bead or other common-mode choke.  Ground the - side after the choke.)

Concerning power-supply confusion due to sparks, that's my guess.  However, if a ZVS circuit stops oscillating for whatever reason (usually a low-Q resonant circuit), it's supply current ramps up "indefinitely", to the short-circuit capability of the FETs.  That would shut down the supply (of course).

The larger counterpoise should help.  Is it on a concrete slab or other surface built directly on the ground, or on a typical indoor raised floor?  The former will have better capacitance to ground.  If the latter, then I think some connection to a water pipe or line ground would be wise.  For the middle section (flyback secondary/MMC/spark-gap circuitry), I'd still recommend some form of grounding of the HV return pin.  Otherwise stray capacitance could cause a voltage spike on that pin when the spark-gap fires, eventually breaking down insulation within the flyback.  If the power supply still has issues after adding a ground plane under it, perhaps grounding the HV return down to the counterpoise would be less problematic.

Wiring the flyback HV output to the MMC instead of the spark gap isn't normal (to my knowledge), but could have advantages in your case.  It does make the inductor string critical.  After spark-gap firing, the MMC voltage resonates negative almost as far as it was charged positive.  Negative voltage on the flyback could be problematic depending on it's internal leakage inductance, as the diodes would all be forward-biased.  However, as you pointed out, the MMC doesn't have the sudden voltage step of the spark-gap.  That makes the slew rate more manageable, lowering the influence of stray capacitance.  So, it's a trade-off of gentle slew rate for higher peak-to-peak voltage across the inductor string.  Thus, if you feed the MMC, having all 120 inductors is more important, to share the higher peak-to-peak voltage, but the physical layout of the inductors isn't quite as critical.  (I'd still avoid the stack configuration, however.)

If I missed anything for which you wanted feedback, please let me know.

One more thought for grounding the flyback HV return:  Some DRSSTC designs use a bidirectional TVS (Transient Voltage Suppressor) between the Tesla primary and ground.  Sometimes a string of TVS devices rather than only one to get more voltage.  DRSSTC H-Bridges are often powered by a VBus that's directly tied to line voltage (diode bridge from line), so cannot ground that circuit.  A TVS string or capacitor conducts high frequency spikes to ground.  (For the DRSSTC case, spikes are usually Tesla secondary-to-primary arcs.  In your case, they are also normal spark-gap operation.)  For your design, a TVS would be better than a capacitor.  It allows for a moderate spike (500V or whatever TVS voltage you pick), but protects the flyback from larger spikes.  If direct grounding to the counterpoise or to the new ground plane still confuses your DC supply, grounding through a TVS might solve that issue while still limiting voltage to something the flyback transformer insulation can handle.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 09, 2019, 03:41:42 PM
For the inductor sets, I'd recommend mounting them side by side or end to end, which ever way extends the rows or columns in the same sequence they are wired within each board.  In other words, input to the inductor string on one edge and output on the opposite edge.  With the inductor boards stacked, the capacitance between boards bypasses inductors for the highest-frequencies of the spark gap.  (This may be less important if driving the MMC directly instead of the spark gap - see farther below.)  My other possible concern is the etched copper on the breadboard used for inductor support.  Is the copper just at the fingers and plating through holes?  Or, does it have any larger traces or planes (intended for distributing power or ground in a typical circuit)?  Such traces could bypass the start-to-end of each inductor array with close-enough gaps to potentially arc over.  Even if not arcing, they would add capacitance that bypass inductors much as the board stack.  6H vs. 12H isn't an issue.  It's that with 6H, the voltage is split among only 60 inductors, so higher voltage per inductor, so higher risk of breaking down the inductor's internal magnet wire enamel.

That was in the back of my mind – how much stray capacitance that might add, and what the effect and magnetic field interaction might be when stacking the two. It was out of convenience more than anything, so can easily be remedied. I appreciate you clearing up those suspicions up for me. I did consider the voltage-per-inductor if I were to split the strings. I'm sure I'm pushing them a bit, but right now I'm only running for a few seconds at a time - not giving anything too much of a beating.

I'm glad you noticed the boards, as well - as that was another thought in the back of my mind. There are no traces or planes, just plated through-holes. I have never had any issues with arcing using these boards in 'high voltage' applications, but I know i should not be using these particular style boards. There's certain "upgrades" I'd like to make before going much further. I used some lexan and drilled holes to make a proper MMC array yesterday. I’ll be mounting all 40 so the capacitance can be adjusted. Will hopefully have that finished up within the next couple days. I'll probably leave the inductors on the boards they're on to avoid desoldering 120 inductors, probably ruining a few along the way. I'm not completely satisfied with my spark gap right now either. Might revert back to the ol' drawer knobs as I tweak the one I worked on so lovingly.

Quote
For simulation, note that the flyback schematic had diodes between secondary winding sections, and the focus tap came after the first diode.  Focus is a DC tap, not AC as in your simulation.  Unrelated thought on focus:  since this flyback has focus coming from a tap rather than the full HV output, there's a better chance of keeping it from arcing when open.  It will be at only 1/3rd of your output voltage.  I'd still recommend tying it to the HV return pin, but perhaps not critical.  (I suspect your small flybacks take focus from the full HV output as do the smallest ones I have.)

That's a good point and something I had done in some simulations, but left out in the example I provided. I've updated this in my current representation.

Quote
For grounding, I'd suggest a ground plane under the entire "low" voltage side, the front half in your image.  Given the relative proximity to the higher-voltage spark-gap components, I'd recommend that the ground plane include a vertical part between the two halves.  Ie. a piece of aluminum bent into an L, with horizontal leg under all the LV circuitry and vertical leg separating the two halves, with some cutout or pass-through for the flyback primary leads.  I'd ground this sheet-metal piece to the line ground at the input to the power supply (at the point the line cord enters your system), and to the negative supply output terminal.  (For added protection against supply confusion, run the supply output wires (both + and -) for a few turns through a large ferrite bead or other common-mode choke.  Ground the - side after the choke.)

I do have a few pieces of 1'x1' steel plates that could be suspended beneath the low voltage side. Is that what you're picturing? Are you suggesting the vertical plane for shielding purposes? Another plate could be attached vertically between the two halves. Then again, a 1x2 sheet of aluminum would be lighter and could just be bent. More decisions.

Quote
The larger counterpoise should help.  Is it on a concrete slab or other surface built directly on the ground, or on a typical indoor raised floor?  The former will have better capacitance to ground.  If the latter, then I think some connection to a water pipe or line ground would be wise.  For the middle section (flyback secondary/MMC/spark-gap circuitry), I'd still recommend some form of grounding of the HV return pin.  Otherwise stray capacitance could cause a voltage spike on that pin when the spark-gap fires, eventually breaking down insulation within the flyback.  If the power supply still has issues after adding a ground plane under it, perhaps grounding the HV return down to the counterpoise would be less problematic.

The counterpoise is placed directly on my concrete garage floor, so ground level.
I’ll make a short video of the different configurations I’ve tried, and the observable effects.
This whole stray capacitance factor is doing my head in.

I feel stupid even bringing this up (I'm sure it's simple enough), but  as you can see from the attached image on my last post, I've placed the entire setup on a stand, which is metal. How, significantly, might this now be affecting operation? I know that anything in the vacinity of the coil will affect the resonant frequency, but how greatly, if frequency is accounted for, can this change the behavior?
Too add to my feeling of ignorance, I'm left scratching my head today after observing what resembled lower-voltage sparks being created at one of the bolts holding the stand together . At the time, there was absolutely no live wiring in contact with the frame, and the stand is sitting on 2x4's on top of the counterpoise (the wood shelf and top are not even attached to the frame). I attached a wire from the bolt to my grounding bar, leading back to the mains ground...still sparks from the bolt. The spark appears when the spark gap fires. What is this voodoo?

Here’s a video where you can see the spark from the stand [go ahead and disregard the BPS here]:
https://drive.google.com/open?id=1HkMU6Wq_A7iXT1Z1vB4ykZQa03OFjEHc

Quote
Wiring the flyback HV output to the MMC instead of the spark gap isn't normal (to my knowledge), but could have advantages in your case.  It does make the inductor string critical.  After spark-gap firing, the MMC voltage resonates negative almost as far as it was charged positive.  Negative voltage on the flyback could be problematic depending on it's internal leakage inductance, as the diodes would all be forward-biased.  However, as you pointed out, the MMC doesn't have the sudden voltage step of the spark-gap.  That makes the slew rate more manageable, lowering the influence of stray capacitance.  So, it's a trade-off of gentle slew rate for higher peak-to-peak voltage across the inductor string.  Thus, if you feed the MMC, having all 120 inductors is more important, to share the higher peak-to-peak voltage, but the physical layout of the inductors isn't quite as critical.  (I'd still avoid the stack configuration, however.)

I have seen the circuit configured in both ways, but, you're right, almost always with the spark gap in parallel. The capacitor-in-parallel configuration seems more common in “spark gap transmitter" schematics than “tesla coil” schematics, but, even then, can be found interchanged.

An issue I was having yesterday – I haven’t conclusively determined the cause – was the varying BPS. I experimented with adjusting voltage/current as much as I was comfortable, but seemed to always have the same result.
Could this have anything to do with the single inductor string possibly affecting the flyback's charging of the MMC or affecting the oscillations in a way that the MMC is noit fully discharging, or possibly overcharging, etc. for some reason?
An inductor string on both the HV out and return, I would think, should help further prevent interference between/isolate the flyback output and the tank circuit as it oscillates. Just a suspicion.

These new variables are making it harder for me to troubleshoot this solo.

Here’s a video of the effect during a quick 18V test: 
https://drive.google.com/open?id=1HpKlL3n0VsK1rMohkfpC5O3ucFM3YPM9
and a better one with view of the spark gap:
https://drive.google.com/open?id=1Ivh2nUFPYYyEWtffFCxdjwl6uGNmo1NY

In the videos, the breaks are not exactly inconsistent, per say, but the speed varies. It will be constant at one speed, then all the sudden start firing much faster, then return back to where it was. It's easy to see the effect on arc formation.

It's not the best example... - I'd like to record the actual spark gap at 120fps, as well, to observe that more clearly.
I might even be able to use some neutral density filters on my lens to actually be able to see the arc in the gap, as well. Hmm...that would be interesting.

Quote
One more thought for grounding the flyback HV return:  Some DRSSTC designs use a bidirectional TVS (Transient Voltage Suppressor) between the Tesla primary and ground.  Sometimes a string of TVS devices rather than only one to get more voltage.  DRSSTC H-Bridges are often powered by a VBus that's directly tied to line voltage (diode bridge from line), so cannot ground that circuit.  A TVS string or capacitor conducts high frequency spikes to ground.  (For the DRSSTC case, spikes are usually Tesla secondary-to-primary arcs.  In your case, they are also normal spark-gap operation.)  For your design, a TVS would be better than a capacitor.  It allows for a moderate spike (500V or whatever TVS voltage you pick), but protects the flyback from larger spikes.  If direct grounding to the counterpoise or to the new ground plane still confuses your DC supply, grounding through a TVS might solve that issue while still limiting voltage to something the flyback transformer insulation can handle.

That’s an interesting idea – something I should probably have known from other projects where I had to deal with transients. Instead of the HV return pin directly to ground, you’re proposing to place TVS diodes in series along the way – am I getting that right? Would it not be better to use MOVs with a higher clamping voltage and tolerance for higher energy/temperatures? Would large enough spikes be expected that I could throw in a little  NE-2 neon in the mix as an indicator?
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 10, 2019, 06:04:28 AM
Wow, a lot of great information.  I'll attempt to comment top-down.  (When I quote, it often ends up jumbled on the reply, so I'm skipping such.)

The inductor boards aren't an issue with only plated holes/pads.  No need to change.  I just saw the edge-connector pads and thought they might be like some old breadboard I have that include power and ground traces distributed across the hole array.  (Stacked boards are a capacitive-coupling issue, not magnetic, in this situation anyway.)

If making a 40-cap MMC with ability to run 20, the two strings need to be well separated.  I'd suggest two separate strings that can be bolted together.  With two strings in place, and one string open at one end, there's your full ~20kV primary voltage from the open string to the adjacent connected string.

If the 1'x1' steel is galvanized (zinc plated), it could work.  Plain steel has a VERY thin skin depth due to being ferromagnetic.  The separate sheets would still need good connections along their edges even if galvanized.  If you don't mind the cheap construction, aluminum foil taped to cardboard works fine.  (My DRSSTC has the control circuitry in a cardboard box lined with foil.)  The ideal would be a complete enclosure (Faraday cage) around the LV circuitry, but two sides is probably sufficient and leaves it much more open for debug.

Counterpoise on concrete floor is great!

I noticed the stand, but initially guessed that just the four posts/legs were metal.  Now I'm thinking that there are square metal frames for the top and shelf, with wood squares for the surface.  Is that correct?  It's the only I can explain the interesting sparks in your first video.  They appear to be sparks of burning steel, not electrical sparks.  That could make sense if the top (square) ring is steel bolted together with marginal electrical connections.  It's inductively coupled to the Tesla primary, inducing enough current in that loop to spark where the steel touches (high-current welding-style sparks, not high-voltage sparks).  The steel legs may be OK (why I didn't comment on your last post), but any closed metal loop near or above the primary is quite problematic.  Could easily explain your inconstancy.  Possibly you could break (insulate) the joints, but it would be better not to have any steel close to the Tesla primary either.  It will concentrate the magnetic field, then cause a lot of hysteresis and eddy-current losses within the steel, lowering overall efficiency.

Inconsistency is likely the steel frame as mentioned above.  Another possibility is that the spark gap doesn't remain conductive long enough, so the Tesla primary L/C doesn't ring down all the way, leaving either a negative or positive charge on the MMC after firing.  Much less likely, however.

For flyback HV return grounding, yes, I'm suggesting a string (or just 1 if it's voltage is high enough - whatever keeps your power supply sane) from HV return to ground, perhaps down to the counterpoise for ground presuming the counterpoise does have some wire to a line or pipe ground.  MOVs have higher capacitance, so would pass more of any high-frequency spike, so won't protect the supply as well.  (I'm hoping the ground plane is sufficient to keep the supply sane, so no TVS diodes are needed.)  You could add a neon bulb - it may arc rather than a normal neon glow for the short pulse if the voltage is high enough, which may fry the bulb.

BTW, if you end up with a stand with any metal parts, my general rule is no electrically-floating metal.  All metal is either part of an electrical circuit or tied to ground.  This is more for ESD issues in equipment (printers in my case), but I'd recommend for almost anything.  If nothing else, it makes behavior consistent, rather than allowing occasional arc-over between metal pieces.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: Mads Barnkob on December 10, 2019, 10:27:54 AM
In regard to counterpoise and grounding, I wrote a long article on the subject and it goes just as well for SGTC as for DRSSTCs.

Grounding rods are almost always a must, the capacitance ratio between your counterpoise/artificial ground plane/metal sheet on the ground has to be around topload 1 : counterpoise 10, in order to have a solid distribution.

Read it all here: http://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/grounding-circuit-protection-and-emi/
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 11, 2019, 01:14:32 AM
After some further research, I'm considering the following determinations here. Bare with me and let me know what you think.

Metal stand aside, for a moment while we consider the spark gap behavior as well as another observation I hesitated to share until I understood further what might be happening - that being, a few instances when a cap in the MMC would actually discharge, I assume arcing to another capacitor, which has never happened before in any of my testing. The leads of the capacitors are spaced a good deal apart, so this was very surprising when it happened. I was able to recreate (more so observe) the condition multiple times, albeit not on 'command', exactly.
Anyway, back to constructive analysis:
While using inductors as a sort of ballast in a charging circuit that incorporates a spark gap, the inductive kick effect, that the inductor(s) are there to mitigate, can increase the level to which the tank capacitor charges. This would be due to the inductors ability to store energy which would otherwise be dissipated in other forms, primarily heat, I assume. The stored energy in the inductor is then released some time later in the charging cycle.
With that in mind, it may also be worth considering that although the breakdown voltage is largely proportional to the electrode spacing, it reduces with increasing temperature. If the electrodes and the air in the spark gap are allowed to heat up excessively the breakdown voltage of the spark gap decreases considerably. In this instance the spark gap fires at a higher rate, but with the tank capacitor at a much reduced voltage. With the time of year and the ambient conditions changing quite a bit from when I first started this project, it may also be a factor to consider. To add to that - obviously, when any system that uses a gap-based "switch" powers up, the gap is cold, and the first breakdown occurs at a higher voltage than when running. If I increase the distance between electrodes, a very high initial voltage needs to occur before the spark gap will start firing. However, once running, the firing voltage is lower and less power is being transferred. As a consequence, the flyback, inductors, and MMC must be rated to withstand the high power-up transient, but the overall performance would depend on the lower breakdown voltage when the system is running.
Another takeaway - with my current, 20 cap series, MMC setup rated at only 24kV, I may be dancing on thin ice. It would seem, regardless, I should probably add more caps to the series, considering the rated output of the current flyback.
To an extent, the above effects may be partially responsible (as well as the proximity of the steel stand) for the observed inconsistency as the spark gap heats up. The increased firing rate of the spark gap would cause further heating and the static gap to become truly overloaded. As can be heard in the video, a characteristic increase in pitch is noticeable in the sound from the spark gap, and it is accompanied by almost total loss of spark output. However, in my case, there was still fluctuation in speed, which might indicate more than one, possibly even more than a few factors are contributing to my inconsistent results. Granted, with a static spark gap, is it difficult to take repeatable measurements as the behavior is different from one supply cycle to the next.

You're right about the stand. There are square, metal, frames for the shelf and top, and now that you point it out, that actually makes complete sense. Being inductively coupled, and a closed "loop" (for lack of a better description), that spark is forming at the "weakest" connection in the "loop" (ie. janky bolt that's not tightened down all the way), which is why that's the only bolt I'm seeing that exhibited. Still very interesting to see that when the stand is grounded. So many things to consider here!

I'm going to remove everything from the metal stand and repeat previous tests. I will be sure to update with new information!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 11, 2019, 06:08:48 AM
Sparks between MMC capacitors sounds like your most serious problem.  Unless leads are extremely close, it would require MUCH more than 1200V to spark.  Do you have a bleed resistor across each cap, all the same value?  That's a requirement for any series-connected MMC.  Otherwise non-uniform corona will slowly redistribute voltage throughout the array until you have a serious over-voltage on one or more caps.  If a lack of bleed resistors isn't the issue, then a close-up image of the MMC array and wiring would help to see what else could possibly be the issue.  I'd also suggest measuring each of the 20 caps to make sure that whatever the issue may be hasn't damaged them.  (Damage usually shows as capacitance drop initially - a few percent, followed by rising leakage current if abuse continues.)

For your concern about 24kV max, I wouldn't worry much.  I've destructively tested a couple of these caps in Tesla use.  It required +-2150 volts for a couple hours to induce failure.  At +-2000V they lasted for a couple days of continuous firing before I gave up waiting and upped to +-2150V.  These same two caps had already been ran continuously at +-1700V for a week.  (When I say "continuous", I'm referring to ongoing 1% duty cycle, 500us of 80kHz repeated every 50ms.)

For the circular (square) current loop of your stand top, grounding doesn't matter.  It's a local loop.

Do you have the fan blowing on your spark gap energized?  If so, I doubt you'll see much voltage change.  What is the gap distance?  My 6kW Marx generator had huge voltage shifts before adding air flow, but was plenty consistent with quite moderate air flow.

One more thought on consistency:  It's possible that the flyback is generating enough current to sometimes sustain the spark within the gap, so delaying any subsequent MMC charging.  That would be especially likely without air flow.  That was a big problem with my Marx generator even with air flow.  The 6kW input had plenty of current to maintain spark gap arcs.  I had to add a circuit to pause the supply for ~1ms after each firing to allow time for the arcs to blow out.

I doubt the inductor string matters much.  It's current just before firing isn't going to be as high as the flyback will generate just after firing.  The only extra current added to the inductor string due to firing is from the energy of the stray capacitance of the flyback output and its wiring, perhaps 1 or 2mJ.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 13, 2019, 05:05:55 PM
Secondary LC removed from metal stand and placed back on previous base, which is just sitting on a cheap wood computer desk, exactly where it was before. Basically zero metal in the desk besides the two drawer rails beneath. The counterpoise was moved out from under the metal stand and placed directly under the desk, centered under the secondary LC.
I moved the shelf with the primary LC and ZVS, power supply, etc. to sit on top of the metal stand for convenience and placed a couple feet away.
So as it sits:
     - The Negative terminal of the ZVS driver is tied to mains ground, as is the metal stand.
     - The HV return and focus pin (7) are joined and ‘grounded’ to the counterpoise at the spark gap. (still not sure about this pin 7 business)
     - The counterpoise, itself, is not attached to earth or mains ground.
     - The bottom of the secondary is attached to the counterpoise.
     - The secondary LC is approximately 4’ above the counterpoise.
     - 6H inductor between the Flyback output and spark gap.
     - 16.5nF MMC

Before a quick “let’s see what happens” sort of test, I measured all 20 capacitors. Each measured between 330-335nf. 
I removed the variable of the new spark gap, and used my old starter gap (ie, two drawer-knobs (nickel [I think]; unplated) on screws through an electrical box.) There is no airflow on this gap, but is more ‘finely’ adjustable
During the first test I used an input of 18V and limited current to 6A. The spark gap fired consistently, with the occasional hesitation.
I didn’t take any other measurements during testing, partly because the MMC experienced the “something” again, after about 30-40 seconds of the first run-time. It doesn’t happen immediately, or in any type of pattern. It seems slightly prone to occur sooner if I let it run a moment, then stop the input for a few seconds and reapply voltage again. Everything is really speculation for me at this point
Note: before testing I  added more solder to all MMC connections. There was no possibility of any faulty connections. The individual capacitors are wired in an odd zig-zag pattern (simply just to fit as many on the small boards as I could), but no leads come within ~½” from each other. As previously mentioned, however, the plated through-holes do give me cause for concern.

Here’s a tour (this is after re-soldering and the “events”) (also: if video looks like garbage quality, try downloading)
https://drive.google.com/open?id=1KLX0-wxhPm45ne9nwg5_x22xdWKtkPja

All resistors are 1M with the exception of a few 1.1M. I ran out of 1M apparently. I have plenty more on the way for the new MMC though.

I went as far as to remove the regulator and essentially go back to the original configuration with the exception of the inductor string and new flyback, of course. I supplied a 24V input (adjusted with a trimmer on the supply. The supply is rated for 10A. The spark gap continued firing consistently, but the circuit still provoked a reaction in the MMC.
I made a video recording at 120fps and still could not catch more than the flash, so I have no idea where it’s actually coming from within the MMC. I thought an arc should have been visible, but this is just a loud pop with no arc or spark to be seen. The flash is really only visible in the video, and even then the origin can’t be determined. It happens very quickly.
What kind of voltages would have to be present here for a flash-over to form? Is it possible the plated through-holes, if enough voltage is present, are providing a path for arcing to occur?

Here’s the best video I could get: (same - if video looks like garbage quality, try downloading)
https://drive.google.com/open?id=1K65AYEjPaQVol-eoSUEmOMvaWGghPmCN

UPDATE***:
I think I've spotted it...maybe. Watch the far left end of the strip of hot glue on the left side. It flips up a bit during each occurrence (it's not stuck to the board anymore). It looks like it might be happening between the boards, beneath that piece of hot glue. Just speculation at this point.

I really thought I’d be back to or close to my previous results by now, with improvements – this is quite frustrating, especially considering this is the same MMC I’ve been using since the beginning of this project and have had zero issues until now. Setbacks, setbacks. All in the name of learning.

I intended to complete my new MMC yesterday, however, some of the capacitors I was planning on using required de-soldering and ended up being more trouble than it’s worth. I’ll probably just buy some more. They’re cheap enough.

I think that’s it for now. I should have some more time to put in tonight for debugging and will report any further findings.

Thanks again!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: MRMILSTAR on December 13, 2019, 08:56:27 PM
Its just my quick observation but all those plated-through holes look like a disaster waiting to happen. You have hundreds of potential flash-over points on that board. I would never consider such a construction method for high voltage circuitry.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: John123 on December 13, 2019, 09:08:28 PM
Could the ozone levels around the coil, hv spiking and UV from sparks actually encourage that prototyping board to flash over at a much lower voltage?
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 13, 2019, 10:50:03 PM
Its just my quick observation but all those plated-through holes look like a disaster waiting to happen. You have hundreds of potential flash-over points on that board. I would never consider such a construction method for high voltage circuitry.

You're right on and that has been in the back of my mind for a while now, and it's my current thought and concern, exactly. It's the only thing that makes any sense to me right now, and, I would imagine, is likely the culprit here.
Perhaps my previous flyback was a bit more wimpy than this guy, so I'm only just now seeing the error of my ways. Against my better judgement, I used these because they were what I had around. You're right though, i shouldn't be using these boards for this application.
I should have the new MMC together this weekend. The caps will be mounted on a sheet of lexan, so should have no chance of flash over or anything.

Could the ozone levels around the coil, hv spiking and UV from sparks actually encourage that prototyping board to flash over at a much lower voltage?

That's an interesting question John. I don't think the TC was running long enough to create high enough levels of ozone to affect it like that. We're talking 10-20 seconds of run time.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: John123 on December 13, 2019, 11:00:18 PM
It's just when I built a marx generator the UV light from the initial spark gap was used to encourage the others to spark in sync, if direct line of sight was blocked then the rest wouldn't fire reliably.

But yeah how many volts are across those prototyping board capacitors? With the prototyping boards I've used you could take a low-high voltage source of say 5-10kV and even that was enough to jump the entire boards width of gaps. Looked like an LED light show project when pulsed.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 13, 2019, 11:19:27 PM
It's just when I built a marx generator the UV light from the initial spark gap was used to encourage the others to spark in sync, if direct line of sight was blocked then the rest wouldn't fire reliably.

But yeah how many volts are across those prototyping board capacitors? With the prototyping boards I've used you could take a low-high voltage source of say 5-10kV and even that was enough to jump the entire boards width of gaps. Looked like an LED light show project when pulsed.

It's definitely over 10kV as this flyback is rated at 24kV max. This one does seem more current hungry, so I can certainly imagine it capable of getting angry enough to flash over. That is part of my predicament as I don't currently have a way of measuring the exact HV ouput, voltage or current. It's so hard to find specs on these flybacks that I can't even compare the rated output to the previous flyback I was using when everything ran like butter...until it didn't.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: MRMILSTAR on December 13, 2019, 11:48:12 PM
One other thought. I see that you are using 1200 volt capacitors. What is the voltage rating of your bleed resistors? Do these have a sufficient voltage rating? I don't know what your capacitors are being charged to and it sounds like you don't either.

With regard to MMC construction technique, perf board is acceptable but not with the plated-through holes.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 14, 2019, 12:57:15 AM
One other thought. I see that you are using 1200 volt capacitors. What is the voltage rating of your bleed resistors? Do these have a sufficient voltage rating? I don't know what your capacitors are being charged to and it sounds like you don't either.

With regard to MMC construction technique, perf board is acceptable but not with the plated-through holes.

I'm not sure I follow regarding the bleed resistors. They don't need to be rated for "1200V" do they? I'm not sure what you're getting at.
Aren't these bleed resistors in place to, for the most part, dissipate residual charge in the capacitor after the system is powered off? They do not need to withstand the 250-300W (max) that the primary LC might see while the SGTC is operating. If there was an issue with the resistors, I would have expected to see at least one of them fail, but that's not the case. Am I wrong?
I was planning on using 1W, 1M, resistors on the new MMC.

______

Just to add to my last "update" post, as well.:
This is a run on the same circuit with the same MMC, with a different flyback and no inductors.
Until now, I've never had any issue with the MMC and DC-input-wise, I'm using nearly half the voltage I was using previously.

Video:
https://drive.google.com/open?id=1BNOaCzaXydjddC9Pn8yb6Wtv-VFpBh03

And since I'm feeling nostalgic - a nice display at 120fps:
https://drive.google.com/open?id=1CkyjDH6enRkY398ecstW0iXVxJOWfRqc
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 14, 2019, 05:26:41 AM
Your existing MMC appears to use 1/4 watt resistors.  Yes, resistors almost always have a voltage rating besides a power rating.  Those 1/4 watt resistors are probably rated for 100V.  They are definitely risky at 1200V.  I've fried a number myself.  The actual carbon film band is surprisingly-short in the center, so they will arc over eventually.  The likely-100V spec. is probably limited by resistance linearity.  An ideal resistor has constant V/I.  At higher voltages, the current may increase faster or more slowly than at low voltage, making the effective resistance change.  The spec. voltage is the range where resistance tolerance is guaranteed as well as where no arc-over will occur.

You can buy high-voltage-rated resistors, which I often use.  For low-accuracy needs, I'll run resistors above rated voltage, but within limits based on my experience.  1W resistors as in your new MMC are likely fine at 1.2kV.  I've used 1W parts to 1.5kV.  The 1/4 watt resistors would probably be OK at 600V (two in series per cap).  I tend to keep 1/4 watt parts under 400V for margin (three in series per cap).

The bleed resistors do discharge the MMC after use, but they also play another important roll.  They keep the 24kV total evenly shared among the 20 caps.

The arc on your existing MMC is clearly at the left center as you said.  Here's a frame from your video:
 [ Invalid Attachment ]

This shows up twice, where the flash straddles two frames.  For the other flashes that fill the entire frame, the frame after shows a tiny remnant of glow at the glue.  It's hard to see without comparing to the next normal frame, but here's a capture:
 [ Invalid Attachment ]

The arc location is where I expected it to be after viewing the MMC layout.  With a folded zigzag, there is 80% of the total voltage across that 15mm gap, with plated holes to help bridge the distance.  If you have room, it would be better to make a long bar of side-by-side caps, 30cm total for those 15mm wide parts, with one lead at each end.  If you must fold it, add a sheet of plastic between the two folded halves to separate them.  (BTW, if your MMC was a single 30cm long bar, the plated-through holes would likely not be an issue.  There would be at most 2.4kV between adjacent wires, or only 1.2kV if the wires made all the zags as I suggested for the inductor array.)

I particularly like your 120fps video!  The good results don't mean the MMC would have continue working.  Such flash-over failures tend to build over time.  Local corona discharge slowly damages (carbonizes) the insulation, slowly making the gap shorter until eventual failure.  The same reasoning likely applied to your first flyback, failing after a period of use.

Good luck with this weekend's tests!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: Mads Barnkob on December 16, 2019, 10:38:43 AM
I particularly like your 120fps video!  The good results don't mean the MMC would have continue working.  Such flash-over failures tend to build over time.  Local corona discharge slowly damages (carbonizes) the insulation, slowly making the gap shorter until eventual failure.  The same reasoning likely applied to your first flyback, failing after a period of use.

I agree on Dave's notes on the MMC, folding two high voltage ends of opposite polarity to meet like that will cause trouble.

It is not only carbon tracks from corona that is a problem, it is rather extreme to start getting burned carbon tracks from just corona, usually that takes some sparks. From corona I would say that it is the ozone that is highly corrosive and it attacks and breaks down many types of plastic materials.
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 17, 2019, 04:52:34 AM
Your existing MMC appears to use 1/4 watt resistors.  Yes, resistors almost always have a voltage rating besides a power rating.  Those 1/4 watt resistors are probably rated for 100V.  They are definitely risky at 1200V.  I've fried a number myself.  The actual carbon film band is surprisingly-short in the center, so they will arc over eventually.  The likely-100V spec. is probably limited by resistance linearity.  An ideal resistor has constant V/I.  At higher voltages, the current may increase faster or more slowly than at low voltage, making the effective resistance change.  The spec. voltage is the range where resistance tolerance is guaranteed as well as where no arc-over will occur.

You can buy high-voltage-rated resistors, which I often use.  For low-accuracy needs, I'll run resistors above rated voltage, but within limits based on my experience.  1W resistors as in your new MMC are likely fine at 1.2kV.  I've used 1W parts to 1.5kV.  The 1/4 watt resistors would probably be OK at 600V (two in series per cap).  I tend to keep 1/4 watt parts under 400V for margin (three in series per cap).

The bleed resistors do discharge the MMC after use, but they also play another important roll.  They keep the 24kV total evenly shared among the 20 caps.

The arc on your existing MMC is clearly at the left center as you said.

This shows up twice, where the flash straddles two frames.  For the other flashes that fill the entire frame, the frame after shows a tiny remnant of glow at the glue. 

The arc location is where I expected it to be after viewing the MMC layout.  With a folded zigzag, there is 80% of the total voltage across that 15mm gap, with plated holes to help bridge the distance.  If you have room, it would be better to make a long bar of side-by-side caps, 30cm total for those 15mm wide parts, with one lead at each end.  If you must fold it, add a sheet of plastic between the two folded halves to separate them.  (BTW, if your MMC was a single 30cm long bar, the plated-through holes would likely not be an issue.  There would be at most 2.4kV between adjacent wires, or only 1.2kV if the wires made all the zags as I suggested for the inductor array.)

I particularly like your 120fps video!  The good results don't mean the MMC would have continue working.  Such flash-over failures tend to build over time.  Local corona discharge slowly damages (carbonizes) the insulation, slowly making the gap shorter until eventual failure.  The same reasoning likely applied to your first flyback, failing after a period of use.

Good luck with this weekend's tests!

Great observation. You were absolutely right. After separating the two halves, I've been able to run without issue for long enough that I'm confident for the time being. I've gotten back to the point where I'm encouraged by the output! I've attached some videos with observational notes.

By the end of this tinker session, I was using 35V input. That's about as hard as I want to push my ZVS or my cheap SWPS. It may be time to step up to a proper transformer...
As far as readings go - I'm really having trouble nailing down ZVS output voltage while the system is on. For some reason, even measuring the DC input proves nearly impossible with my meter as well as the digital voltmeter/ammeter I wired in to monitor the SMPS. Readings are all over the place, fluctuating wildly, except for the frequency readings which tended to average ~37-40kHz (consistent with previous tests with this particular flyback).
Unfortunately I seem to have scrambled my cheap oscilloscope, so will have to wait until i get a hold of decent one or grab another cheap kit.
I have a good assortment of cheap analog meters laying around, so I think I'll wire those in to, at least, get an average measurement on volts/amps on the 'low-voltage' side.
Is there any relatively simple way to modify one of these analog meters to monitor voltage/current on the high voltage side while the system is running?
Note: I was also reading .07A on the ZVS input negative/ground.

After 1 minute of run time, Flyback secondary and core are hot but are not of concern, yet. The ZVS inductor is very hot, as were the FETs. I did not take temperature readings, but plan to next time.
If I continue to use this ZVS driver, I should probably add a fan. You can generally just smell hot-ass electronics.
Inductors remained cool.

Specs I have for the flyback are:
Rated Output Voltage: 24.2kV/30kV (2 sources vary)
Rated Output Current: unknown
Core Dimensions: 15.46mm*41.97mm*74.11mm
HV to Return pin resistance measures: 24.35M
HV to Pin 7 resistance measures: 24.62M
I'm also reading .05nF capcitance from HV to Return.
Schematic attached in earlier post

Currently, I have the inductors split into two 6H arrays - one on the HV output, and one on the return with the spark gap in parallel. I may try swapping the MMC and spark gap and compare.
I still have the occasional fluctuation in the spark gap, which can easily be seen in the 120fps video below and still seems rather random, but short-lived,.
With the current spark gap, I do not have any airflow. Now that I have taken care of some of these variables now, I can focus on the spark gap and will refine the new one I had constructed already. I still need to solder up the new MMC, but that's coming as well.

Finally, I'm getting closer to the results I was aiming for with this build. If anyone would like, I have accurate physical measurements and schematics needed to reproduce it, and I can post them with current schematics and known specs..
After a few more tweaks I may do a proper and entire walk-through of this build process up to this point, if anyone thinks it's worth it, including the final *nice* toroid, as I've taken pics and videos all along the way.

Smaller Topload @ 35V:
/>
Larger Topload @ 35V:
/>
Larger Topload @ 35V (120fps):
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 18, 2019, 06:09:34 AM
Justin,

All my cheap meters are useless with sparks around, even corona sometimes.  Placing them in a Faraday cage helps a lot (aluminum foil lined box w/o top so meter can be viewed).  Tie the foil to the meter negative lead inside the box.  A low-pass filter on the positive lead helps further, something like 10K-100K in series with a >=0.1uF cap across the meter input.  (10K if the meter input impedance is only 1M, 100K if 10M, makes 1% error in the reading.)  I made the same sort of foil/cardboard cage for my cheap Chinese scope, with a few holes for cooling air.  Besides grounding the foil to the scope ground inside the box, I also made a common-mode choke for the AC power cord (several turns of the power cord around an E55 ferrite core).  Sometimes I'll wind scope probe leads around ferrite cores as well, even on my nice Tektronix scope - to protect the scope and reduce channel crosstalk.   Add common-mode ferrites to scope probes at work as well, even without sparks to filter out.  Helps when measuring one higher-voltage signal such as an H-Bridge output and one low-voltage logic or analog signal.  (BTW, wiring for your panel meter is rather close to a HV lead, making interference worse.  Having the LV side on a ground plane as suggested previously may help with meter behavior too.)

Analog meters are great for Tesla projects.  Most list the full-scale current in fine print on the face plate.  1mA is typical.  If you can find one with 0.1mA (100uA), that would be great for measuring your HV.  200uA could be usable.  Just add a string of HV resistors from the HV flyback output to the meter, with meter - lead returning to HV return pin.  Here the resistor voltage rating does matter, as you're after linearity.  Vishay's VR68 series are good for 10kV each.  (Of course, power rating needs to fit as well.)  For measuring Tesla HV with an analog meter, dial down the ZVS input voltage until getting only 1 spark  per second.  Then you'll have some chance of seeing the firing voltage before it discharges.

You could make a small signal-diode bridge rectifier to monitor the 90Vpp-rated flyback tap with an analog meter.

I'm not clear on where you measured 70mA.  That's way to low for much on the ZVS circuit or flyback input.  It's high (but could be accurate) for the flyback secondary return pin.  If it is there, that's likely well more current than the flyback's rated for.

If you can find a PFC inductor from a good sized DC supply (PC supply or whatever), those make great ZVS inductors.

What's the diameter of the flyback ferrite core going through the potted coil?  Or, the cross-sectional area of the rectangular remaining ferrite?  That's more useful for calculating saturation than the larger dimensions.

Resistance measurements on flyback HV and focus pins is generally meaningless.  Meters don't output enough voltage for the forward drop of the HV diodes.

With two separate 6H inductor strings, is the HV return ground (to counterpoise) still at the flyback HV return pin, not after inductors?  I'd still recommend a path somewhere from HV return to the ZVS negative lead.

From the 120fps video, I think the firing variation is from remaining ionization in the spark gap, with random subtle air current changes.  A fan should get you to all louder lower-frequency sparks.  If you prefer smaller high-repetition-rate sparks, then reduce the spark gap separation, but leave air flow on for stability.

I'm presuming you re-tuned when changing toroids.  Looks like about half-way out on your primary coil in the first video.

The relatively long and separated wires running to your Tesla primary are adding inductance, which lowers coupling factor.  Not sure how beneficial higher-coupling is for this design, but I'd consider eventually shortening wires and perhaps routing them closer together with appropriate insulation in the gap.  (In my DRSSTC, even the MMC array halves are adjacent to lower inductance, but with a 10mm thick PP sheet between for insulation.)

Great to see progress!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: jturnerkc on December 19, 2019, 03:38:42 AM
Instead of winding around a ferrite core for a common-mode choke, I had considered throwing on some clip-on ferrite beads for the hell of it. Would those even be worth bothering with?

As far as having the LV side on a ground plane, as suggested previously, I do have the ZVS attached to mains ground, while the flyback is still grounded to the counterpoise from the return pin along with pin 7. The inductor string is placed between that point and the spark gap.


I have a small ground bar where the HV return and TC secondary ground meet and lead to the counterpoise. Connecting the HV return to the ZVS negative would no longer keep the primary and secondary tank circuits electrically isolated from mains ground, as the ZVS negative is already grounded to mains. I had considered attaching a line from the counterpoise to a nearby pipe (granted they are likely connected to mains ground as well, but I can only assume that's better grounding directly back through the outlet) or may just suck it up and drive in a physical ground rod outside the garage…Tying all 3 to a single ground point, especially mains ground, seems to go against nearly everything I've read about Tesla Coil grounding.
I'm not sure I completely follow the reasoning, however I know the proper grounding techniques for TCs has been hotly debated, and I've certainly found evidence of at least some users tying the counterpoise to mains ground. I could give it a try. I wouldn't know where to begin trying to determine if it is for better or worse, though, unless something noticeably changes operationally that I'm able to perceive or measure with my limited equipment.

Regarding analog meters and monitoring, thank you for reminding me you had mentioned measuring the 90Vpp before. I’m not sure why that didn’t come to mind as an option. I suppose only partially because I’d like to know what the actual output voltage is of the flyback in operation. It would be worth monitoring the 90Vpp though, so at least I can have an idea of max input voltage. I have a bunch of AC analogs so may not even need to use a rectifier. I’ll have to dig and see what voltage ranges I have. I have an aluminum enclosure I can install these in as well. I had considered housing the all the LV circuitry inside, but it’s just barely too small to fit the supply and zvs both. I’ve got an eye out for a larger one.
If I were to use a bridge rectifier, what signal-diode would you suggest? Is the standard 1N4148 beefy enough to rectify 90Vpp or is there another more appropriate choice? I'll have to see what I have around. I do have a bit of an assortment that might have what I need.

Quote
What's the diameter of the flyback ferrite core going through the potted coil? Or, the cross-sectional area of the rectangular remaining ferrite? That's more useful for calculating saturation than the larger dimensions.
I'll have to get back to you on the core cross-sectional area.

You're correct about the HV and focus pin measurements - I could not get a voltage drop reading.

I would also agree with your observation of the spark gap BPS. The variations seemed different in nature to the previous issues I was having with the "new" spark gap. I'll add some airflow during my next round of tinker-time.

To answer your question about the two toroids and tuning, 'yes', I did have to re-tune when switching toroids. I moved the primary tap out about 1 1/4 turns due to the frequency change induced by the larger topload.
I’d like to try the oscilloscope method of tuning since I’m pretty much just doing "fine tuning" manually right now. Unfortunately my function generator decided to stop working since the last time I used it. Strange. I have to replace my scope anyways. Probably just get another cheap handheld since it will probably be scrambled at some point like its predecessor.

You had mentioned the length of wiring and you're right, however the wiring is really just temporary at this point and I’ve changed the orientation of some things. The longer wires were just giving me room to move things around a bit. Once I have something more finalized it will all be wired more appropriately and properly housed, etc. I’d also like to place all the LV circuitry, along with the supply in an aluminum enclosure with controls and whatnot that could be operated further away from the rest of the unit. The way you saw the MMC in that video and the other components was before I reconfigured the layout. I now have the MMC in line, essentially just fully unfolded.

I've made extremely slow progress, the last week, getting the new MMC put together and soldered. I'm actually waiting on some additional caps now, as well. I've attached an image. Every lead is separated by 1 inch and each cap is connected with a piece of 3/32" welding rod. I intend to complete wiring in the same fashion/configuration as the inductor array(s).


Considering exchanging the ZVS inductor - I do have that 0-48V SWPS I haven’t fixed yet… I could just cannibalize it. I had my eye on those big electrolytics too, for something else...maybe even a full-bridge DRSSTC later? Hmm...
I’d like to increase the operating range of my ZVS, as well, so the inductor and FETs aren’t getting so hot at 35V input. I just want everything as cool as possible. I’ve removed the ground from the caps and wired them in parallel as is more common. I had considered adding 2 more caps – 2 series strings in parallel – and adding an additional inductor in parallel with the existing to the center tap. I'm just not sure how much more the FETs can take though. I’ll need to wind my own inductors or get a hold of two matching if I go the dual inductor route.
I've also been tossing around the idea of just removing the center tap from the equation completely and just having inductors connected to both ends of the primary. It would hardly take any modification, but I've yet to investigate how the output would be affected in this actual application. More ideas to bounce around.
I'd be curious to hear your thoughts on driver usage and possible changes you might consider, yourself.
Thanks, as always!
Title: Re: SGTC MK1 - An Accomplishment in Progress
Post by: davekni on December 19, 2019, 06:35:28 AM
Yes, the clamp-on ferrite beads will help with the highest frequencies, which are often most problematic.  If the center hole is large enough and the wire long enough, wrap the wire through the bead twice or three times or whatever you can fit.  Inductance (impedance at any given frequency) goes as turns squared.  Three passes through a bead is nine times better!  There are almost no cases where common-mode inductors/beads cause trouble, and many cases where they help.

When you eventually get an aluminum box for the LV electronics, that will be even better than a ground plane.  Ferrite beads are useful where cables enter and leave the box, such as on the line cord.

Grounding of the HV return pin counterpoise would be better with a separate connection down to the counterpoise, away from where the Tesla secondary wires.  That way there isn't common wiring inductance to induce voltage spikes during arcs from the top-load.  The value of HV return grounding is to avoid enough voltage difference to arc over inside the flyback, as it's not expecting significant voltage between HV return and the primary windings.  (Perhaps I'm worrying about that too much.  If you are not connecting to any of the flyback primary windings, then your custom 10-turn primary provides additional insulation.)

If you do decide for a full HV return ground, that could be done by another wire from the line ground (or water pipe) to the counterpoise.  That keeps a bit of wiring inductance between the HV return and ZVS negative.  That bit of high-frequency isolation could be valuable especially before getting a box or ground plane for the LV electronics.

Your counterpoise on concrete is sufficient that you shouldn't experience any performance difference with grounding it.  Any difference is possibly with flyback lifetime or with interference disrupting LV circuitry.  If your copper piping includes an underground run from the water meter, that should be a great ground (unless your house is on very dry ground).  It should be tied to line ground near your breaker panel.  (BTW, if the flyback is generating 70mA, that's probably much higher risk to its survival.  An analog meter in series with the HV return pin, with a cap across the meter, would be a great addition.)

Using AC meters at 50kHz may or may not give reasonable readings.  I've never tried them above 60Hz.  AC meters also don't cover the low end of their range well.  The internal magnetic force is proportional to the square of current/voltage.  Spring and/or magnetic shaping attempts to make the response closer to linear, but that's quite limited at the low end of the range.  In general, I'd use rectifiers and DC meters for AC.  1N4148 should be fine for a 90Vpp bridge.  The peak voltage maximum is 45V, enough below1N4148's 75V rating.  For feeding a DC meter, you could include a small ~0.1uF cap across the bridge to measure peak voltage, or leave the cap off to get average voltage (2/PI times the peak for a sine wave).  Either way, add ~1.2V to the meter reading for diode drop (two diodes).  Yes, without a flyback spec for output voltage, 90Vpp (45Vpeak) just insures you aren't running the flyback over-voltage.  It doesn't directly measure output unless you can come up with a calibration scheme.

The hot ZVS inductor may be due to insufficient current capability or insufficient voltage capability.  So, parallel or series may be what helps.  (By voltage-capability, I mean volt-seconds before saturation, and losses at a given flux density and frequency.)  It's hard to figure out without a scope, so hopefully you have luck finding a new one or figuring out what's wrong with the old one.  The inductor core material makes a huge difference in heating as well.

Do you know the ZVS FET part number?  Do you have the ZVS circuit to share (or did you already)?

Don't remove the center tap!  The inductor requirements are much tougher when feeding the ends.  The signal on the center tap has twice the oscillation frequency, with the shape of a full-wave-rectified sine wave.  Peak voltage is half that of the end points.  The only reason to feed the ends is when a center-tap isn't feasible, as is often the case with induction heater coils.  Even then, it's slightly better (slightly reduced total magnetics size) to generate an artificial center tap with a separate ferrite transformer.  (Common-mode power line filter chokes work well for this artificial center-tap generation.)  All my ZVS circuits are this way - artificial or real center tap.

At the end you say: "I'd be curious to hear your thoughts on driver usage and possible changes you might consider, yourself."  I'm not certain what part(s) you are calling the "driver".  In general, ZVS can be a good fit for supplies that need to ramp from 0V to some maximum voltage on their output (as in charging your MMC).  That's very similar to my Jacob's ladder ZVS, as the output is initially at high voltage, then jumps to almost zero when the arc strikes, then rises again as the arc rises.  The key is coupling factor of the output transformer (flyback in your case).  Coupling needs to be below 86%, but not way below for reasonable efficiency.  The hand-wound flyback primary is about perfect - 81% in the quickie version I made just for this thread.  My Jacob's ladder ZVS output transformer has K = 83%.  K of 86% or higher causes the ZVS to drop out in the middle output voltage range.  Too much damping (low Q) to continue.

I'm anxious to see any flyback HV return pin current measurements, or HV output voltage and spark repetition frequency, which would allow calculating average current.  Your 70mA number, if HV return is what it referred too, is a bit scary for frying the flyback.  If that is the case, two or three flybacks in parallel is probably the best option, since you won't want to reduce current and suffer the resulting reduced spark rate.
SimplePortal 2.3.6 © 2008-2014, SimplePortal