QCW with replaceable ferrite-core primary

[Uspring]

Thank you for checking out the idea that magnetic fields might influence arc shape. I've noticed from other pictures, that an increasing length of burst tends to curl up or meander the arcs near the bottom. Your images seem to point in the same direction. I'm quite mystified about the reason for this, since the effect of both magnetic and electric field of an arc on itself should straighten it out.
I still have a (remote) hope, that low frequencies could allow for branchless arcs, although the mechanism I'm thinking of requires a growth speed proportional to the frequency, i.e. a constant mm/cycle value. That would require even longer bursts at the frequencies you use. But then, these bursts cause curling or meandering.

BTW: You seem to operate the coil much like an SSTC with ZCS, like discussed in another thread.

[davekni]

Uspring,

Yes, you are right, it is running just like an SSTC with ZCS.  I hadn't realized that until this morning after your comment.  My goal had been to run upper pole with ZCS by having high coupling.  Only now I realize that's what SSTCs do.  With H-Bridge primary drive being low impedance, SSTCs are effectively running upper-pole.  (At least for half-bridge designs that have capacitor(s) in the primary, so there actually are two poles.)  The only difference (besides higher coupling) of my design is that I'm using primary current feedback, not secondary as SSTCs do.

Unfortunately I don't have a high-speed camera.  Would love to have one.  I did explore arc development the only way that comes to mind:  keeping ramp parameters stable, but changing the ramp end time.  Each shot is different, but I get a general idea of what it looks like initially, a bit later, etc.  Yes, the beginning is always thinner and much less jagged.  The zigzag develops later as the ramp continues.  The beginning often has small branches that seem to sometimes extinguish later.

Yes, I'd hoped that stretching ramp times by 1/f or even 1/f^2 would make straight arcs.  Exploring that thought is the #1 reason I built this coil.  I still plan to finish it's high-power drive in spite of (I think) disproving that hope.  The first (slowest) ramp that I posted lasts 25ms.  The final version should be capable of 100ms ramps at higher power.  However, that lower-power 25ms ramp will be much like the first quarter of a high-power 100ms ramp.  I hope it will be interesting arcs to display, but have little expectation of achieving straight arcs now.

My guess as to the growth of large zigzags at the bottom is air turbulence caused by the column of hot low-density air surrounded by cold high-density air.  Just had an idea if I feel ambitious enough - wonder if I could talk Portland State University into letting me run this in their micro-gravity drop tower.  That would be fun and a bunch of work.

[Uspring]

Forum member Phoenix sent me a 960 fps video taken with his QCW coil. https://drive.google.com/file/d/1FH6-gjzs18cjbQTu9FdJ7u1WuvL1cjOf/view
A thank you from me for his permission to post the link. It is full of interesting details. Right at the start there is s small branch to the left, which vanishes again. Also some wiggles can be seen at the bottom of the arc, which become larger as time progresses. Near the top of the arc another small branch angles to the left, while the main arc turns right. I believe that to be due to the repulsion of the space charges of the branches. Toward the end of the sequence the wiggles at the bottom of the arcs become even larger.

My guess as to the growth of large zigzags at the bottom is air turbulence caused by the column of hot low-density air surrounded by cold high-density air.

A very plausible conjecture. I've been thinking of another cause, i.e. the mutual repulsion of the space charges that make up the arc. Consider e.g. a loose string. When you pull on it, it will straighten and will wrinkle again as you loosen it.The repulsion in the arc cause an internal pressure or in analogy with the string a negative tension. The (negative) force developed is of the order

1/(4*pi*eps0) * rho^2,

where rho is the charge per unit length of the arc. The above is basically the Coulomb force equation. For typical charge densities of arcs, around 1 uC/m, the force comes out to be about 0.01N. That is not very much, but the mass of air, that is pushed around is only about 1 mg. That would be 1 cm^3 of air. So considerable movement is possible within the duration of the arc.
Similar forces arise due to convection. A very hot cubic cm of air experiences an upward force of also 0.01N.

[davekni]

Your space-charge theory is looking much more likely than buoyancy.  I'm coming up with 0.00001N for buoyancy for 1cc of very-hot air, three orders of magnitude less than your space-charge estimate (which seems reasonable).  Space charge seems to set the upper limit for QCW ramp time.  My low-frequency coil requires more voltage for a given arc current, which increases space-charge issues.  It will still be interesting to see what effect I get from throwing more power at it eventually.  Need to order more SiC diodes to complete my ramp-generator buck converter pair first, and build a new H-Bridge.

The QCW high-speed video is much appreciated!  Yes, the bottom develops zigzags by the end that look much like what I get.

[Uspring]

You're right about the bouyancy force.
It'll be interesting to so what happens, when you lengthen burst and increase power.

Mads wrote:

Number of zig-zag edges to spark length seems to be consistent when looking at 170-400kHz QCW coils, it is just the amplitude of the zig-zags that is changing. It is however harder to count the zig-zags on the very bright and straight arcs of a high frequency coil, so its an observation with a lot of uncertainty.

I believe the zig zags to originate from tiny branches, which vanish again quickly, so they aren't visible, except maybe in high speed videos. With longer burst times the wiggles then increase in size.

[Mads Barnkob]

You're right about the bouyancy force.
It'll be interesting to so what happens, when you lengthen burst and increase power.

Mads wrote:

Number of zig-zag edges to spark length seems to be consistent when looking at 170-400kHz QCW coils, it is just the amplitude of the zig-zags that is changing. It is however harder to count the zig-zags on the very bright and straight arcs of a high frequency coil, so its an observation with a lot of uncertainty.

I believe the zig zags to originate from tiny branches, which vanish again quickly, so they aren't visible, except maybe in high speed videos. With longer burst times the wiggles then increase in size.

Yeah, I agree on that. That is the same for conventional DRSSTC sparks. Here is one from my small DRSSTC 2


I got some 1000FPS 720p recordings of my VTTC, but not even 1000FPS is enough to show growth details. I have not yet uploaded them to youtube as I was not really satisfied with the results, I will try to do some higher FPS recordings another time. I can do the following:

1280 x 720 @1,000fps 
1280 x 360 @2,000fps 
1280 x 180 @4,000fps 
1280 x 60  @8,000fps
640 x 360 @4,000fps 
640 x 180 @8,000fps 
640 x 120 @10,000fps 
640 x 60   @20,000fps

[davekni]

"I believe the zig zags to originate from tiny branches, which vanish again quickly, so they aren't visible, except maybe in high speed videos. With longer burst times the wiggles then increase in size."

Could it be the other way around?  Could the zigzags form, making points that start branches?

[Uspring]

Mads photo sure makes a point, that wiggles and branches correlate.

Could the zigzags form, making points that start branches?

Quite possibly. Phoenixs video shows branches to originate at the tip of the arc. So I think of branches and bends to begin simultaneously. A possible cause might be 2  free and separate electrons near the tip, which form avalanches, which both then connect to the main arc channel. But there could be branches sprouting sideways from a curved part of the arc later, i.e. without the tip being involved. Not that I have seen enough high speed videos to judge.

[davekni]

Finally "finished" my planned high-power drive for this low-frequency QCW coil: dual buck-converter feeding CM600DY-13T bridge.  Moved the coil outside for testing last Friday.  Even at low power, behavior outside is very different than inside on my table.  Most arcs curve towards the ground.  Higher power for longer ramps didn't help.

Here are a couple videos.  I've edited out gaps between firings.  These cover a random range of ramp shapes.  I'm playing with four knobs adjusting a quadratic ramp generator.  Parameters are initial PWM duty cycle, linear ramp rate, quadratic rate, and end time.  Longest ramps here are 55ms with ending power about 100kW: ~320V at ~330Adc, ~530A primary coil peak current.  (Buck converter and bridge are capable of over twice this power, but the primary winding impedance is too high.  Power into the buck converter comes from a 0.2F 450V capacitor bank charged to 400V.  I'm wearing hearing protection now in case of a failure:)

/>

/>
Had a couple strikes to the secondary at the end of the afternoon video, making shorted turns.  Removed a few turns and patched, then added a ground strike ring before the evening video.  The evening (darker) video shows more of the internal discharge problems.  Potting ends a few cm above the bottom.  I though secondary field would be low enough there.  Not the case   There was charring along the bottom edge of potting.  Yellow glow flashes can be seen at the bottom of the coil.


The afternoon video had a 0.5mm copper wire breakout point which was slowly melting down. For the evening I switched to 0.5mm tungsten rod.  Instead of melting, it make bright white flashes from incandescence as in this frame:


No plans to extend this low-frequency QCW experiment.  Perhaps some day the dual buck converter can power a conventional QCW coil.

BTW, H-Bridge construction is described on this low-inductance-bridge thread:
https://highvoltageforum.net/index.php?topic=1324.msg11785#msg11785

[Steve Ward]

Those are some crazy sparks!  Might i suggest carbon "gouging" rod for breakout point?  I'd rate it as "bright, but much less bright than tungsten" and doesn't melt like metals.

Since its already a LF coil, could you try a bigger toroid to push the sparks up?  That seems to be my impression of how to control the spark direction.

Congrats on no transistor explosions, that's a tremendous energy reserve!

[davekni]

Steve, thank you for the compliment!  Yes, I just purchased a few 6mm diameter carbon rods after seeing that you started using those for your HFSSTC.  I'm trying to make a fixed ISM-frequency HFSSTC, at lower power than yours, for 120VAC 15A circuit.

Yes, a larger top-load would likely help keep arcs vertical a bit longer.  Then I'd be down to the ~80kHz of my larger DRSSTC.  The idea of this low-frequency QCW was to test a speculation that if ramp time were stretched in proportion to 1/frequency it might behave similarly to high-frequency QCW coils.  Answer is clearly "no".  Uspring and I have a long discussion here about the physics and I think a reasonable understanding as to why by the end.

The big 0.2F capacitor bank is eventually for my DRSSTC.  I'm working on a crazy experiment for long arcs.  448 BTB16 TRIACS will add MMC capacitance as the arc grows, just over doubling capacitance in 6 progressive steps, with switching timed at zero-crossings.

[davekni]

Since I received an email question about the buck-converter (ramp generator) portion of this final high-power drive, decided to post some info here.  Unfortunately, I don't have any consolidated schematic for the whole system or even the buck converter.  So I sketched the power stage below.  It is a dual buck converter ran 180 out-of-phase to minimize ripple currents.  The "m=6" etc. labels indicate how many devices are in parallel.  Easier than drawing multiple devices:


Gate drive is from small driver boards that are much like half of a UD2.7 driver (without the current limit portion).  I hand-patched them to run both outputs in parallel feeding one side of the buck.  Two identical boards are used.  The comparitor input is fed from a phototransistor fiber receiver.  Here is the hacked gate driver schematic:


BTW, a third copy of this board, unmodified, is used for the primary QCW H-Bridge drive.

I'd designed the buck converter for ramping to 600A DC (for 940A peak primary coil current).  This low-frequency QCW coil primary impedance ended up too high, so DC current didn't get above 330A.

As with my H-Bridge designs, the key to this converter is low-parasitic-inductance.  The power stage is double-sided copper clad with ground (V-_Input) on the back side, V+_Input/Output in the center top rectangle, and switched nodes (IGBT collectors) on the two outer rectangles of top-side copper.




This layout is similar (in that it uses overlapping parallel planes) to my H-Bridge example shown here:
https://highvoltageforum.net/index.php?topic=1324.msg11785#msg11785

Gate drive is also on double-sided copper for interconnect, with 50-pin ribbon cable to the driver boards.  Ribbon cable alternates between emitter and gate, so is 25 wire pairs in parallel, to minimize parasitic inductance.

Finally, here is an overview of the setup all connected.  I'd tried an indoor run initially, before quickly deciding that outside would be more appropriate:


The buck converter power stage is at the top of the image, with the 400A 64uH inductors below.  The inductors are wound with 4 strands of 14AWG wire.  Cores are powdered iron bar cut into four sections each, spaced with 2mm gaps between sections.  The small inductors with Hall-sensors are a bit hard to see, hanging in the air by their wires.  These are home-made 450A current sensors used to keep the two halves of the buck converter balanced and to abort if current gets too high.

Control of the buck converter with ramp generation is done with an FPGA on a separate board not shown in the image.  For FPGAs, I use old head driver boards from work.  FPGA LVDS input structures are used as comparitors to monitor current (voltage from Hall sensors).  FPGA outputs drive LEDs fiber-coupled to the gate driver boards.  Buck chop frequency is fixed 40kHz, except for very low and very high duty cycles where cycles are skipped to maintain minimum on and off times.