Oversized QCW

[davekni]

After experimenting with a low-frequency (100kHz) QCW coil earlier:
    https://highvoltageforum.net/index.php?topic=1268.0
Decided to try a somewhat-more-normal QCW coil.  Larger than typical.  No arc length to secondary height records here.
    5.6uH primary, 4 turns of copper foil tape.  Two interleaved 4-turn windings, in parallel except for separate halves of MMC.
    11.8mH secondary, 170mm diameter, 210mm high on 300mm high core.  255 turns + 4 spiral from 210mm to 300mm.
    0.71 coupling factor.  Ferrite floor and center post.  No potting.  Top of post is NiZn ferrite for low conductivity.
    545mm OD top load.  Ring of 19mm OD copper tube with foil/plastic center.  ~32pF total including coil capacitance.
    64.8nF MMC, 3s72p MMC of 2.7nF 1600Vdc PP caps, split into two 32.4nF halves for the two primary windings.
    H-bridge using pairs of FGH75T65SHDTLN4 IGBTs for each switch, 8 total.  Buffered GDT outputs per this thread:
    https://highvoltageforum.net/index.php?topic=2389.msg17547#msg17547
    485kHz upper pole unloaded.  440kHz at end of normal arc.  413kHz at end of ground strike arc.

H-bridge in its milk-crate home:



With MMCs added.  One folded back for this picture:



With lid and ferrites:



With coils and top load.  Control electronics live outside and aren't shown here.  Plan is to experiment with different control methods.  Initial tests are with ramped Vbus to H-bridge using one side of buck converter from my earlier low-frequency experiments.  Didn't need both interleaved halves of buck since 350A is plenty for this coil.



I'll add more pictures and scope plots in a second post to this thread.  One image to get started:



Vbus ramps to 240V for above image.  Average primary current ramps to 220A, so 345A peak of sine wave.
Peak power to H-bridge is 53kW at end of ramp.  Primary is quite low Q, so around 40% of this power is consumed in losses, mostly in primary foil winding.  That is based on matching LTSpice model to measured parameters.  Foil provided high coupling but also high losses.

Video of some arcs:
   

/>All arcs in this video are with same ramp parameters.  Amazingly varied arcs.  Anyone with QCW experience:  With more refinement of ramp parameters, is there likely a setting where arcs behave more similarly from one pulse to the next?

[davekni]

Images and associated scope traces in this second post to keep sizes reasonable.  Video linked above was made late evening when dark outside, so only indoor room lighting.  Camera fixed to somewhat-low exposure to keep arcs from being too bright.  Pictures below (and the one still above) are mid afternoon so have bright background.  Same ramp settings for all pictures and video.

All scope captures are as follows:
CH2, green, average primary current at 50A/div.  Peak is PI/2 times higher.
CH3, red, Vbus at 50V/div.  Ignore V/div label on images, as 500x probe is not a Tek version with probe identification.
CH4, cyan (light blue), average secondary current at 1A/div.  Peak is PI/2 times higher.
Notice how closely primary and secondary currents track.  Just under 50:1 at start to just over 50:1 at end.  I was initially surprised at this close tracking.  But it makes sense for upper pole with high coupling and matched frequencies.  There is rather little energy in the mutual inductance, and rather little magnetic field in the ferrites.

"Normal" arc that doesn't hit anything:  (Based on similar arcs:  485kHz start to 440kHz end.)



First frame showing main arc fully developed:



Second frame after arc ends.  I think this is real after-glow and not just an artifact of the video camera sensor, based on looking through many more frame sequences.  Often there is still a faint violet glow yet another frame later (on the night run where background isn't so bright).



Arc that hits ceiling.  I suspect that the small current dip at ~17ms is when the arc hits.



First frame capturing arc just before hitting ceiling:



Second frame with full arc:



Arc that hits grounded foil on floor adjacent couch.  Sudden drop in current at 24ms.  Vbus rings after sudden current drop due to L/C resonance of buck output inductor and capacitor.



Arc image for hitting grounded foil on floor:



And a faint afterglow.  Probably frame timing happened to be later relative to arc timing compared to earlier afterglow image.



Arc ran along ceiling mostly out of camera view.  Ended roughly over my head, as I'm sitting in the far left corner.  That prompted me to add grounded poles for shielding me in the corner for evening video.  Current waveform is odd.  Look like it strikes something (current drops) and then breaks free??  Hard to tell since end of arc is beyond image and I got only the brief sense of an arc above my head.



Image for arc extending out of view to left:



Finally, one more ground strike.  This one hit an outlet strip on top of my normal DRSSTC electronics base stored in the corner.  I'd thought of QCW arcs as not all that dangerous.  But resulting char marks on end of outlet strip (and above arc over my head) suggest increased caution.



First frame with arc just hitting outlet strip:



Second frame with fully developed arc and bright glow from outlet strip strike point.  BTW, interesting that it hit the end of the strip which is plastic rather than an exposed hole for ground.  Strip is plugged in, so grounded.  Arc must have gone through joint in plastic or just coupled through plastic capacitance.



End of outlet strip that was hit:



Again, anyone have insights on why all the variability given identical ramp for all posted images and video?

[Rafft]

Congratulations!

Arcs so loopy loop 

Edit:
Still wondering how they become so loopy.

Maybe its because of (very) low impedance primary(and high K) + high inductance sec with very small topload.

PRI(low inductance) and high C
SEC (high inductance?) and small C(topload)

Or maybe it has something to do with the ferrites?

High-ish ramp start ~ end frequency difference (~40khz) clearly indicates low topload C.

[Uspring]

A very cool coil!

Notice how closely primary and secondary currents track.  Just under 50:1 at start to just over 50:1 at end.  I was initially surprised at this close tracking.  But it makes sense for upper pole with high coupling and matched frequencies.

The set of equations I'm using also suggest that tracking. Detuning effects from the arcs don't affect coil operation much, since the operating frequency is always far away from the secondary resonance due to the huge coupling.
But the equations predict a primary - secondary current ratio almost 3 times larger. Could it be that this has something todo with the splitting of the primary?

I find the curvy arcs rather pretty. But you sure have a hard time aiming them. You do have quite long burst times, much longer than the often used quarter wave line voltage ramps. That might make the arc particularly sensitive to the surrounding fields. I'd expect the arcs to follow the local field, i.e. the sum of the top load field and the arc field. Since the arc tip is charged through a long resistance path similar to a transmission line, significant phase shifts between the local field and the top load fields might occur. That might cause an attraction between the arc tip charges and the top load and lead to a bend back.

[Mads Barnkob]

Very cool coil and as with every QCW coil I see, I feel I am missing out from not having build one yet

I think the very curvy arcs are more to do with topload field shaping, or more lack of it. Its a very wide and very thin ring. Its simply not powerful enough to keep the arcs from going back towards itself.

[MRMILSTAR]

Very nice! I also think I'm missing out having never built one.

[davekni]

Maybe its because of (very) low impedance primary(and high K) + high inductance sec with very small topload.

PRI(low inductance) and high C
SEC (high inductance?) and small C(topload)

Or maybe it has something to do with the ferrites?

High-ish ramp start ~ end frequency difference (~40khz) clearly indicates low topload C.

I think 545mm diameter top load is significantly larger than any other QCW coils shown on the forum.  Any larger examples?
Jan's QCW coil uses ferrites.  I can't think of any physics reason why ferrite use matters for arc shape.

But the equations predict a primary - secondary current ratio almost 3 times larger. Could it be that this has something todo with the splitting of the primary?

That seems odd.  My simple LTSpice simulation models show roughly 50:1, so match well.  I keep tweaking parameters trying to match measured waveforms, but they are all around 50:1 current ratio.  Here's my latest version:

This is trying to match arc that doesn't hit anything.  If this version is accurate, just over half of the power is dissipated in primary coil losses, even more than my earlier version that showed 40%.

I find the curvy arcs rather pretty. But you sure have a hard time aiming them.

Yes, curved arcs are interesting.  Would be great if I could find some ramp parameter sets that allowed intentional selection between straight and curved.

You do have quite long burst times, much longer than the often used quarter wave line voltage ramps.

Yes, even the curved arcs are comparatively straight if aborted earlier.  I'd modeled my ramp from Jan's 27ms ramp (then tweaked a bit longer):
   https://highvoltageforum.net/index.php?topic=1073.msg7668#msg7668
Though his does seem to be longer than others.  Phoenix post shows 18ms not counting end decay:
    https://highvoltageforum.net/index.php?topic=1213.msg8861#msg8861
My initial conservative testing was around 18ms too.  However, if I ramped fast enough to get long arcs, I experienced too much branching.  Perhaps I need to explore shorter bursts more thoroughly.

I think the very curvy arcs are more to do with topload field shaping, or more lack of it. Its a very wide and very thin ring. Its simply not powerful enough to keep the arcs from going back towards itself.

Yes, top load is wide and thin.  Wide based on previous discussions suggesting that a large QCW top load would help start the arc out straight by repelling it upwards away from ground (besides minimizing frequency shift).  I'll need to think more about the thin part.  Makes a higher local field strength near top load rim, but the arc is never that close to rim.  I'd thought the only significant requirement for thickness was to avoid breakout along rim.  Since QCW coils run at relatively low top load voltage, 19mm diameter rim is sufficient for that need.  At some point I'll make a thicker (and bit smaller diameter to keep frequency constant) top load for comparison.

[Uspring]

Thank you for the schematic.
I've checked my equations and found a problem with them. I started from energy stored in the tanks and have now realised, that this is a wildly varying quantity due to the large coupling. Also the energy content is a dubious quantity, since much of the field is shared by the tanks, so that energy cannot be attributed separately to them. Anyway, after corrections I can confirm your values. (It's hard to argue against measurements)

I'm wondering a bit about the energy loss in the primary. I would have put the resistance causing the loss in series with the coil. That is because the coil voltage is the sum of the drop along the resistance and the induced voltage. Did you make a primary Q measurement before installing the secondary?
For a 50% loss the series copper resistance would have to be about 0.4 ohms. That seems much more, than can be accounted for by skin depth of your copper foil, assuming the current is distributed evenly across its width (maybe a cm or so). MMC ESR would have to be added to this. But perhaps current density varies across the width of the band. I haven't looked in detail at this. Did you consider using stranded wire?

Regarding curvy arcs: Perhaps Mads is right, that the top loads field doesn't matter much. If you're interested in that, you could try a very long breakout rod (perhaps 1m) or place it off center or point it at an angle. Adding extra capacitance to the top likely won't affect tuning very much, since the coupling is so large.

[davekni]

I'm wondering a bit about the energy loss in the primary. I would have put the resistance causing the loss in series with the coil. That is because the coil voltage is the sum of the drop along the resistance and the induced voltage. Did you make a primary Q measurement before installing the secondary?

I believe almost all of the primary loss is proximity-effect.  I've found that parallel resistance more closely models proximity-effect than does series resistance, though neither is perfect.  Proximity effect is especially high at second pole where primary current is counter magnetic field from secondary current. Copper foil tape is 25mm x 50um at a roughly 38mm pitch.  Foil from the two windings overlap (with polyimide tape insulation between).  The goal is to block all magnetic field penetration through primary winding, as it is solid copper from a side perspective.  That is what makes coupling high.  Blocking field also makes for a lot of eddy currents in foil.  100um foil would be better, but I didn't have much and it is much harder to wrinkle into proper conical shape.  The bottom turn is double-layer 50um tape since FEMM indicated maximum loss there.
I think I measured primary Q once by itself, but don't recall the value.  It is not very useful since frequency is much lower and eddy current loss is higher when "fighting" secondary current at upper pole.  I did measure upper-pole system Q at low power, about 17.  I believe most of that is primary.  MMC ESR is insignificant for this build.

Did you consider using stranded wire?

Yes, and I may go that route if I want to run above 1 pulse per second.  It lowers coupling which lowers upper pole frequency.  Would need to wind new secondary if I want to maintain frequency, or use a lower frequency.  For my existing experiments I have more power available.  Decided to stop at 53kW peak as arcs are already problematic in my living room.  Too much work to set up outside, at least until more packaged and not an experiment.  Then I can to to 120kW or so.

Regarding curvy arcs: Perhaps Mads is right, that the top loads field doesn't matter much. If you're interested in that, you could try a very long breakout rod (perhaps 1m) or place it off center or point it at an angle. Adding extra capacitance to the top likely won't affect tuning very much, since the coupling is so large.

I thought Mads was saying top load field does matter and that mine was not good.  If I move breakout point a significant distance, needs to be outside or at low power.  I already experimented some with breakout point height, but all centered above top load center.  Low might work better if I make a more complex ramp.  Low (close to top load) requires more voltage to start arc.  That starting voltage then causes arc to grow too rapidly and branch.  Thus I need a brief higher voltage, then a voltage drop, and then a ramp up.  Hope to try that some time.  Next major step is too use my optically-isolated probes to measure charge on top.  That presumes I'm successful repairing my long-record-length scope (or purchase another scope).

[Mads Barnkob]

Regarding curvy arcs: Perhaps Mads is right, that the top loads field doesn't matter much. If you're interested in that, you could try a very long breakout rod (perhaps 1m) or place it off center or point it at an angle. Adding extra capacitance to the top likely won't affect tuning very much, since the coupling is so large.

I thought Mads was saying top load field does matter and that mine was not good.  If I move breakout point a significant distance, needs to be outside or at low power.  I already experimented some with breakout point height, but all centered above top load center.  Low might work better if I make a more complex ramp.  Low (close to top load) requires more voltage to start arc.  That starting voltage then causes arc to grow too rapidly and branch.  Thus I need a brief higher voltage, then a voltage drop, and then a ramp up.  Hope to try that some time.  Next major step is too use my optically-isolated probes to measure charge on top.  That presumes I'm successful repairing my long-record-length scope (or purchase another scope).

I do not think a longer break out point would change much, as the sparks are just LONG! You are running out of ceiling height quickly.

I think its a large minor diameter on the toroid that is needed, to form a larger and stronger field, that can reject the spark from going sharply down-/inwards.

[Uspring]

I've found that parallel resistance more closely models proximity-effect than does series resistance, though neither is perfect.

Yes, compared to the serial circuit, it adds a frequency dependent term to the serial resistance, increasing with higher frequency. That is probably what you want as proximity and skin effect show the same behaviour.

The goal is to block all magnetic field penetration through primary winding, as it is solid copper from a side perspective.  That is what makes coupling high.

That is an interesting way to view coupling. Formally it is the correlation coefficient between the fields generated by the coils. It is 1, if the fields are multiples of each other in every point of space. So it is basically a volume effect, with the largest weight, where the fields are highest.

Wrt the curvy arcs: Whatever the top load field may be, I would expect the arcs to head away from the top load, making them more or less straight. Some of them seem to have an almost constant radius of curvature. Curious.

You are running out of ceiling height quickly.

You don't have an oversized coil but an undersized lab space. Coils can never be big enough

[davekni]

I think its a large minor diameter on the toroid that is needed, to form a larger and stronger field, that can reject the spark from going sharply down-/inwards.

Hope to try that eventually.  Though you are accurate that my ceiling height is quite limiting.  Once I'm done experimenting, plan to package control electronics for a more portable unit.  Then outdoor testing becomes more practical.  Before that I plan to experiment with audio modulation, pulse-skip ramping, and phase-shift ramping.  (Comments on the forum suggest pulse-skipping ramp will not be smooth enough.  Probably true.  But I want to see for myself.  I'm not aware of anyone else actually testing pulse-skip.)  Phase-shift ramping will limit power and duration for my given IGBT H-bridge due to hard off switching increasing IGBT power dissipation.  However, I'm not sure the theoretical 120kW of this initial buck-converter version will actually be useful.  Definitely not useful in my constrained indoor setting.

That is an interesting way to view coupling. Formally it is the correlation coefficient between the fields generated by the coils. It is 1, if the fields are multiples of each other in every point of space. So it is basically a volume effect, with the largest weight, where the fields are highest.

Yes, the volume effect makes sense.  Blocking field from passing through primary coil forces more of it through the same volume as secondary, rather than looping locally around primary turns.  That was exactly my goal.  This is all based on FEMM simulations.  However, the gain in coupling is relatively small for a large decrease in Q.  Ferrite core keeps coupling reasonably high even with a wire or tubing primary.  In hindsight I would have made a somewhat more standard primary.  May decide to make such a new primary someday.

[johnnyzoo]

I think its a large minor diameter on the toroid that is needed, to form a larger and stronger field, that can reject the spark from going sharply down-/inwards.

Hope to try that eventually.  Though you are accurate that my ceiling height is quite limiting.  Once I'm done experimenting, plan to package control electronics for a more portable unit.  Then outdoor testing becomes more practical.  Before that I plan to experiment with audio modulation, pulse-skip ramping, and phase-shift ramping.  (Comments on the forum suggest pulse-skipping ramp will not be smooth enough.  Probably true.  But I want to see for myself.  I'm not aware of anyone else actually testing pulse-skip.)  Phase-shift ramping will limit power and duration for my given IGBT H-bridge due to hard off switching increasing IGBT power dissipation.  However, I'm not sure the theoretical 120kW of this initial buck-converter version will actually be useful.  Definitely not useful in my constrained indoor setting.

Pulse skip modulation sounds like a very elegant solution as it preserves soft-switching and doesn't need many extra components. Not sure how hard it is to implement but it's something I might want to try if I were to build a Tesla coil now.

Looking forward to see your results.

[davekni]

Pulse skip modulation sounds like a very elegant solution as it preserves soft-switching and doesn't need many extra components. Not sure how hard it is to implement but it's something I might want to try if I were to build a Tesla coil now.

Looking forward to see your results.

That's next on my list, so shouldn't be too long.

Finished a bunch of top load, arc, and secondary top charge measurements using my optical-fiber isolated scope probes.  Lots more arc images and corresponding arc voltage and current and power data in this thread on DRSSTC top measurement:
    https://highvoltageforum.net/index.php?topic=1950.msg17699#msg17699

[davekni]

Ran high-current turn-off tests on IGBTs used in this H-bridge, posted in this thread:
    https://highvoltageforum.net/index.php?topic=2498.msg18359#msg18359
Added R+C snubbers to H-bridge based on above results.  Using data-sheet transient thermal resistance and above measured Eoff and estimated conduction loss, calculated peak die temperature would be a bit over 200C for my QCW coil running in phase-shift mode.  Decided to test it anyway.  Expected there would be enough margin in data sheet values, and that silicon devices generally don't die immediately before at least 250C.  Failed sooner than I'd expected.  Worked great at 300Vbus.  Was hoping to get to 400Vbus.  However, failed on second shot at 360Vbus.  Three sequential frames from video:



Rolling-shutter of video camera causes flash to show up in only bottom portion of first frame.

Scope capture of failure below:
    Ch1 (black) is average secondary current at 1A/div (2V/A).
    Ch2 (green) is Vbus at 50V/div, position at -4div (bottom of image).
    Ch3 (red) is a debug output from FPGA indicating intended drive voltage as a fraction of Vbus.  Ignore aliasing.
    Ch4 (cyan) is average primary (H-bridge output) current at 50A/div (50A/V).



Notice in second zoomed-in plot, Vbus goes low for ~230us then rings back high.  I'm measuring Vbus at H-bridge.  There's roughly 1m of wire from bulk caps to H-bridge.  IGBTs fail shorted initially.  Energy from 40,000uF bulk caps is sufficient to explode IGBTs and open the short after 230us.  All 4 IGBTs of one half-bridge are split.  Haven't tested other half-bridge yet.

"Explosion" showing at bottom of video frames isn't IGBTs themselves.  Instead, it is a clip lead I'm using to bypass buck converter, feeding bulk caps directly to H-bridge.  The extreme high current caused a bit of the clip to melt and blow apart.

Also, arcs were much straighter (much less curved) on average today.  Expect that is related to somewhat higher temperature and humidity, as there is little intentional difference in ramp characteristics.  (Only a difference in how ramps is generated, by phase-shift instead of by buck-converter feeding Vbus.)

BTW, pulse-skip didn't perform as well.  Might be better if my primary coil Q was not so low.  Was planning to document today in comparison with phase-shift.  Needs to wait for H-bridge repair now.

[Mike]

BTW, pulse-skip didn't perform as well.  Might be better if my primary coil Q was not so low.  Was planning to document today in comparison with phase-shift.  Needs to wait for H-bridge repair now.

This is something I had wanted to play with using my freewheeling controller. I was abusing the current limiting to allow the primary to be much further out of tune with the secondary than usual so that I could allow more streamer growth, the limit was transferring enough energy to allow secondary breakout to start. Starting so far out of tune means you have a very high Q but the constant bus voltage limits how small a current step you can make on the primary. That said the energy transfer is also very poor even after breakout starts so there might be a nice operating point where you can slowly ramp up the secondary voltage, especially if you have a variable current limit which is next on my list to implement, hopefully by PWMing the interrupt line and using a filter for the current limit and peak detect for the interrupt line. Even without that I was seeing noticeably less branching.

[曹靖]

As mentioned in my previous post, in the winter experiment, my QCWDRSSTC arc curved more towards the ground, and after I lit the fireplace and heated the indoor temperature, the arc returned to a straight line. Although I used a step-down converter instead of phase shifting, it did not affect the arc performance. My experiment showed that the indoor temperature had a greater impact on the arc shape. This also showed the same characteristics in your experiment - the rising temperature in summer makes your arc more straight

[davekni]

especially if you have a variable current limit which is next on my list to implement

That's an interesting idea.  Easy way to convert a freewheeling driver into a skip-pulse ramped driver.

As mentioned in my previous post, in the winter experiment, my QCWDRSSTC arc curved more towards the ground, and after I lit the fireplace and heated the indoor temperature, the arc returned to a straight line.

Yes, I remember you saying that.  I'm still amazed at how such a small temperature difference makes a big difference in arc curvature.  My first thought was that humidity might be more significant than temperature.  However, in your case of heating in winter, humidity would be lower when warm.  In my case of comparing April run with June run, inside temperature and humidity were both a little higher for the June run making straighter arcs.
Wish we had more understanding of the physics causing arc path changes.

[Uspring]

I'm also puzzled about the temperature and/or humidity depencies of the arc curvatures. Particularly, because these quantities are scalars and don't directly imply vectorial effects, such as arc directions. But they might modulate other effects. I will speculate a bit:
Candidates for arc direction causes are the ambient electric and magnetic fields. Slowly growing arcs are particularly susceptible to these.Consider e.g. a 2 m long arc grown in 20 ms, such as from Davids coil. It lengthens at about 100 m/s or about 0.1 mm per half cycle of its frequency.
Now suppose, the arc bends with a 1 m radius of curvature. That implies, that for each 0.1 mm step, the angle of the arcs direction has to change by 0.1mm / 1 m = 0.0001 radians. The direction of the electric field near the tip is straight ahead, changed a tiny bit by some externally caused surrounding field. The straight ahead compnent of the field is in magnitude near the breakdown field of air, i.e. 30 kV/cm. But to change this field in direction by 0.0001 radians, an external field of strength of only 0.0001*30 kV/cm = 3 V/cm is necessary. This kind of QCW arc is really a field probe.

The strongest fields surrounding the TC are the top load electric field and the primary and secondary coil magnetic field. The primary and secondary coil magnetic field lines can only cause azimuthal or spiralling direction components, which don't seem to happen. From the videos it is a bit difficult to tell.

The top load electric field is definitely strong enough to influence the direction of the arc, but shouldn't it cause the arc to seek a direction away from it?
I've run a simulation of the arc with my arc model. In the diagrams below the voltages of the top load (red) and along the arc (green) are displayed. In the first diagram the arc voltage near the breakout point is shown. In the second diagram the arc voltage near the tiip is displayed.





The arc resembles a phase shifting transmission line. The phase shift in the second diagram between the voltages is near to 180 degrees, so there the top load will bend the arc towards it. As from the statements above, this tendency is small and will cause only a slight bending, the major directional influence still being the arcs charge right behind the tip, which will tend to keep it straight.

Davids setup is nearly axially symmetric, so the arc has to have a reason to go anywhere except straight up. The effect I've described works only, if the arc is already off axis, since an external field can only bend the arc, if it is not parallel to it. I believe this might be caused by the initial bump in the primary current as shown here: https://highvoltageforum.net/index.php?topic=1950.msg17719#msg17719
This is different than in the diagram https://highvoltageforum.net/index.php?topic=2397.msg18480#msg18480, where the bump is much smaller. The bump might be causing a jump start of the arc, which will give it a more random initial direction from where it can bend from. This is very speculative, but might explain the observed straighter arcs seen with the new driver.
Another difference might arise from the ambient temperature. A higher temperature will lower the air density and thus reduce the breakdown voltage. That will change some arc properties but I'm not sure how that affects arc curvature.

Edit: replaced the first diagram with a corrected version (the original one was taken with a wrong time range)

[davekni]

I've run a simulation of the arc with my arc model. In the diagrams below the voltages of the top load (red) and along the arc (green) are displayed. In the first diagram the arc voltage near the breakout point is shown. In the second diagram the arc voltage near the tiip is displayed.

Your model of R/C phase shift makes sense, but I wonder if modeled R is too high?  Primary and secondary current drop (Q drops) very rapidly when arc hits a grounded object, as shown in previous scope plots.  If arc resistance is high towards the end of the tip, it must drop very quickly as tip hits electrical ground.  Perhaps such rapid resistance drop is reasonable, as such occurs in a normal spark gap.  So perhaps the rapid Q decrease does not provide any real information about pre-contact arc resistance.

I believe this might be caused by the initial bump in the primary current as shown here: https://highvoltageforum.net/index.php?topic=1950.msg17719#msg17719
This is different than in the diagram https://highvoltageforum.net/index.php?topic=2397.msg18480#msg18480, where the bump is much smaller. The bump might be causing a jump start of the arc, which will give it a more random initial direction from where it can bend from. This is very speculative, but might explain the observed straighter arcs seen with the new driver.
Another difference might arise from the ambient temperature. A higher temperature will lower the air density and thus reduce the breakdown voltage. That will change some arc properties but I'm not sure how that affects arc curvature.

That seemed like a good guess.  However, now that my coil is repaired (much more work than I'd hoped given extensive damage to one half-bridge), I experimented with initial ramp.  Back to buck-converter operation inside, with warmer June temperature, arcs are much straighter as with phase-shift before failure.  I tuned to make more of that initial hump down to none.  Made very little difference.  Then tested outside, at roughly same or slightly higher temperature, around 23C.  Much more curved and a bit more branched.  Occasional very straight arcs, but mostly curved down to ground.  I'll post more on this and comparison with pulse-skip once I have time to edit video and grab snapshots from that.

Davids setup is nearly axially symmetric, so the arc has to have a reason to go anywhere except straight up.

I also experimented a bit with precision of centering breakout point.  Made little if any difference.  Even when a bit off-center, curvature direction is quite random, not always towards or away from closest top load edge.

Yet another speculation:  I've noticed that several other QCW coilers report the same behavior I see: More curved arcs outside compared to inside.  Perhaps the key factor is uniformity of air temperature and humidity.  Especially in urban environments such as mine with buildings and trees round, even subtle breeze has randomly varying velocity (speed and direction).  Surface temperature varies due to sun-heated asphalt to lighter concrete to repetitively-cool damp lawns.  Even if arc growth is fast compared to wind speed, wind and surface temperature may leave significant non-uniformity in air temperature.  That may cause arc bending towards pockets of warmer air where breakdown voltage is slightly lower.

[Uspring]

Your model of R/C phase shift makes sense, but I wonder if modeled R is too high?

When I decrease the R in the model, the phase shift for each R/C circuit decreases but also its voltage attenuation. In effect the arc lengthens and that almost perfectly cancels the reduced phase shift. The total shift remains about the same.

That seemed like a good guess.  However, now that my coil is repaired (much more work than I'd hoped given extensive damage to one half-bridge), I experimented with initial ramp.  Back to buck-converter operation inside, with warmer June temperature, arcs are much straighter as with phase-shift before failure.  I tuned to make more of that initial hump down to none.  Made very little difference.  Then tested outside, at roughly same or slightly higher temperature, around 23C.  Much more curved and a bit more branched.  Occasional very straight arcs, but mostly curved down to ground.

and

I also experimented a bit with precision of centering breakout point.  Made little if any difference.  Even when a bit off-center, curvature direction is quite random, not always towards or away from closest top load edge.

Thank you for running an experimental check on these ideas. Doesn't look good for them.

Yet another speculation:  I've noticed that several other QCW coilers report the same behavior I see: More curved arcs outside compared to inside.  Perhaps the key factor is uniformity of air temperature and humidity.  Especially in urban environments such as mine with buildings and trees round, even subtle breeze has randomly varying velocity (speed and direction).  Surface temperature varies due to sun-heated asphalt to lighter concrete to repetitively-cool damp lawns.  Even if arc growth is fast compared to wind speed, wind and surface temperature may leave significant non-uniformity in air temperature.  That may cause arc bending towards pockets of warmer air where breakdown voltage is slightly lower.

Qualitatively, temperature gradients make a lot of sense, since they are also density gradients which affect breakdown voltages. I've tried to get some numbers on the magnitude of the effect and arrived at radii of curvature to be twice the inverse of the relative temperature gradient (if the gradient is orthogonal to the field). So if, e.g., the temperature would rise by 3 C / m, that would amount to about a 1% change in density per meter. Twice the inverse would then be 200 m, which is a lot more than you see. The assumption, that goes into the calculation is, that the arc grows into the direction of the biggest average E/ρ (ρ being the density of the air).