Oversized QCW

[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).