QCW with replaceable ferrite-core primary

[davekni]

Inspired by Jan's impressive ferrite-core QCW coil here:
https://highvoltageforum.net/index.php?topic=1073.0

and his wishing that the primary could be adjusted, I'm starting a project to make a QCW with replaceable ferrite-core primary.  The idea is to pot the inside of the secondary with a cavity for the primary.  However, a plain cavity would form corona discharge between the inside surface of the cavity and the primary.  So, here's the secondary form and cavity I plan to pot inside in a couple days:





The cavity is a Faraday cage with no closed loops.  All the wires connect at only the top.  Wires will all be inside the potting material.  At the bottom, I'll connect ONE of the wires to ground.  That way the internal electric field of the secondary will return to the Faraday cage within potting material.  Internal electric field around the primary will be due to only the primary voltage, low enough to avoid corona.  Magnetic field will be free to pass between Faraday cage wires.

Anyone see any issues before I mix and pour epoxy?  I will be using vacuum to remove as much air as possible, per what I've learned on this forum.

This is the first step of a long project.  Haven't built any of the electronics yet.  My aim is to explore lower frequency and longer ramp times than typical for QCW, to see if slow ramps can compensate for lower frequency.  Somewhere around 170kHz and 100ms ramp.

[Uspring]

Lovely idea to improve on dr. kilovolts design by making the primary removeable with a Faraday cage. I'm wondering, whether an iron core actually increases coupling. Have you any estimates on this?

I'm certainly fond of any research on arc behaviour.

[davekni]

I made a mock-up by winding old power cords into roughly the intended geometry.  Coupling for that was 0.875.  Didn't measure coupling without the ferrite core, but would guess around 0.7 or just under.

Potting had some problems.  Developed air leaks around the top seal (bottom during potting) during vacuum degassing.  Managed to mostly patch those before epoxy cured, but it's not degassed very well.  May still be OK.  I read that QCW voltage is around 50kV.  Even if low-frequency raises that to 100kV, this potting may be sufficient.  Radial distance is 21mm epoxy plus 5.8mm PVC pipe wall.  Axial distance at the top is 30mm epoxy plus 6.1mm polycarbonate.

Uspring, Thank you for your modeling too.  I've done a bit of magnetics modeling, but don't know enough about arcs to simulate those.

[Mads Barnkob]

Interesting project, something that has been talked about, especially the internal primary coils and potting, but the shielding and other measures of actually making it feasible are new to me, looking forward to follow this.

In regard to low frequency QCW, there might be a little knowledge to get out of my failed attempt of making it simple: http://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-iv/

Regarding shielding around magnetic components, perhaps you can find inspiration in the Pearson Current Monitor 1330 I took apart, I was surprised that it was fully copper shielded in 2 layers with only a small slot cut through it:  https://4hv.org/e107_plugins/forum/forum_viewtopic.php?p=1&id=175696#post-175696 (I can make a repost here on hvf with full size pictures if you want?)

[davekni]

Mads,

Thank you for reminding me about your simple line-power-cycle-ramped QCW.  I'd looked through that project when I first found this forum a year ago, but had forgotten that it was also low-frequency for a QCW.  For initial experiments, before I build a proper ramp generator, I'm thinking of trying a resonant ramp generator.  Charged bulk cap array feeds a large series inductor into another bulk cap array, with thyristor for trigger.  I have some MOT primary halves left from my eddy-current levitation demo.  Should make something similar to your line quarter-cycle without risk to other line-powered devices.  (I've seen a vacuum-tube coil powered by isolated line half-cycles as in your demo, behaving much like QCW coils.)

I use a somewhat similar slightly-over-one-wrap of insulated copper foil to shield the first stage of my DRSSTC current transformer.  Likely not needed.  For this QCW primary, copper foil could work around the circumference, but not at the top end.  Most of the primary magnetic field is passing out its top end.  There's no way to make a non-shorted turn in the plane where magnetic field passes through.  That's why the array of wires.

[Hydron]

Hey I've moved the ping-pong/pumpkin shooting stuff to https://highvoltageforum.net/index.php?topic=1274.0 - hopefully I did it right (I think I might have screwed up renaming the topic!).

Dave - some of your posts are relevant to both threads - maybe worth copying/pasting the bits into a new post there, sorry I couldn't figure out how to make two copies of said posts when splitting the thread.

Gotta say the pumpkin discussion led to watching some fun youtube videos - I liked the visual spectacle of rubber band/trebuchet ones more than gas launched though to be honest.

[davekni]

I've now wound a secondary around the potted core and a primary to go inside.  Here's some measurements:
     Secondary winding:
299mm high
160.8mm diameter
27AWG wire
~700 turns
33.19mH w/o ferrites
156.7mH with ferrites
60 ohms DC
     Primary winding (at least this initial version):
270mm high
90mm diameter
28 x 27AWG home-made litz wire
Four interleaved 7-turn windings.  Can be connected in series or parallel or 2S2P.
307uH with ferrites and series (4S) connection.  (19.2uH for 4P connection.)
     Combination:
0.909 coupling with ferrites (0.51 w/o ferrites)
124kHz upper pole frequency w/o top-load, with 7x lower primary tuning (50x capacitance - 2.4uH primary using 4S connection)
106kHz upper pole frequency with 620mm OD top-load

The big surprise (my dumb mistake to miss initially) was all the extra capacitance from secondary winding to internal Faraday-cage shield.  I was aiming for 170kHz with top-load.  Capacitance through the potting epoxy has fairly-high loss too, acting as around 900k-ohms across the coil.  Still shouldn't be a killer issue compared to ~50k arc impedance.

Some images of construction, starting with secondary around potted Faraday cage, up-side-down view:


Primary coil with four interleaved 7-turn windings:


Secondary inside primary Faraday cage:


Ferrite stack, made of a total of 44 E-55 core halves, 3C92 material, plus four flat E-64-50-10 cores for spacing:


Primary around ferrite stack, with leads extending through bottom plate:


Secondary on primary and ferrites:


More ferrites on top, 3F4 material 64 x 51 x 5mm plates:


With 620mm OD top-load, which I won't be using since frequency is already too low:


I had one other problem with the potting besides vacuum issues.  Three connections of the Faraday cage broke, presumably due to stress of epoxy shrinking slightly during cure.  I connected them at the bottom, but with parallel R/C impedance.  The impedance is to avoid a shorted turn in case one of the bad connections reforms with thermal cycling or age or whatever.

Ran a quick test driven more like an SSTC, using a quick half-bridge I constructed for example layout.  Ran it up to ~100kV peak.  At least there's no immediate insulation breakdown issues.

[Uspring]

That's one huge coupling value. If you tune the primary tank to your secondary resonance frequency, the lower pole will be at 77 kHz and the upper one at 350 kHz. Choosing other primary fres's can extend this range. You could name the coil an AFTC (Any frequency Tesla coil).

[davekni]

Yes, coupling is large.  That was one of my design goals - as large coupling as possible.

I must have a mistake in my posted numbers for you to get a 350kHz upper pole option.  124kHz is upper pole with huge (call it infinite) primary capacitance and no top-load.  If I were to move to exactly in-tune, upper pole increases by only sqrt(2) to 175kHz.  However, I want to keep the primary capacitance somewhat large.  That minimizes primary voltage and minimizes flux in the ferrite core.  The 2.4uF in my first test is too large.  As the arc load grew, my simple ZCS driver switched to lower-pole.  (Spice simulation shows this too.)  About 32nF would be in tune with the 4S (28 turn) primary connection (with no top load and no arc).  I plan to drop from 2.4uF to about 0.33uF for the next test.

[Uspring]

I took a k of 0.9, which raises the frequency by a factor of 1/sqrt(1-k) ~ 3. You probably used the k=0.5 as for the coil without core. It should be possible to run your coil at the known frequencies for straight arcs and also explore lower frequencies. Nice 
Exotic experiments would be running the coil with lower and upper pole at the ratio of 1:3 simultaneously. That are the fundamental and first harmonic of the driving square wave. You'd have a square wave TC then. (SQWTC).

[davekni]

Uspring,
The 124kHz measured frequency (w/o top load) is upper pole, so includes the 1/sqrt(1-k) factor already.  This is with k=0.909.  The only tests I ran w/o the ferrite core were to measure secondary inductance and coupling - just for curiosity.  Nothing else is air-core.

[Uspring]

David, I'm sorry to have bothered you with this. I was under the mistaken impression, that the location of the poles don't depend much on coupling, if resonance frequencies of the (uncoupled) tanks are sufficiently distant. That is only true for the lower pole. I had assumed the upper pole in your measurement to be near the secondary fres. Actually the upper pole seems to vary from 1/sqrt(1-k^2) of the secondary fres at very low primary fres to 1/sqrt(1-k), when primary and secondary fres are equal.

[davekni]

No problem!  I appreciate people questioning my results, as I make careless mistakes more often than I'd like.

[Mads Barnkob]

That amount of ferrite blocks o_O

The bottom cross reminds me of how higher coupling is achieved in induction stoves


But what are the purpose of the plane of ferrite blocks living on top of the secondary coil and between that and topload? Surely the ferrite was all there to get the highest possible coupling between primary and secondary coils?

With a mind blowing coupling factor of k=0.909, that is extremely close to k=1 where the coils would take up the same physical space. Really looking forward to see where this project ends

[davekni]

The top ferrites make a small contribution to coupling factor.  They also allow the top-load to be placed close to the top of secondary winding.  However, since I already have too much capacitance, it may make sense to drop the top ferrites along with the top-load.  The ferrites are conductive enough that I need some form of corona ring to avoid arcing from the corners.  Without ferrites I might still need some small corona ring at the top, but it could likely be a non-shorted turn and still sufficiently effective.

Yes, lots of ferrites.  Cost was $158 for all the E-cores that make up the column and cross at the bottom.  The flat ferrites were in my old stash.

[davekni]

My guess had been that slow ramping could compensate for low frequency.  Initial low-power testing shows the opposite.

Until I build the real H-Bridge for this coil, I'm running with a smaller practice bridge I'd built as a layout example.  It uses obsolete 330V TO247 IGBTs, so bus voltage is 250V maximum for a bit of margin.  Primary is 2S2P (14 turns) here with 14uF primary series capacitance (roughly infinite).  Oscillation (upper pole) starts at 115kHz and drops only a bit to ~112kHz by the end of each ramp.  (Due to high secondary parasitic winding capacitance.)  The primary capacitance has 40uH + 1.8ohms in parallel to damp lower-pole resonance, keeping the simple ZCS oscillator running at the upper pole.

Arc shape varies drastically from shot to shot even with the same ramp.  However, it's clear that shorter steeper VBus ramps produce more instances of relatively-straight arcs.  Below are a few examples of arcs and associated scope traces.  (I'm not certain that each scope trace is for exactly the associated image, but they are at least with the same ramp settings.)  The cyan trace is Vbus at 50V/div.  The green shape is primary current at 110V/div (22A/volt at 5V/div).  These plots are zoomed-out, so only the current outline shows up.  For image sizing, the top-load is 30cm diameter.  Each image is of a single ramp manually triggered somewhere during a long exposure.  Also, notice that the scope plots have different horizontal scales, 5ms/div, 2ms/div, and 1ms/div.













The current envelope is interesting.  It seems that more dips in the current trace correlate to more branching.  That makes some sense, as current dropping while voltage is increasing indicates an increase arc loss.  Each loss increase may be caused by a new branch.  It makes me wonder if some active feedback could be useful.  Perhaps if the voltage backed off every time current drops, that could extinguish branches.  Not sure how much effort I want to put into testing that speculation, however.

Finally, I found it interesting that when an arc is relatively-straight and reaches the ceiling 82cm above, it tends to turn sideways and run along the ceiling.  Didn't have the camera pointed quite high enough, but looking at the top of these next two images, the tip of the arc is heading away from the camera along the ceiling.




Next steps will be building the real H-Bridge and the other half of the dual-buck-converter ramp generator and moving the coil off my dining-room table.

[Uspring]

A very interesting experiment   A feature of the arcs seems to be, that they are much more zig zag on a smaller scale and straighter overall. Overall straightness can be explained by the toroid field. But that influence seems to less important on a smaller scale. Perhaps local deviations can come from the coils magnetic field, which might lead to helical forms. Since both the current in the arc and the field change synchronously, the helicity does not change with a reversal of the current. Longer bursts might enhance helicity effects. Do you have a high speed cam to watch slomos? And is it possible that the magnetic field leaks into the arc region?

Still, the lack of straightness is evident compared to the higher frequency arcs. The only real difference I can think of between high and low f arcs is, that the high f arcs are hotter due to the more frequent transfer of space charges.

[davekni]

I expected the stray magnetic field to be small since there is a ferrite plate on top, then a small aluminum pie tin inside the top-load.  Just to be sure, I made a different top today.  Ferrite plate is the same.  On top of that is a large aluminum pie tin encircled by a copper-pipe ring brazed together at the joint to form a solid shorted turn.  Behavior is still roughly the same:


As before, there are plenty of more-branched shots too:

[Mads Barnkob]

The primary capacitance has 40uH + 1.8ohms in parallel to damp lower-pole resonance, keeping the simple ZCS oscillator running at the upper pole.

The current envelope is interesting.  It seems that more dips in the current trace correlate to more branching.  That makes some sense, as current dropping while voltage is increasing indicates an increase arc loss.  Each loss increase may be caused by a new branch.  It makes me wonder if some active feedback could be useful.  Perhaps if the voltage backed off every time current drops, that could extinguish branches.  Not sure how much effort I want to put into testing that speculation, however.

Are you perhaps draining your ramp inverter DC bus energy storage so bad that you have a voltage drop that is recharged by the front stage or mains and that causes the current spikes? The current waveform reminds me of my DRSSTCs when running in a "current limited equilibrium" mode due to too little DC bus capacitance. So it essentially just feeds whatever it can into the arc, there is no peak current storage energy to draw from.

Regarding current dropping, wouldn't it be a higher voltage? As we saw in Hydron's measurements of topload discharges: https://highvoltageforum.net/index.php?topic=117.msg784#msg784

A very interesting experiment   A feature of the arcs seems to be, that they are much more zig zag on a smaller scale and straighter overall. Overall straightness can be explained by the toroid field. But that influence seems to less important on a smaller scale. Perhaps local deviations can come from the coils magnetic field, which might lead to helical forms. Since both the current in the arc and the field change synchronously, the helicity does not change with a reversal of the current. Longer bursts might enhance helicity effects. Do you have a high speed cam to watch slomos? And is it possible that the magnetic field leaks into the arc region?

Still, the lack of straightness is evident compared to the higher frequency arcs. The only real difference I can think of between high and low f arcs is, that the high f arcs are hotter due to the more frequent transfer of space charges.

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.

[davekni]

"Are you perhaps draining your ramp inverter DC bus energy storage so bad that you have a voltage drop that is recharged by the front stage or mains and that causes the current spikes? The current waveform reminds me of my DRSSTCs when running in a "current limited equilibrium" mode due to too little DC bus capacitance. So it essentially just feeds whatever it can into the arc, there is no peak current storage energy to draw from."

By appearance that would make sense.  However, in this case the entire ramp is fed by a bulk cap (the 20mF one I accidentally shorted to cause hearing damage last July - wearing ear protection now).  The cyan trace is voltage out of the ramp generator (bus voltage to the H-Bridge).  Poor choice of colors, but if you look closely it is visible on top of the green current trace.  Voltage continues to ramp smoothly until the ramp ends.

"Regarding current dropping, wouldn't it be a higher voltage? As we saw in Hydron's measurements of topload discharges: https://highvoltageforum.net/index.php?topic=117.msg784#msg784"

Could you explain this more?  After the ground strike, voltage is low (topload shorted to ground through the arc) and currents are also low (low-Q secondary resonance due to arc short prevents secondary current from building up).  Although not plotted, primary current is likely increasing after the ground strike due to lack of secondary load.  For my QCW coil's high coupling and upper-pole operation, primary and secondary currents roughly track.  The dips in primary current plotted should correspond to dips in secondary current caused by dips in secondary voltage due to increased arc loading.  I wasn't measuring secondary current in these runs, so don't have explicit data for that.