SiC PSFB QCW coil

[dr. kilovolt]

Hello everyone,

I have been persuaded by YouTube friends to join this forum and share some of my TC work The biggest mystery of Tesla coils is the plasma physics and resulting loading (R, C) of the secondary coil, this can only be guessed based on experience. So hopefully my data will help with better coil design. 

Parameters of my coil:
Secondary 125mm dia / 150mm winding length, approx. 400 turns
Primary coil is internal and has a ferrite rod core
The coils are filled with polyurethane potting compound to avoid flash-over
Topload 6x25cm toroid
Phase-shifted full bridge of UF3SC120016K4S FETs with 800V bus voltage (charged)
5nF tank cap
3kW 800V PFC front-end with 230VAC input

Measured data:
Lpri=65uH, Lsec=19mH (with toroid), k=0.55
fpri=279kHz, fsec=251kHz
fosc 400kHz start, 350kHz end (not exactly sure, might correct later)
Ipk=100A (bus voltage dropping to approx. 650V) which translates to Qpri~11 and ~40kW peak
Maximum spark length is over 2 meters but haven't measured exactly...

Just after I did these measurements I realized the primary is not designed optimally, it would be better to use a lower inductance and higher capacitance and tune the primary lower (this would keep the peak current but increase efficiency). However this is not possible with the potted primary :-) Originally I had a too low peak current, the only thing that I could do was rewinding the secondary for more turns.

A note on phase-shift modulation: The transfer function from phase difference to output power is not linear, it follows a 1-cosine curve. For this reason, to obtain best results, the phase ramp should compensate for this with inverse function. It is not required, but without the correction longer ramps (and more energy) are required to obtain straight sparks.

Video:

/>
A few photos are in the attachments.


Jan

[Mads Barnkob]

Hi Jan and welcome to HVF!

Impressive performance, is all drivers and PFC homemade or are you using some of the drivers (1-2 SiC afaik) floating around the forums?

Primary coil is internal and has a ferrite rod core
The coils are filled with polyurethane potting compound to avoid flash-over

You built a Tesla Coil there is no Tesla Coil at all! You broke the "air core resonator" rule 

Resin/epoxy filled coils or even put inside a oil tank was something I have thought about for a long time, but never gotten around to actually spend time on making. Great to see it in effect!

One last thing. There was quite a few dr.kilovolt or kilovolt on 4hv and since your name is Jan, are you jmartis2?

[dr. kilovolt]

Hi Mads!

The electronics are my design. Control is analog/CMOS. No MCUs, I don't like them, maybe partly because I can't do them.    The SiC bridge uses a GDT drive. The PFC uses the UC3844 IC, which reachces an adequate PF in DCM mode, but gives me more freedom in choosing the characteristics.

Yes you are right that I should not be posting in this forum as this is technically not a Tesla coil at all.    The potted coils are nice, but highly resistant to modifications and tuning. 

Yes I'm Dr. Kilovolt from 4hv and jmartis2 on youtube.


I forgot to mention in the first post, there's one characteristic of QCW coils which I NEVER noticed on any other coil before, and that is dependence on atmospheric conditions and/or surroundings.  Usually in a normal room with standard indoor conditions, the coil works nicely, producing almost perfectly straight sparks with correct setting of the ramp speed. However there seem to be conditions, related most likely to air humidity / temperature but maybe also to conductivity of floor beneath (e.g. ground floor or outdoors), which make the streamers curve downward, sometimes with such a small radius that the streamers almost tend to hit the bottom of the coil (sometimes they hit the coil base) - see the attached photo for an example. Does anyone have an idea what causes this, or even what are the physics behind this?

[MRMILSTAR]

Very nice coil! I have only built large SGTCs and VTTCs. I haven't gotten on board yet with SSTCs. Of all the SSTC designs that I've seen the QCW types interest me the most. If I ever build a SSTC it will be the QCW type.

[John123]

Yes I'm Dr. Kilovolt from 4hv and jmartis2 on youtube.

Oh god its you! How are you these days? I've been messing with some of your old tv flyback transformer circuits during lockdown.

Nice coil btw, wish I could help out with SSTC technical stuff but I'm not at that level (yet).

[Weston]

Hi Jan,

Thank you for taking the time to write up your coil on the forum! Technology wise it seems to be one of the more "out there" coils and the build quality sure looks excellent. Seeing the work of others motivates me to work more on my coil.

What drove the design choice behind the ferrite rod? The first thing that comes to mind is that I would expect a ferrite rod to allow for a higher Q factor resonator. Is the higher coupling factor due to the ferrite rod also advantageous, or is it merely a side effect? 

Have you taken any waveforms of the tank current throughout a burst? I am getting almost constant tank current on my QCW coil and am curious if others are seeing anything similar.

[davekni]

Very impressive coil!

What's with this "air core" rule?  Tesla started with oil-insulation, then moved to air to get sizes larger than practical under oil.

Did you ever measure coupling factor?  I'm guessing that high coupling is a key factor for performance, along with input power and proper phase (amplitude) ramping.

[Uspring]

Jan, a beautifully working coil. And interesting observations coming with it. You wrote:

The biggest mystery of Tesla coils is the plasma physics and resulting loading (R, C) of the secondary coil, this can only be guessed based on experience. So hopefully my data will help with better coil design. 

Here's an attempt at estimating arc loads. I dunno if it works properly for the frequency of your coil. I've just tested it at 70 and 140kHz. Also there is a lack of measurements for longer bursts as in QCWs. https://highvoltageforum.net/index.php?topic=670.msg4474#msg4474

An idea for the power up ramps: I don't have any experimental data to confirm this, but I believe, that it might be the best to grow the arc at a constant velocity. John Freau's formula for arc length versus power is length^2 ~ power. Then P should be proportional to time^2. Did you experiment with different ramp ups?

These looping arcs are intriguing. You may be on the track of a new phenomenon. I believe slowly growing arc to be sensitive to the surrounding fields. More so than faster growing arcs, where random events near the arc tip cause break down avalanches in the direction, where free electrons happen to be. A bit about this is written here: https://highvoltageforum.net/index.php?topic=973.msg6602#msg6602

The electric field near the tip is caused by space charges right behind it and also by the toroid. Initially ths causes the arc to grow straight and away from the top load. Your coil is special in the way, that it has a rather small toroid compared to the arc length. That means, that the toroid doesn't hold much charge, so that its influence on the arc direction diminishes fast as the arc lengthens. The field near the arc tip is probably influenced to some extent by the nearest grounded surface.

Another effect, that might be interesting: The nature of the arc being a chain of resistances and space charges produces a phase shift between top load charge and space charge near the tip. Now (speculating wildly), the tip space charge might be a 180 degrees relative to the top load. That would cause the arc to be attracted to the toroid, so that it might form a loop. In principle one could imagine a closed loop arc, coming from the top and going back into it. The situation would resemble a transmission line, where the ends are at the same voltage and the middle opposite in phase. Maybe this situation is not stable and collapses into a single arc, but for a short time it might be looping.

I liked your video. I was waiting for the arc to hit the possibly flammable liquids on the left.  I couldn't discern, whether your arcs are meandering or spiraling near the bottom. Spiraling could be explained by the secondaries magnetic field. It should then always have the same helicity for every arc and it should be opposite to that of the secondary winding.

Do you have a high speed camera? It would be lovely to see the arc growing and how it branches, i.e. does it branch at the tip or break out sideways later.

[dr. kilovolt]

What drove the design choice behind the ferrite rod? The first thing that comes to mind is that I would expect a ferrite rod to allow for a higher Q factor resonator. Is the higher coupling factor due to the ferrite rod also advantageous, or is it merely a side effect? 

Have you taken any waveforms of the tank current throughout a burst? I am getting almost constant tank current on my QCW coil and am curious if others are seeing anything similar.

The ferrite core is there mainly for increased coupling. I was afraid of high loaded primary Q with standard coils (my guesstimates were Q>20) and I didn't like that. I ended up with a too low Q on the other hand, without any means of changing the primary parameters.    That was the main reason. Of course the ferrite also allowed a higher unloaded Q or lower losses in other words, which are also advantageous. Without this the internal primary would be uncoolable.

The primary current envelope looks something like the sketch in the attachment. If a spark is done with a cooled down discharge rod, there is a small spike seen at the beginning. (Of course I have taken waveforms but was too lazy to save them, now I'm too lazy to discover the coil and install a current transformer.  )
A note on the current envelope, probably well known to QCW experimenters (but not to me as I really haven't done much QCW before): During a ground arc the current drops. This is probably the opposite to a classic DRSSTC.

Did you ever measure coupling factor?  I'm guessing that high coupling is a key factor for performance, along with input power and proper phase (amplitude) ramping.

The coupling coefficient is 0.55 (as written in the first post). It is very high, but (as also stated) it is not utilized fully. The coil assembly could get an improvement, but that would mean potting another coil. Might try that sometimes though.

Jan, a beautifully working coil. And interesting observations coming with it.
Here's an attempt at estimating arc loads. I dunno if it works properly for the frequency of your coil. I've just tested it at 70 and 140kHz. Also there is a lack of measurements for longer bursts as in QCWs. https://highvoltageforum.net/index.php?topic=670.msg4474#msg4474

Thanks I actually have seen that Your post before, interesting stuff. Sadly I don't have an LTspice. Might try installing it though, having a relatively working arc model is a lot of help.

An idea for the power up ramps: I don't have any experimental data to confirm this, but I believe, that it might be the best to grow the arc at a constant velocity. John Freau's formula for arc length versus power is length^2 ~ power. Then P should be proportional to time^2. Did you experiment with different ramp ups?

Isn't the John Freau formula applicable only for SGTCs with very low duty cycles and continuous streamers? However my thinking with a required energy to grow a given length straight streamer showed a somewhat similar quadratic relationship. If you want to grow a double length arc, the energy dissipated in its top half is the same as in the original arc. However, growing the top half also requires supplying additional energy to the already grown bottom half, which gets fatter. Based on this I guesstimate the energy to grow twice the length of arc to be 3-4 times higher.

I experimented just a little with different ramps. I just settled on what seemed to work best and that is close to linear bridge output voltage ramp (but not a linear power ramp). That would be a disussion for another post, as this one is getting way too long...

The constant velocity growing should intuitively be correct. I think this also means a constant arc tip voltage relative to distant ground. However I don't know how to deduce the required power ramp shape from this.

These looping arcs are intriguing. You may be on the track of a new phenomenon. I believe slowly growing arc to be sensitive to the surrounding fields. More so than faster growing arcs, where random events near the arc tip cause break down avalanches in the direction, where free electrons happen to be. A bit about this is written here: https://highvoltageforum.net/index.php?topic=973.msg6602#msg6602

Very interesting theories, thanks. It is funny how little we know, we can just observe.

I have been working with Tesla coils for maybe 15 years and this is the first time I observe such a behavioral change of a TC just by moving it to a different environment. The influence is quite massive. Sometimes most of the arcs loop with a diameter really less than 1 meter and the ramp parameters don't seem to help much - it is almost impossible to show off the coil  . If an upright arc happens at these conditions, it is usually quite short, while sideways (or even almost horizontal) arcs seem to be much longer. Sometimes a longer ramp helps to straighten the arcs a bit. But other times the arcs are straight all the time or curve with a really large radius, just a too short ramp makes them branch. At time to time the arcs even take an "S"-shape, as depicted in the attachment. So far there is a weak correlation that dry air tends to help straight sparks and temperature does not have much influcence. But really don't know. I have never seen anything like this before.

Surely I have observed the looping arcs before, while I'm sure that a higher frequency supports this. At 2 MHz from a ramped VTTC,  the diameter of the arcs is a few tens of centimeters. However I have never observed the "atmospheric" influence.

The nature of the arc being a chain of resistances and space charges produces a phase shift between top load charge and space charge near the tip. Now (speculating wildly), the tip space charge might be a 180 degrees relative to the top load. That would cause the arc to be attracted to the toroid, so that it might form a loop.

I have always thought of the arc equivalent ciruit as a distrubited RC network, which possibly cannot do such large phase shifts? I also thought of the current propagation speed to be much faster than what would be required for and 180 degree shift. But I'm not really sure...

[Uspring]

Isn't the John Freau formula applicable only for SGTCs with very low duty cycles and continuous streamers?

Yes, and moreover it is established from coils of different sizes, the more powerful ones running at lower frequencies. Differing burst rates, which markedly affect power consumption also aren't taken into account explicitly. Freaus formula is just a starting point. I certainly share your view, that the outer half of an arc twice as long resembles an arc of total length of the outer half. The inner half has to carry much more current and is consequently hotter and more power consuming.

The dependency of the direction of the arc on the environment is puzzling. My guess is, that QCW arc directions depend more on the surrounding fields than those of rapidly growing arcs. The comparatively high fields near the tip of fast growing arcs make the direction of growth more susceptible to local conditions, e.g. spurious free charges near the tip. For a high field, the volume around the tip, where avalanches are possible, is larger and perhaps less pointed forward.
So, if all of the above is correct, QCW arcs tend to follow the field lines of the toroid. If you are outside, with no other grounded planes except the bottom, the field lines will all point away from the toroid when they are near the toroid. Eventually they will all curve down to the bottom. See e.g. the image near the top of this article https://en.wikipedia.org/wiki/Method_of_image_charges
If you are in a room, the field lines will not all go to the bottom, but some of them will end on the walls or the ceiling. Since your breakout point is facing upward, most of the field lines will go up to the ceiling and most of the arcs will follow them. It is maybe not an accident, that you observe the loop outside (no ceiling) and that it loops toward the house wall (a ground plane, where the field lines are headed)

I have always thought of the arc equivalent ciruit as a distrubited RC network, which possibly cannot do such large phase shifts?

The phase shift of a distributed RC network can be arbitrarily large. It comes, though, with an attenuation of voltage. So it might be, that once you have a large shift, there is not anymore voltage left for the arc to grow. The idea is very speculative, but it could be a possible explaination for arcs bending back to the toroid.

[dr. kilovolt]

If you are in a room, the field lines will not all go to the bottom, but some of them will end on the walls or the ceiling. Since your breakout point is facing upward, most of the field lines will go up to the ceiling and most of the arcs will follow them. It is maybe not an accident, that you observe the loop outside (no ceiling) and that it loops toward the house wall (a ground plane, where the field lines are headed)

Then explain the first attached image.  I have also observed severe looping in the garage, which I would say has a standard room dimensions (and walls), but it is not quite hermetical - connected with outside environment by numerous air holes. Once I was getting serious looping, heating up the air to room temperature didn't help much. But what quite helped was opening the garage door and letting outside air in. Even though it was relatively cold (~10 degrees C), the looping decreased a lot.

The second attachment shows also a somewhat weird spark.

-Jan

[Uspring]

Then explain the first attached image.

I can't  And I also don't have the slightest idea how air temperature or humidity could cause arcs to bend or not. Does anybody have a suggestion?

[Mads Barnkob]

Then explain the first attached image.

I can't  And I also don't have the slightest idea how air temperature or humidity could cause arcs to bend or not. Does anybody have a suggestion?

Could it affect how fast the arc channel is cooled down again? Or how hard it has to fight through the water vapor, taking energy out of the channel as it evaporates?

[Teravolt]

HI dr. kilovolt, I like the simplicity and great wood craftmanship of your QCW, it is an assume display. Are you using any ramp wave applied to the uc3844 and dose the tesla have a standard DRSSTC feed back to keep it in tune. Id ask for a schematic or block diagram but I think it is frowned apron, don't know.
I am working on my own QCW but it uses the same old Class D amp to power The DRSSTC but my aim is to be able to modulate the spark with frequency or change the ramp to be like half sin wave and half triangle to get different effects. I also want to make a demonstrator like yours that could fit in a back pack. Do you have and pictures of your ferrites how they are located in your coil and what material are they made from. thanks for your post

[dr. kilovolt]

Teravolt: The UC3844 in the PFC has no ramps on inputs, it is just a DC regulator. The converter runs in DCM all the time if output voltage is above 650V and for a fixed duty cycle this draws harmonic LF current, so this by itself assures a good PF. There is however a feedback from the output which decreases the current limit for lower output voltages than 700V to possibly avoid CCM switching - the turn-on losses of the SiC transistor are much higher if the free-wheeling diode is forced to recover.

The feedback is a standard primary feedback using a PLL to clean up the waveforms and produce exactly 50/50 duty cycle (I really didn't want to pop the relatively expensive FETs on a first quirk).

I would suggest using a classic buck conveter for the modulation and I would do it in my next QCW. The phase-shift is nice in that it possibly results in smaller electronics (but possibly not lower losses), but it has 2 main drawbacks. The first was already mentioned, the non-linear transfer function from phase difference to output power. The second one is that if the bridge fails, it goes with a loud bang, all of the stored energy dumps into it and you end up with charred parts and possibly also the PCB. With two converters effectively in series, a simple protection can be realized which saves the other converter if one fails and avods discharging the storage caps, resulting in a silent failure. Then there's that third advantage - lower losses and noise in the bridge, for modern semis this is not THAT big problem, but still its not enjoyable.

The primary winding has almost half the dimensions of the secondary coils and is filled with I-cores of 3C97 material.


-Jan

[flyglas]

I like your coil and the archived results.
Do you have pictures of the primary coil with ferrit core prior to potting?
I like the idea of putting the primary into the secondary coil.

[Bambinz]

Amazing performance and costruction, probably the best QCW coil that I ever seen.

[dr. kilovolt]

Thanks for the compliments 

Sadly I didn't take photos of the primary construction...

But I have taken some measurements for the interested, see below. The power is average power delivered to the tank circuit. Also I have now somewhat quantified that most of the time the coil produces approximately 2-2.5m long sparks.


Regards, Jan

[Weston]

Thanks for providing a waveform of the current envelope!

Why did you only have the ferrite core go part of the way up the secondary? Is it due to insulation concerns?

For high voltage particle accelerators (or something, this is mostly outside my research area) they build something called "insulated core transformers" where they split the core up into sections and put insulation between them to reduce the insulation requirement between the core and the windings. This paper should be open access: https://accelconf.web.cern.ch/rupac2016/papers/thpsc038.pdf

I was considering building something similar for my chainsaw coil by stacking commercially available ferrite torroids. And possibly making it an "I" shape with extra ferrite on the bottom to protect my control electronics and at the top to stop the topload from acting like a shorted turn.

[Uspring]

I've run a simulation of your coil using the tank parameters given in your initial post. The bridge voltage is ramped up linearly, except for an initial jump to 50V. The max voltage was chosen in order to reproduce the input power at the end of the burst. Burst length was set to 2ms in order to limit computer time. I don't expect any significant differences for longer bursts, since the arc model used cannot reproduce any features such as straight versus wiggly or branched arcs. The LTSpice simulation below shows bridge power (smoothed) and primary current.

The calculation reproduces initial current, i.e. 35A at the 1-2kW power level well. It comes out a bit high at 115A at max power. The overall shapes of the curves look similar to those measured.

The simulation allows to follow the growth of the arc. I've taken the tip of the arc to be, where a current of about 10mA begins to flow in the chain. I've plotted length (cm) versus time (ms).

The graph shows an almost constant growth speed. Very well done. The total length calculated is barely the half of the real arc. I've calibrated the arc length in the model by the distance between breakout point and tip for conventional non QCW DRSSTCs. I believe the difference between prediction and reality is caused by omitting the usually wiggly nature of the arc and also the effect of branching. Ironed arcs are longer. An LTSpice file is appended.
arcsim.zip
Sorry for some annoying error messages. Just ignore them.