This is less a set of specific replication instructions, and more a generalized set of advice/recommendations.

The plasma toroid project can be tricky -- although lots of folks are able to achieve success, there seems to be a lot of sneaky variables than can affect performance which haven't been pinned down yet. Recall that my initial project log was titled "inevitable first prototype failure."

Understand the circuit before you start. "Understand" means more than just nodding along with an explanation, it means being able to reconstruct the circuit topology from memory and explaining why component values are what they are. 

LTspice circuit simulation is your friend.

In addition to this writeup, I highly recommend digging into the linked videos/documents by BacMacSci, Steve Ward, and Humxn. 

The frequency range of 10-15Mhz is fast enough that parasitics and return-current path do matter, so use a PCB and mind layout best practices like having a continuous ground plane. Fortunately, the RF wavelengths are long enough that you don't really need to care about impedance control.

The intense EMI from this device makes oscilloscope probing challenging. Using probes with a ground lead wire and alligator clip will probably show phantom wiggles -- the little spring contact ground has a much smaller loop area and will thus work a lot better. A superior approach is to follow what Humxn did and design scope probe connectors directly on to the board. 

Although it may be possible to get this working without an oscilloscope, without instrumentation you may have to rely on some luck. 

The high voltage parts of the circuit will hurt you if you touch them. These voltages can be high enough to break through PCB soldermask. RF burns can take days to fully develop, so if you feel an ouchy zap, run your hand under cool water even if it "looks fine." 

Primary inductor coil:

Actually measuring the inductance of your coil will help with simulation/analysis. Even if you don't have a dedicated LCR meter, you can do this with a capacitor and an oscilloscope.

The coil will get hot, disproportionate to what you might expect from current flows in DC or 60Hz AC with similarly sized conductors. 

Choke Inductor:

The choke inductor carries DC, not radio-frequency AC. You don't have to worry about self-resonant frequency or hysteresis loss or whatever. A ferrite core inductor is fine. 

Mosfet:

An ultra-beefy component isn't what you want here. Higher maximums directly trade off against other performance criteria. 

• Max amperage: needs to be high enough, but not more. You'll probably be thermally-limited anyway.
• Max drain-source voltage: Same story. Simulation should guide intuition for how high is "enough".
• Total gate charge: Lower is better. This is probably the spec to prioritize since switching losses will massively impact heating, and we're not feeding with a "proper" high-current, square-wave mosfet driver. 
  (Unless you are, in which case, good job.)
  
  
• Coss: lower is better, but I'm still not totally clear on how much it truly matters in this kind of soft-switching configuration. Power engineers, go ahead and @ me on this. 

• On-state resistance: Lower is better but you don't need to go wild. At these frequencies, you'll see way more heating from switching losses than from on-resistance. 
• Rise/fall/delay times: Benefit from being kinda fast. For reference, the period of one cycle at 15MHz is 66 nanoseconds. 

A chunky heatsink will be mandatory. You'll probably need to use mosfets in either the TO-247 (through-hole) or TO-263 / D2PAK (smd) package sizes in order to dissipate enough heat.

Mosfets will immediately die if their max gate voltage is exceeded, so protection with a zener diode pair is a good idea. 

Mosfets will also tend to fail short, so if suddenly your power supply is trying to pass infinity current (or protectively turning itself off), de-solder the mosfet and check it individually. 

Mosfet gate resistors have to pass substantial current. Size components accordingly to prevent overheating -- I messed this up all the way to my "final" design showcased, though it's corrected in the publicly posted PCB files. 

Tank Capacitors:

I had great results with a bank of 1808-sized ceramic chip capacitors. Use NP0/C0G type. Other folks have equal success with more trad "doorknob" type caps. 

In my design, the slot under the capacitor bank is less about high-voltage creepage distance, and more about making sure I could inspect underneath each part to ensure there wasn't any trapped flux, solder balls, etc. I came across that recommendation in an application note... somewhere. 

These kinds of caps can be prone to cracking due to thermal stress, so hot air or reflow works better than a soldering iron. 

If you want to replicate my exact design, I've attached complete files.

A copy of the (unpopulated) board can be ordered directly from PCBWay (preview looks glitched but it fabricates fine; they seem to have a view rendering issue with the central aperture). I have not verified whether PCBWay can perform assembly services for this board. 

Component placement locations are in the interactive HTML BoM, in the /bom subfolder of the KiCAD files.

Reflow or hot-air board assembly is highly recommended. You'll also need to add a short length of 16ga (or similar) insulated wire for the arc-striker. 

The heatsink comes with thermal paste pre-installed, but you'll need an additional kit of the ATS-HK91-RO spring pins to properly attach it to the board.

For assembly, print 2x ea of the _LH and _RH legs, and 1x of the fan shroud and base ring. Preferred print orientation should be fairly obvious. For my builds, I used a light fuzzy-skin on the legs and base ring.

Connect components together using m2 x 4mm heat-press inserts.

I'm done with this project, but here's what you can improve on:

- This circuit really should be driven using an adjustable current-control on the input -- intensity control via adjusting the mosfet duty cycle (by adjusting the bias voltage) is kinda janky, and it's easy to lose oscillation.

- I suspect that use of a sine-wave for gate drive is causing a lot of excess heating. A better solution would be to use a proper mosfet gate driver IC, probably synced via a PLL. 

- MOSFET selection may not be fully optimized.

- Asymmetric on-off gate drive using a diode for turn-off is worth investigating. 

- The flyback + GDT + trigger transformer works OK for arc start, but it's a high BoM cost and reliability isn't perfect. At a minimum you could probably replace the specialty photoflash IC with a simpler flyback driver.

- You don't need to use this kind of self-oscillating class-E driver to create the toroid effect. Try an entirely new topology!

I've expanded on some of these thoughts in Project Log #11: "Design deficiencies, and what you can improve"