sky-guided single-circuitboard plasma toroid

[sky-guided]

Breaking this out into its own thread, so I don't derail alan's


I'm both incredibly proud of this build, and also see definite areas where it could be improved.

Video

Schematic


Full writeup
https://hackaday.io/project/194683-plasma-toroid-sky-guided-pcb-edition
(includes BoM and PCB source files)

Stuff I know I got wrong / you can do better
https://hackaday.io/project/194683-plasma-toroid-sky-guided-pcb-edition/log/232417-11-design-deficiencies-and-what-you-can-improve

Critique and/or roasts are encouraged!

[davekni]

Thank you for posting schematics.  Much easier than viewing schematics in a video.  I haven't had time to look through your details yet.

Beautiful clean build!  (As has already been said.)  In particular I'm interested in your single-turn coil.  Was planning to try that myself some day.  Lower priority now that you have already shown such.  Reduces capacitively-coupled voltage, so presumably starting is mostly by induced loop voltage.  That was my question - is higher capacitively-coupled voltage required to start?  Early designs seemed to work that way, being sensitive to hands or other objects near gas globe.

[sky-guided]

It's actually a two-turn coil -- the second turn is hiding on the bottom layer of the PCB.

Mine can't self-start, so it uses a modified xenon flash tube trigger circuit to initiate capacitive ionization. Reliability is ballpark ~30% when the globe is cold, and 95% on re-starts.

[Anders Mikkelsen]

Here are my quick notes on the design:

1: You have pure ceramic class 2 capacitor decoupling, so beware that capacitance will change a lot with applied voltage, and there's very little damping with ceramic capacitors. I would be worried by voltage overshoot on hot-plugging, from wire inductance. For linear caps this can reach 2x the applied voltage, while with nonlinear caps I've seen over twice this again, which could overstress your low voltage regulator connected to this rail. Either some bulk electrolytics for damping or at least some TVS clamping of this voltage would make it more robust.

2: The ground plane extends under the high voltage output of the ignition circuits. Removing the plane under the transformer and particularly the output side can reduce voltage stress on the PCB and potentially give higher peak voltage out from reducing capacitive loading on the secondary

3: The startup circuit can be made cheaper as you note. I would look into either an off-the-shelf flyback transformer for mains to 5 V applications, for example the Wurth 750310787 driven by a baxandall oscillator, or an ATB322515 driven by some current mode flyback or boost controller

4: For the MOSFET heating, I would try out different parts, and try to use a single transistor if possible. GaN or SiC can work, GaN requires very tight gate drive voltages, which suggest going with a gate driver chip, SiC can be made to work in the standard HFSSTC circuit, and SiC cascodes are even more forgiving on the gate voltages if you want to maximize your chances of success without adding extra work. UF3C065080B3 is a good candidate for example

5: What's the purpose of D8? With 1k in series, it won't do much clamping, but it will shift the bias voltage when the gate drive voltage is large enough for it to start drawing current from your bias circuit. Is this the intention?

6: I would consider hooking up the overtemperature sense to a latched shutdown of the power oscillator.

I'm tempted to assemble a copy of your board myself, as I have a tubulated 1 L flask and a pumping manifold with Xenon here, Then I could also experiment with gas mixtures and pressure.

[sky-guided]

Thank you so much for the detailed notes!

1: I hadn't considered that aspect! Earlier prototypes did have a TVS diode on the voltage input, but that was mostly intended to protect my power supply if something went screwy.

2: I also hadn't considered that. Looks like the ground plane is like half under the output pads, but there's no reason I can't give it more clearance. And... thinking further, I might want to issue a board revision specifically to increase my clearance between the striker and the ground plane.

3: Heh, I actually looked at that exact Wurth transformer at one point, but didn't pursue it much since it seemed overly bulky.

4: lol I was using the UJ4C075060B7S for my first few prototypes but ditched it because I got tired of accidentally crisping $15 mosfets. (Pretty sure I was exceeding gate voltage, because I hadn't added clamping at that point.)

5: I'd have to go double-check my work on this but iirc: I was afraid of smoking the low-wattage zener and in the sims I was showing that 1k did a pretty decent job of clamping to ~18V if the gate drive did a bit over-enthusiastic. Though... if I actually pause to think about it it's pretty clear that ~3mA isn't going to clamp much. Hm.

6: Definitely would have done something like that if I were designing a consumer product. As is, I'm kinda glad I went with the indicator-only approach -- in the version before final, I made the surprise discovery that I needed additional decoupling on the programming resistor due to proximity to the switching currents, else it'd continually alert. I imagine that might've been a fairly annoying bug to track down if it instantly halted oscillation 😅