Dual Class E Reverse Engineering/Engineering

[YSPACE Labs]

I found that a lot of the Aliexpress Tesla coils use a topology known as Dual Class E or Push-pull Class E. It's similar to a Push-pull coil (like Tefa's SSTC4), but there is a capacitor in series with the non-tapped primary, two chokes, and two drain-source capacitors. It's basically two single mosfet class E stages glued together. I've seen some DIY coils using this topology (Magneticitist, Ethan's Lab, and some others (none call it dual class E (Magneticitist calls his DCE coil a push pull))), but there are virtually no details on how to tune it. There are component values, but without knowing primary inductance, it's hard to know how it's tuned.

The advantages of this topology seem to be that it's dual resonant (and it produces big sparks for its input voltage and secondary windings number), doesn't require any special floating gate drive, can run at very high frequencies (possibly up to 4MHz if my JavaTC calculations for the secondary are correct), and is pretty tolerant of things that happen to it. I made a PCB (will upload picture later) to experiment with driving low voltage push pull coils like this.

I'm not sure how exactly to tune it, but it seems like either you tune the series cap to resonate with the primary, tune the series cap so that its capacitance in series with one of the DS caps resonates with the primary, or to use the Class E equations and change the DS caps so that it operates in ZVS. Also, according to Magneticitist, the "dirty double hump" waveform is good for getting large sparks (so maybe it should be detuned in that way).

So I just want to see if anyone more experienced or with one of the aliexpress coils can help me research and develop this LVSSTC (Low Voltage Solid State Tesla Coil) design.

[YSPACE Labs]

Here are a few images of the topology and other things:
PCB design
Schematic of topology

[davekni]

Would you call the typical cheap ZVS induction heater circuits class-E?  I don't know the exact definition of class-E.  If gate drive is 50% duty cycle, could be run almost exactly like these ZVS circuits.  Capacitors from drains to ground behaves the same as one capacitor drain-to-drain as long as gate duty cycle is 50% or higher.  (Same for tuning.  Does change FET current, as ground caps add resonant current to FETs.)

Experimenting in simulation (LTSpice or other) is often easier than modifying real parts on a real board.

If you complete a schematic including gate drive, please post.  That could be interesting.

[YSPACE Labs]

OK. Here is the actual driving portion of my board: . This is somewhat pulled off a circuit I found on Aliexpress (lol) (that schematic is what sparked my interest). Based off my sims, the circuit can function in ZVS class E mode and produce some really high primary currents. The gate drive is really simple since it doesn't need any transformers or bootstrapping to drive a floating MOSFET. It probably doesn't even need the resistors and diodes going to the gates.

Those cheap ZVS induction heaters are not Class E. They use a parallel resonant cap and for whatever reason, it's not as good as having a series resonant cap (I read this on a Discord server, and the person who said it said that it was from the HV forum). If anything, those ZVS drivers are something like Push pull Class D^-1 (Based on a slideshow from Berkeley ECE that I found when trying to figure out Dual Class E). Based on my knowledge (I am not an EE), the chokes and DS caps act as "pumps" that smooth and feed the primary through the impedance matching series cap. I could be completely wrong on this theory though. Richie (https://www.richieburnett.co.uk/tesla.shtml) knows something about this topology and calls it Double-ended Class E.

I still have to run some more simulations on it, but I think you can just use the class E equations to tune it (the DS caps need a little adjustment in the sim).

Here's the Aliexpress schematic:

[davekni]

OK. Here is the actual driving portion of my board:

The two U8 inputs labeled "IN" connect to an antenna?  Or to a CT?  Either way, presuming they are wired together, then the gate signals are complimentary 50% duty cycle.  That answers my main question.

It probably doesn't even need the resistors and diodes going to the gates.

Agree, I can't think of any value for the resistor//diode in series with gate.  No dead time is needed for ZVS circuits (whether class E or push-pull or whatever).  If the diodes were reversed for faster FET turn-on, that would guarantee no time when both FETs are off.

Those cheap ZVS induction heaters are not Class E. They use a parallel resonant cap

If FET gate diodes were reversed to avoid any time when both FETs are off, then this circuit would behave the same if C2 and C3 (DS caps) were replaced with a single cap between drains (as in ZVS inductrion heaters).  Either way, I'd consider this circuit half way between series and parallel resonant.  The primary resonant circuit is primary inductance and C1 in series with C2/3.  Drive is in parallel with C2/3, so not in series with the entire resonant circuit, nor in parallel with entire resonant circuit.

I'm not sure how exactly to tune it, but it seems like either you tune the series cap to resonate with the primary, tune the series cap so that its capacitance in series with one of the DS caps resonates with the primary, or to use the Class E equations and change the DS caps so that it operates in ZVS.

I think above three options accurately describe the range possible for this circuit.  Best efficiency should be with your second option of series cap (C1) in series with C2/C3.  Then FET body diodes don't conduct.  Each drain waveform is a half-cycle while off.  For fixed drive frequency, if C is reduced a bit towards your first option of resonant with just series cap C1, drain waveforms will get shorter and higher, looking more like classical class E.  If C1 is reduced further until it alone is resonant with primary, then drain half-cycles would become infinitely narrow and high (if not for real limits of parts).  If C is increased above your middle option, then drain half-cycles don't complete before FET turns on again, causing excess FET stress and power dissipation.  So likely a bit towards class E waveforms is ideal for margin.

They use a parallel resonant cap and for whatever reason, it's not as good as having a series resonant cap

The big advantage of series resonant is higher coil voltage for a given input voltage.  I'm guessing a parallel drive would work with a low inductance single-turn primary (ie. copper sheet bent into a turn) and parallel plates/sheets of copper connecting driver to primary.  Something I hope to try some day.

[Magneticitist]

If I thought I could give a comprehensive explanation of the circuits I made using this topology I definitely would. I've been waiting for someone smarter than me to come along and really dig into this topology because it seems far too under-demonstrated. It seems like you've already fleshed out how it works. The most impressive iterations I saw were on Chinese websites (hard for me to even re-find again) and that's where I first saw the term Double-E. I've only described one of my coils as Dual Class E, or close to it, and it was more a random discovery than an intentional tuning. This was probably the 'dirty double hump' coil where the only reason I even got so curious was because I was wondering how the heck that thing was running the way it was without getting the mosfets hot. A video I saw of a nice little setup had a circuit that illustrated a choke and the series primary capacitor resonating. I think sometimes I would just use a primary capacitor to help adjust my feedback. A couple setups I tried dual tuning it just like I would a basic drsstc using secondary feedback. For me it was never really about Class E but just getting a higher pk to pk voltage on the primary for low voltage DC input setups that I realized could give nice sword spark outputs. At least, it would sort of appear that way given the spark length per input voltage and particular tweaking of the breakout point. I later realized I could land on a tuning that would allow me a good deal of adjustment on the primary coupling, where maybe just raising it a little higher could give me little swords, and lowering it would give me higher voltage branched output.

When it comes to the split primary circuit (Tefatronix/Skori) I always seemed to see about consistent results.. When switching to the single primary and 2 chokes I couldn't really say I ever saw a consistency. There just seemed to be too many ways to tune the thing. I've ran one of those at like 12-15V with a mason jar coil and single turn primary winding so loosely coupled that I could probably wrap it around my waist. I must have made and torn down about 25-30 of those things before finally settling on less than 10 I have left. When using gate driver chips instead of the feedback inductors I felt like I was kind of just unnecessarily complicating the circuit, but I don't think I tried any dual tuning then. The last one of these I made I thought worked out well and I meant to really get into scoping it and trying to figure out what variables were really coming into play relative to other similar builds I've made.. but then I kind of just shelved it IDK maybe just reaching a point where I thought "ok whatever good enough." I can easily get 6 or more inches of branchy arcs to my hand from it.. but then a similar circuit I could get longer arcs to the hand from and the output is all swords. Both are probably identical outside of the mosfets, pri/sec coils, and probably the drain-source caps or something like that. This is basically coming down to more of a real-world tuning to get the desired outcome as opposed to going straight from a simulation to success. I'm curious to know where your progress is at this point. MSLabs had maybe showed off his QCW setup since this post using a very nice little PCB he designed but I've never actually seen the circuit.

[YSPACE Labs]

Some more results of research and thinking:

Actually this circuit/topology is known as "push-pull class E", and there are some non-paywalled papers online about it. However, none give an easy-to-understand way to tune it. My best guess is that it's literally just two class Es put back to back with the series capacitance and inductance (C2 and L2) of both combined into a single series inductance and capacitance (so the dual class E will have a C2 value half of what the class E design equation come up with and an L2 value twice what the equation come up with). This is still completely untested and unsimulated, so I'm not sure (and you also will need to work backwards from your tesla coil's primary inductance to find C2 and C1). I'm also not an electrical engineer so 🤷‍♂️. I also know (the legendary) Richie Burnett has mentioned "double-ended class E" and there's a 4hv post where he details some of the circuit. It seems like C2 might actually not be a critical value (iirc from him and my sims), but C1 is absolutely critical to get ZVS.

My current theory and a rough idea of how to design it is (pasted from a Discord message) is this:
Transistor charges inductor to a certain energy, and then the capacitor gets charged to a certain energy equal to the inductor's energy (charging should take a little less than 1/4 of the period). Then the DS cap that is charged to like 30v or something and somehow transfers its energy to the inductor, capacitor, and resistor (it's not a resonant tank since the values don't form resonant tank, but it's something else). The energy gets stored in the not-tank circuit and the transistor(s) pump more energy into the tank every cycle (single) or every half cycle (dual).

However, some newer simulations with a single class E show that the series cap should resonate with the primary.

There are definitely EEs who know about this (one that I've talked to thinks it's a terrible topology for some reason dc-dc converters (according to him, it has multiple undesirable properties for dc-dc conversion, but I think some of those things don't really matter for low voltage SSTCs or can be considered reasonable trade-offs (like being hard to tune)). There might also be a connection to Slobodan Ćuk.

Currently, I've found 3 types of dual class E tesla coils from China
Chinesium products:
https://www.aliexpress.us/item/3256805512921612.html
https://www.aliexpress.us/item/3256805035455545.html
https://www.aliexpress.us/item/3256804522839720.html

I also got myself a cheap Chinese pancake Tesla coil from Amazon, and I reverse engineered it:



I think this topology is very useful for low voltage SSTCs because the inductors boost the voltage to the primary, it's efficient, and uses two low side mosfets that are easy to drive. Sadly, it's underrated.

I'm on Discord (Styropyro, Great Scott's Community, and United Tesla Coil Builders (LabCoatz server)) and this forum seems to move more slowly than a Discord chat so it might be better to get some more information and testing done over there and then post a compilation of the results here where it will get indexed on the internet.