sstc 1

i cannibalized an old coilform and toroid i made years ago, because they were some of the very first parts i ever 3d-printed and i wanted to put them to their original purpose. (some might say i started 3d-printing specifically so i could make nice toroids. some might say i have an obsession. i might agree with Some.)

the coilform is extremely small as tesla coils go (which proved to be a problem later), but i wanted something that would easily fit on a desk. mostly just to see if i could pull it off; i built a traditional spark-gap coil back in high school that threw meter-long bolts, and from that i learned that physically larger coils are easier to get good results from.

i originally wanted to wind this coil in #42 wire (totaling about 800 turns), since that's the smallest gauge i had on hand, but it turned out to be a real pain in the ass: i couldn't see where it was most of the time (it's much thinner than my hair), and it would cross itself while winding with the slightest provocation. after a few failed attempts, i decided to give up and use #38, which i've used for many years to great effect.

one thing i discovered: do not use double-sided tape or other adhesive on the start of a winding with wire like this! the wire wants a little bit of lateral play so that it can re-align itself as further turns are laid down by the machine (or by you, if you're insane enough to guide the wire by hand — it can be done…). i had a few false starts at #38 before i figured this out, but once i did, i secured the wire with blue painter's tape instead and things went beautifully.

after winding, i carefully secured the ends of the wire to the coilform with cyanoacrylate and dipped the coil in a thick cut of shellac in ethanol i like ethanol as a solvent for this sort of thing, as i find it smells better than isopropanol and the whole solution is technically edible. shellac is used in, among other things, candy. to secure the turns in place, after which i let the coil air-dry thoroughly.

after drying the coil, i remounted the toroid, along with a screw in the center to (hopefully) serve as a breakout point. spoiler: this didn't work and i wound up using a cut piece of brass wire instead, as it was much sharper.

on the other end, for robustness, i terminated the hair-thin #38 wire in a brass rung, heat-sunk into the end of the coilform with a soldering iron.

this isn't much of a section, because at any given time the primary coil was a few winds of hookup wire positioned carefully underneath the coil, without much structure to it. it turned out that proper primary inductance and primary-secondary coupling was critical to good operation of this coil, so i was always fiddling with it. i'll fix that property the next time i try a class e sstc (because of course there will be a next time, who do you think i am!?).

i built the driver around an irf510 fet (known for being cheap and having reasonably good hf performance) and a ucc37322 low-side gate driver (known for me having one on hand). as best i could, i followed the layout guidelines in the UCC37322's datasheet regarding bypass capacitors and low-inductance layout, but there's only so much you can do, you know.

i built the driver to take an external gate signal input, which i fed for most of this project from my function generator. this allowed extremely easy retuning while the coil was in progress, as well as easy experimentation with interrupter ratios, and i will certainly be doing this again! previously i've built the core oscillator from the get-go, but i now think that's a mistake. running a class e coil with an interrupter at all may also be a mistake if you're not playing music… but see later on this page. ;)

i had most of the parts for the driver on hand, but special-ordered some metalized polypropylene capacitors for the resonant load after blowing up one of my salvage caps. these are the capacitors of choice for tesla coils for a few reasons: pp has relatively low loss at the frequencies we're talking about, they're film caps and as such have low esr and high linearity, and metalized film is "self-healing:" if an arc breaks out inside the capacitor, the sudden discharge current will cause that arc to vaporize metal film near the site of the punch-through, marginally reducing the capacitance but giving enough insulation around the hole that the capacitor will continue to function. (or so i'm told.)

i also custom-wound the rf choke (the two wire-wrapped toroids in the above & below photos). magnetics you should really just make yourself for one-off projects like this, and i have a small supply of ferrite in various sizes and shapes for doing just this. basically the only properties i thought about for the choke were

1. the inductance (sufficiently high), and
2. the current rating.

this is a dc choke, so the current rating is given by the lesser of the wire capacity and the saturation flux density of the core — in my case, most likely the latter. by the definition of inductance, for my estimated 500 mA load,

\begin{align*} \Phi & = LI \\ \implies B & = \frac{\Phi}{\text{area}} = \frac{LI}{\text{area}} \\[0.5em] & = \frac{ 47\ \text{µH} \cdot 500\ \text{mA} }{ 10\ \text{mm} \cdot 4.3\ \text{mm} \cdot 2 } \\[0.5em] & \approx 270\ \text{mT}, \end{align*}

which is pretty safe for a generic ferrite i got these in a lot on eBay with no knowledge of what material they were. the reason i have these cores on hand is for winding gate drive transformers, which i didn't have cause to use here. gdts are real cool things which i'll definitely discuss another time, but i had no use for them here because i only have one power transistor. lol. at low frequency. powdered iron or electrical steel could go higher (≥2 T for modern stuff), but i don't have that on hand, so. it worked as a choke, more or less, but certainly didn't block the heavy 100 Hz current oscillation with the interrupter on, so i did a second driver revision with some large filter capacitors and a larger heatsink on the main transistor.

this is the revision the coil shipped with. once i had the thing tuned properly and in (roughly) its final configuration, i whipped together a 1.47 MHz lc oscillator to hook to it instead of the function generator.

i attached this oscillator after retuning the coil for cw operation, right before the end of the project.

both to prevent interference and provide a predictable topload capacitance (and also a little bit because i didn't feel like running an rf ground line) i stuck this lil' guy in a faraday cage. i initially made one using a metal screen, which was pretty easy and worked well (electrically), but obscured the coil more than i wanted it to.

since part of the goal of a tesla coil is, necesarily, aesthetics, i made a new cage out of steel wire by hand. eventually i'll write a post on these (and do more formal testing than "does an antenna attached to my scope pick anything up outside the cage versus inside it?"), but for now have a photo.

first of all, i had to determine the resonant frequency of the secondary. i did this by either

1. (outside the faraday cage) sticking a scope probe nearby and grounding the base of the secondary with the scope, or
2. (inside the faraday cage) sticking an antenna (just a length of hookup wire) through, grounding the base of the secondary to the cage and the cage to the scope, with the probe measuring the signal on the antenna,

then sweeping the primary with a function generator and looking for a peak. this worked pretty well, as far as i can tell, giving me relatively consistent results for any given setup and correlating well with the observed ideal frequency to drive the coil at for the biggest sparks. the astute reader will notice the steel sheet underneath the coil and wonder whether it detuned things at all. having tested this setup both on the sheet and on the bare wood desk beneath: it does not appear to have.

i did take off the toroid at one point, due to reports that a large top load isn't really beneficial for high-frequency coils like this one, but that increased f_res to something like 10.2 MHz, which was too high for my tastes. so i kept it on.

the main thing to know about primary tuning for these coils is that coupling is critical. class e coils benefit from relatively close coupling: the resonant tank is not left to ring as it is in a drsstc, so you really do want to transfer as much energy from primary to secondary on each cycle as possible, which equates to high mutual inductance. but if you have high mutual inductance, you also have high sensivity to the exact configuration of the primary, secondary, driver, housecat, phase of the moon, etc., because a slight disturbance will change the coupling.

i would recommend scrapping the separate-primary idea entirely and treating the secondary as the series lcr circuit that constitutes the usual class e load, driven from the 0 V end. i haven't tried this, but will be for the next class e coil i build.

once i knew f_res, i had to tune the driver. mostly this involved scoping the gate and drain waveforms at low voltage and adding/removing turns from the primary until the waveform was correct for class e (or close enough — hitting zero \(dV/dt\), the "just-kiss-zero" condition, is hard, but you can pretty reliably hit zero voltage on turn-on). i got pretty close, though:

extending the pulse length allowed some actual resonant rise on the secondary:

that drain voltage spike at the end, i suspect, is partially the current in the choke dropping and partially the secondary backfeeding the primary. it's only a problem if you're running the coil in interrupted mode, but if you are running the coil in interrupted mode you should really think about that! it was the limiting factor to input voltage for me, given that irf510s are only rated for 100 V.

one issue i ran into when switching to a larger form factor capacitor was ringing on the drain:

through simulation and some careful guesswork, i traced this to about 100 nH of parasitic inductance in series with the capacitor, due to the length of the fet pins between it and the cap. this is not the sort of thing you can remove — it's necessitated by the geometry of the circuit, so if you add a ferrite bead, an extra capacitor, etc., it only gets worse.

but if you think about it… that ringing is due to a tuned lc circuit, which requires some harmonic of the input signal to be nearby enough and sufficiently large to cause ringing… and the only place that can come from is the effective square wave at the drain itself. so i added a 10 Ω series resistor to the gate of the fet to slow down its switching times, and…

…there went the issue. there are reasons gate resistors exist. (this is not the only reason, but it was a surprising enough one that i thought it merited some exposition here.)

here's a bunch of shots of the coil running on 12 V: