update on the ~100mhz ringing:

thought about it some more, and did some more poking at LTspice. While the previously highlighted axial resistor might well be inductive, it's hardly conducting any current. Not nearly enough to produce that kind of oscillation.

My new suspect: the parasitics on the mosfet itself (plus a bit of pcb trace) are acting as resonant LC circuit. Turns out this is totally a known phenomenon if external gate resistance is too low. I had been thinking of my 2Ω2 gate resistor as part of a phase-shift RC circuit (that's how it was presented in previous work I'm referencing), but it also is dampening the parasitic LC resonance.

Here's a video of a lovely gentleman going in-depth on mosfet gate ringing and necessary resistor values

https://www.youtube.com/watch?v=6plZ2iOI0d4

should be able to test this tomorrow.

and I'm... having this feeling like I've rapidly departed from "dabbling at some electronics" and entering territory of "this legitimately qualifies as electrical engineering".

it's a funny feeling.

goddammit blew up my last FET.

What I'm tempted to do is order more of them and keep fucking around.

What I should actually do is redesign this board based on lessons learned and include some more self-protection features.

problem: no overvoltage protection on the mosfet gate
plan: add zener/TVS clamping diode

problem: mosfet failure to oscillate creates possible overcurrent condition
plan: current-measuring shunt + op-amp + etc to force mosfet gate low if device overcurrent

problem: layout is kinda bad for parasitics.
plan: make the layout less bad.

swear someday I'm gonna do my best to explain wtf is going on for all the non-electronics critters who so kindly follow me

in the meantime this is raw projectblogging

me: what's the deal with HV creepage (arcing/breakdown) distances under surface-mount capacitors? Seems to violate IEC design guidelines for minimum space

high-spec capacitor mfg: Hey here's some information about this exact question.

https://www.johansondielectrics.com/impact-of-pad-design-and-spacing-on-ac-breakdown-performance

like this is something that seems particular for electronics part manufacturers, even moreso than most industrial products:

they go out of their way to show you *exactly* how best to use their shit.

It makes learning about how to actually make stuff so, so much easier. You might need to wade through a forest of datasheets and application notes, but information is the opposite of obscured.

what people think electronics design is: routing pcb traces, brandishing a soldering iron

what it actually is (much of the time): squinting at poorly-photocopied performance curves in AZ23C51OMGWTF.pdf with calc.exe open on the 2nd monitor.

All right, so, I have this mosfet which I expect to be dissipating a fair bit of power (several watts, potentially >10W). In the previous prototype I used a THT part (SO-247) with a bolt-on heatsink, which worked pretty well.

Ultimately I'd prefer to use a surface mount part, and the device is also available in D2pak-7.

My plan was attach the tab to an oversized pad, drop an array of thermal vias to the ground plane on the opposite side, and mount the entire (single-sided) board's ground plane on a shared heatsink.

The problem is the thermal tab on this device isn't connected to ground, but instead is the pin for Drain. So I can't drop thermal vias to the ground plane.

Is the expectation that I use a top-mount heatsink, dissipate heat through a big power plane for Drain on the board's front, or what?

datasheet https://www.qorvo.com/products/d/da008707

Here we go, a Revision 2 of the board! I think it's ready to send to the fab, but inevitably I'm gonna find some small things to tweak after posting.

Light blue rectangle is the outline of a big heatsink on the back side of the board.
Thanks to @gsuberland and @electronic_eel for their input on thermals :)

new boards already arrived!

ngl I'm a bit spooked by how fast and I
inexpensive it was, even though there's pretty explicitly some kind of promotional/loss-leader pricing going on.

Also these are 0.8mm boards (half of "standard" thickness) which should help with both front-back thermal conductivity and make the two-layer inductor a little more inductor -y.

oh yeah, seen in action here is a hot tip I'm only recently internalizing:

if you're making a thing and it has screw and/or mounting holes, mark the screw size/spacing directly on the object if at all possible.

it is with profound excitement that I'm able to finally unshroud the project I've been working on:

xenon plasma toroid

tried to run the toroid off of my beefier bench supply and blew a mosfet.

It's looking like the driver is probably drawing a lot more current than simulated. The miniware psu (and the USB-PD supply) I used yesterday handles brownout pretty gracefully, but I suspect the bargain bin bench PSU is less graceful.

Part of the higher power draw could be due to messy switching, but if I lower the primary cap values in the sim to 66pF (from the nominal 99pF) I get results more similar to the observed circuit. I don't think the frequency or temperature is high enough to cause that kind of de-rating but who knows, could be as simple as one of the three capacitors has failed.

Going to order some more caps, this time explicitly marked as low-ESL.

Still not sure what's up with the coil getting blazing hot.

...not the exact capacitors I have installed, but similar.

Using a bank of only three is *definitely* asking too much, even if I think things aren't quite as bad as 300°C/W. Note the log scale on x-axis.

Sincerely though, does anycritter have ideas on why the two-turn 1.9uH PCB coil gets so dang hot?

According to the saturnPCB calculator / IPC 2152, I'd expect less than 30°C rise to steady-state even taking skin effect into consideration, based on my ltspcie sims expecting ~12 peak amps at ~13MHz (sinusoidal).

I see like 70°+ temperature rise within a minute.

Guessing there's something going on with inductor energy getting dumped as heat, so resistive load calcs go out the window. But I have no idea how to reason about those possible effects.

lol so ended up just using I^2 *R just given DC resistance of a few tenths of a ohm and amperages in the 6-10A range...

that's multiple tens of watts of heat.

Very rough estimates of thermal mass give many degrees per second rise.

...yep there's a problem.

So the standard trace width calculations just aren't remotely applicable here apparently!

I now slightly better understand why other folks haven't used PCB inductors here.

I suspect I might still be able to donut with much lower amperages though! If I get down below 3A in the inductor, that's only like four watts heat rather than dozens.

More progress on the plasma toroid! More videos! Less overheating!!

Writeup at https://hackaday.io/project/194683-plasma-toroid-sky-guided-pcb-edition/log/227790-tweakin

Higher-quality vid at https://www.youtube.com/watch?v=rEc76ckTJcg

More plasma toroid progress! Things are going super well!

Video demo:
https://youtu.be/fBrOaj5pBa0

Project log:
https://hackaday.io/project/194683-plasma-toroid-sky-guided-pcb-edition/log/228101-version-03-is-calm-cool-and-collected

I'm at a point where I could call this "done" if I wanted.

It's not done yet though ^_^

roflll, I'd been working hard to implement a subcircuit for monitoring current draw to indicate out whether or not the primary inductor was in oscillation, using a shunt resistor + op-amp + comparator etc.

And just now realized I can do the same thing *better* using nothing but a loop of wire and an LED.

Plasma toroid project writeup has been updated, with details of the non-contact indicator light shown in the previous video clip.

"status LED, galaxy brain style"
https://hackaday.io/project/194683-plasma-toroid-sky-guided-pcb-edition/log/228373

ANOTHER plasma toroid project update: Pushbutton Arc Start!

https://hackaday.io/project/194683

This one is a big deal, since I've now hit every core functional milestone. The next steps are putting all of the subsystems together into a single, unified board.

This subcircuit is also neat because it uses multiple different bits of antiquated technology: photo-flash trigger circuits and gas discharge tube surge arrestors.

(demo vid attached here; high-quality available on youtube https://youtu.be/Vi6Pb0WSCZw )

Like I spent literal hours today trying to pick out a heatsink.

Digikey has literally one hundred *thousand* heatsink SKUs from one vendor, and they're all permutations of solution space for the same basic design.

Neither digikey parametric search nor the manufacturer website narrowed things down very well. Even after several different dead ends to figure out what I actually wanted, I basically had to manually decipher the under-documented way this mfg was assigning model numbers.