Re the core: I was under the impression that the core was generally considered live. Is there scope to (with appropriate engineering) argue that the core is insulated from the primary to the extent that the core itself is consistent with Safety Extra Low Voltage?
Hmm, I can't think of anything that would necessitate that. I understand the impression -- but I can't think of a reference where I've seen it claimed, either.
At least from first principles: if everything is properly insulated (REINFORCED type), what's it matter? It's just another conductive bit of metal (or ferrite semiconductor). I don't recall 60950 making any special cases about it, it should be handled like anything else.
There are good reasons to ground the core, of course -- for safety (primary-core insulation can be BASIC type), or for EMI (often the noisy primary is in close proximity to the core, thus some capacitance couples to it). In the latter case, grounding might amount to simply shielding it, which can also be combined with a flux band for example, which offers another kind of EMI shielding too.
So, I don't think that's a problem?
Heh, speaking of shielding, besides a shield over the board, or like a heatsink plate or something (which might be necessary anyway, given the power density these designs can offer), you could use an EMI spring to touch the core. There's lots of kinds available, I bet there's something to handle horizontal contact in a low profile...
Re coating: there definitely are dielectric coatings but I’d rather avoid wet processes.
Does anyone have experience with something like a stick-on film? I know Kapton tape has great dielectric strength but rubs through...
Hmm, I wouldn't say it rubs through? It's pretty tough stuff, though tape being tape, it could rub off from an edge. Or, if you mean with enough time and vibration against a loose (i.e. not cemented?) core, yeah, that'll happen eventually.
Tape has the problem that it's not likely to sit perfectly flat on the board, and any void is a direct creepage path. The insulation is not the film itself, but the adhesive holding it there. (Polyimide is a magnificent insulator by itself, at least!)
Not sure what if anything, standards have to say about tape. It seems a bit dubious to me, but maybe it's acceptable after all.
Hm, I forget if there's consideration for hi-potting something at enough of an excess voltage to satisfy concerns about creepage. That is, test it at a voltage where any creepage path is certain to flash over, and if it doesn't, well the tape must be doing its job.
Would that even make sense, though? What about exposure to a slightly salty, condensing environment, such that creepage paths might become filled in with water?
So, Idunno.
Note that, in conventional transformer construction, tape is applied such that it can be assumed to have gaps between layers -- enough tape width, overlap, and extension over the edges of the windings (using margin tape to reserve space if necessary) is used to ensure the worst case creepage path is still long enough. And then it's filled in with varnish, which is probably not a guarantee by itself -- it's effectively potting when it works, but how do you inspect a winding, y'know?
Best practice would seem to be, do everything you can, starting with absolute minimum requirements (i.e., overlapped tape, leaving margins), and make it better and better from there (the adhesive will seal to some extent, the varnish even more, etc.).
Re extra prepreg: that’s an interesting thought. If it counts as a cemented joint, it might be possible for a PCB manufacturer to laminate it in while doing the board stack.
Oh-- I think you can special-order certain kinds of coatings and laminates, uh, who knows how that's going to break down in terms of cost, labor, special ordering -- but that might end up cheaper than doing whole-ass extra layers on things. Just slap down a square of prepreg around the affected area, run it through the press again, then off to the router. Or... well, with all that special ordering, who knows, a semi-proto 6/8 layer board might still be cheaper. In quantity? Who knows.
Also a good reason to consider using planar transformer components. You can get hundreds of windings on a single panel, and you can stack them vertically in parallel (given enough core height) for arbitrary ampacity. The assembly can be potted, guaranteeing ratings. And it can be soldered into a much cheaper, larger board in the end. But that's also additional labor making whole-ass components, so, tradeoffs of course.
Tim