redpixel wrote:I'm usually a big Centuriontech fan, but seems like this design might have been a bit flawed... However I think it's unfair to accuse him of using "cheap Chinese" switches/power modules, in my experience he always uses quality parts.
...
I agree that it's risky to route mains and a 3.3 V logic-level signal through the same switch though, and it would've been nice if he'd have put an optocoupler or something in-between. It's a bit worrying that the PSU used in the blown-up computer which is on its way to you exxos seems to have been the 2025 version. Will be interesting to hear your findings!
Sorry, missed this post somehow, thanks for digging those part numbers out.
To clarify, I wasn't accusing him of using cheap Chinese parts specifically, that's just where my mind naturally goes when I see something failing repeatedly, because that's usually the pattern with failures I come across. I don't actually know what he used or where it was sourced from, I was just asking the question. A proper branded part like an Alco switch isn't something I'd expect to be the weak link either, so that's useful info to have confirmed.
Good spec on the switch too, 1000 Vrms dielectric strength sounds fine on paper for isolation between the contacts, but that only covers the switch itself in isolation. It doesn't help once you've routed a low voltage logic line right next to the mains tags on the same body, that's a PCB/wiring layout problem rather than a component quality one. Even a well rated switch can arc or flash over to an adjacent pin if the clearance and creepage distances around it haven't been designed with that in mind.
Worth throwing into the mix as well, this is pure speculation on my part, but we don't actually know if those power modules, or any other modules used across the various revisions, produce back EMF when they're switched off. That can rise to several thousand volts depending entirely on how the module itself is designed internally, and it's not something I've looked into personally. Only one Falcon so far seems to have actually died, with no other reports of the switch itself failing as a component, so it's worth considering whether a turn-off spike that isn't being suppressed could be behind it, rather than the switch being at fault at all. Just putting it out there as a possible explanation alongside the flashover theory, not a conclusion.it needs investigating.
Once the Falcon and PSU arrive I'll get the switch out and have a proper look at it myself, be good to confirm exactly what's there rather than going off photos.
Only other thing is I've not been well the past few days, so it may delay me getting into the full inspection, just a heads up in case it takes a bit longer than expected to post an update.