on a scale of 1 to stupid

[solarbird]

So in what might be a very dumb idea I've realised I can block the DC on the UPS ground with a good quality (say, 250V film) capacitor. It'd still pass the AC, but it'd contain the DC leakage until it gets so high that the capacitor bridges and it becomes a DC conductor as well. That would preserve the safety function of the ground pin, at least, in theory.

This is the UPS that occasionally trips GFCI. As far as I can tell, APC expects a percentage of its units to trip GFCI - they're more up front about this in documents meant for Europe, which implies they don't have good control over the issue. But this one's long past warranty.

(In Europe, the policy is not to answer questions but just replace the unit no questions asked... and hope, I guess, that the replacement is leaking less. It does seem to be rather variable. Our longer-term situation, fwiw, is going to be "fewer UPSes." As these old ones give out, we'll move to a single UPS for the tower. I don't like it, but it seems least problematic in the long run.)

On a scale of 1 to what the fuck is wrong with you, how bad an idea is this?

Comments

[wrog]

so this may be a stupid question but is there a chance that the capacitor could defeat the GFCI? (i.e., your UPS develops a ground fault somewhow [say, somebody spills water on it?] and the GFCI won't trip the way it's supposed to because the capacitor is there?).

Or is GFCI basically useless for protecting UPSes from anything? (and therefore pointless to have on that circuit in the first place, and the capacitor is a way easier fix than messing with the house wiring...)

[solarbird]

It puts enough current on the GFCI to trip it occasionally, but not regularly. My thought is that with the cap, those incidents will be limited to AC only (where the voltage being output is much more stable for whatever reason) and let it get past those.

The new circuit is GFCI because a combo AFCI/GFCI breaker was the cheapest, easiest, and most of all, most inspector-friendly way of solving the various code requirements of the two outlets in question. (AFCI required on either one or both - I think both but maybe not because garage, have to check again to be sure - and GFCI required on the one in the garage.)

The more complicated solution would've been an AFCI breaker and single GFCI fixture, but that would've cost more. The cheaper solution would've been no garage outlet and AFCI-only, but I didn't think this would be an actual problem.

Now to get a little further into it: GFCI doesn't actually monitor the ground pin directly. So it won't even know if something is up. It's monitoring leakage by comparing voltage levels across the neutral and hot lines. By defeating the apparently-variable-DC-leakage from some shit component in this thing that I can't even find (because I've checked everything remotely obvious and some things not so and at this point I'm at a bit of a loss) and confining it this way to the UPS, I'm trying to reduce the voltage loss across the hot and neutral pins without defeating the ground pin entirely. I'm thinking (hoping) that the AC voltage is just induction, and the DC is the actual problem.

But by using a cap instead of something dumber (like a ground lift) I'm keeping the ground in place for AC entirely, and it'll just take a tiny bit longer for the capacitor to bridge and start conducting DC if the levels actually get anywhere close to dangerous.

Or that's my idea, anyway.

It's not pointless to have GFCI on UPSes, though. But mostly that's done in data centres where the UPS system prices have another 0 on the end and are smarter/better/more capable and _expect_ GFCI, whereas this thing, while on the high end of the consumer level, is a consumer-level product and isn't really made with the expectation it'll ever see GFCI, because who plugs in computers in the kitchen or bath? But surprise, now they're seeing more GFCI circuits because in new construction these days contractors just fill the box with combo breakers, and they're gonna have to do something about this eventually.

As for the water scenario: that'll create an arc fault as quickly as it'll create anything else, and the AFCI will catch that just fine. It doesn't rely on the ground pin either. If we all had GFCI/AFCI, we could pretty much give up on ground pins altogether. They're both better solutions and make the ground pin kind of redundant. (Not that redundancy is, like, bad.)

(I also also think that there's a problem of too many switching power supplies on the one circuit. If what I've managed to glean from various sources is accurate, individual fixtures have to trip at like 4ma, but breakers have to account for vaguely leaky kit across the entire circuit, and I read that it's generally set to trip at 30ma. And I think this thing is crossing the border sometimes, probably when charging the battery (tho' that's a hunch) and _may_ never reach the trip point on its own. It may have to be on a circuit with a bunch of _other_ switching power supplies. Like those found on computers in a server rack. I haven't tested all these compies. I did remove one power supply that I did test and was Not Being Good, and the rest all have nice modern shit, but still.)

Anyway. That's about what I can manage while roofers are here. Does it make sense?

[wrog]

so today I went to wikipedia to learn about AFCI.

(...It's sort of annoying that, even having had the physics classes at Princeton, including the sophomore level advanced E&M track where they dial the firehose up to 11 as a last filter to discourage people from majoring, it seems I know way less about electricity than I think I should (e.g., having any kind of clue how one would design a circuit that does what GFCI and AFCI are supposed to do....)

... or maybe Physics and Electrical Engineering are different subjects after all. Who knew?...

... and now I'm reminded of my friend, the grad student in Control Theory who burst out laughing at my expressed opinion at the time that classical mechanics was basically a dead subject seeing as Lagrange, Hamilton et al had already mapped out everything that matters 200+ years ago.... "so umm, we're trying to make this Actual Satellite stay pointed in the right direction, and it's kind of a Hard Problem..."

... or maybe I'm just annoyed at the Princeton Physics contingent that landed at Microsoft in the late 1980s [when they bought Nathan Myhrvold's company] and ended up molding a goodly chunk of the corporate culture there -- just in case you were wondering where that rapid-fire think-on-your-feet "quick, tell me how much water runs out the mouth of the Mississippi in a day" interviewing style came from.

... and then there's the guy who taught my intro Quantum Mechanics course (William Happer) who is now a major-league climate-change denialist...)

[solarbird]

That was a hilarious number of edits.

(11 email notifications! One comment. xD )

For whatever it might be worth, I'm quite sure AFCI came about well after your grad school experience. ^_^

It's always disappointing when someone good at something turns into a complete lunatic. I've seen it happen more times than I like.

I've also seen it happen to people I know.

It's terrifying.

[wrog]

That was a hilarious number of edits.

It was the middle of the night and my theory is until someone gets around to replying, it's fair game for editing. And I've had people get upset at me for posting lots of little comments when one big one will do (though I suppose that might be more of a FB phenomenon).

It's always disappointing when someone good at something turns into a complete lunatic.

I'll note I hadn't thus far actually credited him with being good at something, though, to be fair, getting tenure at Princeton is no small thing, and I'm sure if I had questions about atomic physics and lasers, i.e., his actual fields of expertise, I may well be giving his opinions due deference. He kind of sucked at teaching though; though also to be fair, that sort of thing wasn't unusual (out of maybe 50 people on the math faculty, there were 3 or 4 that could actually lecture their way out of a paper bag; it simply wasn't what they hired for; physics dept. tended to care more about this but that's not saying a whole lot), and he was arguably kind of doomed having to take over the course from Peebles, who was going to be a hard act to follow no matter what.

My real beef with him is this was intro Quantum, where ALL sorts of questions come up (what are the axioms? why do commutators matter? why Poisson Brackets? Copenhagen vs. Many-Worlds vs. Transactional?) and he was basically an experimentalist who didn't give a shit about any of that stuff ("here's how you compute this scattering cross-section... and here's how you compute this other scattering cross-section... and here's how you compute yet another scattering cross-section",... because that's what they do in the particle accelerator labs), and I dunno, maybe the curriculum was constrained and they didn't have a whole lot of time for philoposphy wanking, or maybe the problem was with me in that this was the point where I was figuring out I didn't want to be a physics major after all... the cause & effect is a little hard to sort out.

I also can't say I was that surprised -- having already seen Roger Penrose (one of the gravitation/cosmology gods) riffing on Artificial Intelligence and getting basic CS wrong -- at seeing Happer go off the rails on climate science. There's something about physicists -- I guess when you work in a field that's the foundation for Everything Else, it's an easy trap to think you therefore know everything about Everything Else or at least can work out the Stuff That Matters on the back of the envelope, and they're sufficiently bright people that this works out more often than it should -- so that when they get really, truly wedged down some rabbit hole in some field that's Not Actually What They're Good At, there's nobody they're going to trust to lead them back to sanity.

[solarbird]

That's two.

Replying locks edits on Dreamwidth so you can't anymore, which means for you editing is fair game until you can't.

lol.

And yeah, getting tenure at Princeton counts.

[solarbird]

Also, Happer's approach to lecture reminds me of the single most useless and incomprehensible lecturer I had in grad school, whose brutally obfuscating classes on data theory managed to make the simplest fucking CS subfield in the curriculum into a mystifying bombardment of obscurely-muttered content-free demi-sentences. It was like being bored to death by having your skin scraped off with glass shards and I learned absolutely nothing.

[wrog]

AFCI came about well after your grad school experience

well I was talking about my undergrad experience, but in any case, wiki seems to imply that AFCI was first developed maybe 10-15 years ago, so I'm feeling slightly less stupid.

[wrog]

although I'm also told that getting tenure in Physics in the 1960s was way, way easier, what with the boomers swelling the ranks of the undergraduates creating huge demand, with relatively few of them having yet made it through grad school (assuming they were even going to grad school -- the NSF hadn't yet started in with its annual chant of "OMG There's gonna be this huge shortage of STEM people by the early 1990s when the boomers are all going to retire"), and all of that post-Sputnik NSF and DoD funding being handed out like candy.

(file under: why I went into CS rather than math or physics, because CS in the 80s was more like physics in the 60s, and physics in the 80s, well, wasn't -- witness all of the physics PhDs from my generation who ended up at brokerage firms and places like Microsoft (i.e., Not Doing Physics))