According to Andy Hill :
(Doug Miller) wrote:
In article , wrote:

Intermittent ground in the flaky outlet box? Offhand, you'd think that since
a ground isn't necessary for GFCI operation, you wouldn't need one for testing,
but I'm at a loss (at least until the caffeine kicks in) how one could test
GFCI operation without a ground to fault to...

While that might explain the test button on the GFCI not working, how about
the test button on my plug-in tester? That tripped the GFCI _every_time_.

Gah. OK, how 'bout there's already some ground leakage from the neutral to
ground (capacitive coupling or something), the "test" button on the GFCI is set
to fault the hot to ground, and the "test" load is close enough to the hairy
edge that it can't quite swing the differential current back over the "trip"
level. The tester, OTOH, uses a larger load, so it swings it all the way over
to the trip level. Now, how the heck a neutral (even a floating one) could leak
enough current to ground for that to happen is left as an exercise for the
student (meaning, I can't really think of a way).

A GFCI's test button is designed to trip the GFCI even without a ground. It does
this by shorting one of the output leads through a resistor to an input lead, or
something like that (so the output leads are imbalanced).

A GFCI tester cannot do that (because it doesn't have access to the "input leads"),
so it simply runs some current from the hot wire to the ground (and hence it
needs a ground to operate properly).

It seems to me that if the circuit has an open neutral, the GFCI test button
wouldn't trip it, but the standalone tester _would_ trip it.

OK, so I'm blowing a smoke cloud that can be seen for miles. Damned if I can
come up with a reasonable explanation, either...


--
Chris Lewis, Una confibula non set est
It's not just anyone who gets a Starship Cruiser class named after them.