BigWallop wrote:


So, I'll start again. A neutral / earth short circuit will not cause an RCD
to trip.

It does, demonstrably, and without any further faults.

Your mental model of what an RCD does is all at sea, I'm afraid. Since
it's a CURRENT BALANCE device, it's easiest to see what happens in an
N-to-E fault in terms of CURRENT.

By "it's a CURRENT BALANCE" device, what I mean is this: it compares how
much current goes 'out' from its L or 'phase' terminal, to how much
comes 'back' to its other, N terminal (both these on its load side).
While wot-comes-back is within 30mA of wot-goes-out, it's happy; if the
imbalance exceeds 30mA, it trips. This contributes to safety, since that
30mA of current is going somewhere it oughtn't to - just maybe through a
person; 30mA is chosen as it's around the current needed, given typical
body resistance, to be approaching Serious Consequences.

When there's an N-to-E short, there's a second current return path
created: the nature of this return path and the direction of the
imbalance vary according to the source of the 'outgoing' current and the
earthing arrangements.

In the case most relevant to the OP: the RCD is supplying the load (here
a washing machine) with the 'outbound' current off its L terminal, and
wants to see it all come back to its N terminal. With no N-E fault, this
is just what it gets. (Indeed, if the E for the circuit/socket/plug
supplying the washing machine becomes disconnected, the RCD is *still*
in balance, and doesn't trip.) Now, if an N-to-E short happens on the
circuit supplying the washing machine - or inside the washing machine
itself - all the current which went 'out' to the washing machine now has
a choice of routes 'back': the 'original', intended route, back through
the RCD's N terminal (and back along the relevant CU's N busbar, the
black meter tail, and back to the substation/transformer where the
supply's L side comes from too), but also a new route through the 'E'
conductors (the circuit-protective-conductor in the T&E cable supplying
the washing m/c), crucially BYPASSING the RCD's N terminal, back through
the mess of E wiring around the CU, and finding its way back to the
substation and points closer to the installation through the many paths
between the installation's main E terminal and the supply's N. (In the
case of a PME (TN-C-S) installation, there's a local low-resistance path
right at the supply entry point; in the case of a TN-S installation the
supply E and supply N may not be in good metallic contact until we're
back at the substation; in the case of a TT installation there's only
the 'mass of earth' path back to the originating N.)

So, how much of the load current coming out of the RCD's L terminal
'chooses' the 'balancing' path back to the RCD's N terminal, and how
much goes the 'wrong', 'imbalancing' way? It depends completely on the
relative resistance (impedance, if you want to get picky) of those two
paths. For a PME installation, as outlined above, the 'imbalance' path
is all within the house, and is almost as low as the 'balancing' path -
a little higher, since CPCs are usually a bit thinner than N conductors;
so for a 10A load you might find 7A going back the 'right' way and 3A
going the 'wrong' way. The RCD will certainly trip in this case! The
TN-S case isn't that different, except that the nearest link between the
installation's E and the supplier's N isn't within the house, but might
be as far back as the substation/transformer: but both paths are
low-resistance and metallic, so again you can expect a substantial
proportion of the load current to 'choose' the 'imbalance' path, and
again the RCD will trip. At the other extreme, for a TT installation,
with a relatively high earth electrode resistance, the 'imbalance' path
could be at least 100 times higher resistance than the 'balancing' one,
so an on-load N-to-cpc fault isn't nearly so likely to cause a trip.

The analysis is similar for the related FAQ case of 'I was working on a
ring circuit I'd turned off the MCB for, touched its N to its E and all
the RCD-protected circuits in my house went dead'. Here the just-created
N-to-E short provides an alternative path bypassing the RCD's N terminal
for all the loads in the *OTHER* circuits the RCD is supplying, since
the N in the 'isolated' ring isn't isolated, but connected to all the
other N's through the consumer-unit busbar. Andy W's essay covers those
cases in detail, so I won't go on any more.

Whether this is enough to convince you that your 'neutral potential'
model is misleading, I can't tell. I assert again that a 'pure'
current-balance RCD doesn't give a monkey's about the relative potential
of L, N, and the installation's E (to which it's not directly
connected), though obviously you need to have an idea of their
potentials to see whether any current will flow among them. (Old-style
voltage-operated earth-leakage breakers are different, and some 'modern'
electronic RCDs have further refinements beyond the pure 'current
balance' idea; but that's another discussion). I hope it at least gives
you some pause for thought, and might even inspire you to do a (safe ;-)
bench experiment with a small load, a current-balance RCD, and
selected-resistance N-to-E faults ;-)

Cheers, Stefek