In uk.d-i-y, ARWadsworth wrote:
Whole installation RCDs are to protect against fire not human contact.

I thought MCBs protect against fire (overload protection) and RCDs are for
human protection

Both of these shorthand formulations have the potential to mislead. The
specific reason for putting in 100mA RCDs on TT installations (those where
the electricity company provides no earth, and there's a local earth rod)
is to kill the power in the event of a fault-to-earth quickly enough.
The reasoning goes like this: with a good, i.e. low-resistance, path to
earth in the case of a fault-to-earth, a big enough current will flow to
cause any fuse or MCB to blow/break in a Short time to avoid the voltage
on nominally "earthed" equipment being elevated for Too Long a time.
"Short" and its kissing-cousin "Too Long" are defined quite precisely
in the Regs, with the two most relevant times being 5 seconds for fixed
stuff (lighting circs, immersions, cookers) and 0.4s for socket circs -
'cuz hand-held stuff gets plugged in to those. For low-resistance
earth paths, quite enough of a current flies through the fault path to
blow the protective devices in good time; for installations relying only
on a local earth rod, the earth impedance may be too high and variable
according to soil conditions to be reliable. So you stick an earth-leakage
protection device in to assure disconnection times even with a sometimes-
too-high earth rod resistance.

There *is* a fire/cable-heating aspect to these disconnection times: while
there's a fault current flowing, it's massively more than the current
which the earth conductor (that's the Circuit Protective Conductor in
carefully-correct current terminology) could ever carry for a Sustained
Time. Again, the Tables in the Regs - based on the known resistance of
copper cables - account for this quite carefully (i.e. it's not just a
handwave, there's a quite detailed model of how much temperature rise
will happen, incorporating the eminently reasonable assumption that none
of this heat can get out of the cable in the short times we're considering).
If the fault current were to last too long, the cable would be dangerous
for future use (e.g. insulation would've got soft and conductors would
touch or nearly-touch which are supposed to be kept apart by that
insulation) - long before the cable actually Catches Fire in the manner
beloved by Hollywood special-effects people! The Tables also account for
how hot different cable types can safely get while carrying such fault
currents - so, for example, at two extremes of that range are normal PVC
cables (insulation gets unacceptably soft at, what, 90 degrees C?) versus
MICC cables (bare copper live-and-neutral conductors bedded in a mineral
insulation, copper sheath on the outside as the earth, often seen surface-run
in churches and other buildings where you have to surface-run and want
cables as thin and un-ugly as possible: a Win for these calculations both
because the mineral insulation doesn't go soft at wimpy temperatures like
mere water-boiling-point, and because the earth conductor is of bigger
cross-sectional area in these cables than the live conductor).

So: the disconnection time is what the ELCB is there to assure. The
disconnection time in turn is important for two reasons: one, to limit
the duration of the exposure of any person to higher-than-earth voltages
on anything nominally "earthed"; and secondly, to limit the time during
which cables have to carry fault current.

Hope that helps some - Stefek