Terry Pinnell wrote:

So, let's get this straight:

The MCB in position 1 on your CU (just that and nothing else) tripped.
In doing so it cut power not only to the immersion heater, but also to
your computer and some other appliances?

To restore normal operation you reset that single MCB for the immersion
heater, and did not need to touch the RCD itself?

If this is the case then you have a serious wiring error that needs
immediate attention.

Although that's exactly what I recall happened, I'm now beginning to
doubt it even myself! I could swear I didn't have to reset the RCD
switch itself as well. But, as you and others have said, that doesn't
seem to make sense.

Indeed, and its about to even more strange!

The following may be unrelated, but I'd appreciate it if you and
others would consider the following additional information to see if
it offers any new clues.

Earlier this morning I decided to do some methodical leakage tests. My
intention was to introduce a leakage from the live connection of the
circuit under test to a convenient earth, and observe what tripped.
That would at least confirm that I understood what behaviour I was
*supposed* to get from my Crabtree StarBreaker CU. Maybe it would also
reproduce that bizarre behaviour I described. I used a couple of
robust 15k resistors in parallel, theoretically giving me a leakage
around 32 mA (in my UK 240 V circuits).

I decided to start with a circuit which also had an extra RCD unit
over and above the main CU. This is a double socket RCD in the garage,
fed from #3 on the Main Switch (non-RCD) side of the CU. These power

You have two #3's on your photo, one is on the main switch and is
labelled "lighting" and has a 6A MCB (i.e. #3 on the left), and the
second one is on the right hand side of the CU and is labelled sockets.

I trust that the socket you were testing is not fed from the lighting
circuit, and so is in fact powered from the right hand #3?

If this is the case, then that socket *is* on the RCD controlled side of
the CU and not the non RCD side.

(Sorry if this sounds like I am teaching grandma to suck eggs, but I
need to make sure we are using the same terms): The main switch it the
big two module wide thing on the extreme left, with a red switch marked
cat No 100/MI2. This switches power to the whole CU. The five MCBs
adjacent to it are the non RCD protected ones. The three module wide
thing in the middle with the test button is the RCD, and this protects
the next five MCBs to its right.

all sorts of stuff in my garage and shed/workshop. As expected,
momentarily connecting this leakage current immediately tripped the
garage RCD. However, to my surprise, maybe 1-2 seconds later (with the
leakage removed), the *main* switch on the CU was tripped. So my PC
and various household devices and clocks went down anyway, despite my
intention to avoid that ;-)

The main switch (i.e. red thing) has no capability to trip. It is just a
switch. The only way for it to open (assuming it is not faulty) is for
you to open it.

I'd not have expected this to happen, so could someone offer a
possible explanation please? Does it offer more insight into the odd
behaviour already reported? Or totally unrelated?

The behaviour is exactly (well mostly) as expected. I would expect that
either one or both of your RCDs should have opened on the test. You have
an undesirable bit of wiring here, in that the RCD socket in the garage
is itself powered from a RCD protected feed. Unfortunately when you have
cascaded RCDs like this there is no "discrimination" between them, a
leakage current large enough to open a RCD can open either or both, and
you don't know which one will get there first (it will also usually be
inconsistent possibly giving different results each time you test)[1].

Hence the garage socket with RCD should be powered from the non RCD side
of the CU. (As it stands you could remove the RCD socket in the garage
and replace with a normal one, since its own RCD is in effect redundant)

In case it helps, the electrician's scrawled installation notes record
trip times of 18/19 ms for all the RCD circuits, and 7 ms for the
external garage RCD unit. There are various other columns like 'Loop

When commissioning a new install there are various tests that one does
on RCD circuits to check they open fast enough under fault conditions.
The limit for a normal RCD is 40ms on this test. So both yours were ok
when tested. You may not find that you get the same results every time,
but so long as they are fast enough this is ok.


Impedance', 'CPC-CPC ohms', 'Ph-Ph ohms', 'R1+R2 or R2 ohms'; they
mean nothing to me but please let me know if they could help the
diagnosis.

Probably not going to tell us much here, but FYI:

Loop impedance (i.e. resistance), is a measure of the total round trip
resistance between live and earth at (typically) the furthest socket on
a circuit. This tells you how much current will flow if a short were to
occur between live and earth. The lower this value, the higher the fault
current. The higher the fault current the faster the protective device
(breaker or fuse) will open. So a very low earth fault loop impedance is
a "good thing".

CPC-CPC Ohms (Circuit protective conductor - or Earth Wire to its
friends), is a measure of the resistance between both ends of the earth
wire in a ring circuit. A|gain it should be low. The actual values
dictated by the total length of cable assuming all joints are made
correctly. It is one of a number of continuity tests that demonstrates
the "correctness" of a new install.

Ph-Ph phase to phase (i.e. live to live) is the same thing for the live
wire (normally slightly lower than the CPC to CPC since the live wire is
a bit thicker). Test done for the same reason as above.

If you were to refer to the wire resistance table he

http://wiki.diyfaq.org.uk/index.php?... istance_table

you could probably work out the circuit length from the numbers.

R1 + R2 is a resistance measurement of the phase and neutral wires - If
you know the impedance of the earth connection provided by your
electricity supplier, it can be used to calculate the earth fault loop
impedance. (alternatively electricians will have a test meter that can
measure the loop impedance automatically)

Pondering this over a coffee just before hitting Send, I'm wondering
if perhaps some sort of inductive effect might be responsible? Amongst
other things, I have an ancient large DC power supply in my
shed/workshop (30A), permanently connected to provide automatic garden
lighting at dusk. Could the sudden removal of power have had the
effect observed?

Well it might give you a big back voltage on disconnection, but that
would be after the RCD has opened.

If so, my starting choice was poor, and I'll try
another circuit. Or maybe disconnect the two plugs from the garage RCD
unit.

Looking at the bigger picture, if when your immersion MCB tripped, it
was for example the result of an insulation breakdown that allowed a
live to earth short to occur, then it would be entirely possible for
both the MCB and the RCD to open. The same could result even if the
cause of the overcurrent was not a short to earth, but there was also
at the same time a largish leakage to earth (either from that source or
others) that meant the RCD was close to its tripping point anyway.

The fact that the MCB opened would suggest that the immersion is faulty
or there is a wiring error / damage on its circuit. These things don't
last forever so a faulty element would be my first guess. As a first
step I would suggest you need to isolate it from its supply and test it.


[1] There is a way round this using a special type of time delayed RCD
as the upstream device. Often used on TT systems (see the RCD link I
posted before for more on this)

--
Cheers,

John.

/================================================== ===============\
| Internode Ltd - http://www.internode.co.uk |
|-----------------------------------------------------------------|
| John Rumm - john(at)internode(dot)co(dot)uk |
\================================================= ================/