RCD just wont hold!

[Screwi]

I have been to a job with something similar, I found the following:-

Washing m/c with a leak dripping water inside.

and also a Undercupboard light in the kitchen wired to the lighting circuit with a borrowed neautral to a plug socket.

It wasnt enough to trip the rcd all the time, but under the right conditions it did.

 

[datadiffusion]

datadiffusion wrote:
following a full inspection in which all was clear (no insulation problems, rings all OK, etc...).

Did you do the inspection and test yourself or did you get a sparks to do it for you?

Do you have a schedule of test results to refer to regarding the insulation resistance test between each conductor?

All done properly @ 500v - result, at least 200 megs between all legs! L, N, E all 0.06ohm ring continutity, etc... No other work carried out between the test and installation. I am satisfied the wiring is OK (but will recheck now as everything else has been exhausted).

Thing is, the fault appears not to be not restricted to one circuit - and plugging in the appliances one by one did not result in any tripping. But when it does trip, it them wont reset, even if you try it with just one ring at a time energised.

Ah well, soon be sorted im sure!

 

[datadiffusion]

I have been to a job with something similar, I found the following:-

Washing m/c with a leak dripping water inside.

and also a Undercupboard light in the kitchen wired to the lighting circuit with a borrowed neautral to a plug socket.

It wasnt enough to trip the rcd all the time, but under the right conditions it did.

Yep, fair enough, I too have found 'borrowed' neutrals, once found the upstairs sockets feeding the kitchen lights (not sure what happened to the original, but all 'correctly' wired in uninsulated conduit singles

But its the first thing I checked for - certainly not a factor here unfortunately...

 

[mapj1]

Sounds like you could do to acquire a clamp on ammeter for the day, and measure the live-neutral current yourself (or the CPC currents).
The Henry, by the way, may be DI (plug it in via a 2 core lead if in doubt about the earth pin, but I' doubt it very much) but it will have a 50A plus inrush - the winding resistance (on DC) is very low, and the inductance back EMF takes a while to spin up. If it is an N-E leak (even if capcaitive) the inrush could push the trip over the edge.
Also what about filter capacitors in fridges, strip lights etc. you dont need many nano-farads between live and earth to clock up the leakage current to near tripping levels.

 

[datadiffusion]

Thanks thats all useful, having thought about it I too realised inrush must be a large part of the Henry causing trips. Its fairly new so I had no immediate reason to suspect a fault.

Only the one fridge freezer on the entire circuit - all other 'white' appliances on another DB entirely in another part of the house. I would estimate the remainder as being 1 PC, and plenty of tablelights etc.. 95% of which are 12V type fed by plug in transformers. All other loads are resistive.

A clamp on ammeter is something I dont own - but could easily borrow. It would help tie things up for sure.

One thing I have never understood is something I came accross a few years ago. A old fuseboard had everything via a 30mA RCD - no real reason as it was a modern flat with PME. It would trip, but when reset, usually last about 15 minutes before going again. I (finally) traced the fault to an overheated pendant - on the actual bulbholder itself. It had been fitted with a 150W bulb, and so the casing was scorched to the point of crumbling, but it was a modern type fed by 2 core flex only. The heat had turned the copper wires purple, and presumably as a result high resistance, as happens when some idiot uses 0.75mm flex to connect a 3kw fan heater!

I understand the 15 minutes being due to the bulb heating up, but how does a 2 core flex, with no path to earth (suspended by PVC flex in mid air!) cause a leakage trip, rather than an overcurrent? Forgive me if I have missed some fundamental of electrical circuits, but it has bugged me to this day!

 

[mapj1]

If an L-N load seems to be causeing earth leakage, what it is usually really doing is causing a lot of noise - in the case of the frying lamp, I should imagine a 'feezzling' sort of effect. The high frequencies of the fourier components of this current will be passed to earth more readily than the fundametnal 50Hz, through the normal house wiring self capcitance of a few nF (say 100pf/m for normal twin and earth.)
The effect is to increase the overall imbalance in the coils of the RCD (albeit with a very odd waveform), and so to fire it. In this respect the device is really working as an arc fault detector. Some of the better modern RCDs have an addtional functional earth connection for internal supressors that reduce the risk of accidental tripping on such noise and spikey waveforms.
Hope that helps.

 

[datadiffusion]

Ah, I see, so it was the HF bits leaking to earth by capacitance as soon as the flex joined on to some T+E...

That explains a lot - thanks again!

 

[mapj1]

Indeed - this is how switch mode power supplies, with chopping frequencies of perhaps 50 to 300KHz, manage to leak so many milliamps to earth through relatively small filter capacitors - capacitors whose 50Hz leakage would normally be negligeable, may find themselves carrying a significant circulating current.
(Obviously over many cycles of the high frequency the current first added too and then removed from the L-E loop averages out to the expected 50Hz value (and many slower meters only measure this lower figure), but on the scope plus current probe it looks like one of those really hairy caterpillars, except being asynchronous the spikes crawl along the waveform, like a demented chair lift)

This problem is compounded at harmonics of the chopping frequency if the installation wiring happens to resonate. Those of us who design electronic stuff like this for a living have got very wary of filter capacitor current ratings over the years. As said earlier though many of the classier RCDs have internal filtering components that reduce the false triggering to a manageable level.