The importance of RCD selectivity and type

[Rockingit]

I quite often contribute on other's posts about issues with RCD's, thought I'd tell you about what happened to me yesterday as a little bit of a tail and maybe a lesson.....

Huge stadium show lighting rig, fed in three sections, each section being about 300/3 draw, mostly all LED fixtures. Thankfully only on set-up day but one of the feeds tripped - upstream, at the generators. Should never have happened, main ELCB had tripped.

Investigation and much headscratching, mutterings in Frenglish (this is Paris) with the local generator guys..... and this is what was happening: Their lighting rig contains a large number of some fixtures that have come straight from the factory in China. And the electronics in them are so poor that each one dumps about 10-15mA to cpc. Each. So the 30mA RCBO each one is on is happy enough. The control and distribution end of the system has 'divisional' RCM that's not seeing anything at all. Over 7 amps to the cpc cumulatively, non-fault. So why had nothing downstream tripped when the upstream ELCB was set to 3A 500mS?

Because.... wrong type of RCD's. These needed to be A types and they were all standard AC. So much DC going on when I measured it properly that the RCD coils would have just been permanently saturated the same as an MFT does. No option to disable the upstream protection (even if I'd been tempted to do that couldn't anyway due to French regulations) so had to solve the problem another way. With some butchery we split the loads up to spread the dump across more than one device and set the ELCB's to 5A.

The other option I had was the control racks all have bonding points in them (for other occasions), linked directly from the incoming cpc. So I could have also bonded them back to the venue MET and dumped the current that way via parallel path back to the generation PEN, upstream of the outlet monitoring.

And this was a 'perfectly good working system'. The games we play.

 

[pc1966]

I'm intrigued about what was dumping DC in the the "residual path".

Most normal leakage is capacitive and you would expect to IR test L+N to E at 500V with negligible signs of a path. To actually get a DC term you need not just a restive path, but one from some rectified supply, such as the input stages of a SMPSU.

Edit: From the Blue Peter school of "here is one I prepared earlier" we present:

RCD "blinding" by DC

A recent thread was discussing having a delay RCD (type A or AC) for TT before a garage CU with a type B RCD for an EV charger. The question was would this be OK or not? Standard advice is the type B should be direct off supply (possible problem for TT) or fed from another delay type B (at...

www.electriciansforums.net

 

[Rockingit]

I'm intrigued about what was dumping DC in the the "residual path".

Most normal leakage is capacitive and you would expect to IR test L+N to E at 500V with negligible signs of a path. To actually get a DC term you need not just a restive path, but one from some rectified supply, such as the input stages of a SMPSU.

Edit: From the Blue Peter school of "here is one I prepared earlier" we present:

RCD "blinding" by DC

A recent thread was discussing having a delay RCD (type A or AC) for TT before a garage CU with a type B RCD for an EV charger. The question was would this be OK or not? Standard advice is the type B should be direct off supply (possible problem for TT) or fed from another delay type B (at...

www.electriciansforums.net

All Switch Mode PSU's. Hundreds of them. And a gazzillion LED drivers.

 

[Lucien Nunes]

It would definitely be interesting to see the specific cause of the DC leakage from these fixtures. Obviously it's a phenomenon we think and talk about often these days, and it's clear how various fault conditions could produce it. It's less obvious how DC might arise in quantity by design, to the extent that (e.g.) suppressor capacitive functional AC leakage does.

As it happens I was only thinking earlier about DC leakage & shock current scenarios from vintage consumer electronics. In AC/DC (transformerless) valve radios and TVs it was normal to have the signal ground, and hence internal chassis, connected to one side of the mains. To minimise hum pickup at the volume control, the single-pole power switch is normally in the chassis connection, so whichever way you connect the mains lead*, the chassis is live either when the set is on or when it's off. These days we always make the chassis neutral and don't rely on the internal switch. In any case, if you come into contact with the chassis when it's live, the shock current is AC. You can get a DC shock to earth if you touch the HT rail downstream of the half-wave rectifier but to do that you usually have to get significantly inside the set.

But then came the solid-state live-chassis TVs with the chassis connected to full-wave rectified DC negative, so any shock from the chassis is pulsating DC. Back in the day when these sets were in widespread use most houses didn't have RCDs, but it was not unknown to work on them in-situ without an isolating transformer. I fell to wondering whether a type AC would have added anything in terms of safety to the repair man or been completely useless. I probably missed getting quite a hefty belt from my folks' Philips G11 when I was distracted after completing a repair. I started to fold the swing-out panels back in while it was switched on, with my other arm leaning on the VCR which was earthed via the audio output to the HiFi. Evidently I didn't touch any metal bits on the VCR.

* DC footnote. In the days of 3-wire DC, half the houses on a street would have one live outer and neutral and the other half would have the other. Since the HT rail must be positive, in houses with positive live** the chassis had to be neutral, but in houses with negative live the chassis had to be live. A 2-pin mains inlet connector was often fitted so that it could be set to the correct polarity. With it reversed, the valve heaters would glow but there would be no sound or picture. This problem doesn't arise on AC so we can always make the chassis neutral, apart from the pesky SP switched neutral in many sets.

** I.e. what we now call 'line' but because of the historic context, it sounds wrong.