As part of major bedroom refurb I've replaced one pendant CFL with LEDs. 3 circuits. One for general room light (2x 3W/300lm luminaires) and 2 (1W/100lm) ceiling-mounted reading-lights, one each side of the bed, with standard '2 switch' circuits so there is a switch by the bed as well as the door.

The LEDs are all current-driven 350mA units. There are 3 drivers, one for each circuit. The switching is all mains. The drivers are in the loft for future accessibility, with all through-cabling gasketed to preserve airtightness as much as possible. The Luminaires are in a false ceiling below the airtightness membrane.

After much farting about I have all the wiring done and I even got it right - the switches turn on/off the right lights. However, whilst the main room light works fine, both the reading-lights flash on momentarily every few seconds even when 'off'. One with 4-second period, the other 7.

They worked fine when initially tested, and when all three drivers were running off one switch (temporary wiring setup for a couple of weeks until today). But now that the cabling is properly separated and in conduit I have this interesting strobing. What on earth is that about?

The LED/drivers combo for the 1W LEDs are these:
http://www.ledkia.com/uk/buy-led-downlights/1104-adjustable-1w-cob-led-spotlight.html (you have to click the right photo to see the driver - website too fancy for me to give you an actual URL), i.e. 'cheap and nasty' but they work with about 70% efficiency which is good enough given the very low power.

So how does the driver manage to generate some light when its mains input is switched off? And why hasn't it being doing it for the last few weeks? My best guess is that now I have cables properly laid there is some voltage induced from nearby parallel cables and because it's such a low-power driver this is enough for the driver to try and start up? I've not done any measurements yet, because I've spent about 12 hours cabling/wiring this weekend and I'm bored of that. The cables from junction boxes to switch are parallel for about 3m. But those shouldn't have any current in them when all the circuits are off. The incoming mains doesn't run parallel at all, so my theory isn't very convincing to me, but otherwise it makes even less sense. It's also on a short spur so won't be current-carrying unless at least one of the circuits is on.

I suppose I need to disconnect some things and see what it is that makes it stop.

Just thought I'd post in case anyone else had come across this sort of peculiarity before.

Capacitive coupling between permanent live and switch live cores (and neutrals)

Once any real load is applied, it will swamp the coupling current (the halogen above is doing this job)

The simplest retrofit is to put a small mains rated capacitor across the LED fitting. It's like having a minimum load, but without the heat.

You need the capacitance to be several times larger than the capacitance across the switch and its wiring in the off state.

Use something like 10- 100 nanoFarad for this task - wire ended ones are easier for connection.

Wire in parallel with the lamp, not across the switch - you want to form a capacitive voltage divider between the switch and the capacitor you are adding to lower the voltage when off to below the striking voltage for the electronics in the lamp.

Also please check you do have adequate earthing on the lighting system

Regards

Barney

If there's no current in the parallel wires but there is a mains voltage on them then it might be capacitive rather than inductive coupling. E.g., live wire to a switch which is off next to switched live coming back to the ceiling junction box forming a capacitor to allow a small amount of power through, enough to slowly charge the big capacitor in the switch mode power supply.

[PS, cross posted with Barney.]

we have an LED light in the hall that when off flickers almost imperceptibly, but only when the landing lights are on. Landing lights off, no flickering.

I am assuming parallel running cables at some point is the cause in my case (everything goes back to a central point so there is a gathering where they all end up running together) but my runs are all low voltage DC and I would have thought that DC wouldn't suffer from that problem so much as AC. It could I guess be the lighting controller that has a slight leakage - I haven't checked to see if the hall is on the same controller as the landing.

Been like it for five years and I suspect that it will be like it for another five :)

<blockquote><cite>Posted By: barney</cite>The simplest retrofit is to put a small mains rated capacitor across the LED fitting</blockquote>

Across the fitting, or across the driver? - they are several metres apart and it's the driver input that will be seeing the capacitance from the switch cables. Ah yes, you say 'electronics in the lamp', so you do mean 'across the driver'.

An alternative to adding a capacitor is presumably to use a less shoddy driver unit with better rejection of such capacitive switch loads. (I looked inside and it was not pretty - I wasn't expecting much for £2.95 for lamp _and_ driver)

BTW, I realise you plan to connect it but a floating earth wire, even on a circuit which doesn't need earths, is a reason for failing electrical inspections. Somebody I know put in mains-powered smoke and fire alarms and used T+E, because that's what he had, for the interconnection but the electrician who came round to inspect his installation (for building warrant sign off for a major DIY renovation/extension project) snagged it for that reason. Fix was just to connect it to earth somewhere, of course.

The mains-powered alarms I have state that they do not need an earth but they provide a 'convenience' terminal to terminate an earth in case it happens to be present. They do need a three core cable though, because of the interconnect signal.

RCDs and alarms cause confusion for some electricians, I believe.

OK - I suspect said spark might have been confusing things slightly

The requirement is in the big book of fun, Regulation 411.3.1.1 - last sentence - as almost without exception in a domestic setting the protective measure of automatic disconnection of supply is relied on for safety (regardless if the actual item relies on double or reinforced installation)

Note the requirement (for a circuit protective conductor) for "............. run to, and terminated at..........."

Regards

Barney

So, a bit of experimentation reveals that the earth is (not very surprisingly) floating at about 118V from both N and L. And if I earth it to a handy pipe then _one_ of the lamps stops strobing. But not both. Curiouser and curiouser. Switch the 2-ways to the other way round and the 2nd lamp stops strobing too. So without earth they both strobe in one switch position but not the other. With earth only one strobes in one position. So I can conclusively say that floating earths affect the capacitance in switch drops.

Measuring at the drivers (with a high-impedance digital meter) they see about 17V in one switch arrangement, and either 39V in the other (for the one that's not strobing) and gently rising over a few seconds up to 48V in the other. At 50V the driver strikes and the voltage falls to something below 8V (0V? digital meter lag) and the process repeats. So, if you can keep the voltage at the driver under about 45V then everything works as expected.

I've taken the drivers out for a proper look with my scope.

This is the driver: http://szdarkenergy.com/product/detail/118.html which uses a BP9021A: http://www.bpsemi.com/en/Data/BP9021A_EN_DS_Rev%201.0.pdf

It's not obvious to me why the capacitance changes between the two switch positions, even after drawing out the wiring. In the section between the two switches (the bit that changes) you get NNEL in one position and NLEN in the other (or maybe NENL and NELN, depending which wires are connected where. The cable is flat. The two neutrals are connected at the far end of the cable.
Anyone grok this? (largely of academic interest - it clearly does approx double (I have 3m to 1st switch, then 6 - 8 between the switches).

The 17th regs seems to say nothing about cable capacitance, which surprises me. (well, one mention saying that big capacitors must also have a discharge resistor).

Sounds about right

The difference is due to a differing length of conductor depending on how you set the 2 way switches (draw it out and you should see a doubling/halving effect - which your measurements support)

17th edition wouldn't be specifically concerned as

1 - It expects the system to be earthed

2 - It expects proper workmanship etc so doesn't really predict long length of capacitive coupled cabling (the physical arrangement tends to mitigate this)

Keep in mind the wiring regs are only concerned with protecting persons and livestock from shock fire and burns

Regards

Barney

Hardly radical

I've a small office design on the go that powers the LED lighting with power over Ethernet (plus)

Just stuff a 13A power supply into the "hub" and wire out to all the luminaires, switches, etc with Cat 6 cable

Easy to install, and great fun to programme for just about any circumstances that can be dreamt up - there is even a control mode via a smart phone

A number of the switches, daylight sensors, PIR's etc can be wireless to a series of wifi nodes doted about and cabled back to the "hub"

Look for lighting over POE+ on the interweb thingy

Regards

Barney

Posted By: barney

The difference is due to a differing length of conductor depending on how you set the 2 way switches
(draw it out and you should see a doubling/halving effect - which your measurements support)

I did draw it out and the lengths are the same, only the signal ordering in the cable changes (with 'straight-through' wiring. See Ed's handy pic above - both 'offs' have the same cable lengths. If it was back-back then one wire swaps from the L to the S side, which I can see is likely to change the capacitance.

2 - It expects proper workmanship etc so doesn't really predict long length of capacitive coupled cabling (the physical arrangement tends to mitigate this)

But any switch-drop cable is a capacitively-coupled length, and that's hardly unusual. I guess there may be parts telling you not to make those 'too long', but then I'd expect a discussion of what 'too long' meant in terms of pf/m.

Ed well done for deducing the wiring type from my message (And yes I should have used 'S' for switched not 'N' as it's not a neutral).

OK Wookey, that's not what I'm seeing - if you have "conversion method" (the back to back) or conventional method (the straight through) and assuming you haven't wired it all in singles, then depending on the switch positions, either case increases or reduces the length of permanently live and switch live conductors in close proximity to each other

OK - current practice for domestic lighting circuits would tend to loop both lines and neutrals via switches - so the "twin with earth" to the light is usually dead and with a permanently connected neutral to the luminaires - in older language it would be akin to "joint box method"

Using those techniques tends to significantly shorten those cables that carry a permanent line and switched (off) line - the more conventional 3 plate systems will exacerbate the problems which manifests itself with CFL's and LED's

Regards

Barney

Well some more dicking about has produced a set-up that actually works. Thought I should report back.

Adding 120pF capacitors (the smallest mains ones I had lying about) made no difference to the behaviour. Replacing the one remaining original lighting wire (no earth) in the system with some modern T&E made that lamp stop flickering (adding an earth the system in general had already stopped the other one).

So, as already observed, earthing matters with low-power drivers and switch drops. Or at least these drivers. The offending cable was 4 or 5 tinned strands, 2 conductors ~1.5mm total CSA, so it's possible that the capacitance of tinned, stranded wire is higher than single-solid-core, but as both new and old cable flickered without an earth, I reckon it's mostly that having an earth limits the capacitance/voltage rise.

Live and learn.

Posted By: wookeyAdding 120pF capacitors (the smallest mains ones I had lying about) made no difference to the behaviour.

Probably needed a bit more. At 100 pF/m wire capacitance that would only be “shorting out” a small part of the drop to the switch. You'd be wanting a substantially lower impedance than the wire's capacitive coupling so, as Barney suggested, about 10 nF or more rather than 0.12 nF.