I have replaced some 25W light bulbs with 4W LED-filament bulbs with 2700K emission. The light is equivalent if not superior, and 25W incandescent bulbs are becoming increasingly difficult to find.

There's a stroboscopic effect in the lighting but it's barely noticeable if you know how to look (update – it appears to be unnoticeable by nine out of a sample of ten people so far)

The problem is that the bulbs use so little energy to make light, that very little energy is enough to make a noticeable light. And when the switches are OFF, one of the bulbs begins to emit an eerie glow which is of course particularly noticeable at night.

The circuit is pretty simple: I have the neutral wire going to the socket, while the phase wire goes to a first 3-way switch from whence two wires run to a second 3-way switch whose middle connector goes then to the socket.

     N----------+ ( BOX ) +-e--------------------------------------+
LAMP                               /a-------------------b\         |
     P-----------------------SW-1                        SW-2 -f---+
                                   \c-------------------d/

My very first thought was of a dispersion due to phase/neutral inversion, so I checked and the circuit is as described above.

Then I thought one of the switches was defective and leaking, but they're OK – open circuit resistance is to all intents and purposes infinite, voltage loss on closed circuit is zero.

The worrying part began when I tested the wires and found a low voltage in both control wires (a-b and c-d), that I had not yet reconnected. As far as I can tell both wires have both ends disconnected, they run about 15m (45ft) up, down and horizontally down a wall, and both of them show voltage (at both ends; this is somewhat reassuring. Though this be madness, yet there is method in't). The voltage increases if I reconnect the second switch and the wire going to the electrical box up in the wall, even if the wire there is not yet reconnected (i.e. there is continuity from a to e, but the circuit ends at e).

I am now led to believe I'm looking at capacitive and/or inductive coupling, even if it's not clear to me coupling with what. But it's true that I read very high field values in the volume of air next to the wall where the wires are running; it's enough to make a contactless voltage detector trip when about one foot off the wall, in a patch a meter high running the whole corridor. Hence my first question:

Can a phase wire with no load at the other end, but running parallel to a 15m neutral conductor in a section about three meters long, induce around 5-6V in this same conductor?

(This is Europe – 220V, 50Hz; even if being near the distribution node I actually have around 237V, 50Hz).

Now, even if this were actually the explanation, I still need to somehow solve the problem – I don't like having eerie glows in my house at night.

Running the wire(s) elsewhere is not feasible, I'd have to break the wall or run an ugly external wire guide. If the induction source is the phase wire powering the other half of my house (not shown, but it should run in the same general area), cutting it is again not feasible – I think I'd like half my home in the dark at all times even less than eerie glow at night.

Swapping phase and neutral doesn't seem safe, and moreover I think it would be useless. The conductor picking up the stray voltage is the one which is unconnected at any one time, so being disconnected from the neutral or from the phase would change nothing – disconnected is disconnected.

I finally thought that, if I ran a third wire along the other two, and connected this third wire to the ground line, the inductive coupling would split between the two (or all three?) conductors. I would only have half the effect on the "true" control conductor, perhaps even less (the wire being grounded would "siphon" off more). The glow would then disappear or become un(- or less-)detectable (is my theory sound? Or is this wishful thinking?).

…and then with this hypothetical third wire in place, I would have a current going to ground, wouldn't I? Would this current be picked up by the main differential, and thereby increase its chances of tripping?

(My "plan B" solution would be to save all incandescent 25W lamps I have, and use them all for that one socket alone. I know it will work, of course. But it smells of surrender).

Creative attempts

1. FAILED: I unearthed two 1N4148 diodes and soldered them in antiparallel, put them in series with the bulb. I thought this would mean that when AC current is applied, one of the diodes would interdict, the other would conduct, so the thing would behave almost like a conductor, except that I would have a reverse voltage drop on the conducting junction. Not noticeable when the lamp is ON and drawing current; enough to choke conduction when the lamp is OFF.

I had overlooked the fact that this is sinusoidal current, not square wave. For a small fraction of the time, forward voltage is near zero enough that it can't overcome the junction barrier. Result: for that small fraction of the time, twice each 50Hz cycle, the lamp is effectively turned off. Enough to make it very noticeably flicker. Adding a smoother to the circuit does not only "cover the hole", the power must come from somewhere, so it dampens the whole wave – and since luminosity appears to be proportional to the square of the current, the (sufficiently-)deflickered bulb becomes perceptibly dimmer.

Long-term behaviour of gimmick also somewhat uncertain.

1.1 FAILED: same result using an integrated diode rectifier (it has four diodes on the inside – actually, I believe, a more complicated single-crystal setup I hoped it would work better. It turned out to work exactly the same).

2. (not yet attempted) Ariel's idea – a 1MOhm resistor in parallel to the fixture.

Going to get me some resistors and a variable one also, to experiment 🙂

3. (not yet attempted) Apply solution (1) to a higher wattage LED bulb. It should behave like a somewhat-lower-wattage bulb with no flicker and no glow. Probably. The difficult trick would however be to get "somewhat-lower" to closely match the other bulbs wattage.

"Regular" attempts

1. (waiting) Follow Russel McMahon's advice in the answer linked by Ariel: go back to the shop and replace the bulbs (well, actually only the one) as "unfit for purpose sold".

2. (waiting) "Plan B": put back an incandescent light bulb in that one fixture. The four bulbs I have now spared will probably last enough for high-inertia LED filaments to be invented 😀