[graham.d (OP)]

A curious thing. I was lying there counting Neils jumping over a gate when I noticed the light briefly and dimly flash. Better than a Neil flashing I thought. So I noted that the light flashed about every 29 seconds. On a second night the time between flashes was 24 seconds. It seems very consistant on any one occasion though; at least it stayed the same until I went asleep.

What could be the cause of this given that the light was switched off? It is the main light in the room with only one switch (single pole on the "live" wire) on the wall about 15 feet away. The light is connected via a twin cable from the ceiling rose where there is an unconnected earth wire. As a result it is hard to see any current path.

In the UK the neutral line is earthed at a substation (I think) so there can be a few volts of AC on the neutral line. I can only think that this must be enough, combined with the capacitance of the switched-out live wire to provide some tiny amount of power and I guess this could be enhanced a bit by some RF harvesting. However, I still don't know how this could produce the regular very low frequency pulses which I can only think is to do with the electronics in the light bulb. The light bulb is one of the new electronic fluorescent job with some means of making it instantly bright on switch on.

Any ideas?

[imatfaal]

Is the instant on feature some form of capacitor? (the instant on of old tube fluorescents was a capacitor?)  I seem to remember from school physics (I am not a fan of electricity so haven't keep even close to up to date) capacitor with an inductor will act as an oscillator - or have I got that entirely wrong?  Coil of wire in the fitting acting as inductor?  Seems unlikely and you shouldn't have a closed circuit anyway.  I think the period only depended on the form of the capacitor and inductor - so I don't quite think it fits a different period on different nights

[Geezer]

It could be a sheepy discharge. No! Stop it! I'm trying to be serious here.
 
I've noticed that I can cause a CFL (comapact flourescent) to flash briefly if I produce a static discharge when removing my woolly jumper. I have to be quite close to the CFL for that to happen, but it suggests that it does not take much of a field change to make them light up.
 
So, unless you are tossing and turning in bed a lot, it sounds like something near or within the bulb is collecting charge until the potential is great enough to produce a small arc. That might account for the time constant.
 
Not sure if the instant-on has much to do with it. I think they incorporate a halogen bulb that comes on while the fluorescent section is getting up to strength. I eventually stuck a LED bulb in my bedside lamp because I got fed up waiting for the CFL to warm up!
 
 

[Bored chemist]

I rather suspect that the answer is something like this
http://www.cliftonlaboratories.com/relaxation_oscillator.htm

Incidentally, I get the same problem with one of my lights because the switch is a timer switch and it's powered by electricity drawn through the lamp.

[SeanB]

Yes, it is energy storage. the cable forms a small capacitor, and this allows a small current to flow into the lamp itself. This is redtified by the internal diode bridge and charges the low value capacitor inside slowly to a voltage high enough to trigger the lamp to start operating. The lamp attempts to start, but there is not enough power coming in to keep it running, just what has been stored internally. Thus it flashes dimly, then stops as the capacitor is discharged.

This will eventually kill the lamp, as each flash depletes the emissive material on the cathodes of the lamp, and this will eventually cause it to not start. The solution is to either use it with a bedside lamp where the switch will have lower capacitance, and there are not long lengths of cable to couple together.

Try other makes of lamp, and if nothing works then replace with a halogen lamp or a regular incandescent ( if you can find them in the UK now, you might have to look for Rough Service lamps to get them) to stop this.

The reason the lamps take a while to come to full brightness is because the mercury inside is contained as an amalgam in a metal bead, so the lamp has to get hot enough to evaporate the mercury out to get it to a vapour where it provides the discharge path. LED lamps start out at full brightness, and do get dimmer as they heat up somewhat.

[Geezer]

Yes, it is energy storage. the cable forms a small capacitor, and this allows a small current to flow into the lamp itself.

I don't understand why the capacitance of the cable would make any difference, but if the switch was a bit "leaky", that might do it.

[graham.d (OP)]

Thanks for the inputs guys. I also think it is to do with the line capacitance being sufficient to have a sufficiently low impedance at 50Hz to allow a small AC voltage to be present across the bulb. Remember that the Live line is switched out of the circuit at the switch (it is the capacitance of this line to earth that we are talking about). What remains connected is the Neutral line. However, there is only a few volts of 50 Hz AC on this. So you have a circuit with a few volts of AC driving the CFL with the other side of the CFL with only a few 10s of pF to ground. I can see that the energy accumulated in 20 to 30 seconds may well represent the amount of light in the flash (i.e. not a lot), but I don't understand how it is stored or builds up enough to trigger the tube. It must be getting rectified and then charging a capacitor in the electronics. This is what takes the time with the limited current available (because of the series cap). But as the max voltage is only a few volts - I'm surprised this is enough to strike the tube. Is there some more complex circuitry in these CFLs that is used to generate a higher voltage? I'm thinking that there may be enough voltage to fire up this electronics which may have an internal oscillator that then drives some circuits to produce a striking voltage. The fact that the time between flashes varies one night to the next (although is very consistant on any one night) could be due to temperature effects or, more likely, variation in the voltage on the neutral line.

I can see curiosity getting the better of me at some point and this light getting sacrificed. I suppose an interesting test would be to stick a bench power supply across the lamp to see at what voltage it lights up.

[Geezer]

There's an easier way to find out than that. Just isolate the earth wire at the plugtop  [:0]   (or is it a moulded plug?)
 
 

[graham.d (OP)]

There is no plug. It's a main room light with a switch on the wall. However I could experiment when I get time. There would be significant capacitance, although reduced, whether the earth was there or not though. I think we can assume that the circuit must be completed via the wire capacitance as there are no other paths.

[Geezer]

Ah! What sort of voltage do you typically see neutral to earth? I seem to remember a few volts was not unusual, but that was a long time ago. Don't see much of anything here of course because they are tied together at the distribution panel.

[graham.d (OP)]

I will check tomorrow what the voltage is between earth and neutral. I don't remember what it was last time I looked but it is typically a few volts.

[syhprum]

I think the solution would be to connect a resistor of about 100,000 ohms across the lamp to divert the current leaking through the capacitance of the cable or install a two way switch that switches the line feed to the lamp to neutral in the off position (via a resistor for saftey ).

[Geezer]

You could also dump the neutral, and wire it line to earth!
 
Come to think of it, didn't the old 15 amp outlets in the UK switch both line and neutral?
 

[SeanB]

The capacitance is between the live in and the switched live out in a long cable. This makes a small capacitor, and this is used to transfer charge every cycle of the mains to the lamp's internal capacitor ( there for energy storage after the input bridge rectifier so the lamp does not flicker at 100Hz in operation) until the voltage is high enough to trigger the lamp start up circuit, which typically will try to start the lamp when the voltage across this capacitor is around 150 V DC, giving the flash as the energy is dissipated in the brief flash from the flourescent tube of the lamp.

[Geezer]

Ah! Yes - I can see how that might happen with building wiring where there is a long run to and from a wall switch for example. Don't CFLs include a bleed resistor to prevent that?
 
EDIT:
 
Apparently not. I just measured about 15 megohms across the contacts of one.

[graham.d (OP)]

I think our proposed mechanisms may be wrong! I checked the voltage from neutral to earth - admittedly near the meter and with a fairly crap voltmeter - and could not detect any AC voltage at all (probably less than a volt). In any case, with the circuit of neutral having some AC voltage with respect to earth and the eath going to the other sie of the lamp via a small capacitance, I don't see how any significant voltage could be built that is greater than that applied. Also, if you assume a few hundred picoFarads, the leakage time constant of even 15Mohms is only a few milliseconds. I am now more puzzled than ever.

Other possibilities are pick up from RF sources but it will still need some significant capacitance to store any rectified voltage for 20 seconds or so.

[Geezer]

Are ewe sure there are no woolly jumpers in ewer vicinity?

[graham.d (OP)]

Well, as I said, I was counting sheep :-)

[graham.d (OP)]

Syphrum, there is no source of DC. I have assumed that there is almost no leakage between the earth and live wires in around 20 feet of 3 core mains cable (many Megohms), and in any case the live wire is floating because it is isolated by the switch on the wall. The neutral is low impedance and closely approximate to 0V (i.e. Earth voltage). The only "circuit" is from the neutral, through the lamp to the isolated Live wire and that there will be some capacitance from this to the earth wire. As the neutral to earth actually seem to have little or no AC voltage across them (measured with a voltmeter set to AC) how is any voltage appearing across the lamp (AC or DC)? It's hard to believe the switch is leaking.

 [ Invalid Attachment ]

[graham.d (OP)]

Oops. How do I make the GIF bigger?