Wierd one....

I suspect poor quality LED drivers that are allowing significant leakage or capacitive coupling between one side of the mains input and the presumably metal frames of the LED modules.

I doubt that the leakage current from a single light is dangerous, but would be rather more concerned at the cumulative leakage  from several. If not directly dangerous, there would seem to some indirect risk of a shock resulting in a fall.

I would bond the metal ceiling, at several points.

In years gone by, I suspect that metal ceiling grids were fortuitously connected to earth by mains voltage lights, each with a CPC and inserted into the ceiling.

I've seen this also with SELV drivers for 12v undercabinet lighting, where the customer (or in one case myself) has received a painful 'nip' from the supposed isolated DC side wiring. Makes me wonder if it wouldn't be better to earth the secondary and effectively make it FELV?

I think it mite be due to leakage through a capacitor which is wired between the negative side of the rectifier and the low voltage DC output negative it's there to cancel out the capacitive coupling of high frequency noise thru the transformer. Many switch mode supply's bite because of this one unit I've got has nearly 100 volts AC on it DC output leads both are at the same AC potential so there's no current flow  however its presence means that my fone could bite me while it's on charge ouch!!

Many thanks for the replies....


This is really weird.

If I connect one panel in the grid I get 20v

if i connect 2 i get 38v

Etc Etc it keeps going up.....


even weirder...


If I change the mutimeter range to dc volts

I get -2.8v but when I swap the probes over I still get -2.8v


I think I will have to get the manufacturers to visit so that I can show them the fault and get their advise......

The LED drivers are using a switch mode power supply, and being low wattage probably use a design with a single power transistor rather than a pair in push pull. This makes the voltages on the transformer windings and the AC waveform prior to rectification  highly assymetric.

(  optional side track into cheap flyback converter design)

The meter is trying to respond to a waveform that is very far from the simple DC or 50Hz sinewaves forr which it is intended. There is probably a much higher peak voltage than your meter reports, but only for a very small fraction of the time. Also the frequency will be too high for the rectifier in your meter to actually rectify properly.

The meter being unable to respond to the fast pulses decides the signal is lower than it truly is (it will be more than the 20V the AC range reckons too), and switches its internal range scaling resistors incorrectly,, so the analogue to digital converter (ADC) clips on the peaks. Then, due to the adc internal details, clipping is not the same for positive overshoots as negative, as inside the chip is supplied by a battery in a 'negative earth' sort of way and the 'earth' is created from a voltage part way up the battery supply.  In your case it seems the ADC interprets the assymetrically clipped fast waveforms as an average value that is slightly negative - regardless of the input polarity - it is the asymmetry of the overload response inside the meter electronics that cause the 'DC' reading to always be the same way up.



Adding more converters will add more pulses that may reinforce or cancel the effect depending on phasing, so it is hard to predict what will happen.


High frequency "Conducted emissions" as the EMC folk call them are very hard to measure properly, and to do it justice you really need a measurement receiver.

However, I suspect these units either fail EMC standards, or only scrape a pass in some carefully contrived configuration.

Earthing the output may solve the tingle problem, but the switching waveform is clearly escaping as an equivalent voltage source between the mains wiring and the LED dive lines, presumably via the inter-winding capacitances.


To design these things properly requires inter-winding screens on the high frequency transformers, either earthed or connected to the mid point of L and N via a pair of capacitors, and properly chosen  L-R-C filtering on the input.

The pressures of value engineering mean that such things often get left off, and while external chokes and filters can be added, it is not usually possible to be as good as  if it was well designed from the beginning.

Hi Mapj1


Many Thanks for the reply.


If I swap the live and neutral connections the voltage falls away to almost nothing....


But my problem is that if I swap all of these connections will I inadvertently induce a potentially more dangerous problem..????

Or should I just bond all the ceiling grids that I have installed these panels in???

Also bear in mind that the voltages you're measuring may have high impedance sources (due to e.g. capacitive coupling). Multimeters, being very high impedance themselve, tend to register very high voltages; but as soon as you apply any sort of of load the voltage plummets. You could (for example) temporarily connect the frame to earth via a 1K resistor (approximates body resistance) and use a leakage clamp to see what sort of current you get. If low enough, repeat without the resistor.

Earthing the ceiling grid will not do the lights any harm. The LN reversal is interesting, of course no CE marked fitting should care about LN reversal , as  neutral insulation has to be to the same standard as live, but I agree it feels very wrong.

It may be a clue and mean that internally the mains is single phase rectified (one diode in series with one line rather than a bridge of 4 diodes which would treat L and N equally.) It also suggests that the internal high frequency filtering of the rectified mains before it goes into the switch mode supply part is pretty poor in these units.


One of the stranger effects of cheap semiconductors in this century is that it is now worthwhile to rectify the 50Hz mains to a DC and then chop it up again to make a really rough alternating current at a much higher frequency (typically supersonic, anything 20 kHz upwards to perhaps half a MHz or so) so a far smaller and lighter (and so cheaper) transformer can be used than would be needed at 50Hz. The hidden problem is that such things tend to be designed and built by folk who cut their professional teeth on 50Hz  power engineering and laminated iron cored transformers years ago, rather than specialists in what it has become, which is more of a branch of high power Radio Frequency design, and accidental RF emissions, both radiated, (have you tried to find a quite spot on a MW radio recently ?) and like this, conducted, are the result. The product standards and recommended test methods are always playing catch-up to an extent as the practice keeps evolving.

The fact you get -2.8 whichever way round your probes are could be because what your meter is seeing is some sort of high frequency noise which rather than it being proper DC it's confusing your meter. I bet if you held an AM radio near the lights you would hear all sorts of noise  from them

Makes me wonder if it wouldn't be better to earth the secondary and effectively make it FELV?

or if the source still gives "SELV" standard separation from the LV live conductors, it should still qualify as PELV.

I suspect poor quality LED drivers that are allowing significant leakage or capacitive coupling between one side of the mains input and the presumably metal frames of the LED modules.

I wonder where the capacitive coupling is? Is if just between the LV and ELV sides of the PSU - in which case is the frame of the fitting deliberately connected to one of the ELV lines? Or is it capacitive coupling between the LED array etc and the ceiling grid?


I've got my doubts about being able to get a decent bonding effect to a ceiling grid that made up of dozens of separate bits of metal, often painted and just clipped together - I'm wondering it it might be more effective to connect each fitting's metal frame (presuming it has one) to the c.p.c. directly - trapping the problem at source as it were.



Now I think about it some of the LED bulkheads I've used (the type with a strip of LEDs around the inside of a circular heatsink) - has the ELV side metalwork connected to the supply earth terminal - I thought it odd at the time as there seemed to be plenty of separation from the LV side and the driver looked to be SELV - but maybe now I'm starting to understand!


   - Andy.