Lots of failing outdoor LED lamps...very long overhead mains cable.

[treez]

Hello,
We have installed one hundred 25W outdoor LED lamps in a town in Bosnia. Sixty failed in just a few weeks. :-(

They all are fed from a single 240VAC generator which delivers power to the 2500 metres of mains cable. A contactor near the generator is used to switch them all on/off. The lamps are fitted every 25 metres along this cable. The cable is overhead (not buried). It’s aluminium cable. Its twisted, though the installers tell us that in places the twisting is “not so tight”. We suspect there is an awful lot of mains stray wiring inductance.
The strange thing is that they have another run of one hundred of the exact same lamps, but in this case the 2500 metre mains cable is buried in the ground. –None of these lamps with the buried cable have failed despite being installed months ago.
The lamps are controlled by microcontrollers in them which slowly dim the lamps as the night goes past the small hours, so as to save energy when few people are really outside.

We suspect that maybe there is some kind of transmission line effect going on, with the long mains cable. We have advised the installers to fit an RC snubber across the contactor, and another across Live/neutral, just downstream of the contactor. We also wonder that with the “not so tightly twisted” mains cable, maybe noise is getting into the long overhead mains cable, and disrupting the microcontroller.

The lamps are not SMPS based, but instead are sucessively switched linear current regulators. There is no mains input filter in each lamp, and the input stage is as in the attached. (ltspice sim also attached, including ZR431 model)
May I request what are your thoughts on why they have failed?

 

[betwixt]

I think you first need to see what has failed. Is it the LEDs, the micro or the regulator?

Looking at the schematic, my first thought was moisture ingress as you have some very high value resistances, particularly R2 and R3 in a critical position and with high voltages present. It wouldn't take much leakage to upset the regulation across X1. Similarly, the only current regulation relies on current through R66, R67 and R68 although they should be able to share ~115V each as long as good PCB isolation is provided.

The other point is: can the micro part of the circuit (not shown) run reliably from ~340V unfiltered DC pulses.

You need to get samples back for investigation - I would guess your customer would be more than happy to comply

Brian.

 

[c_mitra]

Sixty failed in just a few weeks...

First things first: need a very good post-mortem on ALL the sixty failed module. Please see:

1. where exactly it has failed;

2. the environment (whether they failed at the same time; they were all close by; or any related feature)

Stuff can fail EVEN when everything is within specifications. All specifications are statistical in nature.

 

[treez]

Thanks,
We are going to ask the installer to put small boxes across live/neutral. In these will be a series combo of TVS and 32mA quick blow fuse.
The TVS will be SMCJ400CA.
In another box we will have two series SMCJ250CA TVS's in series with a fuse. (this gives a bit higher voltage detection than a single SMCJ400CA)

....If the fuse blows, then we at least know there was an overvoltage transient
Do you think a lighting strike to one bit of the overhead cable would take out say 30 lamps either side of the strike point?.....i am guessing it would?

 

[KlausST]

Hi,

Do you think a lighting strike to one bit of the overhead cable would take out say 30 lamps either side of the strike point?.....i am guessing it would?

A lightning strike is the first thing that came into my mind. I think that this is very likely.
If I remember right, then there are lightning strike monitoring systems. You may find online data sources for lightning activity for the area where the lights are installed.

A lightning strike is voltage/current referenced to earth GND. Do you have overvoltage protection against earth GND? At least at both ends of the cable?

If there is an lightning strike...it will cause a lot of "common mode" voltage on the cable. And it will cause differential voltage on the cable.
With your protection circuit you take care about the differential voltage on the cable.
But still the common mode problem remains. Imagine your circuit is installed in a metal box...which electrically is referenced to earth GND (mounted on steel armed concrete, or at/in a light pole). Now the high common mode voltage may cause a spark between your PCB (or cabling, or LEDs...) and the metal box.
Thus, besides the overvoltage protection against earth GND, I recommend to use increased air gap (>>10mm) and/or an additional isolation, like an extra plastic housing.

But for sure there may be other problems.
Indeed you wrote a lot of text...but not very much "technical informations" in it.
What is the "240VAC generator"? Why not mains AC? Electronic generator? 1 phase? Three phases? Pure sine wave? 50Hz?

What are the "32mA quick blow fuses"? A 240V rating will not be sufficient when they should protect against incoming high voltage (spikes).

If no lightning strike: Is there the possibility for some resonance effects on the power supply and cabling?

But the first thing is - like others already recommended - to analyse the failed circuits.

Klaus

 

[treez]

You obviously don't have a clue because the circuit works on pulses and turns off when the current reaches 700mA, so a 32mA is not going to tell you anything.

Thanks yes i agree, but this TVS/fuse thing is separate to the product, these are to go across live/neutral to proove to the customer that overvoltages are present......because they think our product is just failing without overvoltage.

If no lightning strike: Is there the possibility for some resonance effects on the power supply and cabling?

Thanks yes we are wondering about voltage reflections due to the sheer length of cable, and perhaps a standing wave being set up, with very high voltage peaks....which blows the MOV module up, (sorry i didnt show the LSP05 MOV module in the above schem of the top post) and then takes out the rest of our cct.

 

[tsan500]

we are wondering about voltage reflections due to the sheer length of cable, and perhaps a standing wave being set up, with very high voltage peaks.

If the generator supplies 50 Hz sinusoidal voltage, then 2500 m is a very short cable. It is not possible to have standing waves. Speed of light is 300 km/s which means wavelength of 50 Hz is 6000 km. Difference to 2,5 km is so big, that it is not necessary take account velocity factor of the cable.

 

[treez]

Thanks, we were thinking of standing waves when there is some LC ringing when the supply is suddenly turned on. Much higher frequency than 50Hz

- - - Updated - - -

120kHz for example, has a lambda of 2500 metres...theres a bit of 120khz ringing when our linear regulators transition from one to the other, and when the mains is suddenly switched on, there may be 120khz ringing of the wiring inductance and some capacitances etc

 

[betwixt]

I think you can speculate forever and never reach the right conclusion.
Standing waves are not a likely cause of the problem, even if there is any resonant effect, for the simple reason that the wave will have common mode peaks and troughs relative to a fixed point. You are not drawing power from that fixed point, you draw it differentially from the power line wherever the light is fitted. So even if the power line did have elevated voltages on it, both wires would be elevated at the same time but there would still be 240V across them. For the same reason, I don't think the fuse/MOV will tell you anything conclusive except that normal transients appear on the wires.

If lightning strikes are the problem, the best protection would be a spark gap from each input line to ground.

Brian.

 

[treez]

Thanks, is it possible for lighting to strike the ground near to outdoor streetlights, but not kill them..just "weaken them", so that they fail some weeks later?

 

[c_mitra]

If a majority of the 60 failed lamps failed together, then the cause is perhaps something catastrophic- lightning strike is perhaps one possible cause. But if the 60 failed ones died progressively (not all together) then I strongly suspect some component being stressed progressively.

If the lightning strikes close (not directly on the overhead lines- then all the lights must have failed), then the induced voltage will be equal on both line and neutral conductors (do not expect a perfectly equal) and that can cause partial failures.

Perhaps the lamps were running close to their specified limits- you need to keep some overhead margin (for safety- prosperous life) but that is only a speculation.

Lightning ls likely to strike the poles and not the conductors themselves (because the poles are at ground potential): just another speculation.

But which component failed?

 

[KlausST]

Hi,

Thanks, is it possible for lighting to strike the ground near to outdoor streetlights, but not kill them..just "weaken them", so that they fail some weeks later?

Yes. Consider it as an ESD pulse. It may harm semiconductor internal isolation barrier...and this may lead to delayed fail.

BTW: The same is true for ESD problems during production/testing.

Klaus

 

[Relayer]

Hello treez,
I wonder if it would be best to try fitting Metal Oxide Varistors as shown below:

Found on [URL="https://www.electronics-tutorials.ws/resistor/varistor.html] THIS[/URL] website.
There are also other examples of circuit configurations.
I hope the above helps in some way.
Regards,
Relayer

 

[mtwieg]

I'm a bit shocked that you've come here to discuss hypotheticals, while not mentioning any details about the failure modes in the sixty devices. It's like spending an hour debating the hypothetical need for an umbrella for the day when you could just poke your head outside and look up.

Surely you are in possession of the failed devices and you've documented which components are damaged, right? And you've asked the user about when the failures occurred (gradually or groups at a time), right? If not, why are you here?

 

[c_mitra]

Surely you are in possession of the failed devices and you've documented which components are damaged...

That is exactly what is called a good post-mortem report. Badly needed for a good start in the real investigation...

 

[treez]

thanks, -the location is still in the throws of "un-civilisation"....if we go there to look ourselves, we are likely to meet a sticky end, and not come back.
We could send them the money so that they return them to us...but they might well just receive the money, and not send them back.

 

[betwixt]

It's not that bad in Wales Treez!

Brian.

 

[c_mitra]

We could send them the money so that they return them to us...

I suggest you do not offer money to return the failed device. Offer them 2% discount on the next purchase (one time offer), or 5% discount on the maintenance charges (one time) or one free lamp for every 50 damaged lamp (your make).

I do not know who invented money but make something like that (above statement) a part of the "small print" in all your future invoices. Just my random thoughts, in any case.

And make that a limited time offer!