The Thunderstruck

Afternoon and evening storms are common here in Central Europe in the summer. It happened that during one of them, lightning hit around 500 meters from my house. I literally saw it while standing outside on the porch (it wasn't funny at all, to be honest). When I went back inside, I noticed that half of the house had lost power. That was the job of surge arresters, shorting the overvoltage spike to ground and causing circuit breakers to trip. And oh boy, how glad I am I've installed these arresters in all electrical boxes and in the outlets powering computers and other IT stuff. Without them, a lot of things around the house would have been fried instantly in this unfortunate moment.

The only device that died was one of the lights. It just stopped working at all, without affecting the other ones connected in the same circuit. So I decided to simply replace it with the new one (I always keep spare lights of all the types I use in the house) and then, of course, check what went wrong. Immediately after pulling the can from the ceiling, I smelled a familiar scent of burned electronics, although both the converter and the light fixture looked perfectly fine from the outside.

A quick check with a multimeter showed that there is an open circuit on the primary side of the converter. After unscrewing two small screws on its enclosure and a short fight with fancy plastic cable clips, the following unpleasant view emerged:

The LED panel was perfectly ok though, producing very bright light when powered from the bench power supply at 18.4V and 300mA. So let's take a closer look at the overcooked converter:

The circuit is a simple AC/DC constant current fly-back converter controlled by the OB3635NCP chip, produced by On-Bright Electronics, dedicated to LED lights. Nothing fancy here, it seems to be a typical implementation straight from the datasheet (although I couldn't find any sheet for this exact NCP version, only for similar chips from the same family). But what could have gone wrong? As you can see in the photos, there are two scorch marks, one around the fuse and another near the diode bridge. Measuring resistances here and there gave the possible answer: a short on the AC side of the bridge. It also looked like some traces were missing, so I decided to unsolder a few components to check what's underneath them:

And here is the reverse-engineered schematic showing the circuit from the AC input to the bridge:

So, what did happen here? Presumably, the voltage spike killed the bridge shorting its input between pins 2 and 4, which then caused a short current high enough to evaporate the trace connecting the bridge to the VR2 varistor, and, in the end, to blow up the F1 fuse. All these shiny spots on the PCB and the plastic cover were remnants of the exploded copper trace. But why did the circuit fail this way? The MB10F breakdown voltage sits at 1000V. The 07D511K MOV has 845V of clamping voltage at 10A, but it reaches 1000V at 100A, and the MOV should theoretically withstand this for at least 40us. Both the varistors survived, though, so I think about two options:

1. The voltage spike was really sharp (roughly 100V/ns rise time), so the bridge passed out before any of the MOVs started to clamp.

2. The overvoltage has been clamped by one or two MOVs, but it lasted for tens or even hundreds of microseconds, rising clamping voltage above 1kV, and the bridge eventually gave up.

We can forget about the fuse, it's the slow-blow (T) one, which needs around 1ms at 100A to act (example from Littelfuse). So when it cut the current, the bridge had already died. Of course, the fuse played its part excellently in the final scene, preventing any further heat or fire damage.

Did the protection circuit work as intended? Fuse - yes, MOVs - not necessarily. Personally, I would select the one with a little lower clamping voltage, e.g., 07D471K has 775V at 10A and 900V at 100A. Also, putting MOVs in parallel without precise matching isn't a really good idea; just one, but more bulky, would be a much better choice here.

Anyway, the whole power supply passed away and wasn't worth the effort to try to repair it. Moreover, with traces missing on the primary side and metal particles sprayed all over the place, it would be simply too dangerous to hide in the ceiling, even if revived to a functioning state.

The Hesitant

I was repainting the ceiling in the kitchen the other day. To ease the job, I removed all the lights, and then, after the paint dried, I installed them back. And for no reason, one stopped to glow. So, again, I pulled it back from the ceiling, but left it connected - pushed the switch - it worked. Maybe something was wrong with the connections? I powered the lights off and double checked the primary side (WAGO clamps) and the bayonet-style connectors between the converter and the LED panel - all looked fine, holding firmly. Powered on one more time - worked, so in this state I started to push it back into the hole in the ceiling and... it stopped working. OK, that was the clear sign it's susceptible to movements. As for the fried one, I mounted a spare, and the indecisive one went onto the workshop desk for further investigation.

After powering up the whole set on the desk, it behaved exactly the same, but by shaking, tilting, and bending all the cables, I narrowed the source of the problem to the DC connection between the power supply and the LED panel. The connector itself is kind of a barrel jack, but with a bayonet lock. You need to push the male part (panel side) into the female part (supply side), and then rotate it slightly to lock in place. Using a multimeter, I quickly found that there is a break somewhere in the positive wire between the converter and the female connector. Solder joints on the PCB looked good, though, so the only point left was inside the receptacle. And unfortunately, I had to cut it with the knife because the metal contacts were put inside a mold made from some hard rubber. The cut revealed the damage very quickly - it was just a cold solder joint:

The repair for this one was very quick. I just resoldered the whole wire, taking it first from the converter killed by the lightning. Easy-peasy. For the wire and the socket already cut in half, I remade them into the 'test adapter' that allows to easily power the LED panels from the laboratory supply:

The Flickering

This one was behaving weirdly for months. At first, I noticed that it was giving slightly dimmer light for a second or two, but it was only occasional. Then it started to dim and flicker more and more frequently, until it became permanently dimmer than the other lights. And that wasn't the end, as a few weeks later, it again started to flicker and darken even more. A single diode could just die. It happens a lot in LED lights, but this case was the sign we're running down the hill, and it would be only worse. Again, spare one placed in, bad one went onto the desk.

First, the power supply. All was good here, producing correct and stable output. I even opened it to check for any loose connections or cold solder joints. Nothing suspicious there.

Next the LED panel. I powered it up, and it started to flicker immediately. OK, so something inside was wrong. I removed all the screws from the back, two mounting springs, the plastic wire holder, and finally pulled up the metal cover to see this:

I was a little surprised at this point because I thought there would be a PCB of some sort with LEDs soldered onto it. Nope. As you can see, the panel is constructed in a way that an ordinary LED strip is glued around the inner surface of the aluminium ring, and there are two pieces of semi-transparent plastic diffusers in the middle. The outer one is 1mm thick and white-ish, the inner one is 2.75mm and clearer. On top of diffusers, there is a thin white metal foil (acting as a reflector) with foam strips squeezing everything together when the back cover is screwed on. Generally, the lack of PCB is not a bad idea here. LEDs mounted around the diffuser create a very even and soft beam of light, without any bright spots. One funny (or maybe not so funny) thing to note: the edge of the bottom diffuser is painted black, probably to avoid some unwanted reflections. And it seems to be literally painted by hand with a marker... yeah, it looks like a lot of manual labor is involved in the manufacturing of these panels.

Back into the case. First, I observed a not very good soldering job where the wires were attached to the LED strip:

Cold joint again? Not this time. Although it looked pretty bad from the outside, it provided a good electrical connection and held the wires in place. Let's power it up:

Ouch, one LED was already done, and the next one was flickering from time to time. Have you already spotted what's going on? Yup, the gap between the strip and the casing:

This small gap prevented the heat produced by LEDs from being transferred freely to the aluminium body, which is not only a mechanical base, but a radiator, too. Consequently, the diodes were boiling themselves until one flew into the land of eternal lights, and the second one was on its way to the same place.

But why did this happen? I can only guess: someone didn't stick the strip correctly to the body, or some other guy struggled with soldering, overheating the thing until the glue came off, or a different person pulled the wire, tearing the strip off, or... Anyway, it looked like a mistake in a manufacturing process, which eventually caused the LEDs to slowly die, one after another.

Was it worth the repair? It would if only I could find the replacement for the strip. For the original one (marked as JC-2835-286*6mm/5B6C) I wasn't able to find any info over the web, seems to be specially made for these small lights. Unfortunately, all the widely available typical strips wouldn't fit - they differ in the number of diodes per unit of length, so the outcome could be the panel working fine, but producing less or more light (and probably in a slightly different colour) than the rest mounted in the same room. Which I found rather annoying.

Finally, the strip went to the e-waste, and the rest of the panel to the box with spare parts. Who knows what will happen in the future with other, currently working, lights?

Resurrection

What can we do with three partially broken lights? Make two new ones! As I already said before, the first one hit by the overvoltage had a dead converter, so its wire and connector were transplanted to the supply of the second light, which had the broken joint in the barrel jack. And the third one's perfectly good power supply has been paired with the LED panel from the thunderstruck one.

All in all, I ended up with not three, but only one damaged set. And also, now I have a bunch of other parts that might be useful in the future. Is this a good outcome? Yes, I count it as a success. Of course, it's only because I know how to repair such things. A person uneducated in this field would probably throw away all of them (and post an angry online review afterwards).

Do I still trust my recessed lights? I think yes, I do. When I'm writing this, already seven years have passed since the installation. Around 30 other ones of the same type work fine. As I've shown in this post, some things might be improved in the design or in the manufacturing process, but after all, I am still happy with my recessed lights. With a few spare ones being kept in the closet, their future is (literally and figuratively) bright!