Until, that is, the CW one began to flicker, then 52 of the LEDs went off - half of the end ones and three of the 6 boards. I stuck it on one side rather than throw it out, because there are a lot of LEDs which might be useful for something. Recently I dug it out and thought I might investigate the fault.
The end cap just pulled off - not glued, just located by pins - and I found that the cap was just holding everything in place. The end board came out followed by each of the six 16-LED boards, and finally a small PCB with power components. The power board was wired to the bayonet cap connector (240V AC) and delivered about 176V DC to the end board and one long board. It was arranged so that the DC was split two ways at the end pcb, each feeding four LEDs and then wired across to a side pcb. Three side pcbs were wired in series and the common return was on the last one. This gives two parallel strings of 52 LEDs wired in series, each fed by the 176V DC.
The whole of one string was off, so the fault was obviously either an open LED or a bad connection. I checked the voltage down the chain from the top until it went from 115V at one LED to almost zero at the next - obviously this was the bad LED. Using the diode test function on my DVM, I tested the LED and it was open both ways. But so was the next. and so was every other one on both strings. It seems that the junction voltage on these LEDs is more than the 2.8V which the meter uses for testing. Checks on other diodes proved that the meter was working fine.
I went around every connection in the lamp, reverting to my experience as a quality control engineer. I found that about 10 connections had already been reworked and a further two were dry - clearly the quality of soldering is questionable. Perhaps if the labour costs are cheap enough, it makes it financially viable to employ remedial QC workers even though the product is bad. After all, it only needs to work long enough to get into the end user's appliance for a few minutes. After my loving ministrations, the lamp was still faulty so I decided that the dead LED was probably the only remaining fault.
I didn't have any spare diodes of this type, so I had to improvise. With 176V across 52 diodes, that's only 3.4V each, so the extra voltage across each of the others when one is bridged is only 66mV or about 2%. Of course the higher voltage means more power dissipated in each one, so shortens the lifespan, but not significantly. Luckily they fail open-circuit; if they failed short-circuit there would be a runaway chain effect, hardly noticeably at first but culminating in total extinction and a bang. I could replace the LED with a suitable resistor or diode but that meant more calculations and anyway, I don't have many resistors in stock.
Having decided that it was worth a try, I just put a short circuit across the faulty one and switched on.
The string was still dead! And now the 115V dropped at the next LED. I was wondering if I was actually causing a runaway effect. But in fact, all it was was that both diodes had died together. I won't keep doing this of course, but I shorted out the second diode too, and everything worked again. It all went back together without problems (or glue) and I now have a fully working bulb again.
All this is a lot of trouble to go to for one light bulb, but I was curious about the construction and the fault. More recently, the advent of chip-on-board (COB) LEDs has improved the performance of LED lighting, although I don't suppose the build quality is necessarily any better. The lesson is probably the old one - that cheap products with no warranty are likely to be a bad investment in the long run. I've had many LED bulbs since this one (in fact, it's hard to find anything else these days) and so far they have been very satisfactory. Nowadays I can buy an LED bulb for little more than the incandescent ones used to cost, so it makes sense to buy from a well-known manufacturer. They have their brand reputation to protect.
No comments:
Post a Comment