Jeff Duntemann's Contrapositive Diary Rotating Header Image

Lazy LED Bulbs

Some months after moving into this house at the end of 2015, we went on a sweep and replaced nearly all of the incandescent bulbs with 2700K LED units. The drop in power usage was obvious from our monthly bills. However, I’ve had a whole lot of bulb failures within those three years. Some bulbs, in fact, didn’t even last out their first year. So much for 25,000 hours of service.

I’ve done teardowns on four or five dead bulbs, and found both dead power supplies and dead LED wafers. One of the power supplies was intermittent: Tap on it with the plastic handle of a screwdriver, and it will suddenly light its wafer again. Tap on it some more, and eventually the light will go out. The solder joints looked fine under my digital microscope. I even reflowed a few of them, but the unit’s behavior did not change.

However, the oddest failure mode we’ve experienced is this: bulbs that take longer and longer to illuminate after you flip the switch.

We have four ceiling can fixtures in our kitchen. During our LED sweep, we replaced the 75W incandescent floods with 75WEQ LED floods. A few months ago, I noticed that one of the bulbs took five or ten minutes to come on after flipping the wall switch. Once lit, the tardy bulb shone at identical light levels as the immediate bulbs did. Before it lit, it remained completely dark. (I.e., it spent no time at partial brightness.)

Well, as the months rolled on, the tardy bulb grew tardier and tardier. When it was up to about half an hour delay, a second bulb in the group of four started coming on late. A month or two after that, a third bulb in the group began delayed illumination. By that time, the first bulb would take almost two hours to light up. However, in every single case, all four bulbs eventually came up to full illumination from full dark.

I admit that once the second bulb started acting up, I put off replacing them to see what would happen. Yesterday I got tired of it, and replaced all four (even the one that still lit up immediately) with identical EcoSmart 75WEQ floods from Home Depot. The new floods produce five more lumens than the old ones but only draw 11.5 watts. (The originals drew 15 watts.)

I’m trying to figure out what sort of electrical failure would cause this. When time allows, I’m going to remove the plastic envelope from the original malefactor and take a close look. All of the bulbs I’ve cut open have used switching power supplies built into their bases. There is another kind of LED power supply: a capacitive voltage drop/rectifier system. (Wiki article here. More discussion here.) If the bulb uses a capacitive dropper, the capacitor is probably electrolytic. Electrolytics dry out over time (though it generally takes more than three years) and I’m wondering if poor-quality capacitors are at the heart of the problem. (Bad caps have caused trouble before. And again.) It’s not a time-constant thing, and in truth I don’t know what it might be, but doing a little probing will be fun.

And if any of my EE regulars know or have other (less wild-ass) guesses, I’d sure love to hear them.


  1. Paul says:

    I have zero ideas on your LED light issues, but I find the topic interesting. Our house is a combination of LED and CFLs, with maybe one or two incandescents still hanging around. Interesting to me, is that we have had much better luck with LEDs than you, though we have replaced one or two. What did not hold up at all were the CFLs. Ninety percent of our bulbs come from Costco. We live in the pacific northwest, so maybe there are some environmental or supply differences. I do find the claims of lifespan on LEDs to be suspect regardless though, as I cannot believe any of those included power supplies will last close to what they claim.

    1. We tried CFLs briefly in Colorado, but they all burned out quickly, and some took forever to light up if the house was cold, which it often was. They were also very blue-leaning, and made an unpleasant sort of light.

    2. TRX says:

      We had bad luck with compact flourescents as well. Their failure mode always involved smoke, twice accompanied by fire.

      Most of the pile of dead LED lamps died because the magic smoke got loose from their power supplies; there’s usually a blackened or melted spot on them.

      A tungsten filament’s failure mode typically doesn’t involve “burn your house down…”

  2. RH in CT says:

    The main problem I have had with LED lights is interference on FM radio. It varies by station, with 90.5 severely impacted but 91.1 (slightly stronger signal) sometimes hardly at all.

    Another problem was an “instant on” bulb that was anything but. Two bulbs were on the same switch, one on each of two hallways that met at right angles. I found it very annoying to have them come on over a second apart.

    My sisters had an interesting problem. They screwed an LED bulb into a fixture controlled by a wall switch. The switch was off, but the bulb glowed dimly. This scared them and they removed the bulb. The switch was a very old dimmer meant for incandescents, and certainly not meant for LEDs. When the switch was replaced with a conventional on/off type everything was fine. I might have had a similar problem at home except that I replaced the old dimmer with an LED-compatible one before changing the bulbs.

    I have a torchiere lamp that takes two bulbs; both are LED. Turn the switch to the first setting and it is bright but flickers horribly. On the second setting is is brighter, with no flicker. I’ve been considering replacing the switch with one that has A/B/A+B/Off, with A a 60 equivalent and B a 100, giving three brightness levels.

    1. Orvan Taurus says:

      I had an LED bulb that was very low power (several years ago the Wal-Mart special with the nigh-tetrahedral arrangement) that remained lit with a conventional switch OFF. At first I thought it might have been lit by the local (and not very far away) broadcast station. Then I went to replace a fluorescent fixture in the kitchen and was horrified to discover how it had been wired: neutral wire was switched. Lamp was ALWAYS ‘hot’ – which explained things. You bet I changed that arrangement. Especially as it was the lamp above the sink… and it’s a good thing I have this ‘paranoia’ about water supply (and drains…) and power supply.

    2. LED bulb power supplies, like all switching supplies, put out a lot of RF racket. One reason I’m setting up a 12V lightning system in my garage/shack is that I’m hoping to reduce some of the hash my IC-736 picks up. We’ll see how far that hash travels once I get a reasonable HF antenna on the roof. The house is full of LEDs, each with its own little broadband transmitter in the base. I suspect it’s a good thing that I’ve never been one for weak-signal work.

      1. TRX says:

        A while back I bought something interesting from eBay: a fifteen-foot roll of 2700K LEDs and a wall-wart power supply, less than $15. It came with an adapter that screws to one end of the strip. You can cut the strip to any length. And it has an adhesive back.

        My word, those are bright…

        The new plan is to stick them to some pieces of 1″ square tube, which I can buy cheap and won’t sag in 8′ lengths, and put them up on the shop ceiling. No fixtures needed. And I can run them all off one large, electrically-quiet power supply – you can buy just the LEDs without the wall wart.

        If they hold up in service – the LEDs appear to be wired in parallel, so an occasional dead diode shouldn’t hurt anything – I’ll be using them all over the place, including over the side windows in the station wagon and in the shell over the back of the truck.

        All of the LED failures I’ve had so far have been due to their power supplies; none of the diodes themselves have failed, as far as I know.

        1. Tom Roderick says:

          Back in about 1984 I built a project that used one red and one green indicator to indicate the status of an alarm that was active 24/7 each color was on about half that 24 hour period when I built it. Over 20 years ago a very close lighting strike induced enough voltage to short the driver transistor for the green LED which has been on continuously since then. Both LED’s are still working, and appear to be the same brightness (as best as I can tell) as when new.

  3. WAG from an EE: might there be something in the capacitor which is dried out to the extent that it needs to warm up before it works? If it’s turned of then immediately on again, is it still “tardy”?

    1. Once the tardy light finally lit up, it would come up immediately if you switched power off and then on again. Only once it went off for more than eight or ten hours would it need time to come on again. I didn’t do detailed tests like that on it, because there were better uses of my time.

    2. Oh–and the phenomenon was the same year-round, irrespective of inside temperature. That is, the tardy bulb took about the same time to light up when the house was warm as when the house was cold.

  4. * I agree with Paul – we’ve had terrible luck with the Costco CFLs. Not sure what’s up with those but they had shorter lifespans than most incandescents.

    * Most of our lights are now LED, and most were purchased from IKEA (the Phoenix valley location down off I-10 and Warner). They’ve been working well – no lifetime issues.

    * I’ve also seen some issues with LED lamps interfering with broadcasts – we don’t take cable and watch only ATSC stations some of which used to be completely swamped out when we turned on a particular lamp. Swapping out for a newer LED bulb solved that problem.

    * No clue what might be causing the slow start you’ve observed. Temperature? Capacitors? Who knows?

  5. Larry Nelson says:

    The new LED lights I put in the garage:
    are great.

    Except for the fact that if the lights are on in the garage the door opener doesn’t work when I drive up with my car. It works with my wife’s car. Aaaargh.

  6. CG says:

    Not sure if this is relevant, but created somewhat similar problems of briefly healing itself: cracked surface mount capacitor which had a higher leakage current when humidity had seeped into cracks, but once device was on, heated up, or dried out, problem “healed” itself — for a while.

    More details: knew someone who dealt with this about 15 years ago. It was a small PCB with some surface mount, some through hole parts.
    – Circuit schematic stayed the same, circuit topology changed slightly. All production line tests still passed with circuit topology change.
    – Company started receiving customer complaints about products not turning on at low end of spec (this product had a wide turn-on range). Was a small percentage of units shipped, but far higher rate than usual for problems on those units.
    – Units returned to company often worked fine when received and tested. Company finally got some that were shipped 2-day or overnight, they did exhibit problem, but as soon as case was opened and small wire was soldered to PCB to give the oscilloscope probe something to hold on to, problem healed itself.
    – Took a bit more investigation to realize problem occurred in high humidity environments and the change to circuit topology was subjecting the SMT capacitors to a higher thermal shock than before. Once that was found, tests were run with new units, problem was replicated, circuit topology change was reversed, more tests were run, so on and so on.
    – But until the problem was found to be a combination of high humidity and cracked SMT capacitors creating a higher leakage current than device could tolerate and still work, the problem was hard to diagnose and created some odd behaviors. Visual inspection of solder joints showed no problems.

    The circuit in question was far simpler than a switched mode power supply, and the symptoms were not quite the same as what you’re describing, but the information might help.

  7. Lee Hart says:

    I’ve had the same sorts of problems with CFL and LED lights. In theory they last forever; in practice, they often die young.

    The failed ones I’ve analyzed have always been the switchmode power supply. Basically, they are poorly made; junk-quality parts, flaky circuits, and mickey-mouse construction. It’s what we get in a world where price matters above everything else.

    Re failures caused by humidity: Most electronic components are adversely affected by moisture. The packaging is supposed to keep water out. But as we keep shrinking packages, there is less and less material to keep it out. Then we surface-mount everything, so there is no strain relief or heat dam between the solder and part, so thermal stress cracking creates microscopic “leaks” to let water in. Then we reduce PCB spacings so even the tiniest amount of leakage current will cause circuit malfunctions. It should come as no surprise that these will make parts fail sooner rather than later.

    LEDs are mighty simple to drive. You don’t really need a switchmode power supply to do it. Just wire as many as you can in series, with a simple resistor in series to take up the last of the voltage drop to match your supply. If it’s AC powered, use a series capacitor or inductor to drop the excess voltage and limit the current; they have negligible efficiency loss. None of these produce any RF, and all of them have extremely high reliability if you size the parts right.

    So build your own LED lights! It’s easy, fun, and they’ll last a heck of a lot longer! 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *