Repaired/worked-around LED "integrated tube" fixtures

[FliesLikeABrick]

In our shop, we have around 75 of these integrated LED "tube" lights (and a couple retrofit LED "tubes" which were put into fluorescent fixtures where the ballast was bypassed), half a dozen of which have completely stopped working after being installed for 5-8 years

Overall they have been super reliable - but here and there one has failed over time. Around 6 in total. Not a terrible failure rate, and it's only about $7-10/ea to replace them - but let's see about fixing them faster and for free or cheaper.Brought this one over to the bench, took an endcap off, and slid the lens out.

The failure mode is that they stopped outputting all light, but were still passing 120v through to the next fixtures.

Let's see if we can fix them instead of immediately deciding to replace them






Looking over them, you can see through the solder mask that the individual LEDs are 5x in parallel, and then those groups of 5 are in series down the length. Looking closely, I was able to see a group of 5 that has a visible defect/failure in the middle. Testing with a multimeter, all 5 of them failed open, therefore the entire string of LEDs in the fixture failed open.Integrated fixtures that pass through to others still did so, since they're passing the input power through (not the output of the driver in the fixture). What probably happened is that one of the LEDs failed open, increasing the current through the remaining 4 by 20% -- a significant increase which killed another one, then the remaining 3 were running 40% overdriven, and so on; until all 5 were dead and the entire fixture then went dark.On the first one I used my multimeter in "diode test" mode to light up each LED group of 5 to confirm the bad ones were the ones that I saw the visible defect in, and that there were no other failures.



I used a heat gun to desolder the first and last in the failed group. I think there are two options for repair - one is more of a workaround: Desolder the first and last LED in the group, and solder in a jumper wire past those 5. I think depending how the driver works (constant current or constant voltage), this could end up running the other groups at a fractionally higher voltage since there is now one fewer group to share the voltage drop. The fixture would therefore put out nearly the same amount of light it did before, because it would be overdriving the remaining LEDs and maybe they wouldn't last as long. Or if the driver is constant current, then the fixture will put out slightly less light but the remaining LEDs would be running normally instead of overdriven.The second option is to procure a bunch of these LED chips, and replace the failed ones. I think at the factory it is possible these LEDs are binned by their voltage drop or equivalent resistance characteristics to ensure groups of 5 can be put in parallel and run at the same brightness, and I'd need to see about buying equally-matched sets (and believe that is true, which would be a gamble if I'm buying low volumes on ebay or something). I'd rather skip that, do the bypass, and assume the driver is either constant current or that the voltage increase across all the LEDs in the fixture is not going to be significant.





Then soldered in a small piece of 18awg wire from a spool we already have.



Here is the garage, where most of the repaired ones ended up (while on the ladder around the shop, I swapped good ones out of the garage leaving this the space for the repaired [or replacement] ones to all go in one room.If you look really hard and in full resolution, you may be able to see a dim spot in some of the fixtures by the far wall - that's where the 5 dead LEDs have been bypassed.5 of them were fixed with the repair described above, and a 6th turned out to just have a bad solder joint where the driver fed the PCB strip that the LEDs mount to.I am not particularly skilled in electronics - this is something that anyone with basic soldering skills should be able to fix, hope it helps and/or is interesting!




Thanks for reading, hope this helps someone

 

[markba633csi]

LEDs should last a long time in theory but it's the implementation where they fall down
They shouldn't just stick them in parallel without some type of balancing circuit or equalizing resistors
I have noticed many LED city traffic lights that have partially failed probably for the same reasons

 

[RJSakowski]

When I designed my LED lighting system back in the Wild West Days in the early 2000's, I specifically addressed the issue of failure when used in a series/parallel configuration. I used a driver called the BuckPuck which put out a constant current of 350 ma, 700 ma, or 1 amp at up to 30 volts. The LED's that I initially used had a rating of 1 watt so the drive current of 350 ma worked nicely. For economy's sake, I drove a parallel combination of either 2 with the 700 ma driver or 3 with the 1 amp driver. Later, I used a lower power/higher output LED for under cabinet lighting where I could drive five series strings in parallel.

If an LED in a series string failed open, it would sense the higher current flow to remaining LED strings in the parallel array which would trigger a crowbar circuit and drop the current. If it failed shorted, the current flow to that LED string would similarly trigger the crowbar circuit and drop the current to all the LED's in the series/parallel array to a safe level. The BuckPuck drivers are also capable of pulse width modulation for dimming although we seldom run them dimmed.

Due to physical constraints, I used a variety of configurations ranging from 6 x 2 to 8 x 5 in series/parallel. The initial installation was done in 2007 and has been working an average of hours a day for the past 17 years. To date, I have had only one failure and that was a BuckPuck driver.

 

[MrWhoopee]

I'm just starting to see some failures in the LEDs I put up in my shop about 5 years ago. May have to explore this.

 

[homebrewed]

Your solution probably is the easiest to implement. An alternative would be to use one of the failed ones as a source of LEDs. You'd pull all 5 good ones in a block, that would preserve the matching you're concerned about.

But that would be a lot more work compared to what you did.