Theories about why my WS2182B LEDs keep dying?

I’m working on some overly complicated under cabinet lights, which I’ve posted about a few times in the past. My latest issue is that LEDs keep dying. Usual symptoms are that the bad LED will be orange while others are white. LEDs after the bad one will either not show anything, or display very erratically. I’ve found that sometimes pressing or twisting the bad LED will temporarily “fix” it, but not always. So far I’ve had to replace about 5, and my supply of spares is running out.

Here is an awful diagram of my set up: http://i.imgur.com/NiKkZub.png . The most recent LED to pop was towards the end of Strip 1. Others have been towards the beginning and middle of Strip 1, as well as towards the beginning of Strip 2 - but those were all using the old power supply. I’m using a Photon to drive them. The photon’s outputs are shifted from 3.3V to 5V with a SN74HCT125N. Data lines both have a 330 Ohm resistor on them.

So far I’ve tried 2 main things. My initial setup didn’t have capacitors across the power supply input, so I added those. Also I found that my power supply was outputting 5.5V (with no/low load, not sure about full load because I stopped using it) and the WS2182Bs supposedly only tolerate up to 5.3V. So I bought a new power supply that is supplying 5.32V under no load, 4.3V under full load (all LEDs full white).

One of my thoughts has been that heat is killing them. Strip 1 is directly over a stove, and is also where most have died. But it looks like the operating temperature can be at least 100C, maybe more depending on what data sheet you look at. And these are in silicone tubes which should be insulating them fairly well. Currently the strip over the stove doesnt have its ends sealed (since I keep having to open it up to fix it), but the LEDs at the ends have never been the ones to die. LEDs on strip 2 near the stove have died, but so far not since I’ve replaced the power supply (knock on wood). I may temporarily install strip 1 somewhere else for a while to rule that out.

My other thought is that the old power supply outputting 5.5V has weakened some of them. The small “Strip 1” is what I’ve used to for individual testing most often (half because it breaks most often, half because it’s easiest to remove). Now even though I’m using a better power supply, some are dying due to past events.

My final thought is that they might just be bad. Making them “work” by pushing on them makes me think that there are bad internal connections. They may be failing on this strip more often due to the added mechanical stress of installing/uninstalling them, although I dont think I’ve been very rough.

I dont know. Any other thoughts?
(sorry about the novel)

How many LEDs are you using and at how many places do you feed the power to the strip?
Feeding at too few places (too many LEDs via one feeding point) might put too much stress on the power traces.

Total of 269 LEDs. Power is fed at the orange lines in the awful diagram: at the beginning of Strip 1 (43 LEDs), at the beginning of Strip 2 (226 LEDs) and also about 2/3 of the way through Strip 2. Without the last power line, the end of Strip 2 will fade from white to orange at the end.

So you are feeding about 4.5A to strip 2 at the beginning and 2x4.5A at 2/3 (when full brighness white) - that feels a lot IMHO.
I don’t know what the trace geometry on your strips is, but I could imagine that they might get rather warm at times.

Have you checked if the “dead” LEDs just got slightly unsoldered? When they start working by moving the strip it might be due to bad contact too.
Do they still behave badly once taken off the strip and only used solo?

Thats probably about right. But Strip 2 has only had 1 LED pop, while strip 1 has had about 4 pop.

BTW - how did you calculate the amps? I’ve read from a couple sources that these should consume about 60mA per LED, but I’ve measured roughly 38mA on full white so far (measured 5A with 130 LEDs at full white). My Adafruit 5V 10A power supply seems to have no issues powering the full strip, while initial calculations made it look like I’d need a good 16A.

I havent tried resoldering the LEDs yet - the solder points on each LED are super small and my awful soldering iron’s wide tip and awful soldering skills would leave a melted mass of former LED.

I’ve only tried a couple solo, and their behavior was the same as when they were part of the strip. I’ll try more as I’m testing the latest fixes to Strip 1.

I took the datasheet's 60mA and with your setup (226LEDs powered @ #0 & #150) I simplified that feed 1 would power #0~74 (75 x 0.06A = 4.5A) and feed 2 would supply #149~75 and #150~226 (2x ~4.5A).

Oh I see what you’re saying now. I thought you meant about 9A total for Strip 2, which would be pretty close to what I measured, but not the 13.5A you intended. I thought you found maybe found a different datasheet or had some fancy math. :slight_smile:

Anyways, I fixed up Strip 1. I tested the one that died, and it worked even worse than when it was part of the strip. With only 1 LED and the Photon the power supply was delivering just over 5.3V so it’s possible that that killed it further. It’s also possible that I further killed it myself while testing for bad solder points. In any case it’s dead, and it didn’t seem to be misbehaving solely due to it’s position in the strip. Bonus, I killed about 10 more when I connected them to a live circuit (oops) completely backwards (OOPS). Don’t do that.

I’m going to try running them for a while with Strip 1 not installed over the stove. If they last longer, either I’ve finally found and replaced all of the “weak” ones, or heat is the culprit.

Did some temperature testing, and my probe got up to 185F on its own, and 165F in some extra silicone tube. Both could have gone higher, but the kitchen was getting too hot. But heat death is a definite possibility. So if the LEDs hold up out of the hot spot, maybe I need to find some heat resistant LEDs or make some drastic changes. I might be able to come up with something that’d automatically turn on a fan, but probably not in a way that I can hide well…