WS2812 LED failure

[Nogtail]

Around a year ago I got myself a reel of LED tape with everyone's favourite addressable RGB LEDs. I used around half a meter on a unicorn vomit inspired project of some description and put the rest in storage. I was recently feeling the call of more retina vaporising rainbows, but unfortunately a year in storage had been harder on the strip than I'd expected. Around one in 5 LEDs have a failed diode. It seems like the red and blue diodes were the first to go, as a lot of the strip is showing cyan, green or yellow when set to white.

Is this a common failure? I've heard of dead LEDs relatively often, although most of the time it seems to be caused by a dead driver chip. This reel has never been powered up and has been stored at a consistent temperature in a sealed bag. As far as I can tell it's a mechanical issue (dicky bond wires?) as pressing on the surface of the LED often brings it back to life.

Does anyone have recommendations for a more reliable addressable LED? I've had more luck with the SK6812 RGBW LEDs so far but haven't used enough of them to see the issues.

[L0R3NZ0_L30Z]

Hi, I never had such a problem with the w2812 LED strip.
Knowing that the red and blue LEDs fail and that they spring back to life when you press them, I think that it could be a pcb trace problem or  something with the solder joints... I would try reflowing them and see if that changes anything. You also could try cutting the strip and see if it was something at the beginning.
I always used the w2812, and it has worked for me. If you buy the same model make sure that you buy them from a different manufacturer.

[coromonadalix]

i would say they were badly soldered ?? or cold solder joint ??

are they enclosed in a rubber jacket ??  mecanical stress  if they have been stored in / like  a spool ??

i've seen  many grade with addressable or rgb led strips coming from c.......   

[AndyBeez]

A test is to provide power (V+ and Gnd) to both ends of the strip, so current spills in both ways. On white, ALL leds are consuming power, power that is also needed to ensure clean data transfer along the strip.

Are there dead Leds (gaps) when exclusively setting red, blue or green? If not, it's likley a supply or data quality issue.

[electr_peter]

I have used W2812 on strips, but newer had a failure described above.

As far as I can tell it's a mechanical issue (dicky bond wires?) as pressing on the surface of the LED often brings it back to life.

Can you take a closer look under microscope? Maybe you will see exact reason of failure and then be able to mitigate it.

[james_s]

If these are actually WS2812 LEDs then it can't be soldering, the driver ICs are integral to the LED package. If it was a soldering issue you wouldn't lose just one color.

[uer166]

I've had maybe 1-2% WS2812 LEDs fail for no real reason on strips, and applying mechanical stress into the failed LEDs seemingly "fixed" some, albeit temporarily. I think Mike had also a similar fallout?
They generally seem crap-tastic and I don't think it's the integrator's fault here.

[james_s]

So far I haven't had one fail, but I have heard that they are very fragile and easily damaged when installing them. They are amazingly cheap so it wouldn't surprise me if they're cheaply made.

[artag]

I've had them fail quite a lot when given a hard life - flexing the strip etc. But as james says, it can't be soldering if only one colour is affected.

I have had both one-colour and data loss - the data loss ones can sometimes be fixed by jumpering the data line from one package to the next, so is probably a cracked track.

I haven't had any fail in storage but bear in mind that as ebay or aliexpress buyers we're probably getting the scrapings of the quality bin. A proper commercial supply arrangement with a manufacturer would be wise.

On one installation that failed a lot I bought the ws2813 leds that have a backup data line. I still got failures, and they were still affected by pressure.

I have seen speculation that the gel lens doesn't bond well to the white case and lets air in which might corrode the bondwires over time.

[tooki]

Around a year ago I got myself a reel of LED tape with everyone's favourite addressable RGB LEDs.
…
Does anyone have recommendations for a more reliable addressable LED? I've had more luck with the SK6812 RGBW LEDs so far but haven't used enough of them to see the issues.

FWIW, the ‘2812s aren’t my favorite. The low PWM frequency is super annoying to me. The SPI-like APA102C is better in that regard, but APA102C clones and successors like the SK9822 and HD107 are much better still.

[szan]

The low PWM frequency is super annoying to me.

We live in a quantum analog world, so PWM is a disaster 
I can show you recorded video of back car lights which just look nasty by superimposing 30fps recording with probably some kind of PWM control on them 
That is why I decided to use linear regulator LEDs output driver and I'd like to change whole room 230VAC LED lights with custom made powered by battery DC, since I'm sure that most of those shop home LEDs blinks at 100Hz since here in Europe we have 50Hz mains, but of course I will check it using light sensor connected to home computer   

[tooki]

The low PWM frequency is super annoying to me.

We live in a quantum analog world, so PWM is a disaster 
I can show you recorded video of back car lights which just look nasty by superimposing 30fps recording with probably some kind of PWM control on them 
That is why I decided to use linear regulator LEDs output driver and I'd like to change whole room 230VAC LED lights with custom made powered by battery DC, since I'm sure that most of those shop home LEDs blinks at 100Hz since here in Europe we have 50Hz mains, but of course I will check it using light sensor connected to home computer 

No need for a video, I can see those awful tail lights with my naked eyes... I've posted in many lengthy discussions on here about PWM frequencies with regards to flicker. (The name for the trail of blinking dots is called the "phantom array" effect, by the way.)

I wouldn't say PWM is a disaster. It's fine if done right. I built a PWM LED dimmer for my under-desk lighting that runs at something like 26KHz. My favorite LED driver chip does PWM at 97KHz. Any of those is far, far above what's needed. (But the "sweet spot" of switching efficiency and no flicker [something like 5-10KHz] is in the audible frequency range, so it's better to go higher.) While linear/CC dimming has some advantages, it's less efficient, not as straightforward to control digitally, and can cause color shifting, since LEDs' wavelengths are slightly sensitive to drive current.

Most household LED lighting use switch-mode power supplies and don't flicker. You don't need to use a fucking battery, just use a regulated power supply... 

[uer166]

The low PWM frequency is super annoying to me.

We live in a quantum analog world, so PWM is a disaster 
I can show you recorded video of back car lights which just look nasty by superimposing 30fps recording with probably some kind of PWM control on them 
That is why I decided to use linear regulator LEDs output driver and I'd like to change whole room 230VAC LED lights with custom made powered by battery DC, since I'm sure that most of those shop home LEDs blinks at 100Hz since here in Europe we have 50Hz mains, but of course I will check it using light sensor connected to home computer 

No need for a video, I can see those awful tail lights with my naked eyes... I've posted in many lengthy discussions on here about PWM frequencies with regards to flicker. (The name for the trail of blinking dots is called the "phantom array" effect, by the way.)

I wouldn't say PWM is a disaster. It's fine if done right. I built a PWM LED dimmer for my under-desk lighting that runs at something like 26KHz. My favorite LED driver chip does PWM at 97KHz. Any of those is far, far above what's needed. (But the "sweet spot" of switching efficiency and no flicker [something like 5-10KHz] is in the audible frequency range, so it's better to go higher.) While linear/CC dimming has some advantages, it's less efficient, not as straightforward to control digitally, and can cause color shifting, since LEDs' wavelengths are slightly sensitive to drive current.

Most household LED lighting use switch-mode power supplies and don't flicker. You don't need to use a fucking battery, just use a regulated power supply... 

IIRC anything above 2kHz should be invisible/undetectable by a human, even including persistence of vision effects. So definitely PWM is perfectly fine when done right. Of course cameras shutter can still pick it apart.

[szan]

Most household LED lighting use switch-mode power supplies and don't flicker. You don't need to use a fucking battery, just use a regulated power supply... 

IIRC anything above 2kHz should be invisible/undetectable by a human, even including persistence of vision effects.

Nobody knows how brain processes information, but people used to use sun light and natural light sources, so even  a few kHz light noise comming from modern LEDs might have some negative impact on how you feel.
When you didn't notice something moving very fast it doesn't mean that it has no impact on visual information processing in brain.
Another issue of moders LEDs is lack of IR radiation, which means that water wapour in air in a house rooms can't be heated, so nowadays people breathe with a bad quality air not energized by infrared radiation 
Of course most 230VAC home LEDs are flickering, but with higher noisy frequencies, while in the case of old light bulbs it was 100Hz flickering at 50Hz mains easy detectable by photoresistor, however those old more power needed bulb lamps had positive effect as a source of infrared light, which is esencial for human health, so modern "more" efficient LED lamps... leave people at homes with negative energy in the air 
 

[james_s]

If it's moving fast enough even flicker in the MHz can be seen by the naked eye, that's pretty much how analog oscilloscopes work. Stationary light sources strobing up into the hundreds of Hz are clearly visible by darting your eyes past them.

[tooki]

IIRC anything above 2kHz should be invisible/undetectable by a human, even including persistence of vision effects.

It’s way more complicated than that. Some people can detect flicker at 6kHz, as found in one study, though 1-3kHz was more typical. (Numbers from memory, don’t hold me to them. But the source is in a post of mine in another thread.)

Others have calculated the speeds needed to eliminate flicker and phantom array effects and arrived at 35kHz.

The upshot is that in addition to individual sensitivity, it also depends on whether the light source is stationary or moving, the human is stationary or moving, and eye movement. The switching characteristics of the light source matter (which is why incandescent 50Hz flicker isn’t as bad as 50Hz LED flicker: the thermal inertia of the filament acts as a thermal LPF). The size of the light source matters. Point sources cause phantom array effects sooner than diffuse sources. Specular highlights in the room matter, as they become point sources. Etc etc etc.

This is why it annoys me when knowitalls here (not talking about you) repeat the “anything above 60Hz is flicker free” nonsense, because it’s nowhere near that simple.

[uer166]

This is why it annoys me when knowitalls here (not talking about you) repeat the “anything above 60Hz is flicker free” nonsense, because it’s nowhere near that simple.

I agree 100%, my take was more of a "typical" use case where it becomes a non-issue for most people, it certainly is more complicated than that, but if it's some burning man art installation with WS2812s, it's not worth caring about it.

What annoys me is also generalizations (and I may have made one here as well, although I was trying to be pragmatic), such as "never use PWM" and spew some pseudo-science BS.

[thm_w]

They were reliably able to tell the difference between 60kHz and 11kHz flickering lights by looking at the trailing dots (phantom array).
Doesn't necessarily mean that 1kHz PWM is not enough, it just means some will see artifacts in certain scenarios (bright light at night and moving around).

https://journals.sagepub.com/doi/10.1177/1477153519852391
https://www.led-professional.com/resources-1/articles/lighting-with-leds-2013-more-than-just-illuminating-objects
https://www.ies.org/research/fires/concerns-in-the-age-of-the-led-temporal-light-artifacts/

Once the flicker frequency was about 1000 Hz or higher, differences in flicker modulation amount were unimportant for acceptability of stroboscopic effects (Figure 4), but the amount of modulation had a large impact on detection of the stroboscopic effects of flicker (Figure 3). In other words, even when many subjects were able to detect stroboscopic effects, most of them did not necessarily find the flicker unacceptable on this basis.

[amaschas]

I wouldn't say PWM is a disaster. It's fine if done right. I built a PWM LED dimmer for my under-desk lighting that runs at something like 26KHz. My favorite LED driver chip does PWM at 97KHz.

Do you have any recommendations for high frequency, high resolution PWM LED drivers?

[tooki]

They were reliably able to tell the difference between 60kHz and 11kHz flickering lights by looking at the trailing dots (phantom array).
Doesn't necessarily mean that 1kHz PWM is not enough, it just means some will see artifacts in certain scenarios (bright light at night and moving around).

You say that as if that’s some trivial edge case that can be safely ignored, but that’s precisely the problem: it’s exactly those artifacts that I, and others, find extraordinarily distracting and annoying.

[tooki]

I wouldn't say PWM is a disaster. It's fine if done right. I built a PWM LED dimmer for my under-desk lighting that runs at something like 26KHz. My favorite LED driver chip does PWM at 97KHz.

Do you have any recommendations for high frequency, high resolution PWM LED drivers?

I’ve used the NXP PCA9634 (which can drive external drivers for higher voltages/currents) and the sibling PCA9624 (built in 40V, 0.8A drivers; recently discontinued, so stock up now if you want them). They do 97kHz 8-bit channel dimming, and can also overlay a 190Hz 8-bit global dimming. I believe it’s also possible to supply your own global dimming PWM signal at whatever frequency you want.

[thm_w]

They were reliably able to tell the difference between 60kHz and 11kHz flickering lights by looking at the trailing dots (phantom array).
Doesn't necessarily mean that 1kHz PWM is not enough, it just means some will see artifacts in certain scenarios (bright light at night and moving around).

You say that as if that’s some trivial edge case that can be safely ignored, but that’s precisely the problem: it’s exactly those artifacts that I, and others, find extraordinarily distracting and annoying.

Its a tail light, can you see the tail light and its not affecting your vision of other objects? Yes? It did its job then.
Seeing a close trail of dots behind it instead of a continuous blur hardly seems "extraordinarily distracting". As I quoted, even for most that noticed the effect, it was not an issue.

Many people get starbursts or halos at night, due to cataracts or other issues, which is far more distracting (as it obscures areas outside of the tail light). But they manage to survive: https://www.allaboutvision.com/symptoms/starburst-lights/

[tooki]

They were reliably able to tell the difference between 60kHz and 11kHz flickering lights by looking at the trailing dots (phantom array).
Doesn't necessarily mean that 1kHz PWM is not enough, it just means some will see artifacts in certain scenarios (bright light at night and moving around).

You say that as if that’s some trivial edge case that can be safely ignored, but that’s precisely the problem: it’s exactly those artifacts that I, and others, find extraordinarily distracting and annoying.

Its a tail light, can you see the tail light and its not affecting your vision of other objects? Yes? It did its job then.
Seeing a close trail of dots behind it instead of a continuous blur hardly seems "extraordinarily distracting". As I quoted, even for most that noticed the effect, it was not an issue.

Many people get starbursts or halos at night, due to cataracts or other issues, which is far more distracting (as it obscures areas outside of the tail light). But they manage to survive: https://www.allaboutvision.com/symptoms/starburst-lights/

How do you know that’s not super distracting to some people? Just because it’s not distracting to you doesn’t mean it’s not super distracting to me and others. You don’t get to decide what other people find distracting!!!

If a distraction draws my attention toward it when it shouldn’t — and those trails of dots very much do — then it is “affecting my vision of other objects” (by affecting perception) and introducing a hazard.

[Prime]

I had lots of crap with them. First they changed the dye and the code had to be changed. There was no revision control and you had to scrutinize the dye to determine which version was in use.

On a commercial run, we had 110 in 5500 fail. At least one colour would stop working. I honestly wouldn't use them in anything critical...

Both batches had a certificate of conformity and were traceable.

[Ground_Loop]

I installed numerous strings of about 100 LEDs each for under cabinet lighting and noticed a stark difference between the genuine neopixels from Adafruit and the cheep knock offs both of which I got off Amazon.  The Neo Pixels are an even white when all colors are on and each color individually is bright and saturated after about six months of full intensity, half day use.  The cheep ones are fading especially the blues.  I have already resigned myself to replacing all the inexpensive ones.  Adafruit per meter costs about 3X what the inexpensive ones do.