The WS2812B-4020 apparently are really unreliable

jostsalathe / colock

A USB powered RGB wall clock inspired by analog clocks

0 stars 0 forks source link

The WS2812B-4020 apparently are really unreliable #2

Closed jostsalathe closed 1 year ago

jostsalathe commented 2 years ago

The first batch ordered at JLCPCB (PCB manufacturing and SMT PCBA of most components) had about 10% of the LEDs dead on arrival. I suspect that the LED chips either have a really high failure rate from the factory or are quite fragile so they don't survive the PCBA or the transportation between China and Europe.

This is generally in line with my previous experiences regarding the WS2812B product line by WorldSemi. The SK68 product family by Dongguan OPSCO Optoelectronics seems to be a lot more robust and reliable. They offer compatible products like SK 6812 SIDE-A on their homepage: opscoled.com/en/product/index.html?sid=13.

They also have Variants that are smaller (SK 6805 SIDE-G) or even smaller (SK 6805-EC3210 R) or add a white LED chip (SK 6805 SIDE-F RGBW-WW-P6). I gravitate to one of the two latter. Additional white would be quite nice but so would be the tiny 3210 package...

jostsalathe commented 2 years ago

I fiddled around with the data timings using this great test firmware on an Arduino Nano while consulting this fantastic article about WS2812B timings but I unfortunately was not able to get successful data transmission to all 120 LEDs whatever I did to the timings.

At this point I give up on that type of LEDs. I think I'll check out SK 6812 SIDE-A.

jostsalathe commented 2 years ago

Some more investigation with a logic analyzer raised the suspicion that there may also be a power delivery problem with presumably the segment interconnect. The contact cross section of the current solution probably is neither that great nor that consistent so #1 should greatly improve on that.

I probably should also lean towards the 5 mA variants of the SK 68 family to reduce the max current draw a bit... Still unsure if the 5 mA refers to a single color or the whole package. I suspect 5 mA per chip which would mean 15 mA x 120 = 1.8 A (or 20 mA x 120 = 2.4A for RGBW variants) which is more or less reasonable, I guess.

jostsalathe commented 1 year ago

JLCPCB offers the following variants of SK 68XX SIDE LEDs:

SK6812SIDE-A (C5378721) is the basis for the current footprint design (44e4420a77ac9b469bbdeaa6783c847538c6f4dc) but is not in stock
SK6812SIDE-A-RVS (C2890037) has the same footprint as SK6812SIDE-A
SK6805SIDE-S (C2909057) has a considerably different footprint and is really flat and wide so I suspect that the individual colors are more spread out in the package and maybe mix less nicely
SK6805SIDE-G-001 (C2909060) is only available for standard PCBA and not for economic

So I'm gonna go with the 12 mA A-RVS variant and just dim the lights more in software... This probably also makes the software more compatible with the APA106 LEDs that probably have a light output more similar to the SK6812 than the SK6805.

jostsalathe commented 1 year ago

The changes are reflected in the new README and Layout of commit eb1d6db0c4d97f9ff2cdd2f8c11dac122732fe76.