0

I created an LED picture on a big PCB with almost 2000 WS2812B-2020 LEDs. The bottom layer is round, and the top layer is 5V and has the LED signal routes.

I use a Pi Zero with a 74AHCT125 to control the LEDs and a 40A 5V power source connected to the PCB with an XT60 plug. I run the LEDs on around 10% of the max brightness.

Sometimes the last couple hundred LEDs behave strangely with random colors and or brightness changes. I think the signal may get weaker by the end, but it's also not happening all the time, just sometimes, which is strange. Also, when all the LEDs show the correct colors, and I disconnect the Pi, this issue stops, and the randomness never happens again, even after days.

Any idea why this happens?

Thanks in advance!

Screenshot of the PCB: enter image description here

Picture of the LED:

enter image description here

Klausmaus
  • 11
  • 2
  • 1
    Do you just have one data line in series for all 2000 LEDs? – vir Apr 03 '23 at 19:44
  • Is the ground plane really only connected at those points? – Voltage Spike Apr 03 '23 at 19:58
  • Usually the problem with neopixels is voltage drop over distance, but it looks like your +5V and ground are *plenty* big and you're only using them at 10% brightness so that isn't the problem. Although with 2000 LEDs maybe it still could be a problem - check the voltage at the far end and find out. – user253751 Apr 03 '23 at 20:04
  • 1
    Next most common is data line interference from EMI or from a broken LED (e.g. due to accidental ESD damage when building the circuit). If the problem always starts at the same LED try replacing that one. – user253751 Apr 03 '23 at 20:04
  • 1
    Finally I notice there aren't any decoupling capacitors. I saw recommendations that each LED should have 100nF in parallel with +5V and ground. You could try adding some in the area where the problem occurs and see if that helps at all. – user253751 Apr 03 '23 at 20:06
  • 2
    One thing that is *not* a problem is data strength fading over distance. That just isn't a thing, since each LED restores the data signal back to full strength. – user253751 Apr 03 '23 at 20:06
  • @vir Yes just one data line, is there a way to have multiple? – Klausmaus Apr 03 '23 at 20:10
  • @VoltageSpike No the whole backside of the PCB is ground and also connected to the XT60 – Klausmaus Apr 03 '23 at 20:10
  • @user253751 No it starts at different LEDs, the datasheet I found says no caps are needed for the new version of the 2020. – Klausmaus Apr 03 '23 at 20:11
  • @Klausmaus It looks like you have two separated ground planes – Voltage Spike Apr 03 '23 at 20:34
  • Try decreasing the data rate. If that fixes it (and faster rates are worse) then you know you have a signal integrity issue. – rdtsc Apr 03 '23 at 20:38
  • I would expect you’d need some bypass caps. 2000 times, say, 10mA per device is 20A with lots of noise created by pwm is likely to cause problems. – Kartman Apr 03 '23 at 20:59
  • If you look closely at your LEDs, do they have a small capacitor next to the control chip? Some do, some do not. If not, you need to add bypass caps yourself. I probably would have added them regardless though, that's a lot of RF energy bouncing around. – user1850479 Apr 03 '23 at 22:08
  • @user1850479 I can't really tell because it's so small. Would you add them to every LED or just every other? – Klausmaus Apr 03 '23 at 22:25
  • @Kartman How many would you use, and what size? – Klausmaus Apr 03 '23 at 22:35
  • Check with a magnifying lens and see if they put one in. Typically every digital chip needs a decoupling capacitor, either on package or on board. You might be able to get away with one shared if the chips are very close, but this isn't good practice. – user1850479 Apr 04 '23 at 00:40
  • @user1850479 I found my old microscope. I added the picture to my post. It doesn't look like it has one, right? – Klausmaus Apr 04 '23 at 01:38
  • Doesn't look like it has a capacitor. See this answer for an image with a cap: https://electronics.stackexchange.com/questions/548666/individual-ws2812b-circuit-do-i-need-a-capacitor-power-distribution – user1850479 Apr 04 '23 at 01:46

1 Answers1

0

probably you have the 5V supply bbouncing around a bit on those long fingers of 5V

Add some capacitors ( the datasheet reccomend one near each WS2812B across 5v and ground ("104" means 100nF)) , as a retrofit you can scratch off the solder mask and solder to the via and the 5V fill.

You probably don't need 2000 capacitors, probably 200 will be enough.

Add some stitching on the back side of the circuit board to connect the 5v stripes together (bridging behind the data line), or use SMD links on the front side. For a retrofit add links on the front, but when you redesign you can print the link on the back and use vias unless there's a requirement to have no holes in the ground plane)

Jasen Слава Україні
  • 31,874
  • 1
  • 31
  • 65