WS2816B non-functional on ESP32 with default 3-step I2S cadence

[gsuberland]

Describe the bug
WS2816B LEDs do not work out of the box with the default 3-step I2S cadence.

Defining NPB_CONF_4STEP_CADENCE before importing the NeoPixelBus.h or NeoPixelBusLg.h header fixes the issue.

To Reproduce
Steps to reproduce the behavior:

1. Wire a single WS2816B LED to the ESP32.
2. Set up a simple sketch using NeoGrbWs2816Feature and NeoWs2816Method.
3. Run it.
4. LED goes random colours, locks up white, or turns off.

Expected behavior
LED works correctly.

Development environment (please complete the following information):

• OS: Win10 x64
• Build Environment: Arduino IDE 2.3.7
• Board target: DOIT ESP32 DEVKIT V1 (esp32 v3.3.5)
• Library version: 2.8.4

Minimal Sketch that reproduced the problem:

#include <NeoPixelBus.h>

NeoPixelBus<NeoGrbWs2816Feature, NeoWs2816Method> strip(1, 4);

void setup() {
  strip.Begin();
}

void loop() {
  strip.SetPixelColor(0, RgbColor(255, 0, 0));
  strip.Show();
}

Additional context

I took a look with an oscilloscope, and the output is incorrect without the NPB_CONF_4STEP_CADENCE define. The signal looks mostly WS2816-like but the timing is sloppy and has some jitter. I found this define mentioned in #828 and presumably the 33/66 pulse duty timing is too loose for the WS2816B to accept.

With NPB_CONF_4STEP_CADENCE defined the LED works as expected. In my opinion, documenting this requirement would be sufficient as a short-term fix.

[Makuna]

Not sure what to do about this short term. We have other chips that don't work with 4 step due to similar issues.

Longer term, cadence will be exposed so that it can be defined by the method rather than this compile type define.

[gsuberland]

I think documenting it somewhere would be a good start, even if it's temporary. I spent a few hours on this thinking I'd messed up my design, or maybe got a fake LED, so even a quick sentence in the methods docs about which LEDs are known to currently need the cadence switch would help others avoid that.

[gsuberland]

I didn't realise editing another project's wiki just automatically committed the results with no approval (I was expecting it to make a PR to the wiki backing files or something), so feel free to revert, but I've added a quick paragraph here to cover the issue and mitigation:

https://github.com/Makuna/NeoPixelBus/wiki/Neo-Methods

[Makuna]

I didn't realise editing another project's wiki just automatically committed the results with no approval

Yeah, for some reason they didn't expose it that way. I should get notifications of changes and I can revert them. Be professional, go to the discussion's areas for questions about best practices if it's a major change.

Some projects have a root folder they put documentation into so it you get the merge and reviews. I kept holding out for them to improve the wiki support. Heck, my project is older than the actual Wiki feature itself 8-)

[gsuberland]

Hah, well hopefully the changes are reasonable enough for now, and lesson learned for next time! :)

Happy to contribute further to docs if needed.