Neopixels: glitchy after the first few LEDs with ESP32-IDF+SPI

[scoates]

Hello. Starting out by saying that wokwi is amazing. There's so much here, and I hope you've proud of the work that's gone into this. I was happy to join the club and send you some dollars.

Describe the bug

There seems to be a bug in the simulated wokwi-neopixel-matrix where it glitches out after the first 6 (or so) ws2812s in the chain.

I've tried this on real hardware, and it works as I'd expect (more than 6 LEDs are able to be addreesed).

https://github.com/wokwi/wokwi-features/assets/71983/54a41cda-ba4e-4720-8e01-7504bec7eb28

(This is running on the result of idf.py flash from the same binary that's installed in the wokwi simulator in the above screenshot.)

This is on ESP32-C6 with IDF and lighting the pixels via SPI (you can see this in the code).

I've also tried this with stringing a bunch of independent neopixels together (in wokwi; the simulator), and it also glitches out, so it's not exactly the panel that is failing; it's probably the SPI timings? I didn't manage to get it to work at all with RMT instead of SPI, so I'm not sure if it's directly tied to SPI emulation. I do know that SPI works on the real hardware.

I don't really know enough about SPI and timings to do much with the logic analyzer, so I don't have that to share, but someone with the right knowledge can attach that to the project easily, I'd think.

To Reproduce I have a repository over here that illustrates the problem. Run this code in the simulator and it fails. There's a UF2 here which can be uploaded into this example sim and it shows the problem (the default code only shows the hello message, use F1 + Upload the downloaded uf2.bin).

Expected behavior I expect the 9 LEDs to light up blue both on the real-world matrix and on the simulated one. But the Simulated one glitches after 6. I limited my test to 9 but it fails after 6, here, even if I try to address all 256. Some are unlit, many are white; many are the wrong colour. It really looks like a timing issue which I experienced on other platforms.

Environment:

• OS: not sure it's relevant, but MacOS, M2 MBP, ESP32 C6 + IDF.
• Browser: happens both in Firefox and in the VSCode Extension.

[scoates]

FWIW, I tried this on the simulator only with the c3 and s3 boards (the h2 board doesn't seem to have the spi capability in IDF), and it failed (glitchy) with even fewer blue neopixels.

[urish]

Thanks for the super detailed report Sean, it's really helpful. Also thanks for all the kind words! ❤️

I was able to reproduce the issue, and am now looking into it.

[scoates]

Thanks for taking a look. Let me know if I can help in any way.

[urish]

Update: I believe I found the cause - it's a combination of too tight timing margins in the WS2812 decoder together with inaccurate timing of the SPI transmitter. Working on a fix.

[urish]

Pushed a fix - can you please test and report?

[scoates]

Yes! All 256; no glitches. Thank you!