WLED: ESP32-C6, -C2, -C5, -H2, -P4 are not supported

What happened?

Unable to flash WLED firmware to ESP32-C6-DevKitC-1 v1.1

To Reproduce Bug

Just try to flash wled firmware to any ESP32-C6 chip

Expected Behavior

I guess we need two versions of firmware

Install Method

Binary from WLED.me

What version of WLED?

WLED_0.14.0-b1

Which microcontroller/board are you seeing the problem on?

Other

Relevant log/trace output

[user@user-nvme esp32]$ sudo esptool.py --no-stub write_flash 0x1000 ./esp32_bootloader_v4.bin 
esptool.py v4.4
Found 2 serial ports
Serial port /dev/ttyS4
Connecting......................................
/dev/ttyS4 failed to connect: Failed to connect to Espressif device: No serial data received.
For troubleshooting steps visit: https://docs.espressif.com/projects/esptool/en/latest/troubleshooting.html
Serial port /dev/ttyACM0
Connecting...
Detecting chip type... ESP32-C6
Chip is ESP32-C6 (revision v0.0)
Features: WiFi 6, BT 5
Crystal is 40MHz
MAC: 60:55:f9:f6:a2:bc
Enabling default SPI flash mode...
Configuring flash size...
Flash will be erased from 0x00001000 to 0x00010fff...
Erasing flash...
Took 0.29s to erase flash block
Wrote 65536 bytes at 0x00001000 in 0.6 seconds (900.7 kbit/s)...
Hash of data verified.

Leaving...
Hard resetting via RTS pin...
[user@user-nvme esp32]$ sudo esptool.py --no-stub write_flash 0x10000 ./WLED_0.14.0-b1_ESP32.bin 
esptool.py v4.4
Found 2 serial ports
Serial port /dev/ttyS4
Connecting......................................
/dev/ttyS4 failed to connect: Failed to connect to Espressif device: No serial data received.
For troubleshooting steps visit: https://docs.espressif.com/projects/esptool/en/latest/troubleshooting.html
Serial port /dev/ttyACM0
Connecting...
Detecting chip type... ESP32-C6
Chip is ESP32-C6 (revision v0.0)
Features: WiFi 6, BT 5
Crystal is 40MHz
MAC: 60:55:f9:f6:a2:bc
Enabling default SPI flash mode...
Configuring flash size...
Unexpected chip id in image. Expected 13 but value was 0. Is this image for a different chip model?

A fatal error occurred: ./WLED_0.14.0-b1_ESP32.bin is not an ESP32-C6 image. Use --force to flash anyway

Anything else?

No response

Code of Conduct

  • I agree to follow this project’s Code of Conduct

About this issue

  • Original URL
  • State: open
  • Created 4 years ago
  • Comments: 58 (23 by maintainers)

Most upvoted comments

The new RMT support is in the NeoPixelBus/Core3 branch. Still not configured to use the new boards yet. Still waiting on an updated IDF to fix issues others have seen with it.

What’s working is Art-Net network output to professional/all Art-Net control boxes.

Fixed that. I said “semi-professional” once but it actually works on all Art-Net controllers, and it’s stupid fast. I was referring to “semi-professional” as more a price point.

It’ll work on all of them, from using WLED itself as an Art-Net endpoint to middle range $150 Chinese controllers all the way up to “this costs a stupid amount of money because it’s meant for stadium shows” controllers.

To say the P4 port isn’t useful isn’t true - it works extremely well with Art-Net, and Art-Net is what you want when you’re running huge displays anyway, as you can have essentially unlimited “pins” (outputs) with Art-Net. Parallel outputs, stupid fast, accurately clocked.

It costs more money, but using the P4 to drive a few local LEDs is kinda overkill even if it worked.

The P4 is for driving a LOT of LEDs… I’ve pushed over 65,000 LEDs with it via Art-Net from WLED.

Nothing on the market does this for $0 software at this rate of speed with full audio reactive effects.

It’s useful - but more on the “I’m a crazy person” level. 😁

OT: wow, if this meme is insulting for you, must be a hard life then.

You write that the esp32-c3 and esp32-s3 boards, the mini ones, are not supported normally yet

Nobody said that. MCUs, C3 and S3 (and S2) are well supported in WLED. The issue is development board implementation by different hardware manufacturers. There are variants that will not play well with officially distributed versions of WLED binaries. In such case you’ll need to compile (and figure out compilation parameters) yourself. There are also poorly designed boards out there. No software can help with that.

do you have some firmware posted, are these boards supported or not

Yes we have. Yes and no. Depends on particular development board. See above.

If yes, do they work well?

Depends on what you consider “well”. There may be connectivity issues if your WiFi is not perfect, you may see performance lag compared to ESP32, etc. As there are numerous development boards available and not all are made equal we cannot support all of them. Currently working “well” are the following (the ones I tested, @softhack007 and @DedeHai may add theirs):

  • Lolin ESP32-S2 mini (single core, low on RAM, direct replacement for Wemos D1 mini)
  • LilyGo TTGO T-01 C3 (ESP32-C3, single core, slow speed, plenty of RAM, direct replacement for ESP01)
  • Espressif ESP32-C3-DevKitM-1 (ESP32-C3, single core, slow speed, plenty of RAM)
  • Lolin C3 pico (ESP32-C3, single core, slow speed, plenty of RAM, direct replacement for Wemos D1 mini)
  • generic ESP32-S3 Zero (dual core, fast, plenty of RAM, bad WiFi performance, gets hot, bad effect performance)
  • generic ESP32-C3 Zero (single core, slow speed, plenty of RAM, poor WiFi performance)

Supporting those newer chips requires updating to the esp-idf v5.1. From what I’ve seen, the migration documentation is pretty good. A potentially valuable bonus is that some of the newer chips, like the S3, now support DMA over the RMT interface. From my experiments, enabling DMA reduces the CPU usage during a transmission to 0.01% compared to about 18% using PIO. I believe this would also reduce the chances of a rogue interrupt causing a glitch proportionally. I’m not sure if this project bumps up into those limits or has a problem with interrupt-caused glitches, but if so, that might be another reason to consider it.

Then be respectful and ask politely instead of posting insulting memes.

Similar for other very brandnew chips: -C2, -C6 , -C61, ESP32-H2, -H4, or ESP32-P4, or -C5.

To put it into some context - market availeability of ESP32-C6 was announced just 3 weeks ago on Jan 6, 2023. Knowing espressif, I bet it will take at least 2 years until stable development environments and open source libraries are ready for any “serious use” on -C6.

We should put it on the backburner, until stable support is availeable from the core libraries we need, for example:

For now, we have enough troubles to make -S3/-S2/-C3 work. So don’t expect support for other new models this year.

Image

@troyhacks Is it complicated to transplant to ESP32C6?

@Cherish-Gww the -P4 port from @troyhacks is based on my work for the esp32-C6 (see a few posts up). So yes it could be ported to C6, but the result will not make you happy right now.

Currently we have no (zero, nada, nothing, nix) usable support for led pins, not for digital leds and not for analog leds, either. That’s due to massive API changes in the new framework versions we are forced to use for C6 and P4. What’s working is Art-Net network output to semi-professional light control boxes.

So there is still a long way ahead of us. The message is, we’re moving.

@AskAlice

The first ESP32-P4 I could physically get into my hands arrived sometime in August of 2024.

I went to Burning Man and got back early September and started work and had “first lights” over Art-Net by the end of September 2024.

I shipped Makuna some P4 boards somewhere along the line and around Feb 2025 we had the first physically attached LEDs on the ESP32-P4 via NeoPixelBus.

So… within 1 month of having one in my hands it was booting WLED-MM and doing things, and within 6 months we had it “fully” working.

So that’s the output of 2 people and cost me personally a few hundred dollars to get some of the earliest ESP32-P4 dev boards available - not to mention 3 solid weeks of all my free time just to get WLED to work with IDFv5 and also figure out how to build the Arduino-ESP32 framework for IDFv5 and integrate the new API needed for these boards into WLED which wasn’t even released officially from Espressif for the P4 and the other new boards.

All this work and I’m probably the only person using WLED on the ESP32-P4 in the wild.

But it was worth it. ❤️

The new WLED-MM P4 code I pushed today also moves everything that touches LEDs into 32-bit counters, allowing for more than 21845 RGB LEDs to be addressed. My tests previously with 65000 LEDs did work… but the 16-bit counters were wrapping. I didn’t notice because I don’t have 65000 actual LEDs to test with. 😆

Image

Image

Image

Poor Akemi only runs at 11 FPS on 65025 pixels… but everything runs just fine. 383 Art-Net universes.

Image

2D matrix widths beyond 255 are also completely possible now too. 😄

Image

My P4 build has been updated to work with locally attached LEDs, thanks to @Makuna !

At the moment only one physical output is working, but that’s likely more my fault and not understanding the code - I’m sure someone more involved with the bus classes could fix it up. This should also work for the C6 and H2, etc.

I believe only 4 RMT outputs are possible on the P4 and likely on the other RISC-V chips. Big (eventual) possibilities of new drivers on the newer hardware tho, like Parallel IO which can push up to 16 pins fast. This is entirely speculation at the moment and no LED drivers exist for this feature.

As an update on progress on the RISC-V chipsets, the ESP32-P4 is running the WLED-MM fork very well with Arduino-ESP32 built on IDF v5.3.2 - but it’s entirely limited to network pixels at the moment.

https://github.com/MoonModules/WLED/tree/P4_experimental

There’s still a few rough code edges, but I have several of these (still hard to find) boards in use in the wild, including one at a venue where it’s driving their Art-Net lighting installation with over 11000 pixels every night with full audio reactive effects - so the build is reliable.

Once Makuna has some NeoPixelBus magic done for the P4 it should just snap in and we’ll have local LED output as well.

@Makuna - I’d be very happy to ship you a few free ESP32-P4 boards to support development, BTW. Just send me your address. ❤️

PXL_20241111_170456134

18432 pixels in this display pushed from a single ESP32-P4 to Art-Net hardware. 😀

It is listed in official KB. There’s few most often used repos with ready to use binaries.

2. with 5ghz i get more speed on the devices but i am forced to use 2.4 because of my esp32

Sorry but -C6 is still 2.4Ghz only. So it would not help in your situation.

So here ist my list of boards I have.

  • generic ESP32S3 8MB flash no PSRAM–> custom buildenv based on [esp32s3] but without -DBOARD_HAS_PSRAM
  • Lilygo/TTGO T7-S3 16MB flash, 8MB PSRAM --> [env:esp32s3dev_16MB_opi]
  • Freenove ESP32-S3 8MB flash, 8MB PSRAM --> [env:esp32s3dev_8MB_opi]
  • Waveshare ESP32-S3 Zero 4MB flash 2MB PSRAM (needs passive cooling) --> [env:esp32s3_4M_qspi]
  • ESP32-S2-DevKitC-1U-N8R2 with external radio antenna (8MB flash, 2MB PSRAM) --> custom buildenv with 8MB partition table (based on [env:lolin_s2_mini])
  • Lolin S2 mini (has low wifi range) --> [env:lolin_s2_mini]
  • ESP32-C3-DevKitM-1 --> [env:esp32c3dev]
  • Lolin C3 mini, Lolin C3 pico (low wifi range) --> [env:esp32c3dev]

plus a “nanoESP32-C6” for experimenting with C6 builds. But that’s absolutely experimental.

Please note that some of the boards need a modified platformio.ini buildenv - usually to adjust for flash size, or add -DLOLIN_WIFI_FIX for the small ones that only have a crystal “radio”. Also I prefer to have boards with a “real” Serial-to-USB chip, because the “USB-CTC” port is too unreliable for me.

I’m looking for the smallest board possible to put a levelshifter on and connect it to the strip. My maximum profile width is 3 cm and the depth is 2 cm. Esp32-c3 is ideal - it just fits, all other boards are bigger unfortunately.

The smallest one I know about is the waveshare S3 Zero - 1.8cm x 2.5cm. It needs the esp32s3_4M_qspi buildenv.

A bit more seriously - there is no schedule for supporting the -C6 in WLED. Personally I have a fork that compiles for -C6, but there is still a looooong way to go. C6 support might come as “experimental” in 0.16 (may be released in one year), and it will be comparable to the -C3 performance wise.

What would be the best way to help out with -C6 support? I should have a few -C6 boards to try out in a couple of weeks, and should be able to contribute some development effort.

Sorry to be that guy, but is there an rough estimate on how long it will take? I’m looking to invest in to these boards, once support is added. Thanks

As you are “that guy” you probably guessed the answer 😉 “it’s ready when its done”.

A bit more seriously - there is no schedule for supporting the -C6 in WLED. Personally I have a fork that compiles for -C6, but there is still a looooong way to go. C6 support might come as “experimental” in 0.16 (may be released in one year), and it will be comparable to the -C3 performance wise.

From experience with other “new boards” like the -S3, I would expect that we’ll need around two years from “experimental” to “WLED runs stable on this MCU”. The same was true for -S3, where we started “experimental support” in mid 2022.

Same for possible support of the “C61” and P4". OK P4 might even take longer because there is no support from arduino-esp32 or other libraries yet.

So if you’re looking at a project or other business case, I’d say don’t bet your money on the -C6.

If you need ZigBee, you could be faster by taking a ZigBee “daughter board” (e.g. XBee module) and then you’d program a usermod that communicates with the daughter board.

It’s been some time since this issue has been created, have there been any advancements with the libraries and such to make the new chips work?

Not enough advances, unfortunately - it looks like my old “at least 2 years” estimate was quite accurate…

If NeoPixelBus would be able to utilize the new RMT driver in esp-IDF 5.1.x, we could make an attempt to get WLED running on the -C6. Maybe this will happen during the next year. For other new chips - nothing in sight.

I’ve been over to the NeoPixelBus repo and looking at Makuna/NeoPixelBus#598 it seems they’re already looking into it.