No lvgl_micropython REPL on the Waveshare ESP32-S3-Touch-LCD-4.3B

[rasjidw]

I'm having trouble getting a working build for the Waveshare ESP32-S3-Touch-LCD-4.3B.

I figured it would be very similar to the Waveshare ESP32-S3-Touch-LCD-4.3, and have read through discussion #142, but still seem stuck.

I can flash and run the vanilla micropython build - ESP32_GENERIC_S3-SPIRAM_OCT-20250415-v1.25.0.bin and connect to the repl with mpremote, and see the 8MB of ram.

I can build lvgl_micropython with:
python3 make.py esp32 BOARD=ESP32_GENERIC_S3 DISPLAY=rgb_display INDEV=gt911
and that works and I can get a repl, but I don't have access to the SPIRAM in that case and don't have enough ram to actually do anything.

If I build with:
python3 make.py esp32 BOARD=ESP32_GENERIC_S3 BOARD_VARIANT=SPIRAM_OCT DISPLAY=rgb_display INDEV=gt911
then I am unable to connect, or at least I don't get a repl.

I'm running on Ubuntu 24.04, and connecting via a USB-C cable.

Been programming in Python for years, but I'm completely new to micro-controlers.

Any advice on how to proceed or diagnose what the issue is would be greatly appreciated.

[kdschlosser]

OK I added an automatic build file for you to be able to get up and running pretty quickly.

You need to clone the repo again

git clone https://github.com/lvgl-micropython/lvgl_micropython
cd lvgl_micropython

you need to make sure that you have the toml library installed into python 3. this only needs to be done a single time...

pip3 install toml

next is to run the build

python3 make.py toml=display_configs/Waveshare-ESP32-S3-Touch-LCD-4.3B.toml

That should compile. Use the instructions for flashing the ESP32 that are output to the console once the compilation has finished.

Once you have the firmware flashed you will need to use a mpremote or a python library called ampy to connect to the ESP32 to be able to upload files to it. You want to upload your script as a file named "main.py". In that script you want the following code.

import display
import lvgl as lv

slider = lv.slider(lv.screen_active())
slider.center()

once you upload the file reboot the board and the screen should come up and a slider should be displayed. If you do not see the slider let me know. You should also be able to touch the screen and move the slider. If that doesn't work or it doesn't work properly let me know.

[rasjidw]

Hi Kevin,

I did a bit of a manual git bisection, and the last commit that gives me a REPL I can connect to is 0999246 ("Fixes RGB rotation..", 2025-06-12).

With that I can get the display and touch screen working, which is fantastic, but of course I'm missing some of the newer changes you have done in the code since then.

Happy to test any changes that would get it working again with the current codebase.

[kdschlosser]

which is interesting because there is nothing that has changed after that PR that would stop you from getting a REPL. What are you using as a build command?

[kdschlosser]

I just noticed a comment you made. It states that if you use this build command..

 python3 make.py esp32 BOARD=ESP32_GENERIC_S3 BOARD_VARIANT=SPIRAM_OCT DISPLAY=rgb_display INDEV=gt911

it doesn't work but if you omit the BOARD_VARIANT it does work...

ALL esp32-S3's have SPIRAM so you should always have access to it. I am thinking that what you may not have is octal SPIRAM and instead you have quad SPIRAM.

If we look at what waveshare states on their website for the specifications for the board it states...

Built in 512KB SRAM and 384KB ROM, with onboard 16MB Flash and 8MB PSRAM

so we know it does have SPIRAM. It doesn't state how many lanes though.

But if we go to their wiki page for the board we see this for the ESP-IDF config parameters...

CONFIG_FREERTOS_HZ=1000
CONFIG_ESP_DEFAULT_CPU_FREQ_MHZ_240=y
CONFIG_ESPTOOLPY_FLASHMODE_QIO=y
CONFIG_ESPTOOLPY_FLASHFREQ_120M=y [Need to be consistent with PSRAM]
CONFIG_SPIRAM_MODE_OCT=y
CONFIG_IDF_EXPERIMENTAL_FEATURES=y and CONFIG_SPIRAM_SPEED_120M=y [Need to be consistent with FLASH]
CONFIG_SPIRAM_FETCH_INSTRUCTIONS=y
CONFIG_SPIRAM_RODATA=y
CONFIG_ESP32S3_DATA_CACHE_LINE_64B=y
CONFIG_COMPILER_OPTIMIZATION_PERF=y
#The following LVGL configuration items are helpful for frame rate improvement (LVGL v8.3):
#define LV_MEM_CUSTOM 1 or CONFIG_LV_MEM_CUSTOM=y
#define LV_MEMCPY_MEMSET_STD 1 or CONFIG_LV_MEMCPY_MEMSET_STD=y
#define LV_ATTRIBUTE_FAST_MEM IRAM_ATTR or CONFIG_LV_ATTRIBUTE_FAST_MEM=y

I can see from the config parameters that you are supposed to have octal SPIRAM

CONFIG_SPIRAM_MODE_OCT=y

and quad flash

CONFIG_ESPTOOLPY_FLASHMODE_QIO=y

[rasjidw]

Just to fill in some details, working with the commit 0999246 ("Fixes RGB rotation..", 2025-06-12).

Running a build with Octal SPIRAM:
$ python3 make.py esp32 clean --flash-size=16 BOARD=ESP32_GENERIC_S3 BOARD_VARIANT=SPIRAM_OCT DISPLAY=rgb_display INDEV=gt911 IO_EXPANDER=ch422g
gives

mpremote a0
>>> import gc, esp
>>> mem = (gc.mem_alloc() + gc.mem_free()) / (1024*1024)
>>> print(f'{mem} mb memory')
6.436035 mb memory
>>> print(f'{esp.flash_size()/(1024*1024)} flash')
16.0 flash

Running a build without Octal SPIRAM:
$ python3 make.py esp32 clean --flash-size=16 BOARD=ESP32_GENERIC_S3 DISPLAY=rgb_display INDEV=gt911 IO_EXPANDER=ch422g
gives

mpremote a0
>>> import gc, esp
>>> mem = (gc.mem_alloc() + gc.mem_free()) / (1024*1024)
>>> print(f'{mem} mb memory')
0.1938477 mb memory
>>> print(f'{esp.flash_size()/(1024*1024)} flash')
16.0 flash

which is not really enough memory to do anything.
So I really do want the Octal SPIRAM.

Will do some more testing over the next few days to try and work out the exact line that causes my board conniptions.

[kdschlosser]

OK so you do in fact have the 8mb of octal SPIRAM. what is eating up 2mb of the memory is the ESP32 loading the entire firmware into RAM on startup. This is done because of an issue with some RGB displays. If you add CONFIG_SPIRAM_XIP_FROM_PSRAM=n that is going to free up about 2mb of RAM. Your display should work with the setting that way.

What I am wondering is why you are not able to compile and run the code in the master branch. That is what is not making any sense to me. I don't see any changes that would cause you to not be able to at least connect to the ESP32 and get a REPL prompt.

The differences between the display you are using and the previous hardware revision is the removal of the second USB port. The display you are using only has USB-CDC available and no UART. The RS485 was moved over to the UART0 RX and TX pins and they removed the external IO to select if you wanted to use the USB-CDC or RS485.

LVGL was upgraded but that should not stop the REPL from loading.

Have you used the TOML file method of compiling? I added the config file for your display which has all of the proper settings in it for your MCU.

The other thing is and this is important because of build caching if is you change commits you need to do a fresh clone and not reuse the clone you already have. It's has to do with how the submodules are handled and the build files that already exist. You cannot simply checkout a different branch of the REPO because it is not going to remove the submodule files and folder and do a pull on those again if the files already exist. If the versions of the submodules do not align properly the build is going to end up being off.

I would also like you to use the TOML build because it is a static thing meaning the compiler settings are not going to change between builds. It being static like that allows me to know that all of the config options are being set like they should be.

[kdschlosser]

One other thing is have you tried compiling and flashing vanilla MicroPython? I am curious if that works without any issues. Don't use one of their precompiled binaries. I would like for you to clone MicroPython and checkout version 1.25.0 and compile that and see if it works. If you are able to connect to the REPL with Vanilla MicroPython then I know it is something in my code somewhere and I need to track down where the problem is. But if it doesn't work then I know the issue is upstream somewhere.

[rasjidw]

I've used the vanilla precompiled binaries, but will try building and flashing 1.25.0 from source and let you know how it goes.

[rasjidw]

And yes, I've been doing a rm -rf to clear out everything before each checkout.

Also, when trying to use the .toml file on the current HEAD, I get:

$ python3 make.py --toml=display_configs/Waveshare-ESP32-S3-Touch-LCD-4.3B.toml
Traceback (most recent call last):
  File "/home/lvgl/lvgl_micropython/builder/toml_reader.py", line 493, in run
    t_data = str(toml_obj)
             ^^^^^^^^^^^^^
  File "/home/lvgl/lvgl_micropython/builder/toml_reader.py", line 187, in __str__
    output.append(str(child))
                  ^^^^^^^^^^
  File "/home/lvgl/lvgl_micropython/builder/toml_reader.py", line 202, in __str__
    output = [str(child) for child in self.__children]
              ^^^^^^^^^^
  File "/home/lvgl/lvgl_micropython/builder/toml_reader.py", line 202, in __str__
    output = [str(child) for child in self.__children]
              ^^^^^^^^^^
  File "/home/lvgl/lvgl_micropython/builder/toml_reader.py", line 277, in __str__
    raise RuntimeError
RuntimeError

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/lvgl/lvgl_micropython/make.py", line 62, in <module>
    build_command = toml_reader.run(toml_path, driver_out_path)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/lvgl/lvgl_micropython/builder/toml_reader.py", line 525, in run
    raise SyntaxError(f'Unable to parse .toml file ({toml_path})') from err
SyntaxError: Unable to parse .toml file (display_configs/Waveshare-ESP32-S3-Touch-LCD-4.3B.toml)

[rasjidw]

I've used the vanilla precompiled binaries, but will try building and flashing 1.25.0 from source and let you know how it goes.

Okay. Built vanilla 1.25.0 from sources ($ make BOARD=ESP32_GENERIC_S3 BOARD_VARIANT=SPIRAM_OCT), and all working as expected.