How to install MicroPython to different flash size? (ESP_ERR_FLASH_NOT_INITIALISED)

[dusan-zervan]

Hello, I am using MicroPython on many different boards, but now I have problem with the new one.. It is LilyGO T-QT Pro with ESP32-S3 and 4 MiB flash size. It has also 2 MiB of SPIRAM, so I have used Generic ESP32-S3 (SPIRAM) image. It is working, but I can't save any file to device. It says an error after boot: OSError: (-24579, 'ESP_ERR_FLASH_NOT_INITIALISED')

It seems like the image is designed for 8 MiB flash size and when I have flashed it, device reports back incorrect size - esp.flash_size() reports 8388608 bytes. I am using esptool to flash, so I tried to use parameter --flash_size 4MB - no success, it is not accepted, because of this warning:

Warning: Image file at 0x0 is protected with a hash checksum, so not changing the flash size setting.

Is there a way to solve that without compilation? (I have never compiled such project.)

[glenn20]

I believe you will need to recompile (unless someone else corrects me - or there are some precompiled 4MiB S3 images around):

You can set the flashsize in ports/esp32/boards/GENERIC_S3/sdkconfig.board and select the default partition table for a 4MB flash:

...
CONFIG_ESPTOOLPY_FLASHSIZE_4MB=y
CONFIG_ESPTOOLPY_FLASHSIZE_8MB=
CONFIG_ESPTOOLPY_FLASHSIZE_16MB=
CONFIG_PARTITION_TABLE_CUSTOM=y
CONFIG_PARTITION_TABLE_CUSTOM_FILENAME="partitions.csv"

Compiling ain't so hard with docker (thanks to @mattytrentini):

git clone [email protected]:micropython/micropython.git
cd micropython
# Edit ports/esp32/boards/GENERIC_S3/sdkconfig.board as described above
docker run -ti --rm -v $(pwd):/micropython -w /micropython espressif/idf:release-v4.4 bash -c "make -C mpy-cross && make -C ports/esp32 submodules all BOARD=GENERIC_S3 -j4"

[mattytrentini]

(Apologies @glenn20, didn't mean to basically duplicate your response; it just hadn't appeared by the time I finally wrote my response!)

[unixb0y]

Worked for me, but not with partitions.csv, instead I used https://github.com/raibisch/ESP32-S3-WROOM-4MB/blob/master/4mb_2ota_partition.csv .

To be exact, I'm using the ESP32-S3-WROOM-1-N4 module.

[edcline]

I am having the same issue, just trying to do a basic prototype board bring-up with S3-WROOM-1U-N4, don't need a production version of mpy. Would you be able to post your binary for a quick test?

[unixb0y]

@edcline yes, I uploaded it here: https://github.com/unixb0y/Micropython-ESP32-S3-WROOM-1-N4

[edcline]

Excellent, thank you...

[mattytrentini]

You're right; I had the same issue.

It's only a straightforward change but it does necessitate compilation. In GENERIC_S3_SPIRAM/sdkconfig.board you need to make the following changes:

CONFIG_FLASHMODE_QIO=y
CONFIG_ESPTOOLPY_FLASHFREQ_80M=y
CONFIG_ESPTOOLPY_FLASHSIZE_DETECT=y
CONFIG_ESPTOOLPY_AFTER_NORESET=y

CONFIG_SPIRAM_MEMTEST=

CONFIG_ESPTOOLPY_FLASHSIZE_4MB=y # This was unset
CONFIG_ESPTOOLPY_FLASHSIZE_8MB= # This was set
CONFIG_ESPTOOLPY_FLASHSIZE_16MB=
CONFIG_PARTITION_TABLE_CUSTOM=y
CONFIG_PARTITION_TABLE_CUSTOM_FILENAME="partitions.csv" # This pointed to 'partitions-8MiB.csv'

I think we should follow-up with two options: a) Add a board definition for the T-QT Pro with this configuration. b) Add variants to allow multiple builds of the GENERIC_S3_SPIRAM for the common flash sizes...

[edcline]

Was this ever completed....is there a build available for the 4MB flash modules?

[FriendlyNGeeks]

Was this ever completed....is there a build available for the 4MB flash modules?

What are you trying to flash, brand; model?

[edcline]

I've got an ESP32-S3-WROOM-1U-N4R2

[dusan-zervan]

Excellent, thank you both! I had to use git clone https://github.com/micropython/micropython.git instead of git clone [email protected]:micropython/micropython.git - I don't know why. But it is working.

However, I have some questions (maybe not very clever, I have never used that):

• I've got 3 files: bootloader.bin to 0x0, partition-table.bin to 0x8000 and micropython.bin to 0x10000. Is there a way to modify some git parameters to combine those files to just one file?
• I've got a "nightly build" 724. Is there a way to get official / stable build?
• Why is it needed to use IDF 4.4? I don't even understand what is it (never mind), but I've tried idf:latest and it failed to compile.
• Perhaps I could also integrate some modules that I need, maybe even my own boot.py file directly to image. Could you direct me to some instructions / tutorial how to do that?

[dusan-zervan]

For my last question I have found information:

• https://docs.micropython.org/en/latest/reference/manifest.html
• https://github.com/micropython/micropython/wiki/Building-Micropython-Binaries

It is good starting point, I will study that later.

[glenn20]

Excellent, thank you both! I had to use git clone https://github.com/micropython/micropython.git instead of git clone [email protected]:micropython/micropython.git - I don't know why. But it is working.

Whoops, my version is for ssh auth to GitHub, which is what I use, but only useful if you have setup SSH access.

However, I have some questions (maybe not very clever, I have never used that):

• I've got 3 files: bootloader.bin to 0x0, partition-table.bin to 0x8000 and micropython.bin to 0x10000. Is there a way to modify some git parameters to combine those files to just one file?
• I've got a "nightly build" 724. Is there a way to get official / stable build?

Yes, from within your micropython folder execute git checkout v1.19 to get the latest micropython release tagged version. Then run the build docker command again. You should probably run the "make clean" docker command first.

• Why is it needed to use IDF 4.4? I don't even understand what is it (never mind), but I've tried idf:latest and it failed to compile.

This is the idf version that is used to build the production esp32s2/3 images. The 5.x idf is only recently released.

The IDF is the integrated development framework provided by espressif, the manufacturer of the esp32 devices. The Micropython esp32 port is built on top of that.

• Perhaps I could also integrate some modules that I need, maybe even my own boot.py file directly to image. Could you direct me to some instructions / tutorial how to do that?

Looks like you found the answer to that one with the manifest file.

[steeley]

I have an ESP32-s3 4R8 and tried everything here and while I can use the flashed firmware I have 0 storage so can't save anything.
Anyone got this config working?
module is this
[(https://www.digikey.co.uk/en/products/detail/espressif-systems/ESP32-S3-WROOM-1U-N4R8/16162634](url)

The module works perfectly in Arduino so my hardware setup is just fine.

[bixb922]

I don't have that specific module, but this utility program; https://github.com/glenn20/mp-image-tool-esp32 allows you to set the flash size without needing to recompile Micropython. It worked well for me for an ESP32-S3 N8R8 and N16R8 with SPIRAM starting with the standard image for octal SPIRAM from https://micropython.org/download/ESP32_GENERIC_S3/

If not, you will have to install Esp-idf and recompile Micropython, see https://github.com/micropython/micropython/blob/master/ports/esp32/README.md

[bixb922]

Are you sure you downloaded the firmware from the section of the download page marked as "Firmware (Support for Octal-SPIRAM)"? The name of the Micropython version in the log seems to indicate that this is the version without SPIRAM support?

[steeley]

Ahh yes - accidentally converted the wrong image. the SPIRAM version seems ok now, but I still get the SHA checksum error - not sure if thats a problem?

[bixb922]

The SHA comparison error could means that the Micropython image arrived corrupted at the ESP32-S3.

Could you please post the command you are using to flash? Try using esptool with -z, the compression adds also a certain layer of transmission checking

Have you tried erase flash before writing?

Do you get the REPL prompt?

Are you using the serial port or the USB native port of the ESP32S3? With the native USB port I get less transmission errors than with the serial port

[robert-hh]

@steeley I never have seen 'real' SHA1 errors after uploading a ESP32 image. So I suppose that the tool you use to adapt for a 4MB flash did not recreate the SHA1 checksum of the firmware file. Then the SHA1 checksum of the file does not match.

[steeley]

thanks for the suggestions - REPL did work OK. I'm using the serial via a known good usb-UART serial adapter cable.
I always erase flash before writing.
Yes- could use the built-in USB on the S3 -in fact I'm going to add a connector for this.

I'll try esptool directly with -z. I was using Thonny to upload which uses esptool but without the -z option I think.

[a-sapuppo]

we use ESP32-S3-WROOM-1U-N16R8 for micropython