Board def for WeAct Studio RP2350B fails updating submodules

[scruss]

I've been trying to create a board def for the WeAct Studio RP2350B core board, mostly by modifying existing definitions and hoping for the best. This is what I've come up with so far: scruss/WEACTSTUDIO_RP2350B: attempted MicroPython config for this board.

I have a dev system set up that I can build more conventional RP2350 boards. When I attempt make BOARD=WEACTSTUDIO_RP2350B submodules, however, I get this error:

Use make V=1 or set BUILD_VERBOSE in your environment to increase build verbosity.
make -f ../../py/mkrules.mk GIT_SUBMODULES="lib/pico-sdk" submodules
make[1]: Entering directory '/home/scruss/c/micropython/ports/rp2'
Use make V=1 or set BUILD_VERBOSE in your environment to increase build verbosity.
Updating submodules: lib/pico-sdk
Synchronising submodule url for 'lib/pico-sdk'
make[1]: Leaving directory '/home/scruss/c/micropython/ports/rp2'
make[1]: Entering directory '/home/scruss/c/micropython/ports/rp2'
Use make V=1 or set BUILD_VERBOSE in your environment to increase build verbosity.
Updating submodules: 
# If you see this error, it may mean the internal step run by the port's build
# system to find git submodules has failed. Double-check dependencies are set correctly.
Internal build error: The submodule list should not be empty.
exit 1
make[1]: *** [../../py/mkrules.mk:259: submodules] Error 1
make[1]: Leaving directory '/home/scruss/c/micropython/ports/rp2'
make: *** [Makefile:77: submodules] Error 2

I've only ever seen this before if core parts of the build subsystem are missing. In this case, they're present, as I can build for other stock RP2350 boards. What am I missing?

The WeAct Studio RP2350B core board has optional PSRAM /CS on GPI00, so we can't use the default UART0 settings for that pin.

[aylen384]

Looks like a regression as it works fine when using the version with the v1.24.1 release tag

I opened an issue: #16870

[scruss]

I agree that make submodules with v1.24.1 gets further for my board def, but building from a working board def (such as BOARD=RPI_PICO2) has no problems.

There's still stuff I don't understand that's missing from my definition that causes the final make BOARD=WEACTSTUDIO_RP2350B to fail:

[ -e build-WEACTSTUDIO_RP2350B/Makefile ] || cmake -S . -B build-WEACTSTUDIO_RP2350B -DPICO_BUILD_DOCS=0 -DMICROPY_BOARD=WEACTSTUDIO_RP2350B -DMICROPY_BOARD_DIR="/home/scruss/c/micropython-1.24.1/micropython/ports/rp2/boards/WEACTSTUDIO_RP2350B"
PICO_SDK_PATH is /home/scruss/c/micropython-1.24.1/micropython/lib/pico-sdk
Target board (PICO_BOARD) is 'WEACTSTUDIO_RP2350B'.
CMake Error at /home/scruss/c/micropython-1.24.1/micropython/lib/pico-sdk/cmake/generic_board.cmake:37 (message):
  Unable to find definition of board 'WEACTSTUDIO_RP2350B' (specified by
  PICO_BOARD):

     Looked for WEACTSTUDIO_RP2350B.h in /home/scruss/c/micropython-1.24.1/micropython/ports/rp2/boards/WEACTSTUDIO_RP2350B, /home/scruss/c/micropython-1.24.1/micropython/lib/pico-sdk/cmake/../src/boards/include/boards (additional paths specified by PICO_BOARD_HEADER_DIRS)
     Looked for WEACTSTUDIO_RP2350B.cmake in /home/scruss/c/micropython-1.24.1/micropython/lib/pico-sdk/cmake/../src/boards (additional paths specified by PICO_BOARD_CMAKE_DIRS)
Call Stack (most recent call first):
  /home/scruss/c/micropython-1.24.1/micropython/lib/pico-sdk/cmake/pico_pre_load_platform.cmake:70 (include)
  /home/scruss/c/micropython-1.24.1/micropython/lib/pico-sdk/pico_sdk_init.cmake:87 (include)
  CMakeLists.txt:85 (include)


-- Configuring incomplete, errors occurred!
make: *** [Makefile:58: all] Error 1

[mattytrentini]

With only a few updates I was able to get a clean build pretty easily. It doesn't yet work on the device - I suspect I need some additional flash configuration (?) - but it builds fine!

Changes are here:

mattytrentini@3f230d6

Update: A couple of other defines for the flash and it now works! Check out mattytrentini@954084c. Note: I haven't tried building for RISC-V yet.

[scruss]

Ah, thanks!

I don't think you want the #define PICO_DEFAULT_UART 0 though, because the SPIRAM's CS line is on GPIO0

[mattytrentini]

I presume it doesn't matter for the default case; when no PSRAM (or additional flash) is installed? GPIO0 is also broken out as with all the other pins so it can be used for other purposes.

But yes, if we were to produce a build that supported PSRAM or additional flash we'd have to change those defines.

To be fair, we can probably choose better defaults so we can use the same for all variants...I'll do that, thanks for the feedback! 😄

[scruss]

And thank you for looking at this and fixing it. I really wish there was a guide to what you need to do to write a board def.
(Then again, there are many board defs stuck in the PR queue, so maybe we need fewer boards, not more)

[mattytrentini]

I have been meaning to write such a guide...and I think we need more board defs - and more people that can review them!

[mattytrentini]

OK, I've tidied the board definition up, tested with RISC-V and raised a PR: #16892

If anyone could test I'd appreciate it!

[scruss]

Thank you! I've only tried it in ARM mode, but all the storage and memory is there:

Board       : WeAct Studio RP2350B Core with RP2350
Frequency   : 150 MHz
Free Memory : 492864
File storage: 14328 / 14336 K

Pin('LED') is where it should be. Status of the KEY button works with:

ks=machine.Signal(machine.Pin('KEY', machine.Pin.IN), invert=True)

Any other in-depth tests you recommend?

[mattytrentini]

Off the top of my head, no, there's nothing obvious to suggest to test. If you have any existing software running on another rp2 board then you could try and port it across - but no matter if you don't! I think it looks pretty straightforward.

It turns out that Damien has been sneakily testing the RP2350B periodically when integrating pico-sdk changes, which is probably why there were no surprises.

Thanks for testing that out! And for providing the starting point for the board definition. 😊

[shariltumin]

In the "boards/WEACTSTUDIO_RP2350B_CORE/weactstudio_rp2350b_core.h" for UART

#ifndef PICO_DEFAULT_UART
#define PICO_DEFAULT_UART 0
#endif
#ifndef PICO_DEFAULT_UART_TX_PIN
#define PICO_DEFAULT_UART_TX_PIN 16
#endif
#ifndef PICO_DEFAULT_UART_RX_PIN
#define PICO_DEFAULT_UART_RX_PIN 17
#endif

will not set default UART0 to tx=16 and rx=17.

>>> import machine
>>> uart = machine.UART(0)
>>> uart
UART(0, baudrate=115200, bits=8, parity=None, stop=1, tx=0, rx=1, txbuf=256, rxbuf=256, timeout=0, timeout_char=1, invert=None, irq=0)

This is because "machine_uart.c"

// UART 0 default pins
#if !defined(MICROPY_HW_UART0_TX)
#define MICROPY_HW_UART0_TX (0)
#define MICROPY_HW_UART0_RX (1)
#define MICROPY_HW_UART0_CTS (2)
#define MICROPY_HW_UART0_RTS (3)
#endif

The "PICO_DEFAULT" for I2C and SPI flags are used in "machine_i2c.c" and "machine_spi.c", but "PICO_DEFAULT_UART" is not used in "machine_uart.c".

For the time being, the easiest way to set the default tx and rx for UART0 is to add the def to the "boards/WEACTSTUDIO_RP2350B_CORE/mpconfigboard.h".

// Override machine_uart.c def
#define MICROPY_HW_UART0_TX (16)
#define MICROPY_HW_UART0_RX (17)
#define MICROPY_HW_UART0_CTS (-1)
#define MICROPY_HW_UART0_RTS (-1)

After a rebuild we will have:

>>> import machine
>>> uart = machine.UART(0)
>>> uart
UART(0, baudrate=115200, bits=8, parity=None, stop=1, tx=16, rx=17, txbuf=256, rxbuf=256, timeout=0, timeout_char=1, invert=None, irq=0)
>>> 

[scruss]

Yes, I'd particularly recommend that, as GPIO0 is busy being the CS line for the SPIRAM.

(Which I haven't managed to get working yet)

[mattytrentini]

Oh, nice catch! I hadn't realised that PICO_DEFAULT_UART wasn't correctly interpreted.

I've made those suggested updates, thanks!

[scruss]

Thanks to Matt's excellent work, there is a board definition accepted in the main repo: ports/rp2/boards/WEACTSTUDIO_RP2350B_CORE and UF2 images are available on the official download page: download/WEACTSTUDIO_RP2350B_CORE

[scruss]

thanks, but that doesn't work either. I dropped an issue into your repo

[fanoush]

If you are interested, please take a look at https://github.com/shariltumin/weact-rp2350b-psram-micropython

Thank you, it seems to work with APS6404L-3SQR PSRAM I got from https://www.aliexpress.com/item/1005007652282144.html (BTW I am not the reviewer that posted photo of this board there)

MicroPython v1.14-4791.g293d17e44 on 2025-03-27; WeAct Studio RP2350B Core w/PSRAM-8MB with RP2350
Type "help()" for more information.
>>> import gc
>>> gc.mem_free()
8670320
>>> import micropython
>>> micropython.mem_info(1)
stack: 452 out of 12032
GC: total: 8677888, used: 7952, free: 8669936
 No. of 1-blocks: 125, 2-blocks: 24, max blk sz: 64, max free sz: 512276
GC memory layout; from 20009740:
00000000: h=MLhhhBDhhBTTBDhTBDBDBh===BTB=hB=BBBh=====BBTB=BTB=BBBTB=TB=BTB
00000400: =Bh===h===========Dh============BBBhSLh=hh========h=============
00000800: ====h===========================================================
00000c00: ====h===========================================================
00001000: ====hhhBhhhBLhShSShh=h=h=hhhBBhhh=hhhhLTThh=h===h==hLhST=Shhh===
00001400: hhh=========h===Sh=Sh=======h========h==hSh=ShShSh========ShShBh
00001800: T==hh=h====h=h========h==Bhhhhh===h=====hhh==============h====BD
00001c00: hh=h=hhhhBh=hhhh==hB..hhh=.......h=...h==.......................
00002000: ....................h========h==========........................
       (461 lines all free)
00075800: ............
GC memory layout; from 1102eec0:
       (8004 lines all free)
007d1000: ....................

Do you have some suggestion/code to stress test the PSRAM so it can be verified it really works fine and is stable?
I am not sure whether it needs capacitor on empty pads next to it or will be stable without one.
As the SPI flash already has it and both are not used at the same time (?) maybe the one for flash could be enough for both?

[shariltumin]

I do not know about the capacitor. It seems to work fine without it. I am not sure if the pads are for capacitor, maybe for 0Ohm resistor. I shorted the pads and it did nothing.

Some tests I hacked and maybe not what you have in mind is at https://github.com/shariltumin/weact-rp2350b-psram-micropython/tree/main/tests

[fanoush]

I shorted the pads and it did nothing.

Maybe you were lucky to not to do it properly? I just measured it and one pad is 0 ohm to VDD PSRAM pin next to it and the other is 0 ohm to VSS/GND in the opposite corner.

According to datasheet https://www.pjrc.com/store/APS6404L_3SQR.pdf#page=21 there should be 1μF capacitor. But it seems to work for me without, probably also because the flash chip on the other side possibly has same requirements and both are connected together (except CS pin) and there is only one QSPI interface so only one memory is busy at the same time.

Some tests I hacked and maybe not what you have in mind is at https://github.com/shariltumin/weact-rp2350b-psram-micropython/tree/main/tests

Thanks again, missed that folder, copy pasted both scripts an both run for quite some time and printed

RAM free: 8665168
CPU: 260000000
T for ba: 1955
T for ca: 1203
RAM free: 2672272
T for ba write: 34341
T for ba read, ca write: 20121
ca[1000]==ba[3900134]==0xcc: True

and

RAM free: 8671712
CPU: 260000000
T for ba write: 36446
T for ba read, ca write: 20444
ca[1026]==ba[3900108]==26: True
T for ba write: 36446
T for ba read, ca write: 20444
ca[1071]==ba[3900063]==71: True
T for ba write: 36446
T for ba read, ca write: 20444
ca[1218]==ba[3899916]==218: True
T for ba write: 36446
T for ba read, ca write: 20443
ca[1002]==ba[3900132]==2: True
T for ba write: 36446
T for ba read, ca write: 20444
ca[1204]==ba[3899930]==204: True
>>>

so hopefully it is OK

[fanoush]

Oh, now I see my version is reported as 1.14 (visible also in post above) I used your binary from https://github.com/shariltumin/weact-rp2350b-psram-micropython/tree/main/firmwares/arm , however your log file https://github.com/shariltumin/weact-rp2350b-psram-micropython/blob/main/tests/test_psram_log.txt shows 1.25 and you get slightly better numbers 17000 vs 20000

EDIT: this is interesting, I tried to run https://viper-ide.org/benchmark.html and it shows
"machine":"WeAct Studio RP2350B Core w/PSRAM-8MB with RP2350","release":"1.25.0-preview","sysname":"rp2","version":"MicroPython v1.14-4791.g293d17e44 on 2025-03-27" so the release says 1.25 and version says 1.14. So unfortunately lower numbers are probably not because of the version, maybe the APS604L is slower or runs at slower clock.