Board definition for Pico 2 clones with GD25Q127C 16MB flash (fixes "no USB / never boots")

[MobileFF]

Some third-party RP2350 boards copy the Pico 2 pinout but use a 16MB GigaDevice GD25Q127C flash (marking 25q127cs16). On these boards the official RPI_PICO2 UF2 never boots: the BOOTSEL drive goes away, but no USB CDC port ever appears. CircuitPython runs fine on the same board.

Cause: RPI_PICO2 runs the flash at 75MHz (PICO_FLASH_SPI_CLKDIV 2). Both boot2 and rp2_flash_set_timing() use that value, and this chip is not reliable at that speed. With 37.5MHz (PICO_FLASH_SPI_CLKDIV 4) it boots. I tried three different boot2 variants at 75MHz (w25q080, a GD25Q-specific one, generic_03h) and all failed, so the clock is what matters.

I've published an out-of-tree board definition (build with BOARD_DIR=, no changes to the MicroPython tree):
https://github.com/MobileFF/micropython-pico2-gd25q127c

• flash clock 37.5MHz, full 16MB (14MB filesystem)
• boot2 based on boot2_w25q080.S, with QE set via 0x31 (Write Status Register-2), as in the GD25Q127C datasheet
• prebuilt UF2 (v1.29.0-preview) in the Releases

Verified on real hardware: boots, USB CDC works, 14MB filesystem, and flash reads at 4/8/12MB do not wrap.
Not yet tested: the RISC-V variant, and faster clocks such as 50MHz.

Since I only have one of these boards, any reports (works / doesn't work, with the flash marking) are welcome. It might also be worth discussing whether the rp2 port should make it easier to override the flash clock for boards like this.

[mac59]

This issue was driving me crazy. Thank you. We have a "Cytron MOTION 2350 Pro REV 2.0" https://www.cytron.io/amp-p-motion-2350-pro board experiancing these symptoms. Works perfectly with circuitpython, but fails to boot after flashing to micropython and power cycling. Would you mind posting the firmware "uf2 file" with the changes. Well be happy to test it and report back.

[MobileFF]

Thanks for testing so thoroughly, this is really useful. Your results point to a different cause than on my board:

• It boots right after flashing and after a software reset, but not after a power cycle or the reset button. After flashing, the crystal oscillator is already running (the bootloader uses it for USB), while a power cycle or reset button press makes it start from scratch. That pattern matches a crystal that takes longer to start than the SDK waits for. CircuitPython's config for exactly this board sets PICO_XOSC_STARTUP_DELAY_MULTIPLIER 64 ("Allow extra time for xosc to start"); MicroPython uses the SDK default of 6. cytron-test1 did not change that (it only changed the flash clock), which fits it failing the same way. Your latest observation (it boots once every 5-10 power cycles, at random) also fits: a startup time right at the edge of what the SDK waits for.
• The flash is fine. CircuitPython lists this board's flash as a W25Q16JV, a 2MB Winbond chip that is happy at 75MHz. The 127MB drive in BOOTSEL mode is just the bootloader's virtual disk, not the flash size.
• Sorry about the v0.1.0 UF2: on a 2MB chip its 16MB filesystem layout starts exactly where it wraps onto the firmware, so it most likely overwrote itself on first boot. Your nuke + reflash fixed that.

Here is a build with only that one change (official Cytron definition, crystal startup delay x64, flash clock back at the normal 75MHz):
https://github.com/MobileFF/micropython-pico2-gd25q127c/releases/tag/cytron-test2

Could you check whether it now comes back every time after a power cycle and after the reset button? And if you can, run the same snippet with 2MB added: for mb in (1, 2, 4): .... If 2MB shows SAME, the chip is 2MB.

If this works, I think two things are worth reporting for the official CYTRON_MOTION_2350_PRO definition: the crystal startup delay, and the flash size (it assumes 4MB, so on a 2MB chip the 1MB filesystem currently sits on an aliased region and only works because the firmware is smaller than 1MB).

[mac59]

success!! "cytron-test2" works.
"PICO_XOSC_STARTUP_DELAY_MULTIPLIER 64" did the trick :)
now the board runs and power cycles successfully every time.

and it appears 2MB is the storage amount.

import machine
BASE = 0x10000000
same = lambda a, b: all(machine.mem32[a+i] == machine.mem32[b+i] for i in range(0, 256, 4))
for mb in (1, 2, 4): print(mb, "MB:", "SAME (wrap)" if same(BASE, BASE + mb10241024) else "DIFFERENT")
...
1 MB: DIFFERENT
2 MB: SAME (wrap)
4 MB: SAME (wrap)

[MobileFF]

Great news, and thank you for the careful testing! That confirms it: on the MOTION 2350 Pro the cause is the crystal startup delay, not the flash. Your wrap test also confirms the flash is 2MB.

I've recorded the result in the repository README (under "Other boards"). You can keep using cytron-test2 as it is. Its filesystem layout assumes 4MB like the official definition, but that is harmless as long as the firmware stays under 1MB, which it is (about 330KB).

Thanks again, your reports made this much easier to pin down.

[mac59]

Thank You for looking into this!! Great work :)
Now we can get on with our project!!
will this information find it's way into a MOTION 2350 Pro port build eventually?

[MobileFF]

Glad it's working for your project! For it to reach the official builds, MicroPython's own CYTRON_MOTION_2350_PRO board definition would need to change (the crystal startup delay, and the flash size of 2MB). The two settings and a working board definition are in the repository under experimental/CYTRON_MOTION_2350_PRO_SLOWXOSC, so anyone can propose it to MicroPython, and an issue there from a board owner, or from Cytron, would also help. Until then, cytron-test2 is fine to keep using.