RP2350 boot mode not working with Mac Sonoma and non-USB power.

[JimMacA]

Kind of a specific problem. When my custom RP2350 board powers up in boot mode, it reliably appears in Windows and Mac Ventura, but not Sonoma. It's a battery-powered box, so when I rewire it to only have USB power, Sonoma sees it. This implies that there's a tight window during boot where the code needs to see USB activity, and that window can best (only?) be met if booting is synchronized to the USB connection. Can anyone point me to the RP2350 boot docs which discuss the timing, not just the sequence of events? Thx.

[dhalbert]

macOS 14.x (Sonoma) had a problem writing very slowly to small FAT drives. Some of the FAT filesystem support was moved to userspace. Before 14.4 this problem was quite serious, and caused I/O errors. In 14.4 and after the I/O errors went away, but writing to drives smaller than 2GB was about 40 times slower than expected. The problem was fixed in 15.2.

See https://blog.adafruit.com/2023/12/05/macos-sonoma-problems-writing-to-circuitpy/ for some background on this.

It sounds like your problem is somewhat different, but it may be related to the macOS FAT filesystem code. I would say that for all of Sonoma, there were serious problems with the OS code related to this, and I'm not surprised that you have seen problems. Do you see these problems in more recent versions of macOS?

There was a worse problem in early versions of Ventura:
https://blog.adafruit.com/2022/10/31/uploading-uf2-files-with-macos-13-0-ventura-apple-microbit_edu-raspberry_pi-circuitpython/

[JimMacA]

Thanks for that. Unfortunately, this hardware is going out to clients who own a variety of Macbooks, and aren't very computer literate, so it will be on me to come up with a robust solution instead of asking them to upgrade their OS. So maybe a switch which disables the battery in boot mode. Ick. I don't have Vbus connected to any GPIO because from what I can make out of the docs, the boot code doesn't monitor the USB Vbus. Can anyone confirm/deny this?

…

On Wed, Jun 11, 2025 at 11:04 PM Dan Halbert ***@***.***> wrote: macOS 14.x (Sonoma) had a problem writing very slowly to small FAT drives. Some of the FAT filesystem support was moved to userspace. Before 14.4 this problem was quite serious, and caused I/O errors. In 14.4 and after the I/O errors went away, but writing to drives smaller than 2GB was about 40 times slower than expected. The problem was fixed in 15.2. See https://blog.adafruit.com/2023/12/05/macos-sonoma-problems-writing-to-circuitpy/ for some background on this. It sounds like your problem is somewhat different, but it may be related to the macOS FAT filesystem code. I would say that for all of Sonoma, there were serious problems with the OS code related to this, and I'm not surprised that you have seen problems. Do you see these problems in more recent versions of macOS? There was a worse problem in early versions of Ventura: https://blog.adafruit.com/2022/10/31/uploading-uf2-files-with-macos-13-0-ventura-apple-microbit_edu-raspberry_pi-circuitpython/ — Reply to this email directly, view it on GitHub <#2990 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BBUCFDB3EO3SIV3SCLTLWHL3DDU5JAVCNFSM6AAAAAB7EEDCM2VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGNBUGA2TKMI> . You are receiving this because you authored the thread.Message ID: <earlephilhower/arduino-pico/repo-discussions/2990/comments/13440551@ github.com>

[dhalbert]

If this is indeed happening in the ROM bootloader, you may want to ask about / search for this in https://forums.raspberrypi.com or https://github.com/raspberrypi/pico-feedback/issues.

I'm not sure what you mean about battery vs USB power. on your board Often a couple of diodes are used to select the higher of the two voltages (battery or USB) to supply to an on-board regulator, and prevent backpowering one or the other.

[JimMacA]

Thanks, I will post there and let you know if anything comes of this.
Just as a dump of where in the rat-hole I am, the original power path looks like this:

USB -> battery charger IC ->mech switch -> vreg -> RP2350
|
battery

This works consistently in boot mode and the unmodified Adafruit TinyUSB msc_ramdisk example, when attached to a Windows computer, and a MacBook running Ventura, but not Sonoma.
Adding a diode from USB power to the vreg input (to the right of the mech switch) means that the RP2350 will always have power when USB is attached. It also violates the max capacitance USB spec for my design. This setup sometimes works with Sonoma. There seems to be a magic amount of capacitance to optimize the chance of it working, implying that this is a startup race problem, but I don't have enough data to be sure.
Another design of mine with identical hardware -- except that it uses an external power supply and an ADum3160 USB isolator -- always works, regardless of power sequencing and USB connection timing.
USB bus analysis with a Beagle shows that when the connection fails, enumeration doesn't even start.
If I had to guess, I'd say there's something going on with the timing of clearing MAIN_CTRL.PHY_ISO, but tinkering with that isn't a solution for me because this thing needs to work in boot mode, and I can't modify that code. At this point, my only hope is to find a cheap hub that works well enough that I can recommend it to my Macbook users.