Default `SupplyConfig` for stm32h family?

[sgasse]

Hi!

We have a few STM32H745XI discovery boards used throughout smaller projects. After updating from an older version of embassy to the latest version, we ran into a curious issue that is probably related to a change in the default supply config:

The issue happens when we flash a board with the latest version of embassy.

• While the board remains connected via USB, every example still works.
• Once disconnected (powered down) and reconnected, none of the examples work (even after reflashing).
• After flashing an example with an old version of embassy and reconnecting USB (power cycle), everything works again.

My colleague @KloolK debugged it a while back and ran into the same issue and solution as described in #2430. So when setting the SupplyConfig to DirectSMPS, we prevent the board from getting stuck after the next power cycle:

    let mut config = embassy_stm32::Config::default();
    {
        use embassy_stm32::rcc::*;
        config.rcc.supply_config = SupplyConfig::DirectSMPS;
    }
    let p = embassy_stm32::init(config);

Right now, if anyone uses a STM32H745XI discovery board or similar with any of the examples, they may run into the tricky situation of everything working seemingly fine one day, but after a power cycle not being able to run any example. My suggestion is to switch the default SupplyConfig to DirectSMPS at least for the feature stm32h745xi-cm7, maybe also stm32h745xi-cm4.

I'd be happy to create a PR but I wanted to make sure that there is no other reason for having the default as it is.
@badrbouslikhin could you comment since you were the last one to edit SupplyConfig?

[badrbouslikhin]

Hi @sgasse,

Just wanted to point you towards the discussion about the SupplyConfig default setting in this PR where they first introduced power supply options: #2237 (comment). There’s a notable comment there that really sums up why SupplyConfig::Default seemed like the right choice initially:

When I connected the BOOT0 pin to VDD, my STM32H746XI (MCU) started in System memory.
The internal bootloader automatically sets the PWR_CR3 register to 0x05000046 which corresponds to the default power supply configuration (SupplyConfig::Default).
This configuration ensures that the MCU will boot properly regardless of the hardware setup.
I think it's indeed the most sensible default setting.

However, what you and @wagcampbell have experienced seems to challenge this perspective.

I'm not convinced that DirectSMPS is the best default, though. According to the Reference Manual for STM32H745XI:

The SMPS step-down converter isn’t available on all packages.

It worked for your STM32H745XI discovery board, but may not be the same for other boards.

I am not sure what would be a better way to handle this besides better documentation.

[sgasse]

Ah alright, thank you for providing some perspective @badrbouslikhin 🙏 I had hoped that the power supply circuitry would be comparable enough for one class of an MCU. Then I agree with you that the only thing to improve the situation may be better documentation.

Would it be worth adding a small section "Troubleshooting" or something in the README here?

Not sure though if this is really the place where people would look first when the board stops working. The fact that it happens only after a power cycle and is not reset by a flash memory wipe with STM32CubeProgrammer gave me personally the impression that I either bricked the board or that something about my laptop setup was wrong ^^

[badrbouslikhin]

@sgasse how about in the examples code + the FAQ page on the Embassy website?