ADC initialize() methods

[rleh]

There are currently four different ADC platform drivers for STM32 in modm.
I would like them to use the same interface to make porting easier.

// STM32F103, STM32F373, STM32F429, STM32F746
template<class SystemClock, frequency_t frequency = MHz(10), percent_t tolerance = pct(10)>
static void
initialize();

// STM32F072, STM32G071
template<class SystemClock, ClockMode mode, frequency_t frequency = MHz(1), percent_t tolerance = pct(10)>
static void
initialize();

// STM32F303
static void
initialize(const ClockMode clk=ClockMode::DoNotChange, const Prescaler  pre=Prescaler::Disabled, const CalibrationMode  cal=CalibrationMode::DoNotCalibrate, const bool blocking=true);

// STM32L433, STM32G474
static void
initialize(const ClockMode clk=ClockMode::DoNotChange, const ClockSource  clk_src=ClockSource::SystemClock, const Prescaler  pre=Prescaler::Disabled, const CalibrationMode  cal=CalibrationMode::DoNotCalibrate, const bool blocking=true);

What do you think about moving the clock mux configuration (which is usually shared between several instances) into the RCC or SystemClock code?

I would also suggest to remove the calibration from initialize() and let the user do it manually.

[salkinium]

Agreed, the initialize methods are very confusing especially with the shared clock(s) among the ADCs. But perhaps a AdcCommon{12|34} class would make more sense? There are more features than just the clock shared among the ADC.

Thinking a little further, the problem of choosing the right ADC (instance, channel) for a pin reminds me of the DMA resource problem, where the specific optimal ADC (instance, channel) choice depends on the whole system configuration. In addition there is also the whole (semi-)automatic chained(?) conversion thing going on here, with a lot of scheduling options (interlacing, injection etc) plus DMA, which is in its complexity definitely beyond a "simple" C++ API. So the AdcCommon and Adc{instance} classes should just provide a basic HAL abstraction to simplify complex ADC configurations, right?

[rleh]

I see the ADC api im modm currently being useful for most simple use-cases. Read one (or multiple) ADC channels using polling or interrupts.
Complex scenarios, e.g. master/slave mode and/or triggering from other peripherals, will always require a deep dive into the ST reference manual (and then writing to registers directly is ok for me).

I like the Idea of adding a AdcCommon12/345 class 👍
This class then configures the ADC common registers and RCC registers, not?

AdcCommon345::initialize<SystemClock, 1_MHz, ...>();
Adc4::initialize<...>(...);
Adc4::read(channel4, ...);

[salkinium]

Complex scenarios [...] will always require a deep dive into the ST reference manual

Agreed.

This class then configures the ADC common registers and RCC registers, not?

Yes, exactly! And also deals with the shared interrupts that are currently also wrongly part of the per-instance API.