Code size report:

   bare-arm:    +0 +0.000% 
minimal x86:    +0 +0.000% 
   unix x64:    +0 +0.000% standard
      stm32:    +0 +0.000% PYBV10
     mimxrt:    +0 +0.000% TEENSY40
        rp2:    +0 +0.000% RPI_PICO
       samd:    +0 +0.000% ADAFRUIT_ITSYBITSY_M4_EXPRESS

OK, apparently this doesn’t work on the nRF9160. Will investigate.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.39%. Comparing base (d694ac6) to head (9705dd4).

❗ Current head 9705dd4 differs from pull request most recent head 821da37. Consider uploading reports for the commit 821da37 to get more accurate results

@@            Coverage Diff             @@
##           master   #13367      +/-   ##
==========================================
- Coverage   98.39%   98.39%   -0.01%     
==========================================
  Files         161      161              
  Lines       21200    21078     -122     
==========================================
- Hits        20860    20739     -121     
+ Misses        340      339       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Builds on all boards now. No idea if it works on the nRF9160, as I don’t have one to test, but apart from the different register its System OFF mode looks similar.

Pins are configured as wake sources using an additional argument sense to the machine.Pin constructor.

When implementing lightsleep/deepsleep for the SAMD port, I went for taking all pins as wake-up source which had IRQ enabled. That avoids another Pin argument. Besides that, an additional argument may better be called wakeup instead of sense and matching state names, no matter how nrf calls it internally.

Rebased on master to deal with the STATIC removal, no other changes required.

Thanks for the contribution.

Is it possible to split up the second commit into two commits:

• one that implements deepsleep
• one that adds the sense argument to Pin constructors

The first part there can be merged easily. The change to Pin needs more thought with respect to the API, because other ports would need to follow it.

Splitting seems like a good way forward, done. Not thoroughly tested yet, will do that in a few days.

I agree that how to configure pins for waking needs more thought and discussion.

Can the deepsleep commit be merged?
It would make for a good improvement on the NRF52x chips - as they are most relevant for low-power stuff (especially BLE).

So, I've been looking at putting my NRF52840 Xiao board into deep sleep. @cwalther's implementation does indeed put the nrf52 to sleep, but it still consumes a lot of current. After looking for a reason on Arduino forums, I found that the Xiao board has a qspi flash memory chip on it, and it needs to be put to sleep. Arduino people have had success using Adafruit SPIflash library that sends command 0xB9 to the qspi chip, then de-initing the qspi interface. In Nordic's documentation, they also show a way to do it with the registers. I don't know if many nrf boards use qspi flash memory, but I feel like for the deepsleep() function to be consistant with other ports, the qspi should be put to sleep also.

I'd propose a change myself, but I'm not well versed in C, and I haven't been able to put the flash chip to sleep only using micropython. I wonder if something in the drivers keeps it from entering sleep mode, since it would prevent micropython from working properly.

Good point. There does seem to be precedent for board-specific deep-sleep handling dealing with SPI flash in some STM32 boards using the MICROPY_BOARD_ENTER_STANDBY macro. So I agree that this could be added here too. However, I would consider that an additional change that can go in a separate pull request (based on this one), and I currently have no plans to work on this in the near future.