Lowpower sleep rfm95 + SAM D21

[MortenGuldager]

I have made a minimalistic board with a Microchip SAM D21 and a RFM95 module.

If I deep sleep the processor before doing radio.begin() I measure a current in the 16uA range, but if instead deep sleep after radio.begin() and a following radio.sleep() I measure ~70uA.

I guess it's some pins left in a less optimal state. Question is if there is a "correct" way to handle this?
If I just change pin-mode at will to get back to 16uA I will probably wreck something in radiolib.

[MortenGuldager]

Well, without really knowing what I were doing I experimented a bit and when I threw in

    digitalWrite(SPI_CS_PIN, LOW);

right after radio.sleep(); it am now seeing 10.5uA when deep sleeping.

Since nothing else is on the SPI bus I guess it is safe. After sleep mcu again takes control over CS and everything appears to be working just fine.

[jgromes]

Generally there's no controlled way to do this from the library side, since from there we don't know what other devices may be on the bus or on the other pins. Changing the CS level for RFM9x (aka SX127x) should be safe-ish, though for other modules it could behave differently. SX126x for example uses CS being pulled low as a signal to wake up. So careful testing is needed, no generic answers here unfortunately.