I've had success getting the nRF24L01 to listen once per second for packets at an average current drain for just the nRF24L01 of 5.35ua. Not bad! The snag I'm running into though is that the listen window, encompassing power-up of the nRF24L01 through Rx mode is 1.6ms. However, if I use the atmega328p to time that 1.6ms, it raises the average current drain to 45ua, which is worse than my keyfinder keyfob's current drain (about 30ua on average). So....I need to find some way to sleep the atmega328p for 1.6ms in order to keep the total current drain low. However, how to do that? The TPL5111 has a minimum period of 100ms, so that's out. The atmega328p's WDT has a minimum period of 16ms, so that's out too. Unfortunately, the atmega328p has no built-in RTC. Even if I were to add an external RTC, I don't know of any that would let me dial-in 1.6ms.
Presently I can think of only two solutions:
Add some kind of external oscillator and feed it to Timer2 in asymchronous timer mode while sleeping the atmega328p. However, I can't seem to find from the datasheet what the current drain of that configuration would be, so it seems like a shot in the dark.
or, possibly use a GPIO pin to charge up an external capacitor and somehow calibrate the charging time so that when the capacitor discharges via some RC time constant it triggers a falling edge interrupt on the atmega328p to wake it up from sleeping after 1.6ms (the sleep time begins after it charges the capacitor). I haven't tried this, but it might actually work. It would notionally be kinda/sorta similar to how one can use an atmega328p to charge up a capacitor in order to determine the capacitor's capacitance value: https://www.norwegiancreations.com/2019/08/using-a-simple-arduino-to-measure-capacitor-value/.
If anyone has any other ideas on how to sleep an atmega328p for 1.6ms, please do post.
[Edit: Regarding #1, I suppose I could try measuring the sleep current with external asynchronous mode engaged but nothing connected to it. If it's acceptably low, I could then hook up an ultra low current 32.768K external crystal at an additional 500na or so. Hopefully there's some way tot turn everything off on the atmega328p except for the asynchronous counter. I'm doubtful as to whether I can apply such fine grained control like that to one of the sleep states (POWER_SAVE MODE), but maybe....? Here's some reason to be hopeful:
Power-save: is similar to the power-down mode, but the Timer2 is still awake and could be operated via an external clock.
Ugh. Except Timer2 is an 8 bit counter...., but that's OK: 1.6ms would be 53 ticks on a 32.768k oscillator, so at least that part should work if the mcu sets timer2 to overflow in 53 ticks and then go to sleep and if that overflow will trigger an interrupt to subsequently wake the mcu back up after 1.6ms.
Well, it's a plausible theory. Just need to test it!
Is it worth the effort though? Even if it worked, an nRF52 module would outperform this type of setup, would be more compact, would be easier/faster to assemble, and would (probably) be lower cost as well. Hmmmmm..... And the nRF53 will be even better hardware still, though the nRF53 does come with the extra overhead of an RTOS....]