I have to make a confession: The µ in µPower is silent... But let me explain my dank joke...

My goal was to get around one year aka 365 days of continuous operation out of 3 AA batteries (not rechargeables). To archive that I read lots of great articles all around the net.

The main culprit seemed to be the power-hungry SD card. Though the SDfat library doesn't officially support switching the cards power supply on and off, it seemed to work for others to reinitialize after power switching. So I decided to use a high-side switch for that purpose (to prevent a floating ground situation).

Thing is: switching it on/off does seem to work sometimes but not always. I mean, switching itself works, the card is working fine after a new SD.begin() call.

But, power consumption is higher in sleep mode than keeping it always on and let the card decide when to sleep! I have to look into that behavior.

Another thing I overlooked when ordering the PCBs is the DS3231. It is supplied directly from the regulator and that's exactly how it isn't done when saving every microamp is what you want (datasheet states up to 0.3mA quiescent current).

Actual power consumption

In sleep mode (power down with BOD and ADC disabled) the whole thing sips 0.350 to 0.380mA from the batteries. This includes the Atmega, 3x DS18B20, 1x Si7021, the RTC and the high-side switch.

That's not too shabby, but I guess I could cut that by at least another 0.1mA when powering the RTC from an Atmega pin.

Battery capacity

The batteries are Ansmann industrial alkaline AAs. They're 1.65V open-circuit and rated for "2700 mAh discharge at 10mA load; 24hours/day; End Voltage (EV): 0.8V"
We need to take into account that we don't have that constant load characteristics here. The MCP1700 does need at least 3.6V @200mA load (SD card open/write can get this high) for proper operation. This results in: 1.2V/cell maximum discharge voltage. I'm wasting capacity here, so I have to switch to 4 cells. That way I can discharge them to 0.9V/cell.

Another problem could be unevenly discharged cells as I do not use any balancing. With thee cells the difference between cells was something like 0.1V max, so I guess that's fine.

Actual operating time

Time will tell... I guess it will be somewhere around 120 to 180 days. That's not a year, I know but I'm ok with that.

The unit does voltage measuring before every SD write to prevent data loss and shuts down if the voltage is too low.

to be continued...