Cool! One question: I thought I read in the riffle documentation that you could only log in multiples of 8 seconds. That seemed liked a weird multiple to choose. You said you were logging at 10 seconds. So did I read something wrong?

Don thinks that eight seconds might be the minimum interval during which the system can be put into deep sleep. So all the sleep stuff happens in multiples of eight seconds. But it must take some time to do some housekeeping (writing data, turning out the lights, getting under the covers), so the timestamps on the data writes from this overnight deployment were mostly close to 10 seconds apart.

When you want to log data every 10 minutes, I guess the housekeeping chores will only happen once every ten minutes, so dividing 10 minutes by eight seconds will get you the proper multiplier. We will see.

I think 8 seconds is the "maximum" interval with a 16MHz crystal. You wake up every 8 seconds and can go back to sleep if there is no work to be done.

George that's a good point. Although the real time clock is running all the time, it is not smart enough to do something as complicated as setting an alarm and waking something else up. Only the Arduino cpu (Atmel 328p) can do that. But for some reason I guess the Atmel is not capable of sleeping deeply for more than eight seconds. That must be wildly inefficient if you want to log data every hour. But maybe waking up and going right back to sleep doesn't use much battery power.

That must explain why the interval between data writes is determined with a for loop:

for (int i=0;i<numIntervals;i++) {
LowPower.powerDown(SLEEP_8S, ADC_OFF, BOD_OFF);
}

That loop must tell the cpu to keep going back to sleep a certain number of times (numintervals) and only after all those eight second naps do some work when you wake up. I do pretty much the same thing some days.

Chris,

I really don't know the actual time but assuming 800usec (nice for the math) to sleep and wake up it would be 800 / 8,000,000 or 0.01% which seems fine to me. This is another of the functions we should instrument and evaluate. I suspect users would appreciate knowing the typical usable battery life with known loads such as the demo data logging sketch.

BTW, there is still some current drawn in sleep mode.

The RTC can be set to shutdown and wake the AT328P, it just uses another pin that could be used for a sensor. @donblair left an exposed pin available on the board, and wire can be soldered to it and attached to one of the header pins (not sure which on) I don't have a riffle in front of me this second but it looks like pin D5 in the riffle schematic. @donblair correct me if I'm wrong/tell folks where that pin can go. https://raw.githubusercontent.com/OpenWaterProject/riffle/master/hardware/riffle-beta-0.1.0/riffle-beta-0.1.3.2-schematic.png

D5 appears to connect the INT/SQW on the RTC to PD5 on the CPU. I would be very interested in how one uses the feature. I'm assuming that it requires configuring the DS3231 and writing code that enables the CPU to respond to the signal. I am a novice with ATMEL chips but would be willing to test any algorithms that are proposed.