Back in November 2011 when I made my first ATmega328 based wireless temperature sensor node I did a rough calculation estimating that it would last 2 years or more on a couple of AA batteries. That first node was decommissioned after a few months but the second version, also based on an ATmega328 is still running and on the original set of batteries since 29 December 2011, exactly 2 years ago today.
That’s with an ATmega328P running at 16MHz with a RFM12B radio and DS18B20 temperature sensor powered from 2 x Energizer alkaline AA batteries (essentially the same as the new SMD Tiny328 nodes but built on stripboard) and is transmitting once a minute. The battery voltage is now down to 2.71V which is well out if spec for a 328 running at 16MHz but the datasheet specs seem to be very conservative and it is still ticking along fine. Previous tests showed that it should be good down to around 2.6V with the DS18B20, below that the sensor’s readings start to become erratic. Again, that’s a good margin less than the minimum of 3V stated in the data sheet.
Two plus years is inline with my original predictions but it is good to see it actually happen in the real world, particularly as this node started out on code that wasn’t quite as optimised for low power use and since I enabled ACKs at the end of August 2012 it makes quite a few retries (it is the furthest from the base station) so in reality it often sends more than once a minute.
My first ATtiny84 based node (the original stripboard prototype for the TinyTX) with a TMP36 sensor is also still going strong after 22 months and is currently at 2.97V, this one is running at 8MHz using the internal oscillator and also running from 2 x Energizer AA batteries. The analogue TMP36 sensor is obviously better for low power compared to the DS18B20 but I found it was a nuisance to calibrate so I’ve stuck with the DS18B20 in all my other temperature only nodes as they are so much nicer to use and 2+ years battery life is plenty, in fact I never even bothered to reduce the sensor resolution I mentioned here, leaving them all at 12 bit.
Here is the graph over 2 years for the 328 based node, I think you can see the slightly steeper decline start once the ACKs were added at the end of August 2012.
Good result I think.