Avergage power consumption is about 1,1mA when updating once every 30 minutes where 1mA is ironically consumed by the coulomb counter. This is strange as the part should have a much lower power consumption. We left it in the system as we want to learn about what happens with a little panel in a non-optimal mounting position. You can see that charge/discharge in the second graph on https://wiekaltistderkanal.de
Another little feature is that we can change the update rate via downlink command - so we can have a high rate when testing and reduce this drastically once everything is stable (to save airtime and power).
"Backend" and visualisation is Node Red.
Data is not only exported to the site, opensensemap,dweet but a local Elasticsearch. I plan to replace the Node Red visualisation with Kibana,
Here is what we like to improve for version 2:
* The red case is great but also asking for trouble in the wild. People like the thing but there are always some idiots... So we'll use something less conspicuous
* Use steel cable ties for the same reason
* Use two temperature sensors to get a depth profile.
* Put the temperature sensors in a perforated plastic pipe that is only loosly affixed to the top. We hope that this makes this more robust against humans and ice as well
* Switch the over-the-air protocol from binary to msgpack (to make it easier to change)
* The current design has a little hole with an inward pipe attached to address condensation. I like to replace this with a membrane to make the thing really waterproof. It is currently mounted 20cm over the waterline and it gets it share of splash from ship swell. Also: the waterlevel is not as constant as we assumed.
* Perhaps carry over an ultrasonic level sensor from another project we're working on.
* replace the Autonomo with something cheaper.
* use a smaller battery
* use cheap metal case
Let me know if you want to know specifics...