This thread is more of a "comment box" to the lovely LadyAda and her minions. This also serves as a touchstone to help guide other users attempting to build dataloggers for the field.
Problem Background
I noticed independently that my devices which use the Adafruit Data Logger Shield fail at input voltages above 12V. Per the comment from @adafruit_support_bill in the thread linked at the top of this post, it may be due to the limitations of the LDO. He suggests that at high input voltages, the LDO is working too hard to dissipate excess energy as heat.
My Devices
The devices I am working with are 6 node devices powered by a solar station. The solar station is a 15W continuous power supply with 100Ah capacity across a pair of 12V batteries and a 60W solar panel. Each node is an Arduino Uno, Data Logger Shield, custom sensor package, and air pump. Each node is pulling a peak of ~0.6A when everything is powered on. This means that the devices can run all night without issue during sunny periods. The solar unit is programmed to automatically shut down below a threshold voltage on consecutively cloudy days (I think it is Vbatt = 11.5V, but my memory is hazy on the precise value). The nodes are powered through the Vin pin through parallel connection to a fat power cable. For the purposes of heat dissipation, it is worth mentioning that the network is deployed in an Oklahoma summer with peak temperature measured by the instruments at ~50C/122F.
Prototyping
I prototyped the nodes at my the desk level using the USB in and 12V benchtop supply. The devices function correctly with both power inputs. This is noteworthy because it rules out wiring problems when powered by 12V at Vin.
Observations at Deployment
When I have checked on my sensors at the field site, I measured the battery voltage and voltage supplied to each instrument. The battery voltage ranged from 12.28V-14.24V based on hand measurement when I was at the site. This is biased for sunny days (who goes and pokes at electronics in the rain?). I had assumed that the devices were functioning correctly. However, when I look at the data, I see I am incorrect.
In this figure, the red and blue points labeled as "Canyon" are the Adafruit Data Logger Shield nodes, while the "Roof" points are other equipment. Notice that the only times the nodes are active is during the two spurts near periods of low temperature. When I correlate these times to local weather data, it becomes apparent that these dates closely follow cold fronts moving through the region.
Why does this matter? Cold fronts bring clouds. Clouds reduce solar flux. Reduced solar flux reduces battery charge. Reduced battery charge (on this system) reduces the voltage supplied to devices. Simply: I have demonstrated that the devices only function during periods of lower voltage than the normal "battery full" charge of ~13V.
What Needs to Happen?
- I need to buy a 12V-9V Buck from Amazon to correct the problem. It's an easy hack fix, but it's a shame that no one sells them locally.
- The Adafruit team should update the learn.adafruit page for the Data Logging Shield. There are now at least n=2 users who wrongly assumed that the Vin of a build using this shield could tolerate voltages higher than 12V based on the 20V tolerance of the Arduino Vin. I'm a bit miffed that I lost a few weeks of prime data collection due to a problem that might have not occurred with more documentation.
- Consider a different LDO on future versions of this or similar boards.