Wouter and I observed clipping in the hourly graphs at around 7200W. Often in combination with gaps in the graph. I also noticed this in the graph of a third user.
Of all cases I have screenshots available if desired. I suspect it is a bug of the sensor board, something must be overflowing or something.
The 7200W limit is due to too strict rrdtool input validation checking. The max validation parameter was set to a value of 2, meaning it would reject measurements with a rate higher than 2Wh/s = 7200W. Corrected in rev 79 and backported to existing rrd databases. An input power range between 0 and 72kW is now accepted.
Thanks for detecting and reporting the issue. Let me know if the fix works for you.
Cheers,
Bart.
Thats great. :) Do I understand correctly that all changes are made server side? Ergo the fluksometer owners don't have to do anything.
I will test the clipping by simulating a high load somewhere this weekend.
Yes, I can confirm that my samples are lost when I use more power then approx 72KW. Guess I have to cancel my plans for starting an iron ore smelting plant...
If the Fluksometer is the only obstacle to you starting this smelting plant, let me know. The solution would only be an integer away. :)
The input validation is meant to protect the rrd database from being corrupted by bogus measurements. The parameters were set a bit too strict. Thanks for testing!
Bogus input? Have you checked Yuri's graphs a few days ago? He had negative power! Now that is an interresting start for a smelting plant!
Yes I know, but smelting plants are very sensitive for solar flames.