After well over a month of smooth running, including the first storm of the season the day, before I had odd behaviour between 0315 and 0615 yesterday morning resulting in unwarranted gusts of up to 105mph!
During that period there were no single second reading that were “out of range”, and few one second step changes that weren’t “reasonable in some circumstances” and thus hard to filter logically as spikes.
Some sort of electrical interference at a frequency & duration that hadn’t cropped up so obviously before seems the almost certain cause.
Taking a close look at the data the, 105 mph gust could have been spotted logically quite easily (a 50mph acceleration between one second and the next!), but just before that there was another peak well over 90mph that was much more subtle, with each step in the chain quite reasonable in itself.
The more I look at the data the more it looks as though hardware (perhaps plus a little extra spike logic) is the best angle of attack.
I’m assuming that the interference is coming in short bursts of highish frequency rather than being sustained lower frequency.
The current low pass filter clips at about 160 Hz in theory (equivalent to 400 mph if sustained, i.e. well beyond any real weather), but that assumes no extra resistance or (probably harder to guarantee?) capacitance in the circuit. Without a scope (and ideally a signal generator?) it’s hard to be certain exactly what it is really clipping at and (aside from reporting invalid data) the last thing I want to have is a system that starts clipping real gusts.
So here’s what I have in mind. Maybe someone with more idea of electronics than I could comment on the plan?
If the weekend is dry (as looks possible) I hope to try adjusting the filter so that the theoretical clip is within what is actually happening and see how theory compares to reality. Then I can maybe raise the limit again to a value that is tighter, but still with a safety margin for real weather.
e.g.
100K and 0.1uF in theory clips at 15.9Hz (~40mph)
200K and 0.1uF in theory clips at 8Hz (20mph)
470K and 0.1uF in theory clips at 3.4Hz (~8.5mph)
If one (or hopefully more) of those seem to give a reasonable match between theory and practice, then maybe a couple of 100K and a 4K7 in series might be a decent “as low as reasonable without getting too close to potentially real data” filter?
24K7 & and 0.1uF in theory clips at 65.5Hz (~160mph!)
How’s that sound as a plan?
PS. So what chance a flat calm at the weekend, eh? 