Phantom rain rate

[mcrossley]

hmm, that is odd, Cumulus just recorded a rain rate of 306.9mm/hr at 22:33 when there was no rainfall recorded? I'm not in a position to look at Cumulus tonight, I'll take a look in the morning.

[steve]

As you no doubt know, all of the rain figures are derived from the station's 'rain counter', so presumably it did something odd. If you look in the diags file, you should be able to see what happened.

[mcrossley]

It was the station throwing a wobbly of course.

The rain counter and temperature values changed, the rain for more than 6 readings. So Cumulus accepted the new value, then it jumped back to the correct reading where it has stayed since.

So Cumulus correctly compenstated the rain total for the rain total counter jumping around, but it looks like the rain rate calculation is still factoring the jumps in?

The jumps of both counters look like single bit flips: the rain counter jumped by 256 (1310.1 -> 1233.3 = 76.8/3*10 = 256) , and the temperature by 64 (10.4 -> 4.0 = 6.4*10 = 64)

It looks suspicusly like the bits continued to be flipped until the station moved on to the next memory location - maybe there is a bad memory block in there somewhere and I'll hit it every cycle round the buffer

[steve]

So Cumulus correctly compenstated the rain total for the rain total counter jumping around, but it looks like the rain rate calculation is still factoring the jumps in?

Because of the way the rain rate calculation is done, it's hard to cater for anomalous jumps up and down in the rain counter automatically, so instead I've provided a rain rate 'spike removal' setting.

[mcrossley]

OK, thanks Steve. I'm not sure how I would generate a valid number for a rain rate spike filter value - a sudden downpour can see the rate jump from essentially zero to a hundred or more.

Anyway all cleaned up, the Cumulus editors make it much easier than before, though there is still some manual file clean up left, and it nicely coincided with the monthly patch reboot. Though I will have to take it down again later today to have a new smart meter fitted.

PS: I'll keep a look out around the 8/9th December for more funnies as that is when the memory buffer should cycle round to the same location again