Bug ? 1.8.8_868

[beteljuice]

My 'Todays Rain (so far)' figures have suddenly zeroed

01/11/09,20:21,5.9,83,3.0,3.0,14.1,266,0.0,10.9,996.4,5.7,21.4,30,2.7,5.9,5.9
01/11/09,20:22,5.9,83,3.0,2.5,8.5,249,0.0,10.9,996.5,5.7,21.4,30,5.4,3.5,5.9
01/11/09,20:23,5.8,83,4.6,4.5,4.5,203,0.0,10.9,996.4,5.7,21.4,30,4.5,4.3,5.8
01/11/09,20:28,5.8,83,4.6,6.6,8.7,188,0.0,0.0,996.4,5.7,21.5,30,8.7,2.9,5.8
01/11/09,20:31,5.8,83,2.9,3.5,8.7,280,0.0,0.0,996.5,5.7,21.5,30,2.0,5.8,5.8
01/11/09,20:32,5.8,83,2.9,4.0,9.8,247,0.0,0.0,996.4,5.7,21.5,30,2.7,5.8,5.8

Base station (LaCrosse 2300) still has a count. Cumulus still has a max rain rate, and the graphs show rain, but no figures in 'output' to console or realtime.txt

update: ... hmm .. looks like my base unit 'total rain' has rolled around the clock again.

Did we ever think a way out of this one ?

[steve]

My assumption has always been that the counter would roll over at a convenient time - i.e. midnight on New Year's Eve (but yes, this still isn't ideal for 9am users), so wouldn't be too much of a problem. I think this is probably true for some but not all station types. What value did it roll over at?

[beteljuice]

my base unit 'total rain' has rolled around the clock again

By that I mean 999.9 => 000.0

[steve]

I'm confused, because the (unoffical) spec I have says that the rain counter wraps at 10000. But that's probably irrelevant anyway. The problem with detecting the wrap is being able to tell the difference between a genuine wrap and a reset. I suppose I could check for the total getting over 9999 (or 999 or whatever it is) and then assume that when it gets to zero it's wrapped rather than reset, but that's only a little bit better than what it does now. Maybe.

Could you paste the bit of the Nov log where it actually wraps?

[beteljuice]

01/11/09,09:35,8.4,97,7.8,9.0,20.6,175,2.2,10.3,993.1,2499.4,22.3,34,13.6,0.9,8.4
01/11/09,09:36,8.4,97,7.8,9.0,20.6,176,2.3,10.4,993.0,2499.9,22.3,34,11.6,1.9,8.4
01/11/09,09:37,8.5,97,7.9,8.9,20.6,177,2.0,10.4,993.0,2499.9,22.3,34,2.0,8.5,8.5
01/11/09,09:38,8.5,97,7.9,8.6,20.6,176,2.1,10.4,993.0,0.0,22.3,34,6.5,5.3,8.5
01/11/09,09:39,8.3,97,7.7,8.1,20.6,175,2.1,10.4,992.9,0.0,22.3,34,8.9,3.3,8.3
01/11/09,09:40,8.5,97,7.9,7.7,20.6,179,2.1,10.4,992.8,0.5,22.3,35,3.4,8.5,8.5
01/11/09,09:41,8.5,97,7.9,7.7,20.6,178,1.7,10.5,992.7,1.0,22.3,35,3.6,8.5,8.5

Actually it looks like it rollsover at 2499.9 !

But now I am confused again !

Why did Cumulus 'lose' the day total hours later ?

01/11/09,20:23,5.8,83,4.6,4.5,4.5,203,0.0,10.9,996.4,5.7,21.4,30,4.5,4.3,5.8
01/11/09,20:28,5.8,83,4.6,6.6,8.7,188,0.0,0.0,996.4,5.7,21.5,30,8.7,2.9,5.8

[steve]

Actually it looks like it rollsover at 2499.9 !

That makes more sense, the documentation doesn't actually say that it rolls over at 9999, just that it has a thousands component. So I guess this is only an issue for most people once every 2 or 3 years; more often for those in the wetter parts.

[steve]

Why did Cumulus 'lose' the day total hours later ?

01/11/09,20:23,5.8,83,4.6,4.5,4.5,203,0.0,10.9,996.4,5.7,21.4,30,4.5,4.3,5.8
01/11/09,20:28,5.8,83,4.6,6.6,8.7,188,0.0,0.0,996.4,5.7,21.5,30,8.7,2.9,5.8

From the timestamp intervals, it looks like the difference between 'live' and 'logger' data, so a restart of Cumulus was involved. I guess the fact that the current counter was less than the start of day figure has caused it to start ignoring rainfall. I'll look at the code. Without the restart, it looks like it would only have lost 0.1 mm.

[steve]

I see what the problem is, but I'm not sure how easy it will be to fix. When it spots that the rain counter has wrapped, it creates a pseudo start of day figure based on the current counter and today's rain so far - hence the generated start of day figure will be negative. This is fine until you restart Cumulus; it uses a negative Start= figure to indicate that the Start= figure has not been initialised, so ignores it and starts counting again from that point.

I guess it should really use an 'uninitialised' value of a very large negative number, rather than anything negative.

[beteljuice]

From the timestamp intervals, it looks like the difference between 'live' and 'logger' data, so a restart of Cumulus was involved.

Yup - sounds about right, probably the time it took me to update Cumulus to _868 !

[beteljuice]

FYI - the new LaCrosse WS2800 / 2810 is like we (or I) originally thought

(Rain)
"From 0 to 999.9 mm (over 1 or 24 hours) - from 0 to 9999 mm (over 1 week, 1 month
and total rain volume"