Cumulus Support

Morning all,

I had a temperature glitch in the readings last night and I have sat and changed all the data / ini files to remedy, but it got me looking at the spike removal settings and I have now changed the values from 999 to what I think will work (i.e. 3.5 for temperature).

Just wondered what values you use that you have found to work well?

These are my spikeremovals for the Fine Offset

Very kind of you, thank you. Not too dissimilar from what I have entered.

This has worked well for me. Took me a while to get it to where I wanted. I never had to change it from default settings till not quite a year ago when my station's rain gauge decided to act up and keep telling me I had several inches of rain in a few minutes on a dry day.

Thank you for your response also. No problems last night, but then I hadn't had for the previous 2 weeks. Very odd these things.

My two penn'th. (and that is all it is worth)

The spike settings need to take into account previous experience of extreme weather in a particular location,
[EDIT: You need a value which is beyond any possible legitimate difference in readings which would appear 48 seconds apart. In other words ask yourself how much can a value really change in just 48 seconds, then make the spike value higher.]

Temperature differences tend to lag quite a bit but I have noticed that occasionally the temperature here can rise or fall by up to 3'C in 5minutes. I have my spike set at 3 times that at 2'C per minute

Wind speeds (gusts) can go from 'fairly light' one minute to a record breaking speed the next. I set gust differences at 35m/s or 78 mph which is twice the highest difference I've had so far.

Average wind speeds are less likely to violently fluctuate from one minute to the next so I set that at 9m/s or 20mph, which again is 3 times my highest observation difference.

Humidity I set at 10% a minute which seems unlikely even with a massive change in weather.

Pressures can drop quite quickly and recently I had a legitimate drop of 5mb/min. I now set it to 10

Rain rates can be huge and I haven't worked out a reasonable figure yet so I leave it at 999.

How much rain can you log in an hour? What figure would be beyond any possible extreme for your location? I set it at 5 inches or 127mm which is well above the UK record of 92mm.

I always tick the 'Log to Error Log' box so I can evaluate it myself later on.

Wow, those figures are exactly where mine are set - excepting rain rate I left at 500, for exactly the reason you say (A lot can fall very quickly, but not necessarily a big total).

Yours and my weather will be very similar, hence why we came up with the same conclusions I guess.

Note that the spike removal settings are independent of your logging interval, either in Cumulus or in the console. They work on consecutive live readings, i.e 48 seconds apart, as far as the station is concerned.

Ahhh, so the settings could be even lower. Noted

steve wrote:Note that the spike removal settings are independent of your logging interval, either in Cumulus or in the console. They work on consecutive live readings, i.e 48 seconds apart, as far as the station is concerned.

I hadn't appreciated that. I assumed the 'spike' was recognised from the recorded data. It makes little difference to me as my recording interval is 1 minute anyway, but if users were assuming longer intervals they may have higher figures than they need. I have edited the post that suggested that recording interval was a factor

westonweather - I agree, same 'same problem - same solution'. My figures have been honed through experience, and our experiences will be quite similar.

Hi,

steve wrote:They work on consecutive live readings, i.e 48 seconds apart, as far as the station is concerned.

Correct, but two words of warning about that statement. Firstly, the station doesn't actually report a "lost contact" until about 8 consecutive messages have been "lost". So it's possible for a gap of over 5 minutes to occur in the data stream without an "error" being logged. Secondly, the FO Radio Controlled Clock versions (e.g. WH 1080) may "go to sleep" for several minutes "on the hour" (to check for time signals), so again there can be a gap of some minutes in the updating of the data.

My final warning is to beware that once a "larger than expected" step occurs in the data, there is the risk that the spike settings may "lock out" all subsequent data changes (until it returns to similar values), so don't set them too low.

Cheers, Alan.

True, and the help warns about the latter point. Some users try to use the spike removal settings for readings which are just within the normal variation of the station, and this is dangerous and not at all the purpose for which they are intended. They are intended for the sudden very large changes in readings which can occur.

I agree that spike figures should catch impossible values rather than abnormal ones. In fact, in the light of Allycat's information, I think I need to look again at some spike values. A few of them might be exceeded in the scenarios he has outlined, although fortunately I hardly ever get a sensor signal drop out.

I get very few spikes logged too, but am I right in thinking that Cumulus does log the figure (or the difference) in the error log when a spike is detected?

Yes, it writes this sort of thing:

10/01/2014 10:57:12.371 : Wind or gust difference greater than specified; reading ignored
10/01/2014 10:57:12.381 : Gust: Old value = 0.0 New value = 40.0 EWgustdiff = 25.0
10/01/2014 10:57:12.382 : Wind: Old value = 0.0 New value = 0.0 EWwinddiff = 25.0