A new error for me, after 5 years!

[dc1500]

Every 12 to 48 hours (random) Cumulus fails to pick up outdoor temperature reading. Sometimes it self rectifies after a random number of hours and always rectifies with a restart of Cumulus. The console displays outside temperature all the time but when Cumulus is restarted it does not reload up this missing temperature data. The log has just kept registering the last temperature when it was working.
Additional information:-
1. All other data picked up ok
2. Have installed latest version of Cumulus with no effect
3. No system changes have taken place before this started to occur.

Any ideas?
Dave

[steve]

Please zip up the diags folder and attach it.

[dc1500]

Hi Steve, after it happening again this morning and examining the last diag file (attached) I may have discovered the problem. I had a problem with spikes several months ago and set the spike removal in the calibration for temperature at 0.5C (10min interval) Of course at this time of year we are getting sometimes 18C change in 6 hours and this value is being exceeded for the first time. It is comparing the original value against each new value so the gap just keeps getting bigger instead of comparing it to the previous value. This is flagged in the diag file. So it seems to just continue to log the old original value until Cumulus is re-started manually. There do however seem to be some instances when it has automatically started logging again but I am sure the period has been longer than that of temp increases outside the preset for the spikes. If this is how it should work I have rectified by changing the spike value to 1C in 10 minutes. Still of course can happen with line fronts etc but that won't worry me as I will see what has happened.
Regards, Dave

[steve]

The spike removal values are applied to consecutive updates of live data, i.e. over a period of 48 seconds. As the help warns, if you set the value too low such that a genuine change exceeds the setting, you will be stuck at the same value until the reading returns to within range again.

The spike removal option is intended to filter out the very large erroneous readings that Fine Offset stations can sometimes produce - actual spikes of several degrees.

[dc1500]

Yes, I understand that - it's just that my spikes were small , just a few degrees so I wanted the lowest value that would work most of the time which I am now closer to.
When you say till it returns to the "normal range again" do you mean back down to the value when it happened or when the spike between two consecutive readings is back below the value entered for the spike? Because looking at the diags the first spike was 0.6C (8.5-9.1) which was ignored because the spike was above the 0.5C set. And the first few subsequent 10 min interval were but after a while e.g 12.07 - 12.17 onwards when 10 minutes apart readings were 14.2 -14.0, only 0.2 C apart. But the diag seems to be comparing the 14.0 against the original 8.5C and ignoring it even though it is back in spec? Surely it should compare it to the reading 10 minutes before to see if it is still a spike or am I missing something?
Dave

[steve]

It's designed to remove large spikes, not variations of half a degree or so. The spikes that Fine Offset stations tend to produce where the temperature goes from 10 degrees then up to 30 degrees and then back down again to 10 degrees again on the next reading. So it compares each reading with the last value that it knew to be good, i.e. 10 degrees in this case. Comparing the next reading with the reading of 30 degrees would be self defeating, because if the spike happened to last for more than one reading (I'm not sure that they do in practice) it would then decide that the second reading of 30 degrees was OK because it was within the specified range of the previous one.

I don't know where you're getting this 10 minute figure from; as I told you, it works on readings 48 seconds apart. Well, actually 10 seconds apart, because that's how often Cumulus reads the data, but the station only updates every 48 seconds.

[dc1500]

Hi Steve, like I said I do understand it's for large spikes and as I said I was getting very short spikes of say 2 deg which was messing up my max and so I used the facility to prevent this and it worked, till now when we have rapid temperature changes so I have upped it. The only reason I quoted 10 mins is that is the datalog interval set as per log files and I was forgetting it is reading data every 10/48 secs thinking it was working on that.(or not thinking actually, as the diag file clearly shows otherwise!) I am in no way criticising you or the software, as you know I have been full of admiration since the beginning. I can now see the reason for not comparing to the last reading in case the spike lasts for a period of more than 10/48 secs. It is not going to be able to correct for small sharp spikes and will confuse them as in this situation of a steadily rising or falling temperature, or at least not for a long time until the original temperature is reached. I just need to find the lowest value likely between two readings that is unlikely to be real and that won't be that high. I guess if it were to be set to return to normal logging after an arbitary time of say 15 mins whatever the differential it would work for any short spike although of course if it were just an incremental false jump that stayed there the data would subsequently be wrong, but the fault would not have been spiking.
Thanks,
Dave