Hello,
i have a bunch of CR10X used for meteorological data gathering. Recently most of them started at the *same time* to compute a very low value ranging from -0.009 to 0.003 in place of zero. The value should be exactly zero because the value is the hourly mean of a 5 seconds flag (0 or 1). The code also computes a 10 min mean of this flag and this value is correct (i.e. 0.000).
I have checked the code, compared between affected and not affected stations and i don't find any explanation. Can someone give me a hint?
Thxs,
LGS
The datalogger works in what is called "single-precision" floating point maths, which can lead to apparent rounding errors, that can lead to "non-integer" results when you might otherwise not expect them. This is not a problem specific to our loggers but a consequence of binary mathematics. There is a little description of this in the logger manuals.
It would surprise me though if working out the average over ten minutes of 0/1 would create such an error though, unless perhaps you are using some form of rolling average rather than a block average.
If you are using a rolling average then it is possible for the rounding errors to accumulate with time. We found this affected the rolling average instruction in the CR10X and recoded the instruction about halfway through the logger's product life, to avoid this. If you are using that instruction you could consider upgrading the operating system or the logger. You can also get the same error if you had written you own code too - to avoid it the secret is to avoid keeping a rolling total which values are forever added to or taken way.