When using the temperature logger TAGs in the field, a lot of new users will react to the timestamps on the temperature measurements as they arrive in TAG Portal.
They see that the "last logging point" was several hours ago and think that the temperautre logger has prematurely stopped. This is usually not the case.
The TAG T1 temperature logger has a clock accuracy of ±1%. This is due to its internal oscillator circuit being affected by the cold and/or hot environment. When the product is cold, its internal clock will tick slightly slower and when it is hot the clock will tick slightly faster. This is a common effect in most electronic timekeeping devices. The result of this is an inaccuray in the logged time.
Allthough the effect seems minimal at ±1%, it can have a substantial impact on the data set. In a worst case scenario, where the absolute maximum clock drift occurs, the data logger can experience up to 1 hour and 41 minutes error over a logging period of 1 week.
The figure below shows an example low-temperature timeseries with clock drift at small scale compared to the expected timeseries. Here, the error is at the maximum -1% over a period of 24 hours. This results in a negative time offset growing over time until reaching 14 minutes and 24 seconds at the time of readout [24 hours and 5 minutes].
For hot applications, the effect will be reversed as the error moves to the other extreme (+1%) and the last expected data points will not have been logged yet at the time of reading, thus making it look like the data log is missing one entry.
If we go back to the cold example and increase the time before readout we start seeing a real impact in the data set.
Here wee see that after 9 days, the maximum negative drift is taking effect and the last log point is lagging 2 hours and 24 minutes (134.4 minutes) "behind".
These examples show the maximum error. These degrees of error only take effect near the extreme ends of the temperature range (-29°C / +50°C). In most applications the error is found to be below ±0.15%.
The table below shows the average clock error from a randomly selection of temperature logger TAGs in the most common operational temperatures.
The table below shows the error distribution within the same test data set.
As we can see, statistically the error is usually in the ±0.15% with some outliers for these temperatures.
Wikipedia does a good job at explaining clock drift further: