...
The periodic spikes in GPSdiff_max up to 1 second that occur at 23:00 local time and last about an hour, are simultaneous with the network transfer of the day's data files from the DSM to the RAL server. These indicate increased sampling latency is happening at these times, which needs to be investigated and improved.
At these times there is also a little bump in NTPFreqOffset. It is likely that this occurs due to increased interrupt load at these times, increasing the latency of interrupt function that is called in response to the PPS interrupts. Increased latency in response to PPS interrupts should cause NTP to think that the system clock is ahead of the GPS clock, but the NTPClockOffset at these times is positive, and the slope of NTPFreqOffset is positive, indicating that NTP thinks the GPS clock is ahead. The bump shows up at cold ambient temperatures, so I don't think it is due to increased heating of the CPU card under the increased load caused by the network transfers.
As expected, the frequency offset shows a temperature dependence in the system clock oscillator. We do not have a measurement of the temperature inside the data system. The top panel in the plot below shows a time series of the ambient air temperature at 2 meters on the tower, along with the NTPFreqOffset, for a cool 3 day period in April. When the ambient air temperatures is below 5 deg C, the system clock oscillator does not show an obvious temperature relation.
...