PDOP —
Timing history
Scale follows the observed range to make slow frequency drift visible.
Kernel thermal zones inside the enclosure. Compare against the frequency trace above: with the case closed, the rubidium and the CM4 share one thermal volume, so load-driven warming can appear as reference drift.
Discrete receiver-reported integer values, not measured absolute error.
Local clock minus each server; ± is Chrony’s NTP root distance. This is a millisecond-scale network cross-check, not a nanosecond calibration.
Receiver-native indicators, not proof of interference or authenticity. Correlate alerts with signal levels, position, and time behavior.
Best carrier-to-noise: — dB-Hz · max-error bound —
Overlapping Allan deviation of Chrony’s frequency estimate for the AR-40A, from the longest unbroken stretch of the last 7 days (— h). Hourly drift-file updates set the floor, so 1 h is the shortest meaningful τ. A mechanical trim, a power cycle or a kernel swap is a real frequency step and inflates every τ for days afterwards.
Allan deviation of the ZED-F9T’s own TCXO against GNSS time, from the last PPP run (—, offset — ppm). Nothing disciplines this crystal, and it is not the 1 PPS the receiver outputs; the rig’s frequency comes from the AR-40A, some three orders better.
Processing and publication dates this station waits on. CSRS-PPP final products appear about two weeks after the observations; the BIPM publishes each month’s UTC(MIKE) file a couple of weeks after month end, so a comparison logged in one month yields a number in the middle of the next.
—