Some results from testing radio clocks, OCXO and software based control loops ...
One of these has one of the CTI OCXOs ingested, the others run on the original design TXCO, slightly improved by a more stable control voltage. The flat box on top of the rightmost unit is a simple GPS module providing a PPS to one of the radio clocks.
This is the frequency counter and the data logger pushing the data to TimescaleDB (some Postgresql based DB server intended for time series data).
This one (a GPSDO assembled by forum member roadrunner) provides the reference clock
The result:
Green trace: Phase deviation of the received DCF77 signal over time. Note the unstable signal during the night time, cause by direct / ionosphere reflected signal fading. At about 15:00, one can notice large phase errors, this is caused by temporary transmitter shutdown (no signal at all, causing random phase values).
Yellow trace: Absolute time deviation of the GPS PPS controlled OCXO PPS output vs. the original GPS PPS output
Red trace: OCXO frequency over time
Looking at the frequency alone, it looks quite good (0.001Hz equals 1E-10 or 0.1ppb), but +/- 500ns over the day still has some room for improvement. Looks like it's caused by the integral part of the control loop.