And the results are in ... Phew ....!
24 hour log of QO100 beacon frequency and a locally generated uplink at 2400.0525 MHz; 800mW to a 18dBi antenna, indoors, pointing at the satellite through a window.
The receiver and uplink were locked to a caesium frequency standard. Uplink was 26dB down on the beacon and set to be 2.5kHz higher; the 2.5kHz was then corrected for in the plots.
Tuning of the receiver was set such that a beacon frequency on exactly 10489.550000 MHz would come out as a tone at 1000Hz, the local uplink 3500Hz. Frequency measurement uncertainty 0.18Hz (65536 point FFT on 12kHz sampling) Measurements of each tone were taken at approximately two-minute intervals.
Results can be seen here:
There is clearly a cyclic variation due to Doppler of around 50Hz peak-to-peak.
The mean is around 1100Hz so, as this offset can be seen equally on both beacon and local reference uplink, it means the satellite LO is 100Hz low.
(This absolute frequency error assumes the 50Hz offset in the Octagon LNB local oscillator, discovered recently and confirmed independently, is EXACTLY 50Hz)
To measure the difference between the two downlink tone frequencies, corrected for 2.5kHz offse, tone measurement uncertainty of 0.18Hz was compensated for by adding a 6th order polynomial trend line to the delta-tone plot.
There is approximately a mean 2Hz difference between the received uplink tone and the beacon. Until known otherwise, this has to be assumed to be in the uplink frequency accuracy. A 2Hz error at 10GHz, (0.2 parts per billion) is considerably in excess of what the HP5061 caesium reference is capable of (even without local magnetic field compensation).
A 0.2 PPB error is probably beyond what might be expected if the beacon were generated from a good quality ovened oscillator, but is typical of the absolute accuracy of a rubidium source not recently calibrated.
A simultaneous test of the Cs reference against GPS timing indicates a mean error of certainly no more than 1 part in 10^12 (the measurement resolution) over the 24 hour test period
Looking the trend line, there a cyclic difference between the two signals in the order of 0.26Hz which is possibly due to the differential Doppler introduced by the separation of the two uplinks, in Germany and the South Coast of the UK. The value needs to be taken with care as this 0.26Hz is an average and little more than the instantaneous measurement uncertainty
One other residual source of frequency error (absolute) is the Elad-FDM duo that uses a Numerically Controlled Oscillator as its LO. This introduces a systematic frequency setting uncertainty of up to 0.016Hz
Short term variations visible on the curves, common to both signals, cannot be accounted for; possibly PLL glitches in the reference source, or propagation anomalies.