Alright, dear Mechatrommer, it was my mistake. After all, you know that
I do not believe in ghosts!

The terms "voltage" as well as "ghost voltage" were not accurate but oversimplified ones. Would it be better if I used the terms "electric potential difference" and "unspecified/phenomenal voltage" instead, respectively? I was trying to start from square #1 since the measuring equipment impedance was very close to the impedance of the circuit under test, something that VK6ZGO explained very well.
The RP2200 probes are not a linear load, since they are loading the test circuit with a combination of a 10MΩ resistance in parallel to a 17±5pF capacitance. In DC their loading is straightforward: 10MΩ! But in AC, things change dramatically as the frequency rises.
For example, RP2200's load at 100MHz is equivalent to a 10MΩ ±2% resistor in parallel to a 72.3Ω .. 132.6Ω one (i.e. 17±5pF parallel input capacitance @100MHz), which equals to a load of an equivalent resistance of 72.3Ω .. 132.6Ω. Of course, this is not a load that can be ignored, since it inserts some significant measurement error to a signal source of 50Ω or 51Ω impedance.
Under these conditions, a reading of 2.32Vpp at the open end of a non terminated 50Ω transmission line using the RP2200 probes would suggest an actual presence of 3.20Vpp .. 3.92Vpp at the test points: A deviation of +38% .. +69% =
+53.5%±15.5 of the actual reading --having, of course, not taken into account any possible parasitic elements (as the probe's ground current loop inductance), or standing waves or reflections due to impedance mismatch of the load, and/or the length ratio of the transmission line. Well, this is not just a
deviation; this is a
faulty reading.
In a similar manner, the DS1000 oscilloscope input impedance at the BNC is 1MΩ±2% with a capacitance of 18pF±3pF. Again, in DC the input capacitance is insignificant; but not in HF (my mega8 16MHz oscillator example) or, even worse, in VHF...
-George
[EDIT]: Calculation errors corrected.
