That's interesting, didn't expect that gravity changes the frequency. But I wonder how the frequency counter measures mHz. Looks like in the video the gate time is one second, so this would be 10e6 cycles per second for 10 MHz. Even 100 mHz more would be still 10e6 (measured) cycles per second. Does it measure fractional cycles within the gate time, too?
The HP counters are reciprocal counters, therefore the basic resolution is constant for all frequencies, depending only on Gate Time and XTAL frequency.
That would be 1e-7 per second.
Additionally, these counters use time interpolation methods, so that the original 100ns is further interpolated by analogue or digital techniques.
This counter series allows 100ps interpolation, that means 1e10 counts per second gate time.
Also, they implement continuous counting and time stamping, which may further increase resolution.
The famous HP 5370B used a digital interpolation scheme and achieved 20ps T.I. resolution, i.e. nearly 1e11 counts per second.
Frank
PS: uploaded AN-200, as maybe by tomorrow, it will not be available @ agilent any more.