[time-nuts] Time Interval Algebra?

[John Ackermann N8UR]

I have what may be a dumb question but I can't get my head around it 
(having lousy high school algebra grades is not a good omen for a future 
time-nut).

I'm doing a time-interval measurement of Rb vs. GPS, using 1pps 
signals.  For convenience, I'm using the Rb as the reference for the 
counter.  The Rb 1pps is going to the counter "start" input, and the GPS 
1pps from a UT+/TAC is going to the "stop" input.

Over many days, the phase record indicates about a -1x10e-12 frequency 
offset.

My confusion stems from the fact that the counter is clocked by the 
device under test (the Rb), not the real reference (GPS).  Does that 
mean that the measured phase is actually twice the actual drift, so my 
-1x10e-12 is actually -5x10e-13?  I think so, but I don't have a lot of 
confidence in that conclusion.

Once that question is resolved, next is what impact, if any, this has on 
the AVAR calculation.  Is it the equivalent of measuring two identical 
units, so you'd divide AVAR by sqrt(2)?  (This I'm not so sure about, 
since true "identical units" would have independent noise, while here 
the "two" devices would be walking together.)

I suppose the real answer is to use a GPSDO as the counter reference to 
effectively have zero offset against GPS, but I didn't think of that in 
time :-).

Thanks for any enlightenment...

John