[time-nuts] What position is measured?

[jimlux]

Tom Holmes wrote:
> Thanks, Jim.
> 
> I assume that neither the satellite nor the receiver knows what the
> variation in the light time delay is, so it must be small enough to allow
> the claimed nanosecond accuracy of the PPS edge. 
> 


Well.. that's the difference between a L1 only and a L1/L2 receiver.  If 
you measure the same signal at two different frequencies, you can use 
that to estimate the total electron content (TEC) of the path, which in 
turn can be turned into a delay correction.

The uncertainties are on the order of meters/few ns, so keeping the 1pps 
to within 10ns is doable with a L1 receiver.




> Although one sat is sufficient for time work, would using more improve the
> PPS accuracy? Seems like having more inputs would help with the light delay
> and other corrections, but it probably is no different than having multiple
> Rb's in the lab (the guy with two is never quite sure and all that).


One sat works *if* you know where it and you are.  In practice, though, 
you look at multiple satellites and solve for position and time offset 
simultaneously.  The "secret sauce" in GPS receivers that distinguishes 
one from another is:
1) acquisition  (how long does it take to find the signal and start 
tracking)
2) how do you best form the estimate of position and clock offset.

Typically it's done with some form of Extended Kalman Filter (EKF) so 
you also wind up with estimates of the covariance matrix.  Whether or 
not that gets shoved out to the user is another matter.

the timing receivers separate the "where am I" and the "what time is it" 
questions.. you do a survey mode to get a precise position, then lock 
that down, and go to timing only mode, essentially averaging the time 
info from multiple satellites (not true averaging, almost always a 
weighted average)