[time-nuts] Synchronisation using GPS

[Magnus Danielson]

From: "Shaun Doughty" <s.doughty at ee.ucl.ac.uk>
Subject: Re: [time-nuts] Synchronisation using GPS
Date: Tue, 9 Aug 2005 11:03:20 +0100
Message-ID: <002f01c59cc9$9233b930$61292880 at willow>

> Thanks, the VLBI stuff looks pretty relevant and interesting, and I notice
> that CNS managed to get an accuracy of a few nanoseconds between receivers
> with the M12+'s that Tom mentioned at a 21.5km baseline, which is
> encouraging too.
> 
> Within the radar network itself we are only really concerned about relative
> synchronisation between several sites, but this doesn't seem to be something
> that is reported on the majority of receiver data sheets - generally I guess
> it's just accuracy to UTC reported, so any information on relative
> synchronisation like the CNS tests is pretty useful for deciding whether GPS
> is a good idea to replace the current wired system being used as we move to
> slightly longer baselengths, and which receiver might be most suitable.

Just to give you the short story with minimal of technicalities:

Consider that you have two local sites and they have their own timescales
UTC(A) and UTC(B) (these names are chosen "arbitrarilly"), they both receive
UTC(GPS) (which is a bit incorrect, but lets just say it is what pops out of
the receiver) then you can locally measure the time-differences t1 and t2 which
we define as

t1 = UTC(A) - UTC(GPS)
t2 = UTC(B) - UTC(GPS)

if you now transmitt either one of these to the other site you can then
caluculate t12:

t   = t  - t  = [UTC(A) - UTC(GPS)] - [UTC(B) - UTC(GPS)] = UTC(A) - UTC(B)
 12    1    2

Notice how the time given from the GPS system cancels out. Most, but not all,
of the estimation error for UTC(GPS) will also cancel given equalent receivers.
This is the simple mathematical background for GPS common view and it will with
a very minimal transfer of information give you relative timing information.

A pair of M12+ and some minor post-processing should make you happy. Just some
additional processing and you can have your slave site track your master sites
phase using this common view comparision. A pair of time-interval counters and
means to control frequency should do it.

Cheers,
Magnus