[time-nuts] Time Transfer

Tom Van Baak tvb at LeapSecond.com
Fri Dec 14 16:06:09 EST 2007

> Also for GPS timing applications (i.e. the GPS receiver is in position hold 
> mode) a single satellite is all that's needed.  Adding more satellites can add 
> a failure alarm function but may not improve accuracy.  In my case multipath is 
> a big problem so a high elevation mask is needed.

Both one-satellite and all-in-view methods are used for time
transfer. But all other things being equal you get the typical
sqrt(N) improvement with multiple concurrent SV in view so
it's more than just a redundancy trick.

Your concerns with multipath are valid; I suspect the more
serious the user the more attention is paid to the antenna
design and placement (e.g., choke ring antennas, remote
sites, dual frequency, etc.).

> An idea: When averaging GPS 1 PPS signals the receiver switches satellites as 
> the currently tracked ones set so there may be a slight change in accuracy 
> because of the satellites being used.  But changing from a satellite that's 
> about to set i.e. that has a lot of multipath to high in the sky sat should 
> make for a better 1 PPS.

The traditional method of GPS common view avoids this by
carefully picking which SV to track at which specific times. Thus
there is no switching in the middle of any 13 minute track.

I'm not sure what the all-in-view time transfer solutions do for
this case. Someone can check FCS, ION, or PTTI abstracts
for a paper on this. Somewhere there must be a nice plot
showing measured accuracy by SV as a function of angle.
This would be a worse issue, except that even since the first
GPS receivers, provision is made for elevation mask angle.

I'm guessing low elevation is more than just a multipath
problem. The ionosphere correction may also be less accurate
at low angles. Perhaps Tom Clark can tell us about this.

Brooke, one cool thing you could do with your Oncore and a
cesium is to change your elevation mask angle, say once a
day, or day and a half, going all the way from 0 to 90 over a
week or two.

That single plot will tell a good story. You might guess that you
will get poorer results near 0 (obstructed view or excessive
multipath) and also near 90 (not enough satellites in view). But
where the sweet spot is, and how many tens of degrees wide
it is will be revealing.


More information about the time-nuts mailing list