[IGSSTATION-660]: Using IGS Clock Products to Monitor GPS Station Performance

["Ken Senior" <senior @ juno.nrl.navy.mil>]

******************************************************************************
IGS Station Mail          15 Oct 07:09:16 PDT 2005      Message Number 660
******************************************************************************

Author: Ken Senior (NRL) & Jim Ray (NGS)

Since 22 Feb. 2004 the IGS official satellite and station clock products
have
been aligned to a new, highly stable timescale realized by an internal
ensemble of the available frequency standards dynamically weighted based
on
their individual instabilities.  (Comparable, unofficial clock products
are
archived starting 29 Oct. 2000.)  All results and a variety of associated
plots can be accessed at https://goby.nrl.navy.mil/IGStime/index.php.

In addition to their value for clock diagnostics, these products can be
used
to monitor the general health of GPS tracking stations.  This is only
minimally true for stations not using H-maser external frequency standards
since the large epoch-to-epoch clock variations limit the sensitivity to
detect all but the most severe types of problems affecting data quality. 
With
H-masers, though, many subtle effects on the pseudorange observables can
be
observed.  This is because the overall clock bias for a given processing
arc
(normally 24 hours) is set by the average code data, while the higher
frequency variations are determined by the carrier phase data.  Assuming
an
average uncertainty of 1 m for code data and 5-min sampling, the formal
accuracy of each clock estimate should be near 120 ps.  A quantitative
test of
the actual clock accuracy can be made by comparing clock estimates at the
boundaries between independent analysis arcs (i.e., at midnights between
consecutive days).  This is analogous to the classic geodetic
repeatability
test for a time series of positioning results.  The test is only feasible
when
H-masers are used since the instabilities of lesser frequency standards
dominate over the day-boundary jumps caused by code data quality.  For
further
background and details, please refer to our paper at:
  https://goby.nrl.navy.mil/IGStime/refs/ray05.pdf

Applied to IGS clocks, we have found a very wide dispersion in station
performances, from RMS clock jumps near the expected level of 120 ps to
values
>1 ns.  Results are now posted in plots (updated weekly) available at:
  https://goby.nrl.navy.mil/IGStime/daybdy/
In a number of cases, there are discrete changes in the magnitude of clock
day-jumps correlated with equipment changes or malfunctions of various
types.
Sometimes the cause of changes in performance is unknown.  Three Canadian
stations (ALGO, NRC1, YELL) show large annual variations, being far worse
in
winters than in summers.  In general, the clock jumps presumably reflect
primarily the level of local code multipath, especially the longest
wavelength
type due to near-field reflections.  It is likely that in most cases there
are
associated effects in the carrier phase data , which are much more
difficult
to detect since they are largely absorbed into the estimated ambiguity
parameters.

The same techniques can be used to monitor data performance at stations
not in
the IGS network, provided that they are equipped with an H-maser frequency
standard.  A time series of clock estimates must be determined with the
precise point positioning method using fixed IGS satellite orbits and
aligned
clocks.  The high stability of the IGS timescale ensures that day-boundary
jumps in the station clock can be detected to better than 100 ps RMS.

Station operators are encouraged to check the clock-jump plots
occasionally.
Here is a brief summary of stations where we have noticed recent changes
in
performance:

  SPT0      -- degraded after about 15 Jul. 2005
  TID1/TIDB -- degraded after about 10 Jan. 2005
  YELL      -- usual improvement in summertimes not seen in 2005


Kind Regards,

Ken Senior

--
Naval Center for Space Technology
U.S. Naval Research Laboratory (NRL)