[IGSMAIL-5965]: change to exclude some IGU clock predictions

[Jim.Ray]

******************************************************************************
IGS Electronic Mail      01 Jul 11:06:47 PDT 2009      Message Number 5965
******************************************************************************

Author: Jim Ray & Jake Griffiths (ACC)

Starting with the IGS Ultra-rapid (IGU) combination today
for 1538_3_12 we have implemented a new option to exclude
clock predictions for certain satellites.  Recall that in
each IGU SP3 product file, the first 24 hr period provides
satellite orbit positions and clock values based on near
real-time observations; the second 24 hr period provides
predicted orbit and clock values extrapolated from the
earlier data.  This change affects only the second half of
predictions.

While the IGU orbit predictions are of very high quality,
normally with errors at the few cm level (IGS Mail 5874 at
http://igscb.jpl.nasa.gov/mail/igsmail/2009/msg00000.html),
the IGU clock predictions are only about the same level as
the broadcast nav clocks.  So real-time point positioning
using the IGU predicted products performs not much better
than BRDC.  (On the other hand, near real-time positioning
using the first half of each IGU SP3 file should perform
nearly as well as the IGS Rapids.)

The change today will eliminate predictions for the most
problematic satellite clocks and should greatly reduce
infrequent disruptions due to unexpected clock resets.  A
prime motivation for this change is the realization that
most users of IGU clock predictions seek a stable time
scale reference, not accurate inter-satellite clock offsets
(which are being developed as a separate real-time product
by the IGS Real Time Pilot Project).

Starting today, IGU clock predictions are excluded for:

* all 7 IIA SVs with Cs clocks -- PRNs 3, 8, 9, 10,
   24, 27, 30 (usually all highly unstable)
* all SVs set UNUSUFN -- presently PRNs 1, 25
* any SV during MX events -- presently none

Clocks on the unhealthy UNUSUFN satellites are sometimes
changed with no notice and MX events usually deal with
clock issues, so eliminating these should improve the
reliability of the IGS clock predictions.  The IIA Cs
clocks are simply much less stable than other SV clocks.
As the very old satellites continue to be replaced, the
number of IGU clock predictions will increase.

We expect that the RMS of the remaining IGU clock
predictions will improve from about 2 to 4 ns now to
well below 1 ns.

An ongoing problem with the IGU combinations is that
the summary reports are incomplete.  Therefore we will
be working to enhance these reports so that they give
a comprehensive account of the dynamic variations that
occur in the combinations, including clock exclusions.