Defects Index | Documentation Table of Contents
00/01/04 DEFECT IN LIBRARY libgpsF
SYMPTOM: Binary ephemerides from the orbit integrator, arc, had large
differences when compared to standard products such as the
IGS ephemeris. Subsequent examination showed that this only
occurred if the ephemeris' initial conditions were after
00/01/01 10:59:08 UTC.
PROBLEM: The computation of the equation of equinoxes was not completed.
CORRECTION: inod.f: change the initial definition of a "last" time to a
large negative number so that it always appears that a significant
time has elapsed on the first call to subroutine eqdot.
FOUND BY: W. Kass and R. Dulaney
FIXED BY: M. Schenewerk
NOTES: This will require that arcic, rotorb and pages to be recompiled.
arc requires (I believe) that the a priori state vector for
each satellite be in a standard inertial frame. To convert
to the inertial frame, computation of several rotation matrices,
including one for apparent sidereal time, is required. Computation
of apparent sidereal requires more detailed knowledge of the
celestial frame which, in turn requires knowledge of the nutation
series etc. Computing this resource intensive, but fortunately,
the change is slow and predictable. To speed the runtimes of
programs requiring this information, the routines were modified
to interpolate between values and only compute new values when a
"significant" time had elapsed. This, of course, requires
initialization which was done by setting the first "last time
computed" to zero. Approximately 10 hours after midnight on
Jan 1, 2000, the reference century used in computing the
nutation series switched from positive to negative. With that
switch a zero "last time computed" would not trigger the
computation of the nod parameters. Thus, the equation of equinoxes
was in error by about 1 sec and the apparent sidereal time was
in error by about the same amount.
January 04, 2000