[daip] comment about JPRECS

[Leonia Kogan]

Hi,

Recentely a problem was found with the subroutine JPRECS: routine to precess 
positions using the Julian IAU 1984 conventions (i.e. J2000 positions).  
Optionally JPRECS makes corrections for relativistic bending of the light 
by the sun, diurnal aberation and polar motion.

1. The description of the JPRECS input parameters says about DELDAT:
   DELDAT is interpolation interval; compute precession etc.
   parameters at this interval and do a linear interpolation (days).
   Actually there is no any interpolation inside of the JPRECS.
   Instead, all calcualtions are carried out for the time JD+DELDAT

   So when FITLD uses DELDAT= 1, the calculations are carried out for 
   the next day and difference can be severall hundresds milliarcsec.

   I am going to leave it as it, adding the explanation and may be default.
   Your opinion???

2. Eric payd attention that using JPRECS twice from mean coordinates to 
   apparent ones and back we do not return back to the same coordinates.
   The difference is as large as 4 milliarcsec.

   JPRECS calculates the coordinate corrections at the following sequence:
   mean => apparent: relativistic bending of the light by the sun =>
                     aberration => precession and nutation
   apparent => mean: precession and nutation => aberration 
                     relativistic bending of the light by the sun

   a) 'mean => apparent': precession and nutation =>
      'apparent => mean': precession and nutation

       Original codes: the direction change is done by change the time sign 
       from plus to minus. But the parameters of precession are calculated 
       as: a*time + b*time*time + c*time*time*time and therefore 
       switching the time sign at the 'apparent => mean' way can not compensate
       the precession and nutation at the 'mean => apparent' way.

       I do not change the time sign at the 'apparent => mean' way.
       Instead I evaluate the inverse matrix and apply it.
       Therefore the total matrix I aply at the two ways is the unit matrix 
       and the coordinates return back precisely.

   b) 'mean => apparent': aberration =>
      'apparent => mean': aberration

       Original codes: At the 'apparent => mean'way , JPRECS uses 
       the relevant formulae without paing attention that the arguments 
       the formulae are apparent coordinates as at the 'mean => apparent' 
       way the arguments are mean coordinates.

       I use the solution of the relevant algebraic equations to solve 
       the problem.

    c) 'mean => apparent': relativistic bending of the light by the sun =>
       'apparent => mean': relativistic bending of the light by the sun

        Original codes: The equation are too complicate to find the solution.
        Solution is found using two iteration. But the accurasy is not enough.

        I use the subroutine GRD used by the routine PRECES.
        Using this subroutine allows return back to the original mean 
        coordinates very simple.        

On Monday I plan to speak with Bill Cotton who wrote JPRECS

Sorry that the report is too long


Leonia