NOTES TO ipac_att FILES RELEASED 1997-05-28

1- This is the third and final release of the improved preflip 
   data, and includes position reconstructions for 30 orbits in 
   two packets from the pre-flip period, namely 03292312 and 
   03310335.  Previous reconstructions of these packets are super-
   seded and should be replaced.  Orbits included are 180-209, how-
   ever not all have been re-fit from previous releases.  Newly 
   fitted orbits are 180, 181, and 193-209, of which 180, 193, and 
   195-203 were previously flagged as dubious.  


2- Note the following changes compared to previous releases:  

      (a)  Orbit 180 could not be solved entirely, but Dr. Minoru 
           Freund obtained a good solution using NIRS data for the
           1100 sec including the galactic center.  Good reconstruc-
           tions for orbits 193 and 195--203, previously marked as 
           dubious, have been obtained.  Orbits 194 & 209, previous-
           ly believed to be good, were discovered to be substan-
           tially in error, due to G-angle changes in the middle of 
           the CC segment, and improved solutions have been 
           obtained.

      (b)  Orbits 204-208 have been fitted in two segments (AB & 
           CC) with an accompanying gain in accuracy, as has been 
           done for the previous two releases (of 1997-04-28 and
           1997-05-13).  The segmentation occurs at point "D", and 
           at the sunrise point of each orbit, the two segments being 
           denoted "AB" for the first and "CC" for the second.  The 
           result of this segmentation is a reduction in cross-scan 
           errors of typically 2-4 arcmin.  

           Most fit segments after orbit 192 have RMS residuals of 
           less than 2.0 arc min cross-scan and less than 1.0 arc min 
           in-scan, and many are substantially better.  Typical 
           attitude errors should be less, by a factor depending on 
           the number of match stars (usually in the range 15-45) used 
           in the fit for each segment, and on the distribution of
           unknown systematic effects.  

      (c)  These typical residuals are still somewhat poorer than the 
           corresponding figures mentioned (1.5 & 0.6 amin) for the 
           previous delivery.  The ripples, up to ~2 arc min in amplitude, 
           seen in the cross-scan residuals (Figure 1) are also larger 
           than for the two previous deliveries.  In Figure 2 the 
           histogram sigmas are larger than before because of the global 
           pattern in offsets  and the presence of several dubious orbits 
           early in the interval, as enumerated below.

      (d)  A few segments remain problematic, particularly some near 
           the beginning of the AB segment.  Orbits 182, and 185--187 
           remain dubious and have been flagged accordingly.  THESE 
           HAVE BEEN MARKED WITH A 99.0' POSITION UNCERTAINTY.

   ****** The users are warned that those time intervals, where the ******
   ****** boresight pointing uncertainties are set to 99.0', should ******
   ****** be used only with great attendant risk.  The uncertainty  ******
   ****** field M U S T  A L W A Y S  be consulted before use!!!    ******

      (e)  Portions of some other orbits are also flagged, typically when
           no or few secure match stars could be found.  This was the 
           rule in particular for the first ~400 sec of each orbit, 
           between point D and sunset.

3- Methods of reconstruction described in previous releases have generally
   been used, and quality checks previously described have been applied.


   Figure captions:

   Figure 1:  Plot of cross-scan (top) and in-scan (bottom) fit residuals 
   for the approximately 1245 match stars used in the 2 packets.  

   Figure 2:  Top:  Histogram of cross-scan residuals dX.  RMS is 1.60 
   arc min.  Bottom: Histogram of in-scan residuals dY.  RMS is 0.79 arc 
   min.