This message concerns an improvement and a bug fix for pileup
correction.  This does not fix all of the problems associated with
the pileup correction.  I'm hoping to make progress in this area

1. NB, if you have been writing spectral fits files with both
pileup_correct and sp_semi_calibrated set, those files should be
discarded.  You can examine the info parameters in the third
extension ( d3 = mrdfits( fitsfile, 3) and examine d3."infoparname")
to see if you have done this.  There was a bug that applied the
diagonal corrections into the count rates on the input to the pileup
correction routine.  That was incorrect and would probably have
resulted in results so obviously suspect as to have been normally
discarded.  However for mild deadtime the effect could have been
small enough to have been unnoticeable.  This bug has been fixed and
the new version will propagate by 23 June 2005.

2. The PILEUP_TWEAK control parameter used to modify the amplitude of
the pileup correction has been changed from its original default of
1.0 to the Brian Dennis and David Smith recommended value of 0.6.
This will propagate similarly to the change above.

If you have any results based on the conditions in 1, you must
reanalyze that data.  For number 2, we believe 0.6 is the superior

Richard Schwartz