Several improved routines associated with hsi_spectrogram have been =
placed in atest.   They effect all of the binned data, i.e. spectrum, =
lightcurve, and image.

An important fix is to a bug introduced into the dropout routine about =
June 24.  Since that time, the bug caused the DP_CUTOFF parameter to be
=
ignored.  The effect was to create large datagaps and excess apparent =
deadtime when the rates were low.  This may have also had a marginal =
effect on imaging with low total count rates, I can't assess how serious
=
it might be.  This was a very nasty bug, introduced when I fixed another
=
problem.  I need to add to the test suite so we can be certain to catch
=
these problems sooner.  As a caveat, many people run the lightcurve gui
=
when initally examining an event, or look at the observing summary.  =
Presently, the standard view of these rates is not corrected for =
deadtime.

There was also a fix to the NPHT out of range bug in that routine.


In hsi_spectrogram__livetime_merge_datagaps.pro there were two fixes:
;Add protection for one sel1 value, ras, 13-aug-02.
;Ensure that ut_off.e is after ut_off.s, ras, 14-aug-02.


In hsi_spectrogram__define there are three changes:

The control of the binning algorithm has been made more robust.  That =
should reduce the time to fix future bugs in this area, and hopefully =
reduce some of the bizarre results you may have experienced from time to
=
time.  One change affects the counts in the last channel of a spectrum =
which should be unimportant for most puposes.  The other change =
introduces support for adjusting the detector calibration as a function
=
of time through a file in SPEC_RESP.  This should be transparent to =
everyone but important for the best quality spectroscopy, particularly =
narrow line spectroscopy. =20

The F_DIV function has been changed to be more like the division =
operator in IDL.  The purpose of F_DIV() is to provide a division =
operation which traps division by zero, and sets the result to 0 or a =
user default. The version online could result in an array as large as =
the denominator array, instead of the lessor size of the numerator and =
denominator arrays.  If you have been using this function, it may cause
=
problems if these two arrays could have different sizes, so beware.

The changes to the other routines are minor.

Richard