Re: [TASS] photometric vs. non-photometric

[Stupendous Man <richmond@A188-L009.RIT.EDU>]

  Andrew Bennett has noticed the uneven coverage of 'tenxcat':

> A cursory examination of tenxcat shows that some sources
> were measured once and some 40 times. More detailed
> examination shows that this wide scatter is:
> 1) Not confined to weak sources.
> 2) Not confined to the edges of the survey area.
> 3) Not just the result of the uneven RA coverage.
> Indeed the scatter persists even for small selected
> areas well within the most heavily covered area and
> with the exclusion of weak sources.

  I suspect that some of this scatter is due to the matching
of sources from night to night.  Images from each night are
reduced locally, yielding list of stellar positions and brightness.
When we import them to the database, each position is compared
to positions of stars already in the database: if a new
source is within 15 arcseconds of an existing source, then

      1. the new source is added to the database as one more
             instance of the existing source

   (and here's the critical part)

      2. the position of of the existing source is updated, to
             become the mean of all instances (including the new one)

  So, it turns out that the position of a source in the database
"wanders" a bit as new positions are added.  This probably isn't
the best way to do things, though I can't think of a better way
at the moment.  Here's the problem:

  Consider three detections of a star:

             - detection 1 is at RA = 0, Dec = 0
             - detection 2 is at RA = +12 arcsec, Dec = 0
             - detection 3 is at RA = -12 arcsec, Dec = 0

  Now, if we use a 15-arcsec matching radius, detection 2 will be
classified as just another instance of detection 1.  The mean position
of this entry will become RA = +6 arcsec.

  When detection 3 is imported, it WILL NOT MATCH to the combined
1-2 entry.  So, it will be put into the database as a new entry.

  If detection 3 is imported before detection 2, of course, the
situation is reversed.

  I know that the TASS database contains many "spurious pairs",
sets of stars which are really instances of the same celestial object,
but which appear as separate entities in the database.  I believe
that some of the objects Andrew mentioned, which appear only a few
times in the midst of a field with stars detected many times,
may be members of such spurious pairs.

  Open question to the TASS mailing list: given a stream of data,
night after night, of stellar positions with some estimated
uncertainty "sigma" arcseconds, how should we identify detections
as instances of the same physical object?  Perhaps we should wait
until N nights have been collected and then merge all at once,
rather than merging 1 night at a time in a gradual process?

                                              Michael Richmond