[Kstars-devel] Some observations regarding USNO NOMAD in KStars

[Akarsh Simha]

Hi

I did some tests on the USNO NOMAD catalog machinery in KStars
today. I'm sending a copy of these to the mailing list both for
documentation / archival purposes and so that others can look it up
when required.

This was prompted because I noticed that the star field in Trixel
12310 (of the "deep" level 6 SkyMesh) did not seem to match with the
DSS image of the same region. This could involve precession and might
be related / a bug introduced by the short-circuiting of precession,
nutation and aberration computations when the simulation time elapsed
since the last such calculation is less than about a minute (see
revision 832469e64f222e6d1cc9da77a7b4587db20c05ba). Or it could just
be that the catalog (with the search criteria used to limit down to a
100 million stars in KStars) disagrees with the POSS plates.

The following are the results of some checks done against trixel 12310
/ a rectangular patch in (RA, Dec) around trixel 12310:

1. (ra, dec) seems to be recomputed when the simulation clock is
   sufficiently slewed.

2. (ra0, dec0) of stars in KStars as dumped from JITUpdate() in a
   rectangular patch near Trixel 12310 agree with the dump of trixel
   12310 from USNO NOMAD 1e8.dat

3. (ra, dec) computed from (ra0, dec0) by accounting for precession
   looks locally like a shift in RA and Dec for the most part, which
   it should. However, there were some outliers observed! This needs
   investigation!

4. The subset of the USNO NOMAD catalog we have was made with the
   following search criteria:
   (is a 2MASS star) AND ( V <= 16.5 OR R <= 16.0 OR B <= 16.0)

   (See http://markmail.org/message/dz7g3wnzl2wmfyyk for more)

   There are noticable differences when different bands of magnitude
   are used to filter between 12 and 16th mag using USNO's NOMAD
   search facility:
   http://www.usno.navy.mil/USNO/astrometry/optical-IR-prod/icas/fchpix

   This could explain the difference between the POSS plates and
   KStars rendition of the same star field. Another possibility is
   that there are NOMAD catalog artefacts. In either case, it looks
   like the fault is not within KStars. Trixel 12310 raised my
   suspiscion because it seems to have a rich band of stars, but that
   artefact exists in the catalog as well, as is verified from the
   plots generated by USNO's website.

5. Despite setting magnitude limits to 30 on trixel 12310, some stars
   that were present in trixel 12310 were never displayed -- or at
   least JITUpdate() was never called on them. But these were very few
   in number.

Puzzles:

1. Some stars seem to have a wildly off value for (ra, dec) - (ra0,
   dec0) from the rest of the stars. It is unlikely that the cause of
   this is proper motion since faint stars have little proper motion
   owing to their distance from the earth. Looks like precession is
   not being applied uniformly to all stars.

2. Heisenberg stars: Some stars seem to suddenly disappear / change
   their positions when you try to zoom in on them. After a couple of
   tries, it usually reaches a "steady state" (an eigenstate :P). This
   will probably not show on my analysis since I dump every star
   that's JITUpdate()-ed, and not those that survive in the "steady
   state"
   I think this strongly suggests that StarBlockFactory's marking and
   block-disposal / reallotment scheme is the culprit.

3. Imprecise precession: More tests need to be done until the
   short-circuiting of precession is verified to work. Puzzle #1 may
   be related to short-circuiting of precession.

Maybe it is a good idea to permit the precession short-circuiting to
be something we can turn on/off from the advanced options.

I'll hopefully have some figures coming up.

Regards
Akarsh