[vsnet-chat 4442] Re: MISAO color coefficients

[Brian Skiff <bas@lowell.edu>]

     Thanks to Yoshida-san for the additional plots and information about his
photometric reductions.
     Just to confirm it for other readers, the plots at the URL given indeed
show the basic process one uses to determine color terms in a CCD or single-
channel photoelectric system.
     The scatter in the individual measurements on each plot is disturbing.
If I got such plots in my own data I would assume it was cloudy and slide it
off into the trash.  I can think of two reasonable explanations for this.
The obvious one is that it was not stably photometric, but presumably this
would not happen on all the nights.
     The second derives from the comment that the photometric reference was
to Tycho (1 or 2?) stars fainter than mag. 8.5---but as faint as mag. 13,
the implication is that the plots involve stars right to the catalogue limit.
Bear in mind that fainter than about mag. 10.5 or so, the errors on the Tycho
photometry increase rapidly, and are of a size to account _entirely_ for the
scatter we see in the plots.
     Another feature common in the plots is a fairly well defined base-line
of points, but with a fair fraction scattering off mostly toward the top.
Given the use of Tycho photometry, I would bet that the stars in the relatively
well-defined band at the bottom are brighter ones where the errors are small;
the rest are the faint ones with lousy Tycho data.  That base-line band 
probably represents the true color term, but it is masked by the scatter in
the data.
     Finally, I notice that it seems the 'k' values seem to have been
determined without omitting the outliers in the plots.  For instance, 'k' shows
a shallower slope when there are lots of points above the main ridge.  It would
be useful to plot the actual fits to see if they lie among the main string of
points.  Certainly don't be afraid to trim out the bad points, which could be
due to crowding or simply an especially bad Tycho  magnitude.
     I would recommend omitting any Tycho star with errors of 0.05 or greater
from the analysis (or perhaps > 0.03 if this leaves enough points to do the
fit---two dozen stars with a good range in color is sufficient).  This will
probably clean the plots up considerably, and make the fits to 'k' more
consistent.  One should use Tycho-2 for this, because Tycho-1 contains a scale
error at the faint end such that stars are registered too bright by several
tenths of a magnitude, even on the internal Tycho system.  Note that the trend
of the plot scatter being mostly upward (i.e. toward smaller delta-magnitudes)
is consistent with the "too bright" systematic errors in the old Tycho-1 data.
The Tycho-2 data also are simply better than in Tycho-1, and reach with good
errors down to between V mag. 10.5 and 11 depending on the field involved.
The errors on the photometry are given with each entry in the source
catalogues, and so the stars should be selected individually using these
errors, rather than simply omitting everything fainter than some fixed value.
     I am sure Arne Henden will provide additional comments.

\Brian

VSNET Home Page

Return to Daisaku Nogami

---

vsnet-adm@kusastro.kyoto-u.ac.jp