RE: Post flat processing

["Gutzwiller, Michael" <mgutzwiller@lanvision.com>]

Arne and Michael had a lot of comments and questions so here goes:

Arne wrote

	Mike G. posted his proposed method of correcting flat fields.
In general,
	it looks good, though I have a few questions and comments.
	(1) I hope that this processing will be used *only* for already
acquired
	    data.  The far better approach in the future is to use a
flatfielding
	    system that doesn't need corrections.
I agree in principle though a proper flat has eluded us all as far a I
know.  Have any of the TASS sites found a good solution?
	(2) The process only works for photometric nights.  What are you
proposing
	    for the remainder of the nights?
I would expect to use the nearest photometric night's solution.
	(3) Why are you dividing the declination into 4 zones?  Do you
see evidence
	    of slope differences between the top and the bottom of the
camera?  If
	    so, I would assume a quadratic or higher-order fit to the
entire
	    declination coverage might be a better approach.
The slope does indeed vary in the lists I have examined.  The changes do
not look quadratic and I worry about the outlier's affects on higher
order solutions.
	(4) While the median linear least squares technique does a nice
job of
	    rejecting outliers, I'd also like to see weighting on the
basis of
	    Tycho sigmaV and Star sigma(mag).  Does that subroutine
permit weighting?
	    If not, you might consider using a more standard least
squares routine
	    and iterate to remove outliers.
It does not take the weighting into account.  I can look into adding
that in.  Most of the Tycho stars are in the 0.040 to 0.060 range while
the Star mag errors are usually in the 0.010 to 0.020 range.  The actual
error is therefore dominated by the Tycho error and by the Tycho color
transformation error for I.
	(5) Color transforms require iteration, since you need
flatfieldcorrected
	    instrumental magnitudes before you can obtain proper colors.
An
	    alternative solution, depending on how many stars you have,
is to
	    use Tycho (B-V) to preselect a set of stars with similar
colors to
	    use in your fits.
I was thinking of using the corrected magnitudes after the fit step to
calculate the color transformations since I had all the data anyway.
	(6) Since this adjustment is subject to change, you need to save
the
	    raw magnitudes in your output list as well as the adjusted
magnitude,
	    unless you can document clearly how to recover the raw
magnitudes
	    after the correction has been applied.
Very true.  My current intent was to keep the raw magnitude and add the
corrected magnitude to the list.
	(7) You may find that the correction is pretty stable
night-to-night,
	    depending on how the original flatfield was created, and
might be
	    able to use the photometric nights to correct the
nonphotometric
	    nights.
This is probably the only way to correct non-photometric nights.
	(8) Michael R. mentioned the problem of correcting the I-band
data.  Note
	    that my transformations of Tycho data yield B,V,R,I
magnitudes, so
	    at some level you do have I-band measures for calibration
purposes.
	    However, you will have to limit the color range, and the fit
isn't
	    nearly as good as for V, which means there will be
additional scatter
	    in your median-lsq fit which may require a larger number of
Tycho
	    stars.  Michael's suggestion of just using Landolt standards
for the
	    I-band correction won't work since the Landolt standards
don't cover
	    the entire declination zone (especially for Glenn's southern
	    coverage).
I do indeed see a degradation in the fit of I vs V data to the Tycho
catalog.  The V data shows a sigma of about 0.060 for V and about 0.090
for I.  This means that our correction for I will be poorer.  Landolt
star will not work since they don't cover enough of the range in
declination.  The Tycho I values may be estimated but they really are
the only data that cover the range.
	(9) There is definitely the chance of systematics arising from
this
	    process, but there are already systematics in the data from
the
	    improper flatfields.  You have to weigh one against the
other.
	    The best procedure IMHO is to get proper flatfielding on
future
	    scans, use those to set up the master photometry list, and
use
	    the earlier scans for differential photometry of any
discovered
	    variable stars, etc.  Or at least give future scans higher
weight
	    in the solution.
I agree.  The process is subject to systematic error, especially near
the edges of the images.  All in all I do think the correction should be
good to about 0.01 to 0.02 mag over the range of the images.

Mike G.