Improved photometry

[Michael_Gutzwiller@AICI.COM (Michael Gutzwiller)]

To all,

I've been trying to get better photometry for my images and have some
improved results.  To deal with the slopes I have been seeing in my
images I refit the instrumental magnitudes with a robust line fitting
algorithm based on the medfit routine in the Numerical Recipes in C.

Essentially this algorithm finds a best fit to a line in a manner
similar to using the median as a robust fit for the mean.  This makes
the fit much less sensitive to outliers in the Tycho database.

If you wish to use the algorithm in the Recipes book, be aware that I
found a bug in it and have submitted a bug report to the authors.

Results:

Based on nine photometric nights I have obtained the following standard
deviations on the measurements:

            refit set   prev results
magbin     num   sigma   num  sigma
7.5-8.0      4   0.029    12  0.057
8.0-8.5      9   0.031    10  0.047
8.5-9.0     10   0.033    34  0.061
9.0-9.5     14   0.038    45  0.054
9.5-10.0    13   0.045    78  0.054
10.0-10.5    4   0.046    99  0.064

The previous results are from Arne's post on April 11 for results from
several nights analysis.  As you can see the refit set has lower sigmas
though not as low as I would like.

It looks like I at least cracked the 0.050 sigma barrier but have run up
against another at 0.030 sigma.  This could be for several reasons.  
First there are several outliers in the Tycho set one of which is
measured consistently .3 magnitudes brighter by TASS than by Tycho and
another that sure looks like a long period variable.  While the line fit
is less susceptible to outliers, its not immune.

Second, TASS should be able to measure magnitudes much better than
Tycho.  Already the TASS average magnitudes for these 54 stars are
measured to better than 0.015 magnitudes.  I envision being able to
create a TASS catalog for use in fitting which should get better results
than the Tycho catalog.

In any event I think we can be confident of at least getting to 0.030
sigma magnitude measurements.

If you would like to see the results, I have placed two files on
storm/incoming, refitpsf.xls an Excel spreadsheet and refitpsf.txt a
text version of the same.


Mike Gutzwiller