Re: Calculation of mag errors

[Stupendous Man <richmond@stupendous.cis.rit.edu>]

  Andrew Bennett went through the calculations of 
uncertainty in the magnitude measurement of a star
of V=12 on one particular image.  He found

> Magnitude s.d. = this x 1.0857 = 0.06698 mags

  I took his numbers for integrated flux from the star
and for the background, did the calculations by hand,
and end up with roughly the same overall uncertainty
of about 0.07 mag.  So we agree.

> The values 0.019 to 0.022 mags in the database are wrong.

  They certainly don't seem appropriate for a magnitude
V=12 star from this particular image.

  I looked at the sky values for other nights close to the
image Andrew mentioned (I looked in TR1 dataset, which has
dates JD 2452606 - 2452658, in Nov-Dec 2002).  
The V-band background sky values are typically 1900 counts per 
pixel at a minimum, so they image Andrew picked doesn't 
have a higher-than-average sky value.  The TR1 images have 
exposure times of 100 seconds.

  I happen to have some data from the Mark IV database on my
machine at the moment, which spans the time range from Nov 2002
to the most recent, late 2004.  When I look at the "uncertainty"
values for stars in the range V=11.5 to V=12.5, I see several
groups, as Andrew mentioned: one group at about 0.02 mag,
another at 0.04 mag, a few up to 0.08 mag.

  In this particular dataset (which covers 1 square degree near the
equator), I picked out all the stars with V=11.5 to V=12.5.
I then plotted the uncertainty versus JD.  Aha.  I see
that the values fall into clear groups by date: between JD 2452200
and 2452548, the values are around 0.04-0.05 mag.   
There's a clear change somewhere in a gap of data
between 2452548 (Sep 30, 2002) and 2452608 (Nov 29, 2002):
after 2452608, the values drop to about 0.02 mag.
Over time, they slowly rise, so that the most recent 
measurements have uncertainties estimated at 0.03 mag.

  Tom may be able to tell us if the nature of the observations
changed between Sep 30, 2002 and Nov 29, 2002.  Perhaps
the exposure time increased, or the lenses were re-focused,
or something like that.

  In any case, I would like very much to have a single sample
image which corresponds to some measurements in the database,
so that I could go through the analysis just as Andrew did
and compare the numbers with those in the database.  Unfortunately,
I don't think I have any of the raw or processed Mark IV images
from this period in my possession at the moment.  If anyone
does have sample Mark IV images from any time in the past three
years, and could send one to me, I would be happy not only to
do the checks, but also to put the image on my web site so
that it would easily be available to others, too.

  It's a good thing that Andrew is bringing this up -- thanks,
Andrew!  

                                  Michael