quick note on DS23 positions

[Stupendous Man <richmond@a188-l009.rit.edu>]

  I'm in the middle of analysis of the DS23 data in Tom's 
"collected.big" file.  A Tech Note describing my results will
eventually appear, but I thought I'd mention a bit about
astrometry now, since the E-mail list is discussing it.

  If one compares the individual positions of stars in the "collected.big"
file versus the mean position for each star, one gets some idea for
the _precision_ of the positions: how well can one repeat a measurement?
There can still be differences from the _true_ positions, but that's
another matter.  Anyway, I've done the calculations, and find for
bright stars (magnitude 8-10 in both V and I)

         stdev of (RA - mean RA)   =   0.30 arcsec  ~   1/25 pixel
         stdev of (Dec - mean Dec) =   0.29 arcsec  ~   1/25 pixel

  Arne wrote (in msg 000487)

>  I expect the
>  final Mark IV product to be ~1/30pixel astrometry (0.25arcsec)

Well, the _precision_ is there now.  I would be willing to put
money that, since the astrometric solution is based on Tycho-2
measurements, the _accuracy_ is not much worse, but I haven't
tried to show that yet.

  Recall that Mark IV pixels are about 7.5 arcsec wide.  If these
measurements have a gaussian distribution, then about one percent
will have an RA value more than 0.9 arcsec from the mean, and
(a different) one percent will have a Dec value more than 0.9 arcsec
from the mean.  If the two measurements are independent, then only
0.01 percent = 1 in 10,000 will have _both_ RA and Dec more than
0.9 arcsec from the mean.

  Remember, though, these are the bright stars only.  I'm sure that
the faint stars have a larger scatter in position; a pure guess
is by roughly a factor of two or so.

  The "twins" John Greaves mentioned were found like so:

> Most instances out of 45 I found by
> matching the 94,500 stars against each other on position with a few
> arcsecs leeway, had only a single observation for the "twin", and delta
> V twixt the two, though as much as half a mag in some cases, was mostly
> barely 0.2. 

  On the other hand, Tom Droege wrote that when he created the 
"collected.big" file, he placed together all objects within 12 arcsec
of each other.  Hmmm.  How could one not include as a match a star
which was within "a few arcsecs", given a 12-arcsec matching radius?
I dunno.

  Anyway, my candidates for creation of twins:

        a. two stars _just_ the right distance apart (around 3 or 4 pixels)
               that they are detected separately one night, but 
               together on all the others (due to changes in seeing and focus)

        b. a honking big cosmic ray in the outskirts of a stellar object
          
  I put at a much lower probability -- without any evidence --

        c. a faint variable object (e.g. CV) around 3 or 4 pixels away from a
               bright steady star; the faint object usually isn't
               detected, but one night it flares up and overwhelms
               the steady star.  The photocenter of the object moves
               to the fainter star's position, for this night only
               (I think this is the situation John Greaves was
                mentioning as a possibility)

  Anyway, that's a brief status report.  

  Arne also wrote:

> I expect the
> final Mark IV product to be ... 0.01mag photometry.

  Preliminary results are that the V-band data may be in this
ballpark, but the I-band data definitely are NOT.  There are clear
indications of systematic errors of up to 0.10 mag as a function
of position on the chip.  Tom, it would help if you would explain
how you chose to do flatfielding for the DS23 measurements:

         - used the night-sky images for each night to create
                    flatfields for each night individually

         - used the night-sky images for one or two good nights
                    to create flatfields which you applied to 
                    other nights

         - used a flatfield lamp to create flatfields which
                    you applied to many nights

  Back to the salt mines ...

                                        Michael Richmond