RE: Photometry

["Michael Gutzwiller" <deepsky@fuse.net>]

As the other Michael mentioned by Michael Richmond, I would like to share my
observations on what is affecting the photometry of the bright stars.  In
investigating the sources of photometric errors both Michael Richmond and
Arne are right.  I have been able to get the floor down to about 0.006 mag
standard deviation when the following conditions are met:

1) The Mark IV system follows a single field for as long as possible to keep
the stars in the same region of the CCD.  This reduces the importance of
flats and makes #2 possible.

2) The same comparison stars are used in each frame.  This is surprisingly
important, even when quite a few comparison stars are used.

3) The night is photometric or near to it.  This is not quite as important
than one might think unless the zero point is really bouncing around.


We may be able to improve on the 0.03 mag floor we are currently seeing if
we can reduce the number of comparison stars to those in the immediate
neighborhood so the same set can be used from different nights or if we can
compensate in some other way as Michael suggested.

Mike Gutzwiller

-----Original Message-----
From: owner-tass@listserv.wwa.com [mailto:owner-tass@listserv.wwa.com]On
Behalf Of Stupendous Man
Sent: Saturday, June 07, 2003 6:10 PM
To: tass@listserv.wwa.com
Subject: Re: Photometry



  Tom wrote:

> I quote from the Mark III paper:
>
> "We find the standard deviation from the mean has a minimum of about 0.03
> mag for bright stars, increasing to more than 0.2 mag for stars near the
> limits of detecton."
>
> I could say the same thing for the Mark IV data.
 ...

> With all these differences, we seem to get the same result.  There are
some
> common features:
...
>
> 4) Both software implementations were written by the same person.

  No, not true.  Mike Gutzwiller wrote the software used to clean
the Mark III images and turn the images into lists of stars with
positions and magnitudes.  I did a little bit of work to add
small color corrections at the very end of the calibration procedure,
but that was it.

> 2) Michael has somehow written into both his codes a floor of 0.03 mag.

  It would have to be "people named Michael have somehow written ..." :-)

> I can even rule out 2).  It is possible to improve the noise floor by
> tracking one field on a good night.  So it can't be the software.

  Agreed.  That is, without any post-facto corrections based on
repeated observations of the same field, the noise floor is more
than 0.01 mag.


  Michael Koppelman asked:

> As someone mentioned earlier, aperture and focal ratio may be a factor
> too. What are the differences in the systems in that regard? How about
> plate scale? S/N?

  The plate scale of the Mark III was 13.7 arcsec per pixel.
The plate scale of the Mark IV is about 7 arcsec per pixel.
So we can use smaller apertures on Mark IV images, including
smaller amounts of light from the sky.


  Doug Welch asked:

> Perhaps you explained this in an earlier message, but is this an internal
> error comparison? That is, is it the RMS scatter of the same stars on the
> *instrumental* system? It is easy to introduce an additional 0.02 or 0.03
> mags from using the (typically good but not perfect) standard systems
> available.

  The figure Tom quotes is an external error, in this sense:
we measure instrumental magnitudes of stars in an image, and use
the photometric solution for the night to turn these instrumental
magnitudes into calibrated magnitudes on an individual basis.
We then compare measurements made from different images, either
on the same night or, more commonly, on different nights, and
end up with a scatter of 0.03 mag.

  Some of this scatter is due to errors in our reference photometric
catalogue, which is the Tycho-2 catalogue.  I would _guess_ that
this portion of the error is larger in I-band than in V-band,
because Tycho-2 doesn't have anything close to an I-band magnitude
(we extrapolate from Bt and Vt).


  Tom agrees with me:

> I had best let Michael reply to this.  I am looking at what comes out of
> the end of the pipeline.  I believe this is "reduced to the standard
> system".  Our reference catalogs are "standard system" reference catalogs
> and I have been suspicious that they were part of the error from the
> start.  As you realize, there is a problem getting I data for such
> catalogs.  It has to be derived from other data, a problematic process.

  Yup.

> It also includes the error caused by measuring a star in different
positions
> in the field which in turn includes the error from using a different set
of
> reference catalog stars.  It is what you get in a practical way.
>
> If one puts a star in a fixed position in the frame and tracks it many
> nights, then one can get below 0.01 mag.  This is not practical for this
> survey.

  Right.  If you _are_ interested in changes of a single star, then
I would bet large amounts of money that you could take the measurements
Tom has made (selecting them from a database, say), and improve
them by

     A) selecting magnitudes for surrounding stars in each image
     B) building an ensemble of these surrounding stars
     C) calculating a magnitude difference (target - ensemble)
               for each image

  This will (I believe) yield much more precise differential measurements,
at the expense of moving them off the standard scale in some small way.

  When I receive Tom's disks, one of the several things I plan to do
is to look for errors as a function of position in the focal plane.
If I can find effects which are statistically significant, I can
create a secondary catalog which contains corrected measurements.
It _may_ be possible to decrease the scatter from 0.03 mag to 0.02 mag,
or maybe even 0.015 mag, at the bright end.  Only time will tell.

                                         Michael Richmond