Clarification on photometry discussion

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

  There has been an exchange of questions on the proper method for 
measuring stellar brightness ...

  Richard said: "It is necessary to use very large apertures".

  Tom asked: "Why?  How will it improve Mark IV results?"

  Richard replied

> I am mainly referring to photometry where several images taken at differing
> airmasses are used in reducing data to the Johnson V magnitude system
> primarily.

> On the other hand if your data reduction methodology relies on reference
> stars included in a single image plus a knowledge of your transformation
> coefficients, then provided the image quality is quite uniform across the
> field you should be O.K. using 'cut-down' apertures.  I shall have a browse
> of the TASS technical papers and see what methodology is followed.

  Exactly right.  The method Tom is following with the Mark IV data
is the second choice: we use reference stars within each frame to 
calibrate the stars in each frame.  In essence, we are doing differential
photometry within a single frame.  As Richard states, in this case,
we don't need the humongous apertures.  

  The reason one needs really big apertures when one is comparing
star A in image 1 to star B in image 2 is that the PSF might have
changed between the two images (due to shifting winds, for example).
If the central 4-pixel-aperture contains 90% of the light of stars
in image 1, but only 85% of the light of stars in image 2, then
comparing small-aperture measurements will incur a 5% error
right off the bat.  Bad.  In a case like this, one typically chooses
apertures which are REALLY big -- 20 to 30 arcseconds; so big
that the amount of light outside the aperture is completely 
negligible in all images.

  Tom mentioned:

T> I have tried different apertures on the theory that we were losing light
T> from the brighter stars.  I found that increasing the aperture 
T> did not change the error for the brighter stars but did 
T> increase the noise for the fainter
T> stars.  This work eventually led to the discovery of the current flat field
T> problem which completely masks any other problem.  
T> When this is solved, then we need to go back and do aperture studies.  
T> I think Michael R. made a pretty good
T> choice in putting a 4 pixel radius aperture as the default in the pipeline.

  Partly luck.  The typical FWHM is around 2 pixels, and a decent
rule of thumb is to set the aperture radius 2-3 times the FWHM.

  Richard went on to say ...

>  and then try
>  comparing precision and accuracy of the photometry WITH and WITHOUT
>  flat-fielding.  One useful test is to image M67 and model the variation in
>  response (un-flat-fielded) across the chip as a polar diagram, i.e. a smooth
>  function of radius from, and angle around the optical axis.  Don't assume
>  the optical axis coincides with the center of the chip however - it usually
>  doesn't!  This type of approach, may I suggest, is an alternative to
>  flat-fielding in the conventional sense.  It of course largely misses the
>  localised 'donuts' and other dust motes but since you are drift-scanning
>  then you are tracking across the chip  ....

  The TASS Mark III _did_ drift-scan, and so one could try this method
on it.  There was a small difference in sensitivity from the top of
the scan to the bottom, which we did correct.

  The Mark IV camera does _not_ drift-scan, but takes snapshots.
The donuts are present and important.  I suspect that our current
procedure of making a night-sky flat is not a BAD thing to do,
though it surely does induce a small systematic error from the 
center to the edges of the field.  

  It would indeed be interesting to compare raw to flatfielding
measurements in the grid exposures.

                                           Michael Richmond