Re: more flatfields

[Chris Albertson <chris@topdog.pas1.logicon.com>]

aah@nofs.navy.mil wrote:
> 
> Chris wrote:
> >I think the true soluution is to take the above "sub-region method"
> >to the limit of where the "number of sub-regions approaches infinity"
> >In other words just accept that our images have continously varying
> >PSFs and gradients and process accordingly.  I have been arguing for
> >using 2D curve fitting with a  varying PSF model for some
> >time now.  I appears to me from recent posts that the assumption of a
> >constant PSF made by the current software cost us about mag. 0.02+
> >error.
>   If one uses an analytic form for the psf, then you could have that
> form depend on declination, and match what you are saying.  You only
> need a 1-D variation due to the drift-scan process on the mark III,
> though a true 2-D model would be required for the mark IV.  DAOPHOT does
> not assume an analytic model, however, and if you look at typical
> seeing-limited profiles, like King profiles, they are definitely difficult
> to model analytically.  

It looks like there is conciderable flexibility in the PSF model DAOPHOT
can use.  It can be purely analytic, analytic with a table of residuals
or if the analytic portion is define as a constant zero, purely table
driven.  The analytic component can have x,y as parameters or not.
Yep, there are about a billion parameters to set.  It may be slow but
still runs faster then real-time.  That is DAOPHOT will reduce a frame
faster then the exposure time so it could keep up.  For the MkIII I'd
try  varorder  =  1  or  maybe 2.

A description of how the PSF model is stored from the HELP file:

    The  output  PSF  image psfimage  is normally a 2D  image containing
    the image header parameters, "XPSF", "YPSF", "PSFMAG"  and  "PSFRAD"
    which  defines  the  centroid,  magnitude  and  size of the PSF, the
    parameters "FUNCTION", "PSFHEIGH", "NPARS", and "PAR#" which  define
    the  analytic  component  of  the PSF, and a single look-up table of
    residuals from the analytic fit, which  has  been  subsampled  by  a
    factor of 2.
    
    If  the  DAOPARS  parameter  varorder  =  -1,  the PSF is fit by the
    analytic function only and psfimage has no pixel file.
    
    If the DAOPARS parameter varorder  =  1  or  2,  then  two  or  five
    additional  lookup  tables  are  computed and psfimage is a 3D image
    with 3 or 6 planes respectively. The first  two  additional  look-up
    tables  contain  the  first  derivatives  of the PSF wrt the x and y
    positions in the image (varorder = 1), and the next  three  contains
    the  second  derivatives  with  respect  to  x  ** 2, xy, and y ** 2
    (varorder = 2).

--Chris Albertson

  chris@topdog.logicon.com                Voice:  818-351-0089  X127
  Logicon RDA, Pasadena California          Fax:  818-351-0699