Re: systematic gradients

[aah@nofs.navy.mil]

Mike G. wrote:
>Unfortunately a poor flat field is not the only source of a gradient in
>declination.  I have noticed in both my and Tom's images a difference in
>focus between the high and low declination edges of the frame.  For a
>fixed size aperture a broader PSF means a relatively lower measured
>magnitude so one side of the frame may systematically measure lower than
>the other side.
  This is a good point.  The varying psf was discussed many moons ago, and
is one feature that I hoped a TASS site would investigate.  The fwhm in
the RA direction will vary, as the VCO rate is strictly true only for one
declination strip through the image.  The declination spread should only
be a function of the lens abberrations.  However, note that what Mike
says also applies to _all_ photometry in the image.  At least checking with
known stars like the Tycho catalog will show any systematic trend, whether
caused by the flatfield or the imperfect photometric extraction.  It really
doesn't matter whether the low-frequency error is caused by the camera system
or the photometric extraction; the main points are to find such systematics,
and then to apply a correction to remove any systematic effect.

>We also need to be careful in which Tycho stars we use.  Some of the
>stars have pretty crude magnitude measurements compared to the stars in
>the Hipparcos catalog.  I have been investigating the Tycho catalog and
>find that only 20% of the entries have both sigvt and sigbt less than
>0.050 magnitudes.  Using all the Tycho entries may give poor results
>compared to using a selected subset.
  There are a number of flags and error entries that should be checked before
using any Tycho star for either astrometry or photometry.  You should
set the zero point with Tycho stars that at least have both bt and vt, and
then weight them according to their magnitude errors.  The Tycho errors
are around 0.05mag at V=10.5 and degrade fainter than that.  For the flatfield
experiment, I'd use several frames from a photometric night and combine the
results to beat down the signal/noise.  We do that when measuring the POSS
plates, both to see the Schmidt astrometric distortions (a so-called
Taffogram), and to find the vignetting function for the telescope.  For
instance, for the UJ survey (2min unhypered plates at the same plate centers
as the POSS-II survey), we have 800+ plates that we stack the deviations from
to come up with both astrometric and photometric difference maps.  The main
thing you don't want is any variation within a given frame that is caused by
clouds or local lights coming on/off.
  Another systematic effect for wide field photometry is the variation in
atmospheric extinction due to the differing amount of air the light traverses
based on zenith distance.  The celestial equator is at a zenith distance
of 40 degrees for a 40 degree latitude site.  The difference in air mass
from the top of the frame to the bottom of the frame is 0.057, and a normal
extinction coefficient for the V bandpass is about 0.2 magnitudes/airmass.
This means that there will be 0.01mag errors based on airmass alone from
the top to the bottom of the frame.  For observers further north, so that
the equator is lower in the sky, or for camera systems that have a wider
field of view, this systematic error will be correspondingly larger.
  In other words, there may be lots of these systematic errors.  As long as
they are constant, they can be calibrated out.

Chris wrote:
>Could you use the above method interativly  In other words
>rather then just using the Tycho stars as a check could we
>use those stars to _compute_ a flat field?
  The current programs use a separate flat field for each frame, based on
median-filtering the current frame (if I remember correctly).  If you then
correct that flatfield based on photometry within the frame, then I think
you get into a particularly nasty loop.  For example, say that the night
is not photometric and you have thin clouds drifting through part of your
frame, then both the calculated flatfield and the photometric correction
are based on sky conditions, not on any physical constraint of the camera
system itself.  The ideal case is to create a single flatfield and a single
correction map that are applied to all frames taken during a given night.
  Astronomers often use dome flats to take care of the high frequency errors
of the flatfield, then take sky flats to get a low-frequency correction to
the dome flats.  TASS could do something equivalent if each site had a
flatfield projection system (like a dome flat) that gave identical flats
from night to night.  Then do a photometric 'taffogram' on one or more
photometric nights to get the low-frequency correction to the projection
flat.  The low-frequency correction is then a fixed item that is applied to
the nightly projection flat (which might change more often due to dust and
other dynamic events).
Arne
--------------------------------------------------------------------
Arne Henden                               Instruments/software/CCDs
US Naval Observatory Flagstaff Station    Cepheids/photometry/IR
P.O. Box 1149                             ftp: 192.68.148.67  
Flagstaff, AZ 86002-1149                  Voice: (520)779-5132
aah@nofs.navy.mil                         FAX:   (520)774-3626