[vsnet-chat 857] Re: PSF-fitting or aperture photometry (Richmond)

[Taichi Kato <tkato>]

Some practical view on PSF-fitting or aperture photometry

   As Michael Richmond wrote, a typical FHWM of 1.5-2.0 pixels is too
undersampled for conventional PSF photometry.

   Some more to add to Richmond's comment:

   For faint stars, aperture photometry with a fixed aperture (to be
optimized for brighter targets) will have low S/N owing to the strong sky
contamination.  Aperture photometry for faint stars therefore requires
a smaller aperture size, something like "adaptive" process.  This process
may introduce a systematic bias between faint and bright targets (if different
aperture sizes are respectively used).  It is a well-known pitfall for
photometry beginners to use a small aperture for faint stars, a large
aperture for bright stars, and calculates differential magnitudes from
resultant total counts.  Another pitfall of faint-object aperture photometry
may be the centerling problem.  Due to the two reasons -- the smallness of
the optimized aperture for faint stars and the noise affecting the centering
accuracy -- aperture photometry on automatically detected faint stars tend
to be biased to brighter values.  For faint objects, aperture centers should
be placed at "prescribed" coordinates, rather than eye estimates or
automatically searching for the peak.  These systematic effects should be
carefully removed.

   Even for undersampled images, I would recommend to use aperture and PSF
photometry at the same time, and compare the results.  PSF centering should
be, as in aperture photometry, at prescribed coordinates, rather than by
fitting results.  This prescription reduced two degrees of freedom, thus
providing less scattered and biases results.

   Some more thoughts on small FHWM:

   Intrapixel CCD response is not uniform (even for "astronomical" quiality
ones).  Owners of CCD equipped with ABGs should extremely care for this,
since ABG pattern on the CCD surface non-uniformly absorbs the light.
Experiences with ABG-equipped CCDs and this small FHWM have shown up to
0.2 mag variation depending on the "intrapixel" location of the image
center (even worse than visual estimates!).  FWHMs at least 3 pixels are
recommended for more accurate photometry.

Regards,
Taichi Kato

VSNET Home Page

Return to Daisaku Nogami

---

vsnet-adm@kusastro.kyoto-u.ac.jp