[vsnet-chat 2087] Re: Plotting light curves

[aah@nofs.navy.mil]

Gianluca Masi asked,
>This consideration offers me another question: to perform aperture
>photometry, can I work in the 3*3 binning mode (with a scale of
>3"/pixel, each binned pixel 27*27 microns), or can this affect the
>accuracy of the results? This to limit further the tracking problems.
  Of course, a question with many caveats!  You are saying that, with
  3x3 binning, your superpixel scale is then 3"/pixel?  And that, with
  no binning, your pixel scale is 1arcsec/pixel?  What is your
  typical seeing?

Normally, 3arcsec pixels cause problems with photometry since there
is intrapixel QE variations from such things as overlying gate structures.
However, when you bin, the single pixel gate structure now becomes the
average of 3x3 (or 9) pixels and is much reduced, as long as the star
image is larger than 3arcsec.  You reduce your read noise by a factor of
sqrt(9) or 3, always an advantage when you are using short exposures and
so are readnoise rather than skynoise dominated.  It takes less time to
read the CCD, and less storage per frame.  The harder question in
my mind is how to center an aperture on your binned star image.  Digital
apertures always take square pixels and make a round circle, usually
by splitting the flux in pixels on the border.  When you make your pixels
big, then the border becomes much harder to fit properly and you start
making mathematical errors in determining the aperture magnitude.  Centroiding
something that falls entirely in one pixel is a nightmare.  Likewise,
if you are using psf fitting, you do not have enough profile structure to
fit the star properly.  Another consideration is dynamic range.  While the
3x3 binned pixels can theoretically hold 9x the flux that a single pixel can,
the serial register and summing well can usually only hold a maximum of
4 pixels of flux, and your signal chain may be set such that ADC saturation
occurs long before superpixel saturation.  In other words, you've concentrated
all of a star's flux into a single pixel instead of spreading it amongst many,
and so may saturate quicker.  Certainly, you can do reasonable photometry with
3arcsec pixels, but you may have to experiment to find the optimal pixel
scale and aperture size.  The rule of thumb is to use 2-3 pixels per fwhm,
and use an aperture 3-5 times fwhm in diameter.  Any time you deviate from
this, you may get great photometry or you may compromise your results -- you
just have to learn what works for your project.
Arne

VSNET Home Page

Return to Daisaku Nogami

---

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