[Aavso-photometry] Re: PT Sampling question

[arne]

BailyHill at aol.com wrote:
> Happy New Year Arne;
> 
> I had a question about the sampling requirement of 2-3 pixels per FWHM.  
> 
> I can see that if one is doing PSF photometry, that this would be a 
> requirement.  One need spacial resolution that meets nyquist(sp?) criteria so that the 
> original data can be reconstructed from the digitized measurements, without 
> aliasing.  
> 
> When doing aperture PT, assuming that one does not get 2 stars in a pixel, 
> not sure why the same criteria would apply.  Seems like we are just summing the 
> counts.  Most of the time the aperture is larger than the airy disk, so we are 
> not sampling nor excluding the outer rings of the airy disk.  
> 
> I suppose its good practice, so that one can use either reduction technique.  
> Are there other reasons?
> 
> Feel free to post to the group, if you feel this is a worthy question.
> 
There are several problems:
  - each pixel is not uniformly sensitive across its area.  There are overlying
gates, antiblooming gates, microlenses, etc. all intercepting a piece of
the incoming light.  By spreading the star profile over several pixels,
you average most of these effects.
  - there are nonuniformities between pixels.  This is one of the biggest
things you remove with flatfielding, but flatfielding is not perfect.
By using many pixels in the star profile, you average out most flatfielding
errors.
  - pixels are square, star profiles are round.  By using more pixels per
star profile, you already make the circularization process easier.
  - if you put all of the light into a single pixel, you decrease your
dynamic range.  A single pixel can only have 16bits of digitization; using
say 8 pixels in the image, you spread the light out so that you effectively
have 19bits of digitization.
  - as you state, more pixels help in fitting analytic profiles, not only
for psf-fitting but also for centroiding/astrometry.

I can come up with a few more, but I think these will give you the basic idea.
Arne