[Aavso-photometry] Large pixel size for photometry: what about defocusing?

[arne]

Matilde Fernandez wrote:
> Dear colleagues,
> 
>   we are working on the design of an infrared camera that will be used
> (among other projects) for photometry, hopefully with precision of a
> few millimag.
> 
>   We are right now discussing about the pixel size.
> 
>   From my experience, a pixel size that matches the average seeing does
> not give good results (as many or all of you already know) and a pixel
> size that is half the average seeing works much better. I have
> made some tests in order to show this effect and the results are plotted
> in the attached figure: it shows the dispersion of the data when the
> ratio seeing/pixel_size increases from 1 to 4. We did not reach very
> high precision because the stars were faint, but the point was to show
> the improvement of the photometry.
> 
>   Now my question:
> since some colleagues want a large field of view for their images, they
> ask for a large pixel size and suggest us to defocus in order to reach
> the photometric precision that we need. I wanted to ask those of you
> that have used this strategy. How well does it work? What precision do
> you reach? How much do you defocus?
> 
defocussing can improve precision, since you can gain greater dynamic
range as each star image occupies more pixels.  The ultimate of this
is MOST, which does pupil imaging.  However, defocussing means crowding
becomes more important, and for faint objects, all of the usual error
sources such as sky noise, dark current, readnoise become major
contributors.  Precisions exceeding 1mmag are possible for bright
objects - I've certainly approached that level before with the 1.0m telescope.

>   I am using (on well focused images) apertures that are only a bit
> larger than the FWHM (Full Width Half Maximum) in order to get 'the
> best' of each star, and this technique has helped me to increase my
> photometric precision. I am afraid that defocused images might not allow
> to use this small apertures, unless the defocused stars keep the shape
> of a Gaussian. Am I wrong?
> 
For any reflecting telescope, defocussed images are definitely not Gaussian.
You may have trouble accurately centroiding, and that will affect your
aperture photometry.  You might consider psf-fitting, as you can accurately
describe the weird profile that way, or you may have to use a larger aperture.
Personally, I am surprised that you are getting higher precision with one
fwhm apertures, as there have been several papers (and my own experience)
that says you need something around 4-5fwhm diameter for best precision
on well-exposed stars.  Certainly for faint objects, smaller apertures are
better since they include less sky, but defocussing will be a problem for
faint objects.
Arne