Re: It's The Optics Stupid!

[Stupendous Man <richmond@a188-l009.rit.edu>]

> Yesterday I had Dan make aperture masks for the I lenses. 
     ..
> In the past, the V camera has always had significantly less spread in 
> errors than the I camera.  See for example, Figure 3 and 4 from TN-88.
> 
> From this short run, the I data has less spread than the V data, and is 
> about half the value of similar runs taken earlier this month.

  Rah!

> Is this a scatter problem???  The flat field should correct for optical 
> gain unless light from a point source produces a different flat than one 
> from a diffuse source.  Optical experts might comment.  Is this a standard 
> problem?  

  Yes, it is.  In theory, if there is some vignetting, then stars (and sky)
at the center of the field receive light from the entire lens, whereas
stars (and sky) near the edges receive light from only a fraction of
the lens.  However, the fraction should be the same for diffuse (sky) and
point (star) sources, so ordinary flatfielding _should_ bring stars 
of identical brightness across the entire frame to a common value.
The scatter in measurements near the corners will be larger, of course,
since fewer photons strike the CCD there, but the average value should
match that at the center.

  However, if there is any scattered light in the system, then 
point sources and the diffuse sky (may) no longer behave the same
way.  Suppose that, say, 30% of the light striking the east side
of the interior of the barrel bounces off the wall and onto the 
CCD, near the eastern edge of the image.  The reflected light 
will probably be spread out over a large area, which makes the 
"sky" value larger.  The light from a star falling near the eastern
edge of the frame, on the other hand, still contains the same number
of photons as always.  In other words, the "sky" appears BRIGHTER
than it ought to, relative to the stars.  If one makes a flatfield
from this frame with scattered light, and then divides a data
frame by the contaminated flatfield, then stars in the eastern area
will appear FAINTER than they ought to ... because their light
is divided by an improperly large "sky" value.  

  So, a systematic source of scattered light will cause stars in
certain regions of a chip to be fainter than identical stars
in other regions.

  It's not clear to me how stopping down the aperture reduces
scattered light.  Let's see ... if you didn't have any sort of 
"blinders" on the front of the lenses in the past, then light 
from a neighbor's house (for example), which is ostensibly far
to the side of the true field of view, might bounce off the 
front surfaces of the lens and the walls of the tube.  In that
case, an aperture mask might help, simply because it would eliminate
some (or all) of the incident light from far to the sides of
the field.

  On the other hand, if you _were_ already using some lens shades,
then, um, hmmm.  I can't think of a simple way to explain the improvement,
unless it might be errors in the shape of the lenses near the edges.
If the lenses aren't properly formed, so that light passing through
the outer edges doesn't come to a focus properly, I suppose it might
give significantly different PSFs in different parts of the frame.
I do see a relatively big shift in the PSF shape across my images,
actually.  The simple aperture photometry used by the current pipeline
can cause systematic errors if the change in PSF shape is strong
enough, and a large fraction of the light falls outside the aperture.

  Tom -- you can check this idea in the following way: examine
old and new (pre- and post-mask) images in I-band.  Pick two or three
stars in the center of the frame and in each of the 4 corners.
Measure the size and shape of the PSF of those stars -- you can fit
gaussians to the profiles, or, heck, just move your cursor around and
count pixels.  If you see that the PSF is more uniform across the
frame in the new (post-mask) images (and I'll bet it is), then I'd
guess a large portion of the blame can be laid at the feet of 
the PSF ... and, hence, of the lenses.

  To be fair, we _are_ using a pretty big field.  It isn't easy
to manufacture and assemble lenses which give good, uniform PSFs
across a 4-degree field.

                                         Michael