Observations:
- When seeing is 0.9" or better, the focus parabola as measured by
mscfocus becomes flat-bottomed (well, actually, it becomes noisy) if a
focus step of <=100 units is used. Under these circumstances,
mscfocus becomes useless, if not actually dangerous -- it is trying to
get best focus by fitting a parabola to noise. Worse, mscexam uses
only the focus value with the smallest FWHM and its two nearest
neighbours; that is, only three points on the focus parabola are input
into the calculation.
- When seeing is worse than this, clean parabolae are generated and
mscfocus works just fine (provided that seeing is stable, of course --
if the seeing is varying while the focus sequence is collected, then
all bets are off.)
- If seeing is ~0.8" and a focus step of <=100 units is used, it
becomes difficult to do "focus by eye" - the focus sequences look
pretty much the same over several steps.
- FWHM is better on the left half of the mosaic than the right, and
there is a smooth transition between the two. (Best FWHM is to be
found near the ends of CCDs 2 and 6, worst in CCD 4). If the seeing
is ~0.8", roughly the same distribution of FWHM variation may
be seen across several 50 unit focus steps. This distribution is also
present in images of worsening seeing, but becomes less pronounced (I
see it in images of well over 1" seeing). Most observers don't see it
because they don't take the time to collect enough focus stars and
play with mscfocus.
- The best seeing I've seen in the last two mosaic observing blocks
has been 0.75" (in R).
Conclusions:
- The optical system is limiting the best delivered seeing to
0.75-0.8" in R.
- The variations in FWHM across the field reflect distortions
introduced by the telescope & corrector.
- The actual seeing can be better than 0.75-0.8".
Under these circumstances, changing the focus
across some range of values (of width dependent on actual seeing) will
change delivered seeing in an unpredictable manner because you're
sampling optical distortions, not seeing, and mscfocus is rendered
useless.
- The well-known "W" shape of seeing variation through focus is an
artefact of the distortions present in our particular optical system.
- Chris's technique of looking for the roundest images in the focus
sequences to find best focus may work well in good seeing, but the
correct method would be to fit the wings of the focus parabola for
out-of-focus image sizes >~0.8". We don't currently have the resources
to impliment this.
Therefore we recommend:
- Focus sequences should cover at least 500 focus units,
probably more, especially in good seeing.
- When the seeing is better than 0.9", there's a good chance mscfocus
is giving you a bad focus value.
Under these circumstances, I can
think of three ways to get decent (if not best) focus:
- focus-by-eye with a large focus step value
- Chris's roundest-image technique.
Do an "mscexam" and hit "m". If the resulting profile looks North-South
elongated like this:
then you must decrease the focus value.
- Collecting lots of stars in a focus image with ~100 unit
steps and aiming at the middle of the flattish/noisy portion
of the through-focus FWHM curves.
|
