Re: [TASS] Flat fielding

[Andrew Bennett <andrew.bennett@NS.SYMPATICO.CA>]

On Fri, 10 Sep 1999 09:17:36 -0700, "Arne A. Henden"
<aah@NOFS.NAVY.MIL> wrote:

>Andrew,
>  We discussed flatfielding to endless lengths about
>a year ago.
Copious newbie apologies for rehashing old ground.

>  The possible gradients in the flats is
>well-known.  The solution that was decided upon was
>to flatfield in the usual sense, and then apply a
>correction factor based on photometry of Tycho stars
>within the field.  Mike G. can discuss his ancillary
>program that does this.
Timid newbie suggestion that somebody who nearly
understands what is going on expands the description
of the software. When these descriptions are written by
somebody who really understands, they leave out a lot of
"obvious" stuff! I need FlatComp for Dummies.

>  You have to flatfield if
>for no other reason than to remove the vignetting.
I presume that the option of using a fixed correction
for vignetting and no separate "flat" image was
considered and rejected?
The current "flat" image introduces some column to column
noise (hopefully negligible? - if it wasn't you would be
taking more flats and averaging, right?) plus the N-S
gradient which is taken out again (hopefully perfectly
or reasonable facsimile?) by Mike G.
If the fixed correction were based on the average of
a sufficient number of images with a sufficiently
perfect (ha!) light box, it would add negligible
column to column noise and N-S gradient but not
remove wandering dust bunnies.

>In general, the flatfielding of a drift scan system
>is far superior than for a stare-mode system since
>everything is averaged over the entire column length
>of the chip.
Yes indeed. And conversely, the Mk IV is a horrible
thing to flatfield. Michael Richmond's M31 (TN0047) is
a thing of beauty compared to my first attempt ... or
my second ... or my final attempt before I quit. And he
describes his processing as "less than perfect". Time
to pull out that perfect light box and generate some
perfect flats!

>  Five percent standard deviation is acceptable,
Never!
>considering that the internal error of any site is
>on the order of three percent and you are combining
>several sites.  This error reflects color transformation
>errors, flatfielding errors, nightly zero point errors,
>errors in the frame zero point from Tycho stars, errors
>in the aperture magnitude calculations, etc.
Techniques such as those described by C. Alard &
R. H. Lupton, astro-ph/9712287, ApJ 503, 325 1998
claim to reduce the relative error for two
measures of the same star to close to the photon
limit. Has anybody tried this with TASS data?

>  What *does* worry me is that the mean magnitudes and
>colors in tenxcat are not asymptotically approaching
>the Landolt magnitudes.
Um. Er. They've been fed in twice and still fail to
emerge at the far end? Statistician needed!

> Likewise, the finding that some
>bright stars are being missed quite often needs to
>be studied.
Understatement. I will so employ my copious spare
time. Don't let that stop anybody else trying. Despite
my frequent statements to the contrary, I am not
infallible.

Andrew Bennett, Avondale Vineyard, Nova Scotia, Canada.