Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.eso.org/~pmoller/FIZZWIG-FAP/Team-meetings/2007-07-12 Дата изменения: Mon Jul 16 15:36:04 2007 Дата индексирования: Sat Dec 22 23:57:44 2007 Кодировка: Поисковые слова: п п п п п п п п п п п п п п п п п п п п п п п п

-----------------------------------------------------------------------------
Dear All,

this is my summary of today's meeting.

Wolfram
-----------------------------------------------------------------------------

We intend to use aperture photometry with aperture scaled by seeing. We
will test how well this works before making a final decision. Probably we
will use aperture radii about 1.7 FWHM*seeing, where the seeing will be
determined on each image.

Our favorite program is Sextractor, because of its good background
subtraction, and because Wolfram's programs are already using it. Carlo
prefers to write new code for the new FORS pipeline, but it will contain
an similar algorithm for background subtraction.

Nando will carry out some tests to determine the aperture correction, and
compare the Stetson magnitude ZP with such aperture compared to Landolt's
ZP. He will use both Sextractor and other programs to see whether there is
any discrepancy. His tests will be finished by mid-July, when Wolfram
has time to re-run the fitting.

After the new fits ( End of July / beginning of August), several of us
will work on the absolute photometry such as color terms and extinction.

We should exclude standard stars which have neighbors close enough to the
aperture to be used for 2.5 arcsec seeing, the exact criteria TBD. Wolfram
will put this into his code. Wolfram will also check whether there is an
effective maximum weight for stars when the flatfield correction is
computed.

Sabine's flatfield versus rotators angle look promising, but more tests
should be done which do not depend on the structure looking like a finger.
These tests should also clarify the significance of the finding. Palle
will write a detailed list of test to be carried out.
-----------------------------------------------------------------------------

Dear all,

Here the resolution of my action item.

The question is this: When we do the counter-rot of the frames, and then
co-add, how do we know that the features we see are true rotating
features, or simply remaining features seen in single frames?

In the perfect situation we have a large number of frames, evenly
distributed on rot-angles. In this situation we trust the law of large
numbers, and believe that the individual features are beaten by root(N)
where N is the number of frames. Since individual features are seen up
to 5-6%, and we are looking for features that are of order 1%, we must
require that N is much larger than 36, and that the histogram is flat.

Therefore, the first thing we need to know about the counter-rot-comb
flats is:
1) N = number of frames used for combination
2) histogram of rot-angles

I fear that in many cases there will either not be enough flats, and/or
the histogram will not be flat. For those cases we need other tests.

One test is to cut each sample into 2 halves, treat each half separately
and then check if they give the same answer or not. The halves could be
either ordered by rot-angle (e.g. 0-180 together, 180-360 together), or
totally random. The best would be to do both.

Another test is (if there is a dominating peak in the histogram) to
drop all the images of that dominating peak and only combine the rest.

I am not sure how much work all of this is, but for a start the
histograms would be good to see.

I think this is enough for now.

Cheers,
Palle
-----------------------------------------------------------------------------

Hi All,

I fully agree with Palle on the importance of watching the histograms. In
addition, I suggest that for any combined flat we always look at and
compare 3 versions next to each other, they should be shown with the same
scale. The three versions are:


1. The mean of all flatfields normalized by dividing by its mean

2. The flatfields divided by their mean, and counter-rotate with the
opposite of the rotator angle (as Sabine and Nando have shown before).

3. As 2., but choose the angle to rotate at random.


A feature which stays fixed on the detector should show up in 1, one which
rotates in 2.

Wolfram
-----------------------------------------------------------------------------