Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.eso.org/~chummel/oyster/lab/imaging.html
Дата изменения: Mon Nov 26 08:49:01 2007 Дата индексирования: Fri Feb 28 11:57:49 2014 Кодировка: Поисковые слова: south pole |
Imaging optical interferometry data is the fun part! See a double star with 3 milliarcsecond (mas) resolution, or image a stellar surface!
Here, we are interested in what Mizar looks like based on the data we just reduced and calibrated, and then in determining the orbit. Imaging expects the least number of assumption to be true about your source, like positivity, and therefore will show you stellar structure no matter what it looks like. However, this requires a lot of good data! Modeling, on the other hand, assumes you know the structural elements for which it returns the parameters. It can work with much less data.
If you are unsure about the quality of your calibration of this data in the last step, just use the provided files to continue.
hds_close
get_data,'1997-05-01.cha'
HDS-file opened; mode=READ.
% Compiled module: STDEV.
GeoParms loaded.
GenConfig loaded.
Warning(DAT_FIND): component not found (METROCONFIG)!
Scantable created.
% Compiled module: UNIQ.
StarTable created; number of entries = 8.
Finished reading catalogs.
Finished reading catalogs.
Number of new diameters added: 6.
Number of new (R-I) values found: 0.
Finished reading catalogs.
% Compiled module: POLY.
Finished astrometry computations.
Reference station set.
Scans loaded.
Warning(VISEST): some diameters = 0 or not found in table!
Finished visibility estimation.
CONSTRICTOR log read.
Observer log read.
Imaging interferometry data presents us with two challenges: the aperture plane (i.e. the complex visibility function) is sparsely sampled, and the baseline phases are "totally corrupted" by the atmosphere while the closure phases, i.e. the sum of three closed loop baseline phases, is free of systematic errors. The two algorithms to take care of these challenges are deconvolution and phase self-calibration.
The deconvolution algorithm we use is the CLEAN algorithm. In each iteration, it correlates the dirty map (i.e. the Fourier transform of the visibility data) with the dirty beam (i.e. the Fourier transform of the uv coverage) and place a so-called CLEAN component (i.e. delta function type pixel) where it found a maximum correlation. It is based on the intuitively understandable assumption that a structure in the dirty map which looks like the dirty beam is caused by a compact component convolved with the dirty beam rather that a complex one.
There is fewer phase information in the closure phases than in the baseline phases. That's why model phases have to be substituted for the missing information in order to convert the closure phases into baseline phases for the Fourier transform. This process is called "phase self-calibration". Since the model changes after each CLEAN, we have to iterate this process.
OYSTER uses the difference mapping algorithm, originally invented at Jodrell
Bank (UK). Basically, this algorithm assembles a model (i.e. image) and a
self-calibrated set of phases in increments, each one preceded by run CLEAN
just for a few iterations on the residual map of the previous cycle. This
allows the user to inspect the residual map for areas of unCLEANED flux, and
possible sidelobe structure resulting from calibration errors.
Continue