Talk:VSOP Automatic Stellar Parameters

The automatic classification should run on the vsop machine. I don't see any other easy options for automation. In that case, we may as well start to coordinate things, check installations, install what is missing etc.

I imagine the procedure will be essentially the same as we do for STELLA, which is:
 * pre-processing spectra:
 * correct to RV=0 using CCF value (FITS header I/O)
 * extract regions to ascii files (FITS I/O)
 * estimate of vsini from CCF (FITS I/O)
 * rebin if high vsini?
 * processing (Carlos tool = black box)
 * post-processing:
 * take Teff, logg, [M/H] and write to wiki page (robot?)

Now, I suggest to use cfitsio and the Perl interface to that (I'm sure there's a Python interface too...). Do you know if that's installed? -- Tdall 15:19, 29 November 2006 (CLST)


 * There is a nice perl module to access fits keywords called Astro::FITS::Header, that interacts flawlessly with CFITSIO through the Astro::FITS::Header::CFITSIO; that comes with it. I'm already using it in my renaming script and it just works.


 * Since it does not come with the standard perl modules, I am sure it is not installed. I don't know which part of perl is installed on the vsop machine, but if we plan to use perl tools also on it, I strongly suggest that we check if the cpanplus module is installed, and install it if not. This will help a lot installation of future perl module if we need some. --

Choice of software is under the responsability of the one who will implement and debug it. I personally provide absolutely no support for Perl, as you may guess. Let me just mention that the interface between fits and python is one single file: pyfits.py (with the help of the numpy module), designed by the STScI and which provide access to the header and the (table) data under the form of an n-dimensional array. There is a priori no need for cfitsio.

But that's out of the point here, I guess. Except that interaction with the wiki is done through PyVSOP. So I could take care of post-processing in Python if it is relevant. --


 * I think it sounds easier to use python, and since I was already planning to learn it... well, let's wait and hear if Carlos has an opinion on this. In any case, I'll try to make some fits i/o scripts in python, just to see if it's really as easy as Cedric claims ;-) --


 * Cedric.... do you have some snippets of python code I can play with? Something that involves fits header manipulation? Or a link or something? -- Tdall 19:54, 5 December 2006 (CLST)
 * Little python snippet:

Pre-processing
A question for Carlos: would there be any problem with feeding the program with stellar templates that are rotationally broadened? My thought was to let it run as normal and reach a solution. Then, in an interval around this solution, make some broadened templates and let it run again (based on the width of the CCF). Do you think this would work, or even; would it make sense? -- Tdall 19:54, 5 December 2006 (CLST)