README
======

This module defines classes, that can be used as SHERPA models.
They fold the model spectrum with the COS line spread function (LSF).
Different classes are supplied for individual gratings, the following models are automatically created when the module is imported:
    
G130M: tabG130M, empG130M
G160M: tabG160M, empG160M
G140L: tabG140L
NUV: tabNUV

The convolution assumes that
    
- the x-axes has units of  Angstrom.
- the x-axes is binned such that one bin is 1 native COS pixel.

Caveats:
- This routine bins the COS LSF to integer pixels (some of the tables  give fractional pixels.
        
Example:
    
    ... load your data from file into wave, flux, err arrays ...
    
    sherpa> load_arrays(1, wave, flux, error)
    sherpa> set_model('tabG160M(const1d.c+gauss1d.g)')
    sherpa> fit()

tabG130M, tabG140L, tabG160M and tabNUV are based on the 
Tabulated Theoretical Line Spread Functions  
(COS ISR 2009-01, Ghavamian et al.).

empG130M and empG160M are based on the 
Tabulated Empirical Line Spread Functions
(COS ISR 2011-01, Kriss).

See [this link](http://www.stsci.edu/hst/cos/performance/spectral_resolution)
for more details.

The code has been tested with Sherpa in the CIAO 4.3, tests are written for py.test [see](http://pytest.org), version 2.2.


Format of input tables:
    
- The header line has `nan` as first element and wavelength values as column
  titles.
- Tabulated theoretical Line Spread Functions: These tables go from -n 
  to +n pixels, in the case of NUV in fractional pixels
- Tabulated Empirical Line Spread Functions: These go from 1 to 2n.
    
The tables are read such that the the resulting LSF is tabulated from
-n to +n assuming that the tables *give a symmetric region around the line center*.

author: Hans Moritz Guenther, hguenther@cfa.harvard.edu