In my previous post, I introduced the FITS format, which can be used to store spectral data. One of the aim of my GSoc project is to implement a reader for this format. The FITS format can hold a very diverse range of information. Also, this format has been around for more than 30 years and has evolved a lot in that time. The basic keywords in this format are well defined, but the new keywords for storing spectral data are not. This has made the format really messy. There are different keywords that imply the same thing as well as same keywords that imply different things. The documentation of this format is incomplete, there are minimal examples and no concrete definition to make things less ambiguous.
To tackle this issue, we decided to tackle the format, one specification at a time. The first target is long slit, multi-object or echelle spectrographs (one-dimensional). The reader being implemented follows the Image Reduction and Analysis Facility(IRAF) standard described here. Basically, the FITS files can store mappings from pixel coordinates to wavelengths (or dispersion). These mappings are defined using functions. To save space, a lot of these one dimensional mappings are stored together in one FITS file. For each sequence of coordinates, the mapping function is defined in the header. This kind of format is called multispec. There are eight possible dispersion function types that can be defined in multispec format:
To tackle this issue, we decided to tackle the format, one specification at a time. The first target is long slit, multi-object or echelle spectrographs (one-dimensional). The reader being implemented follows the Image Reduction and Analysis Facility(IRAF) standard described here. Basically, the FITS files can store mappings from pixel coordinates to wavelengths (or dispersion). These mappings are defined using functions. To save space, a lot of these one dimensional mappings are stored together in one FITS file. For each sequence of coordinates, the mapping function is defined in the header. This kind of format is called multispec. There are eight possible dispersion function types that can be defined in multispec format:
- Linear
- Log-Linear
- Legendre polynomial
- Chebyshev polynomial
- Linear Spline
- Cubic Spline
- Pixel array
- Sampled array
The first three were already implemented in the specutils package. I added the functionality for the fourth function and now I am currently implementing linear and cubic splines. I have been searching everywhere for files that store data in this format, but nothing has been found. Still, for completeness I am trying to decipher the format specification described, and implement a reader in case someone has these files. If you do you files for these formats, please get in touch with us!
Understanding these mathematical models has been very challenging, but fulfilling and interesting. Hopefully, with the contribution of developers, the reader will be ready by the midterms! Until next time...
No comments:
Post a Comment