Builds a paramtric spectral model for the background of any instrument from a large number of example background spectra.
Uses PCA to learn average spectral shape and features and their correlations.
Works on the detector-level, completely empirical (does not go through the response). See Simmonds, Buchner et al. 2018.
- analyse.py bkg1.fits: plot spectrum
- compile.py pack.hdf5 bkg1.fits bkg2.fits bkg3.fits: Combine number of counts from many background spectra into a HDF5 file
- `(concat/combine).py outfile.hdf5 infile1.hdf5 infile2.hdf5 etc ` # take several hdf5 packs and combine them.
- compress.py <cmd> infile.hdf5
- create: make infile.hdf5pca.hdf5 -- find 20 most important pca components
- components: read file above and plot components
- showcomp: for each component show interactive plot
- check: for a few spectra, show its approximate pca reconstruction based on pca components
- repack.py pack.h5 # stacks spectra so they each have at least 10000 counts The output is pack.h5repacked.hdf5
- export.py targetdir make json files