Spectral extraction is implemented in the task omgrism, which runs over all sources in the source list (created by omdetect), cross-identifies the zero- and first-order spectra and decides which of them are usable for extraction. The output of the task is a FITS-file containing the spectra, each in a separate FITS extension. Omgrism allows spectral extraction using three possible algorithms: simple summing of counts above background in the cross-dispersion direction; one-dimensional Gaussian fit; and Optimal Extraction (Horne's algorithm [2]). The implementation of the latter method is still under development. Examples of spectra extracted by omgrism are given in Fig.6 (corresponding to the lower source in Fig.5) and Fig.8 (corresponding to Fig.7).
The coincidence loss correction is currently not applied (the algorithm is under developing). The background counts are scaled to the spectrum extraction region, smoothed along the spectrum and subtracted from it. The wavelengths are computed by using the CAL routine with the distances between the centre of the zero-order image and the current spectrum cross-section as the input for the CAL routine.
