An overview of the reduction

The first stages of the reduction are common to all types of IFU observation. In order to reformat the raw output of the IFU into a datacube it is first necessary to locate the top and bottom of the spectrum from each slice. The spectra are then extracted and formed into another 2-d frame in which the spectra are stacked in the order in which the slices are arranged on the sky. At this stage the spaces between the spectra are removed and the spectra are shifted in the dispersion direction in order to align them to approximately the same wavelength scale (the spectra are shifted by an integer number of pixels to avoid resampling). These positions and offsets should all remain constant, so no calibration data needs to be taken for this stage of the reduction. The data now look very like a conventional long-slit spectrum.

The subsequent sequence of the reduction depends on the type of observation and the recipe used. If the image is a flat-field frame then it is now reduced using the conventional spectroscopy flat-field reduction primitives. If the observation is the first in an object-sky pair then reduction stops at this point. If it completes a pair then sky subtraction is now carried out.

The precise alignment of all the spectra is done simultaneously with wavelength calibration. An arc spectrum is taken and reduced to this stage. Wavelength calibration of each row in the frame is done using the FIGARO iarc routine. The wavelength calibrations measured are then applied to subsequent IFU frames using iscrunch.

Once we have a frame in which all the rows have been scrunched to the same wavelength scale we can cut out the 2-d spectrum from each slice and use each slice as one (y, $\lambda$) plane of our (x, y, $\lambda$) datacube. Small shifts in y are required to ensure correct reconstruction of the image (the x, y plane), but this is constant and a predetermined calibration is used.