The GPI IFS data reduction pipeline meets the following top level requirements:
The data pipeline comprises several main components, all written in IDL:
Reduction tasks are defined by Recipes, which list both the specific reduction steps (primitives) and the data files to apply those steps to. When running, the DRP constantly monitors a queue for new recipes to reduce. During observations, an Autoreducer generates quicklook reduction recipes on the fly as data is taken. After observations, the Data Parser and Recipe Editor can be used to create and adjust additional recipes. Reduced calibration files are stored in a special Calibrations Database allowing appropriate calibrations to be retrieved automatically at each step of data reducion. As the pipeline processes data, it can also display plots or figures on screen, or send data to the GPItv display tool for interactive exploration of images and spectra. A Status Console provides a status display of the data reduction process, and keeps a log of all messages generated during reduction.
The most complex step in the data reduction is the conversion from 2D raw IFS frames to 3D datacubes. Because the light from each lenslet in the IFS is dispersed across many detector pixels, the process of uniquely assigning flux from detector pixels back to GPI’s field of view is complex. This is called ‘Data-cube Assembly’. The overall process is analogous for both spectral and polarimetry modes, though algorithmic details differ.
For each mode (prism+filter choice), a map of the reimaged lenslet array geometry must be made based on observations of calibration lamps (e.g. the locations and wavelength solutions of the ~37000 microspectra). This calibration data is produced from arc lamps for use in the extraction process. Obtaining the necessary lamp data is expected to be part of Gemini’s facility calibrations rather than requiring calibration by individual users. Photometric, spectral, and polarimetric standards must be observed at night in the usual manner. The GPI DRP is designed to reduce all of this calibration data, and to reduce scientific data to the level where an astronomer can begin custom analyses.
The GPI DRP includes primitives for PSF subtraction using Spectral and Angular Differential Imaging, implemented both using least-squares (LOCI) and principal component analysis (PCA/KLIP) algorithms. But it is expected that many uses will wish to use PSF subtraction methods of their own devising on the output data cubes from the pipeline. Likewise the GPI DRP includes primitives for polarimetric differential imaging.