3.3 Calibration models
Author(s): Claus Fabricius
The calibration models for the sky mapper (SM) and astro field (AF) CCDs must be sufficiently detailed and flexible to allow obtaining optimal image parameters for astrometry and band photometry, yet not so complex that it becomes unrealistic to carry out the calibrations. These devices are physically almost identical, but, from the way they are operated, they fall in three groups.
The SM CCDs provide large ( pixel; Table 1.2) images around each source, with the purpose to eventually map the surroundings. The SM CCDs are read in full image mode and with a pixel binning. They have therefore a high readout noise, and are in reality under-sampled. For bright sources ( mag), images saturate, while for the fainter sources ( mag), a further binning is applied before sending the data to ground (Table 1.2). Image parameters for SM therefore have only little or no weight in astrometry and photometry, but the devices must be calibrated to facilitate their future use.
The first AF CCD, AF1, serves the purpose of confirming the detections from SM, in order to censor detections caused by for example cosmic rays. All windows are therefore read with 2D resolution ( pixel; Table 1.2), at the price of a higher readout noise due to the larger number of samples. To save telemetry, the windows sent to ground for fainter sources ( mag) have their samples co-added in each line, and lose again their AC resolution (Table 1.2).
The following AF devices, AF2–9, are the workhorses of astrometry and band photometry. They are the CCDs providing the highest potential and therefore the ones with the more demanding calibration requirements.
A limitation of the calibration models in place at the present stage of the mission is that they are focused on the treatment of isolated point sources, and they may not be fully applicable to more complex sources.
As discussed in Section 3.4.2 and elsewhere data are processed both on a daily timescale within IDT/FL and then cyclically in IDU. While the calibration and processing steps are broadly equivalent between these two systems the algorithms can be quite different. The daily processing must be fast and stable whereas the greater resources available to IDU allow a more sophisticated approach. Indeed, the longer timescale used in IDU permits modelling of the unanticipated features which continue to appear as our knowledge of the spacecraft improves. In practice this had led to a divergence between the two systems. Since the IDT/FL models have been effectively frozen since Gaia DR2 the descriptions in the following section will relate to the cyclic models used in IDU. We refer the reader to the Gaia DR2 documentation for details on the models used in the daily processing.