# 5.1.2 Reference System

Author(s): Dafydd W. Evans

An essential aspect of the photometric data processing is the split between internal and external calibrations. The internal calibrations bring all observations onto the same reference system, while the external calibrations provide the transformation between this internal system and an absolute one that can be interpreted physically. This general principle is applied to both the flux photometry and the BP/RP spectra. The models used for the internal calibration are described in Section 5.4.1. The external calibration model is described in Section 5.4.1.

For the internal calibration of the fluxes ($G$-band and integrated BP/RP), the reference system needs to be set up. No external data is used in the generation of these reference fluxes. The reason for only using data from the satellite is that if ground-based data is used, seasonal and hemispheric systematic effects can be introduced into the system. Also, Gaia has the potential to provide data that has better uncertainties and sky coverage than any current survey.

The internal calibration is carried out in a bootstrap manner illustrated in Figure 5.4. The initialization of the reference system is carried out on a period of data that is not affected by contamination which makes the process more stable. For the definition of this INIT period in Gaia EDR3 see Fig. 5 in Riello et al. (2020). Initially, the reference fluxes are generated for each source by accumulating all the raw (uncalibrated) fluxes from the INIT period and generating weighted mean values. Using these as an initial reference, calibrations are carried out. There then follows an iterative loop where the calibrations are used in accumulating calibrated fluxes to generate a better set of reference fluxes and the calibrations repeated.

This method converges since the observations for the sources have different calibrations applied to them and provided that each calibration is carried out with different sources. Given that there is good mixing between the calibrations and sources, i.e., more than half of the sources are observed in two or more configurations (CCD, Gate, FoV, …), this process should converge quickly.

In previous releases, a Time Link calibration was computed to reduce the impact of systematics due to varying contamination during the initialisation of the reference system. This was not required in the latest processing run as an extended period with very low contamination was available for the initialisation of the reference photometry.

A similar scheme is applied for the instrument calibration of the BP/RP spectra in that there is an iterative loop between the instrument calibration and the source update process which creates the reference spectra.

Further details can be found in Section 5.4.1 and Carrasco et al. (2016).