As explained in Section 5.1.4, the different instrument configurations (i.e. gates, window size and sampling) create effectively different instruments to be calibrated. If a sufficient number of sources were observed in different configurations, thus providing sufficient
mixing between the calibrations, the internal Gaia photometric system would be homogeneous.
In certain cases however, because of the small number of sources available for a certain configuration (for instance bright stars) and because of the accuracy in the on-board magnitude detection (which determines the different configurations activation), the mixing is not sufficient.
While the iterative process used to determine the photometric system should be able to handle this, convergence can be quite slow. We therefore introduced an additional calibration for the band and the and integrated photometry that
helps to speed the iterations, providing a link between the different window class and gate configurations. The links are computed from multiple observations of the same sources in different configurations and applied only at the first stage in the calibration process when deriving the initial set of raw reference fluxes needed to bootstrap the iterative process.
For a more detailed description see also Riello et al. (2018).