9.3.2 Treatment of Astrometric Field windows

A cubic spline interpolation is applied in each observed AF1-9 window and nine corresponding re-sampled (‘synthetic AF1-9’) windows are produced, constituted of (144 pixels AL $\times$ 1 pixel AC) with a pixel size$=$(4.91 mas $\times$ 2121 mas) for $G$ $>$ 16 mag.

For each transit, the signal in the nine synthetic AF1-9 windows are re-centred respectively according to the local plane coordinates (LPC) that position the windows onto the sky. The LPC value and the position of the maximum flux in the synthetic windows are compared and if the difference is $\ge$ 33 mas for quasars and 66 mas for galaxies (typical RMS of the differences over all sources and all AF windows) the window is eliminated.

For each transit, the median of the nine synthetic windows is calculated to produce the ‘median-AF’ window. This 144 pixel-long median-AF window is finally re-binned into a 12 samples window corresponding to the standard AF2-9 dimensions.

Considering the median-AF windows of all transits on a source, an iterative sigma-clipping approach is applied on each sample of the window. Eliminated samples do not participate to the fitting. For the six central samples a 4-sigma clipping is adopted to preserve the signal and for the rest of the samples a 3-sigma clipping is chosen. During this step, standard statistics on fluxes and position of maximum flux are derived. In case too many samples or the two central samples are eliminated, the median window is eliminated.