In CCDs part of the charge released by an absorbed photon is lost during transfer to the readout node. This loss depends on several parameters, in particular on the position where the photon was detected, its energy, the temperature of the CCD, and the saturation of traps by charges preceding along the readout direction. In order to determine how these parameters affect the charge loss of the pn-CCD cameras, more than three billion events were recorded in extensive sets of calibration measurements from February 1998 to January 1999.
For deriving the CTI, each CCD column was split into macro pixels, which were determined by adaptive binning to contain a sufficient number of first singles around the line feature. The charge losses were then obtained, column by column, by cross-correlating a spectral template accumulated from all first singles of a column with spectra from the macro pixels of the same column, to yield a first estimate of the energy shifts. This value was then used to improve the template and the cross-correlation was repeated until convergence was achieved.
In order to obtain a general CTI correction code, a model based on the capture and emission process of electrons in deep level traps was developed. This model, described in detail in [], considers the dependence of the CTI on energy and temperature, and takes the effect of precursors and the presence of noisy pixels into account. It was calibrated with the spatially resolved charge losses obtained with the technique described above, and makes it possible to calibrate not only the spectral response of each pixel, but also the mutual influence of pixels along the same CCD readout column.
