Fig. 14

(a) Non-linearity correction function from Eq. 67 as a result of the minimization of Eq. 86. The fit parameters are provided within the graph. (b) The mean value of the corrected signal in dependency of the variance of the corrected signal and (c) the residual plot of the difference to the fit. (d) To show that the gain non-linearity correction actually works as intended, a second measurement was conducted by using the bracketed repeat-exposure (BRE) method. In contrast to (b), we analyzed the mean value of the image (without dark frame subtraction) as a function of the exposure time. To pronounce deviations from a linear model that we aim to archive, we used the region below 6600 counts to fit a sloped line to the respective data and subtract it. The uncorrected original data (yellow) shows a drop in value with increasing exposure time relative to a linear increase. After correcting for deviations in the beam current (orange) by the BRE-method, the drop is significantly reduced but still observable. Applying both, the non-linearity correction from (a) and the beam current correction, leads to the desired linear behavior, where the difference to a linear fit shows randomly distributed deviations with increasing time. A linear fit across the entire range of the data, shown as the black line, shows no significant slope within tolerances.