Test of the S/N for rising λ. The two lines represent the theoretical shot noise limit and the theoretical asymptotic excess noise of an EMCCD. It seems that the two methods are comparable up until about λ = 1 where the performance of the dead time correction deteriorates as expected. The theoretical curve has been calculated based on the Eqs. (25) and (26). The data set with the lowest λ has been obtained from a set of dark frames. The electrons in this set are mostly parallel CIC electrons but they follow the same statistics and are completely indistinguishable from photo-electrons.