Histogram of pixel readout values of 500 bias frames, in total about 5.25  ×  10⁸ pixels. The best fit model to the data, with and without the sCIC term, is over-plotted. The lower panel represent the relative error between the model with sCIC and the data, defined as δ = (d_i − m_i)/m_i, where m is the model and d is data. Because of the large number of pixels in the histogram, even small effects, which can not be fitted by a simple model, may become evident. A Fourier analysis of the difference residuals shows a clear peak of 62 counts at 16, which is probably from imperfections in the AD converter. The value from the readout amplifier is a 16bit binary number, and a period of 16 means that there is an imperfection in the fourth bit (2⁴ = 16). In the transition from the bell to the exponential arm there is a deviation in a small region, which can be understood as being due to the simplifications done to the equations to make them computationally manageable.