The BRE method

BRE measurements of the non-linearity of the CCD system at the 74 inch Telescope at Mt. Stromlo Observatory were made in 1997 March. We used the 2Kx2K Tek chip (CCD10), with a nominal gain of 2.0e$^-$/ADU, with the B grating (dispersion 0.5Å/pixel) set up to look at the 6000Å to 7000Å wavelength region. A tungsten lamp was used as the light source.

Several sets of repeat exposures were made for the BRE method as described above, with the maximum number of repeats ranging from 10 to 16. For each set of repeat exposures, we plotted non-linearity (relative gain versus measured-counts) using five different regions on the CCD. The data for each plot were then normalised to the same value of relative gain at 10000 counts. Figure 1 shows the data from five sets of exposures with a quadratic fit to all the data points:

$${\rm relative gain} \: = \: \frac{N_m}{N_{t}} \: = \: 1 + 1.68 \times 10^{-6} N_m - 8.8 \times 10^{-12} N_m^2 \: ,$$ (1)

where $N_m$ is the measured counts and $N_t$ is the `true counts' or normalised expected-counts.³

Figure 1: BRE measurement of the non-linearity of CCD10. The figure shows relative gain versus measured counts, normalised to 1 at zero counts. The symbols represent the data from five different sets of exposures taken in 1997 March. The solid line is a quadratic fit to all the data points except those with counts above 64000 which are saturated.

There is a 7% change in the gain between 0 and 64300 counts when the detector becomes digitally saturated (the bias level is $\sim$1200 ADU and the digital saturation before bias correction is 65535 ADU). There are CCDs with non-linearities less than 0.5% across a similar range. In order to convert data to true counts, we divide the measured-counts by the relative gain.