The new MIKE_blue CCD currently begins to go non-linear at a pixel count of approximately 8500 DN and is strongly non-linear by a pixel count of 10000 DN. (This is measured for single pixels .... i.e. binning of 1x1) The limitation appears to be in the pixel full well and not in the electronics, therefore, higher counts can be obtained -up to the digital saturation - by binning.

The folllowing table provides measured gain and read noise values at various count levels for 1x1 binned flat field observations using the internal incandescent lamp. The mean pixel count is measured in a region of about 30x70 pixels and has not had the bias level of about 420 DN subtracted.

Pixel_Count(mean[DN]) Gain(e-/DN) Noise(e-)
1592 0.46 1.68
2715 0.46 1.67
2898 0.45 1.62
3896 0.43 1.58
4306 0.47 1.74
4610 0.42 1.56
4850 0.46 1.67
5308 0.46 1.66
5960 0.44 1.61
7280 0.47 1.72
8086 0.51 1.89
8667 0.54 1.97
9080 0.64 2.33
9849 1.39 5.08
10300 3.28 12.06
11049 9.13 33.4

These values are taken from measurements across several flat field pairs of varying integration times. The pixel threshold value is approximately 11000 DN.

Test flats with an exposure time of 30 seconds have peak pixel values of about 5000 counts and have gain and noise measuremnets across the orders ranging from 0.44-0.50 (1.57-1.88) at count levels of 250-4500.


Bottom Line: MIKE_Blue is effective only for fainter targets. The current configuation may not be able to be used on bright target observations (once the single pixel count levels are above about 4000 electrons.