The Andor iXon-897 BV CCD camera had some properties we need to review
The RON was low (2 ADU/pixel in practise, 1ADU/pixel in the brochure ...)
The bias pattern was strong, as it is on thinned CCDs. The level of the bias was temperature dependent and as the camera was cooled in a thermostat loop the mean strength of the bias pattern had to be adjusted for this. We did this by using only bias frames taken just before and after the science images and scaling a 'superbias' field. If the superbias was representative of the actual pattern we are probably doing well _ but this in itself should be tested. We certainly have the data! All shout: Student Project!
Linearity: We have data that suggest that the CCD was not '99% linear' as all CCD brochures promise. As the evidence depends on the shutter being linear with exposure time we have to revisit this.
We had 'dark bands' matching the width of the Moon, in the readout direction. While we may have compensated for this by doing 'profile fitting' also in the row direction of the image we would like to know what was going on, and choose a future CCD accordingly.
CMOS: they are available in 16-bits (Andor) and colour (in DSLRs: expensive!). Now, what was the benefit of using CMOS instead of CCD? Need to compare linearity and readout speed.
Some aspects of the expensive Andor were of no practical use for us: ability to have EM - that caused the bias average bias to flicker by +/-1 counts (not pixels - the average!). Faster readout modes were available, but never used - they cut into dynamic range or gave more noise. An internal shutter was a possible option but was not chosen - so we had to rely on the dodgy external one! A possible coating and enhancement of the blue-sensitivity was not chosen - choosing it would have left the red-sensitivity unaltered, so why not get it?
Cooling with water was possible but never used as it would be just so much more plumbing to worry about.