I'm working through the Basic CCD Testing section 8.2 of Richard's Image Processing book, applying the tests to my Canon Rebel T5.
I'm thinking that I have to add some wrinkles to the testing:
- separate test series for different ISO's
- Debayer the test images and do each color separately
Other things I should look out for?
One thing I see with a quick test is that the darks look just like the bias shots...
I'm wondering if the Canon is doing some tricks for me that I need to understand.
I'd be grateful for any tips or suggestions .
Canon is indeed playing games with the raw images.
You will probably find, depending on how you convert CR2 images to FITS, that the bias level is 2048 or very close to it. Canon activity adjust the bias level to this value by using reference pixels around the edges of the image.
The dark will "look just like" a bias, but do the numbers. There are very few hot pixels, so to see them, you need to subtract a bias from the dark to see the difference. You'll get a scatter both sides of zero. Most of the scatter is noise, but the hot pixels will form a tail trailing off to high values.
Craig Stark has an interesting article which includes description of the strange behaviour of the dark current in long exposures in Canon DSLRs. It also has instructions on performing some tests. The paper is here: