I have given-up with CHDK DNG and ImageMagick.
I am using Reyalp's tool and looking at the data with ImageJ's histogram function.
It indicates the median and other values and automatically only uses the area of the histogram that contains pixels to draw the screen image.
If CS4 can do that I have not found how.
As I said above, I'm mainly using DNG because I'm testing on the SX10. Whatever you can get to work, as long as it is giving you the linear response expected.
Now, I need to make the histogram as narrow as possible by using a steady, monochromatic, very diffuse light source.
The histogram doesn't necessarily need to be very narrow, you just need to minimize over/under exposure. But a narrow histogram does help to give a wider testing range.
If you can't diffuse evenly over the entire image, you can always crop a central smaller portion from the images that is within the even illumination area and get proper results (as long as the cropping is equal from image to image).
Problem is, even if I find the true shutter values for my A620 the same factors may not apply to other cameras if it depends on how quickly the mechanical shutter can close.
This is true, absolute shutter (and aperture) values achieved are really only valid for the model tested.
Interestingly, I have also found that in really fast shutter settings, there is some variable functionality: For example, when I had previously done some shutter tests, I could reliably get exposure variation down to about 1/32000 (whether it is actually 1/32000 is yet to be determined) and typically beyond that, the exposure stays somewhat constant. But occassionally at faster shutter speeds, a single shot will show up that does have a lower exposure. I find this can be sometimes found if you set a static shutter speed and then take a series in burst mode. This may mean that occasionally, the camera will actually achieve a faster speed, but this is unreliable and can't be depended upon to be consistent.
