I made a calibration file at each f-stop (5.6, 8, 11). This is an image made with the a Leaf supplied diffusor that covers the lens being tested. The shutter on the 24XL was adjusted to maintain consistent exposure. These uncompressed, raw *.mos files are now the calibration files for the 24XL.
As a test, I ran Custom Gain on the calibration file against itself (that is, the calibration file was loaded both as the calibration and as the file to analyze with the fall-off correction set to 100% to remove vignette effects). The output of this was a wonderfully even toned gray image without visible vignetting. (I can post this file, but there's not much to see.) So far, so good.
Next, I used this same calibration file and loaded the problem f8 exposured file as described in the first post, above). The result is an image with no vignette problems, but the cyan-magenta cast is still present. Not so good.
I think this suggests some interaction of lens-subject. The only problem is I used two different subjects, under different lighting, and still had the color cast issue. However, the one common potential problem is that I put the camera/lens very close to the subject in an attempt to totally blur the image, and I'm wondering if this could be the source of the color disturbances.
If the sensor has its own distortions, the calibration file, which is essentially subtracted from an image, should remove these non-linearities. The proof is that the calibration file when run against itself does makes a neutral, flat image. This makes for a good control. The same should hold true for the lens, and it does: vignette effects are removed. However, when shooting a reflective subject at too close of a distance, perhaps the calibration file might not handle it the same as when it is imaging a diffused light (ie, the calibration process). This only thing I can think is to re-shoot at a greater distance, maybe 1m (I think the lens is focusable from 0.6m to infinity), leaving a sharp subject plane, use Custom Gain again and see if the matter is resolved. Other ideas?
Thanks,
Robert