digital paradise said:
I'm trying to understand this graph. It is showing all 4 lenses need to be set from -10 to +10?
No, I just arbitrarily set to W = -10 and T = +10, and moved the zoom ring from the wide to the tele end in small increments taking a shot each time. I then looked at the recorded EXIF for each shot, which shows the focal length and the AMFA applied to each image, and plotted those values for each lens (each point is an image). The point was to test how the camera determines the AFMA value to use for focal lengths between the extreme ends of a zoom lens. It seemed logical that it would be a simple linear regression, but I wanted to confirm that empirically (unlikely, but maybe it would use the W value from the wide end to the midpoint of the zoom, then the T value from the midpoint to the long end, for example). I chose four lenses ranging from a 2x zoom (16-35) to an 11x zoom (28-300).
Practically, when I do AFMA on zoom lenses, I do both ends and some intermediate focal lengths. For 2-3x zooms (16-35, 70-200), I test one or two intermediate FL (e.g. 16-24-35 and 70-135-200 for those lenses, or 24-35-50-70 for a 24-70). For 4x zooms, I test two intermediate FLs (e.g. 24-50-85-105 or 100-200-300-400. You have to draw the line somewhere, so for the 28-300L I tested 28-50-100-200-300. Basically, I want to know that when I set W and T values, the intermediate FLs fall on that regression line. For example, my 24-70/2.8L II needs W = 0 and T = +5, and FoCal indicated +1 for 35mm and +3 for 50mm, which fall right on the regression line. If the lens had needed 0 at 35mm and +1 at 50mm, for example, then shots at intermediate FLs would not have the right AFMA and would be out of focus...and in that case, I'd have returned the lens for another copy and tried again.
Hope that makes sense, but let me know if it doesn't!