Murray O writes:
With computer designed lenses today, extent of focus shift should be quantifiable for each lens at each aperture and distance, I would imagine. Any reason why the manufacturers can't include an in-body software correction for each lens? It would seem to be a relatively simple calculation.
DIGLLOYD: the computer design part is not relevant. Focus shift is something that can be measured accurately with the right equipment, how the lens is designed does not matter, though computer aided design can compute the focus shift, which is what makes the current Nikon prime lens designs so puzzling— focus shift that destroys image sharpness on a 36-megapixel DSLR, and is evident even on a 12-megapixel camera.
My understanding is that Hasselblad already accounts for focus shift in their H4 models of medium format digital cameras.
As I noted in my June 8 entry:
Surely Nikon (and Canon) could build focus shift compensation into the camera’s AF system firmware. To my knowledge, this is not implemented.
This would still raise some issues— manual focus behavior, focusing wide open then stopping down, etc. But for general AF use, an AF system with built-in focus shift compensation would be a strong marketing point, and reassure me that willful organization inertial ignorance about real-world challenges has been beaten back in the right engineering areas.
But since Nikon still takes the self-defeating approach of forcibly stopping down the lens while focusing in Live View (5 years or so later, and yes I did ask two people at Nikon to fix this, way back in 2008), my hopes are not high on seeing improvements to productivity-impairing camera features .
For that matter, Nikon still does not allow the mirror up prior to the count-down in self-timer mode (a simple firmware update would do it), so in my view, idiotic inattention to key usability details still reigns at Nikon, and so my hopes are not high for AF that compensates for focus shift.
Maybe it will be up to Canon to do this first!
Hasselblad does it right
Thanks to Alfred C for sending the a link to The Evolution of Lenses at Hasselblad, which states:
For fine detail of a given orientation, spatial frequency, and distance; the software can predict the contrast levels produced by the lens at any focus position and aperture setting. From these data points, tables of correction coefficients are calculated that allow the autofocus system to compensate for aperture-dependent focus shift.
An ideal lens would concentrate parallel rays of light to an incredibly small focus point, but in reality, spherical aberration turns this point into a fuzzy blob. Light rays entering a spherical lens far from the optical axis are refracted more strongly than those entering near the axis, so there is no common focal point.
Even when spherical aberration is corrected for, this correction will never be perfect; actual lenses will be either under- or over-corrected and modern lens designs will tend to over-correct. When stopping down, rays far from the optical axis are excluded and the point of optimum contrast will shift, resulting in back- focus with under-corrected lenses and front-focus with over-corrected ones.
The TTL auto-focus sensor uses f6.7, so without compensating for focus shift, focusing would be slightly off at larger or smaller apertures. With small apertures, the increase in depth of field is sufficient to mask this effect, but with a fully open lens, a high-resolution CCD might reveal a discernible lack of sharp- ness. Since the offset between the optimum focus position for a given aperture and the position found by the AF sensor with the lens fully open is known, an H System camera can easily compensate for focus shift, ensuring maximally sharp images at any aperture setting.
DIGLLOYD: Why is this critical feature not implemented for Canon and Nikon? Canon and especially Nikon DSLRs have as high or higher precision demands (pixel pitch) than Hasselblad H4 cameras.