It’s sad to see what might otherwise be a very fine lens degraded by such a severe handicap for practical use, at least when high quality results are desired on high-res digital.
UPDATE: I’ll soon post a follow-up study comparing the Nikon AF-S 28mm f/1.8G to the venerable Nikon AF-D 28mm f/1.4D, along with as-focused and optimal-tweak variants for both.
Tim A writes:
A quick note to encourage you to look for another sample of this lens: I ordered it yesterday about an hour before you published these results and was all ready to return it but when it arrived today I stuck it on my 800E and tested it at a variety of apertures, distances, subjects and POF and I must say that as long as I use my centre AF point it is a peach. Testing it with a lens calibration chart shows that mine carries 80% of its DOF forward wide open, but by F5.6 that has gone to 80% rearwards: so a well calibrated example keeps the subject in focus.
I have found that at F5.6 I can make it work well at all subject distances and it will make a good landscape lens. It does have field curvature but in what I consider a good way: it lets me keep the foreground in focus whilst the distant centre is also in focus but the distant peripheries are OOF. apart from a slight weakness at very top corners, I can make a three metre away planar subject fully and sharply in focus. One thing: turn off lens corrections in LR if you are using them: they 'stretch' the data on the vertical thirds and create smearing.
DIGLLOYD: “mine carries 80% of its DOF forward wide open, but by F5.6 that has gone to 80% rearwards” is a confirmation of what I have shown: severe focus shift. It’s also possible that this “forward” depth of field is deliberate on Nikon’s part, knowing the focus shift will push the zone of sharp focus rearward when stopping down.
The next sentence is a contradiction: one cannot calibrate a lens with focus shift except to one aperture. Yes, one can calibrate it so that results at f/5.6 will be optimal (let’s assume for giggles that the autofocus system is precise and accurate), but all other apertures will then be off by varying degrees, especially f/1.8. That is the whole point of why focus shift is problematic! In short, if you shoot at only one aperture, it is relatively easy to deal with. If you shoot at both f/1.8 and f/5.6 (or whatever), then it becomes a significant damage to image quality (e.g., 36MP is blurred to 18MP or so), since one aperture or the other has to be chosen to “calibrate”.
Surely Nikon (and Canon) could build focus shift compensation into the camera’s AF system firmware. To my knowledge, this is not implemented.
My comparison with the Zeiss 28/2 Distagon shows that the foreground image of the Nikon 28/1.8 never gets sharper at any aperture. That is the damage that is wrought to a landscape photo; the zone of sharp focus is thrown rearwards, and for that one must compensate somehow in order to achieve the desired zone of focus. On another test scene (not yet published), a distant scene never gets sharp at any aperture with the Nikon 28/1.8G— negating any benefit of the 36MP D800E sensor over a D3x or D4, and greatly inferior to the Zeiss 28/2 (which BTW also has a small focus shift from f/4 - f/5.6, which explains why it doesn’t improve on a planar target past f/4).
Field curvature: I have no objection to the field curvature in the Nikon 28/1.8G— all 28mm lenses have it to some degree, the Zeiss 28/2 has a lot of it.
Sample variation: I have asked for another review sample from B&H Photo. The lens I tested had more problems than focus shift, but focus shift does not change with the lens sample.
Lens corrections: I have commented on this many times in my lens comparisons. Of course I alway disable all lens corrections, except if I am doing an explicit on/off piece (rare).