As discussed in Making Sharp Images, any assumption of a camera being within accurate mechanical tolerances for just about anything is a bad assumption, whether it’s brand-new or not. Errors these days in the 5-10 micron range show up with high res digital in fairly obvious ways.
In particular, manual focus accuracy is a mechanical issue, with generally quite sloppy tolerances, since both Nikon and Canon pretty much assume autofocus these days. My experience has been relatively lucky, perhaps because I generally shoot the pro-grade camera bodies, but even those are not necessarily perfect. It’s not a given to expect agreement between the two optical paths of a DSLR (to the sensor and to the viewfinder). The same is true of rangefinders. As another example, on several Nikon bodies I’ve used, I can’t get camera viewfinder, the sensor, and a bubble level to agree!
When testing for accurate manual focus, shoot your lens wide open at its maximum aperture, e.g., f/1.4 or f/2 is best (do not focus wide open and stop down to shoot, because focus shift will confuse the issue, see Making Sharp Images). Testing with an f/2.8 or slower lens will make it hard to detect the error, due to depth of field masking the problem.
Reader Scott McLeod experienced severe manual focus errors with his Nikon 700. He writes to describe what he had to do to make focusing accurate. Read his report below.
I just wanted to let you know that my attempt to re-shim the VF screen in my D700 was a total success and a lot easier than I expected. Other than being careful to handle the shims without bending or dropping them and patient enough to go through the procedure several times, it's really no more difficult than changing screens. Actually by far the worst part was getting rid of the tiny specks of dust/lint/whatever that end up on the screen and the base of the prism while it's all apart, but repeated use of the Rocket blower and the LumaMax eventually got it all spotless (shining it through the eyepiece reveals the exact location of any dust on the prism).
FYI the shims are color-coded but for some weird reason the part # sequence does not correspond to an increase in thickness:
1K603-374 - 0.15mm (copper) 1K603-373 - 0.20mm (brass) 1K603-372 - 0.10mm (silver/chrome) 1K603-384 - 0.05mm (copper)
The small blue tool is the wire latch tool (it came with the KatzEye screen I bought but could not get used to), the Pec-Pad is to put the screen on (shiny-side down) while it was out of the camera, and the tea-towel was to rest the body on to stop the shims dropping out while upside down; the steel rule was taped down at both ends and the desk lamp just provided a bit more light on the markings. I used a Tokina M100 as it is extremely sharp, the FL made seeing the ruler markings a *lot* easier, and it has a very long focus throw (at this distance, the widest this lens goes is f/3.5 but that's still a pretty shallow DOF; a 50/1.4 would have been even better but I would have *needed* a magnifying eyepiece to get decent accuracy). I wore non-powdered rubber gloves for the whole procedure and cleaned the tweezers with 100% isopropanol beforehand.
I focused on the 500mm line as best I could in the VF and then switched to LV to check where it was *really* focused (in this case, about 5-6mm further back), repeating the process several times for each shim pack. This camera had 1x -373 and 1x -384 shims in it for 0.25mm total installed in it from the factory. I already determined that since the VF was actually back-focusing, it needed more/thicker shims, so I removed the -384 shim and tried each successive thickness until I got it right (or as close as it could be given the limited range of thicknesses available). I ended up with 2x -373 shims for 0.40mm, quite a large difference from the original. When this camera was sent in to Nikon to get a few hot pixels mapped out they also adjusted the AF and it came back absolutely spot-on - I am pleased to report that it now agrees with the viewfinder!
While this may look/sound daunting, anyone who has replaced a focusing screen could do this job, and (IMO) serious users of manual-focus lenses should check their camera for VF/AF agreement and think about shimming if they need to (go easy on the caffeine beforehand, though :)). Considering I bought 2x sets of shims for $23.68 (as I didn't know what I'd need and there was always the possibility of damaging one), this was the best-value upgrade to this camera I can think of, since I can now *accurately* focus MF lenses, even long ones, and tweaking AF-S lenses is now something I can do with complete confidence in the results.
Reader Shane reports:
I performed a similar operation on a D200. In my case it had severe front focusing issues. When I removed the shims I found it had two of the same thickness, so to "start" I just removed one of them. First try produced the results I was looking for. I suspect that the shims had stuck together during the original camera assembly at Nikon. This does appear to be an issue when working with extremely thin shims, so care should be taken when performing a shim adjustment that this does not happen otherwise results can be confusing.