The AF Fine Tune Itch

Sometimes companies get themselves into strange positions by actually trying to do something useful. AF Fine Tune certainly seems to be one of those things. 

So first some background.

The modern DSLR is an alignment nightmare when it comes to focus. Light goes through a partially silvered mirror that needs to be at just the right angle to another mirror that needs to be at just the right angle to another mirror that needs to be at the right point to a separator lens set that needs to be properly positioned to the autofocus sensors, which also need to be properly positioned in 3D space. Oh, and all that alignment needs to agree with the focus screen, prism, and viewfinder alignment (otherwise you might try to refocus something that was actually in focus), plus the imaging sensor itself. 

On top of this, the geometry of phase detect generates pretty small numbers that need clear differentiation. Then, once the camera has accurately generated that number, it needs to know that the lens will go exactly to the right position it tells it to. So the lens elements also need to be properly positioned with no slop in their positioning mechanism and the focus motor must repeat the exact same move every time it is instructed to.

Oh, and did I mention that everything in the above has to be absolutely repeatable every time you press the focus button (shutter release)? Anyone that knows about system designs knows if you look at a close enough level, repeatability is a “range” not a binary yes/no. 

Thing is, even a really small tolerance error somewhere in all that I’ve described can produce image results where the focus plane isn’t quite where it was supposed to be. As we get higher and higher sampling at the sensor—more megapixels—it gets clearer and clearer when we have small “misses” by the focus system. 

So the camera makers added AF Fine Tune. 

Unfortunately, I think this just added a whole can of worms that the camera makers weren’t expecting: user paranoia. 

No doubt there are camera/lens combinations out there that are putting focus at a wrong position every time. No doubt at all. Generally, those errors still should be small and the errant combinations should fall in a bell curve around the proper position. 

Unfortunately, the first thing that happens is this: the user tries focusing on something, the camera doesn’t focus where they think it should, and thus the user immediately decides to try AF Fine Tune. 

Wrong. Stop. Don’t.

Before you start dropping into AF Fine Tune you absolutely must eliminate any misinterpretation of the focus system as the cause of your problem (e.g. user error). I’ll just say this: I’ve dealt with hundreds—maybe even thousands by now—of “doesn’t focus right” errors in the last 20 years. More often than not, it’s assumptions about what the focus system is doing that are causing focusing errors, not AF Fine Tune. In all that interaction, I can count the number of “really off” systems on my digits (I have the proper 20, so don't jump to the conclusion that perhaps I’m a mutant of some sort ;~).

At the start of every African wildlife workshop my assistant and I look carefully at all the student camera/lens alignments and perform AF Fine Tune if it’s needed. Most of the time, it isn’t, and even when it is it tends to be in modest amounts (e.g. in the +5 to -5 range). Generally if I find a combination within +/-2 I won’t set AF Fine Tune active.

Why? Well, here’s where we start getting to the D5 part: I don’t think that we can absolutely guarantee that the lens always stops right at the value you set, even on the very best gear. Repeatability of autofocus using phase detect—where the lens goes straight to a position—has a very small variability to it. Very small, but there.

On my D5 and 80-400mm, for example, an hour of careful analysis told me that under phase detect it was always hitting in a range that equalled +1 to -2 in AF Fine Tune adjustments. That’s slightly off the expected point, but the fact that the lens was always headed into that same narrow range meant that I shouldn’t at all be worried about it. 

After another hour of using the Automatic AF Fine Tune on the D5 via manual focusing in Live View, I finally decided that the absolute best setting for the lens was probably -1 in AF Fine Tune. Note what I said about my acceptable range, above. I’ve never had to adjust AF Fine Tune with this sample of the 80-400mm on any of my cameras. Of course, I’ve never examined quite this closely before, either. But if it isn’t 0 or -1 on my cameras I’d eat my warranty card. 

I should point out that this alignment was analyzed using a Lens Align target that was absolutely positioned correctly (all red alignment marks visible at the camera), and at a distance of 20x the focal length (I can’t easily do 30x in my studio the way things are currently arranged). Lighting was a mix of ambient and my Lowel lights.

Now, what did the D5 want to set via Automatic AF Fine Tune? Well, anything from +5 to -7, as it turns out. But as I’ve noted before, I don’t believe my sample of the 80-400mm actually focuses correctly via Live View ;~). Virtually every time I let the camera do the focus in Live View, I could zoom in on the Live View display and manually tweak focus to the right point. And yes, I did not have to do that with phase detect autofocus: the 80-400mm was always hitting a much narrower range when operated with phase detect autofocus, as I reported above.

Meanwhile, I’m getting a number of reports from others of a small-to-wide range of results in repeated use of the Automatic AF Fine Tune with the same D5/lens combination. Some lenses seem to generate a very narrow range in Automatic AF Fine Tune, others seem to generate much larger ranges, though I’m not sure exactly how disciplined these folk are in doing their testing. 

So the whole Automatic AF Fine Tune thing isn’t without its own set of issues. 

The trouble is that this just sets off more user paranoia. “Camera didn’t focus. Tried Automatic Fine Tune. Camera focuses different but still wrong. Nikon sucks.” 

Uh, no. I’m not going to blame Nikon at all on this. They’re actually trying to help us, but there’s still some limits to their help. And if you aren’t completely disciplined in doing your tuning, you can simply just get different wrong results. 

So what do I suggest?

First, don’t even try unless you’ve got a way to absolutely guarantee that you’re in alignment to the test target. That’s one of the nice things about the LensAlign target: it has a built-in verification-of-alignment helper. So don’t tell me that your results are wrong or inconsistent unless you can first tell me how you verified that you camera was properly positioned to the target. 

Next, do some repeated phase detect focusing. You should be getting results in a very narrow range to your aligned test target. If you aren’t, then I’d have to guess there’s something wrong with your lens, your support system, or your choice of autofocus settings. But here’s another thing: are you sure you’re not looking at anti-aliasing of some sort? Nikon’s JPEGs, for instance, have a default of almost no sharpening, and they look a bit fuzzy even if focus was achieved correctly.  Short version: if you don’t have repeatable results at this stage, you’ve got something wrong somewhere.

Moving on, we try Live View autofocus the same way. How repeatable is it? 

text and images © 2020 Thom Hogan
portions Copyright 1999-2019 Thom Hogan-- All Rights Reserved
Follow us on Twitter@bythom, hashtags #bythom, #dslrbodies