Autofocus Testing and Microadjustments, Round 2
Today concluded round 2 of serious focus testing. What’s different this time from the last? A new target, a new alignment strategy, and some new results. Last time I looked at autofocus, I had enough problems focusing fast lenses that I was growing concerned that there was either a systematic design flaw or similar error in phase detection AF systems. I was seeing behavior with fast lenses on both Canon and Nikon bodies that would inconsistently focus depending on a verity of situations.
The good news, the new target dramatically improved auto focus consistency on both platforms and the new alignment strategy shortened setup time and aided accuracy. The not so good news is that I’ve determined that there is definitely a problem with my EF 50mm f/1.8 II and it may in fact be a much wider issue; though I guess you could say, what do you expect from a $100 lens.
The Bad: AF with the Canon EF 50mm f/1.8 II
The Canon EF 50mm f/1.8 II–the lens that got me started looking at focus calibrating–is clearly a problem, at least mine is. I have over 500 frames shot with both a EOS 1D Mark3 and an EOS 40D, under carefully and not so carefully controlled conditions that lead me to believe, with quite a bit of confidence, that either my copy or the lens as a whole are problematic. Further, I’ve read a few reports with other photographers expressing similar similar behavior from this lens. In the end, I can only conclude that the chances are that it’s not my copy specifically. However, at only $100 it’s not really worth it for me to examine this further.
Quite simply, while the EF 50mm f/1.8 II was somewhat of a deal when it cost less than $100 I can’t say that any more. At the current price, perspective buyers should strongly consider either spending a few hundred more on the EF 50mm f/1.4 USM (not that I’ve tested it) or be well aware that their lens may inconsistently focus on their camera. If you’re going to go ahead and get an EF 50mm f/1.8 II, I’ve found that shooting in AI Servo provides the most consistent results, but even then it’s a good idea to shoot a couple of extra frames as backup.
The Good 1: Consistency
The second issue that came up in the previous tests was the lack of repeatability with a few lenses other than the Canon EF 50mm f/1.8 II. Specifically two Nikkor primes an 85mm f/1.8 and a 50mm f/1.8. The odd part was that other than the Canon lens, the problem was slightly less predicabel with the Nikkors. It was bad enough that I though maybe the problem was due to play in the AF drive mechanism or worse that turned out not to be the case. However that appears not to be the case.
In the time since the last post and this post I had been working off and on with my glass trying to figure out what was up with the AF inconstancy issue. The biggest reason I had been using Jefferey’s target was for the light gray marks along the center of the target. The theory being even if the target is slightly misaligned they’ll still give an accurate indication of where the lens is focusing. However try as I might I still had issues with fast lenses miss-focusing more than they should using even the 5% target.
The confusing and frustrating issue was that I couldn’t generally get the cameras to focus on the light gray pattern in Jeffery’s target. Which is the test he recommends for determining which density to use.
In a fit of frustration I tried a different target, one that simply had a white field with a black bar. Surprisingly I had very few focus problems with this target. So I went back and tried focusing on the low contrast gray area to see if it really wouldn’t focus. Turns out that at least on the EOS 1D Mk3, and after today’s testing the Nikon D700 (and by extension the D300, D3 and D3x), under some conditions with a fast enough lens the AF system is capable of locking on the to light gray text in the target.
The result, a quick redesign with a white central field and the repeatability of focusing picked up significantly.
The Good 2: Alignment
The final change was in the arrangement of the test setup to eliminate some variables and control some others more easily. The most important aspect of these tests is squaring the camera to the target. Some minor variation is workable but major errors are not.
Typically, the target is placed on the ground, or on a desk, and the camera is angled down at it. In this case it’s difficult to insure that the camera is square to the target, as you have to deal with alignments in several directions. Instead I’ve taken to mounting my targets to a vertical surface, such as a wall or large window. Ideally the target would be cemented to a flat surface such as a piece of glass, to insure it is flat; but again there is some room for error.
With the target leveled and on a vertical surface the the camera can be aligned simply by leveling the camera (I use a Hot Shoe Level(Affiliate Link) ) and adjusting the height until the center of the lens is at the same height as the target. No more trying to square the target and the camera and any number of other things.
There’s no point trying to preform AF adjustments on a lens that can’t repeatedly focus at the same point. Before even attempting to do any kind of micro you need to insure your lens focuses consistently at the same point. I do this by manually focusing to both infinity and the close focusing distance (macro) on the focus ring and allowing the camera to refocus. At a minimum with a new lens I repeat this 4 times, twice from infinity and twice from macro. If there is any discrepancy I’ll repeat it often enough to be able to characterize the problem as either user error or an issue with the lens/camera. If I can’t get at least 95% of my shots to focus on the same place for both directions with a single black line AF target, I pretty confident that the lens is problematic.
Second, target design is important. It’s my experience that several things need to be addressed in the target. First, the actual autofocus point must be clear of any diagnostic markings. It’s not simply enough to assume that there is one predominate mark and any low contrast marks won’t be detected. In addition the target needs a visually clear system to show DoF. Most targets use some kind of simple ruler, that’s simply not adequate. The reason I like Jeffery Friedl’s target is the inclusion of hashed lines of varying sizes. The hash marks make determining DoF significantly easier than trying to guess at fuzzy numbers.
There’s certainly no reason not to test and adjust your lenses to achieve best focus. However the process is not something that should be taken lightly and there’s plenty of room to shake your confidence in your gear.
There are no comments on this article yet. Why don't you start the discussion?