Autofocus Testing and AF Micro-Adjustments
If there’s one thing I’ve learned today, it’s that auto focus alignment tests are not for the faint of heart. Better yet, as I’m slowly becoming convinced, anyone. I certainly wouldn’t recommend even bothering unless you are almost certain that your lens is consistently misfocusing with significant repeatability.
I spent the better part of today working with a fellow photographer testing and adjusting the auto focus on both my EOS 1D and his Nikon D700. We were using the target from here. Our initial objective was to determine how viable testing and adjusting was using a simple target or whether a lens-align should be purchased. Suffice to say, our results were mixed and raised several questions about the overall accuracy and repeatability of these tests.
We started our tests using the 25% target as suggested, with mixed results. I have since repeated the tests with the 10% and 5% targets to insure that I was actually focusing where I was suppose to and not accidentally locking on to the faint gray markings. We also worked as carefully as possible to align the camera to the target so the reading would be as accurate as possible. For starters, this is easier said than done.
Initial testing was done with my EOS 1D Mark 3 with my EF 50mm f/1.8 II and EF 100-400 f/4.5-5.6L IS USM lenses. Our procedure mirrored standard recommended practice. Align the camera, set the lens to infinity shoot the test image look at it on the computer. We started with an unbiased test image to determine which way we needed to work. Then shot a series of images adjusting the AF in 5-point increments. Again the images were reviewed on the computer and the best two were used to determine what settings to shoot between in single point increments.
We found that frequently the first pass through got us close enough that we didn’t feel the need to further fine tune, and in many cases, we only needed the 5 or 10 bias to get things well in hand.
The second round of testing was based in our desire to see how this affected things in the real world. That is, just because everything looked good on the focus target it doesn’t necessarily mean the autofocus was better in the real world. These tests were crude, quick and very effective, as this is where we started seeing problems.
We first spotted the problem when shooting with a Nikon D700 and an AF Nikkor 85mm f/1.8D. Our testing on the focus chart showed the lens back focused a slight bit. We dialed in the necessary correction in this case moving the AF 10 points closer to the camera, and everything looked spot on, at least on the chart. Then we put it thought it’s paces shooting objects around the room. Those “real world” shots were way off and the lens now showed considerable front focus.
At this point calling us confused would be an understatement. Most AF test procedures, regardless of the target they use, state that you should set the lens to infinity between each exposure. They say this is to insure that the AF system is actually acquiring the target properly. Well in the real world, not every AF movement is from infinity to some closer distance. In fact I’d bet that roughly half are from some closer distance towards infinity. In our case, we were finding realistic movements were throwing something off.
Retesting on the focus target showed that what would happen predictably going one direction, say from infinity to 5 feet, wouldn’t predictably happen going the other, in this case 1.6′ to 5′. We could get some lenses to focus differently when changing where they started focusing.
My first suspicion was that the error was due to the sensor locking onto a different part of the target’s bar. I didn’t feel very comfortable with this though as the errors in focus were quite significant, far more than any slight misalignment should have created. We’re talking about going from slightly back focusing to severely front focusing.
Following that, I suspected that the error was created by play in the AF drive chain. After all, these were Nikon AF-D lenses, and there had to be some gears between the focus motor in the body and the part of the lens that actually moved the glass, not to mention there was a coupling involved as well. This was partially disproven though as well, as I could produce the same error using my EF 50mm f/1.8 II (though it’s a micro motor design and may have some play in the focus drive train there as well).
The final option was to retest using AF-S/USM glass. This initially confirmed our suspicions that there was play in the AF drive train, two AF-S lenses aligned perfectly showing no error between the infinity-to-target and macro-to-target focus passes. That was until we got to the AF-S Zoom-Nikkor 70-200mm f/2.8G IF-ED, which exhibited the same differential focusing behavior.
Needless to say, the afternoon was less than conclusive with respect to focus testing and micro adjustments.
To make maters worse, I repeated the tests on my workbench later in the evening. This time I was using the 10% target under the assumption that perhaps the 25% target had too much contrast and the AF was locking on to the gray text instead of the black line. I was surprised to find that in that first retest, my EF 50mm f/1.8 II focused repeatedly on the same spot when starting from either the infinity or macro positions. I then repeated the test a second time and found the lens again focused differently.
I decided to take this a step further in hopes of increasing the consistency. I did this by moving the alignment target into the vertical plane, as square to the camera as possible and using a sloped ruler adjacent to it to make the distance estimations. Even with this configuration, I’m seeing mixed results. While the issue of the point of focus changing seems to be addressed, at least I’m seeing more repeatability regardless of focus direction. There are still problems. For starters, I’m now seeing erroneous results when the lens is taken back to a real-world scenario. In this case, with my EF 50mm f/1.8 II, the focus target is saying that I should have the AF adjust set 15 points. However, shooting a real scene like that nets an out of focus image.
Tonight’s conclusions are brief. For focus tuning, targets and charts are certainly important, but ultimately any lens really needs to be retested in conditions similar to how it’s normally used. Alignment is important but difficult, if you have a grid focus screen or grid lines on newer Nikon bodies, use them it’s very helpful. Though alignment is important, repeatability is crucial, before even attempting to perform AF tuning make sure you’re camera can hit the same point consistently. Double check with real subjects, just because the chart says you’re good to go doesn’t necessarily mean you are. Finely, not specific to autofocus adjustments, I’m finding that pumping the shutter release once when working with very narrow depth of fields considerably improves the accuracy and consistency of focus.