>On this one you will have to trust me if you don't find the >need to try out: I took hundreds of photo to play with all my >lenses, and the variability is negligible compared to the size >of the sharp zone of dots you have to observe on the target. I >therefore strongly believe that the camera (D800) is very >systematic at picking the dot consistently when there is >nothing else around it except white paper. I think most if not >all of the variability usually observed with other methods >comes from the fact that the target has several parts at >different distances (especially a ruler aimed at 45 degrees) >that are picked a bit randomly if they are too close.
I think you've touched on a more important matter. Making it, hypothetically, so easy for the camera to lock focus (by eliminating the potential distraction of adjacent potential targets) doesn't really test the accuracy or acuity of the camera's AF system. In actual photography situations, a chosen AF target or subject is almost always surrounded by nearby potential targets. So it's important, during any sort of AF system testing I think, to create a test bed that in several important respects challenges the camera's AF system as a whole rather than just a single AF sensor.
>I totally agree with you that one has to get rid of the >problems generated by aiming at a target at 45 degrees. What I >am claiming is that by exposing only one dark dot on the >target when focusing, you are not aiming at 45 degrees or any >other angle. With only a small dot to aim at, there is no >reference to any direction. It's just like looking at a point >floating in the air.
The problem I have with your methodology is that it is an approach which shows mainly that the method can work, rather than an approach which challenges the overall accuracy of the camera's AF system.
>What I am saying is that if you >only give the focusing system one dot to analyse, variability >is negligible in practice (at least with my camera and >objectives). And honestly, if my camera was not able to pick >consistently such an easy target, I would not felicitate Nikon > .
Again, my feeling is that testing of an AF system using a method which does not challenge the entire AF system at once, does not provide a result or a basis for AF fine tune adjustment which will be viable in real shooting conditions in which, invariably, the entire AF system is always challenged. By eliminating proximal targets, you may only be demonstrating whether or not the chosen AF sensor point placed on your target actually functions, not how accurate that AF sensor point actually is. I suggest the reason is that the both contrast detection and phase detection use a variety of information proximal to the AF target to make presumably accurate decisions about the viability of the chosen target point and how effectively in can be locked on. Without proximal information (literally, contrast or phase information adjacent to the selected AF target), I don't think the AF system is actually being effectively tested.
Rather than eliminating some test bench variables with your methodology, I think instead the methodology reduces the ability to accurately determine the state of actual AF system calibration.