To @licht96 and other interested parties, I'm late to the game with this post (I needed to find some time to do a proper test).
I tested one of my lenses, FA 43mm Limited, for its AF performance on my K-3 Mark III. My set up and procedure was similar to what I described in a recent article:
K-3 Mark III and Firmware 1.31: Requires New Autofocus Fine Adjustment - PentaxForums.com
In summary, here is my setup, following the OP's list:
Pentax K-3 Mark III, Firmware 1.31
Lens SMC Pentax FA 43mm Limited
Focal length 43 mm
Aperture f/2.8, f/1.9
ISO 125-200
Exposure Time 1/200, 1/400 s
Camera tripod mounted; SR Off
White Balance AWB
Focus Adjustment to center: 0 nominal (set to +6 in one test)
Lights: 2 compact fluorescent bulbs, 23 W, 6500 K, mounted in 20-cm workshop reflectors, 60 cm from target centre, offset 40 degrees left and right from the target axis.
Ambient temperature: 18° C
Distance Focal plane camera to Focus Test target: 200 cm
Camera horizontal and vertical adjusted with internal balance
Target level and aligned with sensor plane
Before each shot the lens was defocused to infinity.
AF mode: AF.S, SEL S (Select single point), single shot.
Here is a picture of the target setup, as seen in my basement workshop (actual test shot 5981, referenced below):
The next figure shows the map of the camera's AF points superimposed on the target within the test scene. For reference, the target has a width of 46 cm. The distance between the camera's sensor plane and target is 200 cm.
In summary, the seven points within the central AF frame that I tested all provided a sharp focus at the baseline optimal AF Fine Adjustment (AFFA) setting of zero. The points on the periphery (3,5,6,9,10, 13) all produced mis-focused shots when using the optimal AFFA setting of zero.
My original baseline AF Fine Adjustment test, performed earlier when I updated to Firmware 1.31 and using the standard central AF point, is shown in the following figure, top graph. The AF consistency was very good and the overall data was tight. The 'Relative Focus Quality' indicates the variation of focus sharpness as the AFFA setting is changed. The optimal AFFA setting is zero for this lens. Note that any image with an RFQ above 8.5 is considered well-focused; there is no apparent difference in image sharpness when viewing at 100% full-scale on an HD 1920 x 1200 24-inch monitor. See the referenced paper above for more details.
The lower graph shows how the optimal AFFA setting changed when using the far left edge AF point (Point 3 on the map), positioned over the centre of the black target cross. The optimal AFFA setting is +6. The lower graph also shows (Run 9) that the centre left AF point (Point 2) provides a sharp focus at AFFA=0, with an AFFA profile that is practically identical to the baseline case.
The next graph shows the focus consistency for two cases. In Run 10 with AFFA=0 and aperture f/2.8, the camera and lens 'nailed' every shot in the 10-shot series. Run 11 shows the shot-to-shot consistency wide open at aperture f/1.9. In this case, the AF precision is also very good. However, the purple fringing and generally softer optical characteristic that the lens produces at f/1.9, lowers the relative image quality compared to the f/2.8 case, which is seen here as a lower RFQ value. Despite the purple fringing, the camera and lens nailed the shots with respect to the focus.
The main part of this test examined the AF results for 13 different AF points, as indicated in the AF-point map above. The results are shown in the following graph. The non-central AF points (3,5,6,9,10,13) resulted in images that were seriously out of focus.
To test the AF of the non-centre points and obtain images for comparison, the camera was moved on its ball head mount and secured to place the point at the centre of the black cross on the target. The lens was manually defocused and then focused using the back AF button; the camera was repositioned again to place the centre of the frame back to the target's centre; then the shot was taken. In this manner, the peripheral AF point was used to focus, while the centre of the target was used to determine the relative focus quality. This avoided any potential degradation of the lens' sharpness towards its edge. With a target distance of 200 cm, a target half-width of 23 cm, and depth of field of 23 cm at f/2.8, any geometrical focusing error introduced by this 'focus and recompose' method should be negligible (as evidenced in Run 15, last graph below).
Finally, in the following graph we compare the focus results when using the left edge AF point (number 3) at two different AFFA settings. At AFFA=0 (Run 14), the relative focus quality is poor, as described above. Resetting the AFFA to +6 results in very good, consistent focus (Run 15). Shot 1 is the reference data from image 5981, which used AF Point 1 (same as AF Point Number 1 in the graph immediately above). Shots 2-6 used AF Point 3 for each of the AFFA sets.
My test led me to conclude that for my FA 43mm Limited lens, the outer AF points beyond the central frame require a different AFFA setting than those inside the frame.
To explore a possible variation in focus accuracy between the central cross-shaped AF sensors and the peripheral line-sensors, I ran several other tests using different sections of the focus target. For example, the AF point was placed just to left of the centre of the black cross to ensure that the horizontal line was within the line-sensor's field. I found no difference in AF performance.
Owing to time limits, I did not test all of the peripheral points to the same level of detail. However, it is believed that the selection of the 13 dispersed points was adequate to provide a reasonable input to the issue raised by the OP in this thread.
I prepared this post somewhat quickly, so I may have left out some details or missed an explanation or two. I'd be glad to receive any comments or questions on the test procedure or results, general observations, or suggestions for improvements.
- Craig