Summary
As a follow-on to my earlier tests concerning the outer AF points, I tested two other lenses and repeated my test of the FA 43mm Limited. My testing did not reveal any overall systemic issues with the AF points in my sample of the K-3 Mark III. At the moment, I believe that the viewfinder AF (i.e., phase detection AF) of my K-3 Mark III works well. New Focus Target
I constructed a new focus target that would allow the camera to remain stationary on the tripod for each AF test shot, rather than using a 'focus and recompose' approach. The target was scaled to accommodate the 13 selected AF points at a nominal target distance of 206 cm or 41.2 times a focal length of 50mm. The target is applicable to other focal lengths by changing the target distance to maintain the same field of view. In this manner, each of the mini-target 'chips' aligns with AF points 1-13 as shown below. The target panel measures 26 x 11.5 inches (66 x 29 cm). (Why 41.2 x FL? Somewhat arbitrary -- I didn't want to make the target panel too wide and I had a convenient piece of foamboard on hand.) The target chips present a clear, unambiguous cross as well as a graphical pattern to aid visual inspection and comparisons. The image below shows the target overlaid with the camera's focus point field.
Incidentally, if you see a Moiré pattern in the figure above, it's an artifact caused by the interference between the fine lines and your monitor's pixel array; the element is actually a series of progressively finer spaced semicircles. Methodology
Rather than using a quantitative method, which was described in my earlier post (ImageJ analysis and plots), I visually inspected each target chip in its corresponding JPEG image, using DxO Photolab 5, e.g., for an AF point that was placed over chip 6, I assessed chip 6 for focus quality. This approach precluded any out-of-focus issues that might have arisen with the earlier approach of using the central chip 1 for all assessments regardless of the AF point. It also alleviates issues with possible slight misalignment of the target panel and sensor plane (although care was taken in setting up the camera and target).
Images were viewed at 100% scale, after adjusting the exposure, applying nominal sharpening (Photolab's Lens Sharpness function), and correcting chromatic aberrations where necessary (the same processing profile was used throughout). In this manner, the images were put through a typical post-processing chain. Most of the images sharpened up nicely compared to their neutral out-of-camera originals.
All of the tests were run with AF Fine Adjustments set to their optimal values, as determined in my earlier calibration exercise after updating to Firmware v1.31. These settings were based on the central AF point.
The AF mode was AF.S, SEL (S). An infrared remote was used to trip the shutter.
In some of the test runs, two shots were taken after focusing the lens manually in magnified Live View (focused in LV, shot in VF AF), which provided reference images of sharply-focused shots. Each of the three lenses was put through at least four independent test runs. For the DA 20-40, I took an additional 30 shots using AF Point 8 and 20 shots using AF Point 6, variably at f/4 and f/5.6. Additional shots were also taken with the DA* 50-135, concentrating on several points that initially seemed to be out of focus.
The relative focus quality ('RFQ') of each shot was classed subjectively as Sharp, Acceptable, or Misfocused. Most of the 'sharp' images were clearly sharp and most of the mis-focused shots were similarly poor. There were instances where it was a 'coin-toss' between acceptable/misfocused or acceptable/sharp; in those cases I tended to be pessimistic and chose the lesser quality. Recall that the images were inspected at full 100% zoom, which is a demanding viewing condition.
After assessing each image, a tally was made of the three focus classes, and a final, overall quality was assigned for each AF point. For example, a focus point attaining three sharp and one acceptable ratings would be deemed 'sharp' while two sharp and two acceptable shots would give an overall rating of 'acceptable'. Results smc Pentax-FA 43mm Limited, f/2.8
The FA 43mm Limited showed results that were practically identical to my earlier tests: the AF points in the inner AF frame all produced sharply-focused images. The outer points (using AF line detectors) generally misfocused.
Although the six outer points are not located at the very edge of the image frame, they appear in the outer quarters. The FA 43 is known to exhibit field curvature (see, e.g., the PF review at SMC Pentax-FA 43mm F1.9 Limited Review - Sharpness | PentaxForums.com Reviews), which might account for the relatively poor AF performance of the outer points at a fixed AFFA setting, especially since the misfocusing appeared to be symmetrical across the AF field. Recall that the lens was calibrated against the very central AF point of the camera, which probably does not work at the outer points for this lens, as suggested in my previous post.
To check the optical performance of the lens, a separate run was done in which two shots at AF Point 3 were manually focused in magnified Live View and five shots were autofocused. The first two shots were fairly sharp -- much sharper than the misfocused others. One AF image was relatively sharp -- a lucky shot, so to say. This seems to confirm that the poor image quality found earlier at this outer point is not likely due to an inherent weaker optical quality near the edge, but rather a focusing issue. HD Pentax-DA 20-40mm Limited
In the case of the DA 20-40mm Limited at 40mm, all of the AF points can be trusted to give at least acceptable focus. Not surprisingly, slightly sharper results were attained in a run at f/5.6 compared to the others at f/4. Wide Angle -- Disappointment and Relief. The DA 20-40 was also tested at 20mm. Initially, I took six shots at each of the 13 AF points -- one in Live View AF and five using VF AF -- wide open at f/2.8. To my surprise and disappointment, most of the shots appeared to be soft, except for the centre point. Then I recalled that at a wide-open aperture of f/2.8, this lens has relatively weak optical performance at 20mm except at the centre, so I was probably seeing the ill effects of sharpness drop-off rather than a focusing error, even at the inner AF points. I followed-up with spot checks at three AF points (3, 5 and 6), stopped down to f/5.6; all shots were sharp. smc Pentax DA* 50-135mm
The DA* 50-135mm (converted to screwdrive) was tested only at 50mm, f/3.2. The AF points in the inner frame gave good results, while the outer points were weaker; two points misfocused more often than not. Conclusion
For the three lenses that were tested, the AF points in the inner frame appear to function very well, producing sharply-focused shots in general. The outer points, which use line detectors, are less robust but produced at-least acceptable shots in the majority of cases. In the case of the DA* 50-135 (at 50mm), two AF points misfocused in most shots. Finally, the FA 43mm Limited may suffer from field curvature, which would require a different AFFA setting for the outer points. Otherwise, the FA 43 focuses sharply with the inner AF points.
My testing did not reveal any overall systemic issues with the AF points in my sample of the K-3 Mark III. The limited testing at a focal length of 20mm confirmed sharp focusing at a wide angle.
My research on the subject of outer AF points led me to numerous web sites, including some early (2012) issues with the Nikon D800, which misfocused systemically when using left-side AF points. An interesting account of this problem may be found at: D800 autofocus problem | hifivoice
- Craig
|