12-24-2015, 09:28 PM
The story is quite simple: Although I adjusted AF for the Sigma 20/1.4 in front of a D810, the LiveView based contrast detection AF knows better in most cases or equally good and never (!) worse. There are about 3 shots where both kinds of AF came to the same conclusion. I didn't show all the samples. But I've seen enough.
It was highly unscientific - without any tripod and probably some blur came for shutter/mirror vibration. I count that as one reason more to abandon mirrors... all shots were done in about ½ hour, just by walking through the streets, take one picture with AF-C, move the AF-point to a suitable place and focus. Then I switched to LV, focused again and released. There's a note in the gallery which tells just what I said. The bicycle was trickier and here I enlarged the AF zone in LV and adjusted manually. The little jewelry tree was also tricky, I didn't see the branch in foreground in the viewfinder, but thought, I was focusing at the face plaquet under the rear branch. That was were I put the AF to in LV, again with manual adjustment. The wall painting with the soldier in a small creek, the light chains on the facia - hopeless to get that with PDAF.
In the gallery is only the first picture with PDAF. Plus a comparison of the AF zone at 100% crop. My conclusion: It's close to pointless in terms of getting the full resolution, the most possible information out of a picture if I (you or others may do better) can't rely on PDAF, adjusted as good as it gets with FoCal. Fast lenses = pricey lenses and in my case high resolving sensor, outresolved by the lens in theory. In real life crashed and smashed by a second best AF device. When I was adjusting the lenses, the spread of focus quality is visible in any FoCal protocol although the software triggers AF and shutter. If I go to check various focus points, I even see which one are weak.
As you said, dave, LV is a pain in terms of speed and of no use in street photography with a DSLR and I fully agree. Of course, DSLRs are not made for street photography with LV. Checking proper focus afterwards? Yes, why not. Unless I don't compare it with the potential of the LV versions, I was always fine with the results and thought, that's as good as it gets. True, for PDAF it is. But there's more in the lenses and I can't see it in most samples.
It was highly unscientific - without any tripod and probably some blur came for shutter/mirror vibration. I count that as one reason more to abandon mirrors... all shots were done in about ½ hour, just by walking through the streets, take one picture with AF-C, move the AF-point to a suitable place and focus. Then I switched to LV, focused again and released. There's a note in the gallery which tells just what I said. The bicycle was trickier and here I enlarged the AF zone in LV and adjusted manually. The little jewelry tree was also tricky, I didn't see the branch in foreground in the viewfinder, but thought, I was focusing at the face plaquet under the rear branch. That was were I put the AF to in LV, again with manual adjustment. The wall painting with the soldier in a small creek, the light chains on the facia - hopeless to get that with PDAF.
In the gallery is only the first picture with PDAF. Plus a comparison of the AF zone at 100% crop. My conclusion: It's close to pointless in terms of getting the full resolution, the most possible information out of a picture if I (you or others may do better) can't rely on PDAF, adjusted as good as it gets with FoCal. Fast lenses = pricey lenses and in my case high resolving sensor, outresolved by the lens in theory. In real life crashed and smashed by a second best AF device. When I was adjusting the lenses, the spread of focus quality is visible in any FoCal protocol although the software triggers AF and shutter. If I go to check various focus points, I even see which one are weak.
As you said, dave, LV is a pain in terms of speed and of no use in street photography with a DSLR and I fully agree. Of course, DSLRs are not made for street photography with LV. Checking proper focus afterwards? Yes, why not. Unless I don't compare it with the potential of the LV versions, I was always fine with the results and thought, that's as good as it gets. True, for PDAF it is. But there's more in the lenses and I can't see it in most samples.