07-24-2012, 06:56 AM
[quote name='Frank' timestamp='1343101319' post='19576']
Hi Wim:
But from various resources I read that the PDAF process in reality works in an iteration way. For example http://graphics.stanford.edu/courses/cs1...cusPD.html, which in the "The autofocusing process" section states that "In practice it usually moves close to the correct position, then takes additional measurements to refine this position". So, I assume that after the lens takes reaction to the command of the camera, the camera does check if the lens has followed its command correctly (otherwise there would be no the "green dot" (or beep) confirmation I think). Could you clarify this?
Thanks,iterary
Frank
[/quote]
Yes, it is correct that PD Af works in iterations. On internet it has been a long belief that it was a look once move once affair, but lots of evidence makes clear it is a process done in steps. The reason why PD AF can get it so wrong still, at times, is because the steps are quite crude for speed reasons, and the last step is not verified again, also for speed reasons. If the camera is certain it is close to focus, it will tell the lens to move a certain distance compared to the last step the lens did. If the lens goofs up this step or the camera calculation for whatever reason is incorrect, you end up with a more or less missfocussed image. Also, some lenses just are missfocussing very constantly on some cameras, even though their lens movements are calibrated. This can have to do with certain lens aberrations, especially if the AF system looks at just a certain range of the spectrum.
Focus shift has nothing to do with how PD AF works though. It is called focus shift when the lens wide open (AF is determined when the lens is wide open) has a different focal plane than when you close the lens down a bit. The focal plane shifts, that is focus shift.
Some cameras can take that into consideration, only when the camera knows about that certain lens of course, and if their AF system is advanced enough and has such considerations implemented.
Hi Wim:
But from various resources I read that the PDAF process in reality works in an iteration way. For example http://graphics.stanford.edu/courses/cs1...cusPD.html, which in the "The autofocusing process" section states that "In practice it usually moves close to the correct position, then takes additional measurements to refine this position". So, I assume that after the lens takes reaction to the command of the camera, the camera does check if the lens has followed its command correctly (otherwise there would be no the "green dot" (or beep) confirmation I think). Could you clarify this?
Thanks,iterary
Frank
[/quote]
Yes, it is correct that PD Af works in iterations. On internet it has been a long belief that it was a look once move once affair, but lots of evidence makes clear it is a process done in steps. The reason why PD AF can get it so wrong still, at times, is because the steps are quite crude for speed reasons, and the last step is not verified again, also for speed reasons. If the camera is certain it is close to focus, it will tell the lens to move a certain distance compared to the last step the lens did. If the lens goofs up this step or the camera calculation for whatever reason is incorrect, you end up with a more or less missfocussed image. Also, some lenses just are missfocussing very constantly on some cameras, even though their lens movements are calibrated. This can have to do with certain lens aberrations, especially if the AF system looks at just a certain range of the spectrum.
Focus shift has nothing to do with how PD AF works though. It is called focus shift when the lens wide open (AF is determined when the lens is wide open) has a different focal plane than when you close the lens down a bit. The focal plane shifts, that is focus shift.
Some cameras can take that into consideration, only when the camera knows about that certain lens of course, and if their AF system is advanced enough and has such considerations implemented.