QuietOC wrote:
I've noticed this fairly often with the Sony FE 85mm F1.8.
Phase-detect focusing relies on pattern recognition in the phase domain. I suspect the shape of the out-of-focus point-spread can cause false positives. Somewhat similar to the focusing problems with undercorrected spherical abberation like the Sigma 45mm F2.8--except with those are more not detecting when they should.
Remember Sony mirrorless can only detect contrast horizontally. Point at horizontal high-contrast. Bumping the ISO up to the second gain also seems to help in lower contrast situations.
Now it's making sense. Fault by design
Thanks
Karl
QuietOC wrote:
I've noticed this fairly often with the Sony FE 85mm F1.8.
Phase-detect focusing relies on pattern recognition in the phase domain. I suspect the shape of the out-of-focus point-spread can cause false positives. Somewhat similar to the focusing problems with undercorrected spherical abberation like the Sigma 45mm F2.8--except with those are more not detecting when they should.
Remember Sony mirrorless can only detect contrast horizontally. Point at horizontal high-contrast. Bumping the ISO up to the second gain also seems to help in lower contrast situations.
Yes many a tree trunk has been my friend to regain focus !
Thanks
Karl
EB-1 wrote:
The Canons that have it are the R1 and R3, and that was done by alternating orientation of verticals/horizontals. OM cameras with quad pixels have it. AFAIK Nikon and Sony don't have it.
snapsy wrote:
@EB-1@ covered the direct technical reason why the on-sensor PDAF pixels can be blind to a severely-OOF subject and not know which way to drive the lens (or by how much). Why many MILCs choose to give up in this scenario and do nothing rather than at least attempt a speculative focus, even if it involves a time-consuming focus rack from infinity <-> MFD or alternatively, a contrast-detect cycle, is what's surprising. This tells me perhaps there are other, more common scenarios that yield the same PDAF pixel blindness, where those time-consuming focus-seeking actions would be more disruptive. For example, perhaps when there are temporary obstructions on a tracked subject (like a passing vehicle) that would yield the same blindness and make such speculative AF movements undesirable. ...Show more →
Makes sense Adam. Most often it happens when trying to get focus on a very OOF subject. Many missed opportunities on fast moving little birds. Lower light too certainly plays into it (far less contrast)
Thanks good to hear from you
Karl
Choderboy wrote:
I think many of us knew this going into mirrorless and adapted.
Most just get used to quickly pointing at something (very) roughly at a similar distance and focusing on that so the lens is far more likely to focus on the actual subject.
To be honest, in my case, barely an adaption. I used the same technique with DSLR but it was more of a speed issue.
The DSLR would usually force lens to move to extent of focus range and back trying to achieve focus.
For Canon, there was even a menu setting Lens drive when AF impossible. Using the point and focus at convenient alternate subject was just quicker than waiting.
As far as I know, Olympus is the only manufacturer to actually implement a work around.
EB-1 wrote:
It's a critical limitation of the MILS, supposedly due to the small effective baselength and inability to detect degree and direction of defocus
EBH
Can you (or anyone else) explain a bit more about what you are referring to regarding the "small effective baselength" and how that causes "inability to detect degree and direction of defocus"?
chiron wrote:
Can you (or anyone else) explain a bit more about what you are referring to regarding the "small effective baselength" and how that causes "inability to detect degree and direction of defocus"?
PD systems work by splitting incoming light at opposite paths and measuring their phase differential when both are combined on a set of detection sensors. For DSLR PDAF the angle/width of this split is large because the two opposing sensors are physically wide apart, allowing for a wide range of phase differentials to be measured. For MILC on-sensor PDAF, the two opposing sensors (ie, PDAF pixels on the sensor) are very close, limiting the range of phase differentials they can measure.
The baseline width is a measure of the distance between the split light paths.
snapsy wrote:
PD systems work by splitting incoming light at opposite paths and measuring their phase differential when both are combined on a set of detection sensors. For DSLR PDAF the angle/width of this split is large because the two opposing sensors are physically wide apart, allowing for a wide range of phase differentials to be measured. For MILC on-sensor PDAF, the two opposing sensors (ie, PDAF pixels on the sensor) are very close, limiting the range of phase differentials they can measure.
The baseline width is a measure of the distance between the split light paths.
Thank you for the clear explanation--makes perfect sense.
But, why don't they move the sensors farther apart on the MILC systems? Is the AF implication not enough of a problem or is there an advantage in the MILC configuration for having the sensors close together?
yes that has been the override that works but ...........
!