SCoombs
Offline
• •
Upload & Sell: Off
|
 p.14 #8 · Z8 extremely unreliable/inconsistent AF in mildly low light | |
jimmy462 wrote:
Hi snapsy,
Late to the party here, thanks for taking the time to make those test videos, I found them very informative....my 2˘ herewith...
What I'm seeing is that both the a7R IV and Z8 are both preferring/prioritizing the brighter, more-contrasty right eye (left eye on screen) of your, er, model. Both cameras utilize Hybrid (combination Contrast-/Phase-Detect) AF systems so there must be some level of computational "weighting" component for the contrast part of their focusing algorithms, and (it seems) that for both cameras that the contrast weighting will override any proximity weighting for their eye AF.
What I'm also seeing is the Z8 trying to acquire the more-proximate (read: more-preferred to the OP's needs) left eye (right eye on screen) of your test model where than the a7R IV never seems to. Clearly two different "AF soup recipes" being employed here.
My initial take-away is that the Z8's eye-detection algorithm has some, er, "awareness" that when it's detecting an eye it also seems to know that there is also a second eye present, whereas the a7R IV never once put an AF box (dancing boxes, yes, one time) around the closer eye and does not appear to have any level of, er, "two-eye awareness" in your tests.
___________________
Establishing a baseline is needed here...
Your tests at 1/160, f/5, ISO 20,000 to create an EV 12 image translates into an Light Value of EV 4.33 at ISO 100, pretty dim conditions. So, a redo of these tests at 1/160, f/5, ISO 100 where the scene lighting is ample/bright enough to provide for an actual Light Value of EV 12 conditions (meaning that the exposures with be neither under- or over-exposed but closely resemble the model's skin exposure in your current ISO 20,000 test scene) might reveal some behavior differences with the Eye AF for both cameras...or, maybe, not.
It appears that the Sony has gone "all contrast detect" in your current test, where the Nikon has is having a battle between Contrast and Phase detection. I'm curious if there is an illumination threshold (which would likely be different for both cameras) where both camera's will weigh more-heavily with Phase detection vs Contrast detection.
Your thoughts and feedback?
Jimmy G...Show more →
I think this makes some sense. The outstanding issue here is that the camera is displaying focus on one eye when the focus is actually on the other (and as it's sometimes showing focus on the closest eye but actually focusing on the far, we know it's not just a proximity thing, as I think you're noting).
In other words, if the subject detection system was choosing the more contrasty eye and then displaying this choice in the form kf the AF box, I don't think we'd have the same sort of problem here. The issue is that it's displaying focus on whichever eye it is or whichever it's been directed to by the user, but actually, as if "secretly," it is going for the other eye. If that's because it's more contrasty that makes sense, but it should still not be reporting focus on the eye it's not going for, AND this is especially true if the user has specified an eye and the camera is ignoring that and focusing on the other eye all while telling the user "okay, I'm on the eye you want."
Another thought: was the switch to other eye programmable control added in 2.0? I can't remember. If so, I wonder if the coding change behind that could.be messing with things a bit. It may even explain why some people haven't seem this problem if they haven't shot in the right circumstances for it since 2.0.
|
That said, sometimes its wisest to know when it's finally time to bring in the manufacturer to have them help you, er, "fix their thing"! Haw!