SCoombs
Well-known member
I have been posting a lot about this elsewhere in more general photography forums, but as I have done more testing and begun to hone in more I figure I may as well ask here, where some of the more technically saavy users I know of tend to reside. Thus far I have mainly noticed this with people, but thinking back over all my wildlife shooting I imagine it may have impacted that as well.
Essentially, in normal light - say EV 5.6 or higher - the camera operates entirely as expected. Below that, I am finding subject recognition to be extremely "misleading:" it will pick up on a subject and narrow in on the eye (following the standard body>face>eye hierarchy), displaying the green "focus confirmation" box on the eye, but the photos will be back focused. Very, very frequently if focused on the front eye, the rear eye will actually be in focus. Sometimes it is the rear of the head. It is very, very rarely anything in front, as if closest subject priority is responsible. Quite the contrary, it is back focused behind the closest subject - but remember, ONLY in light below EV 5.6 or so.
Please note that this is with any and all settings dialed in which will increase viewfinder brightness, since the EVF feed is used for focus. This includes d9 Adjust for ease of viewing and/or d10 Starlight view - I've tested with and without Starlight and there is no difference.
Now using a "dumb" mode like single point or dynamic area modes work much, much better. They are not flawless, but they get in focus results much more consistently. With subject recognition, I have had results range in tests from 30% in focus to 70% in focus at a live event I shot recently to this evening's test where it was more like 10% in focus and the rest back focused badly.
Also note that using wider apertures makes a difference. I have generally found in my testing near perfect performance at f1.8, very, very inconsistent performance at f4 and performance in between at f2.8. This may be unsurprising, but is worth noting.
This afternoon I tested at EV 6-7 and once again performance was almost flawless at both f1.8 and f4. When the light dropped, I did it again at EV4 and the results were extraordinarily discouraging. This was easily the worst performance I have seen as I had something like 90% of the photos out of focus and in particular they were all back focused.
For the evening tests, I set a fully manual flash so make sure the exposures were good enough to make a judgment on as without the flash they were extremely noisy at ISO 20000. Otherwise, everything is operating as usual so the focusing is unaffected.
Also note that while the AF box turns white after every picture, this is only because I was taking my finger off the AF-on button so I could press playback. I did this so that the videos contain an index of every shot by file name. A selection of RAW files from these tests can be found here if anyone wants to see EXIF data. These are mainly out of focus shots, but a few in focus ones are included for reference: https://drive.google.com/drive/folders/1qnD6wU3LGa5GVIwKsTG1dulmLYPknrEg?usp=sharing
Also note that because I was constantly shifting my hands to press the playback button the camera does move around a lot, but I made a conscious effort to try to keep it steady when actually shooting.
I did also shoot a number of shots in dynamic area. It actually made a few back-focusing mistakes as well this time and also had more trouble on acquiring focus than it typically has, but overall it still yielded sharp, in focus results most of the time.
Please also note that I did test this without the flash and I did record a video, just to try to see whether or not perhaps in some weird way the flash was causing the problem, but the results were the same. I have not uploaded the video because it didn't contribute much due to the low quality of the photos, but it did manage to verify that the issues occur with or without flash.
First, here is a video in which I review all of the shots. I stop at each one and, when the shot is out of focus, I circle the mouse around the area that appears to actually be where the plane of focus lies.
Second, here are the three videos shot at f4:
and here is the video at f1.8:
Essentially, in normal light - say EV 5.6 or higher - the camera operates entirely as expected. Below that, I am finding subject recognition to be extremely "misleading:" it will pick up on a subject and narrow in on the eye (following the standard body>face>eye hierarchy), displaying the green "focus confirmation" box on the eye, but the photos will be back focused. Very, very frequently if focused on the front eye, the rear eye will actually be in focus. Sometimes it is the rear of the head. It is very, very rarely anything in front, as if closest subject priority is responsible. Quite the contrary, it is back focused behind the closest subject - but remember, ONLY in light below EV 5.6 or so.
Please note that this is with any and all settings dialed in which will increase viewfinder brightness, since the EVF feed is used for focus. This includes d9 Adjust for ease of viewing and/or d10 Starlight view - I've tested with and without Starlight and there is no difference.
Now using a "dumb" mode like single point or dynamic area modes work much, much better. They are not flawless, but they get in focus results much more consistently. With subject recognition, I have had results range in tests from 30% in focus to 70% in focus at a live event I shot recently to this evening's test where it was more like 10% in focus and the rest back focused badly.
Also note that using wider apertures makes a difference. I have generally found in my testing near perfect performance at f1.8, very, very inconsistent performance at f4 and performance in between at f2.8. This may be unsurprising, but is worth noting.
This afternoon I tested at EV 6-7 and once again performance was almost flawless at both f1.8 and f4. When the light dropped, I did it again at EV4 and the results were extraordinarily discouraging. This was easily the worst performance I have seen as I had something like 90% of the photos out of focus and in particular they were all back focused.
For the evening tests, I set a fully manual flash so make sure the exposures were good enough to make a judgment on as without the flash they were extremely noisy at ISO 20000. Otherwise, everything is operating as usual so the focusing is unaffected.
Also note that while the AF box turns white after every picture, this is only because I was taking my finger off the AF-on button so I could press playback. I did this so that the videos contain an index of every shot by file name. A selection of RAW files from these tests can be found here if anyone wants to see EXIF data. These are mainly out of focus shots, but a few in focus ones are included for reference: https://drive.google.com/drive/folders/1qnD6wU3LGa5GVIwKsTG1dulmLYPknrEg?usp=sharing
Also note that because I was constantly shifting my hands to press the playback button the camera does move around a lot, but I made a conscious effort to try to keep it steady when actually shooting.
I did also shoot a number of shots in dynamic area. It actually made a few back-focusing mistakes as well this time and also had more trouble on acquiring focus than it typically has, but overall it still yielded sharp, in focus results most of the time.
Please also note that I did test this without the flash and I did record a video, just to try to see whether or not perhaps in some weird way the flash was causing the problem, but the results were the same. I have not uploaded the video because it didn't contribute much due to the low quality of the photos, but it did manage to verify that the issues occur with or without flash.
First, here is a video in which I review all of the shots. I stop at each one and, when the shot is out of focus, I circle the mouse around the area that appears to actually be where the plane of focus lies.
Second, here are the three videos shot at f4:
and here is the video at f1.8:
