There have been reports by reviewers in the past, regarding Nikon's infamous "Eyelash-AF" with shallow DOF.
It seems the z8 is not much better, although over at Petapixel, the former DPReview crew of Chris and Jordan did mention
in their Z8 review, that they were delighted that the z8 seems to have solved the eyelash AF issue.
I assume it's not quite fixed at f/1.2 or f/1.4, judging by your experiences.
OhSnap wrote:
Everyone here will lie to you because this is a Nikon forum and tell you that's fine, and it's you, when it's not.
Would you prefer that those of us having no issues at all make up stories claiming that we are having issues? What purpose would that serve? If something doesn't fit your narrative, it doesn't automatically make it untrue.
I know wedding photographers using Z9's and now Z8's that do not have any issues, and they are shooting all day long at wide apertures. Professional reviews have mentioned they were not having issues. I personally cannot replicate the issue on my Z9. I guess that's BS though because we're all a bunch of liars
CanadaMark wrote:
Would you prefer that those of us having no issues at all make up stories claiming that we are having issues? What purpose would that serve? If something doesn't fit your narrative, it doesn't automatically make it untrue.
I know wedding photographers using Z9's and now Z8's that do not have any issues, and they are shooting all day long at wide apertures. Professional reviews have mentioned they were not having issues. I personally cannot replicate the issue on my Z9. I guess that's BS though because we're all a bunch of liars
I was about to write exactly the same!
Why should we believe that an isolated report of improper eye AF performance represents the performance of the camera while many experienced users get truly excellent and consistent results shooting bright fast lenses wide open in a variety of situations.
I have provided concrete guidance about what settings work for me. There has been no reaction…
bernardl wrote:
I am getting excellent consistency with eye AF on the pupil with Z8/Z9 using various shallow AF lenses including 50mm f1.2 S, 85mm f1.2 S, 200mm f2.0 VR and 400mm f2.8 S shot close enough that pupil and eye lash have clearly different focus points. I even consider it a major strength of the Nikon AF compared to what I used to get with my Sony a9II.
For close up I get best results with AF-C set to trigger and focus, erratic with lowest delay setting (1) and all jpg image enhancing set to off.
I get worst results with AF-S even on supposedly static subjects due to the actual movements of subject and photographer. It matters on really shallow DoF situations like these.
Now, you son apparently standing outside has his eyes nearly closed (I assume due to bright conditions?) I tend to shoot in conditions where subject have their eyes opened. This may also have an impact.
Finally, although I personally tend to go to the edge of the gear because I bought bright lenses, the obvious reality is that 99.99% of historically successful portraits by the best photographers in the world were shot with significantly more DoF than what one gets with a 85mm f1.2 S up close. 😀
Jman13 wrote:
I honestly think this might be the real reason. Nikon has 493 PD points, and to maintain speed, rather than truly focusing at the exact pinpoint spot where the eye is detected, the eye detection algorithm is (correctly) analyzing the frame, and simply selecting the closest focus point to the detected eye. If it happens to be over the iris, you get perfect focus. If it happens to be nearest the eyebrow, you get focus on the eyebrow. The points are in a 17x29 grid, and that's good coverage, but does have some small gaps that could cause issues in the right circumstances.
For sports / wildlife / portraits with any sort of normal depth of field, this works great, and results in wonderful accuracy. (And I've noticed it tracks extremely well). But for very shallow depth of field work, it can sometimes result in missing by a bit. I think that if I do find myself in a situation where I'm splitting this razor DOF and need the precision, using Pinpoint AF in AF-S is the way to go, as that appears to use CDAF to confirm focus, and the ability to select position is much finer than the 493 points.
...Show more →
I’ve read the focus point count is fluff and solely the number of “selectable” points rather than a true count of af points, which is a much higher number than what’s listed in Sony and Nikons specs. Anyone know for sure?
Buckeye2604 wrote:
I’ve read the focus point count is fluff and solely the number of “selectable” points rather than a true count of af points, which is a much higher number than what’s listed in Sony and Nikons specs. Anyone know for sure?
This is correct.
Only someone interested in conveying a misleading impression that Canon has a “natural” lead would claim otherwise.
I am getting really tired about those troll posts.
OhSnap wrote:
I've heard the Z8 sucks compared to the Z9 for hitting the eye, and the Z9 was borderline when I used it.
Everyone here will lie to you because this is a Nikon forum and tell you that's fine, and it's you, when it's not.
The shallower the depth of field, the worse it gets. For reference, I was shooting with the 200mm f/2 and 85mm f/1.2.
Who exactly said everything is fine and its the original poster that is the problem.
Yeah, we Nikon shooter only shoot out of focus eyes and we all can't tell sharp from out of focus pictures
and are in denial. I am not sure why you would hang out in a forum where everybody lies.
May be you need to get a life and get out of here. How offensive and insulting can you be?
Buckeye2604 wrote:
I’ve read the focus point count is fluff and solely the number of “selectable” points rather than a true count of af points, which is a much higher number than what’s listed in Sony and Nikons specs. Anyone know for sure?
Right, but that goes for all of these systems; the Canon system uses basically all photosites for PDAF while Sony and Nikon use a small percentage of photosites for this purpose. These individual PDAF sensors are grouped into user-selectable points. The question then is how the camera decides what information to use for autofocus with subject detection. It is quite possible it uses the same groupings for PDAF as presented as user-selectable points. This is one potential explanation for the inconsistent results reported by the OP. Other factors that can influence the results include lens focusing speed and data communication rate (at close distances with shallow depth of field, the lens has to move fairly large focus elements quickly to account for slight variations in subject-camera distance), VR, shutter speed, settings in the AF system, hand-holding technique etc. and yes, of course, algorithms in the camera too.
Only someone interested in conveying a misleading impression that Canon has a “natural” lead would claim otherwise.
I am getting really tired about those troll posts.
Cheers,
Bernard
I believe most of us are trying to help the OP get better results (I certainly am, as I too like photographing people with shallow depth of field) and trying to understand why something occurs is a part of the process. Canon's system has its own advantages and disadvantages, mentioning the differences doesn't suggest one is pushing an agenda or trying to provoke anyone. Engineering is full of compromises.
ilkka_nissila wrote:
I believe most of us are trying to help the OP get better results (I certainly am, as I too like photographing people with shallow depth of field) and trying to understand why something occurs is a part of the process. Canon's system has its own advantages and disadvantages, mentioning the differences doesn't suggest one is pushing an agenda or trying to provoke anyone. Engineering is full of compromises.
I am totally fine with describing how Canon's AF works. My issues are with
- a misleading explanation of how Nikon's AF works,
- the misleading conclusion that Nikon's AF cannot be as good as Canon for eye AF as a result of that misleading explanation.. which goes against trying to help since it besically says that it cannot be helped.
bernardl wrote:
I am getting excellent consistency with eye AF on the pupil with Z8/Z9 using various shallow AF lenses including 50mm f1.2 S, 85mm f1.2 S, 200mm f2.0 VR and 400mm f2.8 S shot close enough that pupil and eye lash have clearly different focus points. I even consider it a major strength of the Nikon AF compared to what I used to get with my Sony a9II.
For close up I get best results with AF-C set to trigger and focus, erratic with lowest delay setting (1) and all jpg image enhancing set to off.
I get worst results with AF-S even on supposedly static subjects due to the actual movements of subject and photographer. It matters on really shallow DoF situations like these.
Now, you son apparently standing outside has his eyes nearly closed (I assume due to bright conditions?) I tend to shoot in conditions where subject have their eyes opened. This may also have an impact.
All of your points are very good. For action photography I'm used to having a slightly faster than middle setting for the delay (I use 2) but I haven't used erratic as I sometimes miss the subject with the focus area momentarily and I don't want the camera to go looking for something in the background during those moments. I haven't used different settings for portraiture simply because I often mix action and portraits in my shooting, but if doing only portraiture with a camera that does have subject eye detection, it makes a lot of sense to try erratic and the fastest delay setting because the subject isn't likely to run away for a moment. If I'm doing a portrait shoot I should try these settings although I haven't run into problems as it is. Maybe the results can be even better.
Excellent point about the bright light and slight squinting which brings the eyebrows closer to the irises and may make it harder for the camera to ignore the contrasty eyebrow detail. I find myself using wide apertures in two situations: one is when there is action and I want the background clutter minimized because I can't control what is happening in the background, example: person walking through a crowded street. However, in these situations I typically have at least one half of the body or full body of the person in the shot and depth of field easily covers eyelids and eyebrows along with the irises. The other situation is indoors in dim light when I want to reduce noise. In that case the subjects normally are not squinting because the lighting indoors is not bright. If I'm doing a formal portrait outdoors I usually have the aperture set to about f/5.6 because in those situations I have control over the subject position relative to the background and I don't need to blur the background to make the image work. I also try to avoid situations where the subjects are squinting because the images look better when the subjects have their eyes open. Finally a lot of people in my country have light-colored hair and the eyebrows are not as contrasty as in the OP's pictures. Thus I'm not running into the OP's problem with such frequency in my own photography.
Finally, although I personally tend to go to the edge of the gear because I bought bright lenses, the obvious reality is that 99.99% of historically successful portraits by the best photographers in the world were shot with significantly more DoF than what one gets with a 85mm f1.2 S up close. 😀
Absolutely! If looking into famous photographers' portraits in books and magazines, many of them have everything in the subject in focus (there is often an interesting background that tells part of the story of the person) and shallow depth of field is used much more sparingly than what many of us do today. Shallow depth of field is like a drug, you can get a rush from it but in the end it may not be what the image needs.
You indicated that you "picked up" a 105/1.4 to rent so an F-mount lens not a z-mount.
I found that all my f-mount lenses benefitted (needed) some AF fine tuning on my Z9. Some needed a significant adjustment. Do not believe those who claim it is mirrorless and so no AF-fine tuning is "needed". There is vast evidence that such responses are not supported by data.
I suggest you test the accuracy of the focus calibration - use a tripod and a simple ruler on a slope and a target next to the ruler.
bernardl wrote:
I am totally fine with describing how Canon's AF works. My issues are with
- a misleading explanation of how Nikon's AF works,
How exactly each of these AF systems work in detail is not public knowledge and I certainly don't have insider information about it. I presented the difference in PDAF point spacing as a possible contributing factor, not as *the* explanation. Some people report that Canon's system is able to detect bird eyes further away than Sony's, for example. Different algorithms can also be a factor but a natural explanation would be (to my mind) simply because Canon's system can gather adequate contrast-to-noise phase information from a smaller patch of sensor area because they use all the photosites for PDAF. Even if the density per se of the Nikon / Sony PDAF sensors is high enough, the quantity of light recorded by these embedded sensors (that cover a few per cent of the sensor area) affects the noise in the phase data and thus in practice the Nikon / Sony system probably has to average PDAF data from a larger area than the Canon to get useable PDAF data for focusing. How they do this analysis is not something I know; I suspect for simplicity's sake, Nikon group the PDAF data into the "points" that the user can select and also use these points to group the data in the algorithm where subject detection (based on all the photosites in the area) and PDAF sensor data are combined to focus on the eye. If this is how they do it, then it may be a contributing factor to the OP's results. The fact that he is getting better results by placing a single point on the iris is consistent with this theory.
Disadvantages of the Canon system is that they need twice as many photosites (what they market as a 45 MP camera really has 90 million photosites) and so there is more data that the camera has to process and one of the side effects is that the cameras tend to overheat more easily than others, also the first DPAF cameras had relatively slow focusing. This is a legitimate design compromise and Canon bet that having all that PDAF data would help them in the long run more than the initially sluggish performance because of lack of processing power, and overheating. Hopefully in the future they avoid these problems altogether.
- the misleading conclusion that Nikon's AF cannot be as good as Canon for eye AF as a result of that misleading explanation.. which goes against trying to help since it besically says that it cannot be helped.
This isn't the conclusion I would make. My own experience with the Z8 is that I get about 99% or higher percentage of shots in focus in the kind of people photography that I have done with that camera, photographing subjects including people on the street, dancing in midnight, stage performances, a wedding, a funeral and a portrait shoot. Most of these images were made with custom group area with human face/eye detection on, and some with 9-point dynamic area. I thought the focusing of the Z8 is exquisite and is way better than I had expected. Thus my own initial internal reaction to the OP was disbelief. But I have to take facts presented as they are and if he is getting better consistency of results with the Canon then there is probably an explanation to it. That doesn't mean the results can't be improved by adjusting settings and technique. I'm certainly hoping that the OP can get good consistency of results with more experience with the camera.
Steve Spencer wrote:
Let me offer a perspective that is perhaps sacrilege here. I think there can be too much emphasis on would ha on the iris. For me when I take a portrait I want not only the iris to be in focus but the tip of the nose as well. That means that the depth of field needs to be at least an inch and with some people 2 inches and these shots have a lot less depth of field than that at least with the sort of magnification presented, which is of course a lot more magnification than would ever make sense for a full image. This is the sort of magnification for an 8 X 10 or 12 X 15 print where you would be putting your face right up next to the print. I think for prints of this size 11 of the 12 shots would have the Iris in focus from a reasonable viewing distance. Still for my tastes the depth of field may be too shallow and I doubt the tip of the nose would be in focus even with reasonable print sizes and reasonable viewing distances. It seems to me that from this distance any reasonable portrait would need a narrower aperture something like f/1.8, f/2, or maybe even f/2.8. Would the iris be in focus with this wider depth of field? I think all these images would be fine for focus with such a wider depth of field for my tastes and I think the solution here is simple. For such shots stop down a bit.
Why does this matter? Because I am not sure that with really shallow depth of field I want the plane of focus to be right at the iris. Some of these shots with the plane of focus just in front of the iris may well turn out better as the tip of the nose might fall more within focus and the whole of the face will have better focus, for my tastes at least. I am just not sure that the focus on the Iris being in focus from extreme magnification is the best way to judge a portrait. I suspect it is not, for my tastes at least.
This is not to say that there are no advantages to a fast lens as in this case if Jordan took one step back, the depth of field would increase a lot and f/1.4 would likely work well and there would be no issues with the Z8's eye AF.
Nor is it to say that Nikon's system isn't a bit behind, say Canon's system. I think the issue isn't accuracy, however, but precision which is always something to consider with any system or instrument. How precise does it need to be? My bathroom scale, for example, seems only to be precise to about a pound. I can step on the scale repeated times and get pretty different numbers. I fine with that as I don't need a really precise measurement of my weight. Weight naturally fluctuates anyway, so precision isn't essential. For the Z8, eye AF seems to be precise to about 1/4 of an inch, which for me would work fine in pretty much all of my photography because I don't need the extreme magnification that Jordan is showing here and would want portraits with more depth of field and more depth of field means less precision is needed. Canon in contrast from Jordan's experience seems to have precision to something like 1/64 of an inch. There is nothing wrong with that and I would take it if there were no other tradeoffs, but for me I just don't value that extra precision as it would very rarely be useful to me. Just like I don't value a bathroom scale that would be precise to the once as that sort of precision I would never find useful.
We can and should differ on how much precision we want in an AF system. Precision is a good thing but precision also always reaches a point where additional precision adds very little. For me the additional precision of the Canon system's eye AF, adds very little for my type of portraiture over the Z8 system's eye AF. Of course other people's MMV.
...Show more →
Absolutely spot on. What 'portraiture' I did professionally, using Hasselblad, Mamiya etc would not have satisfied the client if nose and ears hadnt been 'sharp', within the limits of DoF, viewed on a 10x8 print viewed at a 'normal' distance, not with a magnifying glass. The distance between the planes of the tip of my nose and my ear lobes is more like 3 inches, so although 'interesting' if you've nothing better to do, testing and worrying about less than half an inch between eyebrow and pupil is really not of any real importance. I doubt if the focus screen on the cameras we used and my eyesight would have got as close as that, but it didn’t, and doesn't, matter.
Wonderful that AF has got as good as this really.
I’ve responded to this claim in another thread, and let me just say that I strongly disagree with your assessment that there is vast data that says otherwise. I’ve been shooting with mirrorless for 13 years, reviewing them for 11 years, with almost every mirrorless system and well over 150 lenses, both native and adapted and have never had a focus miscalibration. Not once. Meanwhile I picked up a D750 three months ago with a few lenses and needed to fine tune 3 of 6 lenses. All three of those are perfect on my Z cameras. As is the 105/1.4 when shooting in non eye tracking situations (and even in eye tracking it does hit more often than it misses).
I’m sorry, but I am perhaps one of the more qualified people to say: you’re not correct here. Mirrorless doesn’t require AF fine tuning.
agrumpyoldsod wrote:
You indicated that you "picked up" a 105/1.4 to rent so an F-mount lens not a z-mount.
I found that all my f-mount lenses benefitted (needed) some AF fine tuning on my Z9. Some needed a significant adjustment. Do not believe those who claim it is mirrorless and so no AF-fine tuning is "needed". There is vast evidence that such responses are not supported by data.
I suggest you test the accuracy of the focus calibration - use a tripod and a simple ruler on a slope and a target next to the ruler.
agrumpyoldsod wrote:
You indicated that you "picked up" a 105/1.4 to rent so an F-mount lens not a z-mount.
I found that all my f-mount lenses benefitted (needed) some AF fine tuning on my Z9. Some needed a significant adjustment. Do not believe those who claim it is mirrorless and so no AF-fine tuning is "needed". There is vast evidence that such responses are not supported by data.
I suggest you test the accuracy of the focus calibration - use a tripod and a simple ruler on a slope and a target next to the ruler.
Z-mount Nikkor lenses are much more accurate.
All the sensing and decision making is done in camera, the lenses have motors which are told what to do by the camera. In a dslr the focus sensor is in the base of the cameraand has to be adjusted to the same register as the film/sensor plane, as does the focusing screen. My D7000 was hopelessly out when bought, but when adjusted has been OK ever since. Presumably the need for fine tuning is to do with ray angles striking the sensor differently because of optical design. In a 'mirrorless' camera the AF sensors are on the sensor itself, so it doesn't matter about such things and the lens will simply be told to focus until sharpest point reached, no parallax errors , no register or alignment errors, just the inherent inaccuracy tolerances of the system.
And 'pilot error' of course, which is often evident! So many think it should be 100% accurate all the time even if they don't bother to learn how to use it properly (not necessarily relevant in this case)
I have also believed that mirrorless cameras don't need to be fine tuned. I have never done it, and have never noticed a need for it.
However, the makers of FoCal, an AF fine tuning system, have published an article on the Z9 saying that improvements can be had by fine tuning both F and Z mount lenses for use on your Z(and presumably all your mirrorless cameras). They claim to have collected a lot of data confirming this need. Whether this is an attempt by them to keep their product relevant in the age of mirrorless cameras I don't know. But it may be where agrumpyoldsod gets his information regarding this issue.
Wait, the company that relies on selling focus tuning aids, which will be made obsolete by mirrorless cameras, is claiming you still need to buy their products? Ok.
Not buying it for a second. I have also collected a lot of data. As in hundreds of thousands of images with a huge variety of lenses and cameras.
