A7III, Eye-AF: next episode

After reading RubberDials extensive post where he expresses his findings with the complaints about Eye-AF i've been watchting that thread all day to find some responses. Only to find out that the previous thread is full. I think the discussion was not finished so the discussion deserves a new thread.

I am on the fence of buying the A7III, but still hesitant because of the reported problems with Eye-AF. I think ATL Shooter did a great job in showing why it did not work as he expected. Rubberdials explained why ATLShooter might be wrong. But that's where the discussion ended.

I really hope that there is nothing actually wrong with the camera and I think Rubberdials did a good job explaining that there might be not such a big problem. However, the examples ATLShooter showed us, led me to the impression that there really is something wrong. So to let this discussion bleed to death is not desirable in my humble opinion.

Anyway: link to the old thread: https://www.dpreview.com/forums/thread/4278856?page=8#forum-post-61125299

I hope that discussing this matter further will shed some brighter light on this ( pun intended).

This doesnt give me confidence in buying this camera yet,

ATLshooter wrote:

Soooo, if I'm wrong, how do you reconcile that with the fact that I have an a6500 and now an a7riii that both do eye af perfectly? Even in the same situations, with the same lenses, same everything?

Hi Altshooter - I addressed this in my earlier post that unfortunately was the last in the thread. https://www.dpreview.com/forums/thread/4278856?page=8

What happened with your A6500 or A7rIII do not explain anything about your A7III because they are different cameras.

I note that the depth of focus of the A6500 image you posted was great and that neither of the two A7rIII photos I've seen from you have sharp eyes either:

https://www.dpreview.com/forums/thread/4280294#forum-post-61109691

https://www.dpreview.com/forums/thread/4277930

That doesn't make a lot of sense.

Last saturday I did a whole portrait shoot (including some extreme close ups) using literally nothing but the eye af. Over 500 pictures where I used ONLY eye af. Nearly every shot is in perfect focus, as in 99.99% of them. I think I culled less than 5 due to focus.

Yet, with either 7iii, eye af failed nearly every shot. The closer you got, the worse it was. I could get close enough (but still not min focus distance, I proved that) that it would fail EVERY time.

You only posted four shots taken with Eye AF and of these only 1 was unacceptable - and it was taken close up, where depth of focus is thinner than at medium distance or infinity.

If someone needs to say I was wrong to make themselves feel better about it, then go for it. But I'm using my 7riii the exact same way with zero issues. I literally pulled it out of the box, and used eye AF immediately and posted the results. Yet somehow I was the cause of the eye af failure on the 7iii's? Lol ok. Not to mention Sony has now confirmed they are working on a fix for it. If there's no problem, how can there be a fix?

To the best of my knowledge Sony hasn't confirmed there is a problem. Various people have had these mollifying statements when they have complained but this is a different thing entirely. Companies will often tell you there is a problem and they are working on a fix - this is a standard strategy used by customer service departments. By telling you there is a known problem they essentially close down your argument and defuse your anger. They may return your defective article with no change but many people will think it is fixed.

If there was a known issue Sony would have made an official announcement and offered to take back affected cameras. As far as I know they haven't done this.

Incidentally, no-one thinks you are being disingenuous or lying and no-one is saying you are wrong to make themselves feel better. This community isn't like that. We have some very knowledgeable posters and experienced photographers here and people don't let things hang. If there's a problem it gets resolved or explained.

[Note: the FW update for the 7riii regarding eye af wasn't an issue with the actual use of it. There was a setting that had to be turned on for eye af to work (actually two, and either of them had to be on). That was what they fixed in the FW update, now you can turn those off and still use eye af]

NOTE: This was my post to the earlier thread, which unfortunately was the one that reached the thread limit so nobody was able to reply. Rather than get people to seek it out I've decided to re-post it, so that the affected people can comment and everybody with new issues can see what was said.

To see the comment in context go here: https://www.dpreview.com/forums/thread/4278856?page=8

It is the last post.

++++++++++

jameszhan wrote:

RubberDials wrote:

jameszhan wrote:

A while ago, ATLshooter had 3 huge threads about eye-AF failure on two copies of A7 III. I thought I should be okay because only 8 people were reporting the same problem in this forum.

I was wrong. My copy (serial number 3379xxx) of A7 III has the exact same issue.

I would caution you against collecting a list of other people's problems and trying to present them as a trend or an indication of something. It's almost impossible to know whether people are having similar problems or whether their issues are occasioned by the same things or are user error or something else all together.

I look at these threads often and I have seen no issues of eye-AF failure that you are referring to, let alone cases from '8 people'

You sure you look at these threads often? You must have missed all the reports.

Please read the following threads if you have the time to. If you make a count, there are now more than 10 people experiencing the same issue.

https://www.dpreview.com/forums/thread/4273349

https://www.dpreview.com/forums/thread/4275541

https://www.dpreview.com/forums/thread/4276377

https://www.dpreview.com/forums/thread/4277220

http://www.fredmiranda.com/forum/topic/1540503

http://www.fredmiranda.com/forum/topic/1543271

I don't expect you to read them, but I'm just showing you this is not just an isolated case and there are more and more people reporting the same problem, some of them are in this very thread we are in.

Right then James, although you didn't expect me to I read all of the threads you linked. Took me a while!

As I expected, and why I hadn't seen it - there weren't 8/10 people with the issue there was one - ATLshooter, and now yourself. Several people with no previous posting history chimed in and said 'yeah, mine does that' but unless people provide detail and test shots you can't presume that they are experiencing the same problem. This is self evident - if someone has a headache because they have a brain tumour it doesn't mean everyone with a headache has a brain tumour. The issue is further complicated by the incessant trolling that takes place on Sony forums.

Of course the fact that some people take the opportunity to bash Sony does not mean that there may be no issue, so let's look at that.

I didn't do any controlled scientific testing; I didn't need to because I used the same lens, 85mm f1.4 GM, on two copies of A7 III (one mine, one in-store on display shelf), and mine was clearly front-focusing while the other was not.

Sony mirrorless cameras do not front or back focus, this is an issue that occurs when the focussing sensor is not exactly on the film plane. Mirrorless cameras can mis-focus and they can focus on things that you didn't want them to focus on but they can't front or back focus.

Well, I'm using the term "front-focusing" quite literally, meaning the focus is achieve at a plane before the intended plane, and that is certainly proven possible with mirrorless cameras, though it's not due to the misalignment of the AF module found in DSLRs.

Here's the A7 III front-focusing, courtesy of ATLshooter: https://youtu.be/ueudeH23ZXg

Here is another thread demonstrating A7 III front-focusing (possible cause is a misalignment between the displayed focus point and the actual PDAF pixel on the sensor in A7 III): https://www.dpreview.com/forums/post/61108811

Okay, that's understood. I think a better term would be mis-focussing because front and back focus is a calibration issue and Sony e-mount cameras are self calibrating.

The discrepancy between the displayed focus point and the actual one has been mentioned before. I have no reason to doubt that this is a real issue.

I'm exchanging my copy tomorrow. I'm making this thread just to raise some awareness because I suspect the problem may be more widespread than we thought.

That's very kind of you but if there was a problem I think it would have been seen by reviewers and some of the more dedicated users on here.

It has been, as shown in the links above. But keep in mind, it doesn't seem to affect all A7 III bodies.

Okay. I don't believe that it has been shown, nor in the examples you gave at the start of the thread.

Apologies in advance for talking about ATLshooter in the third person. It will be confusing if I split my post between him and you. ATLshooter, please jump in at any point.

ATLshooter (hereafter referred to as ATL) created three threads about his AF difficulties and claimed that he had two A7IIIs and they didn't eye AF properly, they 'front-focussed'.

ATL posted several pictures to demonstrate this. The first is here:

https://www.flickr.com/photos/139655841@N03/41408864121/in/shares-gHiJp4/

No one could argue with his claim that eye AF has not worked. Neither eye is in acceptable focus.

ATL then links two more pictures. Only one is with eyeAF, the second one. Of it he says:

"shows what eye af does almost 100% of the time. If you download the image and zoom up you will see that only the bridge of the nose is in focus. The eye lashes are definitely out of focus, and the eye brows are even slightly out of focus. Only the bridge of the nose is dead in focus."

Here is the picture:

https://i.lensdump.com/i/8BO4nF.jpg

In fact the eyelashes on the model’s right eye are in focus and this is what the camera has focussed on. The eyelashes on the right eye are NOT out of focus.

He then posts four more pictures.

Two are close up and two head and upper body.

#2 and #4 are with Eye AF

The first shot with eye AF is not that sharp. but Depth of field is much thiner at close distances. ATL doesn’t seem to appreciate this.

The second shot with eye AF (The 4th overall) is pretty sharp although ATL says:

"Fourth shot: Further away using eye AF. Still missed, focused on the bridge of the nose. It's harder to see but the eyes are definitely out of focus."

Here are links to the two photos.

First eye AF photo (2nd overall) https://www.dpreview.com/forums/post/61020370?image=1

Second eye AF photo (4th overall) https://www.dpreview.com/forums/post/61020370?image=3

This is not the sharpest possible, but the eyes are definitely NOT out of focus as ATL claims.

That’s it. Despite filling three threads to capacity complaining about the eye AF, ATLshooter has only actually posted FOUR pictures taken with eye AF. Of those four, one was taken in extreme close up, where the eyeAF has failed. The second was also close up and is poor. Of the other two, BOTH are acceptable, though one is sharper than the other. Neither is as sharp as the spot focus with contrast phase. This is not unexpected behaviour. Without the final contrast detect phase (where the lens cycles in and out) which doesn't happen in eye AF AF-C, the camera is only using phase detect. It is simply not as precise at large apertures where depth of field is very shallow as the full hybrid focussing cycle.

I noted the following about the ALT’s testing. He did not use a tripod. He did not use the same position or composition for any shot. It was pointed out to him that the focus was on the plane of model’s eyelash on more than one occasion, but he continually referred to the focus ‘jumping onto the bridge of the nose’ which cannot have happened. ATL was deriving his eye AF expectations from using the system on aps-c and at smaller apertures.

ALT ignored three posts by Horshack, who is one of the most knowledgeable members of the community who suggested a simple test for confirming the mis-focus was a phase detect issue.

ALT used the fact that he got the results he wanted with the A7rIII (a completely different camera model) to ‘prove’ that both his A7III’s were faulty. This is a logical fallacy.

The following two threads by ALT demonstrate a certain unfamiliarity with the system and perhaps FF photography (He apparently did not know what lateral chromatic aberration was).

https://www.dpreview.com/forums/thread/4280294#forum-post-61109691

https://www.dpreview.com/forums/thread/4277930

Please go test out your A7 III's eye-AF with a fast lens before the exchange/return period expires.

No need to debate whether it's user error or not as it is already been argued to death in ATLshooter's threads. And in my opinion, it is quite clear cut that this is not user error.'

Lol. Sadly when it comes to issues to do with focussing or lens sharpness it usually is user error. Never wise to dismiss it.

Sure.

Note that this two shots are not the only ones I took. I was using my 24-70 GM at f2.8 and I already noticed a bit of softness at the eye with eye-AF, which was why I went into the store today and tested it with a f1.4 lens.

I would like to thank ATLshooter for raising awareness of it. I don't think I would go in store to get a f1.4 lens to test if I didn't know this problem could exist in some copies of the camera. I currently don't have the budget to buy another lens, and if I found out about the problem when I saved up enough to buy a fast prime, it would most likely be too late for me to exchange.

It doesn't look to me as if there is anything wrong with your camera. In the first photo the focus is on the left eye (the left eye of the model) so I'm not sure why you have reproduced a detail of the right eye. The face is at an angle and the depth of focus at maximum aperture is very small.

Eye-AF box picked up the right eye and confirmed focus on it; instead in the final image, the focus is on the left eye, proving a back-focusing issue. I had her face at an angle on purpose so it's very easy to see if the focus is front-focused or back-focused. That's how you check front and back focusing, traditionally with a ruler at a 45 degree angle.

If I said to you I'm going to pick up the green ball but instead I picked up the red ball, which ball would you say I picked up? The fact that the focus is on the left eye suggests the camera focussed on the left eye, rather than the camera focussed on the right eye and somehow mis-focussed.

Eye-AF box picked up the right eye and confirmed focus on it; instead in the final image, the focus is on the left eye, proving a back-focusing issue. I had her face at an angle on purpose so it's very easy to see if the focus is front-focused or back-focused. That's how you check front and back focusing, traditionally with a ruler at a 45 degree angle.

If I said to you I'm going to pick up the green ball but instead I picked up the red ball, which ball would you say I picked up? The fact that the focus is on the left eye suggests the camera focussed on the left eye, rather than the camera focussed on the right eye and somehow mis-focussed.

I can't speak for ATLShooter of course, but I think that what he meant was that even when he pressed the button on the camera that shows where it focused, it showed the eye that was not sharp, altough the other eye was sharper.

What do you think could explain that?

Kenneth16 wrote:

Eye-AF box picked up the right eye and confirmed focus on it; instead in the final image, the focus is on the left eye, proving a back-focusing issue. I had her face at an angle on purpose so it's very easy to see if the focus is front-focused or back-focused. That's how you check front and back focusing, traditionally with a ruler at a 45 degree angle.

If I said to you I'm going to pick up the green ball but instead I picked up the red ball, which ball would you say I picked up? The fact that the focus is on the left eye suggests the camera focussed on the left eye, rather than the camera focussed on the right eye and somehow mis-focussed.

I can't speak for ATLShooter of course, but I think that what he meant was that even when he pressed the button on the camera that shows where it focused, it showed the eye that was not sharp, altough the other eye was sharper.

What do you think could explain that?

The paragraph you quote is in reply to James Zhan, not Altshooter. The point I was making is that whatever eye is in focus is likely the one the camera was focusing on. It's illogical to assume that the camera indicated one eye and misfocussed on it, but the misfocus randomly allowed the other eye to be in perfect focus.

RubberDials wrote:

What happened with your A6500 or A7rIII do not explain anything about your A7III because they are different cameras.

The way eye-AF behaves in other A7III, however, does. Other users are not having the problem we are. Shooting 1.8, 1.4, Native & Adapted lenses, MFD or not. They achieve perfect focus on the eyeball. We don't.

stopandwatchthesky wrote:

RubberDials wrote:

What happened with your A6500 or A7rIII do not explain anything about your A7III because they are different cameras.

The way eye-AF behaves in other A7III, however, does. Other users are not having the problem we are. Shooting 1.8, 1.4, Native & Adapted lenses, MFD or not. They achieve perfect focus on the eyeball. We don't.

Hi there.

Did you read my long post? Without controlled testing it's impossible to say what people are experiencing as there are so many variables. Below for example are two photos used by James to confirm that his A7III had a focussing problem. The first thing that stands out are the different depths of focus present in the shot. Many users, perhaps because they are new to FF do not seem to understand the immensely thin depth of focus at large apertures. The depth of focus becomes even thinner at close distances.

See below for examples

Her right eye was marked by the green box, but it was out of focus

With a different copy, this time her right eye is tack sharp

James says the green box was focussed on the right eye but the right eye is out of focus. That's true. But the left eye is in focus. The plane of focus is very thin and does not cover the whole eye. I suggested to James that the camera has actually focussed on the girl's left eye. But he still is presenting this image of as an example of the camera back focussing.

The picture he uses to confirm his second A7III body is working correctly and his first one wasn't has the girl's head turned more to the front, putting both eyes closer to each other on a plane of focus.

It is easily possible that the box jumped from the right eye to the left eye at the moment of exposure and James didn't see it or it wasn't displayed and that there is actually nothing wrong with either camera.

The examples given are simply not proof of anything. (James please jump in if you feel I have misrepresented anything here - obviously that is not my intention).

Stopandwaththesky, I haven't had a close look at your samples but will if you want. It's worth noting the following:

• Continuous AF is not as accurate as single shot because it is phase detect only.
• Contrast detect AF systems have been known to give a focus confirmation before an image reaches peak sharpness - this is possibly exacerbated with high-contrast subjects and very sharp lenses. Any testing of eye AF should be done on an eye without false eyelashes or heavy mascara as a matter of course.
• Focus is an approximation. There is no such thing as 'correct focus' only 'enough focus' or 'focus where you want it'. Some shots will always have 'better' focus than others, on any system working 100% as intended.
• Do not take your sharpest shot and then assume that anything not as sharp shows the system as faulty - this is an example of the 'Texas sharpshooter fallacy'.
• Eye AF jumps around. If it fails the system resorts to face detect, which might mean you have a perfectly focussed chin.
• Shooting wide open at very large apertures close up is not an intelligent way to get guaranteed in focus shots all the time. (Obviously if that's the style of shots you want to take you have no option.)
• E-mount lenses have linear motors and are capable of very precise placement. It is perfectly possible that ring-mount motors used in adapted lenses in AF-C are just not as precise.

While I appreciate you explaining the limitations of AF and what can be expected, in most of my pictures where the DOF was so shallow that it is easy to tell where the camera focus, the focusing point was on the nose, or the eyebrows, consistently. Which means when I press eye-AF, I can expect my camera to focus on the eyebrow/eyelash, if not the nose. I would swap to manual focus, and peaking was not on the eyeball, but on the bridge of the nose. 
While I'm trying to say is while eye-AF is not 100% accurate, My camera does not respond in the same way as other users'. I have checked youtube videos of people using native and adapted lenses, and in those units, AFC + eye-AF was focusing sharply on the eye . Don't get me wrong, I'm the first one to try as many things as I can before I have to send my camera away as I live in Argentina and it'll be a whole month before I have the camera with me again.  
However, after exchanging a few emails with Sony, they agreed to have my camera reviewed. I'll send it to them the first week of july, so they can tinker with it. If this means my problem could be fixed and I'll have a camera able to delivering pictures in focus with eye-AF, so be it. At worst, I'll be stuck without a reliable eye-AF, or maybe sell it to someone who wants use it for video.

RubberDials wrote:

Stopandwaththesky, I haven't had a close look at your samples but will if you want. It's worth noting the following:

• Continuous AF is not as accurate as single shot because it is phase detect only.
• Contrast detect AF systems have been known to give a focus confirmation before an image reaches peak sharpness - this is possibly exacerbated with high-contrast subjects and very sharp lenses. Any testing of eye AF should be done on an eye without false eyelashes or heavy mascara as a matter of course.

Images with good lighting and no make up of any kind, the camera focused on the front part of the nose, while the eye-AF green box was centered on the eyeball.

• Focus is an approximation. There is no such thing as 'correct focus' only 'enough focus' or 'focus where you want it'. Some shots will always have 'better' focus than others, on any system working 100% as intended

Agreed, but my camera is consistently focusing somewhere else but the actual eyeball.

• Do not take your sharpest shot and then assume that anything not as sharp shows the system as faulty - this is an example of the 'Texas sharpshooter fallacy'.

I don't. When checking all my images, more often than not, images show eyes out of focus.

• Eye AF jumps around. If it fails the system resorts to face detect, which might mean you have a perfectly focussed chin.

Except the subject was still and the focus square was tiny tiny on the eyeball (eye-AF) and not big (FD)

• Shooting wide open at very large apertures close up is not an intelligent way to get guaranteed in focus shots all the time. (Obviously if that's the style of shots you want to take you have no option.)

Alas, I shoot wide open. Artistic preference. This is why I bought the camera, to assist me with my shooting style. If I'd wanted eyes not in focus, I would still be using my Canon 5dmkII!

• E-mount lenses have linear motors and are capable of very precise placement. It is perfectly possible that ring-mount motors used in adapted lenses in AF-C are just not as precise.

A tripod + very still subject should take of that... but it doesn't. At least for me.
Like I said before, I appreciate you breaking down the possible reasons why it could be not working properly. If any, I already tried ruling out every possible bullet point. However, I still believe my unit is not as accurate as other people's.