Focus issues of Fuji GFX50

[cgarnerhome]

Fair enough.  I guess that's why some people rent the systems first.  I guess us early adopters pay some price for our impatience

[chrismuc]

Is it really true that GFX auto-focuses at working aperture? I know that my A7rII does and it causes me headaches as even the GM lens occasionally does focus-hunting if the aperture is f/8 or smaller.

I read that you can focus manually at working aperture by using M mode and pressing DOF button.

I checked:

At apertures from f2.8 to f5.6, the camera closes the aperture to that value and performs the auto-focus at that f-stop as soon as the shutter is half pressed.
At apertures smaller than f5.6, the aperture is first closed to that particular value, so one sees in life-view the world at that chosen aperture but as soon as the shutter is pressed half to focus, the aperture opens to f5.6 (that means auto-focus is carried out at f5.6), as soon as the shutter is fully pressed for the exposure, the aperture is closed to the first value.

So in first case, the slight focus-shift has no impact on focus precision because focusing is carried out at working aperture, in second case focus-shift has very little impact, because only the slight shift between f5.6 and smaller apertures is relevant.

Interestingly, f5.6 seems to achieve enough contrast for the focusing, otherwise Fuji would not have implemented it in such way.

In comparison the Sony A7RII:

w/ Batis 85f1.8
The behaviour is different than described in a previous post of that thread, sorry my bad.
At any working aperture, the camera closes the aperture to that value (life-view at that value like GFX), pressing the shutter half for auto-focus: the aperture is opened nearly completely but not totally, let's say to f2. Shutter fully pressed for exposure, the aperture closes again to the first value.

w/ Loxia 21f2.8
Because the aperture of the Loxia is fully mechanical, it stays closed at the chosen value. So one sees the life-view image at that aperture value and also must manually focus at that value.

[hubell]

The other thing about Lloyd's reviews that make them so valuable to me is that Lloyd doesn't own most of the equipment he reviews. That's true with the GFX and the X1D. When someone owns a camera or lens and "reviews" it, the outcome always seems preordained. I mean how often do you see someone who owns a camera or lens or is an "Ambassador" who has been provided a camera or lens to review give that camera or lens a negative review.

[Michael Erlewine]

The other thing about Lloyd's reviews that make them so valuable to me is that Lloyd doesn't own most of the equipment he reviews. That's true with the GFX and the X1D. When someone owns a camera or lens and "reviews" it, the outcome always seems preordained. I mean how often do you see someone who owns a camera or lens or is an "Ambassador" who has been provided a camera or lens to review give that camera or lens a negative review.

I don't care whether they own the equipment or not, but mostly about what they have to say about it that I can weigh in my mind. With the odd lens to check out, I can always send it back. But with systems like the GFX, the X1d, and even the Pentax K1, a lot of work can be Required. With the K1, I had to spend months finding the lenses that I could use on the system, only to send it back and gradually sell off the lenses. I still have the X1D system and am thinking of keeping it, but time is running out for me to decide. And I can't find an extra battery either.

[Quentin]

"My take on the two Fuji lenses is that the 120 macro is a tour de force..."

Well said, Jim.  It is one of the best lenses I have ever used - and I have used plenty - and a country mile better than the Hassy HC 120mm.  And that is before one adds in the benefit of OIS on a macro lens, incredibly useful if you use hot / LED lights in a studio, or shoot portraits or botanical work in moderate light. The 32-64mm is merely excellent. 

I get the impression some people resent the GFX and look for faults that are not there or exaggerate those that are.

Focus issues?  I have found none.

[Jim Kasson]

The other thing about Lloyd's reviews that make them so valuable to me is that Lloyd doesn't own most of the equipment he reviews. That's true with the GFX and the X1D. When someone owns a camera or lens and "reviews" it, the outcome always seems preordained. I mean how often do you see someone who owns a camera or lens or is an "Ambassador" who has been provided a camera or lens to review give that camera or lens a negative review.

Lloyd buys quite a lot of his own gear if the lists of his used equipment for sale that appear irregularly are any indication. He also has some kind of relationship with Zeiss, as indicated by his trip report last year.

I see no indication that either of those serve to make Lloyd a biased reporter. He is scrupulous in his testing methods, and documents what he finds in detail. I like quantitative testing, but I pay happily to subscribe to Lloyd's qualitative stuff. In fact, most of the time, we end up looking at the same issues from different perspectives. Sometimes we disagree, but that in no way diminishes the respect I have for the man.

He is really a valuable asset for anyone looking for independent testing.

Jim

[Michael Erlewine]

Lloyd buys quite a lot of his own gear if the lists of his used equipment for sale that appear irregularly are any indication. He also has some kind of relationship with Zeiss, as indicated by his trip report last year.

I see no indication that either of those serve to make Lloyd a biased reporter. He is scrupulous in his testing methods, and documents what he finds in detail. I like quantitative testing, but I pay happily to subscribe to Lloyd's qualitative stuff. In fact, most of the time, we end up looking at the same issues from different perspectives. Sometimes we disagree, but that in no way diminishes the respect I have for the man.

He is really a valuable asset for anyone looking for independent testing.

Jim

Exactly!

[Jim Kasson]

I've finished the first pass GFX AF testing with the 63/2.8, and the results are a little confusing to me:

http://blog.kasson.com/the-last-word/fujifilm-gfx-af-accuracy/

Warning: those of you who don't like graphs will hate that blog post. OTOH, you probably wouldn't like to look at 700+ images that are all pretty close to being alike, either. Still, I advice you to ignore this post.

First off, the AF system is delivering very sharp images. That's the good news. The not-so-good (I hesitate to call it bad, considering the quality level that we're talking about here) news is that at f/4 and f/5.6 the spread is wider than one would like, and the average is lower than  one would expect. Not a lot, mind you. But enough to make me curious. Because the spread widens at the two suspect f-stops, it's hard to say that there is a systemic error source such as not focusing at the taking aperture, but I think there might be.

Here are the test conditions:

• ISO 100.
  5000K illumination, bright enough so that the f/2.8 exposure was 1/320 second
  Target distance 3 meters
  AF focus/release/priority: focus
  Rapid AF: off
  Pre-AF: off
  AF-S
  117 AF point mode, center point chosen, default size.
  EFCS on
  Pre-release half-shutter depression time: 750 milliseconds
  Shutter release full depression time: 250 milliseconds
  Camera moved 4 mm between exposures over a total range of 0 to 160 mm (40 moves, 41 exposures per stop)
  Camera motion applied by Cognisys focusing rail, with 1250 msec post-move settle time
  Shutter timing does by the same Cognisys controller
  RRS heavy-duty legs
  Arca-Swiss C1 head

If you think I left something important out, please let me know.

Jim

[Brad P]

Very interesting results, Jim. That is consistent with those noted by Lloyd Chambers.   In that regard, I sent you a private message on this site that you might find interesting and might want to follow up on. 

[Brad P]

I checked:

At apertures from f2.8 to f5.6, the camera closes the aperture to that value and performs the auto-focus at that f-stop as soon as the shutter is half pressed.
At apertures smaller than f5.6, the aperture is first closed to that particular value, so one sees in life-view the world at that chosen aperture but as soon as the shutter is pressed half to focus, the aperture opens to f5.6 (that means auto-focus is carried out at f5.6), as soon as the shutter is fully pressed for the exposure, the aperture is closed to the first value.

So in first case, the slight focus-shift has no impact on focus precision because focusing is carried out at working aperture, in second case focus-shift has very little impact, because only the slight shift between f5.6 and smaller apertures is relevant.

Maybe this has something to do with what seems to be increasingly reported critical focusing problems with the GFX.

1. If Fuji lenses have focus shift across apertures (which seems mostly undisputed), and if it is focusing at F5.6 for all aperatures greater than F5.6, then it almost by definition will not focus critically. 

2. Jim Kasson's latest blog (see a few posts up) notes auto focus problems beginning at apertures greater than 2.8 (i.e., 4.0 and 5.6 included).  So this phenomenon, if true, would not explain what he observes there. But if by chance the camera focuses at 2.8, it would explain what he observes.  Are you sure the camera focuses at 4.0 and 5.6 when those apertures are selected?   If Fuji has programmed the GFX to obtain a faster auto focus by focusing at 2.8 (or up to 5.6), that would come at the expense of critical focus. Maybe that works in the consumer camera market....

3. All this shouldn't seem to affect the mysterious manual critical focus issues a few people have reported on elsewhere.  But it might provide a lead to those inclined to follow it.   Since I'm becoming inclined to bite the bullet and go for the X1D, I lack that inclination. 

[Kevin Raber]

I just received the review GFX 50 with the 63mm and 120mm lenses.  I shot a lot of images under a number of situations on Saturday and found the AF worked just fine.  I shot everything handheld as I was just messing around and getting a feel for things, and even at low shutter speeds made some nice images.  I don't see a lot of differences from the Sony A7rII images but the images are nice.   I will try to put up a quick preliminary report this week showing images at 100%.  It was easy to set a focus point with the joystick.  At this point of the testing, I have to say I am impressed.  Dynamic range and sharpness are real good.  I have a nice example of that.  This will be a fun camera to review.  Time this week will be limited due to travel and a load of meetings.  The camera feels good in my hands.  And, the 120mm lens is very impressive. As with my XT-2 it is a fun camera to shoot with. 

[Jim Kasson]

Looks like the 120 macro is affected by whatever is causing lower-than-expected AF sharpness at some f-stops. It looks like it's only f/5.6 in this case. Also, the levels of sharpness are so high (MTF50 beyond Nyquist in some cases!) that I doubt if anyone would see this effect outside of a laboratory setting, or if they were pixel-peeping and looking for it.

http://blog.kasson.com/the-last-word/fujifilm-gfx-with-1204-af-accuracy/

Warning: there are numbers and graphs in the link above. Some of you may wish to pass it by.

Jim

[Brad P]

Looks like you're on to something here, Jim.  Still excellent graphs and numbers all around as you note, but something is going on that at least we care about. 

Would be interesting to see if manual focus for some reason suffers from the same effect (as at least one widely followed reviewer suggests).  If you can peak at the iris when focusing in live view at F16, for example, and see whether live view is using F16 (or F2.8, 4 or 5.6), that might possibly help suggest answers. 

Living on an island in the middle of the Pacific (with dealers 2,500 miles away), I might be your archetypical most appreciative reader right now.  It sure would be nice to figure out that these fine focus issues are focus aperture related that might be solved by a firmware fix and aren't wired into the hardware.   By the way, the GFX is back in my sights now due to the bokeh issues I've listed in a separate post.

[Jim Kasson]

By the way, the GFX is back in my sights now due to the bokeh issues I've listed in a separate post.

Didn't see the post yet, but that was on my list to test with the HC lenses on the GFX. However, I couldn't get the Fuji H-mount adapter to fire the leaf shutters.

I was thinking that, at just the right f-stop and shutter speed, that the H-lenses might have an apodization effect.

Jim

[Brad P]

Not having them, I tried to look up the specs on the HCD lenses but couldn't find info on the blades, easily at least.  My bokeh concern is about the XCD lenses which, if the H is like them, you might see when testing. At full aperture the XCD appears to produce octagonal bokeh that seems quite displeasing to me with bright or high contrast in the background.  See https://diglloyd.com/blog/2017/20170214_2028-HasselbladX1D-90f3_2-bokeh-OOF-night.html   

The GFX lens line uses a 9 circular blade design in contrast that should produce much smoother, more pleasing bokeh in contrast.  And I see that in GFX image samples on the internet.   The trade off may be for sunstars.  I think I'd rather have smoother bokeh and to me, the type of bokeh I see in the XDC lens line so far (there is a paucity of images on the net with high contrast bokeh as of today) is a huge negative. 

[Jim Kasson]

Would be interesting to see if manual focus for some reason suffers from the same effect (as at least one widely followed reviewer suggests).  If you can peak at the iris when focusing in live view at F16, for example, and see whether live view is using F16 (or F2.8, 4 or 5.6), that might possibly help suggest answers. 

With the 120/4, f/5.6 looks fine with manual focus (the usual warning about numbers applies here):

http://blog.kasson.com/the-last-word/fujifilm-gfx-with-1204-mf-accuracy/

I don't see any manual focusing strangeness, although there are a few things I like to see changed about the manual focusing experience in the GFX. BTW, I could say the same about just about every camera I've used.

I'll test the 63 with MF tomorrow.

By the way, I'm using the stop-down feature of the camera to focus at taking aperture. As an aside, it's a lot easier to focus this lens if you turn on the OSS function. I forgot to turn it off before making the exposures once, and it didn't seem to hurt the results, but I haven't included that data in the above post.

Jim

[Brad P]

Thanks for doing this one Jim. This is an important test for the 120mm given other reviews by good reviewers less scientifically expressing doubts about the MF.  These results are exactly as I'd expect them to be assuming no MF issues.   F5.6 ain't an unimportant stop either. 

Now if you could just get rid of the numbers, graphs, and all that repeatability stuff 

As an aside, I'm not coming to an easy answer on GFX vs X1D, but you breathed life back into the GFX platform in my mind seeing that MF is apparently unaffected by the focusing bug.  I'm going to have to do an old fashioned pros and cons list to sort through it all.  You may have seen that the Fuji 23 F4 and 100 F2 are rumored to be announced on Wednesday, April 19.  Must...be...dispassionate...

[Jim Kasson]

As an aside, I'm not coming to an easy answer on GFX vs X1D, but you breathed life back into the GFX platform in my mind seeing that MF is apparently unaffected by the focusing bug.  I'm going to have to do an old fashioned pros and cons list to sort through it all.  You may have seen that the Fuji 23 F4 and 100 F2 are rumored to be announced on Wednesday, April 19.  Must...be...dispassionate...

Consider also that you can use these lenses on the GFX, but not the X1D;

Otus 55 and 85
Zeiss 135/2 Apo
Nikon 58/1.4
Nikon 105/1.4 (wide open only)
Summicron-M 50
Canon 17 T/S

And probably a lot more.

Jim

[Jack Hogan]

I was thinking that, at just the right f-stop and shutter speed, that the H-lenses might have an apodization effect.

I was thinking the same but apparently you get angular bokeh, which does not compute.  So maybe what people are seeing is not about the shutter?

[ErikKaffehr]

Hi,

Keep in mind that numbers can be nasty, but they don't lie. Human vision is easy to fool. Quite possible that the focusing errors Jim has noted would not be observable in prints. After all, if you compare large prints you need to move your eye from one to other.

The focusing error that Jim has observed is very odd.

Reminds me. In Jim's early tests it semt like it was focusing on the red channel. I don't think Jim found an explanation, but it was also very odd.

Best regards
Erik

Thanks for doing this one Jim. This is an important test for the 120mm given other reviews by good reviewers less scientifically expressing doubts about the MF.  These results are exactly as I'd expect them to be assuming no MF issues.   F5.6 ain't an unimportant stop either. 

Now if you could just get rid of the numbers, graphs, and all that repeatability stuff 

As an aside, I'm not coming to an easy answer on GFX vs X1D, but you breathed life back into the GFX platform in my mind seeing that MF is apparently unaffected by the focusing bug.  I'm going to have to do an old fashioned pros and cons list to sort through it all.  You may have seen that the Fuji 23 F4 and 100 F2 are rumored to be announced on Wednesday, April 19.  Must...be...dispassionate...