So, what was wrong with the X-T4? What did Fujifilm find? Is the camera fixed? Does the AF work as it should? Can I trust it? Read on.
If you have been following this blog you know why the camera was sent in for a complete check over and autofocus calibration. If this is the first of my posts you are reading about the issue, to briefly recap, many Fujifilm users and I have found a systemic autofocusing inconsistency issue with our X camera bodies. The issue is that when you repeatedly focus on a stationary subject, the camera will often focus at different distances. Really, it does. From the emails and comments I've received as well as information I've found on the Internet, the issue seems to be widespread. This focusing error is easily repeatable. When I first learned about the AF inconsistency issue from a friend, I thought it wasn't occurring in my camera. That is, until I thoroughly tested it.
Additionally, the wider angle focal length as well as the smaller aperture, the worse the focusing errors are. For example, using AF-S, if you repeatedly focus on an object 10 ft. away, the camera might focus at the 10 ft. mark the first time (or it might not), but the second time it may focus at 8.5 ft. and the third time 11 ft. In actual practice, there may be more variation than that depending upon the distance and aperture. With wider angle lenses and shorter focal lengths that same 10 ft. target may have a focus error range from 6 ft. to infinity. Seriously. Infinity! As I said, this phenomenon is repeatable in multiple experiments.
When autofocusing using AF-C and half pressing the shutter button the distance scale in the LCD or EVF may initially stop at that 10 ft. subject (or it may stop at another distance) but then continue to move left and right (hunting), changing the focusing distance continuously. In both cases, no other autofocus camera I've ever owned, film or digital, has exhibited this behavior. Also, all those initial tests I conducted before sending my camera in for examination and calibration were done on a tripod in bright light with contrasty subjects and using several different sized focusing squares. So those issues were not factors in this inconsistency issue.
To quickly check this yourself, just add the AF distance scale to the LCD/EVF, repeatedly focus at the same distance (using the shutter button as using the back-button doesn't show focusing distance on the scale) and look to see if the focus distance indicator lands at different distances on the scale each time you autofocus. Also, try several focal lengths as shorter focal lengths are more prevalent for this error. Additionally, stop down to f/5.6 or f/8 and see if the errors increase over wide open. I think I may know the reason for this. Hold that thought.
I hope your Fujifilm camera consistently focuses on the exact same spot each time and you haven’t experienced the frustrations that I have.
That said, besides seeing this in my controlled tests this also translated to real world photography where I shot outdoor scenes in the manner I normally would photograph and found that over 12% of my images were out of focus. Some slightly, some more than slightly.
(DISCLAIMER: I don't consider my testing highly scientific. I try to remove all variables and make the results repeatable, but I'm not claiming that they will stand the most rigorous scientific scrutiny. I highly recommend conducting your own tests.)
If you are not interested in reading the rest of this whole post, I'll give you the answer to the original questions now. No, not completely fixed, only seemingly partially. Better, but accurate focusing is still not 100% as it should be. How's that for ambiguity?
In the past, when I routinely found some of my images slightly out of focus (I make about 3 exposures of the same subject just to ensure at least one is perfect. I started that in the film days because I was always worried a single negative or slide could be damaged) I thought my technique may have been faulty. With my X-T4, sometimes all three image files were in sharp focus but it wasn’t unusual for at least one to be slightly unsharp. I now don’t think it was my technique at all. I now believe the camera has not been focusing accurately. I've been photographing seriously for 51 years and have owned and used almost 100 cameras and more lenses than I can document and I’ve never had this issue in the past, especially with any of the other 29 digital SLRs or mirrorless cameras I've owned and used.
I'm not going to repeat everything about the issue I've previously written but if you are interested in exactly what the AF inconsistency issue is in depth and the testing I conducted to nail down exactly what was happening, please read these blog posts located
here,
here,
here and
here. That will tell you everything you know in great detail.
As a result of these tests as well as numerous private emails and public comments I received from other Fujifilm users having the same issue, I decided I needed to eliminate one last variable—was my camera no longer meeting factory specifications? To find out I sent my X-T4 to Fujifilm North America repair in Edison, New Jersey, included documentation explaining the issue, asked them to completely check over the camera and, if the AF system is out of specification, fix it. That is where the frustrating series of errors on their part come in, but I digress.
Upon its return, the repair sheet, in the technician's handwriting, indicated that he, "Extensively inspected function and performance. Performed firmware upgrade." As an aside, during one of the many phone calls I made to Fujifilm to find out where my camera was, the technician on the phone told me they "tweaked the camera body" and "updated the firmware." I don't know the technical definition of what "tweaked" means, but I hope it means they actually made some sort of adjustments. But I can't know that, can I?
Now that I have it back in my hands once again, I set out, for comparison, to now test the camera in the same manner as I conducted the earlier tests. In a bright indoor area I used a high contrast target with the camera on a tripod, an exact measured distance to the target, squared up horizontally and vertically to the subject to eliminate any errors, etc. so focusing could be as precise as I could make it. In other words, I locked down all parameters so the only variable would be autofocus accuracy. The test target was set exactly at 10 ft. from my camera. As an additional test, I did much the same but outside in bright sunlight at a distance of about 200 ft.
The results of that test are listed below (First Test of the 16-55mm f/2.8). Before I you get to all the results, I also took the camera and lens outdoors and did some, as they say, real world photography in a manner in which I would normally photograph. I wanted to have more data with additional lenses and to know more than just a static test could provide.
For this outdoor test, I again attached the 16-55mm f/2.8 lens (no slouch when it comes to sharpness and the lens that stays on my camera 90% of the time) and tested the autofocus in three different ways—AF-S, AF-C and Manual focus using Instant AF in the AF-C mode (that is normally how I have had my camera set in the past). I would then make three identical exposures in each AF mode, autofocusing between each, using shutter speeds higher than 1/250th second with IBIS set on continuous mode and electronic first curtain shutter. Camera shake and potential shutter bounce would be eliminated. I photographed a number of different scenes in bright sunlight.
I then wanted data about my other Fujifilm lenses so I conducted another series of the same real world photography subjects—not indoor static subject tests as I first did with the 16-55mm lens. I photographed separate and distinct subjects at various distances. Additionally, I shot a variety of focal lengths with each of the zoom lenses. I made three consecutive and identical shots each using AF-S, AF-C and Manual focus using Instant AF in the AF-C mode. I shot mainly at f/5.6 but I also shot some wide open with the wider focal lengths.
The results look promising—better than before I sent the camera to Fujifilm, but not perfect nor as they should be. I see some improvement in some modes of autofocus.
Here are the findings. These tests are not scientific by any means, but informal tests based upon how I use my camera and the kinds of subject matter I photograph.
10-24mm lens @ f/5.6 (at 10mm & 14mm; distances at 3 ft., 25 ft. and 200 ft.)
AF-S 14/15 in sharp focus (93.3%)
AF-C 12/15 in sharp focus (80.0%)
Inst. AF 13/15 in sharp focus (86.6%)
23mm f/2 (15ft., 25 ft. and 200 ft.)
100% in focus in all focus modes
I ran two separate sets of tests with the 16-55mm f/2.8 lens because this is the lens that stays on my X-T4 90% of the time. I wanted extra data for this lens.
16-55mm f/2.8 (first set of tests, indoors with a test target and outdoors, which were different subjects than what tests were performed on the other lenses listed here) (various distances from 5 ft. to 200 ft.)
AF-S 36/36 in sharp focus (100%)
AF-C 25/36 in sharp focus (69.4%)
Inst. AF 28/36 in sharp focus (77.7%)
16-55mm f/2.8 (second set of tests which were outdoor with the same subjects as the other lenses listed here) (15 ft. 75 ft. and 200 ft.)
AF-S 42/42 in sharp focus (100%)
AF-C 38/42 in sharp focus (90.4%)
Inst. AF 42/42 in sharp focus (100%)
55-200 lens @ f/5.6 ((15 ft. 75 ft. and 200 ft.)
100% in focus in all focus modes (100%)
100-400mm lens @ f5/6 (15 ft. 75 ft. and 200 ft.)
AF-S 14/15 in sharp focus (93.3%)
AF-C 14/15 in sharp focus (93.3%)
Inst. AF 15/15 in sharp focus (100%)
Again, these statistics are better, but all of them should be at 100%.
Comment: The really baffling aspect of this for me has been that I am standing somewhere in bright, sunny daylight with a detailed, contrasty subject 15 ft. or 30 ft. or 200 ft. in front of me. I press the shutter button, the lens focuses and the image is made. I press the shutter a second and a third time in exactly the same manner just so I have insurance against a corrupt image file. Then, when I later look at the images in Lightroom, one or maybe two are not in sharp focus! How can that be? It just doesn’t compute. Why doesn't the system grab sharp focus on the easiest types of subjects? The kinds of subjects that any point-and-shoot camera could do repeatedly. But, again, I digress. To continue...
Based on these tests and my previous tests, I noticed a couple of things of interest so I delved into them a bit more.
First, the more I stopped down (more DOF), the wider the variation in focusing error as shown on the distance scale. Also, as I mentioned above, the shorter the focal length, the more error was found (again, more DOF).
Second, and I found this very interesting as well as telling, when conducting the indoor test with the 16-55mm f/2.8 lens, wherever the camera autofocused at any focal length or any aperture, when I switched the camera to manual focus mode with red focus peaking, the 10 ft. target showed focus peaking on its edges. Hmm? Does this mean the camera's focusing scale is the issue? For a short while I thought it might be but when I then started tweaking the manual focusing ring and saw that although the test target had red focus peaking on its edges, it wasn’t in the center of the areas covered by focus peaking. By tweaking the focus ring I could intensify the amount of focus peaking on the target itself. That means the target would have been in the 'general zone of focus, but not precisely in sharp focus as it was not the center or most intense portion of the red peaking. Make sense? In every exposure, no matter where the focus indicator landed on the distance scale, the red focus peaking included my test target (within some portion of depth of field?) but I could intensify the peaking by manually focusing on the target after autofocusing. The same result occurred with the camera set on AF-C. This phenomenon led me to what I think may be happening with the variation in focus distance. Again, hold that thought.
I think we have three things going on here that may have affected what I have seen.
First, when you manually focus a Fujifilm X lens using the focusing ring, the change in distance is not a ‘smooth continuous’ change. Focus distance jumps slightly from one distance to the next. Again, it appears not to be a continuous change of distance. That 'seems' to preclude the ability to achieve the finest, precise and exact focus. I verified this with all of my lenses. Maybe the lens cannot precisely autofocus on a specific distance unless that distance corresponds to one of those distances pre-programmed into the lens? As an aside, I also checked my Nikon Z lenses and the manual focus moves in a similar manner—a bit jumpy, not smooth and continuous. That said, 100% of my Z7II images have been in focus.
Second, Fujifilm has engineered the camera/lens combination to autofocus at the aperture set on the aperture ring, not at maximum aperture as we have been used to for all of our past photography lives. In other words, when you half press the shutter the lens stops down to the set aperture instead of focusing at maximum aperture. With my X-T4 you can hear the aperture blades stopping down when initiating autofocus. Depth of field then plays a role in focusing accuracy. I surmise that once the phase detect pixels recognize a certain level of accuracy (that being with in the DOF), they stop focusing. That precludes achieving ‘exact’ focus on the distance of the subject. Unless depth of field is shallow, focus is, I’ll call it, ‘close’ but not exact. I think Fujifilm is counting on the existing DOF to cover any slight focusing error. That may explain why the anomaly is more frequent with wide angle lenses (shorter focal lengths/more DOF) and when stopped down (again more DOF). The inaccuracies seem to be less frequent with shallower depth of field. In the times of film and DSLRs, our cameras all focused with the lens at maximum aperture. Evidently not with some mirrorless cameras.
Third, Fujifilm’s latest firmware update (v1.5), released on June 29, 2022, seems to have had an impact on focus accuracy. A portion of the firmware update included, "The accuracy of AF function was improved at manual focus mode using "One Push AF." I will interpret that to apply to what the X-T4's menu calls "Instant AF Setting." Personally, I think improvement can be seen across the spectrum of AF modes not just with Instant AF using the manual focus mode. However, I may be wrong.
At this point, now knowing what I know, I think my camera and lenses are ‘usable’ but I still don’t fully trust the AF system, especially with wide angle lenses. In some modes there is an occasional focus error and in those modes I'll give the camera a "pass," as that could happen with any system. I'll consider using it in only the most accurate modes. I really think Fujifilm’s engineering of the system was faulty and ‘sloppy’ from the beginning. I have been told this issue does not occur in the GFX series of cameras, which again I'm told, has a different focusing system. I don't know for sure, just passing on that tidbit.
There you have it. You have to decide what this means to you. I normally don't use AF-S to photograph. I normally use either manual focus mode with Instant AF set to AF-C or the AF-C mode directly with back-button AF. If I were to continue using this camera I would only have confidence in using it with AF-S mode.
I would be very interested if any X-H2s owner has tested his or her camera and found this issue. I'm hoping some have but found no inconsistencies. Fingers crossed as that would give me an option for the future (as well as the X-H2 and X-T5 (as I really don't want to entertain selling my Fujifilm gear. It is my favorite digital camera system and has given me the greatest user satisfaction. If you are an X-H2s owner, test your AF and let us know what you find. Please, as your input would be very valuable.
For now, I'm going to let my Fujifilm gear live in its bag in the closet. A Fujifilm manager from Germany said in a recent interview that there is a possibility that Fujifilm may push down some of the AF improvements from the new X-H2s into older cameras. Of course, the speed and subject detection most likely wouldn't be available due to the limitations of the older image processors, but the AF accuracy might be improved. If that doesn't happen, I suspect I will then painfully divest myself of all of my Fujifilm gear and build my Nikon kit.
One last comment. As many of you know, I took my Nikon Z7II and Nikon 24-200mm f/4-6.3 lens on my recent month long coast-to-coast and Route 66 road trip. I made almost 1150 exposures with that combination and EVERY image at every focal length and every aperture is in sharp focus. Why can't my X-T4 perform like that?
As I said in the title, I intend this to be my last post on this issue. That is, unless something radically changes. I will then report that to you.
UPDATE: One more interesting thing…. Before I sent my X-T4 in for repair, the last file number I made with the camera was 13,087. After it’s return I noticed the first file number I made was 19,886. I can only understand that to mean Fujifilm made 6799 shutter actuations with my camera during the time they had it for repair. I find it curious that they made that many actuations. That’s much more than I would have thought necessary. Also, on the ‘being thankful’ side of things, they did not reset any of the camera many, many settings that I had made. I thought for sure the camera would come back with factory defaults. Nice surprise. Just passing this on because it was unexpected and interesting.
Join me over at my website, https://www.dennismook.com.
What this means to me is that my next camera is going to be a Sony. I am considering the A7 IV, but I may wait to see what the next iteration of that system brings.
ReplyDeleteI forget how long you've had the X-T4, but would Fuji be willing to send you an new version, i.e. not one that has been repaired? I had an experience where my X-T2 (old, I know) no longer shows the correct ISO when Aperture and ISO are both set to automatic. It takes the picture at the correct ISO and you can see it after the fact, but not when the shot is being metered in the EVF. I called them to see if they had an answer and they were baffled. I still use it, but I don't keep the Aperture in the "A" mode anymore!
Thanks for the comment. I’ve had my X-T4 beyond the warranty period so asking for a replacement would be futile. Also, I believe the cause of the error is systemic to the engineering of the AF function itself, so a different camera would likely suffer from the same issue.
DeleteSorry to hear about your frustrations with this, Dennis. Especially since I owned the X-E2S and X-T2 and had excellent AF performance from them. Perhaps the best thing to do is rent an XH-2 to see how it performs with your lenses, though it likely will be several months before the XH cameras are available.
ReplyDeleteThanks, Jon.
DeleteIs there any news yet on whether or not the X-H2 also suffers from this autofocus inconsistency?
ReplyDeleteI have not read nor seen anything where someone has reported the same AF inconsistency issues with an X-H2. That’s not to say someone hasn’t reported something, but I’ve not read any about any issues. I have an X-H2S and I have used it extensively using AF-S, AF-C and with MF using Instant AF in the AF-C mode and have not had any issues with inconsistent AF. I’ve done several tests and all of my images have been in sharp focus. The X-H2 has the same focusing system so I would assume the same experience could be had using it. All I can say is my copy of the X-H2S autofocuses perfectly.
DeleteOwning both, H2 and H2s, the problem is still there and lens dependent with the old 56 1.2 i even got some other strange behaviors like when in MF(!) the lens make small focus adjustement pressing the shutter button. I am in contact with fujifilm trying to understand what's going on, but it's annoying. Years ago i owned a GFX50S and i sold it immediately because in studio (i work as commercial photographer), it was impossible to achieve perfect Manual Focus because the lens was "stepping" from one distance to another in too large adjustement for a perfect macro shot.
Deleteexperience with a real manual MF lenses were so much better, but in the end i switched the GFX50s with a 645Z
Owning both, H2 and H2s, the problem is still there and lens dependent with the old 56 1.2 i even got some other strange behaviors like when in MF(!) the lens make small focus adjustement pressing the shutter button. I am in contact with fujifilm trying to understand what's going on, but it's annoying. Years ago i owned a GFX50S and i sold it immediately because in studio (i work as commercial photographer), it was impossible to achieve perfect Manual Focus because the lens was "stepping" from one distance to another in too large adjustement for a perfect macro shot.
Deleteexperience with a real manual MF lenses were so much better, but in the end i switched the GFX50s with a 645Z
I am not sure when Fuji is going to fix this big problem with their AF-S and wide-angle lenses.
DeleteAt the same time, their distance scale cannot be trusted since it shows different distances for different lenses on the same body, and the shown distances are different between X-T3 and X-T5.
What matters is that I am getting OOF shots that are not supposed to be there. I am unwilling to use all sorts of workarounds that sort of work either. Fuji equipment is not cheap. It would be nice for it to meet the expectations!
There is no Fuji's acknowledgment of the problem, and the problem still persists. Time to think about switching the brand?
I'm finding the same poor, inconsistent AF performance with my X-E4, X-T5 (now sold), and my newly acquired X-H2. Reported to Fuji tech and their response was send your camera in. Knowing what will happen, that isn't going to happen!
DeleteI found that after Fujifilm had provided a firmware update, subsequent to the publishing of this post, my X-T4’s AF was much better. That said, I sold the X-T4 to purchase the X-H2S and X-T5 and, evidently unlike what some others have said, I have had absolutely no issues with inconsistent or inaccurate AF. None. But I can only speak about my two cameras, not all that have been produced.
DeleteHere is something that may help. The way Fujifilm has engineered their AF is that when the AF switch is set to AF-S or AF-C, the lens stops down upon half-press of the shutter, thus increasing the depth of field. The lens focuses at the aperture you have set, not wide open as is done with DSLRs and some other mirrorless cameras. I believe that may be part of the problem with inaccurate AF. However, if you set the focus switch for Manual Focus and, in the menu, set Instant AF for either AF-S or AF-C, the lens will autofocus with the aperture wide open, which will produce a more accurate and consistent AF. I set my Instant AF setting for AF-C and use the back button (depending upon the camera, either program the AF/AE Lock button for AF-On or if there is an AF-On button if your model camera has one) to activate the autofocus. Works great.
Thank you for chronicling this crazy problem. I really noticed it on my XT4 when I got the 18 1.4 and the 8mm lenses. I am trying the fix you have in this post. I am also looking at Leica rangefinders. Go back to a simpler time perhaps.
ReplyDelete