Nikon D850: Apparent Build Tolerance Error in Flange Focal Distance vs Optical Viewfinder, Causes Major Focus Error with Manual Focus Lenses and Focus Confirmation
This post had gotten long and messy so I have reworked it for clarity and added user comments.
Update as of 2017-09-30 13:28 PST: two D850 camera bodies (mine and two others) are confirmed as having this issue. See the comments by Stefan P. Other users report not finding an issue. So at the least, it seems that Nikon has a quality control problem.
Update as of 2017-10-01: the problem is worse than it had seemed: the root cause is that the sensor in the D850 is not parallel to it lens mount flange. Tests with two Sigma Art lenses (85mm and 135mm) confirm excellent uniformity on the Nikon D810, and distinctly poor symmetry on the Nikon D850.
I’ve been using the Nikon D850 loaned to me by a reader. In all of what follows, the open question is whether the D850 body I am using is a one-off out-of-spec body, or some problematic build tolerance issue with most/all D850 bodies. I don’t have an answer to that yet, but by happy coincidence, my own Nikon D850 should arrive Monday Oct 2. My D800, D800E and D810 all were consistent; the D850 breaks that pattern.
As it stands, my Nikon D810 has proven itself for over 3 years to focus all my lenses right where I expect, with this loaner D850 in violent disagreement. When my D850 arrives, I’ll be able to confirm if what follows is a one-off defect, or an issue that if general then IMO would require a recall of all Nikon D850 camera bodies (or Nikon service adjustment). That is, unless Nikon repudiates it and claims that only Nikon brand lenses are supported. Or unless Nikon has intentionally made a flange focal distance change.
I shoot primarily manual focus (focusing helicoid) Zeiss lenses. This particular D850 leaves no room for focusing adjustment across temperature changes with my Zeiss Milvus and Zeiss Otus lenses, which have to racked beyond the infinity mark to the hard mechanical focusing limit, or very close to it.
As to autofocus lenses, it is now my conclusion that “slushbox” autofocus mechanisms are unlikely to be affected by this issue, showing inconsistency among lenses as well as inconsistency across camera bodies—just too mechanically sloppy to be meaningful as to flange focal distance.
Gross error in Nikon D850 flange focal distance
See the discussion of optical paths and flange focal distance further below.
Years ago, I had two Canon 5D bodies. One would focus to infinity at the hard-stop infinity mark with a Zeiss 28/2 lens; the other required backing off that mark a bit, or the image would be soft. That was my first experience in seeing variance (error) in flange focal distance. In that case I suspected a flange focal distance variance of perhaps 30 micros (0.03mm). Flange focal distance on Canon DSLR bodies is 44.0mm, so while 0.03mm does not seem like much, it is an error that results in a visibly blurred image at infinity. But as seen below on the Otus 55/1.4 focusing ring, the variance between my Nikon D810 and Nikon D850 is substantial.
Same lens, but major difference in where the focusing ring (helicoid) needs to be to acquire infinity focus. All of my highly corrected Zeiss lenses show a similar differential between the Nikon D810 (accurate) and D850 (inaccurate)
Here is what I am observing with the Nikon D850 I am currently using, comparing it to the Nikon D810 using the same lenses:
- Focus confirmation (“green dot”) is stated by Nikon to be substantially improved on the D850. But my testing shows that D850 focus confirmation is so far off that I might as well focus using the distance ring on the lens. For example, the D850 might indicate focus confirmation at 60 feet out, but the taken image is actually focused about 15 feet out. Even f/11 won’t fix that magnitude of error. Similarly focusing at very far distance subject with “green dot” focus confirmation results in major blur in the taken image, because focus is much closer.
- All my highly corrected Zeiss Otus and Milvus lenses must racked to the mechanical limit of the focusing range (or very close), which is far beyond the marked infinity mark, as shown below. See below how the D810 focus position is where one would expect for infinity focus, The D850 is so maladjusted that my Otus 55/1.4 cannot quite reach infinity focus in colder temperatures.
- Autofocus with AF lenses seems to be accurate, which directly contradicts the green dot focus confirmation and the physical fact of different infinity focus for the same lens on the D810 vs the D850. I have no explanation except that with what I term “slushbox” AF, software is involved and behavior clearly differs with autofocus vs focus confirmation. Autofocus lenses are “intelligent” and thus may interact to compensate for the flange focal distance, somehow. I do not know how to reconcile these seemingly contradictory behaviors except by camera-lens communication and software differences versus manual lenses.
A mechanical focusing system (helicoid or similar) is invariant and not subject to electro-mechanical slushbox autofocus as with an AF lens (or any software behavior). Thus a lens with a focusing helicoid is the final arbiter of flange focal distance. Thus I want to emphasize that this error has nothing to do with Zeiss lenses in particular.
Zeiss sets infinity focus precisely; setting aside temperature extremes with highly corrected lenses having special glass types, the same lens should focus at the same point on the focusing ring, with little variance for camera bodies within 'spec' for flange focal distance.
The comparison images show above with the Otus 55/1.4 shows that the D810 and D850 are in violent disagreement as to where infinity focus lies. Obviously the Otus 55/1.4 does not change its behavior minute to minute when swapping between camera bodies; the only explanation is a flange focal distance error on the D850. Over the past 3 years, my Nikon D810 has always consistently reached infinity focus at or very near the infinity mark with all of my Zeiss manual focus lenses, and other brands too. The D850 I’m testing violates this contract.
Note that some lenses may be incapable of reaching infinity focus at all, Nikon AIS or Zeiss or any other brand. For example, my Otus 55/1.4 cannot quite reach infinity, at colder temperatures, at least not everywhere in the frame (mild field curvature).
Thus, the flange focal distance on the D850 looks to be in gross error. I estimate that it is off by at least 100 microns (0.1mm), i.e., the flange focal distance might be 46.4mm instead of the required 46.5mm. This is why focus confirmation is way off: the two optical paths are different. For perspective, I consider a deviation of 20 microns to be borderline unacceptable; here we are talking about 5X or greater deviation.
How this camera body ever left the factory is a mystery—it implies poor quality control. If the problem is a general one affecting all D850 bodies, then Nikon has a mess on their hands. The good news is that this should be adjustable by a Nikon service center, or at least I assume that is feasible.
Two men say they’re Jesus, one of 'em must be wrong — Dire Straits.
The only good test is 2 or more bodies, which is what I'm doing with D810 vs D850 using the same Zeiss lenses—they disagree in a large way.
But in deciding whether this is a general Nikon D850 behavior, the most useful test is two or more D850 camera bodies, which either agree in both being way off as I’m seeing, or disagree, showing that one is off but not the other. Either way, it implies poor quality control.
Autofocus lenses are inconclusive with their slushbox AF; they are both inconsistent on the same body and inconsistent in being consistent between the Nikon D810 and Nikon D850 (since a manual focus lens shows what has to be a flange focal difference).
Zeiss or other brand manual focus lenses with focusing helicoids have a fixed mechanical focusing mechanism. The right way to test and see this if this issue affects a particular D850 is as follows:
- Mount a lens with a focusing helicoid on the D850, e.g., Zeiss Otus or Milvus lens.
- Enter Live View zoomed to 100% or 200% and focus precisely at infinity (most sensitive to differences, but other distances can be used so long as focus position is noted).
- Repeat with a Nikon D810 or D800E or D800 or D850.
- Note difference in focusing position (if any) on the lens barrel.
All of my Zeiss lenses which focus past infinity exhibit a marked difference in focusing helicoid position on the D850 I currently am using. All of them reach the limit of their focusing range in order to reach infinity. This is a serious problem if a temperature change requires a bit more range.
In general, it appears that the highly corrected lenses that have a focusing range extending beyond the infinity mark are subject to this problem. I have many more lenses at home, but am currently in the field and thus cannot test a variety of hard-infinity-stop lenses at this point.
Here, “infinity mark” means the center of the infinity symbol. Tests at about 70°F:
- Zeiss Milvus 18mm f/2.8: reaches infinity focus beyond infinity mark at mechanical limit of its focusing range.
- Zeiss Otus 28mm f/1.4 APO-Distagon: reaches infinity focus beyond infinity mark at mechanical limit of its focusing range.
- Zeiss Milvus 35mm f/1.4: reaches infinity focus beyond infinity mark at mechanical limit of its focusing range.
- Zeiss Otus 55mm f/1.4 APO-Distagon: unable to reach infinity focus everywhere in the frame at 39°F, though might be able to reach infinity at some areas of the frame (mild field curvature).
- Zeiss Otus 85mm f/1.4 APO-Planar: reaches infinity focus beyond infinity mark a tiny bit off the mechanical limit of its focusing range.
- Zeiss Milvus 100mm f/2M: focus at infinity is at infinity mark (has a hard stop at infinity mark). But on D810, the focus is before the infinity mark, so behavior is thus consistent with all the other Zeiss lenses.
- Zeiss Milvus 135mm f/2: focus at infinity is past infinity mark, but with some room to go further.
- Nikon 14-24mm f/2.8G: focus is well past infinity mark.
- Nikon 105mm f/1.4E: focus is at infinity mark. [note: lens does not focus optimally with conventional autofocus or contrast-detect autofocus in Live View].
- Sigma 14mm f/1.8 DG HSM Art: focus falls short of infinity mark, about half way through the right side of the infinity symbol.
- Sigma 85mm f/1.4 DG HSM Art: focus is at infinity mark
- Sigma 135mm f/1.8 DG HSM Art: focus falls short of infinity mark, about half way through the right side of the infinity symbol.
In my case I can see not only gross focusing errors with the D850, but comparing it side by side with my Nikon D810 is proof that the D850 is in gross error in its flange focal distance. Live View is usable of course, but if a lens cannot reach infinity focus, that’s a very serious problem, not to mention the gross errors with autofocus or manual focus confirmation (which uses the AF system) or the optical viewfinder.
Moreover I found the Nikon 100/1.4E did not focus optimally either with phase detect AF or in magnified Live View with contrast-detect AF, easily seen by manually focusing and improving the sharpness. But that is a longstanding behavior among AF lenses; I never assume accuracy of AF when testing.
I have seen small flange focal distance errors over the years, but this D850 seems to be off by 10X to 30X greater than anything I have seen in 10 years. I assume it is a gross build error (not a design error), and I also suspect that it can be fixed by adjustment by Nikon service, though the error is so large I am uncertain. How this Nikon D850 ever made it into a shipping box is baffling; it should never have passed quality control.
Background — optical paths
See also Mount Flange to Sensor Parallelism.
In a DSLR, there are two optical paths whose length should match exactly:
- The optical path to the sensor; this is the taken-image path. The distance (“flange focal distance”) should be 46.5mm on a Nikon DSLR, which is why Nikon lenses can mount on Canon bodies (44.0mm) with a lens adapter, but not vice-versa.
- The optical path to the AF sensors in the mirror box, which is where the autofocus system and focus confirmation are done by the camera. This is the AF focusing path.
- The optical path to the focusing screen and thence to the eyepiece; this is the optical viewfinder focusing path, what the eye sees. Manual focus done via eye is highly unreliable no matter how good ones eyes are due to modern focusing screens being unfit for manual focus discrimination.
Paths 2 and 3 are the same in that they both are different from the optical path to the sensor. I’ll lump them together for purposes here, though in theory they could also be at odds with each other.
When these paths do not match exactly, there is focusing error: for example, the autofocus system can claim to be in focus but if the distance to the sensor deviates as little as 10-20 microns, the image will be soft. Many people think this is autofocus error when in fact it could be a build tolerance error (differing distances for the two optical paths). It is one reason that AF Fine Tune is a camera feature; to allow users to adjust for the differing light path length. The other reason is autofocus lens interaction with the AF sensor being off. The only consistently reliable focusing mechanism is via the sensor light path in magnified Live View, particularly far distance.
On modern lenses, infinity focus can be a variable thing due to special types of optical glass which change refraction behavior as temperature varies. Accordingly, while some lenses have a fixed infinity focus mark (many prior-generation Zeiss lenses for example), most lenses these days (Nikon and Zeiss and other brands) have a marked infinity position for focus, but allow focus well beyond that mark.
Zeiss sets infinity focus very precisely, and all my 20 or so Zeiss ZF.2 and ZE lenses have always proven this out on my Nikon D800, D800E, D810. Not so on the D850, where they all focus well beyond the infinity mark racked out to the limits of focusing.
Roger D writes:
My new d850 exhibits the same flawed behavior . Manual focus confirmation is way off . Zeiss Otus focus goes beyond infinity.
[diglloyd: but then recants]
Ok I did some infinity tests and they all look good. The Zeiss OTUS 55 focuses slightly beyond the Zeiss infinity mark but this of course could be just me . I found the focus confirmation to be dead on ..better than my eyesight .. If you can nail a sign at a 100yards at f1.4 ..good enough for me .
When I set the lens beyond the infinity mark it turns to mush, back off too the mark and it looks good . This is at f1.4 and with no calibration of the lens to the body . It would be incredibly rare for me to manually focus at infinity using f1.4 .assuming I would be at f8 or f11 ..I can t see this as a issue. I would confirm that its no easy job nailing focus at f1.4 with a manual lens at distance, target simply isn t large enough. I am not a lens tester and prone to operator error and over confidence but I will have to eat crow on my initial observations .
DIGLLOYD: that seems to suggest a general issue, but the next comment.
Tom S writes:
I just did a (phase detect, 'traditional') infinity AF test with the Nikon 105mm at f1.4 and it's repeatedly focusing right on infinity, at the infinity mark on the lens, 'green dot' and all. I have not calibrated my Nikon AF lenses at all after getting the D850 and have been impressed by the accuracy of the AF, near to far. Will try the Zeiss 135mm later.
Matt K writes:
Your problem description sounds exactly what happened to me when I was trying to adapt Canon 17mm TS-E and 24mm TS-E MK2 to Sony A7R2 using Metabones MKIV smart adapter. While those lenses worked very nicely on A7R1 with the *same* adapter, trying to use them on R2 caused major headache - not only the infinity focusing was way off, but also the pictures fell apart towards image edges due to screwed up field curvature at infinity (or close) distance. Some users have managed to shim their Metabones adapters which has actually fixed the issues.
It may be the BSI sensor causing the problem as the sensor surface is a bit closer to the lens with all the wiring behind the detectors. Nothing else explains how suddenly R2 started failing with those adapted Canon lenses and not only those. I also used an adapted Zeiss 21mm Distagon .ZF2 on A7R1 with excellent results but with the same adapter the infinity went way past infinity mark when attached to A7R2.
Sure, it may be bad quality control, your new unit should tell whether that is the truth or not.
While it sounds weird that BSI sensor would cause a problem there might be something to it. We ran out of copies of Metabones adapters when trying to find one that would work with those TS-E lenses. After five copies still no luck with my R2 or the demo R2 unit in the shop. But every single one of those adapters worked acceptably with a used A7R1 body which was available at the time in the shop as well.
DIGLLOYD: it does wound weird indeed. Surely if this were the case Nikon would mount the sensor so as to produce the same flange focal distance.
Paul B writes:
I’ve just started reading you D850 articles and I am looking forward to you posting more. It should be quite the comparison to the other high-resolution cameras currently available. Which will play a big role in my decision to upgrade to the A7RII, D850, Fuji 50 GFX, or wait for the A7RIII.
Though I was a bit disappointed to read about your infinity focus issues with the borrowed camera. I have done a lot of work with adapting lenses to mirrorless cameras and modifying adapters to make lenses focus properly, so I have seen this sort of thing several times. I feel your overall assessment and the possible causes are correct.
From reading your description and looking at your image of your Zeiss lens on the D850 compared to the D810, I would say the flange focal distance of the D850 (i.e. lens mount to sensor) is long (>46.5mm) and the adjustment range built into your lenses. Rather than an issue with the optical distance to the focus screen/AF sensors.
Did you compare the live view focus distance, to viewfinder focus distance, to the same subject at a closer distance than infinity? If yes, what was the distance to the subject, and the focus distance on the lens using both focusing methods? The answers to these questions can give you qualitative answers to the questions of is the flange distance to the sensor off and is the distance to the focus screen/AF sensors close to the same. This process can be used to determine the flange distance error with some precision, though you need a measured distance from the sensor to the subject.
Concerning the comments by Matt K, I have had similar issues with the Metabones MKIV adapter. In my case the adapter did not hold the lens far enough away from the sensor of my A7II and it did not hold the lens square to the sensor. It sounds like Matt had a similar experience with the A7RII, while the adapter was a closer fit to his A7R.
DIGLLOYD: I agree that the AF sensors are not likely at issue, and that the flange focal distance is the culprit. There is no other explanation possible; a Zeiss (or other brand) manual focus lens does not change its focusing behavior by being mounted on a different camera body.
I did compare both at far distance and medium range. I’ve not had the time yet to do more, and out in the field I do not have a tape measure, but my 60/15' example in my discussion is one example.
The optical viewfinder is too coarse a mechanism to say much and always has been, so it’s not useful in this discussion (the focusing screen is not suitable for manual focus, and it brings visual acuity of the human eye and the diopter adjustment into the equation—not tenable for assessment).
As for lens adapters, I completely agree that these introduce all sorts of possible errors.
Kevin S writes:
I tested infinity focus using magnified live view and have enclosed the following photo. No response necessary.
... also tested it on a D5 and the infinity focus spot is the same
DIGLLOYD: Kevin’s photo is of a Zeiss Otus 55/1.4 focused so that the focusing ring is about 1mm past the infinity mark. The best test is that plus the same test on a prior Nikon body like the D810/D800E/D800. With Kevin’s test on the D5, this suggests good news, that is, that the D850 I am using is just off with its flange focal distance, indicating an issue with Nikon build quality tolerances.
Stefan P writes:
Thank you for your recent information on the Nikon D-850.
I tested my Otus 55mm on it and the results were the same as reported. As a consequence I returned the camera today with the printout addressing these issues. I have 5 of the lenses listed in that blog and am not prepared to keep the camera under the circumstances.
As an aside, several of the camera staff read the blog... there was only one that appeared to understand it.
DIGLLOYD: looks like poor quality control to me: a mixed bag of “OK” and “no OK”.