Sensor Alignment and Focus errors in D850

[Les Olson]

[..]  a register distance that is off by 0.1 mm is not negligible and beyond Nikon's tolerances. It is in particular noticeable on short focal length lenses like utrawides or fisheyes, and especially if the sensor is not parallel to the mount it can affect optical performance.

I said the effect of a 0.1mm difference in flange focal distance on infinity focus is negligible - and an effect on infinity focus is what diglloyd says he can observe.  Plus, any effect of the flange focal distance being wrong would affect the whole range by the same amount: the effect of changing the position of the real principal plane by any given amount would stay the same, just as it is with an extension tube.  An observation that " [...] the D850 might indicate focus confirmation at 60 feet out, but the taken image is actually focused about 15 feet out" cannot possibly be due to a 0.1mm error in flange focal distance.

Now he says the sensor is also not parallel to the mount (https://diglloyd.com/blog/2017/20171001_1530-NikonD850-sensor-lensMount-parallelism.html).  What he says about the effect of the sensor not being parallel to the mount is obviously wrong, and the pictures he shows that purport to prove the claim show nothing of the sort.

There have been previous episodes of people claiming that out-of-adjustment cameras were common and causing a problem - in the case of focus issues there was the one where back/front focus was supposed to be caused by the mirror being at the wrong angle, and you had to adjust it, eg (http://leongoodman.tripod.com/d70focuspart3.html). 

[Ethan]

WoW. so lloyds used a camera which is not his, attached to a non Nikon lens and decided that the D850 is crap because it does not reach "Infinty and beyond"

I mean seriously. Is this guy real. I appreciate it is Autumn and the cyberweather is a bit slow and need to pick up viewership/readership.......paid of course........

I mean is there a remote possibility that the reverse engineering by Zeiss on his particular Otus lens does not marry very well with his and his friend cameras.

Oh, I know, let's ask Thom and Ming. They know everything even before it happens.

Apparently they are now the Holly Oracle Trinity: Lloyd - Thom and Ming.

[David H. Hartman]

It appears one has to pay to see how Digi Lloyd test for sensor/bayonet alignment, i.e. science is not free. I'm skeptical of his methods and I certainly won't pay to find out what they are.

Were I testing a Nikon D850 I'd use a 55~60mm Micro-Nikkor known to be in good condition as these lenses are designed for flat field use. I have multiple copies so I'd test several. Other lenses may exhibit field curvature that would invalidate focusing on one side and examining the other. One needs to focus at the center, and the center only and then examine both corners. 

In the past I've noticed that both individuals and dedicated websites test lenses like 50/1.8~1.2 and 35/1.4 on a small target at short range right were I would expect distortion to be at its worse. Then they report the distortion but that report is worst case distortion and not typical of general use. I'm sure most who read these reports never question there validity.

At this time I'd trust Ken Rockwell over Digi Lloyd to perform such a test. At least Ken would explain his procedure in the hope of donations and click-throughs rather than throwing up a pay now or ...

It's not so easy to setup a valid test. In the case of censor alignment it would seem easy to find a problem where there is none.

Dave Hartman who can't be the early bird or adopter for various reasons.

[Øivind Tøien]

I said the effect of a 0.1mm difference in flange focal distance on infinity focus is negligible - and an effect on infinity focus is what diglloyd says he can observe.  Plus, any effect of the flange focal distance being wrong would affect the whole range by the same amount: the effect of changing the position of the real principal plane by any given amount would stay the same, just as it is with an extension tube.  An observation that " [...] the D850 might indicate focus confirmation at 60 feet out, but the taken image is actually focused about 15 feet out" cannot possibly be due to a 0.1mm error in flange focal distance.

Now he says the sensor is also not parallel to the mount (https://diglloyd.com/blog/2017/20171001_1530-NikonD850-sensor-lensMount-parallelism.html).  What he says about the effect of the sensor not being parallel to the mount is obviously wrong, and the pictures he shows that purport to prove the claim show nothing of the sort.

There have been previous episodes of people claiming that out-of-adjustment cameras were common and causing a problem - in the case of focus issues there was the one where back/front focus was supposed to be caused by the mirror being at the wrong angle, and you had to adjust it, eg (http://leongoodman.tripod.com/d70focuspart3.html).

I was talking about infinity focus too. It does have effect on critical focus. I had a whole long thread in the old forum about my problems in this respect with a D5100  (which also  had non-parallel planes). Some documentation still remains on my site, showing difference between hard stop and tweaked focus with my 28mm f/2.8 AIS wide open, http://otoien.zenfolio.com/p1004553826/e4530c452 , in this case also a 0.1mm misalignment. It could be the difference between being able to set the lens to the hard infinity stop and get good focus stopped down a little, and having to tweak the focus a little every time. (obviously I am am not one of those believing in setting the lens to an assumed hyperfocal distance....) In Astrophotography near wide open the hard stop/infinity mark is not to be trusted and common practice is tweaking focus in live view to get it perfect.

It seems to me he is going through exactly the same thing as I did with my D5100. That said, I agree that extrapolating from a couple of bodies to all bodies being defect is taking it way too far; a lot more data are needed. As with all things mechanical there are tolerances, and occasionally there can be an outlyer beyond the stated acceptable limits. After the D5100 problems mentioned above, I had opportunity to test a few more D5100 bodies, and there were subtle but much more acceptable differences, and they could go in different directions. (Looks like it would be a good project for Roger Cicala at the Lens rental blog getting more data on alignment for the D850...)

[Les Olson]

Some documentation still remains on my site, showing difference between hard stop and tweaked focus with my 28mm f/2.8 AIS wide open, http://otoien.zenfolio.com/p1004553826/e4530c452 , in this case also a 0.1mm misalignment. It could be the difference between being able to set the lens to the hard infinity stop and get good focus stopped down a little, and having to tweak the focus a little every time. (obviously I am am not one of those believing in setting the lens to an assumed hyperfocal distance....) In Astrophotography near wide open the hard stop/infinity mark is not to be trusted and common practice is tweaking focus in live view to get it perfect.

It seems to me he is going through exactly the same thing as I did with my D5100. That said, I agree that extrapolating from a couple of bodies to all bodies being defect is taking it way too far; a lot more data are needed. As with all things mechanical there are tolerances, and occasionally there can be an outlyer beyond the stated acceptable limits. After the D5100 problems mentioned above, I had opportunity to test a few more D5100 bodies, and there were subtle but much more acceptable differences, and they could go in different directions. (Looks like it would be a good project for Roger Cicala at the Lens rental blog getting more data on alignment for the D850...)

The fact that hard infinity stops are not trustworthy is well known, and has nothing to do with flange focal distance or mount alignment.  It has to do with the effect of temperature on the infinity focus point (https://www.bhphotovideo.com/explora/photography/features/who-killed-infinity-focus%3F).   

As far as the lens mount not being parallel to the sensor, it is easy to define what you would see if it were to happen: it is what you get with a tilt lens.  That is not remotely what diglloyd reports seeing.  In particular, with a tilted lens focused at infinity the plane of focus is perpendicular to the sensor independent of the degree of tilt - the degree of tilt only changes the location of the Scheimpflug point - the distance large format users call J (there is a quantitative treatment in Harold Merklinger's book about focusing view cameras http://www.trenholm.org/hmmerk/FVC161.pdf). 

[Øivind Tøien]

The fact that hard infinity stops are not trustworthy is well known, and has nothing to do with flange focal distance or mount alignment.  It has to do with the effect of temperature on the infinity focus point (https://www.bhphotovideo.com/explora/photography/features/who-killed-infinity-focus%3F).   

Yes, many ED lenses are temperature sensitive. But not all lenses are the same. For simple wider manual focus optics of some vintage the infinity point works pretty well, but as I remarked above, there are subtle deviations even then. The point of my remark was that a 0.1 mm axial difference in position of the lens barrel has observable effect on critical focus at infinity. Btw, I found back to the original thread where Bjørn stated that "the register distance (flange-sensor) tolerance is +- 0.05 mm" referring to Nikon bodies of course. So 0.1 mm is well out of range.

As far as the lens mount not being parallel to the sensor, it is easy to define what you would see if it were to happen: it is what you get with a tilt lens.  That is not remotely what diglloyd reports seeing.  In particular, with a tilted lens focused at infinity the plane of focus is perpendicular to the sensor independent of the degree of tilt - the degree of tilt only changes the location of the Scheimpflug point - the distance large format users call J (there is a quantitative treatment in Harold Merklinger's book about focusing view cameras http://www.trenholm.org/hmmerk/FVC161.pdf).

Thanks for the link. Can you point us to the exact figure that describes the situation. One can get a little overloaded by all the details paging through the document... Theory aside, I can only refer to what I observed. Today's optics, and particularly short focal lengths may be more complicated than that used for the view cameras, so different rules might apply. What was regarded as negligible in the old days may not be so today with high res. sensors or the effect is larger if lens is slightly defocused. The effect was moderate but was there, just as 0.1 mm shift has observable effect on critical focus near infinity. The mount was also visibly tilted. https://www.fotozones.com/live/index.php?/forums/topic/45315-solved-repair-woes-d5100/&. I would rather not think more about that traumatic experience with Nikon repair, although it all eventually ended well with an exchange to a new body.

[Les Olson]

At http://www.northlight-images.co.uk/using-lens-tilt-on-your-digital-slr/ there is a version with animations to follow the planes, and examples.

[Øivind Tøien]

At http://www.northlight-images.co.uk/using-lens-tilt-on-your-digital-slr/ there is a version with animations to follow the planes, and examples.

Thanks for the link (although still a bit hard to follow after a quick look without experience with the concept; bookmarked for later reference) Another thought is that if the lens has field curvature, the tilt could bring a defocused perimeter asymmetrically into the frame. Also it seems that if the lens is not quite focused at infinity (but the subject is at infinity) there would be some tilt of the focused image.

[Eric Borgström]

WoW. so lloyds used a camera which is not his, attached to a non Nikon lens and decided that the D850 is crap because it does not reach "Infinty and beyond"

I mean seriously. Is this guy real. I appreciate it is Autumn and the cyberweather is a bit slow and need to pick up viewership/readership.......paid of course........

I mean is there a remote possibility that the reverse engineering by Zeiss on his particular Otus lens does not marry very well with his and his friend cameras.

Oh, I know, let's ask Thom and Ming. They know everything even before it happens.

Apparently they are now the Holly Oracle Trinity: Lloyd - Thom and Ming.

This Zeiss lens worked well on Lloyds D810 as did his other Zeiss lenses. Lloyd might have a point. This specific camera probably has a problem.
/Eric

[Les Olson]

Also it seems that if the lens is not quite focused at infinity (but the subject is at infinity) there would be some tilt of the focused image.

No, there is no such thing as focusing an image, or focusing a lens - although we talk so much as if there was that it sounds odd to say there isn't.

If you have a lens sitting in the light all on its own it forms a three-dimensional image behind itself, called an aerial image.  The whole aerial image is sharp - there is no depth of field.  The image is reversed - left-right, top-bottom, and near-far.  Things at different distances in front of the lens are imaged at different distances behind it, and the exact distance is given by the lens conjugate equation: 1/f = 1/u + 1/v, where f is focal length, u is the distance to the object in object space and v is the distance to the image in image space.  When u is very large - "infinity" as one might say, exaggerating slightly - 1/u is very small and 1/f = 1/v.  That is, objects at infinity are imaged at the focal length (or, the definition of focal length is the distance at which objects at infinity are imaged).  As u gets smaller, 1/u gets bigger, and to keep 1/f the same, 1/v has to get smaller, so v has to get bigger.  So objects closer than infinity are imaged further away from the lens.  (This is why diglloyd's claim that his D850 had a flange focal distance that is too short - "46.4 instead of 46.5" - so he loses infinity focus is 100% solid-gold no-way-out wrong).

When you use a view camera you get an image of things closer (lower u) by moving the lens away from the ground glass (bigger v).  You can't do that with a 35mm camera, because it has a fixed lens mount.  But v is measured from the rear principal plane, and that can be anywhere - in the lens, behind it or in front - and it can be moved by changing the position of lens elements. That is mostly how a 35mm lens "focuses". (There is another way: if you look again at the lens conjugate equation, you will see that if u gets smaller, so 1/u gets bigger, but v does not change, 1/u + 1/v is bigger, so 1/f is bigger and f is smaller.  So when you have moved the real principal plane as far as you can, and you want u to be still smaller, you can move some other lens elements and change f, which is "focus breathing"). 

If you tilt the lens or the film, that means that part of the sensor/film has v bigger than another part.  That means those parts of the film/sensor have sharp images of things at different distances.  It does not mean that one side of the image is "soft" or out of focus - unless you are photographing something two dimensional.  That is why you can use a tilt lens to manipulate the depth of field. 

[Bill Mellen]

No, there is no such thing as focusing an image, or focusing a lens - although we talk so much as if there was that it sounds odd to say there isn't.

If you have a lens sitting in the light all on its own it forms a three-dimensional image behind itself, called an aerial image.  The whole aerial image is sharp - there is no depth of field.  The image is reversed - left-right, top-bottom, and near-far.  Things at different distances in front of the lens are imaged at different distances behind it, and the exact distance is given by the lens conjugate equation: 1/f = 1/u + 1/v, where f is focal length, u is the distance to the object in object space and v is the distance to the image in image space.  When u is very large - "infinity" as one might say, exaggerating slightly - 1/u is very small and 1/f = 1/v.  That is, objects at infinity are imaged at the focal length (or, the definition of focal length is the distance at which objects at infinity are imaged).  As u gets smaller, 1/u gets bigger, and to keep 1/f the same, 1/v has to get smaller, so v has to get bigger.  So objects closer than infinity are imaged further away from the lens.  (This is why diglloyd's claim that his D850 had a flange focal distance that is too short - "46.4 instead of 46.5" - so he loses infinity focus is 100% solid-gold no-way-out wrong).

When you use a view camera you get an image of things closer (lower u) by moving the lens away from the ground glass (bigger v).  You can't do that with a 35mm camera, because it has a fixed lens mount.  But v is measured from the rear principal plane, and that can be anywhere - in the lens, behind it or in front - and it can be moved by changing the position of lens elements. That is mostly how a 35mm lens "focuses". (There is another way: if you look again at the lens conjugate equation, you will see that if u gets smaller, so 1/u gets bigger, but v does not change, 1/u + 1/v is bigger, so 1/f is bigger and f is smaller.  So when you have moved the real principal plane as far as you can, and you want u to be still smaller, you can move some other lens elements and change f, which is "focus breathing"). 

If you tilt the lens or the film, that means that part of the sensor/film has v bigger than another part.  That means those parts of the film/sensor have sharp images of things at different distances.  It does not mean that one side of the image is "soft" or out of focus - unless you are photographing something two dimensional.  That is why you can use a tilt lens to manipulate the depth of field.

Beautifully explained Les!

[Øivind Tøien]

If you tilt the lens or the film, that means that part of the sensor/film has v bigger than another part.  That means those parts of the film/sensor have sharp images of things at different distances.

Emphasis by me; that is the observation that made me draw my conclusions regarding the D5100.

I agree that Loyd is in error in interpretation of the register distance; if the lens had to be focused further past infinity, it is of course to compensate for too long register distance. (In my case compensation was in the opposite direction).

[Les Olson]

Sorry, I must have missed something.  Which of the images in your link show a tilted plane of focus? 

Remember that if the lens is focused at infinity and there is any lens tilt the plane of focus is perpendicular to the sensor (independent of the degree of tilt) and that tilting the lens has a different effect to tilting the sensor - it is the difference between tilting the lens with the front standard of a view camera and tilting the film with the rear standard.  The reason view cameras are expensive is that being sure whether there is tilt and how much requires very precise calibration.  In your case, using a 35mm camera, even accepting for sake of argument that the mount was not straight, how can you know whether the mount and the lens were parallel to the scene and the sensor was tilted, or the sensor was parallel and the lens plane was tilted (or some of both)?

[Jack Dahlgren]

Sorry, I must have missed something.  Which of the images in your link show a tilted plane of focus? 

Remember that if the lens is focused at infinity and there is any lens tilt the plane of focus is perpendicular to the sensor (independent of the degree of tilt) and that tilting the lens has a different effect to tilting the sensor - it is the difference between tilting the lens with the front standard of a view camera and tilting the film with the rear standard.  The reason view cameras are expensive is that being sure whether there is tilt and how much requires very precise calibration.  In your case, using a 35mm camera, even accepting for sake of argument that the mount was not straight, how can you know whether the mount and the lens were parallel to the scene and the sensor was tilted, or the sensor was parallel and the lens plane was tilted (or some of both)?

Please check this. I do not think plane of focus would ever be perpendicular to sensor in any sort of camera that I’ve ever seen.

Also, on view camera, simple geometry tells us that tilting rear standard + front standard rise/fall can result in the exact same relationship between film plane and axis of lens. It is just a question of which is more convenient in a particular situation.

Final thing is that typically view cameras are not and don’t need to be as precise as smaller cameras. The smaller the lens and camera the more precision there must be. Tolerances shrink proportionally. View cameras are cheap these days compared to DSLRs.

[pluton]

...how can you know whether the mount and the lens were parallel to the scene and the sensor was tilted, or the sensor was parallel and the lens plane was tilted (or some of both)?

In the DigiLloyd case, he had lenses which been fine on multiple Nikon cameras up until the bad D850 sample.