Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

Started 4 months ago | Discussions
MayaTlab0
MayaTlab0 Senior Member • Posts: 2,868
Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

Hi, there’s a phenomenon I’ve noticed in some samples taken with the recent Nikon 50mm S that I find quite peculiar - at least to the extent that I’ve never seen anything quite like it before, or at least not with quite such prevalence, and I'd welcome your input regarding what could potentially cause it.

Some samples taken with this lens feature what DPreview called a “bullseye” pattern (top marks to DPreview for being the only publication so far that mentions that issue), even in the centre of the frame, distinct from the usual onion rings one can see in some lenses, but also distinct from the typical ways most lenses look like when slightly defocusing a point light source. Basically slightly defocused point light sources, or small contrasty objects, may exhibit a bright rim mid way between the centre and the edge of the blurred object. It look like this, although the relative brightness differences between the central hotspot, the edge of the defocused object, and the very annoying interesting rim mid-way through seems to vary meaning that they won’t show up in the same way in all pictures :

Photographyblog’s gallery is quite useful to see that phenomenon, it’s easily visible even in quite defocused point light sources : https://img.photographyblog.com/reviews/nikon_z_50mm_f1_8_s/sample_images/nikon_z_50mm_f1_8_s_25.jpg

Here what appears to be the same phenomenon on a more generic subject : https://www.imaging-resource.com/PRODS/nikon-z6/FULLRES/Y-WB-Z6-DSC_1049.HTM

Here an example with less defocused objects. Unlike typical onion rings that wouldn't show up with such a little amount of defocus, in that crop it seems that the bright mid rim dominates (red arrow) - and IMO ruins the overall picture as it's very susceptible to sharpening. Interestingly in that crop the lens also produced a number of defocused point light sources closer to what I would expect (green arrow). While not a point light source, the grass seems narrow and bright enough to have triggered the appearance of the mid-rim (blue). https://www.flickr.com/photos/145268771@N04/46403229932/in/album-72157698622513220/

Here what I think could be an example of the same phenomenon, again with a small amount of defocus. https://www.dpreview.com/sample-galleries/9560366311/nikon-z-50mm-f1-8-s-sample-gallery/5340007587

When the object is only slight defocused, it seems that the mid-rim's brightness dominates the other features of the defocused object (hence the very ugly results in the third sample). When the defocusing increases it seems to start to level. At some point it disappears, perhaps more so if the object is large enough to not excite that phenomenon too much to begin with. In typical bokeh balls test it doesn't show up at all - while onion rings would.

Interestingly, the degree of that problem seems to some extent dependent on the sample gallery. Dpreview’s only show a mild problem. But Asobinet’s gallery on flickr shows it with a much stronger intensity in a large variety of situations and at various focusing distances. In general if I see it popping up in one user’s gallery, it’s likely to pop up in other shots, and if it’s only mild, it seems to remain mild in other shots. But it could simply be a case of shooting style / subject matter, exacerbated by processing, not necessarily sample variation, I don’t know.

I would find it surprising that this would have been a conscious decision to design this lens in such a way, at least in the early phases of the design. Could it be related to an unforeseen production issue, or at least something that didn't show up in Nikon's simulations but did when the lens was actually produced ? Could it be related to the manufacturing of some lens elements (that it shows up in more defocused point light sources in some samples reminds me of onion rings related to CNC manufacturing I believe - but contrary to somewhat inoffensive onion rings the bright mid rim seems to shop up quite violently in just slightly defocused point light sources at times and won't show up in fully defocused objects). Is there anything about the design of this lens that makes it more susceptible to such a problem ?

If it's a production issue, what could / would Nikon do ? Just let it be and learn from it or try to fix it ?

Thanks for your input !

Note : please explain for dummies. Think "me, Jane, you, Tarzan".

Tom Axford Veteran Member • Posts: 6,469
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

Have a look at this article on bokeh.

There is also a more authoritative article by Nasse on DoF and Bokeh.

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

It is also worth remembering that the lens designer has in-focus image quality as top priority in most cases, not the quality of the bokeh.

There are a small number of lenses available which were designed primarily for bokeh.  They are expensive and very slow by current standards (e.g. Sony 100mm f/2.8 STF GM OSS )

J A C S
J A C S Forum Pro • Posts: 15,197
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

Tom Axford wrote:

It is also worth remembering that the lens designer has in-focus image quality as top priority in most cases, not the quality of the bokeh.

It should have both, together with many others. I think that I read somewhere that Zeiss optimizes for about 30 parameters or so.

The examples the OP posts are horrible. Almost like a mirror lens.

Tom Axford Veteran Member • Posts: 6,469
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

J A C S wrote:

Tom Axford wrote:

It is also worth remembering that the lens designer has in-focus image quality as top priority in most cases, not the quality of the bokeh.

It should have both, together with many others. I think that I read somewhere that Zeiss optimizes for about 30 parameters or so.

Of course, but some are given a higher priority than others. I suspect that for the OP's lens low cost was the top priority, then high IQ for in-focus images, and bokeh after that.

MayaTlab0
OP MayaTlab0 Senior Member • Posts: 2,868
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?
1

Tom Axford wrote:

Have a look at this article on bokeh.

Thanks for chiming in and thanks for that link. I had read it a while ago.

There is also a more authoritative article by Nasse on DoF and Bokeh.

Idem. A nice introductory article with good illustrations for people wondering why this or that lens behaves this or that way.

While the latter article is sufficient to help me understand to some degree what I can see for example in the little test I did out of curiosity some time ago with my 50 STM and why I consider it a f2.5-f2.8 lens with bonus apertures if one isn't concerned about bokeh, I don't think that it's quite enough to describe what is going on with this 50 S - or at least some of these 50 S lenses.

Canon 50mm STM, rear defocus, two different source sizes. Not a rigorously conducted test by any stretch of the imagination. Focusing distance around 3m, around the centre of the frame - things obviously look much different off axis, particularly with such a lens, different shutter speeds to keep the blobs' brightness as constant as I could, but brightness still adjusted in post in C1 (so maybe some linearity problems (?), and definitely a noise issue - don't look too much into the fine structure inside the blobs), LED lamp as the light source (so not a superbly uniform spectrum - maybe affects CA ?), a plain mate black cardboard in front of it with two less than perfectly circular holes punched in it (I wanted to see how the difference in object size would affect the defocused behaviour - I assembled the pictures in photoshop but they are to scale if I recall correctly), white balance eyeballed so that the CA looks somewhat logical but not done with much exactitude, you get the idea. But after reading Marianne's thread I was curious to see if I could make my 50 STM's very traditional behaviour when defocusing a point light source more explicitly visible with some basic tools. This sort of result (a brighter hotspot surrounded by a bright ring at the outer edge) is what I see in a lot of classic lenses - and why I tend to dislike their bokeh wide open but love it in the centre, closed down. In the 50 S there's a bright rim mid-way through the defocused ball in some image samples, and unfortunately its location means that it doesn't go away quickly when stopping down.

This thread was also interesting but unfortunately the pictures have gone missing : https://www.dpreview.com/forums/thread/4031515

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

I'm not in any capacity to single out the aspheres' design / manufacturing / use as the cause of that peculiar behaviour. But I'd welcome some input as to what could have lead to these results, and why would a company like Nikon end up with that. The weird thing to me about this phenomenon is that it seems to straddle the ground between "slightly defocused" and "fully defocused" as the mid-rim manifests itself way sooner than onion rings typically would, but persists into the defocus further than typical SA features for a modern lens, at least in some samples.

It is also worth remembering that the lens designer has in-focus image quality as top priority in most cases, not the quality of the bokeh.

I am unsure about this. The 50 S shows very clear signs that in the grand scheme of things it's been designed just like a lot of lenses with a deliberate fine tuning of aberrations to bias the blur in favour of rear transitions. A picture like the one below wouldn't show such a hard edge to the foreground blur vs. the background blur I think if it weren't the case (all right, close focusing distance but still) : https://diglloyd.com/blog/2019/images/2019-0131-MiningGears-NikonZ50f1_8S-f1_8.jpg@auto.html?dglySz=4320x2880&dglyD=-C-g_ZBuHAuMsOm75DOT02u-tJzrtzyRhay2E5_lnbRsc39vBaiMwTaY7e_I2MFLD98x4z87nwB4cpUTRqSOvNIxf0bu_Z82fKvQg-k8UQmb0uhMNzgbdzXHl86hBKIZUTp33cd3tN-pZhu2Hw_v46hrAnFKWlR4Aaa1cY3d00BvIb1t3PxEFka5C77lcy3iA1kYJJKMr58Y_jfS9XHU63nXF1uaKXsAUS_tfUG3KiESA1OofKCqUdFMztQlA90BAdbWcG7zkBZZxe2Aq0Ngi3l2IkNj_X4=

I still think that even at a distance a lot of samples from that lens show some slightly gaussian quality to its rear bokeh typical of such design... at least as long as the subject / context doesn't trigger the issue discussed here.

It's also remarkably homogenous throughout the frame both in and out of focus for a lens in its price range - and that's something I'm increasingly valuing highly as far as bokeh is concerned.

My little finger tells me that it was designed just like most of Nikon's recent primes (it seems that the company has been quite consistent in terms of what they believe makes for good bokeh), that something wasn't properly anticipated at the design stage and that the translation into manufacturing led to this problem - at least with some copies, but that's just me making an uneducated guess and my little finger is sometimes quite wrong.

AiryDiscus Senior Member • Posts: 1,866
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?
1

Tom Axford wrote:

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

If the optics are ground, they are polished after.  The surfaces are "flat" (depart from the ideal profile) to better than 100 nm over the entire clear aperture.  Grinding only gets you to something like 5 microns.  It is not molding to blame, but CNC machines operating on highly structured grids to produce the desired shape.  whether that processing is on the optical elements or a mold is irrelevant.  The traditional method of polishing a sphere is to spin the optic and the tool, which are both "full aperture."  This results in the appearance of random motion over the entire surface, with maybe some very low order control over polishing the middle vs the edge more.  This leads to very smooth shapes.

CNC moves only the tool, and does so on a structured grid, which prints a signature into the part that must be smoothed out or mitigated in the design of the CNC machine or its tool.

AiryDiscus Senior Member • Posts: 1,866
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?
2

MayaTlab0 wrote:

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

I'm not in any capacity to single out the aspheres' design / manufacturing / use as the cause of that peculiar behaviour. But I'd welcome some input as to what could have lead to these results, and why would a company like Nikon end up with that. The weird thing to me about this phenomenon is that it seems to straddle the ground between "slightly defocused" and "fully defocused" as the mid-rim manifests itself way sooner than onion rings typically would, but persists into the defocus further than typical SA features for a modern lens, at least in some samples.

The problem is not the aspheres.

It is also worth remembering that the lens designer has in-focus image quality as top priority in most cases, not the quality of the bokeh.

I am unsure about this. The 50 S shows very clear signs that in the grand scheme of things it's been designed just like a lot of lenses with a deliberate fine tuning of aberrations to bias the blur in favour of rear transitions. A picture like the one below wouldn't show such a hard edge to the foreground blur vs. the background blur I think if it weren't the case (all right, close focusing distance but still) : https://diglloyd.com/blog/2019/images/2019-0131-MiningGears-NikonZ50f1_8S-f1_8.jpg@auto.html?dglySz=4320x2880&dglyD=-C-g_ZBuHAuMsOm75DOT02u-tJzrtzyRhay2E5_lnbRsc39vBaiMwTaY7e_I2MFLD98x4z87nwB4cpUTRqSOvNIxf0bu_Z82fKvQg-k8UQmb0uhMNzgbdzXHl86hBKIZUTp33cd3tN-pZhu2Hw_v46hrAnFKWlR4Aaa1cY3d00BvIb1t3PxEFka5C77lcy3iA1kYJJKMr58Y_jfS9XHU63nXF1uaKXsAUS_tfUG3KiESA1OofKCqUdFMztQlA90BAdbWcG7zkBZZxe2Aq0Ngi3l2IkNj_X4=

I still think that even at a distance a lot of samples from that lens show some slightly gaussian quality to its rear bokeh typical of such design... at least as long as the subject / context doesn't trigger the issue discussed here.

It's also remarkably homogenous throughout the frame both in and out of focus for a lens in its price range - and that's something I'm increasingly valuing highly as far as bokeh is concerned.

My little finger tells me that it was designed just like most of Nikon's recent primes (it seems that the company has been quite consistent in terms of what they believe makes for good bokeh), that something wasn't properly anticipated at the design stage and that the translation into manufacturing led to this problem - at least with some copies, but that's just me making an uneducated guess and my little finger is sometimes quite wrong.

The design isn't optimized for front or back bokeh, or at least not on large scale.  Most optical design forms tend towards undercorrection of spherical aberration.  Undercorrected (as opposed to overcorrected) makes a softer blur in the background than the foreground.    So, the soft background is "the natural way of things" and it would (tend to) require futzing the design into an ugly space to make the front softer instead.

AiryDiscus Senior Member • Posts: 1,866
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?
1

MayaTlab0 wrote:

Hi, there’s a phenomenon I’ve noticed in some samples taken with the recent Nikon 50mm S that I find quite peculiar - at least to the extent that I’ve never seen anything quite like it before, or at least not with quite such prevalence, and I'd welcome your input regarding what could potentially cause it.

Some samples taken with this lens feature what DPreview called a “bullseye” pattern (top marks to DPreview for being the only publication so far that mentions that issue), even in the centre of the frame, distinct from the usual onion rings one can see in some lenses, but also distinct from the typical ways most lenses look like when slightly defocusing a point light source. Basically slightly defocused point light sources, or small contrasty objects, may exhibit a bright rim mid way between the centre and the edge of the blurred object. It look like this, although the relative brightness differences between the central hotspot, the edge of the defocused object, and the very annoying interesting rim mid-way through seems to vary meaning that they won’t show up in the same way in all pictures :

Photographyblog’s gallery is quite useful to see that phenomenon, it’s easily visible even in quite defocused point light sources : https://img.photographyblog.com/reviews/nikon_z_50mm_f1_8_s/sample_images/nikon_z_50mm_f1_8_s_25.jpg

Here what appears to be the same phenomenon on a more generic subject : https://www.imaging-resource.com/PRODS/nikon-z6/FULLRES/Y-WB-Z6-DSC_1049.HTM

Here an example with less defocused objects. Unlike typical onion rings that wouldn't show up with such a little amount of defocus, in that crop it seems that the bright mid rim dominates (red arrow) - and IMO ruins the overall picture as it's very susceptible to sharpening. Interestingly in that crop the lens also produced a number of defocused point light sources closer to what I would expect (green arrow). While not a point light source, the grass seems narrow and bright enough to have triggered the appearance of the mid-rim (blue). https://www.flickr.com/photos/145268771@N04/46403229932/in/album-72157698622513220/

Here what I think could be an example of the same phenomenon, again with a small amount of defocus. https://www.dpreview.com/sample-galleries/9560366311/nikon-z-50mm-f1-8-s-sample-gallery/5340007587

When the object is only slight defocused, it seems that the mid-rim's brightness dominates the other features of the defocused object (hence the very ugly results in the third sample). When the defocusing increases it seems to start to level. At some point it disappears, perhaps more so if the object is large enough to not excite that phenomenon too much to begin with. In typical bokeh balls test it doesn't show up at all - while onion rings would.

Interestingly, the degree of that problem seems to some extent dependent on the sample gallery. Dpreview’s only show a mild problem. But Asobinet’s gallery on flickr shows it with a much stronger intensity in a large variety of situations and at various focusing distances. In general if I see it popping up in one user’s gallery, it’s likely to pop up in other shots, and if it’s only mild, it seems to remain mild in other shots. But it could simply be a case of shooting style / subject matter, exacerbated by processing, not necessarily sample variation, I don’t know.

I would find it surprising that this would have been a conscious decision to design this lens in such a way, at least in the early phases of the design. Could it be related to an unforeseen production issue, or at least something that didn't show up in Nikon's simulations but did when the lens was actually produced ? Could it be related to the manufacturing of some lens elements (that it shows up in more defocused point light sources in some samples reminds me of onion rings related to CNC manufacturing I believe - but contrary to somewhat inoffensive onion rings the bright mid rim seems to shop up quite violently in just slightly defocused point light sources at times and won't show up in fully defocused objects). Is there anything about the design of this lens that makes it more susceptible to such a problem ?

If it's a production issue, what could / would Nikon do ? Just let it be and learn from it or try to fix it ?

Thanks for your input !

Note : please explain for dummies. Think "me, Jane, you, Tarzan".

Much as I gag at him, Rockwell has some sample photos for bokeh which do not show this phenomenon.

https://kenrockwell.com/nikon/mirrorless/lenses/50mm-f18.htm

In such a well corrected lens, there cannot be enough aberration to reshape the light that much, so the issue would have to be an amplitude problem.  Perhaps there are issues with the coatings on some of them - the design is really ugly and full of high ray angles, which makes the coating design a lot harder and the process of applying it no doubt trickier.

Perhaps PB's copy is pre-production and the coating was not finalized at that time.  Though Nikon is really quite incredibly good at coatings, so I would be surprised if they had intermittent issues.

Ayoh Regular Member • Posts: 354
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

AiryDiscus wrote:

Tom Axford wrote:

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

If the optics are ground, they are polished after. The surfaces are "flat" (depart from the ideal profile) to better than 100 nm over the entire clear aperture. Grinding only gets you to something like 5 microns. It is not molding to blame, but CNC machines operating on highly structured grids to produce the desired shape. whether that processing is on the optical elements or a mold is irrelevant. The traditional method of polishing a sphere is to spin the optic and the tool, which are both "full aperture." This results in the appearance of random motion over the entire surface, with maybe some very low order control over polishing the middle vs the edge more. This leads to very smooth shapes.

CNC moves only the tool, and does so on a structured grid, which prints a signature into the part that must be smoothed out or mitigated in the design of the CNC machine or its tool.

For moulded aspheres, I thought the concentric rings were due to machining marks in the metal moulds. Is the case or is there another effect involved? Thanks

SmilerGrogan Contributing Member • Posts: 758
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

So to put it in photography terms, are you saying that the patterns we see in the out of focus highlights are just images of the microscopic bumps, dips, and grooves on the surface of an element in the lens group?

Is it just one lens in the group that that causes the problem? Or are we seeing the sum of all the imperfections on all the elements?

And finally, it sounds like you're saying that the process is not to blame, it's more due to the choice of that particular CnC system or the design of the shaping/polishing protocol.

AiryDiscus wrote:

Tom Axford wrote:

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

If the optics are ground, they are polished after. The surfaces are "flat" (depart from the ideal profile) to better than 100 nm over the entire clear aperture. Grinding only gets you to something like 5 microns. It is not molding to blame, but CNC machines operating on highly structured grids to produce the desired shape. whether that processing is on the optical elements or a mold is irrelevant. The traditional method of polishing a sphere is to spin the optic and the tool, which are both "full aperture." This results in the appearance of random motion over the entire surface, with maybe some very low order control over polishing the middle vs the edge more. This leads to very smooth shapes.

CNC moves only the tool, and does so on a structured grid, which prints a signature into the part that must be smoothed out or mitigated in the design of the CNC machine or its tool.

AiryDiscus Senior Member • Posts: 1,866
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

Ayoh wrote:

AiryDiscus wrote:

Tom Axford wrote:

With modern lenses using moulded aspherical elements, bokeh can be much more complicated than was usually the case with older lenses made only of ground spherical elements.

If the optics are ground, they are polished after. The surfaces are "flat" (depart from the ideal profile) to better than 100 nm over the entire clear aperture. Grinding only gets you to something like 5 microns. It is not molding to blame, but CNC machines operating on highly structured grids to produce the desired shape. whether that processing is on the optical elements or a mold is irrelevant. The traditional method of polishing a sphere is to spin the optic and the tool, which are both "full aperture." This results in the appearance of random motion over the entire surface, with maybe some very low order control over polishing the middle vs the edge more. This leads to very smooth shapes.

CNC moves only the tool, and does so on a structured grid, which prints a signature into the part that must be smoothed out or mitigated in the design of the CNC machine or its tool.

For moulded aspheres, I thought the concentric rings were due to machining marks in the metal moulds. Is the case or is there another effect involved? Thanks

Bolded now....

The marks you see are not to do with molding, but to do with CNC.  Most lenses you see today still do not contain molded elements - in particular this one does not.  Nikon marks this with their "PGM" alphabet soup - precision glass moulding.

The molds are usually silicon or tungsten carbide - metal would deform far too much under the temperatures and pressures involved.

AiryDiscus Senior Member • Posts: 1,866
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?
1

SmilerGrogan wrote:

So to put it in photography terms, are you saying that the patterns we see in the out of focus highlights are just images of the microscopic bumps, dips, and grooves on the surface of an element in the lens group?

Typically, they are, yes. In this case I think they are not / it is unlikely.

Is it just one lens in the group that that causes the problem? Or are we seeing the sum of all the imperfections on all the elements?

It would take a phenomenal eye to fish out wish element(s) it came from. You see "the sum" of them.

And finally, it sounds like you're saying that the process is not to blame, it's more due to the choice of that particular CnC system or the design of the shaping/polishing protocol.

It is categorically the use of CNC. Whether those CNC machines generate their signatures on molds or glass directly is immaterial.

However, to fabricate an aspheric optic without CNC is exceptionally difficult, and even to make a sphere without CNC requires skilled opticians, who are a dying breed. Thus, the trajectory is towards CNC, with added smoothing.

As an example, Leica produces their aspheres on CNC machines produced by this vendor, which are themselves top-notch

https://www.optotech.de/en/precision-optics%20

They are then finished on one of the two premiere finishing technologies,

https://qedmrf.com/en/

(the other is IBF -  ion beam figuring)

And finally smoothed by hand.  This is why you see so little artifacts in the bokeh of Leica lenses, even though they have quite a large number of aspheric surfaces.  They expend tremendous capital to clean up the problems made by CNC, and produce lenses in a volume where this drawn out process is viable.

MayaTlab0
OP MayaTlab0 Senior Member • Posts: 2,868
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

AiryDiscus wrote:

The design isn't optimized for front or back bokeh, or at least not on large scale. Most optical design forms tend towards undercorrection of spherical aberration. Undercorrected (as opposed to overcorrected) makes a softer blur in the background than the foreground. So, the soft background is "the natural way of things" and it would (tend to) require futzing the design into an ugly space to make the front softer instead.

Duly noted. Let's not dwell on that too much here as it's not the main subject, but I've frequently seen the sort of rather typical defocused point light sources my Canon 50mm STM produces in the centre of the frame (cf. above) described as being "undercorrected" as far as the hotspot is concerned and "overcorrected" as far as the brighter rim at the edge is. That's to some extent the terminology Dr. Nasse's article used. Hence why I thought that more modern lenses made a deliberate effort to minimise the "overcorrected" part - which certainly is quite ugly in photos.

Should I understand from your comment that it's wrong to use the "overcorrected" terminology to describe the bright rim on the 50 STM's edge ?

In Dr. Nasse' article, he mentions that the lens with spherical aberration plotted on the graph on the left below produces a defocused point light source with the brightness profile next to it :

Just as you wrote, the SA curve seems to stay firmly in the "under-corrected" direction but the lens still produces a - seemingly mild - brighter rim.

What terminology would be better suited to describe this ?

MayaTlab0
OP MayaTlab0 Senior Member • Posts: 2,868
Thanks for chiming in.

AiryDiscus wrote:

Much as I gag at him, Rockwell has some sample photos for bokeh which do not show this phenomenon.

https://kenrockwell.com/nikon/mirrorless/lenses/50mm-f18.htm

Indeed.

In such a well corrected lens, there cannot be enough aberration to reshape the light that much, so the issue would have to be an amplitude problem.

Sorry Brandon but you're starting to diverge from the "you Jane me Tarzan" or at least Carl Sagan level of vulgarisation :D. I get what a waveform's amplitude is for example but you're probably using the term in a more specific way I'm unlikely to understand or to know how to apply to optics.

If I opened a thread here it's because I sort of ran out of optics for dummies vulgarisation articles that would start to shine a light on that weird phenomenon - but perhaps it's such a specific, rare issue that it doesn't deserve one or that it's too complex to vulgarise without excessively loosing accuracy.

So, what I understand so far :

- the phenomenon is visible on axis, so if it were caused by typical aberrations it would have do to with aberrations present on axis.

- You say that the lens is too well corrected for these aberrations to "reshape the light" that much - I'm understanding here the impact of them - notably SA - on the distribution (my terminology but feel free to correct it) of brightness in defocused point light sources (central hotspot, bright rim, etc.). Basically how aberrations affect where the rays are going. Am I right (probably not quite I guess) ?

- So if it isn't about the distribution of light, it's an "amplitude" problem. And here I'm lost.

Good luck if you dare try to make me understand the "amplitude" bit, at least to some degree. Think of it as a challenge, perhaps ? Are there some articles you'd recommend me reading ?

Is it because it isn't caused by typical aberrations that the results are difficult to predict and appear more random to end users like me ? This is a photo of the 50 S's defocused point light sources at a very close focusing distance from a Hungarian website if I recall correctly. There's no trace whatsoever of the phenomenon discussed here and it looks like what I'd expect :

Perhaps there are issues with the coatings on some of them - the design is really ugly and full of high ray angles, which makes the coating design a lot harder and the process of applying it no doubt trickier.

Since this is such a weird behaviour I was expecting to see some potential causes coming from out of left field instead of "Aspheric lenses".

Should I understand that coatings are designed for specific ray angles and that a design like the 50 S requires a more thorough understanding / analysis of which sort of coating to apply on each surface and how, to prevent such issues from occurring ?

Now let's assume that I'll end up understanding the "amplitude" bit above (big assumption), how would coatings affect it to the extent that this behaviour happens ? Even if I don't get it others may be interested to know.

Perhaps PB's copy is pre-production and the coating was not finalized at that time. Though Nikon is really quite incredibly good at coatings, so I would be surprised if they had intermittent issues.

It might be a pre-production copy but some long time users of the lens on Flickr seem to own a copy that also exhibits that issue. We can see some traces of it in the following shots for example, mild or severe :

https://www.flickr.com/photos/123657498@N08/47661713162/in/pool-z50f18s/

https://www.flickr.com/photos/164582289@N02/33919562498/in/pool-z50f18s/

https://www.flickr.com/photos/ubicfromtokyo/40298615653/

https://www.flickr.com/photos/stephenjae/48392020961/ (this one exhibits the issue at a closer focusing distance).

I think that some evil copies affected by this problem were released into the wild and are wreaking havoc on some people's pictures.

AiryDiscus Senior Member • Posts: 1,866
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?

MayaTlab0 wrote:

AiryDiscus wrote:

The design isn't optimized for front or back bokeh, or at least not on large scale. Most optical design forms tend towards undercorrection of spherical aberration. Undercorrected (as opposed to overcorrected) makes a softer blur in the background than the foreground. So, the soft background is "the natural way of things" and it would (tend to) require futzing the design into an ugly space to make the front softer instead.

Duly noted. Let's not dwell on that too much here as it's not the main subject, but I've frequently seen the sort of rather typical defocused point light sources my Canon 50mm STM produces in the centre of the frame (cf. above) described as being "undercorrected" as far as the hotspot is concerned and "overcorrected" as far as the brighter rim at the edge is. That's to some extent the terminology Dr. Nasse's article used. Hence why I thought that more modern lenses made a deliberate effort to minimise the "overcorrected" part - which certainly is quite ugly in photos.

"under" and "over" corrected spherical aberration refers to the sign of spherical aberration (negative or positive) - nothing more.

Should I understand from your comment that it's wrong to use the "overcorrected" terminology to describe the bright rim on the 50 STM's edge ?

Yes, it is not correct.

In Dr. Nasse' article, he mentions that the lens with spherical aberration plotted on the graph on the left below produces a defocused point light source with the brightness profile next to it :

Just as you wrote, the SA curve seems to stay firmly in the "under-corrected" direction but the lens still produces a - seemingly mild - brighter rim.

What terminology would be better suited to describe this ?

I can't understand what I'm looking at.  Given that I have a degree in optics, that's not a very good sign.  I have no idea what those curves represent.

MayaTlab0
OP MayaTlab0 Senior Member • Posts: 2,868
Re: Nikon 50mm 1.8 S’ “bullseye” bokeh behaviour - design feature or production problem ?
2

AiryDiscus wrote:

MayaTlab0 wrote:

Should I understand from your comment that it's wrong to use the "overcorrected" terminology to describe the bright rim on the 50 STM's edge ?

Yes, it is not correct.

Noted.

In Dr. Nasse' article, he mentions that the lens with spherical aberration plotted on the graph on the left below produces a defocused point light source with the brightness profile next to it :

Just as you wrote, the SA curve seems to stay firmly in the "under-corrected" direction but the lens still produces a - seemingly mild - brighter rim.

What terminology would be better suited to describe this ?

I can't understand what I'm looking at. Given that I have a degree in optics, that's not a very good sign. I have no idea what those curves represent.

Oops. Well we'll forget about Nasse's article then. The one on the right plots the relative brightness of a defocused point between centre and edge apparently.

Anyway thanks for your input.

alanr0 Senior Member • Posts: 2,237
"Bullseye" bokeh - Amplitude or Phase distortion ?
6

AiryDiscus wrote:

MayaTlab0 wrote:

<snip>

Some samples taken with this lens feature what DPreview called a “bullseye” pattern (top marks to DPreview for being the only publication so far that mentions that issue), even in the centre of the frame, distinct from the usual onion rings one can see in some lenses, but also distinct from the typical ways most lenses look like when slightly defocusing a point light source. Basically slightly defocused point light sources, or small contrasty objects, may exhibit a bright rim mid way between the centre and the edge of the blurred object. It look like this, although the relative brightness differences between the central hotspot, the edge of the defocused object, and the very annoying interesting rim mid-way through seems to vary meaning that they won’t show up in the same way in all pictures :

Photographyblog’s gallery is quite useful to see that phenomenon, it’s easily visible even in quite defocused point light sources : https://img.photographyblog.com/reviews/nikon_z_50mm_f1_8_s/sample_images/nikon_z_50mm_f1_8_s_25.jpg

<snip>

Could it be related to an unforeseen production issue, or at least something that didn't show up in Nikon's simulations but did when the lens was actually produced ? Could it be related to the manufacturing of some lens elements (that it shows up in more defocused point light sources in some samples reminds me of onion rings related to CNC manufacturing I believe - but contrary to somewhat inoffensive onion rings the bright mid rim seems to shop up quite violently in just slightly defocused point light sources at times and won't show up in fully defocused objects). Is there anything about the design of this lens that makes it more susceptible to such a problem ?

<snip>

In such a well corrected lens, there cannot be enough aberration to reshape the light that much, so the issue would have to be an amplitude problem. Perhaps there are issues with the coatings on some of them - the design is really ugly and full of high ray angles, which makes the coating design a lot harder and the process of applying it no doubt trickier.

Perhaps PB's copy is pre-production and the coating was not finalized at that time. Though Nikon is really quite incredibly good at coatings, so I would be surprised if they had intermittent issues.

My emphasis.

Lens coating or refractive error

I understand your reasoning, but I doubt a dodgy coating would cause what we see here. If coating problems are to blame, I would expect dark rings to be clearly visible when viewing a diffuse reflecting surface through the front element of the lens. DxO measured the transmission as 1.9 T-stops, suggesting that transmission losses are low -- at least in the sample they tested.

How can we discriminate between these two explanations?

If errors in the surface profiles of lens elements are responsible, then we can apply the principles of curvature sensing. This predicts intensity fluctuations in the bokeh disk which are proportional to the curvature (Laplacian) of the wavefront distortion in the pupil plane, and inversely proportional to defocus distance in the image plane. Near to focus, diffraction upsets this simple picture, however curvature sensing has been used to measure aberrations, and applied to adaptive optics systems for astronomical telescopes and for retinal imaging.

The intensity changes are opposite in sign for front and rear defocus. If phase distortion is responsible, we should see complimentary ring patterns for defocused highlight in front of and behind the subject. In contrast, if coating shortcomings are responsible, there should be a similar pattern for front and rear defocus. Has anyone seen such images on-line?

How much wavefront distortion can we tolerate?

Absent more convincing evidence, or access to a suspect lens for testing, how plausible are phase errors as a cause? The radius of the Airy disk is around 1.2 micron at f/1.8, so we can go some way beyond the Rayleigh or Marechal criteria for diffraction-limited performance, without markedly degrading contrast at 30 cycles/mm or when using 4.3 micron pixels. We can tolerate larger errors in the outer margins of the lens, provided the inner rays are better corrected.

Below is a crop from the PhotographyBlog image linked above, and simulated point spread functions for front and rear defocus, referred to the image plane, of 0.65 mm.

The PB blur disk was approximately 90 pixels in diameter (0.39 mm in the sensor plane).

The simulation used a Fresnel/Huygens calculation, with defocus modelled as a quadratic phase variation across the lens aperture.

PB image from: https://img.photographyblog.com/reviews/nikon_z_50mm_f1_8_s/sample_images/nikon_z_50mm_f1_8_s_25.jpg. Monochrome crop compared with simulated blur at 0.65 mm defocus, f/1.8, 550 nm.

The Zernike simulation imposed a smoothly varying wavefront distortion across the lens. It shows the broad features seen in the PhotographyBlog bokeh, but the boundaries of the bright ring are rather smeared out.

The Quadratic/Cubic fit addresses this by alternating annular rings with constant curvature joined by cubic polynomial regions. The abrupt changes in curvature are responsible for the sharper radial transitions in the bokeh. Note that the first derivative of the wavefront is continuous, so there are no sharply defined ridges in the lens profile.

The Quad/Cubic foreground image shows similar structure with bright/dark transitions inverted compared with the background simulation. This is shown more clearly in the red and green traces below. The blue trace shows a horizontal slice through the PB blur.

Simulated intensity profiles across background and foreground bokeh disks, compared with horizontal profile through PhotographBlog crop. Quad/cubic distortion profile simulated.

The Zernike and Quad/Cubic wavefront distortion profiles used in the simulations are shown below. Peak-Valley distortion is 0.5 μm (0.9 wavelength). RMS distortion is 0.08 μm (1/7 wave).

Simulated radial wavefront distortion.

I don't claim that this is an accurate model of the lens aberrations - the intent was simply to show that a modest amplitude of wavefront distortion is consistent with the observed bokeh. The aberration profile above does result in some focus shift as the lens is stopped down, and this may not be representative.

How does this affect performance? - Lens MTF.

I used a similar Fresnel/Huygens diffraction calculation to simulate the point spread function, this time at the focal plane. The optical transfer function is the Fourier transform of the PSF. This is a monochromatic calculation, including diffraction, but neglecting longitudinal chromatic aberration (LCA), which is present according to Camera Labs' review.

Simulated MTF for Quad/Cubic wavefront distortion, aperture f/1.8 to f/8, 550 nm.

Simulated MTF at 30 cycles/mm is 0.86 at f/1.8 is1 broadly comparable with Nikon's value of 0.82 on-axis. I believe Nikon's published MTF curves include LCA, but not diffraction, while my simulation includes diffraction, but neglects LCA.

Nikkor Z 50 mm f/1.8 S MTF: https://imaging.nikon.com/lineup/lens/z-mount/z_50mmf18s/spec.htm

Bokeh variation with defocus.

As image-space defocus increases, the bokeh disk becomes larger and more uniform.

Simulated bokeh variation with image-space defocus distance. Top line is background defocus, lower line is foreground. f/1.8, 550 nm.

Conversely, with smaller defocus, the annular rings becomes more prominent, before diffraction finally dominates.

Simulated bokeh for smaller image-space defocus. Magnified x4 compared with previous image.

Compare the 0.4 and 0.2 mm defocus simulations with the 3rd and 4th images in the original post.

In summary, the broad features of the observed bokeh can be explained by a rotationally symmetric refractive error (P-V amplitude 0.5 μm), which is consistent with Nikon's nominal MTF at 30 cycles/mm.

-- hide signature --

Alan Robinson

Jack Hogan Veteran Member • Posts: 7,114
Re: "Bullseye" bokeh - Amplitude or Phase distortion ?

Brilliant.  Thank you for the excellent lesson Alan.

Jack

MayaTlab0
OP MayaTlab0 Senior Member • Posts: 2,868
Re: "Bullseye" bokeh - Amplitude or Phase distortion ?

alanr0 wrote:

Lens coating or refractive error

The intensity changes are opposite in sign for front and rear defocus. If phase distortion is responsible, we should see complimentary ring patterns for defocused highlight in front of and behind the subject. In contrast, if coating shortcomings are responsible, there should be a similar pattern for front and rear defocus. Has anyone seen such images on-line?

My emphasis.

[...]

Bokeh variation with defocus.

As image-space defocus increases, the bokeh disk becomes larger and more uniform.

Simulated bokeh variation with image-space defocus distance. Top line is background defocus, lower line is foreground. f/1.8, 550 nm.

Conversely, with smaller defocus, the annular rings becomes more prominent, before diffraction finally dominates.

Simulated bokeh for smaller image-space defocus. Magnified x4 compared with previous image.

Compare the 0.4 and 0.2 mm defocus simulations with the 3rd and 4th images in the original post.

Thanks for your input.

I gather from your post that you're looking for pictures with both fore and aft bokeh near the centre exhibiting a defocus pattern similar to the one above (ie with rings patterns opposite the others between fore and aft defocus), right ?

Jim Kasson has already performed a similar test but it's unclear to me if he either had a decent copy of the lens in that regard or if his methodology wasn't sensitive enough to display such behaviour : https://blog.kasson.com/nikon-z6-7/50-mm-f-2-nikkor-s-transition-point-spread-functions/

Not easy to find such shots among various samples and I won't be able to take pictures with the one copy I should have an easy access to for months but in the meantime :

https://www.flickr.com/photos/thomasrubach/44474353320/ (that's another account for Cameralabs)

https://img.photographyblog.com/reviews/nikon_z_50mm_f1_8_s/sample_images/nikon_z_50mm_f1_8_s_25.jpg

https://www.flickr.com/photos/145268771@N04/32362098138/in/album-72157698622513220/

Let me get back to this thread if I find a more convincing sample.

alanr0 Senior Member • Posts: 2,237
Re: "Bullseye" bokeh - Amplitude or Phase distortion ?

MayaTlab0 wrote:

alanr0 wrote:

Lens coating or refractive error

The intensity changes are opposite in sign for front and rear defocus. If phase distortion is responsible, we should see complimentary ring patterns for defocused highlight in front of and behind the subject. In contrast, if coating shortcomings are responsible, there should be a similar pattern for front and rear defocus. Has anyone seen such images on-line?

My emphasis.

[...]

Bokeh variation with defocus.

As image-space defocus increases, the bokeh disk becomes larger and more uniform.

Simulated bokeh variation with image-space defocus distance. Top line is background defocus, lower line is foreground. f/1.8, 550 nm.

Conversely, with smaller defocus, the annular rings becomes more prominent, before diffraction finally dominates.

Simulated bokeh for smaller image-space defocus. Magnified x4 compared with previous image.

Compare the 0.4 and 0.2 mm defocus simulations with the 3rd and 4th images in the original post.

Thanks for your input.

I gather from your post that you're looking for pictures with both fore and aft bokeh near the centre exhibiting a defocus pattern similar to the one above (ie with rings patterns opposite the others between fore and aft defocus), right ?

Jim Kasson has already performed a similar test but it's unclear to me if he either had a decent copy of the lens in that regard or if his methodology wasn't sensitive enough to display such behaviour : https://blog.kasson.com/nikon-z6-7/50-mm-f-2-nikkor-s-transition-point-spread-functions/

Jim's test should have shown the effect clearly, if it had been present in his sample of the lens. There is structure in Jim's images, but not the rather broad bright annulus shown in the PhotographyBlog image of light fittings.

Jim's images do show the expected contrast inversion between front and rear defocus, but since there are no features comparable in scale with the PB annulus, this may not be immediately obvious.  As far as I can tell, the composite image has been down-sampled and compressed, which makes a detailed comparison difficult.  Contrast-enhanced crops are shown below.

Two of Jim Kasson's bokeh disks. Camera front and rear defocused by approximately 0.77 mm for left and right images respectively. Up-scaled, de-saturated, and contrast increased. Right hand image rotated 180 degrees to compensate for inversion on opposite sides of the focus plane. https://blog.kasson.com/nikon-z6-7/better-nikon-50-1-8s-transition-point-spread-functions/

These images do show that most of the bright features in the background blur disk have corresponding dark features in the foreground blur, and vice-versa.  Note that I have increased contrast and de-saturated colours to emphasize this aspect.

The image-space defocus here of 0.77 mm is not too far from the 0.65 mm I estimated for the PB sample.

The lack of strong structure in Jim's images, and in many of the other review images, suggests to me that PB annulus represents sample variation, rather than a design feature.

-- hide signature --

Alan Robinson

Keyboard shortcuts:
FForum MMy threads