The Naked Truth About Bokeh

[Marianne Oelund]

Unfortunately, Zeiss glass is conspicuously absent from my collection of 50mm lenses, but many of you will find the models I have available for testing to be of interest, so here goes . . .

I am including the results for the 58G for comparison, followed by the Sigma 50 Art, Nikkor AF-S 50/1.4G, AF-S 50/1.8G, an old AI 50/1.8 and the PC-E 45 f/2.8:





The Sigma Art still does use under-corrected SA to soften background bokeh, but much less so than the 58G. The modest use of SA allows the foreground circles to have more interior brightness, which improves foreground bokeh, while still giving the background circles nicely softened edges. This is a nicely balanced design, in my opinion, and is clearly better than the other 50mm lenses shown below it. The only issue I might have, is that the foreground transition circles up to -20mm or -30mm are very ring-like with little interior brightness; in this range the 58G is doing better.

The Nikkor AF-S 50 f/1.4 is just weird. It somehow manages to have bright blur-circle rims in both foreground and background. Further, the foreground rings are not at the perimeter, but a little inside the circle, so they are still present at f/2. Bah! At least the background is looking better at f/2, but I cannot recommend ever using this lens at f/1.4.

The AF-S 50 f/1.8 is actually doing better than the f/1.4. It is putting more brightness into the interior of the foreground circles and doesn't have the pronounced rim brightness around the background circles. There is, however, significant asymmetry due to lens decentering - such is the kind of compromise one should expect with inexpensive optics.

The old manual-focus AI 50 f/1.8 seems to be the ancestor of the AF-S 50 f/1.4G, displaying the same inherent evils. As if that weren't bad enough, it also gives us a good dose of asymmetry and the heptagonal bokeh shape upon stopping down due to lack of curved aperture blades. Almost everything one could ask for, to produce the worst possible bokeh.

The PC-E 45 f/2.8 is nicely behaved, giving us foreground circles that are solidly filled in, and background circles with diffuse edges. Focal-plane sharpness is excellent. The only issue showing up here, is some asymmetry along the tilt axis; I will need to look into that and see whether the tilt centering detent is accurate.

--
Source credit: Prov 2:6
- Marianne

 

[Lcameron1951]

Marianne,

Thank you for this. The comparisons are very interesting to study.

Do you have a 50/1.2 AIS for testing? It would be interesting to see what occurs between 1.2 and 2.8.

And, of course, it would be interesting to see what the 105/135 DC lenses produce. Thanks again.

 

[Marianne Oelund]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring. The SA control is highly effective, and as a bonus, even in the DC-neutral position the bokeh quality is excellent; this is the only lens I've tested so far, capable of producing foreground and background blur circles which simultaneously lack any bright perimeter ring.

The only issue I've uncovered, is an asymmetry in my copy which appears to fade away if the DC ring is rotated well into the "Front" range. But if you overlook that, you can see how well the lens is able to mimic various designs.




A very broad palette to choose from, indeed.

With the DC ring at the Rear 4 position, the lens even does a respectable job of mimicking the 58/1.4G.

--
Source credit: Prov 2:6
- Marianne

 

[YRUNVS]

Many Thanks! I was keenly interested in how the DC performed and much appreciate your analysis.

 

[just Tony]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring. The SA control is highly effective, and as a bonus, even in the DC-neutral position the bokeh quality is excellent; this is the only lens I've tested so far, capable of producing foreground and background blur circles which simultaneously lack any bright perimeter ring.

The only issue I've uncovered, is an asymmetry in my copy which appears to fade away if the DC ring is rotated well into the "Front" range. But if you overlook that, you can see how well the lens is able to mimic various designs.


A very broad palette to choose from, indeed.

With the DC ring at the Rear 4 position, the lens even does a respectable job of mimicking the 58/1.4G.

You've been very productive and there's lots of creamy goodness on tap here.

Other than that, I'm very curious what the front side bokeh looks like in an actual scene, in the focus transition zone, at f/2 with DC = Rear 4.

 

[Rudolf Morf]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring. The SA control is highly effective, and as a bonus, even in the DC-neutral position the bokeh quality is excellent; this is the only lens I've tested so far, capable of producing foreground and background blur circles which simultaneously lack any bright perimeter ring.

With the DC ring at the Rear 4 position, the lens even does a respectable job of mimicking the 58/1.4G.

 

[MisterHairy]

I believe that Marianne mostly works with cameras which do not correct for focus shift (this is true for most Nikons to date) so her results are correct for her (and many others') usage.

Yours are representative of only the most very recent generation of Nikon cameras.

Do you know which cameras this includes?

The D810 seems to have firmware correction for focus shift in modern Nikon AF lenses. Dxxx FX cameras prior to that do not seem to. I do not know about the D750.

The D5 seems to and I believe that the D4s may (I have never used one) but prior Dx cameras do not seem to.

I don't know about any DX cameras but would assume that the D500 does.

 

[jtra]

Rudolf Morf wrote:
Do you also have access to a 100mm/f2 DC lens?

I have tested 105 DC in similar way in my page on bokeh.
http://jtra.cz/stuff/essays/bokeh/

Though I rather use test pattern on monitor: http://jtra.cz/stuff/essays/bokeh/bokeh-test-lines.png captured at angle which gets me information about bokeh in single picture. For 105 DC I have some outdoor samples that include various DC ring positions and include Sigma 85/1.4 for comparison.

I also have nice visual explanations how spherical aberration (SA) changes light distribution in foreground vs. background bokeh.

It might also be interesting to study the blur for off-center geoemtries. I would be most interested in the behavior of the 58g in this respect. Does the desired blur property in the background persist for off-center positions. Does it degrade and if so, how does it degrade for larger offsets?

That is a great question. I have seen lenses usually have fairly nice bokeh in center and somewhat worse bokeh in corners, but not just due to mechanical vignetting. For example my Sigma 50/1.4 non-art shows very nice background bokeh in center, even in DX area, but in sides the bokeh is triangular and it has pronounced edges.

Another thing is that SA is distance dependent. So for example my Sigma 85/1.4 has fairly neutral bokeh at close distance around 3m and edgy background bokeh at around 10m.

I think the apodization is superior method to getting smooth bokeh, but it has some disadvantages (usually no PDAF - I don't know why when AF system is using f/5.6 obstruction where apodization usually does not have much effect).
When apodization is weak (can be seen on lens diagram by looking at profile of gray element) and lens has overcorrected SA the endresult for background bokeh near focus plane is not optimal. You can see it in this duel of Sony 135 STF and Laowa 105 STF: http://phillipreeve.net/blog/stf-duel-laowa-2105-vs-sony-2-8135/
where Laowa has weak apodization and overcorrected SA. Still an interesting lens from Laowa.

 

[Stacey_K]

Isn't that what we are really looking at here, given we are looking at what happens in the first 60mm from the focus plane?

I'm also curious the focus distance used for this test.

Please I am -not- saying anything negative, this is really cool what you posted and helps me understand some of what I see when shooting with the 58 vs other lenses.

I'd also love to see this repeated with some 85mm lenses, specifically see how the old 85 1.4D compares to the new G lens.

--
Stacey

 

[gabriel foto]

Unfortunately, Zeiss glass is conspicuously absent from my collection of 50mm lenses, but many of you will find the models I have available for testing to be of interest, so here goes . . .

I am including the results for the 58G for comparison, followed by the Sigma 50 Art, Nikkor AF-S 50/1.4G, AF-S 50/1.8G, an old AI 50/1.8 and the PC-E 45 f/2.8

Thank you Marianne for a brilliant (and somewhat new) take on lens comparison! This may open doors to understanding a lot more about the general rendering of a lens.

If I understand your tests correctly, they show a pinpoint light source focused, I guess, at a few metres (?) and then with the focus varied between 60 mm in front of and behind the focal plane. This could perhaps, in real life, be compared to for example a portrait of a person, focused on the closest eye, with other parts of the face out of focus.

It would be interesting to know more about how your findings would look if the out-of-focus zone is expanded a lot further, in particular behind the focus plane. Thinking again about a portrait at relatively close range, many such subjects have a background which is completely out of focus, many metres away or even at infinity.

Also, I believe there may be significant differences for any lens depending on the focus distance used in testing, especially perhaps for lenses with floating elements.

(To you optical linguists: I am not sure about the terminology here but the first, small incremental changes in rendering, just in front of and behind the subject in focus could perhaps be the out-of-focus transition, and the character of the far away blur could be called the bokeh? Or would you call this bokeh, all aspects of it?)

I have often wondered how these different parameters interact:
* Field curvature and bokeh
* Focus distance and bokeh
* Chosen aperture and bokeh
* Vignetting and bokeh
* How DOF differs over the image field
* How out-of-focus transitions differ over the image field
* How the bokeh character differs over the image field
* What causes 'swirly bokeh', apart from perhaps astigmatism
* How to focus, for max sharpness or max contrast
* How to compensate or not compensate for focus shift

Marianne, don't you think that your way of testing may deserve being developed into a new standard for lens testing, alongside MTF measurements and the like?

And meanwhile, if you were to describe exactly how your tests are performed, could this method be used in order to make a sort of amateur standard by which members of the thread could add other lenses tested in the same way?

Gabriel

 

[ruifonikon]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring. The SA control is highly effective, and as a bonus, even in the DC-neutral position the bokeh quality is excellent; this is the only lens I've tested so far, capable of producing foreground and background blur circles which simultaneously lack any bright perimeter ring.

The only issue I've uncovered, is an asymmetry in my copy which appears to fade away if the DC ring is rotated well into the "Front" range. But if you overlook that, you can see how well the lens is able to mimic various designs.


A very broad palette to choose from, indeed.

With the DC ring at the Rear 4 position, the lens even does a respectable job of mimicking the 58/1.4G.

--
Source credit: Prov 2:6
- Marianne

This is really good to see. Congratulations and thank you.

Have you ever done the same with the 85mm f/1.4D? I would love to see it side by side with the 135mm f/2DC.



--
________________________________________
Flicker page:

ruifo

Explore ruifo’s 2,742 photos on Flickr!

www.flickr.com

Flickriver: ruifo's most interesting photos

Flickriver - view images as a 'river of photos' and more...

www.flickriver.com

 

[Bernard Delley]

out of focus point spread functions from row of single green pixels spaced 10mm on macBook retina screen. Screen is fully tilted back (45deg) to shorten long side of illuminated area. Image is rotated clockwise to make foreground appear left and background right. reduction ratio to sensor is about 1:15

 

[Marianne Oelund]

Isn't that what we are really looking at here, given we are looking at what happens in the first 60mm from the focus plane?

Yes, it is within the transition zone where we see all of the interesting differences between the lens designs with respect to SA. It extends further than the +-60mm I used in my tests, but that's the limit of my macro stage. I should also mention that for moderate subject distances, the transition zone can extend all the way back to infinity.

I'll be talking about behavior outside of the transition zone later, where blur circles flatten in their intensity profile, and bokeh quality comparisons become dominated by factors such as aperture blade shape, color fringing and exit-pupil eclipsing.

I'm also curious the focus distance used for this test.

I should have mentioned this in the opening post: Most of the tests have been done at 0.9m to 1.2m distance, except for the opening comparison where I had the 58G at 1m and the 85/1.4G at 1.4m to equalize perspectives.

Please I am -not- saying anything negative, this is really cool what you posted and helps me understand some of what I see when shooting with the 58 vs other lenses.

I'd also love to see this repeated with some 85mm lenses, specifically see how the old 85 1.4D compares to the new G lens.

I only have the 85/1.4G and 85/1.8G, and will post a comparison of those. I'd be happy to include an 85/1.4D or others, if anyone is interested in loaning theirs out.

--
Source credit: Prov 2:6
- Marianne

 

[ruifonikon]

Trying to mimic your experiment with these three lenses (still quite far from your rigor, though): Nikkor 85mm f/1.4D + Nikkor 105mm f/2.5 AI-S + Sigma 105mm f/2.8 Macro



--
________________________________________
Flicker page:
https://www.flickr.com/photos/ruimc/
http://www.flickriver.com/photos/ruimc/popular-interesting/

 

[Stacey_K]

Isn't that what we are really looking at here, given we are looking at what happens in the first 60mm from the focus plane?

Yes, it is within the transition zone where we see all of the interesting differences between the lens designs with respect to SA. It extends further than the +-60mm I used in my tests, but that's the limit of my macro stage. I should also mention that for moderate subject distances, the transition zone can extend all the way back to infinity.

I'll be talking about behavior outside of the transition zone later, where blur circles flatten in their intensity profile, and bokeh quality comparisons become dominated by factors such as aperture blade shape, color fringing and exit-pupil eclipsing.

This is very cool stuff. As a bokeh addict, glad to see some actual testing that shows the differences in a repeatable fashion

 

[Stacey_K]

Thanks for these, I love the look I get from my 85mm f1.4D and this shows some of what is going on optically.

 

[Marianne Oelund]

Trying to mimic your experiment with these three lenses (still quite far from your rigor, though): Nikkor 85mm f/1.4D + Nikkor 105mm f/2.5 AI-S + Sigma 105mm f/2.8 Macro

Thanks for the effort.

As a general principle, the dot(s) that you use as the light source needs to be smaller than the finest details you are trying to see in the bokeh circle. Try using just a 2x2 pixel spot, and longer exposures.



--
Source credit: Prov 2:6
- Marianne

 

[Marianne Oelund]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring.

Thought I'd try my hand at mimicking the 58G at f/1.4. I found I couldn't dial in as much Rear DC as I wanted, as there was too much sharpness loss. I ended up using Rear 2.8 instead, which doesn't produce quite as much of the 58G's "aura," but otherwise it looks close to me.

What do you think of this first attempt?






Attempt at duplicating the 58G's wide-open bokeh, using the 135 f/2 DC



--
Source credit: Prov 2:6
- Marianne

 

[Lcameron1951]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring.

Thought I'd try my hand at mimicking the 58G at f/1.4. I found I couldn't dial in as much Rear DC as I wanted, as there was too much sharpness loss. I ended up using Rear 2.8 instead, which doesn't produce quite as much of the 58G's "aura," but otherwise it looks close to me.

What do you think of this first attempt?


Attempt at duplicating the 58G's wide-open bokeh, using the 135 f/2 DC

--
Source credit: Prov 2:6
- Marianne

Interesting comparison. It would also be interesting to see the 58 shot at 2.5 as well. The contrast difference is more than I would have expected.

 

[Bernard Delley]

I must say I'm very pleased with the way Defocus Control works on the 135 f/2 DC. It is like having a lens which you can re-design on the spot by twisting a ring.

Thought I'd try my hand at mimicking the 58G at f/1.4. I found I couldn't dial in as much Rear DC as I wanted, as there was too much sharpness loss. I ended up using Rear 2.8 instead, which doesn't produce quite as much of the 58G's "aura," but otherwise it looks close to me.

What do you think of this first attempt?


Attempt at duplicating the 58G's wide-open bokeh, using the 135 f/2 DC

interesting, that the 135mm DC background bokeh stays much closer to neutral color in the rail comparison than the 58G, which gets greenish.