01-06-2015, 07:16 PM
@JoJu
Take this rim plot:
[sharedmedia=gallery:images:1315]
It is almost exclusively spherical aberration. There are some minor polychromatic aberrations, but what is driving the 500-micron spot size is spherical. If you introduce a componenet behind this that has the same spherical, but with a negative sign instead of a positive sign, you get a perfect system.
Unless Lenscore's optic before the sensor is diffraction limited at a bigger aperture than the test lens, the aberrations it contributes will be significant. The sign and magnitude of aberrations in that optic, then, becomes very important. If you have a lens such as the fuji 56/1.2 or the panaleica 15mm f/1.7 which undercorrect the spherical aberration (most lenses do, these are mere examples) and the optic inside their camera overcorrects spherical, you have a net reduction in the spherical which will boost the contrast (and to some extent resolution) of the result image. Similarly, if you take a lens such as the zeiss 100/2 which overcorrects the spherical aberration and have that optic in the camera also overcorrect the spherical aberration, the result is worse contrast and worse resolution.
It is in this way that their test is flawed, but chances are the optic inside the camera doesn't contribute too too significantly to the overall result. It would only be when comparing diffraction limited or nearly diffraction limited systems that the aberrations of that optic would contribute in an overly significant manner. Of course, those lenses (such as canon's 300/2.8) are very rare, but then perhaps fans of one manufacture or another would use this particular cases to show the merit of their favorite manufacture over another, using such "ideal" data as Lenscore's data to split the hairs.
Interferometric testing would be a very expensive and "wasted" effort on photographic lenses. One must have an expensive reference sphere matched to the focal length being tested - to test the full range of photographic lenses one would need dozens of reference spheres, each costing between 5 and 15 thousand USD - not to mention the often months-long wait time for one to arrive after ordering. This is on top of the purchase of a $100,000-$500,000 interferometer, as well as a couple thousand for the vibration isolation table things should sit on, a few more thousand for the staging equipment to hold everything... You see where this is going vs perhaps $10,000 for a full imatest setup.
A very real argument can be made for time as well - but alas if you want the "ultimate" lens test and to test the lens without influence of any other anything... interferometry would be that test. In a nutshell, all of the optics behind the test optic are removed computationally. What an interferometer does is take the reflection (about 2%) off of the final reference surface and interfere it with the beam from the test surface. Because the difference is what is measured, you isolate the contribution from the test surface entirely.
[sharedmedia=gallery:images:1320]
While the canon equivalent design looks like this:
[sharedmedia=gallery:images:1319]
If we begin with first order analysis (the arrangement of lenses which focus and those which diverge) both are the same - they are telephoto lenses in a P-N arrangement wher P stands for positive and N stands for negative.
But, the construction of the that "P" is different. Canon has opted for a positive lens followed by a cemented achromatic doublet the formation of I don't know of the top of my head, followed by a gauss doublet, followed by another likely achromatic doublet of the same form as the one in front of the gauss doublet. Nikon has opted for a flat protective front element followed by a large positive lens, then an achromatic doublet, then much further down the barrel an extreme steinheil doublet to form the "P" group.
The designs are very different!
Of course one may improve it after stealing it, but they were not good enough to come up with it on their own to begin with...
Before the 55mm otus, not a single manufacture had ever released such a long focal length retrofocus lens for the full-frame sensor size. The otus was being shown cut in half for 1.5-2 years before its release which is more than long enough to produce an "original" copy of that design methodology for that focal length / aperture / angle of view combination. I do not believe for a moment that of all the camera lens designers in the world, the ones at sigma were the first in history to come up with a design methodology implementation in parallel to the engineers at zeiss...
Not defining the limits of your weather sealing is a method of protecting yourself from lawsuit. It is not ideal, but these are not cheap devices to manufacture and having to replace them when people use them outside spec but claim they used them within spec and fried them could be quite expensive.
Regarding "giving nikon a reason to improve" - R&D is very expensive. Sigma, as well as the mirrorless manufactures, have chosen to burn through their bank accounts with massive overspending in R&D while nikon and canon have been more conservative. Arguments can be made for the merits of both business strategies, neither is particularly superior. Nikon has been expanding into alternative technologies, namely the use of fresnel lenses, electronic diaphragms, and fluorine coatings lately. All very worthwhile, but I do question whether they should also be looking into MRF polishing for 5" class optics to further lighten their telephoto lenses, and of course MRF could be used to make more and better aspheres for smaller lenses as well. If they could polish one of the ED elements in the "P" section of the 400/2.8 into an asphere, for example, they could take out one of the large front elements which drive weight, bringing the lens weight down perhaps 500g+ instantly.
Quote:Certainly, you've been here for quite a long time, it is clear you have no affiliation with them.
I don't want to defend LenScore, I just appreciate them as "another source" of collected data.
Quote:My criticisms of their process are perhaps strongly worded, but they are marketing themselves as "the ultimate definitive best-ever lens test which places all lenses on equal footing" and so on. The trouble is, their test is not putting everything on precisely equal footing.
If their process is actually that weak and full of flaws like it appears to be in your reply
Take this rim plot:
[sharedmedia=gallery:images:1315]
It is almost exclusively spherical aberration. There are some minor polychromatic aberrations, but what is driving the 500-micron spot size is spherical. If you introduce a componenet behind this that has the same spherical, but with a negative sign instead of a positive sign, you get a perfect system.
Unless Lenscore's optic before the sensor is diffraction limited at a bigger aperture than the test lens, the aberrations it contributes will be significant. The sign and magnitude of aberrations in that optic, then, becomes very important. If you have a lens such as the fuji 56/1.2 or the panaleica 15mm f/1.7 which undercorrect the spherical aberration (most lenses do, these are mere examples) and the optic inside their camera overcorrects spherical, you have a net reduction in the spherical which will boost the contrast (and to some extent resolution) of the result image. Similarly, if you take a lens such as the zeiss 100/2 which overcorrects the spherical aberration and have that optic in the camera also overcorrect the spherical aberration, the result is worse contrast and worse resolution.
It is in this way that their test is flawed, but chances are the optic inside the camera doesn't contribute too too significantly to the overall result. It would only be when comparing diffraction limited or nearly diffraction limited systems that the aberrations of that optic would contribute in an overly significant manner. Of course, those lenses (such as canon's 300/2.8) are very rare, but then perhaps fans of one manufacture or another would use this particular cases to show the merit of their favorite manufacture over another, using such "ideal" data as Lenscore's data to split the hairs.
Interferometric testing would be a very expensive and "wasted" effort on photographic lenses. One must have an expensive reference sphere matched to the focal length being tested - to test the full range of photographic lenses one would need dozens of reference spheres, each costing between 5 and 15 thousand USD - not to mention the often months-long wait time for one to arrive after ordering. This is on top of the purchase of a $100,000-$500,000 interferometer, as well as a couple thousand for the vibration isolation table things should sit on, a few more thousand for the staging equipment to hold everything... You see where this is going vs perhaps $10,000 for a full imatest setup.
A very real argument can be made for time as well - but alas if you want the "ultimate" lens test and to test the lens without influence of any other anything... interferometry would be that test. In a nutshell, all of the optics behind the test optic are removed computationally. What an interferometer does is take the reflection (about 2%) off of the final reference surface and interfere it with the beam from the test surface. Because the difference is what is measured, you isolate the contribution from the test surface entirely.
Quote:Oh no - I meant everything they publish. They claim it takes about 2 hours to perform each of their tests yet when new lenses or bodies are released within a couple of days results for most compatible lenses/bodies magically appear. Ignoring the time constraints (this would require something like 80-150% of the time available since the announcement of the body/lens...) there is also a question of inventory - does DxO have an inventory of hundreds of thousands of dollars of photographic equipment lying around? Possible, but I still doubt it. Often very old and unpopular lenses appear for brand new bodies, one must doubt that the test was truly done rather than just mathematically simulated...
And right, DxO mark doesn't test all body-lens combinations.
Quote:Of course, but the D3x test camera features an AA filter, as does the 5D2. There will be some difference in the strength of the filter but this won't account for more than a 5-10% error at most. By shifting the 5D2 data so the "mean value" matches you in essence "align" the results so that the contribution of the AA filter is more likely to at most affect things that 5-10%, rather than a 5D2 with a stronger AA filter (-10% result resolution) suddenly making the 5D2 result 19% worse than the same lens on the D3x.
A last bit about your simulation of how Canon and Nikon tests could be interpolated by MP count: Canon uses AA filter, Nikon D750 doesn't - that IS increasing the resolution, but less than 10% (D800 / D800E being the only camera types to be comparable in this aspect)
Quote:I could argue for days about different designs and who 'stole' from who, but the historical designs - such as the sonnar, tessar, double-gausse, ernostar, biotar, heliar, Landscape Lens, bertele's symmetrical wide-angle lenses, Angenieux's retrofocus, etc, are very very old - and they are "requirements" - truly certain combinations of f number and angular coverage cannot be achieved without some of these designs. Most modern lenses are "original designs" - take for example the 400mm f/2.8 lenses from canon and nikon, nikon's newest looks like this:
Well, talking about "stealing" is talking about Japanese (and lately Chinese and Korean) manufacturers who stole a lot knowledge from German manufacturers and after that protected and improved their copy work.
[sharedmedia=gallery:images:1320]
While the canon equivalent design looks like this:
[sharedmedia=gallery:images:1319]
If we begin with first order analysis (the arrangement of lenses which focus and those which diverge) both are the same - they are telephoto lenses in a P-N arrangement wher P stands for positive and N stands for negative.
But, the construction of the that "P" is different. Canon has opted for a positive lens followed by a cemented achromatic doublet the formation of I don't know of the top of my head, followed by a gauss doublet, followed by another likely achromatic doublet of the same form as the one in front of the gauss doublet. Nikon has opted for a flat protective front element followed by a large positive lens, then an achromatic doublet, then much further down the barrel an extreme steinheil doublet to form the "P" group.
The designs are very different!
Quote:I don't mean the protocol - the ring USM motor design, mechanically, was copied to produce "HSM."
In Sigma's case: They improved Nikon's VR quite a bit. I can't judge if they improved also the focus drives but I know they do reverse design - how else could they come to equal AF qualities, at least in terms of accuracy?
Of course one may improve it after stealing it, but they were not good enough to come up with it on their own to begin with...
Quote:Retrofocus is a 'nice' design form to work with, sort of. There are staunch limits on the length of the system, but it effectively works by having a very short/quick focus of light in the front of the lens, then that is spread out so it moves straight backwards in a collimated fashion, and finally the light is focused to the sensor. The entire space between the front focus and the back focus groups is free for correcting aberrations, creating a focusing group, image stabilization group...
I'm just saying Sigma's not only copying, I see improvements in lots of aspects and after all, coming so close to an Otus at ¼ of the price plus implementing an (as Zeiss tries to make it look impossible because of the lack of accuray) AF is a performance of it's own.
Before the 55mm otus, not a single manufacture had ever released such a long focal length retrofocus lens for the full-frame sensor size. The otus was being shown cut in half for 1.5-2 years before its release which is more than long enough to produce an "original" copy of that design methodology for that focal length / aperture / angle of view combination. I do not believe for a moment that of all the camera lens designers in the world, the ones at sigma were the first in history to come up with a design methodology implementation in parallel to the engineers at zeiss...
Quote:It is unique, but it is also a bit of a wash. There is no point testing the lenses on a layered sensor design as it is no more sensitive than "flat" sensors. You also introduce a great deal of complication regarding chromatic aberrations by taking that approach. It's also an APS-C sensor to test full-frame optics.
The lens testing with Foveon sensors is unique and each lens is tested
Not defining the limits of your weather sealing is a method of protecting yourself from lawsuit. It is not ideal, but these are not cheap devices to manufacture and having to replace them when people use them outside spec but claim they used them within spec and fried them could be quite expensive.
Regarding "giving nikon a reason to improve" - R&D is very expensive. Sigma, as well as the mirrorless manufactures, have chosen to burn through their bank accounts with massive overspending in R&D while nikon and canon have been more conservative. Arguments can be made for the merits of both business strategies, neither is particularly superior. Nikon has been expanding into alternative technologies, namely the use of fresnel lenses, electronic diaphragms, and fluorine coatings lately. All very worthwhile, but I do question whether they should also be looking into MRF polishing for 5" class optics to further lighten their telephoto lenses, and of course MRF could be used to make more and better aspheres for smaller lenses as well. If they could polish one of the ED elements in the "P" section of the 400/2.8 into an asphere, for example, they could take out one of the large front elements which drive weight, bringing the lens weight down perhaps 500g+ instantly.