Normally, when I get a new lens, I try to test it corner to corner by shooting an horizon picture where the horizon is tilted corner to corner. Like this one:(http://farm4.staticflickr.com/3742/10388259676_84ecb6a700_z.jpg) (http://www.flickr.com/photos/jerome_munich/10388259676/)
I do that at various apertures so as to see the progression in sharpness. The test also gives information on AF accuracy and vignetting. Basically, it is a deceptively simple yet extremely accurate landscape test. I can also compare different lenses of the same focal between them (e.g. zoom versus prime, at different apertures, etc...).
Today, I tried to compare two different lenses mounted on two different cameras. One is the Hasselblad HC 50mm - II, mounted on the H3DII-50. The other is the Nikon 35mm f/1.4 G, mounted on the D800. The HC 50 - II is an excellent lens, but so is the Nikon (in any case it is the best I have that gives that field of view and mounts on the Nikon). The HC 50 - II is full open at f/3.5 (it does not get much sharper closed down...), the Nikon 35mm is closed at f/4.0 (and it only gets a bit sharper when closed down to f/11, but not that much).
To ease the comparison, I cropped the horizon and rotated the pictures, then put them on top of each other. The D800 picture is scaled up to match the resolution of the Hasselblad. You get this:(http://farm4.staticflickr.com/3713/10388618894_7bd71ea49d_b.jpg) (http://www.flickr.com/photos/jerome_munich/10388618894/)
Clic here (http://www.flickr.com/photos/jerome_munich/10388618894/sizes/o/in/photostream/) to see the full resolution (on flickr).
The top one is clearly better in terms of resolution & detail
... Nikon guys will choose the bottom image.
My personal preference is the top image.
That is because I shot MF.
The bottom photo is clearly a Nikon look. Nikon guys will choose the bottom image.
Evgeny
Well, I'm a Nikon guy, but the tip image is clearly sharper, especially in the corners.
Stitching is a PITA...
possibly investing in a top lens for the D800 would "save" the price of investing in the H body, for someone who wants these results.
If a d800 user needed the higher resolution they would stitch rather than uprez.
Stitching is a PITA...
Stitching is liberating ..., both for resolution and for composition.
Using a longer focal length allows to capture a higher resolution, and stitching will make up for the FOV. Stitching allows to compose more freely, not limited by the rectangular sensor dimensions, but rather by the framing that the subject requires.
What Jerome's example shows is that up-sampling will not benefit resolution, and that sharpness fall-off towards the corners can be an issue when producing large format output. A Raw converter like Capture One offers a sharpening correction for such sharpness fall-off. It could also be done with a duplicate layer that was deconvolution sharpened specifically for the corners, and a circular mask. Stitching would allow to down-sample to gain corner sharpness with a more natural transition across the image.
Cheers,
Bart
Would be interesting if Jerome see's weak corners on lenses that typically are very good across the field like short telephoto and macro lenses.
The test on the horizon is deceptively simple. It actually allows one to determine defects such as centering and, by comparison with a few lenses, lack of parallelism between sensor and mount on the camera. I tested a good dozen lenses on that D800 and even more on my Sony A900. A few on Canon as well. I can definitely say that sharpness corner to corner is much better for wide angle lenses on the Hasselblad H system than on any 24x36 system.
We all know that stitching will beat any MF or even LF camera resolution wise. Everybody has seen the huge Gigapan pictures.
So you see the problem on wide angles but not on longer lenses?
What surprised me somewhat was that it was not just a corner fallof on the Nikon but the degradation of the image was obvious over a large part of the field.
Yeah we should compare a stitched Nikon with stitched medium format....
The tests seems interesting, but it probably mostly shows the degree to which the lens delivers a flat field at infinity. Which is obviously just one characteristic of a lens that is only relevant for some applications.
The 35mm f1.4 was designed to deliver a look and good quality when shot near wide open, it is clearly not a landscape lens designed to deliver flat field at infinity. I am pretty sure that many cheaper Nikon zoom lenses with a less ambitious aperture offer a flatter field at infinity and will therefore perform better in this test.
I am sorry, but I don't think this is a good indicator of the degree of image quality achievable with a D800 vs a Hassy H4D50.
As far as stiching goes, yes it is possible to stitch with MF cameras... but what is the value? There was a time when the quality of the pixels in terms of DR was higher with MF cameras... but it is now well established that this isn't the case anymore. So stitching with a D800 and a Hassy H4D50 will deliver the same image quality, you will just need a bit more frames with the nikon... at 1/6th of the price.
Cheers,
Bernard
I would agree that f/1.4 lenses contain a lot of glass to achieve good performance at large aperture. At medium aperture less glass is probably preferable, for contrast but also because having fewer and smaller lens elements probably makes for more flexibility in positioning those lenses. The best lens I have for my Hasselblad V is the Sonnar 150/4, it contains just five element in three groups. It is very good, but have some color bookeh.
On the other hand, many lenses, also from Zeiss, seem to have rather weak corners.
The tests seems interesting, but it probably mostly shows the degree to which the lens delivers a flat field at infinity. Which is obviously just one characteristic of a lens that is only relevant for some applications.
I am pretty sure that many cheaper Nikon zoom lenses with a less ambitious aperture offer a flatter field at infinity and will therefore perform better in this test.
Yes: landscape photography. I thought landscape photography was the subject of this web site.
I am pretty sure that no other Nikon lens performs better at this particular test, since I have tested quite a few of them (the tests are on my flickr account). It is a very simple test, quite easy to set up. You are welcome to do your own test and prove me wrong.
Yes: landscape photography. I thought landscape photography was the subject of this web site.
The test you did is interesting and I think I agree with your observation, based on the limited number of lenses I use. It seems many 135 lenses cut corners in corner sharpness.
Can we at least measure at f7.1-f9? I know few people shooting landscape at f4.
Cheers,
Bernard
Nowadays at F9 you are approaching or into diffraction on something like the D800. You may flatten out your field and also get depth of field but the image will go soft.
Can we at least measure at f7.1-f9? I know few people shooting landscape at f4.
It is a simple test and it shows two simple, unsurprising facts:
-that 50 mpix is more than 34.
-that MF lenses are generally designed for higher sharpness corner to corner than lenses for the 24x36 format.
MF lenses are generally longer focal length, due to the requirement to cover a larger image circle, and therefore simpler to design/correct than a wider angle lens.
For a better visual comparison one would need to down-sample the larger pixel size dimension to a similar magnification factor and (vertical) field of view, which takes care of focal length and pixel dimension differences in one go.
They are of a longer focal length, but they are not simpler to design due to the requirements of covering a larger image circle. If you scale a lens and its sensor, you get exactly the same results.
You are welcome to do that yourself, but as I said once "if you design the test to reduce differences until the two cameras are equal on all measurable factors, you will find out that the cameras are indeed equal on all measurable factors". Part of the appeal of MF cameras is that they have more pixels.
Is a mistake to do such comparison with autofocus. Should be manual focused.
The AF system of the Hasselblad and the AF live view system of the Nikon are more accurate than manual focus.
The AF system of the Hasselblad and the AF live view system of the Nikon are more accurate than manual focus.
Weak corners is not really the point here. The point is the ability of the lens to render sharply a flat surface at infinity. Flat as opposed to spherical.This is likely correct, and it would be easy to test. Focus manually on the corner(s). Is the center then noticeably soft? If field flatness is the problem, then a possible solution is focus stacking, as an alternative or addition to horizontal/vertical stitching.
That is theoretically impossible.
It is a simple test and it shows two simple, unsurprising facts:I suspect it shows more that 50mm is a longer focal length than 35mm.
-that 50 mpix is more than 34 ....
I suspect it shows more that 50mm is a longer focal length than 35mm.
That is theoretically impossible.
... Why would that be?...
It is interesting that a confrontation with a Hasselblad conclusively demonstrates ... the rather mild shortcomings of the Nikon 35mm system owned by a hundred thousand consumers.
I had in mind that any auto focus system relies on certain tolerances, ie, the system considers it is sufficiently "in focus" when certain parameters are inside a certain range. In other words, if the subject distance is 10', then the system will stop focusing when it finds itself within +/- 10.x'
Manual focus with a live view and a 10x magnification, however, shall be more precise.
This is likely correct, and it would be easy to test. Focus manually on the corner(s). Is the center then noticeably soft? If field flatness is the problem, then a possible solution is focus stacking, as an alternative or addition to horizontal/vertical stitching.
This is quite an amusing way of seeing things... but you are right. Indeed modern digital 135 cameras allows us to take pictures of a quality high enough to be enlarged to huge sizes (at least for landscape pictures). To show these differences, I had to resort to a complicated procedure which is the equivalent of devising a special scene where the smaller camera will fail and enlarging the results to wall size.
I think its is a neat and simple test. Can't wait to see people use it on their "other" MF brands.
Edmund
I also cannot wait to see it applied to other focal lengths. I have a Zeiss 55mm f1.4 on order and would be more than happy to participate in a comparative test against the Hassy 80mm f2.8 that I used to own on my H1. ;)
As far as 35mm lenses, a retest with the Sigma 35mm f1.4 that seems to exhibit at lot less field curvature, while being twice cheaper, would seem fair. I'll fully support the test results then.
Besides, it would also be interesting to see to what extend DxO 8 can improve the performance of the DSLR lenses by applying automatically different levels of sharpening to different parts of the image depending on the local level of sharpness of the lens.
Cheers,
Bernard
I also cannot wait to see it applied to other focal lengths.
Last but not least, I am also not interested to do tests on order, just to please random photographers on the Internet. I won't go and buy the Sigma 35mm or Zeiss 50mm from my own money just to please you, while I have zero interest for these lenses.
Some RAWs for those who are interested
Base ISO, f6.3, focused on the middle (4th) building
IQ180/cambo WRS/Rodenstock 70HR
Nikon D800E/Leica 28-90 ASPH
No issues Jerome.
I am just pointing our the fact that your tests depicts a view of the D800's potential for landscape work that is, IMHO, undervaluing its potential based on a single data point.
I am just pointing our the fact that your tests depicts a view of the D800's potential for landscape work that is, IMHO, undervaluing its potential based on a single data point. I would not have reacted had you simply described it as a test of the field curvature of the Nikon 35mm f1.4 and its poor value as a landscape lens.
But you seemed to extrapolate this as a more generic assessment of the D800.
How about the request made to try refocusing the Nikkor in the corners to see whether you can achieve higher levels of resolution? Are you not interested in that test?
the Nikon is soft through the entire image.
the Nikon is soft through the entire image.
This is likely correct, and it would be easy to test. Focus manually on the corner(s). Is the center then noticeably soft? If field flatness is the problem, then a possible solution is focus stacking, as an alternative or addition to horizontal/vertical stitching.
I am extrapolating this test and the whole collection of lens test I already did to a more generic assessment on all 24x36 wide-angle lenses, actually. I am not limiting myself to the D800. The problem we see with that 35mm lens is more general, you will not find a 35mm with this kind of homogeneity on 24x36 cameras.
No, I am not interested in landscape pictures with sharp corners and out of focus center. Let me suggest something else: since you rave about the Sigma 35mm, what about you getting your D800 out in a field and taking the same kind of photos to show us how that particular lens works for landscape? It is a very simple test and anybody can do it.
There is no magic in bigger sensor, there is no magic in optics, just a very different marketing point.
The pictures have been taken with an HC 50-II. Can we find the equivalent lens for the D800 (or any other full frame camera)?
The HC 50-II:
-is 116mm long
-weights 975g
-is open at f/3.5
-uses 11 elements in its optical design
-costs €3555 without VAT
An equivalent lens for the D800 does not exist, but we can imagine one: we will suppose that we have a reducing machine which can make everything smaller in all dimensions. We make everything 40% smaller, so that we get a 35mm lens. What would that lens be? It would:
-be 82mm long
-weight 355g (the weight is divided by the cube of the distance, so 1.4^3)
-still be open at f/3.5 (aperture is dimensionless)
-still use 11 elements
-still cost €3555 (actually, maybe a bit less if we could benefit from economies of scale, but not less than, say, €2000-2500, because it is still a very complex design).
And there lies the difference. Actual lenses for the 35mm format are very different. You do not have an 11 elements f/3.5 in 35mm format at any price. Nobody would buy them. You have:
-either very fast lenses, so they are optimized for very different criterias
-or slower lenses, but they use relatively simple optical designs with 5-6 elements so are poorly corrected.
That Nikon 35/1.4 G seems quite good in Photozone's test but Photozone ignores field curvature in their tests.
....
And there lies the difference. Actual lenses for the 35mm format are very different. You do not have an 11 elements f/3.5 in 35mm format at any price. Nobody would buy them. You have:
-either very fast lenses, so they are optimized for very different criterias
-or slower lenses, but they use relatively simple optical designs with 5-6 elements so are poorly corrected....
What they do, as in most tests where they give nice curves as a result, is test at a relative close distance because the resolution target is not that big. My resolution target is 2 Km wide. Testing at a shorter distance will give different results, because of the way these lenses are designed (e.g. internal focus).
Moreover, resolution figures are only loosely related to the visual impression of sharpness, photozone gives only 3 figures without specifying where they are exactly taken (and we see from my test that the resolution changes continuously across the frame), etc... Testing lenses is surprisingly complex.
Hi,
50-100X the focal length is OK to test.
The problem with Photozone test is that they refocus the lens for center, border and corner. That means that field curvature is hidden. DPReview and DXO don't have that problem.