Separate names with a comma.
Discussion in 'Native Lenses' started by Halaking, Dec 4, 2015.
Measurements: Lower than expected sharpness | DxOMark
Measurements were available even when $99 sale started - I remember checking it.
That's surprising, because that means that the Oly 25mm/1.8 is also unsharp, because someone compared the two lenses on this forum and the results are near enough to identical. But that doesn't make sense either, because people draw the same conclusion when comparing the Oly 25 to the PL25 in sharpness.
So if A=B, and B=C, then...see what I'm getting at here?
My theory is sample variation.
The only oly vs panny 1.7 comparison i saw was on tree foliage, which is a really hard thing to use for sharpness. Was there another one?
Yet the ePhotzine review (with measurements) says :
"Sharpness is quite another matter and the centre at f/1.7 is already very good, rising to excellent levels by f/4 and maintaining this through to f/8. This falls very slightly to very good levels at f/11, is still good at f/16 but only fair at f/22.
The edges maintain very good levels of sharpness from f/1.7 all the way through to f/11, only dropping to good levels at f/16 and fair levels at f/22. The edge values are commendably close to the centre values throughout and the performance is quite even.
This is a very good level of performance overall, and images should be bright, sharp and attractive throughout the range."
And Photoreview say:
The review lens showed itself to be a stellar performer in the Imatest tests we carried out with the GX8 camera body. JPEG files from this camera came very close to meeting expectations for its 20-megapixel sensor at f/1.7, met expectations at f/2.2 and remained above expectations through to f/8, after which diffraction kicked in.
The highest resolution was measured at f/5, where it was almost 20% above expectations, which is excellent for JPEG files. Edge and corner softening were slight throughout the aperture range and became insignificant from f/5.6 onwards, as shown in the graph of our test results below.
Lateral chromatic aberration was well controlled and effectively negligible at all but the widest lens apertures. The graph below shows the result of our tests with the red line marking the border between 'negligible' and 'low' CA.
We found some vignetting in raw files from the GX8 at the widest aperture settings. It was also present in JPEGs captured with the camera's Shading Compensation switched off. The effect was most noticeable at f/1.7 but barely visible by f/2.5. Switching on the Shading Compensation eliminates it from JPEGs; raw files are easily corrected in conversion software.
Distortion was effectively negligible in both uncorrected JPEGs and raw files. Here, too, in-camera corrections make this aberration a very minor concern.
Because of its wide maximum aperture, the quality of out-of-focus blurring is an important criterion when assessing this lens. When backgrounds were evenly lit, bokeh was very smooth and attractive. However, we found some choppiness and slight outlining when backgrounds contained bright highlights. Examples are shown below.
The review lens was relatively flare-resistant, retaining a high percentage of both colour depth and contrast in scenes when a bright light was included within the frame.
And I say: "Typical DxO"
You may be right about the foliage, so I guess more testing remains to be seen.
The other reason to be skeptical is that the 25/1.4 is not sharper than the 20mm/1.7. The reverse is true, at least according to LensRental's exhaustive testing which I consider far more credible. They tested 6 copies of the lens on two sets of bodies. Given that Dxo claims the "New" 20mm v. II is less sharp than the original despite being completely identical optically also would support my theory of sample variation...
LensRentals.com - Wide-Angle Micro 4/3 Imatest Results
Until there's a side by side comparison we probably couldn't be sure.
But from my experience, PL25 1.4 may seem soft wide open, it sure is very sharp from 2.8 and onwards which is quite hard to beat, also consider the coating to be superior.
So I wouldn't be surprised to find the PL25 1.4 better in every aspect, though the 1.7 isn't bad.
In my real world opinion, the 20mm 1.7 is sharpest wide open, followed by the 25mm 1.7 and lastly the 25mm 1.4 lens. Based on me shooting all 3 wide open at close distances.
I wonder what the DXO results would be if they used camera bodies newer than 2013.
Very little unless they use the 20mp GX8. All the 16mp sensor cameras will be in the same general range. (The differences between 16 mp cameras regardless as to whether or not they have an anti-alias filter has very little a impact on visible differences.)
While I agree with you in principle. They did test the lens with a number of bodies (99% of them 16mp) and did end up with varying results. So I would say the internal processors for the sensors have something to do with it and since it is a brand new lens I would like to see it tested on the latest Olympus and Panasonic bodies.
DXOMark has never been about "visible differences," though.
The only difference between the D800 and the D800E is the removal of the anti-aliasing filter, and according to DXO the Nikon 70-200 f4 gets 21MP resolution on the D800 and 30MP resolution on the D800E. So it appears that the removal of the anti-aliasing filter can in fact make a huge difference (+42% resolution!)
DXO does not list resolution numbers. Their mp numbers are some sort of Effective Megapixels.There's no way of knowing what the actual difference in resolution is. If you want to know what the resolution is, look up test results on sites that report in units of resolution.
One of the more important things to remember is that a given lens on a give body yields system resolution. The only way to isolate the lens resolution performance is to either test the lens on an optical, where no body is used, or to always use the same body. Only Lens Rentals uses an optical bench and they don't seem to have an adapter for µ4/3 lenses. SLRgear uses a GX1 for their µ4/3 testing, and that has a 16mp sensor, so if that gives you a warm fuzzy, use that site. Lenstip uses an E-P1, because it doesn't massage RAW files, and even with a 12mp sensor, differences between very high resolution lenses are observable.
Useful, valid testing only changes one variable. If one wants to know which lens is sharper, then the test results should only be influenced by the characteristics of the two lenses and not the body. For instance, if one wanted to know if the new P25/1.7 was sharper the the P20/1.7, then the test results would have be obtained from the same model camera. The only way to look up, useful test results to compare would is if a test site used the same camera for both lens tests. Knowing that a lens gives 4 lp/mm higher resolution on a 2015 body than a 2009 body doesn't give any useful information about the sharpness of a lens. If lens A is sharper than lens B on a 2009 body, then the same will be true on a 2015 body. Beyond that it's a meaningless exercise in How Many Angles Can Dance on the Head of a Pin.
DXO's Perceptual Megapixels are basically a 2-dimensionalization of the MTF values, which are only a linear value. They used to report lp/mm years ago before they developed it. But yes, I agree that DXO's methodology is not transparent.
I also agree with you that the only measurements that we can find are going to be system measurements. Most test sites (Lenstip, Photozone, SLRGear, etc...) maintain their test body consistent for exactly this reason, which is also why they don't like you to compare between systems because they might be testing their lenses on a 12MP camera with a heavy AA-filter, whereas a 16 or 20MP AA-filterless sensor is avaiable to give "better" results but ones that are not directly comparable. But ultimately, for us photographers, the system measurements are the only thing that matters. I don't think anyone should care if they have an incredibly sharp lens but only have low resolution sensors to take advantage of it. A lens that works very well on a 12MP sensor may or may not show the same performance advantages on a 20MP sensor. It's hard to tell where the lens will run out of steam before actually testing it.
So really, it would be nice if we could see how all the different lenses perform on all the different bodies, as that would at least give us a way to compare between different systems in a practical, apples-to-apples way. Testing all on the same body is the most useful for comparison for users of a single system, but it would still be preferable to test all the lenses on the highest resolution body that's available. But obviously time and practical constraints dictate that's not really possible for most operations...
The difference in resolution between two lenses will be greater on a 20mp camera than a 12mp camera, since a 20 mp sensor has a higher pixel density/spatial frequency. If there is a just barely visible difference of 10% with a 12mp camera, then the difference on a 20mp camera would be 28%. (the difference in the long side of a 12mp to 20mp sensor is 4032 to 5184 pixels, an increase of 28%)
I don't really care what DxO think. What do the real pictures look like?