edwardkaraa wrote:
Thanks for the test photos. But what they really show is that it's impossible to have the DOF cover the entire frame at f/2.8. If you take a photo where both center and extreme corners are at infinity, or close to, you would be surprised of how good the corners are even at f/2.8.
I will try that out at some point. I wonder where infinity starts with that lens in shooting reality. The HfD (sorry, again ;-)) for 2,8 is 7,4 m and gives a near limit of 3,6m. That is not far away for "infinity" .
I was already several times pleasantly surprized by the lenses abilities at 2,8. The focus fades nicely at the near end, and if you don't have an object of interest right there, this is alright for me.
Ok, I had to know. So I´ve just performed a startest with my 21ZE and my 25ZF+Haoda adapter. A startest is the ultimate test of sharpness at true infinity, as the stars are pinpoint light sources. I use Polaris, the polar star as the center star. This gives the smallest problems with star trails when the universe turns around me. There are other potential problems as glare, atmospheric turbulence, far sub zero temps, snow up to the jewels, disturbing lights in the surroundings. But no quake, no tsunami, no nuclear core melt so I Do Not Complain. My thoughts go to our japanese sisters and brothers. Totally horrifying images from the devastated areas.
I wont post any images as they lack all sex appeal. But I will tell you this: My 25 is as sharp as my 21 over the FF of the 5D2 from about f/4. The 21 has a _slight_ edge at wide open. Ok all Photolenstestguruzones out there, get your act together, tell the truth about your limited tests, or shut down the sh!t.
And once again I learn something new and wonderful. Thank you all! Stay safe!
edwardkaraa wrote:
Why is it so difficult to believe that the 25 may be sharper than the 21. It's not like many people own both and have done extensive comparisons. I have no problem believing that, until proof of the contrary
Guys, these are Zeiss lenses we are talking about so that question is easy enough to answer as we have actual measured MTF charts.
At f/2.8, the 21 obliterates the 25. At f/5.6 things even out and on parts of the frame the 25 is actually slightly sharper than the 21. On other parts the 21 is slightly sharper. The only significant difference is in the extreme corners (around 20mm) where the 21 has an edge and the lower tangential curve of the 25 at 40 lpm which is significantly weaker. Simply translated the 21 will produce noticeably better tangential fine detail.
These are not surprising results. The 21 has two tricks up its hat - one is an astonishing performance wide open and the other is the reproduction of fine detail in an image.
With respect to the ZF 21, ZF 25, ZF 28, I rented all 3, and bought one, for now. With the ZF 21, it is marvelous how much detail it captures, and the colors. But for me, while I was shooting with it, I was struggling to keep verticals from converging. For the compositions I like to make, I prefer the elements to be more natural looking, where the lens used is not so conspicuous. Not a knock against the lens so much as the focal length. I compared the 25 and 28 pretty closely, & even though they are pretty close in focal length, and their wide angle distortions are somewhat similar, I bought the 28 because I liked its signature better. Also, at f8, I can get everything from a few feet in front of me to the tree line on the horizon in focus. And the trees in the horizon "pop". Never got that with my 28 (f2 or f1.4) AI-S. Also, I was surprised at how natural the image from the 25 looks. So, I may still buy the 25, but since rumor has it that the new 25/2? will be more like the 21/2.8, I'm waiting to see what it offers. If its only OK (doubtful), I may get the current 25/2.8 - if only to use it to shoot into the sun for the dramatic white out and shadow effects.
Actually, the lower MTF values somewhere beyond 20mm can be seen on the star test images. As we learned from the two Zeiss papers on MTF´s, a lens does not necessarily produce points from light point sources (e.g. stars). Both lenses show nice "gull wing"-shaped stars at the extreme edges. As these are far wider tangentially than sagittally, they will blur the tangential more = lower tangential MTF.
However, when the lens is correctly focused these patterns are so small that they are almost totally hidden by the 21Mp FF sensor. Result: A lens that is perfect all over the frame! As I, of course, managed to miss focus on a few images, the points are spread over a larger area = several pixels, and can be clearly seen. Never underestimate the educational benefits of a juicy blunder!
According to specs the 5DII handles up to 70 lp/mm, so the 40 lp/mm of the bottom pair of lines are definitely resolvable. The best crop cameras handle even more - for instance the 7D resolves over 100 lp/mm. So all the lines on the MTF charts will have a visible impact when using modern DSLR sensors.
Z250SA wrote:
Actually, the lower MTF values somewhere beyond 20mm can be seen on the star test images. As we learned from the two Zeiss papers on MTF´s, a lens does not necessarily produce points from light point sources (e.g. stars). Both lenses show nice "gull wing"-shaped stars at the extreme edges. As these are far wider tangentially than sagittally, they will blur the tangential more = lower tangential MTF.
However, when the lens is correctly focused these patterns are so small that they are almost totally hidden by the 21Mp FF sensor. Result: A lens that is perfect all over the frame! As I, of course, managed to miss focus on a few images, the points are spread over a larger area = several pixels, and can be clearly seen. Never underestimate the educational benefits of a juicy blunder! ...Show more →
If you expose a bit longer you should observe color fringing with the 2.8/25. If not - show us the raw files. Transverse chromatic aberration is the Achilles' heel of retrofocus wideangle lenses, because it does not go away upon stopping down. The 2.8/25 has significantly more TCA than the 2.8/21, and therefore cannot be a better lens.
With current HI-DEF sensors, why not to go for the no-brainer 21 and to crop until 35mm ?
If not, the ZE 28 seems to be a excellent alternative. The 25 even if excellent is called to depreciate due to cessation of its production.
denoir, the curves also show that except for the sagital performance for fine detail wide open, the 2 lenses are actually very, very close, and the 25 is far from being a "weak" or "disappointing" lens not worth half the prize of a 21.
They don't show how either of them performs at f8 and 11, but from what we see at 5,6 the 25 might well surpass the 21 at the outer areas of the frame, leading to a sharper impression across the whole frame as Herb is finding in his images.
toothwalker, the 25/2,8 does show some TCA, I get them especially when I shoot against the light on grass like structures.
No one ever said that the 25 is a "better lens". All we are trying to do here is to figure out where it's strengths and weeknesses are as compared with the 21 and 28.
1. As the MTF charts would suggest, the poor sagittal performance of the 25mm might be quite apparent in the image. T=0.45 for 40 lpmm at 10 mm image height cannot be good news for fine detail rendition or, for a favourite characteristic of mine, object shaping - which needs drawing accuracy in both directions - across the object and along the object.
Commenting on a competitor's lens (unnamed by CZ as they are more gentlemanly than us here) Zeiss have this to say:
"The obvious difference is that the tangential values in the field, e.g. at 15 mm
image height, are much smaller for the zoom lens. The cause of this effect is
evident in the tangential line image which shows the zoom lens to have strong
lateral chromatic aberration." The tangential line sagged in the acompanying MTF chart, a more exaggerated effect than seen in the 25mm under discussion.
I always look for lenses with very high, and very near to parallel lines for this reason - the first for overall contrast and high detail rendition, the second for a high level of correction, from what I understand.
To achieve this with a 21mm is one reason it is so highly regarded. But of course each person will have a different set of priorities.
2. denoir, I am not sure it is so simple. I agree with you as a matter of intuition rather than knowledge. The best lenses do out-resolve sensors (except at sub-optimal apertures and in some parts of the frame), otherwise we would have needed better lenses with each new generation sensor, and B&W film would have needed much better lenses still. I recall theSuede saying as much WRT the 7D - the lenses were better.
In reviewing two lenses in an article, Zeiss use as an example a lens with 160 lpmm, so that gives us some measure, they also disparagingly note a poor lens which 'only' delivered 80 lpmm due to 'stronger aberrations'. I have no idea how they arrive at these figures.
And of course 40 lpmm is an arbitrarily chosen metric. Lenses will resolve more, much more lpmm but at *lower levels of modulation*.
In the same study they state sensor MTF as 59 lpmm for a 12Mp APS-C format and 84 lpmm for a 24Mp FF sensor.
Does anyone know the formula for arriving at this figure for other resolution levels? It would be good to know the data for the rumoured 24Mp APS-C Sony sensor coming in the A77 (can't wait), for example.
My grasp is limited obviously...did I get any of that right, Toothwalker? ;-)
cyra wrote:
denoir, the curves also show that except for the sagital performance for fine detail wide open, the 2 lenses are actually very, very close, and the 25 is far from being a "weak" or "disappointing" lens not worth half the prize of a 21.
Yes and no. Wide open it is much worse, at both 20 and 40 lp/mm lines, both sagittal and tangential. The 25 is still not bad in an absolute sense but the 21 is in a different league at f/2.8.
At f/5.6 it seems to be a very competent performer. You have sharpness across the frame almost to the very extreme edges. The 21 will still produce significantly better finer tangential detail, but otherwise the lenses are pretty evenly matched - the 25 surpasses the 21 on some places and vice versa on some other places.
I would definitely not categorize the 25 as weak. As for the price, well, there's usually a principle of diminishing returns where each minor improvement costs disproportionately much. It is not uncommon with lenses that a 1% improvement leads to a 100% increase in price.
They don't show how either of them performs at f8 and 11, but from what we see at 5,6 the 25 might well surpass the 21 at the outer areas of the frame, leading to a sharper impression across the whole frame as Herb is finding in his images.
No, probably not. The spike increase (i.e astigmatism) in sharpness around 17mm is likely just to emphasize the field curvature of the 25 and the overall image would look less sharp.
It's possible that Herb's 21 is not correctly centered. It's very common. The 21 has an immensely complex design (18 elements or so IIRC) and misaligned copies are commonplace. If I remember correctly Lloyd Chambers had to send back several copies of the 21 ZF before he got one that was correctly aligned. It's one of the disadvantages of the 21 compared to the much more simple design of the 25.
denoir wrote:
According to specs the 5DII handles up to 70 lp/mm, so the 40 lp/mm of the bottom pair of lines are definitely resolvable. The best crop cameras handle even more - for instance the 7D resolves over 100 lp/mm. So all the lines on the MTF charts will have a visible impact when using modern DSLR sensors.
Perhaps, yes. But the point spreads are not eaven in neither brightness nor colour. Longer exposure with higher ISO (I ran at 800 and 3200) would make them somewhat more apparent. Furthermore the resolvable lppmm is dependent of both the pixels and the interpretation of the RGGB-grid. Add orientation, with the point spreads in the corners being at an angle to the pixel grid, which should reduce the sensor-lppmm.
Tooth, yes, the 25 has significantly more TCA than the (near APO) 21, visible mostly around the brightest stars, as is noted in day light photography too. But the small stars are equally sharp with both lenses and in practice over the full field of the sensor. In this regard the 21 should be less prone to lighting conditions than the 25, which it is. And this should affect the IQ much more often with the 25, makking the 21 a "better lens". But in some regards the 25 is at least as good as the 21.
Z250SA wrote:
Perhaps, yes. But the point spreads are not eaven in neither brightness nor colour. Longer exposure with higher ISO (I ran at 800 and 3200) would make them somewhat more apparent. Furthermore the resolvable lppmm is dependent of both the pixels and the interpretation of the RGGB-grid. Add orientation, with the point spreads in the corners being at an angle to the pixel grid, which should reduce the sensor-lppmm.
I'm not sure that I understand what you are getting at. Are you talking about your star test?
Any digital camera newer than 10 years should be able to resolve beyond 40 lp/mm effortlessly. With the best fine grained film you can resolve above 400 lp/mm. The "fine detail" in the Zeiss MTF charts is very conservative. That's why for instance Coastal Optics (high end lenses for scientific use) for instance provide MTF charts with 80 lp/mm and 160 lp/mm lines.
Incidentally, the highest performing lens recorded is the Zeiss 25/2.8 Biogon (ZM mount) that scored a 400 lp/mm @ f/4 in the center of the image which is right at the diffraction limit at that aperture. That means that it is a the theoretical maximum resolving power.
Z250SA wrote:
Tooth, yes, the 25 has significantly more TCA than the (near APO) 21, visible mostly around the brightest stars, as is noted in day light photography too. But the small stars are equally sharp with both lenses and in practice over the full field of the sensor. In this regard the 21 should be less prone to lighting conditions than the 25, which it is. And this should affect the IQ much more often with the 25, makking the 21 a "better lens". But in some regards the 25 is at least as good as the 21.
In some regards, that may be true. But when you talk about sharpness in the sense of resolution, this cannot be judged from the rendering of individual stars. Resolution is about the ability of a lens to discriminate between two closely spaced stars. An unavoidable consequence of TCA is a lowering of the resolution. Just imagine placing a star in the colored blur patch of another star.
philip_pj wrote:
1. As the MTF charts would suggest, the poor sagittal performance of the 25mm might be quite apparent in the image. T=0.45 for 40 lpmm at 10 mm image height cannot be good news for fine detail rendition or, for a favourite characteristic of mine, object shaping - which needs drawing accuracy in both directions - across the object and along the object.
Commenting on a competitor's lens (unnamed by CZ as they are more gentlemanly than us here) Zeiss have this to say:
"The obvious difference is that the tangential values in the field, e.g. at 15 mm
image height, are much smaller for the zoom lens. The cause of this effect is
evident in the tangential line image which shows the zoom lens to have strong
lateral chromatic aberration." The tangential line sagged in the acompanying MTF chart, a more exaggerated effect than seen in the 25mm under discussion.
I always look for lenses with very high, and very near to parallel lines for this reason - the first for overall contrast and high detail rendition, the second for a high level of correction, from what I understand.
To achieve this with a 21mm is one reason it is so highly regarded. But of course each person will have a different set of priorities.
2. denoir, I am not sure it is so simple. I agree with you as a matter of intuition rather than knowledge. The best lenses do out-resolve sensors (except at sub-optimal apertures and in some parts of the frame), otherwise we would have needed better lenses with each new generation sensor, and B&W film would have needed much better lenses still. I recall theSuede saying as much WRT the 7D - the lenses were better.
In reviewing two lenses in an article, Zeiss use as an example a lens with 160 lpmm, so that gives us some measure, they also disparagingly note a poor lens which 'only' delivered 80 lpmm due to 'stronger aberrations'. I have no idea how they arrive at these figures.
And of course 40 lpmm is an arbitrarily chosen metric. Lenses will resolve more, much more lpmm but at *lower levels of modulation*.
In the same study they state sensor MTF as 59 lpmm for a 12Mp APS-C format and 84 lpmm for a 24Mp FF sensor.
Does anyone know the formula for arriving at this figure for other resolution levels? It would be good to know the data for the rumoured 24Mp APS-C Sony sensor coming in the A77 (can't wait), for example.
My grasp is limited obviously...did I get any of that right, Toothwalker? ;-)...Show more →
The first part seems fine. I don't know much about sensor technology. ;-)
denoir wrote:
[
It's possible that Herb's 21 is not correctly centered. It's very common. The 21 has an immensely complex design (18 elements or so IIRC) and misaligned copies are commonplace. If I remember correctly Lloyd Chambers had to send back several copies of the 21 ZF before he got one that was correctly aligned. It's one of the disadvantages of the 21 compared to the much more simple design of the 25.
LC made a test involving many lenses and bodies. After his D3X has been fixed he said he didn't exactly know if one of the 4 samples of ZE provided (at the same time for testing) ZE was really faulty even if somes did not provide exactly the same results.
. 
