alundeb wrote:
Does lateral CA vary with magnification?
Yes.
The findings of PZ (m=0.034) cannot be compared to statements of the manufacturer (m=0). It could very well be that Zeiss put more effort into the variable air spaces of the 2/25, such that its image quality is better preserved at close focus than that of the 2.8/21.
Tariq Gibran wrote:
G. Pierre, I would like to see some tests done at longer distances - landscape vistas, distant trees, mountains, whatever, which might reveal any softness in the cornes and image side borders. This is where the weakness has been shown previously with the 25/2 even stopped down.
cyra, Lloyds does have at least one test of the 25/2 at a much further distance than 1-2m that showed the weak corners stopped down if I recall (I no longer subscribe).
Well, They are already assembled four of the 10 photo-tests that I did the past Monday. I no cut a crop on the central area becasue no are difference between all the AF modes.
The first three photos have an exchange images mousing over them, the fourth you must to select the focus mode in the down texts.
The method is always the same for all: the first with live view focus and that is exchanged with the MF ring full of top-infinity.
The findings of PZ (m=0.034) cannot be compared to statements of the manufacturer (m=0). It could very well be that Zeiss put more effort into the variable air spaces of the 2/25, such that its image quality is better preserved at close focus than that of the 2.8/21.
Thank you. Your article on MTF measurements implicitly say so, but it is not as easy to grasp how dispersion varies with magnification while not with aperture. Do you know how Zeiss measure CA at infinity? That facility is not mentioned in the datasheet for the K8 for instance.
It should not be too difficult to collect some more samples at very little magnification (I hope you agree that maginfication in the order of 0.002 to 0.001 can be regarded as representative of infinity). The ones from Wayne and Thorsten are just single samples and not verified to be representative of the whole frame, as I understand that assymetry may be present.
Well, They are already assembled four of the 10 photo-tests that I did the past Monday. I no cut a crop on the central area becasue no are difference between all the AF modes.
The first three photos have an exchange images mousing over them, the fourth you must to select the focus mode in the down texts.
The method is always the same for all: the first with live view focus and that is exchanged with the MF ring full of top-infinity.
regards
The mouse over is interesting, German. The foreground really gets much worse when you look at the images focused so the focusring hits infinity. It seems that the lens (at least your lens) focuses past infinity when you turn it all the way.
I don't remember where, but that has been said about other Zeiss lenses too - that some can focus past infinity.
Lloyd Chambers also has an example where he tries this kind of thing with the 35/1.4 and comes to the conclusion, that it has to be focused very carefully, he was also getting more CA when not focused properly, if I recall that correctly (my licence expired).
could you post a crop of left and right front corners of the third shot (with the benches), exactly right down to the corner (since the one you have up there doesn't go all the way down)? That would be great.
'The testing distance is not silly, but shooting a plane target at this distance is silly'
Edward's post above nails this, in my view. I would choose an appropiate lens for copy art if that was my interest. The problem is the planar form of the target, how it misrepresents lens performance in 'image space', and the role of curvature.
It's a classic example of measuring procedure efficiency leading and distorting results presentation. I see this all the time in statistical data series evaluation - users will not read the methodology notes and and factor them in - they just want the results so they can draw long bows with them!
An example - the markets get excited because the US employment rate drops...why does it drop? The participation rate falls as more people get so discouraged they stop looking for work. Changes in the PR are the biggest influence on UE but the media and markets are blind to it, and just report the 'headline rate' of UE.
And data always looks convincing - just as with PZ bar charts, with resolution measured to four digits.
'It seems that the ZF improved noticeably on wide open and edge performance at a cost of some CA.'
Yes, I knew this and the background as to why, but I don't mention as many people here use the later vsn - which is also wonderful, and is in keeping with modern Zeiss lenses being better wide open. But it's one big reason I was happy to buy the original.
I would quibble a little with the 'better edges' meme. At 18mm IH (f5.6), the ZF lines are all way apart, at the maximum separation in the chart, and this separation continues until the extreme point of coverage of 21.5mm. It's especially obvious for the critical 40 lpmm - think fine structures, like twigs against a bright sky.
The ZF sagittal does very well in value terms, then plummets (much like the 25/2) whereas the tangential is smoother, at a lower level - leading to differential imaging as a result, quite likely, of CA. In the image centre the CY is also stronger (at f4-5.6). It can be expected to yield more consistent, accurate imaging with a steadier decrease, centre to corner.
This parallelism of MTF lines is best seen in the recent Distagon white paper, where Zeiss publish MTF at f4 for the CY version - the 40 lpmm lines (sag and tan) are identical in value and show ~83% at 5mm image height; I have never seen higher in real world infinity MTF testing, there may be one or two around. Certainly nothing in the current ZE/F range, nor the G range, nor the CY range.
Wayne -- It seems the image circle would only need the couple mm's that we are seeing in vignetting and unrecoverable edge softness. Not to TS extent, just a touch more. I'm not a lens designer, but I wouldn't think it's an issue of the front element or light path, until you get to the rear group -- might need to be slightly further forward, away from the image plane, to give the little coverage needed.
cyra wrote:
The mouse over is interesting, German. The foreground really gets much worse when you look at the images focused so the focusring hits infinity. It seems that the lens (at least your lens) focuses past infinity when you turn it all the way.
I don't remember where, but that has been said about other Zeiss lenses too - that some can focus past infinity.
Lloyd Chambers also has an example where he tries this kind of thing with the 35/1.4 and comes to the conclusion, that it has to be focused very carefully, he was also getting more CA when not focused properly, if I recall that correctly (my licence expired).
could you post a crop of left and right front corners of the third shot (with the benches), exactly right down to the corner (since the one you have up there doesn't go all the way down)? That would be great.
Exactly previous comments to the images (spanish) emphasize this aspect. The whole of this unit and the body used passed the focus an infinite, this is not possible behavior of Zeiss lenses this problem are of all lenses and zooms, for that reason I think it may be a cause that some users have bad results.
The same day a friend carrying a Distagon 21/2,8 and the adjjust to infinit was more short, was impossible to go wrong in the focus but think that the motives more far wolud not a very good sharpness
Regarding the third image, I put the original on the server and the download link below it and you see what you want.
wayne seltzer wrote:
I posted this shot earlier but wanted to post a lighter version with crops which show how good this lens is at f2 across the frame at closer distances.
This lens at f / 2 is excellent in all reasonable distances for this diaphragm. When I say reasonable I mean that the level of longitudinal CAS is admissible. Another issue is that there are some users who want to focus to 50 meters or infinity at f / 2, for those jobs is better to use a telescope :-)
If you display the RGB black letters of your photo you will see, before WB fine tuning, the low level of aberration.
philip_pj wrote:
'
This parallelism of MTF lines is best seen in the recent Distagon white paper, where Zeiss publish MTF at f4 for the CY version - the 40 lpmm lines (sag and tan) are identical in value and show ~83% at 5mm image height; I have never seen higher in real world infinity MTF testing, there may be one or two around. Certainly nothing in the current ZE/F range, nor the G range, nor the CY range.
If I put the MTF for the 2.8/21 CY at f4 from the white paper between those at f2.8 and f5.6 from the published data sheet, there is not much resemblance of these being equal lenses. There is clearly a variance in either lens copy quality or measurement execution. Look at the crossing of the 40 lp/mm sagittal and tangential lines around 17 mm. This cannot just disappear when stopping down to f4 and reappear when stopping down further to f5.6. Or?
alundeb wrote:
Thank you. Your article on MTF measurements implicitly say so, but it is not as easy to grasp how dispersion varies with magnification while not with aperture. Do you know how Zeiss measure CA at infinity? That facility is not mentioned in the datasheet for the K8 for instance.
I don't know the details about their actual measurements, but of course they have the possibility to do ray-tracing for the design ideal.
It should not be too difficult to collect some more samples at very little magnification (I hope you agree that maginfication in the order of 0.002 to 0.001 can be regarded as representative of infinity). The ones from Wayne and Thorsten are just single samples and not verified to be representative of the whole frame, as I understand that assymetry may be present.
I agree that a magnification of order 0.001 is representative of infinity. On the other end of the scale, a magnification of 1 is not:
Here I put the (Y/C) 2.8/21 on a 21-mm tube and placed a ruler in the object plane. The image is badly aberrated even at the employed aperture of f/22. Note that the familiar moderate moustache distortion at infinity has become strong barrel distortion at 1:1, and that lateral chromatic aberration reaches dramatic levels.
In terms of lens performance the magnification used by PZ (m=0.034) is of course much closer to infinity than to m=1, but if the change from m=0.034 to m=1 increases CA from 0.5 to 50 pixels, we cannot assume that the CA measured at m=0.034 is representative of the level of correction at infinity.
Toothwalker wrote:
Here I put the (Y/C) 2.8/21 on a 21-mm tube and placed a ruler in the object plane. The image is badly aberrated even at the employed aperture of f/22. Note that the familiar moderate moustache distortion at infinity has become strong barrel distortion at 1:1, and that lateral chromatic aberration reaches dramatic levels.
In terms of lens performance the magnification used by PZ (m=0.034) is of course much closer to infinity than to m=1, but if the change from m=0.034 to m=1 increases CA from 0.5 to 50 pixels, we cannot assume that the CA measured at m=0.034 is representative of the level of correction at infinity.
I don't know. The curve is rather steep once you go beyond the normal MFD. Inspired by your demo, I tried a similar experiment with my 2.8/21 ZE and a 25 mm tube. The CA was dramatic at maximum magnification (target touching the glass). However, without a tube, the CA at MFD (m=0.2) was down to the normal level, ~1 pixel on the 5DII, in the corner. At PZ test distance, the result was virtually the same, with an error margin of say 20%. At 5m distance it was also indistinguishable from 1m. In the case of this lens we would have to work on a very high precision level to find a difference between m=0.03 and m=0.
The curve is rather steep once you go beyond the normal MFD. Inspired by your demo, I tried a similar experiment with my 2.8/21 ZE and a 25 mm tube. The CA was dramatic at maximum magnification (target touching the glass). However, without a tube, the CA at MFD (m=0.2) was down to the normal level, ~1 pixel on the 5DII, in the corner. At PZ test distance, the result was virtually the same, with an error margin of say 20%. At 5m distance it was also indistinguishable from 1m. In the case of this lens we would have to work on a very high precision level to find a difference between m=0.03 and m=0. ...Show more →
The floating element helps to control aberrations between infinity and MFD, but does not guarantee constant performance. When CA becomes of the order of 1 px or less, the measurement accuracy also starts to play a role, whether that concerns a human staring at a screen or an automated algorithm suffering from possible interpolation errors (something TheSuede wrote recently).
alundeb wrote:
If I put the MTF for the 2.8/21 CY at f4 from the white paper between those at f2.8 and f5.6 from the published data sheet, there is not much resemblance of these being equal lenses. There is clearly a variance in either lens copy quality or measurement execution.
Indeed. I noticed the curve for the Contax-mount Distagon 2/28 shown in that document doesn't quite match the separate datasheet for that lens either. And in this case they are both for f/2, so no interpolation is needed to see they're different.
MTF curves change if the focus is changed, of course. Some Zeiss MTF curves show slight mis-focus at the centre, e.g. the curve that is wrongly labelled f/4 here (it should be labelled f/5.6, and the f/2 curve should be labelled f/1.4). You could do better than the shown performance in the centre by focusing the lens slightly closer, but the periphery would deteriorate due to the nature of its field curvature. In other words, the curve is for a compromise focus between the centre and periphery. I don't know if that compromise is deliberate or simply a consequence of using the same focus (i.e. the best centre focus at f/1.4) for the curves for all f-numbers.
Toothwalker: I copied your test with a 2.8/21 ZF.2 on a Nikon D700, but without an extension tube. At minimum focus distance the lateral chromatic aberration is still admirably low. Here's a full-resolution JPEG. Note: I focused at full aperture on the 2 cm tick mark (tangential structure) at the far-right of the image, and then stopped down to f/16 for the photo. Ignore the uneven lighting: I was only concerned with appropriate exposure behind the 2 cm mark.
I'll venture an opinion here. I've used the ZF 25/2 on a Nikon F2AS shooting Velvia 50: spent about 3 hours one day and about 3 hours another day. I don't think scanning the Velvia to display here would be appropriate, so this will just be informed opinion.
It's a great walk around lens. I think it is like the ZF 35/2 except over a wider FOV. Very strong 3D character, maybe not as contrasty. Colder color balance than 21/2.8, 25/2.8 and 28/2. Very clean output - separates colors very well.
I was trying in all cases to shoot focused at infinity. Ranging from f4 - f11. For me, the lower left and right corners were satisfyingly sharp and in focus when they were within the depth of field. Of course, normally, when shooting landscape or structure scape, the lower corners are going to be in the foreground. The upper left/right corners were sharp and in focus when I tried to be sure they were in focus, which means I had to manually focus on something in the distance where I could line up the split image rangefinder. Of course, shooting ISO 50 film without a tripod (at 1/30, 1/60, 1/125), means I was probably experiencing camera shake at some points.
However, I feel like I can make the corners of this lens sharp, when I want to. Even at f5.6. Would I rather own the 25/2 than the 25/2.8, hard to say. The bright f2 aperture in the viewfinder of a film camera though makes the lens a lot of fun to use. Makes the old version of Velvia 50 sing. Wish they built it 30 years ago.
S Dilworth wrote:
Indeed. I noticed the curve for the Contax-mount Distagon 2/28 shown in that document doesn't quite match the separate datasheet for that lens either. And in this case they are both for f/2, so no interpolation is needed to see they're different.
For the Y/C 2.8/21 there is a third version, in the book "Only Zeiss".
MTF curves change if the focus is changed, of course. Some Zeiss MTF curves show slight mis-focus at the centre, e.g. the curve that is wrongly labelled f/4 here (it should be labelled f/5.6, and the f/2 curve should be labelled f/1.4). You could do better than the shown performance in the centre by focusing the lens slightly closer, but the periphery would deteriorate due to the nature of its field curvature. In other words, the curve is for a compromise focus between the centre and periphery. I don't know if that compromise is deliberate or simply a consequence of using the same focus (i.e. the best centre focus at f/1.4) for the curves for all f-numbers. ...Show more →
The Zeiss focus criterion is to optimize the MTF at 20 lp/mm, in the image center at full aperture.
There is no refocus for measurements at reduced apertures.
Toothwalker: I copied your test with a 2.8/21 ZF.2 on a Nikon D700, but without an extension tube. At minimum focus distance the lateral chromatic aberration is still admirably low. Here's a full-resolution JPEG. Note: I focused at full aperture on the 2 cm tick mark (tangential structure) at the far-right of the image, and then stopped down to f/16 for the photo. Ignore the uneven lighting: I was only concerned with appropriate exposure behind the 2 cm mark.
Thanks. I have to say it is looking very good for the TCA.
grasmuc, Nice shot and show where this ZF25 will shine!
The corner looks good to me for both of your interior and Moto shots. But it may not answer back long distance soft corner issue, since the first one don't have many detail itself on textture to tell and 2nd one is not far away.
Thanks for sharing.
