This topic is in the 2015 archive. You can read it and like its posts here. It is read only.
Archive 2015Zeiss Milvus are out: new range of DSLR lenses!
p.15 #1Zeiss Milvus are out: new range of DSLR lenses!
philber wrote:
If you are shooting a perfectly straight and flat target (0 DOF) for maximum sharpness, yes, as I understand it. That, however, is rarely my case...
From a DOF perspective, flat targets include your typical infinity landscape, which is a fairly common subject. On the other hand, for maximum sharpness you need to stop down, even with Otus, which mitigates alignment errors.
edwardkaraa wrote
So consequently, a little variation in the camera mount or sensor alignment will negate any advantage of using an Otus.From a DOF perspective, flat targets include your typical infinity landscape, which is a fairly common subject. On the other hand, for maximum sharpness you need to stop down, even with Otus, which mitigates alignment errors.
Sep 16, 2015 at 09:02 AM
p.15 #2Zeiss Milvus are out: new range of DSLR lenses!
alundeb wrote:
...
The 25/2.8 and 35/2 are wonderful on the higher rez sensor, but I consider these to be two of the lowest "bs quotient" in the Zeiss line up.i'm confused, you didn't like the ZE 35/2 on the lower res 5DII but you do like it on the d800e? or you just like it better than the nikon 35/1.8 but not better than the EF 35/2 IS?
i've not been sold on any of the ZE/ZF line myself, they don't really match my preferred set of tradeoffs as well as the contax line did.
the new 50 looks interesting as a wide open shooter, where the ZE was nearly unusable. the corner bokeh looks like it could get crazy in a lot of situations though. i doubt i'd ever get one just because of the size. the 85 looks awesome, but way bigger than anything i'd use.Yes! 
(The second posit.)
I prefer the handling, AF/IS, and overall "look" of the 35 IS over the Zeiss 35/2 on the Canon. On the Nikon, I'd rather shoot the Zeiss (used) than buy a new 35/1.8 without IS. Sorry, I didn't differentiate those positions very well.
At the time I bought the ZE 35/2 used, it was running around $700-800, so not a good value compared to the 35/2 IS, which could be bought more recently for $450 used. It wasn't a matter of the sensor being 21MP or 36MP.
My real tipping point came with the ZE 21/2.8, purchased on the same invoice from B&H as the 24 TSE II (that was a costly spree!). Although the TSE cost more ($2200 vs. $1800), its features and lack of terrible vignetting and distortion were a huge plus, in addition to its greater flexibility with movements and its terrific IQ. Otherwise, I liked the ZE very much, but couldn't justify keeping it when they were so close in AOV, as well as owning the EF 16-35L II and the 17 TSE. Recently I find the ZF and ZF.2 models of the 21/2.8 are coming way down in price (used) and might be worthwhile picking up.
I still have a rather tenuous commitment to Nikon, so keep looking for great imaging solutions at the lowest prices. Plus, the ZF lenses can be re-purposed over to Canon in full manual mode with simple adapters. The ZF.2 35/2 at $500 filled that dance card nicely and preforms very well. The real star for me is the ZF 25/2.8, even with it's "defects", selling for just over $400, it is a wonderful lens on that body for landscape and urban shots. On a cost/benefit ratio, the 25/2.8 is close to my top lens in either system for the types of photography I do. So, I haven't written off Zeiss, and no boycott intended.
I may eventually look toward the ZF 100/2 macro to fill that gap in my Nikon performance.I can see where you are coming from, and the situation for the slower Milvus lenses is even worse after we got the Canon 24-70 f/2.8L II and the 16-35 F4 L IS. There is no need to go to Zeiss at those focal lengths for good landscape lenses. I have sold my beloved ZE 21 and also consider letting the 15 mm and the 50 MP go.
Those are certainly amazing Canon lenses, and a bargain for what then can replace in cost and bulk, besides the other integrated features.
For me, it's all a matter of practical considerations, including price and size/weight. Ultimately, it's about the image, and not about brand loyalty or superstitious reverence. I'm finding less to my liking in the Nikon choices (including the flange distance as its Achilles' Heel for Alt lenses) hence my increased interest in using Zeiss ZF on that mount with 36MP sensor.
Gunzorro wrote
sebboh wrote
Gunzorro wrote
I was happy to sell my ZE 21/2.8 and ZE 35/2, as Zeiss seemed to be creatively defining the lens aberrations they infused the lenses with....
The 25/2.8 and 35/2 are wonderful on the higher rez sensor, but I consider these to be two of the lowest "bs quotient" in the Zeiss line up.
i've not been sold on any of the ZE/ZF line myself, they don't really match my preferred set of tradeoffs as well as the contax line did.
the new 50 looks interesting as a wide open shooter, where the ZE was nearly unusable. the corner bokeh looks like it could get crazy in a lot of situations though. i doubt i'd ever get one just because of the size. the 85 looks awesome, but way bigger than anything i'd use.
(The second posit.) I prefer the handling, AF/IS, and overall "look" of the 35 IS over the Zeiss 35/2 on the Canon. On the Nikon, I'd rather shoot the Zeiss (used) than buy a new 35/1.8 without IS. Sorry, I didn't differentiate those positions very well.
At the time I bought the ZE 35/2 used, it was running around $700-800, so not a good value compared to the 35/2 IS, which could be bought more recently for $450 used. It wasn't a matter of the sensor being 21MP or 36MP.
My real tipping point came with the ZE 21/2.8, purchased on the same invoice from B&H as the 24 TSE II (that was a costly spree!). Although the TSE cost more ($2200 vs. $1800), its features and lack of terrible vignetting and distortion were a huge plus, in addition to its greater flexibility with movements and its terrific IQ. Otherwise, I liked the ZE very much, but couldn't justify keeping it when they were so close in AOV, as well as owning the EF 16-35L II and the 17 TSE. Recently I find the ZF and ZF.2 models of the 21/2.8 are coming way down in price (used) and might be worthwhile picking up.
I still have a rather tenuous commitment to Nikon, so keep looking for great imaging solutions at the lowest prices. Plus, the ZF lenses can be re-purposed over to Canon in full manual mode with simple adapters. The ZF.2 35/2 at $500 filled that dance card nicely and preforms very well. The real star for me is the ZF 25/2.8, even with it's "defects", selling for just over $400, it is a wonderful lens on that body for landscape and urban shots. On a cost/benefit ratio, the 25/2.8 is close to my top lens in either system for the types of photography I do. So, I haven't written off Zeiss, and no boycott intended.
I may eventually look toward the ZF 100/2 macro to fill that gap in my Nikon performance.Those are certainly amazing Canon lenses, and a bargain for what then can replace in cost and bulk, besides the other integrated features.
For me, it's all a matter of practical considerations, including price and size/weight. Ultimately, it's about the image, and not about brand loyalty or superstitious reverence.
I'm finding less to my liking in the Nikon choices (including the flange distance as its Achilles' Heel for Alt lenses) hence my increased interest in using Zeiss ZF on that mount with 36MP sensor.Sep 16, 2015 at 09:10 AM
p.15 #3Zeiss Milvus are out: new range of DSLR lenses!
These things are expensive compared to other brands with similar F stops...crazy. I am glad I didn't go Sony. They didn't have good glass, now they have very expensive glass.
Sep 16, 2015 at 03:46 PM
p.15 #4Zeiss Milvus are out: new range of DSLR lenses!
jbranch040597 wrote:
These things are expensive compared to other brands with similar F stops...crazy. I am glad I didn't go Sony. They didn't have good glass, now they have very expensive glass.
... But these are for Canikon. I guess it's time to re-examine the Sony system
These things are expensive compared to other brands with similar F stops...crazy. I am glad I didn't go Sony. They didn't have good glass, now they have very expensive glass.
... But these are for Canikon. I guess it's time to re-examine the Sony system
Sep 16, 2015 at 04:24 PM
p.15 #5Zeiss Milvus are out: new range of DSLR lenses!
To add to what philber wrote below, I've had an email exchange with Brandon at LR after their first results came out, and we talked about why their results don't match Zeiss's, especially why the tangential and radial MTFs diverged so much at the edges in LR's tests when they don't in Zeiss's.
There were several possibilities though he says he wasn't paying too much attention to the Otus when he did the tests.
1. LR uses a Trioptics Imagemaster while Zeiss uses their own K8. From other sources, I've heard that the K8 they use internally is not the same as the one they sell. Take this last claim with a grain of salt. Both LR and Zeiss measure lens rotations, but with different kinds of light and scanning apertures.
2. LR does not take into account sensor cover glass effects with their tests (except for m43 because it's especially thick there), and for something as high-resolution as the Otus, there may be significant effects.
3. LR focuses only on the central point and lets the rest of the field of view fall where it may. Brandon guesses that Zeiss may refocus for field curvature, but this is only conjecture.
So these are differences, but do they cause the different measurements? I don't know, but it doesn't seem unreasonable that different measurement protocols might cause different results. We know not to compare MTF measurements from different test suites (eg. LMA vs. something else), so maybe the same thing applies here?
I'm also a little skeptical of LR's consistency metric. It scales in a weirdly unintuitive way, since higher MTF numbers can lead to higher consistency scores. LR's mentioned this publicly too, so it's not a big revelation. It's also based on just the 30 lp/mm score. That's a very important score, but I'm not sure how well it reflects the performance of the lens as a whole.
Personally, I'm very interested in the 85 Milvus, despite its huge size. From the samples I can see so far, it has really lovely rendering wide open, with its bokeh fringing (ie. SLoCA) artifacts being more pleasing than the 50 Milvus's, which are pretty ugly and jarring in comparison. The 85's MTF numbers stopped down are exceptional, and it looks like it would work well for a landscape lens there and samples I've seen so far would confirm that. So for me, it's a two-in-one lens, for both portraits and landscape.
philber wrote:
What I understand from what Zeiss are saying (I have it second-hand, not first-hand) is that
1) the results are extremely sensitive to even very slight variations in the testing rig and process. A figure of 20 microns has been quoted as being enough to alter results beyond recognition
2) if one is comparing, say an average lens Vs a top-performance lens, the relative performance of the average lens will be affected less by this variation than that of the top-performance lens.
3) top-performance lenses tend to have a performance envelope relative to placement that looks like a high-but-narrow peak. Lesser-performance lenses peak...Show more →
There were several possibilities though he says he wasn't paying too much attention to the Otus when he did the tests.
1. LR uses a Trioptics Imagemaster while Zeiss uses their own K8. From other sources, I've heard that the K8 they use internally is not the same as the one they sell. Take this last claim with a grain of salt. Both LR and Zeiss measure lens rotations, but with different kinds of light and scanning apertures.
2. LR does not take into account sensor cover glass effects with their tests (except for m43 because it's especially thick there), and for something as high-resolution as the Otus, there may be significant effects.
3. LR focuses only on the central point and lets the rest of the field of view fall where it may. Brandon guesses that Zeiss may refocus for field curvature, but this is only conjecture.
So these are differences, but do they cause the different measurements? I don't know, but it doesn't seem unreasonable that different measurement protocols might cause different results. We know not to compare MTF measurements from different test suites (eg. LMA vs. something else), so maybe the same thing applies here?
I'm also a little skeptical of LR's consistency metric. It scales in a weirdly unintuitive way, since higher MTF numbers can lead to higher consistency scores. LR's mentioned this publicly too, so it's not a big revelation. It's also based on just the 30 lp/mm score. That's a very important score, but I'm not sure how well it reflects the performance of the lens as a whole.
Personally, I'm very interested in the 85 Milvus, despite its huge size. From the samples I can see so far, it has really lovely rendering wide open, with its bokeh fringing (ie. SLoCA) artifacts being more pleasing than the 50 Milvus's, which are pretty ugly and jarring in comparison. The 85's MTF numbers stopped down are exceptional, and it looks like it would work well for a landscape lens there and samples I've seen so far would confirm that. So for me, it's a two-in-one lens, for both portraits and landscape.
philber wrote:
hiepphotog wrote
Do you have any insight that you can share to this seemingly less consistency?1) the results are extremely sensitive to even very slight variations in the testing rig and process. A figure of 20 microns has been quoted as being enough to alter results beyond recognition
2) if one is comparing, say an average lens Vs a top-performance lens, the relative performance of the average lens will be affected less by this variation than that of the top-performance lens.
3) top-performance lenses tend to have a performance envelope relative to placement that looks like a high-but-narrow peak. Lesser-performance lenses peak...Show more →
Sep 16, 2015 at 06:17 PM
p.15 #6Zeiss Milvus are out: new range of DSLR lenses!
philber wrote:
What I understand from what Zeiss are saying (I have it second-hand, not first-hand) is that
1) the results are extremely sensitive to even very slight variations in the testing rig and process. A figure of 20 microns has been quoted as being enough to alter results beyond recognition
2) if one is comparing, say an average lens Vs a top-performance lens, the relative performance of the average lens will be affected less by this variation than that of the top-performance lens.
3) top-performance lenses tend to have a performance envelope relative to placement that looks like a high-but-narrow peak. Lesser-performance lenses peak...Show more →
This doesn't make sense to me. Are you an English speaker and the person you talked to German? I'm wondering if this is a lost-in-translation issue (ie they didn't realize you were asking about relative, not absolute performance).
A different type of machine or a differently calibrated machine would account for Roger getting a different MTF measurement on each lens (absolute performance). Roger points that out up front. That makes complete sense to me.
But I don't see how it would affect copy variation (relative performance). While his MTF numbers should be "wrong" (different from Zeiss's), they should all be "wrong" by the same amount and so a variance / copy variation score should be accurate.
Andre Y wrote:
3. LR focuses only on the central point and lets the rest of the field of view fall where it may. Brandon guesses that Zeiss may refocus for field curvature, but this is only conjecture.
I remember reading that in one of his comments.
On this, LR is right and Zeiss is wrong (if the conjecture is correct).
A single MTF graph refocused for every point is basically a lie. It's not much better than the other manufacturer's "calculated" charts (maybe worse). Because it's not a situation that will ever exist in reality (unless you find JUST the right perfect subject, I suppose).
Now, obviously I can see the usefulness of refocusing for the rule of thirds points and maybe even the corners. Not all subjects are in the center. But those are separate measurements that belong on their own separate graphs.
And again, this and the other issues you mention would, I believe, affect absolute but not relative measurements. Though, as you say, LR's methodology for REPRESENTING the relative measurements may not be perfect, but that's really a different issue.
hiepphotog wrote
Do you have any insight that you can share to this seemingly less consistency?1) the results are extremely sensitive to even very slight variations in the testing rig and process. A figure of 20 microns has been quoted as being enough to alter results beyond recognition
2) if one is comparing, say an average lens Vs a top-performance lens, the relative performance of the average lens will be affected less by this variation than that of the top-performance lens.
3) top-performance lenses tend to have a performance envelope relative to placement that looks like a high-but-narrow peak. Lesser-performance lenses peak...Show more →
This doesn't make sense to me. Are you an English speaker and the person you talked to German? I'm wondering if this is a lost-in-translation issue (ie they didn't realize you were asking about relative, not absolute performance).
A different type of machine or a differently calibrated machine would account for Roger getting a different MTF measurement on each lens (absolute performance). Roger points that out up front. That makes complete sense to me.
But I don't see how it would affect copy variation (relative performance). While his MTF numbers should be "wrong" (different from Zeiss's), they should all be "wrong" by the same amount and so a variance / copy variation score should be accurate.
Andre Y wrote:
3. LR focuses only on the central point and lets the rest of the field of view fall where it may. Brandon guesses that Zeiss may refocus for field curvature, but this is only conjecture.
I remember reading that in one of his comments.
On this, LR is right and Zeiss is wrong (if the conjecture is correct).
A single MTF graph refocused for every point is basically a lie. It's not much better than the other manufacturer's "calculated" charts (maybe worse). Because it's not a situation that will ever exist in reality (unless you find JUST the right perfect subject, I suppose).
Now, obviously I can see the usefulness of refocusing for the rule of thirds points and maybe even the corners. Not all subjects are in the center. But those are separate measurements that belong on their own separate graphs.
And again, this and the other issues you mention would, I believe, affect absolute but not relative measurements. Though, as you say, LR's methodology for REPRESENTING the relative measurements may not be perfect, but that's really a different issue.
Sep 16, 2015 at 06:21 PM
p.15 #7Zeiss Milvus are out: new range of DSLR lenses!
Andre Y wrote:
3. LR focuses only on the central point and lets the rest of the field of view fall where it may. Brandon guesses that Zeiss may refocus for field curvature, but this is only conjecture.
my understanding is that zeiss chooses the single point for best focus over the entire frame, but this is often not the best central focus point.
3. LR focuses only on the central point and lets the rest of the field of view fall where it may. Brandon guesses that Zeiss may refocus for field curvature, but this is only conjecture.
my understanding is that zeiss chooses the single point for best focus over the entire frame, but this is often not the best central focus point.
Sep 16, 2015 at 06:41 PM
p.15 #8Zeiss Milvus are out: new range of DSLR lenses!
I think a different aspect of the Otus lenses is that they are assembled and adjusted by hand. It is certainly possible that all the Otus lenses could score excellently buy vary in different regions of the frame. Using Roger's metric for variation they could score a bit higher in variation, but all the variations are excellent. In contrast a different lens could be fairly poor on average, but have a couple of real bad copies but since real bad doesn't vary that much from fairly poor it could have the same score for variation.
Said another way, we need to know not just variation but what type of variation we see and Roger does show that. If you look at the Otus lenses to my eye they show variation but all within the excellent range. This sort of variation in my view is not problematic. Whereas a lesser lens that has variation that leads to poor performance for several copies is a serious issue.
Edited on Sep 16, 2015 at 07:15 PM · View previous versions
Said another way, we need to know not just variation but what type of variation we see and Roger does show that. If you look at the Otus lenses to my eye they show variation but all within the excellent range. This sort of variation in my view is not problematic. Whereas a lesser lens that has variation that leads to poor performance for several copies is a serious issue.
Edited on Sep 16, 2015 at 07:15 PM · View previous versions
Sep 16, 2015 at 07:07 PM
p.15 #9Zeiss Milvus are out: new range of DSLR lenses!
Lee Saxon wrote:
A single MTF graph refocused for every point is basically a lie. It's not much better than the other manufacturer's "calculated" charts (maybe worse). Because it's not a situation that will ever exist in reality (unless you find JUST the right perfect subject, I suppose).
Brandon also mentioned that the Otus had little field curvature, so I think this was more of an off-hand remark. According to him, oblique spherical aberration is the more likely culprit, but this kind of error (along with the other kinds of spherical aberration in the Otus) is sensitive to cover glass thickness, especially at the edges, hence his remark about the cover glass.
BTW, my reasons listed above went from most likely to cause a difference to least.
As for consistency being somewhat constant across different measurement protocols, I don't necessarily agree. If the test rig is sensitive and unreliable for the thing they're using as their consistency metric, things can diverge in an unproportional way, ie. not all measurements will trend in the same direction. For instance, why does the Zeiss 2/135 measure much closer to Zeiss's own measurements than the Otus, when presumably they use the same measurement protocol at the factory for both lenses?
edit: I went back and reread LR's writeup of how they compute their variance score (which we know we shouldn't look at without also looking at the graphs, right?): http://www.lensrentals.com/blog/2015/06/measuring-lens-variance
There are many things one can question in their methods, but the striking thing to me is that they take the worse of either the tangential or radial curves at 30 lp/mm, averaged over 4 quadrants of 10 lenses, to compute their variance score. Let's say their test somehow exaggerates astigmatism (eg. the cover glass issue), causing wide divergence in the two curves, and Zeiss's doesn't. This would lead to variation scores that are wildly different and perhaps even misleading. Bottom line: look at the curves themselves, not the number.
A single MTF graph refocused for every point is basically a lie. It's not much better than the other manufacturer's "calculated" charts (maybe worse). Because it's not a situation that will ever exist in reality (unless you find JUST the right perfect subject, I suppose).
Brandon also mentioned that the Otus had little field curvature, so I think this was more of an off-hand remark. According to him, oblique spherical aberration is the more likely culprit, but this kind of error (along with the other kinds of spherical aberration in the Otus) is sensitive to cover glass thickness, especially at the edges, hence his remark about the cover glass.
BTW, my reasons listed above went from most likely to cause a difference to least.
As for consistency being somewhat constant across different measurement protocols, I don't necessarily agree. If the test rig is sensitive and unreliable for the thing they're using as their consistency metric, things can diverge in an unproportional way, ie. not all measurements will trend in the same direction. For instance, why does the Zeiss 2/135 measure much closer to Zeiss's own measurements than the Otus, when presumably they use the same measurement protocol at the factory for both lenses?
edit: I went back and reread LR's writeup of how they compute their variance score (which we know we shouldn't look at without also looking at the graphs, right?): http://www.lensrentals.com/blog/2015/06/measuring-lens-variance
There are many things one can question in their methods, but the striking thing to me is that they take the worse of either the tangential or radial curves at 30 lp/mm, averaged over 4 quadrants of 10 lenses, to compute their variance score. Let's say their test somehow exaggerates astigmatism (eg. the cover glass issue), causing wide divergence in the two curves, and Zeiss's doesn't. This would lead to variation scores that are wildly different and perhaps even misleading. Bottom line: look at the curves themselves, not the number.
Sep 16, 2015 at 07:11 PM
p.15 #10Zeiss Milvus are out: new range of DSLR lenses!
I want to expand on the point about variability I noted above. Let's imagine two lenses, lens A and lens B, and to keep it simple we test just the 30 lp/mm of four copies of each lens.
The measurements for lens A look like as you go out from the centre:
Centre: 86%, 5mm: 64%, 10mm: 74%, 15mm: 70%, 20mm: 56%
Centre: 75%, 5mm: 75%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 85%, 5mm: 65%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 74%, 5mm: 76%, 10mm: 76%, 15mm: 70%, 20mm: 54%
If you do the math the average MTF across the lenses would be: Centre: 80%, 5mm 70%, 10mm 70%, 15mm 70/%, and 20mm 60%, but there is a fair bit of variation. Yet none of the lenses are poor performers. They are all excellent.
Compare that to this set of lenses for Lens B
Centre: 48%, 5mm: 43%, 10mm: 40%, 15mm: 40%, 20mm: 35%
Centre: 45%, 5mm: 42%, 10mm: 42%, 15mm: 41%, 20mm: 32%
Centre: 46%, 5mm: 45%, 10mm: 44%, 15mm: 37%, 20mm: 33%
Centre: 33%, 5mm: 30%, 10mm: 28%, 15mm: 26%, 20mm: 20%
If you do the math the average MTF across the lenses would be Centre: 43%, 5mm 40%, 10mm 38%, 15mm 36%, 20mm 30% and the variance is not that much higher than Lens A because the top three lenses are very close in performance. Yet looking it is clear that this lens is both mediocre and perhaps more importantly one of the four samples is very bad. The kind of thing you would probably definitely want to send back.
In this example variance of MTF scores is a pretty poor measure of how well lenses are manufactured. What would be much more useful is a confidence interval (I think 90% would work well) around the MTF score at each distance from the centre, but even this would miss that the variance is not normally distributed for lens B. Instead one lens varies substantially (the bad one). Other stats looking for outliers would actually be more useful in noting the number of lenses that are duds that were tested within the set of lens considered. To do such tests I think you would actually need more than 10 copies of each lens (30 would probably be enough), so it would require a lot more testing. In my view as much as I like the tests that lens rentals does, this test of variance in MTF data doesn't really tell us about manufacturing tolerances in the way that the write up seems to suggest. There isn't quite enough data to do such an analysis well.
I guess what i want to say is that the Otus lenses could easily be like lens A above and the variance would not be a concern at all. Whereas other lenses could be like B and the variance would be a real concern. This suggests that the variance number itself does not give enough information to really tell us whether we should be concerned about getting a bad copy of a particular lens.
The measurements for lens A look like as you go out from the centre:
Centre: 86%, 5mm: 64%, 10mm: 74%, 15mm: 70%, 20mm: 56%
Centre: 75%, 5mm: 75%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 85%, 5mm: 65%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 74%, 5mm: 76%, 10mm: 76%, 15mm: 70%, 20mm: 54%
If you do the math the average MTF across the lenses would be: Centre: 80%, 5mm 70%, 10mm 70%, 15mm 70/%, and 20mm 60%, but there is a fair bit of variation. Yet none of the lenses are poor performers. They are all excellent.
Compare that to this set of lenses for Lens B
Centre: 48%, 5mm: 43%, 10mm: 40%, 15mm: 40%, 20mm: 35%
Centre: 45%, 5mm: 42%, 10mm: 42%, 15mm: 41%, 20mm: 32%
Centre: 46%, 5mm: 45%, 10mm: 44%, 15mm: 37%, 20mm: 33%
Centre: 33%, 5mm: 30%, 10mm: 28%, 15mm: 26%, 20mm: 20%
If you do the math the average MTF across the lenses would be Centre: 43%, 5mm 40%, 10mm 38%, 15mm 36%, 20mm 30% and the variance is not that much higher than Lens A because the top three lenses are very close in performance. Yet looking it is clear that this lens is both mediocre and perhaps more importantly one of the four samples is very bad. The kind of thing you would probably definitely want to send back.
In this example variance of MTF scores is a pretty poor measure of how well lenses are manufactured. What would be much more useful is a confidence interval (I think 90% would work well) around the MTF score at each distance from the centre, but even this would miss that the variance is not normally distributed for lens B. Instead one lens varies substantially (the bad one). Other stats looking for outliers would actually be more useful in noting the number of lenses that are duds that were tested within the set of lens considered. To do such tests I think you would actually need more than 10 copies of each lens (30 would probably be enough), so it would require a lot more testing. In my view as much as I like the tests that lens rentals does, this test of variance in MTF data doesn't really tell us about manufacturing tolerances in the way that the write up seems to suggest. There isn't quite enough data to do such an analysis well.
I guess what i want to say is that the Otus lenses could easily be like lens A above and the variance would not be a concern at all. Whereas other lenses could be like B and the variance would be a real concern. This suggests that the variance number itself does not give enough information to really tell us whether we should be concerned about getting a bad copy of a particular lens.
Sep 16, 2015 at 09:08 PM
p.15 #11Zeiss Milvus are out: new range of DSLR lenses!
Lee Saxon wrote:
This doesn't make sense to me. Are you an English speaker and the person you talked to German?
As stated, this was second-hand info. I can't help you beyond that.
This doesn't make sense to me. Are you an English speaker and the person you talked to German?
As stated, this was second-hand info. I can't help you beyond that.
Sep 16, 2015 at 11:43 PM
p.15 #12Zeiss Milvus are out: new range of DSLR lenses!
Steve Spencer wrote:
I want to expand on the point about variability I noted above. Let's imagine two lenses, lens A and lens B, and to keep it simple we test just the 30 lp/mm of four copies of each lens.
The measurements for lens A look like as you go out from the centre:
Centre: 86%, 5mm: 64%, 10mm: 74%, 15mm: 70%, 20mm: 56%
Centre: 75%, 5mm: 75%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 85%, 5mm: 65%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 74%, 5mm: 76%, 10mm: 76%, 15mm: 70%, 20mm: 54%
If you do the math the average MTF across the lenses would be:...Show more →
Did you see the link in Andre's post above yours? They basically are doing a confidence interval. As you say, if they simply averaged the MTF of ten copies that wouldn't really give you a variance score. But they're averaging the MTF of ten copies and then telling you how much the best and worst copies varied from that average. That's as good a method as I can think of for reporting the variance.
As you say it'd be better if they could test even more copies (though check the "Lens Variance by Number of Tested copies" graph which tends to even out by not even 10 but 5), but with many reviewers inexplicably reporting on good or bad manufacturing consistency after seeing two or even one copy (even Lloyd is prone to do this, contrary to his usual careful methodology) ten is a big step in the right direction.
Also, as you say, even an Otus that's x% below average is still excellent, while some lenses the best copy may be just above acceptable. Sure. But that's still a problem. To me a bad copy in the relative sense is a bad copy even if it's still a good lens in the absolute sense. In fact, a better and more expensive lens makes the problem worse: a 10% below normal $500 lens, you've wasted $50 (paid $500 and gotten a $450 value), whereas with the 85 Otus that 10% represents a waste of $450. A $4050 lens might still be great, but if I paid $4500 it ain't great enough.
I want to expand on the point about variability I noted above. Let's imagine two lenses, lens A and lens B, and to keep it simple we test just the 30 lp/mm of four copies of each lens.
The measurements for lens A look like as you go out from the centre:
Centre: 86%, 5mm: 64%, 10mm: 74%, 15mm: 70%, 20mm: 56%
Centre: 75%, 5mm: 75%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 85%, 5mm: 65%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 74%, 5mm: 76%, 10mm: 76%, 15mm: 70%, 20mm: 54%
If you do the math the average MTF across the lenses would be:...Show more →
Did you see the link in Andre's post above yours? They basically are doing a confidence interval. As you say, if they simply averaged the MTF of ten copies that wouldn't really give you a variance score. But they're averaging the MTF of ten copies and then telling you how much the best and worst copies varied from that average. That's as good a method as I can think of for reporting the variance.
As you say it'd be better if they could test even more copies (though check the "Lens Variance by Number of Tested copies" graph which tends to even out by not even 10 but 5), but with many reviewers inexplicably reporting on good or bad manufacturing consistency after seeing two or even one copy (even Lloyd is prone to do this, contrary to his usual careful methodology) ten is a big step in the right direction.
Also, as you say, even an Otus that's x% below average is still excellent, while some lenses the best copy may be just above acceptable. Sure. But that's still a problem. To me a bad copy in the relative sense is a bad copy even if it's still a good lens in the absolute sense. In fact, a better and more expensive lens makes the problem worse: a 10% below normal $500 lens, you've wasted $50 (paid $500 and gotten a $450 value), whereas with the 85 Otus that 10% represents a waste of $450. A $4050 lens might still be great, but if I paid $4500 it ain't great enough.
Sep 16, 2015 at 11:58 PM
p.15 #13Zeiss Milvus are out: new range of DSLR lenses!
Lee Saxon wrote:
The measurements for lens A look like as you go out from the centre:
Centre: 86%, 5mm: 64%, 10mm: 74%, 15mm: 70%, 20mm: 56%
Centre: 75%, 5mm: 75%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 85%, 5mm: 65%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 74%, 5mm: 76%, 10mm: 76%, 15mm: 70%, 20mm: 54%
If you do the math the average MTF across the lenses would be: Centre: 80%, 5mm 70%, 10mm 70%, 15mm 70/%, and 20mm 60%, but there is a fair bit of variation. Yet none of the lenses are poor performers. They are all excellent.
Compare that to this set of lenses for Lens B
Centre: 48%, 5mm: 43%, 10mm: 40%, 15mm: 40%, 20mm: 35%
Centre: 45%, 5mm: 42%, 10mm: 42%, 15mm: 41%, 20mm: 32%
Centre: 46%, 5mm: 45%, 10mm: 44%, 15mm: 37%, 20mm: 33%
Centre: 33%, 5mm: 30%, 10mm: 28%, 15mm: 26%, 20mm: 20%
If you do the math the average MTF across the lenses would be Centre: 43%, 5mm 40%, 10mm 38%, 15mm 36%, 20mm 30% and the variance is not that much higher than Lens A because the top three lenses are very close in performance. Yet looking it is clear that this lens is both mediocre and perhaps more importantly one of the four samples is very bad. The kind of thing you would probably definitely want to send back.
In this example variance of MTF scores is a pretty poor measure of how well lenses are manufactured. What would be much more useful is a confidence interval (I think 90% would work well) around the MTF score at each distance from the centre, but even this would miss that the variance is not normally distributed for lens B. Instead one lens varies substantially (the bad one). Other stats looking for outliers would actually be more useful in noting the number of lenses that are duds that were tested within the set of lens considered. To do such tests I think you would actually need more than 10 copies of each lens (30 would probably be enough), so it would require a lot more testing. In my view as much as I like the tests that lens rentals does, this test of variance in MTF data doesn't really tell us about manufacturing tolerances in the way that the write up seems to suggest. There isn't quite enough data to do such an analysis well.
I guess what i want to say is that the Otus lenses could easily be like lens A above and the variance would not be a concern at all. Whereas other lenses could be like B and the variance would be a real concern. This suggests that the variance number itself does not give enough information to really tell us whether we should be concerned about getting a bad copy of a particular lens.Did you see the link in Andre's post above yours? They basically are doing a confidence interval. As you say, if they simply averaged the MTF of ten copies that wouldn't really give you a variance score. But they're averaging the MTF of ten copies and then telling you how much the best and worst copies varied from that average. That's as good a method as I can think of for reporting the variance.
As you say it'd be better if they could test even more copies (though check the "Lens Variance by Number of Tested copies" graph which tends...Show more →
Lee you missed part of the point I was trying to make. First, they do present variance information, but they do this in separate graphs, when it would be much better to present it on one graph as a confidence interval around the mean. By presenting it on a separate graph it is easy to miss that because the Otus starts at a high level that even the poorer performing Otus lenses are performing much better than even the best performing lenses of some similar lenses. If the variance was presented as a confidence interval this would be obvious.
Second and more importantly we don't know the nature of the variance. As my example above suggests it is possible to have all excellent performing lenses that bounce up and down around a high mean. Each lens is higher at some point from the centre to the edge and lower at others, but all excellent scores. This sort of variance would in practice be non-problematic. In contrast a lens with the same variance score could have 20 percent of the lenses scoring quite poorly all the way across the frame. These two lenses could have the same variance score but we should draw different conclusion about what it means for getting a bad copy. The first lens that simply varies at different points from the centre but has excellent performance at all of the points there are no real bad copies. The second lens clearly has one bad copy. The first there are no manufacturing problems, the second quality control clearly missed a lens.
There are statistics that can tell the difference between these two scenarios, but they would require about 30 copies of each lens to be tested to use them effectively. They are basic outlier analyses. That is what we want to know. How many outliers (and in a particular direction, i.e., negative outliers or bad copies) are there for a certain lens. That is much more important than whether there is 5 or 10 percent variation around the MTF mean. What lens rentals reports is whether there is 5 or 10 percent variation around the MTF mean (and even this is hard to interpret) rather than how many bad copies there are.
I do agree that testing 10 copies is a huge step forward and I like in general very much what Roger and Brandon have done with their testing. It is just this one aspect of their testing in which variance is used as a stand in for what in my view should be outlier analysis that I find problematic.
Steve Spencer wrote
I want to expand on the point about variability I noted above. Let's imagine two lenses, lens A and lens B, and to keep it simple we test just the 30 lp/mm of four copies of each lens.The measurements for lens A look like as you go out from the centre:
Centre: 86%, 5mm: 64%, 10mm: 74%, 15mm: 70%, 20mm: 56%
Centre: 75%, 5mm: 75%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 85%, 5mm: 65%, 10mm: 65%, 15mm: 70%, 20mm: 65%
Centre: 74%, 5mm: 76%, 10mm: 76%, 15mm: 70%, 20mm: 54%
If you do the math the average MTF across the lenses would be: Centre: 80%, 5mm 70%, 10mm 70%, 15mm 70/%, and 20mm 60%, but there is a fair bit of variation. Yet none of the lenses are poor performers. They are all excellent.
Compare that to this set of lenses for Lens B
Centre: 48%, 5mm: 43%, 10mm: 40%, 15mm: 40%, 20mm: 35%
Centre: 45%, 5mm: 42%, 10mm: 42%, 15mm: 41%, 20mm: 32%
Centre: 46%, 5mm: 45%, 10mm: 44%, 15mm: 37%, 20mm: 33%
Centre: 33%, 5mm: 30%, 10mm: 28%, 15mm: 26%, 20mm: 20%
If you do the math the average MTF across the lenses would be Centre: 43%, 5mm 40%, 10mm 38%, 15mm 36%, 20mm 30% and the variance is not that much higher than Lens A because the top three lenses are very close in performance. Yet looking it is clear that this lens is both mediocre and perhaps more importantly one of the four samples is very bad. The kind of thing you would probably definitely want to send back.
In this example variance of MTF scores is a pretty poor measure of how well lenses are manufactured. What would be much more useful is a confidence interval (I think 90% would work well) around the MTF score at each distance from the centre, but even this would miss that the variance is not normally distributed for lens B. Instead one lens varies substantially (the bad one). Other stats looking for outliers would actually be more useful in noting the number of lenses that are duds that were tested within the set of lens considered. To do such tests I think you would actually need more than 10 copies of each lens (30 would probably be enough), so it would require a lot more testing. In my view as much as I like the tests that lens rentals does, this test of variance in MTF data doesn't really tell us about manufacturing tolerances in the way that the write up seems to suggest. There isn't quite enough data to do such an analysis well.
I guess what i want to say is that the Otus lenses could easily be like lens A above and the variance would not be a concern at all. Whereas other lenses could be like B and the variance would be a real concern. This suggests that the variance number itself does not give enough information to really tell us whether we should be concerned about getting a bad copy of a particular lens.
As you say it'd be better if they could test even more copies (though check the "Lens Variance by Number of Tested copies" graph which tends...Show more →
Lee you missed part of the point I was trying to make. First, they do present variance information, but they do this in separate graphs, when it would be much better to present it on one graph as a confidence interval around the mean. By presenting it on a separate graph it is easy to miss that because the Otus starts at a high level that even the poorer performing Otus lenses are performing much better than even the best performing lenses of some similar lenses. If the variance was presented as a confidence interval this would be obvious.
Second and more importantly we don't know the nature of the variance. As my example above suggests it is possible to have all excellent performing lenses that bounce up and down around a high mean. Each lens is higher at some point from the centre to the edge and lower at others, but all excellent scores. This sort of variance would in practice be non-problematic. In contrast a lens with the same variance score could have 20 percent of the lenses scoring quite poorly all the way across the frame. These two lenses could have the same variance score but we should draw different conclusion about what it means for getting a bad copy. The first lens that simply varies at different points from the centre but has excellent performance at all of the points there are no real bad copies. The second lens clearly has one bad copy. The first there are no manufacturing problems, the second quality control clearly missed a lens.
There are statistics that can tell the difference between these two scenarios, but they would require about 30 copies of each lens to be tested to use them effectively. They are basic outlier analyses. That is what we want to know. How many outliers (and in a particular direction, i.e., negative outliers or bad copies) are there for a certain lens. That is much more important than whether there is 5 or 10 percent variation around the MTF mean. What lens rentals reports is whether there is 5 or 10 percent variation around the MTF mean (and even this is hard to interpret) rather than how many bad copies there are.
I do agree that testing 10 copies is a huge step forward and I like in general very much what Roger and Brandon have done with their testing. It is just this one aspect of their testing in which variance is used as a stand in for what in my view should be outlier analysis that I find problematic.
Sep 17, 2015 at 04:32 AM
p.15 #14Zeiss Milvus are out: new range of DSLR lenses!
Lee Saxon wrote:
This doesn't make sense to me. Are you an English speaker and the person you talked to German?
OK, here I have first hand info if it helps. Though it is from Germany, sure is not a native English speaker but it is pretty clear to me.
see below quote
All lenses that leave our house heading for a customer (e.g. out of serial production as a new item, out of service after repair, for demo purposes, or as class B stock lens as a replacement for a lens that cannot be repaired anymore) are adjusted individually on a K8 and have to pass our performance tests (100% control). Our allowed tolerances in performance are very tight – so it plays no role if you use a new lens, a repaired lens or a refurbished one. Buying lenses from us has nothing to do with “lottery”.
The variations that lensrentals “detected” in...Show more →
This doesn't make sense to me. Are you an English speaker and the person you talked to German?
OK, here I have first hand info if it helps. Though it is from Germany, sure is not a native English speaker but it is pretty clear to me.
see below quote
All lenses that leave our house heading for a customer (e.g. out of serial production as a new item, out of service after repair, for demo purposes, or as class B stock lens as a replacement for a lens that cannot be repaired anymore) are adjusted individually on a K8 and have to pass our performance tests (100% control). Our allowed tolerances in performance are very tight – so it plays no role if you use a new lens, a repaired lens or a refurbished one. Buying lenses from us has nothing to do with “lottery”.
The variations that lensrentals “detected” in...Show more →
Sep 17, 2015 at 11:19 AM
p.15 #15Zeiss Milvus are out: new range of DSLR lenses!
zhangyue wrote:
OK, here I have first hand info if it helps. Though it is from Germany, sure is not a native English speaker but it is pretty clear to me.
see below quote
The variations that lensrentals “detected” in lenses out of serial production were caused by focus variations – unfortunately, they are not aware that very small focus variations of about 0.01mm in the image field will have enormous impact on the overall performance in the image especially with extremely high performing lenses, although there is no difference noticeable on the optical axis. The results they provided are the typical variations of one individual copy of a lens focussed and measured several times. Beside this, the MTF measuring equipment used by lensrentals is not really appropriate to find out the characteristics of a photo lens used in the whole visible spectrum.
We invented the MTF measuring machines and the computer based lens performance measurement in 1943 - today, we are the leading experts in provide MTF measuring equipment for lenses in the world, based on this experience of more than 70 years.
I will, respectfully, disagree with most of that. Particularly since we certainly did measure, refocus, and remeasure one lens many times and saw, well, none of that variation. And our focus accuracy is repeatable to 0.0025mm. We actually looked into K8 and K9 machines and found they were rather limited for our purposes. And we do measure both whole visible spectrum and 3 individual wavelength spectrums, although we do generally publish measurements, like most people, using photopic spectrums.
I'm a near-fanboy about Zeiss in general, but someone there is speaking without knowing what they speak of (meaning our testing methods) which is usually a bad idea. And I note that the writer, as most manufacturers do, talks about their tight tolerances without sharing what those were. Tight is not a number and 100% of tight is still not a number. It's a vague verbal description. I do not mean anything negative towards Zeiss because they and Leica are the only manufacturer that, as far as I know, does actual MTF testing on actual copies. But unless they share their numbers (because until they bypass the laws of physics some variation is going to happen) then we have no idea what 100% of some specification means.
On other topics though, I totally agree that outlier analysis is the best way to measure what's being discussed here and that's NOT what I'm able to do. I just don't have the resources and time to measure 50 to 60 (I don't think 30 is enough) copies of lenses to find those numbers. All I can do is some baby steps in that direction.
And I totally agree - look at the graphs, not the numbers.
Roger
Lee Saxon wrote
This doesn't make sense to me. Are you an English speaker and the person you talked to German?see below quote
Earlier quote
All lenses that leave our house heading for a customer (e.g. out of serial production as a new item, out of service after repair, for demo purposes, or as class B stock lens as a replacement for a lens that cannot be repaired anymore) are adjusted individually on a K8 and have to pass our performance tests (100% control). Our allowed tolerances in performance are very tight – so it plays no role if you use a new lens, a repaired lens or a refurbished one. Buying lenses from us has nothing to do with “lottery”.The variations that lensrentals “detected” in lenses out of serial production were caused by focus variations – unfortunately, they are not aware that very small focus variations of about 0.01mm in the image field will have enormous impact on the overall performance in the image especially with extremely high performing lenses, although there is no difference noticeable on the optical axis. The results they provided are the typical variations of one individual copy of a lens focussed and measured several times. Beside this, the MTF measuring equipment used by lensrentals is not really appropriate to find out the characteristics of a photo lens used in the whole visible spectrum.
We invented the MTF measuring machines and the computer based lens performance measurement in 1943 - today, we are the leading experts in provide MTF measuring equipment for lenses in the world, based on this experience of more than 70 years.
I will, respectfully, disagree with most of that. Particularly since we certainly did measure, refocus, and remeasure one lens many times and saw, well, none of that variation. And our focus accuracy is repeatable to 0.0025mm. We actually looked into K8 and K9 machines and found they were rather limited for our purposes. And we do measure both whole visible spectrum and 3 individual wavelength spectrums, although we do generally publish measurements, like most people, using photopic spectrums.
I'm a near-fanboy about Zeiss in general, but someone there is speaking without knowing what they speak of (meaning our testing methods) which is usually a bad idea. And I note that the writer, as most manufacturers do, talks about their tight tolerances without sharing what those were. Tight is not a number and 100% of tight is still not a number. It's a vague verbal description. I do not mean anything negative towards Zeiss because they and Leica are the only manufacturer that, as far as I know, does actual MTF testing on actual copies. But unless they share their numbers (because until they bypass the laws of physics some variation is going to happen) then we have no idea what 100% of some specification means.
On other topics though, I totally agree that outlier analysis is the best way to measure what's being discussed here and that's NOT what I'm able to do. I just don't have the resources and time to measure 50 to 60 (I don't think 30 is enough) copies of lenses to find those numbers. All I can do is some baby steps in that direction.
And I totally agree - look at the graphs, not the numbers.
Roger
Sep 17, 2015 at 12:49 PM
p.15 #16Zeiss Milvus are out: new range of DSLR lenses!
Roger, thanks for comment and detailing about your procedure. I mainly past information I got from Zeiss in recent conversation about my OTUS repair.
I am holding neutral position about Zeiss's statement and LR's measurement. Good news is lenses like OTUS are indeed went through K8 measurement either from repair or from manufacture based on above statement.
As for lens variation, I now realize this is a very complicated terminology which is highly depend on varies factor like machine used, human error, test methodology as well as the definition. I will not treat LR's results or Zeiss statement as is but as a very good information to know.
From my own experience, I haven't notice big variation on any lens I purchased other than a few Nikon AF lens. (the 24-70mm has very different focus profile and is very hard to compensate with focus adjust. only one out of 3 copy I had perform as is over the zoom range and focus distance and I had a bad copy of 851.8G and exchanged one perform wonderful. I bought a new 24-85G and find it perform horrible. Return it and didn't bother to exchange.)
My experience with high end Leica and Zeiss lens have always been good. OTUS55 indeed show its best f1.4 performance I have seen, and 50lux ASPH also meet my expectation. They all seem perform as MTF suggested. 50lux, 35lux R seems even better than MTF suggested.
I guess it is mainly because I use WO only for central upto 2/3 frame and only time I really care about absolute sharpness cross frame: I have well above f5.6 or f8. And I do believe the variation results are especially sensitive to WO cross frame which usually not very applicable for normal use.
I am holding neutral position about Zeiss's statement and LR's measurement. Good news is lenses like OTUS are indeed went through K8 measurement either from repair or from manufacture based on above statement.
As for lens variation, I now realize this is a very complicated terminology which is highly depend on varies factor like machine used, human error, test methodology as well as the definition. I will not treat LR's results or Zeiss statement as is but as a very good information to know.
From my own experience, I haven't notice big variation on any lens I purchased other than a few Nikon AF lens. (the 24-70mm has very different focus profile and is very hard to compensate with focus adjust. only one out of 3 copy I had perform as is over the zoom range and focus distance and I had a bad copy of 851.8G and exchanged one perform wonderful. I bought a new 24-85G and find it perform horrible. Return it and didn't bother to exchange.)
My experience with high end Leica and Zeiss lens have always been good. OTUS55 indeed show its best f1.4 performance I have seen, and 50lux ASPH also meet my expectation. They all seem perform as MTF suggested. 50lux, 35lux R seems even better than MTF suggested.
I guess it is mainly because I use WO only for central upto 2/3 frame and only time I really care about absolute sharpness cross frame: I have well above f5.6 or f8. And I do believe the variation results are especially sensitive to WO cross frame which usually not very applicable for normal use.
Sep 17, 2015 at 03:10 PM
p.15 #17Zeiss Milvus are out: new range of DSLR lenses!
zhangyue wrote:
Roger, thanks for comment and detailing about your procedure. I mainly past information I got from Zeiss in recent conversation about my OTUS repair.
I am holding neutral position about Zeiss's statement and LR's measurement. Good news is lenses like OTUS are indeed went through K8 measurement either from repair or from manufacture based on above statement.
As for lens variation, I now realize this is a very complicated terminology which is highly depend on varies factor like machine used, human error, test methodology as well as the definition. I will not treat LR's results or Zeiss statement as is but...Show more →
I totally agree with everything you say zhangyue. And I encourage everyone to take every bit of data (mine included) and consider it, but not to think that any one is best or better. All those things you mention do occur and getting a 'ghestalt' of everyone's information is the best way to go.
Roger, thanks for comment and detailing about your procedure. I mainly past information I got from Zeiss in recent conversation about my OTUS repair.
I am holding neutral position about Zeiss's statement and LR's measurement. Good news is lenses like OTUS are indeed went through K8 measurement either from repair or from manufacture based on above statement.
As for lens variation, I now realize this is a very complicated terminology which is highly depend on varies factor like machine used, human error, test methodology as well as the definition. I will not treat LR's results or Zeiss statement as is but...Show more →
I totally agree with everything you say zhangyue. And I encourage everyone to take every bit of data (mine included) and consider it, but not to think that any one is best or better. All those things you mention do occur and getting a 'ghestalt' of everyone's information is the best way to go.
Sep 17, 2015 at 05:39 PM
p.15 #18Zeiss Milvus are out: new range of DSLR lenses!
This thread is awesome.
I'm not saying we were on a yacht dancing with bikini babes in a rap video at the beginning of this thing, but it's gotten EXTRA nerdy toward the end. I love it. You guys are awesome.
The diagrams on Zeiss's new Milvus pages appears to show 5 APD elements in the 21/2.8 (but B&H says 4 and the diagram is pretty small for my bad eyes...). Is that the same as the Z version? I don't remember that.
I'm not saying we were on a yacht dancing with bikini babes in a rap video at the beginning of this thing, but it's gotten EXTRA nerdy toward the end. I love it. You guys are awesome.
The diagrams on Zeiss's new Milvus pages appears to show 5 APD elements in the 21/2.8 (but B&H says 4 and the diagram is pretty small for my bad eyes...). Is that the same as the Z version? I don't remember that.
Sep 17, 2015 at 05:48 PM
p.15 #19Zeiss Milvus are out: new range of DSLR lenses!
RCicala wrote:
And I totally agree - look at the graphs, not the numbers.
Roger
I love your graphs, especially the field curvature diagrams.
And I totally agree - look at the graphs, not the numbers.
Roger
I love your graphs, especially the field curvature diagrams.
Sep 17, 2015 at 05:49 PM
p.15 #20Zeiss Milvus are out: new range of DSLR lenses!
zhangyue wrote:
Roger, thanks for comment and detailing about your procedure. I mainly past information I got from Zeiss in recent conversation about my OTUS repair.
I am holding neutral position about Zeiss's statement and LR's measurement. Good news is lenses like OTUS are indeed went through K8 measurement either from repair or from manufacture based on above statement.
As for lens variation, I now realize this is a very complicated terminology which is highly depend on varies factor like machine used, human error, test methodology as well as the definition. I will not treat LR's results or Zeiss statement as is but...Show more →
I think each of us with 30-40 years of experience in photography, we must be able to distinguish and understand lens performances without the need of MTF graphs. However, I do find that Zeiss MTF have matched each and every Zeiss lens I have owned and used (except the ZA lenses, which MTF are by Sony). I have every reason to trust the MTF. Even though I have not used any Leica glass, from our forum discussions and sample photos, I also find the MTF to be very accurate.
Unfortunately I find the results by LR very inconsistent. Some seem to make sense, and some look to be sensationally different. I know because I have owned or still own some of the offenders. Due to this inconsistency, I stopped looking at their MTF.
Roger, thanks for comment and detailing about your procedure. I mainly past information I got from Zeiss in recent conversation about my OTUS repair.
I am holding neutral position about Zeiss's statement and LR's measurement. Good news is lenses like OTUS are indeed went through K8 measurement either from repair or from manufacture based on above statement.
As for lens variation, I now realize this is a very complicated terminology which is highly depend on varies factor like machine used, human error, test methodology as well as the definition. I will not treat LR's results or Zeiss statement as is but...Show more →
I think each of us with 30-40 years of experience in photography, we must be able to distinguish and understand lens performances without the need of MTF graphs. However, I do find that Zeiss MTF have matched each and every Zeiss lens I have owned and used (except the ZA lenses, which MTF are by Sony). I have every reason to trust the MTF. Even though I have not used any Leica glass, from our forum discussions and sample photos, I also find the MTF to be very accurate.
Unfortunately I find the results by LR very inconsistent. Some seem to make sense, and some look to be sensationally different. I know because I have owned or still own some of the offenders. Due to this inconsistency, I stopped looking at their MTF.
Sep 17, 2015 at 10:39 PM