CAs at 1x in High-Pedigree Macro Lenses

[ColSebastianMoran]

Looking for insight on this question: Appears to me that lens designers have made a tradeoff to leave uncorrected CA in some expensive high performing lenses. In what area are they gaining by this tradeoff?

Data: In several high pedigree-macro lenses, testing at 1x for camera-scanning, I see significant CA (color fringing) at edges/corners, along with good to excellent resolution. For example, this is very evident in the Canon 100 f/2.8 L Macro, as shown here.

Here the lesser Canon 50 f/3.5 FD Macro on left has better resolution and no visible CA. Canon 100 f/2.8 L macro on right has good resolution and lots of CA. USAF Glass target, shot at 1x to 50MPx sensor, well off axis, 400% screen grabs:

Here is the 100 L again on Vlad's test target, 1x, 50MPx sensor, extreme FF corner, 400% screen grab:

Virtually the same for the Canon 100mm f/2.8 Macro non-L version, also highly regarded. Also, surprisingly in the highly regarded 60mm f/2.8 Zeiss Macro (C/Y). These were not and are not economy lenses!

This, while other very high performing macros I've tested are quite free of these CAs, for example the more modest 50 f/3.5 Canon FD Macro (left side in first image), the 55 f/2.8 Micro Nikkor, and every Sigma macro I've tried.

So, question is what are the designers gaining by leaving these CAs?

 

[ColSebastianMoran]

Anyone have insight on this? What's the lens designer getting in the tradeoff that results in this CA?

 

[retinax]

Just a wild guess: Maybe they perform better at lower magnifications? Not every macro has to be optimized for 1:1.

 

[Rob-F]

Looking for insight on this question: Appears to me that lens designers have made a tradeoff to leave uncorrected CA in some expensive high performing lenses. In what area are they gaining by this tradeoff?

Data: In several high pedigree-macro lenses, testing at 1x for camera-scanning, I see significant CA (color fringing) at edges/corners, along with good to excellent resolution. For example, this is very evident in the Canon 100 f/2.8 L Macro, as shown here.

Here the lesser Canon 50 f/3.5 FD Macro on left has better resolution and no visible CA. Canon 100 f/2.8 L macro on right has good resolution and lots of CA. USAF Glass target, shot at 1x to 50MPx sensor, well off axis, 400% screen grabs:

I don't really see any difference in resolution between these two.
Of course, I can see the CA.

 

[Freakscene]

Anyone have insight on this? What's the lens designer getting in the tradeoff that results in this CA?

What the designers got was not having to include extra low dispersion glass or floating elements. These are lenses that were highly regarded in their day but are now 50+ year old designs. The Zeiss is optimised for 1:2-1:3 - the C version needs an extension tube to focus to 1:1 and beyond 3m it is pretty ordinary (but the 50/1.4 Planar is exceptional at distance but not close up, so you win some . . . especially if you are Zeiss and picky users need two lenses not one).

Everything is built to a price, even expensive things.

Marty

 

[Chriscrawfordphoto]

What the designers got was not having to include extra low dispersion glass or floating elements. These are lenses that were highly regarded in their day but are now 50+ year old designs. The Zeiss is optimised for 1:2-1:3 - the C version needs an extension tube to focus to 1:1 and beyond 3m it is pretty ordinary (but the 50/1.4 Planar is exceptional at distance but not close up, so you win some . . . especially if you are Zeiss and picky users need two lenses not one).

Everything is built to a price, even expensive things.

Marty

The Canon EF 100mm f2.8L is not a 50 year old design. It is a very recent design. The lenses that gave less CA in ColSebastianMoran's tests were the old designs!

 

[retinax]

The Canon EF 100mm f2.8L is not a 50 year old design. It is a very recent design. The lenses that gave less CA in ColSebastianMoran's tests were the old designs!

If it's a digital era design, CA correction might simply not have been considered necessary because it's easy to correct for digitally?

 

[Freakscene]

The Canon EF 100mm f2.8L is not a 50 year old design. It is a very recent design. The lenses that gave less CA in ColSebastianMoran's tests were the old designs!

I assumed it was the FD macro. If it is the EF 100mm L and it has that much CA it might be decentred. So maybe what they got was cheaper assembly. But Canon changes things a lot. The 100 macro in my lab doesn’t have anything like that much visible CA at 1:1.

Marty

 

[ColSebastianMoran]

I assumed it was the FD macro. If it is the EF 100mm L and it has that much CA it might be decentred. So maybe what they got was cheaper assembly. But Canon changes things a lot. The 100 macro in my lab doesn’t have anything like that much visible CA at 1:1.

Marty

Marty, thanks for the comment. No, it's the 100L that shows the CA, the 50 FD macro is clean.

Before writing, I checked with another photographer who sent me a sample from his 100 L macro that looks exactly like mine.

 

[ColSebastianMoran]

Couple clarifications:

1. Yes, I'm sure that both the Zeiss 60 and the Canon 100 L perform very well at lesser magnifications. 1x is a very tough challenge for lenses.

2. At least for the Canon, "Remove CA" in the Lens Corrections panel does a good job with the CA without loss of resolution.

3. Rob-F commented "Don't see difference in resolution...". Yeah, little difference, but I think Group 6 Element 4 is a little better resolved by the older 50 FD lens.

 

[Darthfeeble]

Could their claim of excellence be somewhat fraudulent??? Gee! Marketing and Accounting rule the world anymore.

 

[Larry Cloetta]

FWIW, on paper at least, the new Nikon Z S-mount 50 and 100 macro lenses seem to be good, perhaps even a step ahead, at both resolution and absence of CA at 1:1. There are tons of tests out there, which mirror this one: https://www.ephotozine.com/article/nikon-nikkor-z-mc-50mm-f-2-8-macro-review-35518/performance
“ CA (Chromatic Aberration) is not a problem, approaching zero throughout the range, both centre and edge. It would be difficult to find any circumstances where it would need any further correction.”

That test was on the 50mm. The 100mm is supposedly even better. Another optical advantage of a larger mount. Nikon’s new Z mount lenses can be fairly criticized for lacking “character” but in terms of objective criteria like resolution and freedom from aberrations they are undeniable advances. (Yes, they’re big. Leica L mount lenses are also big. Peter Karbe acknowledges that the M mount puts limits on optical design that the L mount sidesteps, so the L mount lenses are “better” objectively. Physics is a b…h.)

So, maybe these would be a good option for camera duplication of negatives.

 

[Godfrey]

Many years ago, when dinosaurs still roamed the earth and I was working at NASA/JPL, we needed a few Nikkor lenses for optical ground truth photos of the area the instrument we were flying was covering. To get six 105mm f/2.5 AI-S lenses of matching high-performance quality, we had Nikon send us 20 lenses and we set them up on a test jig.

Out of 20 examples, we managed to get four that were superbly centered and with matching corner/edge quality, and two more that were a little off but still within the range of acceptable we'd set. The rest showed too many aberrations for our needs, although all of them would be fine for non-forensic photographic work.

The moral of the story: Unit by unit sample differences abound. Canon certainly didn't intend that 100mm lens to have so much CA. Luckily, lateral CA like that is very easy to correct in post processing, where longitudinal CA is not.

The other moral: It's an imperfect world. 😉

G

 

[AlwaysOnAuto]

When I got interested in macro, I researched the Nikon lenses mostly, being a Nikon shooter at the time. I read where Nikon worked really hard at making the 55 2.8 a good document replicator. I'm glad to read they are still in there with the other lenses as a good choice. I don't look at my pictures under that magnification so it really wouldn't matter much to me. But it is nice to know they aren't afflicted with the CA those other lenses seem to be.

 

[ColSebastianMoran]

When I got interested in macro, I researched the Nikon lenses mostly, being a Nikon shooter at the time. I read where Nikon worked really hard at making the 55 2.8 a good document replicator. I'm glad to read they are still in there with the other lenses as a good choice. I don't look at my pictures under that magnification so it really wouldn't matter much to me. But it is nice to know they aren't afflicted with the CA those other lenses seem to be.

I've tested probably four dozen lenses at 1x, using a glass USAF target with perfect lines to 200+ lp/mm, looking at resolution and contrast well off axis, to a 55MPx sensor (3.8µ pixel pitch).

With lesser sensor resolution, it can be hard to see differences, but with this sensor, ten lenses stand out above the rest. The 55 f/2.8 Micro Nikkor is one of these.

Interestingly, all ten results look the same at 400% pixel-peeping. This suggests to me that the sensor is probably limiting these ten lenses, some may be better than others, but I cannot tell. Not till a few years from now when we have 100MPx sensors.

Some lenses in that top group:
55 f/2.8 Micro Nikkor.
70 f/2.8 Sigma Macro ART, a current AF lens, $500
50 f/2.8 Sigma EX DG Macro, discontinued but findable in many mounts.
70 f/4 APO Rodagon-D 1x copy lens.
50 f/4.5 Tominon, but not the other Tominon MP-4 lenses, a bargain.
90 f/2.8 Sony Macro, $1000

 

[ColSebastianMoran]

One more puzzle about this lens:
- On the 5DSR, horizontal lines are less resolved than vertical lines. This is throughout the frame, center, edges, and corners, so it's not sagittal vs meridional rays.
- When I mount the same lens on my Sony A7Rii, this doesn't happen.

What in the 5DSR would result is less resolution of horizontal lines?

Example: at near center, 400% screen grabs:

Note: by "horizontal" I mean horizontal with the full image in landscape orientation. As you can see from the Sony shot, this is not an artifact in the test negative; lines in both directions are pretty good.

 

[Godfrey]

From these photos, it looks like you're running into a decent vs poor sensor match to the lens problem.

Because sensors are complex devices with many layers and optics themselves, getting a top notch sensor match when making such critical comparisons is just as important as whether the lens itself performs well.

I'd fit the lens to a film body and do some tests, scan the results, to see what the lens itself actually performs like independent of the particular camera/sensor you're using it with. Then decide whether it is worth keeping to use with whatever camera/sensor setup you have in mind.

I've found discrepancies like this often. It's useful to have a film body around to remove the sensor dynamics when doing lens tests...

G

 

[ColSebastianMoran]

More about my comparison in #16...

This lens on the Sony body is at f/2.8; because lens has electronic aperture control, the adapter cannot adjust aperture. This is pretty amazing performance wide open at center. Not shown, but also pretty good at periphery.

I'm now guessing:
- They traded off CA (correctable) for center performance wide open.
- Less resolution of horizontal lines is due to the AA-Undo filter added on top of the AA filter. This is pure guess.
- Because better on Sony, I suspect this lens will be excellent on any body that doesn't use the AA-Undo filter.

And, I conclude:
- Because adapter won't control aperture, this lens cannot be easily adapted to non-Canon-EOS bodies.

 

[AE_Conrady]

Some hints: Perhaps Canon did trade anything for CA but decided to optimise the lens for a different magnification factor and optimal aperture, which you couldn't test. Older macro lens ("FD") are more optically robust, because of lacking the Internal Focusing which seems mandatory for today's digital macro lens. F2.8 against F3.5 doesn't look like an impressive step here but if one take into account the different focal length (2x), residual SA maybe increased by the same factor. Another item is that you always pair lens with a sensor and you can't seperate the result from each other.

 

[ColSebastianMoran]

Some hints: Perhaps Canon did trade anything for CA but decided to optimise the lens for a different magnification factor and optimal aperture, which you couldn't test. Older macro lens ("FD") are more optically robust, because of lacking the Internal Focusing which seems mandatory for today's digital macro lens. F2.8 against F3.5 doesn't look like an impressive step here but if one take into account the different focal length (2x), residual SA maybe increased by the same factor. Another item is that you always pair lens with a sensor and you can't seperate the result from each other.

Thanks, and I agree. But, what surprises me is the lesser (but still not too bad) performance on the obvious Canon super-resolution body, the 5DSR, versus at f/2.8 on the Sony body. Mystery.