Is it a mistake to put modern lenses on film cameras?

[Derek L]

It's fairly well-known that lenses made for film cameras suffer on DSLR and mirrorless cameras due to their "sensor stack": a few millimeters of glass mounted to the sensor which refracts incoming off-axis light. This leads to bad sharpness at the edges and corners when vintage lenses, which were not designed to accommodate this stack, are mounted on modern digital cameras.

Here are some links that document this:
https://petapixel.com/2014/06/10/sensor-stack-thickness-matter/
https://blog.kasson.com/the-last-word/kolari-vision-thin-stack-mod-on-a7ii-28mm-elmarit-m/
https://phillipreeve.net/blog/rangefinder-wide-angle-lenses-on-a7-cameras-problems-and-solutions/

Now, it seems obvious (at least once it was pointed out to me by jim10219 in another thread—thank you!) that this phenomenon should work the other way. Modern lenses are presumably corrected for the presence of the stack, so there will be defocus in the corners when using them on film cameras. But Googling reveals that no one has written about this before.

Some questions: is this hunch correct? Does this mean a hybrid shooter who wishes to share lenses between analog and digital cameras necessarily must favor one over the other in sharpness when selecting a lens? And, when I bought a 85mm 1.8G for my Nikon F100 a few weeks ago, would I have been better off buying the AF-D version (designed in the 90s) for the purpose of getting sharp results when shooting film?

Thanks in advance for your comments.

Addendum: for telephoto lenses, I suppose not. After all, the AF-D and AF-S G 85mm MTF charts on a digital camera are basically identical except for some extra center sharpness wide open (which isn't caused by the sensor stack, since it's an off-axis effect), so the sensor stack doesn't seem to matter much there. Please refer to these references:
https://www.opticallimits.com/nikon_ff/717-nikkorafs8518ff?start=1
https://www.opticallimits.com/nikon_ff/622-nikkorafd8518ff?start=1

So for my last question, this is some empirical evidence the answer is "no." But for wide angles I can definitely imagine it matters, as documented in the last of the three links above.

 

[Grim Tuesday]

Interesting question. Looking at the database it doesn't seem like even all Nikon cameras have the same filter stack size. So what did Nikon design the 85mm 1.8g for? If it matters, I have found my 50mm 1.8g to be extremely sharp on film.

 

[mshchem]

I use AF-D lenses on my D5 and D850, I get fabulous results. I don't know about the other way around. I suspect that Pro grade lenses made today would work fine on a F100. The beauty of the older lenses is they work with everything (except the marvelous new mirrorless bodies)

 

[Ariston]

I have had no problems using F mount across any of my fikm or digital bodies. I do have some inferior lenses, but they are inferior regardless of the body. The main concern is functionality. Some of my film cameras require the aperture to be on the collar, so G lenses just aren't going to work.

 

[reddesert]

It's fairly well-known that lenses made for film cameras suffer on DSLR and mirrorless cameras due to their "sensor stack": a few millimeters of glass mounted to the sensor which refracts incoming off-axis light. This leads to bad sharpness at the edges and corners when vintage lenses, which were not designed to accommodate this stack, are mounted on modern digital cameras.

Here are some links that document this:
https://petapixel.com/2014/06/10/sensor-stack-thickness-matter/
https://blog.kasson.com/the-last-word/kolari-vision-thin-stack-mod-on-a7ii-28mm-elmarit-m/
https://phillipreeve.net/blog/rangefinder-wide-angle-lenses-on-a7-cameras-problems-and-solutions/

Now, it seems obvious (at least once it was pointed out to me by jim10219 in another thread—thank you!) that this phenomenon should work the other way. Modern lenses are presumably corrected for the presence of the stack, so there will be defocus in the corners when using them on film cameras. But Googling reveals that no one has written about this before.

Some questions: is this hunch correct? Does this mean a hybrid shooter who wishes to share lenses between analog and digital cameras necessarily must favor one over the other in sharpness when selecting a lens? And, when I bought a 85mm 1.8G for my Nikon F100 a few weeks ago, would I have been better off buying the AF-D version (designed in the 90s) for the purpose of getting sharp results when shooting film?

Thanks in advance for your comments.

Addendum: for telephoto lenses, I suppose not. After all, the AF-D and AF-S G 85mm MTF charts on a digital camera are basically identical except for some extra center sharpness wide open (which isn't caused by the sensor stack, since it's an off-axis effect), so the sensor stack doesn't seem to matter much there. Please refer to these references:
https://www.opticallimits.com/nikon_ff/717-nikkorafs8518ff?start=1
https://www.opticallimits.com/nikon_ff/622-nikkorafd8518ff?start=1

So for my last question, this is some empirical evidence the answer is "no." But for wide angles I can definitely imagine it matters, as documented in the last of the three links above.

Where this will matter:
- Non-retrofocus wide angle lenses exchanged between RF and mirrorless cameras. It's possible that a wide angle designed for digital sensors would have a greater telecentricity (see below) in which case it might do ok on a film camera.

Where it won't matter:
- Retrofocus SLR/DSLR wide angle lenses (these have an exit pupil that is moved forward, by their reverse-telephoto optical design, necessary to clear the mirror). Especially if one stops down to, I don't know, maybe f/5.6.
- Normal and telephoto lenses - the exit pupil is far away from the sensor or film, so non-telecentric effects are small and the converging beam is a relatively narrow angle when it hits the glass and sensor, or film.

The exit pupil is the image of the aperture stop formed by the elements behind the aperture. To get a rough idea of where it is, take the lens off the camera and look through it from behind. It helps if you can stop down a little. Now hold the lens at some point along the barrel and rotate it side-to-side (yaw). When you are rotating it about the location of the exit pupil, the image of the aperture will stay in the same place as you rotate it. When you're holding it in front of or behind the exit pupil, the image of the aperture will move side-to-side. This is easier to do than to describe.

On telecentricity: A perfectly telecentric lens has converging beams to the sensor that are parallel to the optical axis, even for off-axis parts of the sensor. Of course, lenses are not going to be perfectly telecentric. Film isn't strongly dependent on the angle that light strikes it, but digital sensors can be, plus they have a cover glass, Bayer filter, etc in front of the actual sensor.

When the non-telecentric converging beam passes through some thickness of glass at an angle, it will introduce aberrations, most strongly astigmatism, I think. (The glass also causes a focus shift even at the center of the field, but the manufacturers presumably locate the sensor+glass to zero out this focus shift.) There can also be chromatic effects, and the combination of these effects and the Bayer filter is probably one reason for the "purple fringing" on digital cameras even with digital-intended lenses, although that seems to be more of a complaint with compacts.

Finally, my guess, just a guess, is that this problem is worse putting an old WA lens on a digital sensor, and not as bad putting a new WA lens on a film body. Because the designers of the new WA lens would have to design the new lens to either (1) introduce astigmatism that would be canceled out at the corners by the glass filter even for cameras with possibly different glass thicknesses; (2) or design it with greater telecentricity; and it seems like the second choice would be superior. Choice 2 would not affect film badly.

 

[Ko.Fe.]

I'm not Nikonian.
On Canon film EOS any L series EF lens or any recent lens EF lens is just sweet on film.

 

[StepheKoontz]

Modern glass works fine on film. The sensor stack issue only rises when you are trying to use old rangefinder WA lenses on a mirrorless camera. I saw this trying to use a canon 35mm f2 on a Sony A7. Something like a Zeiss 21mm designed for a contax IIa would be a disaster on a digital body. But conversely, these work great on film.

 

[mshchem]

Modern glass works fine on film. The sensor stack issue only rises when you are trying to use old rangefinder WA lenses on a mirrorless camera. I saw this trying to use a canon 35mm f2 on a Sony A7. Something like a Zeiss 21mm designed for a contax IIa would be a disaster on a digital body. But conversely, these work great on film.

This makes sense to me. I'm more than found of nice AF-D Nikon lenses because they work with everything I shoot F to D850. I am not well educated on mirrorless equipment, all the adapters are mind boggling. I love my humble Leica M6ttl, I only have 2 lenses 50mm Summicron and a 35mm Cosina made Zeiss "Biogon" f2.8. Both lenses are amazing.

A couple years back I found a truly mint Nikon AF-I 400 f2.8, first motor in lens Nikkor. Heaviest auto focus lens Nikon ever made ( Ken Rockwell). That lens is amazing, no VR but with a tripod and modern DSLR it's amazing. Oh, and my Nikon F bodies look awesome on it as well.

 

[MattKing]

The biggest problem with using modern lenses designed for digital bodies on film cameras is that the modern lenses are designed with the expectation that the body will offer auto-focus, in many cases image stabilization, and in many cases electronic control rather than mechanical control of aperture.
A lot of film cameras are designed with different operational assumptions.
Optically speaking, the light path to the film is simpler and more straightforward, so if anything, things should be at least as good, if not better.
Just watch out for those lenses that are designed to have their focus calibrated using functions that don't exist in the film bodies.

 

[Adrian Bacon]

It's fairly well-known that lenses made for film cameras suffer on DSLR and mirrorless cameras due to their "sensor stack": a few millimeters of glass mounted to the sensor which refracts incoming off-axis light. This leads to bad sharpness at the edges and corners when vintage lenses, which were not designed to accommodate this stack, are mounted on modern digital cameras.

Here are some links that document this:
https://petapixel.com/2014/06/10/sensor-stack-thickness-matter/
https://blog.kasson.com/the-last-word/kolari-vision-thin-stack-mod-on-a7ii-28mm-elmarit-m/
https://phillipreeve.net/blog/rangefinder-wide-angle-lenses-on-a7-cameras-problems-and-solutions/

Now, it seems obvious (at least once it was pointed out to me by jim10219 in another thread—thank you!) that this phenomenon should work the other way. Modern lenses are presumably corrected for the presence of the stack, so there will be defocus in the corners when using them on film cameras. But Googling reveals that no one has written about this before.

Some questions: is this hunch correct? Does this mean a hybrid shooter who wishes to share lenses between analog and digital cameras necessarily must favor one over the other in sharpness when selecting a lens? And, when I bought a 85mm 1.8G for my Nikon F100 a few weeks ago, would I have been better off buying the AF-D version (designed in the 90s) for the purpose of getting sharp results when shooting film?

Thanks in advance for your comments.

Addendum: for telephoto lenses, I suppose not. After all, the AF-D and AF-S G 85mm MTF charts on a digital camera are basically identical except for some extra center sharpness wide open (which isn't caused by the sensor stack, since it's an off-axis effect), so the sensor stack doesn't seem to matter much there. Please refer to these references:
https://www.opticallimits.com/nikon_ff/717-nikkorafs8518ff?start=1
https://www.opticallimits.com/nikon_ff/622-nikkorafd8518ff?start=1

So for my last question, this is some empirical evidence the answer is "no." But for wide angles I can definitely imagine it matters, as documented in the last of the three links above.

I've used modern lenses on my old film cameras and never noticed anything out of whack.

 

[guangong]

It was my understanding that Zeiss developed and patented the concept for lenses designed for digital lenses, but I was never clear about the differing requirements between film and digital. Very interesting thread.
By the way, Zeiss must get some financial benefit from every digital lens made by whoever. Zeiss is primarily a research and scientific manufacturing institution, so that consumer products make up only a very small part of their output. This of course follows the mandate of the Zeiss Foundation...that financially gains must be used for scientific research.

 

[firemachine69]

This makes sense to me. I'm more than found of nice AF-D Nikon lenses because they work with everything I shoot F to D850. I am not well educated on mirrorless equipment, all the adapters are mind boggling. I love my humble Leica M6ttl, I only have 2 lenses 50mm Summicron and a 35mm Cosina made Zeiss "Biogon" f2.8. Both lenses are amazing.

A couple years back I found a truly mint Nikon AF-I 400 f2.8, first motor in lens Nikkor. Heaviest auto focus lens Nikon ever made ( Ken Rockwell). That lens is amazing, no VR but with a tripod and modern DSLR it's amazing. Oh, and my Nikon F bodies look awesome on it as well.

The only real downside I've found shooting modern lenses is that almost all of the newer "pro" models use an electronic diaphragm (which does not work with any film body).

On the note of the D series, the AF is slower (no big deal), but the minimum focus distance is way, way longer.

 

[mshchem]

The only real downside I've found shooting modern lenses is that almost all of the newer "pro" models use an electronic diaphragm (which does not work with any film body).

On the note of the D series, the AF is slower (no big deal), but the minimum focus distance is way, way longer.

Point on film bodies without motor, you can buy a nice F5 for 300 bucks, but your point is well taken. My experience with focusing speed has not been an issue, but since I don't own a high end new pro nikkor lens I'm no expert.
I have had a couple of really bad experiences with AFS lenses, and Nikon USA 's failure to support.
Minimum focus is an issue, but my lenses are fast fixed aperture lens, not sure how this effects minimum focus.
Best Regards Mike

 

[firemachine69]

Point on film bodies without motor, you can buy a nice F5 for 300 bucks, but your point is well taken. My experience with focusing speed has not been an issue, but since I don't own a high end new pro nikkor lens I'm no expert.
I have had a couple of really bad experiences with AFS lenses, and Nikon USA 's failure to support.
Minimum focus is an issue, but my lenses are fast fixed aperture lens, not sure how this effects minimum focus.
Best Regards Mike

Even the almighty F6, nevermind the F5, cannot focus "E" series lenses. I'm just going through this trying to swap out my Tamron 70-200 G2 F/2.8 (also has a electronic diaphragm) for either a VR1 or VR2 Nikon-equivalent. The newer Nikon 70-200 FL is an "E" type of lens, and not film compatible.

 

[David A. Goldfarb]

Haven’t noticed any problems using new Canon EOS and Zeiss lenses on film SLR bodies.

 

[Derek L]

Where this will matter:
Finally, my guess, just a guess, is that this problem is worse putting an old WA lens on a digital sensor, and not as bad putting a new WA lens on a film body. Because the designers of the new WA lens would have to design the new lens to either (1) introduce astigmatism that would be canceled out at the corners by the glass filter even for cameras with possibly different glass thicknesses; (2) or design it with greater telecentricity; and it seems like the second choice would be superior. Choice 2 would not affect film badly.

Thanks for your post! I learned a lot.

I'm trying to reconcile your comments with the Roger Cicala article I posted. Clearly empirically you are right (cf. the 85mm MTF links I posted), but I'm not really sure why.

All of the Nikon lenses in the article I posted have exit pupil distances between 50mm and 100mm. The 85mm 1.4 AF-S G has an 89mm distance. Let's round up and say 100 is a good estimate for 85mm lenses. Now, consider the following chart in the context of the 85 AF-D versus the 85 AF-S G.

.

Look at the green line. It sure seems, according to Caldwell's model, that at the corners a 2mm sensor stack difference should noticeably decrease the MTF at 40 cycles/mm for lenses with that exit pupil distance at f/2. So I would think that, compared on a digital camera, the AF-D (not designed to accommodate the stack) would suffer greatly compared to the AF-S. Yet in the OpticalLimits testing, the resolution of the lenses was exactly the same in the extreme corners and borders of the image at f/1.8.

Clearly there is something I am misunderstanding about the physics of the situation. What am I missing? (I suppose the formula for MTF is not multiplicative...)

 

[mshchem]

Even the almighty F6, nevermind the F5, cannot focus "E" series lenses. I'm just going through this trying to swap out my Tamron 70-200 G2 F/2.8 (also has a electronic diaphragm) for either a VR1 or VR2 Nikon-equivalent. The newer Nikon 70-200 FL is an "E" type of lens, and not film compatible.

Yes you're right. I forgot about the latest cost reduction.

 

[mshchem]

I remember when "Exit Pupil" meant it was time for recess. My favorite subject

 

[StepheKoontz]

My 70-200 f2.8 VRII works perfect on my F5. In fact my only late lens that has issues is a sigma 150mm f2.8 OS macro. It does wonky things when the OS tries to work and the lens focus basically locks up until you reboot the camera.

 

[reddesert]

Thanks for your post! I learned a lot.

I'm trying to reconcile your comments with the Roger Cicala article I posted. Clearly empirically you are right (cf. the 85mm MTF links I posted), but I'm not really sure why.

All of the Nikon lenses in the article I posted have exit pupil distances between 50mm and 100mm. The 85mm 1.4 AF-S G has an 89mm distance. Let's round up and say 100 is a good estimate for 85mm lenses. Now, consider the following chart in the context of the 85 AF-D versus the 85 AF-S G.

.

Look at the green line. It sure seems, according to Caldwell's model, that at the corners a 2mm sensor stack difference should noticeably decrease the MTF at 40 cycles/mm for lenses with that exit pupil distance at f/2. So I would think that, compared on a digital camera, the AF-D (not designed to accommodate the stack) would suffer greatly compared to the AF-S. Yet in the OpticalLimits testing, the resolution of the lenses was exactly the same in the extreme corners and borders of the image at f/1.8.

Clearly there is something I am misunderstanding about the physics of the situation. What am I missing? (I suppose the formula for MTF is not multiplicative...)

First, we should remember Lens Rule #1: The suitability of a lens is dependent on the pictures one wants to make with it, more so than any quantitative measures such as resolving power, angle of coverage, etc.

So for example, a fast telephoto is often used near wide-open for portraits, because the perspective is flattering and the out-of-focus blur isolates subject from background. Given the intended use and shallow DOF, the resolution in the corners may not matter to the final image, and the corners will often be out of focus anyway. The story is different if one is trying to make wide-angle highly detailed corner to corner scenics (but then, stopping down and using a tripod or some sort of camera support are your friends anyway).

If all assumptions are met, MTF should be multiplicative (part of the reason why it is called a transfer function). So for example if you had a lens with an aerial image MTF of 0.5 at 50 cycles/mm, and film with MTF = 0.8 at 50 cycles/mm, and the film was perfectly flat, the total realized MTF should be 0.5*0.8=0.4. In practice I think it doesn't always quite work out that precisely, because the film doesn't behave the way the MTF-measuring optics do, and so on.

I am not an optical designer nor do I run Zemax, but I do know some optics and work with people who do. The plot that Brian Caldwell posted is output from Zemax (an industry-standard optical modeling program). I think what he did was assume an ideal f/2 conical beam incident on the 2mm glass filter, and calculate the MTF of the image aberrated after it passes through the filter at the appropriate field angle. First, note that all of the lines reach about 0.86 on-axis, not 1.0, so even the on-axis beam is theoretically affected; I think this is because the plate introduces spherical aberration. This and all the other aberrations are a strong function of f-number; it would be interesting to see the same plot for f/4.

Next, note that there are two green lines for the 100mm exit pupil, labeled S2 and T2. He didn't explain this, but these are sagittal and tangential MTF. Sagittal means along the radius from center to edge of field, but somewhat confusingly, AFAIK sagittal MTF means MTF measured using a sagittal line pattern, so it's actually image quality in the tangential direction, and tangential MTF is image quality in the radial direction. (This is super confusing and a good reliable web source on it is http://dougkerr.net/Pumpkin/articles/Astigmatism.pdf , which I recommend if you want to understand astigmatism. Many photo blogs explain this badly.)

The green S2 and T2 curves differ, a lot. S2 is better, meaning the images resolve more sagittal (radial) line pairs, ie a point source would be imaged as an ellipse along the radial direction. (Confusing.) Anyway, that's the normal direction of astigmatism - maybe also some coma. It makes some sense because if you imagine the f/2 cone of rays striking the filter at an angle, within the cone, the different rays along the radial direction have more spread in incident angle than the rays along the tangential direction. The actual image quality is going to be the imperfect image produced by the lens, multiplied by the idealized MTF of this filter, and by the MTF of the sensor. Because the lens and sensor are not ideal, this gives some sense of the reduction in resolution, but it isn't a hard and fast guide.

Regarding the on-sensor measured resolution from the opticallimits.com tests of the 85mm AF-D and AF-S G, they come up with numbers of about MTF50 at 3200 LW/PH (line width per picture height) at field edge. Picture height is 24mm and there are 2 lines/cycle, so that's 67 cycles/mm and the MTF is 0.5. Meanwhile, from Caldwell's Zemax plot, the average of sagittal and tangential MTF at 40 cycles/mm is just below 0.5, and that's without including the lens and sensor, so the real MTF would be lower, and lower still at 67 cycles//mm.

What's going on? Well after many years of different people trying to run lens tests with different methodologies, my feeling about comparing experiments to theory and one experimental number to the other is: Don't get bent out of shape about it. Remember Rule #1. It's more difficult to compare numbers absolutely across different test methods than it seems. Generally, relative measurements within the same method should be easier - rank order in the same on-sensor test likely means something, but if lens A measures 50 cycles/mm in one test and lens B measures 65 cycles/mm in some other test, I don't think it's worth trying to draw conclusions.

You can poke through other test methodologies, like for example if you go to imaging-resource.com and compare the 50/1.4 AF-D film-era lens to the 50/1.4 AF-S G, from https://www.imaging-resource.com/lenses/nikon/reviews/ , click on the full frame measures of their "blur number," which is likely derived from MTF somehow, and there really isn't much difference between the lenses. And the two versions of 50/1.8 have a difference wide open but it's gone by f/2.8.

A lot of dialogue about lens resolution on the interwebs, photo blogs, etc, suffers from what Freud called the narcissism of minor differences. The practical applications of resolution in making photographs should be set by the needs of the photograph. St. Ansel made some brilliantly clear shots, but then he also used large sheet film and stopped down a lot.

Sorry for venting on for so long, but aberrations and MTF are a complex topic that is often poorly represented in the photo-Internet media. If you want to read some more about astigmatism, see: http://dougkerr.net/Pumpkin/articles/Astigmatism.pdf If you would like to see actual formulae and charts for the aberrations introduced by a tilted plane parallel plate, try section XI, starting page 40, of: https://wp.optics.arizona.edu/jcwya...8/03-BasicAberrations_and_Optical_Testing.pdf This is a highly technical text (basically an optical sciences textbook). An interesting conclusion you can reach from Figures 45, 46, and 47, which show spherical, coma, and astigmatism introduced by a tilted plate, is that the aberrations are from 10-40x lower at f/5 than at f/2. At f/5.6, all cats are grey. It's a nice aperture and deserves more love.

 

[Sirius Glass]

First, we should remember Lens Rule #1: The suitability of a lens is dependent on the pictures one wants to make with it, more so than any quantitative measures such as resolving power, angle of coverage, etc.

So for example, a fast telephoto is often used near wide-open for portraits, because the perspective is flattering and the out-of-focus blur isolates subject from background. Given the intended use and shallow DOF, the resolution in the corners may not matter to the final image, and the corners will often be out of focus anyway. The story is different if one is trying to make wide-angle highly detailed corner to corner scenics (but then, stopping down and using a tripod or some sort of camera support are your friends anyway).

If all assumptions are met, MTF should be multiplicative (part of the reason why it is called a transfer function). So for example if you had a lens with an aerial image MTF of 0.5 at 50 cycles/mm, and film with MTF = 0.8 at 50 cycles/mm, and the film was perfectly flat, the total realized MTF should be 0.5*0.8=0.4. In practice I think it doesn't always quite work out that precisely, because the film doesn't behave the way the MTF-measuring optics do, and so on.

I am not an optical designer nor do I run Zemax, but I do know some optics and work with people who do. The plot that Brian Caldwell posted is output from Zemax (an industry-standard optical modeling program). I think what he did was assume an ideal f/2 conical beam incident on the 2mm glass filter, and calculate the MTF of the image aberrated after it passes through the filter at the appropriate field angle. First, note that all of the lines reach about 0.86 on-axis, not 1.0, so even the on-axis beam is theoretically affected; I think this is because the plate introduces spherical aberration. This and all the other aberrations are a strong function of f-number; it would be interesting to see the same plot for f/4.

Next, note that there are two green lines for the 100mm exit pupil, labeled S2 and T2. He didn't explain this, but these are sagittal and tangential MTF. Sagittal means along the radius from center to edge of field, but somewhat confusingly, AFAIK sagittal MTF means MTF measured using a sagittal line pattern, so it's actually image quality in the tangential direction, and tangential MTF is image quality in the radial direction. (This is super confusing and a good reliable web source on it is http://dougkerr.net/Pumpkin/articles/Astigmatism.pdf , which I recommend if you want to understand astigmatism. Many photo blogs explain this badly.)

The green S2 and T2 curves differ, a lot. S2 is better, meaning the images resolve more sagittal (radial) line pairs, ie a point source would be imaged as an ellipse along the radial direction. (Confusing.) Anyway, that's the normal direction of astigmatism - maybe also some coma. It makes some sense because if you imagine the f/2 cone of rays striking the filter at an angle, within the cone, the different rays along the radial direction have more spread in incident angle than the rays along the tangential direction. The actual image quality is going to be the imperfect image produced by the lens, multiplied by the idealized MTF of this filter, and by the MTF of the sensor. Because the lens and sensor are not ideal, this gives some sense of the reduction in resolution, but it isn't a hard and fast guide.

Regarding the on-sensor measured resolution from the opticallimits.com tests of the 85mm AF-D and AF-S G, they come up with numbers of about MTF50 at 3200 LW/PH (line width per picture height) at field edge. Picture height is 24mm and there are 2 lines/cycle, so that's 67 cycles/mm and the MTF is 0.5. Meanwhile, from Caldwell's Zemax plot, the average of sagittal and tangential MTF at 40 cycles/mm is just below 0.5, and that's without including the lens and sensor, so the real MTF would be lower, and lower still at 67 cycles//mm.

What's going on? Well after many years of different people trying to run lens tests with different methodologies, my feeling about comparing experiments to theory and one experimental number to the other is: Don't get bent out of shape about it. Remember Rule #1. It's more difficult to compare numbers absolutely across different test methods than it seems. Generally, relative measurements within the same method should be easier - rank order in the same on-sensor test likely means something, but if lens A measures 50 cycles/mm in one test and lens B measures 65 cycles/mm in some other test, I don't think it's worth trying to draw conclusions.

You can poke through other test methodologies, like for example if you go to imaging-resource.com and compare the 50/1.4 AF-D film-era lens to the 50/1.4 AF-S G, from https://www.imaging-resource.com/lenses/nikon/reviews/ , click on the full frame measures of their "blur number," which is likely derived from MTF somehow, and there really isn't much difference between the lenses. And the two versions of 50/1.8 have a difference wide open but it's gone by f/2.8.

A lot of dialogue about lens resolution on the interwebs, photo blogs, etc, suffers from what Freud called the narcissism of minor differences. The practical applications of resolution in making photographs should be set by the needs of the photograph. St. Ansel made some brilliantly clear shots, but then he also used large sheet film and stopped down a lot.

Sorry for venting on for so long, but aberrations and MTF are a complex topic that is often poorly represented in the photo-Internet media. If you want to read some more about astigmatism, see: http://dougkerr.net/Pumpkin/articles/Astigmatism.pdf If you would like to see actual formulae and charts for the aberrations introduced by a tilted plane parallel plate, try section XI, starting page 40, of: https://wp.optics.arizona.edu/jcwya...8/03-BasicAberrations_and_Optical_Testing.pdf This is a highly technical text (basically an optical sciences textbook). An interesting conclusion you can reach from Figures 45, 46, and 47, which show spherical, coma, and astigmatism introduced by a tilted plate, is that the aberrations are from 10-40x lower at f/5 than at f/2. At f/5.6, all cats are grey. It's a nice aperture and deserves more love.

The short translation: For each plot the higher the line is as it moves from left to right the sharper the lens is. That is a perfect lens would have MTF plots straight and level at the height it started all the way across. For most people that is enough for most people to understand.

This is the 80mm "Normal" Hasselblad lens This is the 100mm Hasselblad lens

One can tell looking at the plots that the 100mm lens is much sharper across the whole field of view.

Also each lens manufacturer uses its own proprietary techniques so one cannot use the plot to compare between manufacturers.

 

[Sirius Glass]

For comparison this is the 30mm Fisheye Hasselblad lens and the SWC.

 

[markbau]

The sharpest lens/camera combo I’ve ever had is using my Pentax 67 lenses on a Pentax K-x digital body.

 

[Derek L]

What's going on? Well after many years of different people trying to run lens tests with different methodologies, my feeling about comparing experiments to theory and one experimental number to the other is: Don't get bent out of shape about it. Remember Rule #1. It's more difficult to compare numbers absolutely across different test methods than it seems. Generally, relative measurements within the same method should be easier - rank order in the same on-sensor test likely means something, but if lens A measures 50 cycles/mm in one test and lens B measures 65 cycles/mm in some other test, I don't think it's worth trying to draw conclusions.

Thank you. I read your post in detail a few times and tried to think things through. It contains a lot of juicy facts. Yet, I remain unsatisfied.

What bothers me is that the the discrepancy between experiment and theory here is *huge*. It's not some rounding error, it's not a measurement error, it's an indication that—and this is really the only explanation I can offer—Caldwell's graph has to be mistaken.

Returning to the 85mm example, he predicts the the 2mm cover glass will reduce the sagittal and tangential MTF by about 1/3 and 2/3 in the corners, respectively, at the appropriate cycles/mm. According to the numbers you gave (converting cycles to the line pair measurement used in my link), this should do horrible, horrible things to the AF-D 85/1.8. It should be, say, half as as good as the AF-S in the corners at f/2. Yet it's basically the same.

There is something absolutely fundamental here I am missing and it's driving me nuts.

 

[benveniste]

I can't use many modern lenses on my Nikon FA due to the lack of an aperture ring and I own at least three lenses which are incompatible with my F100 (Sigma 14-24mm, Sigma 100-400mm, Tamron 24-70mm VC), But those lenses which I have used on the F100 have performed very well. To be honest, if I'm looking for the last quantum of sharpness out of my gear, I'm not using 35mm film.