Effects of diffraction

[ErikKaffehr]

Hi,

I made a quick check of the effect of diffraction using three different generation of cameras.

The cameras were:

1) Minolta Dimage 7D, a 6 MP APS-C camera
2) Sony Alpha 700, a 12 MP APS-C camera
3) Sony Alpha 77 SLT, a 24 MP APS-C camera

The lens used for all of the tests a Minolta 100/2.8 Macro. What the diagrams shows is the number of line withs per picture height each camera/lens can achieve at 50% MTF. The images were processed in Lightroom 4 without sharpening.

The Alpha 77-SLT performed worse than expected so the test was remade using the Sony Alpha 16-80/3.5-4.5 ZA lens I mostly use on that camera. The 16-80 outperformed the 100/2.8 macro.

What I see is:

- All combos peak around f/5.6
- Resolution improves with pixel counts
- The Alpha 77SLT may be limited by lens


Best regards
Erik

[bjanes]

Hi,

I made a quick check of the effect of diffraction using three different generation of cameras.

The cameras were:

1) Minolta Dimage 7D, a 6 MP APS-C camera
2) Sony Alpha 700, a 12 MP APS-C camera
3) Sony Alpha 77 SLT, a 24 MP APS-C camera

The lens used for all of the tests a Minolta 100/2.8 Macro. What the diagrams shows is the number of line withs per picture height each camera/lens can achieve at 50% MTF. The images were processed in Lightroom 4 without sharpening.

The Alpha 77-SLT performed worse than expected so the test was remade using the Sony Alpha 16-80/3.5-4.5 ZA lens I mostly use on that camera. The 16-80 outperformed the 100/2.8 macro.

What I see is:

- All combos peak around f/5.6
- Resolution improves with pixel counts
- The Alpha 77SLT may be limited by lens

Erik,

Your results are consistent with the thesis of Bobn2 posted on DPReview. The thread starts here, and Bobn2's most important replies are here and here.

Regards,

Bill

[BJL]

The main conclusion I see in these results (both from Erik's 100/2.8 and from bob2n@dpreview) is that even if sensor resolution is so high as to be irrelevant, the balance of lens aberrations against diffraction gives optimal resolution between f/4 and f/5.6, so that varying the sensor resolution doe not shift the balance much, but most just lowers resolution across the board. It is a fallacy to think that optimal f-stop will be roughly where diffraction matches the Nyquist limit, partly because that argument ignores the third leg of the stool: aberrations. And by the way, without aberrations, optimal resolution with any sensor will always be at minimum f-stop: sensor resolute will only affect how much and when the resolution curve flattens out as f-stop decreases.

However, I note that increasing sensor resolution does seem to move the optimal f-stop a bit to the left. Both Erik's data for there 100/2.8 and bob2n's show some shift from f/5.6 towards f/4. For example, the comparison of the adjacent f-stops, f/4 and f/8, goes from about equal with the 6MP camera to better at f/4 than f/8 with either of the higher resolution cameras.

Because of the importance of aberrations, I avoid complications by looking only at the one lens, not Erik's data for the 16-80/3.5-4.5.


P. S. The excellent results for that 16-80/3.5-4.5 5x zoom lens give a fine counterexample to a whole bunch of lens snob dogma:
- primes are always better than zooms
- lenses of wider zoom range area always inferior to lenses of lesser zoom range (like the 3x limit of constant f/2.8 zooms)
- constant minimum f-stop zooms are always better than variable minimum f-stop zooms
- fast lenses area always better than slower lenses when compared at equal f-stop, and the related idea that
- all lenses are optimal about two stops down from wide-open.

A slow, variable minimum f-stop, wide ranging zoom beating a macro prime, even wide-open: fancy that!

On the other hand, the results does not harm the lens snob dogma that Zeiss designs/makes some great lenses!

[telyt]

P. S. The excellent results for that 16-80/3.5-4.5 5x zoom lens give a fine counterexample to a whole bunch of lens snob dogma:
- primes are always better than zooms
- lenses of wider zoom range area always inferior to lenses of lesser zoom range (like the 3x limit of constant f/2.8 zooms)
- constant minimum f-stop zooms are always better than variable minimum f-stop zooms
- fast lenses area always better than slower lenses when compared at equal f-stop, and the related idea that
- all lenses are optimal about two stops down from wide-open.

I've never met anone who believes this dogma.  Have you?

[Bart_van_der_Wolf]

Hi,

I made a quick check of the effect of diffraction using three different generation of cameras.

The cameras were:

1) Minolta Dimage 7D, a 6 MP APS-C camera
2) Sony Alpha 700, a 12 MP APS-C camera
3) Sony Alpha 77 SLT, a 24 MP APS-C camera

The lens used for all of the tests a Minolta 100/2.8 Macro. What the diagrams shows is the number of line withs per picture height each camera/lens can achieve at 50% MTF. The images were processed in Lightroom 4 without sharpening.

Hi Erik,

Thanks for the test. There may be a bit of a problem interpreting the results for those who are not experienced enough. The point is that the results are influenced by the increased number of sensels, because the results are expressed in LW/PH. A sensor with more (denser input or more) output pixels will automatically score higher! Oversampling is beneficial for same size output comparisons, which is basically what LW/PH does, it normalizes for the same size output (picture height).

When on the other hand you express the results as cycles/mm, then you will be able to calculate the output size differences (the magnification potential). Assuming that 5-8 lp/mm (~cycles/mm) is considered to be good quality output at normal reading distance at adequate light levels, you divide the limiting sensor resolution in cy/mm by the required output resolution, and the magnification factor is the result. You then multiply the sensor dimensions with the magnification factor, and you'll get the output size with identical resolution (although the contrast may need some help to better match the human contrast sensitivity peak). The question then becomes, does the linear increase of vertical sensels still translate to proportional output size differences?

The Alpha 77-SLT performed worse than expected so the test was remade using the Sony Alpha 16-80/3.5-4.5 ZA lens I mostly use on that camera. The 16-80 outperformed the 100/2.8 macro.

While that introduces a variable into the sensor evaluation, it's useful to know that better quality can be achieved, but it doesn't translate to the original goal.

What I see is:

- All combos peak around f/5.6

The aberration versus diffraction trade-off didn't change because same size output was compared. The aberrations were sampled at different densities, therefore the sensors with higher sampling density offer somewhat better (deconvolution) restoration potential, but at the pixel level microcontrast will suffer, depending on the aperture.

- Resolution improves with pixel counts

Yes, especially when expressed with a same size output metric. One could also consider the extra pixels as larger output potential with the same resolution. Besides, f/5.6 is not stressing all sensors the same way at the per pixel resolution. Something like f/18 to f/20 will show what diffraction limited means, and how denser sampling fares.

Cheers,
Bart

[BJL]

I've never met anone who believes this dogma.  Have you?

Only on the internet! There, every one of these over-simplified views is often expressed.

[telyt]

Only on the internet! There, every one of these over-simplified views is often expressed.

I see them expressed most often by those who feel obliged refute them, not by those who believe them.

[BJL]

I see them expressed most often by those who feel obliged refute them, not by those who believe them.

Are you serious? Have you not seen hundreds of posts from people proudly declaring that they only use primes, or that they insist on constant aperture zooms rather than variable aperture zooms because "that's the way all professional zoom lenses are", or bemoaning the fact that camera system X has too many slow, variable aperture zooms and not enough fast, constant aperture zooms and fast primes, or people stating sincerely the old truism that most lenses are sharpest about two stops down from wide-open?

Then you must read quite different forums that I do: please let me know where these wiser forums are!

[ErikKaffehr]

Hi Bill,

The Macro lens I used is quite old and may not be a very good performer by todays standards. The results were not really what I expected, previous testing I have done indicated that resolution increased linearly against sqrt(megapixels) and here it was obviously not the case. So I assumed that lens may be a limiting factor.

So I added the 16-80/3.5-4.5 which has served me well. We should be careful with conclusion, tough, the lens was used about 50 mm which may be close to it's optimum and the image was near the optical axis. The 16-80 performs well in it's reasonable sized "sweet spot" but has awful corners at full aperture, especially in the wide end.

Just to point out, this is not a lens test. The idea was just to demonstrate the effects of diffraction. A lens test would also check edges and corners and other parameters that may affect image quality.

Regrding "Zeissness" it is built by Sony but it has a Zeiss label. It had also a signed QA inspection certificate (without MTF data) from a japanese Zeiss employee and also has double serial numbers.

It seems that Zeiss by and large ignores the corners and strive for high MTF at a reasonably large sweet spot. It seems that most Zeiss lenses share the characteristics.

Best regards
Erik

The main conclusion I see in these results (both from Erik's 100/2.8 and from bob2n@dpreview) is that even if sensor resolution is so high as to be irrelevant, the balance of lens aberrations against diffraction gives optimal resolution between f/4 and f/5.6, so that varying the sensor resolution doe not shift the balance much, but most just lowers resolution across the board. It is a fallacy to think that optimal f-stop will be roughly where diffraction matches the Nyquist limit, partly because that argument ignores the third leg of the stool: aberrations. And by the way, without aberrations, optimal resolution with any sensor will always be at minimum f-stop: sensor resolute will only affect how much and when the resolution curve flattens out as f-stop decreases.

However, I note that increasing sensor resolution does seem to move the optimal f-stop a bit to the left. Both Erik's data for there 100/2.8 and bob2n's show some shift from f/5.6 towards f/4. For example, the comparison of the adjacent f-stops, f/4 and f/8, goes from about equal with the 6MP camera to better at f/4 than f/8 with either of the higher resolution cameras.

Because of the importance of aberrations, I avoid complications by looking only at the one lens, not Erik's data for the 16-80/3.5-4.5.


P. S. The excellent results for that 16-80/3.5-4.5 5x zoom lens give a fine counterexample to a whole bunch of lens snob dogma:
- primes are always better than zooms
- lenses of wider zoom range area always inferior to lenses of lesser zoom range (like the 3x limit of constant f/2.8 zooms)
- constant minimum f-stop zooms are always better than variable minimum f-stop zooms
- fast lenses area always better than slower lenses when compared at equal f-stop, and the related idea that
- all lenses are optimal about two stops down from wide-open.

A slow, variable minimum f-stop, wide ranging zoom beating a macro prime, even wide-open: fancy that!

On the other hand, the results does not harm the lens snob dogma that Zeiss designs/makes some great lenses!

[Fine_Art]

Are you serious? Have you not seen hundreds of posts from people proudly declaring that they only use primes, or that they insist on constant aperture zooms rather than variable aperture zooms because "that's the way all professional zoom lenses are", or bemoaning the fact that camera system X has too many slow, variable aperture zooms and not enough fast, constant aperture zooms and fast primes, or people stating sincerely the old truism that most lenses are sharpest about two stops down from wide-open?

Then you must read quite different forums that I do: please let me know where these wiser forums are!

I only use primes. The other points are bunk. The advantages of primes should be clear:
The designer can optimize with fewer variables
There are fewer elements to cause internal reflections.

The fact that Zeiss can make a great zoom, especially autofocus which they are on Sony, speaks to the skill of the designers, their Schott glass (which is how they made a name for themselves as microscope lens makers) and the quality of their coatings.

[Fine_Art]

Lets put it another way:

The advantage of the zoom, is mostly to the manufacturer, that gets to sell a far more complicated device for more money.

[ErikKaffehr]

Hi,

Zooms have the advantage that you can crop exactly where you stand. I don't mind walking around quite a bit, but many times I find that I cannot move forward or backward. Moving the vantage point would also change perspective.

Obviously, it is much easier to make a good prime than making an equally good zoom lens. On the other hand, essentially any decent lens will be close to diffraction limit at f/8 or f/11. Primes may excel at large apertures, but I'm essentially an f/8 type of kind of guy. The issues you mention are quite valid, I wouldn't argue.

Regarding glass catalogues I don't think Zeiss has a large advantage. Schott is selling to any competitor willing to buy and there are other foundries offering similar lens catalogues.

The impression I have that most Zeiss designs are a bit on the conservative side. Zeiss tries to use designs that don't require the ultimate precision in assembly. Also, Zeiss sets a tighter limit on sample variation than most other vendors.

Check this:
 
from this excellent article: http://www.lensrentals.com/blog/2011/10/notes-on-lens-and-camera-variation

What the image above indicates that the Canon macro is actually outperforming the Zeiss Macro, if you find a good sample. With the Zeiss macro there is much less variation. But most samples of the Canon macro seem to be better than the best sample of the Zeiss macro. The 100/2.0 Macro Planar is a very highly regarded lens.


Best regards
Erik

I only use primes. The other points are bunk. The advantages of primes should be clear:
The designer can optimize with fewer variables
There are fewer elements to cause internal reflections.

The fact that Zeiss can make a great zoom, especially autofocus which they are on Sony, speaks to the skill of the designers, their Schott glass (which is how they made a name for themselves as microscope lens makers) and the quality of their coatings.

[Petrus]

Lets put it another way:

The advantage of the zoom, is mostly to the manufacturer, that gets to sell a far more complicated device for more money.

Let's put it my way: the advantage of zoom is that I need most of the time only one body and lens around my neck, the rest of the kit is in a backpack. In the good old times of primes I always had at least two cameras ready with different lenses, sometimes three, and the camera bag had to be of the shoulder type causing back pain, because I had to be able change lenses on the run. I have been working as a press photographer since -78.

There would be no backpack type camera bags if there were no zooms.

[Fine_Art]

I was speaking historically. Zeiss got its reputation back in the first half of the 1900s by having a superior glass that was not available to other companies. It doesn't matter any more because the widely available low dispersion glasses are better than that old mix. Flourite is the best and also the softest so it is not used that much.

[BJL]

The advantage of the zoom, is mostly to the manufacturer, that gets to sell a far more complicated device for more money.

That would be true if the manufacturers were dealing with a market dominated by photographers for whom one prime lens is as useful as one zoom lens, and who would buy and carry a roughly equal number of lenses, whether all primes or all zooms or some mix. But the lens market is not like that overall: for the great majority of lens buyers, a zoom lens replaces several prime lenses, and leads to spending less money on a smaller and lighter kit.

[NancyP]

Fluorite is still used in microscopy optics designs at times.

[telyt]

Are you serious? Have you not seen hundreds of posts from people proudly declaring that they only use primes, or that they insist on constant aperture zooms rather than variable aperture zooms because "that's the way all professional zoom lenses are", or bemoaning the fact that camera system X has too many slow, variable aperture zooms and not enough fast, constant aperture zooms and fast primes, or people stating sincerely the old truism that most lenses are sharpest about two stops down from wide-open?

Then you must read quite different forums that I do: please let me know where these wiser forums are!

The dogmas you've created in order to debunk them are expressed as absolutes.  Where are you reading these absolute dogmas?  On the forums I read these performance tendancies are expressed as 'more likely', not as 'always'.

As Erik wrote, this test was primarily to show the effect of diffraction.  I don't see much point in drawing sweeping conclusions about percieved internet dogmas from a small sample of lenses of very different ages.

[Rob C]

Primes, zooms, anything that pleases you is just fine.

For myself, it boils down to native suspicion, a bit of common sense and a lot of experience with primes. Primes, as has been pointed out, obviously permit more fine-tuning at their focal length than will any zoom - a zoom has to be a compromise at every length with preferences at the maker's disposition. Primes, if you shoot enough, let you get to know them intimately, so well that you automatically know where to stand in order to get the field of view that you want; you don't even consciously think about it, you just do it. You also learn when you would be better off using a longer or shorter objective than the one on the camera. There isn't the temptation to be lazy.

Also, if you are a pro, you know what the job is going to require before you set to work. You simply take the stuff with you that the work requires. The idea that you might say damn! I should have brought this, that or the other lens instead, is a non-starter. You think the job through beforehand, even if you haven't been cursed with a layout you still know how you envisage the thing. Okay, long trips are a different matter, and I, too, would take pretty much even the kitchen sink; day jobs are not like that, in my experience, YMMD.

Where I do see the appeal of a zoom is for the amateur blessed with time, money and an open mind - or even an empty one - regarding what he is going to do. Maybe he shoots lochs or fiords from a boat and has no control of where that's going to sail; maybe he just doesn't want to spend more money on his hobby than the kit lens that came with the camera - why should he feel obliged to stock up on expensive primes?

So, long experience in primes but what about zooms? I never did feel attracted and no respected reviewer of my period, that I'd read, had any great things to say about tested zooms re. pro photography. Then a couple of years ago, feeling my age and wanting to be abe to have a single camera/lens combination for walkabout shooting, I bought a 2.8/24-70 Nikkor G. Unfortunately, there were and still are no places here that I can find such lenses and try 'em out pre-purchase. So, I ordered the thing and when it arrived, I realised at once that it was a mistake. It was friggin' huge! A walkabout camera/lens combination, did I say? Anyway, I tried it out, on the massive Gitzo, using the D200 that was all I had at that time. It sucked, even on that cropped format! It went back and I had to spend more money buying something else I didn't really need, but that was my way out of the mistake.

But, that's just my experience. I know of other photographers, very good pros, whom use zooms much of the time. Hey ho.

Rob C

[BJL]

The dogmas you've created in order to debunk them are expressed as absolutes.  Where are you reading these absolute dogmas?  On the forums I read these performance tendancies are expressed as 'more likely', not as 'always'.

As Erik wrote, this test was primarily to show the effect of diffraction.  I don't see much point in drawing sweeping conclusions about percieved internet dogmas from a small sample of lenses of very different ages.

My comment was not at all about Erik's data in particular, which just happened to juxtapose a new and quite good zoom lens against an older and no so great macro prime lens.  So certainly I was going off on a tangent from Erik's subject.

Also, please try to see past my rhetorical device of putting in the explicit word "always" to emphasize the tone and style in which many participants in many forums implicitly assume when arguing on the basis of such truisms. Hopefully you know that what I am referring to is people who mindlessly and persistently make oversimplified and unqualified statements like "primes are better than zooms" or "constant aperture zooms are better than variable aperture zooms, as shown by the fact that all pro grade zooms are constant aperture" or "lenses perform best two stops down from wide-open" ... and then offer these simplistic ideas as sufficient arguments for things like disparaging a lens solely because, for example, it is of variable maximum aperture. (Including the funniest case of numerous people stating a preference for any constant aperture zoom design (including f/4) over any variable aperture design (including f/2.8-4 or f/2.8-3.5) due to the inherently inferior "non-pro" status of any variable minimum f-stop zoom lens!)

Not that any of the wise photographic or optical experts in _this_ forum would make such mistakes!