Understanding Lens Diffraction

[Bart_van_der_Wolf]

Bart,

Thanks for the clarification and your method of taking an additional security shot at a smaller aperture.

BTW, did you see my previous post in this thread using your sinusoidal Siemens star. Your comments would be appreciated.

https://forum.luminous-landscape.com/index.php?topic=130733.msg1111937#msg1111937

Hi Bill,

I have not that much to add. You've demonstrated the effect of diffraction in the plane of best focus well. You've also shown that deconvolution sharpening will restore sharpness where there is some left to restore, but there is also a loss of resolution that is unrecoverable.

Your D850 should indeed exhibit a diffraction-limited MTF (for Green wavelengths) when using f/16 (or narrower). Red wavelengths already at approx. f/11, Blue wavelengths at f/22, but Blue and Red contribute much less to luminance which is what the human visual system is more sensitive to.

Obviously, defocus, or DOF, also has a significant effect on resolution, so we usually need to seek a compromise. That is, unless we can resort to techniques such as focus stacking, or make use of hyperfocal distance focusing.

Good photographers have a skillset they can choose from. It's not only about "creativity is everything, and technique be damned". If we have a choice, the combination of creativity AND technique delivers better results in conveying our creative intent (without distractions/limitations caused by poor technique).

Cheers,
Bart

[JaapD]

Hi Bart,

When calculating diffraction-limited MTF should we not take the Bayer demosaicing into account; not taking the actual pixel pitch into the equation but the distance between the individual green pixels, the distance between the individual red pixels etc?

For green light you’d get a pixel pitch * 1.4 (over the diagonal) and for red/blue light a pixel pitch * 2.

Also, there may be an additional effect due to the applied microlenses where the pixel is not sensitive near the edges and all the more near the centre.
What’s your idea on this?


Regards,
Jaap.

[Bart_van_der_Wolf]

Hi Bart,

When calculating diffraction-limited MTF should we not take the Bayer demosaicing into account; not taking the actual pixel pitch into the equation but the distance between the individual green pixels, the distance between the individual red pixels etc?

For green light you’d get a pixel pitch * 1.4 (over the diagonal) and for red/blue light a pixel pitch * 2.

Good point Jaap,

But modern demosaicing algorithms also use the luminance data in the Red and Blue filtered photosites. Therefore I tend to consider the luminance data contained in those photosites diffraction limited as well, and thus consider them each/all Diffraction limited, just like the Green filtered ones.

One of the reasons is that I noticed 'considerable' differences between e.g. ACR, and C1 (or RawTherapee with AMaZE). The latter two produce cleaner conversions with fewer False Color artifacts and higher resolution, when testing with my test chart, like Bill did. The same Raw file produces different results. I have not tested with the latest Detail enhancement option of ACR, which I expect to close the gap somewhat.

This does mean that apparently the specific demosaicing algorithm can have an effect on the combined interpretation of luminance at a specific photosite and its direct neighbors.

Given that in Luminance, the photosites are considered to contribute something like 0.2125*Red + 0.7154*Green + 0.0721*Blue , the Green photosite influence dominates, and Blue is almost negligible, but they all do contribute a still significant enough amount of Luminance data. One could devise a metric with weighted contribution and even include a center wavelength for each filter color, but then one also needs to consider the subject colors. That quickly becomes a complex calculation, which will only marginally change the resulting conclusion based on only green wavelength diffraction.

Also, there may be an additional effect due to the applied microlenses where the pixel is not sensitive near the edges and all the more near the centre.

What’s your idea on this?

Yes, there is some influence on the per photosite diffraction, not so much from the micro lenses but rather the Optical Low-Pass Filter (OLPF). However, experience learns that when we combine multiple kinds of blur, the cascaded result quickly converges to a Gaussian shaped Point Spread Function (PSF). Because the Diffraction pattern also has a non-uniform (somewhat Gaussian) energy distribution, unlike defocus which has a more uniform disk or rounded, or an AA-filter with a more rectangle shaped blur pattern, I expect the Diffraction pattern to determine the resulting distribution more than the microlenses do. Also, the OLPF only has a modest effect on resolution, and they only reduce the likelihood of aliasing, but do not totally prevent it (that would require a more severe permanent loss of resolution for worst case scenarios).

Cheers,
Bart

[Dave (Isle of Skye)]

I agree 100% with Island Dave on this one.

Worrying about those things is a crippling influence that, unless you are making huge blow-ups for hospital corridors etc. mean nothing - at least to me...

Rob

There are two types of people when it comes to choice: maximizers and satisfiers. You are apparently a maximizer. Maximizers won't stop until they find the best among available choices. Satisfiers (like me) stop when they find good enough. Maximizers tend to spend more time deliberating and choosing, Satisfiers choose quicker and move on to do something with the chosen.

Thanks Rob and Slobodan for agreeing with me and my relaxed attitude and advice to such things as diffraction, and how due to modern equipment, these things have now become something of a none issue - well to my mind at least 

Yet it is also worth stating I think, that as we all know, there are all sorts of reason why spending any time looking for mathematical optical perfection and data capture, is an almost unachievable and therefore pointless exercise IMHO. And the reasons for this are many, as there are far too many practical issues to overcome in any system if it wants to remain affordable. Issues such as, does the capture device (the sensor or film) sit on the exact same vertical focal plane as the lens, as well as any other lens mounted onto that body? Probably not, as every manufacturer has to work within acceptable tolerances if they want to make a living, so chances it will not be set at 100% accuracy. Then you have the various elements within the lens, which again will have been constructed under acceptable manufacturing tolerances, so again the lens will probably not be perfect? Then we have a whole host of other issues such flange distances, parallel accuracy of lens mounting plates, different expansion rates of different materials and metals within the lens and body, as well as ambient air temperatures and pressures. In other words the list is endless and so perfection just isn't available, whereas "good enough", most certainly is and that is what I am happy to hang my photographic hat upon.

Anyway, here is a link that I would highly recommend for the seriously techie types (and anyone else), to a very interesting and deeply technical discussion on everything you will ever need to know about how photographic lenses actually perform in the real world. And don't be put of by me saying it is very technical, it is, but it is also very watchable and enjoyable, even for someone like me who is willing to ignore all of this type detail and who prefers to just throw on a lens and go out shooting 

So here you go guys and fill your boots - Oh and for some reason, the video doesn't start playing until after the first 9 minutes, so just move it on and then sit back and enjoy 

You will find the link by clicking HERE

Dave

[petermfiore]

Apart from that seemingly scholastic argument, has anyone actually done any comparison in prints of various sizes to prove the above? At which print size people wouldn’t be able to distinguish 7 Mp from 46 Mp?

This would be interesting... and combining this with viewing distance. This is a big factor.

Peter

[ErikKaffehr]

Hi,

It depends...

I would normally print at A2 (around 16"x23"). At that size, 12 MP may be good enough for a good print. But I am pretty sure could see a difference between 12 MP and 24 MP.

I also noted that I could not really tell a 24 MP image apart from a 39 MP image at A2 size, but going to A1 the difference is quite clear.

One of my lenses is quite weak a bit off axis:


Compare to a better lens:



Printing that image at something like A0 the difference between both of those images (at around 40 MP) was quite obvious at 50 cm viewing distance but not obvious at arms length distance, like 80 cm.

Sharpening can restore detail contrast, so we can get an image softened by diffraction pretty sharp. But fine detail that is lost cannot really be regained sharpening.

For some time, I was shooting with a Hasselblad 555/ELD and a P45+ back. The P45+ resolved 39 MP and my Sony A7rII has 42 MP. The Sony I shoot mostly at f/8 and the Hasselblad at f/11. What I found comparing the two that sharpness wise the two were pretty close.

The best lens I have for the Hasselblad is the 180/4 CF Sonnar while I use mostly a Sigma 24-105/4 Art on the Sony. My general impression is that the two systems deliver about the same in sharpness.

A while ago, I made a very accurate study of MTF on my systems.


Here you can see that the Sonnar 180/4 performs significantly better at f/5.6 than at f/8. Going to f/11 the smaller format Sony outperformed it using the 90/2.8 G macro used at f/5.6.

In the real world, achieving maximum sharpness is not easy and a lot of sharpness lost can be compensated by a bit of aggressive sharpening.

Best regards
Erik

This would be interesting... and combining this with viewing distance. This is a big factor.

Peter

[jeremyrh]

Thanks Rob and Slobodan for agreeing with me and my relaxed attitude and advice to such things as diffraction, and how due to modern equipment, these things have now become something of a none issue - well to my mind at least 

Which part of "modern equipment" do you suppose eliminates the effect of diffraction?  You are of course free to be as relaxed as you like about your photography activities, but the laws of physics don't depend on your mental state :-)

[Dave (Isle of Skye)]

Which part of "modern equipment" do you suppose eliminates the effect of diffraction?

Modern sensor technology such as the BSI sensor, which I am now happily using with the A7R2. Because by shifting the wiring to the back of the sensor, this means that the photosites can be placed at a much shallower depth within the front of the sensor and so the light travelling into those photosites, can still be successfully captured at much more acute angles of entry without any detrimental effects, as often happens when using small apertures,



thereby greatly reducing the effects of diffraction and colour fringing when stopping down - plus when added to the updated CA reduction algorithms in PS, means almost no fringing at all even when shooting at f/16 or even f/22 on my old lenses.

Aslo when shooting with a high density 42mp sensor without AA filters (that also happens to capture something like 14.5 stops of dynamic range), I can also get a much sharper image straight off the card. So my old (and original version) 16-35 L 2.8, is like shooting with a new, super sharp diffraction free lens on the A7R2 even when stopped down, whereas it was almost unusable at most apertures on the 5D MkII I was using it on before.

You are of course free to be as relaxed as you like about your photography activities, but the laws of physics don't depend on your mental state :-)

No but good images often do 

Dave

[ErikKaffehr]

Hi,

BSI will not affect diffraction. Diffraction is simply a property of light.

Essentially, diffraction is determined by the aperture, the wavelength of the light and nothing else.

At some aperture, the effects of diffraction dominate over lens aberration. With good lenses that happens at f/4 to f/5.6. The better the lens, the greater the effect will be.

Here is a great example from Jim Kasson's test of the Fuji GF 100/2 lens:

https://blog.kasson.com/the-last-word/focus-shift-loca-of-fuji-1102-on-gfx/

The lens peaks around MTF 50 of 3500 cy/PH at f/2, at f/11 it delivers still impressive 1700 cy/PH.

Now, look at Nikons 105/1.4, tested on the Nikon D850


The Nikon lens produces best MTF 50 at f/4, around 2200 cy/PH, stopped down to f/11 it reaches 1700 cy/PH. Essentially the same as the Fuji 50 GFX with the 110/2 at f/11.

Just to say, 1700 cy/PH is still pretty good...

Best regards
Erik

Modern sensor technology such as the BSI sensor, which I am now happily using with the A7R2. Because by shifting the electronics to the back of the sensor, this means that the photosites can be placed much shallower within the front of the sensor and so the light travelling into those photosites, can still be successfully captured at a much lower angles of entry without as many detrimental effects



thereby greatly reducing the effects of diffraction and colour fringing when stopping down - plus when added to the updated CA reduction algorithms in PS, means almost no fringing at all even when shooting at f/16 or even f/22 on my old lenses.

Aslo when shooting with a high density 42mp sensor without AA filters (that also happens to capture something like 14.5 stops of dynamic range), I can also get a much sharper image straight off the card. So my old (and original version) 16-35 L 2.8, is like shooting with a new lens on the A7R2, whereas it was almost unusable on the 5D MkII I was using before.

No but good images often do 

Dave

[Dave (Isle of Skye)]

Hi,

BSI will not affect diffraction. Diffraction is simply a property of light.

Essentially, diffraction is determined by the aperture, the wavelength of the light and nothing else.

At some aperture, the effects of diffraction dominate over lens aberration. With good lenses that happens at f/4 to f/5.6. The better the lens, the greater the effect will be.

Here is a great example from Jim Kasson's test of the Fuji GF 100/2 lens:

[snip]

The Nikon lens produces best MTF 50 at f/4, around 2200 cy/PH, stopped down to f/11 it reaches 1700 cy/PH. Essentially the same as the Fuji 50 GFX with the 110/2 at f/11.

Just to say, 1700 cy/PH is still pretty good...

Best regards
Erik


Best regards
Erik

Well I can only tell you what my eyes are seeing, as I have no other way to measure this, nor would I want to waste any time trying to do so.

Just look at the shot I posted several pages ago on this thread, is it sharp enough? Yes I think so. Are there any diffraction or CA issues within it? Well not that I can see. So yes you may be able to show me graphs and scientific papers that tell me the physics are saying one thing, but all I can say is that my eyes are telling me that with this sensor and these old lenses, that CA and diffraction etc, are no longer an issue worth worrying about.

But of course Erik, if you wish to worry about these issues, then that is fair enough and I would not wish to argue with you or anyone else on this, but for me, it simply comes down to the technology having now evolved to the point, where it has become good enough for such technical considerations as these, to have become a thing of the past.

all the best everyone 

Dave

[Bart_van_der_Wolf]

Hi,

BSI will not affect diffraction. Diffraction is simply a property of light.

Essentially, diffraction is determined by the aperture, the wavelength of the light and nothing else.

That's basically the case, where the shape of the aperture makes some difference as to the shape of the resulting MTF curve, but not to the cut-off point where the MTF curve goes to zero due to diffraction.

Once the light has passed the photosite's aperture, nothing changes as far as diffraction is concerned, unless we have a photosite with a very small fill-factor. The fill-factor of a BSI photosite is designed to be large, so no change to diffraction.

Cheers,
Bart

[Slobodan Blagojevic]

...No but good images often do 

Hehe... well played, mate, well played!

[ErikKaffehr]

Hi Dave,

If you refer to this image: , how could I tell? The image is just web size. Still, it may have an oversharpened look.

To a significant effect, loss of sharpness can be compensated by sharpening. So, you can lose acutance and it can be restored using deconvolution sharpening. But, starting with a sharp image would be preferable.

Also, stopping down more than needed mean that you may need to use a higher ISO which will increase noise and reduce dynamic range.

It may seem to make little sense to buy an expensive lens, that performs best at f/4 and use it at small apertures.

On the other hand, it is possible to recover detail until a certain limit with appropriate sharpening.

Here are two real world examples (shot on Sony A7rII with a Sigma 24-105/4 Art), both are 1:1 crops.



In my view the f/8 image is visibly sharper than the f/11 image. That could be compensated by sharpening, of course.



Here I wanted to have long exposure for the water, so I used f/22, but I also wanted a sharp building so I shot another image at f/8. The left image is f/22 the center f/8, the image on the right was f/22 but sharpened by ImageMagic that does deconvolution.

Note that center image has moiré, indicating lens outresolves sensor. That is not visible in f/22 images.

Resolution does not matter for web size images. But my intention is that the images I make will be usable at 40"x60" print size.

Best regards
Erik

Well I can only tell you what my eyes are seeing, as I have no other way to measure this, nor would I want to waste any time trying to do so.

Just look at the shot I posted several pages ago on this thread, is it sharp enough? Yes I think so. Are there any diffraction or CA issues within it? Well not that I can see. So yes you may be able to show me graphs and scientific papers that tell me the physics are saying one thing, but all I can say is that my eyes are telling me that with this sensor and these old lenses, that CA and diffraction etc, are no longer an issue worth worrying about.

But of course Erik, if you wish to worry about these issues, then that is fair enough and I would not wish to argue with you or anyone else on this, but for me, it simply comes down to the technology having now evolved to the point, where it has become good enough for such technical considerations as these, to have become a thing of the past.

all the best everyone 

Dave

[jeremyrh]

No but good images often do 

Dave

Understood. If physics worries you, best to not think too much about it. But don't pretend it goes away when you shut your eyes :-)

[John Camp]

I agree. Diffraction sucks the life out of an image (loss of micro-detail), but there is some mitigation possible with proper (deconvolution) sharpening. Not all images need to be technically perfect, but when given a choice, why not go for the better result?

My rule of thumb is based on that I can see the onset of diffraction losses when the diffraction pattern diameter (at the first zero ring) exceeds 1.5x the photosite pitch. That means that, approximately, pitch x 1.108 is the Aperture number where (green wavelength) diffraction becomes visible as loss of contrast. It can still be mostly compensated for, but it's gradually down-hill from there with narrower apertures until even high contrast detail cannot be resolved anymore. One might as well have used a lower resolution sensor.

MTF will go to zero at the following (circular) aperture:
N = 1 / (cycles_per_mm x wavelength)

One of the problems (in my opinion) with discussions of this type are phrases and words like "sucks the life out of" and "better." I have hanging on my living room wall a very fine print of one of the most famous photos ever made, Adams' "Moonrise," and compared to prints of more or less the same size that come out of my m4/3 cameras printed on a Canon printer, it's not notably sharp. No way it could be, given all the manipulations that Adams put the negative through. But it's not lifeless, and you'd go a long way to to find a photograph that you could call "better."

As almost everybody knows, Adams and other people who advocated straight photography called themselves "Group f64." But f64 was diffraction limited even on 8x10 cameras, which wold suggest something about how much they worried about it. Again, IMHO.

[Bart_van_der_Wolf]

One of the problems (in my opinion) with discussions of this type are phrases and words like "sucks the life out of" and "better." I have hanging on my living room wall a very fine print of one of the most famous photos ever made, Adams' "Moonrise," and compared to prints of more or less the same size that come out of my m4/3 cameras printed on a Canon printer, it's not notably sharp. No way it could be, given all the manipulations that Adams put the negative through. But it's not lifeless, and you'd go a long way to to find a photograph that you could call "better."

Hi John,

As I said, or tried to get across, technical quality usually adds to the creative effort of communicating emotion. An image that lacks an almost tactile representation of the subject matter is usually not a better image. It's an image that can only be viewed from a distance, or at a small size (because there is nothing gained by closer inspection). Given a choice, an image with more/better microdetail will keep on fascinating the viewer. longer. In my view, that's usually better.

As almost everybody knows, Adams and other people who advocated straight photography called themselves "Group f64." But f64 was diffraction limited even on 8x10 cameras, which would suggest something about how much they worried about it. Again, IMHO.

I see that differently, it was more of a necessity to get enough depth of field. It, in fact, was a way to increase sharpness in regions that would otherwise be even mushier, and the loss of resolution in the plane of best focus was an acceptable compromise to achieve it.

It would have been interesting to learn, had Adams had the possibility of focus-stacking like we do today, what he would have done ...

Cheers,
Bart

[fdisilvestro]

Deciding about trade-offs between maximum image quality and achieving the desired image that meets or exceed the requirements of the desired outcome (what others may refer as good enough) is perfectly valid.

Just ignoring the trade-offs because you think that science is BS or that your camera is not subject to the laws of physics does not makes sense to me.

[Martin Kristiansen]

Deciding about trade-offs between maximum image quality and achieving the desired image that meets or exceed the requirements of the desired outcome (what others may refer as good enough) is perfectly valid.

Just ignoring the trade-offs because you think that science is BS or that your camera is not subject to the laws of physics does not makes sense to me.

No one is saying science science is BS. Science indicates there is diffraction, it doesn’t make a value judgement on the importance of diffraction, you do that. You are saying diffraction is more important than depth of field. That is not science, that is your value judgement. Saying that people that don’t agree with your value judgement are unscientific is very poor reasoning indeed. Try at least to understand the arguments from the other side.

It’s like this. There is a trade off between something appearing out of focus due to shallow depth of field and something appearing to be soft due to diffraction. It can happen that a photographer decides he needs deep depth of field and for various reasons focus stacking is not an option. If the need for deep depth of field is greater than the need for micro detail then you stop down and the hell with diffraction. Perhaps the image doesn’t have much micro detail or perhaps the micro detail is unimportant or less important in this instance,  whatever. It is a creative decision. It’s not saying I don’t believe in diffraction, it is a decision that it is less important in this instance.

If most people and the photographic community in general agreed with this whole concept of least diffraction above all else lenses would come Limited to around f5,6 and we would simply use shutter speed and iso to control exposure.

[ErikKaffehr]

Hi,

It is not as simple as that. DoF depends on your definition of sharpness.

Just as an example is this image sharp?


I would guess that most readers would not regard it as sharp. But, it represents a CoC of 26 microns. Most DoF tables define acceptable CoC to be 30 microns.

So with DoF tables near and far limits are pretty unsharp. Best sharpness will fall somewhere in between.

The image shown was shot at f/4, now lets stop down to f/8, that would cut CoC to half:

f/5.6 in focus	f/8 out of focus

Now, let's stop down to f/16 and see what happens:

f/5.6 in focus	f/16 in focus	f/8 out of focus	f/16 out of focus

Stopping down to f/16 does not bring great benefits over f/8, as diffraction now competes with defocus.

Jim Kasson has posted an article on managing blur, worth reading: https://blog.kasson.com/nikon-z6-7/ff-examples-of-optimal-blur-management/

Best regards
Erik

No one is saying science science is BS. Science indicates there is diffraction, it doesn’t make a value judgement on the importance of diffraction, you do that. You are saying diffraction is more important than depth of field. That is not science, that is your value judgement. Saying that people that don’t agree with your value judgement are unscientific is very poor reasoning indeed. Try at least to understand the arguments from the other side.

It’s like this. There is a trade off between something appearing out of focus due to shallow depth of field and something appearing to be soft due to diffraction. It can happen that a photographer decides he needs deep depth of field and for various reasons focus stacking is not an option. If the need for deep depth of field is greater than the need for micro detail then you stop down and the hell with diffraction. Perhaps the image doesn’t have much micro detail or perhaps the micro detail is unimportant or less important in this instance,  whatever. It is a creative decision. It’s not saying I don’t believe in diffraction, it is a decision that it is less important in this instance.

If most people and the photographic community in general agreed with this whole concept of least diffraction above all else lenses would come Limited to around f5,6 and we would simply use shutter speed and iso to control exposure.

[Martin Kristiansen]

Please excuse the sloppy nature of these images.

Shot on a Sony A7Riii on a tripod at 100 iso. 1/250 at the apertures marked. 2.8, 8 and 22 going from left to right. Sony 90mm macro lens manually focussed. Default sharpening applied in LR and no other subsequent sharpening.

Top row of images a crop out of the centre of the ruler showing the area focused on. f28 clearly has the shallows DOF and is the the sharpest aperture of the this. I imagine the lack of sharpness due to lens aberrations. Centre image of top row is the sharpest at f8. Around the set spot for this lens. I imagine little to no meaningful issues from diffraction. Right image shows degradation from diffraction. It is quite clear. Shot at the lenses minimum aperture of f22.

Middle row. A crop from the closest portion of the ruler. Left image shot at 2.8. Totally out of focus, no useful detail visible at all. Middle image of middle row at f8 shows a lot of out of focus and very little detail. Right image at f22 shows quite good focus in general and much clearer than the image shot at f8. In terms of detail while not optimum it it hugely better than the image in the centre shot at f8

Bottom row of images is a reduced full image from which the crops above were taken.

Im pretty sure that's all clear.

Yes its a ruler and yes it could be focus stacked or a ts lens used. That is not the point of this exercise. The point is to show that DOF at f22 beats OOF at F8 despite all the much vaunted superiority of diffractionless images. On a practical level I have a client that spends about $20 000 a year on shooting little items like this. They employ a retoucher and I shoot about 70 to 100 images a day. There is no time for focus stacking in either the capture or post production. I shoot at F22 and we are all happy. They sell lots of stuff and its all good. I have another client that I shoot about $25 000 worth of clothing for. Some flats and some on mannequins and most on models. No depth of field issues I shoot around F5.6 to F8. Sharp due to Eye AF, thank you Sony, and sharp due to no diffraction. Again they sell lots of stuff and we are all happy.

It is not a binary thing. You use what tools and settings you have access to in order to get the job done. If you can get away with ignoring diffraction in the interests of DOF then that's what you do.

Oh and my definition of sharpness is if it looks sharp, simple really.