on landscape The online magazine for landscape photographers
on landscape
facebook twitter rss feed
on landscape Join the conversation

    on Distant Horizons

    It must have been rather fab to research this, and it's terrific to encounter artists involved in the genre.

    - milouvision, 11:20 16th Jun

    on Camera Survey

    It might be interesting to see how the votes would stack if you asked in three months to repeat the survey, I for one missed the original request. It's a very interesting survey and nice to see that Pentax still have such a large following, likewise the medium and large format categories, [...]

    - Douglas Salteri, 08:54 16th Jun

    on Is Adobe Creative Cloud Bad For Photographers?

    Great article Paul! I hope Adobe will continue to develop Photoshop with new features for photographers. I would love some new adjustment layers, how about midtone contrast (a.k.a clarity)?

    - Magnus Lindbom, 07:32 16th Jun

The Diffraction Limit

Science of Photography

How small is too small?

Responses28

I think most photographers will have come across various online resources, books, magazines and blog posts telling them that certain apertures are ‘out of bounds’ and that in order to get the sharpest pictures they need to use a narrow band of apertures (usually f/5.6 or f/8 for 35mm cameras). As mentioned in last issue though, a test of some lenses seen recently where smaller aperture results on the D800 resolved more than the best aperture results on a 5Dmk3 got us thinking. Last week myself, Mark Banks and Joe Cornish met up to run some tests where we used the 36Mp Nikon D800E and a 24Mp Sony A900 (same sensor as the Nikon D3X) at various apertures to find out just how much affect diffraction had on our final prints.

NB: Please note that due to a bug related to a wordpress plugin (Portfolio Slideshow Pro) the slideshow is not working properly in Chrome but does in Firefox – we haven’t checked other browsers but are talking to the developers

Now testing this is fairly simple, we take a couple of cameras and shoot a test target at a range of apertures. The trick is, how do we sharpen them. Well – only trial and error can fix this one and so after using our test target in Joe’s studio, I spent an afternoon trying different sharpening strategies.

I finally settled on the application of two rounds of sharpening, one at a very fine radius and one at a broader radius and then a final unsharp mask to bring contrast back up. Before we show the effects of this though, let’s have a look at the range of apertures for the D800E.

_0006_2.8
_0005_4
_0004_5.6
_0003_8
_0002_11
_0001_16
_0000_22

f/2.8

f/4

f/5.6

f/8

f/11

f/16

f/22

_0006_2.8 thumbnail
_0005_4 thumbnail
_0004_5.6 thumbnail
_0003_8 thumbnail
_0002_11 thumbnail
_0001_16 thumbnail
_0000_22 thumbnail
_0006_2.8
_0005_4
_0004_5.6
_0003_8
_0002_11
_0001_16
_0000_22

f/2.8

f/4

f/5.6

f/8

f/11

f/16

f/22

_0006_2.8 thumbnail
_0005_4 thumbnail
_0004_5.6 thumbnail
_0003_8 thumbnail
_0002_11 thumbnail
_0001_16 thumbnail
_0000_22 thumbnail

Now you can obviously see the effects of diffraction here and the f/22 looks incredibly soft. However, let’s apply our custom sharpening to each of these and see what we can come up with.

_0006_2.8
_0005_4
_0004_5.6
_0003_8
_0002_11
_0001_16
_0000_22

2.8

4

5.6

8

11

16

22

_0006_2.8 thumbnail
_0005_4 thumbnail
_0004_5.6 thumbnail
_0003_8 thumbnail
_0002_11 thumbnail
_0001_16 thumbnail
_0000_22 thumbnail

Now things are looking a little closer but the f/22 shot still looks like it has lost of a bit of contrast. However, the final test is to print these and stick them in front of some unsuspecting punters. So we did. And they couldn’t tell the difference….

Now it’s worth restating that – real live punters couldn’t tell the difference between a shot taken at f/5.6 and one taken at f/22.

We tried this at various print resolutions and for everything from 360dpi to 240dpi our punters couldn’t tell the difference. At 180dpi it does look like the f/22 has lost a bit of ‘bite’ and contrast but the f/16 looks fine. If the photos weren’t shown side by side I don’t think people could have told the difference.

We repeated the test for the A900 and got slightly different results – first of all here are the unsharpened 100% apertures.

SONY DSC
SONY DSC
SONY DSC
SONY DSC
SONY DSC
SONY DSC

22

16

11

8

5.6

4

SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail

Now when we sharpen these files we get a little surprise in that the f/22 result does show more significant reduction in quality. The smaller apertures are also showing more noise when sharpened. This just shows how clean the D800 images were that they could take such strong sharpening without showing much textural noise. However, when printed the f/22 result does well but starts to show some quality loss at about 240dpi.

SONY DSC
SONY DSC
SONY DSC
SONY DSC
SONY DSC
SONY DSC

22

16

11

8

5.6

4

SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail
SONY DSC thumbnail

Our final comparison was to confirm what we had seen that got us thinking about things – what would the f/22 D800 file look like in comparison with a Sony A900 f/8 file? Well, the answer is that they’re pretty close with the D800 getting a cleaner result. Surprising yes?

pk_0001_D800 f-22
pk_0000_A900 f-8

D800 f/22

A900 f/8

pk_0001_D800 f-22 thumbnail
pk_0000_A900 f-8 thumbnail

Finally one of our readers (Simone – tjshot) sent us a link to his research where he used variously well sourced theoretical figures to calculate the potential of various sensors given optimal sharpening. You can look at his research yourself although be warned it is quite in depth. He’s given us permission to reproduce a couple of the most pertinent tables here though.

The following tables give a lot of information but I’ve highlighted the critical parts. The three tables show a 21, 36 and 50 megapixel camera. Each row shows a different aperture. The yellow columns show the theoretical resolution of the lens for 50% contrast (MTF 50) and for 10% contrast (MTF 10).

Now the figures that are normally quoted for cameras are the unsharpened MTF 50 (50% contrast) values which you can see  in pale blue. From this you can see that the value of 26 lp/mm for f/22 is a fraction of the 47 lp/mm for f/5.6 (on the 36Mp camera).

However, if we add sharpening(pink column) we get 60 lp/mm for f/22 compared with 80 lp/mm for 5.6. And finally, if we look at the 10% contrast figures with sharpening (green column), we see that 100 lp/mm for f/22 and 105 for f/5.6!! Hardly any difference. And this is for simple sharpening!

As Simone has shown, the resolving power of a 36mp camera at f/22 does out resolve a 21Mp sensor at any aperture setting. And into the future, a 50Mp sensor will continue to out resolve the 36Mp sensor, even at f/22. This has even been shown in practise as the 7D has the same pixel density as a possible full frame 50Mp camera.

We’ve done a couple of Ad-Hoc tests on ‘pictorial’ subject matter and the ‘theory’ holds up well (I’ve reduced this to 60% to give you an idea of what it looks like in print – you’ll have to trust me that it looks very close). Our own conclusion is that you should be absolutely fine shooting at f/16 and if you need f/22, go for it!

house-comparison-f56
house-comparison-f22

f/5.6

f/22

house-comparison-f56 thumbnail
house-comparison-f22 thumbnail

Summary

We showed that for the D800E, the f/22 result could be sharpened until it was a fairly close match for the f/8 or f/5.6 image, albeit with some additional texture and lower overal contrast. When printed, it was difficult to make out the difference unless they were made at 180dpi or less. f/16 could be sharpened until it was identical in print.

The diffraction at f/22 also served the purpose of removing any moire effects completely (and at f/16 it was very small). When compared the the Sony A900 results we also confirmed that the Nikon D800E at f/22 is sharper than the Sony A900 at f/5.6 – a remarkable result!

It was also interesting that the Sony A900 shot at f/22 wasn’t quite as sharp as it’s f/5.6 f/8 result – the reduction in contrast and increase in noise/texture was enough to reduce the f/22 result somewhat. f/16 did look good though. It seems that you need more resolution and cleaner files to make the most of these f/22 results.

Furthermore we had the theoretical contribution from Simone which which correlated well with our findings in that the contrast at f/22 is reduced by about 20% but the detail is still all there. Simone has also worked out the results for a 50Mp sensor and has confirmed that we should see the same effects with theoretical 50Mp camera outresolving the Nikon D800E even when shot at f/22.

Conclusion

You can’t beat resolution. Modern sharpening techniques have allowed the treatment of f/22 images so that they are almost indistinguishable from f/5.6 & f/8 images. And further resolution will continue to show resolution and contrast improvements even up to 50Mp (and possibly beyond).

It may even be that an 50Mp image shot at f/32 could outresolve a Canon 5dmk3 at f/5.6! The megapixel race is far from over.



Tim Parkin

timparkin.co.uk

Flickr, Facebook, Twitter

28 thoughts on “The Diffraction Limit

  1. Thanks of the very intriguing D800 related articles Tim. Moving from a D3x to a D800 has been worthwhile for me on several levels and now I can obtain slightly more depth of field with little reduction in resolution I am even happier. I have also found that the D800 files respond well to a higher degree of sharpening than I have previously applied. I’m really looking forward to producing some big prints. I think I may have to refer back to your articles on image sharpening though. Can you give more detail on the sharpening process you have been using for the D800 files?

    • Hi John.. I’m still working on settings but smart sharpen in gauss mode at 0.2px – quite high amounts but without visual impariment. Then another smart sharpen at 0.4px with a smaller amount. The end result shouldn’t look sharpened (see my images for examples). Let me know or send me samples and I’ll have a play.

  2. Thanks Tim, I’ll have a play with some images. I primarily shoot with the 24mm and 45mm PCE lenses but I’m aware that the Zeiss lenses will generally resolve more detail. With the potential for increased depth of field they could become more appealing, especially at the wide end. Interesting times.

    Completely off topic I am becoming increasingly interested in the F Stop range of mountain packs. As it seems they can only be found at the Paramo store in Covent Garden could you get Joe Cornish to give his thoughts on his against a Lowepro Trekker. I find the Trekker functional but not as comfortable as a true back pack and wondered if the F Stop range may be the answer.

  3. I think this confirm that the D800/E is close to film because I remember during my days with film, shooting @22 was not a problem as I found on my Nikon D3.

  4. Hi Tim,

    Another excellent piece of research. Your findings are contrary to what I have expected, at least to a significant part.

    Obvoiusly, sharpening is the joker in the DoF vs. diffraction game. I was considering two step sharpening, too. The idea was to sharpen with small radius for detail and a larger radius to increase perceived sharpness.

    Thanks for all the effort you and your friends are putting into this and also for sharing your findings.

    Best regards
    Erik

  5. Hi Tim, This is very interesting. Like Erik, I expected the contrary in general. But it does look like sharpening technology may be able to trump diffraction if done carefully. I wonder if a Raw sharpener like the NIK one combined with an output sharpener may be worth investigating. I know a few people who have good experiences with that workflow.
    I was quietly hoping I wasnt going to be tempted by a D800!

  6. Very interesting article, and great that you reported visual results from prints. Prints always evens out the odds a bit compared to watching at 100% on screen.

    Being able to use very small apertures and sharpen digitally makes digital cameras really strong in the deep DoF game compared to for example large format film. f/22 equivalent aperture on 8×10″ is about f/165 if I’ve calculated right, and you can rarely set it that small, often the limit is f/64 which corresponds to only f/9 on a D800E (!). Medium format digital can usually go to f/32 or more so you have the same advantage there.

    This factor makes it relevant to use a new DoF definition for digital I think, instead of the old “sharp enough for some distant viewing condition for people with average vision” I prefer “so sharp that the DoF edge is indistinguishable from plane of focus even at close inspection, after sharpening has been applied” which leads to smaller CoC, a good one I think is MAX(, ).

    Film required the more relaxed traditional DoF definition since you could not sharpen in the same way, and you could not set as small apertures relative to the film size, but if you want to make as sharp prints as possible with a modern digital system you should obviously use smaller apertures than a DoF table based on the traditional view would indicate.

    • Oops, the CoC definition disappeared in my comment. I wrote that I think a good CoC to use for digital camera Depth of Field calculations when you use it for deep DoF landscape photography is the larger of 2 x pixel pitch size or the airy disk diameter of the given aperture. For small apertures (deep DoFs) the airy disk diameter is typically always the larger one, only if you have a really low resolution camera 2x pixel pitch is larger.

  7. Another comment/question: the f/22 shots show an overall reduction in contrast, the large white fields are darker and the large black fields are lighter. This is a phenomenon I have myself noticed with my lenses but I have never been able to understand why it occurs. Is it really diffraction?

    I don’t see how a large airy disk would reduce contrast of fields much much larger than the airy disk. Does anyone know what this effect is? There’s also quite a lot of grayish “bleed” into the large black square which to me seems much larger than the airy disk diameter (and a bit harder to fix in post-processing, while overall contrast increase is easy)

    • Hi,

      I don’t really have any useful input. I have checked some of my test shots and the effect you are referring to was clearly seen. One explanation I was thinking of was that the Airy pattern has several outer rings which cover a much larger area than the central peak. Clearly most of the energy goes into the central peak. If we have a black patch surrounded by a large white space the integrated distribution from the outer airy rings may explain loss of contrast. One possible explanation.

      I didn’t do the math, it may hold or it may not.

      Best regards
      Erik

      • I’ve asked around a bit and it does seem to be exactly as you say. The Airy pattern has in actuality several outer rings which affects low frequencies. This causes reduction in global contrast and the “bleed” effects. AFAIK normal deconvolution sharpening does not take this into account (too difficult to do), so post sharpening you have sharper edges but still lower contrast and bleed.

        Thus to compensate for diffraction one needs to do some additional step(s). A simple contrast increase probably goes a long way, but that does not eliminate the bleed. Maybe there is some sort of local contrast increase method to employ that would work better?

        • An unsharp mask with about a 20-40px radius and a small amount will boost local contrast well.

          I’m not sure what may cause the diffuse flare apart from the airy disk but I don’t think the airy disk is responsible for it all. More research needed :-)

        • From the description, it seems like a typical veiling glare effect: high frequency details are preserved, but contrast between low frequency ones is low.
          The Airy disk alone is probably not being responsible for it: it should influence only the microcontrast of the image, not the macrocontrast.
          I even doubt it’s connected to the lens: this behaviour was typical of fast lenses from 30 years ago, but only affecting wide apertures; I don’t expect to see it for small apertures.
          In the past, for film cameras, a similar effect was induced by a bad flocking of the exposure chamber: for example it was a typical issue with Kiev medium format SLRs.
          It’s hardly the case with modern DSLRs.
          My take is it could be due to some form of light scattering from the sensor’s optical or AA filters, influencing wide Airy disk diameters more than small ones: like Kaffehr suggested, some sort of integrated scattering of the large Point Spread Function at F22.
          It would be useful to know if the same behaviour shows up for the same lens at F22 on colour film.

  8. What a great article. Thoughtful, well executed, clear, balanced, informative. This is what all web journalism should be like!

  9. Hi Tim – this one’s really useful… every now and then something comes along that’s helpful in other genres. This is the next in that group.

    I’ll stop flouncing around with my architectural images and just stop worrying!!!

  10. It’s all good, but you aren’t helping me resist the temptation to blow my hard earned cash switching from my trusted and (until recently) much loved Canon 5D2 kit for Nikon’s new megapixel behemoth (which admittedly I don’t need but when did that ever stop anyone).

  11. Hi,

    I found this interesting: “Furthermore we had the theoretical contribution from Simone which which correlated well with our findings in that the contrast at f/22 is reduced by about 20% but the detail is still all there.”

    The impression I got that Simone used the methodology suggested by Norman Koren to calculate those values. It just shows how good Normans analysis is.

    I was somewhat surprised with some of the findings, like the differences being more visible in prints at lower PPI, and I actually expected f/22 to limit sharpness. Were the images scaled for different PPI or were low PPI prints simply larger?

    Making real prints and doing a correct blind tests with real punters is a great way to find out what matters.

    Best regards
    Erik

    • I kind of think that blind tests at close range with real picky print experts is better :-) . Small quality differences can only be appreciated by connaisseurs.

      The other day I pointed out that a image on the front page of a small publication was obviously shot with a compact camera and had rather poor technical quality. My non-photographer friends did not understand what I was talking about :-) . One of many times I’ve become aware of that only a small amount of people can see and/or do care about technical image quality in general. But well, I do, so when I see tests I prefer they are done by someone as picky as I am :-)

      • That’s why you have my conclusion included as well ;-)

        And there is a lot of difference between punters assessing a single image against comparing two different images. Yes they will have difficulty placing the single image in a spectrum of good and bad. However even your average punter can look at two different images and make a sensible assessment. They may not know exactly where to look to pick up on the weakness of certain types of files but that is probably a good thing.

      • Hi Tim,

        I looked at resolution in your samples comparing the Alpha 900 at f/5.6 and Nikon D800 at f/22. What I see is that the Nikon D800 image looks smoother but if you look at the converging line pattern in the test target the Alpha actually resolves better. So what I think that we see is that we actually loose resolution but the higher resolution of the Nikon works better with sharpening.

        Anyway, your article is a great eye opener, clearly indicating that “excessive” stopping down combined with adequate sharpening can be a good way to optimize capture.

        Best regards
        Erik

  12. Interesting article Tim, it confirms that effective usable F-stop range is sensibly extended for digital sensors compared to film.
    I’d like to confirm that all the simulation scenarios I proposed are strictly built upon Norman Koren’s analysis; I just used the lens, sensor and sharpening transfer functions he proposed and extended the analysis to a wider F-stop range, to assess the residual performance loss after proper sharpening.
    Theoretical results turned out to be surprising for me and now they are confirmed by empirical tests also, both on actual APS-C and full frame cameras; a denser sensor pitch boosts the effects of sharpening and allows for the use of a wider F-stop range.
    Of course contrast drops significantly with smaller apertures but it’s well recoverable up to F22 for sensor pitch in the 4-5 microns range; smaller stops will involve a serious, mostly unrecoverable, degradation.
    An important thing to keep in mind when considering the effects of diffraction is the difference between shooting a real-life object or a standardized MTF target.
    MTF testing is performed imaging a uniformly-lit sine-wave pattern of given contrast ratio, which may produce a lower contrast transfer than a finite object with hard edges and dishomogeneous illumination: for example a tree against a light background in a landscape shot or a textured object with strong stray light in a close-up shot may retain more contrast than a standard MTF plot of lens/sensor combo would suggest.
    The effect on resolution (MTF 10% values in my simulation) is often negligible, but perception of sharpness (in rough approximation MTF 50% for 35mm sensors) is greatly enhanced.
    On the other side, a dimly lit subject, even with distinct edges and textures, will probably result in lower perceived sharpness than the standard MTF plot of lens/sensor combo would suggest.
    For my simulation I relied on the standard MTF testing parameters (a uniform rolloff for lens MTF) which don’t take into account the increased/decreased contrast resulting from real scenarios.
    In other words, for real subjects the use of a small aperture like F22 + sharpening will probably result in better performance than my theoretical analysis suggests; more so for landscape photography, where high contrast details are common.
    I believe the sample pictures from the article prove it better than words and numbers.

    As a final personal note, previous considerations about real use scenarios should be applied also to many tests of lenses and cameras available online.
    Combined lens/ camera MTF analysis can be a valuable tool to assess the relative performance difference of lens/sensor combinations, especially when the same lens is pitched against different sensors of the same size, but it should not become an absolute indicator of the value of a system, or a lens, as a photographic tool.
    Reliable numerical testing is difficult to perform: DXOmark is among the few sites which seem to do it right, but even then results can be misleading for the casual user, who tends to focus on resolution figures only.
    I’m still amazed about how a Zeiss Distagon 21mm or 25mm manages to fall behind far lesser lenses…
    Some aberrations are functional to the intended rendering of a lens, for example some residual spherical aberration wide open in the close range for a portrait tele (i.e. Zeiss Planar), or some field curvature in the wide-angles, but they are often considered as defects just because thy lower MTF values obtained shooting flat targets.
    It’s always better to take a look at real life samples from a lens/camera combo rather than relying on pure test numbers.

    I link two rather enlightening articles from Zeiss’ H.H. Nasse about the pitfalls of numeric assessment of performance; they provide both qualitative MTF discussion and some good real samples.
    They are clearer than everything I have ever read on the subject and a must for every serious photography addict…

    How to read MTF curves – Part I
    http://www.zeiss.de/C12567A8003B8B6F/EmbedTitelIntern/CLN_30_MTF_en/$File/CLN_MTF_Kurven_EN.pdf

    How to read MTF curves – Part II
    http://www.zeiss.de/C12567A8003B8B6F/EmbedTitelIntern/CLN_31_MTF_en/$File/CLN_MTF_Kurven_2_en.pdf

    How to read MTF curves – Part II – Images
    http://lenses.zeiss.com/content/dam/Pho … images.zip

  13. Pingback: Overcoming My Photo Entekaphobia: The Fear of Shooting at f/11

  14. One way to reduce the outer rings on the Airy Disc is ‘Apodisation’ – a neutral density filter with a radial gradient, so the aperture has soft edges. It’s a technique used in Astronomy, but Sony make a special lens with a grey-glass element !
    http://en.wikipedia.org/wiki/Minolta_STF_135mm_f/2.8_T4.5_lens

    It’s done for bokeh wide-open, but the idea could be adapted to smaller apertures. D800 and ‘Waterhouse Stops’, anyone ?

    Your readers may also be interested in the last section of
    http://www.nikonusa.com/pdf/manuals/dslr/D800/D800_TechnicalGuide_En_v2.pdf
    “Using Diffraction to Combat Color Artifacts and Moiré”

Leave a Reply

Your email address will not be published. Required fields are marked *


*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Notify me of followup comments via e-mail. You can also subscribe without commenting.

On Landscape is part of Landscape Media Limited , a company registered in England and Wales . Registered Number: 07120795. Registered Office: Orchard House, Burnby, York, YO42 1RS