lens nerds: what's up with different lens behaviours with diffraction?

Discussion in 'Open Discussion' started by piggsy, Oct 24, 2015.

  1. piggsy

    piggsy Mu-43 All-Pro

    Potentially a stupid question and I kind of know just enough to form the question here, but, bear with me.

    I had always thought the effects of diffraction were some linear thing linked to the sensor size and pixel pitch, where you would have one aperture that would produce the highest resolution and (generally) a drop in perceptible artifacts, and then there would be a steady decline from that point on in terms of resolution, and a decrease in some artifacts like CA as it went further down. Rule of thumb, kind of thing.

    I started wondering about this when I got the 'Bokina' 90mm macro - always seemed better behaved than the M.Zuiko 60mm, particularly from about F11-F13 onwards. Just seemed to lose less and behave a lot more predictably - the 60 can take on this kind of "torn" looking effect, almost like motion blur, if you give it OOF highlights stopped down a lot just outside the zone of focus.

    Here's photozone's test of the M.Zuiko 60mm 2.8 :


    Wat... wat? That's apparently a real thing it does? I'd never seen that test before and it made me curious to see what was going on with other macros and other formats.

    Been looking at getting a longer lens (Tamron 180 or Kiron/Dine 105 or Sigma 105/150) so I decided to look at some of those (I believe these are all on the same test bed setup / software and all on aps-c/D200/10mp)

    Tamron 180:


    Sigma 105:

    Sigma 150
    (1.4xTC snipped)

    Some questions -

    How transferable are these kinds of results - say - the Sigma 150's F22 result to the Olympus 60mms across formats? Is there something in the design of a lens that you can see on a lens diagram (like, the fundamental type of design) that can predict what happens to CA or resolution as you raise the aperture number? What factors actually cause some lenses CAs to increase drastically as they're stopped down, or res to suddenly crash out beyond the trend line?

    For general shooting purposes, I can understand behavior at F16-22 being somewhat irrelevant but for long macros it's actually something that would be quite handy to know :D 

    (ed - added links)
    Last edited: Oct 24, 2015
    • Like Like x 1
  2. barry13

    barry13 Super Moderator; Photon Wrangler Subscribing Member

    Mar 7, 2014
    Southern California
    Hi, a couple comments:

    Since the 4/3 sensor and pixels are much smaller, the CA numbers would have to be multiplied by the crop factor and the pixel density ratio (16mp/10mp).

    The MTF numbers are probably much harder to translate mathematically across different sensors.

    I don't know why CA gets worse when stopping down the 60mm, but could it be related to diffraction?

  3. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    Diffraction is related to the actual size of the aperture, not the f-stop (which is the ratio of actual size to focal length). The smaller the aperture, the more diffraction you'll get. Therefore, the diffraction due to a certain f-stop will reduce as the focal length increases.
  4. piggsy

    piggsy Mu-43 All-Pro

    Haha ok, now my brain is beginning to hurt D.P. Gumby style. So if the diffraction due to an f-stop change reduces as the focal length increases, does that mean that say, on the 60mm (or any of these same internal focusing macros we're looking at charts for here), you would have one diffraction behavior you could model at say 1:4, and another when it's focused closest at 1:1 as it focus breathes downwards (down to 40-50mm or so on the 60 2.8 I believe) ?
  5. piggsy

    piggsy Mu-43 All-Pro

    Ooof, another thing. OK, so looking at their tests, we have one we're interested in where we can compare the same lens on APS-C and FF and different mpx - the Sigma 150. Here's the test on FX -


    • Like Like x 1
  6. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    That depends on whether the lens changes the aperture size to maintain the indicated f-stop as the focus distance changes. I'm guessing not, so I'd say diffraction does not change with focus point since the aperture size remains fixed.

    Regarding CA - I'm fairly sure it's not affected by diffraction, but I'm ready to be contradicted!
  7. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    Since diffraction kicks in sooner with longer wavelengths you can get a pseudo CA due to diffraction where the airy disk of the longer wavelength light lands on the 'wrong' pixel and cause an aberration to appear, the shorter wavelengths are not yet affected by diffraction as much thus their airy disk is not large enough to land on the 'wrong' pixel.

    As for the f-stop changing it depends on the lens design, if the lens in question does not extend on focusing it's highly likely the focal length lowers as you focus closer thus changing the effective f-stop... it all depends on the lens in question which may be one reason why one lens has better results than another when stopping down (while the physical size of the aperture does not change, the focal length reduces so unless the lens reports a different aperture based on focusing distance and adjusts the physical size to match the lens in question may be shooting at a lower f-number and thus show better results with diffraction due to the reduced focal length).

    All in all, I personally don't worry about it all that much...
    • Like Like x 1
  8. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    That makes sense.

    Amen to that!!
  9. kwalsh

    kwalsh Mu-43 Top Veteran Subscribing Member

    Mar 3, 2012
    Baltimore, MD
    No, that is absolutely wrong. Diffraction depends only on F-stop and is independent of focal length.

    See here: https://en.wikipedia.org/wiki/Airy_disk#Cameras

    You forgot that while the *angle* of the first null in the Airy disk is dependent on the physical aperture size that the *size* of the projected disk on the sensor is magnified for a longer focal length. Thus at the sensor plane the size of the Airy disk is not dependent on focal length but only on F-stop.

    In the macro regime the diffraction does depend on other factors, like exit pupil distance and so forth.
    Last edited: Oct 25, 2015
  10. kwalsh

    kwalsh Mu-43 Top Veteran Subscribing Member

    Mar 3, 2012
    Baltimore, MD
    Yes, that's roughly true. Though only longitudinal CA will be reduced as stopped down, lateral CA will remain constant.

    Key point here is you are most noticing (or at least mentioning) here what happens for OOF. Pretty much all the "basic" optical theory only applies to the image plane - i.e. the focused image. Once you are in an out of focus area much intuition fails.

    However, that said there is an easy explanation for why one lens stopped way down looks worse than another stopped down to the same F-stop. Asymmetric aperture blades. All that diffraction theory is based on a circular aperture. Of course at small F-stops apertures aren't circles, they are a polygon which makes diffraction somewhat worse. Now, they are suppose to be equilateral polygons. Unfortunately in many cases they very much are not, especially at very small apertures. If you get a wacky shaped aperture that will make diffraction significantly worse at a given F-stop and it will make OOF areas look like crap. I have no way of knowing if this is actually what is happening in your case, but it is easy to check. Stop the lenses way down and take a look at the aperture. Is it symmetric?

    The same thing, or similar thing, could be happening to the photozone copy but of course we have no way to know. But that huge drop at F/22 is suspicious.

    As to the CA result increasing that is a testing artifact. The testing method photozone (and many other sites use) is very susceptible to garbage in/garbage out results. A nice write up of the limitations of that test method here:


    And a specific quote from that article regarding rising CA measurements:

    So take all that stuff on photozone with a big grain of salt.

    Photozone, or any other MTF based test chart sites, does not produce data that can be compared across formats at all. It can't even be compared across different cameras in the same format.

    Macro gets more dicey. Now there are a lot of aspects of the lens design (changing focal length, different exit pupil locations) that make the standard "diffraction only based on F-stop" rule no longer applicable. Diffraction depends only on F-stop as long as the lens is focused near infinity. With lots of lens extension (i.e. high magnification, macro) that's no longer accurate. Two lenses with the same focal length and set to the same aperture can actually end up with different Airy disk sizes at the sensor because of differences in the lens design. Focus those same lenses at low magnification (closer to infinity) and those differences will disappear.
    • Agree Agree x 1
    • Informative Informative x 1
  11. Conrad

    Conrad Mu-43 Veteran

    Quite badly. It can be done for comparing resolution (I did it for part of the Photozone data across formats) if you allow for an anti-aliasing fudge factor, but it is not accurate enough to do a prediction from one format to the other. You certainly do not have enough information available to do it for CA: their testing method is too flawed (like kwalsh explains).
    No. A highly skilled engineer would need a full multiwavelength raytracing model with all correct glass and coating properties (wavelength dependent refractive index for every glass type used, usually company secret information).
    I can think of a few, but the Photozone measurement method is too flawed to know what is happening in a specific lens' case. In the case of the Olympus 60, however, the giveaway feature is its wide open behaviour. It's actually one of the worst performing "better-than-kit"-lenses at f/2.8. In other words, the normal aberrations will likely hide the CA at f-numbers below 5.6.
    On the other hand, be careful when extrapolating the Photozone measurements to the macro domain. Modern internal focus macro designs are quasi zoom lenses with two moving groups. Their whole optical formula becomes different at close focus distance. To truly test a macro lens you would need a resolution target with features in the range of 1-100 micrometer.
    You do know that dof is only depending on magnification and f-number (and not focal length)? :) 
    • Informative Informative x 1
  12. Conrad

    Conrad Mu-43 Veteran

    I would have formulated it as "Macro gets more dicey. Now there are a lot of aspects of the lens design (changing focal length, different exit pupil locations) that make the standard F-stop label on your lens no longer applicable. The F-stop as labeled on a lens is only true as long as the lens is focused near infinity. With lots of lens extension (i.e. high magnification, macro) that's no longer accurate."
    • Agree Agree x 1
  13. kwalsh

    kwalsh Mu-43 Top Veteran Subscribing Member

    Mar 3, 2012
    Baltimore, MD
    Yes, I think that is a better way of putting it and emphasizes that really diffraction is down to F-stop in photographic imaging but that F-stop is a bit more complicated than what non-macro shooters are used to.
  14. piggsy

    piggsy Mu-43 All-Pro

    Oh, now that's something I'd be interested in delving into as well - again the rule of thumb I'd always heard was that 1:1 involves two added stops beyond the aperture the lens/camera 'thinks' the lens is really at. My assumption was that this was kind of like "internal extension" - similar to the change or light loss from adding the same amount of extension tubes you'd need to use to reach 1:1? Does that apply further down the line where 2:1 is say four stops beyond indicated? Is that amount of light loss another thing that can be influenced subtly or not so subtly by the lens design?
  15. kwalsh

    kwalsh Mu-43 Top Veteran Subscribing Member

    Mar 3, 2012
    Baltimore, MD
    Yes, the amount of extension is conceptually the same whether it happens internally to the lens or with an extension tube or bellows. However, when using an extension or bellows you know that nothing is changing other than the distance the lens is from the sensor plane. When you turn the focus ring on a modern macro lens there may be a lot of other things happening as multiple lens groups move independently. As a result when using a macro lens that has say native 1:1 support without a tube there is a chance that as you move from 1:3 to 1:2 to 1:1 that other lens parameters like the focal length and pupil ratio are changing at the same time.

    The equation for working or effective F-stop is:

    F*(1+M/P) where F is the nominal F-stop, M is the magnification and P is the pupil ratio. If we pretend P is 1 then we get your rule of thumb that at 1:1 the effective F-stop is two times the nominal F-stop and hence two stops different. At 2:1 You'd multiply the nominal F number by 3 and at 3:1 you'd multiply by 4.

    Yes, see that P above - the pupil ratio.

    Keep in mind the effective F-stop thus affects both light gathering, depth of field and diffraction. Also be aware that changing the design of the lens for a different pupil ratio does nothing to change the limits of depth of field and diffraction, it just changes where they occur on your indicated "nominal" F-stop.

    This post gives a good summary:

    • Informative Informative x 1
  16. piggsy

    piggsy Mu-43 All-Pro

    Ah, that's something that hadn't clicked for me - I had somehow assumed it was more like "the performance of 2.8 with the light you'd get at 5.6" but it's actually genuinely all-the-way-down like setting 5.6 for real in terms of other behaviors too?

    Aaaand I'm now coming to understand why people generally leave it at the rule of thumb part. Holy crap :D 
  17. kwalsh

    kwalsh Mu-43 Top Veteran Subscribing Member

    Mar 3, 2012
    Baltimore, MD
    Exactly, it acts like F/5.6 in pretty much every way. Though of course in the case of DoF that is already extremely complicated at macro - so at high magnifications acting "like F/5.6" still might not be at all like you expect for DoF ;) 

    Yep, through the lens metering is just a whole lot easier than all that arithmetic. Not to mention no manufacturer reports pupil ratio so you'd have to measure it yourself to use the equations...
  18. jyc860923

    jyc860923 Mu-43 Hall of Famer Subscribing Member

    Feb 28, 2012
    Shenyang, China
    I'm thinking of the same thing, so does that mean we need relatively lower f stop to avoid diffraction with wider lenses?
  19. tkbslc

    tkbslc Mu-43 Legend

    Most of these charts you are showing are from different sensor sizes and megapixel resolution. Both of those are key inputs for diffraction limited apertures. FF doesn't hit diffraction until double the f-stop of m4/3. APS-C is about a stop away. And 8MP and 16MP will have different diffraction limits. You are seeing this when you compare lenses tested on different cameras.

    For example, a 16MP 4/3 sensor is diffraction limited at f8, but a 12MP is not.
    A 16MP APS-C is not diffraction limited at f8, but a 24MP sensor is.

    Of course an APS-C sensor needs a higher f-stop for same DOF and higher resolution can offset some diffraction issues, so it's not as easy to just say which is better.

    So in short, you are comparing lenses when you should be comparing sensors. Diffraction is a sensor issue.

    Longer focal lengths increase diffraction at the same rate that smaller apertures increase diffraction. So the larger aperture of a telephoto at the same f/stop is offset by the longer focal length. Hence diffraction limit is the same on a given sensor, regardless of focal length.

    Good write up:

    See note at bottom for diffraction vs focal length.
  20. piggsy

    piggsy Mu-43 All-Pro

    Well, I'm not, really, I was asking to what extent you could transfer these results between formats, and all but one of the first set of tests (the oly 60) were on (AFAIK) the same 10mp APS-C body :D 

    The question about CA and lens design seems to have been answered by kwalsh upthread -

    However as to the actual impact of lens design we still see some stuff that seems worth looking into even on the same camera test bed - whether this is more GIGO is something I don't know, but it's interesting to me anyway. Like - why is the peak resolution of the Tamron 180 at F8, and F5.6 on the Sigma 150? Why is the drop from the peak to F22 20% on the 180, and 28% on the 150? If we look at CA between F5.6 and F22 on the 180 and 150, why does CA increase 29% on the 150, and decrease 89% on the 180? Again possibly answered by the GIGO factor but I was interested as to whether it was a design choice or compromise or what, but certainly it seemed to run counter to the way I thought stopping down a lens worked.

    Not everything is format wars :D 
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.