Diffraction in the real world

One issue that comes up frequently in discussions is that of diffraction, an issue that gets worse with small photosites (high megapixels, small sensors). Low pixel density FF sensors (such as the 5D, D3, D700) are the best to combat the loss of sharpness due to diffraction at small apertures. And the theoretical limits of systems are often quoted.

 

For my own camera, an aged D2X (with high pixel density), a diffraction limit of F11 is cited (to quote Thom Hogan "On a D2x it seems to be around f/11. Up to f/11 and the D2x's acuity appears as you'd expect. Beyond f/11 and you'll likely start saying to yourself that the results don't look like you're getting the full impact of more DOF.").

 

However, this is not what I see in my underwater images. And my only conclusion is that by the time you are shooting through port glass, murky water etc, diffraction is not the biggest of your worries. Even in the lenses sweet spot (F8-F11) you loose acuity for these other factors. So stopping down the issues of diffraction are not so significant?

Yes. If dirty water is the limiting factor, getting closer is only way to improve resolution.

I think the term "limit" in this context refers to the system no longer limited by the resolution of the sensor at F/11 and the lens is more of a limiting factor. The extra megapixels you got from upgrading your D100 is not going to improve your image. Of course if you have a murky medium, that could be the limiting factor rather than the lens.

If stepping down to F/25 does not degrade the image then the sensor is definitely not the limiting factor, and as Craig mentioned you can probably get the same results with a D100 or D2H.

If you know you are diving in these conditions, a wider lens may be a better choice. The image quality will be limited by how close you get to the subject. If the subject is still too small to fill the frame at this distance, you can crop and not lose anything compared to a longer lens that will frame fill the subject. The water is still the limiting factor. I think that's why the film shooter here in northern California used to prefer the 60mm over the 105mm.

... My standard image delivery size is 45x30cm (c. 18"x12") @ 300DPI, which is just under 19 megapixels. With a D2X (12MP) this requires up-rezzing (which is why I want a new camera), which I tend to do in one step in Lightroom, then re-sharpen, lightly in Photoshop (leaving final sharpening to client when image repro size is decided).

I find this hard to understand especially in the context of the rest of the post. Are you saying you are going to buy a new camera, housing, and lenses if you change to FX, just to skip the step of upsizing, with no expectations of increased image quality?

Personally I've noticed nasty, image damaging, diffraction in wide shots (12-24mm) at f/16-22, topside but often shoot macro at those apertures and smaller with no noticeable loss in sharpness.

just what i've observed, perhaps because my macro's are sharper than my wides, or the change in fov makes a difference.....

i don't know much about this stuff, but how much of a roll does lens quality play in diffraction versus sensor size/pixel density etc? surely a piece of pro glass will transmit light better through a tiny aperture opening than a consumer lens?? will the same camera with a cheap lens produce more diffraction?

chris

OT: Is that a tiny nudi under the big nudi

... wish we can resolve what it is :lol:

For my own camera, an aged D2X (with high pixel density), a diffraction limit of F11 is cited (to quote Thom Hogan "On a D2x it seems to be around f/11. Up to f/11 and the D2x's acuity appears as you'd expect. Beyond f/11 and you'll likely start saying to yourself that the results don't look like you're getting the full impact of more DOF.").

The quoted statement is unjustifiably broad. At each point in the image, diffraction and defocus both contribute to the blur. Without knowing how much the subject deviates from the plane of focus, you can't really say at what point diffraction will begin to dominate.

I've used f20 @ 1/200, ISO100 on a D80 with a 105 +4T diopter. It's all dissapointingly soft to me and I really wonder sometimes why I bother shooting these wretched things. I'm really not sure how to get a sharper image.

 

I usually aim for a least f22 for max depth of field this type of thing but hold back from f32 because of a supposed loss in optical quality.

With this shot I might have focussed too far forward but I'm not sure.

Perhaps at least I can discount diffraction related softness as the main problem?

Trevor

A classic example of the complexities of determining the 'problem'. Personally I'd say that adding in a 4T is probably going to be the biggest degradaing element in the optical chain. However I'd also suggest that the type of illumination and the way it is used in images like this is also a critical factor. Diffraction probably doesn't help but in itself isn't the major degrading influence. And as you say focus is itself a problem given the DoF being so shallow. Not easy to produce a really good image of such creatures.

Interestingly, the modern trend in macro lenses is to produce IF (internal focus) versions which produce stunning quality over their 'prime' range (ie infinity to lifesize), but whose IQ (image quality) drops off if any attempt is made to extend the range beyond 1:1. Diopters are probably the 'best compromise' on such lenses to get beyond 1:1, but have their drawbacks. My own experience using the Canon 60EFS is that it works extremely well with a single 12mm of extension tube added but with more IQ drops noticably. I'd suggest that using such lenses outside their range and at very small apertures is probably pushing them a little far. At a guess I'd sau lenses like the older style 60 micro Nikkor will probably perform better if pushed further than 1:1. Would anyone like to comment?

Clarity of discussion:

Saying that a D2X is limited to f/11 is arbitrary. Such an aperture is no more than the point where diffraction reaches a certain level relative to pixel pitch. It does not say that image quality significantly degrades with smaller apertures only that it begins to. Trading slight loss in resolution for more DOF is sometimes a good one though. It seems too few people understand this.

Exactly my point and hence my comment that my personal experience that the trade-off is usually good to f/22 on my DX cameras. I think people get fixated on Craig's diffraction posts as a screed against using small apertures.

The whole point of larger sensors and higher pixel counts ultimately is more resolving power and greater quality at bigger enlargement sizes. Understanding that, it is easy to see how many photographers truly see no need for more pixel count, FX, or discussions of diffraction limits. They are not wrong, but if they assert that those improvements make no difference they are mistaken.

Again, a moment of clarity that is well stated. I am in the camp that sees no need for more pixel count or FX. I am not wrong and I promise not assert that it makes no difference. It's just not in my top two issues when it comes to a final product.

I noticed this post over at DPreview. It links to two other posts that discuss diffraction limits and different sensor sizes.

The author dispels a number of popular myths: that DOF is a function of sensor size, that diffraction limits are a function of sensor size, and that shooting apertures beyond the so-called diffraction limit are not useful. He explains clearly why higher pixel counts are still valuable when you need to shoot at small apertures like we do.

The best treatment of diffraction that I have seen is a site called Cambridge in Colour http://www.cambridgeincolour.com/tutorials...photography.htm It is great.

The Airy disk (named for George Airy) is (loosely speaking) the name for the minimum size feature focussed by a lens.

You can, of course, use any f-stop you want. For Nikon D2X f/8 to f/11 is the point where Airy disk produced by diffraction starts to matter, because it becomes comparable to the pixel size. The Airy disk is the minimum size of anything resolved by the lens due to diffraction. The Cambridge site has a very nice animation of that - you move your mouse over a table to select the combination and it shows you the Airy disk size. Give it a try and this will seem much more natural.

A post above suggests that an article debunks "myths" like DOF is a function of sensor size. Well, only sort of - the thread in DPreview is technically correct but that is not quite what it says. The size of the Airy disk is given by the lens only. However whether this matters depends on the relative size of the Airy disk to the PIXEL size. If you ask the question "what is the smallest aperture before diffraction compromises maximum resolution?" that depends totally on pixel size (not sensor size, not pixel count, just pixel dimensions).

If you go smaller than that aperture (f/11 for D2X) you will get more depth of field, but the overall resolution will suffer. Diffraction is like doing a Gaussian Blur with a size given by the Airy disk - it matters! This is particularly true if you are trying for the utmost in critical sharpness (making very large prints for example).

The traditional way to ask DOF question is "what is the first noticable problem in a print of size X at normal viewing distance". That is quite appropriate if you are only making prints of that size. If instead you have ambitions to make big prints (or small prints made from cropping the image) then you want to know what is the most resolution that you can wring out of the equipment. It's up to you.

However, in many cases you would rather trade off utmost in resolution for depth of field. That is perfectly OK, but know that you are making the tradeoff.

In actual practice many lenses are worse than the diffraction limit - especially wide angle lenses, and especially near the edge. That is true topside - these effects are WORSE yet again for UW photography because the interface between the air in the housing, the lens for the port, and the water acts like an optical element. So, you get in general, worse resolution and other artifacts, especially with wide angle and especially at the edge of the frame.

Lenses frequently have better resolution and fewer abberations as you stop down. This effect can be much larger than the diffraction effect. So even though the best resolution might occur at f/11 or wider, this may be dominated by other abberations that are really bad at wide f-stops and improve at f/16.

Here is a test by Stephen Frink of a Canon lens on the 1Ds Mark II, showing some of these effects http://www.seacamusa.com/canon-lens-14mm-ii.shtml The 1Ds Mark II becomes diffraciton limited between f/11 and f/16. But the f/16 pictures look much better at the edges because the 14mm lens is better there.

This test (again by Frink) shows how the choice of port, and port extender can make a HUGE optical difference in the quality at the edges of the frame for a wide angle lens http://www.seacamusa.com/16-35-test.shtml .

Macro lenses are typically much better than wide angles, both because the lens itself is corrected better, and the port glass is flat.

So, the moral of the story here is that:

- Diffraction can make a huge difference to critical image quality, if you truly need to wring out the most from your equipment.

- But in many circumstances other factors in the picture, or the lens properties may be even more important.

Nathan

Nathan, all you said is true but you did not contradict anything said in ejmartin's posts. You've suggested that his discussion was somehow misleading or incomplete. How so? DOF as a function of sensor size IS a myth. His explanation was correct and complete. Read his presentation in the second link.

Reducing your sensor size will not give you more depth of field, it will only reduce the amount of light you need to achieve it and lower your SNR in the process. The only way to get more depth of field is to reduce the amount of resolution you are willing to accept. Lower your print size. People confuse these issues because (a) they expect a larger sensor to have more pixels and thus create an implicit double-standard, and (b) they associate real-world lens characteristics with the formats themselves.

I linked his posts here because he also supported the argument that started this thread. There is no diffraction "limit", only a knee in the curve. He also said very effectively why you may want higher pixel counts even if you are shooting small apertures. I think that's important to say considering the overemphasis placed on these artificially low diffraction "limits".

Fair enough. I think I said that ejmartin's posts are technically correct - and we agree that they are.

As a personal opinion, I thought they are harder to follow than say Cambridge in Colour. It is true that sensor size, *by itself* does not affect DOF, but I would rapidly follow that up with "but pixel size DOES affect diffraction, and can get in the way of DOF". Reducing sensor size to affect DOF is wrong - as we all agree. However, reducing sensor size while at the same time keeping pixel count up does result in smaller pixels, which then DO effect the smallest aperature you can use before diffraction affects you.

So when the D2X came out, many Nikon fans tried to argue that it was superior to have 12 Mpixels in a reduced size sensor. While it is a good camera, that is not an advantage. More recently Nikon fans are hailing the D3, which has very similar pixel count, but gets lots of benefit in SNR from having those pixels be larger.

It is sort of funny that the meaning of "diffraction limited" in optics is that a lens is very GOOD! It means that the other problems are smaller than the fundamental diffraction affect which cannot be eliminated with fancy techniques since it is due to the basic physcis of light. I agree that there is no "limit", but for every camera there is an f-stop past which stopping down costs you resolution.

You can get increase DOF without hurting resolution as long as you are above that point. So, stopping down from f/2.8 to f/11 does get you more DOF on a Nikon D2X, without hurting the resolution because in that range you are pixel-size limited. Past f/11 you have the tradeoff you mention where DOF costs resolution.

Yes, but there's more. Even when you are "diffraction limited", higher pixel counts will provide increased resolution but at a lower rate of return. Furthermore, when you open up, you will get increasing resolution further with a higher pixel count camera. That's why higher pixel count is more desirable provided you don't need the specific capabilities of the large-pixel version. Underwater shooters are better off with a "D3x" rather than a D3 even if they are shooting at f/22.

If you are claiming 12MP in a DX is not an advantage over 10MP then you are mistaken. It's only that the increment may be too small to matter depending on your perspective.

Nikon "fans" are hailing the D3 for a variety of reasons but I have not heard one claim that the D3 offers better resolution than the D300 nor have I heard anyone claim that the D300 is better than the D2x. I have heard some claim that the D3 is as good as the 1Ds3 but they are clearly delusional. A D2x and a D3 have essentially the same resolving power and are capable of the same diffraction-limited depth of field.

The Cambridge in Colour article explains diffraction very well but it makes no attempt to discuss different sensor sizes. Here is the last paragraph:

Are smaller pixels somehow worse? Not necessarily. Just because the diffraction limit has been reached with large pixels does not mean the final photo will be any worse than if there were instead smaller pixels and the limit was surpassed; both scenarios still have the same total resolution (although one will produce a larger file). Even though the resolution is the same, the camera with the smaller pixels will render the photo with fewer artifacts (such as color moiré and aliasing). Smaller pixels also provide the flexibility of having better resolution with larger apertures, in situations where the depth of field can be more shallow. When other factors such as noise and depth of field are considered, the answer as to which is better becomes more complicated.

Essentially, the article agrees with ejmartin. The article makes one mistake, though. In their two scenarios, total resolution is NOT the same unless the aperture is a pinhole. Sampling a diffraction-limited image at a higher frequency can produce marginally higher resolution.