Pages: [1]   Go Down

Author Topic: DOF vs diffraction in high resolution photography  (Read 698 times)

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6397
DOF vs diffraction in high resolution photography
« on: May 12, 2019, 06:15:12 pm »

With a new wave of cameras offering 100MP or 150MP, and even higher with multi-shot modes, I am wondering again how severe the depth of field vs diffraction restrictions on aperture choice become when one wants to view such images at "full resolution", meaning something like 300PPI prints viewed from one foot away. Every doubling of pixel count goes with a doubling of the print size in area (A4 -> A3 -> A2 ..., or 8 1/2 x 11" -> 11 x 17? -> 17x22" ...) while viewing from the same distance in order to see the benefits of the extra pixels.

The constraint I see is that with each doubling of "image detail" (pixel count, print area),
- the maximum usable aperture ratio that prevents diffraction from erasing the resolution benefits is reduced by one stop
- viewing at the same image size (e.g. standard inkjet print size from one foot) would reduce DOF by "one stop" (a factor of 1.4), but
- the extra 1.4x enlargement expands the size of OOF effects on the print by another 1.4x, so that in total the DOF is halved with this "full resolution viewing": circle of confusion on the print are doubled in diameter.

I will characterize this DOF with an "equivalent f-stop", meaning the f-stop giving the same perceived DOF if one were viewing the image "as a whole" rather than scrutinizing part of a very large 300PPI print from close range; that is, as if one were viewing so-called "normally" from a distance comparable to the image diagonal. And using 36x24mm format equivalence as a point of reference.

With that "normal viewing", f/22 shows distinct diffraction softening in normal viewing of prints from 35mm format, and seems to distinctly hamper sharpness even of an 8 1/2 x 10" 300PPI print, which is about 8MP, so very optimistically, f/22 limits one to at most 12MP worth of detail  At that level diffraction spot (Airy disk) diameter exceeds two pixel widths, so more likely f/16 limits one to about 10MP worth of detail!

Let me call this 12MP "A4 print size", based on a bit over 300PPI. Some DOF limits in terms of "effective f-stop" referred to 36x24mm format are then:

12MP ("A4", 11 x 8 1/2") diffraction limit f/22 or less for full resolution, DOF effective f-stop f/22 or less.
24MP ("A3", 17x11") diffraction limit f/16, DOF effective f-stop f/11
50MP ("A2", 22x17") diffraction limit f/11, DOF effective f-stop f/5.6
100MP ("A1") diffraction limit f/8, DOF effective f-stop f/2.8
150MP diffraction limit f/6.5, DOF effective f-stop f/1.9
200MP ("A0") diffraction limit f/5.6, DOF effective f-stop f/1.4
400MP diffraction limit f/4, DOF effective f-stop f/0.7
"Giga-pixel": diffraction limit f/1.5, DOF effective f-stop f/0.28

This gives me the suspicion that the last four or more of these levels of image detail are limited to use cases like
- reproducing paintings and other "2D" subjects,
- scenes with everything at a great distance like certain landscapes having no foreground subjects of interest, (e.g. hyperfocal imaging of distant mountains), and
- "earth from above" photography (from planes or surveillance drones or satellites), where again the subject is almost "2D", requiring very little DOF.

Does anyone actually working at 100MP or above care to comment, refute, support, or illustrate DOF at this level of image detail?
Stitching counts, since changing to a larger format, including doing this effectively by stitching, has no effect on the resolution vs DOF vs diffraction trade-offs; it just shifts the actual f-stops higher, but not the sizes of the circles of confusion of diffraction spots sizes as seen by viewers of the final prints at a given print PPI.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 9977
Re: DOF vs diffraction in high resolution photography
« Reply #1 on: May 13, 2019, 10:43:47 pm »

Someone needs to do some practical tests to answer your questions, BJL.

I recall about 10 years ago or more, after I'd bought the 10 mp Canon 40D, I was a bit annoyed with myself when Canon released the 15 mp 50D just a few months later. I couldn't justifying buying yet another camera so soon. However, after I began experiencing autofocus inaccuracies with the 40D, because it didn't have any autofocus fine-tuning, I was able to justify buying the 50D.

I was curious what benefit those extra pixels would provide. So I took some test shots using my best quality prime with both cameras, shooting the same subject in the same lighting, with camera on tripod.

The results were rather pleasing. The 50% additional pixel count of the 50D allowed me to get approximately the same level of sharpness and detail as the 40D, at the same point of focus, when the 50D was stopped down one full stop.

In other words, the 50D was noticeably sharper at the same F/stop, and the degree of that extra sharpness was equivalent to the negative effects of diffraction resulting from stopping down one stop.

Of course, when pixel-peeping, the differences in sharpness were not exactly equivalent to one F/stop difference across the whole range. As I recall, the lens I used was sharpest at F4, and the 50D was slightly sharper at F5.6 than the 40D at F4, and slightly sharper at F8 than the 40D at F5.6. At F11 the sharpness of the 50D was the closest to the 40D at F8. At F16 the 50D was not quite as sharp as the 40D at F11, but as close as matters. However at F22, the 50D was clearly not as sharp as the 40D at F16. This was the point, at F22, where the effects of diffraction overwhelmed the benefits of the extra pixel count.

I'm not sure how relevant such a comparison of 35mm cropped formats is, but it's interesting to note that the pixel size of these APS-C and DX formats is roughly equivalent to the Medium Formats pixel densities ranging from 60 mp to 133 mp.

Using the sensor size of the Phase One XF IQ4 as a standard (53.4 x 40 mm), the Canon 40D sensor, extrapolated to the MF size, would be 63 mp. The 50D sensor extrapolated to the same MF size would be 94.5 mp, and the 24 mp Nikon D7200, extrapolated to the same MF size would be 133.4 mp.

I suspect that at F22 a 150mp MF format would show no resolution advantage compared with a 100 mp or even 60 mp MF format. However, at wider apertures, the advantage is, greater DoF when stopped down, without sacrificing resolution at the point of focus, as well as a sharper image at the sharpest f/stop of a good lens.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 935
Re: DOF vs diffraction in high resolution photography
« Reply #2 on: May 16, 2019, 03:17:15 pm »

I'd probably be one stop less optimistic (a standard diffraction calculator is close to two stops less optimistic (f6.3 for 50 MP 24x36), but I'd split the difference - looking at large prints, I use my 46 MP camera very happily at f8, but NOT f11). 46 MP 24x36" is a reasonable starting point for two reasons. First, I have one and have examined the prints from it closely against 24 MP APS-C (which I also have in my bag) and 16 and 20 MP Micro 43 (have owned, and a good friend whom I print for shoots Micro 43 extensively). Second, a lot of LuLa folks have 40+ MP 24x36mm cameras - there are a lot of Sony A7r II and III users here, quite a few of us have D850s and an increasing number of Z7s, and the S1r is beginning to show up.

I agree strongly with your basic point, that resolution can't keep going up without the cameras becoming less and less practical to use. I'd add one additional factor to your calculations - lens performance. It's a lot easier to make a lens that resolves 50 MP at f8 (a good lens) than 200 MP at f4 (an Otus on a good day)!

With good sharpening, I'm VERY happy with prints from 46 MP 24x36mm at 24x36" (it was a revelation how much better that combination is than 24 MP APS-C at the same size), and I can go somewhat larger (I've tried cropped 24x46" with good results). I wish I had a 44" printer - my Z7 makes my Canon Pro-2000 feel small.

With my lenses (f4 S Nikkors that are pretty good at f4, but sharpen up noticeably at f5.6 - I like zooms for landscapes a long way from the trailhead), I have a choice of f5.6 and f8 for best results - I'll use either f4 or f11 in a pinch. If I got myself a 50mm f1.8 S, I'd add f4 (without reservations), f2.8 and maybe f2 to my list of choices.

In practical use, 46 MP 24x36mm feels like it's got about a third of a stop more before running into diffraction (good at f8) than 24 MP APS-C (which softens slightly by f8) and both have a bit more than 16 MP Micro 43 (softens slightly between f6.3 and f7.1).

Of course, the smaller formats actually have more DOF, because those apertures translate to different DOF on different sensor sizes.  If I were willing to accept the print size restriction of a 16 MP camera due to diffraction, I could of course stop the 46 MP camera down to f14 and get the same DOF as the 16 MP camera, but only at a similar resolution (thus a much smaller maximum print size).

I'm not sure that I understand your "DOF effective f-stop" - I would think that stepping back to take in the whole print would mean that more appears "sharp" , not less? Billboards work because they are viewed from a long distance, even though they're not actually sharp? If I'm missing something, it only supports your point more - the calculations below get less practical, not more...

As 24x36" sensors get more resolving, the diffraction limit gets tighter (as BJL says). It also becomes more challenging for lenses, as two things are happening at once. First, the lens has to resolve more and more to keep up with the sensor, and second, lenses are being forced wider than their sweet spot.
A 100 MP sensor would be diffraction limited around f5.6 (only one "good f-stop" on most zooms). Only the finest zooms and good primes will have any chance at all at any stop.

A 200 MP sensor is limited around f4 (and the resolution is so high that only a good stopped-down prime is taking any advantage of it) -  Sigma Art, Nikkor S, Sony G, Canon L or better at its best stop MIGHT capture more detail at 200 MP than at 100 MP.

A 400 MP sensor is limited around f2.8 (and the lens had better say "OTUS" on it to have any chance at all - fortunately, the best stop on the Otus range is f2.8 ).

A gigapixel sensor is diffraction limited at f1 or so - WHAT lens resolves a gigapixel at f1? If there's anything at all, it says "Property of the CIA" on it, is permanently mounted on a spy plane, and is impractically heavy for any other use

By the time you reach 400MP at f2.8, the conditions for a photograph become REALLY limiting. ONLY an Otus, ONLY on the finest tripod, ONLY at f2.8. That's really only for art reproduction and the like, since even "all-infinity" landscapes will be limited by atmospheric effects. All-infinity landscapes are rare - even Ansel Adams rarely had no important foreground detail (I can't think of one, although Monolith - The Face of Half Dome and Mount McKinley and Wonder Lake come relatively close).

Even 200 MP becomes so constrained that it severely limits options for using the full detail of the camera. 100 MP is tricky to handle, but it still has a significant usable range.


Where this gets even more interesting is as the sensor size changes... Going towards medium format, we're mostly OK (although the increased coverage requirement makes lenses trickier, that's counterbalanced by the extra leeway from the bigger sensor).

50 MP small medium format is diffraction limited at f11 or so (gives us one extra choice).
100 MP small medium format and 150 MP big medium format  (Phase size - close to 645 film) are diffraction limited around f6.3 or f7.1 (getting restricted, but still workable). The next generation of 24x36mm sensors (around 60-65 MP) will be very similar to this, as the pixel size is the same.

Going smaller, we start to hit limits - for a while, the ability to build sharper lenses for smaller formats really helps (Fujinons and Olympus PRO lenses are really sharp). Eventually, the required lenses become absurd - a bigger sensor would be lighter and cheaper.

26 MP APS-C behaves like 150 MP Phase, 100 MP Fuji MF and theoretical 65 MP 24x36" (all are the same size pixels, and limited at f6.3-f7.1).  That's still very reasonable since you don't need huge coverage - there are plenty of lenses  that will give you that performance, including pretty much any Fujinon and better lenses from other makers.

APS-C in the 32 MP range would be diffraction limited around f5.6.

APS-C in the 50 MP range would be diffraction limited around f4.5. This starts to be tricky for a lot of lenses - consumer zooms may not open that wide, and are certainly being tested (and failing) on their wide-open performance. The best zooms (red-badge Fujinons) possibly have a single non-diffraction-limited aperture that resolves that well, especially in the center of the frame. Really good primes have a few apertures to choose from.

Micro 43 is limited around f6.3 at 16 MP, f5.6 at 20 MP. Olympus' PRO lenses give quite a bit of flexibility with the existing sensors, although the consumer zooms get quite tricky with maximum apertures close to the diffraction limit, while they benefit from stopping down. Even the PRO zooms have a slight problem at the other end - you can't get much bokeh because the widest aperture (fortunately, those lenses are highly usable wide-open) is still equivalent to f5.6 on 24x36mm. The primes open wide enough (and are usable enough wide open) that they aren't a problem.

A hypothetical Micro 43 sensor at 40 MP would be diffraction limited by f4 -the PRO zooms might well still have one good aperture (the f2.8 models almost certainly resolve that well stopped down one stop, while some of the f4s, particularly the new telephoto, might perform well enough wide open). The best primes have a narrow, but usable range on a 40 MP sensor.

As sensor sizes reduce farther, the corner gets tighter...

A 1" sensor at 20 MP (the common current sensor) is diffraction limited at f4 - the long end of your RX100 VI's zoom can't use the full sensor resolution, the wide end has a lot of distortion correction and the middle of the range uses the full resolution only at a single aperture. A 20 MP 1" camera with a good prime would have a couple of full-resolution apertures.

A 2/3" sensor is diffraction limited at f4 at around 12 MP, f2.8 at around 24 MP. It's JUST possible to build a high-resolution (24 MP) compact with a very good prime that has a single aperture with full performance (remember how good that lens has to be to get 24 MP onto the small sensor). Somebody's probably built one once, although B&H doesn't presently have any. The 12 MP zoom cameras in this class might use their full sensor resolution somewhere in the zoom/aperture range if they have really good lenses.

A 1/2.5" sensor (cell phone) is at best diffraction limited at f4 around 6 MP, f2.8 at 12 MP. The sensor in an iPhone XS has an active area of 5.6 x4.2mm (it's a little bigger, but that's the area used to form the image) - it has the same pixel density as a 440MP 24x36mm sensor (e.g. the impossible sensor that is "Otus-only"). Does anyone think that little phone lens has the center performance of  an Otus (it only needs center performance, because the chip is tiny)? It may not be diffraction per se (the phone lens is fast enough for 12 MP, although not for the higher-resolution phone chips we're seeing), but the required lens performance is just not realistic for a lens that is built to be tiny.







« Last Edit: May 16, 2019, 03:37:38 pm by Dan Wells »
Logged

bjanes

  • Sr. Member
  • ****
  • Online Online
  • Posts: 3348
Re: DOF vs diffraction in high resolution photography
« Reply #3 on: May 17, 2019, 07:29:07 am »


I'd probably be one stop less optimistic (a standard diffraction calculator is close to two stops less optimistic (f6.3 for 50 MP 24x36), but I'd split the difference - looking at large prints, I use my 46 MP camera very happily at f8, but NOT f11). 46 MP 24x36" is a reasonable starting point for two reasons. First, I have one and have examined the prints from it closely against 24 MP APS-C (which I also have in my bag) and 16 and 20 MP Micro 43 (have owned, and a good friend whom I print for shoots Micro 43 extensively). Second, a lot of LuLa folks have 40+ MP 24x36mm cameras - there are a lot of Sony A7r II and III users here, quite a few of us have D850s and an increasing number of Z7s, and the S1r is beginning to show up.

Dan,

Your post and BJL's post make a lot of sense and have important implications for practical photography. Rubén Osuna and Efraín García have an excellent post on LuLa that provides a scientific explanation for these matters and their post is well worth re-reading. I agree that with the D850 f/5.6 is the optimum aperture, but f/8 gives reasonably good results, especially with proper deconvolution sharpening. With such sharpening f/11 and f/16 can be used in a pinch, but f/22 is beyond the pale.

What constitutes proper deconvolution sharpening is an important topic. FocusMagic is the old reliable standby. ACR/LR with a small radius and the detail slider to the right can be used if one is in a hurry, and Topaz Sharpen AI can be used if one has plenty of time. Phase One Capture One has a slider for diffraction removal, but I have not used that software. I would be interested in what others are using.

Cheers,

Bill
Logged

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2153
    • Pieter Kers
Re: DOF vs diffraction in high resolution photography
« Reply #4 on: May 17, 2019, 09:24:36 am »

Tilt-techniques and Focus stacking is the way to go with these limits i guess...
A good implementation in a highMP-camera is worth a lot.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 935
Re: DOF vs diffraction in high resolution photography
« Reply #5 on: May 17, 2019, 02:38:05 pm »

One thing we sometimes miss as we try for the best (and I'm as guilty of this as anyone) is just how good what we're using now is... These 40+ MP cameras, used with IS/VR/whatever else you call a stabilized lens or body, are shooting the equivalent of 4x5" film handheld, with a few hundred shots per film holder. I'm making 24x36" and larger prints from images taken in the backcountry, using 1 kilogram of gear and no tripod - prints that used to require 4x5". The best of 24 MP APS-C is in the range of 6x7 cm or 6x9 cm film. Good  Micro 43 is in the range of 645 film - if you had a 645 camera that shot 20 FPS, fit in a fanny pack, and was nearly indestructible. The Phase One backs are tripod-bound, but hasn't 8x10" always been?

Of course, none of this has movements, and view cameras do. That's where things like tilt and focus stacking come in... Lloyd Chambers is getting interesting results out of the high-res multishot mode on the S1r as well.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6397
Re: DOF vs diffraction in high resolution photography
« Reply #6 on: May 20, 2019, 09:28:59 am »

Dan and Bill,

    Thanks for your responses; I think we all basically agree, so let me just clarify some points from my post.

50MP ("A2", 22x17") diffraction limit f/11, DOF effective f-stop f/5.6

I'd probably be one stop less optimistic (a standard diffraction calculator is close to two stops less optimistic (f6.3 for 50 MP 24x36), but I'd split the difference - looking at large prints, I use my 46 MP camera very happily at f8, but NOT f11).

I agree that with the D850 f/5.6 is the optimum aperture, but f/8 gives reasonably good results, especially with proper deconvolution sharpening. With such sharpening f/11 and f/16 can be used in a pinch, but f/22 is beyond the pale.

For that roughly 50MP sensor in 36x24mm (as for the"D850", close enough) my f/11 is intended as the optimistic upper end of the viable f-stop range, at which point the diffraction is rather dominant over sensor resolution in determining the overall output image resolution, so that further increases in sensor resolution (pixel count) will give no significant increase in that "overall image resolution". There is of course a gray zone; a range of several stops below that where diffraction holds resolution back somewhat, but an increase in sensor resolution can still give some increase in overall image resolution — just not as much improvement as you would be getting if diffraction were absent. The more pessimistic judgements from "standard diffraction calculators" see to be looking at where diffraction becomes a significant factor in resolution, rather than the dominant one. My informal guideline for that is indeed a full two stops lower, so starting at f/16 for a 6MP sensor (as for a 10" wide print at 300PPI), f/11 at 12MP, and so on.

I will characterize this DOF with an "equivalent f-stop", meaning the f-stop giving the same perceived DOF if one were viewing the image "as a whole" rather than scrutinizing part of a very large 300PPI print from close range; that is, as if one were viewing so-called "normally" from a distance comparable to the image diagonal.

I'm not sure that I understand your "DOF effective f-stop" - I would think that stepping back to take in the whole print would mean that more appears "sharp" , not less? Billboards work because they are viewed from a long distance, even though they're not actually sharp? If I'm missing something, it only supports your point more - the calculations below get less practical, not more...

I think I did not explain my weird new DOF equivalence measure well enough! You are right of course that stepping back increases perceived sharpness. What I am trying to measure is what happens when one moves closer to a large, high resolution print, viewing it closer than "normal" for a wider viewing angle (a larger "apparent image size) in order to see all the details.

For example, start with a 6MP f/8 image printed at 300PPI, so 10" wide, viewed from about one foot away for a "normal apparent image size". Then doubling resolution by going to a 24MP image again at f/8, again printing 10" wide (but now 600PPI) and again viewing from the same distance does not change the perceived depth of field and OOF effects. Nor does printing it at 300PPI, so twice as wide (20"), and viewing from twice as far away (two feet) so as to see the whole scene at once with that same "normal viewing angle" and "normal apparent image size".

But if you instead make prints at the same 300PPI (10" wide from 6MP; 20" from 24MP) and view from the same distance so as to see all the detail in each image, then the OOF effects are twice as visible in the larger print from the higher resolution image, so that the perceived DOF is halved—it is as for "normal" viewing of a photo of the same scene taken at f/4 instead of f/8. That is the viewing scenario that my "f/4 equivalent DOF" refers to. Some might refer to it as 100% pixel peeping of images, but I want to refer it simply to how big one displays and views the final images in order to see them in their full glory!
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6397
DOF vs diffraction in high resolution photography—and focus stacking
« Reply #7 on: May 20, 2019, 09:35:41 am »

Tilt-techniques and Focus stacking is the way to go with these limits i guess...
Good point; I was a bit too pessimistic about the limits on use cases. Better to say that when one wants "large format" levels of image detail, one has to accept some "large format" constraints on careful composition and technique. The good news is that with multi-shot techniques like focus stacking, one can achieve images of very stationary subjects more detailed than were ever possible with film in any format.
Logged

Dan Wells

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 935
Re: DOF vs diffraction in high resolution photography
« Reply #8 on: May 20, 2019, 11:37:44 pm »

Yes, I see your new measure - and I like it - high-resolution cameras, well-wielded, are capable of astonishing detail which pulls you in to a print closer than "normal", and which nothing but large format could do in the film days - and that does affect DOF in the direction you describe.

I'm amazed at the large format type results we're getting with the current generation - although DOF is a challenge that will, as you say, require more large-format type techniques in many settings...
Logged

Lightsmith

  • Full Member
  • ***
  • Offline Offline
  • Posts: 186
Re: DOF vs diffraction in high resolution photography
« Reply #9 on: June 01, 2019, 02:33:45 pm »

Diffraction is less of a problem with my D850 and its 45MP sensor than it was with my D3 with itss 12MP sensor. Others have also experienced this in the field. Nikon when it introduced the 36MP D800 camera was cautioning people to avoid apertures smaller than f/8 but while this may have been a measurable point where diffraction would start to increase it still begs the question as to the relative importance of the sensor as compared to the lens for overall image resolution.

I do not agree with the concept of hyperfocal distance for photographing a scene. It assumes that everything in a scene is of equal importance and must be equally sharp and for me the foreground is more important than objects in the distance.

Best way to deal with DOF is to use a tilt shift lens.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6397
Re: DOF vs diffraction in high resolution photography
« Reply #10 on: June 01, 2019, 06:31:56 pm »

Diffraction is less of a problem with my D850 and its 45MP sensor than it was with my D3 with itss 12MP sensor.
Please explain! Are you comparing prints of the same size viewed from the same distance, or is the comparison with the higher resolution image viewed proportionately larger/closer so as to see the extra details that the sensor is capable of resolving? The former comparison will of course give a sharper image with the higher resolution sensor, as will any increase in sensor resolution at equal f-stop, though only to a diminishing degree as diffraction becomes dominant.

Nikon when it introduced the 36MP D800 camera was cautioning people to avoid apertures smaller than f/8 but while this may have been a measurable point where diffraction would start to increase ...
Indeed, f/8 for a 36MP sensor is very cautious; as you say, that might be where diffraction starts to slightly limit the resolution that the sensor is capable of, but is far tighter than the roughly f/13 that my guidelines above suggest as a limit beyond which higher sensor resolution delivers no significant increase in image sharpness at a given f-stop.
Logged
Pages: [1]   Go Up