jamesdfloyd
Film is cheap therapy!
Can someone please explain lens diffraction to me?
Now, I don't really need the full-blown technical explanation, but the practical do's & don'ts.
Every time I read one of the current DSLR landscape oriented how to books, the writer always warns about lens diffraction at f/16 or f/22. Yet, a significant number of DSLR lens go to f/22. Almost all medium format lens go to f/22 or f/32. And Large Format lens go to f32, f/45 or f/64.
After reading many sources, diffraction & stopping down to the smallest f-stop seems to have become a debate akin to RAW vs. JPEG or "great taste vs. less filling". That last one is for us Americans who remember early 1980's beer commercials.
What is the "truth" and what is your practice?
J.D.
Now, I don't really need the full-blown technical explanation, but the practical do's & don'ts.
Every time I read one of the current DSLR landscape oriented how to books, the writer always warns about lens diffraction at f/16 or f/22. Yet, a significant number of DSLR lens go to f/22. Almost all medium format lens go to f/22 or f/32. And Large Format lens go to f32, f/45 or f/64.
After reading many sources, diffraction & stopping down to the smallest f-stop seems to have become a debate akin to RAW vs. JPEG or "great taste vs. less filling". That last one is for us Americans who remember early 1980's beer commercials.
What is the "truth" and what is your practice?
J.D.
john_s
Well-known
Diffraction is physics. It depends on the aperture's actual dimension (i.e. diameter as measured in millimeters). A tiny lens in a P&S will be aperture limited to say f8 where diffraction is already a problem. A large format lens is much bigger because of the large film size, and f8 on that lens is a sizeable opening, so no discernible diffraction. They can go down to something like f64. Medium format is somewhere in between.
Phil_F_NM
Camera hacker
Yes, it's true. After a certain f/stop, according to the design, the image will become a bit less sharp (usually only noticeable by folks like Erwin Puts and pixel peepers.) This is due to the angle that the light rays take around the aperture blades.
The fact that medium and large format lenses stop down to f/32, f/45 and f/64 is a part of the design. While diffraction may come into play, the lenses are a LOT larger in focal length and therefore the physical aperture that the light is bent around is much larger than a 35mm format lens with the same field of view. The other issue with large format is that the photographer can live with a bit of diffraction if it means for a greater field being in focus. The enlargement factor of a 4x5 is much less than that of a 35mm negative. This is why a 100 year old Goerz lens in-barrel can beat the pants off of a new Leica 50mm Summicron (or any 35mm format brand lens for that matter). Just because of the enlargement factor. (**I wish I could get a RF coupled Goerz Red Dot Artar for my Leica, just because.)
REalisatically, f/22 and other small apertures are there because you need them. If you shoot a mechanical RF, your top speed is likely to be 1/1000sec. Unless you're shooting 50ISO, sometimes a bright day will top out your available shutter speeds at smaller apertures and so you'll have to shoot at something like 1/1000 @ f/22. Either that or don't get the shot.
Phil Forrest
The fact that medium and large format lenses stop down to f/32, f/45 and f/64 is a part of the design. While diffraction may come into play, the lenses are a LOT larger in focal length and therefore the physical aperture that the light is bent around is much larger than a 35mm format lens with the same field of view. The other issue with large format is that the photographer can live with a bit of diffraction if it means for a greater field being in focus. The enlargement factor of a 4x5 is much less than that of a 35mm negative. This is why a 100 year old Goerz lens in-barrel can beat the pants off of a new Leica 50mm Summicron (or any 35mm format brand lens for that matter). Just because of the enlargement factor. (**I wish I could get a RF coupled Goerz Red Dot Artar for my Leica, just because.)
REalisatically, f/22 and other small apertures are there because you need them. If you shoot a mechanical RF, your top speed is likely to be 1/1000sec. Unless you're shooting 50ISO, sometimes a bright day will top out your available shutter speeds at smaller apertures and so you'll have to shoot at something like 1/1000 @ f/22. Either that or don't get the shot.
Phil Forrest
Juan Valdenebro
Truth is beauty
It sure exists but I feel it's minimal, and I use all f-stops on any lens except the smallest one, and I never see a real sharpness loss...
Cheers,
Juan
Cheers,
Juan
julio1fer
Well-known
It's a physical effect. Diffraction happens when the aperture opening is small enough for the light waves to spill around noticeably. It depends on the aperture size in mm, not in f stops. The more you close aperture after that point, the softer the image. In the limit you might look at those soft pinhole pictures; they can't be sharp because of diffraction.
Now f stops are just the quotient of focal length / aperture diameter. The larger the focal length, the larger the f number can be without risking diffraction (because the physical opening for light is larger as well).
As a rough guide, I would expect that with a 50 mm normal lens diffraction would be be of the order of the film definition at around f/16.
With the very short focal distances of a digital P&S, definition might be limited by diffraction already at f/8 .
It is a matter of focal length and f number, not of film format - so the limiting aperture with a 135mm lens might be something like f/32, even if the lens is in a 35mm camera.
Now f stops are just the quotient of focal length / aperture diameter. The larger the focal length, the larger the f number can be without risking diffraction (because the physical opening for light is larger as well).
As a rough guide, I would expect that with a 50 mm normal lens diffraction would be be of the order of the film definition at around f/16.
With the very short focal distances of a digital P&S, definition might be limited by diffraction already at f/8 .
It is a matter of focal length and f number, not of film format - so the limiting aperture with a 135mm lens might be something like f/32, even if the lens is in a 35mm camera.
semilog
curmudgeonly optimist
Ken Rockwell has a very good article on diffraction here.
Limitations imposed by diffraction start to creep in sooner or later depending on the resolving power of the film or sensor (units: line pairs or cycles per millimeter). Note that we are NOT talking about film or sensor formats. On digital sensors we are talking about pixel size.
Diffraction becomes limiting at wider apertures for higher-resolution systems. Diffraction is less of a problem for low-resolution sensors (e.g., DSLR's with big pixels; large-format film cameras; fast, grainy films). It is more of a problem for high-resolution sensors (point-and-shoots with tiny pixels; slow, sharp films).
Truly great lenses such as the Olympus Zuiko 150/2 are already diffraction-limited at very wide apertures. Look at this resolution chart. In the case of the Zuiko 150, it's already diffraction-limited at f/2.8, and stopping down (even to f/4) gradually, and then severely, reduces image contrast. Resolution is more than cut in half by stopping down from f/2.8 to 22.
Note that the pixel size of the camera used in this test is the SAME as the Canon 7D. Thus, lenses used for 4/3 cameras, micro 4/3 cameras, or the 7D, to extract the sensor's potential, need to be (a) superbly corrected; and (b) used wide open, or close to it.
Note that there is another implication here: wider-aperture lenses are in principle higher-resolution than small-aperture lenses. Why, then, does stopping down a bit improve most lenses? This is because it is VERY hard to correct optical aberrations at wide apertures. A perfectly-corrected lens (few of these exist in real life) would have highest resolution from wide open, and stopping down would immediately compromise resolution. The 150mm Zuiko is unusually close to this standard.
Limitations imposed by diffraction start to creep in sooner or later depending on the resolving power of the film or sensor (units: line pairs or cycles per millimeter). Note that we are NOT talking about film or sensor formats. On digital sensors we are talking about pixel size.
Diffraction becomes limiting at wider apertures for higher-resolution systems. Diffraction is less of a problem for low-resolution sensors (e.g., DSLR's with big pixels; large-format film cameras; fast, grainy films). It is more of a problem for high-resolution sensors (point-and-shoots with tiny pixels; slow, sharp films).
Truly great lenses such as the Olympus Zuiko 150/2 are already diffraction-limited at very wide apertures. Look at this resolution chart. In the case of the Zuiko 150, it's already diffraction-limited at f/2.8, and stopping down (even to f/4) gradually, and then severely, reduces image contrast. Resolution is more than cut in half by stopping down from f/2.8 to 22.
Note that the pixel size of the camera used in this test is the SAME as the Canon 7D. Thus, lenses used for 4/3 cameras, micro 4/3 cameras, or the 7D, to extract the sensor's potential, need to be (a) superbly corrected; and (b) used wide open, or close to it.
Note that there is another implication here: wider-aperture lenses are in principle higher-resolution than small-aperture lenses. Why, then, does stopping down a bit improve most lenses? This is because it is VERY hard to correct optical aberrations at wide apertures. A perfectly-corrected lens (few of these exist in real life) would have highest resolution from wide open, and stopping down would immediately compromise resolution. The 150mm Zuiko is unusually close to this standard.
Last edited:
tstermitz
Well-known
Yes, it's physics, or optics, I guess. Use "f/8 and be there" and you can ignore everything else! Ken Rockwell's explanation linked above is as good as any. Ken concludes:
In terms of resolution, film has grain in the 2-8 micron range. Digital sensors also have pixels of 2-8 microns: The D90/D300 is 5.5 microns, D3/D700 is 8.5, M9 is 6.8, G10 is 1.7. (Let's not get into the RGBG Bayer pattern or anti-aliasing filters, except to say they contribute more blur) Looking at scanners, 4000 dpi is 6.3 micron and 3000 dpi is 8.5 micron.
This 2-8 micron range isn't a coincidence. Lens resolution ends up covering the same 2-8 micron range, and so do 4000 dpi scanners. So I guess you could say that sensor/film designers and lens designers have co-optimized or co-evolved up to the point where other issues (diffraction or DOF) are the limitations. F/8 and 6 micron is the "sweet-spot" or optimum middle where you have maximum lens resolution, good DOF, and maximal sensor/film resolution.
What does 6 micron mean in terms of Mpixels?
G10 => 14.7 mPixel at 1.7 micron
D300 => 12.3 mPixel at 5.5 micron
M9 => 18 mPixel at 6.8 micron
35mm film => 21.4 mPixel at 6.3 micron (4000 dpi)
Pentax 645D => 40 mPixel at 6 micron
Oops; on the G10 you have to be using f/4 or maybe f/2 to avoid blurring the details due to diffraction. In other words, you probably aren't really getting 14.7 mPixels on the G10.
A lens looking at something (tree branches or newsprint, or whatever) projects that image onto the film plane. No matter how good the lens, there is a limit to the details the lens can resolve which depends on diffraction due to the f-stop. With a smaller f-stop you start to lose details due to diffraction; larger f-stop, less diffraction at the cost of greater depth of field. If you are shooting a flat newspaper, open up to f/1.4 no problem. If you are shooting a landscape or street scene with near and far objects, you have to play the trade off.
In terms of resolution, film has grain in the 2-8 micron range. Digital sensors also have pixels of 2-8 microns: The D90/D300 is 5.5 microns, D3/D700 is 8.5, M9 is 6.8, G10 is 1.7. (Let's not get into the RGBG Bayer pattern or anti-aliasing filters, except to say they contribute more blur) Looking at scanners, 4000 dpi is 6.3 micron and 3000 dpi is 8.5 micron.
This 2-8 micron range isn't a coincidence. Lens resolution ends up covering the same 2-8 micron range, and so do 4000 dpi scanners. So I guess you could say that sensor/film designers and lens designers have co-optimized or co-evolved up to the point where other issues (diffraction or DOF) are the limitations. F/8 and 6 micron is the "sweet-spot" or optimum middle where you have maximum lens resolution, good DOF, and maximal sensor/film resolution.
What does 6 micron mean in terms of Mpixels?
G10 => 14.7 mPixel at 1.7 micron
D300 => 12.3 mPixel at 5.5 micron
M9 => 18 mPixel at 6.8 micron
35mm film => 21.4 mPixel at 6.3 micron (4000 dpi)
Pentax 645D => 40 mPixel at 6 micron
Oops; on the G10 you have to be using f/4 or maybe f/2 to avoid blurring the details due to diffraction. In other words, you probably aren't really getting 14.7 mPixels on the G10.
Last edited:
semilog
curmudgeonly optimist
Oh, it's almost certainly worse than that, since a camera priced like the G10 is not going to be optimally corrected wide open.
Roger Hicks
Veteran
Try it and see. With a test target there will be a slight falloff in resolution at f/11, a noticeable falloff at f/16 and significant falloff at f/22.
Cheers,
R.
Cheers,
R.
peterm1
Veteran
In the case of camera lenses this type of diffraction occurs when light passing thru the lens hits the sharp metal edge of the iris / aperture causing it to bend (diffract) so that it hits the film / sensor in a different spot than it otherwise would have done. Simple as that (and much more complex too of course.) The same kind of thing happens to all waves, including ocean waves - when they hit the edge of a seawall or headland they bend (diffract ) in just the same way.
When the iris / aperture is set to a large setting (ie when the hole letting the light in has a large diameter such as f2.8) this does not matter much as the amount of diffracted light is not all that much compared with the total amount of light passing through the iris. There is some diffracted light present but you do not really notice its effects. But when the aperture is stopped down then a larger proportion of light forming the image has hit the edge of the iris and lands where it "shouldn't" making the image fuzzy. The more you stop it down, the worse this gets.
Its all about the proportion of undiffracted to diffracted light hitting the sensor.
You also have to remember that in small format cameras the aperture size at any given setting is much smaller than in a larger format camera (say compare the lens a Canon G11 to a 35mm full frame lens. In the latter case everything is scaled up. So in the Canon f8 (say) may be the same physical size as f22 in a larger format camera. As another post says - its the mms diameter that counts NOT the actual f stop (which is a relative measure.) This is why many small format digital cameras have a minimum aperture of only f8 - above that diffraction kicks in too much damaging image quality.
And its why, when Ansell Adams used large format plate cameras, he could get away with using aperture settings as small as f64 - in such a large camera this setting is physically not all that small in absolute terms by comparison with say a 35mm format lens.
When the iris / aperture is set to a large setting (ie when the hole letting the light in has a large diameter such as f2.8) this does not matter much as the amount of diffracted light is not all that much compared with the total amount of light passing through the iris. There is some diffracted light present but you do not really notice its effects. But when the aperture is stopped down then a larger proportion of light forming the image has hit the edge of the iris and lands where it "shouldn't" making the image fuzzy. The more you stop it down, the worse this gets.
Its all about the proportion of undiffracted to diffracted light hitting the sensor.
You also have to remember that in small format cameras the aperture size at any given setting is much smaller than in a larger format camera (say compare the lens a Canon G11 to a 35mm full frame lens. In the latter case everything is scaled up. So in the Canon f8 (say) may be the same physical size as f22 in a larger format camera. As another post says - its the mms diameter that counts NOT the actual f stop (which is a relative measure.) This is why many small format digital cameras have a minimum aperture of only f8 - above that diffraction kicks in too much damaging image quality.
And its why, when Ansell Adams used large format plate cameras, he could get away with using aperture settings as small as f64 - in such a large camera this setting is physically not all that small in absolute terms by comparison with say a 35mm format lens.
Last edited:
Gumby
Veteran
Diffraction from using small aperture only matters to those who are oriented toward getting really, really, really, really, really, really, really, really, really sharp pictures... sharper than most people would even care to look at. Diffraction is discussed a lot because it is interesting, often misunderstood, and also because it's the kind of stuff that worry-warts like to worry about. Diffraction is a friend to bokeh... so it is a good thing. As for your other questions: JPEG, great taste, and film.
The "rule of thumb" for diffraction creeping in tended to be a 3mm aperture. That is absolute physical size of the aperture, and is a bit more complex than just focal length divided by f-stop. "BUT" that tends to be a guide for many lenses. So, a 50mm lens at F16 will have about a 3mm physical aperture, and diffraction creeps in. The Nikkor 105/2.5 has about a 3mm opening at F32, and that setting came and went in various generations of the lens.
So it is real, occurs at different F-Stops depending on the focal length and design of the lens, and the 3mm physical aperture size seems to be the consensus for many lens designers.
So it is real, occurs at different F-Stops depending on the focal length and design of the lens, and the 3mm physical aperture size seems to be the consensus for many lens designers.
Pico
-
Does diffraction have different consequences upon digital sensors than on film?
I'm thinking of pixel size and density, the Bayer pattern, and the fact that frequencies (colors) of the light are effected unequally by diffraction.
I'm thinking of pixel size and density, the Bayer pattern, and the fact that frequencies (colors) of the light are effected unequally by diffraction.
tstermitz
Well-known
Digital sensor affected similarly to film
Digital sensor affected similarly to film
Not really. I mean, resolution is limited on film by grain size and on sensor by pixel size. Yes, there are additional issues on sensors due to bayer pattern and anti-aliasing filter, but those can be addressed on their own separately from diffraction. 6 micron cells, 4-dot bayer, anti-alias filter, software reconstruction, diffraction vs DOF, they are all pieces of the whole system.
The demand for ever-more pixels on a sensor is fine, but has to be considered in light of the resolution of the entire system.
Sensors are basically as good as film in terms of resolution, and film is basically as good as the entire system can resolve, including scanning at 4,000 dpi. This means that moving beyond 35mm to medium or large format is the way to get high-rez pictures, and that is true whether we're on film or digital. Well, you can also stitch two images.
The Leica M9 & S2 and Pentax 645D do gain better resolution by removing the anti alias filter.
Digital sensor affected similarly to film
Not really. I mean, resolution is limited on film by grain size and on sensor by pixel size. Yes, there are additional issues on sensors due to bayer pattern and anti-aliasing filter, but those can be addressed on their own separately from diffraction. 6 micron cells, 4-dot bayer, anti-alias filter, software reconstruction, diffraction vs DOF, they are all pieces of the whole system.
The demand for ever-more pixels on a sensor is fine, but has to be considered in light of the resolution of the entire system.
Sensors are basically as good as film in terms of resolution, and film is basically as good as the entire system can resolve, including scanning at 4,000 dpi. This means that moving beyond 35mm to medium or large format is the way to get high-rez pictures, and that is true whether we're on film or digital. Well, you can also stitch two images.
The Leica M9 & S2 and Pentax 645D do gain better resolution by removing the anti alias filter.
ampguy
Veteran
practical explanation
practical explanation
If you're shooting film, or FF digital, and color, don't stop down beyond about f4 or f5.6.
If b/w f5.6 or f8.
If APS-C or APS-H then f2.8, and f4.
if u4/3 or p&s, you're pretty much SOL, just use wide open, and b/w.
Basically photons spread out on the medium if hitting the aperture on it's way to the medium, instead of going straight like it does in the center of the lens. So as you stop down the ratio of straight light to difrracted light gets smaller.
Best easy read I've found is here:
http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm
practical explanation
If you're shooting film, or FF digital, and color, don't stop down beyond about f4 or f5.6.
If b/w f5.6 or f8.
If APS-C or APS-H then f2.8, and f4.
if u4/3 or p&s, you're pretty much SOL, just use wide open, and b/w.
Basically photons spread out on the medium if hitting the aperture on it's way to the medium, instead of going straight like it does in the center of the lens. So as you stop down the ratio of straight light to difrracted light gets smaller.
Best easy read I've found is here:
http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm
Last edited:
Juan Valdenebro
Truth is beauty
I'll tell all those fools designing lenses for the last century, manually and through computers, how blind they have been... It will be easier for them from now on, just designing lenses with half the f-stops... 
Personally I use f/8 and f/11 without any visible problem. I create images, and they require less sharpness than I get with f/8 and f/11... There's a limit for the relevance of technical testing: when photography gets benefits over tests' results, test's results are secondary... (To photographers...)
Cheers,
Juan
Personally I use f/8 and f/11 without any visible problem. I create images, and they require less sharpness than I get with f/8 and f/11... There's a limit for the relevance of technical testing: when photography gets benefits over tests' results, test's results are secondary... (To photographers...)
Cheers,
Juan
ferider
Veteran
As a tech guy: diffraction counter-interacts with other effects, for example focus shift. I have used lenses that peak at 4 or f5.6 and lenses that peak at f8 or f11 (like the 28/3.5 Color Skopar or 35/2 Summicron v3).
As a user: I try to avoid going beyond f11 on 35mm format. Sometimes I need f22 and am glad when a lens can do it. On teles even more.
Meaning, for the OP: it's real, but not a limitation to my photography at all.
Roland.
As a user: I try to avoid going beyond f11 on 35mm format. Sometimes I need f22 and am glad when a lens can do it. On teles even more.
Meaning, for the OP: it's real, but not a limitation to my photography at all.
Roland.
ampguy
Veteran
clarification
clarification
Roland, you might want to clarify that you shoot FF (and b/w mostly?)
Reds become diffraction limited soonest. Most everyone who has posted a u4/3 image with reds, I can see the diffraction.
Juan, if lens makers knew their lenses would be used on u4/3 medium, you're correct, they would have stopped at f2.8 or f4 or color coded to not go past there unless a bigger sensor.
clarification
Roland, you might want to clarify that you shoot FF (and b/w mostly?)
Reds become diffraction limited soonest. Most everyone who has posted a u4/3 image with reds, I can see the diffraction.
Juan, if lens makers knew their lenses would be used on u4/3 medium, you're correct, they would have stopped at f2.8 or f4 or color coded to not go past there unless a bigger sensor.
Juan Valdenebro
Truth is beauty
I want to clarify that for 35mm color film (or B&W), f/4 and f/5.6 are no limit -at all- for getting high sharpness. Anyone can test it.
Cheers,
Juan
Cheers,
Juan
lacavol
Established
I have an old Kodak book and they say there is a loss of sharpness at f16 . But they also say and I quote: "These small changes in performance are of little consequence. These effects on definition are much smaller than those due to slight errors in focusing and to slight camera motion." Spherical aberration is usually helped by decreasing the aperture as angles to the film plane are reduced. Lateral color aberration shows up with different color wavelengths on color film and blur on B&W. A lens is limited by both the aperture and residual lens aberrations. There are other aberrations just too many to list.
When you are making computer chips ( a process much like lithography) then the diffraction problems become greater. To counteract this the masks are usually larger than the final image on the silicon wafer. They also use a shorter wavelength light, usually UV and sometimes gasses with a different refraction index.
When you are making computer chips ( a process much like lithography) then the diffraction problems become greater. To counteract this the masks are usually larger than the final image on the silicon wafer. They also use a shorter wavelength light, usually UV and sometimes gasses with a different refraction index.