How much does diffraction effect actually hurt sharpness?

Most 50mm lenses exhibit their maximum sharpness around f:5.6-8.0. Most of my lenses seem to limit their aperture to f:16, but a late f:3.5 Elmar goes down to f:22.

I have just started shooting B&W in a Leica Standard, which has no rangefinder, necessitating using the DOF scale for "street" shooting. I wonder how much sharpness I'll lose by stopping down to f:22 (8x12 prints)? Not much, or a whole lot?

Diffraction (the circle of confusion) is proportional to the wavelength and the numerical aperture (2.8, 4, 5.6, etc). this takes into account the focal length (the "f" in f/16). The greater the magnification, the larger the circle of confusion on the final print. 35mm enlarged to 8x12" is an 8x magnification, which I assume in the following analysis.

 

Most modern lenses are highly corrected for CA and aberrations, but in general resolution peaks about 2 f/stops from the maximum opening and decreases on either side of the peak. You don't notice diffraction until it is equal to or greater than the uncertainty of the medium. For a fine-grained color film like Velvia in a full-frame (35mm) camera, that might be f/11 or more. The higher the resolution, the sooner the onset of diffraction. Diffraction is limiting in a Sony A7Rii (42 MP, no AA filter) at f/8 and beyond. At f/16, the resolution is about half the peak value, and very noticeable with that camera.

 

A common B&W film like Tri-X has about half the resolution of Velvia (80 v 160 lp/mm). Diffraction is less than the uncertainty of the film as small as f/16 or even f/22. If you hand-hold the camera, camera shake is the largest source of uncertainty until the shutter speed is 1/3F or faster. At a shutter speed of 1/F, the uncertainty is about equal to focus error at the limits of DOF markings.

 

In short, shooting by hand, without a rangefinder, diffraction is of no concern to you.

Edited June 17, 2017 by Ed_Ingold

I wonder how much sharpness I'll lose by stopping down to f:22 (8x12 prints)? Not much, or a whole lot?

 

Hi, I've never done the actual experiment in the real world, but I'd guess that going from f/16 to f/22 the difference will be pretty slight.

 

I have shot a few lens charts, etc., and can clearly see, under a microscope, the detail falling off, but keep in mind that this is with a carefully focused target, a slow film, and a stable camera stand.

 

The theoretical limit on maximum resolving power of a lens (in line pairs per mm) due to diffraction is something like 1600 or 1800 divided by the f-number, so f/22 gives you about 70 or 80 lp/mm max. f/16 theoretically can get to 100 or 110 lp/mm, which sounds like a lot, but when you consider the interaction with typical film, the net increase is a lot less.

 

If you were to use a high resolution document film, focused very accurately, and made sure that hand-held motion blur was minimal, I'd guess that critical viewers would see a difference between f/16 and f/22, with a much more obvious difference between f/11 and f/22. Now if you were talking about something like f/45, the differences would be relatively huge, and I'd say pretty obvious on a print. But f/16 to f/22, I think it's just on the verge of where "system" resolution starts falling off obviously; at this point the finer details of shooting technique probably predominate.

 

Just my two bits, unconfirmed in the real world.

I shoot medium format film RB67. Lens can go down to f/45. I shoot Velvia 50 and Tmax 100.

 

What's the interplay because of the larger film format?

 

Also, since I shoot landscape, I often want max DOF so I stop down to accomplish that. So the question becomes which is more important? DOF or refraction issues? Where's the tradeoff?

What's the interplay because of the larger film format?

 

Larger film is usually enlarged less. Looking at a same-size final print your RB67 only needs about half the (linear) enlargement as a 35mm neg. So it seems permissible for the diffraction blur to be twice as large, meaning that your f/45 is roughly as limiting as a 35mm film camera's f/22. Again, this is based on the same size print.

Everything is better with medium format or larger, except for depth of field. DOF depends on the focal length and aperture, and is about half what you get in 35mm with the same field of view (e.g., equivalent focal length).and f/stop. If you view film with a good 10x magnifier (e.g., Hastings triplet), you see very little difference between MF and 35 mm. Film is film, but MF requires only half the magnification for the same size print.

 

Again, with a 10x magnifier, you see there are no sharp edges in film. Every edge has a fuzzy border which spans 3-5 pixels in a 4000 dpi scan. The uncertainty is nearly Gaussian, so edges and points have a dark core which looks sharp under normal viewing conditions. It is greater than the radius of the Airey disk produced by diffraction until the aperture is f/11 or greater.

 

Digital is different. While the linear resolution is easily calculated from the number of pixels, it is possible for edge detail to change from light to dark in a single pixel. In other words, the acutance is very high, even for subjects with normal contrast (e.g., 6:1). Diffraction is the same, but you see the effects much sooner, at least if the lens quality permits. You find that it is possible to resolve extremely fine detail, one pixel wide, but only with extreme care in focusing, support, and with a remote release. Digital resolution is best measured by edge contrast, expressed in pixels/image height. A good lens will peak at about 4900, two stops from maximum, out of a perfect 5304.

 

Putting wonkiness aside, this has practical consequences. I've experimented with longer shutter speeds for waterfalls, which requires neutral density filters and stopping down. I usually shoot at f/8 or less, but at f/16 my images were decidedly soft throughout, not just the moving water but rocks as well. I used the same tripod, same lens and same technique. Sharpness isn't everything, and the softness lends kind of a "painterly" effect to the scene. Waterfalls don't really look like fog either, so reality is not a firm requirement for landscapes. Images i posted in the thread (Mirrorless Monday, June 12, 2017) taken at 1/4 second and 1/60 second illustrate this issue.

 

Rather than use tiny apertures, I've been experimenting with focus stacking. It is mainly used for closeups, but has applications for landscapes as well. With as few as two exposures, you can achieve a sharp foreground against a sharp background, similar to using a tilting lens board on a field camera. Software like "Helicon Focus" does a good job detecting edge detail and masking, with a minimum of manual intervention.

Luckily, I just found the April 1974 Modern Photography where they actually tested the last Elmar. From a high center resolution of 66 lpm at f:3.5 down to 59 from f:5.6-16, then still 47 at f:22 -- not bad! The edge numbers actually were gradually improving from 26 wide open to 42 at f:16and 22.

No problem shooting Tri-X or XP2 at ISO=400, but I'm shooting EFKE KB25. In bright sunlight that would be 1/12 second at f:22, and the slowest speed on the camera is 1/20. My little experiment in "going back to basics" just hit another snag. The longer I try it, the more impressed I am with those Leica pioneers like Dr. Paul Wolff.

If you don't see image deterioration due to diffraction at f/22, it's because some other part of the chain is even worse, dominating the results.

You should indeed notice the difference. I notice it going from f11 to f16 for a full format lens. I might also be able to pick it out f11 from f8, but this is often more difficult as the increase in DOF often helps to give the appearance of greater overall sharpness.

I'll be printing 8x12, not pixel-peeping. Looks to me not a problem, but I'll try to report.

Sharpness at small apertures varies from one lens to another. The difference in 8x10 prints was noticeable when using an Olympus XA rangefinder camera at f/5.6 versus f/16. But there's hardly any difference at all with my Leica lenses. Same film, same developer, etc.

 

Lately I've been using a Nikon 40mm macro lens to copy slides with a DSLR, and it shows hardly any degradation at all from f/5.6 to f/16. This performance makes sense for a macro lens, which is optimized for good performance at small apertures to provide greater depth of field at close range. I prefer using f/16 to ensure corner sharpness, because slides are usually not perfectly flat in cardboard or plastic mounts. Some old slides are badly warped, so f/16 is required to bring the whole surface into focus. Although f/5.6 is theoretically sharper, much of the slide would be out of focus.

Diffraction - Wikipedia

 

Specifically, d = 1.22*W*N, where W = wavelength (mm), N = numerical aperture and d is the circle of confusion in mm. For the example below, I used the wavelength of green light, 550 millimicrons (mu)..

 

The circle of confusion (Airey Disk) is approximately equal to the resolution of Tri-X at f/16, or Kodachrome at f/22. Because these effects are independent and additive, the onset of diffraction limiting would be about f/11, and significant at f/16. The presence of CA and various aberrations would raise that threshold, which is why you notice it with Leica lenses and not Olympus.

 

Diffraction is a fundamental property of light. If you don't see it in your photos, it is because something else is worse.

 

If you have been following this discussion, you might conclude diffraction is worse with digital. No! Digital has such high acutance that you see the effect of diffraction at a lower aperture, at least for top quality lenses.

 

When copying slides, I use f/5.6 so that most of the image is grain-sharp. It usually doesn't matter of the corners are out of focus, because the subject in most real life photos is in the center, leaving the corners out of focus anyway.

Edited June 20, 2017 by Ed_Ingold

I'll be printing 8x12, not pixel-peeping. Looks to me not a problem, but I'll try to report.

 

I agree, you shouldn't notice really diffraction at that enlargement.

Actually, you can see it. An 8x12" (8x10") print viewed at 10" is the standard for judging sharpness, whether DOF, camera shake or diffraction. The limits are generally defined as a 0.2 mm (0.008") circle of confusion on the print. That's over 2 dots at 300 dpi. You won't see it at 150 dpi, but nothing looks sharp printed at that resolution.

 

If "not pixel peeping" means not looking carefully, you can forgive a whole multitude of things.

 

Depth of field - Wikipedia

Ed, I appreciate your advice, but many of the old slides I'm duping are warped in odd ways, not just at the corners. There might be two people near the center of the frame, but the slide is warped such that both faces won't be in focus at a middle aperture like f/5.6. In other cases, the pictures do have important subject matter near the edges or corners. Rather than go back and reshoot the slides that have focus problems at f/5.6, it's easier to shoot everything at f/16. They're still grain-sharp. I've tested my rig at every aperture from f/5.6 to f/22, and only f/22 shows significant diffraction.

 

My Nikon CoolScan V has less depth of field. It can't always make a sharp scan across the frame of a warped slide. Luckily, unlike some other film scanners, it does have a manual-focus option, so I can put the focus point where I want it. But sometimes the DSLR is better because I can get the whole slide in focus at f/16.

 

Keep in mind I am digitizing some very difficult images. Recently I've been working on Kodachromes from the early and mid-1950s that were poorly stored. Some are badly infected with fungus, dirty, and scratched. Also, using my flatbed scanner, I have digitized many 4x5 negatives so shriveled from "vinegar syndrome" that they were too small to fit the 4x5 film holder. I had to scan them directly on the glass. I have also scanned tintypes from the 1870s so badly faded that they were almost black. Only major effort in Photoshop restored a recognizable image.

 

Coming next: several rolls of b&w negatives that my late father shot in Japan while stationed there with the army in 1946, right after the war. They are uncut rolls, tightly wrapped in paper for 71 years, poorly stored, in deteriorating condition. Different formats, too, including what looks like half-frame 35mm. Another challenge.

 

When duping pristine slides that were shot on modern films, small differences in diffraction at different apertures might matter. For my work, it's not something I worry much about.

Tom, you're lucky that you bought the Nikon Coolscan. I bought the Minolta scanner, which actually gave better results than the Nikon, but then it went KaBoom three times, and then Minolta went KaBoom themselves, and it could no longer be repaired. $600 bucks town the toilet!

All slides in cardboard mounts are cupped to some degree. If you need f/16 or more for damaged slides, that's fine. I'm just saying there is a price to be paid, imposed by mother nature. I have a bankers box full of old family negatives, some nearly 100 years old. They are a treasure trove for their content, if not their photographic quality. I agree that diffraction does little to detract from their resolution, or lack thereof.

 

Tintypes are not negatives in the usual sense. They are mounted against a black background and viewed by reflectance of the silver image. You can do that to some extent with a modern negative, particularly large format, except the emulsion is rough compared to thin, tintype emulsion on polished glass.

 

The Minolta scanner had rather primitive negative handling, and boasted 5000 dpi resolution which was obviously inflated by comparison. It also use a cold-cathode fluorescent lamp rather than LEDs of the Nikon, which seem to last forever.

 

When scanning odd sized film on a flatbed, it helps to lay strips of black matting board to mask them. You could even tape them to the glass with black photographic masking tape, but cleaning up afterward is a PITA. Light leakage around the edges can cause flare, especially as the scanner ages and gathers dust and a film from evaporating plasticizers.

Before buying my Nikon CoolScan, I borrowed a friend's Minolta film scanner for a while. In my opinion (agreeing with others), it delivers superior results with b&w film than the CoolScan. Dust, scratches, and grain are less evident, with no loss of image sharpness. The difference is their light sources. The Minolta scanner has a diffuse light source, whereas the CoolScan has a rather harsh LED point source. It's like the difference between a diffusion enlarger and a condenser enlarger in a darkroom. I've used both and much prefer a diffusion enlarger.

 

Interestingly, my results with slide duping using a DSLR and the Nikon ES-1 slide copier are more like the Minolta's diffusion scans than a Nikon LED scan. The ES-1 has frosted glass behind the slide that diffuses the light source, no matter where it's pointed.

 

The Minolta scanners' weak point is poorer construction. They don't seem to last very long. My Nikon CoolScan just keeps going.

 

Ed, I have also scanned some old box-camera negatives up to 100 years old. Their format wouldn't fit my flatbed film holders, so I made a custom holder out of wood. But the negatives mounted on a black background are ambrotypes, not tintypes. Tintypes are direct positives. They capture amazing detail when scanned and enlarged.

 

Aging flatbed scanners do build up an internal film from evaporating plasticizers, but my Epson V700 is easy to take apart and clean. I do it about once a year.

Tintypes are updated versions of ambrotypes, only on black metal rather than glass with a black background. The image itself is an underexposed negative, visible by reflection from the metallic silver against the dark substrate. Tintypes are best viewed with off-axis light, whereas the light in flatbed scanners is nearly coincident with the sensor. You would probably get better results if you copied them with a camera after carefully adjusting the light to avoid reflection from the glass or substrate. You might need a fairly large, diffuse source with barn doors or grid to control the directionality. As I recall, it's hard to see the whole tintype at once in room light. You need to turn it to see various parts of the image. The book listed below does a good job explaining ways to illuminate difficult subjects.

 

Light Science & Magic: An Introduction to Photographic Lighting: Fil Hunter, Steven Biver, Paul Fuqua: 9780415719407: Amazon.com: Books

Most 50mm lenses exhibit their maximum sharpness around f:5.6-8.0. Most of my lenses seem to limit their aperture to f:16, but a late f:3.5 Elmar goes down to f:22.

 

I have just started shooting B&W in a Leica Standard, which has no rangefinder, necessitating using the DOF scale for "street" shooting. I wonder how much sharpness I'll lose by stopping down to f:22 (8x12 prints)? Not much, or a whole lot?

 

I’d like to know more about what exactly it is that you are shooting (i.e. “Subject”), because I think that you might be addressing this from an unnecessary approach.

 

Maybe a couple of illustrations would assist to convey the point that I making:

 

A “doorstop” using a 50mm with the Subject at 15ft, gives a FoV of about 10ft x 7ft, which is adequate to frame the Key Subject either in Vertical or Horizontal orientation. Pulling the shot at F/11, and in a 10 x 8 print you’re "reasonably safe" from about 12~13ft to over 20ft. I wouldn’t usually be shooting at an SD closer than 15ft with a 50mm lens loaded – and obviously, the farther away the Subject, the relatively ‘more safe’ that you are at F/11.

 

F/11 provides TWO extra stops of Shutter Speed than F/22 and I would much rather have that shutter speed in my pocket to use; because the ability of using a faster shutter speed allows much a better safety net for chasing the Subject or shooting from the hip or shooting ‘Hail Mary’, in these cases Camera or Subject Motion Blur is more likely and access to a faster Shutter Speed is like gold, especially if shooting film, where changing ISO/ASA (mid roll) necessitates the use of other skills and prceedures.

 

When using a 50mm lens and Zone Focusing, I am concentrate more on getting into a shooting position to accommodate my predetermined Subject Distance, rather than stopping down to accommodate a ‘safer’ depth of field. If I have a 35mm lens loaded typically I would be at F/8,

 

But a 50mm lens and for “street work”, I can’t think of many situations where: F/11; a good judgement of Subject Distance; and the speed to get into a good camera position would not be a much more elegant solution, than even considering using F/22.

 

WW

After nearly 70 years of using the "best and newest" photography equipment that money can buy, I decided to challenge myself with using the most basic materials. Besides, I've always thought that the jewel-like chrome "Standard"was the most beautiful of all Leicas. I happened to have a freezer full of EFKE KB25 film which is about as close to the 35mm B&W film of the 1920s as I'm ever going to get. No coupled Rangefinder, no light meter. It sounds like a lot of fun, and what cold possibly go wrong?

The answer is: it nearly drove me crazy!!!! My admiration for Cartier-Bresson just went up exponentially.

I no longer do my own film darkroom work, but the preliminary scans should be back tomorrow, but I won't get the developed film until next week.

I'll try to share the results with y'all.

Heh. My second camera was a 35mm Kodak Pony II (1957-1962), which also didn't have a rangefinder or light meter. And it had a fixed shutter speed of about 1/50 second that was difficult for a beginner to hold steady, especially with the shutter release located right next to the lens. Occasionally I took a semi-sharp picture -- the lens wasn't too bad -- but most were fuzzy from camera movement or subject movement. And most of my pictures were heavily overexposed, because the lens was marked in EV stops instead of f/stops and I didn't know the difference.

The initial scans came today by email, and look surprisingly good.

My initial question about f:22 indicates that, as expected, the images are plenty sharp, with beautiful OOF portions.

As warned by one of the posters, I did have some camera shake in a couple of the pictures.

All-in-all a quite satisfactory result (technically, if NOT esthetically).

I'll give more results next week when the negatives and CD come in the snail mail.

For a pinhole camera, there is an optimal pinhole size, where the size of the hole equals the amount of diffraction, based on the hole to film distance. (You have to vary this size for those zoom pinhole cameras.)

 

Unless you are photographing a completely flat subject, you need some depth of field.

There will then be an optimal aperture, where the circle of confusion due to depth of field equals that due to diffraction.

 

I suspect that the reason most lenses go down to about f/22, is that there are mechanical difficulties getting the assembly to work well over a wide range. Larger aperture lenses, also tend to have a larger minimum aperture.

 

But often, f/22 is at the point were diffraction isn't worse than camera (and subject) motion, and reasonable depth of field.

 

Ignoring lens aberration (which you shouldn't do), you could compute a depth of field scale based on diffraction.

There are several processes affecting resolution, the most important are listed below:

 

• Depth of Field - the limits of visible sharpness, generally taken as 0.2 mm uncertainty of edges in an 8x10" print, viewed at a distance of 10"
  
• Camera Shake - inversely proportional to the shutter speed. roughly equal to the DOF limit for a speed of 1/FL
  
• Diffraction - directly proportional to the numerical f/stop (8, 11, 16, etc) and wavelength
  
• Lens Aberrations - variable, generally greater at wider apertures
  
• Media resolution - Digital acutance is probably better expressed as edge definition rather than a traditional frequency chart, expressed as pixels/image height (Imatest). This method can be used for film scans too.
  
• Format size - Uncertainty is proportional the enlargement factor. Larger film is enlarged less for an 8x10 at 10".
  

These can be expressed graphically as uncertainty (circle of confusion) v f/stop or some other factors, but it's hard to visualize or deal with various scales. Uncertainty from various sources is additive as the root sum of squares.

 

For example, plotting uncertainty against f/stop, camera shake is a horizontal line at a given shutter speed. Diffraction is an inclined line. Using a shutter speed of 1/FL, it intersects the diagonal line of diffraction at about f/11-16. This means that diffraction is insignificant for hand-held shots below this aperture.

 

Media uncertainty (grain, pixel pitch, etc) is an horizontal line which intersects the diffraction line at some point. This is about f/11 for film resolving 100 lp/in. If you use Tri-X, for example, diffraction is insignificant below about f/11. Other things are more important.

 

Digital generally has much higher acutance than film, so the effect of diffraction occurs much sooner. We see the resolution peak at about f/5.6-8 and drop off above this value. At wider apertures, resolution drops due to lens aberration. Applying the math in reverse, if you hand-hold a camera at a shutter speed of 1/FL, the net resolution of the sensor is equivalent to about 6 MP,

 

The last paragraph should help explain why so many people respond that they can't see the effects of diffraction, high resolution sensors or better glass.