A Modest Contribution to Dispelling Diffractiophobia

[Ulophot]

I posted this over at largeformatphotography.info and thought I'd add it here as well.

The matter of diffraction and its effects on image sharpness comes up here periodically. I thought I’d share a little something on the matter.

Some of us work in in large format partly for the purpose of making very large prints, by which I mean 16x20, 20x24, and considerably larger. For those who do, it may be that every scintilla of sharpness is desired to accomplish the intention. Others of us lack the means to make such large prints, probably wouldn’t make them even if we could, and/or are not as concerned with ultimate sharpness as some others may be, for reasons of subject matter, style, or other considerations. I live in this camp, as primarily a portraitist with a maximum print size of 11x14. Nonetheless, I do photograph landscapes and other types of subjects occasionally, and the diffraction question has, I admit, quietly nagged me despite the sage words of more accomplished members assuring that loss of sharpness from diffraction rarely trumps inadequate depth of field in practical work.

Today I finally made a test. Using a still life concoction of suitable items arranged well within the depth of field at f/16, I exposed one negative at f/22 and one at f/45 on my standard film, HP5+. My developer is D-23. My lens is a 1960s or ‘70s 210mm Komura, a lens with a good reputation but by no means a wallet-stripping super-multi-coated Apo-Sharpenar from one of the top German or Japanese makers. I set up my enlarger with my 80mm Nikkor lens from the 1970s and made 5x7 prints from a section of the two negatives at about 5x, i.e., a 20x25-inch enlargement.

Guess what. Examining the prints close-up by eye, I can barely tell which is which. Only in the type on the little card can I sense that the f/45 images is not quite a sharp. The small, serif type, 8- or 9-point, is in the range of 1/64-inch high on the negative—all perfectly clear in the prints. But what really wiped away any doubt, is that some fine, individual hairs frayed from a microfiber cloth are also clearly visible. Remember that I’m staring from about six inches at a teency-tiny area on what would be a 20x24 print. I hope the scans will at least indicate what is visible in the prints. (By the way, I applied no sharpening to the scans and only adjusted the black point slightly since they are on semi-matte paper.)

So, for those of you coming into LF work, or film work at all, I hope I’ve helped to allay some of your concern about this little corner of the technical aspects of making art. Now you can spend more time worrying about composition, a much more challenging subject, in my view.

The detail images were scanned at 1200 ppi from an area about 1 1/4 inches on the 5x7 prints.

 

[ic-racer]

Those are optimum apertures for that lens. Try f64, f128 and f256.

. Try f64, f128 and f256.

 

[ic-racer]

There is an aesthetic to diffraction. Check out The Somnambulist (35mm mostly at f16) and Cape Light (8x10 mostly at f90).

 

[brbo]

BTW, you obviously messed up the labelling of the images, either of the crops of the full size. And one crop has higher magnification than the other.

Otherwise, more like dispelling the physics behind diffraction it looks like you just found out (like many others) diffraction is something you don't have to worry too much about with your intended final print size?

 

[gone]

I heard that Diffractiophobia II in the works, but the book is a lot better.

 

[MattKing]

I've always understood that diffraction was actually related to the physical size of the aperture (and how it relates to the wavelengths of visible light) rather than to the f/stop.
Of course, the visibility of the effects of diffraction are also related to the amount of magnification from negative to print.

 

[DREW WILEY]

You have to attain real film flatness to make any objective comparison, using either an adhesive or vacuum film holder. I think you'd discover that, while f/45 makes very little visual difference in a print enlarged from 8X10 film (having the same effect as f/22 in 4X5), using f/45 for 4X5 format itself will exhibit some distinct diffraction penalty. Some of this is disguised because you are actually helping that other potential problem I just mentioned - film flatness - by giving a little more depth of focus at a smaller stop. A good general rule of thumb is to never shoot 4X5 at a smaller stop than f/32, or 8X10 smaller than f/64 if you're worried about diffraction.

The bigger the film, the more it can potentially sag in a holder. Not much with 4x5; 5X7 sometimes an issue; 8X10 you need to be aware if enlarging more than 3X; and ULF, a big issue. Film base differences also apply. PET polyester or Estar sheet film base is much stiffer than acetate.

Since I have sometimes make big 30X40 inch highly-detailed prints on media capable of extreme detail reproduction, namely Cibachrome before, and Fuji Supergloss now, for that kind of work I use adhesive holders and try to keep even 8X10 color exposures within f/32 to f/45 if possible. That rule of thumb varies a little of course with specific focal lengths.

As far as the work of others goes, and not necessarily one's own priorities, it doesn't impress me a bit when someone refers to some famous photographer who contact printed their work. Contact printers can get away with all kinds of things that might look annoying in even a modest enlargement. And as far as Cape Light and Meyerowitz and his f/90 strategy went - he was deliberately after a soft-edged look, and it distinctly shows even in his early contact prints of those images, even more so in enlargements; book-sized, less, but still there. And frankly, numerous of the well-known color photographers of that era were less than ideal view camera practitioners in terms of focus precision and depth of field management. Some were simply awful, but thought it artsy anyway, especially in their very low contrast nearly muddy Vericolor imagery. Depends on what you're after, esthetically.

 

[reddesert]

I've always understood that diffraction was actually related to the physical size of the aperture (and how it relates to the wavelengths of visible light) rather than to the f/stop.
Of course, the visibility of the effects of diffraction are also related to the amount of magnification from negative to print.

The angle spread of diffracted rays is directly inversely correlated to the physical aperture diameter, but in order to to turn that angle into a spot size on the film, we multiply by the focal length. So the spot size due to diffraction comes out directly proportional to the f-number.

Because larger spot sizes can typically be tolerated in larger formats, the f-number that you can use without worrying too much about diffraction goes up with format size.

To give mathematical formulae, for light of wavelength lambda (say 0.55 microns) , aperture diameter D, and focal length f:

f-number N = f/D.

angle of diffraction: theta ~= 2 * lambda / D.

spot diameter on film: c = theta * f = 2 * lambda * f/ D = 2 * lambda * N.

This is approximate for pictorial use because the diffraction pattern is an Airy disk, not a perfect spot, and because we really care about pictorial effects not mathematical exactness. But for example using this formula we would find that the diffraction spot size is about 0.018 mm for f/16. This means that f/16 is tolerable for 35mm format, but by f/22 you might start to notice. It scales up for larger formats, and down for sub-35mm.

Of course the final spot size you care about depends on print size, but when talking about real images and prints we have to take into account imperfections in the whole system, like focusing inaccuracy, film plane not perfectly parallel, depth of field. Because stopping down compensates for some of these imperfections, stopping down a little more than the "optimal" f-number can still produce pleasing results.

 

[DREW WILEY]

Good principle, but overoptimistic. How often is anything 35mm tolerable at an aperture as tiny as f/16? Maybe 6X7 format ...just maybe. A lot of what constitutes an acceptably confusing "circle of confusion" relies on likewise plasticized notions of "standard viewing distance" (which happens to be about a third of a mile in the case of a billboard), and how much sloppiness once occurred during "typical" enlargement in high school darkroom classes. Glad I didn't learn enlarging in that manner, or spend much time in math either, when I was in high school; I prefer to adjudicate f/stop personal acceptability standards with my own set of eyes instead. But I still land on the quite routine advice of the best RANGE of f-stops which most experienced practitioners abide by.

But portrait photographers seeking shallower depth of field and perhaps softer imagery too, might choose uncharacteristically wider stops than landscape and architectural photographer might, while pinhole photographers and contact printers might bend the "rules" the opposite direction, to exceptionally small stops. All depends. In other words, make up your own rules; but just be aware of what that implies in terms of optical performance.

Also, when making use of older data, like the afore posted Paleolithic Linhof chart, understand that today films tend to have higher resolution and finer grain, and lenses themselves higher average resolution than back in the days they were made by grinding blocks of quartz after chipping those with a rock and an antler. I miss those times, personally; there's nothing quite as good as BBQ giant ground sloth over a campfire, or as memorable as film with grain the size of anti-aircraft shrapnel. But we do need to live in the present, so I'll settle for a can of instant soup on this rainy day, and having TMax 100 in my Fuji 6X9 RF, if it does stop raining before sunset.

 

[Ulophot]

BTW, you obviously messed up the labelling of the images, either of the crops of the full size. And one crop has higher magnification than the other.

Otherwise, more like dispelling the physics behind diffraction it looks like you just found out (like many others) diffraction is something you don't have to worry too much about with your intended final print size?

brbo, while I did not mix up the labelling, you are quite right to point out the difference in crop in the detail shots. My scanner is old enough to require third party software which does not allow precise crop measurements, and perhaps there is some inconsistency in the scan runs; most of my occasional scanning is not images but documents for filing. By the time I got to putting up the post, I was tired enough to let it go. As I noted, the detail images represent a bit more than a square inch of a print that would typically be viewed from two to three feet away.

However, please bear in mind my purpose in sharing the scans along with my qualifying language. I could have added notes on other aspects of achieving sharpness, some of which others have mentioned. Quite a few factors ultimately figure in that I did not address. I hoped, and do hope, that I may have illustrated a simple point overall that will help remove concern for some less experienced photographers or perhaps lead them to make their own tests and view their results as I did mine. That direct examination of the actual prints showed virtually no difference was more than sufficient demonstration for me.

 

[otto.f]

I can barely tell which is which.

? Don’t you see a difference in sharpness in the two Zone Calc photo’s? Or do you share that under contrast difference?

 

[brbo]

brbo, while I did not mix up the labelling, you are quite right to point out the difference in crop in the detail shots.

Texture of the fabric is blown in the full size f22 and crop of f45, it's not blow in full size f45 and crop f22. If that is not the case of incorrect labelling of the images it's most probably because of poor control of the scanning process.

 

[RalphLambrecht]

BTW, you obviously messed up the labelling of the images, either of the crops of the full size. And one crop has higher magnification than the other.

Otherwise, more like dispelling the physics behind diffraction it looks like you just found out (like many others) diffraction is something you don't have to worry too much about with your intended final print size?

to avoid diffraction, which is a real problem, I stay between f/8-11 for 35mm, between f/5.6-8 for MF, and below f/5.6 for LF.

 

[Chuck_P]

and below f/5.6 for LF.

You're saying I need to be concerned about diffraction with 4x5 with my 135mm Sironar-N at f/5.6, which is as low as that lens can get!?

 

[xkaes]

I've always understood that diffraction was actually related to the physical size of the aperture (and how it relates to the wavelengths of visible light) rather than to the f/stop.

The actual f-number doesn't matter, as you say. Still, with the size of the aperture being the same, diffraction will be low at low magnification (ex, 1:10), and high at high magnification (ex, 10:1).

For example, a lens at f16 and 4X magnification is really f80.

 

[DREW WILEY]

You might want to re-read what you just posted, Ralph. It doesn't make a bit of sense.

 

[Ulophot]

? Don’t you see a difference in sharpness in the two Zone Calc photo’s? Or do you share that under contrast difference?

Hi, Otto. Perhaps I should not have scanned the photos, since this seems to causing such difficulties. As noted above, my text referred to examining closely the 5x7 prints I made. I will know next time to be exacting if I print scans from prints. I thought they might be a useful supplement, but evidently I was too casual with the results.

 

[Steven Lee]

The actual f-number doesn't matter, as you say.

Actually, the diagram notation we have was invented for conversations just like this one. It simplifies reasoning about properties of optical systems ignoring the focal lengths and diameters of lenses. Formulas become shorter and more concise. The simplified formula for diffraction (airy disk diameter) is basically the light wavelength times f-stop [1], so having f-stops is superbly convenient here.

This also means that diffraction does not depend on film format. At the same f-stop you get the same airy disk on 35mm or 10x8. We just blow up smaller negatives more, that's all.

[1] https://en.wikipedia.org/wiki/Diffraction-limited_system

 

[reddesert]

Good principle, but overoptimistic. How often is anything 35mm tolerable at an aperture as tiny as f/16? Maybe 6X7 format ...just maybe. A lot of what constitutes an acceptably confusing "circle of confusion" relies on likewise plasticized notions of "standard viewing distance" (which happens to be about a third of a mile in the case of a billboard), and how much sloppiness once occurred during "typical" enlargement in high school darkroom classes. Glad I didn't learn enlarging in that manner, or spend much time in math either, when I was in high school; I prefer to adjudicate f/stop personal acceptability standards with my own set of eyes instead. But I still land on the quite routine advice of the best RANGE of f-stops which most experienced practitioners abide by.

Well, I said "tolerable" (for f/16 on 35mm). Acceptable is different from critically sharp. Normally I stick to f/11 and below on 35mm. But this isn't a hard limit. If one has a compelling reason to use a small aperture, like needing a large depth of field, then it's worth bending that rule. I think this was some of the OP's point.

To put numbers on this, in 35mm terms, the depth-of-field scales on most 35mm lenses assume a circle of confusion c=0.03mm. This corresponds to conventional acceptable sharpness for an 8x enlargement (8x10 print), about 4 lpmm. However, many people feel that if you want to be critically sharp, the DOF scale is too generous and you should compensate by 1 or 2 stops (which corresponds to a circle of ~ 0.02 or 0.015 mm).

Compare that to the theoretical diffraction circle of ~ 0.018mm for f/16. It's similar, but because the diffraction pattern is convolved with the image spot size, these circle diameters add in quadrature (sum of the squares); the blur from diffraction and from out-of-focus adds. So in the ideal world, one would use more conservative DOF markers (if you're at f/11, use the f/8 markers) and stay at f/11 or below on 35mm, say. Of course, these rules of thumb are in competition, so sometimes you have to compromise.

The acceptable f-number roughly scales up with format, for a given final print size, thus if your personal limit is f/11 for 35mm, maybe f/22 for medium format, and f/45 for 4x5. But many people want to print larger from larger formats. OTOH, larger formats also have less depth of field so you might need smaller stops anyway. The blur from out of focus foreground or background can be more distracting than from diffraction. These are rules of thumb, not absolute limits.

 

[Maris]

There is a complication in effects of diffraction on image detail that I first noticed when doing a bit of amateur astronomy using telescopes that are diffraction limited.
Diffraction causes image detail points to spread, Airy discs become larger as diffraction increases. The consequence of this is that fine black details against a bright background, distant twigs against a sky for example, are "injured" by diffraction sooner than fine bright details against a black background.
In the first instance the sky "bleeds" into the twig image and eventually obliterates it. In the second instance the bright line "bleeds" into the black background remains visible even in highly diffracted images.

 

[DREW WILEY]

That's called "hokey-bokeh", or something like that. Some people are willing to spend a lot of money for a little more of either one or the other in a lens. And apparently it can be gamed somewhat by specific lens design.

But if you realllllly want to get into a serious discussion of diffraction limits, that would in optical microscopy. And that's where sophisticated workarounds got developed which have been transferred over to Big Boy non-amateur telescopes, namely, the kind that take national and international budgets to buy.

 

[jimgalli]

The old unproven rule of thumb that actually works pretty well; Write down the numbers 16 - 22 - 32 - 45 - 64 and then underneath the 64 write down 16 - 22 - 32 - 45 - 64 in reverse order. In line pairs per millimeter that's roughly what happens, not that most lenses even got to 64, but a few did. In that scenario f90 would be 11 lppm and f128 would be 8 lppm. OK, go ahead and call me an oversimplified blithering idiot. Big shoulders. Used to it. It's a useful old timers rule of thumb. You knew if f64 was necessary there was some real cost kicking in. If, like me, you read Edward Weston's diaries and then Ansel Adam's letters, you'll discover a letter where Weston is puzzled about one of his shell pictures that he did over and over and over at f256 and it wasn't sharp and in the diary he wrote, movement. Something caused the camera to move. Later, Ansel explains to Edward that at f256 something called diffraction was robbing his sharpness, not movement. News to EW. Then there's the $5 RR that EW bought in Mexico. As good as his other lenses. Probably not but at f64 or f90 diffraction had become the great equalizer between all of his lenses.

 

[DREW WILEY]

Well, Jim, many of EW's contact print images would look pretty miserable enlarged much. But the greater mystery is why someone living out in light so clear even aliens prefer the area as their landing site would become the Grand High Guru of soft focus lenses? Did it begin with trying out an whiskey bottle found at an old Gold Mine dump, and the addiction ending up with Pinkham Smith, or the other way around? Or are you going to leave us guessing? Either way, you've never going to conclusively prove the aliens visited your town using soft-focus lenses.

 

[jimgalli]

I think what draws me to soft focus is the obvious beauty (in the eye of the beholder of course) but it didn't hurt that true soft focus lenses, each with a distinct personality, may be the last man standing in a digital world. AFAIK there's no Pinkham & Smith filter for photo shop that can acheive what a Series IV Pinkham lens can write on 8X10 film. Add to that the fact that the little green men are at my beckon call and they can slip into any place, any time, and pick me up whatever I fancy at the moment and bring it back to Tonopah. You don't really think i paid for all that stuff do you? (With apologies to the OP)

 

[reddesert]

The old unproven rule of thumb that actually works pretty well; Write down the numbers 16 - 22 - 32 - 45 - 64 and then underneath the 64 write down 16 - 22 - 32 - 45 - 64 in reverse order. In line pairs per millimeter that's roughly what happens, not that most lenses even got to 64, but a few did.

The angle spread of diffracted rays is directly inversely correlated to the physical aperture diameter, ...

spot diameter on film: c = theta * f = 2 * lambda * f/ D = 2 * lambda * N.

Jim's rule of thumb and the optics equation I posted relating diffraction spot size to f-number are essentially the same rule. Jim's rule of thumb is about:

lp/mm * f- number ~= 1000.

And for lambda ~ 0.55 microns (green light), the optics equation turns into:

1/(spot size in mm) * f-number ~= 900, (in units of 1/mm).

f-number 16 22 32 45 64
lp/mm . . 64 45 32 22 16
spot size 0.018 0.024 0.035 0.048 0.070 mm

You can see that the predicted spot size is roughly the reciprocal of the lp/mm number, which is about how it works out in practice - the spot diameter allows you to resolve the lp/mm at that frequency at some visible percentage contrast.