Diffraction Test

[AlkemScaleModels]

After last week's discussion about diffraction losses in lenses at small f stops, i decided to do a test to see how much diffraction loss my Canon 5D experienced with a 50mm Macro lens. These three shots are all shot under the same conditions with only the f stop varying. The three shots are at f32, f8 and f2.5. The original images shot in RAW and maximum quality jpg. The original images are 4368 by 2912 pixels. These images are too big to show on an web forum, so they were reduced to 800 pixels wide.

The first three shots show the overall scene.  In these images the best results are obtained with the f32 setting as it is sharpest overall.
  f32
f8
f2.5

The next three images are the actual pixels from the image brought over to the web forum without down sampling to fit. At this setting you can see some diffraction softness in the f32 image. The f8 and the f2.5 are very close, with the lack of depth of field in the f2.5 image making comparisons tricky. Look at the orange peel in the black color of the smoke box and the silver diagonal railing. The softness shows up there.





Note these images are hosted at my webshots account, so you may have difficult seeing them if you don't have a web shots account.

The next step would be to take tests shots with Helicon focus using f8 and f32 to see what the difference is.

[John]

I like the middle f8 the best .

[DKS]

Good choice of subject for a good test. Even more telling than the orange peel of the black paint is the detail painting on the side of the headlamp housing--the gold filigree in the corners is really crisp at f/8. Interesting to see some minor chromatic aberration around bright objects at f/2.5.

[tom mann]

Thank you for putting this together.  Photographic proof that just stopping down isn't sufficient if your image are intended for anything but shrinking down for the web.

Now, what you can do is to retake the shots at Iso100 to reduce as much noise as possible (although your Canon does a good job here).

[AlkemScaleModels]

Thank you for putting this together.  Photographic proof that just stopping down isn't sufficient if your image are intended for anything but shrinking down for the web.

Now, what you can do is to retake the shots at Iso100 to reduce as much noise as possible (although your Canon does a good job here).

Funny, that given the same data I draw a different conclusion. I feel that the softness at f32 was not bad, especially when you consider the overall depth of field. You have to consider the extreme magnification of the lower images.  But it is true for ultimate sharpness you should use optimum lens aperture.  Before Helicon Focus this wasn't an issue as f32 was the only solution. But HF has changed the whole equation.

So the next step is a test with a Helicon Focus shot at f8 and f32 and maybe f16 or f22 so see where the fall off starts. I may also try a comparison of my macro and zoom lens to see what they compare. The nice thing about shooting at f32 with HF is that the program has more sharp image area to work with. That allows a little more leeway in doing the stacking. With a smaller DOF with each f8 image, there is a possibility that there will be areas where there isn't sufficient overlap. You can shoot the images more closely spaced, so you can avoid it if you are careful.

One other thing to think about, the 5D has a 12 Mega-Pixel sensor with full frame CCD, so the photosites are probably slightly bigger than other digital cameras with APS sized CCDs. (30 percent more pixels but 60 percent bigger sensor compared to my previous Canon D60.) I suspect that results with other digital cameras may be more susceptible to diffraction effects as small apertures due to the smaller photosites on the CCD. So you should probably do a test with your own camera to see the delta.

Finally, as to David comments about the subject choice, I agree. Once SMR Trains started offering these locos I was hooked. I really don't have room for a O Scale layout but I am trying anyway.

[DKS]

One other thing to think about, the 5D has a 20 Mega-Pixel sensor with full frame CCD, so the photosites are probably slightly bigger than other digital cameras with APS sized CCDs. (30 percent more pixels but 60 percent bigger sensor compared to my previous Canon D60.) I suspect that results with other digital cameras may be more susceptible to diffraction effects as small apertures due to the smaller photosites on the CCD. So you should probably do a test with your own camera to see the delta.

Yes, the camera (particularly the sensor) being used will certainly make a difference, not to mention the lens. I've just got a 40D with a measly 10 MP sensor (newer point-and-shoot cams have better sensors!), so I would expect to see more diffraction artifacts than you do with the 5D. I do have a decent Canon 50mm prime lens, so that should help a little, but I also have a cheapo wide-angle zoom, so it will be interesting to run some tests.

Edit: just ran a really quick quickie test. 50mm lens on a Canon 40D, 400 ISO with a 12mm extension tube (the tube probably explains the acute chromatic aberrations in the f/1.4 images). No digital adjustment, minimal jpeg compression.









Next, I'll bracket around f/8 to see if there's a sweeter sweet spot.

[asciibaron]

not sure what the goal of all this is...  this seems to be basic photography theory that i learned in college.

for me, Bernie's first stop foreground and then the 2nd shot background put together would be the best "realistic" shot given the period he is modeling and the types of cameras used.

[DKS]

not sure what the goal of all this is...  this seems to be basic photography theory that i learned in college.

for me, Bernie's first stop foreground and then the 2nd shot background put together would be the best "realistic" shot given the period he is modeling and the types of cameras used.

Bernie is not looking for the best way to simulate period photography. He and I are exploring the subject of diffraction, and what is the best f-stop for maximum sharpness as opposed to maximum depth of field (which are different), as well as what impact Helicon Focus has on images taken at various f-stops. It is a continuation of this thread: http://therailwire.net/forum/index.php/topic,18945

[tom mann]

Bernie,

You are right, sometimes introducing a slight blur into the image goes a long way towards realism - and diffraction does that naturally. 

[AlkemScaleModels]

not sure what the goal of all this is...  this seems to be basic photography theory that i learned in college.

for me, Bernie's first stop foreground and then the 2nd shot background put together would be the best "realistic" shot given the period he is modeling and the types of cameras used.

Well, I have to admit, when I went to college, digital photography was the sole realm of NASA satellites. One thing that camera designers and therefore consumers have learned is that digital sensors don't behave exactly like film and that lens designed for film camera bodies might not work well in a digital body. Throw in software post processing and things really have changed. When was the last time you manually spotted a print (note this is a rhetorical question)? I recently found my photo-retouching inks and ended up throwing them away.

Early photographers had some crude lenses, true, but their 8x10 negatives have incredible detail. I love looking at high res scans of their images for details in modeling. We are planning the big truss bridge for my layout and the highly detailed photos are invaluable.

[Zox]

I've heard it said that pinhole "lenses" are the ideal for model photography, since they have infinite depth of field. (Although I imagine that for pinhole photography to work with a digital camera, you'd need an astronomical-grade CCD for the long exposures required.)

So where's the breakpoint between diffraction-troubled small apertures and "perfect" pinholes? I.e. at what point does the aperture go from being "too small for good pictures" to "too large for good pictures"?

(...I know, a silly question, coming as it does from a guy with a 3-megapixel point-and-shoot and no manual controls whatsoever...)

[DKS]

I've heard it said that pinhole "lenses" are the ideal for model photography, since they have infinite depth of field. (Although I imagine that for pinhole photography to work with a digital camera, you'd need an astronomical-grade CCD for the long exposures required.)

So where's the breakpoint between diffraction-troubled small apertures and "perfect" pinholes? I.e. at what point does the aperture go from being "too small for good pictures" to "too large for good pictures"?

(...I know, a silly question, coming as it does from a guy with a 3-megapixel point-and-shoot and no manual controls whatsoever...)

Pinholes may be "ideal" from the standpoint of virtually infinite depth of field (although it can't be infinite because focal length, pinhole size and other factors place a limit on where the DOF begins), but they suffer quite badly from diffraction. It's not a function of a lens aperture having some special aspect that plagues it with diffraction problems; it's that all small apertures suffer the effects, and the smaller the aperture, the bigger the diffraction problem; other problems like vignetting emerge as the aperture gets smaller. Not to mention the practicality of seriously long exposures (I nearly melted a few models sitting under hot lights for so long).

At first I was convinced that pinholes were "ideal" also. I was messing around with pinholes starting in high school using a Polaroid. By college, I had an SLR with which to experiment.



Decades later, I found laser-cut pinholes, which produce much sharper images because of reduced image degradation effects due to imperfections along the edges of the hole, but they still suffered from the effects of diffraction. The broken-down car in the foreground is a couple of inches from the lens, and the boxcar in the background is a couple feet away. The exposure was about a few minutes with a 300-micron pinhole. When this image is blown up, the focus is slightly soft.



Note that these images were produced using the pinhole as an aperture inside a conventional lens, which helps improve sharpness. The pinhole can of course be used alone, but then the image would be softer. More here: http://en.wikipedia.org/wiki/Pinhole_camera

[Zox]

David,

Thanks for the explanation. I am enlightened.

[lashedup]

In nearly every photographic review/test of lenses that I've read over the last 20 years it seems that *most* consumer zoom lenses achieve maximum sharpness at F8-F11. Basically setting the lens in the middle to lower-middle aperture setting is optimum in most cases. So if your lens as an aperture range of F3.5 to F22 then F11 or so would be the mid-way point where light fall off and diffraction are minimize. There are always exceptions to this of course depending on the type of lens, design and other factors. Plus some lens exhibit more light falloff/vignette depending on the aperture setting. So the setting you use is often a best compromise situation.

I think more importantly though is the discussion Bernie eluded to about the intended use of the photo in the end. For web use, sharpness can be brought back in very easily through reduction of the overall size of the source image and tweaking in Photoshop. For magazine or most general print purposes the line screen used in printing is going to offset minor issues in the original photo. The newer direct digital printing techniques allow for higher resolutions but also typically introduce their own limitations that exceed what you're getting out of the camera. In the end color rendition and proper reproduction of the original photo are *still* the primary issues going to print. Moreso in my opinion than the issues being discussed here.

If you guys were doing fine art books, large poster reproductions or anything that would require you to put up the native full-size image out of your camera then I'd be a lot more concerned about minimizing the diffraction issues.

Having said all that, I generally shoot most of my shots at F10 and use Helicon focus to get the best model photography.

- jamie