Luminous Landscape Forum
The Art of Photography => The Coffee Corner => Topic started by: tonysmith on September 24, 2009, 02:00:59 pm
-
I am on the mailing list of The Perfect Picture School of Photography. This is run by Bryan Peterson, who is a popular author of several books on photography. I took one of his courses a couple of years ago.
In some recent emails, Mr. Peterson has insisted it is incorrect that going to the smallest apertures on a D/SLR decreases overall sharpness because of diffraction. He strongly recommends ignoring such advice and using the smallest possible aperture whenever depth of field is important.
I assume this gentleman has credibility, but because this opinion contradicts what most experts say, I am interested in other opinions of his view.
Thanks
Tony
-
Hi,
It's absolutely certain hat DSLR lenses loose sharpness when stopped down to small apertures. This loss of sharpness can be compensated by sharpening to a certain extent. The effect is both measurable and visible.
This was measured on actual photographs: http://luminous-landscape.com/forum/index....ost&id=3387 (http://luminous-landscape.com/forum/index.php?act=attach&type=post&id=3387)
Based on my experience I'd suggest that going beyond f/16 may be detrimental to image quality, f/22 is OK if you really need depth of field and f/32 is bad.
Best regards
Erik
I am on the mailing list of The Perfect Picture School of Photography. This is run by Bryan Peterson, who is a popular author of several books on photography. I took one of his courses a couple of years ago.
In some recent emails, Mr. Peterson has insisted it is incorrect that going to the smallest apertures on a D/SLR decreases overall sharpness because of diffraction. He strongly recommends ignoring such advice and using the smallest possible aperture whenever depth of field is important.
I assume this gentleman has credibility, but because this opinion contradicts what most experts say, I am interested in other opinions of his view.
Thanks
Tony
-
Mr. Peterson has insisted it is incorrect that going to the smallest apertures on a D/SLR decreases overall sharpness because of diffraction.
He's just factually wrong - that's just physics.
The real question is ... for a particular scene and capture and intended output ... does it matter? Does the increased DOF and apparent sharpness around the focal plane make up for the decreased sharpness at the exact plane of focus?
-
Based on my experience I'd suggest that going beyond f/16 may be detrimental to image quality, f/22 is OK if you really need depth of field and f/32 is bad.
Depends, no? Whereas I try not to go above f/16 on the D700, I've heard that on the D300 that same impact is seen at f/11.
-
Everybody breaks a few laws. This guy just chose to do in physics rather than on the highway.
-
His account of the "facts" about diffraction may not be correct, but his advice has some merit (although I would word it a bit differently). Anybody who says you should never shoot a D300 with a smaller aperture than f/11 no matter what (or a D3 at f/13, or whatever) is probably just a pixel-peeping measurebator/theoretician. Sometimes absolute sharpness at the plane of focus is not the most important thing about an image. There are going to be times when having part of the image sharp and part of it out of focus would be undesirable, and getting a more consistent level of sharpness across the entire frame would be preferable. Having said that, shooting at f/22 just because you're too lazy to figure out that f/11 would give you enough DOF is just sloppy craft, and there's really no excuse for it unless calculating the DOF would mean missing the shot.
So I would say you should use the largest aperture that will give you adequate depth of field for your desired composition and intended output.
-
Another physic fact to throw in the equation : due to the AA filter, the effects of diffraction are less than the bare Airy disk calculation would lead to think (about one stop, depends on the filter strength).
But even with the relatively big pixels of my 300d (7.5µm), some loss of sharpness after sharpening is already visible at f/22, and yes I'll call f/32 bad.
Besides that, the interaction between depth of field and diffraction is not very simple, well described here eg : http://www.normankoren.com/Tutorials/MTF6.html (http://www.normankoren.com/Tutorials/MTF6.html)
That said, rather that taking the word of some authority or the other, the best thing to do may be to try it in the real world and indulge yourself with a bit of pixel-peeping.
Your conclusions may vary depending on your sharpness expectancies and needs.
-
In some recent emails, Mr. Peterson has insisted it is incorrect that going to the smallest apertures on a D/SLR decreases overall sharpness because of diffraction. He strongly recommends ignoring such advice and using the smallest possible aperture whenever depth of field is important.
I highly doubt that Mr. Peterson is ignorant of the effects of diffraction. He was probably just trying to explain that out-of-focus blur from insufficient depth of field is often more detrimental than diffraction.
-
Mr. Peterson has insisted it is incorrect that going to the smallest apertures on a D/SLR decreases overall sharpness because of diffraction. He strongly recommends ignoring such advice and using the smallest possible aperture whenever depth of field is important.
The key words here are 'overall sharpness'. Bearing this in mind, Bryan Peterson is absolutely correct. If the subject requires an extensive DoF, the loss of sharpness in the plane of focus, as a result of stopping down, will be relatively small compared to the increase in sharpness at significant distances from the plane of focus, which will be very substantial. If one were to do an analysis of the sharpness of all objects in an image shot at say F22, and averaged the MTF response or sharpness of all those objects at various distances for the whole image, one would find that such avarage sharpness would be far greater in the F22 shot than in the F11 shot. Put another way, the overall sharpness tends to increase as one stops down, but specific sharpness at or near the plane of focus decreases by a smaller degree.
-
An interesting offshoot of this concept relates to the usefulness of the latest camera with yet another increase in pixel count which some folks decry, claiming that current lenses are not good enough to provide any benefit from such pixel-count increases. An example would be the 15mp 50D which closely followed the 10mp 40D. What use is F11 or F16 with a 50D? The additional pixels are wasted because of diffraction, surely.
Not at all. If you do some testing, as I have done, you'll find that the 50D at F16 is as sharp at the plane of focus as the 40D at F11, and the same applies comparing the 50D at F11 with the 40D at F8. The major benefit of the increased pixel count of the 50D is that one can achieve greater DoF than the 40D without compromising sharpness at the plane of focus.
-
An interesting offshoot of this concept relates to the usefulness of the latest camera with yet another increase in pixel count which some folks decry, claiming that current lenses are not good enough to provide any benefit from such pixel-count increases. An example would be the 15mp 50D which closely followed the 10mp 40D. What use is F11 or F16 with a 50D? The additional pixels are wasted because of diffraction, surely.
Not at all. If you do some testing, as I have done, you'll find that the 50D at F16 is as sharp at the plane of focus as the 40D at F11, and the same applies comparing the 50D at F11 with the 40D at F8. The major benefit of the increased pixel count of the 50D is that one can achieve greater DoF than the 40D without compromising sharpness at the plane of focus.
Ray
Which really boils it down to the practical situation, which is that one might as well carry on using lenses more or less in the manner one did on similar formats in the days of film, where the smallest stop in 35mm photography wasn't considered advisable anyway, and many lenses didn't go beyond f16 for that very reason, smaller stops being left to LF equipment where depth was often insufficient without movements, and final, relative magnification generally small enough to hide any ultra small stop failings.
One can watch the monitor until the lights go out, but it's just the prints that really count if your interest is images on paper, and as looking at a screen seems to bear little practical relationship to how something prints at the best of times (definition-wise) because of screen resolution, what the hell!
Rob C
-
I have been interested in this for some time and have been sticking to f/11 on my D300. Recently I shot four images of a leaf on a plant with a vein running through it. f/11, f/16, f/22 and f/32 The difference between the f/11 and the f/32 was slight in the sharpest area of the vein. Not the most scientific experiment but in future I will be using f/16 and f/22 more. Last year a question was asked in the Amateur Photographer magazine UK about diffraction and a good explanation was given. At the end the person answering the question stated that it didn't really matter what aperture was used because if 10x8 was the printing size then the limits of diffraction wouldn't show up at that size in print?
-
A bit of googling and you can find what the man said ... he is clearly not ignorant about diffraction, but erroneously claims that concerns about diffraction were invented by 'pixel peepers' in the digital age.
-
Hi,
It's absolutely certain hat DSLR lenses loose sharpness when stopped down to small apertures. This loss of sharpness can be compensated by sharpening to a certain extent. The effect is both measurable and visible.
This was measured on actual photographs: http://luminous-landscape.com/forum/index....ost&id=3387 (http://luminous-landscape.com/forum/index.php?act=attach&type=post&id=3387)
Based on my experience I'd suggest that going beyond f/16 may be detrimental to image quality, f/22 is OK if you really need depth of field and f/32 is bad.
Best regards
Erik
My experience is similar to Eric's and the issue can be backed up with quantitative data. Eric gave a link to some Imatest results.
Photozone.de (http://photozone.de/nikon--nikkor-aps-c-lens-tests/394-nikkor_60_28?start=1) is an excellent German site that gives Imatest results for a wide variety of cameras and lenses. The link shown is for the Nikkor 60 mm f/2.8 AFS on the Nikon D200. The maximum MTF 50 is at f/5.6, but results are decent down to f/16 and quite poor at f/32. Most lenses for 35 mm formats stop at f/16, but this lens is a macro lens where depth of field is very shallow at higher magnification and f/32 allows greater DOF at the expense of lower MTF.
I did my own tests with this lens on the Nikon D3. I rendered into TIFFs with ACR using the default settings, which include conservative capture sharpening. Shown below are the Imatest plots for f/4 and f/22. Readers are referred to the Imatest documentation for a detailed explanation of how the program works and how the results are interpreted. Briefly, the program uses a slanted edge target with a black-white transition. The steeper the curve as one goes from white to black, the better the MTF, and the MTF can be calculated from this curve. The slope is steeper for f/4 than f/22. The resolution at 50% contrast is 44.5 lp/mm at f/4 and 30.3 lp/mm at f/22. At f/4 the system resolves about 75% of Nyquist, which is typical for Bayer array cameras. Over sharpening can give resolution above Nyquist, but one is not seeing true detail but garbage due to aliasing. Unfortunately, Imatest does not distinguish between the two possibilities.
The blue shaded area to the right of the Nyquist frequency on the X-axis is aliasing. One advantage of f/22 is that the lens is acting as a low pass filter and reduces aliasing. This is analogous to the situation with P&S cameras, which don't have blur filters since the lens performs this purpose at the very small pixel size of these cameras.
[attachment=16806:F4_F22_Results.png]
Shown below are resolution charts from the same images, with the f/4 results on the top and the f/22 results on the bottom. The target was not photographed at the subject distance needed to interpret the resolution in lp/mm, but the results are relative. As the resolution approaches Nyquist, the image breaks up due to alaising, which is prominent at arount 10 on the lower set of bars in each image. On visual inspection, the improved contrast at f/4 is apparent at even the lowest resolution. However, detail is present up to Nyquist even at f/22, but the contrast is low. Indeed, as Roger Clark's (http://www.clarkvision.com/imagedetail/scandetail.html#diffraction) Optical System Resolution Limits chart shows, a perfect lens at f/22 can resolve 35 lp/mm at 50% contrast and 75 lp/mm at the Rayleigh limit, which is about 10% contrast. The Nyquist of the D3 is about 59 lp/mm (118 lw/mm). Cycles/mm is the same as lp/mm, and lw/mm (line widths) is double the former. The 60 mm f/2.8 lens is not perfect, but at f/22 it is safe to assume that it is diffraction limited. The old style USAF and similar resolution bar charts measure MTF at about 10%, but more recent studies have shown that MTF at 50% correlates best with perceived image quality.
[attachment=16807:F4_F22_Res.png]
-
There is a tradeoff between sharpness due to diffraction (which affects the entire image) and sharpness due to DOF (which affects more the farer you get from the focus plane). In general it is not a good advice to stop down as much as you can, specially if you don't need a tremendous DOF.
There is a nice DOF/diffraction calculator by Max Lyons (http://www.tawbaware.com/maxlyons/calc.htm) that implements the physics equations to find the proper balance between diffraction (green is blurrring because of diffraction) and DOF (blue is blurring because of DOF):
(http://img10.imageshack.us/img10/6937/chartanim.gif)
Red line is a suggested CoC.
Here just a sample of the real effect of diffraction (APS-C, Canon 10-22mm@15mm, focus distance about 3-4m):
(http://img3.imageshack.us/img3/4663/aperturaoptima.jpg)
Unless the blurring because of a lack of DOF is greater than this, it's nonsense to go beyond f/11 with this situation. For indoor shooting I simply set f/11 on this lens.
However in this forum somone said something that I found very interesting and made me think. He said sometimes it could be worth, for instance in arquitecture photography, to stop down more, even if that meant a reduction in the average image sharpness, in order to make sharpness more uniform accross the image at all distances.
Perceptually it can be less convenient to be able to detect differences in sharpness accross the image, than having an overall lower but uniform sharpness. This uniformity also helps in postprocessing, specially to choose the optimum global unsharp mask parameters to be applied.
BR
-
There is a tradeoff between sharpness due to diffraction (which affects the entire image) and sharpness due to DOF (which affects more the farer you get from the focus plane). In general it is not a good advice to stop down as much as you can, specially if you don't need a tremendous DOF.
However in this forum somone said something that I found very interesting and made me think. He said sometimes it could be worth, for instance in arquitecture photography, to stop down more, even if that meant a reduction in the average image sharpness, in order to make sharpness more uniform accross the image at all distances.
Perceptually it can be less convenient to be able to detect differences in sharpness accross the image, than having an overall lower but uniform sharpness. This uniformity also helps in postprocessing, specially to choose the optimum global unsharp mask parameters to be applied.
BR
Those are good points, but one should also remember that if one portion of the image is sharp, the viewer may regard the image as sharp. For example, in portrait photography is is usual to focus on the eyes, and if these are sharp the image is usually judged to be sharp. Often a large aperture is used to limit DOF in portraiture. For architecture, when all areas of the image are of interest, I would think that overall sharpness would be desirable.
-
Those are good points, but one should also remember that if one portion of the image is sharp, the viewer may regard the image as sharp. For example, in portrait photography is is usual to focus on the eyes, and if these are sharp the image is usually judged to be sharp. Often a large aperture is used to limit DOF in portraiture. For architecture, when all areas of the image are of interest, I would think that overall sharpness would be desirable.
Agree. Using diffraction as a sharpness equalizer only makes sense in large DOF applications (arquitecture, landscape,...).
-
Of course diffraction is a law of physics, but we can do something about it because image resolution is not totally dependent on lens resolution. Image resolution is a 'system' resolution comprised of both sensor and lens resolution. Increase either one, and image resolution is increased to some extent.
Ultimately, it's the detail in 'real-world' images that counts for the photographer. Some mathematical descriptions matching pixel size with Airy disc size are simply not accurate according to my eyes.
The following 200% crops of an Australian $50 bill featuring the very talented Aboriginal, David Unaipon, demonstrate that the 50D at F16 is the equal of the 40D at F8, for most practical purposes, using the good Canon 50/1.4.
These images have minimal processing, default sharpening of 25 in ACR, and slight adjustment of levels in Photoshop to get the images looking similar. The 40D image has not been uprezzed, but enlarged to 243% on monitor to match the physical size of the 50D image on monitor.
It goes without saying that tripod and remote release were used for these shots, as well as LiveView for focussing. The 50D has a higher resolution LCD screen than the 40D, which makes accurate focussing on such a target as this banknote easier, so I found it necessary to adopt the 'aliasing artifact' procedure when focussing the 40D.
The aliasing artifacts on each side of David Unaipon's face were clearly visible on the 40D's LCD screen and an indication that focussing was 'spot on'.
One could argue that the 50D shot at F16 is superior to the 40D shot at F8 because of its lack of moire. At 200% on monitor there's no additional 'real' detail in the 40D shot at F8, as far as I can see. Have I missed something?
It should be obvious, if the subject were 3-dimensional the 50D at F16 would trounce the 40D at F8 in 'over all' sharpness.
[attachment=16809:50D_F16_..._at_200_.jpg]
-
Agree. Using diffraction as a sharpness equalizer only makes sense in large DOF applications (arquitecture, landscape,...).
And there are other ways of handling depth of field such as focus stacking, which is enabled in PSCS4 and in specialized programs. See this (http://tv.adobe.com/watch/learn-photoshop-cs4/using-advanced-compositing/) movie by Deke McClelland for a demonstration of focus stacking in Photoshop. The focus stacking part of the movie starts at the 2 minute mark. An animation of a 12 image stack performed in Photoshop is here (http://prometheus.med.utah.edu/~bwjones/C853940471/E20090311124658/index.html).
As in panorama work, specialized programs are available for focus stacking, such as CombineZM (http://www.wonderfulphotos.com/articles/macro/focus_stacking/) , Zerene Stacker (http://www.janrik.net/insects/ExtendedDOF/LepSocNewsFinal/EDOF_NewsLepSoc_2005summer.htm) and Helicon Focus (http://www.heliconsoft.com/heliconfocus.html). I haven't tried focus stacking but plan to give it a try in Photoshop. Does anyone have any experience or suggestions with this technique?
-
And there are other ways of handling depth of field such as focus stacking, which is enabled in PSCS4 and in specialized programs. See this (http://tv.adobe.com/watch/learn-photoshop-cs4/using-advanced-compositing/) movie by Deke McClelland for a demonstration of focus stacking in Photoshop. The focus stacking part of the movie starts at the 2 minute mark. An animation of a 12 image stack performed in Photoshop is here (http://prometheus.med.utah.edu/~bwjones/C853940471/E20090311124658/index.html).
As in panorama work, specialized programs are available for focus stacking, such as CombineZM (http://www.wonderfulphotos.com/articles/macro/focus_stacking/) , Zerene Stacker (http://www.janrik.net/insects/ExtendedDOF/LepSocNewsFinal/EDOF_NewsLepSoc_2005summer.htm) and Helicon Focus (http://www.heliconsoft.com/heliconfocus.html). I haven't tried focus stacking but plan to give it a try in Photoshop. Does anyone have any experience or suggestions with this technique?
My trial of CombineZP (new version) was intriguing ... there were issues with color management, but the focus stacking was better than what I was able to achieve in CS4.
i'm totally new to focus stacking ... next up is helicon ...
-
Focusing-stacking software/technology is still immature. I guess it works OK for some types of images; but it has problems in many cases. Not only do you have alignment issues due to lens breathing, but even in properly aligned images you will often get artifacts such as halos. It seems the biggest problems are when you have overlapping planes of focus, such as a foreground element that sticks up in the frame and overlaps with the background.
-
Does anyone have any experience or suggestions with this technique?
I have tried TuFuse (http://www.tawbaware.com/tufuse.htm), which is a particular coding by Max Lyons of this Exposure Fusion algorithm (http://research.edm.uhasselt.be/~tmertens/papers/exposure_fusion_reduced.pdf). This algorithm classifies pixels according to 3 variables: correct exposure, microcontrast (in the near surrounding area) and saturation. By weighting more some of those variables, this algorithm can be easily used for totally different purposes: e.g. HDR fusion (tone mapping) by weighting more the correct exposure, or focus stacking by weighting more the contrast variable.
I did a try and was not bad, but displayed halos in some areas (see the word 'Formas' in the calendar).
However I think I simply was too demanding with the algorithm. IMO these algorithms work well with a quite in-focus set of images, not images with such a strong defocusing as the ones in my test. We also need lenses that minimise changes in the entrance pupil position (perspective), distortion and exact FOV when changing the focusing distance.
Original set of 3 images: shallow DOF with different focus distances:
(http://img101.imageshack.us/img101/2263/pdc1gh4.jpg)
Result from TuFuse contrast fusion:
(http://img101.imageshack.us/img101/3996/pdc2nh6.jpg)
What I find interesting is the discussion about 'when do focus stacking'. At first I simply thought focus stacking could just be an unnecesary improvement in DOF that could be achieved anyway by properly stopping down the lens.
Thanks to Max Lyons forum, I understood the real reason for justifying focus stacking: it is very recommended or it can even become a must, everytime the scene is to be captured with a very high resolution (i.e. devoting many pixels to every area of the scene, typically when stitching). In that case the maximum allowed blurring is so small that the tradeoff between diffraction + DOF simply becomes impossible; any combination of both effects will blurr image details spreading over several pixels because of the high pixel density with respect to image detail.
- For example, if you stitch several shots done at 60mm, it will be very difficult to achieve in every single shot non-blurred pixels at all distances. If we want sharpness in 100% crops the tele lens will need several shots at different focus distances because it needs to be stopped down more to achieve high DOF.
- This would not happen if a single shot with a wide angle is done over the same FOV. In that case the final lower resolution will limit the captured detail before diffraction + DOF does.
Of course the stitched image will allow larger prints than the single shot, both to be observed at the same distance and providing the same degree of detail. After all this is the goal of stitching.
Regards