Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: dwdallam on July 20, 2006, 09:32:31 pm
-
I'd like to keep this thread as specific as possible. Using this defraction calculator, which most of you are familiar with, can someone shed light on the aperture setting where both cameras become defracted?
The test uses the 20D, but teh 5D is not shown. From what I get, the 5D seems to have a huge advantage in aperture sweet sopts compared to the 20D. What do you all come up with?
Scroll down on the page for the calculator:
http://www.cambridgeincolour.com/tutorials...photography.htm (http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm)
-
Forgot to add email notifications.
-
1) Smaller sensors (including film) will always diffraction limit sooner than larger ones due to the greater enlargement factor required to get to the same final image size.
2) Smaller pixels will always diffraction limit sooner than larger ones if you use pixel diameter (or a linear function of it) instead of a fixed numerical constant for CoC.
Combine the two and yes, the 20D will always be handicapped compared to the 5D.
-
One thing to take into account is the increased DOF from a small sensor. My guess is that it is about 1 stop (I have not seen figures, so if someone can supply them, that would be good). On that basis the 20D at F11 will have similar DOF to the 5D at F16. And a 20D at F16 will be comparable to a 5D at F22. I certainly would not want to use a lens at smaller apertures than F22. IMO the results are not good.
I believe that diffraction starts degrading the 20D image at about F16. (In the case of the D200 it is about F13, due to the higher pixel density).
So I don't think the 20D is at a disadvantage for macro work. Apart from the smaller number of pixels of course.
Leif
-
1) Smaller sensors (including film) will always diffraction limit sooner than larger ones due to the greater enlargement factor required to get to the same final image size.
2) Smaller pixels will always diffraction limit sooner than larger ones if you use pixel diameter (or a linear function of it) instead of a fixed numerical constant for CoC.
Combine the two and yes, the 20D will always be handicapped compared to the 5D.
[a href=\"index.php?act=findpost&pid=71349\"][{POST_SNAPBACK}][/a]
I think that Jack's reasoning is correct, but let's look at the situation from a different perspective.
To quote from the Cambridge site: "Are smaller pixels necessarily worse? Not necessarily. Just because the diffraction limit has been reached with large pixels does not mean the final photo will be any worse than if there were instead smaller pixels and the limit was surpassed;..."
In other words, if you had a full frame camera with the pixel size of the 20D, there would be no loss of image sharpness as compared the the 5D even at small apertures.
One situation where the 20D is not handicapped is when you are using cropped images. If you have a 5D and take a picture of a bird at some distance with a 200mm lens (assuming that lens is the only one available) and then crop the image to show the same area that the 20D captures with the same lens from the same shooting position, the 20D image will be better since the 20D has more resolution in terms of lp/mm than the 5D. Ray has shown this in a previous post.
Bill
-
One thing to take into account is the increased DOF from a small sensor. My guess is that it is about 1 stop (I have not seen figures, so if someone can supply them, that would be good). On that basis the 20D at F11 will have similar DOF to the 5D at F16. And a 20D at F16 will be comparable to a 5D at F22. I certainly would not want to use a lens at smaller apertures than F22. IMO the results are not good.
I believe that diffraction starts degrading the 20D image at about F16. (In the case of the D200 it is about F13, due to the higher pixel density).
So I don't think the 20D is at a disadvantage for macro work. Apart from the smaller number of pixels of course.
Leif
[a href=\"index.php?act=findpost&pid=71372\"][{POST_SNAPBACK}][/a]
Well, that is for an 8 by 10 inch print with rather loose standards for resolution. If you repeat the calculation for a 16 by 20 inch print and use 20/20 vision with default viewing distance, the camera is diffraction limited at apertures smaller than f/2. However, at f/2, lens abberations will limit resolution, so you have to stop down more.
Some time ago, I performed resolution tests with my D70 and the 50mm f/1.8 lens and found that the optimum aperture was f/5.6.
(http://bjanes.smugmug.com/photos/17635874-O-2.jpg)
-
Some time ago, I performed resolution tests with my D70 and the 50mm f/1.8 lens and found that the optimum aperture was f/5.6.
But the bar graph clearly shows that the bar for f/2.8 is higher than that for f/5.6; why do you then conclude that f/5.6 is the optimum?
-
I presume that graph is on-axis measurements only? Off-axis results will be very different. See Photo Zone for examples.
Leif
-
One situation where the 20D is not handicapped is when you are using cropped images. If you have a 5D and take a picture of a bird at some distance with a 200mm lens (assuming that lens is the only one available) and then crop the image to show the same area that the 20D captures with the same lens from the same shooting position, the 20D image will be better since the 20D has more resolution in terms of lp/mm than the 5D. Ray has shown this in a previous post.
[a href=\"index.php?act=findpost&pid=71374\"][{POST_SNAPBACK}][/a]
This is indeed true. The same lens from the same position will produce the same DoF whatever the format of the camera. However, if one doesn't crop the image from larger format to the same size as the smaller format, then comparing DoF can be subjectively confused by the fact that the images are different. For example, the image from the smaller format might consist only of a section of a 2-dimensional brick wall, whereas the larger format might contain the same section of brick wall plus a vast vista to the left, stretching to the horizon.
In this situation, the viewer would have no option but to declare that the larger format image had greater DoF.
For this reason, when comparing DoF of the image or print, it is only sensible to compare images of the same field of view. When we do this, the crop factor of 1.6x (in relation to the 20D and 5D) plays a role in most of the calculations. The 5D sensor is 1.6x larger in either dimension, than the 20D. The focal length of lens that produces the same FoV is therefore 1.6x longer (for the 5D) and the f stop # on the longer lens that produces the same DoF is 1.6x larger.
-
But the bar graph clearly shows that the bar for f/2.8 is higher than that for f/5.6; why do you then conclude that f/5.6 is the optimum?
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=71380\")
Because the displayed chart is only for the center of the field. At full aperture, off axis MTF suffers as shown in the MTF plot on PhotoDo. Stopping down to f/5.6 improves off axis performance.
[a href=\"http://www.photodo.com/nav/prodindex.html]http://www.photodo.com/nav/prodindex.html[/url]
-
I prefer the summaries that Klaus on Photo Zone produces. Here are results for a sample of the Nikon F1.8 AFD lens:
http://www.photozone.de/8Reviews/lenses/ni...50_18/index.htm (http://www.photozone.de/8Reviews/lenses/nikkor_50_18/index.htm)
As you can see, edge performance is still improving at F8, and might well improve more at F11, though by then on-axis performance is dropping. Note that Klaus used a D200 for his measurements.
I performed some tests with a 200mm Nikon lens on a D200, and the centre and edge sharpness were superb between F4 and F11, with a tiny softening at F16. I have heard similar comments from others. Klaus has also tested the Nikon 60mm F2.8 micro lens:
http://www.photozone.de/8Reviews/lenses/ni...60_28/index.htm (http://www.photozone.de/8Reviews/lenses/nikkor_60_28/index.htm)
As you can see, that is good to F11, with a slight softening at F16, and a slight edge softening wide open.
"For this reason, when comparing DoF of the image or print, it is only sensible to compare images of the same field of view. When we do this, the crop factor of 1.6x (in relation to the 20D and 5D) plays a role in most of the calculations. The 5D sensor is 1.6x larger in either dimension, than the 20D. The focal length of lens that produces the same FoV is therefore 1.6x longer (for the 5D) and the f stop # on the longer lens that produces the same DoF is 1.6x larger."
Thanks for pointing out the DOF ratio.
Your paragraph is very interesting, but the first part is a little contentious, at least with my style of photography. In practice we might have a 200mm macro lens which we use on a 5D and a 20D. In that case when using the 5D we will move closer to create the 'same' image. Now of course the image is not the same, as the perspective is different. However, DOF is a function of F ratio and image magnification only. Hence in these two situations (photographing a dragonfly that fills the frame say) the smaller sensor will have the larger DOF.
This is probably a significant advantage of smaller sensors for long telephoto photos e.g. birds. Not only does the 20D have a telephoto effect, it also effectively increases DOF.
Leif
-
In practice we might have a 200mm macro lens which we use on a 5D and a 20D. In that case when using the 5D we will move closer to create the 'same' image. Now of course the image is not the same, as the perspective is different. However, DOF is a function of F ratio and image magnification only. Hence in these two situations (photographing a dragonfly that fills the frame say) the smaller sensor will have the larger DOF.
[a href=\"index.php?act=findpost&pid=71419\"][{POST_SNAPBACK}][/a]
Dof is also a function of distance to subject. The closer you are, the shallower the DoF. If you move closer to the subject with your 200mm macro and 5D, in order to get the same FoV, then to get the same DoF as the 20D has from a greater distance with the same lens, you will have to increase the F# on the 5D by 1.6x.
This is probably a significant advantage of smaller sensors for long telephoto photos e.g. birds. Not only does the 20D have a telephoto effect, it also effectively increases DOF.
No it doesn't. If you wish to put a fine point on it and make a sufficiently large print, the 20D shot will have less DoF. If you imagine a 5D sensor filled with the smaller 20D pixels, there'd be close to 22m of them. The 22mp 5D shot cropped to the same size as the 20D shot would be identical in every respect for whatever size print. However, the 12.8mp 5D has less resolving power than the 20D, in terms of lp/mm on the sensor, so the 20D should produce a marginally sharper result than the 5D at the point of focus, but not necessarily sharper away from the point of focus where the CoCs are greater in size than the pixels of both cameras. In this case, at a sufficiently large print size, the out-of-focus parts of the images will be equally unsharp, but the in-focus parts will be sharper in the 20D print and thus DoF will appear to be shallower.
-
Yes, but the crop is a different subject. We're talking about diffraction, not pixel density at a specific mm telephoto range.
I think that Jack's reasoning is correct, but let's look at the situation from a different perspective.
To quote from the Cambridge site: "Are smaller pixels necessarily worse? Not necessarily. Just because the diffraction limit has been reached with large pixels does not mean the final photo will be any worse than if there were instead smaller pixels and the limit was surpassed;..."
In other words, if you had a full frame camera with the pixel size of the 20D, there would be no loss of image sharpness as compared the the 5D even at small apertures.
One situation where the 20D is not handicapped is when you are using cropped images. If you have a 5D and take a picture of a bird at some distance with a 200mm lens (assuming that lens is the only one available) and then crop the image to show the same area that the 20D captures with the same lens from the same shooting position, the 20D image will be better since the 20D has more resolution in terms of lp/mm than the 5D. Ray has shown this in a previous post.
Bill
[a href=\"index.php?act=findpost&pid=71374\"][{POST_SNAPBACK}][/a]
-
I prefer the summaries that Klaus on Photo Zone produces. Here are results for a sample of the Nikon F1.8 AFD lens:
http://www.photozone.de/8Reviews/lenses/ni...50_18/index.htm (http://www.photozone.de/8Reviews/lenses/nikkor_50_18/index.htm)
As you can see, edge performance is still improving at F8, and might well improve more at F11, though by then on-axis performance is dropping. Note that Klaus used a D200 for his measurements.
Leif
[a href=\"index.php?act=findpost&pid=71419\"][{POST_SNAPBACK}][/a]
Obviously, Klaus's full test is preferable to my limited results, but the point of my post was to show that on axis MTF decreases above f/5.6 and this is due to diffraction. Off axis performance increases with further stopping down, but this is related to reduced lens abberations outweighing the effects of increased diffraction. In practical shooting situations the choice of aperture should not be on the basis of diffraction alone, but depth of field and other considerations come into play.
Bill
-
Obviously, Klaus's full test is preferable to my limited results, but the point of my post was to show that on axis MTF decreases above f/5.6 and this is due to diffraction. [a href=\"index.php?act=findpost&pid=71462\"][{POST_SNAPBACK}][/a]
Nevertheless, Bill, your results show the Nikkor peaking at f2.8 (in the centre), falling off a bit at f4, rising again at f5.6, falling again at f8 and f11 and then falling dramatically at f16 and f22. In other words, there's quite a bit of unevenness of performance before diffraction really takes hold. What's particularly interesting is that centre performance in your tests is shown as being equal at f8 and f11.
The result for f2.8 does look suspicious. Off-axis performance is often not relevant when deliberately trying to get a shallow DoF. A lens which is sharpest at f2.8 is unusual.
-
Nevertheless, Bill, your results show the Nikkor peaking at f2.8 (in the centre), falling off a bit at f4, rising again at f5.6, falling again at f8 and f11 and then falling dramatically at f16 and f22. In other words, there's quite a bit of unevenness of performance before diffraction really takes hold. What's particularly interesting is that centre performance in your tests is shown as being equal at f8 and f11.
The result for f2.8 does look suspicious. Off-axis performance is often not relevant when deliberately trying to get a shallow DoF. A lens which is sharpest at f2.8 is unusual.
[{POST_SNAPBACK}][/a] (http://index.php?act=findpost&pid=71466\")
Ray,
You are over reading the graph. With any test there is a certain amount of error. A truly sceintific plot would show bar graphs with ± 2 standard deviation markers so that you could determine if the results of two bars are significant or represent random error. Of course, to get a standard deviation you would have to make at least 10 test runs and I did not do this. I suspect that the minor differences you note are not truly significant and represent random error.
For example, look at figure 5 in the link below. If the error bars on the plot overlap, there is no significant difference between them: the green and blue bars in that example are not significantly different.
[a href=\"http://scene.asu.edu/habitat/data_present.html]http://scene.asu.edu/habitat/data_present.html[/url]
Klaus's results are a bit smoother. He may have averaged several tests. However, it is interesting to note that he notes maximum axial MTF at f/4, which reinforces my point that MTF begins to fall off at a relatively large aperture with a well corrected lens. In Klaus's tests MTF at f/2.8 was better than f/8, so the lens does do well at f/2.8. Of course, he is using the D200, which outresolves the D70 used in my test, so he may see differences that I could not.
Bill
-
Dof is also a function of distance to subject. The closer you are, the shallower the DoF. If you move closer to the subject with your 200mm macro and 5D, in order to get the same FoV, then to get the same DoF as the 20D has from a greater distance with the same lens, you will have to increase the F# on the 5D by 1.6x.
No it doesn't. If you wish to put a fine point on it and make a sufficiently large print, the 20D shot will have less DoF. If you imagine a 5D sensor filled with the smaller 20D pixels, there'd be close to 22m of them. The 22mp 5D shot cropped to the same size as the 20D shot would be identical in every respect for whatever size print. However, the 12.8mp 5D has less resolving power than the 20D, in terms of lp/mm on the sensor, so the 20D should produce a marginally sharper result than the 5D at the point of focus, but not necessarily sharper away from the point of focus where the CoCs are greater in size than the pixels of both cameras. In this case, at a sufficiently large print size, the out-of-focus parts of the images will be equally unsharp, but the in-focus parts will be sharper in the 20D print and thus DoF will appear to be shallower.
[a href=\"index.php?act=findpost&pid=71437\"][{POST_SNAPBACK}][/a]
No DOF really is a function of F number and image magnification alone. (Assuming a fixed circle of confusion.) What you are saying is equivalent. By getting closer, you increase the image magnification, hence reduce the DOF. They are two ways of looking at the same phenomenon.
Yes of course if you take a picture of a subject with a full frame camera, and a cropped sensor camera at the same place with the same lens and F ratio, the DOF will be the same. But that is not the scenario that I described.
I'll explain again. Let's say that I photograph a dragonfly with a 5D and a 200mm lens, such that the insect fills the frame. Now let's say that I try to repeat the photograph using a 20D and the same lens at the same F ratio. Because the sensor is cropped, I must move backwards, until the dragonfly fits in the frame. Now I take the picture. Because the image is smaller, the DOF is larger. Or using your alternative and equally valid viewpoint , because I am further away from the subject, the DOF is greater.
Leif
-
Obviously, Klaus's full test is preferable to my limited results, but the point of my post was to show that on axis MTF decreases above f/5.6 and this is due to diffraction. Off axis performance increases with further stopping down, but this is related to reduced lens abberations outweighing the effects of increased diffraction. In practical shooting situations the choice of aperture should not be on the basis of diffraction alone, but depth of field and other considerations come into play.
Bill
[a href=\"index.php?act=findpost&pid=71462\"][{POST_SNAPBACK}][/a]
I quite agree. Leif
-
Yes, but the crop is a different subject. We're talking about diffraction, not pixel density at a specific mm telephoto range.
[a href=\"index.php?act=findpost&pid=71453\"][{POST_SNAPBACK}][/a]
Amen!
-
No DOF really is a function of F number and image magnification alone. [a href=\"index.php?act=findpost&pid=71480\"][{POST_SNAPBACK}][/a]
Agree totally and that is the complete correct answer -- but keep in mind that subject distance, lens focal and sensor size all factor into image magnification.
-
Agree totally and that is the complete correct answer -- but keep in mind that subject distance, lens focal and sensor size all factor into image magnification.
[a href=\"index.php?act=findpost&pid=71488\"][{POST_SNAPBACK}][/a]
As stated in the rest of my post (apart from focal length).
-
1) Smaller sensors (including film) will always diffraction limit sooner than larger ones due to the greater enlargement factor required to get to the same final image size.
2) Smaller pixels will always diffraction limit sooner than larger ones if you use pixel diameter (or a linear function of it) instead of a fixed numerical constant for CoC.
Combine the two and yes, the 20D will always be handicapped compared to the 5D.
[a href=\"index.php?act=findpost&pid=71349\"][{POST_SNAPBACK}][/a]
To get back on track, not always, Jack. The 20D will not always be handicapped compared to the 5D. The circumstances where it won't be handicapped are as described in previous posts, namely when the telephoto reach is not long enough with the 5D resulting in cropping of the image in post processing to the same size (or smaller) than the 20D image. In such circumstances, the 5D will never have an advantage at any aperture, but the 20D might. My tests have shown that even at a very diffraction limited stop of f22, the 20D can be sharper than the cropped 5D image (but not by much, of course. For normal size prints, both cameras would be equal, in my tests).
-
To get back on track, not always, Jack. [a href=\"index.php?act=findpost&pid=71514\"][{POST_SNAPBACK}][/a]
Sorry Ray. Taken exactly as I wrote my two statements above, they will *ALWAYS* be true.
If you add caveats about when and if you have to crop or increase magnifications, then you might have arguable points, but those extraneous conditions were not part of the statements I made.
Cheers,
-
1) Smaller sensors (including film) will always diffraction limit sooner than larger ones due to the greater enlargement factor required to get to the same final image size.
[a href=\"index.php?act=findpost&pid=71349\"][{POST_SNAPBACK}][/a]
Enlargement is not the proper term for a digital image, which has pixels but no physical dimension. That is why digital camera resolution is rated in pixels/picture height. For example, the Canon 5D has 2912 pixels/picture height and the Nikon D2x has 2848 pixels/picture height (the picture heights are 23.9 and 15.7 mm respectively). Once the images are brought into Photoshop, the sizes of the sensors are not relevant and both cameras can produce equally large pictures at a resolution of 300 pixels/inch on the picture. For a larger picture, you would have to upres.
The Nikon lens would have to have better resolution in terms of lp/mm, but it would have to cover a smaller sensor, simplifying optical design in the case of a DX lens.
-
Once the images are brought into Photoshop, the sizes of the sensors are not relevant [a href=\"index.php?act=findpost&pid=71517\"][{POST_SNAPBACK}][/a]
Okay Bill. I should have added to #1 "assuming pixel pitch remains smaller than CoC"
Then sensor sizes are absolutely relevant if you made the same composition with both and then intend to print out the same size image from both. That simple process circumvents the entire pixel-per-image-height discussion.
Cheers,
-
And 5D vs 20D DoF is very interesting to debunk, also for picture quality when you must crop and upres from teh 5D when printing larger images.
But my question is--CAN YOU SAY HIJACK?
LOL
-
If you add caveats about when and if you have to crop or increase magnifications, then you might have arguable points, but those extraneous conditions were not part of the statements I made.
[a href=\"index.php?act=findpost&pid=71515\"][{POST_SNAPBACK}][/a]
Jack,
And that's why it is not correct to write always. Cropping is a very common activity in image processing. We crop for all sorts of reasons. If you want to make a definitive statement free of caveats, conditions and exceptions, I suggest something along the lines of, "The sensor with the smaller pixels, whatever the size of the sensor, will be affected by diffraction first, when stopping down.
Now dwdallam is right that we are hijacking his thread. His original question was,
can someone shed light on the aperture setting where both cameras become defracted?
The answer is, the aperture at which the 5D becomes diffracted is also the aperture setting at which both camera's resolution is limited by lens diffraction.
As to what that aperture is precisely in terms of f stop# is difficult to say because of lens quality variation and the gradual change, over a number of stops, from mostly aberration limited to mostly diffraction limited. Even at f22 a lens can have a resolution of 72 lp/mm at 10% MTF.
-
CAN YOU SAY HIJACK?
-
Ray:
Speaking of hijacking, I'm not sure why you insist on adding your personal what if scenario on cropping and uprezzing to my comments. Within the parameters I very clearly stated, what I said will *ALWAYS* be true. Period.
As for you and Bill and your insistence on including cropping... I think what you guys are essentially saying is "if you compare crops at actual pixel view, the sensor with the smaller pixels will always -- err make that ususally -- win." And I am sure this will be a critical revelation to a few readers, so thanks for sharing.
,
-
All of this is pertinent, but the question was based on my using the calculator I linked to in the opening post. It seems that the 5D has a much greater latitude of sharp F stops compared to the 20Ds F8-F11. When I used the calculator and input the CoC of the 5D and other specific information, it looked like the 5D would perform as well as the 20D at apertures from as small as 3.5 to F18 (or something incredible like that). Is this what you all get using that calculator?
Hijacking my own post, I think pixel density and resolution are even more important for people to understand. Most people don't realize that doubling pixel density does not double resolution, nor does doubling pixel density double print quality. All things being equal, its all about resolution. That's what I just could not communicate to a friend who has a degree in computer science, and is a photographer. I kept saying that even though you get a roughly 30% increase in pixel density from the 20D to the 5D you don't get a 30% jump in resolution, nor a 30% increase in printed picture quality.