Is there any problem in image quality when shooting @ f/22? And what if I use prime lenses, will this minimize that small f/stop?
Thanks
Posted by Basil Greber on February 3, 2008 at 11:49 AM
You'll lose resolution due to diffraction. When exactly this happens depends on the sensor. It's around f/13 for DX sensors (commong "consensus" says f/13, but resolution tests reveal that good lenses are diffcartion limited on a 10 MP DX sensor already at f/8).
For the 12 MP FX sensor of the D3 (you're apparently interested in getting one), the effect may not be as severe, though.
I am interested to hear that this is sensor dependent. I have older lenses that I used with 35 mm film for many years. In addition to the crop effect caused by those sensors, can I expect image quality differences? I developed some favorite lenses over the years. Should I be ready to re-think things as new sensors roll in? My first digital was 6 mp Fuji S2. My second Nikon (Sony?) 10 mp. Who knows what the next will be (but full-frame would be nice).
Pete D.
Posted by David Bostedo on February 3, 2008 at 12:59 PM
Diffraction is affected by pixel size more so than the particular sensor. Light spreads out more the smaller an opening it passes through. So the blurring effect of diffraction will show up sooner with smaller pixels (from a smaller sensor for a given pixel count, or higher pixel count for a given sensor size). But you will be capturing more detail to begin with on a higher megapixel sensor, so for a given size image it's kind of hard to calculate when you might notice it. I think you'll definitely notice it at f/22 on just about any sensor.
One check for it is to take the lens and camera you are curious about, and simply take the same image at different f stops. (Preferably an image of something like a brick wall, so that DOF isn't an issue.)
There's 2 issues here. Diffraction is caused by the small aperture and is not related to the sensor. Resolution will drop as you move away from the center of the image circle. This is because light rays passing through a small aperture curve away towards the edges. Higher-resolutions sensors will simply allow you to see the degradation better because they are able to capture more image detail in the center.
At f/22 most lenses, both primes and zooms, will show noticeable loss of resolution at the edges compared to f/8-f/11. Still, you should not avoid using f/22 if that's the only way to get significant elements in your composition to look sharp. Having them out of focus will be much more distracting than loss of resolution.
Guy
Posted by Kent Staubus on February 3, 2008 at 1:21 PM
The only time I use f22 is when I'm looking for a deliberately dreamy, blurry look.
Kent in SD
Posted by Dan Borzynski on February 3, 2008 at 2:23 PM
My basic feeling is this: If you need f22, shoot at f22. If you don't, then don't.
Diffraction is a real issue, but sometimes people get more uptight about having a technically perfectly sharp image when it isn't 100% necessary. Photo competitions with anal judges, professional portfolio images aimed at a particular client, ad agencies, okay. But composition, lighting and eye appeal really make at least as much of a difference in the long run as simple sharpness. I am not saying that sharpness should be a secondary consideration, just that sometimes it isn't the only important thing.
Assuming that you are shooting with quality lenses, diffraction is less of an issue than if you were shooting with cheap consumer zooms. The laws of physics aren't any different, but if you are starting with poor optics, the additional loss of sharpness due to diffraction just makes a bad situation far worse.
A secondary consideration is how large do you intend to print at, relative to the size of the recording medium. F16-22 is usually my starting point when shooting large format. There, a 16x20 is only a four fold increase in size for a 4x5 chrome. In 35 mm film or digital equivalent, that is a 4x6 print. Diffraction effects at f22 are almost beyond the ability of the eye to see in a 4x6 print. On the other hand, if I were printing to 40x60, I might be more concerned about it shooting at f32 or f45, even in 4x5.
If in doubt, especially if you are shooting digital, take two shots, one at f8-11, one at f16-22 and see what works at your intended print size.
Posted by Doug Dolde on February 4, 2008 at 12:00 AM
I think those who advocate using f22 are seriously mistaken unless you are shooting 4x5 film.
Have a look at Lloyd Chambers' study, the degradation is clear.
Posted by Mel Ladewig on February 4, 2008 at 1:25 PM
Doug,
First of all, I believe Moshen shoots medium format, so he may fall into some range between 35mm and large format.
There is no doubt that f22 has a significant diffraction impact on 35mm. But I usually find an unintentional out of focus region in image to be significantly more distracting than a slightly soft image. I suspect that you'd find most viewers, including those who might be inclined to purchase a photo feel the same.
I stick by Guy and Dan in saying, if you need f22, then use it and deal with the consequences. If not, then don't.
Mel-
Posted by Basil Greber on February 5, 2008 at 2:24 AM
Guy, I disagree. Diffraction is related to the sensor due to pixel density. A more pixel dense sensor will show diffraction effects more early than a not so pixel dense sensor (that's why most say that a FX sensor has an advantage over a DX sensor here).
Additionally, there are small diffraction patterns overlaid at every pixel (basically every ray of light, going towards all the directions necessary to light the sensor, is affected by it; diffraction causes that ray (or the photons of which it consists) to spread out a little to the sides, lowering definition of the ray, and therefore lowering resolution, or rather, increasing the distance between two points that can be clearly resolved), not a big one on the whole sensor that "makes light curve away to the corners". Consequently, resolution drops over the whole image, not only at the borders.
For the diffraction pattern created by a single ray to reach over the whole sensor, the aperture would be tiny (f/2000 or so, maybe ), and the resolution would go down to 0 (complete blur).
Basil
Posted by Basil Greber on February 5, 2008 at 2:28 AM
Assuming that you are shooting with quality lenses, diffraction is less of an issue than if you were shooting with cheap consumer zooms. The laws of physics aren't any different, but if you are starting with poor optics, the additional loss of sharpness due to diffraction just makes a bad situation far worse.
Diffraction is not a penalty that is kind of subtracted from the resolution capability of the lens. It's a upper limit imposed by physics. So, more expensive lenses will be diffraction limited already at f/8 (check photozone.de's resolution measurements for some good Nikon lenses - they drop at f/5.6 or f/8), while cheapos may go up until f/11 because they started at a lower level and can't reach the diffraction limit earlier. But once diffraction limits resolution, all lenses should resolve approximately the same (there are some that handle it better, and some that handle it a bit worse, but the differences are small).
Basil
Posted by Dan Borzynski on February 5, 2008 at 12:38 PM
Basil, your arguments may be true, but irrelevant. Diffraction does noticeably set in early with high quality optics - I will grant you this. But you also strongly imply that by f22, image quality is essentially the same regardless of the starting point of lens.
This simpy does not agree with my experience.
Yes, at some point, diffraction does take over to render everything soft, but you need to go further than that. I have used both consumer grade lenses and pro-level lenses. At f16-22, there is still a difference. Yes, the top end equipment does have image quality falling off faster than the consumer lenses, if only because the consumer lenses start at a lower point. But there is still a difference. f22 is not enough of an equalizer of performance.
And no, I rarely shoot any 35mm-based lens at f22. The results are usually not all that great. That aperature is almost exclusively reserved for situations where form is more important than details.
Posted by Jim Jirka on February 5, 2008 at 3:40 PM
there was diffraction way before sensors. i tend to agree that it is the physics of the lens, not the sensor.
Posted by Rod Barbee on February 6, 2008 at 3:27 PM
All I know is that I can use my Nikon 24mm at f/22 on a film camera and enlarge images to 16x24 or 20x30 and they look great.
I suppose if one chooses to view these prints at 6 inches or with a magnifying glass, then some problems will crop up. But in real world viewing situations, it just doesn't seem to matter that much. At least with that lens and when shooting Velvia.
Now with my 12-24 on a D200 or D300, I can see, on the monitor at 100%, the loss of resolution due to diffraction at f/22. F/16 looks a lot better.
I guess what it boils down to is this: take some pictures, evaluate the results. Experiment and see what is acceptable to you.
I keep looking in on this thread out of curiousity to see how the discussion is evolving. No one has yet mentioned what the real world implications of looking at a digital file at 100% resolution means (and I don't have much experience since I shoot mostly LF). Does that equal a print of 8 x 10, or 24 x 30, or what? (Yeah, I know, it depends on the camera's resolution - work with me here!! :) The link posted by Doug above is interesting, but the section of the image that is used for comparison is so small that I have no way to calibrate it. Is that detail equivalent to a 40 x 50 print (for example)? If so, you should be shooting large format anyway, or stitching digital images together if you really want high resolution. Ultimately, I agree with the suggestion offered several times that you take some pictures, figure out what the results mean based on your needs and image usage. But any detail on what looking at a file at 100% means in real life would be interesting to me.....
Thanks, CJ
Posted by David Bostedo on February 6, 2008 at 6:45 PM
CJ - Not only does the camera resolution matter for what you see at 100%, so does the size of your monitor. You're right, though, that the focus should be on what it means for certain sized prints, or certain sized web-viewing (the two main ways images are presented these days). I don't know if I can answer it, but I've been thinking about it a lot lately (too much maybe?).
The reason 100% is useful, though, is because it's showing an actual image pixel using an actual screen pixel. So you should be eliminating any funkiness (technical term ) introduced by interpolating the image for screen display. (This is assuming that you're working at native resolution of course - yet ANOTHER variable.)
There's been quite a lot of discussion on other sites about both diffraction effects, and noise. Many of them, like you mention, make use of 100% crops to display noise or diffraction blurring. However, as the resolution goes up, and monitor size increases, a 100% crop is becoming a smaller and smaller portion of the actual image, displayed larger and larger (unless screen resolution keeps going up too).
So if you have a 6 megapixel image, and you take a screen sized 100% crop, you've probably got about 1/6 of the image. If you do the same with a 12 megapixel image, you have half that. If the noise/diffraction blurring is worse in the 12 megapixel image it wouldn't be surprising - since it's sort of equivalent to taking a 200% shot of the 6 megapixel image.
On a pixel per pixel basis, you could say which sensor is cleaner - you can see the noise introduced from one pixel to the next based on the 100% view. You could also possibly figure out which pixel size was more or less sensitive to diffraction (well, physics tells you that) and how sensitive they are for that sensor. But in a print are you better off having more noisier pixels, or less cleaner pixels? More slightly diffraction blurred pixels? Or less diffraction blur-less pixels? And what about the AA filter on the sensor? It blurs all digital images intentionally.
Specifically on the diffraction issue, though, think about this : say we have two sensors - the 12 MP D3 sensor, and the 25 MP D3X sensor (assuming that's what it will be). It's true that diffraction will affect the pixels much earlier in the D3X due to their smaller size. However, all that does is reduce the effective resolution (maybe every pixel gets slightly blurred into it's neighbors). So maybe you get an image with as much detail as the 12 MP sensor. And at apertures that don't experience much, or any diffraction, then you can get the real benefit of the 25 MPs. A 100% crop from the D3X at f/22 might look much worse than that from the D3 (diffraction wise) but in a same-sized print, the D3X will have had a lot more pixels to work with.
So say you now print both at 8x10, and 20x30 - now you've finally arrived at more of an apples to apples comparison. I think, in this case, f/22 may be bad enough to give you noticable diffraction blurring for just about any sensor. However, a higher pixel count, while being more susceptible to pixel blurring, will still have more pixels on any given detail that is captured. This may make it better for the larger print. Either way, you'll have something similar to a lower res image of the same scene. So if you're OK with that, and need f/22 for DOF - go for it.
David - thanks for your reply - I'm not sure if it helped or not!!??!! I think I'll just stick to LF, where f22 is pretty much standard! Seriously, I don't fully understand all of that, but to a large degree you are just reinforcing that you need to do the tests for real world use and decide what you are comfortable with!
Thanks, CJ
Posted by David Bostedo on February 7, 2008 at 11:21 PM
OK CJ - My last post was probably just a tiny bit too long and rambling And I'm not sure if it's right, but maybe this summary will help?
Issue 1 : Viewing at 100%
- Can show pixel to pixel differences, but higher mega-pixel sensors show less of the image at 100% possibly giving apples to oranges comparison
- Print and screen viewing quality will depend on total number of pixels and print size, as it always has - 100% viewing may not tell you much
- So packing more pixels into a same sized sensor may appear to be worse when viewed at 100%, but probably won't show up in a print
Issue 2 : Photos at small apertures
- More pixels will give better resolution at apertures where diffraction is not noticable
- Even at smaller apertures, where diffraction is noticable, this is an effective lowering of the resolution, since some pixels are combined by the diffraction blurring
- Since the resolution is just lowered, you should theoretically get similar results to using a sensor with less pixels where diffraction is not an issue
- So as far as diffraction is concerned, packing more pixels onto a same sized sensor shouldn't really be worse than having fewer, larger pixels
The one area where packing more pixels into a given sensor size is almost always bad is noise performance...otherwise, it seems like shrinking pixels aren't that big a deal.
(I know I'm just repeating a lot, so please stop me if I'm getting out of hand... )
Discussion
17 repliesThanks
For the 12 MP FX sensor of the D3 (you're apparently interested in getting one), the effect may not be as severe, though.
Basil
Pete D.
One check for it is to take the lens and camera you are curious about, and simply take the same image at different f stops. (Preferably an image of something like a brick wall, so that DOF isn't an issue.)
Here's a really good tutorial on diffraction from Cambridge In Colour.
At f/22 most lenses, both primes and zooms, will show noticeable loss of resolution at the edges compared to f/8-f/11. Still, you should not avoid using f/22 if that's the only way to get significant elements in your composition to look sharp. Having them out of focus will be much more distracting than loss of resolution.
Guy
Kent in SD
Diffraction is a real issue, but sometimes people get more uptight about having a technically perfectly sharp image when it isn't 100% necessary. Photo competitions with anal judges, professional portfolio images aimed at a particular client, ad agencies, okay. But composition, lighting and eye appeal really make at least as much of a difference in the long run as simple sharpness. I am not saying that sharpness should be a secondary consideration, just that sometimes it isn't the only important thing.
Assuming that you are shooting with quality lenses, diffraction is less of an issue than if you were shooting with cheap consumer zooms. The laws of physics aren't any different, but if you are starting with poor optics, the additional loss of sharpness due to diffraction just makes a bad situation far worse.
A secondary consideration is how large do you intend to print at, relative to the size of the recording medium. F16-22 is usually my starting point when shooting large format. There, a 16x20 is only a four fold increase in size for a 4x5 chrome. In 35 mm film or digital equivalent, that is a 4x6 print. Diffraction effects at f22 are almost beyond the ability of the eye to see in a 4x6 print. On the other hand, if I were printing to 40x60, I might be more concerned about it shooting at f32 or f45, even in 4x5.
If in doubt, especially if you are shooting digital, take two shots, one at f8-11, one at f16-22 and see what works at your intended print size.
Have a look at Lloyd Chambers' study, the degradation is clear.
http://www.diglloyd.com/diglloyd/free/Diffraction/example-1DsM3.html
First of all, I believe Moshen shoots medium format, so he may fall into some range between 35mm and large format.
There is no doubt that f22 has a significant diffraction impact on 35mm. But I usually find an unintentional out of focus region in image to be significantly more distracting than a slightly soft image. I suspect that you'd find most viewers, including those who might be inclined to purchase a photo feel the same.
I stick by Guy and Dan in saying, if you need f22, then use it and deal with the consequences. If not, then don't.
Mel-
Additionally, there are small diffraction patterns overlaid at every pixel (basically every ray of light, going towards all the directions necessary to light the sensor, is affected by it; diffraction causes that ray (or the photons of which it consists) to spread out a little to the sides, lowering definition of the ray, and therefore lowering resolution, or rather, increasing the distance between two points that can be clearly resolved), not a big one on the whole sensor that "makes light curve away to the corners". Consequently, resolution drops over the whole image, not only at the borders.
For the diffraction pattern created by a single ray to reach over the whole sensor, the aperture would be tiny (f/2000 or so, maybe
Basil
Diffraction is not a penalty that is kind of subtracted from the resolution capability of the lens. It's a upper limit imposed by physics. So, more expensive lenses will be diffraction limited already at f/8 (check photozone.de's resolution measurements for some good Nikon lenses - they drop at f/5.6 or f/8), while cheapos may go up until f/11 because they started at a lower level and can't reach the diffraction limit earlier. But once diffraction limits resolution, all lenses should resolve approximately the same (there are some that handle it better, and some that handle it a bit worse, but the differences are small).
Basil
This simpy does not agree with my experience.
Yes, at some point, diffraction does take over to render everything soft, but you need to go further than that. I have used both consumer grade lenses and pro-level lenses. At f16-22, there is still a difference. Yes, the top end equipment does have image quality falling off faster than the consumer lenses, if only because the consumer lenses start at a lower point. But there is still a difference. f22 is not enough of an equalizer of performance.
And no, I rarely shoot any 35mm-based lens at f22. The results are usually not all that great. That aperature is almost exclusively reserved for situations where form is more important than details.
I suppose if one chooses to view these prints at 6 inches or with a magnifying glass, then some problems will crop up. But in real world viewing situations, it just doesn't seem to matter that much. At least with that lens and when shooting Velvia.
Now with my 12-24 on a D200 or D300, I can see, on the monitor at 100%, the loss of resolution due to diffraction at f/22. F/16 looks a lot better.
I guess what it boils down to is this: take some pictures, evaluate the results. Experiment and see what is acceptable to you.
Thanks, CJ
The reason 100% is useful, though, is because it's showing an actual image pixel using an actual screen pixel. So you should be eliminating any funkiness (technical term
There's been quite a lot of discussion on other sites about both diffraction effects, and noise. Many of them, like you mention, make use of 100% crops to display noise or diffraction blurring. However, as the resolution goes up, and monitor size increases, a 100% crop is becoming a smaller and smaller portion of the actual image, displayed larger and larger (unless screen resolution keeps going up too).
So if you have a 6 megapixel image, and you take a screen sized 100% crop, you've probably got about 1/6 of the image. If you do the same with a 12 megapixel image, you have half that. If the noise/diffraction blurring is worse in the 12 megapixel image it wouldn't be surprising - since it's sort of equivalent to taking a 200% shot of the 6 megapixel image.
On a pixel per pixel basis, you could say which sensor is cleaner - you can see the noise introduced from one pixel to the next based on the 100% view. You could also possibly figure out which pixel size was more or less sensitive to diffraction (well, physics tells you that) and how sensitive they are for that sensor. But in a print are you better off having more noisier pixels, or less cleaner pixels? More slightly diffraction blurred pixels? Or less diffraction blur-less pixels? And what about the AA filter on the sensor? It blurs all digital images intentionally.
Specifically on the diffraction issue, though, think about this : say we have two sensors - the 12 MP D3 sensor, and the 25 MP D3X sensor (assuming that's what it will be). It's true that diffraction will affect the pixels much earlier in the D3X due to their smaller size. However, all that does is reduce the effective resolution (maybe every pixel gets slightly blurred into it's neighbors). So maybe you get an image with as much detail as the 12 MP sensor. And at apertures that don't experience much, or any diffraction, then you can get the real benefit of the 25 MPs. A 100% crop from the D3X at f/22 might look much worse than that from the D3 (diffraction wise) but in a same-sized print, the D3X will have had a lot more pixels to work with.
So say you now print both at 8x10, and 20x30 - now you've finally arrived at more of an apples to apples comparison. I think, in this case, f/22 may be bad enough to give you noticable diffraction blurring for just about any sensor. However, a higher pixel count, while being more susceptible to pixel blurring, will still have more pixels on any given detail that is captured. This may make it better for the larger print. Either way, you'll have something similar to a lower res image of the same scene. So if you're OK with that, and need f/22 for DOF - go for it.
(Hopefully that all made sense... )
Thanks, CJ
Issue 1 : Viewing at 100%
- Can show pixel to pixel differences, but higher mega-pixel sensors show less of the image at 100% possibly giving apples to oranges comparison
- Print and screen viewing quality will depend on total number of pixels and print size, as it always has - 100% viewing may not tell you much
- So packing more pixels into a same sized sensor may appear to be worse when viewed at 100%, but probably won't show up in a print
Issue 2 : Photos at small apertures
- More pixels will give better resolution at apertures where diffraction is not noticable
- Even at smaller apertures, where diffraction is noticable, this is an effective lowering of the resolution, since some pixels are combined by the diffraction blurring
- Since the resolution is just lowered, you should theoretically get similar results to using a sensor with less pixels where diffraction is not an issue
- So as far as diffraction is concerned, packing more pixels onto a same sized sensor shouldn't really be worse than having fewer, larger pixels
The one area where packing more pixels into a given sensor size is almost always bad is noise performance...otherwise, it seems like shrinking pixels aren't that big a deal.
(I know I'm just repeating a lot, so please stop me if I'm getting out of hand...