Bruce n Philly wrote:
Wow.... great stuff..... but.... I didn't read about 100% crops. What then?
I am not talking about comparing an equal crops that pull in an eye. I am talking about 100% crop of a 5D MkX vs a 100% crop of a 5DS. One produces an eye, the other a tear duct.
Why care about this inequality? Bird photography. If you do it, you know what I mean. You will try and crop down to 100% to pull in that rare feathered dinosaur. Let's assume your technique is perfect.
All things being equal and perfect technique, wouldn't 100% crops turn blurry sooner as you stop down your lens on a higher pixel count sensor? Therefore, you do need to care. Again, I am talking about 100% crops. No?
AJSJones wrote:
The smaller the pixel the sooner the Airy disk will exceed it in size and the softening will, under some circumstance such as described above, be perceptible. That's all the table indicates.
That is not what the table indicates because it does not account for the lens. You cannot distinguish between blur from lens aberrations and diffraction softening.
That's all I've been saying. Not rocket surgery: At some aperture, the blur created by diffraction will be bigger than the smallest detail the sensor can resolve...
The table gives the aperture where the diameter of the Airy Disk exceeds an arbitrary figure, such as the diagonal of one or two pixels. However, the spread of light due to diffraction is spread over a circle that exceeds the dimensions of the sensor. The Airy Disk contains about 88% of the light from a point source. So, for a perfect (aberration free) lens, the diffraction limited aperture would be wide open for any pixel size. For a non-perfect lens, one must account for the lens aberrations to determine the diffraction limited aperture.
Using the the diagonal of one or two pixels as a criterion for the "diffraction limited aperture" does not tell you the sharpest aperture with either a perfect or actual lens. It tells you the aperture where the blur due to diffraction is on the same order as the blur due to large pixels. But a perfect lens will still be sharpest wide open (for the portions of the scene within the DOF) and an imperfect lens will be sharpest stopped down to some point which must be found be testing the lens on a particular sensor to account for pixel size and, primarily, lens aberrations.
...and it is one factor limiting detail capture and may be a limiting factor in printing. The contentious issue seemed to be what that aperture is calledThat's why I posted.
The "diffraction limited aperture" as you, and other sites, use the term is meaningless with regards to the visual properties of the photo.
snapsy wrote:
There is a bit of a perverse relationship between higher MP cameras, DOF, and diffraction, at least when wanting to utilize the full resolution benefit offered vs lower MP cameras. Higher MP means a smaller CoC, which translates into a lower perceptual DOF when utilizing the higher MP ability for larger prints. But higher MP means diffraction becomes observable at larger apertures, again when utilizing the higher MP to print larger. This reduces the envelope by which the photographer can maximize both resolution and DOF.
No. Diffraction does not increase or become more "observable" at some apertures on a higher MP system.
The amount if diffraction is exactly, precisely identical, whether you shoot a 12MP full frame camera or a 100MP full frame camera.
The only difference is that with the right lenses you may be able to shoot at a slightly larger aperture and take advantage of potentially higher resolution than you got from the lower resolution sensor.
gdanmitchell wrote: snapsy wroteThere is a bit of a perverse relationship between higher MP cameras, DOF, and diffraction, at least when wanting to utilize the full resolution benefit offered vs lower MP cameras. Higher MP means a smaller CoC, which translates into a lower perceptual DOF when utilizing the higher MP ability for larger prints. But higher MP means diffraction becomes observable at larger apertures, again when utilizing the higher MP to print larger. This reduces the envelope by which the photographer can maximize both resolution and DOF.No. Diffraction does not increase or become more "observable" at some apertures on a higher MP system.
The amount if diffraction is exactly, precisely, identical, whether you shoot a 12MP full frame camera or a 100MP full frame camera.
The only differences is that with the right lenses you may be able to shoot at a slightly larger aperture and take advantage of potentially higher resolution than you got from the lower resolution sensor.
There is no bad news here — only good news.
Dan
Dan you just disagreed with me but then closed out your message with exactly the same thing I wrote regarding diffraction vs aperture.
Fundamentally, the detectability of diffraction in a photographic image depends on the usual suspects, including importantly the print size and viewing distance.
If you print at the same size (say 13 x 19), then the diffraction is independent of sensor size.
But if you print large enough, just below large enough were individual pixels can be distinguished, then you will see more diffraction from the smaller pixel pitch ("denser" sensor) or equivalently you will see diffraction begin to have a visible effect at a smaller f/stop (larger aperture). Of course when you do this, you are also printing larger but viewing from the same distance
However, if you view the larger print from, say, a 5Ds at a proportionally larger distance than you view a lesser sized print from a 5D2, for example, then the diffraction will be equally apparent or not apparent and the f/stop that it becomes apparent at will be the same in both cases.
This is because the viewing distance negates the print enlargement (remember both start from a 36 x 24 mm full-frame sensor). In both cases, the diffraction blur covers the same size on the sensor and cover the same subtended arc as viewed by the eye looking at the prints.
anscochrome wrote:
Clear asterisk or not, you should not assume that the average user of these sites that name these DLA are knowledgeable Fred Miranda users. People see a number-they see it is lower than their old camera-they panic. Further reading stops.
Perhaps it is best that they don't even publish these numbers, since it apparently has no influence as far as usable, real world results are concerned.
You mean "A little knowledge is a dangerous thing"? As I said, caveat emptor. I make no assumptions about the abilities of the reader of the Digital Image (or the Cambridge in Colour) table. If you pick and choose from the information clearly presented to you and get the wrong idea, whose fault is that?
It is only in rare situations that the limit is relevant and those have been described here. It was the question from the OP, clarified here that illustrates the rare case (large prints at the pixel-limit are the other case). If someone knows they will need to use a 100% crop, perhaps because they were focal length limited, or want to make sure the distant leaves and pebbles in the foreground of a landscape are all equally sharp *and* as sharp as possible, this number is a useful reference point. I'd agree with Anders that in the "real world" going to f/11 with the 5Ds is only rarely, if ever, going to cause problems from diffraction.
As an analog/digital anecdote, I learned the hard way that the combination of my Nikon 210 lens on my 4x5 camera and Provia100 scanned on my Agfa2500 and printed on my Epson 7600 at 24x30 would be enough softer when shot at f/45 than at f/32or f/22 to be a disappointment. The desired DoF was there but the overall softening was very evident. A real-world case of balancing DoF with diffraction - my "sensor" was capable of better - and it did much better after I learned the limit
takowasa wrote:
The table ... tells you the aperture where the blur due to diffraction is on the same order as the blur due to large pixels.
Yes.
takowasa wrote:
The "diffraction limited aperture" as you, and other sites, use the term is meaningless with regards to the visual properties of the photo.
dhphoto wrote:
And Circle of Confusion, but I suppose that's sort of covered by visual acuity.
.
The CoC is a reduced parameter, the part that is based on {sensor size to print size ratio, viewing distance, and visual acuity}.
The image below is magnified crops of images of the same target taken with the same lens at 6 different apertures, each one stop narrower, from left to right.
The upper half is from a 36 MP Full Frame sensor, interpolated to 360 MP.
The lower half is 100% crops from a sensor with pixel density corresponding to 360 P Full Frame,
Does anybody want to guess the apertures, and does anybody see a particular diffraction limited aperture?
Monito wrote:
Fundamentally, the detectability of diffraction in a photographic image depends on the usual suspects, including importantly the print size and viewing distance.
If you print at the same size (say 13 x 19), then the diffraction is independent of sensor size.
But if you print large enough, just below large enough were individual pixels can be distinguished, then you will see more diffraction from the smaller pixel pitch ("denser" sensor) or equivalently you will see diffraction begin to have a visible effect at a smaller f/stop (larger aperture). Of course when you do this, you are also printing larger but viewing from the same distance
However, if you view the larger print from, say, a 5Ds at a proportionally larger distance than you view a lesser sized print from a 5D2, for example, then the diffraction will be equally apparent or not apparent and the f/stop that it becomes apparent at will be the same in both cases.
This is because the viewing distance negates the print enlargement (remember both start from a 36 x 24 mm full-frame sensor). In both cases, the diffraction blur covers the same size on the sensor and cover the same subtended arc as viewed by the eye looking at the prints. ...Show more →
Agreed. Where these types of discussions run off the rails is the inordinate focus about how at normalized/equalized print and viewing sizes things like resolution, DOF, noise are the same. I don't understand why that point gets more attention than what the differences are when actually using the higher resolution of the cameras, ie print larger, crop, etc.., since that is purportedly the primary reason one would buy a a 50MP camera to begin with. It would be like buying a Ferrari but then focusing on how it's no faster than a Camry when driven on ice
takowasa wrote: Sneakyracer wroteThe Digital Picture posts the "DLA" (Diffraction Limited Aperture) in their camera reviews. Here is a table they made:The table, of course, is in error. As I noted above, diffraction softening is there right from wide open and increase as you stop down. Lens aberrations exist right from wide open and lessen as you stop down. The point where increasing diffraction outweighs lessening lens aberrations is called the "diffraction limited aperture" and usually lags a stop later in the edges than the center. Thus, the diffraction limited aperture is a function of the lens and where in the frame you are looking, not the sensor size.
That said, the size of the Airy Disk (the disk over which a point source of light is spread, which is what results in diffraction softening) will be the same on all systems for the same DOF....Show more →
I mostly agree with your posts, except for calling such table "an error". You complain about Digital Picture taking the lens aberrations out of the ecuation, but just for the same reason you can not take out the sensor (pixel size). Your definition is flawed because it requires a sensor with an infinite amount of pixels. Since all sensors in the real life are limited in resolution, they "blur" the pictures just like lens aberrations. The pixel size does sets the practical aperture from where diffraction must be, in practice, be considered to start impacting the image. For example, a 320x240 pixels full frame sensor with an excelent lens will likely allow "sharp" F42 pictures (such "pic-thumbnails" won't be a bit less sharp in the center of the image compared to having used F8 or F5.6 instead of F42). But it would be a pity to use the 5DS at F42. That is what the chart is trying to mean (taking its accuracy apart).
If one buys a high MP camera, likely he/she has the intention of getting the highest possible resolution. In order to achieve this, he needs to use a bigger aperture compared to the lower resolution sensors... there is nothing wrong on this assertion. And in fact it can be exactly computed (when the amout of pixels is doubled, diffraction "starts to matter" an f/stop early, in order to "get the absolute best sharpness" of the new sensor).
To summarize, maybe we all likely agree that the following points are true, comparing high resolution sensors with lower resolution ones of the same size:
The higher the sensor resolution, the higher the image resolution at the same aperture (but progressively with diminishing returns, due to diffraction).
Diffraction exists in the whole aperture range, but the effect may be masked by lens aberrations and/or sensor pixel size.
When DOF doesn't matters, there is an ideal aperture, which depends on the lens quality (including the image area center/border), sensor pixel size and diffraction. Such aperture is wider in higher resolution sensors with most current great lenses.
The ideal aperture for a lens/sensor combination is a bit subjective. Tables may approximate the most sharp aperture, but there is nothing like oneself comparing several pictures from a sturdy tripod, in rapid alternance in the monitor to perceive the differences.
Colors matter for diffraction: this is the only non subjective field, thanks to physics (if green areas have some diffraction softness at F8, the same softness is achieved on red areas at about F5.6, and in blue areas at F11: two whole stops from a red-cast area compared to a blue-cast one).
When DOF matters, there is nothing wrong stopping down a high resolution sensor. The picture will always be better, slightly better or at least equal compared to a lower resolution counterpart (being AA filters and sensor tech the same). The improvement mostly depends on the diffraction level... always taking into account the pixel size (at F16, the 5DS likely won't hugely improve the 5D3, but the 5D2 did improved the 5D).
A sensor with a weaker AA filter will resolve more real detail and at the same time introduce more invented detail (a source of potential aliasing), both mostly outside the "diminishing returns" area.
takowasa wrote:
The table, of course, is in error. As I noted above, diffraction softening is there right from wide open and increase as you stop down. Lens aberrations exist right from wide open and lessen as you stop down. The point where increasing diffraction outweighs lessening lens aberrations is called the "diffraction limited aperture" and usually lags a stop later in the edges than the center. Thus, the diffraction limited aperture is a function of the lens and where in the frame you are looking, not the sensor size.
That said, the size of the Airy Disk (the disk over which a point source of light is spread, which is what results in diffraction softening) will be the same on all systems for the same DOF....Show more →
cgarcia wrote:
I mostly agree with your posts, except for calling such table "an error". You complain about Digital Picture taking the lens aberrations out of the ecuation, but just for the same reason you can not take out the sensor (pixel size).
Pixel size does have an effect, but it is minor. I gave an example earlier where the diffraction limited aperture for an 8 MP sensor and a 15 MP sensor were essentially the same. In lensrental's writeup on diffraction, we can see that the diffraction limited aperture shifts about half a stop wider as one goes from 12 MP to 36 MP. In other words, the effect of pixel size on the diffraction limited aperture is negligible with respect to lens aberrations unless the difference in pixel sizes is huge, and even then, lens aberrations are still dominant.
Your definition is flawed because it requires a sensor with an infinite amount of pixels. Since all sensors in the real life are limited in resolution, they "blur" the pictures just like lens aberrations. The pixel size does sets the practical aperture from where diffraction must be, in practice, be considered to start impacting the image. For example, a 320x240 pixels full frame sensor with an excelent lens will likely allow "sharp" F42 pictures (such "pic-thumbnails" won't be a bit less sharp in the center of the image compared to having used F8 or F5.6 instead of F42). But it would be a pity to use the 5DS at F42. That is what the chart is trying to mean (taking its accuracy apart)....Show more →
As I said, and demonstrated, the lens aberrations play a greater role than pixel size when it comes to the diffraction limited aperture.
If one buys a high MP camera, likely he/she has the intention of getting the highest possible resolution. In order to achieve this, he needs to use a bigger aperture compared to the lower resolution sensors... there is nothing wrong on this assertion. And in fact it can be exactly computed (when the amout of pixels is doubled, diffraction "starts to matter" an f/stop early, in order to "get the absolute best sharpness" of the new sensor).
As demonstrated above, from 8 MP to 15 MP, it the difference is at most 1/3 of a stop. From 12 MP to 36 MP, the difference is around 1/2 a stop. The lens aberrations matter more, by far, than the pixel size.
To summarize, maybe we all likely agree that the following points are true, comparing high resolution sensors with lower resolution ones of the same size:
The higher the sensor resolution, the higher the image resolution at the same aperture (but progressively with diminishing returns, due to diffraction).
Diffraction exists in the whole aperture range, but the effect may be masked by lens aberrations and/or sensor pixel size.
When DOF doesn't matters, there is an ideal aperture, which depends on the lens quality (including the image area center/border), sensor pixel size and diffraction. Such aperture is wider in higher resolution sensors with most current great lenses.
The ideal aperture for a lens/sensor combination is a bit subjective. Tables may approximate the most sharp aperture, but there is nothing like oneself comparing several pictures from a sturdy tripod, in rapid alternance in the monitor to perceive the differences.
Colors matter for diffraction: this is the only non subjective field, thanks to physics (if green areas have some diffraction softness at F8, the same softness is achieved on red areas at about F5.6, and in blue areas at F11: two whole stops from a red-cast area compared to a blue-cast one).
When DOF matters, there is nothing wrong stopping down a high resolution sensor. The picture will always be better, slightly better or at least equal compared to a lower resolution counterpart (being AA filters and sensor tech the same). The improvement mostly depends on the diffraction level... always taking into account the pixel size (at F16, the 5DS likely won't hugely improve the 5D3, but the 5D2 did improved the 5D).
A sensor with a weaker AA filter will resolve more real detail and at the same time introduce more invented detail (a source of potential aliasing), both mostly outside the "diminishing returns" area.
I agree with all of the above except for the tables based on pixel size alone being a guide to the diffraction limited aperture -- we absolutely need to account for the lens aberrations where the pixel size is a relatively small adjustment from that value.
Also, there is something wrong with stopping down: not only does that entail more diffraction, but it may well result in greater motion blur / camera shake when a longer exposure time is used to compensate for the more narrow aperture or will result in more noise if the same exposure time is used. In cases where resolution is paramount, the photographer may be well served to consider the very inconvenient technique of focus stacking (which, of course, has problems of its own). That said, for sure, when more DOF is important in a single exposure, stopping down should not be discounted on the basis of diffraction and/or noise. In fact, the effects of diffraction softening are, for the most part, very much overblown.
snapsy wrote: gdanmitchell wrotesnapsy wroteThere is a bit of a perverse relationship between higher MP cameras, DOF, and diffraction, at least when wanting to utilize the full resolution benefit offered vs lower MP cameras. Higher MP means a smaller CoC, which translates into a lower perceptual DOF when utilizing the higher MP ability for larger prints. But higher MP means diffraction becomes observable at larger apertures, again when utilizing the higher MP to print larger. This reduces the envelope by which the photographer can maximize both resolution and DOF.No. Diffraction does not increase or become more "observable" at some apertures on a higher MP system.
The amount if diffraction is exactly, precisely, identical, whether you shoot a 12MP full frame camera or a 100MP full frame camera.
The only differences is that with the right lenses you may be able to shoot at a slightly larger aperture and take advantage of potentially higher resolution than you got from the lower resolution sensor.
Just checking... ;-)
There is no bad news here — only good news.
DanDan you just disagreed with me but then closed out your message with exactly the same thing I wrote regarding diffraction vs aperture.
alundeb wrote:
The image below is magnified crops of images of the same target taken with the same lens at 6 different apertures, each one stop narrower, from left to right.
The upper half is from a 36 MP Full Frame sensor, interpolated to 360 MP.
The lower half is 100% crops from a sensor with pixel density corresponding to 360 P Full Frame,
Does anybody want to guess the apertures, and does anybody see a particular diffraction limited aperture?
The full frame image in the upper half only gains resolution from F16 to F8 (3 crops from the right). Beyond that, the lines continue extinguishing the same, about 1/3 above the center (unless we call "lines" to moire pointing to the wrong direction...). They look more "contrasty", though.
The high pixel density sensor in the lower half continuously gains resolution all the way from the right to the left (this is particularly noticeable looking the lines width horizontally, in their path to the center).
By the way, a 360MP full frame is a mobile phone resolution... with the same lens of a DSLR...?
PS: edited to change F22 for F16 (I counted wrongly :-)
A discussion of diffraction in terms of resolution is limited. Diffraction affects all spatial frequencies, and you lose contrast at larger scales upon stopping down, as clearly illustrated by the experiment of alundeb, There is no pixel density or sharpening method that will undo this damage.
alundeb wrote:
The image below is magnified crops of images of the same target taken with the same lens at 6 different apertures, each one stop narrower, from left to right.
The upper half is from a 36 MP Full Frame sensor, interpolated to 360 MP.
The lower half is 100% crops from a sensor with pixel density corresponding to 360 P Full Frame,
Does anybody want to guess the apertures, and does anybody see a particular diffraction limited aperture?
What limits the resolution on the smaller sensor with the higher pixel density with that lens, diffraction or LP/mm resolving ability of that specific optic?
snapsy wrote:
What limits the resolution on the smaller sensor with the higher pixel density with that lens, diffraction or LP/mm resolving ability of that specific optic?
Do we actually know, at the wider apertures? Isn't the sensor still the limiting factor at f/4 anyway?
The full frame image in the upper half only gains resolution from F16 to F8 (3 crops from the right). Beyond that, the lines continue extinguishing the same, about 1/3 above the center (unless we call "lines" to moire pointing to the wrong direction...). They look more "contrasty", though.
The high pixel density sensor in the lower half continuously gains resolution all the way from the right to the left (this is particularly noticeable looking the lines width horizontally, in their path to the center).
By the way, a 360MP full frame is a mobile phone resolution... with the same lens of a DSLR...?
PS: edited to change F22 for F16 (I counted wrongly :-) ...Show more →
Thanks for your interpretation! The apertures are actually F/4 to F/22 and the lens is the Zeiss Distagon 15 mm F/2.8. The reason I didn't include f/2.8, was that when I made this, it was with the purpose of checking the effectiveness of diffraction as an Anti-Aliasing filter at various apertures and pixel densities. With this, I can interpolate the results to any pixel density that is going to interest me for some time.
If you allow me, I will turn your analysis around and say that the upper half doesn't lose any true, non-aliased resolution until past F/11. That is the A7r. The lower half is the Pentax Q10.
gdanmitchell wrote:
Yet again (he drones on) the news is somewhere between neutral and good for you.
There is no downside.
Dan
Thanx Dan, but edumacate me (as if I can't read!)
Go with me on this.....
You take a picture black on the left, white on the right and an infinity perfect line/transition between.
You have a 5D Mk II...... you crop a pic to 100%..... diffraction sets in at f/11.... You look at the pixels along the b/w line and there is some diffraction blurring. Let's say for arguments sake that the "transition" pixels are 10 pixels wide.
Now swap the body for a 5Ds... because you have more/density pixels, you now crop again down to 100%..... but you are now way deeper. The blurry zone covers more pixel, now 20 pioxels, than the other camera. More pixels are in the blur zone... therefore the need to open up a stop to get the same number of pixels in the blurr zone back to 10.
Remember, my objective is to crop a small part of the pic down to 100%