Higher Megapixels, DOF and Diffraction..

[Moreorless]

I know some of the basics behind these concepts but with the latest round of the megapixel war seeming like its kicking off I wonder how it relates to the maxmimum limates of different sensors.

Do DOF(for an equivalent field of view rather than the same focal lenght) and diffraction work at the same level meaning that effective results are similar at similar pixel densities no matter the sensor size. Basically are different sensor sizes going to reach a maximum pixel density at the same point after which diffraction holds them back?

 

[squishy]

The aperture at which the negative effects of diffraction start to be seen at 100% will get larger as the pixel density increases.
Simply because the extra sensor resolution can show up limitations that lower resolution sensors can't.

For example, I expect if I shot a frame at f/36 on my 40D (10mp) and then the same frame at f/36 on a 7D (18mp), there would be little or no extra visible detail in the shot from the 7D (whereas at larger apertures with decent lenses I would expect more detail).

 

[woof woof]

I don't know if diffraction is a real world worry, and anyway, surely the best option is to have a sensor which is capable of making diffraction visible at all apertures?

At the moment with the stuff that's available today you can see both the particle and wave nature of light but at normal print/image sizes and viewing distances it shouldn't be a problem for image quality unless taken to extreme apertures at which you will probably notice a lack of sharpness and contrast but post capture processing should probably be able to make even these things more presentable.

I have no idea how far current Bayer type sensor technology can go and maybe we'll never know as it's always possible that another technology could emerge which would not make use of pixels and pixel counts would therefore be irrelevant.

 

[Moreorless]

I don't know if diffraction is a real world worry, and anyway, surely the best option is to have a sensor which is capable of making diffraction visible at all apertures?.

The benefits of such a sensor though would surely start to decline as difftraction kicks in more and more.

My point was really is this going to be the factor that means large sensor sizes are here to stay? ISO, DR etc tech may well improve to the stage where a tiny sensor can provide everything thats needed but wouldnt difftraction create a maximum limate on resolving power relatively to size?

 

[HoppyUK]

I know some of the basics behind these concepts but with the latest round of the megapixel war seeming like its kicking off I wonder how it relates to the maxmimum limates of different sensors.

Do DOF(for an equivalent field of view rather than the same focal lenght) and diffraction work at the same level meaning that effective results are similar at similar pixel densities no matter the sensor size. Basically are different sensor sizes going to reach a maximum pixel density at the same point after which diffraction holds them back?

You could write a book on this, they're all separate issues, but the simple answer is don't worry about it.

The concept of DoF is unrelated to the sensor or lens in theory, and in practise. It is based on the smallest level of detail visible to the human eye, and that is the practical limit because even quite basic systems are capable of comortably exceeding that.

Diffraction applies to all optical systems at all times, but as far as we're concerned, it starts at the point where the improvements in sharpness made by raising the f/number (which reduces most aberrations) is overtaken by diffraction which gets worse at higher f/numbers. In practise, it's also affected by format, and it hits earlier with smaller formats. Hence most lenses reach their sharpness peak at around f/5.6 on crop format, and around f/8 on full frame (most compacts don't go higher than f/5.6 because it gets so bad). The very best lenses peak at slightly lower f/numbers, maybe a stop lower, and at a correspondingly higher level.

There is a theory that says higher pixel densities are more prone to diffraction, which is a clash of two different aspects of the system when the circle of confusion coincides, but I have never seen any pictorial evidence of this that has any practical significance.

Lens 'sharpness' is a combination of two factors, resolution and contrast, that move oppositely and in tandem (basis of MTF). Lenses don't just stop delivering resolution, but as it increases so contrast drops. For example, if you shoot a resolution target of black/white lines, at low resolution they will the very clear but as resolution increases so the black lines become lighter and the white lines become darker until they are both the same shade of grey - 0% MTF (50% MTF is a commonly used reference standard).

Pixels are a bit of a red herring. While they get talked about a lot, they are just part of the equation. For example, compare a Canon 7D (18mp, crop format) with a Nikon D700 (13mp, full frame) and the Nikon wins easily, by miles - better sharpness, cleaner colours, less noise, everything. It might have less pixels, but it's twice the size and the lens isn't working nearly as hard and the MTF is higher. And so it goes on

 

[woof woof]

The benefits of such a sensor though would surely start to decline as difftraction kicks in more and more.

No, because the image would have the potential to be huge and you'd have to view either a huge image very closely or at massive zoom on your screen before diffraction became visible at reasonable apertures.

 

[HoppyUK]

This is an interesting test, showing how diffraction hits, Canon 50 1.8 on DPReview. This is the lens on crop format, which peaks around f/5.6. http://www.dpreview.com/lensreviews/canon_50_1p8_ii_c16/page3.asp But notice how it affects the centre first, because that's at a higher level, and there's a point where the edges are still getting sharper while the centre declines. Then the two reach parity, the line flattens and then declines as a straight lines right across the frame - that's pure diffraction.

Same thing on full frame, same lens peaking around f/8 http://www.dpreview.com/lensreviews/canon_50_1p8_ii_c16/page4.asp

 

[woof woof]

Diffraction... In practise, it's also affected by format, and it hits earlier with smaller formats.

In practice maybe but it's not the size of the sensor as such but how format size leads to other choices.

Diffraction is interference. Light has a duality and can act, or appear to act, as both a particle and a wave and when acting as a wave one wave can interfere with another just as it would on water. I don't think that this is anything to do with the format size as such but it will be affected by the aperture which could well be affected by the focal length of the lens which could well be influenced by the format size. The reason compacts show diffraction more is more to do with high pixel counts and the size of the actual hole that the light goes through than the actual size of the chip.

 

[HoppyUK]

In practice maybe but it's not the size of the sensor as such but how format size leads to other choices.

Diffraction is interference. Light has a duality and can act, or appear to act, as both a particle and a wave and when acting as a wave one wave can interfere with another just as it would on water. I don't think that this is anything to do with the format size as such but it will be affected by the aperture which could well be affected by the focal length of the lens which could well be influenced by the format size. The reason compacts show diffraction more is more to do with high pixel counts and the size of the actual hole that the light goes through than the actual size of the chip.

Yes, in practise. But while these affects are all separate in theory, they are all part of one imaging system and they all inevitably impact on eachother.

 

[HoppyUK]

In practice maybe but it's not the size of the sensor as such but how format size leads to other choices.

Diffraction is interference. Light has a duality and can act, or appear to act, as both a particle and a wave and when acting as a wave one wave can interfere with another just as it would on water. I don't think that this is anything to do with the format size as such but it will be affected by the aperture which could well be affected by the focal length of the lens which could well be influenced by the format size. The reason compacts show diffraction more is more to do with high pixel counts and the size of the actual hole that the light goes through than the actual size of the chip.

Your point about compacts - I'm not at all convinced about pixel counts having any meaningful impact and the link to DPReview and the 50 1.8 test on two formats shows the effect that format alone has - same lens, same f/number, but the diffraction point shifts by the crop factor.

And while this theory exists that high pixels counts impact diffraction effects, the fact remain that, whenever we get a new camera with more pixels, if the lens is good enough you can extract more sharpness from it, eg birders with 7D and a tasty L prime can crop harder (several on here do just that). And we have compacts with, relatively, astonishingly high pixel density - say a 14mp compact equates to around 350mp in full frame terms (25x the sensor area).

On the other hand, if you want sharp, the basic rule is start with a big image. The Nikon D700 with a mere 13mp beats any smaller format, easily and absolutely even when they have more pixels, because everything else is upscaled - the whole system is bigger, and that means better in this context.

And there's another point that rarely gets mentioned, which is the anti-aliasing filter which sets the practical limit on sharpness anyway, regardless of pixels. Manufacturers don't like to talk much about that and as far as I know, there are no published figures on anti-aliasing and the 'real world' Nyquist ceiling.

It all gets horribly complicated

 

[Moreorless]

No, because the image would have the potential to be huge and you'd have to view either a huge image very closely or at massive zoom on your screen before diffraction became visible at reasonable apertures.

Surely though thats the point of having those megapixels(or whatever replaces them), in order to be able to zoom in a great deal or to create large prints that retain high levels of detail close up. Obviously not something that everyone is going to want/need but it surely its always likely to be desired/needed by some.

What are reasonabley appatures aswell? I often find myself fighting between what I assume is diffraction and DOF when taking macro and landscape shots.

 

[woof woof]

What I meant was that you'd have to look too closely or zoom in too much to see the problem. Probably

By reasonable apertures I meant nothing too extreme and by extreme I mean something like f22+ but even then it's worth doing if the + outweigh the -. Personally I'm not too worried about diffraction.

 

[woof woof]

Your point about compacts - I'm not at all convinced about pixel counts having any meaningful impact and the link to DPReview and the 50 1.8 test on two formats shows the effect that format alone has - same lens, same f/number, but the diffraction point shifts by the crop factor.

And while this theory exists that high pixels counts impact diffraction effects, the fact remain that, whenever we get a new camera with more pixels, if the lens is good enough you can extract more sharpness from it, eg birders with 7D and a tasty L prime can crop harder (several on here do just that). And we have compacts with, relatively, astonishingly high pixel density - say a 14mp compact equates to around 350mp in full frame terms (25x the sensor area).

At the mo I don't believe that the mp count actually affect diffraction but obviously a sensor capable of high definition will make diffraction easier to spot and I assume that the smaller the format the more tightly packed the pixels are, mp counts being what they are.

I haven't read anything about this on DPR but I just can't see format size affecting diffraction as diffraction is a property of light passing through a hole not what is beyond the hole. However, it stands to reason that a higher mp sensor or a more crowded sensor will demonstrate diffraction at wider apertures simply because it's a higher relative resolution. I remain to be convinced that two different sized sensors with the same resolution would demonstrate diffraction differently.

Smaller formats like compacts have another problem and it's the lenses. AFAIK and I'm sure someone will correct me if I'm wrong, the aperture number (eg f2) is derived from the focal length and the size of the hole, hole size in mm equalling the focal length divided by the aperture (f number)? Therefore (and to make things easy for those of us who can't be bothered to do complex sums...) on FF a focal length of 50mm with f2 = a hole with a diameter of 25mm, but a compact wont have a 50mm lens, it'll have something like a 6mm lens and 6mm at f2 = a hole with a diameter of 3mm. That's a lot smaller than 25mm and I assume that the 3mm hole at f2 on a compact will give you more diffraction than a 25mm hole at f2 on FF.

So with compacts you have the perfect storm.... high pixel counts making diffraction easier to spot and a tiny hole making diffraction more likely. Possibly. Until someone tells me differently.

PS. I forgot to mention that the distance from the aperture to the surface of the film or sensor will matter but I honestly don't know how this will affect the argument of small format v large format.

 

[HoppyUK]

At the mo I don't believe that the mp count actually affect diffraction but obviously a sensor capable of high definition will make diffraction easier to spot and I assume that the smaller the format the more tightly packed the pixels are, mp counts being what they are.

I haven't read anything about this on DPR but I just can't see format size affecting diffraction as diffraction is a property of light passing through a hole not what is beyond the hole. However, it stands to reason that a higher mp sensor or a more crowded sensor will demonstrate diffraction at wider apertures simply because it's a higher relative resolution. I remain to be convinced that two different sized sensors with the same resolution would demonstrate diffraction differently.

Smaller formats like compacts have another problem and it's the lenses. AFAIK and I'm sure someone will correct me if I'm wrong, the aperture number (eg f2) is derived from the focal length and the size of the hole, hole size in mm equalling the focal length divided by the aperture (f number)? Therefore (and to make things easy for those of us who can't be bothered to do complex sums...) on FF a focal length of 50mm with f2 = a hole with a diameter of 25mm, but a compact wont have a 50mm lens, it'll have something like a 6mm lens and 6mm at f2 = a hole with a diameter of 3mm. That's a lot smaller than 25mm and I assume that the 3mm hole at f2 on a compact will give you more diffraction than a 25mm hole at f2 on FF.

So with compacts you have the perfect storm.... high pixel counts making diffraction easier to spot and a tiny hole making diffraction more likely. Possibly. Until someone tells me differently.

PS. I forgot to mention that the distance from the aperture to the surface of the film or sensor will matter but I honestly don't know how this will affect the argument of small format v large format.

Alan, what you appear to be doing is taking a point of view and then looking for evidence to support that theory, which can be misleading. An alternative method would be to observe what is actually happening, and then find out why.

When you say you haven't seen anything on DPReview, what about the links I posted above? Did you look at them? If you move the aperture slider there, you will see what happens on full frame and then on crop format, and note the f/numbers. The two cameras used, 450D (12mp) and 1Ds3 (21mp) are as close as you will get to the same pixel density in a real world example where you can fit the same lens to both.

Yet the differences are plain enough - everything is the same (or very similar) except one sensor is twice the size. As evidence goes, it doesn't get much better than that.

 

[GardenersHelper]

Do DOF(for an equivalent field of view rather than the same focal lenght) and diffraction work at the same level meaning that effective results are similar at similar pixel densities no matter the sensor size.

What are reasonabley appatures aswell? I often find myself fighting between what I assume is diffraction and DOF when taking macro and landscape shots.

For close-ups I use Raynox 250, Raynox 150 and Canon 500D close-up lenses on a Canon SX10is, which has a sensor 6mm across, and a Panasonic G3, with has a sensor 18mm across.

I like to get the most DOF I can for my close-ups so I use small apertures. The smallest aperture on the SX10 is f/8. The smallest aperture on the G3 is f/22. I most often use these apertures for close-ups.

Using these small apertures I lose sharpness because of diffraction, but for my purposes this loss of sharpness seems tiny compared to the big gain in DOF. As an example, these five images show the same subject using the Raynox 250 on the 45-200mm lens on the G3, with apertures of f/5.6 to f/22.

For an equivalent field of view, the SX10 and the G3 are about 3 stops apart in terms of DOF. For example, for a particular scene framed with the same field of view with the SX10 and the G3, I get roughly the same DOF using f/8 on the SX10 as I get using f/22 on the G3. This is true whether or not I am using a close-up lens. The Depth of Field Calculator here illustrates this relationship. It says that:

For a G2 (it doesn't have the G3, but the sensor size is the same), with the subject 10ft away, using f/22 and a focal length of 200mm, the depth of field is 0.48 ft.

(For the G2, a focal length of 200mm is equivalent to 400mm in 35mm terms. For the SX10, a focal length of 71.5 mm is equivalent to 400mm in 35mm terms.)

For an SX10, with the subject 10ft away, using f/8 and a focal length of 71.5mm, the depth of field is 0.47 ft.

Therefore, for an equivalent field of view, the SX10 has the same DOF as the G3 when its aperture is three stops larger than the G3.

(There is another depth of field calculator here that I sometimes find useful.)


....................................

The Diffraction Limit Calculator on this page says that a camera with a micro four thirds sensor like the G3 becomes diffraction limited between f/11 and f/16.

This calculator does not have a precise match for the 1/2.3" sensor of the SX10, so I have used the nearest it does have, 1/2". It says that a camera with a 1/2" sensor becomes diffraction limited between f/4 and f/5.6.

Therefore, the SX10 becomes diffraction limited when its aperture is about three stops larger than the G3's aperture.

...................................

I believe the number of pixels on the sensors has not been a factor in any of these calculations. The DOF calculator I used might have known the number of pixels, but I have got similar results from the other DOF calculator I linked to, and that does not know about the number of pixels. The Diffraction Limit Calculator does not know about the number of pixels. So, for these calculations, the number of pixels on the sensor is not a factor.

Note though that these calculations relate to sharpness (for DOF) or loss of sharpness (for diffraction) that can be perceived by someone with "standard" eyesight looking at a print from a "standard" distance. For pixel-peeping images at 100% on screen, the number of pixels on the sensor presumably would be a factor.

......................

As to what is are reasonable apertures, I think that depends on several factors, some to do with what you are trying to achieve, and some to do with the shooting environment.

Other things being equal, and assuming you like having sharp images, then the nearer you can get to your lens' "sweet spot" aperture for sharpness, then the sharper the image will be.

However, if DOF is a key factor for your image, you should increase the aperture (to get a narrower DOF) or decrease the aperture (to get more DOF) as appropriate, as long as you are content to live with any loss of sharpness that may result (along with any other weaknesses that your lens may exhibit when away from its sweet spot, such as vignetting).

If light levels are poor, and the ISO is as high as you are prepared to go, then you may need to increase the aperture to get a usably fast shutter speed, or use flash (if practical, which of course it would not be for landscape shots). For close-ups, the loss of DOF from increasing the aperture might spoil an image, depending on your taste. If you use flash as your main light source you will be able to use whatever aperture (and DOF) you want, but there are a number of other issues you need to take into account if using flash.

............................

I think that photography benefits from some understanding of the various factors that play into the look and IQ of an image, and the options for capturing the image, and a practical grasp of the trade-offs between those factors.

 

[HoppyUK]

^^^ Interesting post Nick

 

[GardenersHelper]

^^^ Interesting post Nick

Thanks Richard.

(This whole DOF/diffraction business has been a bit of an obsession for me for a while now as I struggled to work out if/how/when to up/cross-grade.)

 

[Moreorless]

As to what is are reasonable apertures, I think that depends on several factors, some to do with what you are trying to achieve, and some to do with the shooting environment.

Other things being equal, and assuming you like having sharp images, then the nearer you can get to your lens' "sweet spot" aperture for sharpness, then the sharper the image will be.

However, if DOF is a key factor for your image, you should increase the aperture (to get a narrower DOF) or decrease the aperture (to get more DOF) as appropriate, as long as you are content to live with any loss of sharpness that may result (along with any other weaknesses that your lens may exhibit when away from its sweet spot, such as vignetting).

If light levels are poor, and the ISO is as high as you are prepared to go, then you may need to increase the aperture to get a usably fast shutter speed, or use flash (if practical, which of course it would not be for landscape shots). For close-ups, the loss of DOF from increasing the aperture might spoil an image, depending on your taste. If you use flash as your main light source you will be able to use whatever aperture (and DOF) you want, but there are a number of other issues you need to take into account if using flash..

What I'd take to be reasonable appatures are those that might be used regularly by your average user on a given sensor size. Using say f/22 on a 1.6 crop camera seems like it would be "unreasbaley" to me but F/11? I'd expect to use that very commonly for landscape and macro shots and thats well past the difftraction limate going from your table.

 

[GardenersHelper]

What are reasonabley appatures aswell?

What I'd take to be reasonable appatures are those that might be used regularly by your average user on a given sensor size.

Fair enough. Without knowing that this was what you had in mind as a reasonable aperture, I was mistakenly thinking about what aperture might be appropriate for particular photographic opportunities. My bad.

Using say f/22 on a 1.6 crop camera seems like it would be "unreasbaley" to me

Using your definition of reasonable aperture, I would think so too.

but F/11? I'd expect to use that very commonly for landscape and macro shots and thats well past the difftraction limate going from your table.

I don't think so. According to the Diffraction Limit Calculator I linked to, a 1.6 crop camera becomes diffraction limited somewhere between f/13 and f/16, so it wouldn't be diffraction limited at f/11.

 

[HoppyUK]

<snip>

I don't think so. According to the Diffraction Limit Calculator I linked to, a 1.6 crop camera becomes diffraction limited somewhere between f/13 and f/16, so it wouldn't be diffraction limited at f/11.

This is where theory and practise diverge - the calculator gives misleading impression. What is meant by 'diffraction limited' at f/13-16?

Because it is evidently true that in practise, diffraction starts to put a ceiling on optimum sharpness much earlier than that, and it is an undeniable fact that diffraction limits potential sharpness from around f/5.6 on a crop camera, and f/8 on full frame.

Witness the links to DPReview I posted earlier, or if you want to see it in more pictorial rather than graph terms, just about any lens test you like on www.the-digital-picture.com That's for most good quality lenses, and one or two really sharp super-primes start hitting diffraction issues maybe a stop earlier, but at a higher level.

You can see this for yourself easily enough, very easily indeed if your camera has a good LCD. Tripod, mirror lock-up, nice detailed subject, and shoot a series through the aperture range - then scroll through the images on max magnification and see which ones are sharpest.

As for when diffraction reduces sharpness to an unacceptable level, well that's down to you, and often the trade off with greater depth of field is worth it, especially for macro.

My rule of thumb is, on a crop camera, not to go above f/11 (f/16 on FF). Diffraction is certainly taking it's toll at that aperture, but the fall off is slight. You'll be hard pressed to even notice it, even in critical working and big prints.

But after that, bad things start to accellerate. F/16 is noticeably worse, and f/22 is frankly pretty dire if max sharpness is the objective. Though you need to decide that for yourself.

 

[Richard King]

wait till you start shooting zoneplates and pinholes

 

[GardenersHelper]

This is where theory and practise diverge - the calculator gives misleading impression. What is meant by 'diffraction limited' at f/13-16?

According to the page on which the calculator appears, "This calculator decides that the system has become diffraction limited when the diameter of the airy disk exceeds that of the CoC [Circle of Confusion]. For a further explanation on each input setting, please see their use in the flexible depth of field calculator."

I used the default settings for Maximum print dimension, Viewing distance and Eyesight, from which I imagine the CoC used in the calculations is derived.

Because it is evidently true that in practise, diffraction starts to put a ceiling on optimum sharpness much earlier than that, and it is an undeniable fact that diffraction limits potential sharpness from around f/5.6 on a crop camera, and f/8 on full frame.

Witness the links to DPReview I posted earlier, or if you want to see it in more pictorial rather than graph terms,

Indeed so, nice link, thanks. This does indeed clearly show sharpness falling from around f/5.6 on a crop camera, and f/8 on full frame.

The calculator claims to be measuring a different characteristic - the aperture at which the diameter of the airy disk exceeds that of the CoC. It also notes that "These calculations only show when diffraction becomes significant, not necessarily the location of optimum sharpness", and notes that optimum sharpness can occur at apertures larger than what their calculations define as diffraction limited.

Given what it claims to be measuring, the caveats referenced above and other caveats in the text beneath the calculator, I don't feel the calculator gives a misleading impression.

As for when diffraction reduces sharpness to an unacceptable level, well that's down to you, and often the trade off with greater depth of field is worth it, especially for macro.

Absolutely.

My rule of thumb is, on a crop camera, not to go above f/11 (f/16 on FF). Diffraction is certainly taking it's toll at that aperture, but the fall off is slight. You'll be hard pressed to even notice it, even in critical working and big prints. But after that, bad things start to accellerate. F/16 is noticeably worse,

I think that pretty much squares with what the calculator says for a crop camera, where it puts the aperture at which diffraction becomes significant between f/13 and f/16. (The calculator draws its line at just over f/18 for 1.3 crop factor and just over f/22 for FF.)

and f/22 is frankly pretty dire if max sharpness is the objective.

Absolutely.

FWIW, maximum DOF is often my objective, coupled with "sharp enough for my purposes" (generally, 1200 x 900 pixels or so on screen and A4 prints, using no more than 50% or so cropping and usually rather light cropping, and occasionally 16 x 12 prints using at most very minor cropping).

Though you need to decide that for yourself.

 

[Moreorless]

I used the default settings for Maximum print dimension, Viewing distance and Eyesight, from which I imagine the CoC used in the calculations is derived.

The default print size is only 10 inches though, surely the whole point of having those extra megapixels is either printing larger than that or cropping to effectively do so?

You go above 13 inches on a 1.6 crop 18 megapixel sensor is limated at F/11, go up to 19 inches and its limated at F/8.

 

[arad85]

the whole point of having those extra megapixels is either printing larger than that or cropping to effectively do so?

It depends which way you look at it. If you're looking at printing purely in terms of pixels per inch then yes, but you can also look at it in terms of magnification of sensor area. That is, for any camera what you're doing is taking a certain sized image projected on a sensor and blowing it up to the desired print size. Whether that sensor has 8Mpix or 18Mpix in the same area, the general quality of the image should be unchanged when viewing it at a "normal" distance. The CoC for a crop sensor is 0.019mm or 19um (micrometre). Given each sensor in a 7D 4.3um and that for something like a 20D is 6.4um, the image should appear the same as the sensor size is below the CoC in both cases.

Discuss....

 

[GardenersHelper]

The default print size is only 10 inches though, surely the whole point of having those extra megapixels is either printing larger than that or cropping to effectively do so?

Yes (as long as putting more pixels onto the sensor doesn't cause the per pixel quality to fall so far it outweighs the benefits of having more pixels).

However, it isn't clear to me what effect the number of pixels on the sensor has on whether diffraction is an issue or not. I put these numbers ...

You go above 13 inches on a 1.6 crop 18 megapixel sensor is limated at F/11, go up to 19 inches and its limated at F/8.

... into the calculator and got the same answers as you. However, I then changed the number of megapixels to 2, and then to 100, and I still got the same answers. So the number of pixels on the sensor doesn't seem to influence the point where diffraction becomes a problem.

It does depend on how large you want to look at the image -diffraction is more of an issue for larger prints/on-screen sizes. And it also depends on how far away you are when looking at the image - many people tend to look at larger images from further away, which makes diffraction less of an issue.

Here's something else I've just noticed. For a 13" print, viewed at the default 25cm, with a 1.6 crop factor and the "default manufacturer standard" eyesight, the calculator says that diffraction becomes an issue between f/11 and f/13. However, change the eyesight to 20/20 and the calculator says that diffraction becomes an issue between f/2.8 and f/4. That is an enormous 4-stop difference! And I would have thought that 20/20 eyesight was a much more reasonable baseline.

 

[HoppyUK]

You guys are in danger of falling into the pixel trap again. It's no good having loads of pixels if the lens can't resolve them.

It's the resolution vs contrast thing. If you compare two prints, even at A4, one shot on full frame and the other from a cropper, it takes one glance to know which was from full frame, even if the crop sensor had more pixels - because the contrast is higher.

Contrast is arguably more important than resolution in the visual impression of sharpness. I'll try and find an example or link that shows this - I know there's a good one somewhere, on the Canon MTF page maybe. I'll have a look.

Edit: it's not a great example, but check the cat on this Canon pdf introduction to lens MTF - page 209 http://software.canon-europe.com/files/documents/EF_Lens_Work_Book_10_EN.pdf The name of the game is to look at the three images, and the one on the right looks least sharp, compared to the one in the middle which has higher contrast. But actually, the one on the right has higher resolution (check around the eyes).

 

[GardenersHelper]

You guys are in danger of falling into the pixel trap again.

That's entirely possible....

It's no good having loads of pixels if the lens can't resolve them.

I certainly had not considered the lens' resolution characteristics.

It's the resolution vs contrast thing. If you compare two prints, even at A4, one shot on full frame and the other from a cropper, it takes one glance to know which was from full frame, even if the crop sensor had more pixels - because the contrast is higher.

Is the difference in contrast a result of different properties of the sensors, and so is true even if the cameras use the same lens? Or is it a property of the lenses, and presumably therefore assuming that the full frame camera is using a better (as in "produces a higher contrast") lens?

If it is a property of the sensors, how does this relate to the point about lenses' resolution characteristics? (Sorry to be thick about this - I'm a bit puzzled about the flow of the argument between the two paragraphs. Or perhaps they are two independent points?)

Contrast is arguably more important than resolution in the visual impression of sharpness. I'll try and find an example or link that shows this - I know there's a good one somewhere, on the Canon MTF page maybe. I'll have a look.

Edit: it's not a great example, but check the cat on this Canon pdf introduction to lens MTF - page 209 http://software.canon-europe.com/files/documents/EF_Lens_Work_Book_10_EN.pdf The name of the game is to look at the three images, and the one on the right looks least sharp, compared to the one in the middle which has higher contrast. But actually, the one on the right has higher resolution (check around the eyes).

Anyway, the point about contrast is very interesting. I had become aware from my PP that increasing contrast/micro-contrast seems to increase apparent sharpness (or something related to sharpness) - nice to see that it isn't just my imagination! And the linked document is a treasure. Looks like it will be a most enjoyable and highly instructive read. Nice one. Thanks.

 

[HoppyUK]

That's entirely possible....



I certainly had not considered the lens' resolution characteristics.



Is the difference in contrast a result of different properties of the sensors, and so is true even if the cameras use the same lens? Or is it a property of the lenses, and presumably therefore assuming that the full frame camera is using a better (as in "produces a higher contrast") lens?

If it is a property of the sensors, how does this relate to the point about lenses' resolution characteristics? (Sorry to be thick about this - I'm a bit puzzled about the flow of the argument between the two paragraphs. Or perhaps they are two independent points?)



Anyway, the point about contrast is very interesting. I had become aware from my PP that increasing contrast/micro-contrast seems to increase apparent sharpness (or something related to sharpness) - nice to see that it isn't just my imagination! And the linked document is a treasure. Looks like it will be a most enjoyable and highly instructive read. Nice one. Thanks.

There's a lot of ground to cover there Nick! Yes, read all that stuff on the Canon link and you'll get a good idea of what lens sharpness is about. Plenty more on MTF if you google around.

With a bigger sensor, two main things are happening. The sensor collects more light, more of the basic raw material - photons - to do more with. Less noise, cleaner colours etc. Crop format is less than half the area of full frame, 4/3rds is only a quarter.

Secondly, and this is the key one I think, becuase the final image has to be enlarged less, the lens doesn't have to work so hard to deliver the same level of resolution, and so contrast is correspondly higher. There comes a point of course where you just don't need the extra quality, depending on your output size mainly, so what is 'best' often invloves a different set of subjective compromises. Cost mainly.

I'm no expert on macro, but if quality is the aim, then full frame will give you more and then the experts play the diffraction/DoF game with focus stacking. Never done it myself, but I know none of this is easy.

Optical contrast is not the same as digital contrast and sharpening, though they can sometimes look much the same to an extent. But if you have a better/cleaner/purer image to start with, then post processing techniques can make that better still. Digital sharpening for example is synthetic in that it senses light/dark transitions and draws an artifical line around them - makes things look clearer, but doesn't increase actual detail, and if you overdo it, it looks pretty naff.

To repeat, if you want sharper images with cleaner colours, start with a big sensor and a good lens. It was always thus, and always will be (basic physics) but there's no doubt smaller sensors deliver amazing results, that are improving bit by bit at a relatively faster rate than full frame because that's where the development effort is mainly going. Just how much of this elusive/expensive IQ stuff you actually need is down to you

 

[GardenersHelper]

There's a lot of ground to cover there Nick! Yes, read all that stuff on the Canon link and you'll get a good idea of what lens sharpness is about. Plenty more on MTF if you google around.

With a bigger sensor, two main things are happening. The sensor collects more light, more of the basic raw material - photons - to do more with. Less noise, cleaner colours etc. Crop format is less than half the area of full frame, 4/3rds is only a quarter.

Secondly, and this is the key one I think, becuase the final image has to be enlarged less, the lens doesn't have to work so hard to deliver the same level of resolution, and so contrast is correspondly higher. There comes a point of course where you just don't need the extra quality, depending on your output size mainly, so what is 'best' often invloves a different set of subjective compromises. Cost mainly.

I'm no expert on macro, but if quality is the aim, then full frame will give you more and then the experts play the diffraction/DoF game with focus stacking. Never done it myself, but I know none of this is easy.

Optical contrast is not the same as digital contrast and sharpening, though they can sometimes look much the same to an extent. But if you have a better/cleaner/purer image to start with, then post processing techniques can make that better still. Digital sharpening for example is synthetic in that it senses light/dark transitions and draws an artifical line around them - makes things look clearer, but doesn't increase actual detail, and if you overdo it, it looks pretty naff.

To repeat, if you want sharper images with cleaner colours, start with a big sensor and a good lens. It was always thus, and always will be (basic physics) but there's no doubt smaller sensors deliver amazing results, that are improving bit by bit at a relatively faster rate than full frame because that's where the development effort is mainly going. Just how much of this elusive/expensive IQ stuff you actually need is down to you

Interesting, and thought-provoking. Thanks for taking the time on this. I think we are pretty much on the same page so I'm not inclined to muddy this thread with yet more nit-picks. Better to spend the time reading that book I think.

 

[Moreorless]

There's a lot of ground to cover there Nick! Yes, read all that stuff on the Canon link and you'll get a good idea of what lens sharpness is about. Plenty more on MTF if you google around.

With a bigger sensor, two main things are happening. The sensor collects more light, more of the basic raw material - photons - to do more with. Less noise, cleaner colours etc. Crop format is less than half the area of full frame, 4/3rds is only a quarter.

Secondly, and this is the key one I think, becuase the final image has to be enlarged less, the lens doesn't have to work so hard to deliver the same level of resolution, and so contrast is correspondly higher. There comes a point of course where you just don't need the extra quality, depending on your output size mainly, so what is 'best' often invloves a different set of subjective compromises. Cost mainly.

I'm no expert on macro, but if quality is the aim, then full frame will give you more and then the experts play the diffraction/DoF game with focus stacking. Never done it myself, but I know none of this is easy.

Optical contrast is not the same as digital contrast and sharpening, though they can sometimes look much the same to an extent. But if you have a better/cleaner/purer image to start with, then post processing techniques can make that better still. Digital sharpening for example is synthetic in that it senses light/dark transitions and draws an artifical line around them - makes things look clearer, but doesn't increase actual detail, and if you overdo it, it looks pretty naff.

To repeat, if you want sharper images with cleaner colours, start with a big sensor and a good lens. It was always thus, and always will be (basic physics) but there's no doubt smaller sensors deliver amazing results, that are improving bit by bit at a relatively faster rate than full frame because that's where the development effort is mainly going. Just how much of this elusive/expensive IQ stuff you actually need is down to you

There seems to be a difference between those concepts and difftraction though.

A larger sensor will always be ahead of a smaller sensor at the same tech level but at some stage might smaller sensor may provide all the quality thats desired by any user? Diffraction on the other hand seems like it presents more of a fixed barrier creating an upper resolution limate for each sensor size.

Thats really why I posted this thread since it seemed like diffraction could have a big impact on the future of photography meaning larger sensors are here to stay rather than potentially being rendered obsolete years down the line.

 

[Richard King]

There seems to be a difference between those concepts and difftraction though.

A larger sensor will always be ahead of a smaller sensor at the same tech level but at some stage might smaller sensor may provide all the quality thats desired by any user? Diffraction on the other hand seems like it presents more of a fixed barrier creating an upper resolution limate for each sensor size.

Thats really why I posted this thread since it seemed like diffraction could have a big impact on the future of photography meaning larger sensors are here to stay rather than potentially being rendered obsolete years down the line.

It's is not to do with the size of the sensor, but the pixel density. If you have a full frame sensor, and the sites are the same size and geometry as found on the crop sensor then nothing has changed. With regards to diffraction, The full frame sensor is only a benefit, if it's pixel density is lower than the crop you are comparing it to

Read this: http://www.billmajoros.com/photoalbum/categories/articles/MegapixelsDemystified.html

 

[Moreorless]

It's is not to do with the size of the sensor, but the pixel density. If you have a full frame sensor, and the sites are the same size and geometry as found on the crop sensor then nothing has changed. With regards to diffraction, The full frame sensor is only a benefit, if it's pixel density is lower than the crop you are comparing it to

Read this: http://www.billmajoros.com/photoalbum/categories/articles/MegapixelsDemystified.html

Yes but more pixels on a smaller sensor means higher pixel density thus the resolution limate imposed by diffraction will be higher for a larger sensor?

 

[Richard King]

Yes but more pixels on a smaller sensor means higher pixel density thus the resolution limate imposed by diffraction will be higher for a larger sensor?

The diffraction will be the same, that's an issue caused with the lens.

 

[Moreorless]

Yes (as long as putting more pixels onto the sensor doesn't cause the per pixel quality to fall so far it outweighs the benefits of having more pixels).

However, it isn't clear to me what effect the number of pixels on the sensor has on whether diffraction is an issue or not. I put these numbers ...

... into the calculator and got the same answers as you. However, I then changed the number of megapixels to 2, and then to 100, and I still got the same answers. So the number of pixels on the sensor doesn't seem to influence the point where diffraction becomes a problem.

It does depend on how large you want to look at the image -diffraction is more of an issue for larger prints/on-screen sizes. And it also depends on how far away you are when looking at the image - many people tend to look at larger images from further away, which makes diffraction less of an issue.

Here's something else I've just noticed. For a 13" print, viewed at the default 25cm, with a 1.6 crop factor and the "default manufacturer standard" eyesight, the calculator says that diffraction becomes an issue between f/11 and f/13. However, change the eyesight to 20/20 and the calculator says that diffraction becomes an issue between f/2.8 and f/4. That is an enormous 4-stop difference! And I would have thought that 20/20 eyesight was a much more reasonable baseline.

What seems to make the difference on that calculator is whether you tick the "Set circle of confusion* based on pixels?" box below the appature box. If you do that then megapixels do have an impact.

The problem I have understanding it is that I'v never been able to fully get my head around how this relates to the circle of confusion so I'm not sure what impact thats having and whether having the box ticked or not reflects real world condictions.

 

[Richard King]

its a lot easier if in your head you separate the optical lens system from the mechanism for collecting the image (sensor/film)

 

[Moreorless]

Its really the collecting mechanism I'm intersted in.

Just to give an example say I have a 40 megapixel crop sensor and a 40 megapixel full frame sensor and the former has tech advanced enough to make image noise etc a non issue. I use different focal lenghts and different appatures to give me the same field of view and depth of field on both, is the crop sensor going to resolve less detail due to diffraction?

 

[ausemmao]

None of this is especially complicated...

There's a difference between diffraction reducing detail and diffraction reducing contrast. Diffraction reducing detail is does not occur until the Airy disk of a point source is many, many times the size of the circle of confusion. Reducing contrast occurs earlier. With enough pixels and some processing power, a lot of that contrast and some additional detail can be recovered.

The perception of sharpness doesn't actually have as much to do with resolution as it does with contrast and edge acuity.

Crop shows diffraction effects sooner because diffraction is a spatial property and at a given level of MP it has greater spatial resolution.

 

[arad85]

I use different focal lenghts and different appatures to give me the same field of view and depth of field on both, is the crop sensor going to resolve less detail due to diffraction?

I think you are mixing things here.

Firstly, a lens only has a single focal plane. Anything in front or behind that focal plane is out of focus. It's just whether you notice it or not as depth of field is only relevant when printing out. If you change the printing parameters (print from a crop of an image, print larger or view closer than "average") you change the depth of field. You can also look at it as changing the Circle of Confusion (CoC) as this is derived from viewing a "standard" print at a "standard" distance (and depending on who is coming up with the CoC, you get different measurements). Note for all practical uses with modern DSLRs, the CoC is bigger than a single pixel (even when measured in a Bayer pattern).

Diffraction limited just means that the effect of diffraction becomes larger than the effect of the CoC. Until you understand how you are going to display that image and hence what DoF it actually has when rendered, you aren't going to know whether the limit is the sensor sized based CoC or the Airy disk sized diffraction limit.

 

[HoppyUK]

Good posts from Andy and Ausemmao

There seems to be a difference between those concepts and difftraction though.

Yes there is, which is what I said right at the top. However, they all operate together in one imaging system so you can't just say it's all about pixels or diffraction or noise or whatever. It's everything together, but as separate properties.

A larger sensor will always be ahead of a smaller sensor at the same tech level but at some stage might smaller sensor may provide all the quality thats desired by any user? Diffraction on the other hand seems like it presents more of a fixed barrier creating an upper resolution limate for each sensor size.

Larger sensors will always be better, they have a basic advantage that technology might nibble away at but those same developments also apply to all formats and the gap is just too large. Where we are today, if you take the best of everything, they all line up in sensor size order and it will take something very different to change that. And sitting on top of all that is the lens, and the fundamental relationship between resolution and contrast, which together is what we call perceived 'sharpness'.

In theory, there is a degree of relative parity between formats. For example, you could take a full frame sensor with 10mp and a 4/3rds sensor with 40mp and they would have the same theoretical resolution potential. You could then shoot at say f/2 instead of f/4 and DoF changes would be equalised (and diffraction) and you could run at two stops lower ISO to put the photons back.

The problem is that sensor technology has not given us a 40mp 4/3rds sensor yet, and even if it did, we don't have lenses capable of matching it. I'm not sure there is a commercial will to push hard in this direction either, when we already have the answer by just starting with a bigger format.

Thats really why I posted this thread since it seemed like diffraction could have a big impact on the future of photography meaning larger sensors are here to stay rather than potentially being rendered obsolete years down the line.

Larger sensors are certainly here to stay. You can fix diffraction issues with lower f/number lenses, in theory, which are easier to design for smaller sensors. Olympus make f/2 zooms and sub-f/1.0 primes are here - there will probably be more, but then you're back up to more bulk and high cost and still chasing your tail, when a bigger sensor to start with just does it all better.

PS Just a comment really. For the purposes of DoF, you only need, in theory, less than 4mp which equates to roughly 30 cycle-per-mm resolution on full frame. We are miles above that with something a like a 7D with 18mp, but lenses are struggling to deliver much by way of contrast at the over 100cpm resolution level this requires.