S2 Depth of Field data?

[SJP]

Advertisement (gone after registration)

If Leica does things by the book then the CoC of the S2 is CoC(ff)/crop factor = 31 micron/0.8 = 38.75 micron. This is assuming you are viewing the images at ca. 1x the image diagonal, this is the way CoC is "defined" together with some assumed "acceptable" level of unsharpness. If they use 33 micron for a FF CoC then it becomes 41.25 micron, but that is a minor point.

 

There is no freedom choice in redefining something which has been decided by convention around 1930.

 

Lens imaging capabilities formally do not play a role at all unless the lens is rather poorly designed - in that case diffraction limitations start kicking in as well, but for any decent lens diffraction limitations should not be a factor anywhere near 40 micron.

 

If you are interested in maximum resolution at 500% then everything changes - except the "definition" of CoC & DoF which is cast in granite.

 

So which CoC are we talking about the "bog standard" or the pixel peeping redefinition of this concept? If the latter who is (re)defining and by what authority?

[Nick Rains]

If you are interested in maximum resolution at 500% then everything changes - except the "definition" of CoC & DoF which is cast in granite.

 

So which CoC are we talking about the "bog standard" or the pixel peeping redefinition of this concept? If the latter who is (re)defining and by what authority?

 

There is no 'definition' of DOF per se, merely a broadly accepted (and historical) starting position based around film, 10x8 prints and a 25cm viewing distance - this is where the 0.033mm CoC comes from. It's a perfectly workable solution to DOF calcs but assumes a conventional viewing distance of 1x to 1.5x the diagonal of the print size.

 

That's fine, but many cameras, the S2 in particular, are capable of far better results than this would indicate, and the prints can stand up to much closer scrutiny. This is not pixel peeping, this is fine print making.

 

CoCs and DOF calcs are merely tools to allow people to understand how much of their print will remain sharp under scrutiny - and you can choose the parameters to suit your needs. For me, 0.03mm is way too conservative, 0.01mm is more realistic.

 

No-one is redefining anything, just using the tools available.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Guest dav_k]

Nick is correct here. The DOF is basically a subjective quantity in photography and any notion of an absolutely correct value for its corresponding CoC will result in endless debate.

 

However, as someone correctly pointed out earlier in this thread, there is an absolute lower limit of what can be used for CoC for digital sensors, and that is indeed the size of the pixel. It is physically impossible for any digital sensor to resolve a feature that is smaller than its pixel size. On film, absolute lower limits of CoC values can sometimes be made smaller since there exist film with grain sizes of only 2 - 3 micron in size.

 

Personally, I like to use a CoC estimate for the S2 that will give me an "optimal" DOF based on its sensor without any interpolation. Namely, I use a weighted average of the native pixel spacing in the S2 sensor in each of the three color channels (red, green and blue). In the S2 sensor, green pixels occupy 50% of the sensor and are adjacent along the diagonals, giving a nearest pixel spacing of sqrt(2) * p, where p represents pixel width. The red and blue pixels occupy 25% of the sensor area and their nearest pixel spacing is 2 * p.

 

So, my "optimal" estimate for CoC is weighted as,

 

CoC = 0.5 * sqrt(2) * p + 0.25 * 2 * p + 0.25 * 2 * p = 1.707 * p

 

Since the S2 pixel size is 0.006 micron, this yields a CoC of about 0.01 mm, which is what Nick has arrived at by examining his prints from the S2.

 

It is true that viewing conditions and print sizes will allow more relaxed values for the CoC. However, once I have captured my images using a CoC based on actual resolution capability of the sensor, I am always in the position to make prints of an already captured image as large as possible and still remain as sharp, where "as large as possible" is only limited by the native resolution of the S2 sensor. Of course, this assumes that the S2 sensor does not out resolve the S2 lenses, which I believe most people would agree on this point.

[01af]

However, as someone correctly ...

... incorrectly ...

 

 

... pointed out earlier in this thread, there is an absolute lower limit of what can be used for CoC for digital sensors ...

No, there isn't.

 

 

... and that is indeed the size of the pixel.

No, it's not. Just why is it so hard to understand? This is a misconception! Circles of confusion and pixels are two different things. So PLEASE stop repeating this popular fallacy over and over, will you?

 

 

It is physically impossible for any digital sensor to resolve a feature that is smaller than its pixel size.

Sure. But that doesn't mean COCs smaller than a pixel wouldn't make any difference. When using a lens with less than infinite resolution (that is, a real-world lens) then the smallest features resolved will always be bigger than one pixel. For the smallest features to approach pixel size, the lens' resolving power must approach infinity—in other words, the COCs' diameters must approach zero. Which is impossible in a real world, obviously.

 

So, on a digital sensor with a given pixel pitch, a lens capable of creating COCs of a diameter of half the pixel pitch will yield a sharper image than a lens capable of creating COCs of a diameter equal to the pixel pitch. And a lens capable of creating COCs a quarter of the pixel pitch will yield a still sharper image. And so on. Of course, you will quickly run into diminishing returns ... still, COCs moderately smaller than the pixel pitch do make a significant real-world difference.

 

 

[...] So, my "optimal" estimate for CoC is weighted as,

 

CoC = 0.5 * sqrt(2) * p + 0.25 * 2 * p + 0.25 * 2 * p = 1.707 * p

 

Since the S2 pixel size is 0.006 mm, this yields a CoC of about 0.01 mm, which is what Nick has arrived at by examining his prints from the S2.

Don't you see what a naive fallacy your calculation is? You're starting out at reasoning about the COC that (allegedly) cannot be improved upon due to the 'pixel pitch limit' and eventually end up at what DOF always is—a compromise that sets a limit for what we might consider 'minimum acceptable sharpness,' depending on viewing conditions.

 

 

It is true that viewing conditions and print sizes will allow more relaxed values for the CoC.

Correct—but on the other hand it is also true that more demanding viewing conditions and print sizes may ask for even more stringent values for the maximum acceptable COC than 0.01 mm.

 

 

However, once I have captured my images using a CoC based on actual resolution capability of the sensor ...

You don't. You never do. No matter how small your COC is (no pun intended )—an even smaller COC will always yield a still sharper image.

[Guest dav_k]

First, no one here ever said that CoC and pixels are the same thing, only that CoC values can be derived from their consideration, if one chooses. It is totally a subjective decision by the photographer to do so, or not.

 

However, as far as the term "pixel" can be said to represent a fundamental sensor element that captures the image data, it becomes the limiting factor to how large a feature may be recorded at the sensor. Sure, you can enlarge the image of the actual object at the sensor with your lens of choice, but ultimately any signal that reaches the sensor that represents something smaller than the size of the individual sensing elements cannot be distinguished. I do not see how you can claim otherwise. It is a scientific fact.

 

As you agree, one can use any value of CoC that one wishes .... and I happen to choose to use properties of the sensor to influence my own personal choice. It works for me. No one is forcing you to choose one CoC value over another.

Edited January 10, 2011 by dav_k

[SJP]

CoC is defined for determining acceptable sharpness under standard viewing conditions, not for determining how far you can go under close inspection.

 

For this you can use Nyquist criteria (only to some extent as 01AF points out), airy disks, diffraction limitations, abberations etc. but not CoC.

 

Using it otherwise is a misuse of the concept and should be strongly discouraged.

[01af]

Advertisement (gone after registration)

... but ultimately any signal that reaches the sensor that represents something smaller than the size of the individual sensing elements cannot be distinguished. I do not see how you can claim otherwise.

Just look at photographs. A basic understanding of math and physics helps, too (but isn't utterly required).

 

 

It is a scientific fact.

Please don't argue with 'science' as a buzzword when you're having no clue about the sciences underlying signal transmission, optics, and photography.

[Guest dav_k]

Just look at photographs. A basic understanding of math and physics helps, too (but isn't utterly required).

 

Please don't argue with 'science' as a buzzword when you're having no clue about the sciences underlying signal transmission, optics, and photography.

 

Unless you are someone like Stephan Hawking or Ed Witten, please don't try to lecture me on math and physics.

 

As a working professional scientist, I believe I qualify to use the word "science" in my discourse. Although my doctoral work was in nuclear physics, I have also spent a significant number of years in the research and development of defense related technologies that specifically include signal processing and related digital imaging science. I doubt that you have similar qualifications for yourself. According to the manner in which you address this forum, it is clear that you are far from being a professional in any scientific field.

 

Even someone with no formal science background has the ability to understand that you cannot measure something that is not within the sensitivity of the measuring instrument.

[Guest dav_k]

CoC is defined for determining acceptable sharpness under standard viewing conditions, not for determining how far you can go under close inspection.

 

For this you can use Nyquist criteria (only to some extent as 01AF points out), airy disks, diffraction limitations, abberations etc. but not CoC.

 

Using it otherwise is a misuse of the concept and should be strongly discouraged.

 

Hi Stephen,

 

I understand your concerns. Can you tell me how you would estimate the depth in an image that would pertain to only the range of maximum theoretical resolution at the sensor (i.e,, assuming a "perfect" lens, and where only Nyquist criteria related to the sensor is relevant)?

 

David

[mjh]

CoC is defined for determining acceptable sharpness under standard viewing conditions, not for determining how far you can go under close inspection.

“Standard viewing conditions” – and there is still room for arguing about what these would be – is what you assume by default. If you have more specific information about the actual viewing conditions, your criteria for acceptable sharpness should be based on these. If I knew that my pictures would be viewed from close up, how could I feign ignorance and assume “standard viewing conditions”?

 

When you find a CoC value (calculated, guesstimated, found in a fortune cookie – whatever) that when fed into the DoF formula gives you results matching the criteria for acceptable sharpness as applied by the actual viewers of your pictures, then that’s OK by any sensible standard. It doesn’t really matter how that CoC value got derived.

[01af]

Unless you are someone like Stephan Hawking or Ed Witten, please don't try to lecture me on math and physics.

Did I? No, I didn't. Instead, I expect you to do your homework yourself. If you really are a scientist, you should be able to. I just couldn't help noticing that your lack of understanding of the basic laws of signal transmission is quite obvious, since you're falling for the usual misconceptions about how two resolution-limited devices (here: lens and sensor) interact when the output of the first is the input to the second. For the average layman that's venial ... but for someone who claims to be a physicist!?

 

 

Although my doctoral work was in nuclear physics, I have also spent a significant number of years in the research and development of defense related technologies that specifically include signal processing and related digital imaging science.

Then why do you believe that sensors can 'outresolve' lenses? Or vice versa? Actually, they can't. Better lenses will always yield sharper images than worse lenses, on any reasonable sensor. Also higher-resolving sensors will always yield sharper images than lower-resolving sensors, with any reasonable lens. So the notion that the sensor's pixel pitch establishes the lower limit for a minimum useful COC diameter is fundamentally wrong.

 

And by the way—digital sensors can, and will, record detail that's smaller than a single pixel. As anyone knows who opens the eyes when looking at photographs.

[Guest dav_k]

Wow, I can't believe that there is such hostility and ignorance in a Leica forum. I was hoping that this would be a place where serious Leica users could share their experiences in an open, friendly and positive atmosphere, with a common goal of being able to enjoy Leica's products. I guess this is my last post here. It's just not worth it.

 

Good luck to everyone else that is able to tolerate this kind of thing.

[c6gowin]

David, I can understand why you feel this way considering the last few posts in this thread. However, as far as I can tell there was only one person contradicting you so please don't let this one person color your opinion of the Leica forum. I see you have 10 posts and 5 of them are on this page alone. I really appreciate your contribution to this thread and I would hate to see you leave because of one other person. I understand why you would, but it would be a shame none the less.

[SJP]

“Standard viewing conditions” <snip> got derived.

 

OK, OK but still I think that making the CoC value "free for all" is a bad idea. See also here: Circle of confusion - Wikipedia, the free encyclopedia there is some argument about the exact FF value for the CoC but 29-33 micron seems to be the accepted range. Backcalculating from the Leica M-lens barrel markings they seem to be using 31 micron, if I got my maths straight.

[mjh]

OK, OK but still I think that making the CoC value "free for all" is a bad idea.

Depth of field is all about acceptable sharpness which is ultimately in the eye of the beholder. Neither Leica nor anyone could convince you that some slightly unsharp detail was still acceptably sharp when you strongly feel that it isn’t. But rather than giving up on the very concept since “It’s all subjective!”, we are factoring the equation into the DoF formulas on the one hand and the CoC definition on the other. That’s a pretty neat separation in that the formulas capture the objective part while the subjective part is in the choice of the CoC value. Once you have arrived at a CoC value that suits you, just plug it into the DoF formulas and calculate depth of field tables to your heart’s content. That’s my view of the matter anyway.

[Nick Rains]

Wow, I can't believe that there is such hostility and ignorance in a Leica forum. I was hoping that this would be a place where serious Leica users could share their experiences in an open, friendly and positive atmosphere, with a common goal of being able to enjoy Leica's products. I guess this is my last post here. It's just not worth it.

 

Good luck to everyone else that is able to tolerate this kind of thing.

 

Hi David

 

Sounds like a recent discussion on another forum, eh! At least ours was civil, if a little intense at times. And I agreed with you at the end, having shifted my perspective on things slightly.

[01af]

... ultimately any signal that reaches the sensor that represents something smaller than the size of the individual sensing elements cannot be distinguished. I do not see how you can claim otherwise. It is a scientific fact.

Just look at photographs. A basic understanding of math and physics helps, too (but isn't utterly required).

Even someone with no formal science background has the ability to understand that you cannot measure something that is not within the sensitivity of the measuring instrument.

Okay. Let's look at a few photographs. It's just some test shots I took five years ago in April (200 mm, 80 mm) and June (35 mm, 18 mm). All shots show basically the same subject. They were taken from exactly the same position (give or take a few inches) with exactly the same camera (APS-C-format six-megapixel DSLR) and with different lenses. The images show the full 6 MP files, downsized to a width of 720 pixels. The square-shaped insets show crops from the original files, as indicated. All images are unsharpened, except the 18 mm shot which received some capture sharpening. The distance to the cranes is approx. 300m/1,000 ft.

 

Look at the cables hanging from the crane's trolley. In the image taken with the 200 mm lens, they are about three pixels wide. So in the 80 mm image they should be approx. one pixel; in the 35 mm image approx half a pixel; in the 18 mm image approx. a quarter of a pixel wide. According to our Doctor of Nuclear Physics, the sensor should not be able to record them in the 35 mm and 18 mm shots. But miraculously, it still does! Glad my camera has no formal science education ...

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

 

 

 

Sure—the imaging of the cables isn't too accurate. In the wide-angle shots, the poor 6-MP sensor had to resort to translating the high-contrast, small-width cables into low-contrast, high-width images of cables. But when looking at the whole image from a reasonable distance, the visual impression of the cables is surprisingly close to the truth after all. In any case, the point is this: The sensor does respond to the cables, even when their width (or actually, the width of their images cast onto the sensor) is significantly narrower than one pixel ... even though David K. claimed this wasn't possible.

 

What the sensor cannot do is this: It cannot properly separate tiny features when there are several of them at distances smaller than one pixel. But that's an entirely different statement.