Talk:Depth of field/Archive 3

Archive 1

Archive 2

Archive 3

Archive 4

New Equations 17 Nov 2008

I am proposing some changes to the Derivations of the DOF Formulas. I find the derivations can be simplified by first deriving (simply) a set of exact equations and then showing how the one approximation affects all the equations. I have used "exact" equations without an explicit caveat about the assumptions that are mentioned as Limitations. I also added a derivation for the "One third rule" that goes directly at the fraction rather than at the related ratio of the distance in front and in back of the focusing distance. I replaced the arguments that allow the approximation of the DOF for macro with ones that are more apt. I also provided a derivation and procedure for the calculation of the DOF limits from the aperture and focusing distance. Using reciprocal distances makes derivation and calculation much simpler. The changes are based on derivations and explanations in Appendix H of Photographic Exposure Calculations and Camera Operation by Prais. Jeff, I like the variables that you used in your paper. (Nice work.) I think they are much clearer in the relationship between variables. They are also very similar what is used in Prais. The presentation of the Depth of Field Formulas in the first part of the article would be helped with some closer connections with the derivations. --Calculist (talk) 22:41, 17 November 2008 (UTC)

I haven't studied the changes in detail yet, but it looks mostly OK. Can you please take care of fixing the heading case to conform with normal wp style (sentence case)? And there's what appears to be an error where you say focusing beyond hyperfocal decreases the near limit. In fact, the near limit increases to the hyperfocal distance as the focus distance recedes to infinity. Dicklyon (talk) 03:43, 18 November 2008 (UTC)

This bit: "The fraction of the depth of field in front of the focusing distance is often and erroneously claimed to be one third of the depth of field" should be done better, and in a different place. This "rule of thumb" is not a claim about exact DOF, and doesn't belong in the exact section; in the approximations sections, it would be better to discuss in a positive way when it is a useful approximation, rather than call it "erroneous", which is what I'd called WP:OR; if you want to cite someone that calls it erroneous, that would be OK, too (this one calls it "certainly incorrect" while honoring it as a useful rule of thumb). But it's sometimes correct, and often useful, to think of DOF as being about twice as much behind as in front, even though it's almost never exactly so; this treatment is typical. Even more useful if you temper with the extreme cases: about equal in front and behind for macro shots, and infinitely more behind when at hyperfocal for further. The notion of a sequence of fractions fits well with Piper's "consecutive depths of field" idea, but doesn't really suit the continuous analysis you provided. Some more sources here will help you find a more positive and typical way to present it; amusingly, a couple of them state it backwards. Dicklyon (talk) 04:15, 18 November 2008 (UTC)

I think some of the additions are helpful, but for the most part, I thought it was more clear as it was. Most of the changes seem to simply reflect personal taste. If they remain nonetheless, they should be cleaned up. A few specific comments:

I question the statement that begins “Since photographers more commonly measure distances in front of the lens”; some photographers undoubtedly do, but many measure it from the image plane because that's how most hand-camera lens distance scales are marked.

The distance H was clearly intended to represent the approximate hyperfocal distance; defining it as special unnamed parameter adds unnecessary confusion. Introducing u_h adds a yet another nomenclature for object distance (we now have D, H, s, and u). Preferably, these distances would all use the same symbol (such as s or, preferably, u, as was done in the material added to Focus and f-number from DOF limits) with appropriate subscripts.

Units don't usually belong in derivations (e.g., under Approximate Depth of Field Equations). With the units mentioned, the equations are incorrect without additional constants.

It's probably OK to put all the approximations in one place.

It's certainly OK to examine the near:total DoF ratio, but it's confusing to look at it that way and later examine near:far (e.g., under Near:far DOF ratio). I agree with Dick that it probably doesn't belong on the new “exact” section.

It seems pretty strange to use H in the context of closeups (which is usually the only context in which DoF is expressed in terms of magnification.

Use of the obscure aperture number in place of the far more common f-number seems pretty offbeat; what's wrong with a WL to f-number as we have done elsewhere?

I see absolutely no purpose to the material added at the beginning of Focus and f-number from DOF limits, and take issue with the accuracy of several statements:

“The depth of field equations are a pair of equations that allows photographers to calculate the limits of the depth of field”

This seems a pretty capricious designation; the term was previously used to refer to all equations relating to DoF.

“It is possible to use these equations to precisely calculate an aperture and a focusing distance from the preferred limits of a field of focus found in the field--even when focusing beyond the hyperfocal distance.”

These equations won't ever give a focus distance beyond hyperfocal.

There also are several spelling errors, numerous typographical and formatting errors, and at least one reference (prais) that is cited but not listed at the end. JeffConrad (talk) 09:08, 18 November 2008 (UTC)

Jeff, thanks for the careful review. I think for now it will be best to revert this major change and then when he has time to work it over, try it again, taking our comments into account. Dicklyon (talk) 15:35, 18 November 2008 (UTC)

After further examination, I don't think it's such a good idea to introduce the approximate equations without including the exact equations in the same place—it's not obvious just how the equations simplify, forcing the reader to go back and forth. Including the exact equations with the simplifications, e.g.,

quantity = exact ≈ approximate

would address this, but seems unnecessarily repetitive. I also don't see the need for a section on “exact” equations; isn't a derivation presumed exact unless a simplifying approximation is introduced? JeffConrad (talk) 06:20, 19 November 2008 (UTC)

Okay, here are some things to think about as I make some of the changes:
One-third rule:
I evidently did not word it strong enough if Dick still thinks that the rule should be used. I don't have a problem substituting wording (for instance, "certainly incorrect" for "erroneous"), but the point is that when you look at the mathematics, the "rule" should be "dishonored". The mathematics, that is, an critical examination of exact formula for the fraction, says, "There is no rule." Another way of putting that is: A "one-half rule", a "one quarter rule", a "one-twenty-fifth rule", and a "40% rule" are all just as useful as a"one-third rule".
The formula says simply that it is not generally true that one third of the DOF lies in front of the focusing distance. It is specifically true that in one specific instance (s = H/3), but by and large it is false. Rules by definition are not true in one of many instances. (That is the definition of an exception.)
The rule is not even an approximation. Is one-third an good approximation to one half? to one twentieth? In what situations does the approximation hold? The claim seems to be "in all situations". That does not sound like an approximation.
(Proof by contradiction) Suppose that the "rule" is valid. Then what DOF are you going to use? The only place that the "rule" is valid is at s = H/3. The DOF at this point is H/4. Are you suggesting using this one value for the DOF at all focusing distances? No? Are you suggesting calculating and using the DOF for each focusing distance? If you are, then you can calculate the limits for each focusing distance, and you find that "rule" is inaccurate. For instance (Prais p296), a 50 mm lens set to f/11 has a 9 m hyperfocal distance. Focused at 8 m, the near limit is 4 m, and the far limit is 75 m. The fraction in front is only 5%. Focused at 1 m, the near limit is 9/10 m, and the far limit is 9/8 m. The fraction in front is 44%. I would not call either of these close to 33% (especially considering the values only run between o% and 50%). The bottom line: There is no "rule" because the "rule" as stated is generally invalid. This is what the readers need to be told.
(If anything can be said, the fraction in front is one half less the fraction of the focusing distance to the hyperfocal distance.)

No, I think you totally missed the point; it's not that you "did not word it strong enough"; I totally understand the extent to which it is not literally true except at a unique distance of about one-third the hyperfocal distance. But we're here to report, not to analyze and criticize. We can report the "rule of thumb" without implying that it's more than a useful approximation, and we can report criticisms of it; but we can't pretend it's a "rule" and insert an original debunking of it. As to whether it's useful or not, that depends on how you look at it; in general, it's correct that the DOF has more behind than in front; the "twice as much behind as in front" rule of thumb is useful as it reminds photographers to focus somewhat closer to the front than to the back of the region of interest. Of course, if you try to apply it literally when the "behind" stretches to infinity, you'll focus at infinity, which is not the right thing to do, but photographers will usually apply it less literally than that. A more accurate rule would be to set the focus at exactly the center of the range on your focus distance scale (assuming that scale is linear in lens extension and hence nearly linear in reciprocal distance); maybe you can find a source for such a rule and report that, too. Dicklyon (talk) 23:28, 22 November 2008 (UTC)

What's wrong with the way the “1/3” rule is currently handled? We state that despite what's often claimed, the rule is only true at one distance, without dwelling on it. I'm not sure deprecation is needed or even appropriate. I personally think the “1/3” rule is nonsense, and never make use of it. But the article is arguably better without my editorial comments.

Setting focus to the point on the distance scale midway between the near and far points to be within the DoF is well established; it's simply the hand-camera implementation of the image-side relationships that are already covered in the article. Unfortunately, this is usually almost impossible with the scales on autofocus lenses.

There's yet another technique for setting focus. For a planar subject at an angle to the image plane, setting focus to the middle of the image area actually sets the harmonic mean of the near and far distances. In other words, exactly what photographers learn not to do is sometimes actually the right approach. To my knowledge, this was first reported by Martin Tai in a post Hyperfocusing the Tulips to photo.net on 25 May 1996. The technique can even be further generalized; see Christopher Jones's article. JeffConrad (talk) 03:12, 23 November 2008 (UTC)

Thinking about the equation for near:total, it can be described in words as the fraction of the DoF in front of the point of focus is exactly half the fraction of the (true) hyperfocal distance behind the point of focus. The way to visualize this is from the point of view of the hyperfocal point to the camera. Picture the point of focus relative to the hyperfocal distance, and move it to half that distance. The relationship between that new point and the hyperfocal distance is the relationship between the point of focus and the DoF. Note well that the new point can move/be anywhere between the hyperfocal point and the camera, so the focusing point can be placed anywhere within (relative to) the DoF. There is not one special value. Making this useful requires values for and visualization of the hyperfocal distance and the depth of field. I think that I would rather work from an immediate, exact calculation or a table of near and far limits. But then, ignorance and the one-third rule is bliss.

${\frac {u_{s}-u_{n}}{u_{f}-u_{n}}}={\frac {(h+f)-u_{s}}{2h}}$ exactly --Calculist (talk) 14:20, 23 November 2008 (UTC)

If you can find a source for that relationship, we can discuss it in the article. Dicklyon (talk) 18:53, 23 November 2008 (UTC)

The relationship is easy enough to derive. But how is it an improvement over what we have, especially the approximate relationship for near:far:

${\frac {s-D_{\mathrm {N} }}{D_{\mathrm {F} }-s}}\approx {\frac {H-s}{H+s}}\,.$

I agree that the “1/3” rule isn't much use, but as I have said, I think we cover it adequately as it stands, debunking it without getting carried away. I honestly can't see either the proposed equation or the one that we have being much use in the field. Because the current equation is simpler, I'd stick with it.

Ostensibly, it's probably easier to think of focusing 1/3 of the way into the scene than to split the foreground and background in a 1:2 ratio. But it's not really that simple. The “1/3” setting (or the proper one) apply to the distance between the near and far limits of DoF; unless something else has been done, these points aren't known. If one simply choses an f-number, the hyperfocal distance can be calculated, and focus is simply set to that distance, and the “1/3” (or whatever) rule is irrelevant. If, alternatively, the near and far limits are chosen, the ensuing calculations (tedious on the object side, simpler on the image side) determine the f-number and focus, so again, the “1/3” rule is irrelevant.

Again, I think the article as it stands covers this more than adequately. JeffConrad (talk) 22:42, 23 November 2008 (UTC)

Ah! You're starting to see how the near:total ratio is easier to visualize than the near:far ratio. To my knowledge, Prais is the first place I have seen the near:total ratio expressed precisely. (It looks like the book came out this year.) He's pretty definite about the invalidity of the "rule" in the book and on a web page. We're pretty much in agreement that the "rule" has limitations. My point is that, given that we now have an precise expression for the ratio (as a function of focusing distance), straight algebra (allowed for making conclusions under Wiki rules) allows us to offer that the "rule" is wrong most of the time. If we, as editors, saw something that failed to use the rules of algebra, we would be within our purview to reject it. Right?

It might be nicer if there was only Dick's more-behind-than-in-front rule because that is always correct. We need to have the near:total ratio in the article because it is precisely an equation for the quantity to which one-third refers. --Calculist (talk) 01:42, 26 November 2008 (UTC)

Notation:
Jeff's paragraph "It seems pretty strange to use H in closeups..." and his statement "The distance H was clearly intended ..." together are a great example of the problems with the current notation. It really needs to be improved. There are a lot of connections that can be illustrated with notation that can help the reader--especially the novice. Here are some things that have worked:

• Use d or d_c for the diameter of the circle of confusion.
• Use D_a or d_a for the diameter of the aperture. It would be hard to get away from Da because N = f/D or N = f/D_a is so frequently seen. Asymmetric lenses create a problem. D_u ad D_v might as the front and back diameters might make a recognizable connection for the reader in this context.
• Use u and v as distances connected by the focusing equation. They look alike, and they are close in the alphabet.
• Use subscripts on all pairs of u and v. Without a subscript someone could interpret simply u to represent the set of three distances, but get confused when focusing distance does not have a subscript.
• Use lowercase letters as much as possible. Uppercase creates emphasis like D for DOF or DOF itself. Just Like Not Capitalizing Each Letter In A Sentence Is Hard To Read, and using all caps in an e-mail is effectively SHOUTING, using UPPERCASE letters is harder to read. We are no longer in eighth grade using a single formula with all caps. The viewer sees lots of formula here.
• Use n, s, and f as the subscripts for near, subject, and far. Assigning a subscript to the focusing (subject) distance is a problem because there is no better word than far for far and f is focal length.
• Use u_h for the hyperfocal distance, and h for the super-parameter f²/Nd that simplifies many of the equations (without approximation). Equate them as needed.
• Use p for the ratio of diameters of an asymmetric lens.
• Think twice about using y as a horizontal distance or a "characteristic" distance. It carries some weight as an arbitrary coordinate (not necessarily simply a distance) in the vertical direction.

The idea is to make it easy to read, to make it consistent, to make connections in context, and to avoid double or imprecise assignments.

I would be more inclined to consider a change of notation is you cited a source that does it that way. Dicklyon (talk) 23:28, 22 November 2008 (UTC)

I could go along with some of the changes, but I agree with Dick that there should be reasonable conformance to common usage—in other words, I would be unpersuaded by citation of one esoteric source. The role of an encyclopedia is to report rather than to proselytize. A few specifics:

• The use of c (or sometimes C) for acceptable circle of confusion is so common that I think d or d_c, no matter how logical you consider them, is a nonstarter.
• I'd be fine with using primarily lower-case subscripts because they're easier to read, especially with HTML-rendered TeX. I'd be careful of using too many upper-case quantity symbols because the math then becomes harder to read, much like putting a sentence in full caps (which is why I prefer c for circle of confusion).
• Most sources with which I'm familiar don't subscript the focused (or subject distance), but simply give it as u. It's one less thing to decode, and the “subject” may not be in exact focus when considerable DoF is required.
• For object and image distances, I prefer u and v, with appropriate subscripts, but as I recall, Dick and some others felt this might be a bit intimidating to many readers. I agree that D_N may introduce unnecessary confusion, but H and s, whether or not the most systematic, are well established in the photographic literature. Consequently, my concern with using H in the manner proposed remains; because of the common usage, the association with hyperfocal distance, exact or approximate, remains. JeffConrad (talk) 03:12, 23 November 2008 (UTC)

Jacobson, Ray, and Attridge in TMoP use u, v, and h. I believe that it is the most technical text in photography. They tend to mimic standards and other research efforts (lots of the time without any additional processing). Even though I find Ray's DoF derivation full of unnecessary complexity and assumptions, it's an appropriate model (except for the capital letters that have no association with the geometry and that make the derivation even more confusing).

I strongly advise against doubling up on the definition of a variable like h. It causes confusion for the reader. --Calculist (talk) 01:42, 26 November 2008 (UTC)

Aperture Number:
I prefer aperture number to f-number for at least these four these reasons:

• "f-number" is not the standard notation. ANSI Z38.4.7.1950 Standard Lens Markings allows "f/", "f", "f:", or "f-stop" followed by a number for the "relative aperture".
• "f/number" is precisely the aperture diameter D_a = f/N. Using f/number as a label doubles up on its meaning potentially causing confusion. (See notation.)
• "f/number" tells the viewer that focal length is the concept on which they should focus, while I want them to think aperture.
• Those who know do the extra work to make the association. I like to be a writer who makes it easy on my readers where possible. (See customer service. :->)

I don't understand; does someone refer to the f-number as an "aperture number"? Who does so? I don't think it's a common practice. Dicklyon (talk) 23:28, 22 November 2008 (UTC)

With all due respect for ANSI Z38.4.7.1950, I take issue with the statement that f-number isn't the standard notation; it's commonly used in more recent standards and elsewhere in the photographic literature. And f-number refers to N rather than f/N. I'm not sure I've ever seen “aperture number”. As Dick and I repeatedly said in similar discussion about “shutter speed”, it is not our prerogative to invent new terminology, even if we thought it would help. In this case, I don't even think it would help. JeffConrad (talk) 03:12, 23 November 2008 (UTC)

Finally, The Big Stuff (Organization of the Whole):
I attempted to recast the organization of this whole section as an exact followed by an approximate section , but I did not have enough time and input about the significance of some parts to integrate everything. Jeff's suggestion of quantity = exact ~ approximate make sense to me for the approximate section. The real benefit is in using the approximate section in this format as the reference from the front sections.
I put the "fraction in front" derivation with the exact stuff because it can be done exactly.
I left the "Near:far DOF ratio" and "... from DOF Limits" separate at the end because I need a sense of how important the near:far ratio is to large format photographer or if it can be replaced with the exact front:dof ratio. Jeff, are you making the point about the s ~ H/3 limitation of the "rule" or do you have another purpose?
I sense that LF photogs use the formulas "from DOF limits" for the image side, so I left part of that section in place. I'm also thinking that it is possible to derive the equations in this part of the section simply from the similar triangle equations at the top of the derivation section to make the connection easier to see. I didn't get a chance to look into the "FG and BG blur" section.
Since proposing my changes, I looked at the asymmetric stuff in vanwalree and in Jeff's pdf and see that the equations for the sym and asym cases from similar triangles are the same type (proportional and linear relations between vn or vf and vs), so the derivation from there is the same. It should be simple to present the similar triangle equations and the final equations for asym lenses in the section for exact results for asymmetric lenses, and put that section after the sym section and before the approximate section. BTW, Prais offers diagrams that show the similar triangle tactic works for points off the lens axis. That makes me wonder whether this analysis based solely on the focusing equation is restricted by paraxial optics as an assumption. Only if the focusing equation assumes paraxial optics, is it the case.

I think it's a tossup between near:total and near:far; the former is probably more common for general photography, and the latter probably more common for closeup work. I don't think format comes into play. If it's a tossup, I'd leave things as they are.

Replace s in near:total with (1+m)f/m for close up work. --Calculist (talk) 01:42, 26 November 2008 (UTC)

Focus and f-number from DoF limits are usually done on the image side in all formats. In LF, the measurement is often direct; in medium format and small format (i.e., helicoid-focused lenses), it's done with the lens DoF scales. Or at least it was before the advent of AF lenses. Until early 2004, some Canon EOS cameras included “Depth-of-field AE” mode, which did much the same thing (the purported “7/17” rule to the contrary not withstanding). Again, these techniques are simply various implementations of the image-side formulas given in the article. And these techniques actually have been put to practical use (perhaps without photographers' being directly aware of it), in contrast to most object-side formulas, which seem to get used primarily in endless academic discussions.

There was some hesitance to include any derivations, because even they probably aren't of interest to most readers (we've had occasional vandalism by folks who apparently aren't math fans). I decided to include a basic derivation because it's quite simple and illustrates that there's little magic involved, in contrast to a lot of nonsense on the web. But we nonetheless had to stop somewhere.

We moved the derivations to the end to make it easier for the average reader to skip them. I had considered including full derivations of the formulas for asymmetrical lenses, but left it out because it didn't seem that relevant to the general audience; the article can't include everything, and this seemed a reasonable cutoff. This article isn't primarily a mathematical treatise; we included numerous external links for the few readers who actually want all the details.

The similar triangle technique indeed works off axis, and the proof would arguably be more elegant if done that way. But we still rely on paraxial optics, because the thin-lens equation itself depends on the small-angle approximations for sine and tangent. JeffConrad (talk) 03:12, 23 November 2008 (UTC)

I Lied. Some small stuff:
Neither the equations I presented in the "... from DOF limits" nor the ones Jeff presented there can "give a focusing distance beyond the hyperfocal". The value of the far limit is not defined in mathematical parlance because infinity is name for a limiting situation and not precisely a value. The two sets of equations are the same equations expressed in different ways, so they should give the same results. However, most anyone will have a difficult time substituting infinity for the far limit in Jeff's version. The reciprocal version that I offer has the added benefit of making the step to the equations for the image side simple by substituting 1/u = 1/f - 1/v.
BTW, my quick derivation of the approximation for N in that subsection suggests that an additional approximation was used. I'm expecting something like N ~ (f2/c)(vn - vf)/(vn vf).
The new sequence of steps is not driven simply by capricious taste. I can offer value and logic for the order usually based on "need to know" (what do you need to know at this point?).
I could not figure out how to put Prais into the references and how to get the formatting to be consistent in the time that I had. --Calculist (talk) 17:12, 22 November 2008 (UTC)

I'm not familiar with your source, but I know these derivations are done many different ways. I had a different approach a year or two ago and Jeff reorganizaed it current way, which has been stable for a long time. I think that if you want to change it, you need to convince the rest of us that it's for a good reason. I agree with Jeff that so far it sounds like just a personal taste difference. Dicklyon (talk) 23:28, 22 November 2008 (UTC)

If it's felt necessary, we could mention the special case of infinity; on the image side, this is simply v_F = f. The “exact” version of the expression for N is exact; no additional approximations were used.

A “need-to-know” order is probably more defensible in the first section that in a derivation, where the smooth flow of concepts that derive from those that precede them is usually more important.

I stand by my comment that most of the changes simply reflect personal taste, and a taste that's a bit outside the mainstream. I'm amenable to some changes in nomenclature if others go along with them, but for the most part, I'd leave things as they are. JeffConrad (talk) 03:12, 23 November 2008 (UTC)

If there is a derivation section, it should at least be structured so that the reader can see the starting equations, the resulting equation, and any equations or approximations used in along the way. The algebra is left to the reader because algebra is logical argument. Several results in the current article are simply stated. They do not belong in the derivation section unless the above pieces can be added. The start of my alternate offering uses this structure: Similar triangle equations ... focusing equation ... resulting equation ... grouping of terms using N and h (definitions not substantive equations) ... resulting equations ... approximations ... resulting equations I'll revise the whole section incorporating the parts that I left out after the holiday so that you can see what I have in mind. --Calculist (talk) 01:42, 26 November 2008 (UTC)

The derivation section already starts with the proportion relationships "from similar triangles". If there's a probem in the section, please point it out more specifically, or simply fix it. Dicklyon (talk) 05:12, 26 November 2008 (UTC)

I've reread the derivation section once again, and for the most part, I don't see may places where the procedures used to reach the final equations haven't been clearly described, especially in the first section. It seems to me that we currently include all that you propose to add. If you have some specifics in mind, it would help if you could mention them. Again, I'd really like to see how what you have in mind fills a gap or improves the article rather than just reflects a personal preference. I think we'd be wise to avoid offbeat stuff such as “focusing equation” when nearly everyone else refers to it as the thin-lens equation.

It does appear that the introduction of the image-side equations under Focus and f-number is a bit abrupt, so I've expanded the treatment a bit to show how we got there. See if this works.

The basic equations for asymmetrical lenses are introduced without proof because the reader is referred to other sources for complete derivations. It wouldn't be that difficult to include a complete derivation (I could just about copy it from my paper), but we could easily end up duplicating almost every section in the derivation. Given the relatively small impact of lens asymmetry in most situations, I think we sensibly chose the cutoff that we did. JeffConrad (talk) 08:05, 26 November 2008 (UTC)

Digitally enhanced DOF: real or just simulated?

I reverted an edit indicating that digital techniques can simulate rather than actually produce increased DOF. Some of the terminology (e.g., “apparent DOF”, “graphically altered”) was a bit off, but the basic question may be valid. Do techniques such as wavefront coding, plenoptic cameras, and focus stacking actually increase DOF? Using the classical optical definition they may not, but should we treat an increased zone of apparent sharpness the same as increased DOF? If we do not, I thing we should apply the same criteria to anything that can't be done with a single focus setting with a conventional camera. I'd consider this question analogous to asking whether techniques such as HDR really increase exposure scale. JeffConrad (talk) 05:47, 22 July 2009 (UTC)

I was the guy who made that edit. I admit it was a bit clumsy, as I am a photographer, not a writer. I take issue with the paragraph as it stands on two points, and hopefully I can express them without bogging the expression down with semantics. The first is that depth of field is an optical property, not a property of an end result, and the article is about, or at least should be about DoF, not imaging, and not cameras or post production effects. The second point is that the ability to simulate such an effect is not dependent on digital technology, as a proficient darkroom technician has a number of methods available to an create increased zone of apparent sharpness as well. Perhaps not as elegantly, certainly not as easily, and with a limited tool set in comparison, but possible nonetheless. The creation or utilization of methods to alter the appearance of an image doesn't somehow alter the laws of optics, which to my mind is the impression that the paragraph conveys. I feel this creates a disservice to readers who may be at a point where they struggling to understand the concept in the first place. (JasBrunner (talk) 03:57, 23 July 2009 (UTC))

In a sense, I agree on both points, which is why I raised the question here. Like you, I tend to view DoF as s a simple optical characteristic, and I also think there is an unwarranted tendency to ascribe magical properties to anything digital (it's only one technique). You'll notice that I avoided mention of “dynamic range” in connection with HDR, because strictly, it's a misnomer in this context (unlike in application to sound recording, there's nothing dynamic at all). I also generally avoid vogue terms such as “tilt-shift photography” (can you imagine Adams ever having used it?) that are most often uttered by people who have no clue about what they are saying. But the train has left the station on both latter counts; as authoritative works such as The Chicago Manual of Style, as well as most dictionaries, remind us, usage is the ultimate determinant of what is “correct”, and I think that is what must guide Wikipedia editors.

My second point is that even if we agree that techniques other than a single exposure only simulate increased DoF, the impact is more far reaching than just the lead section of this article. The concept should be applied consistently throughout this article, and among related articles as well. We do a grave disservice to the reader when one article says one thing and a related article says something completely different; faced with such, the sensible reader concludes that at least one of them may be wrong, or perhaps that Wikipedia is simply a tale told by an idiot, signifying nothing. But I think to insist (or at least strongly urge) that this concept of DoF be applied across a range of articles needs consensus of more than two or three editors; absent such consensus, I think that, at least for now, this article should be left as it stands. Again, I'd like to see what others think. JeffConrad (talk) 07:26, 23 July 2009 (UTC)

I would like to hear from others as well. Consistency is important, and while Wikipedia shouldn't be a tale told by an idiot, facts by consensus don't serve very well either. I agree with your assessment and comparison to HDR, however point out that the HDR effect isn't the best analogy, because not withstanding the endless subjective discussion on acceptable sharpness, one can choose a CoC as an objective measurement and follow the law from there to a conclusion. HDR can't be reasoned out the same way, as it is an arbitrary effect applied subjectively. HDR is a distortion of reality, or perhaps another view of reality, but in no case did the luminance range of the subject change, merely the interpretation. In regard the the question at hand I have some difficulty with the idea that fiddling with the interpretation changes the facts. Photoshopping my blue shirt to red doesn't change the color of my shirt, would be my pedestrian way to express the position.JasBrunner (talk) 14:46, 23 July 2009 (UTC)

How are these concepts represented in sources? If they're consistent, we should be, too. If not, we should try to clearly represent the points of view in sources. Dicklyon (talk) 15:26, 23 July 2009 (UTC)

Sources? ... We don't got no sources ... I don't have to show you any stinking sources! Sources for a Wikipedia photography article? Ya gotta be kiddin' ...

It's hard to say what the sources for Focus stacking say, because ... well ... there really aren't any sources. The sources for Plenoptic camera and Wavefront coding all seem to speak of extended depth of field, but are those sources reliable? Most are written by the developers of the techniques. Many of the authors are trying to commercialize the techniques, so they have a vested interest in the claim.

I think the issue is mainly semantical—is DoF restricted to its classical optical meaning, even in the face of some “sources” who maintain otherwise? And if so, can we cite a source in support? It seems to me that, at present, we don't really have sufficient sources to make a definitive call. JeffConrad (talk) 04:44, 24 July 2009 (UTC)

I've never been a fan of design by committee; things have always worked out better when everything has been done my way. But not everyone has always agreed, so consensus has sometimes been necessary, perhaps even with a few concessions to dimwits. We face the same situation in Wikipedia; if we simply go making changes to a bunch of articles, we're going to end up in nasty edit wars, which won't do anyone any good. I somewhat cringed at the addition of focus stacking to this article, but didn't think I really had solid grounds for removing it. I suppose I could have insisted on a few reliable sources, but if we strictly applied that requirement to all the articles on photography, we wouldn't have any articles on photography. And I think consistency is as important in the application of rules as it is in describing the facts—one either applies rules consistently or there are no rules, however much we may try to deceive ourselves. But this is really a larger issue, beyond the immediate scope.

I agree that with a choice of CoC, we can readily arrive at a DoF—provided we insist that the classic optical definition is immutable. I don't know that we can do that. A definition that seems equally common (and perhaps more common in the popular literature) is the part of an image that appears sharp.

The question here is analogous to that of HDR, though as always, different is never the same. As with many things, whether HDR is a distortion of reality depends on how far it is taken. It can simply be a representation that more closely matches what was perceived in the original. If we are strict, almost any photographic image is a distortion of reality—the colors don't usually match, the brightness usually doesn't match, and the image is usually two-dimensional. So I think the distortion argument may be a bit peripheral.

As somewhat of a purist, I'm not thrilled about taking claims of extended DoF at face value. But I'm not sure we really have a basis for claiming otherwise, and revising articles to suit. We could say that some sources claim that various techniques can increase DoF, but I'm not sure this would go over well, either. It does seem reasonable to make slight revision to this article so that methods for increasing the region of apparent sharpness aren't exclusively digital, but I would think at least one specific example of a viable alternative would be indicated—I'm not sure I've ever actually seen one, aside from a composite that's obviously not a single image. JeffConrad (talk) 05:20, 24 July 2009 (UTC)

Right, and there are no rules in a knife fight (you know that movie?). Here's a source that refers to "the equivalent of infinite depth of field" and a "Miracle cure for depth of field", as if DOF is a disease. It doesn't say focus stacking increases depth of field, but that it's a workaround, basically. So we could go that way, citing that, until someone comes up with something better. (Better than a book? How can that happen?) Dicklyon (talk) 06:15, 24 July 2009 (UTC)

Better than a book? Depends on the book ... if you look hard enough, you can find one to support almost anything. We could probably safely start with retitling “Digital techniques for increasing DOF” to something like “Techniques for increasing the extent of sharpness” (or something less ponderous, with suitable expansion in the text), and a commensurate change in the lead section. We could cite Johnson, but he's arguably a fairly obscure source, at least in comparison with many of the others cited in the photography articles. But this probably is beside the point—I doubt than many would disagree that any of the techniques can increase the zone of apparent sharpness—the disagreement, if any, will arise from the implication that the techniques don't increase DoF. And I don't know if Johnson is much help in that regard.

We could make the changes to just this article, and ignore the disparity with the related articles, but it would seem to me that many readers would think, "Why does this article say one thing and that article another?" And the chickens may come home to roost (much as with California finance). I'd have a bit of a problem with treating focus stacking differently than plenoptic cameras or wavefront coding. As I've indicated, I'm far from sold on the arguments of a few decidedly self-interested papers. But I don't know that we have solid grounds for refuting them, either.

In any event, this article should be internally consistent. JeffConrad (talk) 08:53, 24 July 2009 (UTC)

One possible approach, at least with techniques that combine optics and signal processing (e.g., wavefront coding, plenoptic imaging) is to use the term that the sources for those articles use: extended depth of field. The term has been around for a while, and a web search shows usage in many areas, including scholarly journals and governmental agencies. A book search gives a fair number of hits as well, including the venerable Sidney Ray, who talks of ... gasp ... increased DOF (2002, 232). So my question about whether these folks properly equate their techniques with increased DoF is moot, because it's apparently already been answered. And in any event, the sources don't generally use “increased DOF”, so I stand corrected.

We then go back to focus stacking, to which Extended depth of field redirects. I have a problem with the redirect, because it ignores other equally valid means of achieving the result. Short of writing an article, I don't know of a good quick fix other than creating a disambiguation page that lists the various related articles (and this seems stretching disambiguation a bit).

Going back to this article, I suggest we leave things much as they are, perhaps replacing “increased DOF” with “extended DOF” for consistency with most of the sources. The one remaining issue is “digital”; although there are and have been other techniques (those that Ray describes don't involve signal processing or digital postprocessing), I don't think they were frequently used outside of microscopy (correct me if I'm wrong). So aside from a historical perspective, a strong argument could be made for leaving the mention of digital as it is. If nothing else, digital processing, whether of a single exposure or of multiple exposures, has made the technique available to vastly more people. It's possible that alternatives have been reduced to fringe applications, and WP:DUE might well lead us not to even worry about them. This isn't to suggest that they aren't important; it's simply a matter of the appropriate weighting when the entire topic is covered in one sentence in the lead section. If more coverage is felt necessary, the best approach might be to add another subsection (and perhaps eventually, an article).

Again, after further consideration, I'm for leaving things largely as they stand. We could add a citation of Johnson, but I think the citation would be more appropriate in the main article on Focus stacking. JeffConrad (talk) 21:45, 24 July 2009 (UTC)

So what do we do? I added citation of Johnson to the main article on focus stacking but made no other changes. JeffConrad (talk) 03:08, 27 July 2009 (UTC)

Gibson ref

Jeff, I see you just added the 1975 Lou Gibson ref. Unfortunately, his calculations and conclusions on macro DOF include a bad error, where he got the wrong magnification into his equation at one step, which led him to conclude that you need to do contact prints to get the best DOF, if I'm recalling it correctly. Anyway, it's a great book, but beware relying on it for his DOF derivation. Dicklyon (talk) 01:30, 22 August 2009 (UTC)

Dick, I'm roughly familiar with Ted Clarke's criticism of Gibson's findings, but haven't investigated in detail. I did notice Gibson's Eq (6), p. 97, which essentially says the total DoF (or “depth of detail”, as he calls it) T is given by

$T={\frac {2NC\left(M+1\right)}{M}}$

when I think it should be

$T={\frac {2NC\left(M+1\right)}{M^{2}}}\,,$

which he gives on p. 90. I say “think” because he refers to the total (subject–to–final image) magnification M and projects the CoC C to the subject plane, which is a bit different from what I am accustomed to. Nonetheless, this error (or whatever) has no effect on the optimal f-number (i.e., that which maximizes DoF), though it obviously affects the total DoF. To be honest, I find Gibson's writeup (at least as it appeared in the version I cited) a bit aggravating to follow, simply because one must consult many other pages to find definitions of some parameters, and no support is given for some premises (like Eq. [6]). For that reason, I never really attempted to verify his findings until last night. And I still ignored everything but diffraction and defocus from the camera lens, attempting only to compare his results with Hansma's—so I did not even look at his claim that the magnification must be achieved in camera rather than by enlarging. The results compare quite well for “optimal” f-number, though interestingly, the f-number that maximizes DoF (Gibson's objective) isn't quite the same as the f-number that minimizes the total (defocus + diffraction) blur spot (Hansma's objective).

In any event, my purpose in finally mentioning Gibson was to show that using a root-square combination of defocus and diffraction blur spots isn't exactly a new idea. I'm still not sure it has a solid theoretical basis; I prefer Hopkins's 1955 method using MTFs, on which David Jacobson and I relied in our analyses.

Eventually, I think this article should say a bit more about diffraction, if only to mention that one cannot improve DoF by simply stopping down forever. This has at least two applications:

1. Situations in which extreme DoF is wanted; one simply may not be able to get there from here. Or one may need to employ movements to reduce the focus spread.
2. Smaller CoCs chosen for critical viewing of images captured with high-resolution media. A step or two may work, but diffraction quickly dominates when any significant DoF is required.

JeffConrad (talk) 04:30, 22 August 2009 (UTC)

I agree, the Hopkins 1955 analysis is more correct, and doesn't quite back up the idea of adding the squares of spot diameters from defocus and diffraction; that idea has a solid theoretical basis in convolution, which unfortunately is not applicable to combining a diffraction blur and a geometric blur. The Gibson derivation is aggravating to follow, as you note; I did manage to find the step where he plugged in the wrong M, after corresponding with Ted Clarke about it. Id didn't check to see if he got the same optimal f-number. As long as we're both aware that it's not all quite right, it should be fine. I'd have to get it out and study it again to see what you're saying about the equation above. Dicklyon (talk) 05:50, 22 August 2009 (UTC)

(indent--;) After a further quick look, it appears that Gibson's equation for total DoF is error, though not quite as I had originally guessed. On p. 95, he indicates that total DoF T is given by

$T={\frac {uc}{D-c}}+{\frac {uc}{D+c}}\,,$

where u is the object distance, D is the diameter of the aperture stop, and c is the CoC, referred to the object plane. This formula is correct, if a bit different from the form in which it is usually given today. Combining terms on the right-hand side gives

$T={\frac {2Duc}{D^{2}-c^{2}}}\,.$

Because c² << D², T is well approximated by

$T\approx {\frac {2uc}{D}}\,.$

Noting that

$D=f/N$

and

$u={\frac {m+1}{m}}f\,,$

one arrives at

$T\approx 2Nc{\frac {m+1}{m}}\,.$

So I think Gibson should have used the camera magnification m rather than the overall magnification M (= Em, where E is the enlargement). But shouldn't the denominator be m²? The reference of the CoC to the object plane incorporates a factor of 1/m, so that if the DoF is given in terms of the normal (i.e., referred to the image plane) CoC c′, the DoF is given by the familiar

$T\approx 2Nc'{\frac {m+1}{m^{2}}}\,.$

This error has no effect on the quantity b under the radical on p. 97, so it wouldn't affect the optimal f-number. The error does affect the quantity a, so the value for the DoF would be wrong.

So we can't reasonably use some of Gibson's conclusions. But for the most part, he was conceptually on target, and remains a good reference if care is taken when citing it. Even the root-square convolution of defocus and diffraction blur spots, though weak in theoretical justification, isn't all that bad. Using Hopkins's 1955 method, I get almost the same “optimal” f-numbers as Hansma does using the root-square convolution. And Gibson's approach is essentially the same as Hansma's with a slightly different objective; even if theoretically imperfect, it's apparently much better than ignoring diffraction altogether. And it's infinitely better than much of the nonsense one finds on the web. JeffConrad (talk) 07:15, 23 August 2009 (UTC)

Agreed. And now we have it all on record here. Dicklyon (talk) 15:21, 23 August 2009 (UTC)

Photolithography

Because the wafer surface is the image plane, don't we really mean depth of focus rather than depth of field? JeffConrad (talk) 03:37, 27 August 2009 (UTC)

Tag for inline citations.

Every reference has at least one inline citation, so I don't understand this tag. Perhaps the editor could indicate statements that he thinks require additional support. Absent some explanation, however, I'm going to remove the tag. JeffConrad (talk) 05:12, 10 October 2009 (UTC)

Please use ref tags.--Kozuch (talk) 10:02, 11 October 2009 (UTC)

Also, looking at this talk page, seems like you need to read Wikipedia:Ownership of articles. Thanks for listening.--Kozuch (talk) 10:07, 11 October 2009 (UTC)

The author-date reference system is common in the social sciences, physical sciences, and natural sciences, and the University of Chicago Press recommend it for works on those topics. They also additionally recommend it when a work includes both inline citations and substantive footnotes, to avoid the awkwardness of two different tagging systems for notes. Such a dual system is awkward even in paginated material, but is simply a mess in unpaginated material such as Wikipedia; that’s the main reason I've used author-date here. I don't suggest that Chicago are the only word on the subject (as they readily acknowledge), but they're unquestionably authoritative. More to the point, the Wikipedia guidelines clearly indicate that author-date system is acceptable for Wikipedia articles, and the guidelines also clearly indicate that new sources should follow the existing style. This is simply common courtesy as well as common sense.

It would appear that the tag was added in protest to the author-date system used in this article, which is hardly an appropriate use. In any event, every source has at least one inline citation, including page numbers, so the tag makes no sense, whatever the reason for adding it. Accordingly, I've removed the tag.

I've once again removed the link to Bob Atkins’s page on optimal f-number. There’s nothing wrong with Atkins or the article; the article is unrelated to depth of field. The sources cited in this article do address the tradeoff between defocus and diffraction, and are appropriately included. JeffConrad (talk) 01:03, 12 October 2009 (UTC)

I agree with you both. I understand how to use ref tags, but I can't understand how to maintain an article in this other style. Nonetheless, since Jeff introduced this style back when the article was sourceless, and has maintained and expanded it, we should respect and stick with the style he chose. And while Jeff does exhibit a bit of ownership of some photography articles, it's not excessive or autocratic, and he has earned the right watch over the content that he has done so much to improve over the last three years. I have never known him to reject good contributions from others. Dicklyon (talk) 07:20, 12 October 2009 (UTC)

I've once again removed the tag, because it simply makes no sense, as I've already mentioned. It still appears that the tag was added because Kozuch doesn't like the author-date system of references, which isn't a valid basis. I'll concede that this perception has added considerably to my annoyance. In no way do I suggest that this article is perfect, or a good article, or even that it has all the inline citations that it could. But it's much like the New Englander's response to “How's your wife?”: “Compared to what?”. In comparison with most of the other photography articles, I'd say this one is pretty well sourced. This isn't to say that I find “We're not as bad as our competitors” especially persuasive, but I do think things should be kept in perspective.

This article has been dinged before for lack of inline citations; although many have since been added, the documentation still may not be what it should be. But comments like “article needs further sourcing” or “Many paragrapshs and even sections without a footnote at all” aren't very helpful. Again, if there are statements or even general ideas for which someone thinks support is indicated, please give some idea of what they are. I don't necessarily buy the idea that citations/sentence is a valid criterion for documentation, but rather think it should depend on context. Moreover, many of the statements are quite well supported by derivations of almost every important mathematical relationship, and those that are should not require external support. At one time, I was inclined to get rid of the derivations entirely, but upon further thought, think they are probably better retained simply because there is so much extant nonsense on this topic that the only solid approach is to show the origin so that the user can make his or her own assessment rather than rely on my authority vs. your authority. A good example was Dick's comment that some of H. Lou Gibson's formulas were in error; short of showing exactly what the error was, we have no justifiable basis for making such a statement.

I'll readily concede that I have little patience with tweakipeidans who make edits seemingly solely for the amusement of manipulating the words, and really take issue with edits that are at odds with nearly all published sources. It's possible that sometimes, like Van Jones, I may be aptly described by a technical political term. But at other times, especially when it comes to sourcing, I think I'm pretty easy compared to the merciless Dicklyon, who seems to have a pathological aversion to patent nonsense. Especially when unsupported even by published patent nonsense.

Maintaining author-date citations? How hard can it be? One simply gives the author's last name, the date of the work, and preferably, also the page number. I'll concede that manually coding the links is a bit of a pain; at the time I began doing this, the templates for author-date had a number of shortcomings that have since been largely fixed. One interim approach would be to change the link IDs so that they match those generated by the harvard citation templates; I originally used different names to avoid collisions, but in retrospect, I don't think this was the best choice. I'll work to change this so that future inline citations can simply use the harvard templates.

Now it seems to have been suggested that I've created an unmanageable reference system; if so, I'll volunteer to add the needed citations myself—it's really not that difficult. But it would really help to know where those citations are needed.

It's obvious that in some instances, such as this article, I prefer the author-date system, for reasons I've already discussed at length. Again, it's the predominant system in technical journals, so it's hardly as if I've extracted it from the effluvia of the lodestone, and it shouldn't be all that mysterious. Quite honestly, it's never been my instinctive system, but I've so often found the need for substantive footnotes that I've had to revise quite a number of papers, and have grown quite weary of doing so; mindful of this, I now usually use author-date from the beginning. As the University of Chicago realized long ago. But if it's really the wrong choice here, perhaps we should discuss it. I'm not fond of commingling notes and references, and commercial publishers almost never do it, but it's not without precedent in Wikipedia (see, for example, Contempt of cop).

In any event, I'm quite averse to playing games with inserting tags intended for another purpose. If there are really strong feelings about statements that require additional support, let's offer some specific suggestions. We'll go a lot further in improving this article with a lot less effort and aggravation. JeffConrad (talk) 12:15, 12 October 2009 (UTC)

I am not going to place the template all over unreferenced sections... though it is very sad you dont get this. Look at an FA for how to do referencing right.--Kozuch (talk) 14:42, 12 October 2009 (UTC)

Sounds like Jeff gets it just fine. I even like his characterization of me. The problem I've had with this style is that it requires edits to multiple sections to add a ref, and the structure of those things like span tags and ids is unfamiliar to me, and it's sometimes hard to find what to remove when reorganizing material across articles, etc. Plus I find it harder on the reader to go through an extra level of indirection to find what the source is. My preference is to avoid putting substance in footnotes, and just use refs, but Jeff is right that the MOS says both methods are fine. If there are sections missing sources, just add unreferencedsection tags as needed. Dicklyon (talk) 21:25, 12 October 2009 (UTC)

I think sourcing can become an end in itself. Some editors think that almost every sentence needs a citation, and there are some articles that approach this. But I don't think the MOS even remotely suggests that sourcing need go that far. From WP:CITE:

“Sources should be cited when adding material that is challenged or likely to be challenged, when quoting someone, when adding material to the biography of a living person, and when uploading an image.”

I'm not sure this article falls all that short of compliance. This isn't to say that the article couldn't use some additional citations, but I question whether lack of a citation for every statement or even every section merits the tag that was added; were that the case, 90% of articles would get tagged, and at that point, the tag would cease to have any meaning. If a reasonable case can be made for restoring the tag, I don't have a problem with it. But it's very difficult to address only a general comment, and quite honestly, neither “Please use <ref> tags” nor “though it is very sad you dont get this” is very helpful. It would be far more constructive to indicate even general areas where material needs support, as Dick suggested. But again, the MOS doesn't require a citation for every statement, every paragraph, or even every section—it's substance, not numbers.

There are pros and cons to any citation style, some of which are discussed here.

The notes here address issues that sticklers might reasonably raise, but that are probably of little interest to most readers. The notes could be merged into the text, but after looking again at most of them, I think the text would be cluttered and more difficult to read. Perhaps an even better example is the article Scheimpflug principle; if the material in the notes were incorporated into the text, the article would be a mess. Yet without the qualifications in the notes, some of the statements would be incorrect. But I suppose it's open to discussion, just as is the citation style. JeffConrad (talk) 03:32, 13 October 2009 (UTC)

I've changed the format of the id attributes in the citations and references to match those generated by the {{harv}} templates. And I've replaced the <span> tags for the references with <cite> to get the cute highlighting. Unfortunately, the format of the citations generated is slightly different from the Chicago style used in this article, so every instance would need to be replaced. It would be simple enough (and arguably justifiable) to create Chicago-style templates, which I'd probably call {{harvcomma}} or {{harvCMOS}}. “Harvard” referencing is largely a Britishism not commonly used in the U.S., but since the templates would be only slight variations on those extant, I think the naming scheme should be consistent. But if input in the form of

{{harvCMOS|Ray|2000|pp=52–53}}

isn't seen as an improvement over

([[#CITEREFRay2000|Ray 2000]], 52–53),

perhaps it's not worth pursuing.

For what it's worth, I counted 27 inline citations. Admittedly, they're far from equally distributed. JeffConrad (talk) 05:43, 13 October 2009 (UTC)

I've gradually been adding citations over the last year or so, concentrating on statements that someone might reasonably claim aren't common knowledge in photography. Though perhaps it's not quite up to FA standards, I this article is reasonably sourced. I'm not even sure a technical article like this needs as many citations as one on historical or current events; in any event, I don't think it's strictly a numbers game. (for those counting, there are 21 references with 36 inline citations). Again, if there are particular statements that seem to need support, it would be much easier to address them with some specifics.

Girolamo—you were one of the first to suggest that this article was shaky on sourcing; any more recent thoughts? JeffConrad (talk) 05:12, 15 October 2009 (UTC)

External link to Atkins Digital Depth of Field article

On another issue, I noted that the formula for DoF in Atkins's article Digital Depth of Field is off by a factor of two; compare with Ray (2002, 218). I had mentioned this to him six years ago, but never got a reply, and had forgotten about it. Because the factor cancels out when computing DoF ratios, its omission doesn't affect his conclusions (which are correct), but I nonetheless have a problem with a link that contains an obvious and significant error. I've sent another message pointing out the error; if I don't hear back in the next week or so, I think we should remove the link. We can restore it if the error eventually does get fixed. JeffConrad (talk) 03:32, 13 October 2009 (UTC)

Effect of cropping on DoF

I reverted the edit 29 October 2009 by 68.33.206.177. I assume the edit was good faith, but it was half thought out, both technically and grammatically incorrect, as well as uncommented and unsourced.

Though it's seldom discussed, cropping does affect DoF if the the final image size does not change: the captured image requires greater enlargement, so the CoC decreases. If it's thought to be of sufficient importance, we could add a treatment of this, though offhand, I don't know what to do about sourcing. At the cost of yet more math, we might be able support it with an analysis similar to that used in DOF vs. format size under Derivation of the DOF formulas. JeffConrad (talk) 04:27, 29 October 2009 (UTC)

I think Leslie Stroebel is a reliable enough source, and quite explicit. Or you could cite my draft paper... Dicklyon (talk) 04:52, 29 October 2009 (UTC)

But I think the edit you reverted was agreeing with us. Dicklyon (talk) 05:08, 29 October 2009 (UTC)

I was about to mention the same link. I don't have the 7th ed. (and I've exceeded my Google previews), but it's also covered on p. 134 of the 3rd. ed., which I do have. As for citing any of the current external links, I think we've been pretty selective about what's included; most explain how they arrive at what they get, so they're amenable to self verification by any reader with a mastery of high school math. Though I don't see a problem mentioning them, I'm not sure WP:SOURCES allows citing them. I realize that a great percentage of articles violate this egregiously, but I'm not sure that justifies doing it here. But perhaps I'm being overly strict.

In concept, the edit was saying what we've said here, but it wasn't stated correctly, and I also think that because lead section ideally summarizes the key points in the body of the article, anything in the lead section needs some treatment in the article text, which the editor did not provide. To be fair, I think you've summarily whacked some similar edits (e.g., the Leica M9 in Full-frame digital SLR) that were less questionable than this one. Had the edit been amenable to a simple correction, I'd have just made it—doing so would have taken less effort than a revert. At the risk of ownership, the issue then would seem to be the extent to which an uncommented, unsourced, and technically incorrect edit that requires significant additional material to justify inclusion creates an obligation on the part of others. I'm not sure that it always does.

But it seems as if you think the material merits inclusion, so I'll see what I can come up with. I don't see a current section where this would fit comfortably, so a new section, perhaps just after DOF vs. format size, may be needed. Consistency would seem to indicate a treatment in the derivation similar to that for format size, again with another section. We could, of course, get by with far less if article quality were of no concern. JeffConrad (talk) 05:46, 29 October 2009 (UTC)

Stroebel also says essentially the same thing in Basic Photographic Materials and Processes, p. 153. JeffConrad (talk) 05:56, 29 October 2009 (UTC)

I agree, it was malformed, and I probably would have "summarily whacked" it myself if you hadn't. And I was kidding about citing me, since we have a good source. Since this point comes up a lot, we should say something -- but I sort of thought we did already. Maybe not. It could be a last paragraph in the format-size section, to just indicate that cropping is the same as choosing a smaller format size. So people who argue that a "crop factor" has no effect get another chance to get the point. Dicklyon (talk) 06:37, 29 October 2009 (UTC)

And as I recall, I protested some of the summary whackings ... the attack of the devil's advocates. I suppose in adding this section, I conceded that we probably should cover this, and I think we subsequently convinced ourselves. We may have already covered this, but only indirectly—we didn't explicitly mention “cropped” formats. I've now added explicit mention of cropping to the lead section, DOF vs. format size in the article body, and DOF vs. format size, so hopefully, the bases are covered. I've seen many silly and seemingly endless discussions on this, so the added treatment probably works to good effect. Since I covered this in at most two sentences, I couldn't see adding two new sections. Hopefully, it's not so buried that a reader can't find it.

Stroebel additionally discusses the effect of viewing distance, but I think we already give that sufficient coverage. The lead section is a bit choppy, but I think it may end up unnecessarily longer if it's given much smoothing. Perhaps someone else has some ideas.

As to your paper, I too was half kidding ... the limitations of written language. Though we'd probably strictly be on shaky ground citing most of the linked articles, many of them, including yours, are more useful than some “official” reliable sources because they explain what was done, including simplifying assumptions, in arriving at conclusions. For example, Stroebel (at least in the 3rd ed.), does not make it clear that some of the DoF comparisons apply only over a limited range of subject distances. And these comparisons fail badly near the macro range for the larger format and near the hyperfocal distance of the smaller format. But we're probably better off strictly following the rules. JeffConrad (talk) 08:18, 29 October 2009 (UTC)

The article now has several instances of enlarging an “image” to a “final image”, which I find a bit awkward and potentially confusing. And if it confuses the editor ... For Adams, it was easy enough to speak of the “negative” and the “print”, but that terminology no longer is really appropriate. I have no better alternatives than “captured image” and “final image”. I think the meaning of the latter is fairly clear, and we've used it for some time. The former finds some support in the sources (e.g., Evening, Langford, Langford et. al.), but it's new enough that it probably should be defined at first mention if we use it here. I think “when the captured image is enlarged to the same size final image” quickly gets to be a bit much, so perhaps the former could be used sparingly. In a few places, I think replacing “if an image is taken” with “a picture is taken” might also help preclude confusion. Any thoughts? JeffConrad (talk) 04:12, 30 October 2009 (UTC)

Certainly something needs to be done. I like to think that I've already grasped the practical basics of DoF, but when I read the following (after markup stripping) --

When an image is taken in two different format sizes from the same distance at the same f-number with lenses that give the same angle of view, the smaller format has greater DOF. When an image is taken in two different formats from the same distance at the same f-number using lenses of the same focal length, the smaller format has less DOF.

-- I'm mystified. The first sentence of that is fine. I believe that I understand the second sentence of that, and it's plain wrong. Let's suppose that I have two cameras: one taking 24x36mm images and the other taking what are commonly called 6x9 images (though they're slightly smaller), both with an 80mm lens. For the former camera, the 80mm lens might be marketed as a portrait lens: for the latter, the (very different) 80mm lens as rather wide angle. I believe that (putting aside internal reflections, spherical aberration, etc etc) the negative I'll get from the former camera will be like the central section of the negative I'll get from the latter camera, and with the same DoF.

Now, I'm willing to believe that I'm wrong, but my guess is that I'm mostly right on the issue but have managed to misunderstand or fail to notice some part of the explanation here.

Both sentences are correct as written. Let's not get sidetracked by aberrations, internal reflections, or how a lens is marketed. When a picture is taken from the same distance with lenses of the same focal length set to the same f-number, the magnification is the same, so there's only one variable left—the circle of confusion. Because the 24×36 captured image must be enlarged more than twice as much as the 6×9 captured image, the CoC for the 24×36 image is less than half that for the 6×9 image; accordingly, it has less than half the DoF. I think this is explained pretty well under DOF vs. format size. You are correct that the 24×36 image will be equivalent to the central part of the 6×9 image; when an image is cropped, the DoF decreases, as also noted here. Stroebel said the same thing over 30 years ago (and perhaps earlier), so it's nothing new to the digital era. JeffConrad (talk) 06:32, 30 October 2009 (UTC)

Because the 24×36 captured image must be enlarged more than twice as much as the 6×9 captured image -- It does? Well yes of course it normally does. (See below.) But let's say you and I are in front of the Taj Mahal and I have these two cameras, each with an 80mm lens. We stand still. (We're an odd couple indeed.) With the 35mm camera, I do a head-and-shoulders of you. I also aim my Mamiya Universal Press or whatever straight at your head. The latter photo is compositionally awful in its entirety, so I just use it for head and shoulders..... Hoary (talk) 10:10, 30 October 2009 (UTC)

I'm afraid I don't follow. Aren't we saying the same thing? Assuming we have the same field of view in each final image, the 24×36 and cropped 6×9 images would have the same DoF, which would be less than the DoF for the uncropped 6×9 image. JeffConrad (talk) 20:03, 30 October 2009 (UTC)

Clearly digital image capture is now commoner than image capture on film, and the balance will continue to swing toward digital. However, medium/large-format digital will continue to be esoteric for some time, and the differences in image size among the digital cameras with which readers are likely to be familiar are not so dramatic. For that reason, I think that even for the second decade of this century it might be more helpful to introduce a discussion of the effect of captured-image size via a comparison between (a) APS or Four-Thirds and (b) a medium-format film size that's of the same or similar aspect ratio.

I agree that the most common comparison is between full-frame 35 mm and various “cropped” formats that use the same lenses. But I think the intent of this article is to be generic and essentially format-neutral, so I think the comparison with 4×5 is apt, especially since DoF ratios of various 35 mm formats are already covered in at least two other articles, including Full-frame digital SLR. There actually are quite a few people who still use 4×5, 8×10, and even larger. JeffConrad (talk) 06:32, 30 October 2009 (UTC)

Yes, 4×5 (or above) would be fine, as you say. -- Hoary (talk) 10:10, 30 October 2009 (UTC)

Where you said "the negative I'll get from the former camera will be like the central section of the negative I'll get from the latter camera, and with the same DoF," you would have been correct if instead you had said "the negative I'll get from the former camera will be like the central section of the negative I'll get from the latter camera, and with the same amount of blur at every corresponding point." But this will not correspond to the same DOF unless you use the same sharpness criterion (same Circle of confusion criterion) for both formats.

But it is more consistent, and more conventional, to take a CoC criterion as some fraction of the image size (for a print), or format size (for a negative or sensor). The smaller format, having a more stringent sharpness criterion (smaller CoC diameter), has a smaller DOF. Your reasoning on this is what we were referring to as a common misconception, which we ought to try to clarify, by citing sources that explain it. Yes, cropping really does reduce DOF if you use the conventional criterion; see for example Alan Horder, The Ilford Manual of Photography, fifth edition, Essex: Ilford Ltd, 1958, which gives the CoC criterion as explicitly affected by cropping, "1/1000 of diagonal of that portion of negative used to make print." For more related history, see my paper on DOF. Dicklyon (talk) 05:37, 30 October 2009 (UTC)

Yes, I follow the logic there. Of course in reality people don't use 6x9 or larger negatives merely for central portions that could just as well have been taken with a 35mm camera. (To do so would be perverse to put it charitably.) But the convention isn't an obvious one either. -- Hoary (talk) 10:10, 30 October 2009 (UTC)

I'd say it's uncommon, but hardly perverse. The most obvious reason for doing so would be a grab shot for which there wasn't time to change lenses, which I think is what you were saying above in the example using the Taj Mahal. JeffConrad (talk) 20:03, 30 October 2009 (UTC)

Right; and my paper discusses that issue quite a bit, esp. in the context of digital multi-megapixel cameras. Dicklyon (talk) 16:54, 30 October 2009 (UTC)

It seems to me that we have a source; if Stroebel isn't good enough, no one is. We additionally explain it in far more detail than any source I've seen, in the derivation but also in the main text. From Acceptable sharpness:

“The acceptable circle of confusion is influenced by visual acuity, viewing conditions, and the amount by which the image is enlarged (Ray 2000, 52–53).”

With regard to cropping, it seems pretty simple to start with assuming visual acuity and viewing conditions of the final image are constant, and work backward; the captured-image CoC is the then final-image CoC divided by enlargement. If the image is cropped, it requires greater enlargement. Again, it's really the enlargement, not the format size. I thought this was covered pretty well, but perhaps not. I hate to create a section with only a sentence or two, but if cropping is getting lost in the discussion on format, perhaps that's what we need to do. JeffConrad (talk) 20:03, 30 October 2009 (UTC)

Incidentally, I'm troubled by the way the introduction now allies deep focus with landscape and shallow focus with portraits. There's a lot of truth to this, of course, but I can think of plenty of shallow-focus landscape photos and deep-focus portraits. How about a change to "conventional landscapes" versus "publicity head shots" or similar? -- Hoary (talk) 05:15, 30 October 2009 (UTC)

I agree that not all landscapes have a large DoF, and not all portraits have a small DoF. But I think they're reasonable examples with which the reader can readily identify. And we qualify them with “may be”, which seems sufficient to me. If we say something like “conventional landscapes”, we really need to explain what that means; I think “publicity head shots” is a bit limiting, and there are probably a fair number of people who don't know what that means, either. And of course it means something else entirely to G. Gordon Liddy. If others also see this as a problem, we could remove the examples entirely, but I think at that point we might be better off removing the sentence entirely. I'm not sure that's a good idea. JeffConrad (talk) 06:32, 30 October 2009 (UTC)

Right. The (stereo)typical postcard is deep focus; I wonder if there are any PD postcards, especially the kitschy Alpine ones with Edelweiss or whatever in the foreground, snowy peaks in the back. And I do realize that attempts to be more informative can easily just add to confusion. -- Hoary (talk) 10:10, 30 October 2009 (UTC)

I can't think of any type of photography that always employs either deep or shallow DoF, especially the latter. I wouldn't necessarily assume that head shots always have shallow DoF. So I think almost any illustrative example would be open to criticism. We could try to add some qualifying words, but I'm not sure they'd be much more than weasel words. Another approach would be to borrow from the section Selective focus, and use something like

“In some cases, it may be desirable to have the entire image sharp, and a large DOF is appropriate. In other cases, a shallow DOF may be more effective, emphasizing the subject while de-emphasizing the foreground and background, perhaps giving only a suggestion of the environment.”

It's a bit more abstract, so I'm not sure it's as informative, but it assumes no stereotypes. JeffConrad (talk) 20:03, 30 October 2009 (UTC)

Assessment comment

The comment(s) below were originally left at Talk:Depth of field/Comments, and are posted here for posterity. Following several discussions in past years, these subpages are now deprecated. The comments may be irrelevant or outdated; if so, please feel free to remove this section.

As a new student to photography, I appreciate the examples that were given, instead of just text. It was helpful. Perhaps it is too simplistic, but I'd like another example such as: (This close up photo with a blurred background and clear subject, which was taken with the following settings on a Canon XXX: xxxxx) 67.169.221.47 (talk) 15:47, 30 March 2008 (UTC)

Last edited at 15:47, 30 March 2008 (UTC). Substituted at 14:39, 1 May 2016 (UTC)

Archive 1

Archive 2

Archive 3

Archive 4