Estimating depth of field

Started Feb 21, 2013 | Discussions thread
Anders W
Forum ProPosts: 19,064Gear list
Like?
Re: Estimating depth of field
In reply to Detail Man, Feb 28, 2013

Detail Man wrote:

Anders W wrote:

Detail Man wrote:

Anders W wrote:

richarddd wrote:

Anders W wrote:

richarddd wrote:

Anders W wrote:

richarddd wrote:

One general rule of thumb is to stop down as much as needed to get the desired DOF, even if some of the benefits of more DOF are lost due to diffraction. I don't see anything in this test that's contrary to that general rule.

The camera to first subject distance was about 3' and camera to last subject was about 6', which is much further than I normally think of as macro. Focus was on the first subject.

Is there another test that you would expect to yield a different result? If so, what?

Not really, as long as meeting certain DoF requirements is critical. In landscape shooting, for example, I might frequently be ready to sacrifice a bit in that regard as far as the foreground is concerned for better sharpness in the distance. I would also think, based on other tests that I have seen, that the loss due to diffraction would be more visible than it is in your test.

What other conditions are likely to result in more visible loss due to diffraction?

High contrast along with fine detail. A couple of examples here:

http://www.luminous-landscape.com/tutorials/understanding-series/u-diffraction.shtml

http://www.kenrockwell.com/tech/diffraction.htm

I'd certainly agree that you need a scene with fine detail in order to have a loss of fine detail due to diffraction

Right, but my implicit point was that your choice of test target was perhaps not ideal for the purpose of revealing loss due to diffraction.

Are conditions under which the general rule of thumb I gave above would be wrong?

As I said, not really, if the DoF requirements are critical, such as in a macro shot where it is very important to have some specific subject or part of a subject, e.g., a flower or a bug, entirely within the DoF. But there are other situations, e.g., some landscape shots, where DoF requirements are more gradual (e.g., the more of the foreground that is within the DoF, the better) but not critical and where it might make sense not to stop down further than say f/8 since additional DoF, while valuable would be worth less than max resolution for things in focus or already inside the DoF.

One other question that I have thought of, even before this thread, but to which I have no answer (yet) is whether there are situations where it might be better to try to increase DoF by reducing magnification (e.g., by backing off a little or using a wider FL) than by stopping down further and then crop a bit to obtain the same framing. Both strategies come with their own trade-offs and I am not yet sure which one makes more sense under which conditions. Perhaps the other strategy (to reduce magnification) never makes sense but I am not sure about that at this point. Do you have an idea?

It seems to me that in the strict sense of applying the DOF formulas (where diameter of COC is inversely proportional to the amount of enlargement), sensor/post cropping would not be a benefit.

In cases where the diameter of the diffraction light-spreading (due to diffraction through a circular aperture, in addition to the effects of "pure diffraction") exceeds the diameter of the COC (on the image-sensor surface, and derived from working backwards from the image being viewed with some particular visual acuity at some particular viewing-size and viewing-distance), and additional degradation of spatial-frequency resolution will take place (in an amount greater than the de-focusing that would be expected from the calculated diameter of the COC), as well. Such an effect would combine together with the "pure diffraction" to reduce the MTF.

When the distances between the Plane of Focus and what constitutes the relevant background subject-matter are relatively small, and the ratio of the Camera to Plane of Focus Distances are significantly increased, and the image shot from farther away is then post-cropped (in order to retain the same "subject magnification" at the Plane of Focus), it appears that the mathematical identity that I came up with for calculating Relative Background Blur may well explain the perceived results - whereas analysis by applying the DOF formulas would not predict a difference in the DOF that would serve to explain the effects seen.

Have a look at these posted examples, and my mathematical analysis. Let me know what you think:

http://forums.dpreview.com/forums/post/50659718

Hi DM,

When I try to think a bit about it, this isn't perhaps all that complicated after all. Consider the following example. We are shooting on MFT with a 50 mm lens at f/11 and need two stops more worth of DoF. So what to do? Put on a 25 mm and crop or stop down to f/22?

The change in lens-system Focal Length would require doubling the Camera to Subject Distance in order to maintain the same framing, completely negating the effects of the change in the value of Focal Length, erasing one "stop" of the DOF that you propose increasing, and also changing the "perspective" that exists between any subject-matter at differing distances.

Think again. Subject distance remains unchanged of course (or I would have mentioned it). So framing changes which is corrected by cropping the 25 mm image down to the same size as the 50 mm image. The perspective of the cropped 25 mm image will be the same as that of the uncropped 50 mm image. The situation is perfectly analogous to what we are discussing all the time when comparing FF to MFT.

If we choose the first option, we gain two stops worth of DoF according to well-known principles of equivalence.

But, you forgot to include the constant-framing assumption of "equivalence" (see above).

No I didn't. See above. When comparing sensors, the cropping to achieve the same framing is accomplished by the change of sensor size itself. If the sensor is the same, as in my example, we have to crop afterwards, as I mentioned.

Without the cropping, we would gain four stops but the fact that we have to half the CoC due to cropping brings it down to two.

Two "stops" of DOF, reduced to one "stop" of DOF due to the enlargement of frame-cropping.

No. See above.

At the same time, the cost we have to pay is that resolution drops to half the original value for things that are perfectly in focus.

Right.

If we choose the second option, we again gain two stops worth of DoF. Based on Lenstip test results, the loss of resolution for things perfectly in focus due to diffraction would be almost half but not quite (down from about 50 lp/mm to about 30 lp/mm at MTF-50%).

Sorry. It is unclear to me (from any of the above text) what the "second option" is. Please advise.

The only other option mentioned by my text is that of stopping down further, from f/11 to f/22, using the same lens (50 mm) as the one we started out with.

So it's a close call, but in this comparison the alternative of stopping down further appears to have a slight edge. At the same time, this edge might disappear if the comparison would be conducted for another set of apertures (say f/16 to f/32) or with a higher-resolution sensor (the loss due to diffraction should increase with sensor resolution). Finally, stopping down has the disadvantage of forcing a longer exposure time or a higher ISO so it's not inconceivable that there are situations where the other alternative (backing off or putting on a shorter FL) might be preferable.

What do you think?

Are there conditions under which stopping down results in less DOF due diffraction compared to not stopping down as much?

Yes, as I pointed out in an earlier post, there must even be a point where the DoF effectively disappears altogether if you stop down sufficiently and a point before that where you start losing rather than gaining. But I doubt that these points are practically relevant with the smallest aperture we can practically achieve with most lenses (f/22).

Did you mean to say, "a point where the" [spatial-frequency resolution] "effectively disappears altogether" ? A pinhole camera has DOF approaching infinity and resolution approaching zero.

See here (at the end of the post):

http://forums.dpreview.com/forums/post/50933547

Yes, I can see that you do recognize what is going on. My comment was directed towards the confusing nature of referring to the use of the phrases "the DOF" or "in the DOF" when (I think) that it makes more coherent sense to refer to the overall, composite spatial-frequency resolution.

It does indeed seem to be the case that when the size of an external point-source of light projected onto the image-sensor surface (due to diffraction through a circular aperture opening, and/or lens-system optical aberrations) exceeds the diameter of the chosen COC used in DOF calculations, then it is that diameter which matters.

The (seemingly counter-intuitive, but real) effects of such a situation as described above upon DOF calculations (which assume a constant viewing-size) would be to indicate a higher value of DOF (due to the existence of a larger sized "circle of cinfusion").

This brings to light the fact that spatial-frequency resolution always decreases whenever the "circle of confusion" on the image-sensor surface (by whatever means) increases. Since we are discussing qualities like "sharpness" and "acuity", etc., it makes sense (to me) to speak in terms of spatial-frequency resolution of subject-matter represented in various portion of the image-frame - instead of using the vernacular associated with a calculated DOF based upon a chosen COC diameter that may in many high valued F-Number situations be "trumped" by diffraction (or "trumped" as well by lens-system aberrations even at lower valued F-Numbers).

DM ...

I am not foreign to other perspectives. I just tried to answer Richard's question on the terms it was asked.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +21 more
Reply   Reply with quote   Complain
Post (hide subjects)Posted by
prayNew
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow