DX to FX: effect on DoF?

[joe_c5]

<blockquote>

<p>Joe, don`t you take the circle of confusion into account here?</p>

</blockquote>

<p>Yes. If the focal length and f/ number are scaled with the sensor size, then the circle of confusion must be smaller on the smaller sensor by the same ratio in order to get the same depth of field.</p>

<p>I did not mention it because the user does not have to do anything to scale the circle of confusion, whereas they do have to select the correct lens and aperture.</p>

[jose_angel]

<p>I see it was an irrelevant question. Thanks.</p>

[bjørn rørslett]

<p>One can have equal perspective and field of view, but not magnification (because of unequal focal lengths), when two different formats are compared. Or you can have the equal magnification and perspective (same lens, same shooting position), but not equal field of view. Or you can have equal field of view, but not the equal magnification and perspective (shooting position not the same, but the same lens is used). Usually the underlying assumptions aren't clearly specified and accordingly, one can arrive at almost any conclusion. After all, when two formats *are* different, something *is* different and we cannot equalise over every dimension (had that been attainable, the formats would not have been different in the first place).</p>

<p>Add to this the fact that how we perceive 'DOF' depends on the viewing circumstances as I wrote in my first post.</p>

[epp_b]

<p>Take a look at the second calculator here: http://www.cambridgeincolour.com/tutorials/digital-camera-sensor-size.htm</p>

<p>That should give you your definitive answer.</p>

[joe_c5]

<blockquote>

<p>One can have equal perspective and field of view, but not magnification (because of unequal focal lengths), when two different formats are compared. Or you can have the equal magnification and perspective (same lens, same shooting position), but not equal field of view. Or you can have equal field of view, but not the equal magnification and perspective (shooting position not the same, but the same lens is used).</p>

</blockquote>

<p>Where magnification is used to mean the magnification of the lens only, then this is correct, however at subject distances which are not small compared to the hyperfocal distance the magnification has almost no effect on depth of field.</p>

<p>The total magnification from the actual object to the print does not change in my example above.</p>

<blockquote>

<p>After all, when two formats *are* different, something *is* different and we cannot equalise over every dimension (had that been attainable, the formats would not have been different in the first place).</p>

</blockquote>

<p>There are only three parameters that need to be equalized based on a fourth: the focal length, the f/number and the ISO. They only need to be scaled by the same one number based on sensor size. That does not seem to be difficult to do.</p>

<blockquote>

<p>Add to this the fact that how we perceive 'DOF' depends on the viewing circumstances as I wrote in my first post.</p>

</blockquote>

<p>Perhaps, but in the example above the viewing circumstances for the prints from the two cameras would be the same.</p>

[bjørn rørslett]

<p><em>"at subject distances which are not small compared to the hyperfocal distance the magnification has almost no effect on depth of field"</em> - and you really mean this?</p>

[Dieter Schaefer]

<blockquote>

<p>There are only three parameters that need to be equalized based on a fourth: the focal length, the f/number and the ISO.</p>

</blockquote>

<p>You've lost me there - where does ISO enter the DOF calculation? Or what am I missing here?<br /> It has always been my assumption that for the case were the same size print is to be obtained from an FX and DX image, the latter needs to be magnified more - and that that is the reason to use a smaller circle of confusion in the calculations. In fact, the underlying assumption seemed to me that with requiring a more strict 0.2mm for DX (than 0.3 for FX), the effect would be that after enlargement to the same print size, DOF would be equal for both.<br /> At this point, no assumption has been made as to what the other parameters are: was the same field of view chosen by either moving closer, or using a longer focal length for FX, or was the FOV left different and the images were taken at the same focal length and distance. <br /> Quite frankly, I can't see how a meaningful comparison can be made, since FX and DX are different formats and there is no way to get the exact same outcome without changing one of the parameters that will render the two images different.<br /> Using online calculators for DOF always comes with the problem that the underlying assumptions aren't revealed and aren't obvious - so it's very easy to draw wrong conclusions.</p>

<p>To me these DOF discussions are a fairly moot point - though interesting nonetheless. When DOF really matters to me, I simply select the aperture that gives me the result I want. Which one that is comes from experience - I do not carry a DOF in the field and the - now mostly missing - DOF scale on the lens was at most a coarse instrument to judge how large DOF might be. Never used hyperfocal distance either - if I want a 4mm (on film) subject at a one mile distance to the camera to be resolved, I just make sure that the aperture opening used wasn't larger than 4mm. The most I would ever loose with this method is a bit of DOF for the immediate foreground - never enough to really make a difference.</p>

[joe_c5]

<blockquote>

<p><em>"at subject distances which are not small compared to the hyperfocal distance the magnification has almost no effect on depth of field"</em> - and you really mean this?</p>

</blockquote>

<p>I do, yes. From http://en.wikipedia.org/wiki/Depth_of_field :</p>

<p>DOF ≈ 2Ncf²s² ⁄ ( f⁴ − N²c²s² )</p>

<p>This is an approximation for a subject distance that is large compared to the focal length, and you will notice that there is no “m” in the formula. N is the f/number, c is the circle of confusion, f is the focal length, and s is the subject distance.</p>

<p>At small subject distances, the approximation becomes</p>

<p>DOF ≈ 2Nc(1 + m ∕ P ) ⁄ ( m² )</p>

<p>where m is the magnification and P is the pupil magnification (1 for a symmetrical lens).</p>

 

<blockquote>

<p>You've lost me there - where does ISO enter the DOF calculation? Or what am I missing here?</p>

</blockquote>

<p>ISO has nothing to do with the DOF calculation, it is just to get the same exposure time despite the different f/ number. Without the same exposure time, the images might be different.</p>

<blockquote>

<p>Quite frankly, I can't see how a meaningful comparison can be made, since FX and DX are different formats and there is no way to get the exact same outcome without changing one of the parameters that will render the two images different.</p>

</blockquote>

<p>Again, changing the focal length, f/ number, sensor size and ISO, all in lockstep, will cause the images to be the same. There is therefore a way to get the same outcome.</p>

[Dieter Schaefer]

<p>Thanks for the clarification. In theory, maybe, in practice: changing f/number and ISO don't result in equal image quality - stopping down increases the effects of diffraction and increasing ISO the amount of noise in the image. Not trying to split hairs here - but the differences have just moved to a different area.</p>

<p>I'm not good at optics calculations but am wondering how much the outcome of the formulas you cite are influenced by the model's assumptions - which apparently call for a thin symmetrical lens. Not sure how much that model applies to the complex lens systems we see today. </p>

[bjørn rørslett]

<p>Model predictions are just that, and they are not valid data for testing a hypothesis. You need to do carefully conducted experiments and keep control of the dimensional analysis.</p>

[Dieter Schaefer]

<p>BTW, the wikipedia article you link to provides the ultimate answer to the question of whether DOF is larger or smaller for FX than DX - believe Bjorn provided the same answer earlier already:<br>

<strong>The comparative DoFs of two different format sizes depend on the conditions of the comparison; the DoF for the smaller format can be more than, less than, or the same as that for the larger format.</strong></p>

[sebastianmoran]

<p>Bjorn wrote:</p>

 

<blockquote>

<p>Don't bother [calculating]. 'DOF' is a perceived, not tangible or physical, quantity.</p>

 

</blockquote>

<p>That's a very important point. It's why I have never had much faith in DOF calculators or markings on lenses. When I want DOF, I use the calculators conservatively. When I want shallow DOF, I bracket and test.</p>

<p>On the other hand, I suggest the OP's question is, more or less, "given a specific shot, how much does DOF change in moving from DX to FX?"</p>

<p>That question can be answered easily. To get the same image on FX, you would use a different lens and close down by a little over one stop vs. DX. </p>

<p>An article I thought to be insightful suggests thinking about DOF this way: It depends on the diameter of the lens aperture in mm or inches. So, for lenses of the same angle of view, you multiply the f-number by the ratio of film/sensor sizes.</p>

<p>f/5.6 on DX = f/8.4 on FX, for lenses of the same angle of view.</p>

<p>I found a real DOF surprise using Medium Format: A normal lens f/2.8 shot in medium format (6x9cm) would have the very limited DOF of an f/1 normal lens shot on DX, if you could find such a lens. (My comparison: Mamiya Press 100mm f/2.8 on 6x9 negative vs. 28mm lens on DX. Big difference.)</p>

[joe_c5]

<blockquote>

<p>In theory, maybe, in practice: changing f/number and ISO don't result in equal image quality - stopping down increases the effects of diffraction and increasing ISO the amount of noise in the image.</p>

</blockquote>

<p>The noise from two different cameras with f/ number, focal length, ISO and sensor size all scaled will be much closer to each other than you might think. The photon shot noise will be identical, though the fixed pattern noise and read noise may not be.</p>

<p>Both cameras will also run into the diffraction limit at the same depth of field, even if this occurs at very different f/ number. The effective aperture of both cameras is the same. From my example above, 43mm / 22 = 1.95mm and 7.7mm / 4 = 1.93mm; the difference is just rounding error from when I picked the numbers.</p>

<p>One very significant difference is that lenses are generally much better at higher f/ numbers, even at the lens’s maximum aperture. If shallow depth of field is desired, this may be a big deal. If a wide depth of field is desired, then this may not matter.</p>

<blockquote>

<p><strong>The comparative DoFs of two different format sizes depend on the conditions of the comparison; the DoF for the smaller format can be more than, less than, or the same as that for the larger format.</strong></p>

</blockquote>

<p>I did not dispute that, however it will not be common for the smaller format to have less depth of field, only possible depending on lens and aperture selection.</p>

<blockquote>

<p>Model predictions are just that, and they are not valid data for testing a hypothesis. You need to do carefully conducted experiments and keep control of the dimensional analysis.</p>

</blockquote>

<p>True enough. If one were to do so, I suspect that the images would be highly similar for the conditions I suggested should be identical. Keep in mind that two cameras with identical sensor size, focal length, f/number and ISO could <em>also </em>take photos that were not identical, with the reasons being clearly unrelated to those four parameters.</p>

<p>I am curious what your hypothesis is to explain the differences you seem to expect. The biggest difference I would expect to see would be in sharpness between the two lenses, and even that might only show up in pixel peeping.</p>

<p>I’ll try to post at least one pair of photos tomorrow demonstrating similar photos with a full-frame DSLR and a compact digital camera.</p>

[jose_angel]

<p>Excuse me if I`m a bit dense tonight... could anybody tell me any situation where a D300 have less DoF than a D700?</p>

[joe_c5]

<blockquote>

<p>could anybody tell me any situation where a D300 have less DoF than a D700?</p>

</blockquote>

<p>If you used the same focal length lens with the same aperture at the same subject distance on both cameras, then the D300 would have less depth of field because the identical image on the sensor would be enlarged more in a print.</p>

<p>That could actually happen, but it is common to change either the subject distance or the focal length to get the same subject framing on both cameras.</p>

[john_morris4]

<p>Your lens selection should be perfect with a D700. I predict that you will really enjoy an FX camera with them. You'll likely find that the D700 also focuses more reliably where you want it to, so you won't need to stop down to cover focus errors as often.</p>

<p>If you use the 20 a lot on the D70, you might find that you would like a slightly wide lens for the D700. A variety of 28s that have been discussed here -- I think the 28mm f/2 is quite nice, if you don't mind manual focus. The 35mm f/2 Zeiss is also spectacular.</p>

[jose_angel]

<p>Thanks Joe. Definitely I was really dense. It was too late...</p>

[sunray1]

<p>John M., thanx for your prediction, makes it easier to finally lay down the money when the time comes :-)<br /> Good to hear that I can expect the better AF also to play a significant role in the whole focus/stopdown/DoF-thing.<br /> I actually think that I will enjoy the width of the 20 more on FX, as well of the wider/normal FoV of the 50. The 85 will take the position the 50 has now for portraits, and the 180, although a bit longer, will take the place the 85 has now with the DX D70</p>

[Kent Shafer]

<blockquote>

<p>If you used the same focal length lens with the same aperture at the same subject distance on both cameras, then the D300 would have less depth of field because the identical image on the sensor would be enlarged more in a print.</p>

</blockquote>

<p>I'm probably missing something, but that statement doesn't make sense to me. In digital photography, it would seem that pixels are pixels once they're in the computer, and the size of the sensor that recorded them doesn't make any difference. <em>In principle</em>, couldn't the 12 MP sensors in a P&S camera, a D300, and a D700 all record exactly the same RGB numbers for each pixel in the image of a given scene? In practice, of course, that would be difficult or impossible due to numerous factors—the unavailability of the huge apertures that would be necessary for very shallow DoF in the P&S, the lower noise generally produced by larger photosites on the larger sensor, the differences between different manufacturers' sensors and the electronics that support them, and many, many more.</p>

<p>The <em>number</em> of pixels available to make a print of a given size obviously makes a difference, but I don't see how the size of the sensor that produced them—by itself—is directly relevant.</p>

<p>This contrasts with film photography, where the number of "pixels" (grains) available to form a given image depends directly on the size of the film. With film, comparing film size with print size is very relevant. It's analogous to comparing pixel count to print size in digital photography.</p>

[joe_c5]

<blockquote>

<p><em>In principle</em>, couldn't the 12 MP sensors in a P&S camera, a D300, and a D700 all record exactly the same RGB numbers for each pixel in the image of a given scene?</p>

</blockquote>

<p>Yes, but it would require different focal lengths, where my example with the D300 and D700 was at the same focal length.</p>

<blockquote>

<p>the lower noise generally produced by larger photosites on the larger sensor,</p>

</blockquote>

<p>The larger pixels have an advantage, but it is higher dynamic range due to larger full well capacity rather than lower noise. For the same exposure time, the noise would be about the same because the number of photons captured would be about the same. To turn that into lower noise requires more light or longer exposures, which is sometimes an option and sometimes not.</p>

<blockquote>

<p>This contrasts with film photography, where the number of "pixels" (grains) available to form a given image depends directly on the size of the film.</p>

</blockquote>

<p>Actually, the number of grains depends on both the size of the film and the size of the grains. It is possible to use finer grain film for smaller film sizes. Film grain size is analogous to pixel size, film grain count is analogous to pixel count, and film size is analogous to sensor size.</p>

[jose_angel]

<p><em>"I'm probably missing something... " <small></small></em><small>(Yesterday I missed the same thing, I guess..<em>.)</em><br /></small> Kent, don`t forget that the D300 has a higher pixel concentration. The viewfinder is smaller, the subject size looks the same size (same magnification), but at a 100% size on a computer screen, both images will have the same height&width... making what we see at the DX cropped camera viewfinder like enlargered in comparison to the D700.</p>

[joe_c5]

<p>Here is an example of the same image with two different sensor sizes. The top is full frame digital at 105mm, f/8, 1/1000 sec, ISO 1600, and the bottom is a 2x center crop at around 52mm (the exif rounds to 50mm), f/4, 1/1000 sec, ISO 400.</p>

<p>The resolution is visibly much worse for the 2x crop, but that is expected with the lower pixel count, same anti-aliasing filter, and lens wide open at f/4. If they were actually different cameras instead of a crop from the same camera, the 2x crop camera probably will not lag as badly on resolution.</p>

<p>For this lens the distortion is also different, but the two images are similar enough to illustrate the point.</p><div></div>

[kelly_flanigan1]

<p>*****LOOK at it this way; the lens does NOT know what the size sensor; the size of film; or the shooters goals for the image.</p>

<p>That is why some many are confused.</p>

<p>Unless you have some goals; there can be no depth of field criteria.</p>

<p>All this was well understood 75 years ago; where Kodak in their data books had 1/2000 inch for 8mm cine; 1/1000 inch for 16mm cine; 1/500 inch for 35mm stlll; 1/200 inch for roll film folders. Large Format was often 2 arc minutes then as a criteria. For critical work they told you to use a tighter criteria.</p>

<p>Today folks are a bit dumber; they want a black and white answer to something that has a gradual change; thus assuming types get confused about DOF stuff.</p>

<p>The 50mm F1.4 Nikkor SC for Nikon here has been used on my Nikons 35mm still stuff; the tick marks on the lens are for a 1/30mm COC; like most all Nikon lenses since WW2 to the Nikon rangefinder era.</p>

<p>When I use the same 50mm F1.4 Nikkor SC on a 16mm cine camera; I use a tighter criteria for 16mm; ie 1/1000 inch.</p>

<p>When I use the same 50mm F1.4 Nikkor SC on a 8mm cine camera; I use a tighter criteria for 8mm; ie 1/2000 inch.</p>

<p>When I use the same 50mm F1.4 Nikkor SC on a dslr with a super tight pitch; I use a tighter COC criteria.</p>

<p>The lens really does not know what the purpose it is for the image; you do.</p>

<p>If the same 50mm F1.4 Nikkor SC on Nikkormat was just for 4x6 prints only; I can use a super loose/big COC criteria; bigger than the stock tick marks for 1/30mm.</p>

<p>The problem today is just like those weekly "How big can I enlarge" questions; most folks have NO goals; thus one really has no answer to these type questions.</p>

<p>If you end goal is just some Ebay or web images; and excessive tight COC criteria can looks just the same as a looser one.</p>

<p>If the end product is for a 12x48 FOOT billboard on a Texas Interstate; the viewer might be closest at 200 to 300 feet; one really only needs a VGA image. A photo.netter who dwells on DOF criteria that cannot be detected in the final 12x48 foot print will end up worrying about nothing; what cannot be seen.</p>

[Kent Shafer]

<blockquote>

<p>I'm probably missing something…</p>

</blockquote>

<p>I was. The assumption about same focal length and same aperture. Sorry for going off on a tanget—thanks for setting me straight.</p>

[john_morris4]

<p>Yeah, usually for questions like this, you don't assume same focal length -- assume same framing (or same focal length "equivalent", which is a term that only means anything if you are considering framing).</p>