why would DOF for a lens changes depending on sensor size ?

[Judobreaker]

Because DOF isn't only to do with optics.

Take a full 16MP image and look at it at 100%. Note the objects that are only just out of focus.
Now resize to web size (100px wide for example) and look again. More objects will appear in focus because of the data loss. The objects that were only just out of focus will now appear in focus thus the DOF will have increased.
It's not an extremely big difference probably but it is there.

 

[Dao]

Basically this means that by resizing an image in Photoshop you will alter the DOF.

Photoshop has no optics. How can resizing in Photoshop change the DoF?

When resizing an image and viewing the finial output at the same distance, the action of resizing affect the magnification. A fine point now appears to be a blur point. So part of the photos which "Appeared" to be sharp before are no longer sharp anymore. And that is how DoF is affected by the photo enlargement.


However, if you enlarge the photo and adjust the viewing distance, there is a chance to keep the same DoF.

 

[amolitor]

You're just changing the DoF by changing the target CoC. It's true but it's kind of a degenerate case and not what we're usually thinking about when we're thinking about DoF.

That said, the sensor size thing is arguably a red herring, since it's just a convenient intermediate step between the thing you're taking a picture of, and the final viewable result.

 

[Buckster]

Sorry to push the button again, but I still don't get it, and I really want to.

People keep talking about effective FOV and changing distance to subject or focal length to get the same composition and how that all changes DOF, and I get that. That's crystal clear. It's been crystal clear to me for more than 40 years. That's basic camera operation 101.

What still confuses me is when NOTHING changes but the sensor size. Put a 50mm lens at f/10 focused on a subject 10 feet away on a 7D, and you get one set of DOF and hyperfocal numbers. Put that same 50mm lens at the same f/10 focused on the same subject still 10 feet away on a 5DMKII, and you get a whole different set of DOF and hyperfocal numbers.

Ignore that the subject will appear larger or smaller in the composition and all that - doesn't matter for the point: WHY do the DOF and hyperfocal numbers change when NOTHING but the size of the sensor is different?

Everything else being equal, if I load 35mm film into my RB67, will the DOF and hyperfocal distance be different than when I'm shooting with 120mm?

 

[amolitor]

I *think* it's because the assumed degree of enlargement to the final print is different?

They're coming up with a different Circle of Confusion, and to be honest neither one of them makes much sense to me. They SEEM to have just taken 0.03mm as the reference CoC for 35mm film, and then multiplied by the relative sizes of the sensors appropriately. I'm not sure where they came up with the 0.03mm, though.

ETA: The wikipedia article suggests that the basis for this is a "standard" of 1500 dots along the diagonal of the image. I guess this is sort of reasonable. 1500 dots along the diagonal of the final image is 100dpi for an 8x10 print. It's insane for critical sharpness, but as a measure of 'ok, at this point it's not just soft, it's straight-up out of focus' then it seems maybe ok?

ANYWAYS. At 1500 dots along the diagonal, you come up with a CoC on 35mm film of about that 0.03 number.

For any other sensor, to get the same sort of resolution, you still have to be able to resolve 1500 dots along the diagonal, regardless of how big or small the sensor, so you scale the CoC appropriately.

It strikes me as all rather arbitrary, but I guess a rule of thumb is better than no rule of thumb.

 

[Ysarex84 78]

Sorry to push the button again, but I still don't get it, and I really want to.

People keep talking about effective FOV and changing distance to subject or focal length to get the same composition and how that all changes DOF, and I get that. That's crystal clear. It's been crystal clear to me for more than 40 years. That's basic camera operation 101.

What still confuses me is when NOTHING changes but the sensor size. Put a 50mm lens at f/10 focused on a subject 10 feet away on a 7D, and you get one set of DOF and hyperfocal numbers. Put that same 50mm lens at the same f/10 focused on the same subject still 10 feet away on a 5DMKII, and you get a whole different set of DOF and hyperfocal numbers.

Ignore that the subject will appear larger or smaller in the composition and all that - doesn't matter for the point: WHY do the DOF and hyperfocal numbers change when NOTHING but the size of the sensor is different?

Everything else being equal, if I load 35mm film into my RB67, will the DOF and hyperfocal distance be different than when I'm shooting with 120mm?

It's right there in the illustration you've posted. The math is being run with different values for the circle of confusion; .03mm for the 5D and .019mm for the 7D. That's the variable that changes the results. And as Amolitor is asking; where do they get those values?

The circle of confusion values rest on assumptions and averages; the averages are science but the assumptions are derived from common usage. One of those averages is the acuity of human eye sight. One of those assumptions is that you're going to enlarge the image and view that result from an appropriate distance. If you're practice deviates substantially from those assumptions then you can re-test the whole system and determine a personal value for the CoC that would either tighten or loosen your DOF limits.

Joe

 

[Ysarex84 78]

I *think* it's because the assumed degree of enlargement to the final print is different?

They're coming up with a different Circle of Confusion, and to be honest neither one of them makes much sense to me. They SEEM to have just taken 0.03mm as the reference CoC for 35mm film, and then multiplied by the relative sizes of the sensors appropriately. I'm not sure where they came up with the 0.03mm, though.

ETA: The wikipedia article suggests that the basis for this is a "standard" of 1500 dots along the diagonal of the image. I guess this is sort of reasonable. 1500 dots along the diagonal of the final image is 100dpi for an 8x10 print. It's insane for critical sharpness, but as a measure of 'ok, at this point it's not just soft, it's straight-up out of focus' then it seems maybe ok?

ANYWAYS. At 1500 dots along the diagonal, you come up with a CoC on 35mm film of about that 0.03 number.

For any other sensor, to get the same sort of resolution, you still have to be able to resolve 1500 dots along the diagonal, regardless of how big or small the sensor, so you scale the CoC appropriately.

It strikes me as all rather arbitrary, but I guess a rule of thumb is better than no rule of thumb.

You're exactly right there's a built-in assumption that the original media whatever it's size will be enlarged to a customary standard and viewed in a customary manner. I wouldn't call it arbitrary as much as derived from common practice. It's like in the lab when one of my students sticks his nose in a photo and says, "It's not really sharp," I smack him on the back of his head and say, "get your nose out of it and look at it from where you're supposed to look at it."

Joe

 

[fjrabon]

yes, I think something this thread points out is that there, in some ways, isn't a such thing as 'DoF'. There is an exact plane (well, a curved plane) that is exactly in focus, and everything outside of that is out of focus. What DoF does is measures how fast things go from "imperceptibly out of focus" to "very noticeably out of focus".

When something is enlarged, things that were previously "imperceptibly out of focus" often turn to "somewhat noticeably out of focus". You understand this very well if you've ever zoomed in on your camera's LCD several times to see if you nailed focus.

As Ysarex points out, what that means is that DoF calculators then make assumptions about how much you're going to enlarge your image, given the sensor size.

An important side point is that this often means that DoF calculators can be somewhat off if you crop very much. Usually it's not a very significant amount, but it is there.

 

[Derrel]

As the capture format is made progressively smaller, shorter and shorter focal length lenses are required to maintain an equivalent picture angle of view. This has been known for a loooong time now. On 6x6 (apprx 58x58mm) rollfilm, like that shot by a Hasselblad V-series, or a twin-lens camera like a Rolleiflex or Yashica-Mat, the "normal" angle of view lens has long been a 75mm or 80mm lens; on 24x36mm film, a 50mm lens has been normal; on really teeny-tiny film formats, an incredibly short lens, like say a 15mm lens, can be a "normal" angle of view lens. A "normal" lens is roughly the length of the diagonal measurement of the film format to which it is used with. Not "exactly" the diagonal of the capture size, but typically "roughly in the same ballpark" as the capture size's diagonal measurement.

Kodak years ago recognized a need for a very,very tiny film format, so that they could create a snapshot camera that would require no focusing, and which would give depth of field from very close range, all the way to infinity. Why? The then-current 126 and 110 film catridge-loading formats were popular with snapshooters, and many of them were trying to take pictures of close-range objects,like, say, flowers, babies, puppies, children, kittens,etc.,etc. from inside of three feet...and....their images were coming out out of focus, even though those cameras had lenses that were often operating at a fixed aperture that was pretty small, like say f/8 to f/11. Kodak needed a camera that had almost infinite depth of field.

Kodak decided to invent a brand-new, very small film format size. They went with a fairly short lens. On an 8mm x 11mm sized capture area, which is about what disc format was, a 15mm lens is just a little bit longer than the roughly 13mm diagonal of the film capture area; as I recall, Kodak went with a 12.5mm lens length on their disc cameras, and that gave decent depth of field from about one foot to Infinity. The interesting thing about the whole disc film format is that Kodak actually created an entire, all-brand-new film format, and a film "disc" with its own built-in dark slide--just to solve the problem of insufficient depth of field in snapshot cameras! This entire disc format invention story was related by the late Herbert Keppler in his days as a Pop Photo and Modern Photography magazine editor/author/columnist.

 

[raywzou]

Hi,

Thank you all for the good posts. It took me a while to read through and understand. There is a forth variable (in addition to focal-length + f-stop + camera-object distance) that affects the DOF, the CoC (Circle of Confusion) which is related to the sensor pixel density, or the magnification of the image by the sensor.
And to confirm few other points made.

Judobreaker, thanks for all the response, detailed explanation, and sample pictures. You hit the answer with magnification, and reaffirm the point by saying if not magnified it would be the same. So using say 20D vs 5D, 8mp crop vs 12mp ff (roughly same pixel density I think), same lens, same distance to object, and same f-stop, we would get same DOF. I believe Judobreaker help confirm this, correct ?

BlairWright, Helen B, and Judobreaker, thanks for the info about CoC, a new term I am learning, another interesting thing to read up on later.

Benco - there is another take on it, just take the pictures . I just recently acquired a DSLR camera, and the pictures sharpness and unsharpness is just amazing compared to my old P&S. The unsharpness part was unexpected, this new affect I am getting, want to get more of it, with full body in the frame. And I want to buy the right lens to do it. But after reading up and understanding more things optically, I realize it is harder to achieve the unshapness (Bokeh) of full person with crop than ff, as you would have to stand further away, thus changing the DOF value, reducing the bokeh. Or use short focal, again reducing the Bokeh. And longer focal lens gives better bokeh, no matter the sensor. With Crop longer focal lens makes you stand really far away. So in the end, understand the math, helps with picture technique and purchasing the right equipment. Your way works too of course, just try it in the store and trust the sales guy I just rather figure out the math myself and get exactly the gear I want.

amolitor, thanks for clarifying and multiple post as well. Is that right, the actual physical opening of say Canon 50mm f/1.8 set to f/2.8 on a crop is smaller than the same lens at f/2.8 on a ff ? Another concept that I didn't realize. So if the physical opening is the same, then same DOF ?

Ysarex, amolitor - interesting point of view, so the CoC can be thrown into the magnification bracket of stuff as focal length and distance to object. And that CoC rest on assumptions and averages, subjective measures. The 1500 dot and 100dpi print is interesting information. So if I achieve sharpness with that, image is consider very good.

KmH, Judobreaker, Dao, and amolitor - interesting spin on the photoshop and DOF. As Judobreaker explains and I think it actually helps indirectly explain how magnification affect DOF. If you shrink the image enough, it seems the DOF is now expanded. And DOF is defined as how it "Seems" within focus, front and back of the singular actual focal plane.

Buckster - thank you, you illustracted my point better than I did. And it is reaffirming to know others out there wanting an answer for this, especially an experience pro.

fjrabon and Helen B, thanks for pointing out that the CoC on the DOF calculator is an estimate and maybe accurate up to one decimal or two. And the emphasis on DOF definition, which Judobreaker also mentioned. Perceived as in focus within a specific range of the actual focal plane. And perception can be changed if magnification changes.

Derrel - interesting history of Kodak gears. The point about the DOF changed by change of focal length makes sense, though my OP intended to keep the focal length the same. Thus through all the replies, we discover the 4th variable in this equation, the magnification.

 

[amolitor]

A lens of the same focal length at the same aperture will have the same size hole through which light passes.

The point is that to get the same "picture" onto a smaller sensor at the same distance, you need a shorter focal length lens, so at the same aperture the hole will be smaller. Shorter focal length + same aperture = smaller hole.

 

[zcar21]

 

[Kolia]

Sorry to push the button again, but I still don't get it, and I really want to.

People keep talking about effective FOV and changing distance to subject or focal length to get the same composition and how that all changes DOF, and I get that. That's crystal clear. It's been crystal clear to me for more than 40 years. That's basic camera operation 101.

What still confuses me is when NOTHING changes but the sensor size. Put a 50mm lens at f/10 focused on a subject 10 feet away on a 7D, and you get one set of DOF and hyperfocal numbers. Put that same 50mm lens at the same f/10 focused on the same subject still 10 feet away on a 5DMKII, and you get a whole different set of DOF and hyperfocal numbers.

Ignore that the subject will appear larger or smaller in the composition and all that - doesn't matter for the point: WHY do the DOF and hyperfocal numbers change when NOTHING but the size of the sensor is different?

Everything else being equal, if I load 35mm film into my RB67, will the DOF and hyperfocal distance be different than when I'm shooting with 120mm?

Read this: Diffraction Limited Photography: Pixel Size, Aperture and Airy Disks

The 5D has larger pixels on its sensor, making it "blind" to smaller CoC that would otherwise cover multiple pixels on the 7D.

 

[Helen B]

Sorry to push the button again, but I still don't get it, and I really want to.

People keep talking about effective FOV and changing distance to subject or focal length to get the same composition and how that all changes DOF, and I get that. That's crystal clear. It's been crystal clear to me for more than 40 years. That's basic camera operation 101.

What still confuses me is when NOTHING changes but the sensor size. Put a 50mm lens at f/10 focused on a subject 10 feet away on a 7D, and you get one set of DOF and hyperfocal numbers. Put that same 50mm lens at the same f/10 focused on the same subject still 10 feet away on a 5DMKII, and you get a whole different set of DOF and hyperfocal numbers.

Ignore that the subject will appear larger or smaller in the composition and all that - doesn't matter for the point: WHY do the DOF and hyperfocal numbers change when NOTHING but the size of the sensor is different?

Everything else being equal, if I load 35mm film into my RB67, will the DOF and hyperfocal distance be different than when I'm shooting with 120mm?

Read this: Diffraction Limited Photography: Pixel Size, Aperture and Airy Disks

The 5D has larger pixels on its sensor, making it "blind" to smaller CoC that would otherwise cover multiple pixels on the 7D.

In the cases shown, pixel size is not used for the DoF calculation, sensor size is - based on 0.2 mm for 8x10 and scaled from there. The maximum acceptable circles of confusion stated are larger than the respective pixel sizes: 0.019 mm MACC for the 7D with a pixel size of 0.0043 mm; 0.030 mm MACC for the 5D Mk III with a pixel size of 0.0063 mm. Note that DOFMaster uses the same suggested MACC for all cameras of the same format, regardless of pixel size. Compare the Nikon D800E with the Nikon D3.

 

[Kolia]

Isn't that over simplified in their part to only use the overall sensor size ? How useful is that calculator if the actual limit ends up being our printer's dpi ?