What if 36mp DSLRs were around the corner?

[Sheldon N]

You can say that a camera has possibility of less DoF, since the minimum DoF is limited by the lens maximum aperture / focal length combination which tend to be larger for larger formats.

You could also say that a camera has the possibility of less DOF, simply by virtue of the fact that you can make a larger print with it. For example... the D3X has the possibility of less DOF than the D3S (all else equal), simply because you could make a larger print with the D3X and subject it to greater scrutiny than the D3S. 

[ErikKaffehr]

Hi,

My take on the issue is that DoF simply does not exist...

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures

or

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1

Best regards
Erik

You could also say that a camera has the possibility of less DOF, simply by virtue of the fact that you can make a larger print with it. For example... the D3X has the possibility of less DOF than the D3S (all else equal), simply because you could make a larger print with the D3X and subject it to greater scrutiny than the D3S. 

[Sheldon N]

My take on the issue is that DoF simply does not exist...

Not sure I'm following you on that one... looked at the links but am not sure what you are concluding based on your tests.

[ErikKaffehr]

Hi,

What I mean is that there is only one plane of actual focus, so the images need to be composed so that critical focus is achieved in the important area. DoF scales in general are based on small prints viewed at short distance, and trying to use DoF scales will result in essentially neither subject or background going to be critically sharp.

With the standard definition of DoF every single image in the article is sharp!

The article was really caused by a photographer complaining that his Pentax 645D was out of focus at infinity when he focused on a boat something like 200-300 yards away with a 150 mm lens at f/9.5, so I wanted to find out how things worked. The other observation was that when I shoot a group of people with my Sony Alpha 900 critical focus is incredibly short, much less than I would have expected.

What I'm doing now is to focus on the main subject, and try to stop down if needed. Stopping down beyond f/16 is nothing I like to do. Sometimes I do a focus bracket and merge with helicon focus, but that does not work with all lenses and subjects.

Just a point, if someone buys a 24 or 36 MP camera I presume that he/she would like to make full use of the resolution, and that means exact focus. If the full resolution is not utilized the person would probably be better of with a 12MP camera as it may work better in low light.

Best regards
Erik

Not sure I'm following you on that one... looked at the links but am not sure what you are concluding based on your tests.

[torger]

"DoF does not exist", that's an interesting way of saying that always put your focus on something important in the scene and assume that everything else will be slightly less sharp. It is a good approach, and this is what I actually use in practice most of the time.

For example if hyperfocal say I can focus on 6.6 meters to get sharp from 3.3 and my closest foreground is at 5 meters and is rather low contrast, it is likely that I focus on something interesting in the background instead of doing hyperfocal. Even with my definition (that is don't let CoC exceed airy disc when sharpness is a bit diffraction limited), the edges of the DoF is slightly less sharp than in the focal plane. It is often more valuable to have one detailed key object super sharp and some parts of the picture a bit unsharp than having the whole picture semi-sharp.

In this particular example if the focus is set somewhere between 6.6 and 20 meters near DoF edge is 3.3 - 5 meters, that is due to the slow increase of DoF near edge when increasing focus distance it is often common to find a suitable focus distance that focuses on something important but still give more foreground than focusing on something hundreds of meters away or infinity.

With high resolution cameras which make large detailed prints possible, DoF is a risky concept. The old definition with 25 um CoC will not be adequate on a 36 megapixel 36x24mm sensor. If you really want to use a fixed size CoC for the DoF calculations 11-15 um is more reasonable for small pixel DSLRs, which corresponds to f/8-f/11 airy disc.

[BernardLanguillier]

Another aspect to take into account is lens design. All the models for DoF are based on extremely simplified designs. Actual lenses do behave differently.

The drop of focus as a function of the distance to the perfect plane of sharpness in front and behind the plane of sharpness will therefore be impacted by the design.

Cheers,
Bernard

[torger]

Another aspect to take into account is lens design. All the models for DoF are based on extremely simplified designs. Actual lenses do behave differently.

The drop of focus as a function of the distance to the perfect plane of sharpness in front and behind the plane of sharpness will therefore be impacted by the design.

Cool, I did not know that. Do you know how large differences can be? Is it quite large, or could it be ignored? A 100% perfect model is not feasible to make of course, the CoC=airy disc model is rather approximate too (airy disc varies with wave length, shape is not same as CoC, two equally sized blurs added forms a slightly larger blur, lens resolving power can be considerably more limiting than diffraction in corners etc), but perhaps the lens DoF differences are so big that they should be taken into account in a DoF model at this approximation level.?

[BernardLanguillier]

Cool, I did not know that. Do you know how large differences can be? Is it quite large, or could it be ignored? A 100% perfect model is not feasible to make of course, the CoC=airy disc model is rather approximate too (airy disc varies with wave length, shape is not same as CoC, two equally sized blurs added forms a slightly larger blur, lens resolving power can be considerably more limiting than diffraction in corners etc), but perhaps the lens DoF differences are so big that they should be taken into account in a DoF model at this approximation level.?

There is very little data available unfortunately. You might recall that Nikon used to sell 105 and 135 DC lenses where this effect could be controlled?

http://www.jimgamblin.com/blog/?p=490

Cheers,
Bernard

[Ben Rubinstein]

This following image is just 30.9 megapixels from a stitch. Shot it yesterday. Did the rest of the frames then waited for a bird to land for the final frame (there were of course millions of birds while I was shooting the other frames but had to wait 20 minutes for this one to land). I'd kill to be able to get this level of detail and tonality with a single shot. However I was using a very sharp prime lens for each frame. All the goodness of that lens was being used for every part of that final image. Even though I had to stop down to f20 for the DOF (just enough) the diffraction based on a 12 megapixel sensor wasn't that bad, especially when you had so many frames making up the final image.

If I was to take the same image with a higher megapixel camera the diffraction would be far worse although I'd not have to stop down the same amount due to using a shorter lens but all the resolving power of the lens would be used up in one go and I'm not sure that I would be getting the same resolution at all. Infact when I shot with a 1Ds3 I certainly didn't get the same resolution from a single shot at 21 megapixels as I did from stitching my 5D to the same 21 megapixel final image.

On the other hand and although I'm about to buy one for studio use I haven't done much landscape testing, a Leaf 40 megapixel back seems to come much closer in a stitch vs straight image comparison. In both cases I believe the lens is the limiting factor. When stitching you are using the full power of the lens X6 or X10 (or whatever) rather than X1. No doubt the gap would close rapidly when using a digitar or the like on a well calibrated Tech Cam.

However. Few factors. I never believed that the tonality would be better on the 1Ds3 than my 5D. Not for a second. Impossible, the pixels were tiny in comparison. Boy was I wrong. The 1Ds3 has the best tonality of any DSLR I've ever seen or played with. Including the 5DII by a large margin. Tonality counts, a lot, it counts even when you're shooting a 1962 Takumar 50mm f1.4 wide open like I'm doing at the moment, when resolution matters *bleep* all but tonality is the name of the game. Secondly, you can stitch the higher megapixel camera with just 2 or 3 frames and close any gap that there might have been due to the lens. Lot faster, lot easier, both when shooting and processing. Thirdly and most importantly. The images below were single shots which for various reasons such as time, logistics or location, I was unable to stitch. There is a limit to what size I can print these because I only had 12.7 megapixels and with all the very best technique, lenses and stability, at some point you just run out of printability for fine art. With the higher megapixels you have the choice, no it might not be ideal but it will be far closer to what you would have got with a 1/3 or 2/3 of the megapixel count. You have the choice. I'd love a 1Ds3 for my work but a camera with over 20 more megapixels and significantly smaller and lighter, oh and cheaper than even the current 2nd hand price of a 1Ds3, certainly makes you think!

_

That's not even beginning to talk about studio stuff...

[kers]

Another aspect to take into account is lens design. All the models for DoF are based on extremely simplified designs. Actual lenses do behave differently.
The drop of focus as a function of the distance to the perfect plane of sharpness in front and behind the plane of sharpness will therefore be impacted by the design.
Cheers,
Bernard

I took three photographs the other day with three different 24mm lenses in my room. The were focussed with liveview on the same spot and i used the same aperture of course.
 
The result in difference of the depth of field was striking. It seemed one lens had even more DOF than the other -
probably caused by field curvature and the fact that one lens is a bit sharper than the other, therefore appearing to have less DOF.

[torger]

Ben, really interesting post, and nice photos!

Stitching is like using a larger sensor, exposing one part at a time. I you want large DoF - what you gain (using a large sensor with larger pixels) from being able to use smaller aperture without too much diffraction you lose in the need of using a longer focal length. Mathematically it is exactly the same. An example: for maximum DoF a 36x24mm 36 megapixel sensor with 24mm lens at f/8 is (about) the same as a 36 megapixel 48x36mm medium format sensor with 35mm at f/11.

I think things like tonality is quite hard to judge, I don't really know what to look for. When I do a print I spend quite some time in post-processing to tune the tonality. Poor local contrast can often be quite effectively compensated for, so I don't really know what I actually need here. For landscape it seems like I would prefer a sharp lens with a bit less contrast than a less sharp lens with a bit higher contrast, since I can compensate in post, but I'm not sure. I have used too few lenses to know. What I do note about tonality is that it differs between sensors how much color information there is in the shadows.

[Ben Rubinstein]

Stitching is like using a larger sensor, exposing one part at a time. I you want large DoF - what you gain (using a large sensor with larger pixels) from being able to use smaller aperture without too much diffraction you lose in the need of using a longer focal length. Mathematically it is exactly the same. An example: for maximum DoF a 36x24mm 36 megapixel sensor with 24mm lens at f/8 is (about) the same as a 36 megapixel 48x36mm medium format sensor with 35mm at f/11.

I suppose it's just that with the huge resolution of stitching the lack of DOF becomes more apparent but in that case it would be exactly the same with a higher resolution sensor or in other words, your camera may have 36 megapixels but only up to about f4 or so. That said, if you have to shoot a scene with a D3x at f22 and a D800 at f22 the latter will still give you significantly more resolution for all that it won't be as sharp as it might have been at a lower aperture. Photography is about compromises and sometimes you just gotta stop down...

[BernardLanguillier]

Ben, really interesting post, and nice photos!Stitching is like using a larger sensor, exposing one part at a time. I you want large DoF - what you gain (using a large sensor with larger pixels) from being able to use smaller aperture without too much diffraction you lose in the need of using a longer focal length. Mathematically it is exactly the same. An example: for maximum DoF a 36x24mm 36 megapixel sensor with 24mm lens at f/8 is (about) the same as a 36 megapixel 48x36mm medium format sensor with 35mm at f/11.

Agreed with what you write.

Now, there is something else that comes in the picture, and that is DoF stacking.

With the first example above, it is fairly easy to change the focus point at most areas in the picture to get perfectly sharp images at f8, there is only the area separating the tree from the building behind where DoF stacking would be needed.

Cheers,
Bernard

[ErikKaffehr]

Hi,

Diffraction is relative benign to sharpening as the unsharpness it causes is not a uniformly lit disk but more like a cone: , so although it reduces "microcontrast" it still contains some information. So stopping down is not to bad.

Best regards
Erik

I suppose it's just that with the huge resolution of stitching the lack of DOF becomes more apparent but in that case it would be exactly the same with a higher resolution sensor or in other words, your camera may have 36 megapixels but only up to about f4 or so. That said, if you have to shoot a scene with a D3x at f22 and a D800 at f22 the latter will still give you significantly more resolution for all that it won't be as sharp as it might have been at a lower aperture. Photography is about compromises and sometimes you just gotta stop down...

[bjanes]

None of this seems inconsistent with what I'd outlined earlier.  In the standard measurement for DOF, assuming two images of the same field of view and same aperture, the DOF will be the same.  But at the image plane the digital image will have greater DOF. 

Except that you are not taking into account how sensors scale and have not defined aperture. By aperture, do you mean the f/number (the usual metric in practical photography) or the actual diameter of the entrance pupil? With the same f/number, the smaller sensor will have a greater depth of field in the print, whereas with the latter, the depth of field in the print will be the same. Have you read and comprehended Roger Clark's article that I referenced?

The COC for the image plane needs to be defined in relation to the sensor size. David Jacobson states, "Another rule of thumb is c=1/1730 of the diagonal of the frame, which comes to .025mm for 35mm film. (Zeiss and Sinar are known to be consistent with this rule.)" This accounts for the magnification factor to which you have referred.

Regards,

Bill

[Ben Rubinstein]

Agreed with what you write.

Now, there is something else that comes in the picture, and that is DoF stacking.

With the first example above, it is fairly easy to change the focus point at most areas in the picture to get perfectly sharp images at f8, there is only the area separating the tree from the building behind where DoF stacking would be needed.

Cheers,
Bernard

To be 100% honest, life is too short. I shot a 3 frame bracket of these frames but never used them in the end, I had more than enough information to work with. Once you start playing the HDR/Focus stacking games, particularly the latter, outdoors with changing light and moving foliage, well, you have to be a more patient man than I am.

[bjanes]

I suppose it's just that with the huge resolution of stitching the lack of DOF becomes more apparent but in that case it would be exactly the same with a higher resolution sensor or in other words, your camera may have 36 megapixels but only up to about f4 or so. That said, if you have to shoot a scene with a D3x at f22 and a D800 at f22 the latter will still give you significantly more resolution for all that it won't be as sharp as it might have been at a lower aperture. Photography is about compromises and sometimes you just gotta stop down...

f/4 is a bit conservative for a 36MP full frame dSLR. The pixel size would be 4.9 microns and the Airy disc at f/4 with green light is 5.2 microns. If you consider that diffraction comes into play when the Airy disc is 1.4 x the pixel pitch (6.9 microns), you could use f/5.6 with minimal loss. The Airy disc is 7.2 microns at f/5.6.

Are you sure that the D800 at f/22 would have more resolution than the D3x at the same aperture. The Airy disc at f/22 is 28.5 microns and the Dawes limit at f/22 is 91 lp/mm and the Nyquist for the D800 would be 102 lp/mm and 85 lp/mm for the D3x.

Regards,

Bill

[ErikKaffehr]

Hi,

My test here, done with a Sony Alpha SLT 55, should be quite similar to a 36 MP full frame. The left column corresponds to exact focus using live view. Sharpness seems optimal at f/5.6, red circle is the diameter of the Airy disc and green circle corresponds to CoC (Circle of Confusion) at the given defocus.

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1  (Updated: 2011-11-27, added the link)

The Sony Alpha has a crop factor of 1.5 at 16 MP so it corresponds to 36 MP (1.5x1.5x16 = 36).

Best regards
Erik

f/4 is a bit conservative for a 36MP full frame dSLR. The pixel size would be 4.9 microns and the Airy disc at f/4 with green light is 5.2 microns. If you consider that diffraction comes into play when the Airy disc is 1.4 x the pixel pitch (6.9 microns), you could use f/5.6 with minimal loss. The Airy disc is 7.2 microns at f/5.6.

Are you sure that the D800 at f/22 would have more resolution than the D3x at the same aperture. The Airy disc at f/22 is 28.5 microns and the Dawes limit at f/22 is 91 lp/mm and the Nyquist for the D800 would be 102 lp/mm and 85 lp/mm for the D3x.

Regards,

Bill

[RFPhotography]

You know, Janes, these continual ad hominem remarks really are becoming tiresome. 

[bjanes]

You know, Janes, these continual ad hominem remarks really are becoming tiresome. 

At the image plane smaller sensored cameras have more DOF but in a print not so because those smaller pixels have to be 'magnified' more to make a print of the same size as a larger sensored camera

True.  But those smaller pixels will need to be 'enlarged' more to make that 20" print.  That additional 'enlargement' causes degradation.

None of this seems inconsistent with what I'd outlined earlier.  In the standard measurement for DOF, assuming two images of the same field of view and same aperture, the DOF will be the same.  But at the image plane the digital image will have greater DOF. 

How does one reason with another who refuses to use reason?

Regards,

Bill