[MossyRocks]

Hah, so this explains why I have to set hyperfocal for f5.6 or so when shooting at f8 on my Pentax-M lenses!

I do the same for all of my glass that has the DoF gauge so it is nice to know I'm not the only one.

[photoptimist]

I also find myself in the unfortunate position of not entirely agreeing with @Photoptimist, because I think he's oversimplifying how downsampling works in his example. But having said that, I know from experience that disagreeing with him can end up with me having egg on my face, because he actually knows a heck of a lot more than I do.

I do have to point out to Photoptimist though: The Windows 7 computer that I do my photo editing and viewing on has got no idea what physical screen size I'm using, but only the pixel resolution of the monitor attached. Nor does it have any idea what physical size sensor was used to take any particular photo, but just the pixel resolution. So any relationship between the physical sensor size and the physical viewing size is incidental rather than direct, caused only by the fact that each pixel on the monitor is far bigger than the pixels on the camera sensors (and the pixels in the monitor image are the result of downsampling too). And when we're talking about prints, I think we're better off sticking to non-interpolated prints of the full sensor resolution, because resampling algorithms are a factor better left out of a discussion like this.

LOL!

First, regarding downsampling: Yes, downsampling affects image sharpness and it affects the distances between image features and sizes of features if those distances are measured in pixels. A 2:1 downsampling will reduce pixel distances and sizes by half.

Second, regarding Windows and screens: You are probably right that neither Windows nor its apps know the physical pixel pitch of the monitor. It's possible that some clever apps do know the physical sensor size by reading the EXIF data, looking up the camera model #, and adjusting their automagical algorithms based on sensor size, lens, aperture, etc. But for purposes of DoF, Windows does not need to know this.

The bigger issue is not to lose sight of what DoF is really about. DoF is about a person looking at a flat 2-D image (print, monitor, or projected slide) reproduced from a flat 2-D photosensitive system (a refracting lens in front of some film or silicon array) of a 3-D scene and then judging whether subject, foreground, and background are equally sharp. DoF is not about absolute sharpness but about the perceptual variation in sharpness with variation in subject distance. DoF is about how the system renders a 3-D world into a 2-D image and how the viewer judges the relative sharpness of different foreground and background objects relative to the subject in the plane of focus.

So, we have 4 key parts to the chain from the photographed scene to the photograph viewer:

1) The 3-D scene: A bunch of objects at various distances to the lens and camera. Certain objects are the subject and others might range from the foreground in front of the subject to the background behind the subject.

2) The 2-D photosensitive imaging system: A lens of some focal length and aperture that refracts and projects rays of light back to a 2-D photosensitive medium (film or silicon) of some overall physical size and grain size or pixel pitch. Rays of light from the subject at the plane of focus tend to converge almost to a point on the film or sensor. Rays of light from the foreground objects tend to converge almost to a point BEHIND the film or sensor which means those rays of light are a blurry circle where they hit the film or sensor. Rays of light from the background objects tend to converge almost to a point IN FRONT of the film or sensor and then diverge which means those rays of light are a blurry circle where they hit the film or sensor. For various messy mathematical reasons, the physical distances, physical focal length, and physical aperture all affect the physical size of the blur circle which can be measured in millimeters or microns or whatever.

3) The 2-D image output device: A chain of processes that converts the captured camera image into viewable output of some physical size. The output could be a print, computer screen, TV, projected image, billboard, or whatever. The process may include scanning, resampling, interpolation, optical magnification, and output pixel pitch. Regardless of all the intermediate details, the overall point is that there is a scale relationship between physical distance in the final output and physical distance in the original film/sensor input. That is, 1 physical millimeter on the film or sensor comes out as N physical millimeters in the viewable output.

4) The person: The visual acuity of the person and their distance to the output image determine how they judge the different parts of the image to be more or less sharp.


To get the right answer for DoF, the photographer really needs to start with the person viewing the output image. That viewer might be pixel peeper at close range on a computer screen, an art gallery customer browsing prints on the wall, or they might be driving by a billboard showing a beautiful breakfast of eggs, bacon, toast, coffee, for a restaurant in the next town. Those viewing conditions will determine the acceptable CoC for the 2-D output image. In turn, the 1:N relationship between the camera and the output will determine the CoC on the film or sensor. Next, it's up to the photographer to decide what they want "in focus" versus "out-of-focus." Once the photographer decides the desired range of in-focus foreground-to-background distances (with all else being out-of-focus), they then need to use a DoF calculator and the estimated CoC to crunch the math and get the best aperture and subject distance setting that renders the 3-D scene into the desired 2-D output.

For most "normal" print viewing conditions and most modern high-resolution cameras, the sensor pixel size doesn't really affect CoC. Obviously, the pixel peeping scenario is different. There, the CoC might be only a pixel or two especially for pixel shift or super-resolution images viewed by the most persnickety pixel peepers.

Note to Photoptimist: I'll take the eggs on my face fried sunny side up, with nice crispy brown edges please. No doubt you'll be serving them up to me soon, with some math on the side.

I can only hope that both the near and far edges of this big egg of a post are crisp. But that will depend on your perception, your viewing distance from the post, and how the internet and Windows resamples or magnifies these words.

[Dartmoor Dave]

I can only hope that both the near and far edges of this big egg of a post are crisp. But that will depend on your perception, your viewing distance from the post, and how the internet and Windows resamples or magnifies these words.

Thanks, and I'm happy to go with your beautifully clearly put explanation. You've convinced me that we can meaningfully make a connection between sensor size and final viewing size when it comes to depth of field, and I'm grateful for that.

Mmm. . . eggs. . . yummy. . .

[swanlefitte]

Reading everything again and everything new I think I am understanding much better.
Ultimately pixel pitch really only matters because we are viewing larger images at the same distance, or put another way we pixel peep more. So coc is tied to pixel size because the more pixels the bigger the perceived image if ppi and viewing distance is set. If ppi is not set then we get into sampling.
Does this sound right?

[jack002]

Has anyone pointed out (I read all of post 107) that the crop factor is only a thumbnail "kind of similar" comparison? Useful only to get an idea of what kind of thing to expect?

To illustrate. You have a 50mm lens on a FF camera. You shoot a scene and wish to duplicate it on your APS-C camera (check me here, but that is not 100% literally possible)
The crop factor (I know mine is from what I read for it) is 1.6
So 1.6 is how much more the FF is than the APS-c, so if you do 1/1.6 you get 0.625, and 0.625 x 50 is 31.25
Take that same scene with your APS-C camera with an 30mm lens
Now compare both shots. I think you will find the lines are less bent and straighter on the FF shot.
This is why many are saying a given focal length is the given focal length.
Technically, you are just cropping down a larger image with the crop sensor
[think about how lines look on a wide angle lens compared to a telephoto]

If this was mentioned before, sorry, I skimmed this and didn't see.
If somewhat minor bending of lines is not important to you, then this can all be ignored. If you shoot buildings, you'll care more than if you shoot flowers or landscapes.

[GUB]

Reading everything again and everything new I think I am understanding much better.
Ultimately pixel pitch really only matters because we are viewing larger images at the same distance, or put another way we pixel peep more. So coc is tied to pixel size because the more pixels the bigger the perceived image if ppi and viewing distance is set. If ppi is not set then we get into sampling.
Does this sound right?

No the coc is not tied to pixel sized in the calculator because the calculator criteria is for a given print size whatever the pixel downsampling arrives at for that size.

So just to clarify again I believe the questions we are asking are;

Is the dof calculator technically accurate (for comparative purposes - we know it is a blunt tool)?
And if it is accurate then what is the reason for changing the coc for different formats.?

[GUB]

And my answer is yes it is technically correct
And the coc is applied because the FF image is a higher resolution image.

And if you think about it this is why the calculators were set up like this in the film days. The coc was not applied because of grain size. If this was the case coc would have varied with the iso. It was applied because the larger format simply had more lines (dots) of information on it. (The lpm of the lens when it is applied to different formats).
You guys are not taking the optics into it.

---

Last edited by GUB; 10-07-2019 at 12:48 PM.

[swanlefitte]

No the coc is not tied to pixel sized in the calculator because the calculator criteria is for a given print size whatever the pixel downsampling arrives at for that size.

The coc is recorded at the pixel level. Without sampling the same size sensor gives a larger photo at the same ppi print level the more pixels there are. That means already one must either view at a farther distance or reduce the coc when recorded no?

[GUB]

The coc is recorded at the pixel level. Without sampling the same size sensor gives a larger photo at the same ppi print level the more pixels there are. That means already one must either view at a farther distance or reduce the coc when recorded no?

But that is outside the parameters of the dof calculator.

[bilybianca]

I know that andromeda frames up nicely with my 400mm on APS-C but if I were to switch to full frame I would need a 600mm lens to get a similar framing. If instead I jumped to u4/3 I could get similar framing with a 300mm lens. If I were to jump up to a 645Z because apparently I won the lottery I would need like an 800mm lens to get similar framing. Finally if I was dumb and wanted to try astrophotography with a Pentax 67II I would need something like a 1200mm lens to get similar framing.

Or you can just use your 400mm lens on any of the different cameras, and then crop the picture to APS-c size in post processing. Pixel density aside, a bigger sensor always contains the smaller area of a smaller sensor. Just crop if you want a smaller field of view.

Kjell

[swanlefitte]

But that is outside the parameters of the dof calculator.

The dof calculator is saying on film enlargement by 8x typically. On digital what is 8x? The whole 8"x12" is based on 36x24mm(1"×1.5") film

Digital enlargement without sampling is based on pixel enlargement. 8x300=2400. With 2800 we shrink the image, pixel, or sample and with 2000 we enlarge.
Shrink the image you must shrink viewing distance. Shrink the pixel and you shrink the coc.
The third option is downsampling. Computational manipulation of the pixels does seem to be different. Now I am confused again. At least the confusion has been pinned down.

[GUB]

The dof calculator is saying on film enlargement by 8x typically. On digital what is 8x? The whole 8"x12" is based on 36x24mm(1"×1.5") film

Digital enlargement without sampling is based on pixel enlargement. 8x300=2400. With 2800 we shrink the image, pixel, or sample and with 2000 we enlarge.
Shrink the image you must shrink viewing distance. Shrink the pixel and you shrink the coc.
The third option is downsampling. Computational manipulation of the pixels does seem to be different. Now I am confused again. At least the confusion has been pinned down.

All the calculators I have seen have inputs for; format, subject distance , focal length and aperture and the print size / viewing distance is based on an accepted standard.

[Dartmoor Dave]

The point that answered my own personal objections was:

3) The 2-D image output device: A chain of processes that converts the captured camera image into viewable output of some physical size. The output could be a print, computer screen, TV, projected image, billboard, or whatever. The process may include scanning, resampling, interpolation, optical magnification, and output pixel pitch. Regardless of all the intermediate details, the overall point is that there is a scale relationship between physical distance in the final output and physical distance in the original film/sensor input. That is, 1 physical millimeter on the film or sensor comes out as N physical millimeters in the viewable output.

I can accept that there are ways in which the final viewing size is scalable back to the original sensor size in digital photography, even though there is no optical enlargement taking place in the same way that it does with film, and even allowing for resampling and the fact that digital data itself has no physical size. It's quite a subtle distinction, but it does the job for me within the context of this thread.

[photoptimist]

Reading everything again and everything new I think I am understanding much better.
Ultimately pixel pitch really only matters because we are viewing larger images at the same distance, or put another way we pixel peep more. So coc is tied to pixel size because the more pixels the bigger the perceived image if ppi and viewing distance is set. If ppi is not set then we get into sampling.
Does this sound right?

That's a fair statement of how the modern practice of digital pixel peeping affects CoC. Thus a pixel peeper who gets a camera with more megapixels would change their CoC setting.

However, there are some photographers who still think in terms of a specific print size which can include either paper print or viewing on a large screen monitor with the image scaled (downsampled) to fit. For them, the CoC is defined by the print size and viewing distance which then relates back to sensor CoC which might happen to be many pixel across. A print maker who gets a camera with more megapixels would not change their CoC setting for their existing print sizes but they might look into larger print sizes that imply more magnification and imply a smaller CoC at the sensor.

[GUB]

The point that answered my own personal objections was:




I can accept that there are ways in which the final viewing size is scalable back to the original sensor size in digital photography, even though there is no optical enlargement taking place in the same way that it does with film, and even allowing for resampling and the fact that digital data itself has no physical size. It's quite a subtle distinction, but it does the job for me within the context of this thread.

Naaaa wipe that egg off your face Dave -- I haven't heard a fat lady singing yet. Firstly I agree with Photoptimist it is not to do with pixel size but it still has nothing to do with format either. (except in the practical analogue area)
Let me present an example;
A FF sensor. Lets say we go and put a Nikon lens on and we all know don't we that, because we have payed three times as much for it, it has twice the resolution of the glunky ol pentax one
So we now have a FF image with twice the resolution of a normal FF. Should we double the coc on the calculator accordingly.?
I say yes and so it is not format based but resolution based.
Your pixels were too small in relationship to the coc to have any real input but the practical lpm is much closer to the size of the coc so lpm has a large effect.