Circle of confusion from pixel size for D850

[bovk]

If this does not belong in Camera Talk please move it somewhere else.

I would like to find out what is the CoC looking at the pixel pitch (center to center). I read somewhere that the CoC should cover at least two pixels but being a circle maybe it should be 4 pixels. For D850 the pixel pitch is about 4.3 micrometers.
Could somebody comment on : Is the COC (measured across and diagonally) 8.6 or 12 micrometers. The reason for this is to find the most conservative number for CoC compared to usual 30 micrometers. The attachment has dimensions approximate and 10x larger.

[Kenneth Rich]

Nikon has caused a lot of circles of confusion with its teasers this time, if the opinions on the Mirrorless thread are any indication.

[bobfriedman]

i would check calculation and your units...  https://en.wikipedia.org/wiki/Circle_of_confusion

but listen to this first. https://www.youtube.com/watch?v=Pdq65lEYFOM

[Akira]

I'm not sure if the smallest circle of confusion can be determined like illustrated.

If I understand correctly, the Bayer sensor doesn't simply combine four (R-G-B-G) pixels to function as one single full-color pixel.  Each pixel provides the resolution and the illumination info, and the lacking color info is borrowed from the next pixels of different color.

[pluton]

I would declare several candidate values for the proposed circle of confusion size, then shoot tests.  However:  It seems to me that the useful values for the C of C will vary with any change in the record-reproduce chain, especially the final image reproduction size and viewing conditions.

[David H. Hartman]

My 2 cents is... "20 Microns and Be There!" ...but while trying to research this is bit I notice that Bob Atkins said that 30 microns is the best you can do at an aperture of f/22. This indicates there is a point diminishing returns for stopping down. I use f/11 as a good compromise between DoF and sharpness.

If pixel peeping a D850 or even a D800 image at 100% you'll likely think there is no such thing as DoF. It would make more sense to size the image for viewing on a 4K display and judge DoF at a normal viewing distance for that display.

30 microns is probably appropriate for a Leica III F loaded with Kodak Super-XX and certainly not fitted with a Nikkor lens. 

Best

Dave Hartman

[Peter Forsell]

If this does not belong in Camera Talk please move it somewhere else.

I would like to find out what is the CoC looking at the pixel pitch (center to center). I read somewhere that the CoC should cover at least two pixels but being a circle maybe it should be 4 pixels. For D850 the pixel pitch is about 4.3 micrometers.
Could somebody comment on : Is the COC (measured across and diagonally) 8.6 or 12 micrometers. The reason for this is to find the most conservative number for CoC compared to usual 30 micrometers. The attachment has dimensions approximate and 10x larger.

Many, or even most, of these ’CoC’ drawings circling the web are pretty naive. The Airy disks (’CoC’) are not neatly centered on top of a pixel square. Instead we have a continuous spatial distribution of Airy disks. Another question is where one should draw the edge of the ’CoC’, is it the first minimum of the Airy disk, or the second maximum, or where. Probability distributions are not neatly drawn sharp-edged circles. In practice the center of the ’circle’ is never on top of the center of the square (the probability approaches zero).

[pluton]

Many, or even most, of these ’CoC’ drawings circling the web are pretty naive. The Airy disks (’CoC’) are not neatly centered on top of a pixel square. Instead we have a continuous spatial distribution of Airy disks. Another question is where one should draw the edge of the ’CoC’, is it the first minimum of the Airy disk, or the second maximum, or where. Probability distributions are not neatly drawn sharp-edged circles. In practice the center of the ’circle’ is never on top of the center of the square (the probability approaches zero).

Excellent points made here, Peter.
 

[basker]

Many, or even most, of these ’CoC’ drawings circling the web are pretty naive. The Airy disks (’CoC’) are not neatly centered on top of a pixel square. Instead we have a continuous spatial distribution of Airy disks. Another question is where one should draw the edge of the ’CoC’, is it the first minimum of the Airy disk, or the second maximum, or where. Probability distributions are not neatly drawn sharp-edged circles. In practice the center of the ’circle’ is never on top of the center of the square (the probability approaches zero).

That makes sense to me.

The following are just my bemused ideas on the subject of this thread. I have stifled them as long as I can. I do not claim I am qualified to teach.

It seems to me that fitting any disk, Airy or otherwise, into a square pixel will produce a square. That square would be either R, G, or B, assuming the surrounding pixels are completely dark.

A disk the size four pixels would still produce a square, but with a mixture of colors. The size of the square would be 2X2 or 3X3 depending on how, precisely, the disk is placed.

This came to my attention when I noticed that a properly focused star is more or less square, and poorly a focused star is a collection of adjacent squares of various sizes and diminished brightness. That is the only real situation I can imagine where Airy disks are not overlapping to an extent that they are rendered into a mathematical abstraction.

BTW, I am completely in awe of the remarkable astrophotography shown here on NG.