The Falacy of FF Cameras having Flatter DOF

[NorthernFocus]

A comment in another thread about use of tripods caused me to think a bit about some of the "rules" that we all learn and repeat. Nowadays technology progresses much faster than we can keep up with it. And many rules, conventional wisdoms, etc. should be questioned and/or revisited with current equipment in mind. But we are collectively slow to change so we go on following rules/conventions that may well be obsolete.

One of these conventional wisdoms that I was recently pondering is the "fact" often stated that full frame cameras have flatter DOF(or APS-C sensors have wider DOF if you please). However, if one considers the formula for calculating DOF, the only thing that changes based on camera type is the constant for "circle of confusion". Lens type doesn't alter the equation so the difference is associated purely with the sensor.

These formulae were developed long ago and adapted to use with digital. Early on in the digital era, there were no FF sensors. And when they did arrive on the scene, same gen FF and APS-C sensors had the same or similar pixel counts and therefore the same/similar pixel pitch/density. So the constant used for circle of confusion in the calculations for the two sensor sizes were proportional to the crop factor, namely 0.03 for full frame and 0.02 for APS-C. However, nowadays with many generations of sensors in service, pixel pitch is what is really relevant when trying to evaluate relative DOF characteristics of two cameras. So the circle of confusion used in the calcs should represent the sensor resolution not its size. For example, a Nikon D810 and D7000 have identical circle of confusion and therefore identical DOF characteristics for a given lens.

For those who care...

[William W]

I don't think that I have represented a 'rule' that says something like "full frame cameras have flatter DOF(or APS-C sensors have wider DOF if you please)".

What I have stated often is that if we make two images using two different format cameras: then for the same FRAMING of any Subject and for any given same APERTURE USED, the image made with a 135 Format Camera (aka "full frame") camera will have a smaller Depth of Field than that which is made with an APS-C Format Camera.

What follows from that is - (as just one example using Nikon) - IF you have a D810 and a D7000 and you plonk an 85/1.8 lens on both cameras and make the same FRAMING of any Subject with both cameras and make both shots at (for example) F/1.8, then the shot made with the D810 will have a shallower DoF.

Like a lot of misinformation which sits in the on the www, I think that, the fact I described above has become distorted and manipulated and regurgitated into some chant like "a full frame camera has a shallower DoF than an APS-C Camera" . . . which I believe is clearly wrong and that point is good to clear up.

*

There is also another point in your Opening Post here:

"However, nowadays with many generations of sensors in service, pixel pitch is what is really relevant when trying to evaluate relative DOF characteristics of two cameras. So the circle of confusion used in the calcs should represent the sensor resolution not its size"

I am not too sure about that: but I am not willing to make a definitive comment, on that point, at this time.

WW

[george013]

A comment in another thread about use of tripods caused me to think a bit about some of the "rules" that we all learn and repeat. Nowadays technology progresses much faster than we can keep up with it. And many rules, conventional wisdoms, etc. should be questioned and/or revisited with current equipment in mind. But we are collectively slow to change so we go on following rules/conventions that may well be obsolete.

One of these conventional wisdoms that I was recently pondering is the "fact" often stated that full frame cameras have flatter DOF(or APS-C sensors have wider DOF if you please). However, if one considers the formula for calculating DOF, the only thing that changes based on camera type is the constant for "circle of confusion". Lens type doesn't alter the equation so the difference is associated purely with the sensor.

These formulae were developed long ago and adapted to use with digital. Early on in the digital era, there were no FF sensors. And when they did arrive on the scene, same gen FF and APS-C sensors had the same or similar pixel counts and therefore the same/similar pixel pitch/density. So the constant used for circle of confusion in the calculations for the two sensor sizes were proportional to the crop factor, namely 0.03 for full frame and 0.02 for APS-C. However, nowadays with many generations of sensors in service, pixel pitch is what is really relevant when trying to evaluate relative DOF characteristics of two cameras. So the circle of confusion used in the calcs should represent the sensor resolution not its size. For example, a Nikon D810 and D7000 have identical circle of confusion and therefore identical DOF characteristics for a given lens.

For those who care...

I presume you mean smaller with flatter and bigger with wider.

Dof, or depth of filed, is the range in distance from the camera in which the subject seems to be sharp. Leaves a definition of what is sharp.

A human can experience a circle of 0.25mm on a distance of 1 meter as being a sharp point. It's subjective but it is a defined size. If you want, you can alter it.
So what does that mean for a photo? The reference is a print of A4, let's say 20x30cm. The sensor has to be enlarged 20/2.4=8.3 times. Given a coc of 0.25/8.3=0.03mm. For a FF. And for a Nikon 1.5 crop it will be 0.02. No pixel being involved.
So said, sharpness and dof is not a property of the camera or lens, but of the visible picture on the output device. Assuming a well functioning lens.
It all becomes a little messy when everybody is viewing the picture at 100% on a monitor.
What does change in the camera is the distance in the dof. That's related to the angle of view. And then you can argue about comparing pictures with an equal distance or with an equal framing of the subject or a equal angle of view.

George

[William W]

(My minor editing and arrangement into point form)

. . .

PREMISE of DoF: A human can experience a circle of 0.25mm on a distance of 1 meter as being a sharp point. It's subjective but it is a defined size. So what does that mean for a photo?

1. The reference is a print of A4, let's say 20x30cm.

2. The sensor has to be enlarged 20/2.4=8.3 times.

3. Given a coc of 0.25/8.3=0.03mm. For a FF.

4. And for a Nikon 1.5 crop it will be 0.02.

5. No pixel being involved.

Conclusion: So said, sharpness and dof is not a property of the camera or lens, but of the visible picture on the output device. Assuming a well functioning lens.

Yes. A similar line of logic was what I wanted to produce but I couldn't get my brain to function with this terrible head cold that I have. Thank you, George.

WW

[xpatUSA]

Those deeply interested in DOF and CoC's would enjoy reading this (several times recommended):

http://www.dicklyon.com/tech/Photogr...Field-Lyon.pdf

There we discover methods that cut out much of the obfuscation that exists on the subject, and we can grow to know and love instead the 'Cone of Confusion'. I have written a spreadsheet based on that paper specifically for viewing on my monitor and thus I get lots more DOF that those hypothetical folks who gaze at an 8x10" print a foot from their nose.

[Dave Humphries]

Dan you fooled me,

When I read the thread title I thought "oh, this is going to be about the iPhone 7's apparent ability to falsify* a narrow DoF while using a small sensor and tiny lens" - through intricate processing calculations applied to the images captured by two lenses (and sensors).
* my choice of word
If anyone has not yet heard of this, details are available via Peter's thread here; Apple are at it again
It could be the start of something big.

But that's not it ...


The way I think of DoF with regard to sensor size is that 'crop factor' (cf), which we commonly use to mentally equate angles of view on different format size sensors using a given focal length (e.g. by multiplying focal length by cf), also applies to the aperture to be used for a given DoF.
My mind is a little hazy on this, it being some years since I went through the math and 'proved' this concept to myself, but I think the gist is that using say, an 85mm lens on a D810 and D7000, to match the DoF of the D810 at f/2.8, you would need between f/1.8 & f/2 if the same lens were fitted to the D7000 - if that's even possible!

This is why people move to FF when desiring a thinner (aka flatter) DoF; a full frame body is probably cheaper and lighter than buying an f/1.8 lens instead of an f/2.8 lens, if such is available/affordable, for a given purpose.

They see the thinner DoF on their monitors - at 1:1 while processing their images, so they are happy with the outlay. I only shoot one format (DX @ 1.5 cf), so this discussion is hypothetical for me.

f2.8/1.5 = f1.86 recurring
(where 1.5 = cf of D7000) - and my apologies for transposing the formats, requiring a division by the cf, but that seemed more relevant



However ...

I am inclined to agree with George; the viewing method usurps the relevance of the capture method and any specifications that relate to it, unless they are worse than the viewing method factors in play.

What do I mean?

a) How many of us print large enough for this to be relevant?
b) How many of us separately process a version of the image for on line or display use, at an optimum resolution?
c) Or do we (not me), just publish the big (print version) and leave the downsizing to the web browser?

For on line, projection or 'rolling display' use; by the time the image has been processed and downsized (to say 1920 x 1080), that's over a 3:1 linear reduction of pixels (assuming capture at approx. 6000 x 4000).
If properly output sharpened after downsizing, it will look good - and the apparent DoF will have been increased by the downsizing and sharpening due to their effects on CoC.
That's before we consider the physical display size the image is viewed on (think: mobile phone vs tablet vs 50 inch HD TV) and viewing distance - which we (as photographers) have no control over.

I make a note to myself that as 4K displays become widespread, the downsizing required will be less, so the increase in DoF it causes will reduce and ultimately, I guess we'll get to 6 or 8K displays and then perhaps, the factors Dan mentions may become more relevant - as long as the image is not viewed on a mobile phone or tablet where the physical display size is limited by our need to carry it - and viewing distance and pixel pitch have similar firm limits of what's possible or worthwhile.

Printing is a different ball game - for one thing, we won't be downsizing, possibly the opposite in fact.
So for printing (with a printer of sufficient resolution), the capture parameters, perhaps play a more important role in perceived DoF, all other things being equal (which they rarely are), although we photographers may be able to control some additional factors.


Some thoughts to be going on with?

Given that I (and I suspect many others here) only display images on line, a move to FF to achieve a thinner DoF would appear to be of little or no benefit, given that my subsequent workflow (for display) will largely 'overwrite' the image's captured DoF with that relevant to the display size and viewing distance - likely to be far more.

I may not be correct in all my assumptions, but I'm sure I'll learn from subsequent responses.

Thanks for 'listening', Dave

[GrahamS]

I think things are getting over-complicated here. The shorter the focal length of the lens, the greater the depth of field. To obtain the same angle of view as a 50mm lens on a full-frame camera we need a 35mm lens on an APSC camera (given a crop factor of 1.5X). If the same aperture and focal distance is used on both, the APSC image will have greater depth of field. We can equalise the depth of field on both cameras by using a larger aperture on the APSC camera or a smaller aperture on the full frame camera.

As for the COF or Circle of Confusion, any calculation using this figure for a given lens must now take into account the size of the photo sites of the sensor.

[xpatUSA]

I think things are getting over-complicated here.

< . . >

As for the COF or Circle of Confusion, any calculation using this figure for a given lens must now take into account the size of the photo sites of the sensor.

Why and how, please Graham.

[ajohnw]

DOF is a funny area. As a for instance Zeiss used a much finer figure for their distance scales than the ones introduced largely by the Japanese. So who's idea of sharpness on the usual 10x8 needs to be picked to determine what will be sharp. The calculations for degree of blur are pretty complicated. I've seen them and wish I had a copy of the text. They have to include the final images size and the sensor size and degree of blur at the final image size. Blur is another variable and without stating how much it's meaningless. On ancient figure for decent resolution for any size of final image is 300 dpi. The other one is oh gee look I have an F1 lens on a plate camera. That sort of thing is a bit of an abuse of blur really.

Crop factors are usually worked out on the basis of the diagonal. The wiki for instance reckons the crop factor on M 4/3 is 1.84 - 2. The difference is due to 2 usually being used. I suspect like many things the only way to be absolutely sure of the effects would be to do the sums yourself based on actual magnification. That's what using a crop factor of 2 will do but the view wont be the same as a FF camera.

There is a web page about that actually goes through the workings after a fashion but may be of interest for other reasons.

http://www.trenholm.org/hmmerk/DOFR.html

The book can be downloaded by following the how to obtain link.

What he has done is turned things on it's head. If a tiny object is placed on the focal plane of a lens it will produce an image of it at the distance the lens is focused to. The image scale will depend on the focal length of the lens and the relative distances, just as it does when lenses are used the normal way round. The sums he suggests can also be used to give degrees of "out of focus" or more simply how well they will be resolved so are of more use for determining the degree of sharpness as it doesn't account for the degree of enlargement needed in the final image.

Out of focus blur is all about cone angles. I'd guess pixel counts will crop up but it looks like many cameras have gone well past what lenses can actually do.

John
-

[ajohnw]

A comment in another thread about use of tripods caused me to think a bit about some of the "rules" that we all learn and repeat. Nowadays technology progresses much faster than we can keep up with it. And many rules, conventional wisdoms, etc. should be questioned and/or revisited with current equipment in mind. But we are collectively slow to change so we go on following rules/conventions that may well be obsolete.

One of these conventional wisdoms that I was recently pondering is the "fact" often stated that full frame cameras have flatter DOF(or APS-C sensors have wider DOF if you please). However, if one considers the formula for calculating DOF, the only thing that changes based on camera type is the constant for "circle of confusion". Lens type doesn't alter the equation so the difference is associated purely with the sensor.

These formulae were developed long ago and adapted to use with digital. Early on in the digital era, there were no FF sensors. And when they did arrive on the scene, same gen FF and APS-C sensors had the same or similar pixel counts and therefore the same/similar pixel pitch/density. So the constant used for circle of confusion in the calculations for the two sensor sizes were proportional to the crop factor, namely 0.03 for full frame and 0.02 for APS-C. However, nowadays with many generations of sensors in service, pixel pitch is what is really relevant when trying to evaluate relative DOF characteristics of two cameras. So the circle of confusion used in the calcs should represent the sensor resolution not its size. For example, a Nikon D810 and D7000 have identical circle of confusion and therefore identical DOF characteristics for a given lens.

For those who care...

Interesting Dan. In need of a life I just worked out the actual crop factor of M 4/3 in real circumstances. What I did was assume that a head say 300mm tall was being photographed from 2m to give a certain size image related to the width of the frame. I chose 18mm for FF. The focal length came out at 120mm. I then scaled the 18mm images size based on the width of an M 4/3 frame. Comes out at 8.65mm. Same distance and the focal length needed was 56mm. So on that basis the crop factor is 2.14.

Then say both lenses are F2. The cone angles for both lenses will be exactly the same for items that are in focus so there is no difference in that respect. However the magnification of the lenses is entirely different as one has 2.14 times the focal length of the other. So say something else is 2.1 meters away. The image from the 56mm lens will be at 57.5mm. The image for the 120mm lens will be at 127.27. So in one case the image is 1.5mm behind the focal plain and in the other it's 7,27mm behind the focal plane. Cone angles are the same / similar in this case so the FF camera has a more blur. If then the "head" is printed back to 300mm tall an indication of the changes in the blur would relate to the out of focus distance difference. So in M 4/3 case 1.5mm need multiplying by 34.7=52 and FF 7.27mm by 16.67=121.17. A much bigger result so more blurred.

The reasons for this behaviour is that the sum of the reciprocals of object distance and image distance = the reciprocal of the focal length. The reciprocal gets smaller as focal length goes up so same changes in object distance have more effect.

Phew glad that worked out the right way. I used some noddy web calculator and thin lens formulae. These are approximations but do work for indicating trends. All the general photo sums about are also based on these Sometimes throwing in circles of confusion but it's not as simple as that. I saw a test recently of a sigma 150mm macro lens. Diffraction effects kick in slower than F5.6 and that is where it's peak best even resolution is too. For diffraction really it would be better to read circles of confusion - good name as it all is confusing. At F5.6 a perfect lens should resolve over 60 lp/mm cleanly and with decent contrast. It can manage about 44 at the centre. That actually is rather good for a ff lens especially for a pretty even result across the field.

John
-

[dje]

The concept of circle of confusion involves assessing the smallest amount of optical blur that occurs on a point source that can be detected by the human eye. This relates to the angular resolution of the eye and when the viewing distance of a print is taken into account, angular resolution can be transformed into a circle diameter. The size of this circle in relation to the size of the print can then be used to determine the equivalent circle diameter on the sensor size image. Thus sensor size has an effect as well.

Pixel pitch has an effect on image blur, but this is constant across the image, as is diffraction blur. I think you could say that these effects only have a secondary effect on depth of field/coc, particularly with the fine pixel pitches used today and for reasonably large apertures.

Dave

[Shadowman]

All this talk and not one image as an example, theories and fallacies be darned; let's see some comparisons.

[William W]

All this talk and not one image as an example, theories and fallacies be darned; let's see some comparisons.

Image © WMW 1974~1996 / AJ Group Pty Ltd (AUS) 1997~2016

[Shadowman]

Image © WMW 1974~1996 / AJ Group Pty Ltd (AUS) 1997~2016

There it is, no more arguments unless you are a marketing exec/client. Funny how logos in both are more/less prominent and in the end it all depends on the viewer on which offers the best output. If you were to divide each of these captures into quadrants, what information would be easily evident and what value would it have to specific viewers?

[NorthernFocus]

Isn't it interesting how much discussion can be generated, papers written, etc, about a topic that by its very nature is subjective?

Ted, I did read (well mostly) the linked paper. Plus a couple of his references. It is indeed the best collection of the various theories/methodologies on the topic. Although the author does a good job of comparing the various methods of analysis, ultimately he doesn't take a position on which is most accurate. Or if he did I didn't glean it from a single reading.

There it is, no more arguments unless you are a marketing exec/client....

Actually, the scientific mind should argue that the photos presented are irrelevant. According to any of the various methods of comparing DOF of different capture media. One of the fundamental premises of comparison is "all other things being equal...". These examples were shot with different lenses and at different settings.

So someone (or everyone) answer me this, does the DOF of a given capture change if one crops it in post?

[Stagecoach]

So someone (or everyone) answer me this, does the DOF of a given capture change if one crops it in post?

I say no based on the scientific fact that there is not a 'field' in my DoF calculator for this

[William W]

Well... that one is only one sample of a game I used to play with students.

The more serious exercise was to get them to shoot those ruled markings set at 45 degrees (used for Focus Testing): the students would make equal frames using APS-C and 'full frame' cameras using the same aperture, to test the validity of "Equivalence" apropos DoF.

The reason I kept this one as a sample is because it has a bit more colour and flair than those boring test charts.

If you have a passion for reading, then Joseph James' paper of "Equivalence in Photography" is a informative and easy read.


WW

[george013]

....
So someone (or everyone) answer me this, does the DOF of a given capture change if one crops it in post?

Depends on the defintion of the dof. If it is the A4 print on a distance of 1m, yes, it changes. The crop has to be enlarged more than the original.

George

[george013]

Image © WMW 1974~1996 / AJ Group Pty Ltd (AUS) 1997~2016

It also shows how subjective dof can be. The dof difference is 17mm according the given figures. But the head of the two persons are situated different The man is more en face, the woman turned away. So with a same camera build the distance eye to the Canon sign is different. We're talking only of 17mm.

Another remark about using the dof-calculator. To simplify calculations it's using the focal length where it should use the image distance.

George

[dem]

So someone (or everyone) answer me this, does the DOF of a given capture change if one crops it in post?

Sure it does. DOF is defined for a final printed photo (image size vs viewing distance) and it is difficult to assess by looking at 100% magnification on the screen.

It does not matter if you used a crop sensor with an FX lens, cropped the image in PP or printed a 2 meter poster and cut out a small part of it with a pair of scissors - cropping always leads to blur becoming more visible.