denoir wrote:
That's a good suggestion, I'll do that. I would however like to know what I'm doing before I start - i.e get the theory right
I think you should forget about the single pixels in this case. The samples you showed us are downsized and we have no idea what the single original pixels think about its neighbours...
Look at the whole picture instead; in the case of the APS-C sensor you enlarge the whole captured image to a given size. In the case of your FF image you do the same, and to the same size, no? The APS-C sensor image is enlarged more so the CoC need to be smaller if you want the beholder to see an image equally sharp as the image from the less magnified FF sensor.
If the DOF calculator uses the wrong CoC value the resulting DOF will be wrong.
From 4/3 cameras (crop factor 2) there is about 0.8 stops to APS-C (1.5 crop) and from there it is about 1.2 stops to FF giving us a total of 2 stops between 4/3 and FF.
Theory indeed, but it is very well matched in practical use.
regards,
/Jonas
EDIT: I see I type very slowly...
EDIT 2: Hey, I would also like to thank you for starting this thread. I'm eagerly waiting for the Nex-7 so there has been parts of great interest here. (I'm also eagerly waiting for a FF Live View camera with EVF but that's the next step I guess.)
mawz wrote:
That's exactly what he missed. For identical CoC's denoir's calculations were correct, but acceptable CoC also changes with format, thus theSuede's calculations.
Kind of a moot point, since there are no APS-C 23mm f1.3 lenses. Closest is going to be the Leica 24mm Summilux-M and if you're going to spend like that, an M9 is probably in the budget anyways.
Yeah, don't get me wrong, there are certainly types of shots (especially at wide end of the aperture) that are just better suited for larger formats. If you primarily shoot an M9 at 35mm f1.4, there aren't a lot of options for smaller equivalents. If you shoot primarily at 50mm f1.4 or f2 on 135, then a 35mm on aps-c can get you pretty close. If you spend your time stopped down a bit, there are quite a few things that aps-c can do as well as 135, especially considering aps-c development and improvements are on a higher trajectory than with 135 sensors.
With m4/3, it becomes a little tougher to stretch to 135 or 6x6 equivalents, and so I've personally stuck with aps-c and larger.
theSuede wrote:
If you look at a picture - the finished picture - you think that "sharp" is when something has good contrast at pixel level (at the scale you're looking at the picture with).
Hmm, I think I understand, but it seems somewhat counter intuitive to me. So you don't look at what is projected on the sensor but on the final image. Since the crop sensor is smaller it will have to resolve more (lp/mm) to get the same amount of detail as the larger sensor. Hence per definition the circle of confusion will per definition have to be smaller. Am I right so far?
I think that what I did not think of is that the final image size matters. But it still seems incomplete. What about the lens for instance? Usually as spatial frequency increases MTF drops. If for instance a lens can do (I'm just making up numbers) MTF50 at 40 lp/mm and MTF10 at 80 lp/mm - won't that affect the sharpness?
AFIK DOF is defined as the distance between two lines that appear "acceptably sharp" in an image. What about other factors such as an AA filter?
An AA filter will blur the entire frame and therefor details that would have been just enough sharp to be considered within DOF without an AA filter now become blurry enough not to be considered to be within DOF. Would one say that an AA filter leads to a more narrow DOF? That sounds awfully strange to me...
The AA filter IMO increases DOF by reducing the maximum possible sharpness, thus blurring the line a bit between what is sharp and slightly less sharp, the image requiring extra sharpening as a whole.
As far as I'm concerned, all what I need to know when out there taking photos is some very basic rules. More than that is counterproductive. A 35mm on my a900 will give less DOF than a 24mm on the Nex 7, difference being about 1 stop on similar apertures. Diffraction will kick in one stop earlier on the Nex 7. So for shooting landscapes, I would use f/8 on the a900 and f/5.6 on the Nex 7 for optimal sharpness and similar DOF. That is all what I need to know
Is the box into which the lens fits any bigger in the nex-c3 than it was in the older nex models. just checked nex-c3 images on google and there is not a single pic of the camera with the lens off. Not even dpreview saw fit to do one.
Maybe you could post a pic of your camera showing the inside of the box. It could be made large enough to avoid having to amputate the ears on the Contax G 21mm. In fact I wonder if modifying that box might be a better solution than trimming the protectors on the lens.
theSuede wrote:
If you look at a picture - the finished picture - you think that "sharp" is when something has good contrast at pixel level (at the scale you're looking at the picture with).
Actually I've changed my mind, or rather I've made up my mind. Although I do understand why looking at the CoC from the point of view of a final image is practical, it's also complete bullshit as far as the optical theory goes. No, I'm not saying that you are wrong or that the DOF calculators that use different CoC:s are wrong, but only that such a use confuses the issue (no pun intended).
The circle of confusion is the diameter of the criterion of maximum permissible unsharpness. I find it thoroughly counter productive to using it when discussing an optical system and it will only be relevant when you have a final image in mind - not what is actually being projected on the sensor, film or whatever.
Let me explain. Suppose we have two sensors A & B that have everything identical except size. A is larger than B. When you use a lens (the image circle covers both sizes) to project an image on the sensors the projected image on B will be identical to the center of the projected image on A if focusing distance, subject and lens aperture are identical. Or put in another way if we take a shot using each sensor, when we inspect pixel by pixel we will find that a cropped version of image A will be identical to image B. Pixel by pixel.
Looking at these images (at 100%) we'll determine some maximum tolerable unsharpness and call it the circle of confusion. It will however be identical in both cases. There will be no difference in the CoC regardless if your sensor is an ultra large format one or a tiny 2/3" as long as the pixel pitch of the sensor is the same. This is of course because the smaller image is simply a crop of the larger one and has not changed in any other way.
Now, when looking at things that way - i.e when not considering the various complex issues that occur when you produce a final image (resize, sharpening, viewing distance, eyesight etc etc) and that are way too complex to predict and compare, using the same CoC regardless of format is the correct way to go. And in that case, my initial calculation using the DoF calculator were entirely correct. When you consider only the projected image - which is the equivalent of looking at the image at 100% there will be a difference of about two and a half stops between DOF in "equivalent" shots taken with an ASP-C sensor and a 135 FF sensor.
What's the argument for using the same CoC? The same why we look at 100% crops when we want to see how sharp a lens is. Because resize and sharpen and other operations are bound to affect the final results significantly. We don't say that some lens is sharp on a certain format but not on another. Instead we have MTF charts that are a function of spatial resolution - i.e the MTF are not expressed as a function of line pairs but of line pairs per mm.
So I'm not at all sold on the commonly accepted methodology when it comes to DOF. It seems to me much more reasonable to separate the optical properties of the system to the production (and viewing!) of the final image. A crop sensor does, precisely what the name implies, crop the image. A 135 FF sensor is a crop sensor to a medium format sensor and the latter is the same to a large format sensor. How you crop the image will in no way change the parts you are keeping. Opening an image shot with a 135 FF sensor in Photoshop and cropping the image to the same size as if it had been taken with an ASP-C sensor will give you identical results to an image taken with an actual ASP-C sensor. So why confuse the issue by talking about a later post processed (i.e. resized) image? You certainly wouldn't say that you are reducing the sharpness of an image when you are cropping it, so why would you say that you are reducing the DOF?
According to this theory, if I shoot with a 24mm on FF and shoot with the same lens on a crop sensor (36mm equivalent), I'm supposed to get 2 1/2 stops more DOF from the crop? I don't know about the scientific formulas but in practice, the 24mm on FF will give me more DOF than the 36mm on the crop.
No. If you shoot with the same lens on both at the same aperture, you'll get an identical DOF. Crop is just as the name implies, a crop of a larger image. The content of the cropped portion won't change just because you cropped it.
If you take a shot with a 36mm lens and then switch to a 24mm lens, regardless of format, you'll get the equivalent of about 2.5 stops more DOF - regardless of which format you are using, if you look at the image at 100%.
My point if we exclude resizing, postprocessing, viewing distance and all that other crap that has nothing to do with the image projected on the sensor then DOF has nothing to do with sensor size but only with focal length, distance to focal plane and lens aperture (and to some extent lens design).
As I said, your formula may be correct, but in practical terms, I want to print both photos to let's say A3 size, or view them both on my monitor at 800x1200 resolution. In both cases, the FF photo will give me more DOF. Your theory is correct in absolute terms, but in practice it is wrong, because the end result is either prints or screen viewing.
1) Projected image. In this case only focal length, aperture and focusing distance will matter (and auxiliary stuff as the register distance, lens design etc)
2) Captured image. Here pixel pitch becomes relevant as well.
3) Final output. When you resize for monitor or print to a fixed final size then CoC will be different. This however also depends on viewing distance, print size, eyesight and a bunch of other parameters.
I'm not disputing what Joakim said or what the DoF calculators use. I'm only saying that it's inconsistent with how we usually look at optical stuff. We don't judge the sharpness of a lens on how the output looks after it has been printed. We don't say "this lens is sharp on a FF sensor but soft on an ASP-C sensor". Instead we look at 100% crops or look at MTF charts that are expressed per mm of sensor. So why do we do it differently in the case of DOF?
Well, I guess because most photographers are interested in the end result, and what they can see on the final print. In order to obtain the results you mention in real life, I would have to print the FF 50% larger than the APS-C, which is usually not the case. But as said, what you say is correct on the pixel level.
I guess it wouldn't hurt to have something other than DOF to describe how the degree of blur relates to the distance to the focal plane. For instance one thing that definitely adds to the "FF look" is the fact that most lenses due to field curvature at large apertures produce a lot of blur towards the edges. When you crop the image and just keep the middle, you get more linear sharpness-to-blur transitions with nothing of interest happening at the edges. Your DOF may be identical but you'll still get a very different look as the degree of blur at the edges will be different.
Then you have stuff like differences in lenses. A well corrected lens with little or no SA will have a more rapid sharpness-to-blur transition compared to a lens where SA adds blur of its own.
mawz wrote:
Kind of a moot point, since there are no APS-C 23mm f1.3 lenses. Closest is going to be the Leica 24mm Summilux-M and if you're going to spend like that, an M9 is probably in the budget anyways.
The Canon FD 24/1.4 is available for less. It is a rather big and heavy lens for a NEX though.
denoir wrote:
If you take a shot with a 36mm lens and then switch to a 24mm lens, regardless of format, you'll get the equivalent of about 2.5 stops more DOF - regardless of which format you are using, if you look at the image at 100%.
I agree that the 24mm lens will give more DOF, but I'm not sure I can put the number 2.5 on it.
In trying to keep up with this discussion, one observation is not made very well. A large format image will have less DOF than a small P&S camera, as we all know. But in the above example, if one with the 35mm lens were to back up enough to an equal field of view as the 24mm lens, or the one with the 24mm lens would step closer to equal the 35mm field of view, then the DOF would get greater and lesser respectively, until the difference is, in my own mind, about 1 stop less DOF with the larger sensor. I believe this is in line with edward's quote:
edwardkaraa wrote:
.... A 35mm on my a900 will give less DOF than a 24mm on the Nex 7, difference being about 1 stop on similar apertures.
So, if I want less DOF with my M8 and 50mm Summicron to create those photos with that thin slice of sharpness, should I buy a 50 Summilux? I think I would rather have a M9, and use that Summicron to it's full potential. After I step forward to equal the field of view with the M8, I would still be gaining about a stop worth of DOF, meaning the DOF field is more shallow with the M9.
this is silly, when I talk about dof I care about the actual picture not the individual pixels. who zooms in at 100% to look at dof? when I look at 2 similarly framed shots there usually looks to be a 1 stop difference between FF and asp-c. since I shoot to get pictures not crops, that's what matters to me.
Even as an optics constructor you need to keep track of "scales" in the projection. If we start with the basic nomenclature:
Image side (projection) >> this is the sensor, or the film plane
Object side (target) >> out in front of the lens, 3D reality
Total enlargement >> object size (mm) vs picture size (mm)
Most of the important stuff happens in front of the aperture, on the object side. Behind the aperture you mostly have a bunch of different technical solutions that even if they differ in concept (mirror vs mirrorless, retrofocality ratio and so on) are meant to do the same thing, i.e take care of what the lens "throws" at the sensor/film. Transfer the projection into a picture.
They are not totally decoupled, by obvious reasons. If you use a large format back, you need the lens to be able to cover the entire image (sensor or film size) without vignette, and without severely compromising corner picture quality. You need a larger "throw angle".
Already here we have a practical consideration. As you increase image (projection) size requirements you also increase glass element diameters as you get further out (an in!) from the aperture. When you increase glass diameters, the needed element thickness increases if you want to keep the same surface curvature ratios. Some basic constructions (like the extended planar) don't take this very well - a certain f > Fno > projection circle size ratio needs to be kept. It's almost impossible to make an 85/1.4 planar medium format lens (a reasonably good one at least!). It would be very easy to make a VERY good 85/1.4 APS optimized lens (but it wouldn't be much smaller or cheaper than a "normal" 85/1.4 - just better - as long as you don't need anything outside the 15mm image height, the APS circle radius).
But disregard this for a moment. Just think of the image.
What "makes" an image is as I said in front of the aperture, in the "object plane". The image angle of view is determined by the combined total fl and the sensor size, but to simplify we can say that:
AoV = 2(arcsin(0.5*sensor width)/focal length))
front pupil diameter = f / NA
And those are the two you need to match to get "equivalence". Equivalence falls apart at very low NA (numerical aperture values) because of "technical reasons" behind the aperture. Bad incident light angle efficiency is one of the reasons - a 7D can only assimilate about 1/3 of the light falling at it from the F1.2 angle, you have an almost 1.5Ev loss of light here. This of course dampens the effect the aperture rim light has on the image, by effectively "shrinking" the effective aperture and making the DoF increase. This is one of the compound reasons that makes the 100MP vs Rokkor comparison fail at wide open apertures.
sebboh wrote:
this is silly, when I talk about dof I care about the actual picture not the individual pixels. who zooms in at 100% to look at dof? when I look at 2 similarly framed shots there usually looks to be a 1 stop difference between FF and asp-c. since I shoot to get pictures not crops, that's what matters to me.
In that case, why don't you apply the same principle when talking about lenses i.e a lens being sharp/contrasty relative the sensor size? An ASP-C sensor to produce an image with the same framing as an FF sensor would require that you use a wider lens. This subsequently will require a higher resolving power to get the same amount of detail. Very few lenses that I'm aware of are capable of producing the same contrast at 80 lp/mm as they can at 50 lp/mm.
The main point however is that I think it is an utter folly to start mixing lens and camera optics with post processed output. They are two separate things and should be treated as such. I think the 100 MP vs Rokkor example is a good one of how useless the DOF concept can be. As you pointed out the look and amount of blur was radically different yet according to the simple DOF theory they should have been close. In fact the Rokkor at f/1.2 should have had slightly less DOF, although in reality the 100 MP produced a whole lot more blur outside of the focal plane.
This is of course because of other parameters such as the design of the lens and the fact that you're cutting off the edges on the crop camera where you usually can find field curvature that brings stuff even more out of focus.
So I would say that the DOF concept is rather useless as it is, at least at large apertures where you have all sorts of stuff going on. So why not, like we do with everything else, stick to a strictly optical definition and then interpret it as it applies for different special cases? I think the idea of using post processed output as the basis for evaluation is completely inconsistent with how we usually evaluate the optical properties of a system.
Edit: I see that Joakim beat me to it.
Bad incident light angle efficiency is one of the reasons - a 7D can only assimilate about 1/3 of the light falling at it from the F1.2 angle, you have an almost 1.5Ev loss of light here. This of course dampens the effect the aperture rim light has on the image, by effectively "shrinking" the effective aperture and making the DoF increase. This is one of the compound reasons that makes the 100MP vs Rokkor comparison fail at wide open apertures.
That's very interesting - I didn't know that. Is it simply because the mirror box (or whatever the cavity containing the mirror & sensor is called) blocks off the light i.e is smaller than the max aperture of the lens?
