Nikon D800/E Diffraction Limits

[theguywitha645d]

Fortunately, image quality is a subjective criteria. So basing sharpness simply on MTF or l/mm is not very useful. At any aperture, the resolving power of the D800 will always be in effect, so there is actually nothing to lose. To judge sharpness at pixel level does not have a great deal of meaning as a viewer will not be able to see the detail anyway. So the loss of sharpness because of diffraction does not lead to a softer image as perceived by the viewer and the increase in sharpness through DoF can make the image appear sharper which is far more important than any pixel level measurement of resolving power.

Trying to dissect this problem based on numbers without reference to the human visual system, and beyond simply the resolving power of the human visual system, is really a futile exercise. Our perception of an images counts much more than reducing the problem into the ability to separate lines.

[Wayne Fox]

Fortunately, image quality is a subjective criteria. So basing sharpness simply on MTF or l/mm is not very useful. At any aperture, the resolving power of the D800 will always be in effect, so there is actually nothing to lose. To judge sharpness at pixel level does not have a great deal of meaning as a viewer will not be able to see the detail anyway. So the loss of sharpness because of diffraction does not lead to a softer image as perceived by the viewer and the increase in sharpness through DoF can make the image appear sharper which is far more important than any pixel level measurement of resolving power.

Trying to dissect this problem based on numbers without reference to the human visual system, and beyond simply the resolving power of the human visual system, is really a futile exercise. Our perception of an images counts much more than reducing the problem into the ability to separate lines.

Valid and logical points. When speaking on resolution there are many factors involved including scene frequency, and importance of detail which is subjective and based on what we see in prints.  It's all tradeoffs. Some scenes have no problem printing large from lower resolution cameras, others will fall apart very quickly.  If the scene has no micro detail or important micro detail, or if that detail is just too small for any system to resolve, this may influence our choice of f/stop regardless of diffraction.

But the reality of diffraction on a d800/e is something to be very aware of.  I've tested several lenses now with Reikan FoCal's diffraction test on my d800e and while the software has issues with the aliasing of that camera  it still makes good assessments of the quality of the capture as compared to others at various f/stops, and you can manually click on any point and see the resulting capture.  What I've found is most lenses perform nearly identically from about f/4 up to about f/9.  The image quality falloff is pretty minor at f/9.  After that, it degrades very quickly, and at f/22 you have nothing left of detail ... total mush - unrecoverable.  If the scene has important subtle detail micro(such as the texture in bart's example) and you want to see it in the print, you won't if you shoot at 16 or 22.  You may be able to recover some of it at 11, but really you need to be between 5.6 and 8.

[theguywitha645d]

I am hoping to get a D800 for the studio this summer, so I can test that. I also shoot with a 645d and everyone said you could not shoot above f/11 with that camera because of diffraction. So I did some tests at f/22 and printed them to 36". They are very nice prints. The difference between that and f/16 or f/11 regarding DoF is striking. In comparative terms, the f/22 image is softer, if you are looking for it. Standing alone, it looks sharp. BTW, I am not sure the crop is at the object plane, but this is one of the images printed at 36".

This idea that somehow the imaging system is the only factor in resolving power is really not true. Object contrast also determines resolving power and so if you shoot a flat subject you are not getting the "most" from your system. Our perception of images and image quality is far more complex than simply resolution definitions of them.

[Fine_Art]

That's unfortunate, because it would allow you make a better decision when you can choose between options. Narrower apertures will not only show more sensor dust but also produce more background clutter in Macros, and microcontrast suffers, so if you can avoid it it would improve your overall image quality (you also can use shorter shutterspeeds which helps to reduce handheld motion blur).

That is unfortunately not plausible. At 100% the smaller sensels will have lower contrast than larger sensel cameras at the same aperture, or even lose all resolution at such apertures, and micro contrast detail will be lost whether we care or not. To illustrate I'll show a Macro crop from a larger sensel camera (sorry, I don't have a D800 to demonstrate it with) which demonstrates at f/8.0 what will occur at f/5.6 and narrower on the D800:

The 'Dsd' mentioned is the diffraction spot diameter in sensel widths.

As the theory predicts, until you reach f/16, there is still a lot that can be salvaged.

It's a pitty you can't compare it to an f/16 version, it would have been much sharper, but then maybe you don't care ... It's just too bad that people spend such an amount of money and then throw quality away by stopping down too far. Beyond f/16 there is virtually no fine detail left even in the plane of focus. At f/22 you only get, at best, some 75% of the maximum resolution that the camera is capable of as if shooting with a camera with 56% of the mega pixels, which you could put to good use. And as I said, you'll reduce your risk of motion blur if you shoot a f/16 instead.

Cheers,
Bart

Its also fairly difficult to recover the detail at f16 even with adaptive richardson lucy. If the point is to try to get everyone else taking blurry shots then f16 is great. Otherwise listen to Bart. Try to use f11. You can probably recover a lot from that.

[theguywitha645d]

You do understand that 100% monitor view does not represent any real world viewing condition. To take about sharpness in any absolute terms in regard to pixel pitch does not mean much.

[Wayne Fox]

The medium format camera shows far less diffraction than the 800/e at equivalent f stops.  I have an IQ180 and have tested it to f/32.  Certainly it's soft, but it isn't mush, so a trade off with loss of some micro detail to gain depth of field works.  I have shot at 22 and 32 frequently and due to the nature of the subject and other factors, large prints look very good.

The 800e at 22 is mush.  I'm not talking soft, it's beyond that.  At 16 i think it's much worse than the MFDB is at 22.  However, you are using a wider lens for the same FoV, meaning you have more depth of field at wider apertures. 

And while there is some validity to your point about prints vs. screen, you certainly can use the screen to get an pretty good idea what's going to happen.  The example posted, the question would be what size of print would it take for the detail to become "important" (if at all).  No two scenes are the same, and the value placed on that detail is certainly a subjective position of the photographer.

I'm out of town, but when I get back if I have some time I'll shoot the FoCal target and post results comparing the 800e and the IQ180 at various apertures, as well as the FoCal diffraction chart.

I think the key point is you will lose micro detail if you stop down too far ... there is no way to avoid it.  Your point is also valid in that when making a print perhaps that detail isn't important so the prints look fine and some sharpening techniques will make the prints quite acceptable.

[Ray]

You do understand that 100% monitor view does not represent any real world viewing condition. To take about sharpness in any absolute terms in regard to pixel pitch does not mean much.

I wonder myself just how many contributors on this forum actually realise the full significance of a 100% view of a small crop of an image on their monitor, in terms of equivalent print size.

The result will vary according to the size and resolution of one's monitor. If we take what might nowadays be a fairly typical monitor size of 24" diagonal with a 16:9 aspect ratio, and resolution set at an HD 1920x1080 pixels, then the resolution per inch of screen, both vertically and horizontally, will be about 92ppi.

With this information at hand, one can easily resize any image from any camera in Photoshop, using a resolution of 92ppi, whilst maintaining the same original or native file size, and see what dimensions result.

The 103.4mp image from a D800 (in 8bit mode) would produce a 6ft 8" x 4ft 5" print at 92ppi with neither interpolation nor discarding of image information. (For the benefit of those not familiar with the old-fashioned American system, that's 2032mm x 1346mm).

If the resolution of one's monitor is set to a lower resolution than 1920x1080 pixels, then that equivalent total print size containing the 100% view of a crop, will be proportionally larger.

When doing real-world tests to determine the significance of such minor differences in resolution, according to F/stop used or quality of lens used in relation to a given camera, I'm first interested if there is any noticeable difference at all, at 100% or even 200% on the monitor. If there isn't, then the matter is quickly settled.

If there is a slight but noticeable difference in detail, clarity and resolution in any comparison I make, at 100% or 200% view, then I make a note of that fact and interpret the significance of such differences in terms of output size, or print size.

Such difference may be of no significance whatsoever at the maximum print size from my printer, which is the 600mm wide Epson 7600, if I print the full uncropped image. But what happens if I significantly crop the image before printing?

It's not inconceivable that I might want to make a moderate size print of the 100%, or 200% crop that I see on my HD monitor. I expect my print to match the resolution and detail I see on my monitor.

If I see differences in detail between a D700 shot at F16 and a D800E shot at F16 at 100% on my monitor, the D800E shot being slightly, but noticeably, more detailed, then I understand that such differences would be irrelevant if I were to print the whole scene at A3+ size (or even larger), but such differences would not necessarily be irrelevant if I were to print just the 100% crop I see on my monitor at A3+ size, or A2 size.

[theguywitha645d]

The medium format camera shows far less diffraction than the 800/e at equivalent f stops.

I was not really suggesting the somehow a MFD and 35mm sensor are directly comparable, but more to point out that a limit based on pixel pitch for diffraction was often cited for the 645D, but I could exceed that aperture and still have excellent prints. I think there is simply too much emphasis placed on the effect of diffraction at pixel level. Sure diffraction impacts an image, but most of the time you need a comparison image to see it. When I hang a picture, I do not hang another for comparison.

Also, to mirror Ray's comments about what we are looking at at 100%, on my 24" monitor at 100% with Photoshop, I am viewing a 5mm x 9mm section of the 33x44mm sensor. That is like viewing a 33" x 44" print from 10". I would need reading glasses to do that.

We can all argue at what point diffraction is too much or how large is too large, it really comes down to personal taste, but I find framing this problem using viewing distance is a valuable exercise. I can only see pixel count go up and if pixel-level sharpness is the primary measure then the camera companies are going to have to figure out how to build abberration-free f/1.0 lenses or psychiatrists are going to have a growing number of photographers as clients.

[bjanes]

Some tests that I reported recently in this Thread are not definitive, but are consistent with image degradation becoming noticeable at f/16 and smaller. F/4 to f/5.6 produce the best results with good lenses, but f/8 is quite usable.

Regards,

Bill

[Ray]

Some tests that I reported recently in this Thread are not definitive, but are consistent with image degradation becoming noticeable at f/16 and smaller. F/4 to f/5.6 produce the best results with good lenses, but f/8 is quite usable.

Regards,

Bill

Yes, Bill. At sufficient magnification, image degradation will become noticeable when stopping either down or up from the aperture at which a particular lens is sharpest.

If this were not the case, one would not be able to declare that a lens is sharpest at a particular aperture.

However, it needs to be stressed that the noticeable benefits of either increased DoF when stopping down, or shallower DoF when stopping up, far outweigh the changes in sharpness that may be visible at the plane of focus.

The changes in DoF will likely be noticeable on a very modest sized print, sometimes even a postcard size, whereas the changes in sharpness at the plane of focus will likely be noticeable only on a huge print viewed at a very close distance.

I'm reminded again of Michael's comparison between the Canon G10 and Phase P45 at A3+ print size. Differences in sharpness (at the plane of focus) were not noticeable, even amongst experienced photographers. However, what was noticeable was the difference in DoF, even though the G10 was used at F3.5 and the P45 at F11, which sort of gave the game away because we expect the larger format to have a shallower DoF, especially when the differences in sensor size are as great as the difference between a P&S and an MFDB.

[theguywitha645d]

Some tests that I reported recently in this Thread are not definitive, but are consistent with image degradation becoming noticeable at f/16 and smaller. F/4 to f/5.6 produce the best results with good lenses, but f/8 is quite usable.

Regards,

Bill

I look at those results and it simply shows that shooting at f/16 is very good and will hold up to real-world viewing conditions. So it is really showing that diffraction is a function of format, not pixel pitch, when it comes to viewing images.

[theguywitha645d]

Folks have been so focused, so to speak, on pixels that they really have lost sight of what viewing conditions mean. The often cited 300dpi refer to very specific conditions—300dpi on an 8x10 print viewed at about 10”. If this was made a 16x20 print, then the viewing distance would be 20” and you would only need 150dpi. To say it another way, if you have about 3000 pixels defining the diagonal of the print and you view it from the same distance as the diagonal, you have a photo-quality print.

But this illusion (and photography is an illusion) is very robust. Many people have made prints with fewer pixels and have come out with excellent prints. Viewing distance is not a cliff, but a sliding scale. Zeiss defines the permissible circle of confusion at 1/1500 of the image diagonal. So at 1/2 viewing distance, a print will be acceptably sharp. A 150dpi 16x20 print viewed at 10” should appear acceptably sharp.

The diagonal of a D800 image is about 8800 pixels—lets call it 9000 to make the math simple. So at 1/3 viewing distance, a print from the D800 will reach the 1/3000 condition and at 1/6 viewing distance it will reach the 1/1500 condition. Can you really see a 16x20 print from 6.5 inches away, let alone 3.3 inches? A 40x60 inch print could be viewed from 12” to 24” easily.

DoF is going to have a far greater impact on the image than diffraction ever will. If you are shooting a D800 only at f/5.6, then you are really wasting the potential of this camera as well as limiting yourself over the control of your image.

[Bart_van_der_Wolf]

So it is really showing that diffraction is a function of format,

If with 'format' you mean the physical sensor array dimensions, then yes the effect is different. An f/16 diffraction pattern diameter will have the same dimensions on any sensor array size. However, on a physically smaller sensor array it will occupy a larger percentage of its total surface area. Therefore, upon subsequent output magnification, the physically smaller sensor array will require more magnification, which also magnifies the diffraction pattern more.

That's why f/16 on a nominal 6x4.5cm sensor array with the same number of sensels, will look better than f/16 on a D800. But that's not really the subject of this thread. What is, is that at f/16 on a D800(E) the distinction of microcontrast between pixels will be almost totally lost, and neither output magnification nor sharpening can get it back.

... not pixel pitch,

Well, there is a relationship if both sizes of sensor array have e.g. 40MP.

when it comes to viewing images.

The viewing distance determines how much resolution will be visible, but the differences in output magnification will already have taken effect.

Cheers,
Bart

[Bart_van_der_Wolf]

DoF is going to have a far greater impact on the image than diffraction ever will.

I don't think that is being disputed, certainly not from a creative point of view.

The complicating thing though, is that there is not a given DOF to an image, since the DOF boundaries depend on one's choice of Circle of Confusion limit, which in turn varies with output magnification and viewing distance.

If you are shooting a D800 only at f/5.6, then you are really wasting the potential of this camera as well as limiting yourself over the control of your image.

If there were no focus stacking solutions, or capture limitations which may require to crop, to name a few, then that might be the case. I don't think people will only shoot at f/5.6 when they know that they need to cover a lot of DOF in a single shot. They may however refrain from using f/22, unless they intend to only make small prints ...

Cheers,
Bart

[bjanes]

In addition to Bart's observations this LuLa article by Charles Johnson with a different view by Nathan Myhrvold is of interest. Both of these guys have PhDs in physical science. Because large format cameras need less magnification for a given print size, a larger circle of confusion (COC) is tolerable for the larger format, both for sharpness and depth of field. For a 35 mm format camera, diffraction at the sensor plane is independent of pixel pitch. However, if one wants to achieve the maximum detail of which the sensor is capable, pixel size is important. Nathan states:

"The camera will achieve diffraction limited resolution when the Airy disk diffraction formula 2.44 * N * Lambda is equal to the effective pixel size. In the terms of Johnson’s article, this means that the size of the COF is the pixel size. For green light (where the wavelength Lambda = 550nm) which is where human vision is most sensitive (and is used by Johnson’s article), the most conservative value of COF is equal to the pixel diagonal, which is 1.414 (square root of 2) times the pixel size. For red light (Lambda = 700 nm) and blue light (Lambda = 400 nm) we can use 2 times the pixel size, since they are spaced further apart."

To simplify further, the formula is max f-stop = P x 1.054. The D800 has a pixel pitch of 4.87 microns, so the corresponding f/stop is f/5.1. As one exceeds this critical f/stop, loss of contrast is often more noticeable than the loss of resolution. These considerations derive from the laws of physics and are not a defect in the D800. If you use f/16 on the D800, the results will be no worse than with the D3, which has 8.4 micron pixels.

The commonly held belief is that smaller format cameras have more depth of field than large format cameras, and this is true if the same f/stop is used. However, if the aperture size (in millimeters, not f/number) is held constant, depth of field is the same. See Roger Clark, another PhD in physical science. Small format P&S or camera phone cameras have very small pixels, and require large apertures (low f/stop numbers) to maintain image quality. They don't even offer f/16.

Regards,

Bill

[Bart_van_der_Wolf]

The commonly held belief is that smaller format cameras have more depth of field than large format cameras, and this is true if the same f/stop is used.

Yes, but only when also a shorter focal length is used (because there is also a smaller image circle that needs to be covered with a given FOV).

Cheers,
Bart

[theguywitha645d]

The commonly held belief is that smaller format cameras have more depth of field than large format cameras, and this is true if the same f/stop is used. However, if the aperture size (in millimeters, not f/number) is held constant, depth of field is the same. See Roger Clark, another PhD in physical science. Small format P&S or camera phone cameras have very small pixels, and require large apertures (low f/stop numbers) to maintain image quality. They don't even offer f/16.

Regards,

Bill

So, there are many variables in imaging. Since the variables are used to determine a product of a function, then it is not surprising that those functions would intersect at some point. The problem is they don't intersect at every point. You also must factor in how people use cameras--just to say it is possible to achieve the same result, does not mean a photographer actually uses one format in such a way to imitate another (most photographers work in the format they are in). So in that regard, stating the smaller the format the larger the DoF is not really a false statement, although it is a general one. Especially when you can get to a point where format size is so different that it is impossible to have functions intersect.

[Fine_Art]

The picture I made on the wall behind me is 24x36". @300dpi that is 77.8 MPixels. I did it tiled 3x4 shots. Even with the improved resolution of the D800 you need to take multiple shots. Therefore there is no need to wipe out your detail with high f stops. You stitch the foreground, you stitch the background. Keep your detail. Insisting on throwing it away gets nothing - no tradeoff, its a straight lose.

That is a fairly standard painting size.

[theguywitha645d]

LOL. Is that all? I print single 40MP images out on 44" wide paper stock and 44" is the short edge. The 40MP images aren't even sweating at that size.

[Fine_Art]

LOL. Is that all? I print single 40MP images out on 44" wide paper stock and 44" is the short edge. The 40MP images aren't even sweating at that size.

That doesn't mean anything. Why would anyone want a 150DPI print these days?