Nikon D800/E Diffraction Limits

[Bart_van_der_Wolf]

The f/16 image may look sharp when viewed at long distance, but fine line pattern visible in the red box is very clearly lost on the f/16 image.

Hi Erik,

Indeed, excellent example. Both results will allow to produce good output quality, but the loss of resolution due to diffraction is unmistakable. How important that is, depends on the final use of the output, which includes the output size and viewing distance.

Of course we also have to consider the subject we shoot, a reproduction of a flat object allows us to use wider apertures because the DOF is less of an issue which gives us more control over diffraction, but a subject where DOF is important will push us towards diffraction compromises. It's a trade-off, and a high pixel count offers us at least the benefit of denser/better sampling of the projected optical image.

Cheers,
Bart

[Ray]

What seems to be missing from the more diffraction affected 50D image crop, is the diagonal area detail in the 40D crop, that I marked in Red (the vertical bar) on the attached copy of your image.

No, it's not missing, Bart. Applying a bit more sharpening to the 50D F16 image, the diagonal detail in that vertical strip you've highlighted is actually more accurate than it is in the 40D shot, although fainter.

There are actually about 48 line pairs in that vertical bar. The 40D at F8 shows only about 24 of them.

Below I've attached the more sharpened version of the 50D F16 crop; the whole banknote at a higher resolution showing all detail and lines as they appear to the naked eye; and a reconversion of the 40D shot at F8 with arrows pointing to most of the artifacts and moiré. What a mess. 

However, I'm not trying to make the general point that F16 will always produce better results than F8 used with a lower resolution sensor. This example is an exception because of the nature of the target. After examining numerous comparisons, I'd say that F16 with the higher resolution sensor, such as a D800, will produce, on balance, the same detail as F11 with the lower resolution sensor, such as the D3X; and F11 with the higher resolution camera will produce the same detail as F8 with the lower resolutions sensor, at the plane of focus of course.

I'll try using the D800E on the banknote when I have the time, to see if I can get as much moire as the lower resolution 40D with AA filter.

Cheers!

[erpman]

It does help if we do not have to magnify the existing diffraction to the point that it becomes clearly visible as lost reolution. However, for a given Field of View it won't help to just stitch some more images together, because that will only increase our FOV. While that helps to reach a certain output size with a lower output magnification, it may not give us the FOV we want, it may be too wide. To counter-act that, one typically shoots with a longer focal length with a narrower FOV but unfortunately also a shallower DOF. To compensate for that one could use a narrower aperture, but that defeats the purpose of reducing the visibility of diffraction blur.

Yes, focal length was the missing link...

However, and I´m just speculating here: Is the loss of detail due to diffraction with the longer lens exactly negatively proportional to the gain in detail due to less magnification? Or could it be that the gain in resolution from stitching, makes up for at least some of the increased diffraction, thereby rendering a not flawless but at least better result?

From what I can gather, diffraction is a quite minute phenomenon and has possibly less impact on the total image than the artifacts and softness from extensive interpolation.

I´ve also noticed that diffraction can be an advantage since it makes the transitions between in-focus and out of focus areas smoother when you´re trying to achieve deep DOF.

But of course focus stacking is the most logical method for this, and probably not more time consuming than shooting with a TS-lens either.

[Fine_Art]

Ray, you are trying to confuse things for fun. We all know the 40d or the 50d both have the ability to take a picture with high detail. All those artefacts are the product of your digital manipulation of the file, not the cameras. You down-sampled the hell out of it then blew it up over 200%. The moire was created by your down-sampling not diffraction.

[Ray]

Ray, you are trying to confuse things for fun. We all know the 40d or the 50d both have the ability to take a picture with high detail. All those artefacts are the product of your digital manipulation of the file, not the cameras. You down-sampled the hell out of it then blew it up over 200%. The moire was created by your down-sampling not diffraction.

Not at all. The only downsampled image is the middle one above which was taken from a closer distance in order to show all the real detail. In the original comparison between the 50D at F16, and the 40D at F8, both images had the same default processing in ACR with minor adjustments of levels to get both images looking similar in contrast.

Neither image was interpolated nor downsized but displayed next to each other on the monitor at 200%, in the case of the 50D, and 240% in the case of the 40D, so they would both appear the same size. I then took a 'screen grab'of the results and saved as highest quality jpeg. There is no noticeable loss of quality or detail in the jpeg, and on my monitor at least (1920x1080, 24"diagonal), the jpeg represents very accurately what I saw on my monitor as I compared these images.

[Fine_Art]

My mistake then. I don't understand how you can get that size of aliasing.

[Ray]

My mistake then. I don't understand how you can get that size of aliasing.

I believe it's entirely due to the distance between lens and target. With all the recent concern about aliasing from the D800E, you've probably come across the recommendation to easily avoid the problem beforehand by changing the shooting distance to the problem area, and/or stopping down.

In this case I deliberately searched for the distance which would produce the aliasing, as an aid to accurate focussing. The LiveView screen on the 40D is only 230,000 pixels as opposed to 920,000 for the 50D.

[LesPalenik]

Very interesting and informational review.
Many of the discussed points apply not only to Nikon cameras, but also to other FX and DX models, especially the diffraction issue and "bigger crop" trap.

http://www.bythom.com/nikond800review.htm
 

[Ray]

Very interesting and informational review.
Many of the discussed points apply not only to Nikon cameras, but also to other FX and DX models, especially the diffraction issue and "bigger crop" trap.

http://www.bythom.com/nikond800review.htm
 

Thanks Les for the link to Thom Hogan's review. I find it a very balanced review on the whole, but not without flaws. I'm very puzzled by the following statement from that review.

I'm finding a lot of folk picking up the D800 models fall into the Big Croppers group: they're buying a D800 because it allows them to crop dramatically. As in pulling out 12mp pieces from a 36mp source. I'm not sure why these folk think that's any different than owning a D7000. If you're always cropping that much, the D7000 is actually slightly better than the D800 in terms of pixel density (very slightly: 16mp DX on the D7000 versus 15mp DX on the D800), plus you'll save enough money to buy some lenses that'll let you crop less.

This seems a very confused statement to me. I think Thom's put his foot in it here. He hasn't thought it through. Maybe he's just getting old, as many of us are, so please don't anybody think there are any hard feelings in the following comments. I make them purely for the sake of clarity.

(1) The difference between a 15mp and 16mp cropped-format camera is very trivial. I'd describe it as less than the resolution difference between the Canon 40D and 50D at F16, and less than the resolution difference between the D3X and D800 at F16.

(2) Any FX lens attached to a D800 becomes a much more valuable and flexible lens than it does attached to a D7000, whatever the quality and focal length of the lens.

I own just two Nikkor lenses, the 14-24/F2.8 and the 24-120/F4 zooms. On the D7000, these two lenses become effectively a 21-36/F2.8 and a 36-180/F4 (in terms of 35mm format FoV). However, in terms of FoV at both the equivalent image quality and sometimes even better image quality, by the D7000 standards, these lenses when attached to the D800 effectively become a 14-36/F2.8 and a 24-180/F4. For example, when a 24mm shot from the D800 is downsampled to an equivalent 16mm shot from the D7000, same FoV, the D800 shot is a better quality image, without doubt.

In other words, I'm getting the best of both worlds; the wider angle advantage of the larger format, plus the longer reach of the cropped format.

So what happens in the case of someone who always uses telephoto lenses for birding, and is not interested in the wide-angle aspect of full-frame? Let's consider two photographers out shooting with their sharpest lens, a 400/F2.8 monster which is necessarily a prime if it's the best lens available. One photographer has a D7000, the other has a D800. Both these lenses on the D7000 are effectively 600mm, as they are on the D800 in DX mode. (Okay! On the D800 not quite 600mm; only 581mm. Big deal!)

The point that Thom seems to have missed is that the D800 user effectively has a 400-581/F2.8 zoom, with just one prime lens. Not only that, it's a zoom that has the specifications of a high quality prime lens at all focal lengths between 400mm and 581mm (at least in the central area). It's a zoom lens without precedence. No amount of money could buy such a lens.

Nevertheless, I don't wish to imply that Thom's advice here is not without some merit. If you really are only interested in getting the longest reach with your Nikkor telephoto lenses, then a 24mp D3200 might be a better option than a D800, provided you are not too fussed about SNR and DR.

It so happens that DXOMark have now published their test results for the D3200. It's very clear that the D3200 images, at equal print size, downsampling the 24mp to 16mp, will be noisier and will have lower DR than the D7000 images.

If one wishes to crop the D3200 image to extend the telephoto reach beyond that of the D7000, ie, compare equal size but unequal FoV images, the differences in SNR and DR will be even greater. We're looking at differences in SNR, at 18% grey, of 1/2  to 2/3rds of a stop worse compared with the D7000 and D800E, and differences in DR of up to one stop worse, and even greater. At ISO 200 the DR of the D800E pixel is over one stop better than the D3200 pixel. That's significant. Even at ISO 6400, in case you want a really fast shutter speed to catch that lizard catching a fly, the D800E pixel has almost one full stop better DR than the D3200 pixel.

Sorry, Thom, but the truth must prevail.  

[jgbowerman]

I have been testing a D800E using a Hartblei 40/4 TS. There is no question in what I am seeing on the screen at 100%, both sharpness and contrast fall off substantially at f/16. I want DOF, but not at the cost of losing substantial contrast and detail. I have not printed any files as yet, but for now and depending on the composition, I'll be sticking with f/8 or f/11 for landscapes. F/8 is really outstanding, so I'll frequently bracket these two apertures and decide in PP on which one to keep.

[ErikKaffehr]

Hi,

What you see is what I would expect. Diffraction is benign to sharpening. Try using smart sharpen, advanced, with gaussian and more precise checked. Increase radius until you get optimal sharpness. I would guess that you would need radius around 0.7 for f/8 and around 1.3 at f/16.

It may also be possible to combine an f/8 image with an f/16 image if things are not moving to much. Then there is of course the option to focus bracket.

There is another downside to using small apertures, you can also get unsharpness due to subject motion, so you may need to increase ISO.

Best regards
Erik

I have been testing a D800E using a Hartblei 40/4 TS. There is no question in what I am seeing on the screen at 100%, both sharpness and contrast fall off substantially at f/16. I want DOF, but not at the cost of losing substantial contrast and detail. I have not printed any files as yet, but for now and depending on the composition, I'll be sticking with f/8 or f/11 for landscapes. F/8 is really outstanding, so I'll frequently bracket these two apertures and decide in PP on which one to keep.

[stevesanacore]

Hi,

It actually has to do with the angular aperture, and as such both the actual aperture size and the focal length are in play. However, since our aperture numbers (F-number) are a ratio (f/#) between focal length and aperture size, diffraction is constant (as is the angular aperture) at a given F-number.

That is because the image (and the diffraction) requires less output magnification for a given output size. The f/16 on the 35mm image was magnified much more. The f/22 - f/32 required much less magnification so the actual diifraction patterns stayed small enough to not affect output sharpness too much.

For the 1DsMk3, f/11 is probably the sweetspot where corner resolution has improved enough and center resolution has fallen enough to provide even sharpness across the image-circle, I know it does on my TS-E 24mm II. Optical theory predicts that center resolution will start to be visually impacted by diffraction at apertures narrower than f/7.1 on the 1DsMk3. It's not the optical quality, which probably is second to none, but pure physics.

Cheers,
Bart

Excellent explanation, thank you.

[ErikKaffehr]

Hi,

Bart says that according to theory onset diffraction can be seen on 1DsIII at f/7.1.

I made a series of test shots with a 16 MP APS-C camera which has similar pixel pitch to the D 800/E and I'd suggest that diffraction is clearly visible at f/8. Se left column of the linked page: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1

On the other hand, diffraction is benign to sharpening. We are loosing edge contrast which can be regained with proper sharpening. Stopping down to far would reduce resolution, too. Lost resolution cannot be restored.

My view is that it's OK to stop down, when needed, but to make best use of lenses we need to use medium apertures.

Best regards
Erik

Excellent explanation, thank you.

[Bart_van_der_Wolf]

Hi,

Bart says that according to theory onset diffraction can be seen on 1DsIII at f/7.1.

I made a series of test shots with a 16 MP APS-C camera which has similar pixel pitch to the D 800/E and I'd suggest that diffraction is clearly visible at f/8. Se left column of the linked page: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1

On the other hand, diffraction is benign to sharpening. We are loosing edge contrast which can be regained with proper sharpening. Stopping down to far would reduce resolution, too. Lost resolution cannot be restored.

Hi Erik,

Allow me to add a small, but not unimportant, nuance though.

As soon as the diffraction pattern becomes significantly bigger than one sensel (say 1.5x the sensel pitch, or at f/5.6 for 'luminance wavelengths' with the D800/E), the signal of neighboring sensels is going to be affected. One could look at it as cross-contamination between neighbors. That will reduce the MTF response at the highest spatial frequencies, which means that lower contrast high frequency detail is going to suffer most. In fact it may drop below the noise threshold, and is thus lost. No amount of sharpening can restore that.

Fortunately, although higher contrast micro-detail will also be reduced in output modulation, that modulation may still be high enough to allow restoration with deconvolution sharpening.

Ultimately, the diffraction caused by narrower apertures will become so severe that even the highest contrast micro-detail will be lost. So, the first unrecoverable losses start with low contrast micro-detail already at moderate apertures, and gradually the higher contrast micro-detail will also suffer at narrower apertures, untill all micro-detail is lost.

It is therefore difficult to pin an absolute (F-)number to the resolution losses (and restorability), without also including the input contrast level of the micro-detail. Only the absolute diffraction limit (for a given wavelength and f-number) is a physical fact, since any input contrast level will be reduced to zero.

Cheers,
Bart

[Paul2660]

This has been a good discussion.  I was curious on Thom's full review on the D800 and went back over it last night.

He brings up some interesting points on the effects of diffraction on both the 800 and 800E.  It seems that from his
testing, that both cameras seem to start showing the diffraction effects by F8 and he warns about using an aperture larger than
F11.  He also points out that past F5.6 the differences in the images between the 800E and 800 are so close that it gets
pretty hard to tell the differences.  Others such as Jack Flesher at getdpi.com and Lloyd Chambers have also made this point.  From
the shots I was able to take with both cameras I would tend to agree on this.  Obviously this is a huge area for individual opinions.

A point that Thom Hogan brings up is that he found the 800E to be superior at the wide open apertures.  F 1.4 to F5.6.  The lack
of the AA filter seems to bring out a slightly clear image with less issues around the edges of the subjects.  The 800 from his
testing can't seem to get to the same degree of detail and sharpness in these aperture ranges.  This fact interests me as I prefer the
wider apertures for night work.

Paul

[ErikKaffehr]

Hi,

I don't think it is really a question of opinions, more like "your mileage may vary", depending lenses, aperture choice and weather you are sensitive to aliasing artifacts or not.

Best regards
Erik

...
 Obviously this is a huge area for individual opinions.
...

Paul