D800 vs. D800E: Diffraction according to Technical Guide

[Mike MacDonald]

Hi,

The Nikon D800/D800E Technical Guide says the following on page 16 and I'm wondering what it really means:
"The D800E offers better resolution at apertures where diffraction (page 13) is not an issue. The effects of aperture may therefore be more noticeable than with the D800, and care may be required to avoid loss of definition due to diffraction."

Is it saying?:
1) The D800E has such great resolution (pop) at larger apertures, so at smaller apertures you'll probably notice it more than the D800, but it will be NO WORSE THAN THE D800.

OR

2) When the D800E is stopped down to smaller apertures, the image quality is WORSE than the D800.

I have a feeling it's #1, but I have a D800E on pre-order and I want to be certain.

Thanks,
    Mike

[ErikKaffehr]

Hi,

That means essentially that the resolution advantage that the D800E has over the D800 will be reduced/lost when stopping down. Think of diffraction as a very strictly enforced speed limit. The car may be capable of going faster but can't becuase of the limit.

Best regards
Erik

Hi,

The Nikon D800/D800E Technical Guide says the following on page 16 and I'm wondering what it really means:
"The D800E offers better resolution at apertures where diffraction (page 13) is not an issue. The effects of aperture may therefore be more noticeable than with the D800, and care may be required to avoid loss of definition due to diffraction."

Is it saying?:
1) The D800E has such great resolution (pop) at larger apertures, so at smaller apertures you'll probably notice it more than the D800, but it will be NO WORSE THAN THE D800.

OR

2) When the D800E is stopped down to smaller apertures, the image quality is WORSE than the D800.

I have a feeling it's #1, but I have a D800E on pre-order and I want to be certain.

Thanks,
    Mike

[Mike MacDonald]

Erik,

Thanks for your help.

So, I'll stick with the D800E!

Mike

[ErikKaffehr]

Hi,

What actually happens is that the 800 and the 800 E have the same resolution, that is given by the sensor is called the Nyquist limit. The AA-filter reduces contrast at Nyquist limit to around 20% (on most AA-filtered sensors). Without the AA-filter the contrast at edges may be higher, but that needs optimal technique. Not stopping down to far, tripod, MLU, dead on focus etc.

This article demonstrates the effects of diffraction quite well: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1

The left column shows a correctly focused image with aperture going from f/4 to f/16. Stopping down increases sharpness but after f/5.6 (or so) diffraction limit kicks in.

Diffraction is benign to sharpening. So much of the edge contrast can be regained with aggressive sharpening, preferably using deconvolution, I would use smart sharpen a small radius, relatively high amount, "gaussian blur" and "more accurate". Technically, that would mean that we would approximate the diffraction kernel with a bell shape of similar size.

So it may be better to stop down a bit more and sharpen aggressively than keeping aperture large and not having enough DoF, but for maximum sharpness I would aim for f/5.6 or f/8 with any decently good lens.

Best regards
Erik

Erik,

Thanks for your help.

So, I'll stick with the D800E!

Mike

[Christoph C. Feldhaim]

Erik,

wouldn't it help to take the f-stop into account when chosing the sharpening radius?
I mean, e.g. at f 5.6 we have an Airy disc first diameter in green light of about 1.35*5.6=7.56µ.
With a pixel pitch of 4.88µ for the D800 example what would be the best rationale to chose a radius?
Would the ratio Airy Disc Diameter / Pixel Pitch help chosing a sharpening method / radius?
Would it make sense to sharpen red green and blue channel differently ? (Though not convenient, but could be scripted)
With PKS I usually don't think much about sharpening, but with my diffraction sensitive small compact (S95)
I sometimes ask myself which sharpening rationale would be best and how would it relate to the used f-stop.
Ideally I wished I could just use a table: F-Stop X1->Radius RX1, F-Stop X2->Radius RX2, etc ..
Usually I try to sharpen optically at 100% and then I double check at 33% (Printsize) for creative sharpening.
But this is capture sharpening and a different thing.

Ideas?

[marcmccalmont]

I also have a 800E on order, my thinking is I can adjust the "digital artifacts" with aperture
f5.6 no AA filter, f8 AA filter so to speak. Can't do that with the 800. I was unaware I was shooting my IQ180 diffraction limited at f8 but I have been getting really sharp captures. I'll need to shoot both the IQ180 and D800E at f8 and larger apertures to see what improvement there is, of course DOF will suffer but 2 of my lenses on the IQ are T/S so I should be able to compensate a bit.
Marc

[marcmccalmont]

Erik,

wouldn't it help to take the f-stop into account when chosing the sharpening radius?
I mean, e.g. at f 5.6 we have an Airy disc first diameter in green light of about 1.35*5.6=7.56µ.
With a pixel pitch of 4.88µ for the D800 example what would be the best rationale to chose a radius?
Would the ratio Airy Disc Diameter / Pixel Pitch help chosing a sharpening method / radius?
Would it make sense to sharpen red green and blue channel differently ? (Though not convenient, but could be scripted)
With PKS I usually don't think much about sharpening, but with my diffraction sensitive small compact (S95)
I sometimes ask myself which sharpening rationale would be best and how would it relate to the used f-stop.
Ideally I wished I could just use a table: F-Stop X1->Radius RX1, F-Stop X2->Radius RX2, etc ..
Usually I try to sharpen optically at 100% and then I double check at 33% (Printsize) for creative sharpening.
But this is capture sharpening and a different thing.

Ideas?

Focus fixer extracts the aperture information from the EXIF data
Marc

[Bart_van_der_Wolf]

wouldn't it help to take the f-stop into account when chosing the sharpening radius?

Hi Christoph,

Yes, when diffraction blur starts to dominate the total blur (also from residual lens aberrations, AA-filter, and Bayer CFA demosaicing), the so-called Point Spread Function (PSF) would require a wider sharpening radius to compensate for the increasing proportion of diffraction blur in the blur mix as we use narrower apertures.

I mean, e.g. at f 5.6 we have an Airy disc first diameter in green light of about 1.35*5.6=7.56µ.
With a pixel pitch of 4.88µ for the D800 example what would be the best rationale to chose a radius?
Would the ratio Airy Disc Diameter / Pixel Pitch help chosing a sharpening method / radius?

Yes, that's how it would work for diffraction blur alone. Of course we have a mix of blur components with different PSF shapes, so that complicates the matter a bit, but in general we can get a long way by using a Gaussian shaped PSF with an increasing sigma as the blur to sensel pitch ratio increases. An initial assumption of a Gaussian with 1/3rd the width of the diffraction pattern ratio seems to be a logical starting point.

Would it make sense to sharpen red green and blue channel differently ? (Though not convenient, but could be scripted)

No, not when we sharpen after demosaicing. The demosaicing will create an almost identical MTF for Red, Geen, and Blue. It would be more efficient to sharpen for Luminosity.

The D800E will probably not respond as gracefully to sharpening (unless diffraction dominates), because the inevitable aliasing artifacts will become more pronounced. But then it would require less sharpening at moderate apertures anyway.

Cheers,
Bart

[Christoph C. Feldhaim]

... An initial assumption of a Gaussian with 1/3rd the width of the diffraction pattern ratio seems to be a logical starting point. ...

Thanks Bart, for the clarifications.

One last (hopefully) question:
Do I get it right, that  with a pixel pitch of 4.88µ and an Airy disc diameter of 7.56µ would you suggest
(7.56/4.88) / 3 = initial radius for the Deconvolution Gaussian Algorithm?
Just to have is 100% clear.

Cheers
~Chris

[Bart_van_der_Wolf]

Do I get it right, that  with a pixel pitch of 4.88µ and an Airy disc diameter of 7.56µ would you suggest
(7.56/4.88) / 3 = initial radius for the Deconvolution Gaussian Algorithm?

Hi Chris,

Yes, that would seem to be a reasonable starting point to tackle Diffraction blur, one would need more if there are also significant residual lens aberrations, and not all AA-filters are the same. How much more would have to be established with additional tests. And because 7.56 micron would be close to the onset of visible diffraction (which starts at 1.5x the sensel pitch), the small 0.5-0.6 radius would seem appropriate for diffraction.

Cheers,
Bart

[Christoph C. Feldhaim]

So - for my S95 with its pixel pitch of 2.03 µ this would result in a radius for the Gauss deconvolution of

f 2.0 -> Airy Disc < 1.5*pixel pitch = forget it
f 2.8 -> 0.62 px radius
f 4.0 -> 0.90 px radius
f 5.6 -> 1.24 px radius
f 8.0 -> 1.80 px radius

And for the D800E with a pixel pitch of 4.88µ:

Up to f 4.0 -> ratio below 1.5 = forget it
f 5.6   -> 0.52 px radius
f 8.0   -> 0.74 px
f 11 -> 1.01 px
f 16 -> 1.48 px
f 22 -> 2.03 px

Arf
End of exercise.
 

[Mike MacDonald]

Hi,

What actually happens is that the 800 and the 800 E have the same resolution, that is given by the sensor is called the Nyquist limit. The AA-filter reduces contrast at Nyquist limit to around 20% (on most AA-filtered sensors). Without the AA-filter the contrast at edges may be higher, but that needs optimal technique. Not stopping down to far, tripod, MLU, dead on focus etc.

This article demonstrates the effects of diffraction quite well: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/49-dof-in-digital-pictures?start=1

The left column shows a correctly focused image with aperture going from f/4 to f/16. Stopping down increases sharpness but after f/5.6 (or so) diffraction limit kicks in.

Diffraction is benign to sharpening. So much of the edge contrast can be regained with aggressive sharpening, preferably using deconvolution, I would use smart sharpen a small radius, relatively high amount, "gaussian blur" and "more accurate". Technically, that would mean that we would approximate the diffraction kernel with a bell shape of similar size.

So it may be better to stop down a bit more and sharpen aggressively than keeping aperture large and not having enough DoF, but for maximum sharpness I would aim for f/5.6 or f/8 with any decently good lens.

Best regards
Erik

Erik,

My application is much different and very real world in that there is NO WAY I can shoot at any aperture larger than f/11 and I'll probably spend a lot of my time at f/22, maybe f/19 or f/16. My line of work is landscape photography of the prairie (the flatlands) and my approach to making the image look dramatic and 3-dimensional is to push the lens in VERY CLOSE (20 inches or less from sensor/film plane) with a super-wide angle lens. I feel that I get great results using my current Pentax 645NII camera with the 33-55mm lens set to 33mm and f/32 (curved field lens, too). After scanning, the image has 70 to 100 megapixels depending on dpi of scan and that's plenty for making large prints without even interpolating. However, I frequently print at 24x32 and the largest print that I've printed myself on my Epson 9800 is 43x56". However, I have had a 4000dpi scanned transparency (about 70MP) printed into a 6x9 foot mural at a nature center, but I've never seen it for myself. They're super happy and that's probably what counts the most, but I wonder how I'd feel about it. In any case, the equivalent f-stop in the full-frame 35mm format is about f/20.6 and the equivalent focal length is 21mm, if my calculations are correct. (The dimensions of 645 image is 41.5 x 56mm, just in case you're doing the math, too.)

So, in my real world situation (see some samples here: Outdoor Photographer Magazine Article - November, 2011 or at my website ChicagoNature.com), I'm assuming that I should still get the D800E. But, I never pixel peep with film scans, so it's hard to determine what I'll end up with when I shoot with the D800E. I'm also getting the fine Nikon 14-24mm lens, which I'm sure is much sharper than my medium format lens. The 24mm TS lens is not wide enough for my purposes. Plus from what I've been reading (and please correct me if I'm wrong), tilt/shift doesn't work in some situations like when you have subjects "sticking up in the foreground" because when the lens is tilted, the shape of the depth of field range is triangular (if viewed from the side) and comes to a point at the ground somewhere under the camera, near the tripod feet. So the tops of those near subjects will be out of focus because that's where the DOF triangle is narrowest. Therefore, with my prairie shots, with everything sticking up in front of me, I'm thinking that it will not be as effective as many say. Therefore, what should be my workflow under these conditions? Should it be something like this?:

1) Shoot as usual using the Hyperfocal Distance in order to get everything in focus, even though diffraction will be harsh. (And don't compromise on DOF by opening up a stop or two because the picture will look unacceptably soft at the extremes. True or not? You're the expert on this topic!)

2) Once in Photoshop (RawTherapee, etc.), do some Deconvolution Sharpening (please recommend software) to restore some of the details, resolution, sharpness or whatever it's actually called.  I don't know exactly if that'll work because I've been told that DS it's quite limited in what it can do. But, it seems like you're able to get it to work. I don't know if things differ when shooting landscapes as opposed to flat surfaces, like the dollar bill at the link you provided.

3) Work my magic in Photoshop using Shadow/Highlight Detail, then a version of what's referred to as Local Contrast to restore fidelity, life, and depth to an image by using USM at the approximate settings of 20, 50, 0 (strength, radius, threshold) followed by BLENDIF to protect the highlights and shadows from clipping, and then go from there, ending with sharpening using FocalBlade (which allows for independent control of surfaces and edges, among other things).

After all this, will I end up with a picture that's at least as good as what I'm getting with 645 film (Astia 100F or Velvia 100/F)? This is the question. All I'm looking to do is get an equivalent image quality along with the cornucopia of other benefits like digital provides!

Thanks for any advice you can give, especially if you lay it out in a way that I can easily understand

Mike

[Alan Smallbone]

Focus fixer extracts the aperture information from the EXIF data
Marc

Marc,

It seems they have not been updating this program, looks like it may do some deconvolution but it seems they stopped development in 2008, or is the website just not updated?

Alan

[hjulenissen]

Is focus stacking an option for you? Seems to me that it should easier to merge several sharp, medium-DOF images into one than to make one unsharp image into a sharp one.

-h

[ErikKaffehr]

Hi,

To begin with, if you and your customers are happy with your work it just fine. If it ain't broken, don't fix it, they used say.

Another point I'd make that I'd say that there is nothing wrong with pixel peeping, the pixels are what will eventually form a printed image. On the other hand much of the advantage that can be seen in actual pixels view can disappear in print. So we want to have as good pixels as possible but we can get away with less than optimal pixels quite well.

Here is what I can say about your points:

1) Working with hyperfocal distance is fine, but keep in mind that all depth of field calculations are based on a predetermined circle of confusion. The normal tables are based on 1/30 mm. On a Nikon D800 this would correspond to about 38 pixels lumped into one. In my MF days I used to use the markings for f/8 when using f/16. According to normal DoF tables this image is sharp: but optimal sharpness is this:

Now, the normal DoF tables are based on valid assumptions, like viewing relatively small prints at about 10" viewing distance and assuming that viewing distance increases linearly with print size.

The approach I tend to take is to focus on the most important part, stop down as much as I care and accept residual unsharpness on the less crucial parts. But that does backfire sometimes. I recently took an image with both full frame and APS-C and preferred the APS-C image because of better DoF. In this case wind caused motion was the problem, I wanted to keep shutter time down.

This article discusses DoF in digital in some detail: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/29-handling-the-dof-trap?start=2 . I tried stopping down, Scheimpflug and stacking and all worked.

2) Resolution cannot really be recovered with deconvolution sharpening, but it is probably edge contrast that matters most and that can be improved. What I found works best for me is Advanced sharpening in Lightroom with Gaussian blur option and More Accurate checked, see enclosed screen dump. I have also used "Topaz InFocus". If you want to improve on defocused zone of the image than "Lens blur" would be used instead. And I would adjust radius to match blur. I'm not sure what radius actually means, so I'd just play around with the radius slider until I get good results.

Normally I would sharpen on a different layer and use layer masking to blend different parts of the image. I don't want to oversharpen an image that is already sharp.

4) Michael Reichmann was shooting Pentax 67 on Velvia and jumped ships when Canon 1DsII arrived. I was shooting Pentax 67 on Velvia, but started using digital as a "light weight" alternative, using "bridge cameras". Later I went back to DSLRs. In the recent years I made some comparison between my Pentax 67 and my digital stuff and my conclusion was that digital came out on top. Except resolution of high contrast detail. The latest test involved both amateur scans and drum scans at 6096 PPI.

From what I have seen I would expect that you would be quite happy with any 20 MP camera assuming decent lenses. Nikon is obviously an attractive choice, with the optional non AA-filtered version and the availability of fine manual focus Zeiss lenses.

My two articles on digital versus film may be worth checking out:

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/59-sony-alpha-900-vs-67-analogue-round-2

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/16-pentax67velvia-vs-sony-alpha-900

My test is not so resolution oriented. Tim Parkins test looked much more into resolution and it was really a great test.

As a final comment I would mention Focus Stacking. I have used Helicon Focus but it can be done with excellent results in Photoshop CS5 without additional tools.

Best regards
Erik

1) Shoot as usual using the Hyperfocal Distance in order to get everything in focus, even though diffraction will be harsh. (And don't compromise on DOF by opening up a stop or two because the picture will look unacceptably soft at the extremes. True or not? You're the expert on this topic!)

2) Once in Photoshop (RawTherapee, etc.), do some Deconvolution Sharpening (please recommend software) to restore some of the details, resolution, sharpness or whatever it's actually called.  I don't know exactly if that'll work because I've been told that DS it's quite limited in what it can do. But, it seems like you're able to get it to work. I don't know if things differ when shooting landscapes as opposed to flat surfaces, like the dollar bill at the link you provided.

3) Work my magic in Photoshop using Shadow/Highlight Detail, then a version of what's referred to as Local Contrast to restore fidelity, life, and depth to an image by using USM at the approximate settings of 20, 50, 0 (strength, radius, threshold) followed by BLENDIF to protect the highlights and shadows from clipping, and then go from there, ending with sharpening using FocalBlade (which allows for independent control of surfaces and edges, among other things).

After all this, will I end up with a picture that's at least as good as what I'm getting with 645 film (Astia 100F or Velvia 100/F)? This is the question. All I'm looking to do is get an equivalent image quality along with the cornucopia of other benefits like digital provides!

Thanks for any advice you can give, especially if you lay it out in a way that I can easily understand

Mike

[marcmccalmont]

Marc,

It seems they have not been updating this program, looks like it may do some deconvolution but it seems they stopped development in 2008, or is the website just not updated?

Alan

They keep promising a 64 bit version but nothing for 2 years perhaps others can recommend a better deconvution program? It just works so well water drops look real, chrome bumpers look like metal etc so I keep using it
Marc

[BernardLanguillier]

They keep promising a 64 bit version but nothing for 2 years perhaps others can recommend a better deconvution program? It just works so well water drops look real, chrome bumpers look like metal etc so I keep using it
Marc

Have you tried Raw Developper? It is a very interesting raw converter with built-in deconvolution sharpening. Mac only at the moment.

The software has not have any major upgrade recently, but cameras are supported pretty quickly as they are released and the files are great.

Regards,
Bernard

[Scott O.]

Diffraction is something we have had to deal with forever, it has just become a more frequent talking point with the advent of the 36mp sensor, which it seems will emphasize depth of field/focusing issues.  It is the single thing I am most curious with as I anxiously await my D800e.  If it becomes an issue I suspect I will spend more time with PC-E lenses and maybe even give focus stacking a shot.  I am trying very hard not to dwell on the potential problems of this camera and just focus (!) on what is potentially a huge upside.  Time will tell...

[torger]

I think many are a bit too afraid of diffraction. I'd say if your pixels are not softened by diffraction, you have too low resolution on your sensor :-). Sharp aliased pixels are fun when pixel-peeping, but really it is better if the sensor can extract all information that is delivered to it. In the future I think sensors will be so high resolution that megapixels are irrelevant, and there will be more focus on specific lens resolving power etc, as it should be.

A image with a bit softer pixels also print better at lower ppis I think (no jaggies artifacts when nosing the print), so if you have sharp pixels it is preferable that pixel count is high enough so you can print at high ppis.

[BJL]

I think many are a bit too afraid of diffraction. I'd say if your pixels are not softened by diffraction, you have too low resolution on your sensor :-).

No need for the smiley, that is the simple truth. Diffraction will always be there to exactly the same extent once you have chosen your composition (camera position, FOV, and DOF) so if you are not to some extent diffraction limited it is simply because the fixed effect of diffraction is overwhelmed by other far worse and more avoidable impediments to resolution, like inadequate sensor resolution, lens aberrations, or focusing imprecision.

No one says that you should avoid lower f-stops because your resolution is then sensor limited.