LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E

Started May 9, 2012 | Discussions
falconeyes
Contributing MemberPosts: 762
Like?
LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
May 9, 2012

Ever since Nikon announced that the D800 will come in two flavors, D800 with a Bayer-AA filter and the D800E without one, people wondered how large the difference would be and what version to get.

If you want to know what's really in the boxes and how it really differs, you have to measure it.

Which is exactly what my friend Dieter Lukas from panobilder.de and myself did and now want to share with you. You obviously need both a D800 and D800E to study the differences

In an attempt to send a simple message, we crafted the headline:

D800: E = 0.5px sharper

which means that about 0.5 pixels less sharpening radius in a 100% Lightroom sharpening operation yields about similiar results.

Of course, there is much more to it. We actually succeded to measure the Bayer-AA filter's isolated MTF, an industry-first (in the open web).

The finding is that the Bayer-AA filter strength is only about 75% of what to expect from a full strength filter. The side effect is that D800 images sharpen up to the full Nyquist frequency but at the same time, do not fully rule out the risk of false color moiré.

Eventually, our work provides examples of ISO test charts sharpened up to the same level for both D800 and D800E, including the corresponding MTFs shown below:

There is an example of false color moiré in the wild as well:

(crop, cf. the blog article for the full size image)

If you have an interest to understand the impact of the Bayer-AA filter in the D800, or its omission in the D800E, you may want to have a read.

-> http://falklumo.blogspot.de/2012/05/d800-aa-filter.html

(link provides access to full report and a printable version too)

Enjoy

Dominique Dierick
Veteran MemberPosts: 4,467Gear list
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to falconeyes, May 9, 2012

At least five minutes ago, the URL to the article on the page is missing or wrong:

http://www.falklumo.com/lumolabs/articles/D800AA/D800AA.html

Well done. I actually use the USM values you came up with when PP'n D800. Depending on the lens, a second step 30%, .3 or .4 makes it a bit sharper, still without distracting artefacts. The possibility to relax on the USM with the D800e may be of interest for lower light photographers as noise is less sharpened too.

-- hide signature --

Nikon NPS member

 Dominique Dierick's gear list:Dominique Dierick's gear list
Fujifilm X100S
Reply   Reply with quote   Complain
davidstock
Regular MemberPosts: 118
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to Dominique Dierick, May 9, 2012

Great tests! These are the results I would expect, based on my own comparisons of NEF files and other online resolution tests. But it's good to see the subject rigorously examined this way in the lab.

I do think that sharpening could be further optimized for both cameras, using full-bore deconvolution sharpening instead of a combination of deconvolution and unsharp mask (which is what detail setting 70 is). In ACR I suggest a smaller radius, a smaller amount, and detail at or near 100. It would also be interesting to see what RAW Developer, with a more sophisticated approach to deconvolution sharpening, would do.

Still, these are valid tests, and will hopefully help destroy some of the mythology associated with a.a. filtration. Thanks!

Reply   Reply with quote   Complain
Goodfellaslxa
Junior MemberPosts: 47
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to falconeyes, May 9, 2012

The actual caption from above the picture of the red cloth/apple is:

"In the real world, we found false color moiré from the D800E not to be problem. We only spotted it in a very few shots so far and here is one rare example where it occured to us in the wild:

(crop, cf. the blog article for the full size image)

Reply   Reply with quote   Complain
falconeyes
Contributing MemberPosts: 762
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to Dominique Dierick, May 9, 2012

Dominique Dierick wrote:

At least five minutes ago, the URL to the article on the page is missing or wrong:

http://www.falklumo.com/lumolabs/articles/D800AA/D800AA.html

Thanks for spotting this. Fixed now. The correct link should have read:
http://www.falklumo.com/lumolabs/articles/D800AA/D800AAFilter.html

Reply   Reply with quote   Complain
falconeyes
Contributing MemberPosts: 762
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to davidstock, May 9, 2012

davidstock wrote:

I do think that sharpening could be further optimized for both cameras, using full-bore deconvolution sharpening instead of a combination of deconvolution and unsharp mask (which is what detail setting 70 is). In ACR I suggest a smaller radius, a smaller amount, and detail at or near 100. It would also be interesting to see what RAW Developer, with a more sophisticated approach to deconvolution sharpening, would do.

You're right. But maybe I should clarify:

The 70% applies to the amount and we left the detail slider at its 25% default position (except for one MTF plot showed for comparison).

For the D800E, we found that 0.5px radius is too high with a shot under ideal conditions, in order to use detail near 100 (at least for high amounts). ACR would need to provide access to lower radii what it doesn't do. But of course, combined with other sources of blur your advice can be valid.

Maybe, we should make both RAW files available for download for your own experiments.

Reply   Reply with quote   Complain
Kaj E
Veteran MemberPosts: 8,866Gear list
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to falconeyes, May 10, 2012

Thanks for the interesting work.

I noticed that you would have liked to sharpen at a lower radius than possible in Lightroom (and ACR). You could consider Capture NX2, where a radius of 5 corresponds to 1 in PS, so 1 would correspond to 0.2 in PS etc. The amount on the other hand is the other way around 100% is 500% in PS (if I remember correctly)

I like Capture NX2 algorithm for presharpening.
--
Kind regards
Kaj
http://www.pbase.com/kaj_e
WSSA member #13

It's about time we started to take photography seriously and treat it as a hobby.- Elliott Erwitt

Reply   Reply with quote   Complain
davidstock
Regular MemberPosts: 118
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to falconeyes, May 10, 2012

I would love to get ahold of the RAW files.

I believe (and have seen in other tests online) that deconvolution sharpening is well suited to detailed D800 files.

Adobe doesn't indicate clearly how their sharpening dialog works. It's a bit of a black box. From what I understand, a 25 setting on the detail slider is mostly unsharp mask, which can cause halos and other artifacts. As that slider moves toward 100, more deconvolution sharpening is supposed to be applied.

Much more control over deconvolution is available from programs like RAW Developer or, in Photoshop, from the Topaz InFocus plug-in.

Reply   Reply with quote   Complain
falconeyes
Contributing MemberPosts: 762
Like?
Access to raw files as requested
In reply to davidstock, May 14, 2012

davidstock wrote:

I would love to get ahold of the RAW files.

Upon specal request and for a limited period of time, I provide access to the two NEF files. The URL is to be found in a recent comment to the original blog article.

Reply   Reply with quote   Complain
Kaj E
Veteran MemberPosts: 8,866Gear list
Like?
Re: Access to raw files as requested
In reply to falconeyes, May 14, 2012

1. URL does not work.

2. Can not save target as.

Any advice?
--
Kind regards
Kaj
http://www.pbase.com/kaj_e
WSSA member #13

It's about time we started to take photography seriously and treat it as a hobby.- Elliott Erwitt

Reply   Reply with quote   Complain
TOF guy
TOF guy MOD
Forum ProPosts: 10,547Gear list
Like?
Re: Access to raw files as requested
In reply to Kaj E, May 14, 2012

It works but you have to remove the curved bracket, the | sign, and enter the image number one at a time.

So in fact there are two links to type.

At least it worked for me. Fast connection too.
--
Thierry

 TOF guy's gear list:TOF guy's gear list
Nikon D800E Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF-S Nikkor 24-70mm f/2.8G ED Sigma 150mm F2.8 EX DG OS Macro HSM +3 more
Reply   Reply with quote   Complain
TOF guy
TOF guy MOD
Forum ProPosts: 10,547Gear list
Like?
Re: Access to raw files as requested
In reply to falconeyes, May 15, 2012

Upon specal request and for a limited period of time, I provide access to the two NEF files. The URL is to be found in a recent comment to the original blog article.

Thank you. I was the one who made this request. Very appreciated.

I'll post a few observations on your site.
--
Thierry

 TOF guy's gear list:TOF guy's gear list
Nikon D800E Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF-S Nikkor 24-70mm f/2.8G ED Sigma 150mm F2.8 EX DG OS Macro HSM +3 more
Reply   Reply with quote   Complain
Kaj E
Veteran MemberPosts: 8,866Gear list
Like?
Re: Access to raw files as requested
In reply to TOF guy, May 15, 2012

TOF guy wrote:

It works but you have to remove the curved bracket, the | sign, and enter the image number one at a time.

So in fact there are two links to type.

At least it worked for me. Fast connection too.
--
Thierry

Thanks Thierry,

Worked for me too.

-- hide signature --

Kind regards
Kaj
http://www.pbase.com/kaj_e
WSSA member #13

It's about time we started to take photography seriously and treat it as a hobby.- Elliott Erwitt

Reply   Reply with quote   Complain
Marianne Oelund
Veteran MemberPosts: 4,314Gear list
Like?
Inadequate optics
In reply to falconeyes, May 15, 2012

falconeyes wrote:

Of course, there is much more to it. We actually succeded to measure the Bayer-AA filter's isolated MTF, an industry-first (in the open web).

Are you really sure about that?

What you did succeed in measuring, is the combined resolution of the 24-70 at 50mm f/4, with the D800 and D800E sensors. Unfortunately, the result has not remotely approached what the D800E sensor is capable of. It also does not approach what the D800E sensor combined with a much better optic, is capable of.

I've examined the raw data directly, and have found that it contains no examples at all, of the true D800E sensor edge response. At half the sampling frequency, we should still be seeing D800E MTF holding up to about 0.8, given sufficiently sharp optics at f/4. Further, there is significant variation in sharpness across the test chart, again due to limitations of the optics used.

An additional compounding factor is that the D800E sample has a 5.5% lower magnification, which puts the D800E at a bit more of a disadvantage in areas of the chart where there are line pairs.

Reply   Reply with quote   Complain
davidstock
Regular MemberPosts: 118
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to falconeyes, May 15, 2012

It might be interesting for some to see these resolution tests done by BartvanderWolf based on d800/d800e resolution test files. There are useful indicators for where the nyquist limit is and also what happens when the files are sharpened:

http://www.luminous-landscape.com/forum/index.php?topic=65927.60

starting near bottom of page.....

Reply   Reply with quote   Complain
Lance B
Forum ProPosts: 26,783Gear list
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to davidstock, May 15, 2012

davidstock wrote:

It might be interesting for some to see these resolution tests done by BartvanderWolf based on d800/d800e resolution test files. There are useful indicators for where the nyquist limit is and also what happens when the files are sharpened:

http://www.luminous-landscape.com/forum/index.php?topic=65927.60

starting near bottom of page.....

There is almost no difference between the two, looking at thise samples on that forum.

-- hide signature --
 Lance B's gear list:Lance B's gear list
Nikon D800E Nikon AF-S Nikkor 16-35mm f/4G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED Nikon AF-S Nikkor 24-70mm f/2.8G ED Nikon AF-S Nikkor 500mm f/4G ED VR +12 more
Reply   Reply with quote   Complain
falconeyes
Contributing MemberPosts: 762
Like?
Re: Inadequate optics
In reply to Marianne Oelund, May 15, 2012

Marianne Oelund wrote:

falconeyes wrote:

Of course, there is much more to it. We actually succeded to measure the Bayer-AA filter's isolated MTF, an industry-first (in the open web).

Are you really sure about that?

Yes.

Marianne, I appreciate much of your work but sometimes, you really don't look deep enough into (or appreciate) other people's work.

What you did succeed in measuring, is the combined resolution of the 24-70 at 50mm f/4, with the D800 and D800E sensors. Unfortunately, the result has not remotely approached what the D800E sensor is capable of. It also does not approach what the D800E sensor combined with a much better optic, is capable of.

Of course, a better optics gives marginally better results. Much better optics (in the center) don't exist. But this test is NOT limited by the optics in any way. The lens delivers good contrast at the Nyquist frequency which is all we need to factor out the Bayer-AA MTF. After factoring out, the MTF becomes independent from the other MTFs in regions where the combined MTF stays well above zero.

I've examined the raw data directly, and have found that it contains no examples at all, of the true D800E sensor edge response. At half the sampling frequency, we should still be seeing D800E MTF holding up to about 0.8, given sufficiently sharp optics at f/4. Further, there is significant variation in sharpness across the test chart, again due to limitations of the optics used.

Of course, there is variation across the field. The center is all what matters.

You seem to know little about MTFs. The monochrome microlens digitization MTF at 0.25cy/px is ~0.9, the diffraction MTF is ~0.9 and even the best and legendary lens would have a ray-optical MTF below ~0.95 at that frequency. However, an existing lens with its sweet spot around f/4 like very good Nikkors (except the 200/2G) would be more like ~0.9. This combines to ~0.7. Add to it the degrading interpolation performance of demosaicing and you arrive at something closer to 0.5. The value we obtain for the D800E of 0.4 is rather close (it's closer to 0.5 w/o applying the lens profile correction which we applied on purpose as it facilitates to factor out the Bayer-AA MTF).

After applying any kind of standard raw converter processing (like contrast etc., everything which makes RAWs not look flat and dull) you'll measure MTF value at 0.25 cy/px of something between 0.7 and 0.8, depending on the raw converter. Maybe, you have been confused by our methodology to turn all such options off (again, on purpose).

More than anything else, the MTF value at 0.25 cy/px measures the software used for conversion. Remember that this isn't an optical bench measurement.

An additional compounding factor is that the D800E sample has a 5.5% lower magnification, which puts the D800E at a bit more of a disadvantage in areas of the chart where there are line pairs.

Did you read the paper? The effect was discussed and our method is scale invariant.

Moreover, the most critical factor when factoring out an MTF is defocus. Because it varies heavily. If you can point to any pair of RAW samples where this critical factor is controlled, then please provide an URL. You won't find them on Imaging Resource and similiar sites.

You emphasize irrelevant details (our lens used was fit for the purpose!) and ignore the relevant details.

Reply   Reply with quote   Complain
rhlpetrus
Forum ProPosts: 22,874Gear list
Like?
Something you could provide
In reply to falconeyes, May 15, 2012

Interesting, very good test. I tried to match the overall detail from IR's RAWs using CNX2, by reducing USM with D800E or increasing it for D800 but could not get exactly same results, it always looked like the D800E's images were better overalll (had to look at various parts of image to compensate possible focus issues.

It'd be interesting to have a similar setting like the IR's fabrics and use your accurate focusing method and post it for people to try and see if they can match your findings visually, meaning similar detail by using different USM settings.

Thanks for the good work.
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Nikon D7000 Nikon 1 V1 Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF Nikkor 35mm f/2D Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +3 more
Reply   Reply with quote   Complain
rhlpetrus
Forum ProPosts: 22,874Gear list
Like?
Re: LumoLabs: In depth analysis of Nikon D800 AA filter vs. D800E
In reply to Lance B, May 15, 2012

Lance B wrote:

davidstock wrote:

It might be interesting for some to see these resolution tests done by BartvanderWolf based on d800/d800e resolution test files. There are useful indicators for where the nyquist limit is and also what happens when the files are sharpened:

http://www.luminous-landscape.com/forum/index.php?topic=65927.60

starting near bottom of page.....

There is almost no difference between the two, looking at thise samples on that forum.

I think looking at these charts won't tell the whole story. Reichmann's tests show some real differences in real life settings. Anyway, it's for us to pick which is the better one. I'd say either are awesome tools, I"ll likely get the D800E later in the year (or, depending, get the D600 if it materializes and the sensor is similar in performance).
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Nikon D7000 Nikon 1 V1 Nikon AF-S DX Nikkor 12-24mm f/4G ED-IF Nikon AF Nikkor 35mm f/2D Nikon AF-S Nikkor 70-300mm f/4.5-5.6G VR +3 more
Reply   Reply with quote   Complain
Kaj E
Veteran MemberPosts: 8,866Gear list
Like?
IR NEFs reprocessed 100% crops
In reply to rhlpetrus, May 15, 2012

rhlpetrus wrote:

Interesting, very good test. I tried to match the overall detail from IR's RAWs using CNX2, by reducing USM with D800E or increasing it for D800 but could not get exactly same results, it always looked like the D800E's images were better overalll (had to look at various parts of image to compensate possible focus issues.

It'd be interesting to have a similar setting like the IR's fabrics and use your accurate focusing method and post it for people to try and see if they can match your findings visually, meaning similar detail by using different USM settings.

We know that IR is not the most accurate in their testing. The JPEGs at IR are over-sharpened both for the D800 and D800E.

Just for fun i took the 100 ISO samples for the D800 and D800E and reprocessed them in capture NX2. The samples are identified with letters A and B, the same letter always from the same camera. Converted them to Nikon neutral profile and did sharpen both but differently because of the AA-filter on the D800 (something i am used to from all my previous DSLRs). Then made 100% crops of each.

Which is D800 and which is D800E?

1A:

1B:

2A:

2B:

3A:

3B:

So what do you think, which is D800 vs. D800E A or B respectively?

-- hide signature --

Kind regards
Kaj
http://www.pbase.com/kaj_e
WSSA member #13

It's about time we started to take photography seriously and treat it as a hobby.- Elliott Erwitt

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads