Help... I just don't know enough!

[tashley]

I started a thread over at DPReview's Nikon Lens forum asking for suggestions for Nikon glass for the D800e I have on order. A polite and useful discussion ensued during which one poster (Apaflo) asserted something about the D800e that also applies to MFD and which I have not heard before. I am not sufficiently technical to asses it but is is at least interesting and quite possibly important to anyone shooting without an AA filter.

In essence he is saying that removing the AA filter introduces aliasing artefacts and that when you make smaller prints (small depending on the size of the sensor) you will be downsizing away the resolution advantage that no AA gives you but still suffer from the aliasing artefacts that you gained when you got rid of the AA filter.

There's a link to the thread here and I would really appreciate anyone with profound technical insights who had the time to look at it and give an opinion!

Re: D800 switcher from Canon needs glass advice: Nikon SLR Lens Talk Forum: Digital Photography Review


Thanks!

Tim

 

[yaya]

The E is supposed to produce sharper images than the non E

Which means it'll probably require less sharpening before going to print

Which means it'll probably suffer from less sharpening-rlated artefacts than the non E

By aliasing I'm assuming that the guy means moire, and you buy a 36MP camera so that you can have less moire without having to use an AA filter

I say it all with a big question mark since the camera is not shipping yet...

If I was to buy a new FF Nikon (I'm not as I'm happy with my D5100) I would go for the E for sure...

 

[pophoto]

The E is supposed to produce sharper images than the non E

Which means it'll probably require less sharpening before going to print

Which means it'll probably suffer from less sharpening-rlated artefacts than the non E

By aliasing I'm assuming that the guy means moire, and you buy a 36MP camera so that you can have less moire without having to use an AA filter

I say it all with a big question mark since the camera is not shipping yet...

If I was to buy a new FF Nikon (I'm not as I'm happy with my D5100) I would go for the E for sure...

yaya, yes the D5100 is a small entry DSLR. So while it has the same sensor as the D7000, I'm pretty certain the D800 E or non-E will be a huge step up! Although certainly a lot heavier

Also I think what tashley is referring to is Moire, although anyone today working with RAW processing software will hopefully have that corrected. The D800E as far as details go will surpass what the D700 will give you straight even after down-sizing to comparable 12MP size, consider that super-sampling.

More detail will be caught and hopefully Moire is a non-issue and if anything, just a small one that can be easily corrected! Certainly once it arrives we will ALL know! However, I wouldn't be worried about it, and that does become the issue there's always Ebay if you do

 

[Paul2660]

I believe what the poster is referring to is color aliasing, or the Christmas Tree light effect. This occurs when you don't have a AA filter on any back including a Phase One. You tend to see it when a very small subject is brightly lit like grass in the sun. Here is a link that describes it in more detail:

Photoshop: Fixing CCD Color Aliasing

The AA filter's slight blur reduces the effect.

You can still get Morie with a AA filter, example, Canon 1Ds MKI or MKII both IMO were bad at getting frequency Morie on fabric or bird feathers. Both cameras had a pretty extensive AA filter.

The more the resolution the greater the chance of seeing Color Aliasing. Capture One does a very good job of removing it from Phase files and Nikon has already made a statement about how they will have some software component of their Capture Software that will help remove similar effects.

Paul Caldwell

 

[MGrayson]

:lecture: mode!
To make things easier, let's forget color and imagine a B+W sensor. Consider the image right before it hits the sensor. It has information at all frequencies up to that of blue light (a few thousand per millimeter, but I could be misremembering). In any event, it's higher frequency than we will be sampling. So imagine a curve showing information as a function of frequency with a tail dropping off to zero at very high frequency.

Here's the tricky part! This data is now sampled at the frequency of the sensor. The sensor can only know about information at lower frequencies than its upper limit. So what does it do with the higher frequency data in the image? It wraps that data around to the lowest frequencies! (Much mathematics suppressed here. Be grateful!)

That low frequency data cannot be removed! You can suppress colors, smudge patterns, but you can no longer filter it out.

An AA filter removes very high frequency data from the image before it is sampled.

This has the advantage that there is nothing hitting the sensor above its sampling frequency, so no information pollutes the low frequency parts of the image. The disadvantage is that, filters being imperfect, we have suppressed some information at the high frequency range. Sharpening is an attempt to boost that high frequency signal back up.

So why the dislike of AA filters when they prevent such glaring messes as pattern and color Moire? There are two problems.

1) The AA filter/sharpening round trip is imperfect, and we are left with artifacts.

2) The low frequency aliasing is usually localized in the picture, whereas AA/Sharpening affects everything. And we like our pictures as sharp as possible everywhere.

But there are two things which are not problems with the process, but are problems with us.

1) We forget that the AA filter is only half the process. We must look at the AA/Sharpening round trip to see what is in the image. Yet we look at the unsharpened image as if it were some holy document.

2) And the biggest issue, and what I think is responsible for the "pixel peeping" label, is that when we view an image at 100%, we are really focusing on that part of the frequency spectrum closest to the upper limit. We look at thin lines one pixel wide and judge a picture's technical perfection by that small part of the spectrum. Well, I do, anyway.

Thank you for reading this far.

Best,


Matt

 

[D&A]

Matt, an excellent and interesting explaination, but I felt you left out the punch line...LOL! In other words as this relates to the D800 vs D800e and its 36MP resolution, what would you surmise in terms of the native Raw files from each and advantages and disadvantages prior to applying any software corrections based on what your explained. I understand the theroretical aspects but as they apply to what we might expect in some worst case sernarios in taking identical images with each camera, is something that I think would interest most. In my case, not as it relates to the video portion of the camera, but "still" images. Thanks!

Dave (D&A)

 

[MGrayson]

Dave,

I have no idea! It depends on how well the RAW converter works, and what the camera is used for. Another way to limit aliasing is by stopping down, which uses diffraction to limit high frequency information. It's very usage specific.

We'll see when examples from actual cameras used by photographers in different situations show up.

Best,

Matt

 

[johnnygoesdigital]

wouldn't the quality of the lens also effect the frequency of the data? I had the M9 with wonderful Leica lenses, and only had artifacts, but no moire. Does higher resolution with smaller pixel pitch, using the best lenses, reduce the chance of moire/artifacts?
I would like to use the D800/e for a mix of landscape and portrait, so I ultimately decided to order the D800 instead, as moire can be found at the pixel level and not completely eliminated with software. Can the assumption be made that it's easier to "sharpen" a D800, photograph to be more like the D800E in resolution, than to have an issue with moire/artifacts? Thanks-

 

[MGrayson]

Better lenses at wide apertures should just increase the color Moire and other aliasing artifacts. Seeing these artifacts depends on the scene. If there is little information at higher than sampling limit frequencies, then you won't see artifacts. If, on the other hand, you are shooting patterned cloth at the right magnification, then there may be a large amount of too-high frequency information, and that will cause very visible artifacts. Given the Bayer color pattern algorithms, I don't know of a frequency that would create only pattern but no color Moire. But then again, I don't know how RAW converters deal with Moire. They may recognize certain color patterns and correct for them.

--Matt

 

[Shashin]

For all practical purposes, the D800 is really sharp and the chances of moire are reduced. A spacial frequency filtering (unsharp mask) can make a D800 and D800E image very very close.

And pixel peeping becomes more and more ridiculous with more pixels. I did a little calculation at what I was looking at when I have an image from my 40MP 44mmx33mm sensor at 100% on my 24" monitor in Photoshop. I am looking at a 9.5mmx5.5mm section of the chip. Or, in a print viewing condition, I would be viewing a 44" by 33" print from 11" away.

I would not be worrying about the AA filter.

 

[D&A]

I would not be worrying about the AA filter.

Well thats the $64,000 question (as they used to say). I honestly don't think Nikon made the decision to offer the D800e simply as a marketing decision. I have a gut feeling that in more than just rare photographic circumstances, will the D800e display a certain gain in fine detail that simply can't be compensated by in the alternative D800 body, simply by increasing post process sharpening of the D800 files. Likewise, there will be occasions where moire' may rear its ugly head with D800e files and post prcessing softeware may or may not fully eliminate it. This is aside from use of these cameras for video.

I'm also not certain that we can fully translate the use of 40MP and higher MFD cameras and their being relatively free of issues without use of an AA filter vs. a 35mm full frame 36 MP CMOS chip with it's partiular sized pixel.

Shashin does make a good point about the relevance of examining at 100% vs. what is actually observed in a large format print....although I will say that often exmination of a file at pixel level has often given me a relativly good indicator of how one file's "detail" will print up (and look) vs. another file. It's somewhat relavent, but conversely its importance is often overblown.

Dave (D&A)

 

[johnnygoesdigital]

Matt and Shashin-

Thanks for the great input.

 

[Shashin]

I honestly don't think Nikon made the decision to offer the D800e simply as a marketing decision.

I think this was something that was worked out among the marketing, engineering, and quality control departments--I was fortunate enough to work with the product teams at a Japanese camera manufacturer, so I can imagine the process. Nikon has seen public perception of AA filter and they decided on this compromise solution of giving customers who prefer not to have an AA filter that choice. But likewise, if they find that moire is a significant enough problem, there is a way out, even if that is just telling the customer that we told you so--returns are a significant problem, especially with a product like this.

And then Nikon get to test the waters and see if dropping the AA filter has a positive effect on sales as well as complaints. If the D800E is a hugh success, good sales and few returns/complaints, Nikon will be more inclined to remove AA filtering from cameras.

But most photographers that really get into the photographic process tend to overestimate the significance of the specs and such. If you make an exhibition of photographs from high-end cameras, I doubt anyone can spot which photo came from which camera, let alone, which had AA filtering and which did not. That is not to say as a photographer if I were given the choice, I would not consider it. And I pixel peep too.

 

[tashley]

Thanks to everyone for such an enlightening discussion. Especially to Matt for his impressive array of graphs. Crikey!

I am mildly disappointed that this thread got moved to the Nikon forum though: it might at first sight look as if it's about the D800 but I'm fact it was intended to be a question very much for the MF crowd: is removing an AA filter something that only makes your images look great at 100% on screen or in very large prints, risking aliasing and the pp required to remove it, or does it genuinely improve an image printed at more normal sizes, for example a double page magazine spread?

 

[MGrayson]

I, too, was surprised that it got moved here. This thread may belong in a more general forum, but there's certainly nothing Nikon specific in the discussion.

--Matt

 

[Shashin]

Well, the Fun with the 645D got moved from the medium-format section to the Pentax section and there is slightly more connection with MFD with this.

I have been shooting with the 645D for a year and a P25+ back for three. Neither have an AA filter. I do studio work with the P25+--documents, portraits, tabletop objects. The 645D is mostly outside doing landscapes and such--both urban and rural. I have yet to see moire--now I have said that, every image from this day on will show it.

 

[tashley]

I get a lot of moire on the M9 but it's nearly always in repeating mechanical objects like rails and gratings. It can actually be a real pain.

 

[D&A]

I have been shooting with the 645D for a year and a P25+ back for three. Neither have an AA filter. I do studio work with the P25+--documents, portraits, tabletop objects. The 645D is mostly outside doing landscapes and such--both urban and rural. I have yet to see moire--now I have said that, every image from this day on will show it.

I think (as others have expressed) that the smaller the sensor (or should I say the smaller the pixel pitch), the increased likelihood of certain types of images displaying moire. Thats why I think sensors like those in MFD cameras, even cropped MFD cameras like the 645D etc. have the propensity to display less moire than those with 35mm full frame sized sensors, besides the increase in actual # of pixels on their sensors. Whether increasing the resolution to 36MP on sensors the size as those contained within cameras such as the D800e helps reduce moire as compared to their lower MP full frame 35mm counterparts is also being discussed. Then there is the issue of CMOS vs. CCD with and without the AA filter and does the type of sensor make a difference. One would think not, but who knows?

I agree, the decision for Nikon to go ahead and market two types of D800's, is part marketing, part testing the waters for possible future "increased" sales of having other DSLR's without a AA filter and part that it does offer a small but readily detected increase in resolved detail but not without the occasional potential issues we've been discussing.

Dave (D&A)

 

[pophoto]

It is generally said that when the repetitive patterns exceed the resolution of the camera, that Moire will be apparent. One way to solve the issue with the camera is to zoom in to get a tighter crop, but what if this isn't the image that you want!
So a higher resolution cameras will see less likelihood of Moire.

While it seems to be the concern here for the D800, I think there are other things to consider like Noise. However, from some early examples it proves otherwise too. Performing against the D700 in Noise equally as well or better!

I think I read somewhere that the sweet spot Noise/high ISO and IQ in general for pixel pitch is 5 microns. Anything less is regarded has Noise territory! Also I think most MF backs have from 9microns the AptusII 22MP to generally 6/7 microns for the larger MP MF digtial backs of 40MP up. But we know most MF cameras cannot shoot at very high ISO either!

Generally, the larger sensor has larger capacity to collect photons. The analogy is a large mansion with 3 bedrooms will store much more than a small house with 3 bedrooms, you cannot change or argue that! Hence, cleaner higher ISO images or greater dynamic range as a result. Obviously there will also be advancements in technology over time in sensor technology but say differences in size of sensors of 10% might be something negligible difference, but 50% or more (FF vs MF) might be something that can never be matched and so on. Also below 5microns seems to rapidly degrade imagery whereby approaching 2-3 microns is seen in many pocket camera qualities.

Although it may be argued that pixel pitch isn't telling of everything, I think it can be regarded as rough guide, similarly, 50mm FF lens of aperture f/0.95 will have a shallower DOF compared to a 50mm f2.8 lens say!

Again, I regard it as fun facts, but a good rule of thumb assessment of quality. So while we can make sense of a lot of theory with examples to prove them to be true, there isn't a be all end all explanation for everything, otherwise that would be too easy!

Unfortunately seeing is believing (cameras )! We'll just have to wait and find out for ourselves on a case by case basis!

 

[weinschela]

My only basis of decision here is that I have used Leica M8 and M9 whopich have no AA filter and produce razor sharp images (obviously tge glass counts). Moire has not been a problem. I therefore ordered a D800E.