« D800E Day 4: ER a.k.a. DR | Main | Veterans Day »

Monday, 12 November 2012

Comments

So it sounds like the moire is the result not of the camera but of the algorithm used to resize the image. Yes?

As a matter of fact, on the OMD you can sometimes use moire to confirm focus on certain fabric or screen type subjects. If you see moire shimmering in the viewfinder, it's in focus.

But it has to do with the interaction between the subject and the pixels in the EVF. Doesn't show in the picture.

Prediction: You'll be a D800E owner by week's end.

Thanks Mike, you really provided an excellent answer to my question. Now I can put the D800E on my fantasy wish list. If only I could get one with fantasy money....

Must be something wrong with me. I, um, kind of *like* the screen shot. (The first one.)

p.s. I've just gone through a little of this testing a D600 and if it's any consolation I've been arguably a far LESS rigorous tester than you...

Actually internet lore simply says that you will get moire. Not how much you'll get. No AA filter and the resolution isn't high enough to make diffraction into an AA filter.

Careful with those straw men. In winter if they get too close to the fire...

Mike,

I'm not fully sure that your d800e shot shows what I would call moiré, at least not as camera-testers or photographers mean the usage of the term. In terms of camera testing, moiré refers to the interaction between the subject and the sensor when the frequency of the subject's pattern is very slightly different than the sampling frequency of the sensor.

As I interpret your shot that's trying to show moiré from the d800e, it looks to me like you've got a photograph of the "subject moiré" that comes from the interaction of the two screens, not an interaction between one screen and your sensor. I'm seeing much more of a circular moiré pattern where the two screens interact than I am below or above the secondary screen. There might be something of a moiré pattern below the secondary screen in the single screen, but it's not too strong.

As is often the case, Wikipedia (http://en.wikipedia.org/wiki/Moir%C3%A9_pattern) has a good explanation of how moiré can show up "in real life", not just in digital photographs. My favorite example of it in real life came when I used to drive the freeways of Southern California: as you drive under the overpasses, most overpasses will have fencing on either side and as you drive past, changing the angle of your observation, the moiré pattern changes. In real life. Looking with only your eyes, not a sensor. I think your "torture test" is a photograph of "real world" moiré that you would also see with your own eyes.

In contrast, your Olympus shot looks like an example of sensor moiré to me, as it looks like you're shooting through a single window screen.

Don't forget that diffraction or small errors in focusing can be enough of a blur to counteract sensor moiré, just as the anti-aliasing filter adds a little bit of blur.

The takeaway point, for me at least, is that I wish that other camera manufacturers would allow us the option of cameras without anti-aliasing screens and let us take the chance on moiré.

When the moon hits your eye like a big pizza pie....

aliasing artifacts occur when you sample a high frequency signal at a sampling rate lower than the frequency of the signal. in the photographic world this means when you take a picture of fine details that exceed the resolution of the camera sensor. AA filters act as low pass filters that prevent any details at spatial frequencies higher than what the sensor is capable of recording. the type of moire you are showing has absolutely nothing to do with what an AA filter is designed to prevent. what you are seeing is aliasing patterns due to downsizing. when you drastically reduce the resolution of an image you (obviously) reduce the maximum spatial frequency of that can be shown in the image. if you want to see the type of moire caused by the d800e's lack of AA filter you need to actually exceed the the sensors considerable resolution. that basically means backing away from the screen so far that the sensor cannot separate the lines of the screen. you can only see the aliasing artifacts for what they are at 100%. when you downsize you are resampling the data creating more confounding artifacts (one way to avoid this is actually including a blurring step prior to downsizing).

the biggest problem found with sensors without AA filters is color moire where you get a colored aliasing pattern appearing over a repeated pattern (like fabric or a window screen that exceeds the sensors resolution). if you know what you are doing, such moire patterns are easy to provoke even with a lot of cameras that have AA filters (AA filters can't physically be perfect and the strength of AA filters seems to vary between manufacturers and models). people who have to deal with them regularly (eg fashion photographers) will, of course, know when they'll occur and how to deal with them.

ahem. that's the "wrong kind of moire" (not on the sensor, but on the display/jpg). mike? is that you? :/

Thanks for all these reports, Mike, you got me curious. I am going to try the '800e with the 50/1.4 G for a little while. Never shot with that many pixels or lack of AA filter, so I'm excited now.

Thomas,
Right, but good luck with that. You'll note that this post talks about moire and not about aliasing. The D800E has so much resolution that true aliasing is going to get buried by the stochastic printer ink dots at any size I could think of printing. Like I said, I couldn't get anything to show true sensor aliasing for love or money, and I'm tired of looking for it. The pixel peeping is making my eyes tired.

Ken Rockwell has a nice pair of pictures demonstrating that moire is in fact worse with the D800E. My claim here is that this is however NOT a practical problem of any magnitude.

Mike

To create moiré intentionally (with any DSLR) just find the reproduction ratio where a repeating subject pattern matches the pixel pattern and move back / zoom out a bit (or slant the subject a bit to make the pattern tighter). I've occasionally used such shots to test the demosaicing algorithms of RAW converters. (Btw. appearance of moiré depends a lot on the RAW converter used.)
I've seen (admittedly tiny) patches of moiré in ripple patterns in a river as well as (larger patches) in feathers ...
And finally I've yet to see someone demonstrate an actual advantage for the "E". Comparison shots supposedly showing "a small but clear advantage in resolution by the D800E" were just lacking some USM (0.5 px radius, 10%) in the "non-E" shot. And used at that rate I don't thing I'll run out of USM anytime soon ...

So for anti-aliasing you do not agree that less is Moiré ?

Or to paraphrase Bob Dylan "ain't gonna work on jaggies harm no Moiré"

It's not just a sensor or AA filter thing. Moiré shows when a high resolution image is seen in small sizes. Fortunately it goes away when the picture is seen in large sizes, but thumbnails and the sizes at which photographs are usually seen on websites like Flickr can show it quite badly. A small example: I have a fascination for photographing at metro stations and airports, and all photographs that include escalators show moiré at the steps. (And yes - my camera, the Olympus E-P1, has an AA filter.) However, it does not show in prints, nor when seen on a computer in large sizes, but it always makes me wonder what people will think when they see the small images...

Darn you. Adorama has refurb D800s for $2500. I was going to wait for the next generation to upgrade the D700, but now I'm thinking about returning the NEX 6 and getting a D800 and using the D700 for the climbing camera.

Hi Mike
I remember well discovering your web presence via a mention in your UK Black and White magazine column.
One of the early articles you published on the web was about you having a print of an apple made at a large size which was shot on a Minolta camera with if I remember rightly was 5 MP or may be a bit more.
Well I was looking for a DSLR to make the leap from my Olympus E20 and having read your Minolta D7 review in B&W, seen the apple print and read how Sony were going to go big style into cameras in partnership with Zeiss lenses, I took the plunge into an A100.
I was happy with the A100 and even happier with the A700 and eventually the A900 and A850.
I can without much effort produce A2 prints, with a bit of effort A1 and have had 2mtr prints made from 3 stitched images, all from the Sony FF cameras, but the thought of an image from the 800 E at two and a half the size of your screen or as I have read 6 feet leaves me asking why.
I understand that industrial, architectural, fashion, product, landscape photographers etc. might want to exhibit at that size but for most amateur and professional reproduction purposes it is really overkill.
Also I do not shoot or want to shoot video so superfluous facility as are most of the other facilities beyond manual and aperture priority.
I do not shoot black cats in coal holes or speeding bullets so most of what the A900/850s can do fits MY bill.
I feel it is a shame that all the manufacturers cannot offer us a Nikon F2as/ A900 camera which is solid reliable and delivers the goods.
OKAY I can hear the dissenters listing all the A900/850s faults but it is a good solid camera that delivers if you use it properly, especially with Zeiss glass.
My hope is that my A900s(2) and 850(1) keep going as long as I can.
I love real glass bright viewfinders and whilst my OMD EM5's viewfinder is good I prefer the A 900.
Now of course all this is personal but whilst you have been guilty in the nicest possible way of introducing me to Minolta Sony I will not be going back to Nikon even though I have fond memories of the F/F2As.

David

To further push moire further down in the list of the things to worry about (and despite what Ken Rockwell writes about it) you can address moire in post production pretty easily. If it happened often, it would be a pain. But not an issue if it crops up occasionally.

I think the camera manufacturers have done a magnificent job getting us to demand the removal of AA filters. You can imagine the howls of protest about moire if cameras did not have AA filters by default.

"but the thought of an image from the 800 E at two and a half the size of your screen or as I have read 6 feet leaves me asking why."

David,
It's all about prints and who needs bigger ones. At 300 ppi, the 36-MP D800 still only produces a native 24-inch-wide print. In the art world these days (where bigger = better) that's not much, and advertising photographers routinely need more--that's why they use MF backs. We might be getting towards "overkill" territory for most amateurs, but then, most amateurs do not buy $3k camera bodies, either.

I was perfectly happy with the 3-foot-wide print Ctein made for me from a 24-MP A900 file--to me it stood up as close in as my eyes can focus anymore.

Mike

Ah.! Ctien the guru.
Was it his print of the apple I remember?
The Robert White dealership in London produced as part of an exhibition a 6 feet wide print from an 800e married to Zeiss 50 macro with very VERY little tweaking.
As you say Mike it is down to ones eyes but proper processing is paramount.
Do you remeber the Kodak ad for Ektar colour negative film which reproduced a 35mm frame as a 48 sheet poster and a very large illuminated panel in Time Square.

Regards

David

I find it very touching when such an experienced guy gets so much excited about the new gear :-)

During the first week I had a D800 I took a picture using a lens that on the Internet is not considered the sharpest, used on a real scene, not a test, and got moire. Now I might have a computer science degree, know something about sampling and see moire potential everywhere, but every week I get moire in my pictures. Since I bought a D800 instead of the E version since I can't stand moire and it can't be removed without serious blurring, the moire I get is fairly mundane and not a problem. Now I readily admit that whether moire will be a problem depends on one's style of photography and technical skill, so it shouldn't be assumed to be a universal constant. But when moire gets into pictures, and in some styles of photography it's easy to get, it will require sacrifices to remove, sacrifices involving gaussian blur.

About moire in nature, I distinctly remember one autumn morning with a Nikon D300 where I got moire from very small ripples in water. I believe the converter was Capture NX, another converter may have rendered it differently and lost sharpness. Not a big deal, but moire can indeed occur in nature.

Back in the early 90's, the print shop I worked in started converting to Digital from the older film plate making. Film screens were crossed to create halftones of the photos. I would think you knew this from your years in the industry. The critical issue with the screens is when they crossed the angle of the lines could not be less that 30 or 33 degrees. This induced the pattern in creating a photo. The screens had pin locators to prevent the angle from being altered. I would suspect that the same can be said for a line of pixels trying to form a line of a building or a series of steps in a stairway.

The camera sensor may not be actually creating the pattern at 100%. That is where it matters when you go to print it. A monitor display will probably have a different resolution than the picture and will artificially induce the pattern. With photoshop, the odd number size display at less that 100% will usually show the pattern on the monitor, even though it is not there at 100%. With the small Pica sizes in inkjets, that pattern should not even show in a translation of dot lines from pixel lines.

"We might be getting towards "overkill" territory for most amateurs, but then, most amateurs do not buy $3k camera bodies, either."

So what are the market breakdowns for the different types? You could look at it with Cell Phones and P&S included. Or you could exclude this group and just stick with DSLR sensor size as the market separations. Does mirroress skew the market demographics?

I got Moiré patterns from the ripples on sand dunes with my Nikon D3 and a 35mm lens. But only in a small area where the ripples receded into the distance and became--from the camera/lens POV---very small. As you have stated, the Moiré patterns were easily lost by the printer dots at any thing but a very large print size.

Something to consider before making a final decision...

"DON McCULLIN’S ‘JOURNEY INTO DIGITAL’"

http://cpn.canon-europe.com/content/Don_McCullin.do?utm_source=newsalert-november-2-2012&utm_medium=email&utm_campaign=Newsletter

For shame, Mike. Are you telling me you don't have a couple of extra Gordy's wrist straps around? http://www.electricedge.com/gordy_s_straps/
Every time I buy a camera, the first thing I do is order a custom wrist strap from Gordy. In fact I bought 3 just in the last week. Best wrist straps on the planet.

This whole exercise reminds me of when I rented a Mamiya 6. The rental period gave me a million reasons to get a 6, but I never did. I predict that if Mike spent one of his five days "testing" the camera, he won't end up buying one--it's just not a tool that compels him to use it.

I have a D800E and I do occasionally get moire (the full resolution, Raw file, type of real one). What I have noticed is that I do not get the moire in any patterns that I would expect with. Where I get moire is what looks like flat surface to me. My explanation for this is that if I can SEE the pattern, it is far too big to challenge camera resolution. So very fine patterns, which are not visible to my eyes, may show moire.

The above is a generalization, I have see few exceptions to this. In one case a very minor bit of moire on bird feathers, and on some clothes. In all cases the objects were significantly farther from camera, and the pattern was not certainly visible to the eye.

Relatively distant Aspen tree trunks create moire on my Leica with 24mm. What should be a regular, vertical pattern with white and dark shows up rainbow-colorful.

I got a Pentax K-5IIs just recently, a replacement for my aging K10D. It, too, lacks an AA filter.

As luck would have it, I got a prime example of moire on my third frame with the camera; an inconsiderate pedestrian with a violently patterned shirt. I got a second example last week, shooting a tunnel with a striped wall pattern receding into the distance.

In neither case was it an actual problem, though. The shirt could be fixed simply by smoothing out the color channels to avoid the oil-slick effect; the tunnel will show interference at whatever size I rescale it to anyhow, so no point in fixing the original.

Now, if I had been a fashion-guy, and interested specifically in that shirt it would have been a real problem. But I'm not, and moire turns out to be a non-issue for me even when I managed to get it in my shots.

Meanwhile, I really appreciate the camera. I love being able to shoot at iso6400 without worrying about noise; getting jpegs out of the camera good enough that I now rarely bother with RAW anymore (that was not the case with the K10D); and the compact size, especially with a small prime, means I can bring it in my bag without having a separate camera case for it.

Mike,

not really sure what you're wishing me good luck with? trying to produce moire?

i certainly agree that moire with the d800e will be a non-issue for most photographers (however your test does nothing to show this). moreover, the ones for whom it will be an issue will already know how to deal with it. what i don't understand is what you are trying to show with your test? none of your examples say anything about the importance of an AA filter or lack there of. what they do address is the importance of using proper downsizing methods to avoid aliasing (moire is a type of aliasing incidently). that doesn't really have anything to do with the d800e any more than any other modern digital camera though.

if you wanted to show that moire isn't really an issue with the d800e a more useful thing to do would be to take a picture that actually had some sensor aliasing in it (instead of shots that had no chance of producing it) and downsized them to show at what print size the aliasing would be unnoticeable.

in your addendum that you added after my original comment you said: "You're just not going to see sensor aliasing with a camera with this much resolution unless you're a nut about it—you'd have to work very smart and very hard to bring it out in a print (which is the only place it can matter, because when do you need to show an image at an exact resolution online, at a size that greatly exceeds the size of any monitor?)."

as an inhabitant of some of the geekiest corners o' the tubes, i feel compelled to remind you that this statement is not really true. the most obvious type of sensor aliasing is color moire, which can produce large swaths of color where there was none in the original scene. this false color can easily remain quite visible in small prints and web sized images. such false color can often be produced in a typical cityscape, so it doesn't take much skill or thought to produce.

Chicken Invaders, Ha! No one else I know that hasn't been introduced to the game by me has had the pleasure of wasting many satisfying hours cracking eggs and toasting those chickens! Great to see another random fellow gunner out there.

Mike, dark and rainy is great photo weather!

Re: Moire in nature--I can get moire from woodgrain, feathers, and hair. Even at f11 on a 5D Mark II. I could probably get it from rushes, grasses, and reeds if I shot them more often, and some arthropods if I tried hard to find just the right distance and aperture.

Outside of nature, any building with parallel lines can produce moire, depending on your shooting angle and focal length. Any sharp edge can also produce aliasing artifacts, if the conditions are right.

Dear Folks,

I'd like to point out that the resolution gains from not having an anti-aliasing filter are typically modest. It's kind of like going from a 12 to a 15 megapixel camera. That's nice but it's not earthshaking, and a lot of the time you won't even notice the difference. Generally, there's no reason to get super-excited about it, one way or the other.

(Yes, I am aware of the strikingly different comparison photos on Luminous Landscape. I assert there is something wrong in that test. The D800 sections don't merely look fuzzier, they look SMEARED. An AA filter does not typically produce that kind of difference in appearance.)

pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================

If you do end up buying a D800 or D800E, I would recommend that you spend some time examining results from both before deciding, or alternatively, that you also rent the D800 for a week. I initially ordered the D800E, but given that I only really wanted a 24MP D800, and given the extraordinary pixel-level sharpness of the plain D800, I changed my mind; the first time in recent mind where sense won out over desire when it comes to my photography habit. The D800 is so sharp that I had to significantly weaken my sharpening routine, developed for my D3 (multi-step routine, and although there are more steps now to account for the higher resolution, each step had to be weaker to avoid artifacts).

But are you really sure that you will carry such a camera? It is large, heavy, expensive and obvious. I love the results, but I do keep an E-PL3 around for times when I want less bulk and weight. I plan to replace it with an E-PL5 soon, and hand the E-PL3 to my girlfriend, who finds that she rarely uses her otherwise beloved D5000, because of size. She has been eyeing the E-PL3 for a while now, and admiring the results I can squeeze out of that little camera (mostly photos of our daughter).

Ach, now I forgot to mention one thing: I find the pixel-level detail of the D800 more pleasing, if less sharp appearing, than that of the D800E, and it sharpens more cleanly and smoothly. There is something artificial-looking about the latter, which could be false detail.

Also, check out Ming Thein's comparison of the D800E to the Leica S2, for some moire.

Ctin

" The D800 sections don't merely look fuzzier, they look SMEARED. An AA filter does not typically produce that kind of difference in appearance"

Knowing nothing about the d800 other than what I read, my understanding is that the AA filter is made of a birefringant filter which would "smear" the image rather than apply a gaussian fuzz.

It makes me think that there would be an oppertunity for overs piling by using an active Kerr cell in that position. Or just use one of those sensor wiggling systems.

Haven't read everything here but I suspect you couldn't get any moire the whole time because that 35/1.4 lens is not up to the task. A few zooms are way better in resolving power as tested by Roger Cicala.

I have not yet ordered a new UpStrap for my D800 and almost lost the camera for not having done so. I slung it on it's Nikon strap but being used to more grip didn't pay enough attention to my stance. It slipped and fell down some concrete steps. Nikkor lens hood is wrecked and the camera now has some more unique character marks. Luckily it still works.

Note to everyone, do yourselves a favour and try out the UpStraps.

The comments to this entry are closed.