Why don't we need an anti-alising/low-pass filter in front of the sensor anymore?

Golf007sd

The best way to hedge against a shot that may show up with some level of moire is: have a second body that has the AA on it's sensor and if possible take your laptop with you on location and tether. As for my recommendation on the second body buy a D4. :P

God I'm such a Geek. #-o

DaveyJ

moire is a problem for me on Nikon DSLR long before I pre-ordered the D7100. When I get that camera I will look out for that issue to be sure. I have frequently noted that effect in the past and where possible, just junk that problem image. I shoot usually in JPEG Fine, large usually. I understand if I do get moire problems that even less than perfect fixes are out with the D7100 if I do not shoot in RAW. So all around elimination of the AA Low Pass Filter looks like a gamble.

donaldejose

Lots of discussion about the presence or absence of moire but no answer yet as to why Nikon is removing the AA filter on its three most recent sensors. Any ideas? Size of pixels? Software in camera? Just discovered it occurs so infrequently the trade-off of sharper photos all the time is worth some moire rarely? :-?

When the D800.800e choice presented itself some "experts" recommended most people get the D800 unless you shot landscapes or wildlife most of the time where regular repeated patterns would not occur. Now Nikon seems to be saying: in effect forget the D800, just shoot with the D800e all of the time. Is there something different with eliminating the AA filter on a D800 than there is with eliminating it on a D7100?

Ade

I hadn't seen John Wright's work before (thanks for the new name) - there is moire in quite a few of his images and you can pick out some of the editing.

Which of John Wright's images have sensor moire?

The only way to evaluate moire is at 100% of the unprocessed image. I wasn't aware that John Wright has been making "quite a few" of his D800E RAW files available online.

Msmoto

You folks are too funny....but I am busy with some more "poet" processing..

heartyfisher

Lots of discussion about the presence or absence of moire but no answer yet as to why Nikon is removing the AA filter on its three most recent sensors. Any ideas? Size of pixels? Software in camera? Just discovered it occurs so infrequently the trade-off of sharper photos all the time is worth some moire rarely? :-?

When the D800.800e choice presented itself some "experts" recommended most people get the D800 unless you shot landscapes or wildlife most of the time where regular repeated patterns would not occur. Now Nikon seems to be saying: in effect forget the D800, just shoot with the D800e all of the time. Is there something different with eliminating the AA filter on a D800 than there is with eliminating it on a D7100?

I guess you didn't believe my Matrix proposal ! :-)

BTW the D800e still has an AA filter albeit a reduced one while the D7100 does not have one at all. so the situation may be "worse" on the D7100 except that it has a higher pixel concentration/resolution.

One other explanation I can think of is that because the AA filter in effect blurs the image, it is no longer needed bec the majority of the lenses that the consumer would put on the D7100 will not be sharp enough for moire to manifest. ie the D7100 out resolves most lenses.. and diffraction above F11 blurs the image enough that any lens will not need the AA filter anyway at those higher apertures.

Therefore if for a particular shoot where you want to be sure that there is no moire then just use the higher apertures if you don't need the thin DOF or shoot wide open if you do.

donaldejose

All we have so far are guesses. I suspect Nikon has a specific scientific reason. But no one has yet been able to establish what the reason is. Perhaps as the D7100 is reviewed and written about someone will be able to get Nikon to tell us why they removed the AA filter as no longer needed. I had hoped someone would be able to point to a definitive answer from some Nikon source.

PB_PM

@donaldejose I think it's far more likely to be a marketing ploy, rather than out of any engineering breakthrough.

TaoTeJared

One other explanation I can think of is that because the AA filter in effect blurs the image, it is no longer needed bec the majority of the lenses that the consumer would put on the D7100 will not be sharp enough for moire to manifest. ie the D7100 out resolves most lenses.. and diffraction above F11 blurs the image enough that any lens will not need the AA filter anyway at those higher apertures.

Sorry Heartyfisher but that statement is false. Almost every lens made today can resolve 150+ lines/mm, sensors are not even close to that yet. To equal Velva 50 resolution which most lenses resolve on just fine, you are in the multiple 100's of MPs for that. (Sorry don't feel like doing the actual math.)

heartyfisher

One other explanation I can think of is that because the AA filter in effect blurs the image, it is no longer needed bec the majority of the lenses that the consumer would put on the D7100 will not be sharp enough for moire to manifest. ie the D7100 out resolves most lenses.. and diffraction above F11 blurs the image enough that any lens will not need the AA filter anyway at those higher apertures.

Sorry Heartyfisher but that statement is false. Almost every lens made today can resolve 150+ lines/mm, sensors are not even close to that yet. To equal Velva 50 resolution which most lenses resolve on just fine, you are in the multiple 100's of MPs for that. (Sorry don't feel like doing the actual math.)

Yeah I don't like maths too :-)

I had a look around to see what I can find regarding what you say.. ( I learnt alot about reading MTF charts and ppi and dpi! however that didn't help with the resolution factors we were discussing :-) ) The most useful info I found to substantiate my deductions can be found on the DXO labs lens tests.. If you look at some of the lens tested most of the Zooms are in the 5 to 6 PMPix range and even the macro lenses and primes are only in the 10 PMPix range.. the DXO PMPix is what they call the Perceived Mega Pixel resolution capability of the lens. So it would seem that any increase in MP above 10 is really wasted since most lenses cant resolve past 8 anyway.. and the 24MP on the D7100 would seem to be so much of an Overkill that it would make sense that the AA is really not needed.

http://www.dxomark.com/index.php/Lenses/Camera-Lens-Database/Nikon/Nikon-AF-S-DX-Micro-NIKKOR-40mm-F28G/(camera)/680

PB_PM

That doesn't make any sense. If lenses couldn't resolve more than 10MP (which is utter nonsense), detail would not noticeably improve between the D700 and D800, yet it clearly does.

Even if lenses could not resolve more than 10 MP, what does that have do with the AA filter?

TaoTeJared

DXO PMPix is DXO's own creation. If you have seen my other posts on DXO, you know that I do not believe they actually test lenses (especially on this "resolving" metric), but how sensors actually work with different lenses. It is a bogus measurement for anything other than being able to say "this lens will be slightly sharper than this other lens on the same body."

A good example - I grabbed the 85mm 1.8G - Switch between bodies (D7000 & D3x) and the score changes. By their description, you should loose a bit of "perceivable mp" but on this lens it seems to somehow add mp to a sensor. That is one of the main reasons I believe DXOs lens tests need to be updated.
http://www.dxomark.com/index.php/Lenses/Camera-Lens-Database/Nikon/Nikon-AF-S-NIKKOR-85mm-F18G/(camera)/680/(cameraname)/Nikon-D7000

Since they obviously don't keep every body and every lens laying around, they don't have tests on each. For instance they do not have the D800e testing any of the lenses. I am guessing the D800's have exaggerated their "errors" or have heightened the visibility of the limits of the tests that didn't noticeably show before.

heartyfisher

@ TaoTeJared : I dunno .. that PMPix doesn't seem bogus to me.. .. seems pretty sensible and pretty much inline with what I perceive to be quite detailed and thorough quantification of sensor and lenses by the company.

I had a look at the 85mm nikkor in your example for both the D7000 and the D3x .. ( what great lens that is! )
The PMpix for the D7000 was 11 and the PMPx for the d3x was 17.. I don't see anything wrong with those figures. Can you please explain what you mean by "By their description, you should loose a bit of "perceivable mp" but on this lens it seems to somehow add mp to a sensor." ?

@ PB_PM : "Even if lenses could not resolve more than 10 MP, what does that have do with the AA filter? "
What I am suggesting is that if the sensor can resolve 24MP and the lens can only resolve 10 MP then a line pair in the image will never be sharp enough to cover only 2 pixels ie what should be one black and one white pixel becomes one black and one red or green or blue causing moire. so if a lens can only resolve to 10 MP then one pixel will be spread over 24/10 pixels therefore reducing the occurrence of moire. note that only the best glass is producing 10PMPix most of the zooms are in the 5-7 PMPix ie one resolvable MP becomes 24/5 pixels. also the PMPix is also measured at the sharpest apertures of the lenses therefore at less optimal apertures the PMPix is probably even less. It would seem to me that this supports my deduction that the lenses are not sharp enough to result in Moire at the high pixel densities we now see in the D7100. Only the very sharpest Pro Prime lenses would manifest this Moire at their optimal apertures. Considering the target market for the D7100 i think nikon saw that most would not be paying on average $2-3k per lense to put on a $1K camera... that seems logical right ?

donaldejose

I do suspect the answer to my original question has to do with pixel density on the senor (which is what heartyfisher is also trying to express) but I don't know the math or the resolving power of lenses. My concern with heartyfisher's attempt to find an explanation is that if lenses can only actually resolve about half of what the sensor can resolve why would Nikon waste all the time, effort and money producing 24mp sensors when a 12mp senor would be able to capture all a lens can resolve? The "extra" 12 megapixels must be doing something to improve the image. Clearly, there is added detail in images produced by sensors over 12mp. That detail must get onto the sensor by the ability of the lens to resolve it. Therefore, I think "ordinary consumer" lenses must be able to resolve much more than 10 mp can see. I can believe that about 24mp in DX and about 50mp in FX may be some sort of practical limit on sensor density but I have to admit those numbers are about as solid as a fart in the wind.

MikeGunter

Hi all,

I'm not convinced that there's no problem, yet. The camera isn't on the market; there's no pictures or videos from a production model to make any judgments.

There are two different media to determine moire, not just still, but video as well. if video doesn't concern you, then you don't need to comment on it, but those who do need to consider the professional aspects of broadcast production work which do not and will not accept moire frames in video.

Last year alone, my poor man's D7000 produced footage for CNN, ABC, PBS, Outdoor Channel, and more. If I upgrade to a D7100 without an AA filter that produces moire, that isn't an upgrade.

My skepticism has an honest reason.

My best,

Mike

donaldejose

Yes, skepticism is warranted until proven otherwise; both for stills and video.

heartyfisher

@mikegunter : I dont think you need to worry about moire on the d7100 if our assumption regarding the removal of the AA filter is correct. And even if it is not correct. Video is only 2mp therefore each video pixel is made up of 12 raw pixels on a 24 mp camera. In those situations I cant see that moire will be an issue at all.

MikeGunter

Hi all,

@heartyfisher

I dont think you need to worry about moire on the d7100 if our assumption regarding the removal of the AA filter is correct. And even if it is not correct. Video is only 2mp therefore each video pixel is made up of 12 raw pixels on a 24 mp camera. In those situations I cant see that moire will be an issue at all.

I'm old enough to stop assuming a long time ago, besides, I've no clue how those pixels will be processed, I know that DSLRs have real and significant problems with aliasing right now.

I hope I'm wrong, and I'm quite willing to be wrong. I'd like to see some video from the D7100 of straight lines, perhaps of an escalator going up and down - a lovely test if every there was one - some other real world example, but not some theoretical talk made up of whole cloth - the camera is, at moment, in pixels itself - there isn't a production model in anyone's hands that's showing what it can do until next week it seems, and until something real shows up, I'll remain at least skeptical.

My best,

Mike

heartyfisher

That doesn't make any sense. If lenses couldn't resolve more than 10MP (which is utter nonsense), detail would not noticeably improve between the D700 and D800, yet it clearly does.
...

The lenses clearly can resolve above 10 mp . The example from the 85mm 1.4 clearly shows this. I think the confusion comes from the fact that for a given lense the pmpix for fx and dx sensors are different. Again using the 85mm 1.4 example on dx that lense resolves 11pmpix while on the d3x it resolves to 17pmpix. I think once you understand why that is so it will be clearer.

Ade

@mikegunter : I dont think you need to worry about moire on the d7100 if our assumption regarding the removal of the AA filter is correct. And even if it is not correct. Video is only 2mp therefore each video pixel is made up of 12 raw pixels on a 24 mp camera. In those situations I cant see that moire will be an issue at all.

There are many causes of moire. Exceeding the sensor resolution is only one factor.

For DSLR video, the high megapixel sensor must be downsized and then compressed in real-time. But the downsizing algorithm -- such as line skipping -- can itself produce moire. And on top of that, the compression method may introduce jaggies and other digital artifacts into the video stream.

Therefore, regardless of AA filters or effective sensor resolution, with DSLR video there is always a risk of moire and other artifacts.

Having a strong AA filter does not necessarily mean less moire. Conversely, having a weak AA filter (or none at all) does not necessarily mean increased moire. The total video system design needs to be considered.

Practically speaking, the only way to determine the amount of moire a DSLR will produce is by hands-on testing and evaluation.

aquarian_light

Wow... just wow... all of this.... OP, nikon took off the AA filter because AA filters are an invention of the digital age. Film never had them. Others are partially right right, it does produce noticeably sharper images. All these guys who don't have one are just speculating. My D800E has been put through near 6k frames, like a few have said, and in those 6k frames maybe 65% have been of people with clothing, and not one has shown moiré. I even once tried to test it and force a moiré. I lined up several different kinds of fabric and had different kinds of lights set up and took images from all different angles. I couldn't even force a moiré. I suspect that Nikon understands this. You might find one moiré in 10,000 frames, and even then it would be a throw away anyway because you focused on the models clothing instead of her eye.
So all these people throwing around facts and b***s*** are just pixel peepers. In real life, a working pro looks at prints, not pixels. And the D800E will show you better prints than the D800. Might be a small improvement sure, but knowing what look for and where to look on the prints, the extra cost was absolutely worth it.

MikeGunter

Hi aquarian_light,

How does your video look, especially lines?

My best,

Mike

donaldejose

aquarian_light. Fine, good example and good experience. But why is the AA not needed anymore? Could it have been taken off the 12mp D700? Off the 12mp D3? Off the 16mp D4? Or with 24mp on DX and 36mp on FX have we reached a point in pixel density where moire is so unlikely to appear we can enjoy the benefit of increased sharpness and not worry about moire production? From the responses I have seen on two forums where I have presented this question the primary factor in no longer needing the AA filter seems to be pixel density. Do you agree with that? If so, will we likely see a high megapixel D4X without an AA filter? Likely see a 24mp D400 without an AA filter? Likely see the AA filter removed from the D3300 and D5300 if they both stay at 24mp? Likely see the AA filter remain on a 16mp D4s? If pixel density it the main factor than those previous questions reasonably follow the analysis.

aquarian_light

Hi aquarian_light,

How does your video look, especially lines?

Sorry, don't do much video and haven't tested it. I do stills 99.9% of the time.

aquarian_light. Fine, good example and good experience. But why is the AA not needed anymore? Could it have been taken off the 12mp D700? Off the 12mp D3? Off the 16mp D4? Or with 24mp on DX and 36mp on FX have we reached a point in pixel density where moire is so unlikely to appear we can enjoy the benefit of increased sharpness and not worry about moire production? From the responses I have seen on two forums where I have presented this question the primary factor in no longer needing the AA filter seems to be pixel density. Do you agree with that? If so, will we likely see a high megapixel D4X without an AA filter? Likely see a 24mp D400 without an AA filter? Likely see the AA filter removed from the D3300 and D5300 if they both stay at 24mp? Likely see the AA filter remain on a 16mp D4s? If pixel density it the main factor than those previous questions reasonably follow the analysis.

1-first lets talk about how moiré happens. On the sensor there is what's called a Bayer away. A red pixel, a blue pixel and a green pixel. The processor then blends these three pixels to one output pixel. Moiré occurs when there is detail landing on the sensor that is finer than the actual pixels. Say a point of lands on the blue, but not the red or the green. This confuses the processor and creates patterns that aren't there. Now, the point of light needs to be bright enough and repetitive enough to create a pattern of this confusion in the processor. AKA a brightly lit piece of cloth that produces a moiré pattern, won't in a dimmer light or in light of a different color because it removes the pattern of confusion on the sensor. Now inorder to prevent this confusion from ever happening, the AA filter applies a micro blurr to the image so that a point of light that may only it one sensor pixel of one color, is ever so slightly so that point of light hits all three colored sensor instead of just one.

2-Now a camera is not in need of one because in any realistic situation there is not going to be any point of light small enough and bright enough to excite this confusion in the processor. So yes, pixel density has ALOT to do with it. 12MP dx or 16MP fx are not dense enough to realistically avoid moiré, where 24MP dx and 36MP fx are. So it is very likely that anything going into the future at or above these resolutions are not likely going to have an AA filter, because the micro blur it adds does nothing but reduce sharpness. That and it is exceedingly difficult to manufacture AA filters that micro blur on such a small scale as the individual pixels on modern sensors.

TL;DR Yes, pixel density has everything to do with it.

Ade

See my post above your first one. The cause for moire in video is not necessarily the same as moire in stills. One cannot automatically infer the presence of one from the other.