Canon EOS sensors, and technology

[jrista]

First, you put too much weight on DXO's numbers. As far as their sensor tests go, they do not actually measure "sharpness" or anything like that. It's actually extremely difficult to objectively test a sensor in terms of sharpness, as you have to use a lens to do so, in which case your not testing a sensor, your testing a sensor and lens combined, which totally changes the outcome (and the reasons why you get that outcome). The other problem with lens+sensor tests is they are bound by the least capable component...if the lens is the weak point, then no matter how good the sensor is, your output resolution is limited by what the lens is capable of...you can never resolve more than the lens resolves, period. Similarly, if the sensor has limited resolution and the lens is a powerhouse (like the Zeiss Otus 55mm f/1.4), then your output resolution is limited by the sensor...you can never resolve more than the sensor resolves, period. That makes determining how sharp a sensor is a very muddy issue, one that cannot be definitively pinned down. Hence the reason DXO measures things like SNR and dynamic range and color sensitivity in it's sensor tests...that's all they CAN measure.

This is not news. Lets leave DxO out. I am sorry I even mentioned them for this discussion. I mixed some info and used them as a point of reference. Its something that can happen with the amount of info I go through. Please accept my blunder as a simple error as DXO is often the reference point for sensor quality, and I understand it is not of sharpness of the sensor, and more with DR and ISo, and the related.

[/quote]

It's not a problem to use sources like DXO as a point of reference. It just helps to have all your facts strait before doing so, so you don't mislead or confuse or otherwise sidetrack readers with incorrect or inconsequential information. ;P

Regardless of what DXO has to say about the D800 or D800E sensors, the removal of an AA filter does not increase image quality. Actually, in all too many cases (quite possibly the majority of cases), removal of the AA filter is guaranteed to REDUCE image quality, thanks to increased aliasing in general, moire specifically. This is clearly evident by all the numerous standardized image tests done with cameras over the years...while sharpness has increased in some newer cameras by a small amount, so too has moire. DPReview has plenty of examples where the removal of AA filters in Nikon cameras, or even just the weakening of the AA filter in many brands (including Canon) has greatly increased the amount of moire that occurs. (A great baseline for comparison on DPR is the 7D...it has an appropriately strong AA filter and doesn't suffer from moire at all. You can compare any newer camera with a sensor that is supposedly "better" than the 7D because of the removal or weakening of the AA filter...those images will be sharper, but they are usually riddled with moire.)

while sharpness has increased in some newer cameras by a small amount, so too has moire

The moire is subjective. I'm not too interested in the DPReview samples showing loads of moire issues. I have plenty personal samples I can stand by to tell you otherwise. Many samples in those cases are looking to show moire, and samples of it.

Actually, moire is a concrete, immutable artifact of repeatable patterns near nyquist interfering with the sensor grid. It not only leaves behind funky color and monochrome patterns...they are neigh impossible to correct in post...there IS no full moire removal in any RAW editor for a very good reason: It's impossible. You can reduce color moire, however depending on the tool, you might end up with color desaturation, blurring, etc. as a result. Even after removing color moire, the underlying monochrome moire pattern remains, and it cannot be removed (at least, not without significant blurring.)

We aren't talking about a subjective factor if IQ here, we are talking about a detrimental, and permanent, factor of IQ that gets introduced when the AA filter is removed. The DPReview sample images are not intentionally trying to show moire...they are simple sample shots of their standardized test scene. Moire occurs in their samples as a CONSEQUENCE of weak or missing AA filters. You can't simply brush moire and aliasing to the side and call it a non-issue...it is a critical issue to a great many photographers.

removal of the AA filter is guaranteed to REDUCE image quality

Be more specific. As with this statement, in this discussion you are saying that fullframe or larger sensors that are not using AA have lower image quality. How do you figure?

I explained it in the text you failed to quite.

If the things you photograph have no regular/repeating patterns, and do not contain any elements with clearly defined edges, then increased aliasing due to having no AA filter is not an issue. There are not very many forms of photography where that actually turns out to be the case...landscape photography is probably one of the very few. Even say insect macro photography, for example, will suffer from the removal of the AA filter...things like antenna, feelers, legs, wing veins, anything thin, strait, with high contrast to it's surroundings will end up with clearly aliased edges, and not even a highly optimized AHD demosaicing algorithm will be able to hide that fact.

The underlined falls under EXACTLY what I shoot on a regular basis, and I, with all the respect I have for your knowledge as I have read much of your posts, I think you are simply flat wrong about this. I have worked with about 20 digital camera systems in the past 24 years. I certainly don't have the understanding of sensors, and electro engineering you do, or even in the realm of it. I know I have shot just about everything there is to shoot, and I specialize in macro work WITH dealing of " thin, strait, with high contrast to it's surroundings ". I uesd the Kodak 14mpixel SLRc camera, and if it didn't have issues with handling light, I would continue using it. The images from that didn't suffer the things you claim. Nor do the MF backs, tossing the optional AA filter aside. (never used one to this day). Has moire EVER happened? Yes. Can I remember it being a problem or can I even count on my 10 fingers vs over 400K frames (with half using filter free cameras)? NO.

It's fine to have personal preferences. To base the entire discussion of "Canon EOS sensors and technology" solely on your personal preferences kind of makes it difficult to have a coherent discussion. Your personal preferences should really be left out of an objective discussion of the fundamental technology behind sensors, otherwise were just in the muddy territory of subjective muck, and anyone can make any argument to justify their own personal opinions. I personally try to remain objective when discussing technology, and leave my own personal preferences out of the discussion.

Regarding whether moire is a problem on MF cameras, Leicas, etc. If you do a few quick web searches, you'll find that they are a huge problem. There are countless threads on the subject, dating back many years, with MF and Leica users (and increasingly Nikon users) complaining about how bad the moire and aliasing can be on their incredibly expensive cameras. The solution, for many, is to use the lens so act as the AA filter. Either stopping down beyond the diffraction limited aperture of the sensor, or slightly defocusing, etc. One way or another, people have to deal with moire and aliasing if it occurs. If you have to constantly perform a very slight defocus, that makes using an autofocusing system very tedious. Concurrently, having to stop down more than you really want to in order to force diffraction blurring to soften the image is also less than idea.

You say you have used a lot of cameras over a lot of years. I'd be willing to bet many of them were film cameras, in which case moire was never a problem thanks to the random distribution of grains. When it comes to digital cameras, until recently, lenses, while good, were never as good and sharp as they are today (at least, in the DSLR world...for MF, most lenses have always been rather exceptional.) The softer lenses of the past helped to deal with the problem of missing or weak AA filters. Today, we have a convergence of several things that can only lead to significant problems with moire and aliasing: Radical improvements in lens quality, pushing their maximum resolving power to new limits; sensor resolution increasing at a slower pace than lens resolution; removal of AA filters. This is kind of a perfect storm...some manufacturers are apparently doing everything in their power to make moire a very serious problem for a lot of DSLR photographers, which will ultimately put them in the same boat as Leica and MFD owners: Having to defocus or stop down to force blurring and use the lens as an artificial AA filter.

Did you discuss the bold area I highlighted above (about the ratio between lens to sensor) a bit more in detail someplace? This is likely the feature I'm looking for to be optimal, and likey what the D800E, and A7R have factored in. It is my next criteria for my future camera/sensor purchase.

If you mean the fact that output resolution is based on the convolution of lens+sensor, I've discussed it so many times all over this forum, it shouldn't be hard to find a topic with all the details. The detail in an image (raw file) is the result of a complex convolution of real-world details. In mathematical terms, assuming gaussian-like blurring behavior (which is reasonable), output resolution is roughly equal to the root mean square (RMS) of the input resolutions. Well, to be more specific, the size of the blur kernel that represents the output image is approximated by the RMS of the blur circles of the lens and sensor.

So, if your lens blurs by 3µm and your sensor has 5µm pixels (the lens resolves more detail than the sensor), then the output blur is SQRT(3µm^2+5µm^2), or 5.83µm. Notice that the output resolution is lower than BOTH your sensor and lens. If you improve your lens resolution as far as possible, let's say 0.7µm blur circle (the wavelenght of red light), your output blur is 5.04µm. Your maximum resolution is limited by the sensor...no matter how good your lens is, you can never resolve more detail than the sensor is capable of. This goes the other way as well. Let's say your lens blur is 3µm and your sensor has 2µm pixels. Your output blur is 3.6µm. If you reduce your sensor pixels to 800nm (0.8µm), your output blur is 3.1µm. You can never get any better resolution than your worst performing component.

That's why I always say the whole notion of sensors or lenses "outresolving" the other is more myth than fact. In one sense, I understand why people think about it that way. In reality, the two work together to resolve your image...without both, you have no image, so there really isn't one outresolving the other. The real fact of the matter is your output resolution is never as good as the potentials of your lens or sensor, and your output resolution can never be higher than the least capable of the two. Further, lenses have non-linear performance...as you stop the aperture down, their performance drops. It's tough to say a lens outresolves a sensor in general...at what aperture does it "outresolve"? And by how much? Enough to matter? Or is the lens just outresolving by a tiny bit? When you stop down to f/8, is the sensor outresolving? These questions really don't matter...the thing that really matters is how the output image looks, and regardless of which thing you change, more resolution is pretty much always a good thing, sometimes a neutral thing, but never a bad thing.

The D800/E sensor is definitely higher resolution than the 5D III, for example...however Canon lenses outperform most Nikon lenses, so in most cases, the better Canon lenses paired with the lower resolution 5D III outperform, by a small margin, the D800/E. Even DXO's own lens data shows that. The D800 sensor will certainly make the absolute most out of Nikon lenses, but until Nikon improves their lens designs, the D800 does not actually perform better, in the real world, than the 5D III. Ironically, it is thanks to that very fact that moire with the D800E is not a bigger problem than it is...the lenses soften detail enough that moire tends to occur minimally. The day Nikon lenses perform as well as Canon lenses, however, keep your eyes and ears peeled: The wrath of the moire-hating D800E user will be heard around the world. ;P

 

[Phil Indeblanc]

Otherwise more distant future.....I think it maybe a mix of how the Lytro can focus

Doesn't the lytro "focus" mathematically? If you're happy with light field algorithms, why not deconvolving AA filtered images?

"Here is a sample conversation, or reply to my original post... [/snip]"

But your original post comes across as very self-centered and demanding.

Very logical and true reply. I apologize for those Canon fans that took it personally, and I too am a Canon fan of course..And thank goodness you read into my sarcastic approach to all this, as I am a realist outside of the tech field, and accept whatever hits the market floor for us mortals to enjoy.

 

[Phil Indeblanc]

you'll only obtain this across a specific aperture range, and (of course) when using impeccable technique."

In studio and product, that's all there is. A margin of about half F for max sharpness. In the HR optics on MF the F blades don't refract up to about between 11 and 16 ). But using a 35 would increase dof. I haven't really tested the limits of Canon glass. I usually used the Leica with the 35 bodies for studio work more than 5 years ago.

 

[jeffa4444]

Jrista
Your points about removing AA filters are spot on, moire and alaising will always be present and higher performance lenses will definately show the problem faster I would not buy any camera without one. However AA filters themselves are very difficult to produce especially in large volume and keep consistent. The AA filter is part of the optical system and taken into consideration when designing lenses and well designed and manufactured AA filters now have a minimal impact on resolution. Lateral chromatic abberations are still visable in most Canon lenses including L lenses and this will have an affect on apparent sharpness particularly further out to the edges of the frame and Canon will need to address this going forwards especially as Zeiss roll out more of their Otus range I see this as more of an issue rather than the affects of the AA.

 

[Phil Indeblanc]

JRISTA, I will need to get back to work before I can take all the info, as well as others posts. Yes, there are a couple that I didn't fully read from the 2nd page. I'll will get to them, as I appreciate them, regardless.

 

[jrista]

Jrista
Your points about removing AA filters are spot on, moire and alaising will always be present and higher performance lenses will definately show the problem faster I would not buy any camera without one. However AA filters themselves are very difficult to produce especially in large volume and keep consistent. The AA filter is part of the optical system and taken into consideration when designing lenses and well designed and manufactured AA filters now have a minimal impact on resolution. Lateral chromatic abberations are still visable in most Canon lenses including L lenses and this will have an affect on apparent sharpness particularly further out to the edges of the frame and Canon will need to address this going forwards especially as Zeiss roll out more of their Otus range I see this as more of an issue rather than the affects of the AA.

Canon's shorter focal lengths, under ~200mm (including the 70-200) do need improvement in the corners. The worst of Canon's lineup are their wider angle lenses. The 24-70 II improve things, however it's corner performance (as even indicated by Canon's MTF charts) is still quite poor. I don't know why Canon has such a hard time with wide angle corners, but it's their Achilles heel, for sure. I think that is one of the main reasons Sigma has been making such major strides...they found the weak spot in the biggest photography manufacturer in the world, and have been exploiting it as much as they possibly can.

These days, I'm less concerned about a manufacturer's ability to produce AA filters of consistent quality, and more concerned about their ability to produce them strong enough. Sadly, I think the (uneducated) demands of the consumer for no AA filters are winning out in this arena, despite how non-beneficial that is for IQ. People want "sharp out of camera", and don't seem to understand the consequences of the tradoff that is REQUIRED to make that happen. An appropriately strong AA filter that minimizes moire to the point where only the strongest interference patterns make it show up is what we really need. I'd rather have slightly soft out of camera without moire, as I can easily sharpen in post, than have razor sharp out of camera with a bunch of aliasing and moire.

The AHDD, or Adaptive Homogeneity-directed Demosaicing algorithm used in Adobe ACR/LR is highly optimized. It is capable of interpolating in such a way as to utilize the raw luminance information in EVERY pixel, and only really suffers the resolution loss when interpolating the color channels. That means were getting the vast majority of the resolution our sensors are capable of with modern RAW editors like Lighroom, and only really suffering some loss in resolution and color fidelity in the color channels. That doesn't much matter, though, as we aren't as sensitive to softness in color as we are to softness in luminance detail.

Not to mention the fact that most cameras offer far more resolution than a growing majority of photographers need, what with publication on the web at relatively small sizes (compared to those required for print) being the primary means of sharing photography...from your average instagram and facebook uploads to your avid amateurs to your professionals.

 

[Don Haines]

On the subject of cameras and technology.... I took this shot with an SX-50 on the way home today... I put it into "green box mode" and the image was taken handheld at full zoom (1200mm equivalent) plus another 4X digital zoom..... in other words, handheld at the equivalent of 4800mm! Yes, I know it's a cheap P/S and I know the picture is not national geographic quality... but for a toy camera at 4800mm handheld I think that this is amazing. A lot of research went into making this work as well as it did...

P.S. I couldn't tell it was a wood duck until I saw it through the camera EVF....

 

[Phil Indeblanc]

I don't know why Canon has such a hard time with wide angle corners, but it's their Achilles heel, for sure. I think that is one of the main reasons Sigma has been making such major strides...they found the weak spot in the biggest photography manufacturer in the world, and have been exploiting it as much as they possibly can.

I think Tamorn has the upper hand on wides, 17-35, and 24-70

I think the (uneducated) demands of the consumer for no AA filters are winning out in this arena, despite how non-beneficial that is for IQ. People want "sharp out of camera", and don't seem to understand the consequences of the tradoff that is REQUIRED to make that happen.

I don't think there are general consumers asking for any mods of an AA filter to be removed. I think maybe some pixel peepers,and perhaps those that actually need it in certain markets that see the gain from it.

That means were getting the vast majority of the resolution our sensors are capable of with modern RAW editors like Lighroom, and only really suffering some loss in resolution and color fidelity in the color channels. That doesn't much matter, though, as we aren't as sensitive to softness in color as we are to softness in luminance detail.

You can get cleaner results in sharpening with less muddiness in Capture One with some sensors. LR maybe able to, but takes more work. I have tried, but I'm not at the point of formulating captures with LR settings from memory. I take each image individually, and C1 gets it done cleaner off the bat(*In the files I have worked on).

Not to mention the fact that most cameras offer far more resolution than a growing majority of photographers need, what with publication on the web at relatively small sizes (compared to those required for print) being the primary means of sharing photography...from your average instagram and facebook uploads to your avid amateurs to your professionals.

This too is a very general scope of things.

So far your points are correct for the general market you talk about. It doesn't have much to do with the reason I started the thread, but it does make things clear for me. Besides the overwhelming information you have a great understanding of....
There is the correct tool for every job. For anything specialized don't expect all niche markets to be covered by one. I guess I can buy a $1700 A7R(plus adapter) and see if it is worth replacing a MF back(at least).


I just don't know what areas the A7R will start to bleed into as a replacement to have less gear. I can easily see it be a PointShoot replacement for the Canon-M that I like, and maybe take over the 5D M2? I doubt it. With the adapter, and if lenses get freed up, who knows? My mind is working while I wait for the 5-7 days it takes for my shutter to be replaced on the 5Dm2 for about $350. I'm not complaining, in fact I will take Canon service over Nikon any day, and besides the Sony cell phone, I haven't gotten anything Sony for a long time.


Moreorless,

Its interesting as now with MF going into 50-60+ mpixels, and the sites being as small as 5-6 microns, AND them going towards CMOS, If compared using the same lenses...What else is there?

I do agree the difference is not a major significance for most types of shooting. It does effect stones facets, watches, etc.

 

[Phil Indeblanc]

P.S. I couldn't tell it was a wood duck until I saw it through the camera EVF....

4800MM! That is pretty cool. Reminds me when I had the Oly C2500L, then the UltraZooms came in, and I had the UZ730, I still have a 740 in a drawer. But 30+optical zoom was rather great. Sure the mpixels were around 4!

 

[jeffa4444]

Its funny here everyone wants more resolution & more DR. The guys shooting for our movie screens or TV are trying to break it down to get a more organic look. a look that they feel film used to give them however VFX want it as clean as they can get it there is no squaring the circle.

 

[moreorless]

Whilst removing an AA filter(depending obviously on how strong that filter was) can aid sharpness a little(with certain trade offs) I think the lack of one became a bit of an easy way for people to explain the superior performance of MF digital.

Just pointing to a lack of an AA filter as some kind of magic bullet(especially when all mainstream 35mm and smaller sensors had one) was a lot easier than going into the indepth reasons of why a larger format will generally offer better performance even if the number of MP's on the sensor are similar, most obviously that lenses won't be pushed anywhere near as hard.

 

[Phil Indeblanc]

Otherwise more distant future.....I think it maybe a mix of how the Lytro can focus

Doesn't the lytro "focus" mathematically? If you're happy with light field algorithms, why not deconvolving AA filtered images?

The question was a regarding future cameras, and that would be for general shooting purposes, not product.

I remember reading some time back that digital cameras can be made with a single pixel. a very large single pixel being the light gatherer. I wonder what happen to this approach?

 

[gargamel]

[...]

I think we are on the edge of a shift in digital cameras.

We need to step back and ask "why mirrors". In the days of film, you needed the mirror and optical viewfinder to know what you were looking at and we needed focusing screens to know if we were in focus.... Then came digital sensors and we treated them like film... because that is what we were used to.

A digital sensor is NOT film. It has different strengths and different weakness.... and the mirror is no longer the only way to see through the lens. A decent mirrorless camera (and there are several on them out there) will be designed to the strengths of digital technology. They already do many things better than DSLRs, but a great mirrorless camera will have to do everything better. Right now, the two big stumbling blocks are focusing and viewfinders.

[...]

Exactly, plus, for some photographers, battery life can be an issue with mirrorless cameras. Especially travel and nature photographers, who cannot easily charge batteries when they are "on tour", should consider this. Mirrorless cameras suck power for EVERYTHING. With a DSLR you can look through the OVF and check composition even when your camera is switched off. With an EVF the cameray needs to be on, all the time, and the mini tv set, called EVF, needs power.

Since my old 5D1 died I am looking around, what my next camera might be, and as I don't own a lot of expensive glass, I am open for other systems, too. While many Canon shooters go for Sony, it seems, I got even more interested in Fuji's X system. However, I found that I can get only about 300 photos with an X-M1, when I was able to take about 800+ photos with a 6D (had both for a rent for one day).

Regarding (not only) AA filters, Fuji is lighting the way. As has been said here, Moire (sorry, no diacritic characters on the keyboard I am using right now) was no problem with film, due to its amorph distribution of grain. No geometric pattern, so no interference with other geometric patterns. Simple as that. And it's exactly this, what Fuji is trying to achieve with its X-Trans sensors.
So, while I agree with most of the arguments that AA filters are useful, these arguments apply more (if not only), if the pixels on the sensor are arranged in a geometric pattern, that could interfere with a geometric pattern in the image, which is less likely with an X-Trans sensor than with a "classic" sensor, where the pixels are arranged in the Bayer pattern.

BTW, while better lenses may increase Moire, the way out, apart from other, more chaotic arrangements of senser pixels, is to increase the megapixel count. Only when they reached 24+ MP, Nikon and Sony dared to make cameras without AA filters. Why? Because, the finer the structures and the smaller the areas that can possibly cause Moire are, the less of a problem they will be in the resulting images.
Well, I am not a techie, and maybe some expert can give more detail, and maybe someone with more didactic experience can explain this a lot better than me, but I hope, I was able to give correct facts. (Otherwise, please correct me!)

Finally, for those of you who want to have AA filter or not as an option, the Pentax K-3 should be a very interesting piece of technology....

gargamel

 

[jrista]

Speaking of battery life in DSLRs. When I attach my battery grip to my 7D and use fully charged batteries, I can get well over 2500 shots (which is usually about the limit for what I get on a shoot these days, not because of no battery power, but because by then the sun has set, or the animals/birds moved on, or something.) I rarely use my batteries below the 50% mark when using the grip, so it's possible I could get quite a bit more than 2500 out of one pair of fully charged batteries.

 

[scottkinfw]

Make it stop already!

 

[RLPhoto]

MF is used not solely for its IQ. While it's IQ will always be better to that tiniest nth degree, it is better but we will soon see that gap widen with new CMOS sensors now being made in MF.

MF has the very best optics and is less demanding on the glass too. I remember my beat up scratched 80mm zeiss 2.8 being sharper than any 35mm lens because of the lower magnification.

MF backs can be attached to technical cameras with movements you can't get with a sensor buried in a body.

MF also has high true sync speeds with leaf shutters. A lustful trek for any strobist.

MF gear may give a client an impression of professionalism. Not always true but it not a camera you mom could have.

Is it worth 30K for a top end system? It depends but I am slowly purchasing my body, lens, And an older back to get those sync speeds.

What I'd like to see is a series of lenses from canon with a leaf shutter that works with any canon camera in live view. I can get the super sync speed without having to buy MF gear.

 

[jeffa4444]

RLPhoto

All those points are valid and true. The other point is depth of field which is shallower than on 35mm which means focus is more critical.

 

[Phil Indeblanc]

I opt for Medium Format because what I shoot demands it. My customers demand it. Life would be MUCH easier if I didn't have to bust out a 4x5/MF dB for a shoot. Work would be much smoother.

Are you using a digital back on your 4X5? If so, which one and do you like it?

Sorry for the long time gap. I haven't been getting email notifications.

I currently use the Phase One P25 often. I also use a H25backup, and have used a Hasselblad CF39, Phase One P45, and a more current IQ180

I'm sorry and understand I sound stubborn about this. It is only because I see this segment getting zero attention from Canon, who I have over $20K into.

On the flip side I DO see SONY, and Nikon address these areas. I also see Pentax 645z in action, and ALL MF dB's in the same common denominator. Even Sigma(As there is no need)...So Canon is alone. It is now so commonly spread, that having no AA as become a marketing tool. When working with difficult subjects like reflective product, and you have ZERO absorption surface. Everything is reflective high polish you can see how big a difference AA vs no AA is.

I think mirrorless is surely the direction, and yes, an in-between mirror space gap can make it much easier. I forget the brand years back, I briefly worked with a camera years back that would take filters on the lens mount.