A tongue-in-cheek missive to those who want a high MP 5D3 :)

[neuroanatomist]

I guess I have a higher opinion of where science can go and a lower opinion of where it is.  Whether you're talking about the discovery of quasicrystals for which the discoverer was practially barred from scientific circles because it went against "the laws of physics" (but has since been awared the nobel prize) or the more assinine "truth" that the 4-minute mile is (was) physically impossible, science grows and evolves and changes as we continue to push the boundaries.

For example, what about metamaterials for lenses?

I have very high opinions of where science and imaging technology will go.  Can the diffraction limit be overcome?  Yes.  In the 19th century, Ernst Abbe (friend of Carl Zeiss and one of the pioneers of microscopy and optics) postulated that no amount of glass refinement or lens design could escape the limit of resolution for visible light, which is about 0.5µm.  Today, we have superresolution microscopy that breaks that limit (one approach to which actually does use lenses with metamaterials). 

But, I think it's likely that by the time we progress to breaking the diffraction limit, the dSLR will have gone the way of the dodo.  Picture Canon's 'wonder camera' combined with a plenoptic camera, then fit that into a cell phone - that's just one step along the road...

[K-amps]

About 6 years ago, the air conditioning on the roof of the building I was working in had a problem with a bearing. It produced a very high pitch noise that I could hear and gave me a headache but nobody else in the office could (I was the youngest...)

Dare I ask what your thoughts are on 96kHz audio? I've got a few SACDs and the appropriate equipment to deliver 192/96 to the amp and out of the speakers...

Exactly! I suspect that if a 5DX were to have a foveon style sensor with colours arranaged vertically rather than in a horizontal 2x2, we'd find companies doing the Sigma thing and calling a camera with 15MP a 45MP camera.

Dilbert: We live in a world where the youth of today embrace the MP3... so 96kHz was pretty much dead on arrival. The beauty of SACD was not only the 96/192kHz sampling, but the ditching of discrete paralell bits and going for a stream of data. It was just a different way of processing data, not merely an upsampled engine that CD used. I too have a few SACD's lying around, they are not bad,  the pricing killed it as well.

Going back to the topic: Looking at apple/ google, buying off companies seems to be the best way to get Patents... however I suspect the fact that the Fovean did not dazzle people like they expected it to may have something to do with it...

[LetTheRightLensIn]

1. People use TCs with lenses on the 7D (over 40! MP FF equivalent) and still pull in more than detail than without using a TC.

2. Even a mediocre lens scores more resolution, near the borders, on an 18MP 7D than on a 10MP 40D.

3. Some of the SONY sensors that are far denser than the 5D2 sensor actually have much higher dynamic range per sensor area than the 5D2 does.

4. It's not hard to notice the extra detail you get on distant wildlife between using 12MP FF equivalent and 45MP FF equivalent even if you insist on poo-pooing more detail for landscapes for whatever unknown reason.

4b. Not everyone can afford an 800mm lens but if you want to buy a copy for all of us....

4c. Even at 800mm it's easy to have the subject not fill the frame

5. The crazy density 7D has a little bit BETTER SNR at middle gray per sensor area than the much lower density 5D2.
 
so just sayin'

(not that I want them to go too crazy with it and be able to get at least 6fps and a crop mode would be nice so as to not fill up HD with half wasted space for distant wildlife)

[LetTheRightLensIn]

I can't tell who's right and who's wrong about max. megapixels / resolving power / resolution, because I am not a mathematician, nor a scientist.

What I CAN tell you is my personal experience.

I have owned a 350D (Digital Rebel XT) / 30D / 40D / 1D-mkII / 1D-mkIII / 5D-mkII / 7D / 1D-mkIV

My lenses are:

24-70 f/2.8 L
24-105 f/4 L - IS
16-35 f/2.8 L II
70-200 f/2.8 L
100-400 f/4.5-5.6 L - IS
200mm f/2.8 L
100mm f/2.8 L - IS Macro
35mm f/1.4 L
50mm f/1.4 (Sigma)

I shoot approximately 175.000 pictures per year!

My experience is that the 7D with it's 18 megapixels demands the MOST of all lenses. I get better results with the same lenses on my 1D-mkIV

It is my opinion that more megapixels on a sensor makes it demand more of my lenses.

To be totally honest, I have made my most beautiful pictures with the 350D (Rebel XT) and the 40D. Both sensors have excellent design characteristics, and get the very most out of all of my lenses. They just won't perform equally as good on any of the other bodies.

I feel that Canon would really need to re-design ALL of the above mentioned lenses in order to perform up to professional standards on the 7D / 5D-mkII / 1D-mkIV and the new 1D-X

That's just my two cents worth.

Sure a 7D demands more than a 10D, of lenses and 100% micro-contrast might be a touch lower. BUT scale down the 7D to 10D dimensions and tell me which shows better 100% view micro-contrat. Scale up the 10D to 7D dimensions and tell me which shows more detail.

Only normalized comparisons make sense when trying to say one thing is better than another.

[LetTheRightLensIn]

Not to hate, but from all I read, it seems people want ISO capabilities far beyond their wildest dream. How much high ISO ability does a person really need? How perfect do one's pixels really need to be? What ought a person to expect out of one's camera? Honestly, a tripod, a broader aperture, and a little patience can make up for a whole lot of ISO...

And as far as DR is concerned, what do one really actually need 37 stops of DR for? What about ND grads? What about HDR? Why about exposure stacking and masking? Or fill flash and reflectors? Honestly, my 7D RAW files have piles of DR, when I expose to the right.

Print film ISO 50 has wild DR and low noise... could this work as a temporary stop-gap?

something bright and dark all randomly all throughout the frame and sometimes clouds are zipping through really fast or trees blowing all over or someone is walking so ND grads or exposure stacking don't solve everything evne when you do have a tripod with you. Sometimes you can sort of piece together scenes that had bright and dark all over and motion, but not always, and it can take a ton of time to rescue even modestly tricky scenes like that to any reasonable degree.

[LetTheRightLensIn]

Thats like the argument that nothing can travel faster than the speed of light. For years science has claimed that as an unmovable barrier but now there are experimental results that show that is simply not true.

Just because present sensor and lens technologies have certain theoretical limits I believe it is a mistake to think that this can never be "side stepped". I know that lens design is just a case of pouring money at the problem and sensor design may well develop in future in directions we can only dream about.

experiments which most, including those who carried them out, suspect will eventually be found to have been flawed (although most hope they won't, since that would be interesting)

[LetTheRightLensIn]

I just see these ISO people wanting a 1Dx for the 5d3 price.

I just see these AF people as wanting a 1D X for the 5DIII price.   

Seriously, while we'd all like a 5DIII pro-level AF ('cuz Nikon does it, right?), higher ISO, better DR, fully-sealed build, and we'd all like it for the price of a 60D, it's just not going to happen.  It's not even going to happen for the price of a 5DII.  With the 5DII, Canon used the same sensor as the 1DsIII flagship - but does anyone really think the 5DII was a 'baby 1Ds'?  No - an the 5DIII will not be a 'baby 1D X.'  Either the 5DIII will be 'handicapped' in some (many) ways compared to the 1-series, or it will cost much more.

once upon a time canon did see fit to make such bodies as the EOS 3, don't forget....

[contrastny]

Interesting, but all of this is getting confusing... I might wait for the 5D mkIII to come out and buy the 5D II, hopefully at a discount and pair it up with a Zeiss 18mm, since I mainly shoot landscapes.

[Lee Jay]

Canon's newest L-series lenses (mark II's and newer releases from the last few years) can resolve up to around 45mp worth of resolution in a FF image circle.

How does 737MP strike you (18 * 4^2 * 1.6^2) with the old 400/2.8?

http://forums.dpreview.com/forums/read.asp?forum=1029&message=37493247

[tjshot]

But I completely and totally reject the argument that its possible to ever have a sensor with too many megapixels. Yes, extreme MP counts create problems for the scientists and designers to overcome and I'm confident that they will overcome them.

When those 'problems' involve basic principles of physics, like diffraction, overcoming them may prove difficult...

Yes, diffraction imposes an upper limit on pixel density. No, we are not there yet, and a 35-45 MP FF sensor would yield benefits in fine detail and maximum print size.

That's actually the case.
On a previous post I proposed a simulation of performance increase for different sensor densities, assessing the effective performance boost (both resolution and contrast) stepping up from 21 to 36 an then 50 mpxls.
http://www.canonrumors.com/forum/index.php/topic,1902.0.html
There is also discussed how the concept of a sensor "outresolving" a lense is definitely prone to misconception.
In a few days I'll be posting a similar scenario based on real glossy prints (Epson R1900) from artificially generated targets simulating the same sensors; it could be useful to assess how camera performance is related to effective increase in print quality.
Test results confirm that with proper sharpening applied, despite diffraction, a 50 Mpxls sensor will perform at ISO A2 print size exactly the same as an actual 21 Mpxls one at ISO A3 size and a 36 Mpxls sensor at 16x enlargement (i.e. betwen A3+ and A2).

[Edwin Herdman]

But, I think it's likely that by the time we progress to breaking the diffraction limit, the dSLR will have gone the way of the dodo.  Picture Canon's 'wonder camera' combined with a plenoptic camera, then fit that into a cell phone - that's just one step along the road...

I don't think light field cameras have anything to offer over traditional sensor tech, for almost all applications.  They certainly throw away much of the advantages of high resolution (and, correspondingly, would only aggravate the high sensitivity problems of sensors as small as those found in cell phone cameras, because to get anything like a reasonable final image size you need many, many more pixels to waste in capturing the light field).  Plenoptic technology is silent in the mirrored / mirrorless camera debate, as well (the concept can work in either physical system just as well or as badly).

[jrista]

I have very high opinions of where science and imaging technology will go.  Can the diffraction limit be overcome?  Yes.  In the 19th century, Ernst Abbe (friend of Carl Zeiss and one of the pioneers of microscopy and optics) postulated that no amount of glass refinement or lens design could escape the limit of resolution for visible light, which is about 0.5µm.  Today, we have superresolution microscopy that breaks that limit (one approach to which actually does use lenses with metamaterials). 

But, I think it's likely that by the time we progress to breaking the diffraction limit, the dSLR will have gone the way of the dodo.  Picture Canon's 'wonder camera' combined with a plenoptic camera, then fit that into a cell phone - that's just one step along the road...

Just to make sure I'm clear...when you say "diffraction limit", are you actually talking about the "diffraction cutoff frequency"? I ask because of this statement:

"...postulated that no amount of glass refinement or lens design could escape the limit of resolution for visible light..."

The diffraction limit is a limitation of an entire imaging system ultimately affected by the resolution of the imaging medium and the aperture of the lens. Given your reference to the wavelength of green light, I think you actually mean to refer to the diffraction cutoff frequency, which would be the point at which the resolution of a lens approaches the wavelengths of the light its imaging. As a simple matter of physics, that would be the hard limit, and literally unbreakable in any normal optical system.

Superresolution, even optical solutions, rely either on oscillations and multiple images, or information fabrication, and while they can break the diffraction limit of any given imaging system, I don't know that they can actually image detail beyond the wavelength of light (unless they fabricate it, which leads to a debate unto itself.)

[AprilForever]

Canon's newest L-series lenses (mark II's and newer releases from the last few years) can resolve up to around 45mp worth of resolution in a FF image circle.

How does 737MP strike you (18 * 4^2 * 1.6^2) with the old 400/2.8?

http://forums.dpreview.com/forums/read.asp?forum=1029&message=37493247

Impressive!!! I eagerly await the 7D mkII and the more megapixels hopefully to come...

[neuroanatomist]

Superresolution, even optical solutions, rely either on oscillations and multiple images, or information fabrication, and while they can break the diffraction limit of any given imaging system, I don't know that they can actually image detail beyond the wavelength of light (unless they fabricate it, which leads to a debate unto itself.)

I routinely do brightfield optical microscopy at the Abbe limit (1.4 NA oil objectives and 1.4 NA oil condensers).  Near-field scanning can image detail smaller than the wavelength of the illumination (for fluorescent samples), because the illumination is effectively compressed into a source that is smaller than the wavelength of visible light. 

But you're right that most practical implementations of superresolution (e.g. STED) rely on software-intensive processes and image combination.  Still, a relevant analogy is the use of HDR to increase the dynamic range of current sensors. 

[K-amps]

... Still, a relevant analogy is the use of HDR to increase the dynamic range of current sensors.

Not picking on you John, but Isn't HDR a misnomer? I mean what we are really doing is "reducing" the dynamic range to display on our monitors... in reality it is Dynamic range compression, is it not?