Just for fun: What was technologically impossible years ago, today it is feasible.. One simple example is the tinny cameras used in Pillcam[*], an ingestible capsule device equipped with a miniature video camera to visualize the small intestine. Also the silicon technology used to build processors and other digital products has evolved a lot during the last decade, allowing faster processors with less power that were impossible according to the engineers from few decades ago.

[*] A month ago I started to work in Given Imaging in the development of new products, that hope that will also make a good contribution to our world :)

(**Mod edit: deleted off-topic comment about moderation**)

I am glad Mongo brought it up and I am 100% with him on this matter. I will repeat once more that diffraction calculations/tables are made taking into consideration the PERFECT ( diffraction limited )lens . The only way from there is DOWN and you can not get better results with the lenses you have . In other words , diffraction not starting until f22 in your 2 lenses can not be justified by science & optic theory . In a 13 Mp APS-C DSLR , it starts at after f8 ( or at f11 in 10Mp ) ,it is not lens dependent and it really is not open to discussion. The figures are from generally accepted diffraction tables which I can provide links to.

You may just be not noticing it if you don't need that much resolution in your work.

I'm sure you are tired and I'm not sure what all the DOF calculations are for. Most of us know this or can download the App for our phones.

"Most of the cell phones and many of the small sensor compacts have very "fast" lenses.."
Actually most of them are F4 to F8. The iPhone was f3 which was very rare. Sony did have some 2.8 and the various android phones are going faster f-stops for better low light. None of them are anywhere near the F/1 you suggest.

Cellphones still are way out of line what Heisenberg limits suggest. It looks like you are good at plugging in variables on equations, so I'm sure you can figure that out as well. Use the iPhone 4 specs- that will give you the answer.

Diffraction: Take a 60mm or a 105mm diffraction doesn't start till about f/22. That is a prime (no pun intended) example of engineering success over a physics theory.

Mongo - you are still ignoring the reality, camera and lens makers have been out engineering the basic application of the famed Heisenberg limits for years that people are trying to prove the limits of lenses, sensors, peanut and butter sandwiches are just about here.

I don't understand everything you guys are talking about, but it is very interesting.

As I understand Momgo's position. A 24 mp DX sensor is at (or over) the limit of what will be possible in DX format. The end of the line, so to speak. And software cannot fix it because the data won't be there. As Mongo says "But if your points of light are bigger than two of your sensors it doesn't matter how well the software fixes the data because the data is not there." Isn't that were some "binning" could come into play? If the software treats 4 pixels as one "super pixel" under certain conditions doesn't that allow the new "super pixel" to be larger than the point of light?

TaoTeJared,

The size of the airy disks is directly related to the f/stop of the lens, you can get dramatically higher resolution rates by moving closer to f/1.

Most of the cell phones and many of the small sensor compacts have very "fast" lenses for this very reason (well and to get around the S/N issues with small sensors).

The limit is lines per mm, not a MP count so if you get near f/1 your MP count can go WAY up. Many also have horrid lenses that have so many aberrations that diffraction is not an issue. I have macro photos that are at f/35, the contrast sucks and they probably have the real MP count of around 2MP but they are still great pictures. The rub is that to get better S/N you go with a larger sensor with larger pixels but that also decreases your DOF (mostly due to the requirement of a higher focal length), so you have to stop down in order to get the same DOF.

f-stop is the focal length of a lens divided by the diameter of the entrance pupil or:

f-stop = focal-length / diameter

so where a f/2.8 200mm lens will have a max 72mm diameter lens the iPhone 4 has a 3.85 mm f/2.8 lens with a 1.4mm diameter lens.

The G12 with a 10MP 1.7" sensor is diffraction limited at around f/3.5

A 12MP 1.6 crop factor APC camera is diffraction limited at f/10

A 24MP full frame 35mm is diffraction limited at about f/11

A 24MP 1.6X crop APC sensor sized camera is diffraction limited at f/7

This is assuming a "perfect lens". It is not cliff the sensor falls of either, it is the point where mostly contrast starts to be reduced and the amount of recorded information goes down.

This is just based on the physics, this does not mean that a new camera will not have improvements that will make the image nicer, that can happen just with software or improving the S/N ratio. But if your points of light are bigger than two of your sensors it doesn't matter how well the software fixes the data because the data is not there.

This is why I HOPE that Nikon stops at 18MP for DX, an 18MP DX sensor will start to be diffraction limited at around f/8, this also happens to be the sweet spot for a lot of lenses.

Interesting bit, a 80MP Phase one Back with a 53.7mm x 40.4mm sensor becomes diffraction limited at around f/11 but it's normal lens is 80mm.

At f/11 with a subject at 15 feet or 5 meters it will have a DOF of around 12 ft or 4.6 meters.

A D300s at f/11 with a 35mm lens at f/11 will have a DOF of around 85 feet or 25 meters.

That 80MP sensor would need to stop down to about f/28 to have the same DOF which is well past it's diffraction limit.

Of course if you are taking fashion shot that 80MP will work great because you do not need a lot of DOF. And the landscape guys will do well if they stay below F11 and deal with not having things withing 50 feet in focus but everything past that is (not like they would use a normal 80mm lens)

Really the big issue is that MP really don't matter as much as most people thing and even with my D300s some of the pictures I take at high f-stop values do not have anywhere close to 12MP of data.

But I am really bored with this talk, so I would suggest going out and looking at the MTF of lenses on the D3X when it goes above f/11, you can see it drop like crazy on most lenses.

Best of luck and I hope you are out there making some great images!!!

Mango - that is a good explanation and I agree MIT has many good documents (on damn near anything) to read and there are many others as well. They are not easy to understand but as long as you are brighter than just a Smart A.. poster that has nothing to add than you should be able to learn it.

The crux of declaring Heisenberg limit argument that no one has been able to explain - If there is a limit, why can digi cams and cell phones which have a tremendously smaller light wells and, not the top notch lenses, can create great images.

The fact that Cannon came out with a (I believe read it right) 128mp sensor, Sony has 36mp-46mp sensors in development, and digi cams are topping 16mp I have a hard time applying Heisenberg calculations to sensors yet. I don't question the physics behind it or am saying it is not true, just that the application to sensors is off somewhere. If it was true, we would see companies scrambling to create something different across the board which is not happening.

BTW - good topic donaldejose.

I agree with TaoTeJared

{ "was a technology limit"
This is my point. Everyone calls it a technology limit looking in the rear view mirror...}

There are amazing tech already available but not commercialised yet.. eg black silicon and nikons own patent for 3 sensors per pixel area( using micro prisms to deflect the required colours onto the sensors.) which would provide an immediate jump(at least X 2) of resolution.

re: the combining of pixels.. seems like a logical extension and use for a 24mp camera. there are already some nice algorithms that work at a low level to dynamically combine groups of 4 (blue, green, red, green {aka bayer]) 6, 13 etc.. would make sense that nikon could be incorporating these into their software / hardware.

Super Shooter said:
NERD FIGHT! Stand back everyone. Give 'em room! My money's on Mongo because he actually knows what he's talk'n 'bout. Go get 'em Mongo! :)

I wish it was wrong, and there will be improvements in some places and I am sure
Nikon software will improve but yes, it is Nerd stuff :)

Things get really odd, look up "double slit" which produces a more complex pattern then the single slit diffraction we have in out cameras.

Basically things work like waves on the ocean, where waves will cancel each other out, but when you get small, lets say with electrons, if you shoot electrons through the slits one at a time they exhibit the same patterns, thus they are canceling themselves out and it is as if they, individually, go through both slits at once and each slit or no slits all at the same time.

MIT has a project called open courseware, they have most of their classes on there. If you want to become a nerd too the physics courses will explain these observations.

Note that being diffraction limited is not the end or the world, almost all telescopes work at the diffraction limit, but it does limit the amount of data you can collect per mm for a given fstop and thus DOF.

I have no information but I would guess that Nikon may pull the AA filter off and perform that function in software to get better sharpness than the math should allow.

In reference to BSI, our current sensors have the "wires" in front of the diode that detects the photons, when Thom is talking about a back side illuminated he is talking about a sensor that has the diode in front of the wires, this should help with ISO/Noise because the detection area will be larger but it will not fix the diffraction issue.

If you look at the D7000 jpegs it is obvious that the newer software is better, I have mixed feelings about the raw files, but it is not a "pro" camera.

FYI I do expect the D400 to be better than the D300(s) just not due to MP.

Talking about duality, I like the tech but I know it has little to do with the quality of an image.

NERD FIGHT! Stand back everyone. Give 'em room! My money's on Mongo because he actually knows what he's talk'n 'bout. Go get 'em Mongo! :)

TaoTeJared said:
"was a technology limit"
This is my point. Everyone calls it a technology limit looking in the rear view mirror but at the time people were declaring Heisenberg limits on 12mp along with physics diagrams how "it could never happen" and "images would be awful". When Nikon starts to throw red flags up, then I will think there is a limit. Most of this just comes down to how you count MP anyway.

"Second, Binning is a very common jargon term"
Ehh - Sorry, no one speaks of binning the Cheerios into containers so they stay fresh - not even developers. It is Jargon though - just for very few people. Using Jargon outside of the group who uses it muddles and stifles any good conversations on the topic. If your non-engineer friends would look at you strange if you used it, it is not common.

BSI == back side illuminated
Thanks.
---------------------
On the actual topic, if you can calculate the "Heisenberg limit" than you can calculate your way out of it (or account for it) with software.

Honestly I think we will be seeing more hexagon, or other shapes of the sensor wells until Digital finely hits a resolution limit where size does matter. Personally I think we will start to see a reemergence of sensor sizes even up to 8x10 size.

Nikon will be rich if they can find software to fix the airy disk/diffraction issue. But somehow I think if they had the answer already they would have worked it into their visible light microscopes first.

Not only would they make bank, one or more of their engineers would be a Nobel laureate.

It will be very interesting to see what new technology that D400 sensor and image processor does use. I do hope it is "magical" and can combine both clean high ISO sensitivity with high definition low ISO sensitivity by using different processing at different ISOs.

"was a technology limit"
This is my point. Everyone calls it a technology limit looking in the rear view mirror but at the time people were declaring Heisenberg limits on 12mp along with physics diagrams how "it could never happen" and "images would be awful". When Nikon starts to throw red flags up, then I will think there is a limit. Most of this just comes down to how you count MP anyway.

"Second, Binning is a very common jargon term"
Ehh - Sorry, no one speaks of binning the Cheerios into containers so they stay fresh - not even developers. It is Jargon though - just for very few people. Using Jargon outside of the group who uses it muddles and stifles any good conversations on the topic. If your non-engineer friends would look at you strange if you used it, it is not common.

BSI == back side illuminated
Thanks.
---------------------
On the actual topic, if you can calculate the "Heisenberg limit" than you can calculate your way out of it (or account for it) with software.

Honestly I think we will be seeing more hexagon, or other shapes of the sensor wells until Digital finely hits a resolution limit where size does matter. Personally I think we will start to see a reemergence of sensor sizes even up to 8x10 size.

TaoTeJared said:
1 - What the F is BSI? Never heard of that before.
2 - Fuji has been doing this for years
3 - There is an old article I read on this probably 5-7 years ago and it not new. Their use of the word "bin" & "binning" has driven me nuts that it has stuck and people keep using it. There is binning in data mining but it has nothing to do with hardware or something solid that you can touch. Bin = Noun = trash can (Throw it away in the trash bin.) Binning - Not a real word = (someone's poor abbreviation of the word combining?) used for: Combining pixels or combining data. You can combine pixels or bind them together. Binding them together creates x. Pixels are a receptacle bin for light.

Now that is off my chest -

Sensors do this already with the RGB, combining/calculating 3 pixels for one space of color. Fuji has been utilizing the idea of combining pixels to gain dynamic range, same or different size for years with great results. At some point I think sensors will start to experiment more with this or start moving towards a Foveon type of route. Mongo's response is what drives the Sigma Foveon debate and everyone crazy.

There are multiple ways of gathering more light for low noise (and the actual path Nikon has taken) - 1) larger receptacle that has been used with FX - 2) Better software to handle it - 3) Improved sensor design that gather the light better into a smaller place. Nikon has gone step by step through all three of these over and over.

Many keep calculating the max DX can get is 24mp but that is with what the technology existed yesterday and not what tomorrow will bring. Who knows how designs innovations can change to increase IQ.

A sensor design from 8 years ago vs today (6mp vs 16mp): The smaller receptacle of the 16mp has created sharper and less noisy images with improved image quality. No one can argue that was not the case. Kind of contrary to what everyone has been saying with the 12mp vs 24mp argument that has going around lately huh?

OK my turn to nitpick :)

First, 6MP to 16MP was a technology limit, 16 to 24 starts to be a physics limit, more than that it starts to become a quantum mechanics limit. I am not sure that having three sensors gets rid of that, but to be honest I do not understand Heisenberg and the whole wave duality concept enough to know.

Second,

Binning is a very common jargon term;

It has been used in electronics manufacturing for decades, if you are making a resister you put the ones that test within 1% in one bin, 5% in another. If they test out at 10% with the spec they go in a third "bin". These containers are sold as three different products.

Noun form of bin == box you put things in (not a trash container as you defined it)
Verb form of bin == to put thing in box
binning == gerund or present participle form of the verb form of bin

Third, BSI == back side illuminated

1 - What the F is BSI? Never heard of that before.
2 - Fuji has been doing this for years
3 - There is an old article I read on this probably 5-7 years ago and it not new. Their use of the word "bin" & "binning" has driven me nuts that it has stuck and people keep using it. There is binning in data mining but it has nothing to do with hardware or something solid that you can touch. Bin = Noun = trash can (Throw it away in the trash bin.) Binning - Not a real word = (someone's poor abbreviation of the word combining?) used for: Combining pixels or combining data. You can combine pixels or bind them together. Binding them together creates x. Pixels are a receptacle bin for light.

Now that is off my chest -

Sensors do this already with the RGB, combining/calculating 3 pixels for one space of color. Fuji has been utilizing the idea of combining pixels to gain dynamic range, same or different size for years with great results. At some point I think sensors will start to experiment more with this or start moving towards a Foveon type of route. Mongo's response is what drives the Sigma Foveon debate and everyone crazy.

There are multiple ways of gathering more light for low noise (and the actual path Nikon has taken) - 1) larger receptacle that has been used with FX - 2) Better software to handle it - 3) Improved sensor design that gather the light better into a smaller place. Nikon has gone step by step through all three of these over and over.

Many keep calculating the max DX can get is 24mp but that is with what the technology existed yesterday and not what tomorrow will bring. Who knows how designs innovations can change to increase IQ.

A sensor design from 8 years ago vs today (6mp vs 16mp): The smaller receptacle of the 16mp has created sharper and less noisy images with improved image quality. No one can argue that was not the case. Kind of contrary to what everyone has been saying with the 12mp vs 24mp argument that has going around lately huh?

I don't see a point in touting binning as a feature. Couldn't one just downsize the picture in post?

if you take 4 pixels and count them as 1 that is binning

So a 24MP sensor, which is probably too large to be of much use in a DX camera would become a 6MP camera, with a much higher signal to noise ratio.

I don't see it happening, unless Nikon thinks the MP race is that much more important then image quality, but I think even that will lose.

People won't accept the 50% reduction in lines with a 4X reduction of MP if they care about MP in the first place.

Of course this is just my opinion.

Here is a good article on it.
http://www.microscopyu.com/tutorials/java/digitalimaging/signaltonoise/

Anyone with expertise on this subject?
Can binning and BSI produce both increased light sensitivity and increased resolution?

I am guessing the in-camera computer could be set to join or read or "bin" a number of pixels at high ISO settings to produce increased light sensitivity (say 4 pixels binned in a 24 megapixel sensor means essentially a 6 megapixel sensor at high ISO) and to unjoin or read separately or not "bin" each of the 24 mega pixels to produce increased resolution at low ISOs. Sounds like a fantastic idea to me if it works and why wouldn't it?

See these comments by Thom Hogan:

When I wrote earlier that the D400 would be 24mp, I got a lot of emails asking if I had written the number wrong (or if I was just plain crazy). Neither. First, we know that a 24mp DX sensor exists (or is about to exist in production form, from Sony). Second, back in 2003 I pointed out that the math said we'd get to about 24mp on DX before we exhausted the easy-to-see gains and started outshooting the best existing lenses. Third, at 16mp the Nikon would be trailing its two primary competitors in that market. Fourth, there's the "it's a mini D3x" notion that many will have when they see the D400. So, yes, we'll go there. 24mp is a done deal at some day in the future, so if that Sony sensor is good, the future is just about here. (Beyond 24mp I think things get much more fuzzy, and that's not just a pun on diffraction impacts.)

Now for the part that, I'm guessing, could be Nikon's surprise in August: binning. Binning isn't new. The D1 was a 10.4mp sensor that was permanently binned (four underlying photosites to one pixel). Nikon has played with binning since then, too, with the oddly binned D1x (two side-by-side photosites to one pixel). The answer to the "why 24mp question" suddenly becomes simple if you add binning to the mix: you potentially get a D400h and D400x all in one package. I do know that Nikon and Sony have been working on binning recently. That would make some sense for both stills and video and could lead to a jack-of-all-trades DX body that excels at everything: low noise, high resolution, video without artifacting, etc. And it would explain pursuing 24mp in DX and 36mp in FX.

But I'm only guessing at the binning thing. Without binning, and especially without BSI, we just have more pixels which means more data and power lines cluttering the sensor, meaning less light efficiency. Thus, even with the normal expected gains of a new generation of sensors, 24mp wouldn't really get us much further than the 16mp DX capabilities we already have, if any. Plus 24mp by itself really is pushing the full recording of diffraction into most image data.