This is a question that has been bothering me for a while... at what point it would be completely useless to increase the resolution of a full frame 35mm format dSLR since the lens can't provide enough resolution to support the amount of megapixels (just out of interest, I'm not planning to buy FF body anytime soon).

Lets take Canon 1Ds series as an example. The Mk 2 => 3 jump was from 16Mpix to 21Mpix (if I remember it right). Does Canon have lenses that can support this high resolution of the sensor? And I'm not talking about using photos for web display, but really pushing the limits of the photos by cropping or printing really large prints.

And what about Nikon? There naturally are the speculations of the D3x (like always) that perhaps would be 24Mpix (since Sony seems to be coming up with a pro camera with that sensor).

--
Remember to give constructive critique and help making dA a place that helps people to grow as artists. Don't know how? Read [link]
--
Nothing personal. I just say if I disagree and enjoy a nice debate.

Pretend you are dealing only with perfect lenses. They don't exist, but that's what you can do the math from. A pinhole is fairly close to a perfect lens and the same math applies.

What you've got is circular aperture diffraction. That determines how far you can stop down on a perfect lens - flaws in the glass determine how far open you can be. What we're going to do is find out how far you have to open up before you are not out resolving the media. If you get an answer like f/2.8, it means that if you stop down to f/4, you're introducing blur from diffraction.

So, the calculation is for one wavelength. [link] I'm going to use the frequency for a low pressure sodium vapor lamp - 589.3 nm (actually, two lines - one at 589.0 and one at 589.6, and this is 99% of the light that comes from the light - very monochromatic sources). This is a bit of a cheat because each frequency of light creates disk is a different size. I picked 'yellow' because it was about in the middle of the spectrum. If you use red and blue, you will get two different numbers.

I'm going to 'cheat' and not do all my work here and instead use the javascript calculator on the link above. I'm using 598.3 nm, 25000 micrometers (25mm) for the aperture and projected on a screen that is .5 meters (50mm) away.

Click on 'Displacement y' on the page and the size of the Airy's disk is given. Airy's disk is named after Sir George Biddell Airy (has nothing to do with 'air' and looks like a bullseye. The center is the brighest, and then it drops off, and then there is another ring around that, and then it drops off, and another ring, and it drops off again. It keeps going away, but only the first three are substantial. The wiki page - [link]

The first order (center) has a displacement of 0.0019270110014311418 cm. 0.019 mm. Guess what, thats very close (in terms of orders of magnitude) to another number we so love - the circle of confusion. For 35mm film that is 0.026 mm.

Lets try something.... lets go stop down to f/32 and see how big that first order maximia is to 0.26. Smaller hole, larger disk size. So, f/32 would be 50mm/32 = 1.5625mm = 1562.5 micrometers. And click the displacement y link again...

0.03083218186195728

Well, isn't that interesting - if you stop down to f/32, the size of the blur in the center of Airy's disk is slightly larger than the circle of confusion - it would be slightly soft - exactly what people see. And that is why lens makers don't let you stop down beyond f/32 on a 35mm camera.

That is shooting at f/32. A lens isn't sharpest at f/32, it is sharpest somewhere in the middle.

For an appxomiation of resolution in terms of lines per milimeter resolveable...
f/45 - 35
f/32 - 50
f/22 - 70
f/16 - 100
f/11- 140
f/8 - 200
f/5.6 - 280
f/4- 400
f/2.8 - 560

The above is from [link]

So, lets look at that f/32 number again. 0.030 mm. You go from black, to white, to black - thats resolving two lines in 0.1 mm, or 50 lpmm. That matches again.

So, in theory - Canon and Nikon already have lenses that do the job. Ever notice that EF-S lenses don't let you stop down to f/32? You could go with an even finer sensor, and make sure you shoot primes and don't stop down beyond say, f/8 and you will be able to fit more pixels on the sensor.

That f/8 sounds familiar.... thats the limit on point and shoot cameras. Lets take another look at that.

For a Canon G6, the CoC is 0.006mm. At f/8, you've got 200 lpmm, and Airy's disk is about 0.005mm across.

So, there's your answer - you can make increasingly fine sensors as long as you are willing to not stop down as far. Would you be willing to use an 80-200 f/2.8-5.6 (thats the entire range of the lens! ) ?

There are other problems too that aren't the optics. If you make a lens that is that packed full, that large, it heats up very fast. So you would have lots of noise even at the lowest ISO. And switching to even smaller photosites, you aren't going to have as much light hitting each photosite, so the sensitivity goes down, so you shoot longer, and the noise goes up.

--
Michael (shagie)

Thanks, I'll have to go through the answer with better time later.

--
Remember to give constructive critique and help making dA a place that helps people to grow as artists. Don't know how? Read [link]
--
Nothing personal. I just say if I disagree and enjoy a nice debate.

I think kodak tech pan goes to 600lpmm.

--
ENLARGE YOUR PEN. IS IT LARGE. PLEASE SPEND ALL YOUR MONEY ON PEN ENLARGMENT
I buy french fries a lot and then eat them, I spend most my time at home trying to put porn on YouTube*Baha05

Unfortunately Tech Pan was discontinued in 2004, and is now exceedingly hard to get hold of...

--
"They who would give up an essential liberty for temporary security, deserve neither liberty or security" - Benjamin Franklin

I had the same question; and as I also had access to two cameras with suitable pixel pitch I decided to simply try it out: [link]

Carefully with line pairs, lines and pixels! To add one missing bit of information, typical DSLR with cropped sensor and resolution in the 8 to 10 MP range have about 160 px/mm, the same number results for full frame at 20+ MP. So in theory the current sensor resolution limit is at about 80 lp/mm.
But of course to make use of the theoretical resolution possible with large apertures you need lenses which are more or less limited only by diffraction and not by optical errors at these apertures.
Additionaly, slightest focus errors kill resolution very fast, as well as camera shake, subject movement or with shots over some distance even air movement. Shooting handheld over a meadow in the sun into a dark wood the result is limited neither by the quality of the gear nor the diffraction...
Reaching more than 40 to 50 lp/mm on film/sensor isn't easy to do.

Just out of curiosity, how do you know all this stuff? Your knowledge far exceeds any average photographer's I know in person.

--
I own a camera, and it has a three beer limit on it.

I heard Pan F+ is also decent.

--
ENLARGE YOUR PEN. IS IT LARGE. PLEASE SPEND ALL YOUR MONEY ON PEN ENLARGMENT
I buy french fries a lot and then eat them, I spend most my time at home trying to put porn on YouTube*Baha05

Techpan is indeed a very impressive film [link] but it didn't do 600 lpmm in normal situations. [link] That shows up to 320 lpmm in a 1000:1 contrast setting (shooting alternating black and white lines in a brightly lit room). A more natural setting of 1.6:1 you could get 125 lpmm to 100 lpmm out if it depending on development.

--
Michael (shagie)