that's done by assuming a given signal to noise ratio that a serious photographer can live with and converting that to a well size, and also accounting for diffraction effects at the edges of wells caused by well depth. that well size turns out to be a little smaller than the 14MP sensor in the Pentax K20D. expand that to a full 35mm film frame and you get around 40-50MP. if you want to really know all the details, read http://www.stanford.edu/~pcatryss/documents/2005_SPIE-EI_Roadmap.pdf. brush up on your physics, math, and sensor design before diving in.
Herb...
Jammy Straub wrote:
I've read somewhere the max resolution you could squeeze from a 35mm sized sensor using a bayer pattern area is theoretically around 40-50mp. I can't remember where I read that though or what the math behind it was. Anyone familiar with that, Andre?
16.4MP, would do me very nicely on FF, but I wouldn't baulk at 24MP i the noise were good and the lenses were up to delivering tangible increases in detail. If the D3 had of been 16.xMP it would have been perfection IMO.
chemprof wrote:
Right on Herb. As I have said - my D200 will print excellent quality prints up to 16x24 in. These are prints that live up to very close scrutiny, however, I would say that anything much larger than that does NOT satisfy me. I find the D300 (so far - I haven't printed anywhere near as many prints from this camera yet) to be similar in print quality.
There is no way that I would be satisfied with even a 30x40 in print from these cameras, unless it was a reproduction of some piece of modern art with no fine detail in it... ...Show more →
And suppose you want to shoot landscapes.
Right now with state-of-the-art, money-is-no-object glass, wide angle lenses for small format (DX and FX) sensors still produce coma and chromatic aberration that's greater than the width of one 1 pixel at the edges. Which means that just adding more, smaller pixels buys you nothing because you're already inside aberration envelope of the best glass money can buy.
I don't know if this is a limitation of optical physics or engineering due to the short lens-to-film-plane distance they have to maintain for DX/FX cameras, but medium format doesn't suffer from this problem. So even if you're the kind of geeky pixel-peeper who likes to examine 4'x6' photos with a loupe, you've alreay hit an optical wall with higher-resolution sensors. Higher megapixel counts might make sense on bigger sensors, but they're already past the limits of the optics on smaller sensors.
And that's another thing - next time you look at the huge print ask yourself what limitation you notice FIRST - pixelation or optical limits? I notice the optical limits. Even with today's 12MP sensors, can you reliably take a picture that can resolve two pieces of detail (say, two hairs in a portrait or two slats in a picket fence) separated by one pixel of the sensor? It's very hard to resolve detail at the pixel scale that's not blurred across the pixels. Part of this is due to the lenses and part of this is due to the AA filter, but next time you read a lens-resolution test on a test-website ask yourself why even the best lenses don't seem to reach the maximum theoretical limit of the sensor being tested. Is Nikon going to come out with new, sharper glass to go with their 25 MP sensors?
Several years ago point-and-shoots passed the maximum useful pixel count for their tiny sensors, but the market wanted ever more pixels so they kept going anyway - 8MP, 9, MP, 10 MP. We're seeing the same thing with DSLRs.
Well I'm making do with my D2H, but I would like it if it were an 8mp camera. I think 8-10 is a good resolution for what I do. MY PC would die horendously on a 24MP file, but I can see times when I wouldn't mind using one, landscapes mostly. If the D3X is a slower high res camera at a similar price to the D3, then given the option I'd probably go for the D3X, because the speed of the D3 isn't important to me. Different things for different people.
I think a gigapixel should be enough (1,000 MB). The Gigapan allows people to take pictures which are more than a gigapixel in size, effectively turning a single photograph into a panoramic experience, around which the viewer can navigate on a computer. You should be all set.
plnelson wrote:
I understand they want it, I just can't figure out what they'll DO with it. The new sensor is 6236 x 4124 pixels. If you printed that at 300 DPI it would be 20.7 x 13.7 inches. No one will be looking at a print that big close enough to see 300 DPI of detail! So most of that great resolution will basically be wasted!
I was a medium format photographer (RB67) for years and it's not the sheer resolution that justifies MF. It's two other things -
1. You get wide optically flat perspective with lens focal lengths that can deliver optimal optical performance. Ever notice the problems they've had making really good, distortion-free, CA and coma-free wide-angles in the 10-20 mm range? NOBODY makes a really excellent one! The best of them have coma and CA and distortion. But it's easier to make a lens in the 50-80mm range at a MF - lens-to-film-plane distance that's practically perfect, so MF can give you wide angle pictures that are sharp and distortion-free corner to corner.
2. Bigger negatives shrink grain. But modern digital cameras have noise levels that are invisible at ISO 100, 200, and usually 400.
So basically, #1 cannot be solved by adding megapixels. It can only be solved by either using a bigger sensor than 24x36mm or by some breakthrough in lens design. And #2 is already solved.
Edited by plnelson on May 16, 2008 at 08:38 PM GMT...Show more →
Actually, #1 has been solved... by Olympus. Too bad they missed the boat on #2... but I'm still amazed at how well the 5 MP files from my old E-1 hold up to enlargement.
I'm seriously thinking about going back to using a 4/3 camera for casual shooting, and adding a 645 with a digital back for more serious landscape shooting, now that I have a 24" printer.
I've made beautiful gallery-quality (literally for a gallery) 13x19" prints from my 6MP D100. This works because the bigger a print is the farther away people stand when they're looking at it. Now I shoot with a D300 which has 12 MP and lower noise so the pixels are even less visible. If I made a 28"x42" print, sure, it would be only 100 DPI but at the distance anyone would look at a print that big they still wouldn't see the individual pixels.
Congratulations! You are on your way to being a photographer, not just a technician. Technical details tend to look concrete and are relatively easy to wrap your mind around, so you get a lot of technical focus (particularly from relative beginners). Wrapping your head around your vision is harder to discuss, partly because there are few absolute, final things like numbers....
The best for me would be a D2H with 8MP. Not that I feel limited too often with 4MP at the moment, very very rarely I just need to crop a bit more than I can. On the other hand, I never felt the need for more than the 6MP my D70s has either.
And I have some very nice A4 and 12x8 prints from both cameras.
For more MP than that, I'd rather have a medium format camera but mainly for the tonality, not so much for the MP. A 16MP 503CWD would do nice. Anyone care to send me one?
CA is trivially fixable in post processing. as you have pointed out in other posts, modern lenses have very high resolution on an optical bench and are mostly sensor limited. the Zeiss lenses are more than up to dealing with a 24MP sensor and probably would perform adequately at 40MP. most of Nikon's wide primes are not up to the job and only a few of their high end zooms would work well. that's because they are old and designed to a price point that is a lot lower than Zeiss does. it has nothing to do with inherent limits in optical design.
irrespective of all these, gigapixel prints like i saw this weekend are done by stitching. more high quality megapixels make that a lot easier and less necessary at the same time.
Herb...
plnelson wrote:
And suppose you want to shoot landscapes.
Right now with state-of-the-art, money-is-no-object glass, wide angle lenses for small format (DX and FX) sensors still produce coma and chromatic aberration that's greater than the width of one 1 pixel at the edges. Which means that just adding more, smaller pixels buys you nothing because you're already inside aberration envelope of the best glass money can buy.
May 19, 2008 at 11:00 AM
Andre Labonte Offline • • • • • • Upload & Sell: Off
Most of the times the answer depends not just on the lens aberrations (some of which are correctable) but on the plain, old CoC which determines the size of details a human can discern in an image.
Indeed, viewing a 25" image from 1m distance corresponds to the the standard CoC of 35 microns. For people with 20/20 vision it can be as small as 12 microns. Still, it's larger than the 5.7 micron pixel size of a 12MP DX camera. Increasing the MP is not going to improve the visual perception of an image.
the CoC assumes a certain final magnification ratio and we are talking prints that are way beyond that.
Herb...
gfiksel wrote:
For people with 20/20 vision it can be as small as 12 microns. Still, it's larger than the 5.7 micron pixel size of a 12MP DX camera. Increasing the MP is not going to improve the visual perception of an image.
Dr. Evil: ONE MILLION PIXELS!
Number Two: Don't you think we should ask for *more* than a million pixels? A million pixels isn't exactly a lot of pixels these days. A Nikon D80 alone makes pictures with over 10 million pixels!
Dr. Evil: Really? That's a lot of pixels.
[pause]
Dr. Evil: Okay then, we hold Nikon ransom for...
Dr. Evil: One... Hundred... BILLION pixels!
I was at a show a couple of months ago, an 8x10 landscape photographer. Being a former large format photographer I wanted to see the images. All were scanned and printed via an IRIS printer (didn't anybody tell him they aren't very archival) on 100% cotton water color paper, heavily textured. Well the show was disappointing. The images were gorgeous until you got close, the 8x10 detail just sucked you in, just when you were ready to get the finest of fine detail the paper texture took over, bummer, heck I could have gotten finer texture with my D200. I suggested to use a smooth water color paper to max out the detail and he said nothing like that came up to his standards? A very curious experience.
4ftx6ft and the same 25 inches. as pointed out on the other version of this thread over in Printing and Postprocessing, landscape prints are printed large and examined closely, just like paintings.
Herb....
gfiksel wrote:
What kind of magnification are you talking about - size and distance?
The question really should be: How many megapixels is too much?
I remember a while ago when the same type of race was on regarding computer video display cards, resolution and bit depth. After 24 bit color, the race was pretty much over. Why? Because we reached the resolving power of the human eye. Most eyes can't resolve more then the 24 bit color. OK - Some peoples eyes can resolve a 32 bit color depth. Because 32 bit color was beyond most people, we don't see video cards being developed that have a greater color depth. So 32 bit color was too much because it exceeds the color resolving power of the human eye.
The same thing happened with display resolution, even todays HDTV display maxes out at 1920*1080.
So, if you are displaying an digital image on a computer monitor, most of today's digital sensors already have too many megapixels and some have to be thown away to make the image small enough to fit on the display.
As other posters to the thread have noted, the higher resolution sensors are showing the limit to the optics associated with some of the lenses available today, so the new sensors have too many megapixels for the optics of some lenses.
People also note that larger images often need more megapixels in order to withstand close scrutiney when people walk up to the images. But there is also a limit to the field of view of the human eye. When you reach that point, walking closer to an image causes only part of the image to be viewed by the human eye. This means that if the optical issues can be resolved, that sensor megapixel count will find its limit when it can produce images that exceed the resolving power of the human eye, when the eye is at a distance from the display where the field of view of the human eys just starts to clip the edges of the image. That's when it will really stop. OK - It will probably go slightly past that, but not much.
...
? A very curious experience.