Just so there's no doubt who's winning and has the tiger blood, Canon yesterday announced that it's got an APS-H (29.2x20.2mm) CMOS sensor with—gulp—250 megapixels, approximately five times the resolution of the 5DS. [UPDATE: I should have written "five times as many pixels...." As Mike Plews noted in the Comments, resolution doesn't double along with pixel count. Ctein concurs: "Resolution goes as the square root of the number of pixels. Five times as many pixels means 2.3 times as much resolution." —Ed.]
Canon Global says, "Canon is considering the application of this technology in specialized surveillance and crime prevention tools, ultra-high-resolution measuring instruments and other industrial equipment, and the field of visual expression."
...Cameras to take pictures with, is what that last means. Other sites have quoted Canon saying the DSLR sensor it is developing will have 120 megapixels.
This is 16 megapixels, which is generally good enough for me. Or so I think now. Of course, I was pretty happy with my 6-megapixel Konica-Minolta 7D once upon a time, too, and now you can't even buy a new camera with such a miserly paucity of pixels. I guess I have to admit that the landscape is always changing and always will change; but, really, will I one day have to write something like, "The older D4385 Mark IV has only 250 megapixels..."? Perhaps it's inevitable.
More likely, the megapixel competition, like other such races from insufficiency to excess, will soon enough simply be sidestepped. I guess in practice I have already done so, since I own a full-frame 24-MP camera but seldom feel the need to use it.
The aesthetic problem is that super-sharp, highly detailed rendering is only one "mode" or strategy of pictorial representation. For some photographs it works, for some it doesn't. But it's getting to be all we have. When photographs have more hyper-realistic detail than the human being with the best possible eyesight can see with the naked eye, that will be technically miraculous, but aesthetically just one effect...among many potential ones. The water will find its way around that dam. Just as not every single painting is trompe-l'œil.
"A version of an oft-told ancient Greek story concerns a contest between two renowned painters. Zeuxis (born around 464 B.C.) produced a still life painting so convincing that birds flew down to peck at the painted grapes. A rival, Parrhasius, asked Zeuxis to judge one of his paintings that was behind a pair of tattered curtains in his study. Parrhasius asked Zeuxis to pull back the curtains, but when Zeuxis tried, he could not, as the curtains were included in Parrhasius's painting—making Parrhasius the winner." [Wikipedia.]
Mike
(Thanks to Mathew)
Tech Ed. Ctein adds: In anticipation of the inevitable flood of "why do we need so many pixels" comments, readers might want to reread "Why 80 Megapixels Just Won't Be Enough..."
Which doesn't even factor in that those of us who used negative film saw no sin nor deficiency in cropping, sometimes to extremes, when it made for a better photograph. Perfect in-camera compositing is no special virtue, just a style.
Original contents copyright 2015 by Michael C. Johnston and/or the bylined author. All Rights Reserved. Links in this post may be to our affiliates; sales through affiliate links may benefit this site.
(To see all the comments, click on the "Comments" link below.)
Featured Comments from:
LeftCoastKenny: "I've felt for a long time that photographic sharpness is highly overrated. Looking at a photograph that is so sharp you can see the rods and cones in the back of someone's eye from forty yards away may appeal to some but it simply rubs me raw. In the last few days I've taken to photographing things intentionally out of focus, and finding simple light and suggested shape very appealing. In a way, it is an antidote to the jangled nerves I get looking at knife-sharp images."
Kevin Purcell: "The pixels in the Canon APS-H (crop factor x1.3) 250-MP sensor are 1.4 micron, so about cellphone sized. So nothing too novel there (though Canon probably aren't doing this in 180 nm process). They're probably monochrome pixels too.
"But It's not (just) the pixel count that's important. It's this:
The new Canon-developed CMOS sensor, however, despite its exceptionally high pixel count, achieves an ultra-high signal readout speed of 1.25 billion pixels per second, made possible through such advancements as circuit miniaturization and enhanced signal-processing technology. Accordingly, the sensor enables the capture of ultra-high-pixel-count video at a speed of five frames per second.
"The sensor delivers read five 250-MP frames per second at a data rate of about 15 Gbps. The real issue is how you deal with the data coming off the sensor at that rate (and deal with the heat from the DSPs and the sensor).
"The sensor size is interesting: APS-H. That's bigger than APS-C but smaller than 135 (36x24mm). It's the largest size you can make a sensor in a single 'masking step.' This might be needed as they might have a problem stitching more than one step with 1.4 micron pixels. It also makes me think they're actually producing this for a customer. I suspect a customer has worked with Canon to specify this sensor.
"I would speculate that the customer might be interested in flying these sensors with a 'normal-ish' lens on a UAV to make an 'simplified' GORGON STARE (a.k.a. Wide Area Airborne Surveillance System) -like system. The first WAASS is said to use 192 different cameras with increment 2 using 368 cameras. A 250 megapixel sensor should reduce the number of cameras needed.
Gorgon Stare 1
Gorgon Stare 2
Reaper Drone pod
"In systems like this you really want to put the image processing on the UAV so it can keep track of 'things' on the ground then send the digested information back to base. This data can be kept in a database so you can ask questions about movements on the ground ('What path did this guy just walk?' 'Who left this building in the last 10 minutes?' 'Where did this car come from?' 'When did it arrive?' 'Where did the occupants go?').
"This sensor is less about making photographs than extracting information from the environment. It's not coming to a camera near you soon!"
Richard Newman: "One question is; What are you using the higher resolution (related but not identical to bit count) for? If you want very large prints, then more may be better, but if you aren't planning on 90x120-inch prints, 30–50mp may be (more than) enough.
"For me, the advantage of higher resolution is in the ability to create good images by cropping the initial frame. On numerous occasions I have been unable to get as close as I would like, and have had to include extraneous elements in the frame. Post processing these starts with cropping. Unfortunately, with lower resolution, this can result in 'grainy' images. With more resolution I have more flexibility. On occasion I have been able to extract three or four good images from a single frame, while maintaining good detail, color and contrast."
James Bullard: "A thought in response to your observation that not all photos need to be that sharp: as a fairly regular user of Topaz filters I know that at least the Simplifier filter (and perhaps others) achieve their effects by averaging and grouping (albeit in irregular patterns). When using such filters you are, in a real sense, lowering the resolution of the image. The difference being that you are doing so stochastically as opposed to the regular pattern of a grid of squares."
Simon: "I can remember giving medium format a go back in the '80s. I parted with it fairly soon as it had more detail and tonal range than I wanted, happily returning to 35mm. The level of detail in every digital camera is astonishing but I agree is not the only pictorial approach."
Bernard: "The reason to use 250 megapixels isn't resolution, it's smoothness (for lack of a better term; no doubt there is one in Japanese). It's the same reason why portrait photographers used medium and large format in the film era: they didn't want their subjects' flaws to be sharper, they wanted a smoother look.
"The same principle applies in other fields. A sauce that's been simmering all day has more depth than one that's put together in ten minutes. A full orchestra has a fuller sound than a quartet.
"I'm hoping that the next generation of ultra-high resolution sensors finally puts the 'pixel peeping' era to rest. There will be no need to examine pictures at 1:1 on a monitor. We can finally all zoom out and look at the big picture, as it were."
Michael Perini (partial comment): "Cameras reached the point of sufficiency a while back for most peoples' needs. Most modern cameras are very very good by any historical imaging standard. That's wonderful. But it shouldn't stop us from seeing how good we can get. When I say good, I mean technically good, not aesthetically good which is a separate matter. So I say let it run—let's see how good we can get. No one is forcing us to use the technology unless we choose to."
I could really be wrong here but since sensors are two dimensional wouldn't the inverse square rule mean that you would need around 200mp to double the resolution of a 50mp chip?
This should be handled by the TOP brainiac division, not me. Just vamping here.
Posted by: Mike Plews | Wednesday, 09 September 2015 at 10:58 AM
I hope that one of the happy by-products of ever-increasing pixel densities will lie not in higher resolutions, but rather more useful post-processing...
The ability to place multiple (perhaps dozens of) pixels behind a single microlens could lead to hyper-local noise reduction techniques, oversampling or creative ways of extending dynamic range.
Put another way: file sizes needn't continue to grow along with megapixel counts. What we think of as a single image pixel might ultimately be constructed from the input data of many sensor pixels - algorithms to interpret Bayer array data, or alternate topologies like the Foveon sensor, are only scratching the surface of what's possible.
Posted by: Dave | Wednesday, 09 September 2015 at 11:01 AM
The first digital camera I ever used was a flatbed scanner, and while all of its pixels were crap and its resolution (lp/mm, not megapixels) was poor, and it was prone to producing strange artifacts on a pixel level, massive downsampling provided absolutely beautiful final files.
When I was first considering buying a DSLR, in the early 2000's, I decided that 12 megapixels would be barely adequate. I regarded--and still do--my first DSLR, at 22 bayer matrix megapixels, as a low resolution (in megapixels, not lp/mm) camera. The D800 and A7 go in the same bucket.
The 5DS is starting to get where I'd like to be in terms of pixel count, but the prospect of a 120 megapixel full frame camera (especially if it's functionally identical to the 5D Mark III) has me drooling. I won't gain many (if any) lp/mm, but all those pixels will eliminate demosaicing artifacts on images destined for screens (even 8k!), and if noise is the same or even a little worse than the current state of the art, my prints and screen images will look better than they do today.
Posted by: James Sinks | Wednesday, 09 September 2015 at 11:18 AM
I am not the first person to say this, but I really think we're in the midst of defining the new versions of "35mm" and "medium format".
16mp to me holds the place of the best 35mm film: Small, nimble, but big enough for most every use.
36mp and up feels (and looks) like the best of medium format, with the same constraints: big, heavy, something you only bring out when you need the big files.
(my new 24mp? 645, of course.)
Posted by: david adam edelstein | Wednesday, 09 September 2015 at 11:37 AM
A few back-of-the-envelope calculations yield the result that this 250 megapixel sensor has pixels about the same size as the diffraction disc for an f/1.4 lens, and since the diffraction disc gets larger at smaller apertures, it doesn't require much in the way of stopping down to wind up with an image that's much less detailed than the sensor spec would lead you to think. At least in terms of pixels per unit area of sensor, this thing is close to the useful limit.
Posted by: Bill Tyler | Wednesday, 09 September 2015 at 11:41 AM
Mike,
In this mindless mega pixel race the real losers are we, the end users and buyers of cameras.
Let us face it, the vast majority of photos taken in these times never get printed. They are just viewed on the screen and sadly very often just deleted. That means, all those bucketfuls of mega pixels are a big waste of resources and sometimes a nuisance. What the end user needs is a compact and light camera with good low light performance. I hope, one of these days the camera makers will realise that and stop this mad mega pixel race.
Ranjit Grover
Posted by: Ranjit Grover | Wednesday, 09 September 2015 at 11:46 AM
I "learned" a long time ago never to buy the "first off the block." I have proved this with my 2 (mind you I am married to a photographer) EOS 5ds and 5dsr Canons. It is inevitable that the mfr., will come out with a beefed up version (such as the Fuji X-100 now replaced with an X-100s).
The worst part is that many of my plugins struggle with such large files, e.g. Topaz will ALWAYS send a C++ error from the Canon CR files.
The upside is the files are incredibly sharp and detailed. They go with us this October (with our Fuji's) to France.
And, YES! 50mp is pleanty! I doubt that I will even look at the 120...well maybe just look.
My two pesos, Mike
Posted by: Hugh Smith | Wednesday, 09 September 2015 at 11:49 AM
...like other such races from insufficiency to excess, will soon enough simply be sidestepped.
I'm not so sure in this case. This has been stated before, and yet, here we are again. :)
Posted by: Richard | Wednesday, 09 September 2015 at 11:50 AM
Building on your closing illusion, I once had a friend describe the quality of the paint job he had recently applied to a car as,"So slick flies won't land on it because they think it's wet." Which has nothing at all to do with the megapixel race, but your closing words brought it to mind. So, OT alert on my comment.
Posted by: Jim Allen | Wednesday, 09 September 2015 at 11:56 AM
I read about the Canon sensor on other sites yesterday and some commenters pointed out that the press release didn't indicate it was a color sensor. And I think that from a resolution standpoint it isn't any finer-pitched pixel-wise than sensors used in phones, or at least in the ballpark. Certainly the readout speed is impressive, and would benefit all photography.
I'm surprised that the use cases include things like reconnaissance photography or things like that, as that level of resolution I should think requires an extremely stable platform for the entire image chain (and what lenses can resolve detail at that level? I dunno, maybe some sort of fly-like compound lenses with smaller image circles can be put in front.)
I do think we'll see higher pixel-count sensors, but I suspect those pixels will be used for computational photography usage and we will still be working with image files in the same pixel counts ranges we use today.
Patrick
Posted by: Patrick Perez | Wednesday, 09 September 2015 at 11:59 AM
Hey Mike,
I am looking for a used Cray Super Computer so I can deal with
the files out of this camera. The last Cray used price I saw was
$500,000!
Posted by: Dana Thomas | Wednesday, 09 September 2015 at 12:06 PM
Canon announces vaporware to try and obscure the fact that it's now in third place in the DSL/DSLR wars, behind both Sony and Nikon. It's not what you say you're going to do, Canon... it's what you actually do (or didn't do for users when you had the chance). August, 2005: 5D, great! 5DII better, 5DIII @ 22MP,no 4K, zebras, etc.? Over 15 years?!? Meh. I own them all, use two of them daily, and just ordered the Sony aR7II. Finally, a camera maker who gets me!
Posted by: Joe Boris | Wednesday, 09 September 2015 at 12:10 PM
Dear Bill,
If this were in a Bayer array, the "optimum aperture" would be more like f/2.8. Which is not all that different from the optimum aperture for super-sharp photographs in 35mm, which was typically f/4-f/4.5. Notably, the depths of field would also be similar.
Just think of this like loading up your film camera with TMAX 100 or Ektar 25. You often (usually) didn't make full use of the film's superior resolution. But it was sure nice to have it when you wanted to.
Finally, resolution is not the simple sum of the parts. It's more mathematically complicated. Improving one part of the chain improves overall sharpness. You'd have to stop your hypothetical 250 Mpx camera down to f/8 before you'd stop seeing any gain in sharpness from the ultra-high-resolution array.
'Course, it ain't gonna be no low light champ. But one must suffer for one's art, n'est-ce pas? [g]
pax / Ctein
Posted by: ctein | Wednesday, 09 September 2015 at 01:15 PM
A very high resolution sensor has one potential advantage, assuming nearly perfect lenses and favorable situations.
While one can reduce the level of detail and sharpness of a very high resolution image through later post-processing, the opposite is not true. You can't enhance detail that's not captured in the first instance by a lower-resolution image. Theoretically, then, the higher resolution sensor potentially provides greater flexibility for some circumstances where hyper-sharpness may be needed for practical reasons or might be artistically appropriate.
That said, anything beyond 50MP or 60MP would seem wasted at best for most non-surveillance or non-scientific uses. 50MP to 60MP sensors already allow 30-inch wide images at 300 dpi native non-interpolated resolutions. Very few people have either the ability to print images larger than that or the space to display them. Our eyes are not really capable of discerning significantly higher dpi resolutions. One well-publicized test several years ago by the New York Times found that people could not discern whether very large prints were made with 5MP, 10MP, or 16MP cameras.
Unless you're making very large prints 44-inches wide or bigger, really high resolution sensors thus may not provide any practical benefit, and even then only when subjects are static and near optical infinity. I found that to be true in my own experiments just a few days comparing normal 16MP images made with an Olympus E-M5 II with the same images taken immediately afterwards using that camera's 64MP high-resolution RAW mode with a heavy tripod, remote release, and the same excellent lenses used with very careful magnified manual focusing and optimum apertures.
At the higher resolution and 100% magnification, depth of field and nearly imperceptible subject motion were frequently limiting factors, reducing the fine detail crispness of important parts of many image, even with the shorter focal length lenses used with Micro Four-thirds cameras and view camera-like slow, careful technique. The same camera and lens combinations produced crisp corner-to-corner sharpness with flat test charts but not in the real world.
I doubt that most people using very high resolution cameras would remain confined to using tripods with static subjects, nor routinely be making 44-inch plus prints of subjects near optical infinity and thus would not benefit from markedly higher resolution sensors.
Posted by: Joe Kashi | Wednesday, 09 September 2015 at 01:35 PM
I believe there is such a thing as "human sufficiency," a point in tech where actuals humans don't need anything more, and won't readily pay for it. Consumer computers now aren't much faster than they were five years ago, because for most people, they don't need to be. iPad-style devices are seeing slowing sales, because people are finding iPhone 6-sized devices to be good enough for most of the things they'd do with an iPad. There are certainly people who can always use more (I'm thinking of Ctein, with his super iMac,) and who use the most elaborated versions of heavy-duty software, but they're rare, and they don't drive the economics of producing faster and faster devices. I believe the reason that Adobe and Microsoft are going to subscription models for their software (Lightroom, Word, etc.) is because people won't pay for stand-alone "improvements," so they need another way to create a steady cash-flow. I think the way we go to "faster" and "better" from here will more resemble a drift, than the lurches of the past. We may get to 120-megapixel sensors simply because they become so cheap that there's little advantage to using 50mp sensors, not because the mass of people either need or want them.
Posted by: John Camp | Wednesday, 09 September 2015 at 02:43 PM
I have owned a 40mega-pixel Pentax 645D and currently own 24mp and 36mp Nikon dslrs. These are all pretty wonderful cameras with - among other things -extremely good detail resolution. But yesterday I made a 24" print from a file shot several years ago with the lowly 12mp Nikon D300. The print quality is excellent; the resolution of detail is excellent; colour, contrast and dynamic range are excellent.
Though I love my D600 and D800 I have to ask myself - what made me think I needed more than 12mp, and what practical, tangible (visible) benefits resulted from the move up to 24 and 36mp?
Posted by: David Paterson | Wednesday, 09 September 2015 at 02:47 PM
At the other end of the scale. The new 12Mp iPhone 6s and 6s+ will shoot 4k video!! Plus Live Photo, that records 1.5 seconds of video before and after the photo. I've seen Live Photo like shots in web editions of the LATimes and Sports Illustrated (the future is not-so-still photos).
BTW you will be able to edit 4k video on you iPhone.
BTW2 i've never shot a photo/video with my iPhone 4s, so please don't call me a fanboy.
The photo world is changing rapidly, no thanks to hidebound CaNikon.
Posted by: c.d.embrey | Wednesday, 09 September 2015 at 03:24 PM
Extremely high resolution photographs (more meaningfully, extremely high detail photographs) sometimes offer a view of the world that can't be obtained any other way. This has been true since the early Jurassigraphic era. If you're not familiar with Charles Fontayne and William Porter's 1848 Daguerreotypes of the Cincinnati waterfront, they're well worth a very close (sic) look: http://1848.cincinnatilibrary.org/showplate.php?id=1. Their level of detail clearly contributes to making them so compelling.
Fontayne and Porter had to invest a lot of effort to create their panorama. Today, it's remarkably easy. During a recent visit to Seattle and Vancouver, I had a lot of fun making handheld waterfront and skyline panos with my Fujifilm X-T1, combining them with Lightroom's new "photo merge" capability, then post-processing the composite as though it was a single raw capture. Admittedly, some of them exhibited visible stitching errors—probably as attributable to my sloppy handheld technique as to Adobe's algorithms—but the process was entirely automated and essentially effortless, and the results from my little 16 mPx Fuji were quite satisfactory.
Not that I wouldn't pop for an ultramegapixel camera if it was as compact and light as my Fuji.
Posted by: Chris Kern | Wednesday, 09 September 2015 at 03:38 PM
"The aesthetic problem is that super-sharp, highly detailed rendering is only one "mode" or strategy of pictorial representation. For some photographs it works, for some it doesn't. But it's getting to be all we have."
In my visits to the National Landscape Photographer of the Year exhibition in the UK, I am always troubled by the almost universal obsession with a kind of technical quality that includes - amongst other attributes - high resolution and sharpness. I can admire the technical skill required, but when seen en masse it makes me long for some landscape work that adopts a different aesthetic. (Such work exists, of course, but doesn't seem to meet the expectation of that competition and exhibition, at least).
Posted by: Nick Hunt | Wednesday, 09 September 2015 at 03:47 PM
Didn't Ctein once do a post on why 80MP won't be enough? Would be interesting to hear his thoughts.
My first thought on hearing the news from Canon was about the quantity of information, not necessarily resolution. Perhaps over simplified, but 250MP with a conventional RBGG filter array might reduce to the equivalent of 60MP with full color information at every pixel. Or there might be ways to enhance dynamic range, improve tonal rendition, reduce noise and so on.
I have little or no need for 250MP of resolution, but I'll take improved color and tonality any time I can get it.
Posted by: Gato | Wednesday, 09 September 2015 at 04:33 PM
In other news, Apple's new camera (I hear it's hooked up to a phone of some sort) has 12 mega-pixels, up from the paltry eight mega-pixels of its predecessor. The Wall Street Journal adds ...
The step up in hardware, plus some extra software secret sauce, means improved low-light performance, better detail and enhanced color. It also delivers even bigger panoramas (though they were pretty huge to begin with). On the video side, the new camera can shoot 4K video. The iPhone 6S can also edit that 4K video, which is crucial.
http://blogs.wsj.com/digits/2015/09/09/apples-iphone-6s-vs-iphone-6-the-key-differences/?mod=WSJBlog
Proving once again that better photography requires more than just more.
Posted by: Speed | Wednesday, 09 September 2015 at 04:55 PM
I'd like more pixels, but I'd prefer more dynamic range.
Posted by: Auntipode | Wednesday, 09 September 2015 at 05:01 PM
Rather than trying to cram more pixels into the 35mm size sensor, I would rather see larger sensors that compare to medium format and large format film. I really love the way the larger lenses render images with an amazing transition of tones. Depth of field transitions from sharp to soft in larger formats are also pretty amazing.
Posted by: Joseph Brunjes | Wednesday, 09 September 2015 at 06:38 PM
Canon stated that the expected uses would be for (aerial?) surveillance and technical imaging (possibly with advanced pattern recognition software) for manufacturing, say , chips.
This 250 MP sensor has no relevance to actual photographers. It does have relevance to government (spy agencies) and industry. If you consider that any camera would be fixed in place (industrial) or floating in space (satellite), the technical issues involving camera motion degradation go away.
Posted by: NancyP | Wednesday, 09 September 2015 at 06:44 PM
Less than 10 years ago when the average high-end digital cameras were recording 8mp-12mp we heard the same howling when Canon said it was heading to 21mp-24mp.
Today Apple bumped their new iPhone to a 12mp sensor.
With apologies to Tina Turner, "What does need have to do with it?"
Just sayin', guys. Just sayin'
Posted by: Kenneth Tanaka | Wednesday, 09 September 2015 at 07:15 PM
... And they're still losing marketshare to 8-12 MP smartphones.
Posted by: MarkR | Wednesday, 09 September 2015 at 07:38 PM
Hell yea! Let the race continue, more is more(.)
As a panoramic photographer that uses computations to stitch multiple images to form a single one, more is really more.
The largest print I've produced to date can be seen in NYC. At the 50UN real estate show room. It's a Duratran (think slide film) printed at 12 feet tall by 40 feet long at 155 ppi. The original 16 bit tiff was a bit over 100 gigabytes. You can stand next to it and not see any pixels.
When I visited the showroom I got what is probably one of the nicest compliments my work has ever gotten. One of the managers of the showroom, on the first floor, said they sometimes forget they are not on the 23 floor where the image was taken.
Posted by: Robert Harshman | Wednesday, 09 September 2015 at 07:43 PM
This is not the first analog to digital migration.
Does anyone remember the tag line "Perfect Sound, Forever" as applied to then new Compact Disks. 16bit 44,1k with the A/D & D/A converters of the day was neither perfect, nor forever.
It took quite a while to get better, and it took bit rates and sampling rates that would have seemed ridiculous when we began.
Some think that digital sound never surpassed analog sound.
My ears are no longer good enough to know.
But I do think with regard to photography it is too soon to tell if we have reached any limits yet.
In the consumer market, convenience always wins and the bids for an audio standard above CD quality was killed by MP3 ,AAc which are generally lower in quality than CD's
Cameras reached the point of sufficiency a while back for most peoples needs. Most modern cameras are very very good by any historical imaging standard. That's wonderful. But it shouldn't stop us from seeing how good we can get.
When I say good, I mean technically good, not aesthetically good which is a separate matter.
So I say let it run, lets see how good we can get. No one is forcing us to use the technology unless we choose to.
Does anyone know why no one but MF CCD's has done true pixel binning.? It seems like a 60mp / 15mp sensor with 4x ISO at 15mp would be really interesting.
Posted by: Michael Perini | Wednesday, 09 September 2015 at 08:23 PM
Mike,
I think two points need to be addressed in this discussion. First, many folks use 300 dpi for native printer resolution. That may be true for some printers, but not all. The Epson 3880, for example, has two "native" resolutions: 360 and 720 dpi. Second, it seems to be widely held that 36 (or 50) MP is all one "needs" for moderately large prints -- such as those that can be made by the aforementioned Epson 3880 (about 23 inches long by 17 inches wide).
How many megapixels do we need to make an "optimum" print that is 18 x 12 inches? Here's how I think about it. 360 dpi corresponds to 180 line-pairs/inch (or about 7 lp/mm). The 3880 can easily print 7 lp/mm on glossy paper, and you can see the line pairs on close viewing without magnification. 180 lp/in x 18 inches = 3,240 line-pairs on the long dimension of the image. That is "information" that the printer can put down on paper and that you can actually see (up close).
What sensor resolution do you need in order to reliably capture 3,240 line pairs with high contrast? I have argued elsewhere that you need at least 4 rows or columns of photosites in order to record one black-and-white line pair with high contrast (http://philservice.typepad.com/f_optimum/2014/09/limits-of-resolution-1-sampling-frequency.html). In effect, we need to oversample at image capture. So, in order to record 3,240 line-pairs with high contrast we need 12,960 photosites on the long dimension of the sensor. Assuming a 3:2 aspect ratio, that means 8,640 photosites on the short side, and approximately 112 MP total "resolution".
Since we are making an 18 x 12 print, we must print at 720 dpi in order to avoid down-sampling (12,960 / 720 = 18 inches). (Remember, 720 dpi is the other "native" resolution of the 3880).
Before anyone complains that "we don't take pictures of line-pairs", I'll just say that we need some quantitative measure of "information" in order to get beyond hand-waving argumentation. I think it is an open question whether the sort of capture oversampling I'm describing here will yield results that can be seen in the print -- I've never seen a rigorous test. The answer will almost certainly depend on the actual subject matter, the quality of the lens, the amount of diffraction blur, and the distance from which the print viewed.
(Note: With a 36 x 24 mm sensor, 3,240 line-pairs in the long dimension corresponds to 90 lp/mm. I don't know if there are any full-frame lenses that can resolve that well with high contrast.)
Posted by: Phil Service | Wednesday, 09 September 2015 at 08:26 PM
The Oct issue of Shutterbug has a test of the Canon 5 DS which is "only" 50MP. They did resolution tests with the 50/f1.8 lens Canon provided. At f5 resolution was 5,497 lines, but it retained that kind of performance only between f4-7. Above that it dropped to 3000 lines and was only 3500 lines at f3.2. I suppose that can be taken to mean that sensor resolution is no longer the gating item.
Posted by: Jim | Wednesday, 09 September 2015 at 09:55 PM
As I recall in the film era lot of photographers including a lot of professionals that thought 35mm was good enough, a few photographers that used medium format and even smaller number of photographers who thought that 8 x 10 was just about right.
I do not recall any photographers that used 35mm cameras complaining that their arm was being twisted to buy an 8 x 10 camera.
I really don't get why people complain about equipment they have no interest in being offered for sale to people who are interested.
One might as will complain about how many minutes it takes a Bugatti Vyron to wear it's tires out at 250 mph
From this amusing article http://m.caranddriver.com/reviews/bugatti-veyron-2011-bugatti-veyron-164-super-sport-review
"A set of the Super Sport’s special Michelin tires costs $42,000 and may last 10,000 miles if you’re careful, though they last only 15 minutes at the car’s top speed (at that pace, however, the 26.4-gallon tank is sucked dry in just 10 minutes, and there’s no place on Earth to safely go that fast that long anyway, so no worries). At the third tire replacement, Michelin requires that you also swap out the $69,000 wheels—coincidentally, the only wheels that fit those tires—to ensure a proper bead seal."
Posted by: hugh crawford | Wednesday, 09 September 2015 at 09:58 PM
will I one day have to write something like, "The older D4385 Mark IV has only 250 megapixels..."?
To use the American expression (I think), "You should live so long."
Perhaps one day we'll write, "He lived a full life and died at the age of 500Mpixels."
Seriously, I used to marvel at the depth and velvety smoothness of large format (4"x5" and up). When film grain was absent and great lenses were used, the images were glorious. I would hope that these super resolution sensors will give us that magic in digital.
Posted by: Peter Croft | Wednesday, 09 September 2015 at 10:13 PM
"now you can't even buy a new camera with such a miserly paucity of pixels."
But you can use one, to excellent effect.
Posted by: Jeff Glass | Wednesday, 09 September 2015 at 10:21 PM
The more digital data you have, the more analogue it gets.
Oversampling is fine, and this is effectively what this is. Far more 'bits' than is necessary to resolve the number of lines that the lens can actually resolve...
So, what's the point? Well, no such thing as false detail. No moire, no aliasing, just the progressive onset of optical blur. Analogue resolution.
250 MP is 5X the 5DS, but in resolution terms is it only actually 2.2X. This is interesting because whereas most lenses good lenses can now just about exceed Nyquist on a 50MP sensor at around MTF20, I think this will reverse that. Lenses will never quite reach Nyquist, even at very low contrast levels. Hence, zero aliasing.
Will it make better prints? Well, resolution is not the only arbiter. Smaller pixels won't make much difference to noise, and the large you print, the more visible its effect becomes. That won't change.
Nor will the fact that resolution is already limited in a practical sense by mechanical vibration.
Moreover, in print terms, we are already into oversampling. A 16MP print 24X16" viewed from 24" has a PPI resolution close to the limit of human vision. A 50MP image at the same print size is therefore already being 'downsampled' by the brain.
I suspect the number of people printing much larger than that is limited, and besides, larger prints are viewed from further away...
Even so, one area (other than moire) that may benefit is the resolution of colour detail, particularly small point-details against a coloured background (a distant poppy field for instance). This is one serious Bayer shortcoming compared to film, at least for landscape photographers.
Using more data is actually a very cheap and pragmatic way to achieve better quality. I haven't worked it out, but each pixel may be small enough to allow 12bit or even 10bit sampling as well (pixel level SNR won't be good enough to require more than that) which will help to reduce ADC noise and speed up data throughput.
So, I applaud Canon. This isn't crazy, given that we can still expect sufficient improvements in storage densities and processor speed. In fact, it's probably cheaper than adding an AA filter to a sensor. After all, it's just lots of 1" sensors stuck together.
I just don't think people should expect vastly more useful resolution than they already get with a 5DS. What you will get is a print which is totally free of digital and demosaicing artefacts, even at large print sizes.
Posted by: Steve Jacob | Wednesday, 09 September 2015 at 10:40 PM
Since 2000 when I got my first digital I have gone through quite a few digital cameras of higher resolutions and other advances, but for the last 3.5 years I have stuck with 16mp (4 different cameras with updated features and other tech, but same 16mp sensor) and have been happy with that. 3.5 years. Quite a long time in the high tech world, but I'm content with it. I also still shoot a lot with a 12mp camera. Sure, more megapixels will come and someday whether I care about more than 16mp or not I will get a new camera that has other things I want and end up with more megapixels too.
Posted by: Henry Richardson | Wednesday, 09 September 2015 at 10:51 PM
Dear Mike Plews,
You are correct-- resolution goes as the square root of the number of pixels. Five times as many pixels means 2.3 times as much resolution.
pax / Ctein
Posted by: ctein | Wednesday, 09 September 2015 at 11:29 PM
Eventually technology will render imaging devices obsolete; then we the viewer shall have such images rendered in our brains via the chip so implanted in our bodies at creation.
The description named/called megapixels are simply the explanation of how we view what is around us. It is not for us to improve upon; it is already perfect for our requirements!
Posted by: Bryce Lee | Wednesday, 09 September 2015 at 11:33 PM
"The water will find its way around that dam. Just as not every single painting is trompe-l'œil."
This is already happening. Many people intentionally degrade the color, tonality, and image quality of their photos today with software "filters." For examples see Snapseed, VSCO Film, Instagram, and many more.
Posted by: Bruce Mc | Thursday, 10 September 2015 at 03:08 AM
Having migrated from a 39MP MFDB to a Sony a850 to a Nikon D800 and finally to 16MP MFT Olympus cameras (EPL-5, EM-1, EM-5 II), my favorite cameras by far are the Olys--great optics, lovely pixels, splendid dynamic range. The diminutive form factor of these cameras is an asset. Sure, noise often becomes noticeable at ISO 400. Fortunately, I like the "grain" and work with it rather than fight it. And every so often, when I get an itch to use a view camera, I reach for this. http://forum.luminous-landscape.com/index.php?action=dlattach;topic=103579.0;attach=130234;image
Posted by: Bob Rosinsky | Thursday, 10 September 2015 at 09:04 AM
Bah, "640K ought to be enough for anyone."
8-)
Posted by: Doug Sundseth | Thursday, 10 September 2015 at 10:27 AM
Dear Phil,
I'm not sure what "high contrast" means, quantitatively... Or why it matters, really, when talking about resolution. It's part of the MTF, it's not a defining characteristic of resolution.
That said, back in the film days, I could routinely exceed 100 lp/mm on film in 35mm film / lens tests, which indicated the lenses were doing considerably better than that, when you combined lens, film, and focus sources of blur.
Several prime lenses were clearly close to diffraction limit at f/4-5.6. In other words, they delivered circa 300 lp/mm. Not at "high contrast" of course, but at clearly-visible contrast.
Newbies keep inventing what they think are meaningful metrics, like the limiting resolution for high contrast. They're wrong. Ignore them.
pax / Ctein
Posted by: Ctein | Thursday, 10 September 2015 at 12:08 PM
With that high resolution who is going to buy any more another lens than a Ultrawide and pic out the frame you need on your computer? Is this the death of lens making?
Posted by: Peter Weimann | Thursday, 10 September 2015 at 12:09 PM
I suppose in some ways that more data is more "analogue," but to many people these days analogue implies appealing imperfection, not smoothness. If you have the ultimate sensor and lens that can digitally record all the pop, hiss and crackle of reality, all of which is data, some in the loose sense we call noise, or perhaps just the uncontrollable, chaotic part of creation (to get poetic), it's shooting from the techno-geek end of the art spectrum, but still can be art of course, and may or may not be interesting to look at once you frame it. Okay, back to pixel peeping...
: )
Posted by: John Krumm | Thursday, 10 September 2015 at 12:11 PM
I do exclusively landscapes and bought the 5DsR for that purpose even though I had a perfectly acceptable Nikon D800 (which I will continue to use, even if the controls are contrary-wise different). I am impressed with the 5DsR. When following through with proper techniques, the 5D results are astounding. I've only printed at 20X30 but everyone stops and looks at these prints. That is, I confess, the reaction I'm looking for. And I only give away prints, not sell them. Fun fun in my old age!
Posted by: Malcolm Leader | Thursday, 10 September 2015 at 06:57 PM
A few years ago, an article appeared in the journal Nature, entitled:
Multiscale gigapixel photography
The camera consisted of of 98 14-megapixel sensors, arrayed around a common objective, and each with its own lens to collect its part of the image. I was particularly struct by the last line of the abstract:
"Ubiquitous gigapixel cameras may transform the central challenge of photography from the question of where to point the camera to that of how to mine the data." I think that we may be there already!
The abstract is available online at :
http://www.nature.com/nature/journal/v486/n7403/full/nature11150.html
David
Posted by: David | Thursday, 10 September 2015 at 11:52 PM
@Phil Service:
You ask the question "How many megapixels do we need to make an "optimum" print that is 18 x 12 inches?" and reason your way mathematically to the following result: "...approximately 112 MP total "resolution".
By applying that same reasoning, we must then conclude that nearly no-one, ever, has produced an "optimum" 12x18 inch print from a digital camera - the only exception being the few who stitch really huge image files.
I'm sorry, but I don't really get that :-)
Hanging on my wall is a 12x12 inch portrait that stands up to as much scrutiny as anyone would ever subject it to in a "real world setting". You can move in as close as you want. I even had one professional photographer ask, if it was made using a Hasselblad. It was shot using a Nikon D70 and a very, very sharp Micro-Nikkor lens, then cropped to a square, leaving a total of 4 megapixels.
I honestly believe that we have reached the point of sufficiency a while ago...
Posted by: Soeren Engelbrecht | Friday, 11 September 2015 at 02:35 AM
Dear Soeren,
You are talking apples and oranges. "Sufficient" is not the same as "optimum" any more than "good enough" is the same as "best." Please read my column, linked to under Mike's posting.
35mm film was quite sufficient for the needs of most pros. That didn't mean that medium and large format weren't observably better. And, in that vein, it was essentially impossible to produce a visually perfect 8x10 from 35mm film (that's objectively and experimentally provable fact). That didn't prevent it from being "sufficient."
Most pros would have been happy to get that boost in quality if it hadn't exacted huge penalties in size, cost and convenience. Digital vastly reduces those penalties.
pax / Ctein
Posted by: Ctein | Friday, 11 September 2015 at 12:04 PM