Slashdot Mirror
111-Megapixel CCD Chip Ships
georgewilliamherbert writes "EETimes is reporting that Dalsa has shipped a record-breaking 111-megapixel CCD image sensor to customer Semiconductor Technology Associates. The chip was paid for by a U.S. Navy SBIR project. At four inches across, a bit big for camera phones, but the 10560x10560 format will probably get professional digital camera users drooling."
I'd doubt many professional photographers are drooling over this. The market, at least in terms of commercial photograpgy, is about at its limit of need, in terms of the 32+ megapixel cameras. Manufacturers are now pushing the envelope for satelite and other advanced imaging. In most commercial applications, the current state of the art in terms of cameras combined with transfer and storage requirements is more than sufficient.
The problems that prevent digital sensors from blowing away film are that pixel densities that approach film resolution are too noisy, and digital sensors don't have the ability to handle as wide a range of light intensities as film does.
Some estimates put it at 300-500 megapixels, but it's really relative; the brain doesn't process all the eye sees.
'For we walk by faith, not by sight.' II Corinthians 5:7
According to the quick google search I just did, somewhere in the neighborhood of 576 megapixels.
. html
http://clarkvision.com/imagedetail/eye-resolution
I don't know how reliable these data are, though. There seems to be considerable hand waving between what the eye records and what the brain "sees" in that link.
I found this page interesting. Here's a quote:
Digital X-Rays involve several orders of magnitude less radiation exposure than film X-Rays. That, and the instant development allowing you to know right away if you need to take another shot, are what make digital X-Rays worthwhile. The resolution is more than adequate for either digital or film X-Rays.
[bquote]It reminds me of a story I saw (on PBS or Discovery Channel) about modern medicine in developing countries. People will pay extra for a "digital X-Ray", even though the cheap equipment produces a digital image that has far less resolution than a plain old film X-Ray. But it's "digital", so it must be better.[/bquote]The advantage of "digital x-ray" is that you don't have all those wonderful film processing chemicals around, the results are near instant, and it requires less radiation compared to traditional film x-rays, and convienence. The hospital near my house is 100% digital. As soon as the image is taken it is uploaded to a server where both the radiologist and doctor can look at it, whether they are at the hospital, at the doctor's office next door, the hospital across town, or half way around the world if need be.
The best part about this announcement isn't the 100 megapixel size. Photographers can already buy large format digital backs for view cameras with 300 megapixel resolution (albeit for a hefty price). But they use multiple CCDs and require external power supplies and HDDs. This new chip opens up intruiging possibilities for a self-contained high resolution camera that requires much less power to operate. Still, a CCD of that resolution will generate raw image files of about 350 megabytes each, so portability will necessarily be compromised to a degree by storage requirements.
Thanks for trying, but you're not really saying anything. A 500 megapixel image printed on a 10 kilometer by 10 kilometer screen and viewed at a distance of 1 meter will be easily distinguishable by the human eye. As another poster pointed out, it's actually not area that matters but angles, or if you use area you should consider the viewing distance also.
The link to the SBIR page appears to be defunct due to bookmarking data called from a session. I wasn't about to ask the submitter to give me his cookie and I tried finding info about the Dalsa project on the SBIR site, but wasn't having any luck, so here's a press release from the company that built it.
It sounds like the interest for the navy is along the lines of astro-navigation, but I'm not really sure. It's definitely not something general photographers need or even want. It's kind of pointless if your lenses aren't comparably impressive, or if you're not printing it out at a couple feet in size and to be displayed in a way that someone would get close enough to appreciate the quality. Plus once you take all that data, then you have to store it. I'm not sure how RAW images are stored, but if my math serves, a 24 bit BMP at that size would take about 300 MB per image.
Actually, slide film has higher dynamic range than print film (in addition to having superior grain and vastly superior tonal range), however slide film is much more sensitive to exposure (you can be off by a full stop or two with print film and most people won't even know). A result of this is that many more modern cameras which are designed for print film don't meter for crap when using slide film, while my 1982 Contax RTS II meters beautifully with Kodak E100 or Provia 100F. In fact, I've given up bracketing for all but critical shots because 95% or more of the time the default metering gives a beautiful exposure.
That thing must cost an arm and a leg. The failure rate of chips goes up exponentially with size and at 4 inches across yields must be next to nothing.
--aiee
Some sites have great explainations and demos.
This has nothing to do with image sensors, but does have some bearing on "what can eyes really see".
This issue is a bit more complicated than you think.
This is because of the difference in how high ISO speeds work in digital vs film. High-ISO film is more sensitive to light because the photosensitive grains are larger -- the digital equivalent would be bigger pixel sensors. Digital cameras implement high-ISO mode by increasing the amplification on the pixel sensors, which makes them more sensitive to light, but also more sensitive to noise. If you were to average adjacent pixels in your digital image, you'd have the effect of high-ISO film: less noise, but lower effective resolution.
"They redundantly repeated themselves over and over again incessantly without end ad infinitum" -- ibid.
The eye has around a hundred million nerve inputs, so the per frame resolution can't be higher than that. However, the real resolution of that is actually considerably lower (the signals are essentially downsampled on their way into the brain).
A high speed head mounted display (sufficiently close to the eyes) with only 2-3 megapixels would probably be sufficient to completely satisfy your eyes.
http://health.howstuffworks.com/eye2.htm
"Who is the Journal of Quantum Physics going to believe?" --Stephen Hawking
The fine article appears slashdotted, so I don't know if they cover this. The application which leaps to my mind for this detector is astronomy. Astronomers will pay big money for a better detector - I've seen a US$200k chip (2k x 2k pixels in about 1990, for use in the Sloan Digital Sky Survey camera.) Even at these prices, it is cheaper to get the same quality upgrade by improving your detector than by building a bigger telescope.
Astronomers run their CCDs at liquid nitrogen temperatures (to reduce thermal noise), and for UV astronomy they use "thinned" chips (etch/grind away the back of the chip so you can illuminate it from that side - otherwise too many photons are lost before reaching the light sensitive volume.) I'm not sure what other features astronomical CCDs require which might not be present in this chip. Pixel size shouldn't matter too much (except in its effect on noise) as you can design your camera to scale the image to suit the detector.
Quattuor res in hoc mundo sanctae sunt: libri, liberi, libertas et liberalitas.
It's not a noob question, but it does try to liken things that are not alike. Unfortunately, the human eye and cameras are different beasts that tend to frustrate nearly every attempt at comparison. This is in large part due to the fact that when most people say "the human eye" they actually mean the "eye-brain system," which is far more complicated than just the eye, which is itself already complex enough to do plenty of the frustratin'.
In any case, the issue with throwing the brain into the mix is that it does a lot of "post-processing" on the images that stream in from the eye and give us a mental picture much different from what the eye itself is actually able to pick up. Also, the eye has different kinds of vision--in the center of the field of view, in a very narrow range in fact, we see with acuity. Outside that very narrow range, our brain fills in a lot of the details that we think we see from moment to moment, but is actually not being "seen" in the same sense as what's in the center of view. (Of course, this comment will inevitably beget the philosophical discussion: what does it mean to "see," exactly?) If you doubt that your eyes only see with acuity in a fairly tight circle around the direct center of your field of vision, try this experiment: pick up a book, open it to a random page, and fixate your eyes on a word somewhere in the center. Now, see how many words you can read around that word without moving your eyes to look directly at those words. The words you can make out fall in your acute vision field. (You'll find that if you move the book farther away, you can read more words because they fall within the same angle--this works up until it gets so far away the overall level of acuity you enjoy isn't high enough to make out any of the words at all.) The rest of your field of view is in your non-central field (I'm callng it). Your peripheral vision is comprised of the part of your field of view for which your brain does not bother filling in any detail--you're only vaguely aware of it in the visual sense provided it's not moving.
What our non-central vision lacks in acuity it makes up for in motion detection. That's why hunters often say when you first spot prey in the distance that's fairly well camoflauged with its surroundings as it moves about, don't look directly at it, but look slightly to the side. That way, when it starts moving again you'll see it and you can put it in center vision again, but once it stops, look off to the side again. Stargazers often use this trick as well--if you look directly at a faint star, after a couple of seconds you'll question whether it's actually where it was just a moment before. But if you look slightly off to the side, your eyeball moves around and twitches enough that it creates apparent "motion" of the faint star you're trying to see and you can pick it up again. (Incidentally--this is the reason why our eyes in are constant motion...if you've ever tried to make your eyes exactly still you know how difficult it is to keep from twitching them constantly. It's because our brain requires that motion to keep the motion detecting parts of your eyeballs feeding the detail your visual cortex craves. You'll also find that if you are able to keep your eyes at all from twitching for an extended period, 10 or 15 seconds, you'll find that the level of detail in your non-central vision starts to fall off, sometimes even fading to black...this isn't very noticable until you start twitching again and suddenly see color and detail spring back.)
Anyway, the point is, no matter what one says about the eye in relation to a camera, someone will be bound to argue (and, in some sense, almost certainly be right). It's kind of a useless endeavor to try to get a megapixel rating for the eye, or figure out what it's dynamic range is, etc. A more fair comparison would be hooking a camera up to a computer, then periodically having the camera move slightly and snap a shot, then the computer takes it and stitches it into a composite of the entire scene comprised of s
but have you considered the following argument: shut up.
This is an intresting development and one that will have future implication on imaging in general. A 4x5 neg or tran as a great scan is around 45-50MP so this CCD would exceed film res and possibly optical res as we define it today but the more intresting question is how this new res effects output technologies(print). I posted a podcast on this very subject some days ago....if you are interested just goto itunes/ podcasts and seach for hybrid arts and listen to "future technologies". Cheers!
It's interesting that this came up, since last week I was reading an article on the resolving power of the human eye as it pertains to photography and how to choose output resolutions. Short answer: 50 lines per degree of view. From there you can do some right-triangle trig to figure out how many line pairs should be perceivable for some output format based on how close you're going to be to it. For an 8x10 image, the author says 2300 pixels in the long dimension or 230 PPI would cut it (I didn't double-check his math). I tend to wonder if you don't have to introduce a factor of two in there somewhere, since to reproduce a "line" of resolution seems like it ought to require two pixels.
Of course, that's an oversimplification; hence the long answer. The human eye doesn't have a fixed number of "megapixels" that you could easily convert to a measurement of a photo or really even of another camera. First, you have the problem that the eye's "resolution" isn't evenly spread across the field of view: it's concentrated near the center, and thinner out in the periphery. This is why if you concentrate and try to pay attention to something that's not in the center of your field of view (that you're not looking directly at) it won't be as clear as when you look directly at it. (The exception is in very low light: your indirect vision is better at night vision.) However your brain reassembles the image and makes you think that you're seeing one great-big full-res panorama, when in reality at any one time you're only seeing a small part in "full rez" with the rest of your field of vision at something less, but with the full version available on-demand (by looking at it).
If you could actually do a 'screen grab' of the image your eyes were actually feeding into your brain, at any particular time, I think it would be a lot lower-quality than many people suspect. Almost without question, it would be lower quality than many photographs of the same scene. The depth of field is short, the resolution is concentrated in the center, as is the color, and there's a hole in the dead center of the image because of your optic nerve's placement on your retina. Your sense of sight works as well as it does, in large part, because of all the caching and postprocessing that's done transparently by your brain to the incoming information stream.
Really, when we compare a photo to our "sight," what we're really comparing is the photo to our brain's recollection of how it saw a particular scene, which might be very different from what our eyes actually took in, and further still from the 'objective truth' (if you believe in such a thing, that is) of what actually was there at that moment. The easiest example is color saturation: we tend to see and remember things as being far more colorful than they actually are: an "accurate" photo will therefore look dull compared to memory, so we compensate by oversaturating our photos to make them look more 'realistic.'
It's only possible to make comparisons between our eyes and mechanical cameras, and between our overall sense of sight and recording systems, for very limited cases. Even to answer a relatively simple question like "what's the eye's maximum megapixels?" completely would probably stretch the boundaries of currently understood optometry, neuroscience, and psychology.
"Ladies and gentlemen, my killbot features Lotus Notes and a machine gun. It is the finest available."
The quality of a digital X-ray is as good as the old ones. You won't fail to make any diagnoses because of the changes. The advantages, however, are:
1. Cost - much lower
2. Radiation - much lower
3. Image manipulation - increases diagnostic yield in a variety of ways
4. Transmission - to other specialists, near instantaneously (depending on connection speed - usual rate-limiting factor is getting someone in front of the receiving screen to interpret the images)
5. Can't lose them (not quite true, but easier to back up)
6. Near instant results (check that the film doesn't need repeating after developing)
7. Can be viewed by multiple people at the same time (invaluable - no more x-ray on ward, can't be reported by radiologist)
So really, there's no question. Digital all the way. I am a doctor (emergency, so I look at probably around 100 x-ray images or more per day), having worked with both old and news flavours of x-rays - and there are no actual advantages with the higher resolution of the older type.
This idea was invented by Shampoo.
PPI != DPI
You're printer quotes 1200 (or 2400)dpi as Photo quality, however the number of dots does not equal the number of pixels
Each pixel could take up many "dots".
e.g. ypu want to print a 10" x 8" image, you print-shop asks for a 300ppi image minimum, which would mean it would have to be a 3000x2400 pixel image
more here
http://www.rideau-info.com/photos/mythdpi.html
Or at least so I hear...somebody over at Luminous Landscapes ran a comparison of a PhaseOne P45 39-megapixel back against drum-scanned 4x5 Velvia 50. These are guys whose standard print size is 30"x40", so fine detail is pretty crucial to them, as is color accuracy. Bottom line? The film had a slight edge, but not enough to offset the huge increase in convenience and versatility of digital. Granted, the P45 alone lists for $32,990 at Calumet, plus another $6-10,000 or more for the camera and lenses, but apparently over the 3-year warranty period it works out to ~80% the cost of a view camera, lens, film, lab fees, and drum scanner maintenance.
:)
I know what I want for Christmas this year
Facts do not cease to exist because they are ignored. - Aldous Huxley