[filmscanners] RE: Newish Digital Tech
>...but why would a PMT > get more > > light, > In semiconductor sensors, however, many, perhaps most, of the > photons that hit the junction do absolutely nothing, so they're much less > sensitive. Hi Paul, Id believe that PMTs have a much lower noise floor than CCDs and that is the reason for the much higher dynamic range, and obviously better shadow detail. Is that what you are talking about? As far CCD sensitivity...CCDs have a minimum number of photons before they can register, but I believe that once that level is reached, most of the photons that hit the sensing area are being counted (accumulated)...I don t believe they are doing nothing? Regards, Austin Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] RE: Newish Digital Tech
> From: Austin Franklin > > My second statement is based on images Ive seen like this: > > http://www.pbase.com/image/10390721 That image looks totally strange, unlike any of the other images in the set, so I have no idea why it looks as it does. However, this one: http://www.pbase.com/image/10906254 is exactly the sort of image that would be _much_ noisier on my DiMage 7. I have no experience with high-end digicams, though. -- Ciao, Paul D. DeRocco Paulmailto:[EMAIL PROTECTED] Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] RE: Newish Digital Tech
> From: Austin Franklin > > What about the sensor layers below the top one, and then below > that? There must be some decrease in transmission (not just due > to the top filter, but due to the sensor, material whatever...). > All the Bayer sensors measure %100 of the light that particular > sensor is filtered for, where the Foveon gets decreased light > to the inner two sensing areas. The principle is that the top layer only absorbs red light, converting it to an electrical signal, and passes the green and blue unattenuated. The second layer absorbs the green, converting it to electricity, an passes the blue, which is absorbed and converted by the third layer. > Given this (and a few other issues that relate, like accuracy of > sensing the frequency), I would speculate that the Bayer pattern > sensor would have higher color fidelity (with respect to accuracy > of color) than the Foveon. If you think about how the eye works, what's important in an RGB sensor is that each color filter have a sloped spectral response that overlaps the adjacent color filters, and that the peaks of each response line up with the peaks of the eye's response. This guarantees that any perfectly saturated color of light (e.g., laser light of any visible wavelength), will produce a different weighting of the three outputs. In the X3 chip, the skirts of the filters are more gradual than in conventional chips, but this is something that can be compensated for accurately using arithmetic after the fact. > Also, this issue about %100 of the light keeps being stated by most > everyone who champions the Foveon. Even if it were true, its > a matter of > significance. I do not believe it is a significant issue at all. People > can claim it is, but no one has yet to make any sense as to why, or shoe > evidence of it being so. The low light performance of the Foveon is not > very good, in fact. I believe if it is an issue at all, it is a > VERY minor issue. All other things being equal (which they're not, yet), it could prove to be a major advantage. As I said in another post, my $350 2MP Digital Elph has less noise than the 5MP DiMage 7 I paid $1300 for. The full frame high-end digicams have lower noise still, all because the pixel sensors are bigger. -- Ciao, Paul D. DeRocco Paulmailto:[EMAIL PROTECTED] Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] RE: Newish Digital Tech
>-Original Message- >From: [EMAIL PROTECTED] >> >Hum. Do you have a source for that? I dont believe that is >> >true, and will >> >have to think about your assertion. What is the source of the random >> >variations? I know there is some randomness in reception of photons, >> >simply because of atmospheric dispersion, and other causes... >> >> I think he is talking about the quantum noise which is introduced >> due to the >> Poisson probability distribution of the Photons hitting the detector. The >> resulting error in the signal is proportional to sqrt(2). > >Hello Robert, > >If he is, thats called shot noise and its = sqrt(S), where S is the >signal in electrons. It certainly is the noise that limits CCD >performance. Yep, shot noise = photon noise. And yes it's sqrt(S). So when the signal is twice as big then the noise only increases by sqrt(2) which improves the SNR. Robert Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] RE: Newish Digital Tech
> From: David J. Littleboy > > Exactly. All Foveon does is hide the artifacts so you can't recognize them > when you look at the image. In the res test charts, it "resolves" a 9-band > image as 9 bands for some frequencies, 7 bands for other frequencies, 5 > bands for other frequencies; all with the same contrast. This is seriously > unacceptable. Without an anti-aliasing filter, it's not a camera, it's a > random data generator. Oh, come on. No digicam meets those standards. The only way to eliminate aliasing in a pixelated sensor is to put a diffuser over it, and I don't think anyone does that. There's certainly no intrinsic reason why such a thing would be easier to do on a Bayer chip than on the Foveon. One of the advantages of the Foveon, as I already stated, is that the aliasing matches for all three colors, so you don't get colored moire from a monochome texture. > So? Most serious photographers take RAW images and postprocess on their > PC/Mac. Several seconds per image is no big deal. Also, special-purpose > hardware is always orders of magnitude faster than > general-purpose hardware. I shoot raw, too, but only because it's the only way to get all 12-bits of A/D resolution out of the camera. I'd much rather shoot JPEG2000 files, because I wouldn't have to wait ten seconds for each file to be written to the Microdrive, and I'd be able to fit several hundred more files on it. > Again, all the cameras use special-purpose hardware. Hardware is cheap and > fast. The processing time issue is completely bogus. Are you saying that the typical digicam has more DSP horsepower than my 1.7GHz Pentium 4? I suppose it's possible that all digicams have some really clever custom parallel processing chip, but I doubt it. -- Ciao, Paul D. DeRocco Paulmailto:[EMAIL PROTECTED] Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] RE: Newish Digital Tech
> From: Austin Franklin > > What ever the number is, it's a "so what" though...red herring as it were. > The Bayer pattern sensor is only designed to sense the light of a > particular > color at a certain spatial point. Images from Bayer pattern > imaging sensors > have not shown to be inferior to those of scanning cameras, or the Foveon, > with respect to same sensor area comparisons. Typically, scanning cameras > will be better as they have a much higher resolution. The point is that the more photons you capture, the less noise you get. Assuming the Foveon wastes the same percentage of its area on interconnect as a Bayer pattern chip, the Bayer chip will filter out all the "wrong" color photons that hit each sensor, while the Foveon will actually use them. Obviously, implementation matters. But I see no reason to expect that a Bayer pattern chip is intrinsically better--it's just been worked on longer. -- Ciao, Paul D. DeRocco Paulmailto:[EMAIL PROTECTED] Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] Re: Newish Digital Tech
David J. Littleboy wrote: > This is all interesting, and probably quite accurate, but it isn't what > my eyes tell me. I've downloaded several full size captures taken from > the Foveon chip (via the Sigma camera) and I find the images from the X3 > superior to similar images on a Canon product using a CCD and Bayer. > > > Yes, they _appear_ to look better, but you don't know what is real data and > what is artifacting. You get a twig pattern where the D60 only showed fuzz, > but the twig pattern won't be the actual twig pattern in the scene. The > artifacts look just as good as real data. I just don't think a system with > that level of artifacts is going to be acceptable. With similar pixel counts > and an anti-aliasing filter, I suspect the Foveon sensor might be slightly > better, but then the question of color rendering will arise. Is the Foveon > trick really good enough to separate out the colors as well as the on-chip > color filters? > > I guess no one should use Genuine Fractals, bicubic interpolation, or any other interpolation format, unsharp masking, or any other sharpening algorithm either then, because they all corrupt the data accuracy. What's important to me, is not if on a microscopic scale a data point is in the correct location or not, but if the final image looks clearer, and more accurate in terms of color and contrast, with less obvious artifacting. You know what? I also don't care is the grain pattern in film is accurately laid down as long as the image looks like what I photographed, or what I wished to accomplish with that film. Again, I'm speaking about making digital color image capture, not lab test results. If someone says to me, well "it may look like... but..." I know the issue may be one of practical use versus looking at the "arcane", to use a term that has been showing up a lot the last few days. It looks to me that Bayer images are better than > scanned images on a per-pixel basis. Even the sharpest Provia 4000dpi scans > printed at 300 dpi are a lot softer than D60 images printed at 240dpi. It > seems to me that the main problems with Bayer cameras are sensor size and > pixel counts. > > What problems bother you? Color fringing, smearing, diagonal convergence problems. Cost per color resolution. > A 9MP full-frame US$2,000 camera would put 35mm film out of > business for all practical purposes. I'll retire my 645, pick up a GSW690 for landscapes, and do everything else with the 9MP dSLR... Fine, where do I get it? The Foveon is here now, and available, and within a year chances are much cheaper and smaller version will be available. Are there 9 MP color CCD/CMOS SLR cameras for under $2000 now without a long lag time between when I hit the shutter and when the image is captured? The closest I see is the Canon 1DS at $8000. It claims to have a 55ms lag time equivalent to film, but most less expensive cameras are considerably slower. > > The images I looked at were > not from a Foveon site where one might expect some bias to slant the > image content toward that chip's best case scenario, but in reviews and > commentary of the Sigma camera. > < > > They look nice and sharp, but there are color rendition and blown highlight > problems with the SD9. This isn't fair to the X3, since the SD9 is such a > dog. > Well, if anything that shows that the X3 is even a better chip than the current technology is showing it as. > > Perhaps, each technology will end up having its niche market, or perhaps > something will eclipse both, but if I had the choice, right now, I'd be > much more tempted toward the Foveon technology for digital images > capture. Further, the X3 is literally the first consumer generation > of that chips technology while CCD has had many years to mature. > < > > The bit about it being the first generation is its largest problem. The SD9 > is a really badly implemented digital camera: compared to how well the Sony > F717 is done, and how well the Canon and Nikon dSLRs are done, it's a > pitiful joke. > For some reason, and my guess would be internal industrial pressure by some of the CCD/CMOS manufacturers, these companies did not pick up on using the X3 chip. Thank goodness Sigma had nothing to lose because they were not a manufacturer of CCD/CMOS digital cameras. By making this camera, warts and all, available to the market, they have forced a much more serious look at the X3 technology by other manufacturers. > > Unless > there is a known limitation within the technology of the X3 chip > technology, I'd expect evolutionary improvements as has occurred with > the CCD chips. > <<< > > My reaction to the hype has been excessively negative, but until they come > up with a decently implemented camera with an anti-aliasing filter, I'm not > hopping on the bandwagon. > > Skepticism can be a good thing. I like it when a smaller company challenges the current state of the market and
[filmscanners] RE: Newish Digital Tech
Karl, > That's not how the Foveon chip works. There are no filters. They > are taking > advantage of the fact that different light frequencies have > different depth > penetrations into silicon. Well, yes and no...but anyway, filtering HAS to take place, or you could not distinguish between RGB. > Essentially there is not going to be as clean a > differentiation between the amount of light at the R,G and B > sites, What EXACTLY is the sensing mechanism? Do you know, and if so, can you describe it? > and they > are relying on subractive calculation to compute the R and G > values. Speculation, or do you have a resource for this information? Regards, Austin Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
[filmscanners] Re: Newish Digital Tech
This is all interesting, and probably quite accurate, but it isn't what my eyes tell me. I've downloaded several full size captures taken from the Foveon chip (via the Sigma camera) and I find the images from the X3 superior to similar images on a Canon product using a CCD and Bayer. From a practical standpoint, what I noticed is that most of the inherent problems I have observed with CCD based digital camera captures, have been eliminated with the X3. The images I looked at were not from a Foveon site where one might expect some bias to slant the image content toward that chip's best case scenario, but in reviews and commentary of the Sigma camera. Perhaps, each technology will end up having its niche market, or perhaps something will eclipse both, but if I had the choice, right now, I'd be much more tempted toward the Foveon technology for digital images capture. Further, the X3 is literally the first consumer generation of that chips technology while CCD has had many years to mature. Unless there is a known limitation within the technology of the X3 chip technology, I'd expect evolutionary improvements as has occurred with the CCD chips. To me, the fact that the X3 chip leapfrogs so many of the original problems in the early CCD technology, gives it great promise. It should also lead to cheaper cameras since less is required in terms of calculating electronics within the camera, since the raw image is pretty much a finished product. Art David J. Littleboy wrote: > "Arthur Entlich" <[EMAIL PROTECTED]> wrote: > > >>From the sample images I have seen on the web, the Foveon X3 chip is a > tremendous improvement over the artifact ridden and slow to capture > bayer pattern CCD for digital camera capture. > > > That's exactly the oppposite of my conclusion. The X3 in the SD9 is used > without an anti-aliasing filter, and a quick glance at the resolution charts > shows that while it resolves nicely up to about 1000 lph, above that its a > mass of aliasing artifacts. It shows a lovely strong response to a 1550 lph > pattern, but it gives exactly the same response at higher frequencies as > well, and fails to resolve patterns between 1100 and 1550 lph. I doubt that > people will find it acceptable for serious work since you simply can't trust > any detail it reports. > > > The X3 chip has numerous advantages in digital image capture use. Since > no interpolation is required, the capture is very fast. > > > There's no speed problem that I've ever heard about with Bayer sensors. > Speed problems in digital cameras are always due to handling the large files > after capture. If anything, X3 will be worse, since raw files will be three > times larger than Bayer raw files. > > >> Changing >> > resolution allows for direct translation of the image at lower res since > each pixel is complete for all colors, and artifacting is pretty much > eliminated. > << > > No one is interested at shooting at 1/2 resolution. > > >> Further color accuracy is superior. >> > << > > Color rendition is problematic in the SD9... > > > If Sony and Kodak allow it to happen, this technology > can literally alter the nature of digital capture. > << > > It simply can't provide significantly better resolution than Bayer. To > eliminate aliasing, you have to use an anti-aliasing filter. That reduces > resolution to 70% or so of Nyquist. Maybe the X3 could get away with a > slightly less aggressive AA filter than Bayer sensors, but the difference in > practical resolution is going to be very small. > > The bottom line is that the limits on resolution that Bayer imposes is about > the same order as the limits on resolution aliasing imposes, so X3 will only > be superior for certain extreme cases and, for example, B&W imaging using > only the red channel. > > David J. Littleboy > [EMAIL PROTECTED] > Tokyo, Japan > Unsubscribe by mail to [EMAIL PROTECTED], with 'unsubscribe filmscanners' or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or body
