Has Foveon future?

Discussion in 'Digital Cameras' started by ThomasH, Oct 17, 2003.

  1. ThomasH

    ThomasH Guest

    The time passes and seemingly with the exception of Sigma,
    nobody is really eager to replace the color pattern sensors
    with the Foveon chip.

    Sony has even clearly indicated its decision to stick with
    the color pattern sensors by releasing its original 4-color
    chip with RGBE pattern (E=Emerald, similar to desaturated
    cyan) promising to cover a wider color gamut.

    Nikon's latest chip for the D2H is also a color pattern
    cheap, so is the chip in the latest Olympus 3/4 E1 DSLR.

    The concept of Foveon appeared sound, what went wrong?

    ThomasH, Oct 17, 2003
    1. Advertisements

  2. ThomasH

    JPS Guest

    In message <>,
    The camera only uses Sigma lenses, which means no image stabilization is
    possible, and less quality lenses to chose from.

    The output is noisier than anticipated in theory.

    The camera has no anti-aliasing filter, and the files look a bit too
    sharp out of the camera.

    The benefits of having 3 color samples at each cell site are highly
    JPS, Oct 18, 2003
    1. Advertisements

  3. It wasn't sound; the human eye's luminance resolution is about 3 times it's
    chrominance resolution. So Bayers sensors provide nearly the exactly right
    ratio of luminance to chrominance information. Foveon provides information
    you can't see. Not particularly useful.

    They managed to fool a few people by leaving off the antialiasing filter, so
    Sigma sold a few cameras to folks who flunked, or never took, an intro to
    digital signal processing course.

    David J. Littleboy
    Tokyo, Japan
    David J. Littleboy, Oct 18, 2003
  4. ThomasH

    Tony Spadaro Guest

    Basically the pictures are not impressing anyone, and without impressive
    pictures it's real hard to convince anyone to buy a whole new camera
    system - especially Sigma.
    Tony Spadaro, Oct 18, 2003
  5. Actually, the pictures _are_ impressing people. Most people don't notice the
    gross aliasing on near horizontal and near vertical lines, so the images
    appear very sharp out of the camera. Looking at the images, my guess is that
    the sensel area is quite small relative to the size of the pixel (the pixel
    spacing), and combined with the lack of an antialiasing filter, that gives
    extremely sharp extremely high contrast images, that, per pixel, look a lot
    sharper than the mainstream dSLR images. If you don't mind all your detail
    being in the wrong place by up to 1/2 the pixel spacing, it's quite amazing.

    David J. Littleboy
    Tokyo, Japan
    David J. Littleboy, Oct 18, 2003
  6. ThomasH

    Matti Vuori Guest

    (George Preddy) wrote in
    How can a _sensor_ be too low res for its specs? More likely the lenses
    just can't cut it.
    It is not necessarily a "strategy". Creatind a DSLR is a tremendous effort
    and every manufacturer has to start with one model. Do a count: how many
    removable lens DSLR's does Olympus have now? And estimate how many they
    will have in five years?
    Matti Vuori, Oct 18, 2003
  7. ThomasH

    Chris Brown Guest

    In addition to this, I'm given to understand that the actual frequency
    seperation of each sample is less than ideal for creating RGB output,
    leading to questionable colour accuracy and performance in some situations.
    I am given to understand that this isn't so bad provided you are willing to
    trade off a bit of dynamic range at the top-end and expose slightly less
    than you might otherwise want to. The upshot is that, although the Sigma SLR
    produces impressive resolution for a 3 megapixel camera (albeit at the
    expense of some aliasing), the colour accuracy can leave something to be
    desired. YMMV.
    Chris Brown, Oct 18, 2003
  8. SNIP
    Oh boy, here we go again.
    The camera is only 2268x1512=3429216 pixels or 3.3MP.

    The rest of your comments sound like you're trying to justify your purchase,
    so I'll leave them for what they are, ill-informed marketing blah-blah at

    Sorry to disappoint you, but I suggest reading some of the earlier threads
    in this group. You might learn enough to avoid basing you purchase decision
    on a folder next time.

    Bart van der Wolf, Oct 18, 2003
  9. ThomasH

    Todd Walker Guest

    Aren't worth Joe Schmoe's time? So I take it you are suggesting that
    only the pros who know what they are doing are buying the SD9 huh?
    Interesting. So where are all these pros with their SD9s anyway? Most of
    the portrait photogs, wedding photogs, press photogs, etc. that I know
    of are using Canon and Nikon. Show me some that use Sigmas.

    And 10.3 mp? You have swallowed Sigmas bullshit marketing hook, line,
    and sinker haven't you? It is a 3.3 megapixel camera -- that's it. Try
    printing an 11x14 on an SD9 and a 10D or D100 then tell me that the
    Sigma is 10.3 mp.
    You're gonna have to explain that statement because it makes absolutely
    no sense -- "too low res for its specs?" What are you talking about? Is
    that out of the Sigma marketing handbook too?
    The 14n is a pathetic piece of crap unless you shoot in a studio at ISO
    Wow, now you are really reaching. Please tell me about a high end
    prosumer point and shoot that has better image quality than a 10D, D100,
    or S2. And the fact that you put the image quality of the SD9 in the
    same class as the 1Ds just shows that either you work for Sigma (or
    Foveon) or you are just trying to justify your purchase.
    If this really is true, why do they need to come out with their own
    digital body when they make lenses for every popular mount from other
    They sell plenty of lenses to Canon and Nikon owners who don't want to
    shell out the money for manufacturers' lenses.
    You make it sound like they are selling SD9s faster than they can make
    them, which doesn't seem to be the case.
    I don't think the big manufacturers are too concerned about them.

    Todd Walker
    Canon 10D:
    My Digital Photography Weblog:
    Todd Walker, Oct 18, 2003
  10. ThomasH

    John Bean Guest

    Cheap shot David, you disappoint me.
    John Bean, Oct 18, 2003
  11. ThomasH

    Mark Herring Guest

    Ditto all the comments about how you dont really need the supposed
    advantage of the perfectly registered color. ie the Bayer pattern
    works just fine. When I first started with 640x480 images---yes--the
    firnging was terrible. Wtih my 2MP Canon it's simply not an issue.

    One thing that lawys bothered me was that---since the sensor uses
    silicon properties to get the spectral separation---it would be hard
    to control the indivdual bandpasses.? This would maybe make it more
    difficult to get the right color balance.

    I think spatial performance is the "long pole" in sensor design. To
    get the best true resolution you need lots of pixels with minimal
    cross talk and the highest possible fill factor (ie minimum dead
    space) The Foveon archtiecture does not appear to be consistent with
    these goals.

    digital photos, more and better computers,
    and never enough time to do the projects.
    Private e-mail: Just say no to No
    Mark Herring, Oct 18, 2003
  12. ThomasH

    Tony Spadaro Guest

    In my experience - no one has seen any improvement in quality who is not
    directly involved in selling Sigma or Foveon.
    Tony Spadaro, Oct 18, 2003
  13. ThomasH

    Tony Spadaro Guest

    It' is a 3 megapixel camera -- this business of multiplying pixels based
    on colour information alone is nothing more than an advertising lie.
    Tony Spadaro, Oct 18, 2003
  14. (George Preddy) wrote in
    The Foveon sensor in SD9 is a 3.4 Mpixel sensor! It has
    3.4 M color sítes. So - it *is* a 3.4 Mpixel sensor. Everything
    else is just pure nonsens.

    A Bayer with 3.4 M "one color" sites is more difficult to rate.
    But - the optimal number of pixels for this camera is 3.4 M.
    If you use less you lose information and if you use more you
    are just upsampling. It is therefore reasonable to call this
    sensor also a 3.4 M sensor. Although I agree that it is not
    fair to a true RGB sensor.

    But - to be absolutely fair - the Foveon is not a true RGB
    sensor. It detects three rather broad spectrums and calculates
    the RGB values. This causes noise and also some errors in
    color rendition.
    Uh? What are you on? It is Foveon that is pulling your socks off.

    Roland Karlsson, Oct 18, 2003
  15. ThomasH

    Chris Brown Guest

    A somewhat emotive use of language, which betrays your prejudices.
    Antialiasing is not "bluring". The lack of an antialias filter is, perhaps,
    one of the SD9's bhiggest deficiencies, even though it does lead to images
    which look superficially sharp. Other SLR manufacturers, most notably Canon,
    produce images which are not oversharpened out of the camera, and are
    therefore able to withstand a lot of image processing. Such cameras are
    probably not suitable for those who wish to avoid processing their images.
    On the other hand, it does provide for significantly greater latitude in
    what can be achieved with some competence in Photoshop.
    That is a bizzare statement to make, given that the the 10D and D60 have
    higher resolution sensors that the 1Ds (i.e. the photosites are more densely
    Your comments here are bizarre - they simply fail to match the observed
    results that people are getting from the 1Ds, which quite probably provides
    the best image quality that can be obtained from any 35mm lens system. You
    do the credibility of your point of view no good by making such patently
    absurd statements.
    Most DSLRs produce images that need sharpening simply because sharpening is
    the last thing you should do when processing an image. The artificially
    sharp/aliased images that the SD9 produces are not an advantage, given that
    it's targeted at the same sort of market that the likes of the D100 and 10D
    are targeted at.

    Futhermore, given the well known improvement in the perceived sharpness of
    Canon DSLR images which can be gained from the application of a little unsharp
    masking, relying on unprocessed Canon images as the basis for any claims
    of superior image quality from the SD9 is not remotely convincing - it just
    demonstrates poor experimental method, and raises the question of just how
    confident you really are in your own claims.
    Chris Brown, Oct 18, 2003
  16. ThomasH

    Chris Brown Guest

    If you seriously think that the SD9 images are not the result of
    interpolation of the colour values returned by the sensor, then may I
    suggest you download a copy of Dave Coffin's dcraw utility, which is capable
    of outputting raw, uninterpolated images from any of the cameras it
    supports. If you genuinely believe that the values you are getting in your
    final images are the raw, uninterpolated "RGB" values from the sensor, then
    I suspect you're in for quite a shock when you see what an uninterpolated
    SD9 image really looks like.
    That's nice for them. Phil Askey's independent image testing, on the other
    hand, reveals that they compare favourably with *unprocessed* images from a
    Canon D60. A good showing for a 3 megapixel camera, albeit at the expense of
    aliasing, but far, far short of the exaggerated claimes you are making here.
    Chris Brown, Oct 18, 2003
  17. ThomasH

    Chris Brown Guest

    Indeed it is. Check out the comments from someone who actually wrote a raw
    convertor which handles the SD9 images on:


    He appears to have discovered that the colours aren't actually terribly well
    separated, and getting actual RGB information out of the Foveon sensor is
    nontrivial. Now Foveon is interesting technology, but the comments from the
    Sigma/Foveon faithful about how only Foveon captures "true colour", and so
    forth, in addition to being remarkably tedious aren't even remotely true -
    the colours that Foveon captures are quite a poor match for the desired
    result (i.e. RGB), and the resultant images are the product of significant
    amounts of colour interpolation of the raw sensor data, i.e. just what the
    Foveon Faithful deride Bayer sensors for.
    Chris Brown, Oct 18, 2003
  18. (George Preddy) wrote in
    Eh - where would a point and shoot with 6 Mpixel be on this chart?

    Roland Karlsson, Oct 18, 2003
  19. Whatever you call it, it's necessary. Leaving it out is not acceptable,
    whatever the underlying sensor.
    But "adding sharpness" is exactly what leaving out the antialiasing filter
    does. Digital imaging cannot accurately (or "infinitely sharp(ly)")
    represent the world, and if you ask it to represent the world better than it
    can, you get artifacts, called aliasing. You can't "blur after the fact"
    because there's no way to tell the difference between artifact and real
    data. (Aliasing consists of shifting frequencies that can't be represented
    to frequencies that can. Once you've shifted the frequency, it's an in-band
    signal that can't be differentiated from real in-band signals.)

    As I said, the SD-9 is for people who don't understand digital signal

    David J. Littleboy
    Tokyo, Japan
    David J. Littleboy, Oct 18, 2003
  20. (George Preddy) wrote in
    The SD9 is a good camera and the Foveon technology is interesting.

    But your mathematics is even more interesting.

    The 3.4 Mpixels Foveon is a 3.4 Mpixels sensor.
    Why should 3.4 Mpixels be 10 Mpixels?
    Don't compare it to Bayer. Forget about Bayer.
    Why should the 3.4 Mpixels Foveon then be a 10 Mpixels sensor?

    I understand that you have a problem with Bayer sensors.
    That a 10 Msite Bayer is called a 10 Mpixel sensor.
    I can understand that. It is very doubtful that this is correct.

    But forget about Bayer. Why should the 3.4 Mpixel Foveon then
    be called a 10 Mpixel sensor?

    Roland Karlsson, Oct 18, 2003
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.