purple fringe - due solely to Bayer filter?

Discussion in 'Digital Photography' started by digiboy, May 31, 2004.

  1. digiboy

    E. Magnuson Guest

    It may or may not be depending on the quality of the demosiacing
    algorithm and what you think is "correct" (perfect point sampler?)

    And your theory does not explain why it would be purple.

    Actually, the SD9 doesn't while the SD10 does.
    (and both suffer from lens-related fringing)
    See the examples I posted.
    One theory (the one I favor)

    There are two factors: CA and blooming. However, the CA is not the
    normal red/green CA, but abberations of near IR or UV. How severe
    this appears depends on two things: lens abberations and the near-UV
    or IR sensitivity of the sensor.

    On set of images that led me to this theory was some side-by-side
    photos of the SD9 and SD10 both with the same Sigma 18-50 DC
    lens. (Alas these are no longer online). Both had a dark edge against
    a bright sky, with a distinct purple fringe but the SD10's was smaller
    and darker. One of the differences between the SD9 and SD10 is that
    the "dust filter" of the SD10 appears to to more non-visible light
    filtering (from experiments of people who have removed theirs.)

    Here is one explanation:
    E. Magnuson, Jun 1, 2004
    1. Advertisements

  2. digiboy

    E. Magnuson Guest

    You are going to have to be more specific.
    The *in-focus* high contrast areas look fine. The out-of-focus
    areas look a little unpleasant, but that's a lens effect.
    E. Magnuson, Jun 1, 2004
    1. Advertisements

  3. digiboy

    E. Magnuson Guest

    How many of your films have you examined at the same level of
    magnification? (E.g. 12-16" away from a 30x20" elargement, if you are
    using typical monitor [~100dpi] viewing a 3000x2000 pixel image 100%.)
    E. Magnuson, Jun 1, 2004
  4. Actually the Foveon works better with longitudinal chromatic aberration
    (the tendency of non-apochromats to focus red light further into the
    film emulsion than blue). Since the top layer sensor is blue, middle
    green and furthest in is red, the Foveon matches the most likely
    wavelength related focus errors in a well-corrected achromat or modest
    apochromat in the centre of the field.

    David Kilpatrick, Jun 1, 2004
  5. digiboy

    E. Magnuson Guest

    Another great pseudo-scientific theory why X3 might be better.
    Let's see if it can withstand some back of the envelope calculations.
    According to some Foveon publications, the middle absorbtion depths
    of light in silicon are:

    * blue 0.11um
    * green 0.9um
    * red 1.2um

    So for there to be *any* practical advantage, it would have to be in
    the 0.3um depth difference between green and red. Now the pixel pitch
    of the SD9 sensor is a whopping 9.8um. So even with the completely
    ridiculous assumption that 0.3um of depth means 0.3mm larger defocus,
    it's all smaller than a a single pixel! *You cannot possibly see this
    effect.* Even if you get to the smallest CCD, the pixel pitch is still
    2.2um (or 7x bigger than any possible effect.) And this ignores the
    issue silicon is not a perfect filter and that the "red" layer of the
    X3 picks up a significant amount of green and some blue as well (and
    green picks up significant amounts of red light and so on.)
    E. Magnuson, Jun 2, 2004
  6. digiboy

    Searching_ut Guest

    To begin with, my apologies for the snide remark. I come to the newsgroups
    to try and learn, but find it difficult to do in this newsgroup because of
    SG, who makes any worthwhile discussions impossible if he can find a way to
    interject. To channel frustration your direction simply because you
    mentioned the "S"swear word was wrong. The bad part is that in checking up
    on that individual from people who know of him in person, I find he probably
    believes the stuff he says, and is in fact trying to share his "Infinite"
    wisdom. I liked it better when I just thought he was immature and bored.

    As for film, the fact that you never experienced CA is interesting. It could
    be you just used better equipment than I had, or maybe it's different
    methodology. I had a vivitar 70-210 lens that would almost always suffer CA
    in certain conditions. Winter scenes where the trees were without leaves
    coupled with bright cloudy skys almost always gave me some CA, especially on
    the smaller branches. It wasn't generally all that noticable when I had 4X6
    prints done, but crops and enlargments would make it quite noticable.
    Reading some of the other posts, it would appear maybe there are different
    causes for fringing.

    As for the whether or not a bad lens should make the problem worse on a
    Foveon sensor? I would think it should suffer the problem less. For a given
    surface area, it uses half the number of sampling points vs say the common
    6mp bayer sensor cameras. I would think that could actually make the Foveon
    sensor less sensitive to small refraction errors of the lens vs a sensor
    that had more desely packed pixels. I guess to an extent it would depend on
    how big the actual sensing locations were, and how much of a gap there is
    between sampled areas. I would think this could make the sensor less
    critical of lens quality for sharpness as well.

    Whatever the causes, my own experiences would indicate that the lens still
    has a big impact. On my DRebel, I get purple fringing much more often and
    consistantly when I use my 70-300 vs my 28 to 105, even in the range between
    70 to 105. This would indicate to me that the lens plays a significant role
    in the problem.

    Whatever the causes, it will be interesting to see where technology comes in
    to play in resolving it, and where technology takes photo imaging in the

    For what it's worth

    Searching_ut, Jun 2, 2004
  7. digiboy

    Larry Guest

    The purple fringing is more likely due to lens problems than it is to types
    of sensor be it film/bayer/foveon..

    I agree with you Jeff, this "theory" needs a re-think.
    Larry, Jun 2, 2004
  8. digiboy

    Skip M Guest

    don't see any evidence of CA in this image, are you sure you don't have a
    monitor problem?
    Skip M, Jun 2, 2004
  9. digiboy

    Phil Wheeler Guest

    Phil Wheeler, Jun 2, 2004
  10. digiboy

    Paul Howland Guest

    Me neither
    Paul Howland, Jun 2, 2004
  11. digiboy

    Ray Fischer Guest

    1 pixel wide.
    So what? No camera design is perfect. All create artifacts.
    Can you live with the artifacts?
    Any conclusion you draw without knowing what you're looking for is
    going to be nonsense. The biggest-by-far cause of fringing is the
    Ray Fischer, Jun 2, 2004
  12. First, if the camera has an anti-aliasing filter, any high contrast
    transition is spread over a 2 pixel width minimum no matter how perfect
    the lens.

    Second, what does that second sentence mean? At any one pixel, one
    colour is measured in a Bayer sensor while the other two are estimated
    from nearby pixels. There can be some error in this process, but the
    error is limited because the rate of change of the image is also
    limited. But I don't understand "try to make up for another channel
    having more data".

    Dave Martindale, Jun 2, 2004
  13. digiboy

    Ron Hunter Guest

    From what I have read, it has more to do with the lens and its
    coatings, and what is being photographed.
    Ron Hunter, Jun 6, 2004
    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.