Re: 25 Reasons to Aviod the SD-10 (was 15 Reasons to Aviod the SD-10)

Discussion in 'Digital Cameras' started by Paul Howland, May 9, 2004.

  1. Paul Howland

    Bill Funk Guest

    Your naivete is rather amazing, for someone who professes to know so
    If course there are such people. You don't *actually* expect me to
    believe that the Democratic party is better than the Republican party,
    just because there are people who switch from being Republicans to
    Democrats, do you? (Or, for those who are Democrats, reverse the
    analogy; six of one...)
    Bill Funk, May 23, 2004
    1. Advertisements

  2. Paul Howland

    Bill Funk Guest

    Laurence, do you believe this?
    Bill Funk, May 23, 2004
    1. Advertisements

  3. Paul Howland

    E. Magnuson Guest

    Yes. Satire of a ridiculous statement.
    Not a single insulting word (but alas,
    I slipped from the true path later)
    E. Magnuson, May 23, 2004
  4. Nonsense, it only captures 3 poorly separated color channels, and makes RGB
    out of that.

    Bart van der Wolf, May 23, 2004
  5. Paul Howland

    E. Magnuson Guest

    If we can assume that "luma resolution" is what is typically measured
    by the various review sites resolution charts, then

    SD9/10: 1550/1550
    10D/300D: 1600/1450
    1Ds: 2400/2000

    No published numbers for the 1DMkII yet, but simple extrapolation yeilds:

    Now if you want to play word games and claim this is not "significant"
    when it's enough to command a market premium, go ahead. It's your
    If you have any cites that detail the "in-camera" pre-raw processing
    of any of these cameras, we'd all love to hear it. The only yardstick
    we unambigously have is the output results.
    I don't get your point. Either you are referring to facts not in
    evidence (pre-raw processing) or you are comparing to a feature
    (in camera JPEG) that does not exist on the SD9/10.
    If you are attempting to argue that the learning curve for SPP is
    smaller than for ACR/C1, then I'll agree. Any other point you're
    try to make here escapes me.
    Yes, opinions *do* vary. Primarily they vary on the definition of
    "detail". The biggist, stickiest wicket is the now infamous "detail
    beyond Nyquist". I will agree that from a pure photographic
    standpoint that sometimes this can contribute to the aesthetics of
    certain images. Also, as I state repeatedly, the best evidence of
    superiority is side-by-side images under close to identical conditions
    so that we can all make up our minds. I don't recall seeing any of
    So you want to keep your credibility at the same level as his?
    I agree, you both need to provide more cites. I've said so to Steven
    and to also you even before you came over here.
    Where is the one that has the 10D sensor fill-factor?
    Even if the 10D pixel is smaller than the SD9/10 sensor,
    you (should) know know better than to equate pixel size
    to image quality for completely different pixel constructions.
    It's as silly as a strict adherence to "bigger is better" for
    E. Magnuson, May 23, 2004
  6. Paul Howland

    E. Magnuson Guest

    To be honest, it's pointless arguing semantics as to whether this is
    "interpolation" or not. What has been published is that the RGB is
    calculated from the measured values by multiplying with a 3x3
    correction matric with strong off-axis correction factors. Foveon
    authored papers admit that this has some drawbacks (including
    increased noise) versus directly measuring RGB in close to the ideal
    tri-stimulous space. Now Foveon thinks that the other advantages they
    gain are worth this price. (One side note: attempting to apply this
    type of correction to actual X3F files usually gives rather poor
    results [c.f. the results from dcraw.c]. The real processing applied
    by SPP is obviously much more involved.)

    A considerable amount of processing is applied to the measured data
    for both systems. By the time that ink hits the paper even more
    processing is done including (almost always) interpolation. Why are
    we argueing about this?
    E. Magnuson, May 23, 2004
  7. Paul Howland

    Skip M Guest

    You're one who said his numbers were wrong. I've already done battle with
    the bonehead, I'm done with that. I generally don't get into an argument
    about the cameras, partly because I feel the SD10 is a reasonable
    alternative to the 10D, since they are similarly priced. But George's
    comments about Canon vs. Sigma glass are demonstrably wrong, and, as one of
    several on this list who either own or have owned both, can compare them
    directly, besides the ratings that have been published show him to be way
    off the mark, vis a vis his howler that the Sigma 50mm f2.8 is the best
    rated lens ever built. It's a good lens, no doubt, but several lenses out
    rate it. (See Photodo or Photozone ratings.)
    Skip M, May 23, 2004
  8. Paul Howland

    Skip M Guest

    That, sir, is libelous. But coming from you, who can't post an image that
    you can prove is your own, it is merely laughable.
    Skip M, May 23, 2004
  9. No, it's pretty clear that *you* don't understand the sensor or the
    process. If you were to do a Google search and catch up on 6 months of
    discussion that you've missed, you'd find references to

    - the Foveon sensor spectral response curves, which are nowhere close to
    the ideal RGB responses for a 3-colour camera used for photography

    - what unprocessed Sigma raw data actually looks like

    - the processing necessary to give even moderately plausible colour,
    as performed by the dcraw program (whose source is available).

    Most of the people you're arguing against have been here and reading
    for that 6-month period, and know these things. On the other hand,
    your own knowledge about the camera doesn't seem to extend beyond
    Foveon's marketing literature.
    And just how would you know that? Spatial interpolation isn't
    absolutely necessary to obtain 3 colours at each pixel location, which
    he said already. But considerable processing *is* necessary to get
    reasonable colour and noise, and that processing apparently involves
    surrounding pixels. So it's really arguing semantics whether this
    is called interpolation or something else. It certainly is *not* true
    that the Foveon sensor returns each pixel independent, ready to use.
    It is *not* true that the Foveon simply and naturally calculates pixels
    without any "guessing" - there's at least as much mathematical fudging
    about as Bayer demosaicing.

    Dave Martindale, May 23, 2004
  10. And I'd strongly argue that those SD9/SD10 numbers are wrong. In the
    first place, no sensor can resolve more lines than there are rows in
    the sensor, and the Sigma has only 1512 rows.

    If you look at the full-size resolution test image of the SD9 from the
    dpreview web site, you can see that it really only resolves 1100 or at
    the very most 1200 lines per picture height cleanly, then it starts into
    moire and aliasing patterns. At 1500 you get a clean pattern again as
    the target's lines align with the pixels in the test pattern, but this
    depends on exactly how the pattern and sensor happen to align. A rating
    of 1500 should mean that the reproduction is clean all the way from zero
    up to 1500, and it clearly isn't.

    The SD10 is much better behaved, and looking at its resolution chart it
    is very clear that the resolution is about 1100-1200 l/ph. No stretch
    of the imagination could call it 1500.

    Look at the actual test images, not the numbers that Phil attempts to
    read from them!

    Dave Martindale, May 23, 2004
  11. Paul Howland

    E. Magnuson Guest

    Yes, I almost added that. But then again all 3 sets of numbers were
    determined as a judgment call by the reviewer. If you correct one you
    have to correct them all and I wanted to use numbers well known to
    Laurence w/o editing. The main reason I think the numbers are adequate
    is that they do reflect the relative photographic useful resolution in
    the opinion of that same reviewer (and myself.)
    I'm well aware of that. Note that Phil also does not reduce the
    resolution of CFA cameras that exhibit some color moire or other
    artifacting. Look at his 14n review.
    My point was simply that even using Phil's numbers, there are two Canon
    CMOS sensors with more luma resolution. If you want to argue 3, that's fine
    but I want no part of that argument on either side.
    E. Magnuson, May 23, 2004
  12. Paul Howland

    Bill Funk Guest

    Ok, I'll buy that, sort of.
    This means that when a Bayer sensor says "xMP", they mean "xMP",
    because each pixel remains in place, and the color is added to the
    This is where you err.
    The data pulled off is not some strange quantum stuff, but the exact
    same electric potential that's pulled off Bayer sensors.
    That data represents color that is filtered by the silicon, whether
    you want to admit that or not.
    In the Foveon chip, each data point for each spatial receptor (pixel)
    reads data from the same surface point, but from 3 different points in
    the silocon beneath that point.
    What results after processing is a pixel with a close approximation of
    the color as it falls on that point.
    Hardly. That's just not possible, since the data is pulled from three
    discrete points in the silicon, which acts as a filter.
    Bill Funk, May 23, 2004
  13. Paul Howland

    JPS Guest

    In message <>,
    That's true of every SD9 image taken with sharp optics, focused on fine
    detail. Every single one, without exception. It *IS*, as AFAIAC, the
    reason that the Sigma has any fans based on the way the images look.
    Some people like sharpness, regardless of its nature. Do a
    small-radius, high-power USM on a Bayer image taken with sharp optics,
    and then do a nearest neighbor downsizing from 6MP to 3.43MP, and you
    will get a similar effect, although going to about 1.6MP is more
    SD9-like (4 to 4.5MP is more SD10-like).
    Phil is another lover of false sharpness. He has stated in the Sigma
    forum that he *likes* aliasing. He said that he would rather see crisp
    green edges than a green blur, in a distant tree.

    I would like to see crisp detail in distant trees, too, but I know that
    there aren't enough pixels, and the fairy-tale ends right there.
    JPS, May 23, 2004
  14. Fair enough. My own statements are based on my own judgements of
    resolution, which always seem to be more conservative than Phil's. To
    me, a rating of N l/ph means the camera cleanly resolves all the way
    from 0 to N, not that N is the highest resolution that can be said to be
    But I would. I don't consider the pattern to be accurately resolved if
    there are significant false colours present.
    Ok. No argument at all about the other two.

    Dave Martindale, May 23, 2004
  15. Yes - this is important to understand. Phil is a very
    respectable reviewer that maintains a very good site.
    His opinions are generally very good indeed. But he, as
    all of us, sometimes is in error or at least have some
    strange notions.

    I do not agree with Phil regarding the benefit of Foveon
    lack of anti aliasing filter, the way he describes chromatic
    aberration, his lens resolution tests and his dynamic range
    tests. But, I think thats about it. Most of the other stuff I
    find very accurate and of great value.

    Roland Karlsson, May 23, 2004
  16. No they mean that's how many pixels they are.

    When you claim that Sigma has more than 3.43MP, you are erroneously
    substituting sensors for pixels. Please educate yourself.

    No, full colours are not sensed. Something vaguely like red, something
    vaguely like green and something vaguely like blue are sensed att
    different layers of the silicon chip due to the (rather inefficient)
    filtering properties of the silicon itself.
    Sort of a destaurated RGB, provided by the inherent (but poor)
    filtering properties of the silicon.

    The problem is, however, the need for complex post-processing (or
    interpolation) to restore those desaturated sort of RGB datum points
    to something that resembles the initial colours.

    grant kinsley, May 23, 2004
  17. Paul Howland

    JPS Guest

    In message <>,
    I'm sorry, I didn't paraphrase that too well. He didn't say that he
    liked aliasing, just that he likes some of the effects (without
    mentioning the aliasing per se).
    JPS, May 23, 2004
  18. Paul Howland

    Lionel Guest

    Kibo informs me that (Laurence Matson) stated that:
    EOS 10D (6.3MP), 300D (same sensor as the 10D), 1Ds(11MP+), 1DmkII(8MP).
    (IIRC, the 1D is a CCD, not a CMOS sensor, but I may be mistaken.) All
    of these have significantly higher resolution than the 3.4MP of the SD9
    & 10.
    Yes, & what's your point? TTBOMK, the in-camera NR on the Canon CMOS
    sensor cameras consists solely of dark-frame subtraction on long (>0.5
    second) exposures, & pixel sensitivity mapping & compensation. I don't
    think anyone sensible would describe either of those processes as
    'agressive noise reduction'.
    Eh? - I get excellent results from Canon RAW, & so do many other people.
    It's not terribly difficult if you know what you're doing.
    (Please pardon the dust spots on that image.)
    Lionel, May 24, 2004
  19. Paul Howland

    E. Magnuson Guest

    You asked. I did not bring it up.
    First of all *you* were the one who first said "luma resolution"
    (preserved in the quotes above.) Don't try and change the frame
    of reference to color resolution now that the numbers don't
    back up your original statement.

    What do you want, Laurence? I'll agree with you that X3 sensors have
    better resolution where there is little luma info. I'll agree that
    for many real world photographs they produce prints that are roughly
    on the same level as a 6MP mosiac sensor. I'll even agree that
    depending on your personal taste in what makes a good image, they might
    be better. However, if you make a bold statement where the best facts
    available contradict it, I will call you on it.
    See the thread about how SD10 white balance and how raw is raw. Most
    cameras probably perform some analog processing (and maybe even some
    digital processing) for ISO, noise reduction (dark frame subtraction
    if nothing else), and other signal conditioning before writing the raw
    I'm always open to constructive criticism. I'll even send you some of
    my X3F files if you want to show me how the master does it.

    But again, what's your point? Are you claiming that someone who has
    mastered each post-processing chain will always produce better images
    with the SD10 than a 10D? If so, all you have to do is to show me
    these results.
    Good. Now we are down the to point where reasonable people can
    disagree. I'm actually agnostic on this issue. I feel that sometimes
    is helps and sometimes it hurts. I guess that puts me in the center
    getting flak from both camps.
    All you have to do is ask.
    Uh, Laurence, you do remember that not only did I read that thread the
    same day it was posted, I even asked Mr. Merrill questions which
    he mostly ignored or handwaved because they do not fit his thesis.

    So where again is that 10D fill-factor number?
    I agree. Alas, the photos he references are no longer online so
    you can't see how dubious his comparisons were. And in this
    group, anyone who cites the outbackphoto color resolution charts
    is not going to be held in very high esteem.

    Are you *sure* Preddy did not hijaak your account? I expected you to
    clear up some of the misconceptions about the Sigma cameras and by
    providing a voice of reason, with decent photos to go with it, show
    that there are Sigma owners who like to make good photographs and not
    rehash the same old "it's better technology on a pixel level"
    folderol. Believe me, the arguments have been hashed and rehashed
    here thousands of times here just as on dpreview and no one is likely
    to change his/her mind just because you or Mr. Merrill says so.
    E. Magnuson, May 24, 2004
  20. Paul Howland

    imbsysop Guest

    On 24 May 2004 05:47:36 -0700, (Laurence Matson)

    big snip ..
    This is as interesting as all the rest you post ...

    Non-existent gallery specified in URL. Displaying richardbmerrill's
    root gallery.
    Richard B Merrill | profile | all galleries | recent tree view |

    click on thumbnails for full image
    imbsysop, May 24, 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.