SD9 SD10

Discussion in 'Digital Photography' started by CBM, Jan 21, 2004.

  1. CBM

    CBM Guest

    Has there been any new reviews on the SD10 performance verses the SD9
    As I have read the noise level should be better,but what about bright spot
    blow out on night scenes, blue sky to blue, and color spots on bright
    CBM, Jan 21, 2004
    1. Advertisements

  2. CBM

    Nils Rostedt Guest

    There is a preview on My impression, based on viewing a
    number of sd9/10 images, is that the noise level is better in most images,
    but the use of "FillLight" seems to accentuate noise. The color issues
    remain although there is a slight improvement in some images. In some images
    the microlenses actually seem to slightly increase the CA defects as well as
    color fringing of highlights, but this may also depend on the quality of the
    lenses used. Edge sharpness is slightly decreased, although the resolution
    is similar.

    Just my $0.02
    Nils Rostedt, Jan 21, 2004
    1. Advertisements

  3. CBM

    CBM Guest

    That's what I thought, Dam good Idea on using each pixel for all the
    colors, but as in allot cases, needs more thought, more research, as it
    could prove to be the best Image IC for pictures. Just like Sony years ago
    was one of the first to Stripe the color CRT but after some development. Now
    most TV do it if it has a CRT.
    Keep up the research, new comes from it.
    CBM, Jan 21, 2004
  4. CBM

    AArDvarK Guest

    That's exactly what I thought, they ruined the x3 with those
    lenses. The yellow stands far more than in SD9 images and
    they are not as sharp. SD9 images are better. They need to
    re-engineer that chip if it is to have the lenses.

    AArDvarK, Jan 22, 2004
  5. There are lots of full sized shots from both here...

    I don't think you are likely to see another Foveon sensor seriously reviewed
    in a popular pro forum or magazine, since Canon/Nikon/Fuji will immediately
    pack up and take their money home. The color resolution problem of Bayers
    has become too well known, so these sites are in a serious bind: they can't
    test Foveon without doing a full color resolution test instead of B&W, and
    they can't introduce a full color resolution standard without instantly
    losing all of their Bayer sponsors.
    The last three are non-existent 9, they were all phenomenom unique to the
    original firmware only, and were upgraded away about 1.5 years ago. The SD9 review is useless in this respect.

    Here is a a good demo of that. First, dpreview's obsolete image...

    Compare that to actual SD9 night performance...

    The "blue sky problem" is simply WB, as you can see from the strongly blue
    toned pure grays in the dpreview sample where the claim was originally made.
    dpreview did not allow any WB adjustment, to make it fair with a JPEG
    shooter since WB is permanently set with JPEG. The SDs never permanently
    set WB, so it's absurd. As is dpreview's "point" that no WB bracketing is
    offered with the SDs. That said, the SD's dynamic range is so astounding,
    you actually can capture blue skies instead of the usual pale-blue-to-white
    skies of any Bayer image where the foreground is properly exposed with clear
    shadow detail, like these SD9/10 images...

    All those "issues" were simply bones thrown to inferior Bayer cameras. Much
    like his moving the SD9 15 feet farther away for his resolution tests,
    erroneously thinking that a 1.7X vs 1.6X crop factor actually changed
    optical magnification, as he states in the text. And the SD9 still
    clobbered the D60.

    So with that in mind, the true differences between the 10 and the 9 are...

    - The 10 has better light sensitivty than the 9 due to microlenses. These
    capture more of the obliquely angled rays towards the edge of a digital
    sensor. It's very important to distinguish this method of noise reduction,
    increasing the signal strength, from Canon's approach of algorthmic noise
    reduction, which reduces apparent noise at the expense of unwanted
    degradation of the signal. Foveon's appoach increases it's massive full
    color absolute resolution advantage even more, thus increasing the s/n
    ratio, while Canon lowers absolute resolution to increase the s/n ratio.
    Anyway, the SD10 has legitimately higher ISO performance than the 9 by about
    1 stop. That said, you can recover that stop plus reap lots of other
    benefits, by upgrading lenses with the money you save from buying a $750
    SD9. If money is no object, the SD10 wins with the same lens.

    - The addition of microlenses also seems to make the 10, ever so slightly,
    softer than the 9. Almost imperceivable, but Sigma users are extraorinarily
    picky about having only pristine sharpness for every keeper, typically.
    That said, both the 9 and the 10 blow every other DLSR away.

    - Generation 2 Foveon Pro 10M sensor with improved color rendering and lower
    power consumption.

    - The 10 has reworked AWB routines to completely eliminate the tendancy for
    AWB to produce yellow skin tones in bright yellow sunlight. WB has nothing
    to do with actual color capture, mind you, since a RAW file's WB is forever
    fully adjustable. The difference is simply one of convenience, not color.

    - Single battery source, the 10 doesn't need two C123a's. This is a double
    edge sword. The 123a's are definitely a lot less convenient and they cost a
    little more, $20 for a 20 lot. On the other hand, the 9 has more power. On
    the third hand, the 10 is a little more power efficient. All told, the 10
    seems to get around 10% fewer shots than the 9 using the normal batt
    tray--maybe 250 shots per charge instead of around 275. The difference is
    less significant with a Power Pack, which will last for 500-1000 shots with
    either model, the low end representing about 25% MLU use, the high end
    representing none. Same thing when using 2 CRV3s instead of 4 AAs, then
    either model will get around 600-800 shots. 4 CRV3s in a Power Pack should
    be good for 2000+ shots, though I've never seen anyone post a figure. The
    9's two 123a's last for 1000-1200 shots.

    That's about it, the two are very similar in absolute performance, but the
    10 gets you there quicker/easier with an extra ISO stop. The 9 will
    definitely outperform the 10 if the money saved gets you into EX glass.
    George Preddy, Jan 22, 2004
  6. because there aren't any.
    which is absolutely the correct thing to do - frame the test chart to fill
    the field of view of the camera.
    by rendering anything above 1512 lph - which is the sensors absolute maximum
    resolving power (since the sensor is 1512 pixels high) as sharp detail in
    the wrong place - ie rendering a mass of alias artefacts.

    Because the fill factor of the sensor is around 100%, rather than the measly
    30% which gives rise to its false sharpness and exaggerates ists already
    terrible alising problems.

    .... since that is the only benefit to using a sigma. Flase sharpness instead
    of correct imaging.
    so the SD10 will still produce yellow sking tones in many lighting
    situations which are not WB problems at all - they cannot be fixed ofter the
    Braindead Preddy, Jan 22, 2004
  7. CBM

    AArDvarK Guest

    question ... how do you define "false sharpness"?
    AArDvarK, Jan 22, 2004
  8. Sharpness which is the result of aliasing rather than existing in the
    original scene.

    If you look at the resolution test on you will see that the SD9
    continues resolving as "sharp" details beyond 1512 lpph. This is false
    sharpness, because the sensor is only 1512 pixels high, so it cannot
    possibly resolve these details.

    The correct thing to do in these circumstances is to render these details as
    blurred, the same way that your eyes do. This is why the Sigma pictures look
    so artificial. Well, that and the colour rendition problems.
    Braindead Preddy, Jan 22, 2004
  9. CBM

    CBM Guest

    In reality it looks like the Foveon sensor is a step towards the right
    approach. There may be a few flaws but all new ideas lead to improvements.
    Look at Sonys idea in using four colors in their sensors to make a better
    picture, its a start. I bet right now Sony, Nikon, and others are looking at
    the Foveons chip to further develop something similar,as from the theory of
    its operation it is right on.
    CBM, Jan 22, 2004
  10. Agreed. It needs an anti alias filter though, or at least as massively
    higher pixel count than 3.4 MP.

    And it needs better colour seperation.

    Manufaturing difficulties and marketing decisions mean that the current
    Foveons are a poor choice.
    Braindead Preddy, Jan 22, 2004
  11. CBM

    Mark Herring Guest

    If it is due to aliasing, it can be done mathematically in the spatial
    frequecny domain. If you start with a plot of the modulation vs
    spectral frequency for the scene date as it arrives at the sensor,
    then--knowing the sampling frequency, you can determine what portion
    of that curve is folded over (aliased).

    A more subjective definition: Any modulation in the image which gives
    the impression of sharpnes or detail but does not derive from real
    information in the scene.
    Mark Herring, Pasadena, Calif.
    Private e-mail: Just say no to "No".
    Mark Herring, Jan 22, 2004
  12. CBM

    Lionel Guest


    Check the EXIF information on this shot, folks - it's been run through
    NeatImage to remove the noise!


    Hey George, are Sigma now supplying a free copy of NeatImage with the
    SD9 to fix its godawful-noisy low-light performance?
    Lionel, Jan 22, 2004
  13. NI doesn't remove huge white ovals around point souces at night. The fruit
    look better without the blemishes shown by the SD9, so an 50% NI pass was
    used to reduce the optical detail. Here's the original, in case anyone is
    dumb enough to believe that NI replaces huge white circles around bright
    lights at night with beautiful 18 point starbursts...

    BTW, Canon's noise in-camera noise reduction is much more agressive (and
    primitive) than NI.
    George Preddy, Jan 23, 2004
  14. Nature.
    George Preddy, Jan 23, 2004
  15. CBM

    Lionel Guest

    George, you've proved yourself to be a liar, yet again. You've presented
    an SD9 image that's been post-processed by a very good noise-reduction
    program, *without mentioning it* as an example of the cameras low-light
    performance. It doesn't surprise me greatly though, as it's the only
    possible way that you could make a high-ISO SD9 image even slightly
    comparible to a 10D image of the same ISO setting.

    You're a pathetic, lying weasel, George.
    Lionel, Jan 23, 2004
  16. CBM

    AArDvarK Guest

    And there is nothing in the info besides ""
    that proves it was shot by an sd9/10, it's a simple four
    line EXIF. Nothing about it on the page either.

    AArDvarK, Jan 23, 2004
  17. Good catch, I was confused too much by the aliasing artifacts (artificial
    parallel starburst rays) to even notice the additional fraud.

    Bart van der Wolf, Jan 23, 2004
  18. Indeed, if not depicted as the human eye would, FALSE.

    Bart van der Wolf, Jan 23, 2004
  19. I posted both. The NI image has less detail, which in this rare case is
    arguably nicer, though many probably won't agree, since so many fruit
    blemishes are imaged by the SD9. No one vain, including fruit, wants to be
    photographed by the SD9. The optical resolution is brutal.

    Aren't you going to explain why you claimed NI removes the huge white ovals
    in the dpreview image? Surely you've run that image through NI by now, and
    realized you were wrong.
    George Preddy, Jan 23, 2004
  20. Sure there is, both images are posted on that site...

    Lionel didn't do his homework.
    George Preddy, Jan 23, 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.