D3 - new high iso samples

Discussion in 'Digital SLR' started by frederick, Aug 31, 2007.

  1. frederick

    frederick Guest

    frederick, Aug 31, 2007
    #1
    1. Advertisements

  2. frederick

    Tony Polson Guest

    Tony Polson, Aug 31, 2007
    #2
    1. Advertisements

  3. frederick

    cjcampbell Guest

    cjcampbell, Aug 31, 2007
    #3
  4. frederick

    Toby Guest

    Toby, Sep 1, 2007
    #4
  5. frederick

    cjcampbell Guest

    I think a D3 costs at least 5 children.
     
    cjcampbell, Sep 1, 2007
    #5
  6. frederick

    Matt Clara Guest

    Unless strong like bull...
    ;-)
     
    Matt Clara, Sep 2, 2007
    #6
  7. frederick

    Toby Guest

    Well, perhaps I'll have to sell my soul to the Devil to make up the
    difference...

    Toby
     
    Toby, Sep 2, 2007
    #7
  8. How comes "ISO 200" is out of focus and not sharp, while "ISO
    1600" is?

    BTW, the highlights in the "2" of 1:2,8 are seriously blown.

    -Wolfgang
     
    Wolfgang Weisselberg, Sep 3, 2007
    #8
  9. frederick

    acl Guest

    The shutter speed is higher for the ISO 1600 shot?
    Instead of whining about non-issues, take a look at the individual
    channels of these images and you'll really find something to whine
    about. Or maybe not :)
     
    acl, Sep 3, 2007
    #9
  10. He should have been using the Mk III with its superior dynamic range.






    Rita
     
    Rita Ä Berkowitz, Sep 3, 2007
    #10
  11. Other than the white balance being dizzy (blown red channel, underexposed
    blue), which is something one sees a lot in low-light shooting, what's the
    issue???

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Sep 3, 2007
    #11
  12. frederick

    acl Guest

    Well, if you look eg at the "16" numeral and switch between R, G and B
    channels, you'll see what I mean. Or maybe not.
     
    acl, Sep 3, 2007
    #12
  13. The red channel is grossly softer than the green channel, which is in turn
    grossly softer than the blue channel.

    Perhaps implying strong noise reduction in the green channel, and extreme
    noise reduction in the red channel? Or maybe that the lens is a complete and
    total dog?

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Sep 3, 2007
    #13
  14. frederick

    acl Guest

    I'd go with the lens theory, personally; but something is wrong. Or
    maybe it's the automatic CA removal, I hadn't thought of that... But
    it looks very strange. Anyway, I found this thread in dpreview
    http://forums.dpreview.com/forums/read.asp?forum=1021&message=24669958
    where he posts images at ISO 1600 up to 25600. I only looked at the
    1600 and 3200 briefly, I have to say, it's very impressive... The
    jpeg artefacts have done a lot more damage than the noise (try to
    sharpen it)! Is the 5D like this? I didn't get that impression when I
    played with raw files from one, but then again, maybe Nikon's jpeg
    simply has better NR.

    A raw file would be nice (except that dcraw wouldn't be able to dump
    the data, probably)...
     
    acl, Sep 3, 2007
    #14
  15. frederick

    frederick Guest

    I'd go with the lens theory too.
    IIRC exif shows 85mm used @ f4.5, with maximum aperture
    available f4.4 (at that focal length).
    So, it's a consumer grade variable-aperture zoom set fully
    wide, probably the 18-135mm.
     
    frederick, Sep 3, 2007
    #15
  16. Yes. Probably the lens.
    The red channel should have _less_ NR applied to it. Oops. My logic's
    backwards.
    I've not looked all that closely, but with the 5D, the noise is dealable
    with up to about ISO 12,800. At ISO 25,600, the speckles get out of range of
    the noise reduction.

    Here's the 5D, ISO 3200, underexposed and pushed 3 stops. (ISO 25,600)

    http://www.pbase.com/davidjl/image/75374090/original

    But I don't have any third party noise reduction at hand, just what you get
    with Lightroom and Photoshop. Here's what I got by cranking NR in Lightroom
    1.0 and sharpening in PS 7.0. (Four crops processed individually.) I should
    try again now that I've got PS CS3.

    http://www.pbase.com/davidjl/image/75359389/original

    The D3 is a full two years after the 5D, so I'd expect it to be slightly
    better. (People expecting, or thinking they're seeing, a stop or more
    improvement are dreaming or halucinating, respectively, I think, though.)
    Canon claims that there's wasted space between the 5D microlenses, and
    Roger's numbers indicate that there's room for improvement in the read noise
    as well. Since Canon is cashing in the space between the pixels as more
    pixels, not less noise, historically, Nikon has a shot at having produced
    the best per-pixel low-light camera in dcam history. If they get it right.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Sep 4, 2007
    #16
  17. frederick

    acl Guest

    It looks strange, though. But other samples I've seen don't show this.
    That's also quite impressive! Looking at the D3 ISO 25600 shot, I
    guess it started out in the same ballpark as this (before being noise
    reduced to within an inch of its life).
    Well Lightroom has very ugly NR then... About sharpening, copying the
    image layer, selecting the least noisy channel (eg green), running
    "Find Edges" on it and then using "Brightness/Contrast" is a good way
    to get an edge mask to sharpen only the edges (to avoid the artifacts
    you get).
    Well, I don't know what the effective fill factor of Canon sensors is.
    Presumably it shouldn't be too hard to find out using Roger's results
    for full well capacity... But there may be a lot of room for
    improvement there if you reduce the need for pixel-level circuitry. A
    stop in total shouldn't be impossible with incremental improvements...
    Anyway, the main advantage I see in Canons is that the increase in
    noise as ISO is increased is less quick than with the previous
    generations of Nikons. Judging from the ISO 25600 samples from the D3,
    this has been fixed. But we'll see.
     
    acl, Sep 4, 2007
    #17
  18. frederick

    acl Guest

    Well maybe not, after all!
    http://forums.dpreview.com/forums/read.asp?forum=1021&message=24673062
    and look at the last shot. Look eg at the red car... Ouch! Maybe it's
    the automatic CA removal...
     
    acl, Sep 4, 2007
    #18
  19. Ew. That's ugly. Really ugly. And it's ISO 200. I'm beginning to think that
    Canon doesn't have anything whatsoever to worry about: this camera simply
    can't make a sharp image. Or maybe Nikon can't make a sharp lens. One or the
    other. (Actually, the problem unbelievably abject user incompetence: it's
    shot at 50mm, f/2.8 and 1/4000: it would probably be quite sharp if it were
    shot at f/11 and 1/250. No matter how bad the lens was.)

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Sep 4, 2007
    #19
  20. frederick

    acl Guest


    No way that's the lens: look at the lampposts, for example. I would
    guess it's the automated CA removal. They're probably still fine-
    tuning it. Or screwed up!
     
    acl, Sep 4, 2007
    #20
    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.