Nikon D3: 23.9 ???

Discussion in 'Nikon' started by Mike -- Email Ignored, Dec 7, 2007.

  1. Why is the sensor of the Nikon D3:
    23.9 x 36.0 mm
    rather than:
    24.0 x 36.0 mm
    ?
    Seems peculiar to me.
    Mike.
     
    Mike -- Email Ignored, Dec 7, 2007
    #1
    1. Advertisements

  2. Makes it more like film?

    It's really 23.999mm but they rounded down?

    Matches Canon?

    <G>

    David
     
    David J Taylor, Dec 7, 2007
    #2
    1. Advertisements

  3. Mike -- Email Ignored

    Frank Arthur Guest

    ..1mm = .0039 inch which is less than the thickness of a human hair
     
    Frank Arthur, Dec 7, 2007
    #3
  4. We don't know, and there is very little possibility that
    we ever will.

    However, speculation is cheap. The Canon 5D is
    23.9x35.8mm, for probably the exact same reason (with
    slightly different specifics).

    When a senor is designed, there is necessarily some
    specific number of rows that are minimal. The reason
    for that has to do with the number of data channels
    being used.

    Hence, to add 1 more more of sensors across the top is
    not a matter of just one row, but of 3, 4, 5, 8 or
    whatever. Nikon, Sony, Canon, etc do not give us access
    to enough information about each sensor model they use,
    so we don't know what it is. But the effect is that
    different sensors have slightly different geometry, and
    also slightly different pixel counts.

    It is probably true that for Nikon to have added more
    pixels would have required that the size be something
    like 24.3x36.0mm, which they did not want.

    In the end though, it simply does not make any
    difference at all. 23.9mm is just not different than
    24.0mm.
     
    Floyd L. Davidson, Dec 7, 2007
    #4
  5. Mike -- Email Ignored

    Don Wiss Guest

    Exactly. For jpeg compression it needs to be divisble by 8. The pixel count
    of the D3 is 4256 x 2832, which is divisible by 32 and 16 respectively.

    Don <www.donwiss.com> (e-mail link at home page bottom).
     
    Don Wiss, Dec 8, 2007
    #5
  6. But JPEG compression has exactly zero to do with it.
    This would be entirely a *hardware* phenominon, and
    probably relates to how many read channels exist between
    the sensor and the analog amplifiers that feed the ADC.

    JPEG may well use 8 pixel wide blocks, but it merely
    starts on one side and goes until the end. I guarantee
    that you can size a JPEG image to be an odd number of
    rows or columns... (Try it!)
    However, that is the size of the image. The size of the
    sensor is significantly more. 4256x2832 works out to
    12,052,992 pixels, which Nikon says is "12.1 MP". They
    list the sensor as "12.87 million". That is
    significantly larger!
     
    Floyd L. Davidson, Dec 8, 2007
    #6
  7. Mike -- Email Ignored

    C J Campbell Guest

    Oh, foo. It does not even use the whole sensor anyway. The sensor has
    12.9 million pixels, but 12.1 million effective pixels. The rest are
    used for demosaicing, determining true black, or simply not used at
    all. So the part of the sensor that is actually used to record the
    image is even smaller than 23.9 x 36.0 mm.
     
    C J Campbell, Dec 8, 2007
    #7
  8. Mike -- Email Ignored

    Scott W Guest

    This depends on where you get the jpeg file from. On my 350D if I use
    the cameras jpeg it is divisible by 8 in both x & y, but if I use a
    raw converter (depending on the converter) I get a few more pixels and
    it is no longer divisible by 8.

    Scott
     
    Scott W, Dec 8, 2007
    #8
  9. No it does not depend of where you get it from. It
    depends only on if you actually do read what as said,
    and try it. You *can* size a JPEG image to be an odd
    number of row or columns. Even if it originally came
    from a source that generates images that do just happen
    to be divisible by 8. The point is still, JPEG just
    does not care.
    Well, that's what I said! It does not *have* to be
    divisible by 8. Nobody is saying it can't be.
     
    Floyd L. Davidson, Dec 8, 2007
    #9
  10. That is a characteristic common to all digital cameras.
    The exact specifics differ from one camera to another,
    but they all do that sort of thing

    The Nikon D1, for example, had a 24x16 sensor with 2.6MP,
    according to almost everything you'll read.

    In fact it was 23.7mm x 15.6mm and the image size was
    2000x1312, but the "effective" pixel count was 2012x1324
    (2,624,000).
    I don't know if that is true or not. I suspect that
    when they say 23.9x36.0 they mean the part of the sensor
    that is used. They might not though.

    Now, here's one for you. The D3 generates a 4256x2832
    JPEG image. That is what the camera's menu says when
    selecting the image size for a JPEG, and that is what it
    produces too.

    But the raw data has 4288x2844 pixels (according to the
    EXIF data), and in fact an external RAW converter will
    generate an image of 4288x2844. Hence, Nikon trims off
    a 32x12 border off. (I haven't checked to see if this is
    centered or if the entire area is top/left, or what.)
     
    Floyd L. Davidson, Dec 8, 2007
    #10
  11. Irfanview's lossless 90 degree rotations of jpg images require it to
    be divisible by 8, and if it isn't they truncate it.
     
    Chris Malcolm, Dec 8, 2007
    #11
  12. It would be very tricky to align the sensor exactly in position to
    within a small fraction of a pixel, so the sensor should always be at
    least one row and column of pixels bigger. It would be interesting to
    check the symmetry of the trimmed borders.
     
    Chris Malcolm, Dec 8, 2007
    #12
  13. Mike -- Email Ignored

    Guest Guest

    as you say, 'almost everything you'll read.' :)

    nikon now reveals that it actually had four pixels per pixel making it
    a 10.8 megapixel sensor, and bjorn roslett has a close-up photo of it.


    <http://nikonimaging.com/global/technology/scene/12/index.htm>

    I guess that it's now safe to reveal that the D1 image sensor, with
    specifications noting a pixel count of 2.7-million pixels, actually
    had a pixel count of 10.8-million pixels. The technical reason for an
    actual pixel count four times greater than that indicated publicly
    lies in the need to achieve high sensitivity and a good
    signal-to-noise ratio. Unlike current cameras, for which final pixel
    counts account for individual pixels, we had to include multiple
    pixels in each pixel unit with the D1. 

    <http://www.naturfotograf.com/images/B030217207.jpg>
     
    Guest, Dec 8, 2007
    #13
  14. Chris Malcolm wrote:
    []
    I think you'll find that the extra rows and columns are used for
    dark-current measurement etc., so there will be enough pixels to get a
    meaningful value (i.e. not just one or two pixels). The position of the
    borders will depend on the mask inside the sensor, rather than the
    alignment of the sensor to the mechanics of the rest of the camera. It
    still doesn't explain why the active area can't be precisely 36 x 24mm,
    though...

    Cheers,
    David
     
    David J Taylor, Dec 8, 2007
    #14
  15. Mike -- Email Ignored

    Robert Coe Guest

    : >>On Dec 7, 3:36 pm, (Floyd L. Davidson) wrote:
    : >>> >On Fri, 07 Dec 2007, (Floyd L. Davidson) wrote:
    : >>>
    : >>> >>Hence, to add 1 more more of sensors across the top is
    : >>> >>not a matter of just one row, but of 3, 4, 5, 8 or
    : >>> >>whatever.
    : >>>
    : >>> >Exactly. For jpeg compression it needs to be divisble by 8. The pixel count
    : >>>
    : >>> But JPEG compression has exactly zero to do with it.
    : >>> This would be entirely a *hardware* phenominon, and
    : >>> probably relates to how many read channels exist between
    : >>> the sensor and the analog amplifiers that feed the ADC.
    : >>>
    : >>> JPEG may well use 8 pixel wide blocks, but it merely
    : >>> starts on one side and goes until the end. I guarantee
    : >>> that you can size a JPEG image to be an odd number of
    : >>> rows or columns... (Try it!)
    : >>This depends on where you get the jpeg file from.
    :
    : > No it does not depend of where you get it from. It
    : > depends only on if you actually do read what as said,
    : > and try it. You *can* size a JPEG image to be an odd
    : > number of row or columns. Even if it originally came
    : > from a source that generates images that do just happen
    : > to be divisible by 8. The point is still, JPEG just
    : > does not care.
    :
    : >>On my 350D if I use
    : >>the cameras jpeg it is divisible by 8 in both x & y, but if I use a
    : >>raw converter (depending on the converter) I get a few more pixels and
    : >>it is no longer divisible by 8.
    :
    : > Well, that's what I said! It does not *have* to be
    : > divisible by 8. Nobody is saying it can't be.
    :
    : Irfanview's lossless 90 degree rotations of jpg images require it to
    : be divisible by 8, and if it isn't they truncate it.

    Which belies the term "lossless", doesn't it
     
    Robert Coe, Dec 24, 2007
    #15
  16. Mike -- Email Ignored

    crownfield Guest

    -: > Well, that's what I said! It does not *have* to be
    -: > divisible by 8. Nobody is saying it can't be.
    -:
    -: Irfanview's lossless 90 degree rotations of jpg images require it to
    -: be divisible by 8, and if it isn't they truncate it.
    -
    -Which belies the term "lossless", doesn't it

    in one sense.
    in the other, you can rotate them for the rest of the year,
    and there will be no further loss.

    I do not fret the loss of the outside row of pixels
    when the rest of the image is lossless.

    -
     
    crownfield, Dec 24, 2007
    #16
  17. Mike -- Email Ignored

    C J Campbell Guest

    I suspect that I already knew that, but you probably already knew that
    I knew, didn't you?
     
    C J Campbell, Dec 24, 2007
    #17
    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.