Sony rumor: triple CCD camera with prism beam splitter (a-mount,1.5x crop)

Discussion in 'Digital SLR' started by Alan Browne, Jan 20, 2011.

  1. Alan Browne

    Alan Browne Guest

    I see a rumor floating around that Sony have developed test versions of
    an A-mount (alpha) 1.5x crop camera with prism beam splitters and 3 CCD
    sensors (RGB).

    This would definitely outclass the Foveon for color properties.

    Resolution is not all that great (6 Mpix), but the ISO is reported to be
    a stunning 819,200 (13 stops over ISO 100). One could do portraits by
    moonlight - hand held.

    If they could get this up to 12 Mpix or so, or 20 Mpix or so in FF, it
    could be very interesting.

    http://photorumors.com/2011/01/14/s...-a-mount-with-3-sensors-and-iso-up-to-819-200
     
    Alan Browne, Jan 20, 2011
    #1
    1. Advertisements

  2. Alan Browne

    RichA Guest

    The alignment (and assembly of course) of the images will be done by
    the processor. No way could you do it mechanically without a
    precision-machined, metal (even then, unlikely) body. This is tri-
    colour and it should produce the most accurate colour images yet.
    Basically, the same method used in scientific monochrome CCD cameras.
     
    RichA, Jan 20, 2011
    #2
    1. Advertisements

  3. Alan Browne

    peter Guest

    6 Mpix would provide reasonable resolution for the vast majority of work.
     
    peter, Jan 21, 2011
    #3
  4. Alan Browne

    Joe Kotroczo Guest

    6 Mpix per sensor? Marketing will call it an 18 Mpix camera...
     
    Joe Kotroczo, Jan 21, 2011
    #4
  5. Alan Browne

    Munster Guest

    What if it is madeof plastic?


    Tee Hee Hee <smirk> Ho Ho Ho!
     
    Munster, Jan 21, 2011
    #5
  6. Alan Browne

    Alan Browne Guest

    To those who understand resolution, it won't matter.

    To those who don't understand resolution, it never mattered.
     
    Alan Browne, Jan 21, 2011
    #6
  7. Alan Browne

    Ofnuts Guest

    As I understand it a classic 18Mpix bayer-matrix sensor has 4.5M red and
    and 4.5M blue photosites.
     
    Ofnuts, Jan 21, 2011
    #7
  8. Alan Browne

    Guest Guest

    it does, and it also has 9 million green, which adds up to 18 million
    spatially separate locations, or pixels.

    a 3 chip camera does not have three times as many pixels because
    they're split up onto different chips or multiple layers as with
    foveon.
     
    Guest, Jan 21, 2011
    #8
  9. Alan Browne

    Guest Guest

    there's no need to double anything.
     
    Guest, Jan 23, 2011
    #9
  10. Alan Browne

    Guest Guest

    same as with bayer.
    spatially a 6 mp 3-chip camera is the same as a 6 mp bayer sensor,
    because there are 6 million spatially separate locations on the sensor.
     
    Guest, Jan 23, 2011
    #10
  11. Alan Browne

    Guest Guest

    it calculates *colour* information. the luminance is measured at every
    spatial location. contrast and detail is not reduced either.
    now look at samples of images from that camera of subjects that move.
    there is no free lunch.
     
    Guest, Jan 23, 2011
    #11
  12. Alan Browne

    Robert Coe Guest

    : On 1/20/2011 3:57 PM, Alan Browne wrote:
    : >
    : > I see a rumor floating around that Sony have developed test versions of
    : > an A-mount (alpha) 1.5x crop camera with prism beam splitters and 3 CCD
    : > sensors (RGB).
    : >
    : > This would definitely outclass the Foveon for color properties.
    : >
    : > Resolution is not all that great (6 Mpix), but the ISO is reported to be
    : > a stunning 819,200 (13 stops over ISO 100). One could do portraits by
    : > moonlight - hand held.
    : >
    : > If they could get this up to 12 Mpix or so, or 20 Mpix or so in FF, it
    : > could be very interesting.
    : >
    : > http://photorumors.com/2011/01/14/s...-a-mount-with-3-sensors-and-iso-up-to-819-200
    :
    : 6 Mpix would provide reasonable resolution for the vast majority of work.

    Can I quote you to the editors who believe otherwise? My wife and I once
    submitted some photos to a calendar contest, and the 10Mp cameras we had at
    the time were barely able to meet the required resolution, even with no
    significant cropping.

    Bob
     
    Robert Coe, Jan 23, 2011
    #12
  13. Alan Browne

    Robert Coe Guest

    : In article <4d39a3bb$0$4858$>, Ofnuts
    :
    : > > 6 Mpix per sensor? Marketing will call it an 18 Mpix camera...
    : >
    : > As I understand it a classic 18Mpix bayer-matrix sensor has 4.5M red and
    : > and 4.5M blue photosites.
    :
    : it does, and it also has 9 million green, which adds up to 18 million
    : spatially separate locations, or pixels.
    :
    : a 3 chip camera does not have three times as many pixels because
    : they're split up onto different chips or multiple layers as with
    : foveon.

    It would seem that the most significant point is that with each color on its
    own chip, you can handle the green channel adequately without doubling the
    number of sites assigned to it.

    Bob
     
    Robert Coe, Jan 23, 2011
    #13
  14. Alan Browne

    Alan Browne Guest

    Yes.

    But with the three sensor arrangement, there will be 6 M of each R, G, & B.

    If each sensor is, say, 12 bits, then each colour element will be 36
    bits at every position in the 6 Mpixel image.

    Spatially, the resolution will be better than a Bayer RGB sensor as
    there will be no interpolation. But it would not be as good as, say, an
    10 Mpix Bayer, never mind an 18 Mpix bayer (spatially).

    I'm surprised Sony don't debut this at 10 Mpix per colour. If indeed
    the rumour is true at all.
     
    Alan Browne, Jan 23, 2011
    #14
  15. Alan Browne

    Alan Browne Guest

    That's one important advantage.

    http://upload.wikimedia.org/wikiped...chroic-prism.svg/300px-Dichroic-prism.svg.png

    In the above example, the blue seems to have the least lossy path (2
    reflections, 2 crossings): 4.

    Green: 5

    Red: 6

    (Not counting F1 in the diagram).

    There's probably different amounts of loss for crossings v. reflections.

    The physics of it probably make getting green to the least number of
    interfaces too difficult.
     
    Alan Browne, Jan 23, 2011
    #15
  16. Alan Browne

    Alan Browne Guest

    No. There is only one sample of one colour at each spatial location.
    So there are only 12 (or 14, whatever) bits of real information for that
    sampling point.

    With the split, there is 36 bits (or 42, whatever) bits of real
    information for that same sampling point.
    No. The 6 Mpix bayer creates information based on interpolation. This
    reduces contrast and detail. (Or rather: creates equivocal "information").

    Look at samples of images from the Hasselblads that takes 4 separate
    images, moving the bayer array by 1 pixel width and/or height to record
    R, then G then B at each spatial location. (A 4th exposure is for
    registration verification). The "4-shot" images show finer detail than
    the bayer shots with the same sensor.
     
    Alan Browne, Jan 23, 2011
    #16
  17. Alan Browne

    Alan Browne Guest

    Of course it is. Look at the Hasselblad results.
    No issue with moving subjects with the prism, however. A hair less
    light perhaps (1 bit worth, maybe 2), but definitely more contrast and
    fine detail.

    It really comes down to information quantity and quality. A Bayer is
    compromised in both senses but is cheaper and occupies less volume.
     
    Alan Browne, Jan 24, 2011
    #17
  18. Alan Browne

    Guest Guest

    bayer doesn't have a black filter over 50% of the pixels, so that's
    irrelevant. nice straw man though.
    unless the colours are spectrally pure (which they aren't, except in a
    lab, maybe), there will be a response on all pixels. there's also some
    overlap in the bayer filters, so even in that extreme case, there may
    be a response with the other two colours.

    in other words, in real world conditions of real world subjects, all
    pixels respond.
    nope. bayer is already resolving about where aliasing becomes an issue,
    something that affects all sensors. it can't go much higher without
    artifacts.

    the thing is that with foveon, the aliasing artifacts are not rainbow
    patterns so they're less noticeable and some people think the false
    detail is real detail. plus foveon images have more sharpening than
    bayer usually does, which gives the impression of more detail.
     
    Guest, Jan 26, 2011
    #18
  19.  
    Wolfgang Weisselberg, Jan 26, 2011
    #19
  20.  
    Wolfgang Weisselberg, Feb 3, 2011
    #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.