Kodak DCS Pro SLR/n: New 14Meg Full Frame Sensor

Discussion in 'Kodak' started by deryck lant, Feb 12, 2004.

  1. deryck  lant

    Alfred Molon Guest

    Alfred Molon, Feb 14, 2004
    #21
    1. Advertisements

  2. deryck  lant

    JPS Guest

    In message <>,
    It would take about 100 MP in a 36*24mm sensor to out-resolve the
    sharpest lenses.
    --
     
    JPS, Feb 14, 2004
    #22
    1. Advertisements

  3. deryck  lant

    KBob Guest

    Well let's see if that makes any sense. If a 14n with 14 MP can
    resolve 65 lp/mm, and if the highest resolving 35mm lenses can reach
    about 240 lp/mm, that's a factor of (240/65)^2, and that would
    actually be closer to 200 MP! There are some problems with this idea,
    however. These resolutions above 160 lp/mm are not common, and are
    mostly associated with Leica Summicrons, Micro-Nikkors and a
    (precious) few other prime lenses, and only exist over a couple stops,
    and only in the center region of the image circle. To add to this,
    only specialized B&W thin emulsion high-contrast films will reveal
    these resolutions, and practical photography is virtually impossible
    owing to the very limited latitude involved.

    You will find it challenging enough to fully utilize 65 lp/mm with
    most modern lenses when you consider these through their range of
    useful apertures, and a majority of zooms will show severe problems
    with sensors of far less resolution. I'd guess that with the
    retrofocus lenses available for 35mm, anything over about 20 MP would
    be diminishing returns for most practical purposes.

    Evidently Kodak is sticking with their non-antialiasing idea,
    believing that the blur circle of most lenses will do the job for
    them. The 14n behaves very strangely when used for IR, and this may
    be related to the problem--maybe not. Has anyone else tried this?
     
    KBob, Feb 15, 2004
    #23
  4. deryck  lant

    Alfred Molon Guest

    "Can resolve" means absolutely nothing. What matters is the MTF curve of
    the lens, sort of a frequency response curve of the lens. If the MTF is
    below a specific level at the Nyquist frequency, there won't be much
    aliasing.

    If you check the lenses MTF curves of the lenses at www.photodo.com or
    at other site you'll note that the MTF drops substantially above 40
    lp/mm. For good lenses it's probably still too high at 62 lp/mm (the
    pixel density of the Kodak 14MP CCD), so that you'd need to use a bad
    lens to have a sufficiently low MTF above 62 lp/mm.

    But anyway, I remember that most 35mm lenses have very low MTF values
    above 100 lp/mm - that would be 5 micrometer pixels, or 7200x4800 pixels
    on a 36x24mm CCD, or 34 MPixel.
     
    Alfred Molon, Feb 15, 2004
    #24
  5. deryck  lant

    JPS Guest

    In message <>,
    Alright then, I was basing this on the sharpness of one of my lenses,
    which is known to be sharp, but not the sharpest.

    When I shoot the 60-spoke sinusoidal wheel on my 10D, I get clarity down
    to about 42 to 44 pixels with almost all of my lenses, meaning the
    sensor is the limiting factor. When I put a 2x and a 1.4x TC, stacked,
    on my Canon 300mm, resolution drops to 52 pixels. From that I
    calculated that I have 2.37x more resolution in the lens than the 10D
    sensor can resolve. I only tested at one aperture, though, and I don't
    rember what it was, so I may have been able to get it a little lower
    than 52 pixels.
    Funny, you haven't mentioned telephoto primes.
    My 300mm f4L IS is only a $1100 lens, and it outresolves the 10D sensor
    by 2.37x. Up to 6.3MP * 2.37^2 = ~35MP with the 1.6x crop fact, and
    multiply that by 1.6^2 and you're at 90MP already. The f2.8 version is
    considered sharper yet, as is the 400mm F2.8L IS.
    They believe wrong, if that's what they believe. As long as any sharper
    lens is readily available, the resolution of the sensor is too low to
    overcome aliasing. 14MP in a 36*24 frame is *not* high optical
    resolution. It's actually less than the 6MP DSLRs.
    --
     
    JPS, Feb 15, 2004
    #25
  6. deryck  lant

    KBob Guest

    Yes, you're correct on this of course. And most modern lens designs
    are optimized for contrast at the expense of some resolution. This is
    especially evident with the Leica Summicrons, that now turn in lower
    ultimate resolution figures while producing better appearing pictures.
    I tested several of the newer ones and they consistently limited out
    at 160 lp/mm, noticeably inferior to the older collapsible types in
    this respect.
     
    KBob, Feb 15, 2004
    #26
  7. Which is 61.6 - 58-8 lp/mm on sensor resolution (for the 10D), so on average
    60 lp/mm. The theoretical Nyquist frequency of the 10D sensor alone is 67.8
    cycles, meaning the lenses+AAfilter+sensor form a pretty good combination.
    The theoretical Nyquist frequency of the 14n sensor is 63.3 cycles, but it
    has significant signal above that which will cause aliasing with even
    average lenses.
    Because it is not removable, the AA-filter is probably the most limiting
    factor. However, because of the AA-filter being non-removable, you are
    essentially correct that the "sensor" is limiting the resolution. I also
    base that on my experience with the very good 100mm Macro lens which gives
    me more than 76 lp/mm on film when evaluated with a 5400ppi scanner. A
    balanced combination of "sensor" and lens, seems to be able and reach
    something like 80 lp/mm on a medium 100:1 contrast feature.
    Which equates to 49.8 lp/mm. I assume the 300mm without stacked extenders
    was closer to the 44 pixel diameter, so the combined performance was further
    limited by the extenders in this case.

    SNIP
    In my view, it is the AA-filter that limits the resolution. The 14N seems to
    have (aliased) signal up to 80% of its theoretical limiting spatial
    resolution.

    SNIP
    Correct.

    Bart
     
    Bart van der Wolf, Feb 15, 2004
    #27
  8. deryck  lant

    KBob Guest

    I hope you're not forgetting that the 14n has a CCD twice as large as
    the usual 6 MP types--the sensor density is about the same, so the
    resolution at the FP should be similar and it is. All the more reason
    for giving perforamce in terms of LPH (lines per picture height).
     
    KBob, Feb 15, 2004
    #28
  9. SNIP
    I'm not forgetting that ;-) In fact, that is the reason I mentioned "on
    sensor resolution ", meaning that output magnification differences are not
    yet applied.
    picture height).

    Yes, although it makes it a bit more difficult to say anything meaningful
    about lens influence on the resulting resolution. For that, on sensor
    resolution is a simpler to grasp concept.

    Bart
     
    Bart van der Wolf, Feb 16, 2004
    #29
  10. No, I was pointing out that microlenses won't help this chip; they
    would be irrelevant. It really does pay to understand the technology.
    Extrapolating is a bit silly when there are samples at 400 and 640 on
    the Web. The ISO 400 shot I saw looked acceptable; certainly smoother
    than ISO 400 color negatives.

    Hmm. Or that they valued capturing highlights more than other digicam
    makers. Whatever.

    Anyway, that leaves the wide dynamic range; I think it's more like 14
    stops by default, looking at PhotoDesk. Reducing this to a range more
    like that of other cameras can reduce the noise, as well.
    Oh, sorry, that's what I get for shooting from the hip. I mean Michael
    Schoenfeld; see <http://www.bacall.com/MichaelSchoenfeld/Default.htm>
    and especially <http://www.pbase.com/mlsphoto/14n_samples>. A good
    photographer who is enthusiastic about his DCS 14n and even more about
    its replacement.

    And please don't judge the noise on the screen; the word is that
    prints tell a different story.
     
    Stephen H. Westin, Feb 16, 2004
    #30
  11. deryck  lant

    deryck lant Guest

    deryck lant, Feb 17, 2004
    #31
    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.