The issue of fall-off on full-frame digital sensors

Discussion in 'Digital SLR' started by DD, Nov 23, 2005.

  1. DD

    DD Guest

    I don't know too much about the technology that goes into a DSLR, but I
    have read a bit about how the light hitting the sensors at angles and
    causing fall-off is the real reason why companies like Nikon have not
    pursued the full 35mm frame sensor route.

    Would it not be possible, using in-camera software, to correct the
    problem of fall off in much the same way as using a dust-ref photo
    corrects for dust on the DSLR sensor?

    Example; you put a 20mm lens onto your DSLR and shoot a picture of a
    uniform surface such as the sky. The software can then apply a
    correction to the sensor by amplifying the receptors that are on the
    fringes of the frame to the same level as those in the "DX" region. The
    camera stores the information for this lens and when you re-attach it,
    in the case of a CPU lens it will know what corrections to apply. In the
    case of the non-CPU lens you simply recall it from camera memory in much
    the same way as the D2X and D200 do at present when using manual focus
    lenses for matrix metering.

    Would something like this work? Could it be the method Canon have
    adopted on their full frame DSLRs?
     
    DD, Nov 23, 2005
    #1
    1. Advertisements

  2. DD

    223rem Guest

    The problem is that the signal received by photoreceptors
    off the optical axis decreases with increased eccentricity.
    Amplification cannot selectively increase the signal, it will
    also amplify the noise. You'll replace falloff with increased
    noise/signal ratio towards periphery.
     
    223rem, Nov 23, 2005
    #2
    1. Advertisements

  3. DD

    DD Guest

    I hear what you're saying about the haphazzardness of the fall off, but
    surely that too would be quantifiable and not totally random over a
    constant area, such as a photo receptor?

    Could the noise issue not also be addressed with software?
     
    DD, Nov 23, 2005
    #3
  4. DD

    Colin D Guest

    Light fall-off, or vignetting, varies to some extent with aperture as
    well, so the degree of darkening isn't constant. A number of raw
    converters and image processors will compensate for fall-off, and there
    are a number of plugins for PS that do the same. I think that is
    probably preferable to the camera doing it, and if you're shooting raw,
    the camera won't do it anyway.

    I have started to use DxO, pricy but it seems good, it corrects for lens
    distortions, lateral color, noise, and vignetting automatically, from
    lens look-up tables according to the exif data for the image. It
    converts directly from raw to tiff or jpg, your choice, and does about
    one frame per 20 or so seconds on a 3.00 GHz machine. Will convert an
    average shoot while you make a cup of coffee. just FYI.

    Colin D.
     
    Colin D, Nov 23, 2005
    #4
  5. DD

    DD Guest

    I have downloaded one of the DxO modules but haven't installed it yet. I
    might do so at some point during the next month or so.

    My thoughts on the technical problems with full frame sensors and light
    scatterings are that there are a limited amount of variables to be
    considered for it to be made technically feasible to address each of
    them with some form of in-camera data processing. On the aperture issue
    for instance, that data can be gained for each aperture value of a given
    lens at a given ISO. Simply plugging the lens onto the flange would
    instantly transmit those parameters to the cameras CPU, thus allowing it
    to compensate for whatever variations in light intensity are needed for
    the lens in question.

    Sure it will probably be a lot of kludge work, but is it *that* far-
    fetched an idea?
     
    DD, Nov 23, 2005
    #5
  6. DD

    Jeremy Nixon Guest

    The falloff isn't the main issue; that happens anyway, even with film.
    Light hitting the sensor at an angle causes softness and can lead to
    problems with color channel registration. This is an entirely separate
    issue; it's due to the sensor, while light falloff is due to the lens.
    The falloff will vary with each lens, and with both aperture and focus
    distance within each lens, as well as focal length for zooms. Yes,
    it can be corrected for, if it's bad enough to be noticed. It's usually
    not that big a deal, and is easy to correct visually if it happens.
     
    Jeremy Nixon, Nov 23, 2005
    #6
  7. DD

    colinco Guest

    Simply plugging the lens onto the flange would
    instantly transmit those parameters to the cameras CPU, thus allowing it
    to compensate for whatever variations in light intensity are needed for
    the lens in question.
    [/QUOTE]
    Th Olympus E lenses communicate to the bodies for corner shading
    compensation.
     
    colinco, Nov 23, 2005
    #7
  8. DD

    nv Guest

    In
    yes, but how 'cheaply', please?
     
    nv, Nov 23, 2005
    #8
  9. DD

    bmoag Guest

    Before you trust Dx0 completely with your images, a la jpeg processing, you
    should analyze what it is doing and see if you agree with the "corrections."
    While Dx0 can seem a convenient all in one solution you should examine
    exactly what it is doing with regard to barrel/pincushion distortion and
    chromatic aberration. Rigid application of a "look-up" table may not be
    appropriate for the image at hand. And why shoot raw, instead of jpeg, if
    you are going to hand over image processing chores to the rigid algorithms
    of a piece of software? I tried Dx0 with a zoom whose distortion
    characteristics I know and while it was interesting it was hardly worth $150
    for what I can do with two clicks of the mouse in CS2 (which costs the same
    and is obviously more useful). However if Dx0 was priced like the plug-in it
    actually is, say $50, I would buy it for occasional use.
     
    bmoag, Nov 23, 2005
    #9
  10. DD

    Skip M Guest

    You may have hit on Canon's next "killer ap" that gets all of us with FF
    bodies to upgrade, since there's really no need to keep up with the pixel
    race anymore.
    I'm not kidding, Dallas, with my limited knowledge of electronics and
    programming, I see no reason that such a thing couldn't be done, just like
    noise reduction, in the processor.
     
    Skip M, Nov 23, 2005
    #10
  11. DD

    Kyle Jones Guest

    Yes, and they do warn about resulting noise in the manual, so the camera
    is likely applying some aplicfication curve to the corners. None of the
    lenses I have exhibit any light falloff that I can see so I haven't been
    able to test it.
     
    Kyle Jones, Nov 23, 2005
    #11
  12. DD

    Jeremy Nixon Guest

    I don't know what you mean.

    How much does it cost to adjust a couple of settings in Camera Raw?
     
    Jeremy Nixon, Nov 23, 2005
    #12
  13. DD

    Rich Guest

    If you want to spend the money, you can buy radial gradient filters to
    counteract this. Trouble is you'd need a whole bunch of them.
    -Rich
     
    Rich, Nov 23, 2005
    #13
  14. DD

    Colin D Guest

    I don't quite follow what you're saying here. DxO isn't a plugin, it's
    a complete stand-alone package. It's not a matter of trusting DxO with
    my images, as it does not alter in any way the original RAW files. It
    also gives me the option of saving the processed file as jpg, tif, or
    dng, and the distortion/CA corrections are specific not only to the
    particular lens, but to the focal length a zoom is set to, from the exif
    data (I have the profiles for three of my most-used lenses). I have
    checked the straightness of things like edges of buildings and the
    corrected images are pretty well spot on, and CA is corrected pretty
    well, except maybe for things like tree branches against bright sky, but
    that phenomenon is usually sensor blooming and not CA as such.

    RAW files can be bottle-necked by the conversion software, and that is
    probably the main reason why jpeg is used, despite the limited afterwork
    and correction for exposure and color balance that can be done compared
    with RAW processing. RAW remains the most versatile form of image
    capture we have, and for me at least DxO after-processing of RAW files
    gives me the best quality/time outcome I can get, and certainly the jpeg
    output from DxO is far better than the camera-produced jpeg files.

    Other DxO capabilities are adjusting the 'lighting', i.e. the histogram
    for optimal tonal range, USM sharpening, highlight recovery from
    overexposed RAW files - if one or more channels are blown, it uses
    whatever highlight data remains in the least blown channel to correct
    the image, and it does work.

    All these features are optional, you can turn off any you don't want,
    though the automatic mode produces excellent images from RAW files at
    about one per 15 seconds or so, and they are as good as I can get from
    PS one at a time.

    As far as 'handing over image processing chores to the rigid algorithms
    of a piece of software', what do you think the jpeg conversion in the
    camera is doing, if not precisely that? The algorithms in DxO are far
    and away more complex than could be coded into the limited firmware of a
    digital camera, and if DxO takes about 15 seconds to process a RAW file
    on a computer with a gig of ram and a 3.00 gigaHertz processor, how long
    would it take in a power-limited camera, and what would it do to
    shooting speeds? There is just no comparison between in-camera
    processing of jpegs and DxO.

    Your quote 'it was hardly worth $150 for what I can do with two clicks
    of the mouse in CS2 (which costs the same and is obviously more useful)'
    surprised me somewhat. Where can you buy CS2 for $150? Maybe that
    figure is for educational pricing in the States, but I doubt even that.
    Educational pricing here in New Zealand is over $nz 400, and the retail
    version is about $nz 1,300. That's a helluva lot dearer than DxO.

    Colin D.
     
    Colin D, Nov 24, 2005
    #14
  15. DD

    Stacey Guest

    Skip M wrote:

    I thought you said it would be YEARS before you even considered upgrading?
    It's been what, less than a month and you're already thinking of reasons to
    replace your 5D? And you think you're going to make it a year without doing
    it? :)

    I still got money that this time next year you'll be telling us how great
    your 7D is..
     
    Stacey, Nov 24, 2005
    #15
  16. DD

    wilt Guest

    I hear what you're saying about the haphazzardness of the fall off, but
    surely that too would be quantifiable and not totally random over a
    constant area, such as a photo receptor? Could the noise issue not also
    be addressed with software? <<

    The problem is 'noise' is mixed in with th real 'signal' and not easily
    suppressed. In you have TWO frames of an image, 'signal' is fixed and
    constant while 'noise' is random. So you can compare two frames and
    keep the constant and filter out the random. But since you have only
    ONE frame, not two, in the typical photographic image, removal of noise
    remains problematic.

    The issue for Nikon is NOT the conscious decision to not go FF because
    of noise, but simply that few companies even can make a FF sensor, and
    Nikon is not one of them yut!
     
    wilt, Nov 24, 2005
    #16
  17. DD

    Skip M Guest

    I'm speaking in general terms. One software upgrade like that isn't going
    to get me to switch. And I think it's pretty safe to say that Canon won't
    include such an ap in the next generation of FF bodies.
    My 5D has the two things that I've long said I missed about the 1.6x crop
    bodies, spot meter and 35mm film-sized sensor. 12+mp is enough res for my
    purposes, in fact 8mp was. So, why change?
    Like I said, you might as well send me that $20 now. ;-)
     
    Skip M, Nov 24, 2005
    #17
  18. DD

    Tony Polson Guest


    No, it will be the EOS 3D, with the 16 MP sensor from the 1Ds Mk II,
    but in a lighter more compact body with the same standard of
    environmental sealing as in the EOS 3 film SLR.

    It will cost what the 5D costs now. That's the 'killer app' for me.

    ;-)
     
    Tony Polson, Nov 24, 2005
    #18
  19. DD

    Stacey Guest

    Skip M wrote:

    Nah, when they have a 16MP body with better firmware in a "20D" type body
    that everyone is raving about, you'll be getting out the visa card again..
    We still have a LONG way to go till you "win" that one..
     
    Stacey, Nov 25, 2005
    #19
  20. DD

    Tony Polson Guest


    So will I, and so will literally millions of other people.
     
    Tony Polson, Nov 25, 2005
    #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.