20D or 5D

Discussion in 'Digital Photography' started by Rob, Aug 23, 2005.

  1. Rob

    Mark² Guest

    I guess in my brain it goes back to my original question...of whether the
    sensor itself is more sensitive, or if they are just going through the
    process described.
    If it's truly more sensitive, you could push that true 50 IS0 and retain the
    same quality as pushed 100 on other sensors (or there abouts...).

    Don't mind me... My brain is fried today.
    Mark², Aug 27, 2005
    1. Advertisements

  2. Rob

    JPS Guest

    In message <[email protected]>,
    Is "sensitive" the word used to describe low native ISO? That sounds
    kind of counter-intuitive to me. Sensitive should mean quality
    recording at high ISOs, IMO.

    Anyway, we're not talking about a 1-stop ND filter; we're talking about
    actually shooting at ISO 50, with reduced headroom. That's not really a
    lot to ask for, from many DSLRs, as they already have quite a bit of
    headroom to begin with.

    I just placed my Gretag-Macbeth Color Checker card against the wall,
    after incident metering for ISO 32. I then shot the color checker with
    my Canon 10D in manual mode with those settings (1s at f/1.4) with the
    camera set to ISO 100. The brightest square (the white one) registered
    no higher than 3787 anywhere in the RAW data. Clipping in the 10D set
    to ISO 100 occurs anywhere from 3997 to 4006, depending upon the
    vertical line of pixels a pixel is in (strange but true). I had no
    problems using Adobe Camera RAW getting an image that looked like the

    The guy in the other posts who was demanding "100% reflectance" really
    had no idea how undemanding his demand was!

    The reason why digital images clip if you're not careful is not that
    there is little headroom, per se; it's because the metering has no idea
    where "middle grey" is in the image, and will unknowingly pull the image
    to ISOs that are too low to salvage the image. JPEG is more
    unsalvageable than RAW.

    I have no idea what context you're working in here.

    JPS, Aug 27, 2005
    1. Advertisements

  3. Rob

    Jeremy Nixon Guest

    ISO 50 would be less "sensitive", not more. But it would mean that they
    were able to pull usable data from the signal a full stop lower than
    before (more like lower signal to noise ratio than higher sensitivity),
    which would be a good thing -- but if that were the case, they wouldn't
    hide ISO 50 behind a custom function, they would advertise it.
    Jeremy Nixon, Aug 27, 2005
  4. Rob

    JPS Guest

    In message <>,
    Well, you have to know what you're doing to take proper advantage of it.
    The amount of headroom available at ISO 100 is actually plenty, to the
    person who meters for 100 with an incident meter or grey card, or uses a
    "sunny-f/16" approach in bright daylight. Exposures vary much more
    wildly with automatic exposure, due to the weighted approach used for
    exposure, and the firmware-pulled ISO 50 is more likely to lead to blown
    JPS, Aug 27, 2005
  5. Rob

    Mark² Guest

    Sorry guys..

    I had several related thoughts in my head at the same time which I didn't
    connect at all well in my writing. I was...and am...to describe my thought
    well... Perhaps I'll return when I'm coherent.

    I've been working on a still-image video creation for the last 5 days,
    almost without stopping.
    My brain is mush, and the various "breaks" I've taken to scan the NG haven't
    left me in a good mind-set for accurate thought communication...

    Ignore my last several posts until I can explain what the heck I was
    referring to.
    Believe it or not...it related to something that made sense...
    I just didn't...well...make sense.
    Mark², Aug 27, 2005
  6. So, you are saying that Canon cameras can be overexposed by one stop at the
    lowest ISO setting?

    This seems to contradict the measurements of Roger N. Clark
    (in http://clarkvision.com/imagedetail/dynamicrange2/)
    who saturates the sensor of a 1D mk II at +0.3 stop.
    Philip Homburg, Aug 27, 2005
  7. Rob

    JPS Guest

    In message <[email protected]_id.stereo.hq.phicoh.net>,
    If you're using a grey card or incident meter, yes. If you're just
    setting the EC to +1 and blindly pointing it at mixed sun/shade scenes,
    then you will blow the RAW highlights on some images. Different models
    vary in their headroom. My 10D has far more headroom than my 20D in the
    RAW data at ISO 100 with both cameras metered externally and manually
    (same f-stop and shutter speed). The 20D makes an exception to its
    general linearity of RAW data in the "extra" highlights at ISO 100 only,
    which is missing about 0.2 stops, which it compensates for by ramping
    the top zone to make it reach 4095, while keeping the midtones and
    shadows linear.
    I don't see any real Exposure Index vs RAW value data there, at various
    ISOs. All his output seems to be gamma-adjusted and white-balanced,
    which is quite an abstraction from the RAW data. If you apply just a
    gamma adjustment to RAW data, everything is very dark; it is
    conventional to clip away RAW highlight in each channel, based
    relatively, between the channels, on white-balancing.

    I'd have to have a 1D mkII in my own possession to see what it's really
    all about. There is a dearth of knowledge about all this stuff, as
    people tend to draw conclusions from abstractions like JPEGs and TIFFs.
    I'd like to see a library of RAW data based on external metering of
    standard targets like the Kodak grey wedge and the Gretag-MacBeth Color
    checker, or even blackframes, but everyone seems to treat RAW data like
    a black box that only Adobe, David Coffin, and Michael Tapes, and the
    camera manufacturers have any business looking at.
    JPS, Aug 27, 2005
  8. Rob

    JPS Guest

    In message <>, I,
    Here's a visual example. This color checker was shot in manual mode
    with the Sekonic Incident readings for ISO 100, with the 10D set to ISO
    100, on my terrace in blue-sky shade (making the blue channel very
    strong and the red channel very weak). The color channels of this RAW
    data were interpolated to full-res before downsizing the image, to get
    rid of the checker-board pattern in un-interpolated RAW data. The RAW
    linear range 0 to 4095 is mapped to 0 to 255 in the JPEG, without gamma
    correction. See how dark the white square is:


    It has well over a stop of headroom. One I did last night in
    incandescent like didn't clip with the incident meter set to ISO 32,
    with the 10D set to ISO 100. The meter perhaps has a yellow or green
    JPS, Aug 27, 2005
  9. Before trying all kinds of weird color temperatures, it might be useful
    to start with more common ones, such as D50.

    I don't have absolutely no idea how a Sekonic lightmeter responds to overly
    blue light.
    Philip Homburg, Aug 27, 2005
  10. Rob

    JPS Guest

    In message <[email protected]_id.stereo.hq.phicoh.net>,
    What is "weird" about daylight shade, or warm indoor lighting?
    I've never heard that terminology before. Is that sunlight?
    Am I the first to use one in shade?
    JPS, Aug 28, 2005
  11. Rob

    JPS Guest

    In message <[email protected]_id.stereo.hq.phicoh.net>,
    The color of the white square in the JPEG is not the color of the light,
    if that's what you thought. Real-world white objects in white light are
    cyan to green in most CFA DSLRs.
    JPS, Aug 28, 2005
  12. It doesn't matter. It was specified that the shot was not taken in
    normal daylight.
    Philip Homburg, Aug 28, 2005
  13. Yes, a reference 5000K sunlight.
    Probably not. But in the film world, you add a color conversion filter and
    you add some stops for the loss of light. Usually it is smart to bracket
    in those conditions. But you may get away with holding the CC filter in
    front of your lightmeter's sensor.

    ISO is defined at a specific color temperature (usually 5000K). In all other
    situations (without CC filters) you may end up with completely different ISO
    Philip Homburg, Aug 28, 2005
  14. Rob

    JPS Guest

    In message <[email protected]_id.stereo.hq.phicoh.net>,
    How much of a difference do you think that would make? Between my two
    extremes (incandescent and shade), the ISOs as measured for
    reflectance-only for the white square was perhaps ISO 30 vs 40, both far
    from ISO 100.

    I just tried it now with my 550EX flash, in manual mode, based on the
    guide number; f/22 at 8.23 feet from flash to target with 105mm zoom.
    The green channel predominates here, with blue at about 80% of the green
    RAW values, and red at about 50%. The average green value in the white
    square is about 1520. The blackpoint of the 10D at ISO 100 is about
    126, so that leaves us 1494 RAW values for the 90% reflectance square
    That would put 100% at 1660 above blackpoint. The 10D clips at ISO 100
    anywhere from 3997 to 4005 or so (depending on pixel column), so the
    maximum above blackpoint is about 3875. 1660/3875 = 0.43, or ISO 43, if
    the standard is 100% reflectance. Now, I shot this in a white hallway
    28" wide from 8.2 feet, so there is some extra light bouncing off the
    walls, so the actual value is actually lower, although the practical
    value in a narrow white hallway is about right.
    JPS, Aug 28, 2005
  15. http://www.brucelindbloom.com/CIESpectralCalculator.html
    graphically shows the spectral contribution of several "Reference
    illuminants" as per CIE.
    D50 is somewhat similar to a light source of 5000 Kelvin color
    temperature, a bit warmer than 'average' daylight, the native color
    temperature of calibration target references.

    Bart van der Wolf, Aug 28, 2005
  16. First of all, if you want to demonstrate a 1 stop headroom, you simply
    overexpose a stepwedge by one stop and compare it to a normally exposed

    Playing the numbers game doesn't say much.

    I don't think that playing with flashes and relying on the guide number is
    a good way to get repeatable results.
    Philip Homburg, Aug 29, 2005
  17. Rob

    JPS Guest

    In message <[email protected]_id.stereo.hq.phicoh.net>,
    Based on what reference? A stop more of what? The camera's metering?
    An incident meter? One of the grey squares?
    RAW data *is* numbers.
    I suppose it's just a coincidence that my Sekonic meter, sunny f/16, and
    the 550EX in manual mode all give very similar results?

    If you think a camera's metering is more accurate than the GN of a
    speedlight flash, you put a lot of faith in the wrong things, IMO.
    JPS, Aug 30, 2005
  18. No, I think that a standalone light meter is designed to be accurate (around
    Philip Homburg, Aug 30, 2005
  19. Rob

    JPS Guest

    In message <[email protected]_id.stereo.hq.phicoh.net>,
    And just how much do you think the flash, or shade, would affect the
    reading? No frequency should vary more than about 1/3 stop between
    "D50", and the meter isn't even narrow-band by any measure, just
    slightly weighted.

    What maximum RAW channel value do you think I would get on the 10D if I
    had a D50 light source and an accurate meter calibrated to it, for 100%

    Under no lighting situation that I've ever used my 10D with external
    metering (incident or grey card), has 100% reflectance come anywhere
    near the RAW maximum.

    I state with 100% confidence that you can set your external meter to ISO
    50 with the 10D set to ISO 50 (or 20D set to ISO 64), if you use
    absolute exposure, and your brightest highlights are matte-reflective
    only. If you need to capture detail in glare, or specular highlights,
    then you might not get away with it (although the red and blue channels
    will still have headroom to record greyscale highlights). Reflectance
    of matte whites is usually 90% or less, and you may be able to push the
    exposure more in some lighting.

    Remember, the reputation that some DSLRs for blowing out highlights come
    foremost from JPEGs that clip away RAW highlights, and even when the RAW
    data itself is clipped, it is because the camera's AE increased the
    exposure by a few stops because of weighted averages from the metering
    (like a sliver of sky surrounded by shady city buildings on both sides
    of the street). Incident metering rarely blows highlights and often
    leaves headroom unused.
    JPS, Aug 30, 2005
  20. Rob

    Stacey Guest

    Yet were designed with film use in mind?
    Maybe 1.6X crop models?
    Bingo and yet even then they aren't "stellar" preformers even on those
    Stacey, Aug 31, 2005
    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.