Dynamic range questions

Discussion in 'Digital Photography' started by Charles Schuler, Jan 8, 2004.

  1. 1. Is the dynamic range of Canon digital SLRs about the same? About 9
    f-stops at ISO 100?
    2. Is this about the same as 35 mm slide film?
    3. How much better is the best B&W film?
    4. Are the responses of the Canon sensors linear or do they follow an
    5. Are there any digital cameras that are clearly superior in this arena?

    Charles Schuler, Jan 8, 2004
    1. Advertisements

  2. Charles Schuler

    Robertwgross Guest

    I'm not sure where you got those numbers for dynamic range, but it will vary
    somewhat according to the brand and model.

    The numbers that I have seen lately are these:
    Canon 10D sensor has 7-8 stops of range.
    Slide film has 5-6 stops.
    Negative film has 10-11 stops.

    That last one seems a bit of a stretch to me, but then I don't shoot much
    negative film anymore. If I were forced to guess for myself, I would guess the
    numbers are 8, 6, and 9. Note that Canon uses a proprietary CMOS chip for its
    image sensor, and many other good cameras use a CCD sensor, so that will
    dictate differences in dynamic range.

    Conversion from RAW to TIF can use tricks to maximize the effective dynamic
    range, but if you shoot JPEG, you're stuck.

    ---Bob Gross---
    Robertwgross, Jan 8, 2004
    1. Advertisements

  3. Charles Schuler

    Alfred Molon Guest

    Hmmm... a DSLR should have more than 7-8 stops. Just a guess, but for
    instance the RAW files of the Olympus 5050 are already 12 bit, and a
    DSLR should have more dynamic range than that.

    The response of a CCD sensor should be linear.
    Alfred Molon, Jan 8, 2004
  4. Charles Schuler

    Pygmi Guest

    The number of bits in conversion is just a number.
    7-8 stops for modern DSLRs is something that I have seen too.
    For example for Nikon D1/D2 if I remember it correctly.
    (Yes, I know they have very different sensor types...)

    Pygmi, Jan 8, 2004
  5. Charles Schuler

    Robertwgross Guest

    I don't think that we want to confuse the sampling resolution with the dynamic

    ---Bob Gross---
    Robertwgross, Jan 8, 2004
  6. Charles Schuler

    Flycaster Guest

    I've noticed no relationship, linear or otherwise, between what I understand
    as dynamic range and color bit depth, *if* that is what you are inferring.
    Flycaster, Jan 9, 2004
  7. Charles Schuler

    JPS Guest

    In message <3ffdf098$>,
    Color bit depth puts a hard limit on dynamic range, though. You can't
    have 13 stops of dynamic range with a 12-bit ADC, even if there were no
    noise at all.
    JPS, Jan 9, 2004
  8. Charles Schuler

    Robertwgross Guest

    Are you sure about that?

    The resolution of the A-D converter simply dictates the fineness of the color

    It would be possible to have 13 stops of dynamic range (from lightest to
    darkest) with a 4-bit A-D converter, although it would not be very good

    ---Bob Gross---
    Robertwgross, Jan 9, 2004
  9. You could, actually, but the pictures would look awful. Just have 1 bit
    represent more than 1 stop. Outputting the resulting image in a way that
    resembled reality would probably be more of a challenge, however.

    - jz
    Jeff Zawrotny, Jan 9, 2004
  10. Charles Schuler

    Flycaster Guest

    That may well be (I'm not qualified to answer) but I *will* tell you one
    thing: Give me a DSLR with a true dynamic range of 10 stops (no Photoshop
    "tweaks" required) and I would be a *very* happy camper. My D60 can't do
    this, that I know. Even with some fairly knowledgable PS work on my part
    using RAW files, about the best I can do is maybe, just maybe, 7 *effective*

    Just out of curiosity, how is everyone quantifying this number, or is it
    just an educated guess as it is on my part?
    Flycaster, Jan 9, 2004
  11. Charles Schuler

    JPS Guest

    In message <>,
    Now you're talking about a different number-space. You could use a
    gamma-adjusted number system, I suppose, but that usually isn't done in
    the RAW data.
    JPS, Jan 9, 2004
  12. Charles Schuler

    Don Stauffer Guest

    There are two different types of dynamic range, with two different

    One definition is ratio of signal from brightest area that is
    unsaturated (highest value in digitizer output) to black level value.
    This is a large signal dynamic range.

    The other is the ratio of noise level to mid-level exposures to the
    largest unsaturated value. This is essentially a signal-to-noise ratio

    Film is very low noise for larger areas of exposure, but the fog level
    for unexposed areas limits the large signal dynamic range for some
    films. That is, not all films are equal when it comes to dynamic
    range. Many have quite a curved response curve.

    Dynamic range of CCD cameras are determined by capacitor well size.
    Well when full contains a certain number of electrons. Smallest change
    in signal one can see is one photo-electron. So number of electrons in
    full well is essentially dynamic range. Larger CCDs (larger pixels,
    anyway) tend to have a larger dynamic range than smaller ones because of

    Response curve of CCD itself is very linear, probably more so than most
    films. However, software in camera frequently fits final output to a
    non-linear curve.

    Does this look complex, when you maybe expected straight-forward
    answers? Probably so, but the subject of dynamic range and signal/noise
    in electronic cameras is indeed a very complicated subject.

    Keep in mind that printed photos, regardless of whether from film or
    digital file, are MUCH more limited in dynamic range than either film or
    digicam. However, having lots of dynamic range is still good, as you
    can select which areas of image to print. That is, printing of digital
    images must make same creative choice as film- print for shadow detail?
    print for highlights? Compress contrast and try to fit all areas into
    final print?
    Don Stauffer, Jan 9, 2004
  13. Charles Schuler

    jpc Guest

    A few years back when I had just bought my oly 3020Z I set up this
    experiment. I created a test scene: a white box with a dark towel
    inside it placed on two sheets of poster paper, white and dark red, a
    black ashtray on the border between the two sheet of poster board and
    a wratten fiter envelop in front of the box. I also added a bunch of
    other object to give me a range of diifferent reflectivities in areas
    of different illumination.

    I then set up the camera and illumination so that with a ten second
    exposure the brightest part of the white box was just going into
    saturation. I then took a series of pictures while varing the shutter
    speeds from 10 to 1/800 of a sec. Finally I usied the levels
    adjustments in photoshop to see at what exposures the various objects
    disappeared into the noise.

    At 1/100 of a sec or 1/1000 the correct exposure I still had a
    recognizable picture. I could see the white box, the highlights on the
    edge of the dark towel and the outlines of the wratten filter envelop
    and ashtray. And at 1/400 of a sec or 1/4000 the correct exposure I
    could still see the variations of illumination on the top of the white
    box althought at this point if I hadn't known where to look I might
    have said I was looking at only noise.

    So if you allow me to stretch the defination of dynamic range a bit I
    can say my camera has a dynamic range of 12 stops since I can extract
    pictorial information over a 4000 to 1 range of exposures.

    As for qualifying a more normal defination of dynamic range, besides
    that of the camera you have to consider the dynamic range of your
    output device, say 7 stops on your typically slightly misadjusted
    monitor, and mayber 9 or 10 stops if you use a very good printer, inks
    and paper.

    Finally there is your viewing device, your eyeballs. I've never been
    able to see a difference between zone 0 and 1 on a zone system test
    target, and now after digging out my book on the zone system I'm hard
    pressed to say I can see the difference between zone 1 and 2 even
    under ideal illumination.

    So a ten stop camera/ print system would be wasted on me since I have
    eight stop eyeballs.

    jpc, Jan 9, 2004
  14. Charles Schuler

    Robertwgross Guest

    This seems to address digital cameras that do have a CCD sensor. Many cameras,
    such as the Canon 10D, do not have a CCD sensor.

    ---Bob Gross---
    Robertwgross, Jan 9, 2004
  15. Charles Schuler

    Alfred Molon Guest

    Well, no you have to divide the max. capacity by the noise floor, i.e.
    if a cell can contain 50000 electrons and the noise floor is 10
    electrons, the dynamic range is 5000:1 or approximately 72 dB.
    Alfred Molon, Jan 9, 2004
  16. Charles Schuler

    jpc Guest

    Ignoring the thermal shot noise there are two main source of noise in
    both CMOS and CCD cameras, the electronic noise or noise floor and the
    photon shot noise or quantum noise do to the random emission of photos
    by the light source. The quantum noise is proportional to the sqrt of
    the number of electron in the well. With 50000 electrons in the well,
    with by the way would reguire very large pixels, the S/N ratio would
    be about 220 not 5000

    jpc, Jan 9, 2004
  17. Charles Schuler

    Alfred Molon Guest

    How do you get 220 ?
    Alfred Molon, Jan 9, 2004
  18. Charles Schuler

    jpc Guest

    Square root of 50,000 minus a little because of the noise from the

    jpc, Jan 9, 2004
  19. Charles Schuler

    Alfred Molon Guest

    Looks like we have different definitions of dynamic range. It should be
    the ratio of max. signal and noise floor, not a square root.
    Alfred Molon, Jan 9, 2004
  20. Sqrt(50000)=224 (=0.4%), and it varies with the number of photons so
    with 50 photons it would be 7 (=14%) near the noise floor. But photon
    noise is not a characteristic of the camera.

    As you said, the camera dynamic range is often expressed as the ratio
    of maximum well depth, say 50000, and noise floor which could be
    something like 20-50, depending on many factors. That does assume an
    ADC and supporting electronics that have adequate capabilities.

    Bart van der Wolf, Jan 9, 2004
    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.