Technical - A gray card with 1 stop steps

Discussion in 'Digital Photography' started by Lionel, Feb 1, 2004.

  1. Lionel

    Lionel Guest

    Due to some of the interesting discussions we've had here about the
    dynamic range digital image sensors & film, I'd like to do some
    experimenting. What I'd like to do is shoot a target with a series of
    gray patches, ranging from black to white in *accurate* 1 stop (or 1/2
    stop) steps. The wider the range, the better. Ideally at least 10 stops,
    but the more, the better.
    Any ideas on how I might obtain such a target? (As cheaply as possible!)
     
    Lionel, Feb 1, 2004
    #1
    1. Advertisements

  2. Lionel

    Stewy Guest

    I had a test chart made by Kodak for balancing colour film (you photograph
    the first frame with the chart.
    This and a colour wheel is at a friends High School at the moment but I
    could try and get it.

    Another way would be to make your own test strips in Word or Excel (they
    have various percentages of grey in their text formatting dialog box)

    Let me know if you need a jpeg scan of the Kodak chart

    (the Lycos address is a spam trap - reply here)
     
    Stewy, Feb 1, 2004
    #2
    1. Advertisements

  3. Kodak makes gray scale targets (Kodak Color Separation Guide and Gray Scale
    (Small) CAT 152 7654, there's also a large one). Amazon.com may have it if
    you can't find it locally.

    The bad news is that you do not get accurate 1/2 or 1 stop steps, because
    the full scale contrast varies with the level of the incident illumination,
    i.e. the brighter the lighting, the larger the range from white to black.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Feb 1, 2004
    #3
  4. Lionel

    Paul H. Guest

    Get a can of flat white enamel (titanium dioxide or similar), a can of flat
    black enamel, and a bunch of small paper cups. First make three cups of
    paint,

    A. white
    B. equal parts white and black
    C. black

    Second iteration

    a. A, above
    b. equal parts from A and B
    c. B, from above
    d. equal parts from B and C, from above
    e. C, from above

    and so on, until you get a number of samples greater than or equal to the
    number you want [ 2^N+1, where N is the iteration], then paint your patches.
    Of course, you can't have an *arbitrary* number of patches between pure
    white and pure black, each 1-stop apart from its nearest neighbors. :)
     
    Paul H., Feb 1, 2004
    #4
  5. []
    Are you saying this is a problem because the ambient light varies?

    Surely, within one exposure the 1-stop (or whatever) steps _must_ be
    accurate, because they are ratios, not absolute values.....

    Cheers,
    David
     
    David J Taylor, Feb 1, 2004
    #5
  6. I'm not sure it's a problem, but it seems that the contrast in a scene
    varies with the intensity of the illumination.
    Try spot metering a gray scale card in subdued room light, normal reading
    light, and close up under a bright spotlight.

    Truth in advertising: I was surprised when I tried it. Maybe one of our math
    whizzes can explain it...

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Feb 1, 2004
    #6
  7. Well, if you are seeing errors doing that, it sounds to me like:

    - the reflectivity is sensitive to the direction of the applied light
    (likely).

    - the card is sensitive to colour temperature, i.e. the greys are not true
    grey (but they never are, of course....).

    - your spot meter is non-linear (perhaps check it with ND filters?).

    Cheers,
    David
     
    David J Taylor, Feb 1, 2004
    #7
  8. Lionel

    Lionel Guest

    Very kind of you to offer to go to so much trouble, Stewy, but what I
    need is an accurately rendered reference. It would be very simple for me
    to print such a strip myself, but the tonal accuracy would be totally
    dependent on the printer, print drivers, etc.
     
    Lionel, Feb 1, 2004
    #8
  9. Lionel

    Lionel Guest

    [etc]

    Excellent idea, Paul. I should've thought've that myself.
     
    Lionel, Feb 1, 2004
    #9
  10. You cannot get 10 stops - that would mean a density of ~ 0.2 - 3.2.
    Blacker than 2.0 is hard to get for reflecting surfaces. So - the
    best target would maximally have ~ 6 stops.

    An *accurate* target is *expensive*. I don't even think they make one
    with *accurate* 1 stops. What they do is making a target the best
    they can and then they callibrate it, giving you the actual meassured
    reflectance.

    Even a callibrated target needs to be carefully used. How you lit the
    target affects the reflected light. The light needs to be even. The
    target is probabaly glossy (to get maximum blackness) so the light must
    come from a large angle. Etc ...

    What is dynamic range for a camera? Very hard to define! There are
    several problems. Ambient light hitting the lens decreases the contrast,
    and thereby increasing the dynamic range. So - shall everything except
    the target be black? Nope - that is not how common scenes look. The best
    is probabaly that everything, except the target, is medium gray. And
    what is the darkest things you can record? Small detail will be harder
    to resolve if they are dark, but larger areas is rather easy. So - the
    dynamic range for resolving things is higher for large things. But,
    there is another definition, and the most common one. You don't care
    about resolving power, you only care about the dynamic range for large
    patches. Then you get the highest value.

    Taking a photo of a reflecting target is not a good method for
    meassuring dynamic range. You shall use a transparent target. There
    exist transparent targets, but they are small. So you must do your
    experiments in macro mode. You have to use a very even light source.
    This is rather hard to do really.

    There is an easier method, one that is commonly used, but I think
    this metod is bogus. You can take several pictures with varying exposure
    of the same target. This will *not* meassure the dynamic range of
    the camera. Possibly it meassures the dynamic range of the sensor.
    But - the camera might do some strange compensations when you change
    the exposure time and the lens has not enough stops to do the experiment.

    It is not easy to meassure, or even define, dynamic range.


    /Roland
     
    Roland Karlsson, Feb 1, 2004
    #10
  11. possible!)

    What you want is a Kodak Greyscale. Check any professional photo
    store, they should have some in stock. You can also get colour
    swatches from Kodak (Colour Control Patches) as well. My grey card
    has both of these glued to it, and it was required to the first shot
    on every roll of film when I was in college.
     
    The Black Sheep, Feb 1, 2004
    #11
  12. Lionel

    fake name Guest

    very interesting question Lionel, and I am suprised no one came even
    close to the correct answer (in my humble opinion, of course).

    here is how I'd do this experiment: use *transmission* insted of
    reflection!

    So, find a gray film with 50% transmissin. You could even measure the
    transmission with your camera's spot meter, the exposure difference
    should be exactly one stop when you put this film in front of the lens.

    A camera with a needle rather than a LCD indicator would be better,
    since with the LCD you only know if you are off by 1/2, 1 or 3/2 stops,
    nothing in between.

    to make it even more accurate, put 4 layers, make sure your pucture is
    *overexposed* by 2 stops without film and *underexposed* by 2 stops when
    you put the film in front of the lens.

    You might try to do this film using a inkjet printer and tranparencies
    (for presentations). try different gray levels in word or photoshop.

    Good, no you have the 50% (1 stop) film (plastic foil).


    take a piece of flat glass (or use one of your livingroom windows, if
    you're single, that is, otherwise you might have trouble with your wife,
    LOL :)

    draw a square, let's say 10x10 cm.

    now put a white paper 10x10 cm to make sure you get a uniform light
    pattern (otherwise you would see what's behind the window, trees,
    houses, etc).


    take a piece of your 50% film 9x10 cm, put in onto the white paper.

    you have now a small strip of let's say 0 stop and a big strip of (9x10)
    of let's say -1 stop.

    take another piece of film, 8x10, lay it ober the previous film, so that
    now you have 0, -1 and -2 stops.

    keep going, with smaller and smaller film strips, until the last one is
    1X10 cm, and that would be your -9 stop strip.

    There is theoretically no limit to how many 50% film layers you can
    overlap, so you could even go to -20 stops if needed!


    How does it sound? let us know how it worked. BTW: the deadline is 6 PM
    ET, before the superbowl starts :)



    Good Luck!
     
    fake name, Feb 1, 2004
    #12
  13. Lionel

    Mark Herring Guest

    I'm guessing flare, scattered light, or equivalent. If you measure a
    reflectance ratio in low light, and then turn up the light, the effect
    of any flare or scattering is going to change the **apparent** ratio.


    **************************
    Mark Herring, Pasadena, Calif.
    Private e-mail: Just say no to "No".
     
    Mark Herring, Feb 1, 2004
    #13
  14. Lionel

    jpc Guest


    I've done this using a Kodak photographic step tablet no 2. It has 21
    transmission steps where the optical density ranges from .05 to 3.0
    in transmittance--your 10 stops

    To use it, you set up a uniformly illuminated background.--a white
    gift box in diffuse sunlight worked well for me. Then you set up a
    shield to mask the target and cut out any reflected illumination--in
    my case I made a box from black poster board. Then you photograph the
    step tablet against the illuminated background, collect your
    transmittance values and plot out your curves with your favorite
    spreadsheet or plot program.

    The best program I've found to get the transmittance values is ImageJ.
    It a freeware scientfic image analyse program written by NIH and can
    be downloaded at http://rsb.info.nih.gov/ij/. Also download the line
    analysis plugin that on the same site.

    The most useful part of the program in the line profile function which
    show the pixel intensities along any straight line you draw on your
    image. This is particularly useful in determining wither your
    backgrouind is evenly illuminated. It the line profile is flat except
    for noise it is correctly illuminated, it the line profille slopes up
    or down, it isn't. And with the plugin, it will give you the numbers
    you need for your plot.

    The bad news is the cost of the tablet. I bought mine for $25 but only
    because my local photo store bought one on special order a few years
    before, whoever wanted it never came around to pick it up, and the
    owner wanted to get rid of the thing. The regular price for the
    calibrated version was about $150 and the uncalibrated version about
    $60 if I remember correctly.

    If you can find a Kodak Projection Print Scale--$2.70 when I bought it
    many years ago--that might also work. It's nowhere near as uniform and
    accurate as the step tablet and only has a tranmittance range of 5
    stops but if they still make the thing or something similar it will be
    far cheaper.

    A third alternative that I haven't tried is to buy a Wratten neutral
    density filter. You could then cut it up into strips and sandwich them
    together so at top you have one layer, below it two layers and so on.
    If you can locate some partially silvered mylar in an arts and craft
    store that might also work although you would need to do a few
    calculations to work out your optical densities. ImageJ would be very
    useful there since with a line profile you can see your step
    intensities easily

    If you aren't doing this in the raw mode, don't expect to see straight
    line plots. My somewhat antiquated olympus- a 3020Z has 11 contrast
    settlings. They determine how the 12 bits of the chip A/D are mapped
    into the 8 bits of the camera output. Each one gives me a different
    plot of pixel intesity vs optical density.

    jpc
     
    jpc, Feb 1, 2004
    #14
  15. Lionel

    brougham5 Guest

    I'd look for a MacBeth color chart. If my memory serves me correctly, they
    contain a row of neutral grays ranging from black to white. I'm not sure
    how many there are or if they are exactly "x" stops apart.
     
    brougham5, Feb 1, 2004
    #15
  16. Depends on what you mean by "turn up the light". If you have a single
    light source, and simply increase the light output (e.g. by increasing
    voltage on an incandescent) while leaving everything else the same, all
    measurements should increase by exactly the same ratio. The flare will
    increase by the same amount as everything else. Light behaves very
    linearly. If the meter now shows different ratios than it did, the
    meter isn't accurate.

    On the other hand, if you "turn up the light" by adding additional
    lights in different positions, that can well add flare - which can
    change apparent contrast.

    Dave
     
    Dave Martindale, Feb 1, 2004
    #16
  17. What, precisely, do you want to do that requires this degree of accuracy.
    Perhaps there is another, easier way to get where you want to go.
    Create a B/W Gradient in Photoshop and print it at best quality on matte
    paper.
    Then Shoot that image with your Digicam.
    Load it into PS and do a side by side comparison.
    Check the readings on the Info palette.for quantitative data
    This will test the Dynamic range of your camera. If this is what your
    ultimate objective is.
    Bob Williams
     
    Robert E. Williams, Feb 1, 2004
    #17
  18. A matte paper only have about 4-5 stops. Exactly how many
    differs. But ,,, it is at least fewer stops than the
    dynamic range of most cameras.

    /Roland
     
    Roland Karlsson, Feb 1, 2004
    #18
  19. SNIP
    Assuming the white has no "whiteners", you could also get a couple of paint
    sample strips. If they are marked with reflection codes (for paint
    blenders), it'll save some time.

    If you decide to go the alchemy way ;-) you may just mix by weight (cup on a
    kitchen scale).

    However, this all will result in a still limited density range. White may be
    95% reflection (D=Log10(1/0.95)=0.02 if your very lucky) and black perhaps
    D=2.4, meaning 8 stops. So ultimately, you'll have to do some exposure
    bracketing.

    Alternatively, you might just exposure bracket on a spectrally neutral gray
    card, under not too high illumination so you can use long exposure times,
    rather than different apertures. This will allow to use infinity focus for a
    close up of the card, which suppresses surface structure and takes out
    extension factors for magnification differences.

    Bart
     
    Bart van der Wolf, Feb 1, 2004
    #19
  20. Lionel

    Don Stauffer Guest

    No can do. You can only get a limited dynamic range in a physical
    printed card. The brighest reflectance white paper is nearly 100% (in
    mid to high nineties, anyway).

    However, the problem is the darkest black. The darkest photo papers,
    when fully developed, have a black reflectivity of about 2%. I think
    the blackest printing inks (from printing press) are even higher.

    Even professional anti-glare blacks are at best about a percent. So
    even using one of them still limits you to about a 100:1 dynamic range.
    More normal photo paper and ink printing limits you to about 50:1. This
    is far less than 10 stops.

    While paper is nearly 100% white, to get really black you must make a
    light trap. While it is not that difficult to make one, it is a fairly
    deep device, not thin like a card. Basically it is what is known as a
    black body cavity.
     
    Don Stauffer, Feb 1, 2004
    #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.