How does ISO setting work?

Discussion in 'Digital Photography' started by David Arnstein, Jul 15, 2005.

  1. The digital cameras that I have seen have a feature that allows the
    user to choose an ISO setting. I would like to know what this feature
    does, since I suspect that it is useless.

    My concern is that setting a high ISO number simply
    1. Causes the camera to take pictures that are quite dark
    2. Post processes the picture by increasing the brightness, in
    software.

    If this is the case, then I prefer to increase the brightness myself,
    in Photoshop.

    On the other hand, if the ISO adjustment on the camera actually
    changes the physical properties of the photosensor, then it's a
    different story.

    I'd like some feedback on this before I spend my time experimenting
    with different ISO settings.
     
    David Arnstein, Jul 15, 2005
    #1
    1. Advertisements

  2. David Arnstein

    Your-Nice Guest

    iso settings are to do with how sensitive your sensor is to light , the
    higher the iso setting i.e 400-800 1600 means you could take a photo in a
    dimmer room than iso 100 or 200.
     
    Your-Nice, Jul 15, 2005
    #2
    1. Advertisements

  3. David Arnstein

    tsingh Guest

    ISO does not change the physical properties of the photosensor, it
    simply brightens the image as you suggest. There is, however, one
    crucial difference. It amplifies the signal before digitizing it, and
    this keeps down the quantization noise.

    -T
     
    tsingh, Jul 15, 2005
    #3
  4. David Arnstein

    Marvin Guest

    Not quite correct. The ISO setting doesn't change the sensor at all. It changes the
    amplification in the readout. I

    The picture won't come out dark unless oyu use the wrong ISO setting. Most digicams let
    you use auutomaic ISO setting as part of a point-and-shoot setting. It works properly
    most of the time.

    No. it increases the amplification by the camera's electronic hardware.
    You can do that up to a point. If the exposures are really bad, Photoshop or any progam
    can't do the job well.
     
    Marvin, Jul 15, 2005
    #4
  5. Exactly. Since we only get 12-bit A/D converters (yet) in dSLRs, using an
    analog amplifier before the A/D converters should provide better images.

    FWIW, here's what's in effect an ISO 100 vs. ISO 3200 comparison (no noise
    reduction) in the Canon 300D, where the ISO 3200 is implemented by
    underexposing 5 stops at ISO 100 and postprocessing.

    Note that using ISO 100 in the 300D for ISO 3200 photography results in
    images that are far better than _any_ film could ever dream of.

    http://www.pbase.com/davidjl/image/45495842/original

    (Note that this does not say that "one should just shoot and ISO 100 and
    postprocess", since it doesn't compare in-camera settings and
    postprocessing.)

    One practical reason to use the ISO settings in the camera is that the noise
    reduction in some of the RAW converters is very good. (In particular, I've
    been very pleased with Rawshooter essentials.) This saves a step and speeds
    up one's work flow a lot.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Jul 15, 2005
    #5
  6. David Arnstein

    JPS Guest

    In message <db8u3k$l19$>,
    That should not be a problem. If anything, high ISO images should be
    brighter if they are different in brightness, either because there is
    not enough light for a low ISO image, or the cameras has run out of
    shutter speeds or wide aperture to keep the high ISO image from
    over-exposing.
    Some cameras achieve their highest ISOs this way, and get the lower
    range by varying the gain applied to the sensor capture before it is
    turned into numbers. Perhaps some do all of their ISOs this way, but I
    don't know how many.
    The normal way does not change the sensor at all; the sensors are fixed
    in their sensitivities. The difference between ISOs is usually achieved
    by amplifying the signal by different amounts, causing different ranges
    of sensor voltages to map to the range of RAW data, usually 0 to 4095.
    Experimenting will tell you more than other people usually can, and more
    than what the manufacturer will tell you.

    If the camera has manual exposure, set it up on a tripod or table to
    take an image with a full range of tones, and make a normal exposure at
    the camera's highest ISO. Now, keep the f-stop and the shutter speed
    the same, but set it to the lowest ISO. Boost the exposure level of the
    low-ISO image in software to match the high-ISO. If the high-ISO image
    has less noise and more detail, then the camera may be using
    amplification in hardware, and then you are clearly better off using
    high ISO than under-exposing a low ISO image (or having too long a
    shutter speed at a low ISO). Even if the camera is using amplification,
    if it is JPEG-only, it may be that the camera kills shadow detail when
    making a JPEG, to hide noise, in which case you will get poor images if
    you boost the exposure, anyway. You can also try the different ISOs
    with the same image, and automatic exposure. If the quality
    deteriorates rapidly as you go to higher ISOs, then there is probably no
    amplification. If the quality is only a little bit worse at each higher
    ISO, there is probably amplification. It is also possible that there is
    amplifiaction, but it has a lot of noise and distortion.

    This is a complex issue, with usually no details coming from the
    manufacturer.
    --
     
    JPS, Jul 15, 2005
    #6
  7. David Arnstein

    Matti Vuori Guest

    Could you please post in one group only...
     
    Matti Vuori, Jul 15, 2005
    #7
  8. David, is that true for all dSLRs? I thought even some high-end
    P&S's (e.g. Sony's F717) had 14-bit converters.
     
    Andy Sullivan, Jul 15, 2005
    #8
  9. I love usenet and I endeavor to be a "good citizen" in this space. I
    am very interested in your post.

    Would you please explain your reasoning. Here is mine:

    1. My post is relevant to all three news groups.
    2. By cross-posting, I arrange that an individual using a decent
    newsreader will only see my post once, even if he habitually reads
    all three news groups.

    I look forward to your response Mr. Vuori.
     
    David Arnstein, Jul 16, 2005
    #9
  10. I think Matti's post equates to, "I'm not using a decent
    news reader and don't know how Usenet works."
     
    Andy Sullivan, Jul 16, 2005
    #10
  11. David, please excuse the huge snip but I'd like you to focus on something.
    It's a great demo shot and clearly shows how the S/N ratio is noticeably
    better with the 1/4 s exposure as compared to the 1/125 s exposure. But
    that's kind of obvious, isn't it?

    A third shot with the camera set at ISO 3200 would look about the same as
    the 1/125 s exposure ... right? If not, then please explain how
    amplification before A/D as opposed to amplification after A/D makes a
    significant difference. I am familiar with quantization noise, by the way,
    but am looking for something from you that is more practical than
    theoretical. I mean to say that your test shots are rather dramatic and
    before we get bogged down in technical mumbo jumbo I'd like to see an ISO
    3200 equivalent. I am voting for no dramatic difference.

    Thanks and, again, it's a great demo shot!
     
    Charles Schuler, Jul 16, 2005
    #11
  12. David Arnstein

    JPS Guest

    In message <>,
    Not exactly. It could be a small difference, or a tremendous
    difference, depending on EC. If the RAW data is digitized so as to
    saturate the ~4000 levels in the ISO 100 image, and you use that as a
    reference point for the other exposures, the amount of noise is not
    tremendous relative to signal, at any ISO, as it would be if the image
    were not fully exposed (digitized, actually). I use the term
    "digitized" because it is different from film, where absolute exposure
    *is* the exposure. With digital, higher ISOs are usually used with
    partial sensor exposures, but images come out a lot better even when the
    ISO is high, if the RAW levels are fully utilized for that ISO.

    The fact is, most "existing light" exposures, especially in situations
    that call for high ISO, don't even come close to using the full range of
    RAW levels, especially when specular highlights and direct light sources
    play into the metering. The average RAW value above blackpoint with
    auto-exposure, indoors, with incandescent light on my 20D is something
    like 150 to 200 out of 3966 usable levels. Even on that high-exposure
    image I posted the other day of the painted letters on the wood blocks,
    95% of the pixels were less than 600 RAW levels above blackpoint (almost
    3 stops below the clipping point).
    Quantization noise is not just theoretical; in practice it is clearly
    worse than sensor noise, stop for stop. Here's an example (strange
    color is due to the fact that the image was not color balanced; this is
    the RAW color balance of a black fabric and black plastic buckle).
    Technically, the ISO 1600 image was exposed at -2 EC with a grey card,
    in manual mose, and only the ISO setting was changed for the ISO 100.
    The actual presentation, however, is probably boosted by a stop or a
    little more, so what you're seeing appears as EIs 200 and 3200, with a
    missing bit each, but the relative effect is still there (rendering them
    as black would not make the difference as obvious).

    http://www.pbase.com/jps_photo/image/40038800
    There is no gain-based ISO 3200 on the camera, AFAIK.


    --
     
    JPS, Jul 16, 2005
    #12
  13. David Arnstein

    ASAAR Guest

    At the risk of saying something foolish I'll try to answer that.
    The two shots should look about the same, because the shutter speeds
    (and exposure) were the same for both. The main difference is that
    the shot taken using the actual ISO 3200 will have its brightness
    boosted by 32x (5 bits worth) by the camera's amplifier. Since (as
    David said) the camera uses 12-bit A/D converters, there may
    sometimes (but not always) be some isolated highlights that after
    32x amplification would exceed the capacity of the 12-bit A/D
    converter and cause clipping of the signal, producing blown
    highlights. In effect, you're losing the highest 5 bits of picture
    information, and now have in effect an image with a smaller, 7-bit
    dynamic range.

    Since the picture taken with the pseudo ISO 3200 was really shot
    at ISO 100, it won't have the 32x amplification boost, and the
    12-bits worth of sensor data will be retained and can be sent to the
    computer. If the brightness is amplified by a program that doesn't
    have the camera's 12-bit limitation, you'll get fewer or no blown
    highlights.

    On the other hand, if there were no 'bright spots' in the scene,
    there would be no highlights to blow and so there shouldn't be a
    noticeable difference between the two shots. And then if you decide
    that you don't care about any bright highlights (whether they exist
    in the scene or not) that would have utilized the top 5 bits of the
    A/D converter, you could set the camera to ISO 100. What you'd then
    have would be the same image as the one taken at ISO 3200, but with
    an additional 5 bits worth of shadow detail, a deal worth its weight
    in photons. :)
     
    ASAAR, Jul 16, 2005
    #13
  14. David Arnstein

    JPS Guest

    In message <>,
    That gives equivalent S/N in the analog sensor.
    How is that any different than the the ISO 100 shot? The only
    difference is that the ISO 100 shot is more likely to bloom. They will
    both clip on the same highlights, with the same relative exposure (EC).
    At the pixel level, yes. At lower spatial frequencies, more levels can
    exist via dithering or averaging, albeit noisily.
    Not at all!

    Only 7 bits will be used in any kind of practical way.
    I understand this stuff very well, but you are confusing the hell out of
    me. I can't decide which image you're talking about, a lot of the time.
    I don't know which image you're referring to, but both images will clip
    at about the same point with default transfer curves in the RAW
    conversion, and the image with ISO 3200 set on the camera will have 5
    bits less posterization than the image taken at ISO 100 and
    under-exposed by 5 stops (4, actually, as ISO 3200 isn't 32x gain; its
    16x gain and 2x arithmetic, on most DSLRs).
    --
     
    JPS, Jul 16, 2005
    #14
  15. Yes. But I think you are missing the point. Have you ever looked at a scan
    of ISO 800 film? It's really really gross (incredibly grainy/noisy).

    The original complaint here was that since setting the ISO doesn't change
    the sensor, digital isn;t as good as film, where you can use any film you
    want. I put up that image to point out that dSLR digital at ISO 100 is a
    better low-light camera than any film, whatever the ISO.
    Yes. I realize that I haven't actually spoken to the question at hand. My
    point is simply that ISO 100 dSLR digital is a better low-light system than
    any 35mm film.

    One advantage of shooting correctly exposed ISO 1600 in the camera is that
    you can profile the noise characteristics and set up your RAW conversion to
    clean up the noise to your liking, even if the noise isn't all that much
    lower than underexposed ISO 100.

    (I'm currently being amuzed by the noise reduction in RSE. It's a phase,
    presumably it will pass. Just a slight touch of RSE noise reduction and ISO
    800 images become quite clean.)
    Well, the 300D only goes to ISO 1600<g>. You are right that I need to do a
    4-stop boosted ISO 100 shot vs. and in-camera ISO 1600 shot. My niece is
    using my 300D, so that won't happen immediately.

    Ah. I see that John Sheehy has already done my homework for me.

    http://www.pbase.com/jps_photo/image/40038800/original

    Note that John's point here is that were the A/D converter "perfect" (or at
    least better than the sensor), then ISO 1600 should look exactly like ISO
    100. So your "I am voting for no dramatic difference." would be exactly
    right were the A/D converter better.

    By the way, there's another subtext going on here. I am of the opinion that
    low noise is a very good thing and that sacrificing noise for pixel count is
    not something that should be done blithely.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Jul 16, 2005
    #15
  16. David Arnstein

    JPS Guest

    In message <db9vum$veb$>,
    That's with 12-bit digitization; if the current DSLRs had clean 16-bit
    digitization, even using current sensor technology, the difference
    between film and digital would be big, and totally in favor of digital
    (except that some films, especially at larger sizes, have greater
    spatial resolution).

    Those images I posted recently at ISO 1600 (the paper clip, $20 bill,
    wooden alphabet blocks) are what you'd get if you under-exposed ISO 100
    by 4 stops! You could set your camera to sunny f/16 at ISO 100, and get
    decent shots in deep shade, and excellent shots in the sun. Exposure
    latitude would be tremendous.
    --
     
    JPS, Jul 16, 2005
    #16
  17. David Arnstein

    ASAAR Guest

    Funny, but I find your explanations inscrutable at best.
     
    ASAAR, Jul 16, 2005
    #17
  18. David Arnstein

    ASAAR Guest

    Addendum. Now that I've seen David's latest message, I have to
    add that what he says still seems logical and clearly stated, and
    still consistent with what I was trying to say, albeit I may not be
    speaking the standard jargon that you're used to using. In
    contrast, reading your messages often is an adventure, full of
    twisty passage, all alike. And based on the various responses in
    your reply, it seems like you're jumping to conclusions that I must
    be mistaken so you aren't even trying to understand what I've tried
    to describe. I think that you understand less than you believe you
    do.
     
    ASAAR, Jul 16, 2005
    #18
  19. Following that line of reasoning, isn't your post relevant in
    rec.photo.digital.zlr as well?

    But your reasoning is flawed. If something is of general interest, it
    should only be in rec.photo.digital. The aim of splitting the newsgroups
    was to reduce the number of postings, and help people focus on their own
    interest better. There was no intention to replace the rec.photo.digital
    group.

    It would be better to restrict your posting to just one group if possible.

    Thanks,
    David
     
    David J Taylor, Jul 16, 2005
    #19
  20. Thanks for putting that up, John. Proves your point very nicely (and is a
    case where a picture is worth a thousand words!).

    So we need more bits in the ADC (and perhaps better shielding in the
    pre-ADC circuitry!) if the camera is to capture the full dynamic range of
    which that particular sensor is capable?

    Cheers,
    David
     
    David J Taylor, Jul 16, 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.