Under-exposure Versus Higher ISO?

Discussion in 'Digital Photography' started by Mardon, May 5, 2006.

  1. Today Kaz Kylheku attempted to dazzle everyone with this
    profound linguistic utterance

    > But, since photography is an art, you might just want to
    > experiment and subjectively evaluate the image quality,
    > right?


    Definitely.

    > An under-exposed picture doesn't take advantage of the
    > dynamic range.


    This is not at all a criticism of the OP, but what you say is
    also true of over-exposed images, which is what Mardon's
    pictures look like to me.

    You should always sample signals using
    > enough gain so that you take advantage of the sampling
    > resolution. If you can can sample a signal to a number
    > between 0 and 255, say, but the instrumentation gain is set
    > such that there is no value above 63, then you are down
    > from 8 bits to 6. Quantization from analog to digital is
    > also noise! Even if you perfectly sample a noise-free
    > signal, the sampling introduces noise of its own, and
    > sampling 6 bits introduces more distortion than 8 bits.
    >
    > If you crank up the sensitivity so that you can use 0 to
    > 255, you can always divide everything by 4 to get it back
    > in 0-63 range. That also cuts the noise amplitude by 4. So
    > in other words, even if the greater sensitivity brings in
    > more noise, it's probably worth it.
    >
    > I say probably, because the way amplification works is
    > complex. Amplifiers use negative feedback to reduce gain,
    > which also has the effect of reducing noise. Cranking up
    > the gain means that noise grows in greater proportion than
    > the signal, because the noise-reducing effect of negative
    > feedback is diminished. The amplifier is less stable, less
    > linear and the signal/noise ratio goes down.
    >
    > In order for the tradeoff to be worth it, you have to get a
    > bigger decrease in quantization noise than an increase in
    > analog noise.
    >
    > And that assumes that the two noises are equal from a
    > subjective, esthetic point of view, which they are not.
    > That brings us back to experimentation.
    >
    >> I know that the answer to this question can be complicated
    >> by suggesting flash and other such things to increase the
    >> light but that's not my point.

    >
    > Among "other such things" would be a lens with more light
    > gathering power.


    --
    ATM, aka Jerry

    "I came, I saw, I conquered" - Alexander The Great
     
    All Things Mopar, May 5, 2006
    #21
    1. Advertising

  2. Mardon

    Scott W Guest

    All Things Mopar wrote:
    > > I understand your analogy. In fact, I did exactly what you

    > describe and got exactly the results you predict.
    >
    > I hope to hear more from the person I asked this question of so
    > I can "calibrate" their expectations of ISO vs noise in the
    > context of this thread.


    Ok here are a bunch of shots from the 20D and the XT at a number of
    ISOs.
    These have been converted from raw with no noise filtering in place. A
    bit of
    filter could make them look somewhat better but detail would be loss.

    http://www.pbase.com/konascott/isotest

    As I said before I would be pretty happy with the XT at ISO 400 and the
    20D at ISO 800.

    But a lot depend on what I am going to do with the photos, if printed
    as 8 x 12 the higher ISO numbers would probably be ok.

    Scott
     
    Scott W, May 5, 2006
    #22
    1. Advertising

  3. "Mardon" <> wrote in message
    news:Xns97BA7E3B073F5mgb72mgbhotmailcom@140.99.99.130...
    > Suppose that I'm using a Canon 20D and shooting in RAW, with
    > the slowest shutter speed and largest aperture that I can use for
    > a given situation. I'm at ISO800 ant the photo is still being
    > under-
    > exposed. How bad must the under-exposure become before it's
    > better to up the ISO to 1600?


    It is better to let the camera apply "analog gain" before
    quantization, than to boost the result in post-processing. Analog gain
    will produce less noise than postprocessing.

    > Same question if I'm already at 1600. How bad must the
    > under-exposure be before I'm better off to up the ISO to 3200?


    Since you shoot Raw, never use ISO 3200. It does the same as
    postprocessing an ISO1600 on the 20D.

    Bart
     
    Bart van der Wolf, May 6, 2006
    #23
  4. Scott W wrote:
    > wrote:
    >
    >>Ed Ruf wrote:
    >>
    >>>One thing to consider is are the extreme higher isos
    >>>real. Or are they just iso1600 pushed with analog amplification?

    >>
    >>They're not really real : they're digital ;o).
    >>
    >>A sensor has only one 'real' sensibility (generally the lower one), the
    >>others are made by a signal amplification - that, more or less,
    >>amplifies also noise.
    >>
    >>The only differences i can see are :
    >>- some S/N optimization algorithm may also be applied after
    >>amplification, but would it be better in-camera than what you can do in
    >>post-processing with those neat ninjas of denoiser softs?
    >>- if the analog/digital conversion is made after amplification (likely,
    >>isn't it?), the higher-ISO signal contains more signal levels than the
    >>underexposed one... but I'd think that in the high ISO (and 16-bit)
    >>case, both are limited by noise?
    >>
    >>So, my personal conclusion would be 'six of one and half a dozen of the
    >>other' between underexposing and cranking ISO setting (while RAW of
    >>course)...

    >
    >
    > A quick test with a Rebel XT shows a lot less noise in the ISO 400 shot
    > then the 100, when both are taken at the same aperture and shutter
    > speeds.
    >
    > So at least between iso 100 and 400 it is much better to use iso 400
    > then simply underexpose
    >
    > Scott
    >


    The answer to the question has to do with the sensor gain and
    the sensor read noise. Both of these are summarized in tables
    1 and 2 at:
    http://www.clarkvision.com/imagedetail/digital.signal.to.noise

    For an XT (350D), the gain at iso 800 is 1.3 electrons/DN
    (DN = 12-bit digital number from the camera). The gain
    is 0.6 electron/DN at iso 1600. These is no effective point in
    digitizing beyond 1 electron. Then add read noise. The read
    noise of the 350D is 4.9 electrons at iso 800 and 3.7 at
    iso 1600. These is a modest improvement at iso 1600, but
    by the time you have enough electrons for your image,
    the image is photon noise limited (once you have a few tens
    of electrons). Thus for practical images there would barely
    be a perceptible improvement from 800 to 1600.
    If you were shooting faint sources (low light night images,
    the differences would be a little more apparent, but still
    small).

    The advantage of iso 800 is you have double the dynamic range
    as that at iso 1600, and almost the same noise.

    Roger
     
    Roger N. Clark (change username to rnclark), May 6, 2006
    #24
  5. Mardon

    Neil Ellwood Guest

    On Fri, 05 May 2006 14:54:35 +0000, Mardon wrote:

    > Suppose that I'm using a Canon 20D and shooting in RAW, with the
    > slowest shutter speed and largest aperture that I can use for a given
    > situation. I'm at ISO800 ant the photo is still being under-exposed.
    > How bad must the under-exposure become before it's better to up the
    > ISO to 1600? Same question if I'm already at 1600. How bad must the
    > under-exposure be before I'm better off to up the ISO to 3200?
    >
    > I know that the answer to this question can be complicated by
    > suggesting flash and other such things to increase the light but
    > that's not my point. I'm hoping to get a simple response about the
    > final picture quality that can be achieved when evaluating the trade-
    > off between under-exposing versus using a higher ISO while shooting
    > RAW.

    Why would we suppose this?
    Why don't you get a 20d and try it?
    --
    Neil
    Delete 'l' to reply
     
    Neil Ellwood, May 6, 2006
    #25
  6. Mardon

    Mardon Guest

    Neil Ellwood <> wrote:

    > Why would we suppose this?


    To be helpful.

    > Why don't you get a 20d and try it?


    I already own one. That's why I asked the question. I want to know
    which method will produce the best final image.
     
    Mardon, May 6, 2006
    #26
  7. Mardon wrote:
    > Neil Ellwood <> wrote:
    >
    >> Why would we suppose this?

    >
    > To be helpful.
    >
    >> Why don't you get a 20d and try it?

    >
    > I already own one. That's why I asked the question. I want to know
    > which method will produce the best final image.


    I think his point would be: Why don't you (also) do this and post the
    results? If others have or will do this, then the answers may add up to
    something useful.

    --
    John McWilliams
     
    John McWilliams, May 6, 2006
    #27
  8. Mardon

    Mardon Guest

    John McWilliams <> wrote:

    >> I already own one. That's why I asked the question. I want to
    >> know which method will produce the best final image.

    >
    > I think his point would be: Why don't you (also) do this and
    > post the results? If others have or will do this, then the
    > answers may add up to something useful.


    I'm not good at doing 'rigorously' controlled experiments. I
    already gave my personal impressions in an earlier post to this
    thread as follows:
    "My impression based on my everyday experience is that if a RAW image
    is underexposed by more than a full f stop, then it's probably better
    to go to the higher ISO. If the underexposure is less than a full
    stop, I think it's better to shoot at the lower ISO and use Noise
    Ninja or some other noise filter to reduce the noise. (BTW, I find
    Noise Ninja better than the noise reduction filter built into PS
    CS2.) I'm not sure of this conclusion , however, and wanted to find
    out what others think"
     
    Mardon, May 6, 2006
    #28
  9. Mardon wrote:
    > John McWilliams <> wrote:
    >
    >>> I already own one. That's why I asked the question. I want to
    >>> know which method will produce the best final image.

    >> I think his point would be: Why don't you (also) do this and
    >> post the results? If others have or will do this, then the
    >> answers may add up to something useful.

    >
    > I'm not good at doing 'rigorously' controlled experiments. I
    > already gave my personal impressions in an earlier post to this
    > thread as follows:


    O.K. totally fair; I am not, either.

    > "My impression based on my everyday experience is that if a RAW image
    > is underexposed by more than a full f stop, then it's probably better
    > to go to the higher ISO. If the underexposure is less than a full
    > stop, I think it's better to shoot at the lower ISO and use Noise
    > Ninja or some other noise filter to reduce the noise. (BTW, I find
    > Noise Ninja better than the noise reduction filter built into PS
    > CS2.) I'm not sure of this conclusion , however, and wanted to find
    > out what others think"


    JPS has both the ability to do the testing - I think he already has- as
    well as write about it. Haven't seen him here recently, but I look
    forward to what he writes- maybe this time I will grasp more of it....

    Based on a loose understanding of reports and tests, I tend to set to
    ISO 200 as default, and don't go over 800 ISO. In some sports shooting,
    I will set to M, at 160 or 200 shutter, and wide open. When it gets
    darker , I'll stick on the 580EX, set to 250 and close down a stop or
    two. I am certainly open to better ways to do it.

    --
    John McWilliams
     
    John McWilliams, May 6, 2006
    #29
  10. Mardon

    Neil Ellwood Guest

    On Sat, 06 May 2006 10:21:03 +0000, Mardon wrote:

    > Neil Ellwood <> wrote:
    >
    >> Why would we suppose this?

    >
    > To be helpful.
    >
    >> Why don't you get a 20d and try it?

    >
    > I already own one. That's why I asked the question. I want to know
    > which method will produce the best final image.

    Then try it .
    --
    Neil
    Delete 'l' to reply
     
    Neil Ellwood, May 7, 2006
    #30
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Replies:
    0
    Views:
    739
  2. Paul

    Higher ISO = increased noise

    Paul, Mar 1, 2004, in forum: Digital Photography
    Replies:
    10
    Views:
    591
    imbsysop
    Mar 2, 2004
  3. Matt

    What Creates Noise/Grain At Higher ISO Speeds?

    Matt, Nov 14, 2004, in forum: Digital Photography
    Replies:
    65
    Views:
    2,053
    Roger N. Clark (change username to rnclark)
    Nov 19, 2004
  4. Gary Edstrom

    Which is better: Higher ASA or Longer Exposure?

    Gary Edstrom, Dec 20, 2004, in forum: Digital Photography
    Replies:
    35
    Views:
    703
    Jon Pike
    Dec 25, 2004
  5. Peter Potamus the Purple Hippo

    Re: Mozilla versus IE versus Opera versus Safari

    Peter Potamus the Purple Hippo, May 8, 2008, in forum: Firefox
    Replies:
    0
    Views:
    761
    Peter Potamus the Purple Hippo
    May 8, 2008
Loading...

Share This Page