Is ISO range an electronic componet set by camera manufacturers?

Discussion in 'Digital SLR' started by Dave, Jan 3, 2006.

  1. Dave

    Dave Guest

    When I look at the specifications on different digital cameras, some have an
    ISO range of 60-200 or some have a range of 100-1600.

    Is this something the camera manufacturer can control by the electronics it
    puts in the camera?

    Just wondering how the ISO range is determined.
     
    Dave, Jan 3, 2006
    #1
    1. Advertisements

  2. There are basically two ways to determine the ISO value of a device. Those
    are called 'noise-based' and 'saturation-based'. Typically you want
    saturation-based for digital cameras.

    For saturation-based ISO measurements you measure how much light can be
    faithfully recorded by the device. Too much light and the output starts
    clipping (burned-out highlights). The more light the device can handle,
    the lower the ISO.

    With a digital camera, it is easy to amplify the signal that comes from
    the sensor (either with analog amplifiers, or by scaling the digital output).
    This means that the camera starts clipping at a lower light level, which
    in turn means a higher ISO.

    So, the more you amplify the signal, the higher the ISO. There is almost no
    limit. This means that the important ISO number is the lowest one. That is
    the one that shows how much light a camera can record (useful in bright sun
    light with the lens wide open)

    Amplifying the signal that comes from the sensor is of course limited by
    the amount of extra noise you get. In theory, you can use noise-based ISO
    measurements to determine the maximum ISO of a camera. Noise-based ISO
    measurements simply look for the highest ISO setting that still has a
    reasonable amount of noise.

    Unfortunately, the 'reasonable amount of noise' in ISO measurements is not
    really acceptable to many people.

    In the end, the highest ISO settings of a camera should be studied for noise
    patterns and color accuracy. Limit yourself to the highest setting that
    still results in an acceptable image.
     
    Philip Homburg, Jan 3, 2006
    #2
    1. Advertisements

  3. You forgot about dynamic range. This might create impression that the
    lower ISO, the better camera. Which in almost all cases is wrong.


    B.
     
    Bronek Kozicki, Jan 3, 2006
    #3
  4. Actually, it is just the opposite. Most good digital cameras are
    photon noise limited for the given quantum efficiencies. That
    means dynamic range is controlled mainly by two things:
    read noise and total electrons collected. Dynamic range =
    total electrons (the full well electrons)/read noise in electrons.
    You get maximum dynamic range and highest signa-to-noise ratio with
    lowest iso.

    Having said that, the larger the pixel size, the more the
    electrons the well can hold, so the larger pixel size cameras
    generally have higher signal-to-noise ratios and higher
    dynamic range.

    Details:

    Dynamic Range and Transfer Functions of Digital Images
    and Comparison to Film
    http://www.clarkvision.com/imagedetail/dynamicrange2

    The Signal-to-Noise of Digital Camera images
    and Comparison to Film
    http://www.clarkvision.com/imagedetail/digital.signal.to.noise

    Digital Cameras: Does Pixel Size Matter?
    Factors in Choosing a Digital Camera
    http://www.clarkvision.com/imagedetail/does.pixel.size.matter

    Roger
     
    Roger N. Clark (change username to rnclark), Jan 3, 2006
    #4
  5. .... with the lowest ISO given sensor can handle. Not "with the lowest
    ISO" as it might create impression that sensors used in most compact
    cameras (with minimum ISO 64 or less) have better dynamic range than eg.
    Canon 20D, which (I believe) is not true. This is why second part of
    your post is important, ie. where you explain dependency between dynamic
    range and electrons collected. Interestingly, larger pixel size allows
    higher maximum ISO, thus following is quite typical: the higher max. ISO
    with acceptable picture quality, the better dynamic range at the minimum
    ISO of given camera. Or I believe so ;)


    B.
     
    Bronek Kozicki, Jan 3, 2006
    #5
  6. Dave

    JPS Guest

    In message <qinuf.60957$>,
    The sensor itself can capture a certain amount of photons in each sensor
    pixel before it fills up. The lowest ISO of a camera is nominally based
    on the lowest sensitivity that will leave approximately 3 stops above
    "average exposure" or "middle grey" before the sensor wells fill up.

    Cameras may play with this figure, though, going for lower lowest ISOs
    with less highlight range, or higher lowest ISOs with more headroom.
    The higher ISOs on amy given camera can be achieved in two ways; the
    sensor signals can be amplified before they become numbers, so that only
    1/2, 1/4, 1/8 of the full sensor charge converts to the same range of
    numbers. Where the manufacturer draws the line is usually based on
    where they perceive the point of no further gain is possible; amplifier
    noise or a sensor noise floor may prevent any usable gain, especially if
    the hardware that reads the sensor is of low quality. Some cameras will
    use math instead of amplification to get to their highest ISOs. Most of
    the Canon DSLRs, for example (and probably lots of others) have an ISO
    3200 mode that really is ISO 1600 with the digitized numbers doubled,
    just to make the output brighter on what would otherwise be an
    under-exposure.
    --
     
    JPS, Jan 3, 2006
    #6
  7. Some cameras will
    Yes, it is a simple left-shift of the bits which multiplies by two. ISO
    3200 is just a gimmick ... much like digital zoom.
     
    Charles Schuler, Jan 3, 2006
    #7
  8. Dave

    JPS Guest

    In message <>,
    Both ISO 3200 and digital zoom both work nearly as advertized when
    shooting JPEG, as JPEG would have lost the extra stop of highlights that
    ISO 3200 throws away, anyway, and the center crop is better preserved at
    2x despite the JPEG algorithms. So, for they JPEG shooter, they are
    somewhat real.
    --
     
    JPS, Jan 3, 2006
    #8
  9. Given that this is the digital slr systems forum, I disagree. They are both
    gimmicks. My assumption is that folks here know at least the rudiments of
    post-processing.

    Of course, they "work" but so does a dull knife.
     
    Charles Schuler, Jan 3, 2006
    #9
  10. Yes, that is correct, and a good clarification.

    Roger
     
    Roger N. Clark (change username to rnclark), Jan 4, 2006
    #10
  11. Actually, there is a third way, which is what is used on most CMOS
    cameras for reasons I'll explain. This method is to design the chip
    with different sized capacitors in each pixel which can be switched
    independently to integrate and store the signal. A large capacitor
    produces less voltage for a given amount of photocharge resulting in a
    low ISO figure, while a small capacitor produces a greater voltage for
    the same photocharge, resulting in a higher ISO. Only a few individual
    capacitors are required to produce the complete range of ISO values from
    100-1600 by switching one or more on at a time.

    It is important to distinguish between this approach and either of the
    two that you have previously mentioned because the readout noise is
    independent of the pixel noise when capacitor switching like this is
    used. Readout noise generally dominates dark noise in the pixel
    significantly. In your first method, the readout noise is added to the
    signal voltage from each pixel independent of the ISO setting. As a
    result, the amplification required to achieve the higher ISO results in
    the readout noise also being amplified. Consequently the total noise at
    high ISO is simply scaled from the total noise at low ISO (as is the
    signal).

    In this alternative method, the readout noise is added to the signal
    voltage for each pixel *after* it has been scaled for ISO by the
    selection of the appropriate pixel capacitance. As a result, the total
    noise is proportionally lower at higher ISOs than the gain required for
    the signal alone. Consequently amplification further down the signal
    chain to simulate higher ISO results in a significantly poorer SNR than
    selection of the correct ISO at the sensor.

    This is one reason why CMOS sensors hold their performance at higher ISO
    better than CCDs, even though the SNR of CCDs at low ISO can be better
    than is possible with CMOS - this third method of ISO control is not
    *generally* available in CCDs (although an alternative method of
    achieving extremely high ISO is possible with CCDs, should money be no
    object).
     
    Kennedy McEwen, Jan 4, 2006
    #11
  12. Of all the technical literature I have read, this method has never
    been mentioned. It also seems to have a problem of taking up
    too much space on the chip. CMOS already has reduced active
    area due to the support electronics at each pixel (which is
    not generally an issue if a micro lens is used to focus the light
    onto the active area). Do you have some references?
    Roger
     
    Roger N. Clark (change username to rnclark), Jan 5, 2006
    #12
  13. I'll see if I can dig up the references for you. I believe it was Canon
    who originally published some papers on the concept several years ago
    and it was so significant we adapted it for our infrared detectors,
    where it is now standard across the industry even though it is less
    beneficial in that field because of the high background photon noise
    that is present all the time. It doesn't use that many transistors to
    span the range 100-1600ISO - only 4 - and the total capacitor area is
    almost identical as the non-switched area, it is just segmented.
     
    Kennedy McEwen, Jan 6, 2006
    #13
    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.