Re: The Sigma-Foveon pixel rationale

Discussion in 'Digital Photography' started by Dave Martindale, Apr 1, 2004.

  1. (Dave Haynie) writes:

    >You have to, for the simple fact that your "equal exposure" senario is
    >the way cameras are actually used. Not the "equal output" senario. For
    >any given EV level, you base your exposure on effective ISO, shutter
    >speed, and lens opening. Nothing else. You have no other natural
    >mechanism to force an extra stop's worth of exposure in a head to head
    >comparison of cameras.


    So you rate the RGB sensor as having an ISO speed that is half that
    of the CYGM sensor. Bingo: it automatically gets double the exposure.

    >Think of it this way: IGNORE THE SENSOR TYPE. Pretty much everyone
    >will, unless they're in this discussion. When you test a camera
    >against another, head to head, you're going to set comparable
    >ISO-equivalents, similar exposures (unless you're benchmarking the
    >accuracy or cleverness of their auto-exposure systems), and you shoot,
    >and you report on noise. There's no place to stop and say, "whoa, this
    >is an RBG camera, so I get to drop from ISO200 to ISO100 and double
    >the exposure". Just doesn't play, given the way everyone uses cameras.


    I disagree. I'll happily test an RGB camera with a minimum ISO of 50
    against a CMY camera with a minimum ISO of 100, each at their lowest
    ISO. That's the condition that gives the best results from each camera,
    so why not compare them?

    Yes, I know that *at ISO 100* the RGB camera will be noiser than the CMY
    camera. But I have the choice of using ISO 50 with the RGB camera, and
    when I do use it, the image is less noisy than the lowest ISO on the CMY
    camera. This makes me happy. This might make me choose the RGB camera.

    I may not know whether the camera has a RGB or CMY sensor. All I know
    is that the min-ISO 50 camera takes slightly better pictures than the
    min-ISO 100 camera, both set to optimum ISO.

    In fact, I'll bet most people don't know what ISO they shoot at - they
    just use the default or "auto". They don't know what aperture and
    shutter speed the camera used either, unless they look at the EXIF data,
    because they're using auto exposure too. If they shoot in bright light,
    they won't even be aware that the RGB camera is using longer exposures
    or wider apertures.

    Dave
     
    Dave Martindale, Apr 1, 2004
    #1
    1. Advertising

  2. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <c4gih1$e4t$> on Thu, 1 Apr 2004 08:05:53 +0000 (UTC),
    (Dave Martindale) wrote:

    >... I'll happily test an RGB camera with a minimum ISO of 50
    >against a CMY camera with a minimum ISO of 100, each at their lowest
    >ISO. That's the condition that gives the best results from each camera,
    >so why not compare them?


    Because (as I've explained in some detail previously) it's an apples and
    oranges comparison.

    There's nothing magic about ISO 50 with RGB, and no reason to think the CMY
    sensor couldn't also have an ISO 50 mode with less noise than at ISO 100 or
    than the RGB sensor at ISO 50.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 1, 2004
    #2
    1. Advertising

  3. "John Navas" <> wrote:
    > (Dave Martindale) wrote:
    >
    > >... I'll happily test an RGB camera with a minimum ISO of 50
    > >against a CMY camera with a minimum ISO of 100, each at their lowest
    > >ISO. That's the condition that gives the best results from each camera,
    > >so why not compare them?

    >
    > Because (as I've explained in some detail previously) it's an apples and
    > oranges comparison.
    >
    > There's nothing magic about ISO 50 with RGB, and no reason to think the

    CMY
    > sensor couldn't also have an ISO 50 mode with less noise than at ISO 100

    or
    > than the RGB sensor at ISO 50.


    You've apparently misunderstood my previous note: there is a lowest
    sensitivity that the underlying sensor can operate at. There's a minimum
    noise level that the pixels themselves can produce that you can't improve
    on. It's due to the amount of charge the sensor can store and transport for
    each pixel. If you use a lower sensitivity, your highlights blow out sooner.

    That minimum noise level is unrelated to the type of CFA you use. It happens
    at different ISOs for RGB and CMY.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Apr 1, 2004
    #3
  4. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <c4i1sh$76t$> on Fri, 2 Apr 2004 06:28:57 +0900, "David J.
    Littleboy" <> wrote:

    >"John Navas" <> wrote:


    >> There's nothing magic about ISO 50 with RGB, and no reason to think the CMY
    >> sensor couldn't also have an ISO 50 mode with less noise than at ISO 100 or
    >> than the RGB sensor at ISO 50.

    >
    >You've apparently misunderstood my previous note:


    I don't think so.

    >there is a lowest
    >sensitivity that the underlying sensor can operate at.


    Actually the real/raw sensitivity; i.e., without gain.

    >There's a minimum
    >noise level that the pixels themselves can produce that you can't improve
    >on. It's due to the amount of charge the sensor can store and transport for
    >each pixel. If you use a lower sensitivity, your highlights blow out sooner.


    You actually just have less dynamic range, and thus lower S/N.

    >That minimum noise level is unrelated to the type of CFA you use. It happens
    >at different ISOs for RGB and CMY.


    Interesting, but irrelevant, since you haven't established what that level
    might be -- you've just assumed it to be ISO 50, whereas it might actually be
    (say) ISO 5. As I wrote, "there's nothing magic about ISO 50."

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #4
  5. John Navas <> writes:

    >>... I'll happily test an RGB camera with a minimum ISO of 50
    >>against a CMY camera with a minimum ISO of 100, each at their lowest
    >>ISO. That's the condition that gives the best results from each camera,
    >>so why not compare them?


    >Because (as I've explained in some detail previously) it's an apples and
    >oranges comparison.


    It's testing both cameras at their optimum ISO. I'd call that apples to
    apples. Are you saying you wouldn't use the slower ISO speed if you had
    a RGB camera provided with it.

    >There's nothing magic about ISO 50 with RGB, and no reason to think the CMY
    >sensor couldn't also have an ISO 50 mode with less noise than at ISO 100 or
    >than the RGB sensor at ISO 50.


    If the sensors are different, then they might have different
    characteristics. But, all other things being the same, a CCD with a CMY
    sensor is more sensitive than a CCD with a RGB sensor, by roughly a
    factor of 2. The Sony datasheets for the pair of sensors you pointed
    out make this clear.

    The "natural" ISO rating for a sensor is one where scene white causes
    the CCD's charge well to be close to full, but not saturated. This
    takes more light for an RGB sensor, so its natural ISO is lower than a
    CMY sensor based on the same underlying CCD chip. If ISO 50 properly
    exposes the RGB version of the sensor, ISO 50 will overexpose the CMY
    one, clipping the highlights - so the CMY camera will not provide that
    ISO setting. On the other hand, if the CMY camera *can* expose at ISO
    50 without saturation, the RGB camera can really operate at ISO 25 or
    32. There will always be a difference in "natural" ISO when there is a
    difference in sensitivity.

    And when both cameras are operated at their optimum ISO, giving the same
    output voltage from their respective CCDs, the CMY camera will have more
    noise because of the subtraction inherent in extracting RGB from CMY.

    Dave
     
    Dave Martindale, Apr 2, 2004
    #5
  6. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <c4itul$10q$> on Fri, 2 Apr 2004 05:33:09 +0000 (UTC),
    (Dave Martindale) wrote:

    >John Navas <> writes:


    >>Because (as I've explained in some detail previously) it's an apples and
    >>oranges comparison.

    >
    >It's testing both cameras at their optimum ISO.


    It's testing both cameras at an arbitrary differential ISO, since you haven't
    established any such actual "optimum ISO" values.

    >I'd call that apples to
    >apples.


    Fair enough -- but I wouldn't.

    >Are you saying you wouldn't use the slower ISO speed if you had
    >a RGB camera provided with it.


    I'm saying I would want to use a given ISO speed in a given situation
    regardless of sensor.

    >>There's nothing magic about ISO 50 with RGB, and no reason to think the CMY
    >>sensor couldn't also have an ISO 50 mode with less noise than at ISO 100 or
    >>than the RGB sensor at ISO 50.

    >
    >If the sensors are different, then they might have different
    >characteristics. But, all other things being the same, a CCD with a CMY
    >sensor is more sensitive than a CCD with a RGB sensor, by roughly a
    >factor of 2. The Sony datasheets for the pair of sensors you pointed
    >out make this clear.
    >
    >The "natural" ISO rating for a sensor is one where scene white causes
    >the CCD's charge well to be close to full, but not saturated. This
    >takes more light for an RGB sensor, so its natural ISO is lower than a
    >CMY sensor based on the same underlying CCD chip. If ISO 50 properly
    >exposes the RGB version of the sensor, ISO 50 will overexpose the CMY
    >one, clipping the highlights - so the CMY camera will not provide that
    >ISO setting. On the other hand, if the CMY camera *can* expose at ISO
    >50 without saturation, the RGB camera can really operate at ISO 25 or
    >32. There will always be a difference in "natural" ISO when there is a
    >difference in sensitivity.
    >
    >And when both cameras are operated at their optimum ISO, giving the same
    >output voltage from their respective CCDs, the CMY camera will have more
    >noise because of the subtraction inherent in extracting RGB from CMY.


    If you are using an RGB sensor, unless and until you reach your "optimum ISO"
    (which hasn't been established), if I'm using a CMY sensor I'll have the
    option of matching you on ISO with lower noise, or shooting at higher ISO with
    less noise than you at the same ISO.

    Even if you could show my CMY sensor to have an "optimum ISO" at a level high
    enough to make a difference in a given situation, there would still be many
    other situations where I would still have an advantage because that ISO level
    is needed to get the desired results.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #6
  7. John Navas <> writes:

    >>It's testing both cameras at their optimum ISO.


    >It's testing both cameras at an arbitrary differential ISO, since you haven't
    >established any such actual "optimum ISO" values.


    I haven't shown specific ISO values for specific sensors. But exposing
    so whites reach a large fraction of the full-well capacity is well known
    in CCD literature. It's how you make the best use of the dynamic range
    of whatever sensor you have. It's optimum because if you increase the
    ISO, you reduce the signal and reduce S/N, while if you decrease the ISO
    you get the whites saturating and perhaps blooming.

    So any given CCD has an optimum exposure at a particular number of
    electrons per pixel, which is equivalent to a particular voltage at the
    output. Since a CMY filter passes more light than an RGB one, the
    overall sensitivity of the CMY sensor/filter combo *must* be higher than
    a RGB sensor/filter using the same sensor. And thus the optimum ISO for
    the CMY sensor must be higher.

    >I'm saying I would want to use a given ISO speed in a given situation
    >regardless of sensor.


    In most situations, there is some room to trade off among ISO, aperture,
    and shutter speed.

    >If you are using an RGB sensor, unless and until you reach your "optimum ISO"
    >(which hasn't been established), if I'm using a CMY sensor I'll have the
    >option of matching you on ISO with lower noise, or shooting at higher ISO with
    >less noise than you at the same ISO.


    The second half of that isn't clear. As long as the RGB camera isn't at
    its lowest ISO setting, you have the option of shooting at the same ISO
    with somewhat less noise, or shooting at about 1 stop higher ISO with
    somewhat more noise. I agree that's useful.

    On the other hand, if the RGB camera is used at its lowest ISO, it will
    give less noise than the CMY camera at any ISO that does not clip the
    scene whites.

    >Even if you could show my CMY sensor to have an "optimum ISO" at a level high
    >enough to make a difference in a given situation, there would still be many
    >other situations where I would still have an advantage because that ISO level
    >is needed to get the desired results.


    OK. If higher ISO is important, a CMY sensor is one way of improving
    ISO. But if the absolute lowest noise is important, regardless of ISO,
    then an RGB sensor is one way of reducing noise. (Larger photosites
    can be used to improve either ISO or noise).

    Dave
     
    Dave Martindale, Apr 2, 2004
    #7
  8. "John Navas" <> wrote in message
    news:YD1bc.3208$...
    > [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]
    >
    > In <c4i1sh$76t$> on Fri, 2 Apr 2004 06:28:57 +0900, "David

    J.
    > Littleboy" <> wrote:

    SNIP
    > >there is a lowest sensitivity that the underlying sensor can
    > >operate at.

    >
    > Actually the real/raw sensitivity; i.e., without gain.


    Then why do you insist to use that, and THEN add filters with a different
    transmission amount (integrated over their spectral transmission), as a
    basis? The more rational thing is to base the sensitivity on an equal
    exposure level AFTER the filter layer has done its function. Knowing that
    you understand the concequences of doing so, there must be 'some' reason you
    don't want to consider that.

    SNIP
    > >That minimum noise level is unrelated to the type of CFA you use. It

    happens
    > >at different ISOs for RGB and CMY.

    >
    > Interesting, but irrelevant, since you haven't established what that level
    > might be -- you've just assumed it to be ISO 50, whereas it might actually

    be
    > (say) ISO 5. As I wrote, "there's nothing magic about ISO 50."


    That's rehashing the same unanswered issue. Why not base the sensitivity,
    e.g. threshold exposure to generate a detectable signal above the noise
    floor, on the signal actually reaching the photosensitive sensor area, i.e.
    after the filter layer?

    In general, although filter transmissions can vary between spectral bands to
    e.g. better match the Human Visual System, the RGB filters transmit roughly
    half of the incident radiation when compared to similar CMY filters. The
    only sensible thing to do in that case is to double the exposure on the RGB
    filters in order to better utilize the full potential well depth.

    Bart
     
    Bart van der Wolf, Apr 2, 2004
    #8
  9. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <406d71f3$0$559$4all.nl> on Fri, 2 Apr 2004 16:00:18 +0200,
    "Bart van der Wolf" <> wrote:

    >"John Navas" <> wrote in message
    >news:YD1bc.3208$...
    >>
    >> In <c4i1sh$76t$> on Fri, 2 Apr 2004 06:28:57 +0900, "David J.
    >> Littleboy" <> wrote:

    >SNIP
    >> >there is a lowest sensitivity that the underlying sensor can
    >> >operate at.

    >>
    >> Actually the real/raw sensitivity; i.e., without gain.

    >
    >Then why do you insist to use that, and THEN add filters with a different
    >transmission amount (integrated over their spectral transmission), as a
    >basis?


    Because that's what we're comparing, that's the way the two sensors are built,
    and that's how they will be used in the real world. The sensitivity of the
    completed sensor and how it's used in the real world are at least as important
    as the real/raw sensitivity of the unfiltered semiconductor.

    >The more rational thing is to base the sensitivity on an equal
    >exposure level AFTER the filter layer has done its function.


    I don't agree.

    >Knowing that
    >you understand the concequences of doing so, there must be 'some' reason you
    >don't want to consider that.


    Since light is often a limiting factor, we usually work at a given ISO that's
    matched to the task at hand. Otherwise there would never have been a need for
    any slide film faster than (say) Kodachrome 25, yet faster slide films (e.g.,
    Kodachrome 64) became very popular because the extra speed made them more
    versatile, even when slower film could be uprated with push processing.

    If all you're going to do is photograph static subjects in sufficient light on
    a tripod, then indeed speed becomes relatively unimportant, which was of
    course the niche for Kodachrome 25. But if you're going to photograph motion
    and/or in a wide range of light and/or handheld, then speed becomes an issue,
    and faster films (e.g., Kodachrome 64) give you better results, even at the
    expense of increased grain (much like digital sensor noise) and decreased
    sharpness (not a factor with digital sensors). If faster films behaved more
    like digital sensors (i.e., no loss of sharpness), they might have been even
    more popular.

    >SNIP
    >> >That minimum noise level is unrelated to the type of CFA you use. It happens
    >> >at different ISOs for RGB and CMY.

    >>
    >> Interesting, but irrelevant, since you haven't established what that level
    >> might be -- you've just assumed it to be ISO 50, whereas it might actually be
    >> (say) ISO 5. As I wrote, "there's nothing magic about ISO 50."

    >
    >That's rehashing the same unanswered issue.


    The real unanswered issue is, what are the actual so-called "optimum ISO"
    (real/raw sensitivity) values of each sensor? That's needed to make a
    persuasive case for more light -- if, for example, the so-called "optimum ISO"
    of the RGB sensor is 6, then no real case can be made at all, because that's
    too slow to be practical in the real world.

    >Why not base the sensitivity,
    >e.g. threshold exposure to generate a detectable signal above the noise
    >floor, on the signal actually reaching the photosensitive sensor area, i.e.
    >after the filter layer?


    Because (as I've noted several times) light is often a limiting factor. That
    the RGB sensor can attain slightly better S/N at one particular ISO is
    interesting, but hasn't been shown to be terribly useful or important. The
    difference is going to be pretty small (since this is at so-called "optimum
    ISO") and probably insignificant.

    >In general, although filter transmissions can vary between spectral bands to
    >e.g. better match the Human Visual System, the RGB filters transmit roughly
    >half of the incident radiation when compared to similar CMY filters. The
    >only sensible thing to do in that case is to double the exposure on the RGB
    >filters in order to better utilize the full potential well depth.


    That would be all well and good if light were unlimited and/or exposure
    control infinite, but those are conditions that don't exist in much of the
    real world.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #9
  10. "John Navas" <> wrote in message
    news:888bc.3273$...
    > In <c4itul$10q$> on Fri, 2 Apr 2004 05:33:09 +0000 (UTC),
    > (Dave Martindale) wrote:

    SNIP
    > >It's testing both cameras at their optimum ISO.

    >
    > It's testing both cameras at an arbitrary differential ISO, since you
    > haven't established any such actual "optimum ISO" values.


    As described in the ISO 12231 ( Photography - Electronic still-picture
    imaging - Terminology)
    "ISO speed
    < Speed > numerical value calculated from the exposure provided at the focal
    plane of an electronic camera to produce specified camera output signal
    characteristics using the methods described in ISO 12232. The ISO speed
    should correlate with the highest exposure index value that provides peak
    image quality for normal scenes."
    Which is more specifically defined in ISO 12232 (Photography - Digital still
    cameras - Determination of exposure index, ISO speed ratings, standard
    output sensitivity, and recommended exposure index).

    You seem to refer to, not an ISO speed which produces a specified camera
    output signal (peak image quality), but a so called: "recommended exposure
    index (REI)".
    That is defined by the ISO as a:
    "specific exposure index value recommended by a DSC provider as a reference
    for adjusting photographic accessories, as defined in this standard
    NOTE - REI provides a practical exposure index value for setting the
    reference exposure index of light meters, studio lighting, etc, but images
    taken using this exposure index value do not necessarily provide the best
    image quality."

    That would equate to using the same exposure time and aperture, regardless
    of the signal level it generates in the sensor (unless the recommendation is
    to use a different REI for otherwise similar RGB vs CMY sensors). What you
    describe is therefore not ISO
    speed, but comparing equal scene luminances, using the same REI for
    different sensors, which "do not necessarily provide the best image
    quality".

    Bart
     
    Bart van der Wolf, Apr 2, 2004
    #10
  11. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <c4jan2$43e$> on Fri, 2 Apr 2004 09:10:58 +0000 (UTC),
    (Dave Martindale) wrote:

    >John Navas <> writes:
    >
    >>>It's testing both cameras at their optimum ISO.

    >
    >>It's testing both cameras at an arbitrary differential ISO, since you haven't
    >>established any such actual "optimum ISO" values.

    >
    >I haven't shown specific ISO values for specific sensors. But exposing
    >so whites reach a large fraction of the full-well capacity is well known
    >in CCD literature. It's how you make the best use of the dynamic range
    >of whatever sensor you have. It's optimum because if you increase the
    >ISO, you reduce the signal and reduce S/N, while if you decrease the ISO
    >you get the whites saturating and perhaps blooming.
    >
    >So any given CCD has an optimum exposure at a particular number of
    >electrons per pixel, which is equivalent to a particular voltage at the
    >output. Since a CMY filter passes more light than an RGB one, the
    >overall sensitivity of the CMY sensor/filter combo *must* be higher than
    >a RGB sensor/filter using the same sensor. And thus the optimum ISO for
    >the CMY sensor must be higher.


    True, but fails to respond to my point, which is that you haven't established
    any such actual "optimum ISO" values.

    >>I'm saying I would want to use a given ISO speed in a given situation
    >>regardless of sensor.

    >
    >In most situations, there is some room to trade off among ISO, aperture,
    >and shutter speed.


    I don't agree. Since light is often a limiting factor, we usually work at a
    given ISO that's matched to the task at hand. Otherwise there would never
    have been a need for any slide film faster than (say) Kodachrome 25, yet
    faster slide films (e.g., Kodachrome 64) became very popular because the extra
    speed made them more versatile, even when slower film could be uprated with
    push processing.

    If all you're going to do is photograph static subjects in sufficient light on
    a tripod, then indeed speed becomes relatively unimportant, which was of
    course the niche for Kodachrome 25. But if you're going to photograph motion
    and/or in a wide range of light and/or handheld, then speed becomes an issue,
    and faster films (e.g., Kodachrome 64) give you better results, even at the
    expense of increased grain (much like digital sensor noise) and decreased
    sharpness (not a factor with digital sensors). If faster films behaved more
    like digital sensors (i.e., no loss of sharpness), they might have been even
    more popular than slower films.

    >>If you are using an RGB sensor, unless and until you reach your "optimum ISO"
    >>(which hasn't been established), if I'm using a CMY sensor I'll have the
    >>option of matching you on ISO with lower noise, or shooting at higher ISO with
    >>less noise than you at the same ISO.

    >
    >The second half of that isn't clear. As long as the RGB camera isn't at
    >its lowest ISO setting, you have the option of shooting at the same ISO
    >with somewhat less noise, or shooting at about 1 stop higher ISO with
    >somewhat more noise. I agree that's useful.
    >
    >On the other hand, if the RGB camera is used at its lowest ISO, it will
    >give less noise than the CMY camera at any ISO that does not clip the
    >scene whites.


    I agree that's also useful, at least in theory. I just think it's much less
    useful, since it's only at the actual "lowest ISO" for RGB, which you haven't
    established. If, for example, the so-called "optimum ISO" of the RGB sensor
    is 6, then it's not meaningfully useful, because that's too slow to be
    practical in the real world.

    >>Even if you could show my CMY sensor to have an "optimum ISO" at a level high
    >>enough to make a difference in a given situation, there would still be many
    >>other situations where I would still have an advantage because that ISO level
    >>is needed to get the desired results.

    >
    >OK. If higher ISO is important, a CMY sensor is one way of improving
    >ISO. But if the absolute lowest noise is important, regardless of ISO,
    >then an RGB sensor is one way of reducing noise. (Larger photosites
    >can be used to improve either ISO or noise).


    That the RGB sensor can attain slightly better S/N at one particular ISO is
    interesting, but hasn't been shown to be terribly useful or important. It's
    only a nominal 4.5 DB difference, which is barely noticeable, and probably
    insignificant, since it's at so-called "optimum ISO" where S/N is at its
    greatest.

    In summary, I personally think a nominal full stop difference in sensitivity
    and a bit lower noise at most ISOs is more useful than a bit lower noise at
    one particular ISO where noise is already at its lowest point. You may not
    agree, but that's just a difference in our perspectives.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #11
  12. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <406d82f6$0$575$4all.nl> on Fri, 2 Apr 2004 17:12:54 +0200,
    "Bart van der Wolf" <> wrote:

    >"John Navas" <> wrote in message
    >news:888bc.3273$...
    >> In <c4itul$10q$> on Fri, 2 Apr 2004 05:33:09 +0000 (UTC),
    >> (Dave Martindale) wrote:

    >SNIP
    >> >It's testing both cameras at their optimum ISO.

    >>
    >> It's testing both cameras at an arbitrary differential ISO, since you
    >> haven't established any such actual "optimum ISO" values.

    >
    >As described in the ISO 12231 ( Photography - Electronic still-picture
    >imaging - Terminology)
    >"ISO speed
    >< Speed > numerical value calculated from the exposure provided at the focal
    >plane of an electronic camera to produce specified camera output signal
    >characteristics using the methods described in ISO 12232. The ISO speed
    >should correlate with the highest exposure index value that provides peak
    >image quality for normal scenes."
    >Which is more specifically defined in ISO 12232 (Photography - Digital still
    >cameras - Determination of exposure index, ISO speed ratings, standard
    >output sensitivity, and recommended exposure index).
    >
    >You seem to refer to, not an ISO speed which produces a specified camera
    >output signal (peak image quality), but a so called: "recommended exposure
    >index (REI)".
    >That is defined by the ISO as a:
    >"specific exposure index value recommended by a DSC provider as a reference
    >for adjusting photographic accessories, as defined in this standard
    >NOTE - REI provides a practical exposure index value for setting the
    >reference exposure index of light meters, studio lighting, etc, but images
    >taken using this exposure index value do not necessarily provide the best
    >image quality."
    >
    >That would equate to using the same exposure time and aperture, regardless
    >of the signal level it generates in the sensor (unless the recommendation is
    >to use a different REI for otherwise similar RGB vs CMY sensors). What you
    >describe is therefore not ISO
    >speed, but comparing equal scene luminances, using the same REI for
    >different sensors, which "do not necessarily provide the best image
    >quality".


    Not true. Read what I wrote more carefully and in its entirely.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #12
  13. Dave Martindale

    Dave Haynie Guest

    On Thu, 1 Apr 2004 08:05:53 +0000 (UTC), (Dave
    Martindale) wrote:

    > (Dave Haynie) writes:


    >>Think of it this way: IGNORE THE SENSOR TYPE. Pretty much everyone
    >>will, unless they're in this discussion. When you test a camera
    >>against another, head to head, you're going to set comparable
    >>ISO-equivalents, similar exposures (unless you're benchmarking the
    >>accuracy or cleverness of their auto-exposure systems), and you shoot,
    >>and you report on noise. There's no place to stop and say, "whoa, this
    >>is an RBG camera, so I get to drop from ISO200 to ISO100 and double
    >>the exposure". Just doesn't play, given the way everyone uses cameras.


    >I disagree. I'll happily test an RGB camera with a minimum ISO of 50
    >against a CMY camera with a minimum ISO of 100, each at their lowest
    >ISO. That's the condition that gives the best results from each camera,
    >so why not compare them?


    Because my CMY camera has an ISO equivalent setting of 50. You get no
    automatic boost in exposure. And my ISO 50 is quieter than yours.

    >Yes, I know that *at ISO 100* the RGB camera will be noiser than the CMY
    >camera. But I have the choice of using ISO 50 with the RGB camera, and
    >when I do use it, the image is less noisy than the lowest ISO on the CMY
    >camera. This makes me happy. This might make me choose the RGB camera.


    Or the CMY camera with an ISO 50 setting.

    I'm open to discussion, but how about something other than a straw-man
    argument.

    >In fact, I'll bet most people don't know what ISO they shoot at - they
    >just use the default or "auto".


    On consumer P&S cameras, I'm sure that most folks don't much of
    anything about their settings. Who cares? They also tend to produce
    lousy photos even with a really good camera. And their CCD noise won't
    be germaine given that they're only ever producing 4x6 prints. Noise
    at low light levels will be unimporant, since they don't have tripods
    or even a good optical image stabalizer.

    The fact that we're here, having this discussion, implies we actually
    want better tools, and know how to use 'em. So again, that's a
    straw-man.

    Dave Haynie | Chief Toady, Frog Pond Media Consulting
    | Take Back Freedom! Bush no more in 2004!
    "Deathbed Vigil" now on DVD! See http://www.frogpondmedia.com
     
    Dave Haynie, Apr 2, 2004
    #13
  14. Dave Martindale

    Dave Haynie Guest

    On Fri, 2 Apr 2004 05:33:09 +0000 (UTC), (Dave
    Martindale) wrote:

    >The "natural" ISO rating for a sensor is one where scene white causes
    >the CCD's charge well to be close to full, but not saturated. This
    >takes more light for an RGB sensor, so its natural ISO is lower than a
    >CMY sensor based on the same underlying CCD chip. If ISO 50 properly
    >exposes the RGB version of the sensor, ISO 50 will overexpose the CMY
    >one, clipping the highlights - so the CMY camera will not provide that
    >ISO setting.


    Sure it will. They're just make sure to also provide adequate exposure
    range to ensure the CCD won't become useless in bright light. No one
    is actually taking the CCD out of the camera, out from behind the
    shutter and iris, and fretting about it being too sensitive.

    It's also quite possible to include a few ND filters, in your gadget
    bag or in the camera itself, just as you'd get in a pro-class large
    CCD camcorder. No one's building a more sensitive sensor, then
    fretting about it being too sensitive.

    Dave Haynie | Chief Toady, Frog Pond Media Consulting
    | Take Back Freedom! Bush no more in 2004!
    "Deathbed Vigil" now on DVD! See http://www.frogpondmedia.com
     
    Dave Haynie, Apr 2, 2004
    #14
  15. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <> on Fri, 02 Apr 2004 19:17:08 GMT,
    (Dave Haynie) wrote:

    >On Thu, 1 Apr 2004 08:05:53 +0000 (UTC), (Dave
    >Martindale) wrote:
    >
    >> (Dave Haynie) writes:

    >
    >>>Think of it this way: IGNORE THE SENSOR TYPE. Pretty much everyone
    >>>will, unless they're in this discussion. When you test a camera
    >>>against another, head to head, you're going to set comparable
    >>>ISO-equivalents, similar exposures (unless you're benchmarking the
    >>>accuracy or cleverness of their auto-exposure systems), and you shoot,
    >>>and you report on noise. There's no place to stop and say, "whoa, this
    >>>is an RBG camera, so I get to drop from ISO200 to ISO100 and double
    >>>the exposure". Just doesn't play, given the way everyone uses cameras.

    >
    >>I disagree. I'll happily test an RGB camera with a minimum ISO of 50
    >>against a CMY camera with a minimum ISO of 100, each at their lowest
    >>ISO. That's the condition that gives the best results from each camera,
    >>so why not compare them?

    >
    >Because my CMY camera has an ISO equivalent setting of 50. You get no
    >automatic boost in exposure. And my ISO 50 is quieter than yours.


    I don't profess to speak for Dave, but I think he's assuming that our CMY
    cameras just happen to have a natural minimum sensitivity of ISO 100, and thus
    won't even have a lower noise ISO 50 setting. That's an interesting
    hypothesis, but it hasn't been proven, and I think it would be a remarkable
    coincidence. Unless and until that hypothesis is proven, I think we are
    making the more reasonable assumption.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #15
  16. John Navas <> writes:

    >>Then why do you insist to use that, and THEN add filters with a different
    >>transmission amount (integrated over their spectral transmission), as a
    >>basis?


    >Because that's what we're comparing, that's the way the two sensors are built,
    >and that's how they will be used in the real world. The sensitivity of the
    >completed sensor and how it's used in the real world are at least as important
    >as the real/raw sensitivity of the unfiltered semiconductor.


    You seem to be contradicting yourself here. If the sensitivity of the
    completed sensor is what's important, then the ISO rating of the sensor
    should be based on that, and thus the exposure that the sensor receives
    in practice should be based on properly exposing the completed sensor.
    And that means that the RGB sensor will receive more exposure than the
    CMY one, in order to yield the same output voltage.

    Dave
     
    Dave Martindale, Apr 2, 2004
    #16
  17. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <c4kjts$pb$> on Fri, 2 Apr 2004 20:54:20 +0000 (UTC),
    (Dave Martindale) wrote:

    >John Navas <> writes:
    >
    >>>Then why do you insist to use that, and THEN add filters with a different
    >>>transmission amount (integrated over their spectral transmission), as a
    >>>basis?

    >
    >>Because that's what we're comparing, that's the way the two sensors are built,
    >>and that's how they will be used in the real world. The sensitivity of the
    >>completed sensor and how it's used in the real world are at least as important
    >>as the real/raw sensitivity of the unfiltered semiconductor.

    >
    >You seem to be contradicting yourself here. If the sensitivity of the
    >completed sensor is what's important, then the ISO rating of the sensor
    >should be based on that, and thus the exposure that the sensor receives
    >in practice should be based on properly exposing the completed sensor.
    >And that means that the RGB sensor will receive more exposure than the
    >CMY one, in order to yield the same output voltage.


    I don't agree. Read the rest of what I wrote more carefully.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 2, 2004
    #17
  18. "John Navas" <> wrote in message
    news:Ycfbc.3370$...
    >
    > In <406d71f3$0$559$4all.nl> on Fri, 2 Apr 2004 16:00:18

    +0200,
    > "Bart van der Wolf" <> wrote:

    SNIP
    > >Then why do you insist to use that, and THEN add filters with a different
    > >transmission amount (integrated over their spectral transmission), as a
    > >basis?

    >
    > Because that's what we're comparing, that's the way the two sensors are

    built,
    > and that's how they will be used in the real world.


    We, apparently, disagree. My real world camera's exposure meter will
    indicate or set an exposure level to match the ISO speed of the sensor
    (which is different for otherwise similar RGB vs CMY sensors). The ISO (not
    REI, see my other post on that) speed will produce a similar signal level in
    the sensor, and not half of it when RGB filters are used instead of CMY.

    SNIP
    > The real unanswered issue is, what are the actual so-called "optimum ISO"
    > (real/raw sensitivity) values of each sensor? That's needed to make a
    > persuasive case for more light -- if, for example, the so-called "optimum

    ISO"
    > of the RGB sensor is 6, then no real case can be made at all, because

    that's
    > too slow to be practical in the real world.


    I think you'll see that for similar design (photosite area, doping, etc.)
    sensors have a similar sensitivity to incident light, just like the native
    spectral sensitivity is mostly a given due to the properties of Silicon. If
    all else is the same, the integrated spectral transmission of the CFA
    filters is roughly 50% less for RGB versus CMY. The ISO speeds attached to
    those sensors, uses the same criteria, namely to produce a certain signal
    level will then cause very similar noise characteristics, which results in
    different ISOs for otherwise identical sensors. HOWEVER, the postprocessing
    needed to get RGB signals will then increase the noise in the CMY sensor.

    Bart
     
    Bart van der Wolf, Apr 3, 2004
    #18
  19. Dave Martindale

    John Navas Guest

    [POSTED TO rec.photo.digital - REPLY ON USENET PLEASE]

    In <406dff92$0$574$4all.nl> on Sat, 3 Apr 2004 02:04:34 +0200,
    "Bart van der Wolf" <> wrote:

    >"John Navas" <> wrote in message
    >news:Ycfbc.3370$...


    >> Because that's what we're comparing, that's the way the two sensors are built,
    >> and that's how they will be used in the real world.

    >
    >We, apparently, disagree.


    Indeed.

    >My real world camera's exposure meter will
    >indicate or set an exposure level to match the ISO speed of the sensor
    >(which is different for otherwise similar RGB vs CMY sensors).


    It should actually set an exposure level for a given ISO regardless of the
    actual sensitivity of the sensor; e.g., 1/100 second at f/16 at ISO 100 for
    sunny lighting -- when a given ISO is set, the exposure will be the same no
    matter what the sensor.

    >The ISO (not
    >REI, see my other post on that) speed will produce a similar signal level in
    >the sensor, and not half of it when RGB filters are used instead of CMY.


    At a given ISO, the signal level in the RGB sensor won't be the same as for
    CMY -- due to lower sensitivity, RGB will need greater amplification. The
    signal will be the same after amplification, but not before.

    >SNIP
    >> The real unanswered issue is, what are the actual so-called "optimum ISO"
    >> (real/raw sensitivity) values of each sensor? That's needed to make a
    >> persuasive case for more light -- if, for example, the so-called "optimum ISO"
    >> of the RGB sensor is 6, then no real case can be made at all, because that's
    >> too slow to be practical in the real world.

    >
    >I think you'll see that for similar design (photosite area, doping, etc.)
    >sensors have a similar sensitivity to incident light, just like the native
    >spectral sensitivity is mostly a given due to the properties of Silicon. If
    >all else is the same, the integrated spectral transmission of the CFA
    >filters is roughly 50% less for RGB versus CMY. The ISO speeds attached to
    >those sensors, uses the same criteria, namely to produce a certain signal
    >level will then cause very similar noise characteristics, which results in
    >different ISOs for otherwise identical sensors.


    Interesting, but irrelevant, since you didn't address my point regarding the
    actual so-called "optimum ISO" (real/raw sensitivity) values of each sensor.

    >HOWEVER, the postprocessing
    >needed to get RGB signals will then increase the noise in the CMY sensor.


    Other way around -- the amplification needed to bring effective sensitivity of
    RGB up to CMY results in more relative noise for RGB as compared to CMY.

    --
    Best regards,
    John Navas
    [PLEASE NOTE: Ads belong *only* in rec.photo.marketplace.digital, as per
    <http://bobatkins.photo.net/info/charter.htm> <http://rpdfaq.50megs.com/>]
     
    John Navas, Apr 3, 2004
    #19
  20. Dave Martindale

    Guest

    (Dave Haynie) wrote:

    > (Dave Martindale) wrote:
    >
    > >The "natural" ISO rating for a sensor is one where scene white causes
    > >the CCD's charge well to be close to full, but not saturated. This
    > >takes more light for an RGB sensor, so its natural ISO is lower than a
    > >CMY sensor based on the same underlying CCD chip. If ISO 50 properly
    > >exposes the RGB version of the sensor, ISO 50 will overexpose the CMY
    > >one, clipping the highlights - so the CMY camera will not provide that
    > >ISO setting.

    >
    > Sure it will.


    How many times/ways does Mr. Martindale and Mr. van der Wolf have to
    say it? The "ISO" is a physical property of the sensor - how it
    responds to light, the slope of the line that you get when you plot
    sensor output with light input - not the light itself, the shutter
    mechanism and/or its settings, the scene in question or even the
    battery status of the Spirit rover on Mars.

    > They're just make sure to also provide adequate exposure
    > range to ensure the CCD won't become useless in bright light.


    You can play whatever user interface games you like (e.g., the Canon
    10D and its "3200" ISO setting), but the underlying physical reality
    of the sensor remains unchanged: CMY filters permit 2x the light to
    the CMOS/CCD wells as does an RGB filter'd form of exactly the same
    silicon. ie, the ISO is, by the definition posted by Mr. van der
    Wolf, 2x larger.
     
    , Apr 3, 2004
    #20
    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. David J. Littleboy

    Re: The Sigma-Foveon pixel rationale

    David J. Littleboy, Apr 1, 2004, in forum: Digital Photography
    Replies:
    2
    Views:
    383
    Dave Martindale
    Apr 1, 2004
  2. Dave Martindale

    Re: The Sigma-Foveon pixel rationale

    Dave Martindale, Apr 1, 2004, in forum: Digital Photography
    Replies:
    2
    Views:
    401
    Dave Martindale
    Apr 1, 2004
  3. Dave Martindale

    Re: The Sigma-Foveon pixel rationale

    Dave Martindale, Apr 1, 2004, in forum: Digital Photography
    Replies:
    2
    Views:
    339
    Dave Martindale
    Apr 2, 2004
  4. DM

    Re: The Sigma-Foveon pixel rationale

    DM, Apr 2, 2004, in forum: Digital Photography
    Replies:
    30
    Views:
    888
    David Kilpatrick
    Apr 6, 2004
  5. George Preddy

    Re: The Sigma-Foveon pixel rationale

    George Preddy, Apr 4, 2004, in forum: Digital Photography
    Replies:
    3
    Views:
    341
Loading...

Share This Page