Closer to perfection (current camera sensors)

Discussion in 'Digital Photography' started by Me, Mar 10, 2012.

  1. Me

    Me Guest

    Some interesting work by Bill Claff (and others) with an interactive
    chart on "photographic dynamic range" of (mainly) Nikon slr camera
    sensors, dating back a few years:
    http://home.comcast.net/~nikond70/Charts/PDR.htm

    I've taken a screenshot here: http://i41.tinypic.com/1zz75hd.png which
    shows "pdr" for D200, D300, D7000, provisional (no data yet for D800 at
    <ISO400, and limited number of raw files) for D800 in "DX crop" mode,
    and "ideal dx".

    Ideal DX would be defined by a Bayer type sensor with 100% quantum
    efficiency and no read noise.

    * There's only about 1/2 stop "possible improvement" left (see D800 in
    "dx crop mode vs "ideal dx") with bayer sensors.
    * Getting there (to "ideal") requires perfect dyes in RGB filters,
    perfect microlenses, no read noise. I don't think it can happen - we're
    only going to see small gains in future.
    * between the D3s and D4, there's a slight gain in PDR at low ISO from
    read noise reduction, but little else - it stays about the same, but
    with an increase in pixel density.
    * We'll never see a pdr gain of the size of gain between the D200 and
    D300, or D300 and D800 (dx crop mode).
    * Megapixels can still be increased.
    * BSI isn't likely to offer much improvement at current pixel densities
    perhaps it might be able to offset losses if pixel densities increase
    significantly - but I'm guessing that sensors will be oversampling long
    before BSI is worth the effort with dslr sized sensors.
    * Foveon (or another technology such as Nikon's patent with dichroic
    mirrors reflecting RGB to sensels in a Bayer array) might offer some
    improvement in QE.
    * dx has hit the wall compared to Nikon FX (and Canon FX - except at low
    ISO where read noise still limits pdr). Current FX models exceed "ideal
    dx" pdr, so situations such as the D7000 exceeding the pdr of the D700
    aren't likely to be seen again - dx can't "catch up".


    Bill Claff's methodology could be wrong (or I've misinterpreted it), but
    from my use of some of the cameras concerned, it seems spot on.
    Me, Mar 10, 2012
    #1
    1. Advertising

  2. Me

    RichA Guest

    On Mar 9, 9:57 pm, Me <> wrote:
    > Some interesting work by Bill Claff (and others) with an interactive
    > chart on "photographic dynamic range" of (mainly) Nikon slr camera
    > sensors, dating back a few years:http://home.comcast.net/~nikond70/Charts/PDR.htm
    >
    > I've taken a screenshot here:http://i41.tinypic.com/1zz75hd.pngwhich
    > shows "pdr" for D200, D300, D7000, provisional (no data yet for D800 at
    > <ISO400, and limited number of raw files) for D800 in "DX crop" mode,
    > and "ideal dx".
    >
    > Ideal DX would be defined by a Bayer type sensor with 100% quantum
    > efficiency and no read noise.
    >
    > * There's only about 1/2 stop "possible improvement" left (see D800 in
    > "dx crop mode vs "ideal dx") with bayer sensors.
    > * Getting there (to "ideal") requires perfect dyes in RGB filters,
    > perfect microlenses, no read noise.  I don't think it can happen - we're
    > only going to see small gains in future.
    > * between the D3s and D4, there's a slight gain in PDR at low ISO from
    > read noise reduction, but little else - it stays about the same, but
    > with an increase in pixel density.
    > * We'll never see a pdr gain of the size of gain between the D200 and
    > D300, or D300 and D800 (dx crop mode).
    > * Megapixels can still be increased.
    > * BSI isn't likely to offer much improvement at current pixel densities
    > perhaps it might be able to offset losses if pixel densities increase
    > significantly - but I'm guessing that sensors will be oversampling long
    > before BSI is worth the effort with dslr sized sensors.
    > * Foveon (or another technology such as Nikon's patent with dichroic
    > mirrors reflecting RGB to sensels in a Bayer array) might offer some
    > improvement in QE.
    > * dx has hit the wall compared to Nikon FX (and Canon FX - except at low
    > ISO where read noise still limits pdr). Current FX models exceed "ideal
    > dx" pdr, so situations such as the D7000 exceeding the pdr of the D700
    > aren't likely to be seen again - dx can't "catch up".
    >
    > Bill Claff's methodology could be wrong (or I've misinterpreted it), but
    > from my use of some of the cameras concerned, it seems spot on.


    The 5DII rates better than the D4. Sure.
    RichA, Mar 10, 2012
    #2
    1. Advertising

  3. Me

    Me Guest

    On 10/03/2012 9:33 p.m., RichA wrote:
    > On Mar 9, 9:57 pm, Me<> wrote:
    >> Some interesting work by Bill Claff (and others) with an interactive
    >> chart on "photographic dynamic range" of (mainly) Nikon slr camera
    >> sensors, dating back a few years:http://home.comcast.net/~nikond70/Charts/PDR.htm
    >>
    >> I've taken a screenshot here:http://i41.tinypic.com/1zz75hd.pngwhich
    >> shows "pdr" for D200, D300, D7000, provisional (no data yet for D800 at
    >> <ISO400, and limited number of raw files) for D800 in "DX crop" mode,
    >> and "ideal dx".
    >>
    >> Ideal DX would be defined by a Bayer type sensor with 100% quantum
    >> efficiency and no read noise.
    >>
    >> * There's only about 1/2 stop "possible improvement" left (see D800 in
    >> "dx crop mode vs "ideal dx") with bayer sensors.
    >> * Getting there (to "ideal") requires perfect dyes in RGB filters,
    >> perfect microlenses, no read noise. I don't think it can happen - we're
    >> only going to see small gains in future.
    >> * between the D3s and D4, there's a slight gain in PDR at low ISO from
    >> read noise reduction, but little else - it stays about the same, but
    >> with an increase in pixel density.
    >> * We'll never see a pdr gain of the size of gain between the D200 and
    >> D300, or D300 and D800 (dx crop mode).
    >> * Megapixels can still be increased.
    >> * BSI isn't likely to offer much improvement at current pixel densities
    >> perhaps it might be able to offset losses if pixel densities increase
    >> significantly - but I'm guessing that sensors will be oversampling long
    >> before BSI is worth the effort with dslr sized sensors.
    >> * Foveon (or another technology such as Nikon's patent with dichroic
    >> mirrors reflecting RGB to sensels in a Bayer array) might offer some
    >> improvement in QE.
    >> * dx has hit the wall compared to Nikon FX (and Canon FX - except at low
    >> ISO where read noise still limits pdr). Current FX models exceed "ideal
    >> dx" pdr, so situations such as the D7000 exceeding the pdr of the D700
    >> aren't likely to be seen again - dx can't "catch up".
    >>
    >> Bill Claff's methodology could be wrong (or I've misinterpreted it), but
    >> from my use of some of the cameras concerned, it seems spot on.

    >
    > The 5DII rates better than the D4. Sure.
    >

    If that's what you read from the chart, god help you if you try to make
    sense of a histogram on your camera LCD.
    Me, Mar 10, 2012
    #3
  4. Me

    Me Guest

    On 10/03/2012 10:48 p.m., Alfred Molon wrote:
    > In article<jjefu7$dg0$>, Me says...
    >> Ideal DX would be defined by a Bayer type sensor with 100% quantum
    >> efficiency and no read noise.

    >
    > No, an ideal sensor would be a full-colour sensor, not a Bayer one. But
    > at the moment full-colour sensor technology is not good enough yet.
    >

    I think I mentioned that. Yes, an "ideal" foveon outperforms an "ideal"
    bayer sensor.
    Me, Mar 10, 2012
    #4
  5. Me

    nospam Guest

    In article <>, David J.
    Littleboy <> wrote:

    > Which is to say, I don't get the folks complaining about the Bayer array +
    > AA filter technology. It produces amazing images, and trying to do better is
    > a fool's errand, since all you get is worse color/noise performance (real
    > life Foveon) and horrific artifacts (from leaving out the mathematically
    > required AA filter).


    and colour shifts because foveon is not actually measuring red, green
    and blue. the 3 layers need to be converted to rgb. worse, the colour
    shifts vary depending on the subject and exposure, which makes it a
    royal pain in the ass to fix.
    nospam, Mar 10, 2012
    #5
  6. Me

    nospam Guest

    In article <>, Alfred
    Molon <> wrote:

    > > Ideal DX would be defined by a Bayer type sensor with 100% quantum
    > > efficiency and no read noise.

    >
    > No, an ideal sensor would be a full-colour sensor, not a Bayer one. But
    > at the moment full-colour sensor technology is not good enough yet.


    and not likely to ever be, plus if you oversample with bayer, full
    colour sensors are not needed.
    nospam, Mar 10, 2012
    #6
  7. On 3/10/2012 7:49 AM, David J. Littleboy wrote:

    >
    > Which is to say, I don't get the folks complaining about the Bayer array +
    > AA filter technology. It produces amazing images, and trying to do better is
    > a fool's errand,


    on a single ship: three chip cameras with dichroic filters are much better.

    Doug McDonald
    Doug McDonald, Mar 10, 2012
    #7
  8. Me

    me Guest

    On Sat, 10 Mar 2012 09:59:38 -0600, Doug McDonald
    <> wrote:

    >On 3/10/2012 7:49 AM, David J. Littleboy wrote:
    >
    >>
    >> Which is to say, I don't get the folks complaining about the Bayer array +
    >> AA filter technology. It produces amazing images, and trying to do better is
    >> a fool's errand,

    >
    >on a single ship: three chip cameras with dichroic filters are much better.


    Would that be a sailing ship or a cabin cruiser?
    me, Mar 10, 2012
    #8
  9. Me

    John A. Guest

    On Sat, 10 Mar 2012 11:04:19 -0500, me <> wrote:

    >On Sat, 10 Mar 2012 09:59:38 -0600, Doug McDonald
    ><> wrote:
    >
    >>On 3/10/2012 7:49 AM, David J. Littleboy wrote:
    >>
    >>>
    >>> Which is to say, I don't get the folks complaining about the Bayer array +
    >>> AA filter technology. It produces amazing images, and trying to do better is
    >>> a fool's errand,

    >>
    >>on a single ship: three chip cameras with dichroic filters are much better.

    >
    >Would that be a sailing ship or a cabin cruiser?


    Doesn't matter. Once you split your chips between any two or three
    ships it's well-nigh impossible to get the image registration right.
    (The parallax for 3D imagery, however, is amazing.)
    John A., Mar 10, 2012
    #9
  10. Me

    nospam Guest

    In article <jjftp3$lri$>, Doug McDonald
    <> wrote:

    > > Which is to say, I don't get the folks complaining about the Bayer array +
    > > AA filter technology. It produces amazing images, and trying to do better is
    > > a fool's errand,

    >
    > on a single ship: three chip cameras with dichroic filters are much better.


    only if you don't mind the effect of the beamsplitters on lenses, the
    extra weight and the extra cost for 3 chips rather than 1 and the
    precision to keep everything in perfect alignment.

    other than that.
    nospam, Mar 10, 2012
    #10
  11. Me

    nospam Guest

    In article <>, Alfred
    Molon <> wrote:

    > > and not likely to ever be, plus if you oversample with bayer, full
    > > colour sensors are not needed.

    >
    > Nonsense. Besides oversampling with Bayer is a very poor idea. If you
    > want to oversample you better use full-colour pixels.


    actually it's a very good idea.
    nospam, Mar 10, 2012
    #11
  12. Me

    Ray Fischer Guest

    nospam <> wrote:
    >In article <>, Alfred
    >Molon <> wrote:
    >
    >> > and not likely to ever be, plus if you oversample with bayer, full
    >> > colour sensors are not needed.

    >>
    >> Nonsense. Besides oversampling with Bayer is a very poor idea. If you
    >> want to oversample you better use full-colour pixels.

    >
    >actually it's a very good idea.


    Foveon religious zealots are as irrational as any other religious
    zealot.

    --
    Ray Fischer | None are more hopelessly enslaved than those who falsely believe they are free.
    | Goethe
    Ray Fischer, Mar 11, 2012
    #12
  13. Me

    nospam Guest

    In article <4f5c0089$0$12041$>, Ray Fischer
    <> wrote:

    > Foveon religious zealots are as irrational as any other religious
    > zealot.


    but far more entertaining.
    nospam, Mar 11, 2012
    #13
  14. Me

    Ray Fischer Guest

    Rich <> wrote:
    >Doug McDonald <> wrote in
    >> On 3/10/2012 7:49 AM, David J. Littleboy wrote:


    >>> Which is to say, I don't get the folks complaining about the Bayer
    >>> array + AA filter technology. It produces amazing images, and trying
    >>> to do better is a fool's errand,

    >>
    >> on a single ship: three chip cameras with dichroic filters are much
    >> better.

    >
    >That's what they use in scientific experiments, astronomy.


    It's not what they use in scientific experiments or astronomy. They
    use cooled monochromatic sensors and rotate filters in front between
    exposures.

    > You can't use
    >Bayer because it can't acccurately reproduce colour.


    You can't use Foveon because it can't acccurately reproduce colour.

    --
    Ray Fischer | None are more hopelessly enslaved than those who falsely believe they are free.
    | Goethe
    Ray Fischer, Mar 11, 2012
    #14
  15. Me

    Bruce Guest

    Rich <> wrote:

    >"David J. Littleboy" <> wrote in
    >news::
    >
    >>
    >> "Me" <> wrote in message
    >> news:jjfhuv$ifp$...
    >>> On 10/03/2012 10:48 p.m., Alfred Molon wrote:
    >>>> In article<jjefu7$dg0$>, Me says...
    >>>>> Ideal DX would be defined by a Bayer type sensor with 100% quantum
    >>>>> efficiency and no read noise.
    >>>>
    >>>> No, an ideal sensor would be a full-colour sensor, not a Bayer one.
    >>>> But at the moment full-colour sensor technology is not good enough
    >>>> yet.
    >>> >
    >>> I think I mentioned that. Yes, an "ideal" foveon outperforms an
    >>> "ideal" bayer sensor.

    >>
    >> I suppose. But a real Bayer gets quite close to an ideal Bayer,
    >> whereas Foveon needs 100% transparency to the bands not detected in
    >> the top two layers, and that's never going to happen, not even close.
    >> Also, getting high QE out of Foveon in the detection layers is going
    >> to be way harder.
    >>
    >> Meanwhile, as I keep saying over and over again, Bayer is flipping
    >> amazing. 12MP FF Bayer makes 35mm film look sick at 12x18. There's
    >> just no comparison. And 36MP FF Bayer is going to match 6x9 film at
    >> 16x24. This is friggin' amazing: a measly 24x36mm of silicon competing
    >> with 56x92 mm of film.
    >>
    >> Which is to say, I don't get the folks complaining about the Bayer
    >> array + AA filter technology. It produces amazing images, and trying
    >> to do better is a fool's errand, since all you get is worse
    >> color/noise performance (real life Foveon) and horrific artifacts
    >> (from leaving out the mathematically required AA filter).
    >>

    >
    >Is...the sky...really falling on Nikon?



    No, the sky is falling in on Littleboy and his ilk.

    Problem is, they are too busy peeping at pixels, searching for the
    elusive moire and aliasing to notice the sky is falling. ;-)
    Bruce, Mar 11, 2012
    #15
  16. Me

    Me Guest

    On 10/03/2012 3:57 p.m., Me wrote:
    > Some interesting work by Bill Claff (and others) with an interactive
    > chart on "photographic dynamic range" of (mainly) Nikon slr camera
    > sensors, dating back a few years:
    > http://home.comcast.net/~nikond70/Charts/PDR.htm

    ....
    > Bill Claff's methodology could be wrong


    Oops. Indeed he did made a mistake with data for 5dII/III (now corrected)
    There's also now low ISO data now for the D800 on the link above.
    Me, Mar 11, 2012
    #16
  17. Me

    nospam Guest

    In article <>, Alfred
    Molon <> wrote:

    > > You can't use Foveon because it can't acccurately reproduce colour.

    >
    > The problem with Bayer is that you don't capture the full colour
    > information in each pixel and end up running into trouble when
    > dimensioning the AA filter.


    yet camera makers figured it out.

    > It's a huge mess and one of the reason why
    > cameras using a Bayer sensor have poor pixel level sharpness.


    it's not a mess at all and 'pixel level sharpness' is meaningless
    twaddle.
    nospam, Mar 11, 2012
    #17
  18. On 3/11/2012 7:20 AM, Neil Ellwood wrote:

    >> In article<jjftp3$lri$>, Doug McDonald
    >> <> wrote:


    >>> three chip cameras with dichroic filters are much
    >>> better.

    >>

    >
    > This sounds like a system that was around almost 70 years ago... -
    > Technicolor.
    >
    >
    >

    Correct. But it is still used for top end broadcast TV cameras.

    The main problem is, as someone said, the requirement for a
    large lens-sensor spacing .... in glass. It is also
    difficult to get low f/numbers. The pro broadcast TV cameras typically
    have a single lens with a humongous zoom range. Remember that
    real TV is really low resolution (in actual fact never more
    than 720x1440 despite having 1080x1920 pixels ... "Kell factor" ...
    because TV refuses to use 1080x1920@30p.)

    Doug McDonald
    Doug McDonald, Mar 11, 2012
    #18
  19. Me

    PeterN Guest

    On 3/11/2012 5:57 AM, Bruce wrote:
    > Rich<> wrote:
    >
    >> "David J. Littleboy"<> wrote in
    >> news::
    >>
    >>>
    >>> "Me"<> wrote in message
    >>> news:jjfhuv$ifp$...
    >>>> On 10/03/2012 10:48 p.m., Alfred Molon wrote:
    >>>>> In article<jjefu7$dg0$>, Me says...
    >>>>>> Ideal DX would be defined by a Bayer type sensor with 100% quantum
    >>>>>> efficiency and no read noise.
    >>>>>
    >>>>> No, an ideal sensor would be a full-colour sensor, not a Bayer one.
    >>>>> But at the moment full-colour sensor technology is not good enough
    >>>>> yet.
    >>>>>
    >>>> I think I mentioned that. Yes, an "ideal" foveon outperforms an
    >>>> "ideal" bayer sensor.
    >>>
    >>> I suppose. But a real Bayer gets quite close to an ideal Bayer,
    >>> whereas Foveon needs 100% transparency to the bands not detected in
    >>> the top two layers, and that's never going to happen, not even close.
    >>> Also, getting high QE out of Foveon in the detection layers is going
    >>> to be way harder.
    >>>
    >>> Meanwhile, as I keep saying over and over again, Bayer is flipping
    >>> amazing. 12MP FF Bayer makes 35mm film look sick at 12x18. There's
    >>> just no comparison. And 36MP FF Bayer is going to match 6x9 film at
    >>> 16x24. This is friggin' amazing: a measly 24x36mm of silicon competing
    >>> with 56x92 mm of film.
    >>>
    >>> Which is to say, I don't get the folks complaining about the Bayer
    >>> array + AA filter technology. It produces amazing images, and trying
    >>> to do better is a fool's errand, since all you get is worse
    >>> color/noise performance (real life Foveon) and horrific artifacts
    >>> (from leaving out the mathematically required AA filter).
    >>>

    >>
    >> Is...the sky...really falling on Nikon?

    >
    >
    > No, the sky is falling in on Littleboy and his ilk.
    >
    > Problem is, they are too busy peeping at pixels, searching for the
    > elusive moire and aliasing to notice the sky is falling. ;-)
    >


    Aw Brucie, is moire bad for your sales?


    --
    Peter
    PeterN, Mar 11, 2012
    #19
  20. Me

    Ray Fischer Guest

    Alfred Molon <> wrote:
    > Ray Fischer
    >> You can't use Foveon because it can't acccurately reproduce colour.

    >
    >The problem with Bayer is that you don't capture the full colour
    >information in each pixel and end up running into trouble when
    >dimensioning the AA filter. It's a huge mess and one of the reason why
    >cameras using a Bayer sensor have poor pixel level sharpness.


    That's a non-argument.

    1) Bayer cameras actually work very well.
    2) "pixel level sharpness" is utterly irrelevant.

    A 12MP Bayer sensor has as many "full-color pixels" as does a 4MP Foveon
    sensor with the added benefit of greater control of color fidelity and
    much easier manufacturing and higher resolution.

    Results count, not techo-whining.

    --
    Ray Fischer | None are more hopelessly enslaved than those who falsely believe they are free.
    | Goethe
    Ray Fischer, Mar 11, 2012
    #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. lmnop

    Getting closer to to rip/burn decisions

    lmnop, Nov 27, 2004, in forum: Computer Support
    Replies:
    15
    Views:
    768
    Toolman Tim
    Nov 27, 2004
  2. Annika1980

    D60: CLOSER THAN YOU WANNA BE !!!

    Annika1980, Jul 22, 2003, in forum: Digital Photography
    Replies:
    4
    Views:
    335
    Roger Halstead
    Jul 24, 2003
  3. mimi
    Replies:
    62
    Views:
    1,680
    Urra Dipschitt
    Jun 23, 2005
  4. joe

    Closer!!!

    joe, Nov 17, 2004, in forum: Digital Photography
    Replies:
    3
    Views:
    430
    =?ISO-8859-1?Q?BenOne=A9?=
    Nov 19, 2004
  5. steve

    closer...closer....RAM-based system

    steve, Sep 12, 2006, in forum: NZ Computing
    Replies:
    7
    Views:
    419
    Dogboy
    Sep 15, 2006
Loading...

Share This Page