MP's vs Effective MP's?

Discussion in 'Digital Photography' started by ArtKramr, Oct 7, 2003.

  1. ArtKramr

    ArtKramr Guest

    What is the difference. MP's sound fine. Effective MP's sounds like stuff made
    up by an ad copywriter.

    Arthur Kramer
    344th BG 494th BS
    England, France, Belgium, Holland, Germany
    Visit my WW II B-26 website at:
    http://www.coastcomp.com/artkramer
     
    ArtKramr, Oct 7, 2003
    #1
    1. Advertising

  2. ArtKramr

    Jim Townsend Guest

    ArtKramr wrote:

    > What is the difference. MP's sound fine. Effective MP's sounds like stuff
    > made up by an ad copywriter.
    >


    Actually effective megapixels is surprisingly.. Truth in advertising.

    Lenses are round and sensors are square. Focusing a round image on a squre
    sensor results in the corners being missed. Imagine putting a penny on a
    postage stamp and how it works should be clear.

    A sensor may be able to produce 5.2 megapixels, but the fact that the corners
    don't see any image cuts that down a bit.. They have to crop a bit to
    eliminate vignetting. (Darkness in the corners).

    So a 5.2 megapixel sensor will only rusult in only 5 effective megapixels.
     
    Jim Townsend, Oct 7, 2003
    #2
    1. Advertising

  3. "ArtKramr" <> wrote:

    > What is the difference. MP's sound fine. Effective MP's sounds like stuff

    made
    > up by an ad copywriter.


    It's a (slight) mistranslation of a technical term that would sound just as
    dizzy if a better translation were available. It's the number of pixels that
    actually measure light in the the sensor. However, not all of those are used
    for imaging. Since it's a bigger number than the "recording pixels", the
    twits in the advertizing department prefer it. Said twits really much prefer
    the total number of pixels, which includes a lot of pixels that aren't used
    at all, but that blatant false advertising got silly, so everyone agreed to
    state the effective pixel count.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 7, 2003
    #3
  4. ArtKramr

    Rafe B. Guest

    On 07 Oct 2003 18:36:41 GMT, (ArtKramr) wrote:

    >What is the difference. MP's sound fine. Effective MP's sounds like stuff made
    >up by an ad copywriter.



    There are layers upon layers of misrepresentation when
    it comes to pixel counts in digicams.

    The most blatant is calling a sensor "six megapixels"
    when in fact it's got six million sensing elements distributed
    among three colors. But this misrepresentation runs
    across the board (Ie., not specific to Brand X or Brand Y.)

    Another layer of misrepresentation is the whole "effective
    pixels" game. The idea is that Brand X pixels are somehow
    better or worth more than the other guys, so Brand X invents
    a multiplication factor with which to fudge the numbers.

    IIRC, Fuji was one of the first to play this game when they
    first released their "super CCD" with the hexagonal photo
    sites.

    It's gotten to be a lot like power ratings in audio gear.


    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Oct 8, 2003
    #4
  5. ArtKramr

    Crownfield Guest

    Jim Townsend wrote:
    >
    > ArtKramr wrote:
    >
    > > What is the difference. MP's sound fine. Effective MP's sounds like stuff
    > > made up by an ad copywriter.
    > >

    >
    > Actually effective megapixels is surprisingly.. Truth in advertising.
    >
    > Lenses are round and sensors are square. Focusing a round image on a squre
    > sensor results in the corners being missed. Imagine putting a penny on a
    > postage stamp and how it works should be clear.
    >
    > A sensor may be able to produce 5.2 megapixels, but the fact that the corners
    > don't see any image cuts that down a bit.. They have to crop a bit to
    > eliminate vignetting. (Darkness in the corners).
    >
    > So a 5.2 megapixel sensor will only rusult in only 5 effective megapixels.


    so my s2 pro,
    using normal nikon lenses, which make a great 24x36 mm image,
    falling on a 15.57x23 mm sensor will not cover the corners?

    think again.
     
    Crownfield, Oct 8, 2003
    #5
  6. ArtKramr

    MikeWhy Guest

    "Crownfield" <> wrote in message
    news:...
    > so my s2 pro,
    > using normal nikon lenses, which make a great 24x36 mm image,
    > falling on a 15.57x23 mm sensor will not cover the corners?
    >
    > think again.


    Well... Zuiko says they won't. Or at least mumbled something about light
    drop-off in the corners from the sensor self-shadowing. ;-)
     
    MikeWhy, Oct 8, 2003
    #6
  7. ArtKramr

    Crownfield Guest

    MikeWhy wrote:
    >
    > "Crownfield" <> wrote in message
    > news:...
    > > so my s2 pro,
    > > using normal nikon lenses, which make a great 24x36 mm image,
    > > falling on a 15.57x23 mm sensor will not cover the corners?
    > >
    > > think again.

    >
    > Well... Zuiko says they won't. Or at least mumbled something about light
    > drop-off in the corners from the sensor self-shadowing. ;-)


    what about the coverage, same lens,
    Kodak 14N with 24x36 mm sensor?
     
    Crownfield, Oct 8, 2003
    #7
  8. ArtKramr

    MikeWhy Guest

    "Crownfield" <> wrote in message
    news:...
    > MikeWhy wrote:
    > >
    > > "Crownfield" <> wrote in message
    > > news:...
    > > > so my s2 pro,
    > > > using normal nikon lenses, which make a great 24x36 mm image,
    > > > falling on a 15.57x23 mm sensor will not cover the corners?
    > > >
    > > > think again.

    > >
    > > Well... Zuiko says they won't. Or at least mumbled something about light
    > > drop-off in the corners from the sensor self-shadowing. ;-)

    >
    > what about the coverage, same lens,
    > Kodak 14N with 24x36 mm sensor?


    This was in context of their 4/3 system; tiny, puny, insignificant stuff
    compared to the Kodak's (or even Hasselblad's) monster. Since I see you
    wanting to make a serious discussion of this, self-shadowing probably isn't
    an issue until they bring the lens as far into the case as Olympus is doing,
    and probably not at the sensor sizes they're talking about otherwise. Not
    being an optical engineer, I can only regurge what the popular press has to
    say, and I suspect it's mostly hyperbole based on slim but real fact. How
    much light fall-off can you tolerate before compensation artifacts become
    objectionable? I dunno... hence the emoticon previously. (Ummm... Having
    forgotten what the original discussion was, I agree with your statement
    above. I look forward to less expensive quality glass that takes advantage
    of the smaller coverage.)
     
    MikeWhy, Oct 8, 2003
    #8
  9. ArtKramr

    Crownfield Guest

    MikeWhy wrote:
    >
    > "Crownfield" <> wrote in message
    > news:...
    > > MikeWhy wrote:
    > > >
    > > > "Crownfield" <> wrote in message
    > > > news:...
    > > > > so my s2 pro,
    > > > > using normal nikon lenses, which make a great 24x36 mm image,
    > > > > falling on a 15.57x23 mm sensor will not cover the corners?
    > > > >
    > > > > think again.
    > > >
    > > > Well... Zuiko says they won't. Or at least mumbled something about light
    > > > drop-off in the corners from the sensor self-shadowing. ;-)

    > >
    > > what about the coverage, same lens,
    > > Kodak 14N with 24x36 mm sensor?

    >
    > This was in context of their 4/3 system; tiny, puny, insignificant stuff
    > compared to the Kodak's (or even Hasselblad's) monster. Since I see you
    > wanting to make a serious discussion of this, self-shadowing probably isn't
    > an issue until they bring the lens as far into the case as Olympus is doing,


    it seems that someone who puts a sensor into a camera,
    and then can not use the whole sensor
    because of design problems is really not good at his job.

    could it be that the angle at which the light hits the sensor,
    with those lenses, is acute enough to cause some fall off?

    this would apply to non film lenses,
    and lenses which are designed to have a very small lens to film
    distance.

    > and probably not at the sensor sizes they're talking about otherwise. Not
    > being an optical engineer, I can only regurge what the popular press has to
    > say, and I suspect it's mostly hyperbole based on slim but real fact. How
    > much light fall-off can you tolerate before compensation artifacts become
    > objectionable? I dunno... hence the emoticon previously. (Ummm... Having
    > forgotten what the original discussion was, I agree with your statement
    > above. I look forward to less expensive quality glass that takes advantage
    > of the smaller coverage.)
     
    Crownfield, Oct 8, 2003
    #9
  10. ArtKramr

    MikeWhy Guest

    "Crownfield" <> wrote in message
    news:...
    > could it be that the angle at which the light hits the sensor,
    > with those lenses, is acute enough to cause some fall off?


    That's the gist of it. Film emulsion isn't "rough" enough to be a problem in
    the same way. The sensor, however, is at a non-zero depth below the surface
    of the chip. Hence, the more acute rays away from the lens axis begins to
    self=shadow.

    > this would apply to non film lenses,
    > and lenses which are designed to have a very small lens to film
    > distance.


    This is where a better understanding of optics would help. Zuiko says the
    light rays in their 4/3 series lenses are closer to parallel, less oblique.
    I think I can see how this is possible, but it seems to me you don't get
    nothin for nothin. Any way, it's probably as much marketing as it is real
    science and fact.
     
    MikeWhy, Oct 8, 2003
    #10
  11. ArtKramr

    Jim Townsend Guest

    Crownfield wrote:


    > it seems that someone who puts a sensor into a camera,
    > and then can not use the whole sensor
    > because of design problems is really not good at his job.
    >


    They can build a camera that uses the whole sensor. The Canon 10D.. Nikon
    D100, and Digtal Rebel all do.

    Actually a lot of light falls outside the sensor..

    But this is a bad thing.. If the light focused from the lens falls outside the
    sensor, you wind up with cropping. This is really bad if you want wide angle.
    And just by coincidence, wide angle is very hard to come by in the DSLR's I
    listed.

    The best approach is to utilize as much as the sensor as you can without going
    outside the boundaries and introducing cropping and a smaller field of view..

    Since the light from the lens is round and the sensor is square, there's no way
    to do this without missing the corners.. You just can't fit a round peg in a
    square hole..


    > could it be that the angle at which the light hits the sensor,
    > with those lenses, is acute enough to cause some fall off?
    >
    > this would apply to non film lenses,
    > and lenses which are designed to have a very small lens to film
    > distance.
    >
    >> and probably not at the sensor sizes they're talking about otherwise. Not
    >> being an optical engineer, I can only regurge what the popular press has to
    >> say, and I suspect it's mostly hyperbole based on slim but real fact. How
    >> much light fall-off can you tolerate before compensation artifacts become
    >> objectionable? I dunno... hence the emoticon previously. (Ummm... Having
    >> forgotten what the original discussion was, I agree with your statement
    >> above. I look forward to less expensive quality glass that takes advantage
    >> of the smaller coverage.)
     
    Jim Townsend, Oct 8, 2003
    #11
  12. ArtKramr

    Guest

    In message <>,
    Rafe B. <> wrote:

    >There are layers upon layers of misrepresentation when
    >it comes to pixel counts in digicams.


    >The most blatant is calling a sensor "six megapixels"
    >when in fact it's got six million sensing elements distributed
    >among three colors. But this misrepresentation runs
    >across the board (Ie., not specific to Brand X or Brand Y.)


    But as far as human perception is concerned, it isn't that much of a
    misrepresentation. The luminance is mostly right for those 6 million
    pixels, and unless you blow the image way up so that you're viewing a 25
    PPI printout in your face, you can't really appreciate color information
    that has a spatial frequency as high as the luminance. You wouldn't be
    seeing much more detail if there were really three separate sensors
    there for red, green, and blue. The Bayer mask sounds horrible at
    first, but in reality, it works rather well, especially with low-noise
    DSLRs.
    --

    <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
    John P Sheehy <>
    ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
     
    , Oct 9, 2003
    #12
  13. ArtKramr

    Rafe B. Guest

    On Thu, 09 Oct 2003 00:23:52 GMT, wrote:

    >In message <>,
    >Rafe B. <> wrote:
    >
    >>There are layers upon layers of misrepresentation when
    >>it comes to pixel counts in digicams.

    >
    >>The most blatant is calling a sensor "six megapixels"
    >>when in fact it's got six million sensing elements distributed
    >>among three colors. But this misrepresentation runs
    >>across the board (Ie., not specific to Brand X or Brand Y.)

    >
    >But as far as human perception is concerned, it isn't that much of a
    >misrepresentation. The luminance is mostly right for those 6 million
    >pixels, and unless you blow the image way up so that you're viewing a 25
    >PPI printout in your face, you can't really appreciate color information
    >that has a spatial frequency as high as the luminance. You wouldn't be
    >seeing much more detail if there were really three separate sensors
    >there for red, green, and blue. The Bayer mask sounds horrible at
    >first, but in reality, it works rather well, especially with low-noise
    >DSLRs.



    Mmmm. Yes, I know that's the theory, but isn't it
    ironic that a 4000 dpi film scan of a 35 mm frame
    gives 20 million *real* pixels (ie, RGB triplets) while
    my 10D gives 6 million interpolated triplets.


    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Oct 9, 2003
    #13
  14. ArtKramr

    Guest

    In message <>,
    Rafe B. <> wrote:

    >Mmmm. Yes, I know that's the theory, but isn't it
    >ironic that a 4000 dpi film scan of a 35 mm frame
    >gives 20 million *real* pixels (ie, RGB triplets) while
    >my 10D gives 6 million interpolated triplets.


    Those aren't "real" either, in the full sense of the word. Scanned film
    is a collection of lots of very innacurate pixels.
    --

    <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<>
    John P Sheehy <>
    ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
     
    , Oct 9, 2003
    #14
  15. ArtKramr

    Rafe B. Guest

    On Thu, 09 Oct 2003 10:36:38 GMT, wrote:

    >In message <>,
    >Rafe B. <> wrote:
    >
    >>Mmmm. Yes, I know that's the theory, but isn't it
    >>ironic that a 4000 dpi film scan of a 35 mm frame
    >>gives 20 million *real* pixels (ie, RGB triplets) while
    >>my 10D gives 6 million interpolated triplets.

    >
    >Those aren't "real" either, in the full sense of the word. Scanned film
    >is a collection of lots of very innacurate pixels.



    Don't get metaphysical on me, now.

    Scanned film is hobbled, I know, by a second optical
    process, which is what gives digicams a big leg up.

    Even so, in terms of real digital content, there are
    60 million measurements in my 35 mm film scans
    (20 million pixels x 3 components) versus 6 million
    in my 10D captures.

    Obviously, the folks pushing Foveon sensors think
    this is a big deal, and use it as another means of
    inflating the "effective MP" count of the sensor.


    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Oct 9, 2003
    #15
  16. "Rafe B." <> wrote in message
    news:...
    > On Thu, 09 Oct 2003 10:36:38 GMT, wrote:
    >
    > >In message <>,
    > >Rafe B. <> wrote:
    > >
    > >>Mmmm. Yes, I know that's the theory, but isn't it
    > >>ironic that a 4000 dpi film scan of a 35 mm frame
    > >>gives 20 million *real* pixels (ie, RGB triplets) while
    > >>my 10D gives 6 million interpolated triplets.

    > >
    > >Those aren't "real" either, in the full sense of the word. Scanned film
    > >is a collection of lots of very innacurate pixels.

    >
    >
    > Don't get metaphysical on me, now.
    >
    > Scanned film is hobbled, I know, by a second optical
    > process, which is what gives digicams a big leg up.
    >
    > Even so, in terms of real digital content, there are
    > 60 million measurements in my 35 mm film scans
    > (20 million pixels x 3 components) versus 6 million
    > in my 10D captures.


    So crop out 6 million of those measurements and print them at 8x12, and then
    tell me how much better each of those measurements are than dSLR pixels...

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 9, 2003
    #16
  17. ArtKramr

    Rafe B. Guest

    On Thu, 9 Oct 2003 21:25:03 +0900, "David J. Littleboy"
    <> wrote:

    >
    >"Rafe B." <> wrote in message
    >news:...
    >> On Thu, 09 Oct 2003 10:36:38 GMT, wrote:
    >>
    >> >In message <>,
    >> >Rafe B. <> wrote:
    >> >
    >> >>Mmmm. Yes, I know that's the theory, but isn't it
    >> >>ironic that a 4000 dpi film scan of a 35 mm frame
    >> >>gives 20 million *real* pixels (ie, RGB triplets) while
    >> >>my 10D gives 6 million interpolated triplets.
    >> >
    >> >Those aren't "real" either, in the full sense of the word. Scanned film
    >> >is a collection of lots of very innacurate pixels.

    >>
    >>
    >> Don't get metaphysical on me, now.
    >>
    >> Scanned film is hobbled, I know, by a second optical
    >> process, which is what gives digicams a big leg up.
    >>
    >> Even so, in terms of real digital content, there are
    >> 60 million measurements in my 35 mm film scans
    >> (20 million pixels x 3 components) versus 6 million
    >> in my 10D captures.

    >
    >So crop out 6 million of those measurements and print them at 8x12, and then
    >tell me how much better each of those measurements are than dSLR pixels...



    OK, I'm being a bit of a devil's advocate here.

    But honestly, when printed at 8x12" there isn't an observable
    difference between a 10D capture and 645 film scanned
    at 4000 dpi.

    (Printed on a Canon S9000; the 645 scan was downsampled
    to 420 dpi at the print, while the 10D image was upsampled to
    300 dpi at the print.)

    What I am struggling with is a way to explain or rationalize
    the apparent 10:1 advantage that digicams have over
    scanned film, strictly according to "information content."

    Or, if you will, a 3:1 advantage simply in terms of pixel count.

    Is film inherently poor at capturing data (relative to modern
    CCD/CMOS sensors) or are the losses incurred mostly in
    the scanning process (ie., the 2nd optical step.)


    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Oct 9, 2003
    #17
  18. "Rafe B." <> wrote:
    > "David J. Littleboy" <> wrote:
    > >"Rafe B." <> wrote in message
    > >> >>Mmmm. Yes, I know that's the theory, but isn't it
    > >> >>ironic that a 4000 dpi film scan of a 35 mm frame
    > >> >>gives 20 million *real* pixels (ie, RGB triplets) while
    > >> >>my 10D gives 6 million interpolated triplets.


    Oops: I see that I misinterpreted your "ironic". Sorry<g>. We're in
    agreement after all.

    > >> Even so, in terms of real digital content, there are
    > >> 60 million measurements in my 35 mm film scans
    > >> (20 million pixels x 3 components) versus 6 million
    > >> in my 10D captures.

    > >
    > >So crop out 6 million of those measurements and print them at 8x12, and

    then
    > >tell me how much better each of those measurements are than dSLR

    pixels...
    >
    >
    > OK, I'm being a bit of a devil's advocate here.
    >
    > But honestly, when printed at 8x12" there isn't an observable
    > difference between a 10D capture and 645 film scanned
    > at 4000 dpi.


    I think you've gone overboard here. If you're subject matter is dense of
    detail, 645 can be peruaded to look better at 11x14 than the 10D can at
    8x12. (But it's so much easier to get shots at the limit of 10D performance
    out of the 10D at every shot, whereas 645 is a god-awful pain in the butt,
    that it's really not worth the bother.)

    > (Printed on a Canon S9000; the 645 scan was downsampled
    > to 420 dpi at the print, while the 10D image was upsampled to
    > 300 dpi at the print.)


    Some of that is because you are scanning negative and not slide film. You
    should see a small difference, but only if the material really justifies it.

    > What I am struggling with is a way to explain or rationalize
    > the apparent 10:1 advantage that digicams have over
    > scanned film, strictly according to "information content."


    We really do agree, though: your 10:1 is pretty much exactly the right
    ratio*.

    The zeroth problem here is that your eye is three times more sensitive to
    luminance detail than color detail, so the three measurements per point vs.
    Bayer makes no perceptual difference. None, nada, zilch. You can't see it.

    > Or, if you will, a 3:1 advantage simply in terms of pixel count.


    But the main problem is that film scans just plain look crappy compared to
    digital at the same pixel magnification.

    *: At 100% pixels, 4000 dpi scans are seriously disgustingly bad when viewed
    at 100%. Carefully downsampled to 60% (1/3 the total pixel count), they look
    a _lot_ better, almost as good as digital capture. That corresponds to a
    2400 dpi scan, and from 645, that's a 3500 x 5280 file, or 18MP. That's a
    _lot_ bigger than 6MP.

    This requires a sharp slide that's scanned at critical focus over the whole
    frame. My 8000 is noticeably softer when focus is off by 15 "focus units".
    With the glass carrier, I can get everything within +/- 10 units, but only
    by taking a lot of focus measurements. I suspect a lot of people who scan
    aren't focusing that accurately, or are focusing on negative films, which
    simply don't record the detail.

    > Is film inherently poor at capturing data (relative to modern
    > CCD/CMOS sensors) or are the losses incurred mostly in
    > the scanning process (ie., the 2nd optical step.)


    IMHO, the former. The "detail" that people find with microscopes is always
    of extremely high contrast targets. For subjects of normal contrast, film's
    MTF is too low, and it can't render that detail with enough contrast to be
    usable.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 9, 2003
    #18
  19. ArtKramr

    Rafe B. Guest

    On Fri, 10 Oct 2003 00:06:26 +0900, "David J. Littleboy"
    <> wrote:

    >
    >"Rafe B." <> wrote:


    >> But honestly, when printed at 8x12" there isn't an observable
    >> difference between a 10D capture and 645 film scanned
    >> at 4000 dpi.

    >
    >I think you've gone overboard here. If you're subject matter is dense of
    >detail, 645 can be peruaded to look better at 11x14 than the 10D can at
    >8x12. (But it's so much easier to get shots at the limit of 10D performance
    >out of the 10D at every shot, whereas 645 is a god-awful pain in the butt,
    >that it's really not worth the bother.)



    You posited an 8x12 print earlier, so that's what I commented
    on. Point being -- at that print size, the printer resolution is the
    limiting factor, IMO. I think it's obvious that at significantly
    larger print sizes, the hi-res MF scan will (or at least should)
    win out.


    >> (Printed on a Canon S9000; the 645 scan was downsampled
    >> to 420 dpi at the print, while the 10D image was upsampled to
    >> 300 dpi at the print.)

    >
    >Some of that is because you are scanning negative and not slide film. You
    >should see a small difference, but only if the material really justifies it.


    I had a look at some of my recent Provia scans. I'm not
    overly impressed by the sharpness. Certainly the lack of
    grain is nice, but I see no noticeable improvement in detail
    over my scans of Reala.


    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Oct 10, 2003
    #19
    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. tenplay
    Replies:
    4
    Views:
    461
    Moz Champion
    Jul 14, 2004
  2. guohan  lu
    Replies:
    0
    Views:
    6,205
    guohan lu
    Nov 19, 2005
  3. Chris
    Replies:
    2
    Views:
    2,741
    Chris
    Dec 3, 2003
  4. Rick
    Replies:
    0
    Views:
    536
  5. fredforcash

    not new but very effective

    fredforcash, Mar 7, 2004, in forum: Computer Support
    Replies:
    11
    Views:
    616
    Th Rt Hon Charles Farquaharson Esq
    Jul 5, 2004
Loading...

Share This Page