D-SLR Sensor Resolution and Sensor Size Comparison Size Matters!

Discussion in 'Digital Photography' started by Steven M. Scharf, May 14, 2004.

  1. I put a chart on http://sigmasd10.com that shows the differences in
    resolution and actual sensor size of various digital SLRs. I have reproduced
    it below (use mono-spaced font).

    Model Megapixels Sensor Size (mm)
    -------------------------------------------------
    Sigma SD-9, SD-10 3.4 21x14 (58% of full frame)
    Nikon D2H 4 24x16 (67% of full frame)
    Olympus E1 5 18x14 "4/3"
    Canon EOS-300D 6 23x15
    Canon EOS-10D 6 23x15
    Nikon D70 6 24x16 (67% of full frame)
    Nikon D100 6 24x16 (67% of full frame)
    Fuji S3 Pro 6 23x16
    Pentax *ist D 6 24x16 (67% of full frame)
    Canon EOS-1D Mark II 8.2 29x19 (80% of full frame)
    Canon EOS-1Ds 11 36x24 (full frame)
    Kodak DCS SLR/n 13.5 36x24 (full frame)
    Kodak DCS SLR/c 13.5 36x24 (full frame)

    For reference, the APS frame size is 30x17 mm

    The number of megapixels relates directly to picture quality and the ability
    to do large enlargements. For very large enlargements (> 8 x10) you'll want
    to choose a camera with at least 6 Mp. For professional work that requires
    20"x30" enlargements you'll want at least 8 Mp.

    Sensor size also matters. The cheaper cameras use smaller sensors, cramming
    more pixels into a smaller area, because these are less expensive to
    manufacture. You get more sensors per wafer when you have a smaller sensor
    size.

    Ironically, the larger geometry fabrication processes can yield lower noise
    sensors. One reason for the noise issues on the Sigma cameras, is that the
    Foveon sensor is fabricated with relatively small geometry (0.18 micron).
    Contrast this with the Pentium IV which is fabricated with 90 nm and 0.13
    micron technology, and the Foveon sensor seems to use a large geometry. But
    compare it with the fabrication technology used by Canon and Nikon, and the
    Foveon X3 uses a much smaller geometry.

    You have to strike a balance between megapixels and noise. Kodak boasts a
    13.5 Mp sensor, but the noise levels make it less desirable than the 11 Mp
    sensor in the Canon EOS-1Ds. OTOH, the relatively noisy Foveon X3 sensor
    proves that a lower megapixel sensor can still be made to have a lot of
    noise.

    Ideally, we'll see more full-frame sensors, rather than going to smaller
    geometries. We've seen what happens when the number of megapixels is
    increased without going to a larger sized sensor (by going to smaller
    pixels). But the cost difference of full frame sensors, versus smaller
    sensors such as used by Foveon, is significant.

    The 4/3 idea is very disturbing. If it takes hold, it will mean a sharp
    distinction between professional level D-SLRs and amateur and Prosumer level
    D-SLRS. One good thing about Canon, is that they tend to let their higher
    end technology percolate down into their lower cost products, after a while.
     
    Steven M. Scharf, May 14, 2004
    #1
    1. Advertisements

  2. Steven M. Scharf

    Lourens Smak Guest

    This is not entirely true. not all megapixels are created equal. For
    example, when shooting at F2.8 with a wideangle on the 1Ds, the corners
    will have very little actual image information. A very good F8 shot with
    a 6MP camera might contain more image detail in the corners. ( and with
    the E-1, F2.8 performance is already disturbingly good...)

    Same with scanning; I find that my normal 4000dpi 35mm scans at about
    18MP, have similar amount of image-detail as a 6MP digital capture. Of
    course the scans are much bigger, but they don't contain much more
    "image". (a bit more, depending on the film used)

    To double image resolution you need 4 times as many megapixels. the
    difference between 5 and 8MP isn't that large. More is better, I agree
    with that, but the total picture just isn't that simple. A good 5MP will
    beat a 8MP with a lousy lens anytime.
    ??? Please explain...
    I have created images (professionally) with a 1Ds, D1x, and E-1, and I
    fail to see this sharp distinction.

    Lourens
     
    Lourens Smak, May 14, 2004
    #2
    1. Advertisements

  3. Steven M. Scharf

    gsum Guest

    True, the Canon is better all round than the Kodak but
    it depends what you want to do with the camera. The Kodak
    displays no noise at ISO 80 and, as it has no microlenses,
    has minimal chromatic aberration. For landscapes, the Kodak
    is unbeatable at the moment.

    Graham
     
    gsum, May 14, 2004
    #3
  4. Steven M. Scharf

    dylan Guest

    "when shooting at F2.8 with a wideangle on the 1Ds, the corners
    will have very little actual image information. A very good F8 shot with
    a 6MP camera might contain more image detail in the corners."

    Can you explain a bit more, are you saying a 35mm frame doesn't have detail
    in the corners ?

    Thanks
     
    dylan, May 14, 2004
    #4
  5. Steven M. Scharf

    Lourens Smak Guest

    No but some lenses (most wides actually) are quite soft in the corners
    and most suffer from severe light fall-off too at wide apertures. The
    difference can be as big as 2 stops between center and corner.
    Obviously, both these effects makes one lose detail.
    For example the 16-35 L lens from Canon isn't very good at F2.8 in the
    corners, (dark, soft and lack of contrast) and the quality-difference
    from center to corner is really huge at 16mm/F2.8

    With telephoto things are a bit different but many tele-zooms do have
    light fall-off. Sharpness is usually better across the entire frame.
    (with less degradation towards the corners)

    Of course, shooting at F8 will be better but usually this isn't an
    alternative because of the light-level. And if you can use the 6MP @ F8,
    you could also use an 11MP camera @ F8...

    ;-)
    Lourens
     
    Lourens Smak, May 14, 2004
    #5
  6. Steven M. Scharf

    dylan Guest

    Thanks for your explanation.

    I would say the 'No' at the beginning should be 'Yes' because you are saying
    the corners do have less detail with wide-angles when covering a 35mm frame,
    particularly at wide aperture :)

    Cheers
     
    dylan, May 14, 2004
    #6
  7. So stop down. And compare the same angle of view. Talking about "losing
    detail" seems odd: the E-1/10D can't do a 16mm equivalent at f/2.8. The
    E-1's widest is 22mm, and the 10D with the 12-24 is a 19.5mm f/4.5 lens that
    needs to be stopped down quite a bit. Since none of these cameras gets
    anywhere near 16mm, however bad the 1Ds is at 16mm, it's infinitely better
    than the E-1/10D.

    By the way, I bet the E-1's 11-22 is a mess in the corners at 11mm and
    f/2.8.
    So? The widest lens for the E-1 is 22m: the 11-22mm zoom. I bet that with
    both at f/2.8, with the E-1 at 11mm and the Canon at 22mm, If you made A4
    prints from both, the Canon would be significantly better than the E-1. Even
    at the corners.

    In any reasonable comparison (same angle of view, same f stop), the 1Ds is
    going to capture more detail than either the E-1 or a 6MP dSLR. If
    performance flags in places the E-1/10D can't even go, you're still way
    ahead.
    No, that's the cup half-empty approach. You should be glad you've got f/2.8
    even if it's funky, because there are a lot of images that don't depend on
    the corners. (I'm thinking of buying a fixed-lens camera with a 28mm
    equivalent lens that's only f/5.6.)
    Yes. The "lose detail" concept is bogus.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, May 14, 2004
    #7
  8. Hello Steven

    Excellent post.

    One note Re: your chart: your sensor size column
    percentages might be a bit misleading, on the large
    side, for folks. That's because you're giving percentages
    along just one dimension of the sensor, rather than
    the area of each sensor. For example, the Nikon D70 sensor,
    at 24x16, is (24/36)*(16/24) = 4/9, or approx. 44% of full frame,
    not 67%.

    -- stan
     
    Stanley Krute, May 14, 2004
    #8
  9. Steven M. Scharf

    Guest

    You'd lose that bet.
    The 8-16mm is due out this year, probably at Photokina.

    BTW the so-called 6mp cameras crop to the same or smaller than the 5 mp
    E-1 on most paper sizes.
     
    , May 14, 2004
    #9
  10. The smaller the sensor the better, provided it takes 35mm format
    lenses, because optically the lens is dramatically better in the
    middle. Wiegh that consideration with the total sensor count,
    obviously optical pixels is better. That is why Sigma, having both
    the highest number of optical pixels (interpolation is meaningless,
    only optical MPs matter) and the smallest sensor size, is several
    steps ahead of anything offered by any other manufacturer.

    The Kodak 14n (3.3MP optical; 13.5MP interpolated) has almost as many
    MPs as the Sigma SDs (3.43 optical; 13.72MP interpolated), but it is
    full frame which is makes the optical quality totally unacceptable by
    comparison. Assuming the same MPs even the optical resolution over
    the respective image areas, croppers are dramatically better cameras.

    IOWs, Sigma builds the only pro quality DSLRs currently available.
     
    George Preddy, May 14, 2004
    #10
  11. That is optical MPs, the SDs both natively output 13.72MP at 75%
    interpolated.

    All the rest in the list are only 75% interpolated resolutions...
    Smaller sensor size is dramatically better assuming the same lens, it
    is small sensor pitch (the lateral distance between two) that is cheap
    and inferior. You are hopelessly confused, but don't feel too bad,
    most people are also lost when it comes to these fairly complex
    concepts.

    Simga uses all 3 dimensions to pack sensors vs only 2D for all Bayers,
    so Sigma's snesor pitch (higher is better) is utterly enormous
    compared to all Bayers (including full frame) even after Sigmas
    massively higher sensor count. This is why Sigma images are totally
    noise free at ISO 100, while Canon is awash in noise at ISO 100, and
    Nikon can't even do ISO 100.
     
    George Preddy, May 14, 2004
    #11
  12. Steven M. Scharf

    Robertwgross Guest

    Steve, it certainly seems to me that there is a new candidate for Chief
    Technology Officer of the Nordic Group. (G.P.)

    ---Bob Gross---
     
    Robertwgross, May 14, 2004
    #12
  13. http://www.dilbert.com/comics/dilbert/archive/dilbert-20040511.html

    Bart
     
    Bart van der Wolf, May 14, 2004
    #13
  14. He's already employed full-time:
    http://www.dilbert.com/comics/dilbert/archive/dilbert-20040511.html

    Bart
     
    Bart van der Wolf, May 14, 2004
    #14
  15. Alternately, take your pick from one or a combination of:

    http://www.winternet.com/~mikelr/flame84.html
    http://www.winternet.com/~mikelr/flame63.html
    http://www.winternet.com/~mikelr/flame47.html
    http://www.winternet.com/~mikelr/flame16.html

    At times I fear 'we' are outmatched. In the presense of a master
    http://www.winternet.com/~mikelr/flame78.html
    we are all doomed! ;)

    -Greg
     
    Greg Campbell, May 15, 2004
    #15
  16. Steven M. Scharf

    Paul Howland Guest

    Haha what nonsense George. Keep 'em coming. Still no example
    photographs of your own for us George? A scan from a disposable camera
    would do, if that's all you have.
     
    Paul Howland, May 15, 2004
    #16
  17. So it's 67% of the dimensions, but 44% of the area. Most people would
    say that's 67% of the size, not 44% of the size.

    If you switch lenses from 50 to 100 mm focal length on an SLR, is the
    image twice as large or four times as large? If you look through 7x35
    binoculars, would you say that things look 7 times larger or 49 times
    larger?

    Normally, "size" refers to scale factor, or linear magnification, not
    area.

    Dave
     
    Dave Martindale, May 15, 2004
    #17
  18. Steven M. Scharf

    DJ Guest

    Goerge's Earth is flat, and he lives on the under-side.
     
    DJ, May 15, 2004
    #18
  19. Most people would say 44% of the area is 44% of the size.
     
    George Preddy, May 15, 2004
    #19
  20. Looks like another person was fooled by Bayer manufacturers and didn't
    know. Bayer sensors were interpolated.
     
    George Preddy, May 15, 2004
    #20
    1. Advertisements

Ask a Question

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

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.