What is holding back full size sensors? Just marketing?

Discussion in 'Digital Photography' started by Bay Area Dave, May 24, 2004.

    Bart van der Wolf, May 27, 2004
    1. Advertisements

  2. Bay Area Dave

    bagal Guest

    Yup - good point Roger and well said

    It doesn't matter how super-duper the sensor is it does not operate in

    It needs an operating system and that in itself has its own parameters and
    physical limitations.

    Then it has to talk to devices within and outwith the camera itself

    In order to get the best overall structure everything has to be matched.

    In a car, for example, taking the camshaft from a Merc, in a BMW mounting
    with a Ferrari fuel system may look like the best bit of sherry picking.
    But that's all it is, cherry picking. It wouldn't be an effective engine.
    Similarly so with digital cameras - everything has to match, co-ordinate and
    work. The creative thought gone into with these devices is really rather

    Good point!

    das bagal
    bagal, May 27, 2004
    1. Advertisements

  3. Depending on the job, you can use CCDs with perhaps a lot of defects if
    they are of the right kind, in the right place.

    On the cost side, about 10 years ago, a 2k x 2k Tek CCD cost from $80
    each for one that would tell you that the sun was out, if it was in a
    good mood, to $450,000 for a A grade unit with a max of 3 pixels with
    off gain and no more than 1 dead pixels at the outer edges.

    That was before you thinned it...

    Paul Repacholi 1 Crescent Rd.,
    +61 (08) 9257-1001 Kalamunda.
    West Australia 6076
    comp.os.vms,- The Older, Grumpier Slashdot
    Raw, Cooked or Well-done, it's all half baked.
    EPIC, The Architecture of the future, always has been, always will be.
    Paul Repacholi, May 27, 2004
  4. No, size has nothing to do with sensitivity, except due to changes in
    the overlying layers. What does change is the well capacity of the

    When the pixel gets close to its max capacity, it gets non-linear due
    to charge spill into neibours and leakage. If it is an anti-blooming
    type, this can get bad at 1/4 to 1/3 of max capacity. Some of your
    noise sources scale with pixel size as well, but some do not and that
    favours big wells.

    A possible way out, is to have a BIG reduction in pixel size, then sum
    them. By making the pixels so small that you had at least 2 or 3 in
    the image core of the best lenses, you can get rid of the anti-alias
    filter. Or, more correctly, use the lens as an AA filter, then add
    pixels to reduce noise and get your data size reasonable.
    Interpolation can be used here with very little impact on image

    Paul Repacholi 1 Crescent Rd.,
    +61 (08) 9257-1001 Kalamunda.
    West Australia 6076
    comp.os.vms,- The Older, Grumpier Slashdot
    Raw, Cooked or Well-done, it's all half baked.
    EPIC, The Architecture of the future, always has been, always will be.
    Paul Repacholi, May 27, 2004
  5. It depends on how you look at it. If you always expose so the whites
    are the same percentage of full-well capacity, then the larger sensing
    area and larger capacity cancel each other, and the sensitivity remains
    the same (more or less) - but the signal to noise improves.

    On the other hand, if you expose the large-pixel sensor to about the
    same number of electrons, you get higher sensitivity proportional to the
    pixel area. Now you're not using the full capacity of the wells, but
    you get the same signal to noise.

    So: a larger pixel size gives you better S/N at the same sensitivity, or
    better sensitivity at the same S/N. It all depends on your

    Dave Martindale, May 27, 2004
  6. Everything except a cropping 16MP Foveon sensor, which would then be
    much higher resolution and shoot center glass to boot.
    You don't understand these issues.
    Georgette Preddy, May 28, 2004
  7. Obviously since the current 13.72MP Foveon has already obsoleted
    It would be best to up the MPs but not the physical size as long as it
    doesn't introduce any noise (the current 13.72MP Foveon is noise free
    at ISO 100).
    Georgette Preddy, May 31, 2004
  8. I've wondered this myself... if its possible to produce a physically
    small sensor that can handle 8-megs, when why cant they use the same
    size "pixel sites" in a larger format to create a huge meg sensor in
    full format.

    I'm guessing that one reason as others have stated is that total fail
    rates of a sensor might be higher due to defects in the substrate...
    but then I think... so why doesnt that also equate to processors.

    So one question is, just what are the physical dimentions of say a P4
    "chip" never mind the packaging and the pins... but the actuall chip
    cut from the substrate that is stuck in the middle of all that plastic
    and pins that the outside world sees.

    asuming that its dimentions are not much different than a 35mm frame
    (hard to imagine!) then why are they relitavly cheap compared with
    camera sensors....

    As far as I can see, with a computer its posible to replace the
    processor (except when intel decide to change the physical design to
    kill the replaceables, requireing a mother board upgrade) so generally
    for one purchase of a "computer" there are a couple of potential
    processor upgrades that people may go through, so while they may only
    buy one "computer" every 2-10 years, they may go through 1-4
    processors, mass sales and production always reduces cost, sadly with
    digitals (excluding digital backs for MF/LF's) you cant pop the chip
    and plug a new one in.

    Actually that would be a cool idea!

    I'm not sure the actual +/- position that a film can have as its
    dragged through the camera infront of the lense, but if it there is a
    lee-way then surely it must be possible to design the socket for the
    digital sensor so that it could be poped out and replaced, again with
    MF/FF backs there must be some tolerance for slight miss-alignments in
    the same way that film is not totaly flat and might be under different
    tensions within the camera.

    We have some amazing inovations when you get away from the standard
    format of the SLR, such as phone cams, reasonable quality credit card
    shaped and sized digitals... basically the moden day "spy cameras"
    where what is known is dumped and people try to come up with something
    totaly radical such as micro film cameras.

    Having now used a DSLR I dont think you can beat the basic design and
    methods... but why not go back to the film idea of poping the back
    open where you would stick a film in it, but instead of a film you
    push a chip into the back as each new sensor comes out.

    I know with film there is the lead in material so that you dont have
    to worry about dust etc... so why not have a sensor encased in a
    moveable "front" so that you could put your thumb right on top, push
    it home, and when you close the back that slides out of the way... pop
    the back open and its slides to cover the sensor... you could also
    make it so that if the camera senses you're changing the lense it
    slides in to protect it from dust (would need a manual override to
    clean it if dust gets between the sensor and cover)

    This would also open up the second hand market, so that you could sell
    on your "used" and out of date sensor to those just starting out, in
    the same way that people sell older cpu's, the body of a camera is
    like a computer, depending on how badly its been treated it can be as
    good as new, its just a shame with a camera you have to dump the whole
    thing just to upgrade the sensor and buffer. True there does come a
    point where you do need to upgrade the whole thing to take into
    account of the surounding "bits" but 1 or 2 sensor/buffer upgrades
    should be handlable by most cameras.

    Oh well, just a thought... and i'm sure has already been paitented
    like my last musing... (sigmas prisum 3 sensor camera, lol)
    Jonathan Wilson, May 31, 2004
  9. Bay Area Dave

    Tom Scales Guest

    Unfortunately, you make an assumption that is wrong. According to the Intel
    docs I just read, the chip itself (p4) is roughly 16mm x 16mm.

    The is 256 sq mm.

    A full frame sensor is 24x36 or 864 sq mm, or more than three times the

    Much, much harder to manufacture.

    Personally, I don't care that much. Why are we stuck on that dimension? My
    D100 works just fine.

    Tom Scales, May 31, 2004
  10. Bay Area Dave

    Skip M Guest

    One reason some of us are stuck on that dimension is that we have wide angle
    lenses that at 1.5 or 1.6x become something just short of normal. In fact,
    speaking for myself, that is the only reason.

    Skip Middleton
    Skip M, May 31, 2004
  11. Bay Area Dave

    JPS Guest

    In message <>,
    There is only 3.43MP worth of useful data in a 13.7M bitmap from an X3F

    A 6.29MP 10D image has 6.29MP of useful luminance data. The following
    image is a duplicate of a 1:1 (100%), pixel-for-pixel crop from a 10D
    image. The top is the original crop, from a 6.29MP bitmap. The bottom
    is it duplicated, and then reduced to 50% to drop the entire image to
    1.5MP, as you claim it only has, and then brought back up to size again
    (200%). Both are sharpened aggressively. Only a lunatic or idiot would
    say that there is no more detail in the original (6.29MP crop) than the
    bottom (1.5MP crop). The top clearly has detail that justifies the
    image having a 6.29MP luminance resolution:

    The Sigma cameras seem to have a median filter that is applied on a
    curve; the less contrast there is between an individual pixel and its
    neighbors, the more the contrast is suppressed. Fairly high contrast
    between pixels is boosted a little further yet. That's how the images
    look to me. You don't see a lot of pixel-sized noise at ISO 100, but
    you do see cartoonish textures, and with the SD9, you can clearly see
    blotches of wrong levels in the green and blue channels in smooth
    gradients or areas of solid color.

    And the data is no good for anything more than 3.43MP. 13.7 is fine if
    you're going to rotate the image or perspective-crop it. I would
    convert a 10D image to 25MP if I wanted to rotate it or perspective-crop
    it as well.
    JPS, May 31, 2004
  12. Bay Area Dave

    Bill Funk Guest

    There is no way you can justify that sentence.
    Not even Foveon or Sigma makes that claim.
    Bill Funk, May 31, 2004
  13. It does. That is one (of several) reasons why the chip makers are
    constantly trying to reduce chip sizes; given a fixed number of surface
    defects, the reject percentage drops dramatically as the chip size
    Much larger production runs?
    David Littlewood, Jun 1, 2004
  14. I agree this is probably the most obvious reason. I also suspect that
    (other things being equal) larger sensor pixel size means lower noise
    and better tonal quality.
    David Littlewood, Jun 1, 2004
  15. This is one of the biggest reasons. The other is less obvious. The s/n ratio
    is lower with the larger size pixels, which require larger size sensors
    fabricated with larger geometry. On processors, manufacturers live with the
    leakage (noise) because even though the percentage of leakage is higher, the
    total power is lower. On sensors, the total power of the sensor is less of a
    concern since the sensor is on for such a short time, but leakage power,
    which translates to noise, is a big concern.

    The small, high-megapixel sensors are noisy, i.e. what's on the Sony F828
    and the Canon Pro 1. Even some small low-megapixel sensors are noisy, such
    as the Foveon X3.

    On the other side of the table, the full-frame sensors are difficult to
    manufacture and yield is low.
    Steven M. Scharf, Jun 1, 2004
  16. It keeps changing. www.tomshardware.com often lists die sizes for new
    CPU implementations as they appear. In general, even the most
    sophisticated CPUs are *much* smaller than even the "APS" sized DSLR

    Or just find an older Athlon package. That blue-black blob visible on
    the top is the actual silicon die. I just measured one I happen to have
    here, and the die is less than 12 mm square.

    Dave Martindale, Jun 1, 2004
  17. Bay Area Dave

    Bert Hyman Guest

    In Bay Area Dave
    There's a review in the July "Sky & Telescope" of the SBIG STL-11000M/CM
    which has a 36 x 24.7 mm CCD sensor at 11M pixels (4008 x 2745).

    It uses the Kodak KAI-11000M/CM chip and only costs about $8K.

    Bert Hyman, Jun 1, 2004
  18. Hell, I've got that in the piggy bank!
    Brian C. Baird, Jun 1, 2004
  19. Bay Area Dave

    Skip M Guest

    True, but, again speaking for myself, all things, i.e. noise, being equal, I
    still would hesitate to buy a less than 36x24 sensor because of what it does
    to my WA lenses, particularly my fisheye. Distortion, what distortion?
    Skip M, Jun 2, 2004
  20. Bay Area Dave

    Ray Fischer Guest

    In theory you could make a camera with a 6cm by 6cm sensor.
    After all, semiconductor wafers are 300mm in diameter.

    How much are YOU willing to pay for a camera?

    Ray Fischer, Jun 2, 2004
    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.