What is holding back full size sensors? Just marketing?

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

  1. But adding to it, is that 36mm is bigger than the max mask size most
    wafer steppers can do, so you need CCD specific fabbing gear.

    --
    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, Jun 2, 2004
    #61
    1. Advertisements

  2. And single sensors much larger than this do exist, or have been
    manufactured. I think I saw one a while back that was close to 5 X 7
    inches and they used an array of these things. It was in one of the
    very large telescopes, but I've forgotten which one. I don't remember
    if this one was monochrome or color. I've heard some very large color
    ones have been made, but I can't even begin to put an accurage WAG as
    to how much one would cost. It'd be a heap and a bunch though.

    Roger Halstead (K8RI & ARRL life member)
    (N833R, S# CD-2 Worlds oldest Debonair)
    www.rogerhalstead.com
     
    Roger Halstead, Jun 3, 2004
    #62
    1. Advertisements

  3. Bay Area Dave

    Dave Haynie Guest

    Yup. But due to the serial nature of CCDs, it's far easier for a
    defect to kill the whole chip. If it's just the imaging photodiode
    that's killed, they certain can address that single pixel loss in
    software. But if the "charge bucket" is dead, the chip is dead.

    CMOS can be far more forgiving, owing to its parallel nature: one
    totally dead sensor site doesn't affect the others. Not surprisingly,
    many of the larger CCDs out there are CMOS.
    Actually, no.. given the same defect density, doubling the size of the
    chip usually results in less than half the yield. Let's imagine a chip
    with 40 small dice or 20 large dice. The defect density is such that 5
    out of 40 small dice fail. That will likely still result in 5 out 20
    failures for the larger dice. So that's an 87.5% yield in the first
    case, a 75% yield in the second case.
    And actually, also "no". Wafers are circular, and so you try to stuff
    as many rectangles into that circle as possible. In my pretend
    example, if 40 small sensor fit, you probably only get ~16 of the
    larger in the same wafer. So now your yield is down to 68.75%.
    Flipping it around, you get 11 good dice from a large chip wafer, 35
    good dice from a small chip wafer. So it's more like 3x the expense to
    double the size. That really does tend to be how chip economics work.
    Not sure what you're going for there, bosco. Larger pixels make
    absolutely perfect sense. If you double the pixel size, you double the
    light-gathering capability. Thus, you get to run with less gain on
    your ADCs, and so you can eliminate random noise, at least for a wider
    range of effective ISO. And larger pixels cause your CPU not a cycle's
    worth of extra processing.

    More pixels would, of course, but that's very desirable anyway. CPU
    power is cheap, sensor pixels are expensive.

    As well, after a certainly number of pixels on a sensor, you'll have
    to get larger sensors simply to avoid diffraction -- a pixel element
    smaller than the wavelength of the light you're trying to capture is a
    problem.
    The various full-frame 35mm DSLRs on the market, not to mention the
    various >35mm sized sensor backs for MF cameras, would tend to prove
    that it's an acceptable thing, at the right level in the photo market.
    You probably won't do well selling a large, expensive CCD in a D70 or
    Digital Rebel body, but they'll sell just fine in a pro body.
    CPU's not the issue. For this class of processing, you're talking
    about a $15-$25 CPU, depending on just how much data crunching is
    needed. You'll probably spend half again for the RAM buffer on a
    high-end machine. It's the hundreds you spend for the sensor, for the
    pro-class mechanicals, the $50+ for the color LCD, etc. that's
    building up the price of the full-frame cameras. Not the CPU.


    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, Jun 3, 2004
    #63
  4. Bay Area Dave

    Dave Haynie Guest

    Well, not knowing anything about chip design is part of the problem.
    ALL major CPU designs, for example, assume zero defects (well, many
    memory structures allow for redundancy; definitely in your typical
    DRAM chip, not sure about caches in typical CPUs, that would make
    sense). The current Pentium 4 "Prescott", for example, isn't "free"
    and it's about a 112mm^2 die with 125 million transistors. It may even
    have similar yield issues in theory, but when you're making CPU, it's
    on a super-cheap commodity CMOS line, on 12" wafers, etc. And you can
    fit about eight of these in the space of one full-frame CCD. And the
    CCD, of course, has to live in a far more expensive analog/hybrid chip
    process, not good old cheap CMOS.
    Well, actaully, they are in a CCD. A charge-coupled device is like an
    analog electronic bucket brigade: each charge storage unit passes its
    charge on to the next. No problem if you lose a photodiode in the
    array (just pass on a black charge), but if the charge storage unit
    itself is dead, your bucket brigade is dead.

    In CMOS, it's certainly better: each cell can be independently
    addressable. This is one reason CMOS is being pursued by, well, pretty
    much everyone (Canon's been using it, Nikon's now using it, with
    in-house developed processes). The other advantages for CMOS sensors
    are the ability to integrate on-sensor processing (build the CPU right
    in there with the imager), the fact you're on a low-cost, commodity
    chip process, etc.

    You can't solve it all with redundancy, though. After all, it's a
    visual thing. If a row in a DRAM fails, maybe a substitute at the edge
    of the memory array can be fuzed in place of it. You can't exactly
    steal pixels from the edge of an imaging chip to fill in for a dead
    one in the middle.


    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, Jun 3, 2004
    #64
  5. Bay Area Dave

    Dave Haynie Guest

    Actually, I've mentioned this. It's a pretty standard issue in digital
    video, where the size of the image really doesn't change much (well,
    1/3Mpixel for standard DV, 1Mpixel for HD-720p, 2Mpixel for HD-1080i).
    Of course, light gathering is also an issue when it's motion; not just
    a noise issue, but the fact you can't get away with a 1/10sec shot.
    That's actually why I think they Olympus/Kodak/Whomever "4/3" system
    is a bad idea - they're locked into a 1/2 sized sensor, forever. Even
    the cheaper consumer DSLR sensors are a bit larger (they seem to have
    settled around "APS-sized", which is a bit over 50% of the imager
    size. The current tech is a compromise in pixels, light, or money. It
    isn't all three -- if you have the money today, you can get more
    sensitivity and more resolution. That's only going to get better.


    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, Jun 3, 2004
    #65
  6. That is, in fact, precisely half the defect rate (12.5% versus 25%), as
    the previous poster said.
     
    David Littlewood, Jun 3, 2004
    #66
  7. Bay Area Dave

    bagal Guest

    <everything snipped>

    It may be elsewhere in this newsgroup - something about a 500 Mp imaging
    device

    Now then, to go back ontrack.

    On of the constraints of slapping sensors on the back of a CPU may be that
    the extra enegry generated and dissipated in the form of heat may (will?
    does?) introduce noise artifacts and image blobs

    You never know, it may be a plain old cooling system that is required - a
    simple low tech solution to allow hi tech solutions to aim higher.

    So, any reviewers out there wishing to influence a review may find it
    helpful to stick one camera in the freezer and the other in a warm spot,
    then leave them for a while before capturing images (to observe temperature
    induced image corruption.)

    (influence is a bad word in the above - perhaps the meaning is clear if the
    words are ill arranged?)

    But if temerature is an issue in still digital images is the effect
    replicated in digital movie machines?

    If not, why not? If yes, discuss ...

    das Bagal
     
    bagal, Jun 4, 2004
    #67
  8. I believe the medium format backs like the leaf do have built in
    cooling, not sure the specifics, but at the least a fan, lol

    With a "bulky," camera adding a bit more bulk is a no-brainer... you
    just do it... while not strictly true a MF camera is usually on a
    tripod, there are powerfull studio strobes/flashes and lights so heat
    is a problem, and as the camera doesnt have to be moved all over the
    shop and is heavy anyways, adding a few more ounces is negligable.

    Also the back and the storage device seem to be seperate as well
    (large storage HD device) which is connected to the back via a cable,
    so even tho the total weight is quite high, it is spread across the
    camera and say the users belt clip.

    Power is also a big issue, studio set ups are fine... there is a plug
    socket somewhere and potentially you can run the AC/DC converter
    directly into the camera while working... out in the field if there
    was a fan and the battery went from 2-3 hours down to 15mins-1 hour..
    its a huge difference, or an even bigger weight due to a large (like a
    laptop sized) battery.

    What might be interesting in the future is how the manufacturers get
    away from the idea of what a camera should look like (already done in
    the non-slr's) and how that could be linked to storage systems...
    perhaps a lightweight slr style body and lens mount with a sensor and
    fan and nothing else... bluetooth to a seperate storage device that
    could be cliped to the belt, or perhaps held by someone else for when
    you take pictures on a rollercoaster, as the camera would only have
    body, lens, mirror, sensor and comunications chip it would be far more
    resistant to shakes and knocks, and would require very little power...
    and with in theory a back up medium as large as the latest drives
    (they wouldnt even need to be Micro drives, or laptop sized) then you
    could have a system capable of storing huge amounts of shots in the
    field... with a clever system capable of winding down the drive when
    not in use, buffering enough data while the drive powered up to start
    saving the data, the life span of the batterys would be fairly good...
    imgaine doing 20-40-120gigs shooting in one day with two batterys (one
    for the camera one for the drive/storage setup)

    As always... if someone has done this, just goes to show what museing
    can do, if not.. they cant paitent it cos I just put it in the public
    domain, lol
    jon
     
    Jonathan Wilson, Jun 5, 2004
    #68
  9. Bay Area Dave

    bagal Guest

    groovio Jon BT = less cables = good very good :)

    OTOH - have a look at my postings with links to JPEGs (I appreciate the
    knowledge and wisdom of other posters including the original links in their
    post)

    JPEGs?
    started in the early 90's ... it is time for a revision but the whole
    digital consumer industry seems to be snagged into them. It would be grand
    to see an industry wide standard that addresses image compression (12 years
    is a very long time in digital world)

    dB
     
    bagal, Jun 6, 2004
    #69
    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.