when will "true" medium/large format digital be affordable?

Discussion in 'Digital SLR' started by Scott Speck, Oct 8, 2006.

  1. Scott Speck

    Scott Speck Guest

    The following message is highly speculative, and is based on my limited
    understanding of digital sensors as well as my online reading in various
    photography forums and websites. That said, here goes...

    In playing around recently with medium format film, I was astonished at the
    improvement of enlarged image quality over 35mm. I also own and enjoy using
    both a Nikon D50 dslr and a point/shoot Panasonic FZ30, and I shoot 95% of
    my photos in digital form, at a minimum. However, seeing the huge
    improvement with the larger film size over 35mm, it makes me wonder what a
    truly medium-format-sized digital sensor (with the same pixel density as
    that in a "dx" or "35mm sized" digital slr), or even a 4x5" sensor, could
    do, assuming you had enough data storage to handle those huge images. I
    realize that "scanning backs" exist out there that can cover a 4x5 frame,
    but I'm sure that birds or waves or blowing leaves would appear to move
    across the frame during such a scan.

    I also recently read about a 4x4 INCH digital sensor (150 mpixel) that costs
    $200,000, and I realized that this is far beyond my financial reach.

    So here's my question -- is there anyone out there "in the know" on digital
    detector technologies who can say when a 4x4 CCD will cost $1000 instead of
    $200,000? I realize that might be impossible to answer, but I'm just
    tossing it out there, wondering.

    On the other hand, as digital technology evolves, is a bigger sensor always
    going to be better? For example, let's say that super-quiet ccd's could be
    created such that point/shoot-sized 50 mpixel CCD's could look great, even
    at high ISO. Would this obviate the need for physically larger sensors? I
    realize that, at some point, as pixels are made smaller (even super-quiet
    ones), one will eventually reach the diffraction limit of the optics in
    FRONT of the detector, and I'm wondering how close we already are to those
    limits with the digital sensors in the best cameras reachable by a
    photographic hobbyist.

    Also, I realize that if you make a mosaic of smaller ccd's to have more
    physical size, you'll still have the "blank regions" between the separate
    ccd's. Can sensors now be made that could have such narrow gaps (say 1 or 2
    PIXELS) between their edges that this gap would be inconsquential, so that
    you could build a 2x2 or 4x4 array of cheaper full-frame digital sensors and
    get pretty much the same result as a single huge sensor?

    Scott
     
    Scott Speck, Oct 8, 2006
    #1
    1. Advertisements

  2. Scott Speck

    tomm42 Guest

    The first 6mp cameras came out in 1995 by Kodak for $28,000. That
    camera line ended in 1992 and then list was $7600, but it was and is
    the only camera using a Nikon F5 as a body. Used is still around $1000
    for a DCS760 in good condition, they were very good cameras.
    Hassleblad is now 39mp for $40000 or so, the problem now is a matter of
    scale can they sell enough of these to inspire a less expensive back.
    Unluckily Hasslebald has released their own back and no longer will
    accept backs from other manufacturers. But Mamyia and Pentax have
    medium format cameras coming out, the Mamiya at least is 22mp for $12K,
    but not available in the US. It may take 10 years to bring prices out
    of the stratosphere. But 35mm based cameras are already as good as
    6x4.5 cameras and most 6x6 cameras. You need a film based 'blad with
    good lenses or 6x7 and up to beat a Canon 1Ds MkII, ad most APS cameras
    are just behind that. For 4x5 you need a scanning back, they work great
    in the studio BTW. Some folks have been sucessful using them outside,
    but you still need a static image, don't think you can generate any
    volumes with your 4x4 inch sensor if they start at $200K.
    The 39mp Hassleblad comes close to 4x5, but still you can develop 1000
    sheet for the price of the 'blad, and it doesn't have swings and tilts.
    Only time will tell.

    Tom
     
    tomm42, Oct 8, 2006
    #2
    1. Advertisements

  3. Scott Speck

    tomm42 Guest

    Sorry
    Meant 2002 for the last year of the DCS760.

    Tom
     
    tomm42, Oct 8, 2006
    #3
  4. Scott Speck

    Steve Wolfe Guest

    Never. Until we have Star Trek-style transporters, the cost of making 16
    square inches of semiconductor is never going to be cheap. Investigate how
    those things are manufactured, and you'll see. You could tile multiple
    small sensors, but still - you're talking about 16 square inches, and that
    is COSTLY.

    steve
     
    Steve Wolfe, Oct 8, 2006
    #4
  5. Never say never....
    A high school buddy of mine made millions in developing storage, (MTI it
    is listed in NASDAQ) has since retired. He said his company was selling
    20 gigs of storage (about 15-20 years ago) for half a million! Can buy
    a 200 gig hard drive now for $100. Was going to say "what happened to
    Cray computers, but they are going to deliver the first petaflop 'puter,
    I guess they are still in business...
    http://investors.cray.com/phoenix.zhtml?c=98390&p=irol-newsArticle&ID=873357&highlight=
     
    Wolfgang Schmittenhammer, Oct 9, 2006
    #5
  6. tomm42 wrote:

    36x49mm sensors are available for about 32,000 dollars.
    they will work on Blads and 4x5 cameras.

    PRICE: $29,999.99
    Includes: Hasselblad H2D SLR camera with 80mm lens,
    viewfinder and 39MP single-shot digital back.


    Phase One
    P45 Digital Back for Hasselblad H Series With Three Year Warranty
    PRICE: $32,999.99

    $17,000 Phase One
    Specifications:

    Imager:

    * Color filter: Red, green, blue
    * Pixels (active): 6496 x 4872
    * CCD Size (active): 44.2 x 33.1mm
    * Pixel Size: 6.8 x 6.8 micron
    * Image ratio: 4:3
    * Image output: 48 bits (16 bits per color)
    * Antiblooming: 8 f-stops

    Digital Image

    * Color depth: 16 bit per color
    * Dynamic range: 12 f-stop

    Sensitivity

    * ISO: 50, 100, 200, 400, 800

    Camera system:

    * Capture rate: 45 frames per minute
    * Writing speed to CF: Up to 20MB/sec., depending on CF card
    * Battery type: Lithium ion
    * Battery lifetime: 250 captures/4 hours
    * Exposure time: Several minutes to less than 1/10000
    * IR filter: Mounted on CCD
    * Orientation: Automatic image rotation (Patented)
    * Power: 8-33 V DC

    If you chose the right back,
    it will take either single shot exposures for moving subjects,
    or 4 exposures for still lifes,
    so each pixel location gets all four exposures.

    Ixpress 528c

    4*Res upgrade option: a new level of quality for studio photography
    Get ultra-high-end image quality for still-life subject photography,
    with image sizes from 96 to 528 MB.

    This unique upgrade provides moiré-free operation in both 4-shot and
    16-shot modes.
     
    bob crownfield, Oct 9, 2006
    #6
  7. First hard drive, for an intel development system,
    was 20mb, for $15,000.

    Was going to say "what happened to
     
    bob crownfield, Oct 9, 2006
    #7
  8. Scott Speck

    Steve Wolfe Guest

    Never say never....
    Yeah, storage is cheap. But look into the costs associated with
    semiconductor manufacturing. Over the last 20 years, they have dropped, but
    not by a terrible amount. Seriously, do your own reading on what goes into
    those things, and you'll see for yourself.

    steve
     
    Steve Wolfe, Oct 9, 2006
    #8

  9. Maybe not, and maybe not necessary. How close is the pixel size in the
    current generation of sensors to the theoretical minimum? If it is
    possible to make the pixel size 1/2 or 1/4 of the present size, this
    would result in an image which has four or sixteen times the pixel
    density....

    ns
     
    Noel Stoutenburg, Oct 9, 2006
    #9
  10. http://www.clarkvision.com/imagedetail/does.pixel.size.matter

    Regarding of manufacture of large sensors, the cost for a wafer
    run is pretty high. The last number I heard for an 8-inch silicon
    wafer run was about $10,000. Electronics have become cheaper
    by miniaturization. There is some fraction of defects per
    wafer, so the larger the chip, the higher percentage that
    fail, driving costs up further.

    With digital camera sensors, you can't shrink below
    a couple of photons in size, and even then you
    won't collect many photons (see above web site).
    Photons from the sun are finite, and DSLRs are collecting
    only a few tens of thousands of photons in a typical
    picture, and small P&S cameras much less
    (e.g. less than 10,000).

    Roger
     
    Roger N. Clark (change username to rnclark), Oct 9, 2006
    #10
  11. Scott Speck

    frederick Guest

    I assume that you mean "per pixel" in above.
     
    frederick, Oct 9, 2006
    #11
  12. [Not Roger] - yes, that's a per pixel value. Roger obviously meant "in a
    typical pixel".

    David
     
    David J Taylor, Oct 9, 2006
    #12
  13. Are you sure? You can get a LOT more transistors for $1000 now than
    you could get 26 years ago. Current micprocessors have hundreds of
    millions of transistors in them, compared to the tens of thousands of
    1980.
     
    Toni Nikkanen, Oct 9, 2006
    #13
  14. Yes, I meant per pixel. Sorry for the confusion.

    Roger
     
    Roger N. Clark (change username to rnclark), Oct 9, 2006
    #14
  15. Scott Speck

    J. Clarke Guest

    With regard to the cost of storage, yes, it has come down, but many years
    ago the cost of 3.5" drives hit a bottom--not sure where it was, under $100
    though. The capacity keeps getting higher but the price of an entry-level
    drive is not decreasing. The reason is that there is a certain irreducible
    minimum level of materials and labor required to make the device.

    The same is the issue with large sensors. It doesn't matter how many
    devices you put on it 16 square inches of silicon is still 16 square inches
    of silicon.
     
    J. Clarke, Oct 9, 2006
    #15
  16. Scott Speck

    frederick Guest


    But you could compare a sensor of a particular size with a transistor or
    a certain size over time. A power transistor for say audio power
    amplifier applications would be a good example. A little less expensive
    over 30 years, but not by a factor attributable to Moores Law etc. as
    that relies on increasing miniaturisation.
    So for sensors, they are already at the point where increasing
    miniaturisation is not desirable, so big gains in economy aren't likely.
     
    frederick, Oct 9, 2006
    #16
  17. But, if you can put a gazillion pixels on a 23.6 x 15.8 mm sensor,
    wouldn't that be desirable? And then would a large format even be
    necessary? Just wondering.

    Tony V
     
    Tony Verhulst, Oct 9, 2006
    #17
  18. Scott Speck

    frederick Guest

    No - it wouldn't be desirable.
    Diffraction is explained here:
    http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm#

    Aside from diffraction which would render huge pixel counts of no
    benefit, noise is an issue, and also if you have dealt with large MP
    images (I do with stitched >30mp images) then dealing with them with a
    current home PC is a pain in the neck. (Sure - for the latter you can
    expect that PC processing power / storage increases to make up for it)
    If you look at sample images from the latest small sensor 10mp P&S
    cameras, the results aren't very good - even at low ISO aggressive NR
    robs detail, and at normal working apertures (even typically widest
    aperture at long zoom f/l) they are diffraction limited.
    For an APS-c sized sensor, then over about 15mp isn't going to be much
    practical use, and even if they get to that they have to deal with
    noise. The latest Olympus 4/3 10mp camera has about the same pixel
    density as 15mp or so on APS-c. When sample images are seen, that's the
    sort of performance you'll get from 15mp APS-c. I think it won't be
    very good.
     
    frederick, Oct 9, 2006
    #18
  19. Scott Speck

    Paul Furman Guest

    Smaller pixels have less dynamic range, highlights blow & shadow detail
    is lost. Calculating the quantum physics of how many photons will fit in
    a tiny pixel, we are fairly close to the practical limit. When small
    pixels fill up, the charge bleeds over to adjacent pixels causing purple
    fringing, there is less range for adjusting ISO sensitivity, more noise,
    etc. You never know what other trick they might come up with like
    microlenses or fuji's second set of mini pixels between the main set to
    get another level of the dynamic range, but there seem to be some major
    barriers.
     
    Paul Furman, Oct 9, 2006
    #19
  20. Scott Speck

    Prometheus Guest

    No, the noise would be worse.
    Yes, you could have a larger number of large pixels.
     
    Prometheus, Oct 9, 2006
    #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.