2 questions regarding Bayer sensors

Discussion in 'Digital Photography' started by Georgette Preddy, Jun 2, 2004.

  1. The concept was good, but the implementation was lacking. The original
    Foveon studio camera was great for what it was, a three sensor, low ISO,
    camera that was tethered to a workstation. But it was expensive to
    manufacture, couldn't be used in the field, and didn't keep up with the
    Bayer technology which was advancing rapidly, and soon passed it in
    resolution.

    Hasselblad tried, and gave up, on getting a Foveon prism based system to
    work with their bodies, it had too many issues to have a chance at
    commercial success.

    No one else adopted the X3 sensor because the digital SLR makers wanted a
    higher-resolution solution. It'd be difficult to sell an SLR with a 3.4
    Mpixel sensor when you're selling 4, 5, 6, & 8 megapixel resolution point
    and shoots! Foveon tried to get into the top tier and second tier of SLR
    makers, but no one bought the whole story about how a 3.43 Mpixel sensor was
    really 10.2 megapixels; nice try, but Canon and Nikon engineers aren't
    stupid. Foveon had to find a company that already had an SLR film body that
    they could cheaply retrofit with the X3 sensor, and that had no reputation
    in SLRs to worry about. Sigma was the only company with nothing to lose. The
    first model, the SD9 was bizarre. It was rushed out, and really showed its
    film roots with a bizarre two battery power system. I don't know for sure,
    but I'd bet that Sigma was paid by Foveon to develop the SD9 and SD10. This
    type of arrangement is not uncommon. Sigma hoped to drive lens sales, Foveon
    hoped to drive senor sales.

    Foveon has given up on the professional market; they simply have no
    high-resolution sensor with which to compete. Canon now manufactures the
    world-class EOS-1Ds with a full frame, 11 Megapixel sensor. It's more than 3
    times the resolution of the Foveon sensor in the Sigma SD10, and it doesn't
    suffer from the color, noise, and dynamic range issues. Similarly, Canon and
    Nikon both manufacture "Prosumer" 6 megapixel digital SLR cameras. Foveon
    has given up on the Prosumer market; they have no sensor for Prosumer
    cameras. Even in the consumer market, Canon dominates with the Digital
    Rebel, which while not as full features as the Prosumer models is still much
    higher resolution than the Sigma SD9/SD10, and it doesn't suffer from all
    the issues of the SD9/SD10.

    Foveon has almost an insurmountable task. They can no longer even come out
    with a 6 megapixel sensor and have any chance of success. Their only hope is
    to leapfrog from 3.43 megapixels up to something like 12-14 megapixels.
    Kodak is already at 14 megapixels, but with issues.

    Sigma is predicted to exit the camera business, both film and digital, this
    year (at least under their own name with their own lens mount), and
    concentrate on lenses for other makes of SLRs. However they are
    manufacturing a 14 Mpixel SLR for Kodak, that takes Canon lenses.

    A licensing company that bought the rights to the Polaroid name has
    announced a 1.5 Megapixel Foveon based camera, but it is highly doubtful if
    it will make it to the market. It was just delayed from June until
    September. Three more months for prices to fall, and resolution to increase.
    Even a $100 Canon A60 is higher resolution than the Polaroid! The Polaroid
    brand for digital is essentially toy cameras, so maybe at $75-100 they can
    sell some of these, but there won't be any big volume.

    I wouldn't discount the possibility that some other company tries to make a
    multi-sensor/pixel sensor-chip in the future, but they will have to solve
    the issues that vexed Foveon.

    See http://sigmasd10.com for more information.
     
    Steven M. Scharf, Jun 5, 2004
    #81
    1. Advertisements

  2. OK - some time ago my kids and their cousins went
    to a photographer to make a group photo. The photgrapher
    took maybe 50 photos, half of them from a very weird angle.
    So - there was 25 reasonable ones to choose from. Among those
    25 photos there was only 2 of them where all 12 people
    looked good. Those two were not the two best technically.
    But, choosing a technically better photo where not all looked
    good would have been the wrong choice IMHO.


    /Roland
     
    Roland Karlsson, Jun 5, 2004
    #82
    1. Advertisements

  3. Yepp. To compete you must be able to make new sensors all the time;
    bigger, better, whatever. You must be able to keep up the pace.
    You cannot just make one sensor and then try to survive on that one.
    A small company like Foveon is therefore doomed to fail. The big
    ones have much more money and momentum. National Semiconductor
    is one of the founding companies, but (as far as I know) has shown
    no interest in buying it all and start in big scale.


    /Roland
     
    Roland Karlsson, Jun 5, 2004
    #83
  4. Groups are always a compromise. I never did more than 3 people in the
    studio just because it was so small. Groups were done either in the
    home or in a park. Always tried to do groups full length - no point to
    chopping off people's bodies. Tried to sell large prints and did
    fairly well at it.
     
    Randall Ainsworth, Jun 5, 2004
    #84
  5. True, I didnt mean to sound as tho I was critisising. I guess its the
    nature of the type of work, in a studio setup the only thing you have
    to worry about (bar system fails) is the subject(s) looking in the
    right directions, not blinking or pulling funny faces :)

    There is a studio/wedding photographer guy locally who's work I very
    much admire, however there is very little difference between one
    sitting and another... very much mum on a chair, dad stood behind,
    baby on mums lap kind of thing. I guess the bigest change between one
    sitting and the next is the backdrop being changed to compliment the
    clothes the people are wearing. In his wedding work however when not
    the "standard" shots, he does get a higher fail rate... he uses MF for
    all the official standard shots/poses and baring a cock up gets each
    shot in the can with no fails. He then uses digital for some of the
    more individual "candid" looking, which is where his loss rate goes up
    because its far less controled, but can produce some of the most
    natural looking shots even tho there may be some technical errors. He
    does have a very good eye for seeing obvious artifacts in the
    background, a skill that I feel I am improving upon in leaps and
    bounds... some of my non-studio work in uncontroled environments in
    the early days had all the usual mistakes like trees sprouting from
    peoples heads, forgetting that a mirror reflects the background as
    well as the subject... tipical snap shot mistakes :)
     
    Jonathan Wilson, Jun 5, 2004
    #85
  6. True, I didnt mean to sound as tho I was critisising. I guess its the
    The studio is all about control.

    But more than that, you quickly find out that you can't sell the exotic
    or the bizarre on a daily basis. You'll be lucky to sell the creative
    on a regular basis. People want what other people have. If that's a
    head shot with an old Masters background and some decent lighting, then
    that's what it is.

    You see these nice creative wedding shots in the magazines but you'll
    play hell selling them. People want to see the faces...they want the
    traditional wedding party on the altar, cake shots, etc.
     
    Randall Ainsworth, Jun 5, 2004
    #86
  7. Georgette Preddy

    Ray Fischer Guest

    Ray Fischer, Jun 5, 2004
    #87
  8. Legally actionable.
    Agreed, and the tiny full color MP count of the Canon 1Ds should place
    it in the $150-$200 category (body only).
     
    Georgette Preddy, Jun 6, 2004
    #88
  9. Georgette Preddy

    JPS Guest

    In message <rmkwc.22205$>,
    The overall concept is not Foveon's at all. Everyone knows that
    full-RGB sensing at each pixel would give more accurate luminance and
    higher color resolution, and that's probably what most people would
    assume that digital cameras worked that way if not told about the
    technological problems or Bayer mosaicing. There have simply been
    barriers to implementing full-RGB, and Foveon claimed to have a
    solution, but not a high-quality one.
    --
     
    JPS, Jun 6, 2004
    #89
  10. Georgette Preddy

    Bill Funk Guest

    On what legal theory?
     
    Bill Funk, Jun 7, 2004
    #90
  11. This of course from George/Georgette Preddy who doesn't even know
    his/her/it's own gender. He/she/it has posted under so many aliases
    he/she/it doesn't know his/her/it's real name.
     
    Darrell Larose, Jun 7, 2004
    #91
  12. National is wisely concentrating on their analog business.
     
    Steven M. Scharf, Jun 7, 2004
    #92
  13. Hopefully someone like Canon can come up with a workable solution to this.
    I've heard that they are working on it.
     
    Steven M. Scharf, Jun 7, 2004
    #93
  14. To solve the biggest problems with the Foveon sensor, they need to increase
    the size of the photo-diodes. The noise problems of the Foveon sensor, and
    the Sigma cameras, are directly related to the photo-diode size. The color
    accuracy problem is going to be very hard to solve; you'll be unlikely to
    ever be able to completely overcome the color contamination issue, so the
    Bayer sensor will continue to produce better color. Increasing the size of
    the photo diodes will mean going to a semiconductor fabrication process that
    is optimized for camera sensors; the 0.18u National Semiconductor process is
    excellent for many devices, but not the optimal choice for camera sensors.

    Ironically, until 1993, National used to partially own a fabrication
    facility in Israel, Tower Semiconductor, that has since become a leading
    manufacturer of advanced CMOS sensors. For very low-noise sensors, you can't
    use a 0.18u process, at least not yet. See http://towersemi.com

    See: http://www.photozone.de/7Digital/digital_3.htm

    Very good explanation of sensors; Preddy should read this as a primer so he
    stops embarassing himself:

    Regarding the Foveon X3 sensor:

    "The effective output resolution is really just 3.4 mio. pixels (10.2 mio. /
    3 layers) and that's not too impressive from the perspective of today's
    Bayer sensors."

    And

    "As a nice side effect the photodiodes COULD also be much (3x) bigger thus
    resulting in less sensor noise - POTENTIALLY that is which is not the case
    as of now."

    Canon could easily solve both the spatial resolutio problems, and the noise
    problems, that are two reasons that the Foveon sensor hasn't been adopted by
    any first or second tier manufacturers. Solving the color accuracy issue is
    going to be very hard. It may be that they have to go to RAW only mode, and
    do tremendous amounts of post-processing, as Sigma does now, with not the
    greatest results.
     
    Steven M. Scharf, Jun 7, 2004
    #94
  15. Georgette Preddy

    bagal Guest

    <quite a lot of snipping happened pre-posting>

    These are great inks Steven - thankyou. I am sure that I am learning at
    quite a fast rate partly because of the information that really is out there
    and the links posters include within their message. Thanks to all.

    I have no hangups at all about pixels as picture elements (first introduced
    to them at Uni in a Digital Image course which I will say very little about
    :)

    One of the links posted somewhere leads if not directly to an image composed
    of coloured jellybeans. Yep - an edible picture.

    Now jellybeans have colour and shape and are usually consumed because of
    their lovely flavours and fruity color. In the picture composed of
    jellybeans the jelly bean is a pixel. In this case it is a pixel with
    chirality and what is worse, the jellybean can be rotated in the plain soe
    our pixel has a defined shap and handedness but in the picture it may be
    placed on one of 2 sides and rotated through any number of degrees subject,
    of course, to multiples of 2*pi

    I don't really know why I mention this other than to say pixels are very
    vague things in general and yet in specific cases may be very well defined
    indeed.

    I did read somewhere of a reference to sensels (sensor elements) this would
    seem to solve some of the terminology eg a 10Msensel sensor providing XYZ
    megapixel output

    Somehow I think the topic will rage on - my own particular hobby horse is
    digital iumage burnout which incidentally I consider to be far more
    important that this sensor has more totestosterone than the other.

    Yep, highglight burnout and its twin lolight blobs seem far vmore real world
    issues

    das B
     
    bagal, Jun 7, 2004
    #95
  16. Georgette Preddy

    JPS Guest

    In message <ZXWwc.21$>,
    Well, you have to just break away from the default metering of the
    camera. Discover how it works, and work around it. Use RAW mode, if
    available.

    I shoot RAW almost all the time with my 10D unless I'm shooting fast
    action in moderate-contrast environment, or I accidentally left my extra
    memory cards home. It is pretty rare that anything burns out in RAW
    mode. The red channel has 3.8 stops above medium grey, and the blue and
    green channels have 3.2 stops above medium grey, as opposed to 2.2 stops
    in all channels in a JPEG.

    It takes something like a Great Egret in the sun with most of the frame
    capturing shaded areas for me to get burnt-out highlights, and that
    would usually only be a fraction of the white Egret.
    --
     
    JPS, Jun 7, 2004
    #96
  17. Georgette Preddy

    bagal Guest

    lol - it doesn't seem to happen in my shots

    even in a white-out in the mountains

    focus & balance seem spot on

    but there again, i am using a P&S Fuji :)

    My original concerns were raised by viewing a link in which the photographer
    was very pleased with the still life. However, the still life contained
    quite a bit of highlight burnout

    seemed a shame that a high level (and high cost too!) DSLR produced so many
    burnouts

    In fact, similar effects can be seen on manufacturers websites :-( (sample
    images bit)

    I think there is a later thread about ID'ing sensors so consumers may know
    which ones to go for or which to avoid

    i hope this help clear things up a wee bit - it would be a shame if I gave
    you the wrong impression

    das B
     
    bagal, Jun 8, 2004
    #97
  18. Georgette Preddy

    JPS Guest

    In message <[email protected]>,
    I believe that the P&S cameras do a better JPEG. The 10D uses a gamma
    of 2.2 throughout the entire range for JPEGs when the contrast is set to
    0. Almost every P&S I've used seems to change the gamma at the extremes
    to fit more dynamic range into the image. I posted the exactly curve
    here for the Sony F707 a couple of years ago, after measuring it.
    Well, it was a very beautiful image, otherwise. The color of the wood
    was superb, there was no noise, and quite a bit of dynamic range. The
    burnt-out area was partial (some detail remained in it), and didn't
    affect all channels, and didn't ruin the image for me.
    Well, everyone has an opportunity to work around the arbitrariness of
    the metering system. Fact is, you can expose the sensor however you
    want, and determine how it is digitized with the ISO. Everything else
    is just a roundabout way of achieving that. The violin picture could
    have been shot differently. It could have:

    1) been RAW, if it wasn't.

    2) possibly been converted without burnout if shot in RAW
    (and still can be, since, fortunately, you can't save over a RAW
    file).

    3) Could have been exposed less
    Maybe other people judge images by their overall merit, rather than by
    whether or not there are burnt-out areas. Some people might actually
    like it, especially when it doesn't render large areas completely
    255,255,255 without any detail.
    There may be nothing wrong with the sensor, and everything wrong with
    the arbitrary exposure paradigms used, and gamma-adjustments.
    --
     
    JPS, Jun 8, 2004
    #98
  19. I agree with you, the F707 is sharper, higher resolution, and lower
    noise than the 10D for JPEG, and probably for most situations when RAW
    is used too. The 10D's overly aggressive NR cannot be turned
    off--this reduces real resolution dramatically in all situations while
    the F717, in some situtaions, does ok. It is a real shame Canon still
    doesn't provide a usuable RAW workflow.

    The current F717 is much less expensive than the 10D too, but
    obviously if you want pro image quality both the F717 and the 10D are
    way too sensor poor.

    Foveon dominates the entire digital field with its native 13.72MP
    output when interpolated to Bayer's 25% optical standard, see this
    comparison with the prosumer-Bayer-leading F717...

    http://www.pbase.com/imageprocessing/717_sd9_s602z
     
    Georgette Preddy, Jun 8, 2004
    #99
  20. The person posting under the name of George or Georgette Preddy (and
    other pseudonyms) has an ungrounded but zealous faith that current
    implementation of the Foveon chip is superior to all other chip
    technologies. He will cite portions of reviews to ostensibly support his
    claims, and will repeat, ad naseum, complete lines of out-of-context
    material, as well as manufacturing some pure bull manure.

    Moreover, "Mr." "Preddy" has claimed to be a photographer (pro!), but
    cannot bring himself to post a single picture with EXIF info that he
    shot himself, in spite of repeated requests and challenges to do so.

    Apparently he loathes anything related to Canon and loves everything
    about Sigma cameras and lenses. His "claims" may be ignored, and he is
    doing Sigma, and anyone related to the Foveon chip, no good at all by
    arousing ire, increasing the N/S ratio, and generally spamming this
    newsgroup.
     
    John P McWilliams, Jun 8, 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.