Foveon Wars

Discussion in 'Digital Photography' started by Ken F., Oct 31, 2003.

  1. Ken F.

    Ken F. Guest

    OK, even as a lurker I'm getting a little tired of the Foveon wars.
    I see little more than a bunch of zealots arguing along the line of
    "my camera's better than your camera."

    Hell, why don't you just start debating whether Budweiser is better
    than Coors or whether Ford is better than Chevy? Statements such as
    "Sigma sucks" reflect ignorance and poor upbringing. Now if you would
    like to say "Sigma has never produced a technically satisfactory
    product" and provide an explanation, that's fine, but to extrapolate
    to "Sigma couldn't produce a technically satisfactory product - even
    if they licensed the technology from someone else" is brand name
    bigotry at its best. (and it's what the big brand names want you to
    believe - so that there is no "need" for actual technical comparison)

    I don't have a dog in this fight, but I am interested in technical
    details or even the "perceived" technical details such as "how does
    the image look." When an image looks clearer to me I really don't
    care about someone's argument that it CAN'T look clearer. I don't
    care much for Mr. Preddy's zealousness (nor do I understand the reason
    for it), but after hearing much of what has been said (and comparing
    images) I would have to side with him.

    There are (full resolution) test images on an independent site that
    will help with this.
    First let's look at a Canon 10D image of 3072 x 2048 pixels.
    See if you can read the number which appears on the dumpster:
    http://www.steves-digicams.com/2003_reviews/10d/samples/IMG_0082.JPG

    Next let's try to do the same thing with a Sigma SD10 image of only
    2268x1512,
    (which, based upon it size, we might expect to have much lower
    resolution.)
    http://www.steves-digicams.com/2003_reviews/sigma_sd10/samples/IMG09640.jpg

    Now I didn't take these images and I don't own either a Sigma or Canon
    but find the the Sigma image to be far clearer. I can read the
    dumpster number.
    Are there other problems? Yup, dirty spots coming from somewhere.
    Is the color off? I don't know, I wasn't there, but it certainly
    doesn't make me want to scream "That's not accurate!"

    In any case, if necessary I could envision a quick Photoshop color
    adjustment, but I can't see gaining resolution through PS
    manipulation.

    My impression is that the Foveon sensor holds great technical promise
    for high resolution digital imaging. Whether it will fulfill that
    promise entirely may be dependent on whether it survives in the
    market. And of course that may also be dependent on the inertia in
    "established" brands and technologies, as well as the possibility of
    something even better coming along and leapfrogging Foveon entirely.
    It's likely that if Canon came up with a new and better sensor
    Canon's market position would bolster the product reception in the
    marketplace substantially.

    It is certainly possible that even objective technical superiority can
    be prevented from succeeding the marketplace by brand name zealots.
    That is, for better or worse, what marketing is all about.

    If your mantra is "I'd never own anything but Canon," well, you never
    will.
    But if you add the disclaimer "until something better comes along" AND
    keep your mind open to new products you may have the opportunity to
    experience something better. But, of course, this is all irrelevant
    if the next great innovation should come from Canon. Then you can just
    feel smug and say "I told you so."

    KF
     
    Ken F., Oct 31, 2003
    #1
    1. Advertisements

  2. And what the technologically literate have been saying is that there are
    good mathematical, engineering, and physiological reasons that it holds only
    marginal promise at best.

    (1) Digital sampling theory tells us we can only sample bandlimited signals.
    Sampling a signal that is not band limited results in aliasing artifacts. An
    antialiasing (AA) filter is required for reliable correct digital imaging.

    (2) Since AA filters are not perfect, an AA filter strong enough to elimnate
    artifacts will also reduce practical resolution to well under the Nyquist
    frequency.

    (3) The human eye is far more sensitive to luminance detail than color, so
    the Bayer compromise (lower color resolution) is undetectable by human
    viewers.
    Something already existed prior to Foveon coming along that was better than
    Foveon: the Canon CMOS Bayer sensor.
    You're right. Canon's CMOS sensors are (marginally) better than the
    competition, and they're doing quite nicely in the market.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 31, 2003
    #2
    1. Advertisements

  3. Ken F.

    lcopps Guest

    The Sigma picture is taken closer or more zoomed into the subject.
    Aside from that the brick mortar is much more clear on the sigma
    picture. From the samples I have seen on this group, there is little
    doubt the Foveon is a much superior color technology. I only wish it
    could find its way more into the mainstream. Canon I would think would
    be especially interested since the Foveon is also a CMOS sensor. A small
    company living on Venture Capital pretty much shoots its wad developing
    the initial product and relies on optimistic sales forecasts to fund
    further development. Tech companies are for the most part grossly under
    funded from the outset. I cannot believe that this will be the end of
    their multi-layer silicon technology. If they go out of business, the
    patent will fall into receivership and be sold to a competitor most
    likely. Canon could make better use of this technology since they are
    already committed to CMOS sensor technology.
     
    lcopps, Oct 31, 2003
    #3
  4. Ken F.

    Searching_Ut Guest

    Not really interested much in the which is better debate, I already played
    with both, and made up my own mind in that area. Since you say you're
    interested in the technical aspects however, I thought I might point out
    that your comparison has a major flaw when it comes to comparing
    resolution/detail. Notice the images are framed quite differently, and the
    cannon photo is displaying considerably more area. I didn't take the time to
    do a pixel level calculation, but visually it appears the Sigma image is
    roughly half the size area wise. You'd have to frame the shots the same to
    make the comparison worth much. Having the same dumpster in both shots, same
    lighting etc would also be needed.

    The ability to compare apples to apples is the main reason I still make my
    purchases at brick and mortar camera shops instead of on-line. To me doing
    my own hands on comparison is worth more than hundreds of reviews. I often
    find that in looking at the data in the reviews, I miss things (Like the way
    a sample image was framed) that can make a considerable difference in the
    outcome of the photo.

    For what it's worth

    Jeff
     
    Searching_Ut, Oct 31, 2003
    #4
  5. (Ken F.) wrote in
    The 10D is a 6 megapixel camera. Preddy is claiming the Foveon sensor can
    compete with the 1DS which is an 11 megapixel camera. He also says that
    Foveon and Sigma plan to start marketing the 3 megapixel Foveon as a "10
    megapixel" sensor.

    I don't have a problem with Sigma or the Foveon sensor. I think it's
    impressive enough that their 3 megapixel sensor can compete with 6
    megapixel Bayer sensors. Sexing the story up to "13 megapixels", though,
    irritates me immensely.
     
    Tony Whitaker, Oct 31, 2003
    #5
  6. Ken F.

    Darrell Guest

    What is Coors and Budweiser???
     
    Darrell, Oct 31, 2003
    #6
  7. Ken F.

    Q Guest

    Darrell - typed:
    Crap American beer? Actually, the Czech Budweiser is fine. As for the
    Foveon, it is an impressive achievement but isn't ready yet IMO. Sigma
    would not be my 1st choice of partner but can they afford to be fussy?

    The low-end digital sensor market will soon quadruple if mobile/cellular
    phones with cameras take off which I predict they will - big time.
    Probably not a market that Foveon can capitalise on though.

    Funnily enough, I worry about the optical resolution/autofocus accuracy
    of my C4040 more than the pixel count's contribution to picture quality.
     
    Q, Oct 31, 2003
    #7
  8. Ken F.

    Chris Brown Guest

    Fizzy solutions of sugar, ethanol and a few nasty tasting trace substances
    that is chilled to an extent that, should you actually have the desire to
    ingest the foul stuff, your taste buds are numbed to the extent that it's
    almost bearable. They are claimed by some to have some sort of relationship
    to beer. HTH!
     
    Chris Brown, Oct 31, 2003
    #8
  9. Ken F.

    Mike Engles Guest


    Hello

    What you say may be quite correct, but as far as I can see and so has
    the original respondent, the Sigma images look the sharpest.

    I also do not own a digital camera, but follow these threads to get an
    idea of what is happening.

    Mike Engles
     
    Mike Engles, Oct 31, 2003
    #9
  10. Ken F.

    Darrell Guest

     
    Darrell, Oct 31, 2003
    #10
  11. Ken F.

    Browntimdc Guest

    Pale imitations of Milwaukee's Best :)

    Tim

    --

    "The strongest human instinct is to impart information,
    and the second strongest is to resist it."

    Kenneth Graham
     
    Browntimdc, Oct 31, 2003
    #11
  12. Ken F.

    Darrell Guest

    American Beer is an oxymoron ;)
     
    Darrell, Oct 31, 2003
    #12
  13. Ken F.

    PTRAVEL Guest

    Unfortunately, comparing the two photos doesn't provide any useful
    information, nor are your conclusions based on them valid, as discussed
    below.

    First, the two shots were taken at different at different times of day and,
    it would appear, at different times of the year -- take a look at where the
    shadows fall. Clearly, this is going to effect color rendition.

    Moreover, they were quite some time apart -- the Canon 10D shot shows
    satellite dishes on the roof of the lower buidling, whereas the Foveon shot
    does not. There are also two different dumpsters, which account for the
    difference in color (and, possibly, readability of the numbers) -- the Canon
    10D dumpster looks considerably older and more weathered. Also, the low
    building was either re-painted for the 10D shot, or had weathered in the
    Foveon shot.

    In short, it's impossible to compare color rendition based on these two
    far-from-identical photos.

    Next, as someone else noted, either the lenses used differed in focal
    length, or the Canon 10D shot was taken further away which, too, will impact
    such things as relative sharpness of the numbers on the dumpster and
    rendition of the bricks.

    There are some interesting differences worth noting, however, though bear in
    mind that what is being viewed are JPEGs -- the compression process, itself,
    may have introduced some artifacts and distortion.

    If you zoom in on the street sign, you will see pronounced stair-step
    artifacts in the Foveon shot which are not present in the Canon 10D shot.

    Take a look at where the overhead wires cross the white window frames.
    Shots from both cameras exhibit stair-step artifacts, but they are far more
    pronounced in the Foveon shot.

    If you look at the windows under the overhang on the side of the 2-story
    building, you'll notice somewhat better shadow detail in the Canon 10D shot
    than in the Foveon. This may be the result of the lighting, as opposed to
    differences in the cameras, but it is a difference of note nonetheless.

    Accordingly, your conclusion that the Foveon is "sharper" is really
    meaningless, since (1) the differences are attributable, at least in part,
    to different focal length lenses and/or different distances from the
    subject, and (2) any apparent sharpness advantages in the Foveon are at the
    expense of stair-step digital artifacts.

    It's also impossible to tell anything about color rendition of the cameras,
    since the two shots are of different subjects (the buildings were repainted
    and the dumpsters differ), taken at different times of day and, possible,
    different times of year. It can be seen, though, that the Canon 10D has
    slightly better shadow detail.

    One thing is clear, though: both cameras take impressively good pictures.
    Isn't that all that matters/
     
    PTRAVEL, Oct 31, 2003
    #13
  14. Ken F.

    Lisa Horton Guest

    Wrong. There are actually quite a few excellent American beers.
    They're just not usually made by the big companies that make the far
    more popular beer flavored soft drinks.

    Lisa
     
    Lisa Horton, Oct 31, 2003
    #14
  15. Ken F.

    Darrell Guest

    I will raise a local Canadian micro-brewed beer to all you that know the
    difference ;)
     
    Darrell, Oct 31, 2003
    #15
  16. Ken F.

    Xiaoding Guest

    I see Professor Theory has surfaced yet again...son to be
    bitch-slapped by Mistress Reality.

    "
    Humans cannot detect the limited color in digital cameras? It's the
    first thing I saw, and I'm pretty sure I'm human. In fact, the piss
    poor color produced by digital is it's greatest weakness. I don't
    give a crap about what some "theory" says. Most theory is
    misconstrued and mis-applied by most folks anyways, because they don't
    take time to learn the caveats and limitations of the theory. 255
    gradations of anything is crap. Film gives me UNLIMITED
    gradations...infinite. Foveon is an improvement over the Bayer in
    this respect.

    If Foveon gets over this little hump,it is the sensor of the
    future...unless something better comes along. But Bayer sensors both
    blow and suck.
     
    Xiaoding, Oct 31, 2003
    #16
  17. Ken F.

    Brad Guest

    Coming from a 'nuck? Canada is the most over-rated beer country on the
    planet. I'm sure they've got some decent micros like everybody else, but
    their staples are barely a step up from the parent country's.
     
    Brad, Oct 31, 2003
    #17
  18. SNIP
    That's correct, but on further examination you'll see that the sharpness is
    exaggerated beyond what was in the original scene. That may sound as a nice
    feature, and sometimes it is, but you have no control over it. In fact, the
    potential benefit becomes a drawback on slanted edges. They will show
    stairstepping that's making the image look worse than real life, and you
    can't avoid it because aliasing must be filtered before it is captured.
    That's the right attitude. There is some useful content after you filter the
    noise...

    Bart
     
    Bart van der Wolf, Oct 31, 2003
    #18
  19. Ken F.

    eawckyegcy Guest

    I see nice technical points being made by the educated, and ignored by
    the dimwits. I haven't seen much "my camera is better than your
    camera" from anyone except the local Sigma/Foveon lunatic. Consider
    your strawman construction permit hereby revoked.
    Virtually all of the relevant variables were changed from image to
    image (looks like about 6 months of time between them). Did you even
    look at the EXIF header? It takes seconds to do.

    m$ jhead IMG_0082.JPG
    File name : IMG_0082.JPG
    File size : 3114724 bytes
    File date : 2003:10:31 20:34:23
    Camera make : Canon
    Camera model : Canon EOS 10D
    Date/Time : 2003:04:01 15:26:24
    Resolution : 3072 x 2048
    Flash used : No
    Focal length : 38.0mm (35mm equivalent: 60mm)
    CCD width : 22.66mm
    Exposure time: 0.0040 s (1/250)
    Aperture : f/10.0
    ISO equiv. : 100
    Metering Mode: matrix
    Jpeg process : Baseline
    m$
    m$ jhead IMG09640.JPG
    File name : IMG09640.JPG
    File size : 3589053 bytes
    File date : 2003:10:31 20:34:55
    Camera make : SIGMA
    Camera model : SIGMA SD10
    Date/Time : 2003:10:23 11:37:27
    Resolution : 2268 x 1512
    Flash used : No
    Focal length : 50.0mm
    Exposure time: 0.0025 s (1/400)
    Aperture : f/8.0
    ISO equiv. : 100
    Metering Mode: matrix
    Exposure : program (auto)
    Jpeg process : Baseline

    Note the difference lens, f/ratio and exposure times. In all cases,
    the odds are weighted in the SD10's favour: f/8 vs. f/10, 1/400 vs
    1/250 (if hand-held), 50mm vs. 38mm lens. More on the lens below.

    Did you also notice that the numbers on the dumpster changed in the
    intervening months between the images? Recognizing numbers at the
    extreme limits of resolution depends alot on the numbers in question
    (as anyone who has taken an eye examination knows). Looks like the
    SD10's numbers are mainly in the "easy" set, and those in the 10D's
    are generally harder.

    Also in the SD10's favour is that there are more pixels given to the
    numbers in the SD10's image than in the 10D's image. (Shock! Dismay!
    How could this possibly be?!)

    But even ignoring all of that, just pretending that none of it could
    have any effect at all, it is pretty clear that the focus of the 10D's
    image is in general very soft. According to another tool (I am slowly
    developing), the photographer AF'd, and sensors #1 and #4 (counting
    from left to right on the horizontal) were used. Given my experience
    with the AF on the 10D, I am not surprised things turned out a bit
    mushy. Especially when the JPEG "sharpening" option was set to a
    miserly +1.

    However, even all _that_ aside, what bothers me most is that for the
    10D a cheapie 28-135mm zoom lens was used, but (by the looks of it) a
    50mm prime lens was used on the Sigma. (Praise Allah indeed for the
    Canon EXIF MakerNote!)

    Conclusion: the SD10 may be the greatest camera ever built by mere
    mortals but these images are hardly the definitive proof.

    Recommendation: if you want to be a hyper-critical pixel police
    freak, better to do the job to completion. You otherwise run the risk
    of becoming another George Preddy...
     
    eawckyegcy, Oct 31, 2003
    #19
  20. Because the test was rigged: see the other respondents.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 31, 2003
    #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.