SD10: a raw file fact about White Balance

Discussion in 'Digital Photography' started by David Kilpatrick, May 23, 2004.

  1. This is not a plug for the SD10, or a knock either. Using many different
    cameras, I've found that on the whole 'white balance' is rarely a
    physical alteration to the raw file (it is always a physical alteration
    to TIFF or JPEGs produced in camera). With .NEF, .MRW, .ORF and probably
    also with Canon files, white balance is just a 'tag' added to the raw
    file. The camera always records using the full sensitivity of the sensor
    across the spectrum, the WB setting may alter overall exposure to permit
    the necessary channel adjustments, but it doesn't change the raw file.

    When you come to import the picture, if you have made an error and set
    tungsten in daylight or vice-versa, you can simply change the white
    balance in the raw file converter. I have not noticed any loss in
    quality when doing this.

    With the Sigma SD10, the situation is quite different. If you set
    Tungsten, you might as well be loading tungsten film. The actual raw
    file is altered in camera to have a correct channel balance for tungsten
    light. I've just shot a set of pix in daylight using tungsten balance,
    in error, and found that no amount of adjustment can correct them. On
    the other hand, I've found that auto or custom set white balances on the
    SD10 are very accurate. There is adjustment of course, with quite a wide
    range, but you can't just change a wrongly balanced file back to normal
    as you can with most raw conversions.

    What I do not know is whether the camera uses differential gain in the
    three layers when you change white balance, or whether it has a
    processing stage in between the usual A-to-D and the .XF3 file creation.

    It's another example of how different the concept is, and how
    differently it works. I don't necessarily like it. It can be useful to
    change the white balance when importing the image, but the SD10 really
    doesn't allow that.

    David Kilpatrick, May 23, 2004
    1. Advertisements

  2. David Kilpatrick

    Helge Olsen Guest

    This leads me to think that the SD10's "raw" files aren't indeed raw. The
    whole point of raw is to NOT alter the result in any way from what's
    recorded by the sensor. The sensor should not give any thought of white
    balance in order to record an image. IE. the Sigma cameras do not record
    true RAW files. They just waste space by not compressing altered images. So
    much for "professional"...

    Helge Olsen, May 23, 2004
    1. Advertisements

  3. David Kilpatrick

    Mark M Guest

    If what you are saying is accurate, then the Sigma...
    ....does NOT have a truly RAW mode at all.

    If true, this is a horrendous oversight (or intentional, but poor design).
    Mark M, May 23, 2004
  4. David Kilpatrick

    E. Magnuson Guest

    One, I think that David may be mistaken. This is an area where
    Laurence M can speak with authority if we've not chased him away.
    Be careful here. The line is not clear cut. What about ISO
    sensitivity? The "sensor" seems the same number of photons for the
    same EV; the ISO is usually just changes to the analog amplifier gain.
    There may be other signal conditioning (like noise reduction) that is
    similarly performed on the analog signal before conversion. There may
    be some digital processing done. AFAIK, no camera maker makes any
    disclosure about exactly how raw is raw. Perhaps we should call it
    "rare" or "al-dente" instead.
    E. Magnuson, May 23, 2004
  5. David Kilpatrick

    Alfred Molon Guest

    Can the XF3 files be imported directly into the RAW converter of
    Photoshop or are they only compatible with the Sigma converter ?
    Alfred Molon, May 24, 2004
  6. Agree. The main purpose of any format is to be a good
    carrier for the photo. No format is entirely RAW. All
    formats are digital - the RAW info is analog. All RAW files
    - except Foveon, I think - are processed for noise reduction.
    And - as you say - ISO sensitivity might be done by different

    Roland Karlsson, May 24, 2004
  7. XF3 files can be imported directly into PhotoShop CS.
    Peter A. Stavrakoglou, May 24, 2004
  8. David Kilpatrick

    E. Magnuson Guest

    I'd bet Foveon data is processed for noise reduction as well.
    I can't prove it, but that's my reading of some of the hints
    dropped in talks and papers. It would also make sense because
    of the analog background of some of their people.
    E. Magnuson, May 24, 2004
  9. SNIP
    That's a correct evaluation of what a Raw file comprises. It may have some
    on chip noise reduction applied (e.g. dark field reduction), but the file
    data is essentially Raw, ADC quantized data (e.g. still in Bayer CFA
    "color", or "as captured"). Other camera settings are only tags, used as
    defaults in the Raw converter firm/software or in-camera JPEG creation.

    There will probably be little difference in the hit on S/N ratio, perhaps
    differential gain being marginally more accurate (avoiding cumulative
    integer rounding errors unless, less likely, rational number arithmetic is
    used throughout). Subsequent color balance conversions are in any case to be
    avoided if accuracy is a concern.
    Interesting observation, thanks for sharing.

    Bart van der Wolf, May 24, 2004
  10. SNIP
    XF3 seems to be compatible with the latest (v2.2) Adobe Raw Converter (or
    rather ARC is compatible with XF3s).

    Bart van der Wolf, May 24, 2004
  11. It's not true. WB is forever, fully, and completely adjustable. Just
    like all RAW modes.
    Georgette Preddy, May 24, 2004
  12. What do you suppose the forever changable WB setting in SPP is for?
    Georgette Preddy, May 24, 2004
  13. Probably a difference in analog gain between colours. If you know
    you're shooting in tungsten light, you know the red exposure is going to
    be a lot greater than the blue exposure, so it makes sense to use more
    gain for blue in order to make better use of the A/D converter range.
    On the other hand, if you choose the wrong type of white balance and
    that results in clipping one channel, that's not recoverable later.

    There's no reason a Bayer camera couldn't do this too, but it sounds
    like they do not in your experience.

    In the case of a sequential-exposure camera or a trilinear array sensor
    (e.g. flatbed scanner, film scanner), the three exposures can actually
    be varied so that all three channels are making use of their full
    dynamic range. That should be possible with a 3 CCD camera with
    electronic shuttering too, but not with a single-sensor design like
    Bayer or Foveon X3.
    You should still be able to make smaller changes in white balance.

    Dave Martindale, May 24, 2004
  14. Alfred Molon

    Why are the messages on this group hidden from public view?

    David J Taylor, May 24, 2004
  15. In the Photo Pro Software, you can change the White Balance, as you can
    in Photoshop CS. Photoshop CS shows a minimal effect on the file when
    this is done, and it is clear the data doesn't provide enough
    information. In Photo Pro 2, the change from 'Incandescent' (Sigma's
    term for Tungsten) to 'Sunlight' (their term for normal Daylight) has a
    more balanced effect, but with a loss of apparently 1.5 stops in
    brightness - indicating that it is not a true reprocessing of absolutely
    raw data. When the 'Auto' adjustment is applied to the file, I notice
    that as well as 0.3+ exposure, a white point of 1.1 is set instead of
    the usual zero for a correctly exposed file. In total, the software has
    adjusted a histogram with approximately 1.5 stops too little exposure
    for a pure white to render as white. This is consistent - pretty much
    exactly - with the raw file data being recorded in 'Incandescent' mode
    with different actual colour channel levels, from the same shot recorded
    in Sunlight mode.

    I guess I will have to test this by making identical exposures now, but
    it seems clear how (and maybe also why) the files and software work this
    way. Yes, George is correct in that the software does allow adjustment
    for changing the colour balance, but it warns you that this will 'alter
    the XF3 file' - not that it will 'tag it with new header information but
    leave all the data totally unchanged'!

    The adjustment I've just made to a file appears to discard brightness
    from the blue channel (primarily). That is a SORT of fix for incorrect
    daylight balance setting. It does about the same as exporting the file
    as 16-bit unchanged into Photoshop, then auto level balancing the image.

    As anyone who has printed colour negs will know, you can overcome
    mismatched illuminants without using filters if you give controlled
    overexposure - approximately 1.5x for tungsten film used in daylight
    (not that much tungsten neg exists now) and 2x for daylight film used in
    tungsten. This provides the extra density in the under-sensitive layers
    to yield a minimum shadow point, while the over-exposed layers can be
    adjusted back with enlarger filtration.

    A typical 12-bit raw file is in some ways like a generously exposed
    negative, permitting rebalancing because all colours have full
    information present. The Sigma file is more like a correctly (minimum)
    exposed negative.

    As said before, there are good and bad sides to this, and I'm not trying
    to judge either way. But I am trying to unearth a bit of technical stuff
    with accuracy and objectivity.

    David Kilpatrick, May 24, 2004
  16. It was the Photoshop RAW converter which alerted me to this. The way it
    failed to show the normal changes when altering the daylight balance, in
    contrast to Sigma's software which apparently does the job (but only if
    set to Auto), was entirely different to the way other files with the
    same shooting error have been adjustable in CS camera raw.

    David Kilpatrick, May 24, 2004
  17. Georgette Preddy wrote:

    Emergencies. As far as I can tell it is not destructive, although it
    warns that file data will be changed. It seems to be possible to
    repeatedly change, save, and re-change the file without damaging the
    data - but there is no doubt the original file when shot using
    Incandescent contains physically different data from the original file
    shot using Sunlight - not just different tags.

    It's definitely accompanied by a density change, and the process uses a
    lot of the 12-bit leeway of the file.

    David Kilpatrick, May 24, 2004
  18. Dave Martindale wrote:

    It does offer a good range of colour adjustment outside the basic white
    balance choices, but the white balance changes available (as in other
    replies I've made) are not the same as those made by other raw file

    David Kilpatrick, May 24, 2004
  19. David Kilpatrick

    Alfred Molon Guest

    You can access the archives by signing in (it's free).
    Alfred Molon, May 24, 2004
  20. I would prefer not to have to join a group to see its archives, thanks.
    To me, it says: "this is private, it is a club, it is not for you".

    Why not make the archives public for all to see, like many other groups?

    David J Taylor, May 24, 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.