Bias value of Canon DSLRs?!

Discussion in 'Digital Photography' started by Marc Wossner, Jun 8, 2008.

  1. Marc Wossner

    Marc Wossner Guest

    Hi ng,

    does anyone know how large the bias or offset is that Canon applies to
    its DSLRs in order not to clipp negative voltage fluctuations to zero?
    I read it must be subtracted from the average raw value of an image in
    the course of analysing the signal to noise ratio.

    Best regards!
    Marc Wossner
    Marc Wossner, Jun 8, 2008
    1. Advertisements

  2. It can vary a little bit, especially with long exposures at high ISOs,
    but generally, the 14-bit cameras use 1024, the non-rebels and the 300D
    use 128, and the other rebels (and also the 10D at ISOs 1600 and 3200 use
    256). I don't recall what the 1Ds cameras do at ISO 3200, but the mk1
    and/or mk2 may use 256 instead of 128.
    John P Sheehy, Jun 8, 2008
    1. Advertisements

  3. Marc Wossner

    Marc Wossner Guest

    Interesting, no fixed value. Does Canon state somewhere what bias they
    use in what camera (I havent found any reference on their website) or
    can it be calculated from some raw values?

    Best regards!
    Marc Wossner
    Marc Wossner, Jun 8, 2008
  4. Just take a black 'exposure" and see what the average is.

    Canon and most every other company thinks we're a bunch of morons with no
    right to any knowledge of how our cameras work under the hood.
    John P Sheehy, Jun 8, 2008
  5. Marc Wossner

    Marc Wossner Guest

    Sorry, Im not so deep into that.
    I guess a black exposure is a dark frame with short exposure time and
    the lens cap on but what do you mean by "the average"?

    Best regards!
    Marc Wossner
    Marc Wossner, Jun 8, 2008
  6. Marc Wossner

    Archibald Guest

    I guess you were just called a moron. (I'm one too.)

    Archibald, Jun 9, 2008
  7. The average RAW value of all the unexposed pixels in the black frame.
    John P Sheehy, Jun 9, 2008
  8. Marc,
    how do you expect a fixed value to be achieved in practice?

    Remember this is an ANALOGUE offset. For example, if the "target"
    offset is 128 on a 12 bit ADC then the total offset is only 3% of the
    entire analogue range and needs adjustment with a precision of only
    0.25%. Analogue signals are simply not that precise or consistent and
    will vary by far more than this with age and over the operating
    temperature range of the unit.
    Kennedy McEwen, Jun 9, 2008
  9. With the newer Canons, it seems to be calculated on the fly. The 10D
    "guessed" the exact offset, and was often a bit off.
    John P Sheehy, Jun 9, 2008
  10. Since it is analogue it can't be "calculated", whether on the fly or
    otherwise. What is being offset is the analogue voltage input to the
    ADC, the digital output is only the consequence of that offset.
    Kennedy McEwen, Jun 9, 2008
  11. The ADC itself may be exceedingly accurate in the levels that its zero
    and full scale are set, but that isn't the problem - it is offsetting
    the analogue input by a fixed and consistent amount, not only over
    temperature but over millions of different circuits for their entire
    The only time a "dark frame" is used with the Canon cameras is when long
    exposure noise reduction is enabled in CF-02.
    Kennedy McEwen, Jun 9, 2008
  12. The camera still has to decide what that offset will be, to get the analog
    zero signal mean to correspond to the target digitized value. The 10D was
    inconsistent, the newer cameras are not, so the newer cameras are probably
    looking at some unexposed pixels, or doing something else like a self-test
    or factory calibration to get the desired results.
    John P Sheehy, Jun 10, 2008
  13. Hence, his "if desired".

    You're never too smart to slow down and read!
    John P Sheehy, Jun 10, 2008
  14. Which, in the case of the Canon cameras is not the case for the majority
    of the exposure range.
    Quite, you should try it sometime!
    Kennedy McEwen, Jun 10, 2008
  15. I don't believe it has to "decide" anything of the sort.
    You're sure about that?
    For many of the reasons given already, a factory calibration would be
    likely to achieve anywhere close to the precision necessary - 1 part in
    4096. Analogue electronics just isn't that stable.

    A self test would be "observable" to the user, especially at long

    Unexposed pixels might be possible, but again, I doubt this would
    achieve the expected consistency. Unexposed pixels are outside the edge
    of the frame - a look at the black level of pixels inside the edge of
    the frame, especially at long exposures, shows a significantly different
    mean black level from the central area.
    Kennedy McEwen, Jun 10, 2008
    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.