Page 1 of 7 123 ... LastLast
Results 1 to 10 of 68

Thread: Myth busted: Canon's 14-bit snake oil

  1. #1
    Senior Member
    Join Date
    Dec 2008
    Location
    Vancouver, Washington, USA
    Posts
    1,956

    Myth busted: Canon's 14-bit snake oil






    Myth: 14-bit raw files are better than 10-bit (or 12-bit) raw files. ("Smoother tonal transitions, more natural gradations, etc.")


    Fact: In current Canon cameras, there is no benefit whatsoever, just slower FPS, larger files, and typical Canon snake oil.


    This myth is repeated in a lot of Canon's marketing materials. For example, Canon's brochure says: "The EOS 5D Mark II also employs precision 14-bit A/D converters to process the output of the imaging sensor. This ensures smoother tonal transitions and more natural gradations."


    • In reality, the tonal transitions and natural gradations are never limited by the raw bit depth, but by noise.

    • The amount of noise is determined principally by the amount of light.

    • If you are shooting in low light, the amount of light is determined by the sensitivity.

    • If you are shooting in ample light, the amount of light is determined by the dynamic range.

    • The maximum dynamic range is limited by bit depth.




    From this, we can see that there really is a relationship between tonal transitions and bit depth: if you are shooting in ample light and are at the maximum dynamic range allowed by your bit depth, the only way to improve tonal transitions and gradations is to increase bit depth and dynamic range at the same time.


    Here is a demonstration of how bit depth relates to raw levels. It's a 40D high ISO image that shows how noise erases the benefit of higher bit depths:


    http://thebrownings.name/images/2009-03-bit-depth/John-Sheehy-14-bit-gaborstrapbits.jpg


    There is another factor here: some of the dynamic range in Canon cameras is not useful to normal photographers. For example, on the 5D2, there is a about 12 bits of real dynamic range stored in a 14-bit file. But if you use more than 10 bits worth of dynamic range, the shadows will be contaminated by pattern noise. The only people who aren't affected by pattern noise are those who stack multiple photos of subjects that are in the exact same position in every photo, such as astrophotographers. For everyone else, the last four bits are just wasted disk space.


    Many of Canon's 14-bit cameras barely have have enough usable dynamic range to justify 10 bits, let alone 12. When Canon upgraded their raw files from 12-bit to 14-bit, all they did was tack on a waste of space. Competitors are really leaving them in the dust now. The Nikon D7000 and Pentax K-5 both have true 14-bits worth of usable dynamic range, and they are a lot cheaper than the 5D2, 1D, and 1Ds which all have less than 12 bits of usable dynamic range.

  2. #2
    Senior Member bob williams's Avatar
    Join Date
    Dec 2008
    Location
    Central New Mexico
    Posts
    1,983

    Re: Myth busted: Canon's 14-bit snake oil



    Hmmm,


    Daniel, this is very enlightening and more importantly makes sense-----I would love to see the "canon engineer
    Bob

  3. #3
    Senior Member
    Join Date
    Dec 2008
    Location
    Vancouver, Washington, USA
    Posts
    1,956

    Re: Myth busted: Canon's 14-bit snake oil



    This is just one example of the dozen or so terrible design flaws that I know about in Canon firmware. As I see it, there are several possibilities.


    One is that the Canon firmware engineers are all brilliant, and they hate to hoist such terrible camera designs on the poor unsuspecting public, but they are terribly constrained by some corporate problems.
    • Maybe they have Marketing Department Overlords, and when Marketing says they want to sell "14-bit", they force it on all the raw files, whether it's wasteful bloat or not.
    • Or perhaps it's some sort of middle manager or tenured employee who thinks 14 bits are beneficial, and either wont listen to reason or there is a culture of "don't point out the flaws of your boss".
    • Or possibly the paperwork to fix some of these design flaws is just too much overhead, so it's easier to leave the flaws as they are.



    I've been a software engineer at Fortune 500 companies before and seen all these problems cause terrible flaws. When you first start there, you fight against the Machine and try to do good work despite all the road blocks, but your spirit will break eventually.


    Another possibility is that most of the engineers on the firmware team are numpty divots that couldn't code their way out of a paper bag. I know that the sensor designers are brilliant for sure, but the firmware team is definitely wanting. Maybe they got sacked and the firmware was outsourced to Elbonia.

  4. #4
    Super Moderator Kayaker72's Avatar
    Join Date
    Dec 2008
    Location
    New Hampshire, USA
    Posts
    5,565

    Re: Myth busted: Canon's 14-bit snake oil



    "Numpty divots".

  5. #5
    Senior Member DLS's Avatar
    Join Date
    Dec 2008
    Location
    British Columbia
    Posts
    258

    Re: Myth busted: Canon's 14-bit snake oil



    Thanks a lot for the technical insight Mr Browning. Your passion is much appreciated too. I second Kayaker72..."Numpty Divots"....awesome, ha ha, ha.

  6. #6
    Senior Member DLS's Avatar
    Join Date
    Dec 2008
    Location
    British Columbia
    Posts
    258

    Re: Myth busted: Canon's 14-bit snake oil



    Quote Originally Posted by Daniel Browning
    This is just one example of the dozen or so terrible design flaws that I know about in Canon firmware.

    Does that mean that the fix is just a matter of a firmware update? And, holy crap, what arethe other 11 design flaws I wonder??

  7. #7
    Senior Member
    Join Date
    Dec 2008
    Location
    Vancouver, Washington, USA
    Posts
    1,956

    Re: Myth busted: Canon's 14-bit snake oil



    Quote Originally Posted by DLS
    Does that mean that the fix is just a matter of a firmware update?

    Yes, but the only time I've seen Canon do anything *substantial* in a firmware update was the 5D2, so I'm not getting my hopes up.


    Quote Originally Posted by DLS
    And, holy crap, what arethe other 11 design flaws I wonder??

    I'll have to make another post some time. Here's a little taste:
    • Clipping 1/3 stop of highlight data on some ISO settings
    • Mangling the histogram of raw images by 1-3 stops.
    • Disabling the ability to use HTP in the very circumstance when it should always be on.
    • Deleting 4-5 stops of highlight data for no reason (depending on ISO)
    • Clipping highlights just to hide the sensor's poor angular response.
    • The 7D mazing debacle.
    • Crippling AutoISO when it's most useful.

    • <meta http-equiv="content-type" content="text/html; charset=utf-8" />
      Hiding important features such as the dead pixel remapper and shutter count
    • Braindamaged sRAW.



    There's so much braindamage in Canon's software development that it's almost overwhelming. 2-year-old startup companies with just a handful of employees have built new cameras from *scratch* that have far, far better software (e.g. RED ONE) and none of the Canon braindamage.

  8. #8
    Senior Member DLS's Avatar
    Join Date
    Dec 2008
    Location
    British Columbia
    Posts
    258

    Re: Myth busted: Canon's 14-bit snake oil



    So are you going to stay with Canon? Or try someone else?


    Quote Originally Posted by Daniel Browning
    Braindamaged sRAW

    So I guess its full res RAW or Jpeg - don't bother with sRAW1 or sRAW2?

  9. #9
    Senior Member
    Join Date
    Dec 2008
    Location
    Vancouver, Washington, USA
    Posts
    1,956

    Re: Myth busted: Canon's 14-bit snake oil



    Quote Originally Posted by DLS
    So are you going to stay with Canon? Or try someone else?

    Well, camera choice depends on a lot more than just how braindamaged the software is. For example, Canon gives you 21 MP full frame for three times less money than Nikon. For me, Canon is still the best option right now, despite their many and varied flaws.I definitely recommend Nikon for other people, but it so happens that Nikon makes a few big mistakes in areas that are important to me -- astrophotography. Canon's mistakes only affect all other types of photography -- for astro stuff, Canon doesn't have any braindamage. In any case,it would be a big hassle and expense to switch, but I'm considering it.


    Quote Originally Posted by DLS
    So I guess its full res RAW or Jpeg - don't bother with sRAW1 or sRAW2?

    No, they have some drawbacks (aliasing artifacts, can't demosaic, low resolution, etc.), but in many ways they're better than JPEG if you need smaller files. The reason they are braindamaged is that the files could be two times smaller, 10 times higher quality, with even less processing power required. Canon invented this *terrible* way to reduce filesize. It's like they're flying a huge cargo plane full of garbage with 5 passengers, and when they need to save weight, they throw out the passengers (resolution) instead of the garbage (raw levels that exceed the maximum precision of photon-shot-noise-limited quantization steps, among other things).



    <meta http-equiv="content-type" content="text/html; charset=utf-8" />

  10. #10
    Administrator Sean Setters's Avatar
    Join Date
    Dec 2008
    Location
    Savannah, GA
    Posts
    3,360

    Re: Myth busted: Canon's 14-bit snake oil



    Wow, this post has been highly informative...and entertaining, actually...

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •