Page 1 of 6 12345 ... LastLast
Results 1 to 10 of 60

Hybrid View

  1. Collapse Details
    C300 MKII and FCPX Gamma shift with since latest firmware upgrade 1.0.6
    #1
    Senior Member
    Join Date
    Jun 2013
    Location
    Dubai & London
    Posts
    189
    Default
    Hi,


    Since I have upgraded to the latest C300 MKII firmware 1.0.6 I am now getting a gamma shift, raised blacks when I playback the footage or scrub a clip in FCPX ( also latest version on El Capitain)

    I narrowed it down to the firmware upgrade as the culprit as when I viewed the footage ( all recorded on the previous firmware 1.0.5 ) the footage plays fine but all the footage from projects since the latest firmware produces this gamma shift / low contrast look which is very annoying.

    The other way to get rid of it is to transcode the clips out of FCPX in some converter and then it plays fine but FCPX was fine before so I would love to know if it is possible to go back to the previous firmware, any ideas would be much appreciated.


    Reply With Quote
     

  2. Collapse Details
    #2
    Senior Member
    Join Date
    Aug 2014
    Posts
    164
    Default
    Cane,

    In the new firmware there is an interesting note about additional metadata being added to Canon Log 1/2/3 footage indicating the video is full range (0-255) as opposed to limited (16-255).

    "10. Metadata related to the XF-AVC range has been corrected.
    - The range of Canon Log/Log2/Log3 has been corrected to full range."

    As I've seen in a lot of NLE's, detailed in this post I did here, the footage was being misinterpreted causing clamped (and clipped) highlights and shadows. So data was being lost essentially on import.

    So, without knowing Final Cut X specifically, I would imagine the image you are now seeing is actually proper compared to the clamped and more conttrasty image you were seeing previously.


    Reply With Quote
     

  3. Collapse Details
    #3
    Senior Member
    Join Date
    Jun 2013
    Location
    Dubai & London
    Posts
    189
    Default
    Hi Macaholic,

    Before the latest C300 MK II firmware upgrade i.0.6, it seemed fine, but now I get a gamma shift every time I play the footage so when it's rendered it raises the blacks but when un-rendered it plays with a normal contrast look.

    Which one is right?

    Can I set the "data" levels in FCPX so it doesn't always play like this, every time I press stop/space-bar or scrub through the clip it's really distracting with the gamma constantly shifting, I don't get why the firmware update would do this unless it's a bug?

    In Canon XF utility player, it plays at the normal contrast level and it matches the same as when I scrub through the footage in FCPX, so if this is the case, then FCPX seems to be interpreting it differently and maybe trying to correct it as you mention but whats the fix?

    I can't seem to find the previous firmware 1.0.5 version on the Canon website either, right now I would love to go back to that as it was totally fine!

    I pasted the images below of what happens, this is meant be C-LOG 3. I can still work with it by adding more contrast or crushing the shadows a tad more than before but it's thrown my workflow off as I had LUTS for other edits I could just copy paste but now i need to re-adjust the blacks and overall contrast...

    Looking at the below images, I thought the more contrasty look (still LOG) would've been true to the C-LOG3 and the raised blacks/flatter look more like C-LOG2, I wonder if anyone else is experiencing this problem?


    Screen Shot 2017-01-13 at 2.55.14 PM.jpg
    Screen Shot 2017-01-13 at 2.54.38 PM.jpg
    Last edited by cane141; 01-13-2017 at 08:31 AM.


    Reply With Quote
     

  4. Collapse Details
    #4
    Senior Member
    Join Date
    Aug 2014
    Posts
    164
    Default
    Quote Originally Posted by drummondb View Post
    And yes, C300mk2 footage shot in C-log 1 responded to the LUTs the same way they always have on C100 footage.
    I know C100 Mark II footage in the .MP4 format had the same issue - as in it was being read wrong in Premiere and the blacks were being crushed (again we are talking pretty slightly, but usually exacerbated by a LUT). Can't speak to the C100 MK I AVCHD though as I never shot that.

    Again the simple and most sure test is to run a few frames of black in Canon Log 1 or 3, import it like you normally would, and see where it falls on your waveform, just above 26 or just below? This is something I do when on-boarding any new camera system now.

    Quote Originally Posted by cane141 View Post
    now I get a gamma shift every time I play the footage so when it's rendered it raises the blacks but when un-rendered it plays with a normal contrast look.
    That's odd - does your waveform drop when you are playing vs. when you are paused?

    Quote Originally Posted by cane141 View Post
    Can I set the "data" levels in FCPX so it doesn't always play like this
    I wasn't able to find anyway to dictate the data range in Final Cut Pro when I used it. I was on an older version as I won't update to Seirra just yet.

    Quote Originally Posted by cane141 View Post
    In Canon XF utility player, it plays at the normal contrast level and it matches the same as when I scrub through the footage in FCPX, so if this is the case, then FCPX seems to be interpreting it differently and maybe trying to correct it as you mention but whats the fix?
    I would be inclined to say that it XF Utility is going to be showing the image scaled correctly. If you can get the final cut preview window right next to the xf preview window and they are at all different then I would say there is a problem.

    In my version of Final Cut there is no way to change the waveform to code scale or smaller IRE values than 25, so it's impossible to check for certain like you can in Premiere.

    In your two images Cane, the bottom one looks correct to me for Canon Log 3. In fact bringing the two into DaVinci, I can see that the bottom has blacks all above 128 (since there is no pure black portion in the scene this is what I would expect) and image 1 has blacks plunging down to the 70's - something that is not technically possible in an unadulterated Canon Log 3 curve.

    Canon Log 2's black point is actually 96, so while the mids and highs are more compressed then Canon Log 3/1 giving visually a flatter image, the black point of Canon Log 2 being 96 will technically result in ever so slightly blacker blacks out the gate. Don't confuse that for crushed blacks/less information though, as the curve is actually showing us detail from the 96-128 region.
    Last edited by Macaholic; 01-13-2017 at 04:18 PM.


    Reply With Quote
     

  5. Collapse Details
    #5
    Senior Member
    Join Date
    Jan 2011
    Posts
    251
    Default
    Quote Originally Posted by Macaholic View Post
    "10. Metadata related to the XF-AVC range has been corrected.
    - The range of Canon Log/Log2/Log3 has been corrected to full range."
    Sigh... If that's what Canon has done in this new firmware, then they have screwed up. Canon Log 1/2/3 are *not* full range. They use standard BT.709 level range encodings, though with an extended range of highlights above 100% (superwhites). It's all documented in Canon's white papers.

    On Macs, in FCPX (and in DaVinci Reslove for H.264 files; not sure about XF-AVC), this is going to be a problem, because Quicktime automatically compensates when the full_range flag is set in the recorded media. If you record externally, no such flag would be set. And on Windows, DaVinci Resolve doesn't know how to read the full_range flag, and so everything is interpreted according to BT.709 encodings (automatically for most formats, or if you set data levels to Video), which will be correct for all variants of Canon Log. So this new problem will only be visible if you use FCPX (and maybe DaVinci Resolve) on the Macintosh with internally recorded video files.

    You can test by recording black and measuring the black level in your NLE's video scope.

    Canon Log has black at 7.3% RGB. If you see black at 12.5%, it's wrong.
    Canon Log 2 has black at 3.5% RGB. If you see black at 9.3%, it's wrong.
    Canon Log 3 has black at 7.3% RGB. If you see black at 12.5%, it's wrong.

    Canon Log White Paper: On page 10 in figure 9, 0% reflection corresponds to 7.30597% video out. Also you can see that they use BT.709 level encodings, because 0% video out has an 8-bit code value of 16. In a full_range encoding, 0% video out has a code value of 0.

    Canon Log 2 White paper: On page 7 in figure 6, 0% reflection corresponds to 3.54% video out. As with Canon Log 1 you can see that they use BT.709 level encodings, because 0% video out has a 10-bit code value of 64. The math all works out for BT.709 level encodings: 0% reflection has a 10-bit code value of 95. (95-64)/(940-64) = 3.5%. In a full range encoding, code value 95 is 95/1023 or 9.3%.

    It would be really good if some of you could help out with getting sample files recorded with firmware 1.0.6 and with 1.0.5, so I can analyze and compare and be able to go back to Canon with exactly the right information about what they need to fix. A short clip recorded with the lens cap on will be sufficient.
    Last edited by balazer; 01-17-2017 at 08:06 PM.


    Reply With Quote
     

  6. Collapse Details
    #6
    Junior Member
    Join Date
    Nov 2016
    Posts
    21
    Default
    Here is a sample file with lens cap on, firmware 1.0.6, Canon Log 3 : C.Gamut, preset.

    https://www.dropbox.com/s/8u4vda7pg6...CANON.MXF?dl=0


    Reply With Quote
     

  7. Collapse Details
    #7
    Senior Member
    Join Date
    Jan 2011
    Posts
    251
    Default
    Quote Originally Posted by Alex P View Post
    Here is a sample file with lens cap on, firmware 1.0.6, Canon Log 3 : C.Gamut, preset.

    https://www.dropbox.com/s/8u4vda7pg6...CANON.MXF?dl=0
    Hi, Alex.

    Thanks for uploading. It was an elephant in a room, not black. I analyzed the clip, and it does have the H.264 VUI video_full_range_flag set. That would appear a bug in Canon's firmware, but I'd still like to analyze an all-black clip so I can be absolutely sure before I reach out to Canon.
    Last edited by balazer; 01-19-2017 at 08:33 PM.


    Reply With Quote
     

  8. Collapse Details
    #8
    Junior Member
    Join Date
    Nov 2016
    Posts
    21
    Default
    Hi balazer,

    Sorry for the elephant, wrong clip -that clip was not even Clog. Here is the correct clip - https://www.dropbox.com/s/0wa8p9ibga...CANON.MXF?dl=0

    As Sunnyjimfilms stated the shifting behaviour in FCPX is fixed with the latest update. However FCPX stil interpretes the C-log clips in what looks to be the "wrong gamma". See post #24 in this thread.


    1 out of 1 members found this post helpful.
    Reply With Quote
     

  9. Collapse Details
    #9
    Senior Member
    Join Date
    Aug 2014
    Posts
    164
    Default
    Quote Originally Posted by balazer View Post
    Sigh... If that's what Canon has done in this new firmware, then they have screwed up. Canon Log 1/2/3 are *not* full range. They use standard BT.709 level range encodings, though with an extended range of highlights above 100% (superwhites).
    When looking at a 10 bit code value scope in say DaVinci Resolve, setting the footage to Full Range lands a black capped frame of Canon Log 1/3 at 128 - and that is the stated black point of Canon Log 1/3. When setting it to video levels, that level is brought down well below where the curve maps black. The curve literally does not allow for any data to fall below 128.

    The REC Out terminal also fed to a 8 bit scale shows a live capped off shot at 32.

    Everything I've seen leads me to believe the Canon Log's are encoded at full 0-255, while Wide DR is 16-255 (Video Levels with Super Whites) as you refer to balazer.

    But that line in the white papers did give me doubt - I think it is just worded a little odd.


    Reply With Quote
     

  10. Collapse Details
    #10
    Senior Member
    Join Date
    Jan 2011
    Posts
    251
    Default
    Quote Originally Posted by Macaholic View Post
    When looking at a 10 bit code value scope in say DaVinci Resolve, setting the footage to Full Range lands a black capped frame of Canon Log 1/3 at 128 - and that is the stated black point of Canon Log 1/3. When setting it to video levels, that level is brought down well below where the curve maps black. The curve literally does not allow for any data to fall below 128.
    Scopes in DaVinci Resolve don't show you 10-bit code values. The scale on the scopes is pure fiction. Resolve's pixel format is 32-bit floating point RGB, with reference white at (1,1,1) and reference black at (0,0,0). The top of the scopes is 1 and the bottom of the scopes is 0. At no point does Resolve show you the actual code values of the video file. The file gets decoded and interpreted and mapped into the 32-bit floating point values, and yes, that interpretation depends on how Data Levels is set in Clip Attributes. On the Mac, it can also depend on what interpretation Quicktime does when decoding the video file, before any image data is even passed to Resolve.


    Reply With Quote
     

Page 1 of 6 12345 ... LastLast

Tags for this Thread

Posting Permissions

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