Announcement

Collapse
No announcement yet.

Pocket 4K red clipping

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    quote by CaptainHook:

    There are two parts to BMD Film, the log curve (gamma) and the "colour space" (gamut). The log curve is always optimized for the sensors dynamic range so the curve you get for cameras with different sensors will vary. Some will have more contrast than others. The gamut in previous versions would vary across cameras as well but with Gen 4 it is the same for any of our cameras using it and like Gen 3 it is more saturated to make it faster/easier to grade.
    All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/

    Comment


    • #17
      I have also tried V4 color science with the BMPCC. The clipping is much more pronounced in the BMPCC4K files. It's not just an overexposed problem, because others colours are much better handled than the red. The saturation map method can fix this but with a counterpart on a prores file : you lose saturation in others area of your image.

      Comment


      • #18
        Originally posted by slothorp View Post
        I have also tried V4 color science with the BMPCC. The clipping is much more pronounced in the BMPCC4K files. It's not just an overexposed problem, because others colours are much better handled than the red. The saturation map method can fix this but with a counterpart on a prores file : you lose saturation in others area of your image.
        I agree, something is off and will probably get fixed. Let's hope they already know about it.

        this picture from All2human over in the footage and frames thread exhibits the problem on the tail light of the motorbike in the far distance. https://www.flickr.com/photos/merumontanus/30473709277/

        looks like a hot pixel but it's just the red from the tail light and I don't see how it'd be over exposed?

        Comment


        • #19
          You are shooting with a 10 stop Prores video codec. Log or not that is all the DR you get. Clipped highlights look ugly and are not recoverable. The DR of this scene exceeds the DR of the codec and you have to experiment with exposure to find the best compromise between clipping highlights and clipping blacks. The red blooming is characteristic of 4:2:2 sub sampled chroma clipping.
          Raw gives you much more room to play with this kind of scene.
          Two screen shots from Resolve showing the scopes and the correct color managed settings for BM4K log. Straight log to the timeline and log to REC709.
          BMPCC4k log over.jpg

          BMPCC4k log to 709 over.jpg

          Comment


          • #20
            What about my screenshots? They are Raw 3:1 where v1 looks completely different to v4?

            Comment


            • #21
              The problem is much less visible with others cameras, as we can see with Ryanite screenshots, even in V4 colour science. It affects only the red channel. And the fact that it can be manageable in resolve (with counterparts) shows that you can't use others NLEs. That is why I think that it's not just a common exposure problem. It's a BMPCC4K problem.

              Comment


              • #22
                Here is a screen shot with scopes of the extended video gamma segment. Straight REC709 to the timeline as in your original settings. You can see the hard clipping in both RGB parade and vectorscope. The arrows pointing to white lines on the RGB scope graticule indicate REC709 reference white at 940 and reference black at 64.
                Try some similar experimental shots and bracket your exposures +2 to -2 in 1/2 stop increments to get a better idea of what works. Shoot at the camera's native 800 ISO setting.
                Shoot some short raw clips too. Prores is just video.

                BMPCC4k 709 over.jpg
                Last edited by razz16mm; 10-19-2018, 04:29 PM.

                Comment


                • #23
                  It is not an overexposure issue. Yellows, blues, greens are perfectly handled when overexposed : desaturated with a creamy hallo around. The reds are oversaturated with harsh boundaries, like it was digitally painted.

                  Comment


                  • #24
                    Originally posted by razz16mm View Post
                    Here is a screen shot with scopes of the extended video gamma segment. Straight REC709 to the timeline as in your original settings. You can see the hard clipping in both RGB parade and vectorscope. The arrows pointing to white lines on the RGB scope graticule indicate REC709 reference white at 940 and reference black at 64.
                    Try some similar experimental shots and bracket your exposures +2 to -2 in 1/2 stop increments to get a better idea of what works. Shoot at the camera's native 800 ISO setting.
                    Shoot some short raw clips too. Prores is just video.
                    [ATTACH=CONFIG]26956[/ATTACH]
                    ehh how is prores "just video" prores can handle a very similar amount of dynamic range, shoot in film log and with the prores codec should not create this effect.

                    Comment


                    • #25
                      Originally posted by Ryanite View Post
                      ehh how is prores "just video" prores can handle a very similar amount of dynamic range, shoot in film log and with the prores codec should not create this effect.
                      Log curves compress sensor linear DR into smaller display DR color spaces. They allow you to preserve values and details that would otherwise be clipped and lost. Prores 10 bit 4:2:2 is pretty much a standard REC709 compliant video codec.
                      Either way, even the sensor has its limits. What you are seeing is strictly the result of exposure produced hard clipping as the scopes clearly show, both in log and in REC709. It is baked in in Prores and you don't have any recovery room.
                      Linear raw file DR is only limited by the camera sensor. Not so with Prores in any of its forms. But the sensor is clipped here too it looks like.

                      Comment


                      • #26
                        If it is clipped, it's clipped - period.

                        You can do some sort of highlight recovery, but that's just a band aid - just don't clip it.
                        Raw gives you a bit more wiggle room though.

                        Having said that, aggressive monocromic LEDs can be a problem on any camera, no matter what color.
                        But since we don't have blue brake lights or traffic lights, the problem shows up mostly with red.
                        Blog: http://frankglencairn.wordpress.com/

                        Comment


                        • #27
                          There are two points I wish to answer.

                          First, when it is clipped with a BMPCC, it doesn't react the same way than with a BMPCC4K. Actually, it reacts a much better way as you can see in those exemples taken from raw files, same night and same optical setup :

                          BMPCC raw v1_1.1.1.jpg
                          BMPCC raw v4_1.1.2.jpg
                          BMPCC4K raw _1.2.1.jpg
                          Attached Files

                          Comment


                          • #28
                            Second, the issue is only concerning the RED lights. No problem with the greens or the yellows.
                            You can correct it by using the color transform tool and the saturation mapping. But it affects the entire image, including the skin tones.

                            BMPCC4K prores ev_1.3.4.jpg
                            BMPCC4K prores corre ev_1.3.3.jpg

                            Comment


                            • #29
                              New camera always means new problems and risks. It's always like to be a beta tester. Best scenario is that BM team will fix it in firmware. Worst scenario is send it back to factory to recalibrate sensor. (Do you remember Pocket v1 "white orbs" problem?) My guess it is something wrong with Color Science v4 and it should be reported to developers and fixed.

                              Don't you interested to share those side by side BMPCC v1 vs P4K DNG samples? Or some other samples with light bulbs in the dark to compare dynamic range/noise limits?

                              BTW i found that old thread "Ursa Mini 4.6k Blue LED light looks Magenta" https://forum.blackmagicdesign.com/v...ic.php?t=54377 if you download DNG you can see similar problem with wired single color clipping.
                              Last edited by shijan; 10-19-2018, 07:04 AM.
                              All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/

                              Comment


                              • #30
                                You can find prores and dng in this link :
                                https://www.dropbox.com/sh/d6p9k043v...h43D4Aw-a?dl=0

                                Comment

                                Working...
                                X