Announcement

Collapse
No announcement yet.

BMPCC 6K lab test - dynamic range, rolling shutter, latitude and more

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

  • BMPCC 6K lab test - dynamic range, rolling shutter, latitude and more

    Hi Guys,

    I have conducted quite intensive lab tests with the new BMPCC 6K. You can find a write up here:

    https://www.cinema5d.com/blackmagic-...-shutter-more/

    Let me know if you have questions!

    best regards Gunther

  • #2
    Very nice, thanks. Great that you prefer a conservative approach to measure DR, with Alexa at 14 stops as a reference.
    I see you tested the Mini Pro 4.6, did you maybe happen to have also tested the original BMCC or Pocket/Micro with this test?

    Comment


    • #3
      Originally posted by stip View Post
      Very nice, thanks. Great that you prefer a conservative approach to measure DR, with Alexa at 14 stops as a reference.
      I see you tested the Mini Pro 4.6, did you maybe happen to have also tested the original BMCC or Pocket/Micro with this test?
      Thank you - I tested also the original BMPCC (pocket cinema camera) - it came out at 11.2 stops.

      Comment


      • #4
        Originally posted by GMC View Post
        Thank you - I tested also the original BMPCC (pocket cinema camera) - it came out at 11.2 stops.
        Thank you!

        Comment


        • #5
          I was inspired by those tests and take attempt to exam BMMCC in near similar conditions. DNGs processed with workflow described here https://forum.blackmagicdesign.com/v...543725#p537852 Noise reduction was applied before expose boost.
          My background is simple black fabric, so it is darker and so more extreme test than Cinema5D examples, you can compare things by Color Checker patches.
          Let's be honest, even P6K can't provide usable underexposed 5 stops. It looks too noisy and became plastic mess if you add SNR. So we see here is 5 years old native HD camera with Fairchild Imaging dual gain sensor fights against gigantic 6K/4K modern Sony sensors in downscaled to HD mode. I must admit again that 3 stops under expose is pretty usable with BMMCC. If only we can fight FPN it could be extended even further...
          P.S. I only can imagine what newer Fairchild Imaging dual gain sensor is capable of. Along to 4K it technically waaaay less noisier (Dark Current: 2 e-/sec compare to BMMCC sensor: 25 e-/sec, Ursa 4.6K sensor: 15 e-/sec) (Read Noise 1.0 e- RMS. BMMCC sensor: 1.2 e- RMS, Ursa 4.6K sensor: 1.5 e- RMS)
          It is better to see images at full size in new tab:








          And same examples, but with noise reduction and fIlm emulation LUT:






          Last edited by shijan; 09-03-2019, 08:12 AM.
          All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/

          Comment


          • #6
            Very interesting, thanks for sharing.

            The latitude test is lacking overexposure though. I think footage from the alexa can't be pushed up this much, but if you overexpose your shot it survives without any issues. Also with log, but particularly with cameras that shoot RAW, I think you have to look at both sides.
            My YouTube channel

            Comment


            • #7
              Technically digital sensors don't have such a thing as overexposure. They also don't have highlights rolloff. They capture light in linear gamma and hard clipped in one single point. Next it is usually converted to LOG gamma specially designed for each sensor model. Arri sensor just can capture a lot of dynamic range and Arri LOG gamma designed to shape huge amount of that dynamic range to highlights. So in final when you shoot ProRes with baked log gamma and process image with Arri Color Science you have very large amount of data from middle grey to highlights clipping point.
              All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/

              Comment


              • #8
                Thanks Shijan - interesting to see. Yes, the problem with all Blackmagic Cameras so far was fixed pattern noise or stripes in the images. That seems to have gone with the latest BMPCC 6K

                Comment


                • #9
                  Originally posted by shijan View Post
                  Technically digital sensors don't have such a thing as overexposure.
                  Well, true, but then they don't have such a thing as underexposure either. And yet, it's still useful to run both overexposure and underexposure tests (with respect to some reference, usually the manufacturer's definition of "correctly exposed"), to see how much over-under range you have. I mean, these tests are definitely useful (much more than the DR chart tests), I'm just saying I'd like to see the overexposure side too.
                  My YouTube channel

                  Comment


                  • #10
                    Originally posted by Samuel H View Post
                    Well, true, but then they don't have such a thing as underexposure either. And yet, it's still useful to run both overexposure and underexposure tests (with respect to some reference, usually the manufacturer's definition of "correctly exposed"), to see how much over-under range you have. I mean, these tests are definitely useful (much more than the DR chart tests), I'm just saying I'd like to see the overexposure side too.
                    I really like LensProToGo's exposure recovery tests that they post on their Youtube page. I believe they will be doing the BMPCC6K soon.

                    BMPCC4K:

                    https://www.youtube.com/watch?v=gHMNx_F2oMc&t=236s


                    Ursa Mini Pro 4.6K G2:
                    https://www.youtube.com/watch?v=8t4Wz-9CFV0


                    You can compare against other cameras (RED Monstro, RED Gemini, Alexa Mini, Fuji X-T3, Lumix S1, etc.) from their playlist:
                    https://www.youtube.com/playlist?lis...UU8AmdX_6l3WXY
                    http://instagram.com/teddy_dem/
                    http://vimeo.com/luciddreamsfilms

                    Comment


                    • #11
                      I am probably missing something here, but I really can't the point of analyzing image recovery from just one way (underexposure or overexposure) Does not cameras have different data spread across the full DR?
                      Also I have some thoughts about pink tint that is present on recovered p4k images. Does that color tint is really an evidence of a lack of color information? I mean yes it's a defect, but If we are recovering the image with wrong exposure, why can't we balance the tint? Who knows may be the image with adjusted tint turns to have more accurate colors than the image with correct tint in the first place? This is just my thoughts, I'm totally not an expert here.
                      Last edited by core_50; 09-04-2019, 05:19 AM.

                      Comment


                      • #12
                        Ok, here is BMMCC overexpose test just for fun. I also add two additional underexposed images here, because i use different lighting direction in this test to provide extreme bright light. As before i use REDLog3G10/REDWideGamutRGB as timeline gamma. Other gammas and Log to Rec conversion methods may provide different look of highlight rolloff.









                        And the same with Highlights Recovery turned OFF:





                        All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/

                        Comment


                        • #13
                          Thanks for the additional tests, shijan

                          Originally posted by TeddyDem View Post
                          I really like LensProToGo's exposure recovery tests that they post on their Youtube page. I believe they will be doing the BMPCC6K soon.
                          ...
                          playlist:
                          https://www.youtube.com/playlist?lis...UU8AmdX_6l3WXY
                          Very nice, thanks for sharing! I checked Geoff Boyle's evaluations last year (link, link, link), but he uses much more higher-end cameras that I don't have access to, this is closer to what I can get my hands on

                          So, UMP G2 breaks down at +5 and -4 (it can recover nicely from +4 and from -3, with color shifts but without any other major issues).
                          The p4k is just breaking by a bit in the highlights at +4 and too noisy at -3, which means it has around 1.5 or 2 stops less latitude than the UMP G2.
                          Interestingly, at ISO 3200 it breaks at +5 but it's already noisy at -1, which means slightly reduced DR. I tested this myself and found increased latitude at ISO 3200, everybody said I had done something wrong, and every other test I see says that too, but then I'm shooting in real life and it's ISO 400 that gives me slightly reduced DR. Weird. No idea if there can be sensor variation in this, maybe I got a bad one that shows more noise at ISO 400, or a good one that's cleaner at ISO 3200.

                          The RED monstro (link) is fine-ish at -5 but breaks at +4, where the UMP G2 would be perfectly fine. That means it has one stop more latitude than the UMP G2, not two, as the underexposure test would suggest. As I said: you have to test both sides. The Z7 only starts to show noise at -5 but it's already dead at +1!

                          The S1/S1R shows ugly noise at -4 and completely breaks at +3, so it's a match to the p4k, but with again with different over-under latitude distribution.
                          I already knew the XT-3 was the DR king among hybrid video-stills cameras, and these tests confirm it: it breaks at -5 (at -4 noise is not too bad but color is going to take some work) and at +4, almost matching the latitude of the UMP G2 (but again with a different distribution).

                          Most positive surprise was the Canon R/RP, which held much better than I expected. And most negative one, FS5 II, which is super noisy already at -2 (but maybe the black tshirt is playing a big role here, on the p4k test it was green) and completely breaks at +3.

                          Edit: of course, I had to do my thing and spend an hour watching all tests and writing down numbers...
                          There are different numbers for each camera because how much noise and of what kind you're ready to put up with is different for each person, and because different tests can yield different results (given Bayer color filter arrays can be different, latitude can change differently in different cameras when you change white balance, for example)

                          DR_tests.jpg
                          Last edited by Samuel H; 09-05-2019, 01:48 AM.
                          My YouTube channel

                          Comment


                          • #14
                            LenProToGo's exposure recovery tests for the BMPCC6K are up:

                            https://www.youtube.com/watch?v=mZUCPIuD4rA
                            http://instagram.com/teddy_dem/
                            http://vimeo.com/luciddreamsfilms

                            Comment


                            • #15
                              Oh crap. That's one stop better than the p4k in the shadows, and half-a-stop to one stop better in the highlights. Almost matching the UMP. Now I want a new camera. And it requires new lenses. Oh why. T____T

                              DR_tests.jpg

                              Edit: I could put the magicbooster inside the EF mount and go back to using mainly my vintage Leica-R set, with a minor 1.1x crop that still leaves my widest (24mm) at an almost-wide-enough 26.4mm. And by shooting 6k BRAW 10:1, the bitrate would be the same as what I'm getting now with 4k cDNG RAW 4:1. Or I could use Q5, which will be in that ballpark too. The softness of BRAW compared with cDNG won't bother me becasue it's 6k, so, overkill to begin with. I can't possibly focus my vintage lenses well enough even at f/4 if I'm shooting 6k, can I? The whole pack would be heavier because of the lenses and I'd need a new -bigger- gimbal too... Or maybe with the longer lenses it will balance fine with the motor on the right... Oh my, why am I pondering all this... XDDD
                              Last edited by Samuel H; 09-14-2019, 01:44 AM.
                              My YouTube channel

                              Comment

                              Working...
                              X