BM RAW workflow

Hi guys can anyone make a comparison on what the BM Raw Q0, 3:1 etc represent? For example does the BM Raw Q5 5:1 equal to ProRes HQ? How does everything compare?

Thanks

Ivan you can download Nate Porter’s clips and judge for yourself. Or read the posts in the other BRAW threads that are active.


Sent from my iPhone using Tapatalk

I guess it depends on what you're trying to compare? File sizes? Image quality, DR? Rendering speed?

Was curious to see the differences between compressed BRAW and ProResHQ myself, so I pulled some aggressive keys on clips of windblown trees in motion against the sky. I had to push it hard and colorize the sky to really show the difference between the two.

Below is just a quick comparison between ProResHQ and BRAW 12:1 using a color key from the same framing shot in both codecs on Ursa Mini Pro. The screengrabs are from Resolve, of 400% zoomed clips in highlight mode with identical keys on both shots:

ProResHQ v BRAW 12 to 1.jpg (255.11 KiB) Viewed 1265 times

To be very clear, I would never grade a shot this way. I'm only pushing the clips hard and displaying in highlight mode to reveal differences. However, displayed this way its quite easy to see how the color data suffers with the 4:2:2 sampling of ProResHQ resulting in visible macroblocks.
Compressed raw debayered to 4:4:4 necessarily has cleaner edges than the subsampled chroma of 4:2:2, which from the numbers alone anyone could easily conclude, but its always better to have some empirical testing to confirm.

The conclusion for me is I feel quite confident that 12:1 BRAW is quite fine to shoot with despite the 12:1 compression. I used to shoot ProResHQ without worry, and now I'm going to go ahead and shoot in BRAW 12:1 without worry.

Just so I'm not only showing ugly crazy grades, here's a still from an interview setup yesterday shot 4.6K BRAW 12:1

Screen Shot 2018-09-18 at 10.01.44 PM.jpg (119.22 KiB) Viewed 1265 times

Was curious to see the differences between compressed BRAW and ProResHQ myself, so I pulled some aggressive keys on clips of windblown trees in motion against the sky. I had to push it hard and colorize the sky to really show the difference between the two.]

Nice test! 1:3 of the storage, 12bit vs 10bit, and less compression artifacts...who can say no. My plan was to shot compressed DNG with SSD in longer production for 12bit, but now we have BRAW so we only need 500gb space to do the same as 2TB. This is a reason itself to buy the new pocket camera. It seems like the DNG still got advantage when it comes to clipping, but in minor detail. BRAW even run smooth, multi camera, on a laptop… after grading 

However, displayed this way its quite easy to see how the color data suffers with the 4:2:2 sampling of ProResHQ resulting in visible macroblocks.
Compressed raw debayered to 4:4:4 necessarily has cleaner edges than the subsampled chroma of 4:2:2, which from the numbers alone anyone could easily conclude, but its always better to have some empirical testing to confirm.

Pretty nice example! These artifacts are from DCT macroblocks, not directly subsampling related though. Raw Bayer data is essentially 2:1:1 hardware downsampling in RGB space and if one wanted, 4:2:2 can always be upsampled to 4:4:4 with better results because there is just more data. In practice there are other things at play ofcourse, like actual compression of 422 data and related artifacts, as seen here, and most 422 stuff is already subsampled from 444.

Thank you very much, Jamie. The pictures are very telling. Sure we don’t view video at 400% but I do think it matters to see those extremes which would even be apparent at 200%. We are working with photosites that can be a few microns, perhaps a hundred times smaller than the finest brush stroke in painting. And so much beautiful art has been accomplished playing with our perception of colour and detail. When you pixel peek at traditional art, there’s no beauty left except perhaps admiration at the artist’s skill that transforms those strokes into his/her vision. The artist has one advantage and that is the mixing of colour paints and mediums. We are the ones working with crude tools now, quantized bits and electron volts. Of course we want to make the most of the paucity of the tools we have.

For my purposes, I’m staying with ProRes 444 for its safer handling of moiré on the URSA Mini 4.6 K camera. But on the BMPCC4K, at this point, BRAW 12:1 or Q5 for most shooting purposes and Q0 when you have a reason or desire or capacity to put your best foot forward. The “reason” could be client based or likely subject matter but the level of required scene detail is the driver. This is such a stunning turnabout from everything we learned about CinemaDNG. The level of quality provided by 12:1 and Q5 is obtainable without feeling guilty that you’re cheating on what matters most.

ProRes 422 is a cheat and it seemed good enough until you start to look under the covers. I rarely used it and once I have the BRAW option, the default codec shifts to it. The exception being if there’s a fear of uncontrollable moiré where ProRes 444 in camera is the safer option such as at a wedding where the wedding party isn’t going to have a full dress rehearsal like you might have in a theatrical play or even a costume shoot in a controlled narrative. When in doubt, trust ProRes 444 and then expect to work a little harder on grading in post. Otherwise, BRAW.

Rick, are you seeing a moire difference between ProRes4444 and BRAW?

cDNG certainly has more moire and aliasing, but I’m not seeing a moire difference when between BRAW and the 4:4:4 flavors of ProRes

I believe I see a difference especially on the fold of the grey top around the stomach area. That’s the worst affected. It’s a rainbow on CDNG, maybe a quarter as much on BRAW to the point it may not be an issue. And in ProRes it’s almost eliminated so many might not see any.


Sent from my iPhone using Tapatalk

From other shared samples i found that BRAW don't produce Cross Hatching effect, so they may also produce less moire.