[B]Raw comparisons

EDIT 24 Oct 2018: I revised the grade of both videos to facilitate comparison among the raw formats. Images of DNG and BRaw settings now appear at the end of this post.

Amidst the excitement about Blackmagic Raw (a.k.a. "BRaw"), it is hard to find attempts to evaluate the various Raw and BRaw options side by side. Would a thread dedicated to that be useful?

Getting a firm handle on this is probably not simple. The formats will differ in their relative advantages in different shooting conditions, with different subjects, for different purposes (e.g. natural vs. FX pipelines), and towards different delivery formats.

So, in case this is interesting, may I offer this, which shows the same outdoor scene for ten seconds at each [B]Raw option, shot 2k 16:9 windowed at 120fps:


Look at the video description for more technical details.

The same day I also shot this, though clouds passing in front of the sun make it harder to compare [B]Raw settings:


My preliminary impression is that DNG looks nice for these scenes, with BRaw Q0 and BRaw 3:1 in the same league as DNG 3:1 and DNG 4:1. For a lot of work I could shoot BRaw Q5, but I don't see a use for BRaw 5:1, and I'm not crazy about BRaw 8:1 or BRaw 12:1. This impression is very tentative, though.

These scenes were brightly lit. It would be interesting to see low light shooting. And it would be interesting to see how these formats hold up in keying situations such as green screen work.

Here are how the raw settings for DNG and BRay were set for the grading of both these videos.

DNG settings.png (127.1 KiB) Viewed 2302 times

BRaw settings.png (124.63 KiB) Viewed 2302 times

Owen, I’ll take a look on my 5K iMac. Even on the iPhone HD I think I can see some difference between Q0 and Q5 for example. I thought Q0 was a little better than 3:1. But overall none of the codecs look bad. Really good of you to submit these real world comparisons.


Sent from my iPhone using Tapatalk

…Even on the iPhone HD I think I can see some difference between Q0 and Q5 for example.…

Rick, I agree. Same on my iPhone 7+. I would like if there were a Q2 or Q3.

The Q0 and Q5 designations must have something specifically related to the quantization parameters. It certainly seems there’s room for a Q2/3. The Q5 is scary that it can be decent with even greater savings in storage than 12:1 compression. Q0 on the other hand should give the best results when one simply cannot use uncompressed CinemaDNG raw. I’m really looking forward to trying both extremes. For short pieces, Q0 may be affordable, but when doing a wedding shoot over 12 hours, I’m hoping Q5 will look good. If I pickup a 512GB or 1TB Wise Portable SSD, I’ll never have to give a day’s recording media much thought with Q5, assuming the quality is there. BMD has always given us good tools so I’m fairly confident that Q5 will be suitable for events.

OK Rick and whoever else is interested, I revised the two videos in the first post of this thread so that they have grading and raw decoder settings that better facilitate comparison among raw options. The end of the first post shows the decoder settings used for the revised grade.

The lighting in the second video (IMAGINE and the lake) kept changing due to clouds. But the sun stayed away from the clouds during the first (fountain) video. That helped me notice the difference in gamma and saturation produced by DNG and BRaw decoding. In response to this, I graded the three DNG options identically by copying the DNG 1:1 grade forward. Then I graded the BRaw Q0 clip to resemble the DNG 1:1 grade. Finally, I copied the BRaw Q0 grade to all subsequent BRaw clips. This is all reflected in the revised grade now in Vimeo.

For pixel peepers, here are links to download one full resolution frame grab from each clip of the fountain video. Resolve raw debayering was set to full resolution (2k 16:9), and decoding was set as depicted at the end of the first post of this thread. These TIFs are revealing, and for me serve as a data point leading me to use BRaw Q0 for visually demanding work and BRaw Q5 for informal or ENG work.

IMPORTANT: To fetch each picture, you might need to copy each URL and paste it into your browser's address bar. This depends on how your browser handles such URLs.

01 DNG 1-1.tif https://www.dropbox.com/s/tj10nu91lv46eci/01%20DNG%201-1.tif?dl=0
02 DNG 3-1.tifhttps://www.dropbox.com/s/c1waatznbpeojq0/02%20DNG%203-1.tif?dl=0
03 DNG 4-1.tifhttps://www.dropbox.com/s/luuke08hb3rgj2t/03%20DNG%204-1.tif?dl=0
04 BRaw Q0.tifhttps://www.dropbox.com/s/ntbwt6iatm5szqb/04%20BRaw%20Q0.tif?dl=0
05 BRaw Q5.tifhttps://www.dropbox.com/s/kuwc2rxlqny44kx/05%20BRaw%20Q5.tif?dl=0
06 BRaw 3-1.tifhttps://www.dropbox.com/s/nhkxdsn4rsk96fg/06%20BRaw%203-1.tif?dl=0
07 BRaw 5-1.tifhttps://www.dropbox.com/s/kd9wjy6sxrhlxca/07%20BRaw%205-1.tif?dl=0
08 BRaw 8-1.tifhttps://www.dropbox.com/s/40yait6q4irdc4y/08%20BRaw%208-1.tif?dl=0
09 BRaw 12-1.tifhttps://www.dropbox.com/s/dfyyo4089ddtn91/09%20BRaw%2012-1.tif?dl=0

Great effort, Owen! Your conclusion about using Q0 for “visually demanding” is important. The shoot I completed this week using ProRes 444 was certainly visually demanding in my opinion. If your findings are correct, dreams of using BRAW to save large amounts of storage may only be realistic in some situations.

Your conclusion echoes my thoughts about “putting your best foot forward” meaning we should be promoting the best tools we have when the quality of the work is important. However we need to cognizant of other factors, not the least of which is effective use of the available media storage for the recording durations required. So sometimes circumstances may dictate Q5.

Q0 can rival lossless raw for complex scenes, so minimal storage savings. Will be good when I get some personal experience with this presumably for my December shoots, subject to BMD.


Sent from my iPhone using Tapatalk

Q0 is very adaptive, a quite complex picture of foliage against the sky filling the frame and in focus needed about one third more than 3:1. A CU of a person with the background OOF needs about as much as 5:1.

Thanks for doing this, Owen.
I've been meaning to do my own testing to decide which I'd like to use for various situations and this test makes me want to do it even more. The BRAW seems cleaner, but I'm not sure if that noise reduction is making it feel softer than even 4:1 cdng raw. I also see blockiness on edges in all of the images. So I really need to figure this out for myself and decide what I think is best to use. Other friends are doing 12:1.

Uli, that’s why Q0/Q5 appeals to me. It has the capability of using more storage or less storage depending upon the detail of the scene. Although I often do my tests with very detailed content, and that’s a good thing, my client work doesn’t require quite as much detail and that’s where I’m hopeful of having more detail only where it’s needed and a net savings in media storage. Maybe one day, there will be other quantization levels too. Beauty commercials, especially hair, will look better in Q0, while many other long-form subjects may be fine with Q5. “Happy days are here again!”


Sent from my iPhone using Tapatalk

So I really need to figure this out for myself and decide what I think is best to use. Other friends are doing 12:1.

I was comparing data rates in an excel table that i made and when i got it going well I put the resulting calculator online. it includes BMPCC 6k, & 4k, & UMP with all of their braw formats.

braw.info

I made it before buying the 6k, but now i've got it Im doing more tests of my own with the picture quality. and spoilers I love it!.