I finally had time to do my own testing. I did this with a DNG raw frame from Evans DNG exposure test (i did ask for the frame Marco used but no response).
I tried to follow Marco's testing patterns except i used Resolve instead of Lightroom (i personally don't like the raw conversion from Lightroom and use Capture One 99% of the time for stills anyway). Here's what i did:
• Used frame one of Evan's exposure test since the outside "appears" clipped.
• Exported a jpg of this 'as is' for reference.
• Exported an 8Bit H264 version of the frame which came out at just over 37Mbps (less than Marco's 50Mbps).
• Lowered the exposure on the DNG raw file and did some quick curve and Lift/Gamma/Gain adjustments to represent something i might aim for (not looking to create an HDR type image since i'm not a fan). Created a jpg from this.
• Attempted to do the same with the 8bit H264 37Mbps version. Exported JPG.
• Then i reset the exposure/etc (Base Mem) on the DNG to start from scratch and applied the BMDFilm log preset. Exported as 8Bit H264 and this time the frame was at 12Mbps. MUCH less than 50Mbps which should be a disadvantage here.
• Attempted to do a similar recovery as i did for the DNG raw. Exported JPG.
So i have 4 JPGs to post. The reason i'm doing this is because i think Marco's demonstration was EXTREME and i believe he made a mistake by not using a log curve as well and in his follow up post by using an inverted "s curve" instead of a proper log-type curve (like so -
http://mikemost.com/?p=251). I'm VERY aware of the advantages of 10bit log and 12bit raw but i still think GOOD things are possible with a nice 8bit log file (not on par with 10bit or 12bit raw, but still..).
DSLRs are currently the worst examples of 8bit, yes. But all 8bit != DSLRs. Given the choice, of course 12bit raw or 10bit log is highly preferable to 8bit, and part of the 'success' of my example is the very nice way BMCC handles it's image/dynamic range and the strength of the resolve output/compression. I'm just posting because i wanted to test this for myself (as we always should if we have time/resources) and to provide perhaps a bit of 'balance'.
Click on the JPGs for the large versions if desired:
The DNG raw with no adjustments in Resolve:
The 8bit NO LOG H264 37Mbps recovery attempt (no denying it's awful):
The DNG raw "recovery" in Resolve (to my preference what i would use as a starting point for grading):
The 8Bit Log (BMDFilm) H264 12Mbps attempt (not as nice as the raw but respectable and useable IMHO):
Hope some find that useful. I'm not a DSLR "apologist" and i don't think they're currently capable of this demonstration either, but i DO think 8bit Log didn't get a completely 'fair' run by Marco either. But i understand his enthusiasm for the BMCC, 12bit raw, and 10bit log since they ARE fantastic to work with in post. What i took away from this, is that a GOOD camera with a solid (uncompressed/un-lineskipped etc) 8bit log output is pretty good. Not 12bit/10bit log good, but not necessarily miles away either.