Posted January 10 2013 - 09:11 AM
There seem to be a lot of misconceptions here about 4K, its actual use and usage as well as the technical necessities and the environment around it, such as the color space and the bit depth.
To begin with, "4K MASTERING" as advertised by SONY merely implies that the master used for the Blu-ray encode was crafted in a 4K environment and workflow THE WHOLE TIME, with the resulting encode being a downscale from 4K to HD1080 first and then encoded (for instance, via AVC). Most masters related to 4K so far ARE NOT crafted in a 4K workflow the entire time down the line to the end result, but rather only very partially - meaning that, while the film elements would be sanned at 4K, the actual workflow that follows would be in HD1080, or, for new productions with SFX, mostly in 2K; although this is changing.
Also, 4K DOES HAVE a very relevant and noticeable effect DUE TO THE USE OF OVERSAMPLING on the quality of any (correct) downscaled HD1080 master compared to a native HD1080 transfer. It exhibits and allows for much better color rendition, detail, more accurate sharpness and finer delination in textures as well as has the advantage of dramatically reducing - on a couple of scanners even elliminating - aliasing. This is why we also use 4K in certain projects, whenever we can. LOA was mentioned and it was claimed that 4K workflow (the scan at Fotokem was 8K) had no effect on the resulting master, but rather the well treated HD1080 environment. Let me make this very clear: a native HD1080 scan of LOA would never have resulted on that kind of detail and rendition due to many factors, some of which I already mentioned.
Note that BARAKA and CHITTY CHITTY BANG BANG were NOT MASTERED ENTIRELY in the 4K environment. BARAKA was scanned at 8K resolution at Fotokem, the scan files were then downscaled to HD1080 for further work. CCBB was, as far as I know, scanned 4K and worked down the line in HD1080 as well.
However, the claim that "4K MASTERING" workflow and downscale would result in something significantly different than 4K scanning and downscaling to HD1080 for mastering is, indeed, misleading - and if you look very closely, the suggestion being made in the public relations campaign targets specifically HD1080 and 2K native masters yet mentions with no word existing methods that are pretty similar such as 4K scanning and HD1080 or 2K workflows. There is of course, a reason for this.
As for the "Superbit" comparison: it does not apply. The SUPERBIT DVDs were, indeed, a bit (pardon the pun) of a hustle. The quality was actually not tied to the bitrate alone, and could have been easily achieved by other means (2-Disc sets with the Extras on Disc 2) on "normal" DVDs). So yes, it was a marketing "gag". Also note that there were some "Superbit Editions", which utterly failed the very own program: PANIC ROOM was released with some 30%+ of empty space - resulting in visible artifacts on the film's MPEG-2 encode. Apparently, some extras were planned but striken at the last moment, and someone came up with the (less than bright) idea to issue the DVD under the SUPERBIT program, forgetting or ignoring that the already finished glas master contained a significantly reduced encoded file.
Anyway, with regard to banding and color space and codecs (yes) in that environment:
Blu-ray has a working specification of 4:2:0 Color Space in 8bit per channel at YUV. Does this affect or result necessarily in banding ? No.
For the record: Most masters are 4:2:2 YUV or 4:4:4 RGB HDCAMSR or DPX or, if pre-formatted for encoding, already prepped (with codec) PRORES 4:2:2 YUV mov files. All of the mentioned are delivered in 10bit, mostly HEAD range, few in full that need to be converted.
12bit is only available in Post and / or scanning and also used for Digital Projection for the x-y-z Color Space. It is not used in broadcast and/or digital media such as Blu-ray.
Banding comes actually into play when the color range (8bit at full range 0-255, at HEAD 16-235) is manipulated in such a way that the detail between the tone of each respective color is stretched (mostly with the aim to accomplish a more vibrant picture). This is similar to the stretching of the gradation and/or gamma curve. When one looks into the histogramm after performing this action, one can see that entire parts of the spectrum have been eliminated. This results in "effects" such as solarization, banding, posterization. However, this can happen just as easily (and often does) already in the 10bit range stage during mastering - and the encode, depending on its quality and the way it was handled, put - sometimes significant - artefacts "on top". Here, the 8-bit realm and the 4:2:0 Color Space do not "help" but can "add" to the problem. But, they ARE NOT THE CAUSE.
Does a 10bit realm (0-1023 in Full Range, 64-894 in Head Range) help ? Yes, of course. Does 4:4:4 RGB color space help ? Yes, of course. But if the color timing and mastering is not done correctly, both are just as "helpless". The main issue/objective is to keep the color spectrum intact.
TLEFilms Film Restoration & Preservation Services