There exist various HD-SDI device to record 10-Bit 422 video data. 10 bits is useful for shallow gradients especially when expanded (steeper contrast curve) by grading, while 422 gives better detail, that can matter when pixels are big (e.g. when close to a big screen or when digital zoom employed in post). In any case, such recorders tend to compress less than on-board camera systems, or in some cases not at all, improving the quality. But to what extent can the various NLEs cope with this? From my web searches it seems that the answer is “sometimes”. For example some NLEs will accept 10-bit only in their own favourite formats, otherwise they discard two bits, interpreting the footage as 8-bit. One might (naively) have thought the way to be sure was to experiment – but there is plenty of room for confusion when doing experiments, for example Avid’s color correction tool allegedly only displays to 8-bit resolution even when it is importing/processing/exporting at 10-bit. Other “loopholes” may exist, like it seems (if I understand it correctly) that if you AMA or import 10-bit ProRes then Avid only sees 8-bit, implying one needs instead to transcode ProRes->DNxHD externally (e.g. via MPEG StreamClip?) and import that. But even that might not be possible, as one post suggested DNxHD 10-bit encoding could only work from Avid, not external apps. Furthermore, whereas all ProRes formats handle 10-bit, for DNxHD, only formats with an “x” suffix do; the only one I know of is DNxHD 220x. There exist further subtleties/loopholes/pitfalls, hence more research to be done on this… and I’ll tread very carefully…
Benefit of 10-bit 422:
- Blog Author:
- Personally I have experienced cases where more bit depth would have helped, and have seen examples demonstrating the benefit of this and of 422 colour subsampling. but I trawl the web anyhow for other peoples’ opinions.
- www.dvinfo.net/forum/sony-xdcam-ex-cinealta/135394-ex1-8bit-vrs-10bit-uncompressed-test.html (postings from 2008)
- Jim Arthurs:
- I’ve been have great success with my EX1 doing blue and green screen work, both using the native codec on location, and going live into my NLE through HD-SDI 8bit uncompressed when working in the studio. Even so, I have been wondering lately what I might be missing by not taking full advantage of the EX1’s 10bit HD-SDI output. Specifically, would there be noticeable and practical improvements in the quality of the image if I worked in 10bit vrs 8bit for capture? We know as certain fact that the EX1 is 10bit, and those are active useful bits, not just a cheap padding up from 8bit.
- So, I set up a test. I photographed a white card lit for a very shallow gradient of light across the surface, and defocused the camera to help smooth the image. My hope was to force an image that would display banding (you know, like you see in a pale sky), and a very narrow gradient of 45 to 55 IRE should be subtle enough to do it. I recorded this into my NLE as both 10bit uncompressed and 8bit uncompressed clips. I then used Digital Fusion to save out png’s of the resulting clips, being VERY CAREFUL to maintain the clips native colorspace throughout the export. The 10bit PNG is padded up to 16bits on save, and the DF flow was in full 16bit float colorspace. The 8bit PNG was saved out from an 8bit color space project.
- In Photoshop I ramped the levels of each file until that 10 IRE gradient range was blown out to white on one side and black on the other (staying 16bit in Photoshop). The same was done in 8bit space on the 8bit file. I then dropped the 10bit file down to 8bit and split screened it with the 8bit version.
- Interestingly, The 10bit isn’t particularly more subtle than the 8bit one. As a note, the area in the frame blown to white is just imperfection in the lighting. When expanding a 10 IRE recorded gradient out to black and white you’re going to see everything.
- Conclusion (from people on that thread):
- No obvious advantage
- BUT was the processing chain (including the NLE) 10-bit capable?
- Jim Arthurs:
- x
Apple QuickTime Level & Color Issues (Relevant to both DNxHD and ProRes):
- http://michaelkammes.com/editorial/getting-final-cut-pro-projects-andor-media-into-avid-media-composer/
- In addition, you are rendering things out via Quicktime. Quicktime has its own set of issues with codecs and colors. If you Google it, there are a ton of sites specifically discussing ProRes and QT, especially with AE. For example:
Avid:
- The only DNxHD formats for 10-bit are those with an “x” after them, currently I only know of DNxHD 220x
- http://www.avid.com/static/resources/US/documents/dnxhd.pdf
- Avid DNxHD 220x – for 10 bit sources.
- The pdf document distinguishes this from DNxHD 220 (no “x”) in some parts, not in others.
- (The other DNxHD formats are only for 8 bit)
- Avid DNxHD 220x – for 10 bit sources.
- http://michaelkammes.com/editorial/getting-final-cut-pro-projects-andor-media-into-avid-media-composer/
- DNxHD is 8bit, unless the number has an x after the number, i.e. DNx220 vs DNx220x.
- if you want a FAST IMPORT, it has to be 601/709, if thats not important, you can do either 601/709 or RGB.
- http://www.avid.com/static/resources/US/documents/dnxhd.pdf
- community.avid.com/forums/p/95986/549421.aspx (2011)
- Avid supports ProRes via the Quicktime codec so anything it can play, Avid can import or use via AMA.
- I think that only tells part of the story.
- I don’t believe AMA supports Prores above the HD frame size. I also don’t think it supports Prores 4:4:4:4.
- And lastly, all QT source material other than DNxHD is ingested as 8-bit color by MC.
- {That would apply to ProRes then}
- http://community.avid.com/forums/t/61664.aspx?PageIndex=1 (2008 to 2011)
- Here is the response I got from the lead codec developer here at Avid:
- DNxHD does support 10-bit. What you need to do is render to the Avid DNxHD QuickTime codec that is provided on your installation disk or can be downloaded from the Avid website. Render to that and select 220x. The resulting QuickTime file will be “fast imported” into Media Composer with full 10-bit quality.
- However, all other QuickTime imports, other than through Avid’s QuickTime codecs, will come is as 8-bit.
- Also, … you can import 16-bit TIFF files into Media Composer as 10-bit
- Today, QT Animation will import as 8 bit stored in a 10 bit container..
- In summary:
- MC only accepts 10bit video if it comes via either a 16bit TIFF sequence (a file more than twice the size it needs to be with limited meta-data tracking) or ‘fast imported’ DNxHD 10bit wrapped QuickTime.
- All the other 10bit quicktimes and other 10bit files (DPX, CIN etc) can’t be imported into MC with out ending up with 8bit video…
- …when bringing in 10-bit video … the CC tool is showing … an 8-bit scale. According to this thread (http://community.avid.com/forums/p/68078/380976.aspx), Avid always grades internally to 12-bits.
- Example 10-bit-preserving process in AVid:
- I tested this and was successful. Here’s the steps I took.
- created 1080p24 project in AE. Created gradient from 1023-924 RGB left to right in 16bpc mode. I was able to clearly view the individual steps in the gradient incrementing by 1.
- rendered using DNxHD RGB 1080p24 175X. Imported into MC as 709. Went into CC mode. I was able to see banding and steps shown in the eyedropper were much wider and were 8-bit steps from 235-217.
- made two RGB exports, both QT references, one with CC (no changes) and one without.
- imported these two files into AE and was able to see individual steps in the gradient incrementing by 1, however a few steps were missing, obviously due to compressing to 709 and back to RGB.
- So it appears that MC5 is able to maintain 10-bit footage in and out.
- {However, the Color Correction tool only displays at 8-bit:}
- While in CC mode I droped the saturation to -100 and switched to full screen mode which made the banding much easier to see. Then I switched between Yellow, Green/Yellow, Green and Green(10) quality modes. This clearly made a difference but even in Green(10) the banding was there. The Y waveform showed the stair steps nicely, especially in full screen mode. This also showed that in Green vs. Green(10) a few extra steps were added back in but you were nowhere close to seeing the approx. 100 steps that were there.
- I had heard there was a way to switch to 10-bit mode in CC, but no one has revealed how. This would seem to limit the ability of CC mode when working with 10-bit video. IMHO you really need to see 10-bit values to know how you are effecting 10-bit video.
- I tested this and was successful. Here’s the steps I took.
- Here is the response I got from the lead codec developer here at Avid:
- {Re the Atomos Ninja & Samurai}
- I just got my ninja and to my surprise MC5.5 will not AMA link to 1080 clips!!! The linked clips are just a mess of streached images and scanlines! It will however work with 720 clips just fine. (Sept 2011)
- As far as I could see after a quick preliminary research no timecode (or at least no way to set and lock timecode).
- Samurai (will have) Timecode and genlock amongst other things.
- http://community.avid.com/forums/p/96816/554911.aspx (2011)
- (repaired text) …it appears that the DNxHD Quicktime codec does not function properly outside of Composer so there is no way to make real 10 bit DNxHD X Quicktime files elsewhere.
Apple (ProRes & FCP):
- All ProRes formats handle 10-bit:
- http://images.apple.com/support/finalcutpro/docs/Apple-ProRes-White-Paper-July-2009.pdf
- (All ProRes formats, even the “draft quality” ProRes 422 Proxy format, support 10-bit)
- http://michaelkammes.com/editorial/getting-final-cut-pro-projects-andor-media-into-avid-media-composer/ (2010)
- ProRes is always 10bit. From ProRes LT to ProRes 4444, the sample size is always 10bit.
- The analogy of DNx145 to ProRes 422 was based on data rate, as well as perception of likeness and quality. The debate between which is better will never end. However, these 2 are generally considered to be close enough to be on par with one another.
- http://images.apple.com/support/finalcutpro/docs/Apple-ProRes-White-Paper-July-2009.pdf
- FCforums.creativecow.net/archivethread/8/292652 , a discussion dating back to 2003 (possibly obsolete (?) but even if so then still useful as a question-prompter)
- << if you capture in 10 bit and do cuts only the material will stay 10 bit; but if you apply any kind of effect, it gets knnocked down to 8 bit. FCP and all NLE’s (I think) are 8 bit. Theres hope for the future. >>
- << Some NLE systems are 10-Bit through and through for the most part those systems are pretty expensive. >>
- << Combustion, Digital Fusion, and Shake are able to work with 32 bit per channel files. >>
- <<
- Every component of FCP that does not require rendering already supports any bit depth you throw at it.
- You can capture in 8, 10 or 16 bit (10 or 16 bit require Kona, Cinewave or Decklink); you can capture standard or Hi definition.
- You can edit in 8, 10, or 16 bit; you can add real time 8 or 10 bit effects (no hardware supports 16 bit in real time afaik). You can add titles in real time in 8 or 10 bit.
- You can export in 8, 10 or 16 bit an edited timeline.
- As long as you do not render you stay in the native bit depth of the capture.
- It’s always been that way. Only the render “engine” is limited to 8 bit 4:2:2 colorspace.
- xx
- x
Migrating Media (and Projects) from FCP to Avid:
- http://michaelkammes.com/editorial/getting-final-cut-pro-projects-andor-media-into-avid-media-composer/
- Very nice article with schematic diagrams explaining cost-free methods and why, despite those, Automatic Duck is worthwhile.