Archive for the ‘10-bit depth’ Category

Video with 10-bit Channels: Update: Sony Vegas (12) Handles it OK

Saturday, August 3rd, 2013

In my post http://blog.davidesp.com/archives/573 I reported that when I imported 10-bit footage to Sony Vegas 10, even in 32-bit mode, it still appeared to act as if only using 8-bit footage.

Since then, I re-did the test for Sony Vegas 12.  I applied a 10-bit recording from a CineDeck, in QuickTime (.mov) containing Cineform 10-bit, taken from SDI output of a Sony EX3 camera.  In that case, Vegas 12, with 32-bit (Video Levels) mode, did correctly make use of the 10 bits, as verified by bands on the vectorscope on dark areas of high-alpha-increased underexposed footage.

One thing I noted was the tendency of 8-bit to round-down.  Consequently, when switching project to 32-bit mode (hence 10-bits used), the dark levels visibly (and on vectorscope) brightened.  I guess better compatibility would in that case be obtained (in that NLE or by pre-processing the footage) by first subtracting “4” from the 10-bit levels, since the “2^4” represented by those extra 2 bits would in 8-bit format not exist, hence effectively would be rounded-down.  Or maybe the offset should be 1 less than this, i.e. subtract 3 (depends how the rounding gets done).

Workarounds would be to either pre-process the 10-bit footage (to subtract the offset) or else, less conveniently, apply a levels effect to increase the input minimum level by that amount.  But would be awkward and may or may-not work, depending on Vegas 12 nuances.  Something to be tested!

Adobe Premiere (CS6): Maximum Bit Depth & Rendering

Monday, July 8th, 2013

I knew basically what these were about:

  • Max bid depth to make use of all the information in a more-than-8-bit video file, such as a 10-bit recording.
  • Max render quality to employ higher-quality but slower scaling algorithms – only relevant when scaling of course.

However, there are options to set them in the Sequence and also in the Render.  Like others, I wanted to know firmly (not just by guesswork) how/when to use these.

The answers appear to be:

  • Their values in the Sequence settings only affect the preview, not the render.
  • Their values in the Render dialog override their values in the Sequence.

Hence:

  • In the Sequence, one would tend to leave them disabled, other than temporarily for quality check or comparison.
  • In Render dialog, one might tend to have them initially disabled, for render-speed, then enable them later on for final quality-check and production.

(more…)

Sony EX3 Noise & Bits-Resolution

Monday, November 5th, 2012

It looks to me like it is worth recording from a Sony EX3 in 10-bit when there will be subsequent Neat Video -type temporal denoising in post.

I tried a quick-and-dirty experiment, confirming that, despite the relatively high noise of the Sony EX3 (as compared to mainstream broadcast video cameras), high bitrate 10-bit 4:2:2 recording offers a greater potential than 8-bit 4:2:0  when the Neat Video type of temporal denoising (motion-compensated, I think) is applied in post.

I have yet to dig-down into this, e.g. to see how it would be affected by dropping down to “8-bit but still high-bandwidth” recording, hence I can only conclude that the combination of high bandwidth, 10-bit and 4:2:2 is beneficial.

The experiment:

  • Make an extremely low-light recording on the EX3, in 1080 50i mode.
  • Import it to a SD resolution project in Adobe Premiere.
    • No “scale to project size”, hence pixel-for-pixel, with the HD clip therefore appearing to be “zoomed”.
  • Compare the original to a copy that had the following affects applied:
    • Fast Color Corrector
      • Input-range (0, 1.6, 114), to brighten the (deliberately) under-exposed image.
      • Increase Saturation to 200.
    • Remove Noise (Neat Video)
      • Temporal noise reduction only, radius 4 (frames).

Make the comparison via Preview:

  • Set resolution to 100%, image size to Full.
  • Render the result, i.e. so timeline had green lines not red.
  • Compare by eye.
    • The unprocessed 8-bit (XDCAM-EX) and 10-bit (Cineform-High) recordings appeared identical i.e. very noisy.
    • The denoised 8-bit looked slightly better but the denoised 10-bit looked very significantly better, indeed just about usable.

Sony EX3 Noise & Bits-Resolution & Green-Screen

Sunday, November 4th, 2012

It has been said ( I believe by Alister Chapman ) that there are only marginal benefits from recording XDCAM-EX to more than 8 bits, due to the relatively high noise of this camera, as compared to more typical broadcast cameras.

In my experience, while it was a wonderful step-up from my Z1, certainly it’s recordings are noisier than I’d like, leading me to pretty-process certain footage (using Neat Video denoising plugin to my NLE).  And as a recent project with reasonably well-lit green-screen illustrated, it’s noise in shadows can be a particularly nuisance (much time in post experimenting to work around this).

So I wondered:

  • Even if marginal, to what extent is 10-bit beneficial to EX3 recording?
  • For the EX3, when recording 10-bit, it is also 4:2:2, surely a benefit to chroma keying and resizing (reframing, stabilising/deshaking/tracking).
  • Could the benefit depend on editing workflow?  For example:
    • What if subsequently de-noising (like I mentioned)?
    • Some NLE’s do bits-dithering, hiding the quantisation/banding that would otherwise be apparent from having only 8 bits.

I need to do my own experiments, but for now, here (below) are some results from web-searching…

(more…)

EX3 SDI Output

Sunday, November 4th, 2012

 Worked, but config was not as straightforward as I first (naively) assumed:

The big “Gotcha”:

  • Must first Disable the iLink (IEEE 1394, small FireWire) interface. Otherwise SDI won’t work at all.  I guess EX3’s SDI & iLink might share some circuitry?
    • In EX3 Menu, OTHERS category (last i.e. final category):
      • [i.Link I/O] :Disable.

Then, in EX3 Menu, VIDEO SET category (3rd category), then:

  • [YPbPr/SDI Out Select] : HD
  • [YPbPr/SDI Out Display] : Off

This worked fine in practice.

FurtherTips:

  • Under EX3 OTHERS Menu-Category:
    • With EX3 [Country] = [NTSC Area]:
      • “HQ 1080/60i” gives [1080 interlaced 59.94fps 4:2:2 YUV10].
      • “HQ 1080/30p” gives [1080 progressive 29.97fps 4:2:2 YUV10 ].
      • “HQ 1080/24p” gives ??? (Cinedeck accepted it only at 30 fps)
    • With (correspondingly) [PAL Area]:
      • “HQ 1080/50i” gives [1080 interlaces 50fps 4:2:2 YUV10]

(more…)

Cineform and Alpha Channels

Friday, August 31st, 2012

The full (paid) version of GoPro-Cineform Neo (as I have) does support alpha channels.

(A colleague initially thought otherwise – but that impression turned out to be based on info from old forum threads)

(more…)

Video with 10-Bit Channels

Thursday, July 5th, 2012

If I had a 10-bit video recording such as from the PIX 240, would I know what to do with it, in order to make full use of the 10-bit information?  This question is important, because it cannot be assumed that this is simply a case of inputting it into any arbitrary nonlinear editing system (NLE) – not all NLEs preserve the extra information – and even for those that do, the workflow and configuration must be set up appropriately.  And even having got that right, how can we verify all is working as expected?  Can the NLE’s own effects and waveform monitors etc. be trusted to preserve the extra bits?

Having discovered some sample 10-bit footage at http://www.sounddevices.com/products/pix240/sample_files/ (as reported at http://blog.davidesp.com/archives/570), I was prompted to do some experiments in a few NLEs.   I based the experiments on the following two DNxHD files, as recorded by a PIX 240, both 1920x1080p29.97 and around half a minute in duration.

  • dnxhd220mb_8_29.97.mov = 8-bit
  • dnxhd220mb_10_29.97.mov = 10-bit

The comparison was based on an area of sky at the top-left of frame (in each case), with its (limited) levels-range mapped to full video range, so as to make 8-bit quantization-banding appear.

Conclusions (as far as I can tell from experiments):

  • Adobe Premiere:
    • Propagates the 10-bit footage’s information, achieving better image quality than for the 8-bit footage.
      • However this only happens when correctly configured and then only for certain effects.
    • The Fast Color Corrector levels-mapping appears to introduce some kind of dithering.
      • Hence while the expected banding is visible for 8-bit footage, it is slightly “blurred” on the Waveform Monitor and the resulting image looks more ragged than banded.
      • Nevertheless, the 10-bit footage through this same process has no such banding at all, and resulting the image looks obviously better.
      • None of the cases at http://blogs.adobe.com/VideoRoad/2010/06/understanding_color_processing.html apply here since no blur effect was used.
    • The result of Fast Color Corrector levels-mapping on 10-bit footage result also looks slightly brighter than that on 8-bit footage – presumably a mapping-inconsistency in Premiere?
    • Some other non-obvious pitfalls exist when making such comparisons:
  • Sony Vegas 10
    • Ignores the extra information in the 10-bit footage, evem for Project Settings of 32-bit.
  • Avid Symphony 6
    • AMA appears to truncate to 8-bit, at least it seems so based on what appears in Avid’s Waveform monitor.
    • Import of the given DNxHD-220 to Avid-Import-DNxHD-220 appears to give same result.
    • I assume I am missing something here, some knowledge and/or step and/or monitoring method…

The configurations I used within each application:

  • Sony Vegas 10:
    • Project Properties
      • 1920x1080p29.97. Not automatically readable by Vegas from the DNxHD format.
      • Pixel Format: 32-bit floating point (video levels)
    • Waveform Monitor via: Video Scopes > Waveform
    • Sky-range mapped to full range via: Sony Levels FX
  • Adobe Premiere CS 5.5:
    • Computer had a non Mercury Engine compatible GPU hence software-only graphics / effects processing.
    • Waveform Monitor via: Reference Monitor > YC Waveform
    • Sky-range mapped to full range via: Fast Color Corrector > Input Levels
      • (Prior to that tried various “Levels” effects but they did not work properly in this context)
    • Sequence Setting: Maximum Bit Depth (else levels-resolution was truncated to 8-bit)
  • Avid Symphony (hence presumably also Media Composer) 6

Training: Den Lennie’s “Music Video” Experience

Thursday, October 27th, 2011

I attended, working on one of the camera units.  Had a great time, learnt lots, at all sorts of levels.  Even how to make good use of the Movie Slate application on my iPhone!  Link: http://www.fstopacademy.com/

NLE Handling of 10-Bit Recordings

Friday, September 23rd, 2011

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…

(more…)

10-bit AVC/H.264 Encoding with 4:2:2 for Broadcast

Sunday, September 4th, 2011

Explained in http://x264.nl/x264/10bit_01-ateme_pierre_larbier_422_10-bit.pdf and further documents at http://x264.nl/.  Would like to know what applications recognize this format though, and how would it be incorporated in a workflow involving standard NLEs (e.g. Avid MC or Sony Vegas).  Would it have to be transcoded to e.g. DNxHD first, and if so then how (what app etc.)?