FS7: Is the HDMI Output REALLY Uncompressed?

basspig said:
AtticusLake, if XAVC is just h.264, then why is it that CS6 can play h.264 UHD with no frames dropped, and gives a 'general error' when importing XAVC?
Click to expand...

Because XAVC is H.264 in an MXF container, and CS6 can't parse MXF. The H.264 inside is perfectly standard H.264 -- I know, because I use ffmpeg to re-wrap it in an MP4 container WITHOUT RECOMPRESSING, and it loads into CS6 just fine. It's not the H.264 that's the problem -- it's the container. (Note that I'm using an X70, not an FS7, which I don't (yet) have. But the principle is the same.)

basspig said:
And on an NLE that accepts XAVC, or even VLC Media Player, the h.264 plays smoothly, but the XAVC plays a few frames and freezes? I know it's a flavor of h.264,
Click to expand...

That is true. XAVC is both a wacky container AND a profile of H.264 -- in other words, a particular set of H.264 settings (of which there are many). So the H.264 in XAVC is more complex than the H.264 in, say, AVCHD. (Greater bit rate, bit depth, colour sampling, image size, more complex encoding methods, ...)

If your software is choking whole playing the clip, then my guess is it's just too much stuff to process. If its UHD and your computer is older, and particularly if you don't have a graphics card which the NLE can use for acceleration, then I have to say this isn't a big surprise.

basspig said:
but there must be several levels of complexity and XAVC is probably the very top, requiring dedicated ASICs to properly decode it in realtime. Obviously, the FS7 has them, because it can play back it's own media smoothly, but not my NLE.
Click to expand...

Nothing REQUIRES dedicated ASICs -- any video can be decoded in software 100%. The issue is speed. To do it in realtime requires sufficient processing power, and hardware support (whether an ASIC or FPGA or DSP or whatever) can help with that. Certainly more complex codecs require more processing power than simpler ones.

But an NLE SHOULD be able to decode the video and put it on your timeline correctly, if it decodes it at all, and then export it to the video format of your choice completely correctly (though it might take a long time). The only question in terms of processing power is whether it can play back at full speed. If it can't, that will of course make editing quite a pain.
 
Are you able to view the HDMI in any other way directly?

Could it be settings on the shogun? If you have a very high compression (low quality) then it follows that the prores codec would probably compress in a similar way to xavc?
Was there a lot of motion in that first scene or a static shot?

Can't speak for that combination but the fs700+odyssey or even A7s+odyssey is nothing like that, something is up somewhere...

cheers
Paul
 
The postie has just delivered my XQD media so now I've done an HD back to back with my FS100 and FS7 and neither look as bad as the Basspig examples under 300% zoom.

We've no idea how busy Basspig's image was but in my test clips I could only see a bit of chroma noise in the darker tones of the image and that was after some serious contrast was put through it.

It wouldn't at all surprise me if Sony sent compressed garbage out of the HDMI as they have a long history of pulling strokes like this to segment their own market but in this case I think there's a more prosaic answer to be found.

So far the FS7 has exceed my expectations apart from a few quirks, anyone know how to stop the shutter speed changing on it's own? I haven't worked what actions I'm doing causes the shutter to change but it certainly wasn't changing the shutter speed!!! FFS, presumably this is the "shutter bug" I'll need to read up on?
 
I managed to rewrap XAVC to MP4 (took about 20 seconds to rewrap a 14 second clip because the only online example I could find recodes the audio to AAC--I'd like to find out how to rewrap and leave both the video AND audio intact). CS6 does accept the file. And it does play, though in 14 seconds, it dropped 183 frames. But it's better than DaVinci, as I get an update every fraction of a second or so. But the command line is complicated and my attempts to delete the portion that merges channels and recodes to AAC failed:

K:\FS7 Footage\XDROOT\Clip>K:\ffmpeg-20150720-git-9ebe041-win64-static\bin\ffmpe
g -loglevel error -y -i "302_0003.MXF" -map 0:0 -filter_complex -vcodec copy -me
tadata "date=2014-11-17 00:24:47" "302_0003.mp4"
[AVFilterGraph @ 0000000002fbd240] No such filter: '-vcodec'
Error initializing complex filters.
Invalid argument





I did some more testing today, just a static shot of a boulder in the woods.
Prores HQ is much improvement over 422.. can barely see any macroblocks.
DNxHR medium quality has some visible blocks. The DNxHR HQ setting has the least of the three, and I suspect a bit more detail than the Prores HQ.

I do note that the contrast is must poorer on the DNxHR shots. Range is from 7-100 IRE, whereas the Prores shots stretch from 0-0109 IRE and look much more saturated.

So Prores HQ is probably the best option, given the contrast and color information. At normal viewing conditions, no artifacts to be seen. I had to look at 400% magnification to spot macroblocks, but they were well blended.

So I guess the Shogun is not dead yet. But they better hurry up and add DCI 4K capability to the HDMI input, or I may be tempted to buy a 7Q+.
 
basspig said:
I do note that the contrast is must poorer on the DNxHR shots. Range is from 7-100 IRE, whereas the Prores shots stretch from 0-0109 IRE and look much more saturated.

So Prores HQ is probably the best option, given the contrast and color information. At normal viewing conditions, no artifacts to be seen. I had to look at 400% magnification to spot macroblocks, but they were well blended.

So I guess the Shogun is not dead yet. But they better hurry up and add DCI 4K capability to the HDMI input, or I may be tempted to buy a 7Q+.
Click to expand...
Sounds like the DNxHR is Legalizing the video while the ProRes is leaving it in full Extended range. I'm not surprised as all of Avid is always hearted toward broadcast application. Big question is whether the signal is compressed or truncated. If it is compressed then it can be stretched back out, but if truncated then that is info lost and never to be seen again.

The Odyssey7Q+ allows one to adjust for Extended or Legalized signal, both on input as well as for setting LUTs. I do not believe this is true at all for the Shogun. We also record True 4K (aka 4K DCI) and have done so for quite some time.
 
basspig said:
I managed to rewrap XAVC to MP4
Click to expand...

Good job!

basspig said:
... my attempts to delete the portion that merges channels and recodes to AAC failed:

Code: 
[FONT=fixedsys]K:\FS7 Footage\XDROOT\Clip>K:\ffmpeg-20150720-git-9ebe041-win64-static\bin\ffmpe[/FONT]
[FONT=fixedsys]g -loglevel error -y -i "302_0003.MXF" -map 0:0 -filter_complex -vcodec copy -me[/FONT]
[FONT=fixedsys]tadata "date=2014-11-17 00:24:47" "302_0003.mp4"[/FONT]
Click to expand...


If you want to delete the audio finagling part, then get rid of "-filter_complex". The argument following it is the actual specification of the filter; you have "-vcodec", which isn't valid.

However, I think you need the audio finagling, because an MP4 file shouldn't contain the PCM audio that the camera produces; AAC is more likely to be compatible. Also, merging the two mono tracks into one stereo one should come in handy -- that's actually what the "-filter_complex "[0:1] [0:2] amerge" " bit does.

This is the command I use:

Code: 
ffmpeg -loglevel error -y -i "Clip0001.MXF" -map 0:0 \
    -filter_complex "[0:1] [0:2] amerge" -acodec aac -b:a 256k -strict -2 \
    -vcodec copy  -metadata "creation_time=2014-11-17 00:24:47" \
    "Clip0001.mp4"

I've tried using "-timecode" to set the timecode, but no joy. :-(

HTH...
 
Mitch, I think you're right about the legalized levels of DNxHR. I spent some time looking at the highlight parts of both CODEC's images and noticed that the Prores highlights are flat, lacking in detail, while the DNxHR contain some detail up there. So either the Shogun is doing something different when it prepares the data for Avid's compression, or the compression algorithm itself scales back the highlight curve rather than clipping it.



AtticusLake, the only problem is I don't want anything transcoded at all. I want to keep the audio quality undamaged and I don't want to incur the performance penalty of recoding it. If I could simply rewrap it with no changes to the compression, that would be the best option. If it's not possible to do that in MP4 container, perhaps there's another container that can?

Audio track assignment is no problem in CS6--I simply set the audio track preferences before importing the clip.

The other thing I would have to work out is how to make this work transparently, or batch process the files as they copy over from the memory stick to the SSD drive. I would need a *.MXF type wildcard to copy all of the video, run it through FFMPEG, and write the results to the SSD drive.


CS6 is amazingly efficient. Playback of UHD is smooth and maybe I drop one frame in ten minutes of play. The converted XAVC to MP4, not so good. Not nearly as smooth as h.264 downloaded from Youtube 2160p footage. It takes a lot more horsepower to decode this footage. CS6 dropped 180 fames out of 336 total. DaVinci Resolve drops nearly all of them, displaying few enough to count on fingers of one hand.

I continue to research my upgrade NLE plans. Someone on a tech forum made a good argument for using a dual CPU motherboard. Dual CPUs have twice the memory channels (twice the bandwidth), so two 6-core Xeons could possibly outperform a single 12-core Xeon. Options get very narrow with Dual CPU motherboards though. Fewer PCI slots and fewer brands competing in this arena. But dual 6-core today, upgradeable to 18-cores in a year or two, as prices fall. Decisions, decisions..
 
Today's Test results:

There IS a BIG difference between Prores 422 and Prores HQ!

Here is HQ compared with XAVC recorded in camera:

XAVC:
https://www.dropbox.com/s/4ptatt1dna93ykg/XAVC test.PNG?dl=0




Prores HQ
https://www.dropbox.com/s/cz1at5h3r5lmtsp/Prores HQ test.PNG?dl=0


Some of the fine detail is lost in XAVC, but not a lot. Very hard to see difference with naked eye on 4K monitor, even at 12" viewing distance. But the loss is there. DNxHR (at it's highest setting), is similar quality to Prores HQ.

My observations:

Slight 'false contouring' noted in XAVC version. Edges look more accurately defined in Prores HQ version.
On 4K monitor at 1:1 pixel mapping, there is a slight, perceiveable softening of the texture of the rock's surface in this frame, when XAVC layer is enabled. Prores HQ layer is slightly sharper to the naked eye. You have to be close to the monitor to notice it though.
 
I've just been examining some footage of a swaying tree against a blue sky and cloudy background.

I was surprised to see the XAVC-I codec massively favour the moving tree as it has massively macro blocked the sky and clouds. It was impossible to grade any contrast back into the sky as it just turned to blocky noise territory.

Shooting just sky and clouds moving in the wind showed only minor macro blocking and graded fine. It looks like XAVC-I favours moving objects much more than I was expecting.

I was shooting Slog3 in CineEI mode.
 
There's little question that for complex scenes with some expanses of gradients and flat color, the highest level of Prores that your external device can record is beneficial. Just hope you have gargantuous storage space!
 
I'm certainly going to get a ProRes recorder after discovering XAVC's dirty secret the only question now is which one...

Storage is cheap.
 
You must log in or register to reply here.
Back
Top