moovie bitrate

Started by escho, January 30, 2013, 11:50:49 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

escho

Hallo

Please have a look at this here:

picture 1: CBR 1x

edgar@linux-pusc:~/testordner/test4/exif> exiftool MVI_5869.MOV | grep "Y Cb Cr Sub Sampling"
Y Cb Cr Sub Sampling            : YCbCr4:2:2 (2 1)

edgar@linux-pusc:~/testordner/test4/exif> exiftool MVI_5869.MOV | grep "Avg Bitrate"
Avg Bitrate                     : 42.3 Mbps


picture 2: CBR 3x

edgar@linux-pusc:~/testordner/test4/exif> exiftool MVI_5870.MOV | grep "Y Cb Cr Sub Sampling"
Y Cb Cr Sub Sampling            : YCbCr4:2:2 (2 1)

edgar@linux-pusc:~/testordner/test4/exif> exiftool MVI_5870.MOV | grep "Avg Bitrate"
Avg Bitrate                     : 90.5 Mbps


I have 2 questions:

1.
I thougt, subsampling of fullhd is YUV420. But in the exif-tag I can see YUV422. What is correct, 420 or 422?

2.
A little clip with CBR 1x shows a bitrate of 42,3 Mbps.
The same situation recorded with CBR 3x shows a bitrate of 90,5 Mbps. Why is this only 90,5 Mbps and not 42,3 Mbps x 3 = 126,9 Mbps?

Edgar
https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

a1ex

1. 420

2. "up to 3x" means any value between 0x and 3x.

1%

Your tool must be showing incorrect tags. All my videos show 4:2:0 sub-sampling.


escho

Yes, looks like exiftool shows a wrong value here. Decoding the mov-file to pngs with ffmpeg shows correct 420.

I found in the usermaual, that the CBR works by adjusting QScale. Thats the reason for the unexpected bitrate, I think.
But I also found this: "If buffer usage gets too high, ML will pause all CPU-intensive graphics". Does that mean, that a higher bitrate gets higher buffer usage? If that is the case, can this not be used for avoiding unexpected stopping of the record, if bitrate gets too high?

Edgar
https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

1%

Canon controls rate with the qp of each frame. It predicts what qp to set next via the numbers 'cbr" scales. qscale is just their version of qp. So if you use qscale mode its like fixed qp recording.

ml's cpu usage is trivial for this. real problem is the bandwith for writing to the card. you get 20MB/s tops on digic IV . starts getting hairy around >120 sustained rate. i have quality following write rate and it gives pretty good results but nothing is stop free.

escho

Thanks for explanation

Edgar
https://sternenkarten.com/
600D, 6D, openSUSE Tumbleweed

wwjd

So, is the GENERAL CONSENSUS that using 1.4x bitrate with audio is the way to go?  It adds SOMETHING, works with most cards?
I'd like to sqeak as much quality out of my T3i as possible.
And is 2x and 3x VISUALLY the same, but maybe the files are uselessly bigger in 3X?
Thanks for the info!

1%

you have a 600D so you can use my builds which set qp and go by buffer usage. record with separate wav too. i think the stereo part is fixed now. I have to check again how it syncs up.

visually its all the same till you try to post/grade it. 6d, 5d3, 600D, all the files look the same to me unless i lower quality on purpose :)


wwjd

Quote from: 1% on February 02, 2013, 04:46:45 AM
you have a 600D so you can use my builds which set qp and go by buffer usage. record with separate wav too. i think the stereo part is fixed now. I have to check again how it syncs up.

visually its all the same till you try to post/grade it. 6d, 5d3, 600D, all the files look the same to me unless i lower quality on purpose :)

Thanks 1%.  I'm a newb here and don't understand what you mean by "Use My Builds"?  Is there a preset list or something?
I am definitely color grading everything, so if higher bit rate helps, I'm all over it.

1%

https://bitbucket.org/OtherOnePercent/tragic-lantern

you have to have a exfat formatted card to record over 4gb. if you don't you might get corruption. the bin before last that code got eaten so you're safe on fat32 but limited to 4gb