If we can allow B and P frames to grow as large as they need then would a longer GOP have the same look as GOP1? Or is there a belief that there is something inherent in AVCHDs encoding of B and P frames that the quality will always be less then an I-frame?
I remember @cbrandin saying that a B or P frame could never improve on on the quality of the preceding I-frame, so what if that I-frame had used a low bitrate because there wasn't much detail in it? e.g. The preceding I-frame was during heavy movement, and the blurring caused the bitrate to drop and then the following B/P frames were a static shot. Is there really no way a B/P frame can add detail? They certainly must be able in other encoding schemes, because it's possible to have lossless long GOP codecs.
camera comes in about an hour, classic soligor glass ready to go. Will be testing GOP1 tonight, and shooting 2 shorts and starting on our feature this upcoming month
Well here it is, the latest Driftwood INTRA GOP1 176MH (15-18% better i frames) plus 1080i60 INTRA. This is Experimental so try it out and let me know how you get on. We're upto nearly 800,000 sized i frames now :-) Stable so far on three different lenses. Panny 7-14 ASPH, Panny 14-140, FD 50mm 1.2 ASPH. Yet to test the Voight. The 1080i modes are precarious at the moment and the 720p modes have been successful but wont work whilst 1080i modes are as they are so still work to be done. BYSI Yes there's some weird looking tryouts on the FBs at the moment.
Driftwood PURE Intra ver2 - 176M GOP1H - 1080i60 100M GOP1 - seti.zip
Although the 1GOP festival is just going to rise in public, I stick as I stick, since I improved a little my 3GOP AQ3 setting.
Since I set 24L to 50M, it may also be able to be used. Although tested with swaying the Stray's crazy chart and ss=1/800 cleared, how is it?
About 1GOP ... I am also satisfactory by driftwood's great work for the time being. :-)
Although it can record by 24p ordinarily in this setting, it is displayed as why or "Mode = 1080/24pN (actual frame rate is half)". I do not know the reason.
Don't be too sure about the 1/2 framerate thing. StreamParser simply reports whatever is in the TS header. If that information is incorrect because of a patch parameter StreamParser will report it anyway. It would be good to confirm that with an editor or some other program. Framerate is contained in a couple of different headers, StreamParser uses the TS header version, editors typically use the version stored in one of the AVC headers - if there is a problem, they might differ. @driftwood is right about the bitrate - if the average bitrate is above the max bitrate something is going to give - either the framerate is lowered, or you get the "cadence" issue.
@bkmcwd Thanks for the new 66M 3GOP AQ3! I will test this with my new extreme tests with Stray's new extreme dense chart. (very nice chart). Friday, I will tests your setup under low light and extreme tests for your new
Also, I see that there's an added "p" frame at the start. Does this fuck up the frame order?
driftwood, Thanks again for the long hours of testing. I'm looking forward to running some tests on this INTRA GOP1 176MH (15-18% better i frames). 800MB i-frames is a good improvement.
Man, if we could only get them to hit at 1.0 - 1.1M Do you think it's still possible?
Also, comparing the 88M 12GOP AQ2 MAX VARIATION (stray's built from chris's max variaition 66M), how much detail loss in your GOP1 176M with i-frames at 800K?
Friday, I will tests your setup under low light and extreme tests for your new INTRA GOP1 176M.
@proaudio I'm confused. 1byte=8bits , ok ? So 1.0Mbyte = 8Mbits , ok ? So,24 frames * 8Mbits (I frame ) = 192 Mbits/s ? Correct me if I'm wrong. So it is possible with Gh2 Hack :-) .