bond
26th December 2005, 11:22
the framedropratio option is now supposed to add n-vops also when b-frames are used, when no b-frames are in current queue
now syskin meant that using packed bitstream or not shouldnt make a difference, but in my tests this isnt that way:
there is a difference between using pb and not using it regarding real n-vops placement as it seems with pb more n-vops are placed in the stream than without pb (i am of course not talking about the pb dummy n-vops) in my test sample
can anyone reproduce this?
another thing:
syskin explained to me that the placement should works as follows:
PBI turns into PPI (closed gop) which in turn changes into PNI, so in *theory* any PP changes into PN if drop ratio is high enough, but if difference between p-frames is so small, the second P will almost certainly be B
so i wondered what ~ framedropratio will actually lead to a N, as small differences lead to B and high differences to P?
in my not packed test encode (not perfectly identical black frames) the very last frame was only a N (PBN) only with fdr100 (with packed bitstream there were more, none at the end, but i think thats a bug as written above)
now syskin meant that using packed bitstream or not shouldnt make a difference, but in my tests this isnt that way:
there is a difference between using pb and not using it regarding real n-vops placement as it seems with pb more n-vops are placed in the stream than without pb (i am of course not talking about the pb dummy n-vops) in my test sample
can anyone reproduce this?
another thing:
syskin explained to me that the placement should works as follows:
PBI turns into PPI (closed gop) which in turn changes into PNI, so in *theory* any PP changes into PN if drop ratio is high enough, but if difference between p-frames is so small, the second P will almost certainly be B
so i wondered what ~ framedropratio will actually lead to a N, as small differences lead to B and high differences to P?
in my not packed test encode (not perfectly identical black frames) the very last frame was only a N (PBN) only with fdr100 (with packed bitstream there were more, none at the end, but i think thats a bug as written above)