Multiformat 128 kbps Listening Test
Reply #689 – 2005-11-29 16:33:21
run-in differences are mitigated by setting abchr-java to start playing 1 second into the sample. If nero's abr is like lame's abr, it doesn't work the same way as wma vbr 2-pass. That is, it would be a 1-pass abr, so there wouldn't be the same problem with encoding just short samples. Neros 1pass ABR will have relatively massive run-in discrepancies with the samples encoded insitu. If wmas 2pass targeting is unnacceptable Neros ABR is definitely out of the question.Congrats for the winner! And now back to the samples... The vote was really flawed -you all know it. Now my last attempt at waving the truth in front you guys: Suppose 3 phrases joined: A, B and C It would be wonky for the prepass calculated vbr setting not to be global, assuming it is, this equation would be true: phraseA_Bitrate*phraseA_Duration +phraseB_Bitrate*phraseB_Duration +phraseC_Bitrate*phraseC_Duration =target_Bitrate*Total_Duration (=total bit allocation) Next define Demandrate, a kind of passage complexity estimate from the encoders preferences, high for passages which would demand more bits, low for passages which would demand less. phrase_Demandrate=phrase_Bitrate/target_Bitrate phrase_Bitrate=phrase_Demandrate*target_Bitrate Substituting phrase_Bitrate for its Demandrate expression in previous equation... (phraseA_Demandrate*target_Bitrate)*phraseA_Duration +(phraseB_Demandrate*target_Bitrate)*phraseB_Duration +(phraseC_Demandrate*target_Bitrate)*phraseC_Duration =target_Bitrate*Total_Duration devide both sides of equation by target_Bitrate to leave: phraseA_Demandrate*phraseA_Duration +phraseB_Demandrate*phraseB_Duration +phraseC_Demandrate*phraseC_Duration =Total_Duration That equation illustrates (by being linear) that the bitrate demands of A and B (Phrase_Demandrate*phraseDuration) remain in proportion with each other, nomatter what the demand of C. To be fair C's demand only needs to be normalised because if it is greater than (A+B)/2, A+B is deprived, if it is less A+B is boosted. C can be normalised most easily by removing it altogether, alternatively it could be a sample of the same DemandRate as (A+B)/2 The longer C's duration, the greater its DemandRates effect on A and B's bit allocation, at 0 it ceases to have an effect. The difference between doing the 2pass bitrate targeting on only A+B, and doing it on A+B+C where C is the clipped portion of the parent tracks is at least as transient to the test results, as the difference between the test samples used, and the tracks people will actualy encode guided by the tests results(!!) That's me done. Good luck [span style='font-size:7pt;line-height:100%']edit: speling[/span]