canadian_fbi
24th April 2002, 06:55
i think darkavenger is gone so this question may go unresolved, but i'll ask it anyway.
has anyone played around with the lowpass setting in headac3he for vorbis encoding? i was messing around with it last night, trying to see what the effect of a more restrictive lowpass setting would be on quality and filesize, when i noticed something a little odd. at a setting of 4.99, it defaults to 47.7 kHz, and in fact stays above 20.5 kHz all the way to 4.00. however, i converted a couple of sample oggs encoded at 4.99 to wav and did a frequency analysis of them, and it looks like there's a lowpass cutoff at 18 kHz. it was pretty striking with a couple of different samples, all of which had default lowpass settings of at least 20 kHz.
so is there something in the ogg encoder that automatically puts a lowpass in at 18 kHz at q<5? (haven't tested above that). i don't think i can hear above 18 anyway, but it makes the value displayed by headac3he pretty misleading if that is the case, and it takes away some possible room for cutting out unnecessary frequencies.
has anyone played around with the lowpass setting in headac3he for vorbis encoding? i was messing around with it last night, trying to see what the effect of a more restrictive lowpass setting would be on quality and filesize, when i noticed something a little odd. at a setting of 4.99, it defaults to 47.7 kHz, and in fact stays above 20.5 kHz all the way to 4.00. however, i converted a couple of sample oggs encoded at 4.99 to wav and did a frequency analysis of them, and it looks like there's a lowpass cutoff at 18 kHz. it was pretty striking with a couple of different samples, all of which had default lowpass settings of at least 20 kHz.
so is there something in the ogg encoder that automatically puts a lowpass in at 18 kHz at q<5? (haven't tested above that). i don't think i can hear above 18 anyway, but it makes the value displayed by headac3he pretty misleading if that is the case, and it takes away some possible room for cutting out unnecessary frequencies.