> ?Pierre, what do you mean by the size of the PCM? Do you mean the frame > length? If so, I am confused, because the frame length explicitly depends on > the bitpool parameter. So, the frame length will change. You provide nsubbands*nblocks samples as input. This cannot change. > Moreover, I think that's the whole idea of an adaptive encoding. SBC > encoding masks sounds that are not audible to the human ear because they are > overshadowed by other sounds. Depending on how much the encoder can mask the > frame will be smaller or bigger. My understanding might be wrong. So, please > correct me. Oh now I get it. I am afraid you can forget about this idea of variable bit rate encoding with SBC. What you can do is change the bitrate if there are too many transmission errors, etc. This means adapting the bitrate to the link. But you cannot modify the bitrate within the encoder to improve the quality. First this might not be supported by the link. And second, SBC isn't a very good encoder, it's a simplified version of MPEG1-layer1, somewhat worse than the old Philips digital cassette if you are old enough to know how bad this was (1990?). There's no real computation of a masking threshold. It pretty much iteratively allocates bits to the subbands with the most energy and when the bit allocation stops, well you get what you get. If I remember well, the allocation algorithm is used both in the encoder and decoder so your freedom to improve it is next to nil. Yet another reason why we should have MP3 direct streaming... - Pierre
