sorry I didn't tell you before I commit the new quantize-pvt.c. The new routine uses same algorithm to treat "subblock_gain" that we treat the scalefactor. We didn't consider in the scalefactor when quntizing because the factor is multiplied to xr and l3_xmin before quantizing. I adapted this algorithm to the subblock_gain and now we can neglect the subblock_gain when quantizing. --- Takehiro TOMINAGA // may the source be with you! -- MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
