I've started testing Lame with synthetic input to test some lame properties.
One is a dual tone sweep, a slow araising sweep from 16 Hz to 14 kHz and
a second arising from this frequency up to 18 kHz.
CBR (-b160):
adds some clicks at the end of every fast sweep
VBR (-V4):
sounds really worse. All kinds of errors on a very high level.
May be also coding errors (it sounds like a lot of bit errors AND
masking errors).
VBR (-V0):
sounds equal to CBR
VBR (-V4 --noath):
sounds equal to CBR
VBR (-V4 -b112)
sounds equal to CBR
VBR (--athlower 80):
sounds equal to CBR
So I expected it is a ATH problem. But reducing input by 24 dB don't enforce
or reduce the problem. Tests on -48 dB are not possible because of the
limited input precission, quantization noise becomes important and makes
listening tests inpossible.
--
Frank Klemm
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )