> We'll have to check the normalization of the two codes.
> I think we all define a db value for silence in the same way: the
> energy of a sine wave with the smallest possible amplitude that a CD
> can handle. But this value will depends on the normalization of the
> MDCT, and we can only compare the two codes if they use the same
> normalization.
>
> Mark
Reading Bosse Lincolns Paper I don't see much difference between them.
But here is another thing we may think about. The ATH curve was
approximated from experiments on a number (how many?) of subjects
(what age?). The formula we have implemented seems to be
approximating the results for subjects of age 20. Maybe we should modify
the formula for the virtual result of subjects aged 0?
I'm thinking of the following change I haven't tested yet.
ath=(3.640 * pow(f,-0.8)
- 6.500 * exp(-0.6*pow(f-3.3,2.0))
+ 0.0005 * pow(f,4.0)); /*<-- changed regarding high frequencies */
Robert
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
