I�m not sure that a substraction of 5dB -- or any other
constant x -- from all masking thresholds is the right way to go.
We could vary this value along the Absolut Threshold of Hearing.
The advantage of this method would be a prevention of distortions
at frequencies where the ear is most sensitive.
Here is a little gnuplot script. The plotted function would give a maximum
reduction of nearly 9dB at around 3kHz and no reduction at frequencies
lower than 475 Hz and higher than 9kHz.
------------------------------------begin-of-script----------
min(a,b)=(a<b)?a:b
e=exp(1)
ATH(f)=3.64*(f/1000)**(-0.8)-6.5*e**(-0.6*(f/1000-3.3)**2)+10**(-3)*(f/1000)**4
set logscale x
plot [f=20:22050] -1*min(0.0,0.75*ATH(f)-5)
------------------------------------end-of-script-----------
Robert
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
