> 2) ATH
>
> Without Leonid\'s fix and without -Y -X1 it produces audible artifacts.
I tried to divide ATH and useful values was from 10 to 100 (it was depending
on type of music) - I know it seems too much (20 to 40 dB).
>
>
> for (i=start ; i < end; i++) {
> freq = samp_freq*i/(2*576); /* freq in kHz */
> ATH_l[sfb]=Min(ATH_l[sfb],ATHformula(freq));
> }
>
> ++ ATH_l[sfb]=ATH_l[sfb]/20;
>
> }
> same for short blocks
>
> If I let ATH unchanged, joe satriani sounds worse (more flanger) than
with lame v3.1x
>
I don't know if you know Bosse Lincolns Paper:
"An Experimental High Fidelity Perceptual Audio Coder
Project in MUS420 Win 97"
There you can read, that after some listening tests the "setting
didn't seem to keep high enough frequencies for transient sounds".
And it's supposed to replace the MDCT scaling in ATHformula
ath-=96;
with
ath-=114;
Further in calc_xmin, I think we have to multiply enmax with
its appropriate ratio, but I'm not sure about that.
Robert
PS: Because ATH is mentioned even in the iso Doc, I think Fhg
uses it, and we can't outperform Fhg with it.
--
Sent through Global Message Exchange - http://www.gmx.net
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
