> From: Takehiro Tominaga <[EMAIL PROTECTED]>
> Date: Fri, 07 Jan 2000 13:26:54 +0900
>
> >>>>> "R" == Robert Hegemann <[EMAIL PROTECTED]> writes:
>
> R> Mark wrote:
> >> I'm still not sure the gspi35_1.wav problem is caused by the
> >> ATH. I think it is actually caused by other flaws in the psy
> >> model. My reasons are:
> >>
> >> 1. using --athonly -V5 doesn't have the problem (in this case
> >> lame is choosing the bitrate so there is no audible distortion
> >> based only on the ATH)
> >>
> >> 2. any change which reduces the masking, such as using
> >> MAXNOISE for the noise calculations, or replacing the spreading
> >> function with something more reasonable, will also fix the
> >> problem.
> >>
> : <snip>
>
> Uum, I found this ATH problem using
>
> %> lame -v gspi35_1.wav
>
> The 3rd glockenspiegel sustain sound has many artefact.
>
Yes, but how does not mean that it is an ATH problem? The masking
used with 'lame -v' is the ATH + psymodel. Do anything to reduce the
masking (remove the psymodel, change the spreading function, remove
the ATH or set MAXNOISE) and you get more bits allocated and no
artifacts.
Using VBR to encode every frame so that over=0 using just
the ATH will show no problems (and give very large files), so
I'm pretty sure the ATH is ok. Using VBR to encode every
frame so that over=0 using the psymodel() maskings will
show many flaws of our psy model :-). Thus when the ATH
thinks there is no audible distortion, it is usually right.
When the psymodel thinks there is no audible distortion
in the compressed signal, it is often wrong.
gspi35_1.wav actually has a few short sections where the
left & right channels are similar but with a different phase, so maybe
I have to retract my statement about this type of case being
extremely rare. gspi35_1.wav is not as extreme as Robert's
case, but the side channel has as much as 2x the energy in the
mid channel.
FhG encodes these passages with mid/side stereo but allocates about
the same number of bits to mid and side channel. Lame is confused and
is toggleing back and forth a lot between mid/side and l/r, and when
lame uses mid/side it is reducing the number of bits on the side
channel.
My feeling is that if mid and side are going to get the same
number of bits, why not use regular stereo? But FhG doesn't
seem to agree with this.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
