The psymodel split

[Gabriel Bouvigne]

Looking at current source code, I noticed that several parts of the
psymodel have been extracted for vbr.

exemple: vbrpsy_attack_detection
This one looks very similar to the regular one, except some additionnal
tuning. What I am wondering is why is it not part of the "regular"
psymodel? Is it a lack of time, a matter of convenience, or are the
changes incompatible with the cbr/abr modes?

Considering that about 90% of the vbrpsy_attack_detection is identical
to the regular attack detection (btw I have the feeling that the fs/4
lowpass might be wrong), my opinion is that we would better use the same
piece of code as much as possible.

Regards,

--
Gabriel Bouvigne
www.mp3-tech.org
personal page: http://gabriel.mp3-tech.org

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev

[Robert Hegemann]

Hello Gabriel,

----- Original Message -----
From: "Gabriel Bouvigne" <[hidden email]>
To: <[hidden email]>
Sent: Monday, May 19, 2008 10:12 AM
Subject: [Lame-dev] The psymodel split


> Looking at current source code, I noticed that several parts of the
> psymodel have been extracted for vbr.
>
> exemple: vbrpsy_attack_detection
> This one looks very similar to the regular one, except some additionnal
> tuning. What I am wondering is why is it not part of the "regular"
> psymodel? Is it a lack of time, a matter of convenience, or are the
> changes incompatible with the cbr/abr modes?

The main reason is, I wanted to keep the non-VBR-new part as-is.
If it turns out, after doing some listening test, that the rest of the code
can live with my modifications for VBR new, then we can use it there too.

> Considering that about 90% of the vbrpsy_attack_detection is identical
> to the regular attack detection (btw I have the feeling that the fs/4
> lowpass might be wrong), my opinion is that we would better use the same
> piece of code as much as possible.
>
> Regards,
>
> --
> Gabriel Bouvigne
> www.mp3-tech.org
> personal page: http://gabriel.mp3-tech.org

Ciao Robert


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev

[Gabriel Bouvigne]

Robert Hegemann a écrit :
>> Looking at current source code, I noticed that several parts of the
>> psymodel have been extracted for vbr.

> The main reason is, I wanted to keep the non-VBR-new part as-is.
> If it turns out, after doing some listening test, that the rest of the code
> can live with my modifications for VBR new, then we can use it there too.

Aren't you afraid of ending up with 2 "separate" models to maintain? For
tunings I understand the separations, but some parts like the mask_add
changes are fixes, and as such would also need to be applied to cbr/abr.


--
Gabriel Bouvigne
www.mp3-tech.org
personal page: http://gabriel.mp3-tech.org

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev

[Robert Hegemann]

>>> Looking at current source code, I noticed that several parts of the
>>> psymodel have been extracted for vbr.
>
>> The main reason is, I wanted to keep the non-VBR-new part as-is.
>> If it turns out, after doing some listening test, that the rest of the
>> code
>> can live with my modifications for VBR new, then we can use it there too.
>
> Aren't you afraid of ending up with 2 "separate" models to maintain? For
> tunings I understand the separations, but some parts like the mask_add
> changes are fixes, and as such would also need to be applied to cbr/abr.

IMO the mask_add, as it is used by CBR/ABR/VBR-old, does not make
much sense. But I didn't want to change it for 3.98 mostly because I did not
have the time to explore all the consequences a change would have for
CBR/ABR
and VBR-old. So decided to let it as-is and fix it for VBR-new, which I
was working on. I would suggest to review it when working on version 3.99,
after releasing version 3.98 stable.

> Gabriel Bouvigne
> www.mp3-tech.org
> personal page: http://gabriel.mp3-tech.org



Ciao Robert


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev

[Gabriel Bouvigne]

Robert Hegemann a écrit :

> I would suggest to review it when working on version 3.99,
> after releasing version 3.98 stable.

+1



--
Gabriel Bouvigne
www.mp3-tech.org
personal page: http://gabriel.mp3-tech.org

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev