Re: [Lame-dev] ath5

[J.D. <jo...@ce...>]

Roel VdB wrote:
> 
> Judging by the form ATH5 will have an even much larger bitrate demand than
> the types 0-3.

ATH type 5 is based on collected data, not a bit rate target with
a particular encoder or encoder setting that is subject to change.

> by your own admittance it's still premature and not usefull. It has

No, please don't over generalize.  Certain ATH curves may reveal 
excessive bit rates in combination with certain LAME switches.
For some samples, even the default ATH curve yields excessive bit
rates.  The ATH curves are simply based on experimental data.

> even higher bitrate claims than the one currently in LAME and I think
> it would have been much more sensible to come up with a new ath once
> the need/possibility arises.

The need depends on the objective.  Use of ATH type 5 is optional,
just as a myriad of LAME's other features.

> 
> but, having a setting like "--r3mix" isn't very sensible also from
> some point of views. :)

For it's purpose, "--r3mix" is sensible.  There are many
objectives, purposes, or points of view.

> I plan to change type 3 as lame progresses, independently from type 2
> plans.  I don not wish other types to be dependent on my personal
> preferences.  That both are generated from the same formula is
> just a practical issue. Just as "f  = Min(18.0, f);" is no limitation
> for now.

I believe that the ATH type 3 curve represents more than just your 
personal preference.  As a consequence of pushing up the curve,
its development may have found a more accurate balance between
the high and low frequencies.

By the way, the "f  = Min(18.0, f);" does affect the function of
"--adapt-thres-type 2".  It may reduce the effectiveness of the 
auto-leveling for samples with much content above 18 kHz.  I
suspect that typically only rare cases may suffer due to this
quirk.

Kind regards,

- John