Re[2]: [Lame-dev] adaptative ath

[Roel V. <vd...@yu...>]

Hello Gabriel,

Wednesday, April 18, 2001, 10:50:54 PM, you wrote:

>> I don't have much to add, but maybe you can use that at the moment
>>
>> -V1 + "+ 6; //std --athlower -6" is considered transparent for the
>> freqs <12khz.  Verified by Hans Heijden, which means something...
>>
>> Maybe a point to start.  Would be silly IMO to start
>> with the same LF bitwaste which was there in <=3.87.

GB> Yes, I'm aware of this experiment.
GB> The first thing is that to my mind this should influence the ath shape as it
GB> has nothing to do with the shape, only the level. So I don't think the ath
GB> function is the right place to put the level adjustment.
GB> The second thing is that this is something I'm taking into consideration. Of
GB> course it would require more testing, but it might be an indication that the
GB> base level of the curve is not optimal.

thanks for your comments.

When I did this test, I started from the assumption the base level of
the curve would be too low, just to compensate the poor type0 HF
realism trying to avoid HF artifacts.  Asuming LAME had evolved as
such in the long time of listening tests determining the -V scale.

Since the LF's @ ATHlower -6 yield what most people consider
transparent it might be a wise idea to start from this point and
hallmark (?) as a -V1 or -V2 in the -V scale. (since this should be
about transparent to most)

If you leave the level like it is now then people will end up with
something like --athlower -6 -V* for their optimal results.

You have a good idea about adjusting the curves, but I see no reason
why not to use the athlower knowledge as a starting point.

Highly unlikely people will be testing with --athlower -6 in their
commandline and imo very likely people settle for the adaptive ath
wich is their liking of size, but will "waste" LF bits.

Do you get my point, or am I missing something?

nice work,

Roel