Re: [Lame-dev] substep 2 and outer loop

[Takehiro T. <tak...@in...>]

>In outer loop:
>     if (++age > 3 && best_noise_info.over_count == 0)

>I noticed that increasing the maximum age value is reducing the bitrate
>difference between substep 2 and no substep. Perhaps the maximum age value
>should be x2 when using substep 2?

Oh, good point. I agree.

>Perhaps the value of maximum age could be set to 20 (or similar number) when
>q1 or q0? It seems to me that it's providing the same amount of saving as
>the optimal huffman, but not slowing down things in the same proportions.

Agree and not agree :)

IIRC, current -q mapping is shown below.

-q3: old(before LAME 3.93.1) -q2. default.
-q2: -q3 + best huffman in the inner loop.
-q1: -q2 + only 1 sfb to amplify in the inner loop.
-q0: -q1 + substep shaping

Beside the substep method 1 which is usually decrease the bitrate too much,
I think "substep method 2" is safe to use. So I propose to change like below.

-q3: same.
-q2: -q3 + substep shaping method 2 and increase max_age.
-q1: -q2 + only 1 sfb to amplify in the inner loop.
-q0: -q1 + best huffman in the inner loop.

Any ideas ?
-- 
Takehiro TOMINAGA // may the source be with you!