I own one test CD where are parts of classical music recorded at different
levels (0, -20, -40, -60 and -80 dB). I tried -60 and -80 dB with and
without --noath and I can hear only discrete crackling sound (using
headphones and amplifier at max volume ;-\ ). With --noath it is much better
(original wav is even better).
I know with ATH is assumed that range is between least volume humans can
hear and pressure that give you pain, in addition limited by 16bit dynamic
range (in my test conditions, full 16bit range sound would sure push my
ears into my skull). So if I want to listen music at constant volume level,
modifying ATH is useless. If I want to play with volume knob it is not.
Slavo
------------Original Message---------
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]On Behalf Of Mark Taylor
Sent: Monday, February 14, 2000 7:13 AM
To: [EMAIL PROTECTED]
Subject: Re: [MP3 ENCODER] ATH matching to VBR quality
> From: "Slavo Kopinec" <[EMAIL PROTECTED]>
> Date: Fri, 11 Feb 2000 11:55:33 +0100
>
> I nominate this small change:
>
> quantize-pvt.c, line 267:
>
> replace
> ath -= 114;
>
> with
>
> ath -= (122-gf.VBR_q*2);
>
> to ensure that in high VBR modes (V0, V1) nothing is muted out. (br. is
> increased by near 10kbps in V0)
>
> Slavo (alias macik)
>
Have you seen any indication that the ATH is too large?
I know decreasing the ATH will increase the bitrate (and result in
better quality), but decreasing any of the allowed masking
will have the same effect. Thus I would rather decrease the
masking values computed in psymodel() then decrease the ATH.
Unlike the normal masking thresholds, I "believe" in the ATH.
That is, if we get the adjustment for the db of "silence" correct,
then I think encoding so that all distortion < ATH, will still
sound perfect.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
