[LAU] digital volume sounds better at 0 dB?

Ray Rashif schivmeister at gmail.com
Sat Jun 19 02:28:18 UTC 2010


> The meters are probably completely useless.

+1

Though alsamixer, depending on your device, sets a threshold for
distortion somewhere near even -6dB (mine is when the numeric
indicator hits 90), the actual distorting point can be a little higher
in amplitude (but from experience not lower).

I think the notion of a "Digital Unity Gain" comes from the fact that
many in-the-box mixing engineers have a tendency to bounce to -0.3dB,
or any figure below 0dB depending on the standard operating level of
your target medium (like broadcast or live). This is to leave some
sort of safe headroom for when the digital material is being
processed/mastered to output through various devices, and how the 0dB
in digital is NOT 0dB in analogue.

> Yes it is. Alsamixer controls the volume in the mixer of the sound-card.
> JACK is all in software. Consider the following (somewhat simplified)
> diagram:
>
>  JACKified-audio-app
> (eg mplayer, ardour, jkmeter)
>         |
>         v
>     JACK-server
>         |
>         v
>     Soundcard  ## hardware-mixer controlled by alsamixer
>         |
>         v
>   Amp & Speakers
>
> When you adjust the volume in mplayer, mplayer itself scales the volume.
> mplayer does not interact with the soundcard's hardware-mixer. (Well,
> you can configure mplayer to do either or both, but the default with
> mplayer/JACK is to use mplayer's integrated "amp".)

Yep, though MPlayer with ALSA will control the soundcard by default
(one can add -softvol for an internal "amp").

I like to think of my software audio levels (relative to output
devices) in terms of hardware mixers, where you have:

1) the trim pot (the amount of raw signal you let in)
2) the fader (controlling how much attenuation/amplification applies
to the signal; always at unity gain initially)

My PCM levels would then be my trim pot, and the rest are faders :)


--
GPG/PGP ID: B42DDCAD


More information about the Linux-audio-user mailing list