Perfect Volume Control?, What dial do we turn? (high qualitty digital volume reduction?) Options

[puntloos]

Here's my next question in my 'Quest To Perfect Digital Audio'-series.

Assume:
Winamp (personal reasons, but the question applies to foobar2k too)
Play a bit-perfect copy of a CD source
All volume controls on 100%
Playback to digital out (straight to an external DAC)
and..

Current output is too loud! (Ouch!)

The question is:
Will turning down the digital volume (either winamp or windows master volume) impact the quality disproportionately negatively?

More specifically: turning down the volume will force some part of the digital audio chain to resample the output audio before it gets sent to the DAC. A lot of quality loss can be incurred (or prevented) by doing (or not doing) things like upsampling to higher kHz'es or bit depth, dithering/anti-aliasing etc etc etc.

Is there any way to determine what happens in my audio chain? I realise that it's probably best to just attenuate volume by my analog hardware (preamplifier) but the digital way is much more convenient for me. Does anyone have the detailed stats on:

- Does the windows mixer do stuff itself, or is it purely controlling the soundcard vendor's driver
- What does the vendor driver do? Are all vendors equal or do you have to have a good one?
- Does winamp hurt things? Is the MAD plugin a viable choice? Some other plugin? Kernel streaming?

Thoughts would be welcome!

[2Bdecided]

Firstly, changing the bitdepth is not called upsampling or downsampling.

Secondly, working in floating point, and dithering the output at 24bits, you lose nothing by reducing the volume of a 16-bit signal input by up to 48dB. In theory! You have 48dB "to work with" before losing data, but it's less than this since the dither noise will add. Further, if the original 16-bit dither+quantisation was noise-shaped, but the dither you apply to the 24-bit version isn't, then you have less still.

Beyond this, your D>A and associated electronics will have a fixed noise floor. By reducing the volume of what is contained in the digital input signal, you are sinking further into that noise floor. Whether this is relevant depends on the noise floor, the original recording, how much you reduce the volume, and yourself.

If the listening volume ("analogue" gain setting) is such that the noise floor (with dither) of the D>A is inaudible, then in theory no audible harm can be done by any digital reduction of the volume (e.g. what you lose in the digital domain by reducing the volume by 80dB would be inaudible had you chosen to reduce the analogue volume by 80dB!)

In practice, a less than perfect DAC could do damage (e.g. via distortion), but then in practice the signals which you attenuate digitally the most (i.e. the ones which are loudest in the first place) probably have least to lose.

In short, don't worry about it. Unless, if course, you find the noise floor of your 24-bit D>A is similar (or worse than!) that of a 16-bit recording - in that case, abandon this idea!

Cheers,
David.