I think that this statement at first can seem stupid but if I think a little...

If I have 16 bit of resolution (65536 values possible) I also have a "banwidth" of 96 dB...If I set the interval from 0 to 65535 and assume that the value 0 is equal to -96 dB, than I can easily say that 65535 is 0 dB.

That also means that value 1 is equal to -90 dB, value 3 is equal to -84 dB, value 7 is equal to -78 dB and so on, having that for -6 dB the value is 32767. The conclusion is that if I have an audio sample that have a range of 6 dB it would be better to have the maximum volume (from -6 dB and 0 dB) to have the maximum details possible (the sound wave can have a range of 65535-32767=32768 "steps"), if I decide to lower volume of, for example, 6 dB, my resolution will be of 16384, that's half of the original range.

In the end my question is: using tools such as replaygain can reduce sound details? Is it possible to avoid this by activating the "dither" option in playback?

Thanks for answers and sorry for my bad english...

You as well as nowadays music almost never makes use of all the available bandwidth. Typical quiet rooms have a noisefloor of at least 40dB. The threshold of pain is at about 120dB. Thus, even if you damage your ears, you will not make use of all the bandwidth. Plus, nowadays music barely makes use of any dynamic range at all (compression). The bottleneck regarding soundquality typically isn't bitdepth or samplerate, but instead (in this order): 1. the music itself, 2.loudspeakers, 3.room acoustics, 4. room noisefloor.

edit: removed first sentence because i misread your original post (i was asuming 3dB, but you were talking about 6dB)

The first reply is telling you not to worry about it.. You probably won't be able to notice the difference. Regardless, the answer to your question is yes, reducing the level in the digital domain does reduce the available bit depth. Details drop off the bottom. That is, sounds at the lowest level will disappear, or at least lose detail, if any such sounds exist to begin with.

You cannot avoid this reduction by dithering. Dithering can help resolve the lowest level details better, or at least make them sound better by increasing contrast, but that only applies to what is actually there in the source, not what you've thrown out by reducing the bit depth.

Bear in mind that this is not something limited to the digital domain - if you lower the volume with an analogue source, some previously audible details will also sink into the noise floor.

Actually the correct terminology is a dynamic range of 96dB. But yes what you say is basically correct, if you lower the volume of a recording then you will somewhat reduce the available dynamic range of that piece. For most people it's not really an issue, but it is the reason why many people who have no interest at all in higher bit depths (24 bit and 32 bit) for playback will still prefer to use them for recording an processing stages.

Anyway if you think about it properly then what you are doing by lowering the volume is just choosing to use some of your available 96dB dynamic range for that purpose. Having some tracks (or portions thereof) louder and some quieter is after all what dynamic range is for.

Yes, by lowering audio volume you will loose some resolution as well as degrade S/N ratio. But 96dB is pretty much, so you can sacrifice some... As for details lost by lower resolution - you be the judge.

You might want to increase bit depth to 24bit or 32bit so you don't sacrifice anything (other than disk space).


Regards,
Goran Tomas