If I run a streamer via analoge into a power amp and control the volume exclusively in the streamer, wouldn’t that be in the digital domain?
Probably, but it the streamer has an analog output it might have an analog volume control.
What I‘m getting at are cut off bits. I‘m fully aware that there is a certain headroom for that. But where does this end?
Each bit represents 6dB. If you have 16-bit audio and you lower the volume by -12dB you are only using 14-bits (the two most significant bits are zeros).
If you have 24-bit audio you can reduce by -48dB and you still have 16-bits.
The same thing happens during quiet parts of the song/program but you can't hear a loss of quality because the quantization noise remains below audibility. Because of that,
I say it's OK to use digital volume control for "everyday volume control" and you'll have all of the bits when you turn it up. But you may not want to use it for "gain control" or "permanent attenuation" if you're never going to turn it up and get all of the bits.
Low resolution makes the quantization noise more audible. If you want to know what low resolution audio sounds like, open a file in
Audacity* and export it as an 8-WAV file. Quantization noise is like a fuzz on top of the signal. Like regular analog noise, quantization noise is more noticeable with quiet sounds but unlike analog noise it goes-away complete with dead-digital silence.
Most people who talk about resolution have no idea what it sounds like... The perception of quantization noise is not intuitive.
The real issue is if you re-amplify. And analog volume control has similar issues. If you attenuate by a lot and then re-amplify, any noise that's generated after attenuation gets amplified also and you end-up with worse signal-to-noise ratio.
Also "traditional" analog volume control potentiometers sometimes have differences in the left & right pots and the channels may go out of balance, especially at lower volumes. (There are "stepped" analog attenuators that avoid this.)
In my understanding a preamp or preamp section in an integrated amp attenuate the analoge signal.
They usually have available gain if/when you need it but in most cases we are attenuating, especially if we're not listening at maximum volume.
But noise floor is one thing. What about dynamics and the audiophile-knows-what?
Dynamics - I like the term "dynamic contrast" is the difference between the loud and quiet parts and the comes
from the program. It's complicated because there are long-term dynamics like a song or symphony that starts out quiet and ends loud or micro-dynamics like a loud cymbal crash, and everything in-between. Most modern recording have dynamic compression to make "everything constantly loud" (
The Loudness War). A lot of use music lovers feel that it's overdone, it makes music boring, and we just turn it down.
Basically, everything is linear and the program dynamic aren't hurt or compressed by the equipment unless you over-drive an amplifier into clipping, which is a "very bad kind" of compression. But clipping is simply heard as distortion rather than a loss of dynamics.
Note that
dynamic compression is totally unrelated to
file compression like MP3 which is used to make the file smaller. MP3 is lossy compression and it throws-away information (it tries to throw-away details we can't hear anyway) but it does NOT compress the dynamics.
Dynamic range relates to
the equipment, storage format or transmission. It's limited by noise on the quiet-end and the maximum signal level on the loud-end. The dynamic range of the equipment far-exceeds the dynamic contrast in the program (ignoring "infinite" fade-outs, etc.) but it's still very important because
we don't want to hear the noise at all!, no matter the dynamic range or signal-to-noise ratio.
With digital, it's just a range of numbers... How high you can "count to" with a given number of bits, with 0dBFS (zero decibels full scale) representing the "digital maximum". And how low you can go before you count down to zero.** A 24-bit file has bigger numbers than an 8-bit file but everything is automatically scaled to match the DAC so the 24-bit file isn't louder. And as above, when you get close to zero, the loss of resolution is heard (if heard) as quantization noise. Floating-point digital, for all practical purposes, has infinite dynamic range. IIRC, 32-bit floating-point goes from about -750dB to +750dB (with a value of 1.0 representing 0dBFS). Of course, that dynamic range only exists in the digital domain.
* If you don't already have Audacity "download without Muse Hub" so you don't accidently sign-up for anything extra. And go into the setting and turn off dither, because dither is added noise that's supposed to sound better than quantization noise.
** If you include zero, which is digital dead-silence, we could say that digital has infinite dynamic range because an amplitude of zero is minus infinity dB. But that's not helpful in the practical world. We want to know the lowest non-zero amplitude or the quantization noise.
