Reverse-engineering my speakers' API to get reasonable volume control

I know it is tangential, but this about his old system caught my attention:

  With that system, I could set the amplifier’s analog volume knob such that the max volume out of the streamer corresponded to my actual maximum preferred listening volume, giving me access to the full range of Spotify or AirPlay’s volume controls.

Assuming an analog input, this might result in a noticable quality reduction at low volumes.

In this case, I was using the optical out from a WiiM mini into a Yamaha amp. I don't know much about digital audio, but I know that I was able to control the volume of the WiiM's digital output with that setup.

On the other hand, I use a Schiit Asgard at my desk, where I have it connected to my Mac via USB-C. In that setup, I have no control over the volume level going in to the Asgard. MacOS just disables the software volume control when I'm using that audio output.

Reducing excessively before the DAC and high gain after the DAC is far more likely to result in quality reduction, due to quantization error. Having reasonable levels before the DAC and just the right amount of gain after the DAC (e.g., via an amplifier's attenuator setting) is the best possible scenario. So TFA's prior setup may have been superior in this regard, depending on how the digital volume control on the new speakers is implemented (i.e., before the DAC, or as a VCA after the DAC).

Where this breaks down is if the analog signal path (after the DAC) consists of something noisy after the attenuator. Passive attenuation (like built into an amp, or the master fader of a mixer, etc.) won't add noise, but something active like an outboard EQ would. The attenuation to set desired max level must be completely last (before power amp) to avoid noise.

Tangential fun fact: amps have a fixed gain, because it's hard to make a variable gain without distortion [1]. The volume knob doesn't control amplification, in fact it controls an attenuation stage, because it's easier to make variable attenuation with low distortion.

[1] that's why there were so many different "classes" of amps, they're all making different tradeoffs about how they're doing the amplification.

That helps explain why the "volume" as represented on disk in the debug bundle as "attenuation" and was measured in negative dB.

We really don't want to be touching the digital signal. The state of the art is to change the DAC reference level, putting the DAC output at the sweet spot for the analog stage for any given ultimate output level.

I think for a WiiM mini to control the volume on the digital output, it would need to scale down every sample. This is probably fine over some range (it has a 24 bit output, so putting the volume at two thirds, would still result in 16 bits, the same as CD). But I'm curious what would happen at very low volumes, e.g. if you're down to only 4 bits.

According to Claude, the attenuation stage is before the power amp stage. Does that mean worse SNR whether the volume is controlled using the volume control or via the input?

(Ignoring the additional quantization issue with a scaled digital input.)

Each bit is ~6dB, so 2/3 perceived volume is still 23 bits. 8 bits is 48dB, which is less than 1% of the original volume, and still using 16 bits

And this is why all computer audio should be 24 bit internally. A lot of the newer pro apps are actually using 32 bit floats

Class D doesn't have any attenuation. This is a big factor in their greater efficiency.

You could put the attenuation stage after the power amp stage but it would require big beefy resistors that could absorb a lot of power. They'd get hot and the whole thing would be very inefficient.

But hey, very low distortion.

Too late to edit my original comments. I am clearly mistaken about the relationship between amplitude and perceived loudness.