I need to understand how you take a digital stream and run it through the muse parametric eq. Do you convert it to analog first, adjust the levels, and encode it back into a digital signal? Or do you some how take the digital signal and and adjust it in it's native format/encoding? I really need to know what the roon server is doing specifically from a design perspective. I need to know if the adjustments I'm making will degrade the raw digital stream. Especially if I add 3 to 6db of gain. I use Tidal. Thx. John
Roon server is an app, i.e. just software, so everything it does in terms of DSP (which stands for Digital Signal Processing) is crunch numbers. This means that the signal stays in the digital domain - all the way up to your DAC. During this processing, samples don’t normally stay in the encoding with which they are stored in files, since that is not suitable for the transforms they need to go through. They are converted to a format called floating point, which can accommodate all the existing encodings without any loss of quality. Before being send to the DAC, the samples are converted back to the best available encoding the DAC supports. That is usually 24 or 32 bits, so there’s no audible degradation. Now, if there was any quality degradation with PEQ, that would be because of the curve itself, not because of the internal format transformations.
(In order to have an analog path, you would need some kind of analog hardware, capable of applying arbitrary EQ, controllable by software. Not only is analog hardware not nearly as versatile as software, that would severely limit the kind of hardware Roon runs on. Obviously, that is not the case, considering that Roon server can run on pretty much any computer with Linux, Windows or MacOS installed, and even on certain NAS devices, which have no internal analog audio components whatsoever.)
In addition to what @Marian said, it’s first converted specifically to (huge) 64-bit floating point, then the DSP calculations are made, then it is converted down to the DAC’s 24 or 32-bit input, meaning that mathematical conversion errors are pushed far outside of audible areas. Fellow forumer @DrCWO performed a detailed analysis of this for Roon’s DSP volume, which you can read here: