all dsp disabled is the best option. just a simple unadulterated rbcd stream and the mscaler will take care of everything.
@mastercheif
The potential for transparency is maximized when you send the source bits untouched to the MScaler+DAC. Even the 64-bit + dithered processing by Roon will be detrimental to transparency. It's audible for sure.
I fully understand that using DSP will reduce "transparency", it's plainly audible when I do volume matched comparisons between DSP engaged vs disengaged.
This is a difficult concept to explain, and I can only speak for myself, however you may be surprised to find that "transparency" does not always mean "the most faithful reproduction of the original music performance". "Transparency" in the sense that Rob (I think) uses it or in general conversations about DACs is equivalent to "reproducing the analogue signal that was in the ADC". While AD/DA conversion is a critical component in the audio/music reproduction chain, there are other factors to consider as well. Here's an illustration with commentary of what I'm getting at:
Musicians playing instruments in a room -> Microphone -> Preamp / Analog mixer -> A/D converter -> Mixing / Mastering -> File.
This file contains audio information such as:
"at sample 435784237 the amplitude in the left channel is equal to +4bits"
Easy enough. What happens when you try to play back this file?
File -> D/A converter -> PreAmplification -> Amplifier -> Speakers / Headphones -> Air -> Ears
As you can see, there are many sources of potential distortions and other nasty bits that can interfere with your ability to even hear your pristine D/A conversion.
The simple answer here is to not buy gear that distorts the audio, and I practice this methodology to fullest extent possible. This is why I'm seriously considering driving my speakers from a TT2 (or waiting for DX).
However, money can only get you so much audio gear. The "Air" part of the reproduction chain is where things can really get difficult. I live in a small apartment in NYC and my listening room is long and narrow. I have a huge room mode at 70hz and nasty side wall reflections. I've mitigated these somewhat with the purchase of bass traps and absorption panels from GIK Acoustics, but they can only do so much to cheat the physics of the room.
Getting back to the File -> Ear interface, let's consider again:
"at sample 435784237 the amplitude in the left channel is equal to +4bits"
What if this amplitude/time curve is somewhere around the nasty 70hz room mode in my room? In that case, the resulting sound that one would be hearing would be more like:
"at sample 435784237 the amplitude in the left channel is equal to +10bits"
The distortions that the room imposes on the audio heard by the listener can be far more detrimental to a faithful reproduction of the music performance than a slight loss in transparency resulting from the use of DSP.
This is of course assuming that the DSP is done in a competent manner. I use
https://www.audiovero.de/en/acourate.php to generate the FIR convolution filters because it allows for the fine tuning of nearly every parameter involved in the filter generation process, allowing one to tailor the level of correction to their needs instead of a static heavy-handed approach. The kindle book "Accurate Sound Reproduction Using DSP" is a vital resource in wrapping one's head around the science involved with DSP room correction and how to use Accurate, it's really the "missing manual".