I also learnt that my upstairs FiiO R7 Android 10-based server had zero issues with my new .dsf files whereas Audirvana Studio on my downstairs iMac couldn't even see the same folders much less play them. It took converting the files to .dff which rendered them visible and playable. Being relatively new to native high-res DSD streaming, that was a new lesson learnt.

Here is the FiiO setup with native DSD128 file playback. Because AES/EBU can't handle that file density—it will transmit DSD64 just fine—my 2m USB cable couldn't reach from the R7's usual place next to my seat where I can easily navigate its touchscreen. So the compact SD-card server piggybacked the Nagra instead. With the mouse-over loupe function, you'll see that the playing track packed a file density of 11'289kbps. Contrast that to CD's 1'411kbps. DSD128 is nearly 10 x fatter.

Here comes my downstairs…

…HifiMan Susvara headfi setup. Source was my usual music iMac with Audirvana Studio in exclusive hog mode streaming local files off 4TB external SSD. This output USB into a Singxer SU-6 ultra-cap powered USB reclocker. To A/B against our Cen.Grand DSDAC 1.0 Deluxe, the Singxer fed each DAC its own S/PDIF signal in parallel. The DACs then both hit a Vinnie Rossi L2 Signature linestage with Elrog ER50 direct-heated—and here direct-coupled—power triodes in a grounded-grid circuit; plus remote volume and source switching. The L2 fed a Kinki Studio THR-1 lateral Exicon Mosfet amp, its own passive pot bypassed. Remotely switching preamp inputs, I could compare on the fly with the highest-resolution conventional headphones of our collection. Given S/PDIF's 192kHz ceiling, for native DSD here I was limited to 64.

To segue into Sonic Street on Susvara, let's open the gates with a trick question. Is there such thing as vintage/modern DSD sound as there is a vintage 300B aesthetic from bottle purveyors like Western Electric, TJ/Full Music and Takatsuki, a modern equivalent pursued by Elrog, Emission Labs and KR Audio? Before you flag me down insisting that it's the circuit and not tube brand which sets the sound, I mean rolling tubes in the same circuit. It's what I did when I still owned SET amps. It did change the sound and roughly fell into what I refer to as vintage/modern camps. The former was softer, more billowy, decay rich and midrange focussed, the latter firmer, more damped and extended in both directions. In short, the modern glass borrowed more from the generic 'transistor' aesthetic. Obviously no such classification exists. It's simply a useful crutch when discussing sonic flavours so everyone is on the same page.

As to a page from my own black book, here is Audirvana's 'audio' settings window with the Cen.Grand and Nagra DACs connected. Audirvana embeds two different up/resampling engines called r8brain and SoX. For an even comparison, I swapped USB leads to feed both DACs PCM resampled to DSD128 via either r8brain or SoX, then native unprocessed PCM. I wanted to compare on-Mac vs off-Mac resampling. That gave me three modes per converter¹. As you'd expect if you've experimented along those lines, different up/over/resampling algorithms do not sound alike. The same goes for signal routing software like Audirvana, JRiver, Roon & Co. Such software makes a difference yet is far cheaper than hardware. We obviously need hardware which to run such software on.
________________________

¹ Actually, with the Cen.Grand settable to 128, 256, 512 or 1'024, I would have had a lot more but did that dance when I first bought it. Here I ran it just at DSD1'204.