No microphone feed, no converter, no session anywhere produces it directly. So where do these files come from, and what are you actually getting?

DSD, or more precisely PDM (Pulse Density Modulation), works by impressing an audio signal onto a running bit clock. The result is a continuous stream of 1s and 0s whose density is proportional to the amplitude of the original signal. There are no samples in it, and no digital values in the sense a PCM word stream contains them. When a DAC integrates that bitstream, it removes the clock and what remains is the audio.

That's what makes DSD unusual: it behaves like an analog storage medium, but one that can be stored and retrieved by a computer without loss. Tape and vinyl both impose their own artifacts, magnetic domains in one case, physical groove geometry in the other. DSD's artifacts are far fewer, and once the levels are named 1 and 0, the file survives copying indefinitely.

If there are no samples, nothing is being upsampled. What actually happens is modulation: a new bitstream is produced at the new rate. "Remodulation" is the term you'll see, though strictly it's just modulation, performed on a signal that has already been through the process once.

Every edition we sell is produced directly from the master the label supplies, and never from another edition we made. Each rate is a separate pass from that same master, so no edition carries another of ours stacked on top of its own.

What varies is the master itself. Most music that reaches us has been through post-production in DXD (352.8 kHz PCM), whether it was recorded in DSD or in DXD. What a label sends as its master may be that DXD edited master, or a DSD deliverable at whatever rate the label works in. A proportion of labels send the DXD edited master. Editions produced from it are a first-generation modulation, the original generation of the produced music. Editions produced from a DSD master are a second or later generation, and carry the character of the modulation that came before them.

Converting DSD to DSD directly is less lossy than routing it back through PCM and out again, which is why we take that path where the source is already DSD.

A higher rate contains no additional program bandwidth. The music in a DSD 512 edition is the music that was in the master, and nothing has been added to it.

What changes is where the modulation noise sits. Every doubling of the rate moves the noise envelope one octave higher, widening the clean band between the top of the music and the onset of that noise. This is the room a DAC's reconstruction filter needs: with more of it, the filter can be gentler, with less phase disturbance inside the audio band. That's the whole benefit, and whether it's audible depends on your DAC and your system. (And, your ears ;))

Most DACs convert an incoming PCM stream into PDM internally, using the chip's own on-board processing, before the sigma-delta modulator performs the actual conversion. That on-chip resource is limited compared to what's available to an offline 64-bit computer. Performing the conversion in advance, outside the DAC, has repeatedly proven in listening comparisons to sound more analog and more realistic than feeding the DAC the original PCM. Some high-end designs address this by bypassing the chip's internal interpolation with a dedicated processor alongside it.