DSD is often sold as a pure, unbroken 1-bit chain: the microphone feed captured as DSD and delivered to you untouched, "closer to analog" than PCM could ever be. It's a lovely story. For most DSD releases, it isn't what happened — and understanding why makes you a much smarter buyer.
The problem: you can't actually work in 1-bit
A DSD stream is a torrent of single bits whose density encodes the wave (see DSD vs PCM). That structure is wonderful for playback and hopeless for production. Almost everything an engineer does to a recording —
- adjusting a volume fader,
- applying EQ or any effect,
- crossfading or editing between takes,
- mixing multiple microphones or tracks together —
requires multi-bit arithmetic. Multiply a 1-bit sample by 0.8 to turn it down and the answer is no longer expressible in one bit. You simply cannot mix or edit a true 1-bit stream. DSD is a delivery format, not a working one.
The workaround: DXD
So the industry built a high-resolution PCM workspace for DSD production: DXD (Digital eXtreme Definition) — 24-bit PCM at 352.8 kHz, eight times the CD rate. Introduced around 2004 by Merging Technologies for its Pyramix workstation and Sphynx converters, DXD gives engineers the multi-bit headroom to edit, mix, and master, while staying at a high enough rate to preserve the character of the capture. The finished work is then modulated to DSD for release.
In a typical hybrid workflow, the editor works on a DSD timeline, but any segment that's actually processed is converted to DXD on the fly, worked on, and re-modulated back to DSD; untouched segments stay DSD. If there's real mixing to do, the whole project runs in DXD.
Why engineers minimise the round-trips
Each modulation back to DSD runs a high-order noise shaper, and each pass adds a little more ultrasonic noise — a "photocopy of a photocopy" effect if you do it repeatedly. The SACD specification (the "Scarlet Book") even caps how much accumulated high-frequency noise a release may contain, which is exactly why careful studios keep conversions to a minimum and do their dithering only at the final export.
So what about "pure DSD"?
It does exist — just narrowly. A pure-DSD recording means capturing straight to DSD with essentially no mixing or processing: a well-placed pair of microphones, a live take, out to disc. That suits solo piano, small jazz combos, choral and chamber works, and the audiophile labels (Blue Coast, parts of the 2L and Channel Classics catalogues) that build their identity around it. For anything needing multitrack mixing, overdubs, or heavy editing — i.e. most music — a pure 1-bit path is simply not possible.
What this means for you
This feeds directly into the big question: is DSD really better? And if you're curious how the 1-bit stream is handled at the other end, see inside your DAC.