It's tempting to read "Native DSD" on a DAC and picture the 1-bit stream flowing untouched to the analog output. Reality is messier, and it varies a lot by chip. None of this should scare you off DSD — but it's worth knowing what your converter is really doing.
ESS Sabre: through the PCM domain
In many ESS Sabre designs, a DSD input is converted into the PCM domain internally and run through the chip's own multi-bit delta-sigma modulator, with a selectable low-pass filter to tame DSD's ultrasonic noise. It skips the usual PCM upsampling filters, so it's legitimately a DSD mode — but the 1-bit stream is not handed straight to the analog stage. "Native" here is a softer claim than it sounds.
AKM: a more literal "Direct DSD"
Many AKM (Asahi Kasei) designs offer a DSD Direct path that routes the 1-bit stream to an analog switched-capacitor filter, bypassing PCM conversion and the digital DSP blocks. That's about as close to "the bits go to analog" as consumer DACs get — and a genuine architectural difference from the ESS approach.
R2R / ladder DACs: 1-bit doesn't fit
Ladder (R2R) DACs reconstruct sound from binary-weighted resistor networks, which inherently need multi-bit values. A 1-bit DSD stream doesn't fit that model at all, so an R2R DAC either uses an FPGA to convert DSD into high-rate multi-bit PCM for the ladder, or routes DSD down a separate dedicated 1-bit analog filter path. Either way, "DSD support" on an R2R DAC means something quite different again.
Multi-bit delta-sigma arrays
Some high-end designs (Benchmark is the well-known advocate) use parallel arrays of multi-bit converters rather than a 1-bit core, which lowers noise and eliminates the idle tones that 1-bit conversion is prone to. This is the hardware embodiment of the engineering critique discussed in is DSD really better?
| Architecture | What it does with DSD |
|---|---|
| ESS Sabre | Converts to PCM domain → its own modulator + low-pass |
| AKM "DSD Direct" | 1-bit stream → analog filter, bypassing PCM/DSP |
| R2R / ladder | FPGA-converts to multi-bit PCM, or a separate 1-bit path |
| Multi-bit Δ-Σ array | Parallel multi-bit converters; lower noise, no idle tones |
Getting in is separate from being decoded
Don't confuse the transport with the internal architecture. Whether DSD arrives as DoP (the norm on iOS) or as a native USB DSD stream is a separate question from what the chip does with it afterwards — and DoP is bit-transparent either way, as covered in DSD on iPhone.
Why DSD is fussy about clocks
One genuine DSD-specific quirk: because amplitude is encoded as pulse density over time, DSD reconstruction is unusually sensitive to clock jitter — timing errors translate fairly directly into distortion. It's why serious DSD DACs invest in low-noise clocks, reclocking buffers, and galvanic isolation from the noisy USB source.
The takeaway
You usually can't tell any of this from the spec sheet, and "native DSD" means different things on different boxes. The good news: ESS, AKM, and well-executed R2R DACs all sound excellent in practice. Implementation beats architecture, and listening beats the label. Choose a DAC because you like how it sounds — not because of which DSD path the marketing highlights.
That's the last of the deep dives. Loop back to the foundations in DSD vs PCM, or the verdict in is DSD really better?