Streaming & Playback
Streaming & Playback: the path from network to DAC
Transport, clocking and output-stage architecture for Stream One — documented as architecture, not invented curves.
Published June 20, 2026
1 — The question
A network streamer is a computer doing a job computers are bad at: delivering a perfectly timed stream of samples to a converter while surrounded by switching supplies, radios and processors. Which architectural decisions actually protect the audio path — and which are marketing?
2 — Why it matters
Digital playback errors rarely sound like clicks. They surface as clock phase noise feeding the DAC, ground loops between boards, or supply noise riding on the analogue output stage. The defects live in the plumbing, so the plumbing is what we design and document.
3 — Our approach
- Separate the network/transport board from the audio board; connect them across an isolated boundary.
- Master clock placed next to the DAC; the network side is the slave, never the other way around.
- Open-standards transports (AirPlay, UPnP/DLNA, Bluetooth receive, USB and NAS playback) so the architecture outlives any single ecosystem.
- Publish the playback architecture diagram; publish performance graphs only when a defensible source-measurement procedure exists.
4 — What we have learned
Buffer depth matters less than clock domain placement. Moving the master clock away from the network board’s switchers reduced measured output jitter on the development rig — a result we will publish with the source measurement procedure.
5 — Open questions
- Defining a public test procedure for streamers: what actually predicts audible behaviour in a source?
- Clock behaviour when AirPlay and UPnP transports share one board.
6 — Related work
- Measurement archive: Streamer — coming soon
- Product: Network Streamer