Audio networking

Networked audio arrived in installed AV years before networked video, and it is now the default. A single network cable to a ceiling microphone can carry many channels of audio plus power and control, replacing a bundle of analog mic cables. Behind most of these systems are a few technologies — and one critical idea: everyone shares a clock.

The main technologies

Because Dante is so common, this lesson uses its terms. The ideas — devices, channels, subscriptions and a shared clock — apply to the others too.

Devices, channels and subscriptions

Each device offers transmit channels (audio it can send) and receive channels (inputs that can listen). Routing means subscribing a receive channel to a transmit channel, usually in software such as Dante Controller, by device and channel name rather than by IP address. Name devices clearly — “BRD-CeilingMic-1” beats a factory default — because subscriptions follow the names.

Audio travels as flows of packets. By default, Dante sends unicast flows to each receiving device; for one-to-many distribution it can use multicast, which brings in the IGMP considerations of Chapter 8.

Clocking: one leader for everyone

Digital audio is a stream of samples taken at a fixed rate — 48,000 per second at 48 kHz (Chapter 1). If two devices' sample clocks run even slightly differently, one gradually produces more samples than the other consumes, and the audio clicks or drops out. So networked audio devices elect one clock leader, and all others lock to it using the Precision Time Protocol (PTP), sent over the same network.

• An election chooses the leader automatically. You can bias it by marking one reliable, always-on device — often the DSP — as preferred leader.
• Devices must run at the same sample rate to exchange audio. A 96 kHz device can't subscribe to a 48 kHz one.
• PTP messages are why QoS matters: they need consistent, prompt delivery (Chapter 8's DSCP table).

Latency settings

Each receiving device holds incoming audio in a small buffer, set by its latency setting, so packets that take slightly different times to arrive can still be played out in step. Dante's default is 1 ms. Lower settings suit small, fast networks; larger networks with more switches between devices may need a higher setting. Packets that arrive later than the setting allows are lost, which you hear as clicks. Audinate publishes guidance on which settings suit which network sizes.

Redundancy

Many Dante devices have two network ports, primary and secondary. Connected to two fully separate networks, they send every audio stream over both, and a receiver carries on seamlessly if one network fails. The two networks must never be connected to each other.