Typical Dante Network Design and Recommendations

There are four typical topologies for Dante networks.

NOTE; The Secondary Dante Jack(s) must be programmed to "Redundant Network" before wiring the system for redundancy. If they are left on "Switched Port", and wired for redundancy, the Dante card may become corrupt. If a factory reset is performed on the units, it will revert them back to "Switched Port". Make sure to disconnect all Dante connections before performing the reset.

10 or Fewer Units Without Redundancy

For systems with up to 10 units without redundancy, connect your PC to the Ethernet port on one unit, then daisy-chain the remaining Ethernet ports. Then, daisy chain the Dante ports. No specialized Dante switches or configuration is required. Units should be in Switched Port mode.

More Than 10 Units or More Than 100 Meters Between Units

For systems with more than 10 units or distances between units more than 100 meters, connect your PC and the Ethernet ports on all units to one Ethernet switch, and connect the Dante primary ports to a second Ethernet switch. Units should be in Switched Port mode.

Two Units with Redundancy

For systems with two units with redundancy, connect your PC to the Ethernet port on one unit. Then, connect the Ethernet ports on both units together. Connect the primary Dante ports on both units together, and connect the secondary Dante ports on both units together. Units should be in Redundant Network mode.

More Than 2 Units with Redundancy

For redundancy in systems using more than two units, additional networking hardware is required. Connect your PC and the Ethernet ports on all units to Ethernet switch 1. Connect the Dante primary ports on all units to Ethernet switch 2. Connect the Dante secondary ports on all units to Ethernet switch 3. Units should be in Redundant Network mode.

For complete fail-safe operation, at least two uninterruptible power supplies, or UPS, should be used. One UPS for units' primary power, PC, and Ethernet switches 1 and 2, and a second UPS for units' auxiliary power and Ethernet switch 3.

Latency

The Dante Network Latency is set in the Network I/O Manager found in the Tools Menu.

Network latency increases with the number of switch hops. Use fewer than 2 hops for the best latency. Up to 10 hops total are allowed but with increased latency. Always use the largest practical value for the latency. In many installations latency isn't critical, like for example when sending audio to an acoustically isolated room. In these cases, choose the highest latency to reduce overall network traffic and minimize the chance audio drop-outputs on overburdened networks. If low latency is critical, select the value that corresponds to the number of switch hops in your entire network. If daisy-chaining two units together or connecting them with a single gigabit switch, 0.25 mS is typically safe. Otherwise, choose 0.5 or 1.0 mS depending on your network topology.

In Dante, variation in latency in the network is compensated for at the receiver. Each receiver has an Rx latency setting (again, set in the Network I/O Manager for Symetrix Units). This setting defines the latency between the timestamps on the incoming audio samples and when those samples are played out.

A typical default latency for most Dante devices is 1 mS. This is sufficient for a very large network, consisting of a Gigabit network core (with up to 10 hops between switches) and 100 megabit links to Dante devices. Smaller, Gigabit-only networks can use lower values of latency (down to below 200 µS).

Note: The Dante Latency is a "system-wide network latency setting". This means, there is no additional latency or skew each subsequent switch hop, i.e., the latency is not additive. As long as the (system-wide network) latency set is long enough to traverse the entire network path, the packets will get to the end of the chain before the buffer expires. Then, Dante uses its clock alignment mechanism (Precision Time Protocol) to exactly line up the outputs down to the sub-sample level. You can compare the word clocks between any two Dante units and they should be within 100 nS of each other, or less than 1% of a sample period (verified with up to 10 units). So the outputs at multiple points in the daisy chain will all be sample aligned, each with the specified latency from the sender.

Note 2: The actual Dante latency will be 3 samples or 0.06 mS longer than the time indicated in the Dante latency settings in Composer. This is due to the sample buffering our DSP does and is unavoidable.