Corporate AV teams continue shifting from HDBaseT and point-to-point fiber to Dante AV Ultra for campus-wide distribution. The protocol now supports 4K60 video plus embedded audio at under 2 ms glass-to-glass when the network is properly tuned. In one 40-zone deployment covering four buildings, the integrator used Audinate’s Dante AV Ultra encoders at each source and decoders at displays, paired with Cisco Catalyst 9300 switches running PTP boundary clock mode.
Latency stayed consistent because the team placed a GPS-locked grandmaster on a dedicated timing VLAN rather than letting edge switches compete for the role. Each video stream consumed roughly 850 Mbps, so the design limited any single 10 GbE uplink to six simultaneous streams. Bufferbloat appeared during early tests when audio and video shared the same VLAN; moving audio to a second VLAN with strict priority queuing dropped jitter from 340 µs to 85 µs.
Installation economics reflected the extra configuration time. The project required 22 days of network commissioning instead of the 12 days originally budgeted for a simpler multicast setup. Switch licensing for advanced PTP and IGMPv3 snooping added $18,000, yet the integrator avoided a second fiber backbone by reusing existing 10 GbE links once traffic classes were isolated.
Clock Hierarchy and VLAN Segmentation Practices
Clocking stability depended on defining three strata: a primary grandmaster in the main data center, two boundary clocks per building, and ordinary clocks at every encoder and decoder. The team disabled automatic BMCA on all endpoints and statically assigned clock roles through Dante Controller templates. This approach eliminated the 4-second dropouts that occurred during switch reboot tests when multiple devices attempted to claim master status.
VLAN segmentation followed a simple rule set: one VLAN for Dante AV Ultra video, another for Dante audio and control, and a third for IT data. Multicast boundaries were enforced at the distribution layer so that a single 4K stream did not propagate beyond its intended building. The integrator documented every port with its VLAN membership and QoS trust settings, then exported the configurations into a revision-controlled repository used by the client’s operations staff.
Future campus expansions will likely add Dante AV Ultra sources in conference rooms and collaboration spaces. Integrators planning similar projects now budget for additional boundary clocks and pre-stage VLAN templates so that new zones can be brought online without re-testing the entire PTP domain.
QoS policies assigned strict priority to PTP messages on the timing VLAN while mapping Dante AV Ultra video to assured forwarding and embedded audio to expedited forwarding. DSCP remarking at the edge prevented any IT-sourced packets from disrupting AV classes. IGMPv3 snooping with a distributed querier kept multicast domains limited to each building, and stress tests confirmed that 40 concurrent 4K streams never pushed any 10 GbE uplink above 65 percent utilization.
After handover, custom scripts poll Dante Controller via API every minute to report clock offset, stream status, and buffer occupancy. These values feed directly into the client’s existing NMS, enabling operations staff to address drift before lip-sync errors become visible. Final measurements showed 1.6 ms average glass-to-glass latency and 1.9 ms worst-case across all 40 zones.
The same VLAN templates and boundary-clock configurations are now pre-staged for upcoming 8K encoder rollouts. With current uplinks running at roughly 60 percent capacity, the network retains headroom for higher-bitrate streams without new fiber. The documented revision-controlled switch files have already allowed three additional conference-room zones to be commissioned remotely in a single day.
Installers began by running Fluke DSX-8000 cable certification on all existing Cat6a runs and 10 GbE fiber links before loading the Cisco 9300 configs. Each encoder received a static PTP role assignment via Dante Controller XML import rather than manual entry, cutting per-device setup from eight minutes to under ninety seconds. Port profiles were cloned across switches using a Python script that referenced the master spreadsheet of VLAN IDs and DSCP markings.
During commissioning, teams captured PTP sync packets with a dedicated monitoring port on each distribution switch and reviewed offset graphs in Wireshark before enabling video streams. This step caught two mislabeled fiber strands that would have created a timing loop. The final sign-off checklist included verification that no endpoint exceeded 150 µs offset under full load.
