SMPTE ST 2110 — Professional Media Over IP

SMPTE ST 2110 is the industry standard for transporting professional media (video, audio, and ancillary data) as separate, synchronized IP streams over standard Ethernet infrastructure. Unlike earlier AV-over-IP approaches that encapsulate an entire HDMI or SDI signal into a single stream, ST 2110 treats video, audio, and metadata as independent flows — each with its own IP multicast address, RTP stream, and timing reference. This separation enables the flexibility of IT infrastructure (standard 10GbE/25GbE switches, IP routing, software-defined networking) with the determinism and synchronization quality required for broadcast production. ST 2110 is the successor to SDI in professional production facilities, large-venue AV systems, and broadcast trucks, and is increasingly specified for high-end installed AV where uncompressed quality and low latency are required.

ST 2110 Suite Architecture

ST 2110 is a family of sub-standards, each addressing a specific media type:

A complete ST 2110 system deploys all relevant sub-standards simultaneously. A camera sends three separate streams: a 2110-20 video stream, a 2110-30 audio stream, and a 2110-40 ancillary stream. A production switcher receives and independently routes each.

ST 2110-20 — Uncompressed Video

Bandwidth Requirements

ST 2110-20 carries uncompressed progressive or interlaced video in RTP packets. Bandwidth is determined by resolution, frame rate, and bit depth:

A single 4K60 10-bit ST 2110-20 stream requires a 10 Gbps Ethernet link dedicated to that stream. Multiple streams or 4K12 require 25GbE or 100GbE uplinks. This is the primary reason ST 2110 infrastructure uses 10GbE/25GbE leaf-spine switching, not the 1GbE switches common in AV-over-IP systems.

Pixel Sampling and Color Space

ST 2110-20 supports:

• YCbCr 4:2:2 (most common): Standard broadcast color space, compatible with all professional monitoring
• YCbCr 4:4:4: Full color sampling, used for graphics and compositing
• RGB 4:4:4: Full linear RGB, used in production workflows and LED processing
• 10-bit and 12-bit depth: 10-bit is standard; 12-bit for HDR mastering workflows

ST 2110-21 — Traffic Shaping

Raw uncompressed video generates extremely bursty traffic — a 1080p60 frame is transmitted in the inter-frame gap (~16.7 ms), creating a large bandwidth spike per frame. ST 2110-21 defines sender profiles that shape the output to avoid congestion:

• Narrow (N) sender: Traffic spread evenly over the frame period. Minimal burstiness. Requires more sender buffering. Preferred for switch infrastructure.
• Wide (W) sender: More bursty, up to 1.5× the average rate. Requires larger switch buffers.
• Network Compatibility Model (NCM): Traffic shaping optimized for IP network switches.

Switch selection for ST 2110 must account for sender profiles — verify the switch buffer size supports the sender profile used in the system.

ST 2110-30 — PCM Audio (AES67)

ST 2110-30 defines PCM audio transport and is fully compatible with AES67. Each audio stream is an independent RTP flow with its own multicast address. Key parameters:

• Packet time: 1 ms (most common), 0.125 ms, 0.25 ms, 4 ms — match across all devices in the system
• Sample rate: 48 kHz (standard), 96 kHz supported
• Bit depth: 24-bit standard
• Channels per stream: 1–8 channels (common: 2-channel stereo streams, or 8-channel multichannel streams)
• Clock sync: IEEE 1588 PTP (same grandmaster as ST 2110-20 video)

Audio streams are independent of video streams. A single camera may produce one video stream (2110-20) and four audio streams (2110-30): stereo mix, embedded mic 1, embedded mic 2, timecode audio. The production switcher routes each independently.

See glossary/aes67 for AES67 interoperability details and networking/dante for comparison with Dante audio networking.

ST 2110-40 — Ancillary Data

ST 2110-40 carries ancillary data that traditionally rode in the SDI VANC (Vertical Ancillary Data Space): timecode, captions (CEA-608/708), AFD (Active Format Description), and custom metadata. Ancillary data is mapped to RTP packets with DID/SDID (Data Identifier/Secondary Data Identifier) values preserved from SDI.

In AV integration, 2110-40 is critical for:

• SMPTE timecode (LTC, VITC) synchronization across the system
• Closed captions passing through the video chain
• Teleprompter feeds and prompter metadata

PTP — IEEE 1588 Precision Time Protocol

PTP synchronization is the foundation of ST 2110 timing. All ST 2110 devices must lock to a common PTP grandmaster. Without sub-microsecond clock synchronization, the independent video, audio, and ancillary streams cannot be correctly reassembled at the receiver.

PTP Requirements for ST 2110

• PTP domain: All devices must use the same PTP domain (commonly domain 127 for broadcast)
• Grandmaster accuracy: ≤ 1 µs accuracy required; GPS-disciplined grandmasters (Meinberg, Microsemi, Imagine Communications) provide nanosecond accuracy
• Switch support: Switches must be PTP-aware (boundary or transparent clock) — a switch that does not forward PTP accurately introduces timing errors that cause audio/video sync failures
• Redundant PTP: Production systems use dual redundant PTP paths with SMPTE ST 2059 (ST 2110 timing reference) for failover

See glossary/ptp for full PTP architecture coverage.

NMOS — Networked Media Open Specifications

NMOS (developed by AMWA — Advanced Media Workflow Association) is the open API standard for discovering, registering, and connecting ST 2110 devices on a network. Without NMOS, ST 2110 connections require manual SDP (Session Description Protocol) file exchange and static multicast group management — impractical at scale.

Key NMOS Specifications

IS-04 (Discovery and Registration): Devices register themselves with an NMOS Registry. Operators query the registry via REST API to discover all senders and receivers on the network. The registry provides a real-time inventory of all ST 2110 sources and destinations.

IS-05 (Device Connection Management): REST API for making and breaking connections between senders and receivers. IS-05 is the equivalent of a routing matrix command for ST 2110 — instead of pressing a button on a hardware router, a control system sends an HTTP PATCH to connect a sender to a receiver.

IS-07 (Event and Tally): Tally and event data exchange — camera tally lights, production switcher state, GPI/GPO.

IS-08 (Audio Channel Mapping): Channel-level routing within audio streams.

NMOS-compatible products include: Sony cameras and monitors, Grass Valley production switchers, Nevion SDN controllers, Matrox Media Gateway, Riedel Medior Nodes, and others. Verify NMOS IS-04/05 support in product specs — ST 2110-capable hardware does not automatically include NMOS.

Infrastructure Requirements

Switching

ST 2110 requires managed 10GbE (minimum) or 25GbE/100GbE Ethernet infrastructure. Key switch requirements:

• Cut-through forwarding (not store-and-forward): Reduces latency from ~10 µs to ~1–2 µs per hop. Required for tight timing margins in production systems.
• IGMP snooping v3: ST 2110 uses SSM (Source-Specific Multicast) — IGMP v3 is required (not IGMP v2)
• PTP boundary/transparent clock support: Verified PTP accuracy through the switch fabric
• Jumbo frames (9000 byte MTU): Some ST 2110 video frames exceed 1500 byte MTU; jumbo frames reduce fragmentation overhead
• Low jitter: Switch port-to-port jitter must be < 1 µs for PTP accuracy

Recommended switches: Arista 7050 series, Cisco Nexus 93180/9336, Juniper QFX series. Consumer and prosumer managed switches (Netgear, Ubiquiti, Cisco SG series) are not suitable for ST 2110.

Redundancy

Professional ST 2110 systems implement SMPTE ST 2022-7 seamless protection switching — dual redundant network paths (separate switches and cabling) with the receiver selecting the best stream from two parallel paths. No traffic interruption on single-path failure. Requires all devices to support 2022-7 and two physically separate network paths from every sender to every receiver.

ST 2110 vs. SDI

ST 2110 enables flexible remote production (REMI) — cameras at a venue send ST 2110 over a WAN to a production center anywhere in the world. This is not practical with SDI.

Common Pitfalls

• Using store-and-forward switches instead of cut-through. Store-and-forward switches buffer the entire frame before forwarding, adding ~10–50 µs of latency per hop. In a deep spine-leaf topology with 4–6 switch hops, store-and-forward latency exceeds the PTP synchronization budget and causes audio/video timing errors that manifest as lip sync drift. Fix: verify all switches in the ST 2110 path use cut-through forwarding; check the switch spec sheet — not all 10GbE switches support cut-through.

• PTP grandmaster not GPS-disciplined in production systems. A free-running software PTP grandmaster (Linux with ptpd) drifts over time and introduces clock wander that causes audio clicks, video frame slips, and synchronization failures hours into an event. Fix: use a GPS-disciplined hardware PTP grandmaster (Meinberg M300, Microsemi TimeProvider 4100) for any production system; verify grandmaster accuracy in PTP statistics (offset < 100 ns).

• IGMP v2 configured on switches in an ST 2110 system. ST 2110 uses Source-Specific Multicast (SSM, RFC 4607), which requires IGMP v3. IGMP v2 only supports Any-Source Multicast (ASM) and does not carry the source IP in join messages. With IGMP v2, switches cannot properly constrain multicast to interested receivers, causing all 2110-20 streams (potentially 12 Gbps each) to flood all switch ports. Fix: configure IGMP v3 snooping on all switches; verify with show ip igmp snooping groups or equivalent.

• Missing NMOS registry causing manual SDP management at scale. Without NMOS IS-04/05, connecting 20 cameras to 10 monitors requires manually distributing SDP files and entering multicast group addresses on each device. Any change requires manual reconfiguration of all affected devices. Fix: deploy an NMOS registry (AMWA reference implementation, Sony, or Grass Valley) before commissioning; verify all devices support IS-04 registration and IS-05 connection management.

• Insufficient switch bandwidth for multiple 4K streams. A 10GbE uplink saturated by a single 4K60 stream (12 Gbps) leaves no headroom for additional streams or retransmission. A leaf switch with 8 cameras (each 4K60 = 12 Gbps) requires 100 Gbps uplinks to the spine — not 10 GbE. Fix: calculate total bandwidth before specifying infrastructure; use 25GbE for server-to-leaf connections and 100GbE spine uplinks in 4K systems.