Field notes · Broadcast engineering

A Practical Guide to SMPTE ST 2110 in Major Live Productions

SMPTE ST 2110 is the de facto standard for IP-based live production, but the spec itself doesn't tell you what to do at 3 a.m. when a venue's PTP grandmaster drifts and four world feeds start tearing. These are notes from rolling out 2110 on OBS and HBS productions at the Olympic Games and other top-tier live events.

What ST 2110 actually is

ST 2110 is a family of standards that carries video, audio, and ancillary data as separate elementary streams over IP:

  • ST 2110-20 — uncompressed video essence.
  • ST 2110-30 / -31 — PCM audio (AES67) and AES3 transparent transport.
  • ST 2110-40 — ancillary data (closed captions, AFD, timecode, SCTE).
  • ST 2110-21 — traffic shaping for senders (narrow / narrow-linear / wide).
  • ST 2110-10 — system timing, built on PTP (IEEE 1588 / SMPTE 2059).

The decoupling is the whole point: a single UHD source can be carried as one video flow, sixteen mono audio flows, and an ANC flow, each routed independently across a leaf-spine fabric.

PTP is the production

In SDI, sync came from a black-and-burst or tri-level reference and you never thought about it. In 2110, PTP is the reference, and a misbehaving grandmaster takes the whole compound off air. On OBS deployments we standardise on:

  • Two GPS-disciplined PTP grandmasters with BMCA failover, locked to SMPTE 2059-2 profile.
  • Boundary clocks on every leaf — never end-to-end transparent clocks across the spine.
  • Dedicated PTP VLAN, with PTP-aware switches and hardware timestamping enabled.
  • A passive PTP monitor on a span port, alarming on offset > 500 ns or path delay variance.

The single biggest cause of "mystery" 2110 faults at a venue is PTP, not the media network. Build the monitoring before you build the routing.

Network design: ST 2022-7 by default

Run two physically separate networks (red / blue), and require every sender and receiver to support ST 2022-7 seamless protection switching. A 1080p59.94 video flow is around 2.97 Gb/s; a UHD flow is ~12 Gb/s. Plan for:

  • 25 GbE to edge devices, 100 GbE spine, non-blocking fabric.
  • SSM / PIM for IGMPv3 source-specific multicast — never any-source.
  • Strict ACLs separating media, control (NMOS), PTP, and management VLANs.
  • Pre-calculated multicast address plan; don't let devices pick their own.

NMOS: discovery, connection, audit

ST 2110 says how bits move; AMWA NMOS says how you find and connect them. At minimum you want:

  • IS-04 for registration and discovery (run two registries, mDNS for fallback).
  • IS-05 for connection management — this is what your broadcast controller drives.
  • IS-08 for audio channel mapping (essential the moment you have MADI-style sources).
  • BCP-002-01 / BCP-004-01 for grouping hints and receiver capabilities.

On HBS productions the broadcast controller (e.g. VSM, Cerebrum) becomes the single source of truth — operators never touch IS-05 directly. Make sure every device's NMOS implementation actually honours activation timestamps; some vendors silently round to the next frame.

Gotchas from Olympic-scale deployments

  1. Sender packet pacing. A "narrow" 2110-21 sender on paper is sometimes "wide" in firmware. Validate every model on a packet analyser before the venue load-in.
  2. ANC timing. 2110-40 packets must arrive ahead of the corresponding video frame. Mux-then-route workflows from SDI islands often break this; insert ANC at the IP edge instead.
  3. Audio shuffles. A 16-channel embedded SDI source becomes 16 individual 2110-30 flows. Without IS-08 mapping, an operator will eventually patch the wrong commentary pair to air.
  4. Multicast joins at scale. Receivers joining and leaving hundreds of flows during a rehearsal can overwhelm IGMP snooping on older switches — pre-warm joins where possible.
  5. UHD HDR pass-through. Mixing HLG and PQ sources on the same 2110 fabric is fine, but downstream converters need explicit colour-volume metadata; don't rely on VANC alone.

A pre-show checklist

  • PTP offset on every device < 1 µs, path delay variance logged for 24 h.
  • Red / blue network parity verified by pulling each leg under load.
  • Every sender flow validated against ST 2110-21 with EBU LIST or equivalent.
  • NMOS registry failover tested with live flows up.
  • Broadcast controller salvo tested end-to-end, including audio shuffle.
  • Recording and multiviewer paths confirmed as 2022-7 protected.

When SDI is still the right answer

2110 is the right architecture for a main international broadcast centre with hundreds of sources and complex routing. For a small ENG flyaway or a single OB van doing one event, 12G-SDI is still faster to rig, cheaper to crew, and impossible to misconfigure with a typo in a multicast address. Pick the standard that matches the scale of the production, not the brochure.

Written by Evangelos Karomenos — international broadcast engineer (TOC / MCR / Vision) with credits including OBS and HBS at the Olympic Games. See full profile →