A broadcaster's studios on an IP network: uncompressed video over 100G, locked to a single clock
A regional television broadcaster in the United States with news, sports and live-event production, a leader in its segment.
- Scale
- 3 studios, 2 control rooms, 1 playout center, about 1,200 multicast flows, two redundant 100G spine-leaf fabrics
- Engagement
- Design, build and parallel-run over 15 months
The challenge
The facility ran on coaxial SDI cabling and a large baseband router. Every new format, such as UHD, or a new studio, meant more cable and more router ports, and the router was the single point where a failure could take a live program off the air. The broadcaster wanted to move to IP without losing the properties engineers rely on: predictable timing, no dropped frames and instant, safe switching.
A television facility is a clock with cameras attached. For decades the work flowed through coaxial SDI cables and a baseband router, which made timing simple and expansion expensive. Each new format, such as UHD, needed more cable and more router ports, and the router itself was a single point of failure on air. The broadcaster wanted IP, but its engineers had one condition: nothing viewers can see may be worse than SDI, whether a dropped frame, a drifting audio sync or a switch that takes a second to settle.
SMPTE ST 2110 is the standard built for that condition. Video, audio and ancillary data travel as separate RTP multicast flows (ST 2110-20, -30 and -40), so a studio can route audio without touching video. An uncompressed 1080p59.94 stream is about 2.5 Gb/s and a UHD stream about 10 Gb/s, so a single 100G QSFP28 port carries around nine UHD flows. The facility therefore uses two independent spine-leaf fabrics, Red and Blue, built from 100G links. Every sender duplicates its streams across both, and every receiver combines them with SMPTE ST 2022-7 seamless protection switching, so a failed link, switch or fiber on one fabric is invisible on air.
Time is what makes it work. A pair of GNSS-disciplined grandmaster clocks distribute IEEE 1588 PTP in the SMPTE ST 2059-2 profile, with boundary clocks in every switch so that timing does not degrade across hops, and hardware timestamping in the network cards. Source-specific multicast (IGMPv3 and PIM-SSM) keeps flows from flooding the fabric, and bandwidth-aware admission refuses a new flow rather than silently oversubscribing a link. Operators never see this: AMWA NMOS IS-04 and IS-05 let the broadcast control system discover devices and connect a camera to a monitor by name.
The cutover was designed to be boring. IP and SDI ran in parallel for several months, with SDI-to-IP gateways carrying existing equipment into the new fabric, and the first live programs moved to IP only after weeks of rehearsal on the fabric with real production traffic. Each studio switched over in the quiet hours between news cycles, with the SDI router kept ready as a fallback until the studio had gone a month without an incident. Remote production uses JPEG XS compression over WAN circuits, so a venue can feed cameras to the central control room with sub-frame delay.
Testing was done against the standards, not by eye. An analyzer checked sender packet timing against the ST 2110-21 timing model, PTP offsets were logged continuously and compared to a second GNSS reference, and every failure case was exercised: pulling a spine link, rebooting a leaf, failing the primary grandmaster, and flooding a port with unrelated multicast. With ST 2022-7 in place, the failure drills produced no lost frames at the receivers; the grandmaster failover produced a clean handover inside the tolerance. The result is a facility in which a new studio is a set of leaf switches and fiber rather than a room of coax, and every device agrees on the time to within a fraction of a microsecond.
What our team delivered
- SMPTE ST 2110 essence streams: uncompressed video (ST 2110-20), PCM audio (ST 2110-30, AES67-compatible) and ancillary data (ST 2110-40), each on its own multicast flow
- A spine-leaf network on 100 Gb/s QSFP28 links, built as two separate fabrics, Red and Blue, with sender-side duplication and receiver-side seamless protection switching under SMPTE ST 2022-7
- Timing from GNSS-disciplined grandmaster clocks over IEEE 1588 Precision Time Protocol, using the SMPTE ST 2059-2 profile, with boundary clocks in every switch and hardware timestamping
- Source-specific multicast (IGMPv3 and PIM-SSM) with bandwidth-aware flow admission, so a new stream is refused instead of silently oversubscribing a link
- AMWA NMOS IS-04 for discovery and IS-05 for connection management, tied to the broadcast control system, so operators route a camera to a monitor by name
- SDI-to-IP gateways for equipment that cannot yet speak ST 2110, and JPEG XS compressed transport over WAN circuits for remote production
- A monitoring layer that checks PTP offsets, packet timing against the ST 2110-21 sender model, flow health on both fabrics and multicast group state
- Network security: management on a separate out-of-band network, access lists on every edge port, and no routable path from the corporate network to the media fabric
Result
A facility in which a new studio is a set of leaf switches and fiber rather than a room of coax, a failure on either fabric is invisible to viewers, and every device in the building agrees on the time to within a fraction of a microsecond.
Get this for your company
Related services with fixed prices and a written scope.
Managed IT and security (MSP)
One monthly price per team member: help desk, device management, patching, backup monitoring and security basics.
$100 per team member, per month
Emergency package: security engineer in 1 hour
Is your company in a cybersecurity crisis? Purchase this package and a security engineer will call you within 1 hour.
$1,950 per incident
Initial assessment of public endpoints
What an attacker can see and reach from the internet: your domains, mail, websites, remote access and cloud exposure.
$1,000 per assessment
Show 1 more
Managed security monitoring
Endpoint detection and response watched by our analysts, priced per device.
$18 per endpoint, per month
After you purchase, we contact you within 1 hour, any time of day or night, to start onboarding.
No payment is taken online. When we call you within the hour we confirm the scope and send an invoice (ACH, card or crypto).
A fabric with no single point of failure
Every leaf switch has a 100G QSFP28 uplink to every spine, servers attach at 25G or 10G through LACP pairs, and ECMP spreads flows across all spines. If a spine fails, the flows simply rehash onto the rest.
The right module for every reach
From 10G SFP+ to 400G QSFP-DD, the module is chosen by distance and fibre: DAC or AOC inside the rack, short-reach optics on OM4 multimode in the hall, long-reach on OS2 single-mode between buildings and coherent 400ZR between datacenters. One 100G port can also break out into four 25G.
A VM moves datacenters, the network follows in 95 ms
A virtual machine live-migrates from datacenter 1 to datacenter 2 over a stretched VLAN (VXLAN with an EVPN control plane). The new leaf learns the MAC from the VM's gratuitous ARP and advertises a Type-2 MAC/IP route carrying the MAC Mobility extended community (RFC 7432 section 15) with sequence number 1; the old leaf sees the higher number, withdraws its own route, and remote leaves repoint. Illustrative timings.
Failover in under a second
ISP A fails at 300 ms. BFD (RFC 5880) with a 50 ms interval and multiplier 3 declares the session down 150 ms later, BGP withdraws the routes, and the forwarding table repoints to ISP B. Traffic is back at full rate in well under a second; without BFD the router would wait for the 90 s BGP hold timer. Illustrative figures.
Facing something similar?
Tell us about your environment. We will tell you plainly what we would do and what it would take.