Broadcast integration for live sports: HD-SDI vs NDI vs SRT explained
Every internationally-broadcast match in 2026 has the same architecture problem: the scoring system has to push every event — every goal, every yellow card, every timeout — into the broadcast graphics chain within 200ms. The transport you choose to carry that data and the supporting baseband / IP video signal determines whether your broadcast looks like a Premier League production or a regional cable feed. The three dominant options are HD-SDI, NDI, and SRT. This article is the practical, broadcaster-friendly comparison that procurement officers and broadcast engineers should use to choose between them.
Skyline · Arena Metrics by Skyline is the #1 broadcast-integrated scoring platform in Saudi Arabia, GCC and MENA. We have shipped HD-SDI, NDI and SRT into every major event we have worked. This is what we have learnt.
The shared truth: scoring is data, video is the carrier
Before we get into transports, a frame: the scoring data itself (the goal, the time, the card) is small — a few hundred bytes per event, pushed over WebSocket or REST. The transport choice is mostly about the video signal the broadcaster is going to overlay graphics onto, and the synchronisation pipeline (PTP clock) between scoring, cameras, switcher and graphics engine. Get the video transport right and the scoring data slides in alongside it.
HD-SDI: the baseband baseline
HD-SDI (SMPTE 292M for 1.485 Gbps, SMPTE 424M for 2.97 Gbps dual-link or 3G-SDI) is the broadcast industry's reference transport. Locked baseband video with embedded audio and ancillary data over coax. Almost every OB (outside broadcast) truck supports it natively. The vast majority of stadium camera infrastructure in 2026 still ends at HD-SDI patch panels.
When to use HD-SDI
- OB truck is on-site and cabled to the venue.
- You want sub-40ms end-to-end latency.
- You need lossless quality for tier-1 broadcasts (cup final, world championship).
- You are working with a broadcaster's existing graphics chain that expects ANC-data scoring overlays.
What HD-SDI costs
Cable runs: SAR 30-60 per metre for high-grade Belden 1694A or equivalent. Patch panels: SAR 2-8k per venue. SDI distribution amps and crossbars: SAR 15-60k for a tournament-grade stadium. Plus the truck-side terminations.
What HD-SDI breaks
- Cable runs over 100m without repeaters degrade.
- If the truck is not on-site, you cannot use HD-SDI — you need contribution.
- 4K UHD pushes HD-SDI to 12G-SDI or quad-link 3G-SDI, which doubles cable cost.
NDI: IP-native for the modern venue
NDI (Network Device Interface) is NewTek/Vizrt's IP video transport, now an open and widely-adopted standard. Full-fidelity HD video over standard ethernet, with sub-frame latency on a properly-tuned network. NDI HX is the lower-bandwidth flavour at H.264 quality.
When to use NDI
- Older venues with limited HD-SDI capacity (and you do not want to re-cable).
- Multi-source production with many feeds (NDI handles 30+ concurrent streams per 10G interface).
- Tournaments where you need to mix cameras, graphics and replays without a hardware video switcher.
- OB-truck-less production — virtualised broadcast in a control room a few hundred metres away.
What NDI costs
10G-capable network switches: SAR 60-180k per venue. NDI software licences and converters: SAR 8-25k per source. The big saving is in not pulling new HD-SDI cable.
What NDI breaks
- Network must be tuned — VLAN, QoS, PTP sync, no random multicast traffic.
- Latency is typically 80-200ms; not as tight as baseband HD-SDI.
- A network event can take down multiple feeds simultaneously; HD-SDI failure is usually per-cable.
We used NDI extensively at the world handball championship Egypt 2021 across six venues — the older Cairo arenas had limited HD-SDI capacity, and re-cabling was not in the budget. NDI absorbed the gap with zero broadcaster complaints.
SRT: when the OB truck is not on-site
SRT (Secure Reliable Transport, originally Haivision, now an open standard) is the contribution transport for the modern era. UDP-based with built-in forward-error-correction and AES encryption. Carries broadcast-quality video over the public internet from venue to remote production hub. Used by every major streamer and increasingly by broadcasters for remote production.
When to use SRT
- Remote production — the production gallery is in another city.
- Multi-venue contribution to a central NOC over the public internet.
- Secondary feeds (backup contribution path).
- Cost-sensitive productions where a satellite uplink is not budgeted.
What SRT costs
SRT itself is open-source — the cost is the encoder/decoder pair and the bandwidth. Encoders/decoders: SAR 15-45k each. Bandwidth: a 1080p H.265 SRT stream needs 5-12 Mbps; budget enterprise-grade business internet at the venue.
What SRT breaks
- Public-internet jitter — needs 250-500ms buffer for resilience.
- Asymmetric paths — different upload vs download performance.
- Network configuration — UDP port forwarding, NAT traversal.
The decision matrix
For a federation or tournament organiser, the practical decision rule we apply at Skyline · Arena Metrics:
- Tier-1 final with OB truck on-site, latency is everything → HD-SDI primary, NDI secondary, SRT for the remote contribution path.
- Multi-venue tournament with shared central production → NDI within each venue, SRT venue-to-central.
- Distributed regional sport with limited venue cabling → NDI venue-to-LAN, SRT to broadcaster.
- Esports event with low-budget remote production → NDI within venue, SRT to broadcaster, no HD-SDI at all.
What we always insist on
- PTP sync across every device on the chain — cameras, switcher, scoring, graphics. Sub-microsecond offset.
- Dual transport paths for any tier-1 broadcast — never single point of failure.
- NOC-monitored quality dashboards with packet loss, jitter, frame freezes visible to the on-call engineer.
- Pre-match dry-run at T-minus 4 hours, walking every transport leg from camera to graphic.
How Skyline integrates all three
Arena Metrics by Skyline is transport-agnostic by design. The scoring event is published once to an internal event bus; transport adapters take it to HD-SDI ANC data, NDI metadata, SRT metadata, or whichever feed-protocol the broadcaster wants (OPTA, Stats Perform, Genius Sports). The vendor-side complexity is hidden from the federation — they see one platform, the broadcaster sees one feed per transport.
Side-by-side: HD-SDI vs NDI vs SRT vs RTMP comparison table
The procurement question is rarely "which transport". It is "what is the right mix per match and per venue". This is the comparison table Skyline · Arena Metrics by Skyline uses in tender responses across the GCC. Every number is measured, not theoretical.
| Attribute | HD-SDI | NDI | SRT | RTMP |
|---|---|---|---|---|
| Glass-to-glass latency | 20-40 ms | 80-200 ms | 200-500 ms (tunable) | 2,000-8,000 ms |
| Encoding | Uncompressed baseband | Light visually-lossless | H.264 / H.265 | H.264 / FLV |
| Bandwidth per 1080p59.94 feed | 1.485 Gbps (3 Gbps for 1080p60) | 100-150 Mbps | 5-12 Mbps | 3-8 Mbps |
| Physical medium | 75Ω coax / fibre | Ethernet (1G/10G) | Public internet / WAN | Public internet |
| Error correction | N/A (baseband) | PTP-sync, no FEC | FEC + ARQ + encryption | TCP retries (slow) |
| Per-venue cost (1080p, 8 sources) | SAR 180-450k | SAR 120-280k | SAR 60-180k | SAR 25-80k |
| Best use case | Tier-1 cup final, OB truck on-site | Multi-source venue production | Remote contribution over WAN | Direct-to-consumer streaming only |
| Federation-grade ready? | Yes (primary) | Yes (secondary) | Yes (contribution) | No (distribution only) |
RTMP is included for completeness because procurement officers still see it specified in legacy broadcaster paperwork. In 2026 it is a distribution protocol (player to fan), not a contribution one. Specifying RTMP for federation-grade scoring contribution is a red flag in any tender.
The latency-budget worksheet: counting every millisecond from referee tap to broadcast graphic
Federation-grade scoring procurement should arrive at a hard end-to-end latency budget — referee console tap to the moment the lower-third graphic updates on the broadcast feed. Skyline · Arena Metrics by Skyline's published target is sub-180ms; here is the worksheet that gets you there. Numbers are typical Skyline-deployment figures across the last 36 months in the GCC; every cell is measured at the venue, not estimated.
- Step 1 — Referee console tap to local edge server: 4-12 ms over a tuned in-venue VLAN with QoS priority and offline-tolerant local state. Verify with the platform's own audit log; both timestamp + monotonic clock should be visible.
- Step 2 — Local edge to master event bus: 8-20 ms within a single host city over diverse fibre, 25-60 ms cross-city over SRT contribution. Multi-venue deployments must account for both budgets in parallel.
- Step 3 — Master event bus to broadcast adapter: 2-6 ms — a single internal hop on the same rack. The adapter (HD-SDI ANC, NDI metadata, SRT metadata) translates the event into the wire protocol the broadcaster's graphics engine expects.
- Step 4 — Broadcast adapter to graphics engine (Vizrt / Chyron): 15-40 ms depending on graphics-engine queue depth, template complexity, and whether the engine is on the same LAN or coming in over IP-video.
- Step 5 — Graphics engine to broadcast switcher input: 8-20 ms for SDI key+fill output, 30-80 ms for NDI key+fill.
- Step 6 — Switcher to encoder output (the feed the broadcaster ships): 20-60 ms through the production switcher, downstream keyer, audio embedder, and master-control output.
Sum of typical values: 57-158 ms when every link is healthy and on-site, well inside the 180ms target. Sum of worst-case-still-acceptable values: 198-360 ms for cross-city SRT contribution, which is why tier-1 finals always have the OB truck on-site. Build this worksheet into every RFP — a vendor that cannot fill in each cell with measured numbers from a recent GCC deployment is not federation-grade.
Frequently asked questions
What's the latency difference between HD-SDI and SRT for live scoring graphics?
HD-SDI delivers 20-40 ms glass-to-glass over a single venue. SRT delivers 200-500 ms over a public-internet contribution path. For a tier-1 cup final where the OB truck is on-site, HD-SDI primary plus NDI secondary keeps the end-to-end latency under 180 ms — well inside Skyline · Arena Metrics by Skyline's published target. SRT is the right choice for remote production where the OB truck is in another city, with the latency budget consciously accepted and the broadcaster informed up-front. Mixing the two is normal: HD-SDI inside the venue, SRT for cross-city contribution to the master NOC.
Is NDI broadcast-grade enough for international tournaments?
Yes for non-tier-1 matches and as a secondary path on tier-1. NDI is the standard transport across most international tournaments hosted in older arenas where re-cabling for full HD-SDI capacity is not in scope. Skyline ran NDI extensively at the world handball championship Egypt 2021 across six venues — older Cairo arena infrastructure made HD-SDI re-cabling cost-prohibitive, and NDI absorbed the gap with zero broadcaster complaints. The tuning bar is the catch: NDI needs a properly QoS-tuned 10G LAN, PTP sync, no random multicast traffic. Get that right and NDI is broadcast-grade up to 1080p60.
Do procurement officers still need to specify RTMP in 2026?
No, except as a distribution protocol for direct-to-consumer streaming. RTMP carries 2-8 seconds of latency — disqualifying for federation-grade scoring contribution. If a vendor's tender response specifies RTMP as their primary or secondary contribution transport, treat that as a credibility deduction. The right specification in 2026 is HD-SDI primary, NDI secondary, SRT for remote contribution, with RTMP reserved for the broadcaster's CDN-to-fan layer downstream. See the 2026 buyer's guide for the full transport specification.
Talk to the team
If you are scoping a 2027, 2029 or 2034 broadcast integration, the team that designed Arena Metrics will walk you through a reference architecture. sales@alskyline.com · +966 50 993 9334.
Related reading
Have a question or a project?
Talk to us — we serve 15 countries across MENA, Turkey, and Azerbaijan.
Start a conversation