Published 12 May 2026
Getting live video out of a field, a stadium car park, or a temporary compound is a problem that broadcast engineers have been solving for decades. Satellite uplinks work, but they cost thousands per hour and need line of sight to the bird. Fixed fibre works brilliantly if someone thought to install it six months ago. For everything else, bonded cellular is the answer, and SpeedFusion is how we build it.
This article is a practical sizing guide. It covers how to calculate the bandwidth you actually need, how to account for SpeedFusion overhead, how many SIMs and carriers to provision, which hardware fits which use case, and how to survey a location before you commit to a deployment. It is written for broadcast engineers and technical producers who need to spec this properly, not guess.
Start with the codec, not the pipe
Every sizing exercise begins with the same question: what are you sending, and at what quality? The codec and encoding profile determine the baseline bitrate. Everything else is overhead on top of that number.
Here are the real-world bitrates you should plan around for a single contribution feed:
| Codec / Profile | Resolution | Typical Bitrate | Notes |
|---|---|---|---|
| H.264 High Profile | 1080p50 | 8 - 15 Mbps | Most common for contribution. 10 Mbps is a sensible target for broadcast-quality output. |
| H.264 High Profile | 1080i50 | 6 - 12 Mbps | Interlaced is still standard in some OB workflows. |
| H.264 High Profile | 720p50 | 4 - 8 Mbps | Good fallback for constrained links. Looks acceptable on most platforms. |
| H.265 / HEVC Main | 1080p50 | 5 - 8 Mbps | Roughly 40% more efficient than H.264. Requires HEVC-capable encoder and decoder. |
| H.265 / HEVC Main | 2160p (4K) | 15 - 25 Mbps | Viable over bonded cellular, but pushes the limits of what you can sustain on 4G. |
| SRT (with H.264) | 1080p50 | 8 - 15 Mbps + 3-5% overhead | SRT adds its own retransmission overhead. Account for it separately from SpeedFusion FEC. |
| SRT (with H.265) | 1080p50 | 5 - 8 Mbps + 3-5% overhead | Most bandwidth-efficient option for high-quality contribution. |
These numbers assume a constant bitrate (CBR) or capped variable bitrate (VBR) encoding mode. If your encoder is running in uncapped VBR, the peak bitrate can spike significantly above the average, sometimes by 50% or more during high-motion scenes. For broadcast contribution over cellular, CBR or capped VBR is strongly preferred. Uncapped VBR can cause buffer overruns in the SpeedFusion tunnel during exactly the moments when picture quality matters most: fast action, crowd shots, pyrotechnics.
A practical rule: take your target codec bitrate and add 10% for transport overhead (RTP/UDP headers, SRT framing, or MPEG-TS container overhead). That gives you the raw application bandwidth. For a single 1080p50 H.264 feed at broadcast quality, budget 11 Mbps of application throughput as your starting point.
Contribution feeds vs. return feeds
Most broadcast deployments are heavily asymmetric. The contribution feed (camera to studio or MCR) is the primary traffic and it flows upstream from the field location. Return video (programme feed, talkback, comms) flows downstream to the field. This matters because cellular networks are asymmetric too, and the asymmetry works against you.
On a typical UK 4G connection, you will see roughly 3:1 or 4:1 download-to-upload ratio. A connection delivering 40 Mbps download might only provide 10 - 12 Mbps upload. On 5G NR (sub-6 GHz), the ratio is often similar or slightly better, perhaps 3:1. On mmWave 5G the ratio improves, but mmWave coverage in the UK is currently limited to a handful of urban centres.
This means that when you are sizing a SpeedFusion deployment for broadcast, you are sizing for upload capacity. Download is almost always abundant. Upload is the constraint.
Return feeds are typically low-bitrate: a programme return might be 2 - 4 Mbps, IFB/talkback is negligible (64 - 128 kbps per channel). These fit comfortably within the download capacity of any bonded cellular link.
SpeedFusion overhead: FEC and WAN smoothing
SpeedFusion is not a free abstraction. The bonding, Forward Error Correction (FEC), and WAN Smoothing features all consume additional bandwidth. If you size your link for the raw codec bitrate and ignore the overhead, you will run out of capacity at the worst possible moment.
Bonding overhead
SpeedFusion bonding wraps traffic in a GRE-like tunnel with sequence numbering and reordering. The tunnel header adds approximately 60 - 80 bytes per packet. For a typical 1300-byte video packet, that is roughly 5 - 6% overhead. This is unavoidable and always present when the tunnel is active.
Forward Error Correction (FEC)
FEC is where the real overhead cost sits, and where the real value sits too. SpeedFusion FEC works by sending redundant data across multiple WAN links. If a packet is lost on one cellular connection, the FEC data on a different connection allows the receiving end to reconstruct it without waiting for a retransmission.
The overhead depends on the FEC ratio you configure:
| FEC Setting | Bandwidth Overhead | When to Use |
|---|---|---|
| Low | ~25 - 30% | Stable cellular with low packet loss (<1%). Typical for urban deployments with good signal. |
| Normal | ~50% | The default and the right choice for most broadcast work. Handles moderate packet loss and jitter. |
| High | ~100% (doubles bandwidth) | Hostile RF environments, rural locations with marginal signal, mission-critical feeds where any glitch is unacceptable. |
For a single 1080p50 H.264 feed at 11 Mbps (including transport overhead), here is what the total upload requirement looks like with different FEC settings:
| FEC Level | Tunnel Overhead | FEC Overhead | Total Upload Required |
|---|---|---|---|
| Low | ~0.6 Mbps | ~3 Mbps | ~14.6 Mbps |
| Normal | ~0.6 Mbps | ~5.5 Mbps | ~17.1 Mbps |
| High | ~0.6 Mbps | ~11 Mbps | ~22.6 Mbps |
Most broadcast deployments should run FEC at Normal. It provides good protection against the kind of transient packet loss that cellular networks produce (cell handovers, congestion bursts, interference) without consuming so much bandwidth that you cannot sustain the feed. Drop to Low only if you have thoroughly tested the location and are confident in the RF environment. Use High for genuinely hostile conditions or when the production absolutely cannot tolerate any visible artefacts.
WAN Smoothing
WAN Smoothing is a separate feature from FEC, though they work together. It sends duplicate packets across multiple WAN links and uses whichever copy arrives first. It is extremely effective at reducing jitter, which matters for live video because jitter causes buffer underruns at the decoder.
WAN Smoothing at its maximum setting effectively multiplies your bandwidth consumption by the number of WAN links you are smoothing across. If you are smoothing across three cellular connections, you are sending three copies of every packet. For a 11 Mbps stream, that is 33 Mbps of aggregate upload across all links, plus tunnel overhead.
In practice, you rarely need maximum WAN Smoothing if you are already running FEC at Normal. The combination of FEC Normal plus WAN Smoothing at Medium provides excellent resilience without tripling your bandwidth bill. Test this combination during your site survey and adjust based on measured results.
How many SIMs and carriers
The number of SIM cards and the mix of carriers is one of the most important decisions in the deployment, and one of the most commonly got wrong. The thinking should follow a specific logic.
Carrier diversity is not optional
Using four SIMs from the same carrier gives you four connections to the same mast, the same backhaul, and the same core network. If that carrier has a problem, all four connections fail together. SpeedFusion cannot bond its way out of a single point of failure.
For broadcast, you want at least two carriers, ideally three. In the UK, the three distinct physical networks are EE, Three, and the VMO2 (Vodafone/O2 shared) network. Using SIMs from EE, Three, and Vodafone or O2 gives you genuine infrastructure diversity. If one network has a localised outage or congestion event, the remaining two keep the feed alive.
Working out the SIM count
Start with your total upload requirement (codec bitrate plus FEC overhead plus tunnel overhead). Divide that by a conservative per-SIM upload estimate.
For UK 4G, a conservative per-SIM upload estimate is 5 - 8 Mbps in areas with decent signal. In urban areas with heavy congestion, it might be 3 - 5 Mbps. In rural areas with good signal and light loading, it could be 10 - 15 Mbps. You should always size against the conservative end.
For a single 1080p50 H.264 feed with FEC Normal (total ~17 Mbps upload):
- At 5 Mbps per SIM: 4 SIMs needed (with minimal headroom)
- At 8 Mbps per SIM: 3 SIMs needed (comfortable)
- Recommended: 4 SIMs across 3 carriers (two SIMs on the strongest carrier, one each on the other two)
For dual contribution feeds (two cameras, two independent encoders):
- Total upload requirement: ~34 Mbps
- At 5 Mbps per SIM: 7 SIMs needed
- Recommended: 8 SIMs across 3 carriers, which means you need hardware with at least 8 modem slots
Always provision at least one more SIM than your calculation says you need. Cellular is variable. The SIM that was giving you 12 Mbps during your site survey at 10am might give you 4 Mbps at 3pm when the local schools finish and everyone starts streaming video. Headroom is not a luxury in broadcast; it is the difference between a clean feed and a frozen frame on air.
Antenna considerations
The antenna system is where most broadcast cellular deployments either succeed or fail. The router itself is just a box with modems in it. The antenna determines what those modems can actually achieve.
Internal vs. external antennas
Internal antennas (built into the router enclosure) are adequate for bench testing, vehicle-mounted installations where the router sits under a roofline, or situations where the router is near a window with good line of sight to multiple masts. For field broadcast deployments, external antennas are almost always required.
A good external MIMO antenna mounted at 3 - 5 metres height will typically gain 8 - 15 dB over the router's internal antennas. That translates directly into higher modulation rates, faster upload speeds, and more consistent connections. In cellular, the link budget is everything.
Omni vs. directional
Omnidirectional antennas pick up signal from all directions and are the right choice when you do not know where the nearest masts are, or when you need to bond across masts in different directions. Most broadcast deployments use omni antennas because the setup time is faster and the operator does not need to aim anything.
Directional panel antennas offer 6 - 10 dB more gain than an equivalent omni, but they need to be pointed at a specific mast. If you are doing a multi-day deployment at a fixed location and you have identified the best masts during your site survey, directional antennas will give you noticeably better performance. The trade-off is setup time and the need for someone who knows how to aim them using signal metrics, not guesswork.
MIMO order
Modern LTE and 5G modems support 2x2 or 4x4 MIMO. Each MIMO stream needs its own antenna element. A 4x4 MIMO antenna for a single modem has four connectors. If your router has four modems each running 2x2 MIMO, that is eight antenna connectors total. Make sure your antenna system matches your modem configuration. Mismatched MIMO (running a 4x4 modem with a 2x2 antenna) wastes half the modem's potential throughput.
Cable runs
Coaxial cable between the antenna and the router introduces signal loss. At 2.5 GHz (Band 7, commonly used by EE and Three), typical LMR-400 cable loses about 0.22 dB per metre. A 10-metre cable run costs you 2.2 dB. Thinner cable (RG-58, RG-174) loses far more and should never be used for runs over 2 metres.
Keep cable runs as short as practical. If you need height, it is almost always better to mount the router near the antenna and run an Ethernet cable down to the encoder, rather than mounting the router at ground level and running long coaxial cables up to the antenna. Ethernet has negligible loss over 100 metres. Coax at cellular frequencies does not.
Upload vs. download: the cellular asymmetry problem
This deserves its own section because it catches people out repeatedly. Cellular networks are designed for consumers who download far more than they upload. The radio frame structure in both LTE and 5G NR allocates more resource blocks to the downlink than the uplink. Carriers configure their networks this way because it matches the traffic profile of 95% of their subscribers.
Broadcast is the opposite. You are pushing high-bitrate video upstream, which is exactly what the network is not optimised for.
Practical consequences:
- Speed test apps show headline download speeds that are irrelevant to your use case. Always test upload specifically.
- A site that shows 80 Mbps download and 15 Mbps upload on a single SIM is a perfectly viable broadcast site. A site showing 150 Mbps download and 3 Mbps upload is not.
- 5G often improves download dramatically while upload gains are modest. Do not assume that "5G coverage" means your upload problem is solved.
- Some carriers offer "upload boost" or business-grade APN configurations that rebalance the uplink/downlink ratio. Ask your carrier account manager. On EE business accounts, for example, certain APNs provide improved upload allocation.
Site survey methodology
A proper site survey for a broadcast cellular deployment takes 60 to 90 minutes and follows a structured process. Turning up on the day and hoping for the best is not engineering. Here is the procedure we use.
1. Desk research (before visiting)
Check mast locations using Ofcom's Sitefinder or the MAST database. Identify which carriers have infrastructure within 2 km of your location. Check the terrain profile between the site and the nearest masts using OS mapping or Google Earth. A mast 500 metres away on the other side of a hill is worse than a mast 3 km away with clear line of sight.
Check for known events or activities that might affect network loading. A broadcast from a festival site will compete with 80,000 phones on the same masts. A broadcast from a rural estate will have the masts largely to itself.
2. On-site signal survey
Bring a router with SIMs from all three UK networks. At each candidate position on site, measure:
- RSRP (Reference Signal Received Power): target better than -95 dBm for reliable broadcast. Below -105 dBm, you will struggle for upload.
- RSRQ (Reference Signal Received Quality): target better than -12 dB. Poor RSRQ indicates interference or congestion even with strong signal.
- SINR (Signal to Interference plus Noise Ratio): target above 10 dB for good upload throughput. Below 5 dB, throughput drops sharply.
- Band and cell ID: record which band each carrier is using. Band 20 (800 MHz) gives good coverage but limited bandwidth. Band 3 (1800 MHz) and Band 7 (2600 MHz) offer more capacity. Band 1 (2100 MHz) is common for 3G fallback.
3. Upload throughput testing
Signal metrics tell you what is possible. Throughput tests tell you what is actually happening. Run sustained upload tests (not burst speed tests) on each carrier for at least 60 seconds. Use iperf3 to a known server, not a consumer speed test app that measures peak throughput over 10 seconds.
Run the test at different antenna heights if you are evaluating mast positions. Run it at different times if possible. A survey at 9am on a Tuesday will give different results from the same survey at 2pm on a Saturday.
4. Bonded throughput test
Once you have individual carrier results, configure the SpeedFusion tunnel and run a bonded upload test. This is your real-world number. Compare it against your calculated requirement (codec bitrate plus FEC overhead). If bonded upload throughput exceeds your requirement by at least 30%, the site is viable. If the margin is less than 30%, you need more SIMs, better antennas, or a different deployment position.
5. Document everything
Record signal metrics, throughput results, antenna positions, cable routing options, power availability, and any RF interference sources. Take photos. Mark antenna mount points. Note the GPS coordinates of your test positions. This documentation is essential for the deployment crew who may not be the same people who did the survey.
Real deployment examples
These are anonymised but based on actual deployments we have delivered. The numbers are real.
Rural sporting event, single camera
Location was a rural estate in the Midlands, 1.8 km from the nearest mast (EE, Band 20). Three network had a mast 3.2 km away. Vodafone coverage was marginal (RSRP -108 dBm).
Requirement: single 1080p50 H.264 feed at 10 Mbps to a London MCR. Four-hour production window.
Solution: Peplink HD4 MBX with four SIMs (two EE, one Three, one Vodafone as fallback). External 4x4 MIMO omni antenna on a 5-metre pneumatic mast. SpeedFusion tunnel to a FusionHub VM in AWS eu-west-2 (London). FEC set to Normal. WAN Smoothing at Medium.
Results: sustained bonded upload of 22 Mbps across three active connections. The Vodafone SIM contributed very little (1.5 Mbps upload) but was kept active for diversity. Feed ran clean for four hours with zero visible artefacts. Peak single-SIM upload from EE was 9 Mbps; Three delivered 7 Mbps.
Urban multi-camera, music event
Location was a city-centre venue with 5,000 capacity. Four masts within 800 metres but heavy congestion expected during the event.
Requirement: two independent 1080p50 H.264 feeds at 10 Mbps each. Eight-hour production window across two days.
Solution: two Peplink MAX Transit Duo units, each with two modems and two SIMs (one EE, one Three per unit). Each unit drove one camera feed via its own SpeedFusion tunnel. External directional panel antennas aimed at the two strongest masts (identified during site survey the previous week). FEC High on both tunnels because of anticipated congestion.
Results: during sound check (light network loading), each unit achieved 18 - 25 Mbps bonded upload. During the main event, per-SIM upload dropped to 4 - 6 Mbps as the audience arrived and loaded the masts. Bonded upload held at 12 - 16 Mbps per unit. FEC High consumed significant bandwidth but protected the feeds. Both feeds ran without interruption across both days. Codec bitrate was reduced from 10 Mbps to 8 Mbps during the main event as a precaution. The production team saw no quality difference on the output.
Coastal OB, limited coverage
Location was a harbour on the south coast. Nearest EE mast was 4.5 km inland. Three had no usable coverage. O2 had a mast 2.1 km away with reasonable line of sight across the water.
Requirement: single 1080p50 H.265 feed at 6 Mbps for a live web stream. Three-hour window.
Solution: Peplink MAX Transit Duo with two SIMs (one EE, one O2). High-gain directional Yagi antennas aimed at each mast, mounted on a scaffold tower at 6 metres. SpeedFusion to FusionHub in AWS. FEC Normal. H.265 encoding chosen specifically to reduce bandwidth requirement given the constrained upload.
Results: EE delivered 5 Mbps upload. O2 delivered 7 Mbps upload. Bonded upload was 10 Mbps, giving comfortable headroom for the 6 Mbps H.265 feed plus FEC overhead. Feed quality was excellent. The directional antennas were critical; with the router's internal antennas, upload was below 2 Mbps on both networks.
Hardware selection
Peplink makes a range of routers. Two models handle the vast majority of broadcast deployments.
HD4 MBX
The HD4 MBX is the flagship mobile router. Four cellular modems, each supporting Cat-20 LTE and optional 5G modules. It is the right choice for demanding broadcast deployments: dual contribution feeds, 4K, or locations where you need maximum carrier diversity.
Key specs for broadcast sizing:
- Four independent modem modules with dedicated antenna ports
- Up to 4x4 MIMO per modem (16 antenna connectors total for cellular)
- Hardware-accelerated SpeedFusion throughput up to 400 Mbps
- Supports FEC, WAN Smoothing, and Hot Failover simultaneously
- Dual Ethernet WAN ports for adding wired connections to the bond
- GPS for automatic location logging
- 12 - 48V DC input, suitable for vehicle or battery power
The HD4 MBX is the standard choice for OB vans, flyaway kits, and any deployment where the feed is mission-critical and the budget supports it.
MAX Transit Duo Pro
The MAX Transit Duo Pro is a dual-modem router in a compact form factor. Two cellular modems, each with 2x2 MIMO. It is lighter, cheaper, and perfectly adequate for single-camera deployments where the RF environment is reasonable.
Key specs for broadcast sizing:
- Two cellular modems (Cat-18 LTE or 5G, depending on module)
- 2x2 MIMO per modem (four antenna connectors for cellular)
- SpeedFusion throughput up to 200 Mbps
- Compact enclosure, mounts in a small Peli case for flyaway use
- Single Ethernet WAN for optional wired connection bonding
- Wi-Fi for local equipment connectivity
- 12V DC input, runs from a standard V-mount battery for several hours
The Transit Duo Pro is the workhorse for single-camera ENG (electronic news gathering) kits, web streams, and deployments where weight and size matter. A Transit Duo Pro, a compact encoder, an external antenna, and a V-mount battery fit in a single backpack. That is a complete bonded cellular broadcast uplink you can carry on a train.
FusionHub: the other end of the tunnel
Every SpeedFusion deployment needs a peer at the receiving end. For broadcast, this is typically a FusionHub Solo or FusionHub Essential VM running in a cloud data centre. AWS eu-west-2 (London) or eu-west-1 (Ireland) are the standard choices for UK broadcast because they offer low-latency connectivity to London-based MCRs and playout centres.
The FusionHub terminates the SpeedFusion tunnel, reassembles the bonded traffic, and forwards the clean video stream to your MCR or streaming endpoint via the data centre's fixed-line connectivity. From the MCR's perspective, the video arrives from a stable IP address with consistent throughput. The cellular bonding is invisible.
Size the FusionHub VM for the aggregate throughput of all tunnels it will terminate. A FusionHub Solo handles up to 150 Mbps, which is sufficient for most single-site broadcast deployments. FusionHub Essential supports higher throughput and multiple concurrent tunnels for multi-site operations.
The sizing formula
Putting it all together. Here is the step-by-step calculation for any broadcast SpeedFusion deployment:
- Determine codec bitrate. Choose codec, resolution, and encoding profile. Get the CBR/capped VBR target in Mbps.
- Add transport overhead. Add 10% for RTP/SRT/MPEG-TS framing. This is your application bandwidth.
- Multiply by number of feeds. Two cameras at 11 Mbps each = 22 Mbps application bandwidth.
- Add SpeedFusion tunnel overhead. Add 5 - 6% for tunnel encapsulation.
- Add FEC overhead. Multiply by 1.3 (Low), 1.5 (Normal), or 2.0 (High).
- The result is your total upload requirement.
- Divide by conservative per-SIM upload estimate to get the minimum SIM count. Add one SIM for headroom.
- Ensure at least two, ideally three, distinct carriers in the SIM mix.
- Select hardware with enough modem slots to hold all SIMs. HD4 MBX for four or more. Transit Duo Pro for two.
- Validate with a site survey. If measured bonded upload exceeds your requirement by 30% or more, the deployment is viable.
Common mistakes
After years of deploying these systems, the same errors keep appearing. Avoid these and you will avoid most broadcast connectivity failures.
- Sizing for download instead of upload. Consumer speed tests emphasise download. Broadcast needs upload. Test upload specifically.
- Ignoring FEC overhead. A 10 Mbps feed does not need 10 Mbps of upload. It needs 15 - 20 Mbps depending on FEC settings.
- All SIMs on one carrier. Four EE SIMs is not diversity. It is four connections to the same single point of failure.
- Using internal antennas at field locations. Internal antennas are for desk testing. External antennas are for production.
- Not surveying at the right time. A Saturday afternoon at a stadium is a fundamentally different RF environment from a Tuesday morning. Survey at a time that matches your production schedule.
- Running uncapped VBR over cellular. CBR or capped VBR. Always. Uncapped VBR creates bitrate spikes that overwhelm the uplink at the worst moments.
- Forgetting the FusionHub. The receiving end matters. A poorly provisioned FusionHub VM or one in the wrong region adds latency and can become a bottleneck.
- No fallback plan. If bonded cellular is your primary uplink, what happens if three of four SIMs lose connectivity? Have a reduced-quality fallback profile ready: drop to 720p, reduce bitrate, switch FEC to Low. A lower-quality feed is better than no feed.
Conclusion
Sizing a SpeedFusion deployment for broadcast is an engineering exercise, not a guessing game. Start with the codec bitrate, add the overhead layers systematically, calculate the SIM count, select the right hardware, and validate everything with a proper site survey. The technology works. It works reliably enough that production teams trust it for live-to-air broadcasting. But it works because someone did the maths first and tested before transmission day.
If you are planning a broadcast deployment and want help with sizing, site surveys, or hardware selection, get in touch. We stock the full Peplink range including the HD4 MBX and MAX Transit Duo Pro and can configure SpeedFusion tunnels and FusionHub instances as part of a turnkey deployment.