Published 12 May 2026
For decades, the satellite truck was the only serious option for live outside broadcast contribution. You needed a guaranteed uplink from a location with no fixed infrastructure, and satellite delivered that. It was expensive, it was heavy, it required a trained operator and a clear line of sight to the bird, but it worked. Nobody got fired for booking the sat truck.
That calculus has changed. Not because satellite got worse, but because cellular got dramatically better, and bonding technology matured to the point where four mediocre 4G connections can outperform a single satellite uplink on every metric that matters to an OB engineer: cost, setup time, latency, and operational flexibility.
This article is a technical walkthrough of how bonded cellular works for outside broadcast, what hardware and software you need, where it excels, and where satellite still has the upper hand. It is written from the perspective of teams that have deployed this kit in the field and know what the spec sheets don't tell you.
The cost argument
Let's start with money, because that is what got most broadcast operations interested in cellular bonding in the first place.
A satellite uplink truck for a single-day OB typically costs between £3,000 and £8,000 depending on the provider, the location, and the satellite capacity required. That figure covers the truck, the operator, the fuel, the satellite time, and the insurance. A multi-day booking scales roughly linearly. A week-long event can run £15,000 to £30,000 for satellite contribution alone, before you account for the production crew, cameras, or anything else.
A bonded cellular solution using a Peplink HD4 MBX with four embedded cellular modems, a FusionHub Solo instance in the cloud, and four multi-network SIM cards costs around £4,000 to £6,000 for the hardware (one-time purchase), roughly £200 to £400 per month for data across four SIMs depending on your data plan, and £50 to £150 per month for the FusionHub cloud instance. After the initial hardware spend, your per-event contribution cost drops to whatever data you consume that day, which for a 10 Mbps CBR stream running eight hours is roughly 36 GB per SIM if the load is evenly distributed. On most unlimited or high-cap plans, that is a rounding error.
Put differently: the cellular bonding kit pays for itself after one or two events compared to satellite truck hire. After that, every deployment is nearly free at the network level.
There are caveats. You need to factor in antenna hardware, rack mounting or flight-case integration, and the time to configure everything properly. But even with those costs included, the total capital outlay for a broadcast-grade cellular bonding rig is less than a single week of satellite truck hire at most UK venues.
Setup time: minutes, not hours
A satellite truck needs a level parking area, a clear view of the southern sky (in the northern hemisphere), time to deploy the dish, time to acquire the satellite, time to establish the uplink, and time to confirm the booking with the satellite provider. Best case, you are looking at 30 to 45 minutes from arrival to a confirmed uplink. Worst case, if the venue has trees, buildings, or other obstructions, you are repositioning the truck or calling for a different orbital slot. That can burn an hour or more.
A bonded cellular rig powers on and starts connecting immediately. The HD4 MBX boots in about 90 seconds. The four modems register on their respective networks within another 30 seconds. The SpeedFusion tunnel to FusionHub establishes in under 10 seconds once the modems are up. From cold start to a working bonded tunnel with all four paths active, you are looking at two to three minutes. If the unit was in standby rather than fully powered off, it is ready in under a minute.
For a roaming OB crew doing three or four hits per day from different locations, that difference is transformative. No repositioning. No dish alignment. No coordination with a satellite provider. Power on, confirm the tunnel is green, and start sending pictures.
How SpeedFusion bonding works for broadcast
SpeedFusion is Peplink's tunnel bonding technology, and it is worth understanding how it works at the packet level, because it explains why bonded cellular can deliver broadcast-grade reliability from connections that are individually unreliable.
The HD4 MBX has four cellular modems, each connected to a different mobile network operator. In the UK, that typically means one SIM each on EE, Three, Vodafone, and O2 (or their respective MVNOs). Each modem establishes an independent connection to the internet. SpeedFusion builds an encrypted tunnel from the HD4 MBX to a FusionHub instance running in a data centre or cloud environment. This tunnel spans all four cellular connections simultaneously.
Traffic is distributed across the four paths using one of several algorithms. For broadcast contribution, the most relevant mode is "Bandwidth Bonding with WAN Smoothing." In this mode, SpeedFusion splits outbound packets across all available paths to maximise aggregate throughput, while also sending redundant copies of each packet across multiple paths. If a packet is lost or delayed on one path, the copy arriving via another path fills the gap. The receiving end (FusionHub) reassembles the stream in the correct order and forwards it to the destination.
The WAN Smoothing feature is the critical piece for broadcast. Cellular connections are inherently variable. A single 4G connection might deliver 30 Mbps one moment and 8 Mbps the next, with intermittent packet loss during congestion or handover between towers. For file transfers, that variability is tolerable. For a live video contribution feed, it is catastrophic. WAN Smoothing absorbs that variability by maintaining redundant packet paths, so the output stream at the FusionHub end is clean and consistent even when individual cellular links are misbehaving.
You can configure the level of smoothing. "Low" sends minimal redundancy and maximises usable throughput. "High" sends each packet across all available paths, which uses more data but survives the loss of entire cellular connections without any disruption to the output stream. For critical live broadcasts, "High" is the correct setting. The data cost increase is worth it for the reliability guarantee.
SRT and RIST: the transport protocols that make this work
Bonded cellular gives you a reliable IP path from the OB location to the cloud. But you still need a transport protocol between your encoder and your decoder (or playout system) that can handle the characteristics of that path. This is where SRT and RIST come in.
SRT (Secure Reliable Transport) was originally developed by Haivision and is now an open-source protocol managed by the SRT Alliance. It is designed specifically for live video transport over unpredictable networks. SRT uses ARQ (Automatic Repeat reQuest) to retransmit lost packets, and it has a configurable latency buffer that determines how much time the receiver will wait for retransmissions before declaring a packet lost. For a bonded cellular path with SpeedFusion WAN Smoothing, an SRT latency setting of 200 to 500 ms is typically sufficient. That gives the protocol enough headroom to recover from transient losses without adding noticeable delay to the contribution feed.
RIST (Reliable Internet Stream Transport) is a newer protocol developed by the Video Services Forum (VSF). It offers similar ARQ-based recovery to SRT but uses a standards-based approach that some broadcasters prefer for interoperability reasons. RIST supports both "Simple Profile" (basic ARQ) and "Main Profile" (encryption, authentication, tunnelling). Either works well over a SpeedFusion bonded path.
The choice between SRT and RIST often comes down to what your encoder and decoder support. Most modern broadcast encoders support both. If your workflow is Haivision-centric, SRT is the natural fit. If you are working with a mixed vendor environment or need to interface with legacy IRDs, RIST may offer smoother integration.
The important point is that you are running these protocols over the SpeedFusion tunnel. The tunnel handles the bonding, smoothing, and path management. SRT or RIST handles the video-specific error recovery on top of that. The two layers complement each other: SpeedFusion removes most of the packet loss before SRT or RIST even sees it, and the transport protocol cleans up whatever remains. The result is a contribution feed that is, for all practical purposes, as reliable as a dedicated fibre circuit.
Encoder integration
The HD4 MBX sits between your encoder and the internet. From the encoder's perspective, it is just a network gateway. You connect your encoder to the HD4 MBX via Ethernet (or in some configurations, directly to a VLAN on the router), configure the encoder to push an SRT or RIST stream to the public IP address of your FusionHub instance, and the SpeedFusion tunnel carries the stream across the bonded cellular paths transparently.
This means the bonding solution is encoder-agnostic. It works with any hardware or software encoder that can output an IP stream. Common pairings in the field include Haivision Makito X series, LiveU Solo (used purely as an encoder with external connectivity), TVU Pack encoders, Matrox Monarch series, and software encoders running on laptops using OBS or vMix with SRT output.
For single-camera live hits and news gathering, a compact encoder like the Haivision Makito X4 paired with an HD4 MBX in a small flight case gives you a complete broadcast contribution kit that one person can carry. The total weight is under 8 kg including batteries. Try carrying a satellite dish in one hand.
For multi-camera OB productions, the HD4 MBX can handle multiple simultaneous streams. A 10 Mbps CBR stream for the main programme feed plus a 3 Mbps stream for a confidence return or comms channel is well within the aggregate throughput of four bonded 4G connections in most UK urban and suburban locations.
FusionHub as the cloud decode point
FusionHub Solo is Peplink's virtual appliance that acts as the tunnel endpoint in the cloud or data centre. It terminates the SpeedFusion tunnel from the HD4 MBX, reassembles the bonded stream, and outputs clean IP traffic to whatever system needs to receive it.
For broadcast contribution, FusionHub typically runs on a virtual machine in a cloud environment like AWS, Vultr, or a colocation facility. The VM needs a public IP address, adequate bandwidth (at least 50 Mbps for a comfortable margin on a single HD contribution feed), and not much compute. FusionHub Solo runs on a single-core VM with 1 GB of RAM without breaking a sweat.
The architecture looks like this: your encoder at the OB location pushes an SRT stream to the FusionHub's public IP. The stream travels through the SpeedFusion tunnel across four cellular paths. FusionHub reassembles it and presents it as a clean SRT stream on its local network interface. Your decoder, playout server, or MCR ingest system connects to FusionHub and receives the stream as if it were arriving over a dedicated circuit.
One significant advantage of the FusionHub model is location flexibility. You can spin up FusionHub instances in different regions. If you are covering an event in Edinburgh, you might run FusionHub on a London-based VM, giving you a London delivery point for the contribution feed. If you are covering something in Munich, you spin up a FusionHub in Frankfurt. The HD4 MBX at the OB location simply points its tunnel at whichever FusionHub instance is closest to the delivery point, minimising the IP path between FusionHub and the final destination.
For organisations that do regular OB work, maintaining two or three FusionHub instances in different geographies (London, Amsterdam, Frankfurt, for example) gives you a pan-European decode infrastructure for a few hundred pounds per month total.
Antenna selection for OB vehicles
The antenna is arguably the most important and most overlooked component of a cellular bonding OB rig. The HD4 MBX has four cellular modems, each with two antenna ports for MIMO operation. That is eight antenna connections total. How you connect those eight ports determines how much throughput you actually get in the field.
There are three common approaches, and each has trade-offs.
Omni-directional MIMO panel antennas (roof-mounted)
The most common OB vehicle installation uses two or four roof-mounted omni-directional MIMO panel antennas, each containing two antenna elements for 2x2 MIMO. These are low-profile (typically 20 to 40 mm tall), weather-sealed, and permanently mounted to the vehicle roof. Brands like Poynting, Panorama, and Peplink's own Mobility series are popular choices.
Advantages: no setup required. You park the vehicle and the antennas are already working. 360-degree coverage means you do not need to know where the nearest cell tower is. Excellent for roaming operations where speed of deployment matters more than squeezing every last megabit from the connection.
Disadvantages: lower gain than directional antennas, typically 2 to 5 dBi. In weak signal areas, you are leaving throughput on the table.
Directional panel antennas (mast-mounted or tripod)
For fixed OB positions where you will be stationary for several hours or days, a directional panel antenna on a short pneumatic mast or tripod can dramatically improve throughput. A directional panel with 8 to 11 dBi gain pointed at the nearest cell tower will outperform an omni antenna by a significant margin, particularly in fringe coverage areas or at crowded events where the cell towers are overloaded and you need every dB of signal advantage.
The trade-off is setup time. Someone needs to identify the nearest cell tower (apps like CellMapper or OpenSignal help), point the antenna, and fine-tune the alignment. This adds 10 to 15 minutes to your deployment, which partly erodes the speed advantage over satellite. But for a multi-day event where you set up once, the throughput gain is worth it.
Hybrid approach
The approach we see working best for most OB teams is a hybrid: permanent omni MIMO antennas on the vehicle roof for immediate connectivity on arrival, plus one or two directional antennas on telescopic masts that the crew deploys once the vehicle is parked and the production is being rigged. The omni antennas give you working connectivity within minutes for comms and testing. The directional antennas come online 15 minutes later and boost throughput for the live transmission window.
The HD4 MBX supports this neatly because you can assign specific modems to specific antennas and prioritise paths in the SpeedFusion tunnel configuration. Modems on the directional antennas get priority; modems on the omni antennas act as supplementary paths and automatic fallback.
MIMO considerations
Every modem in the HD4 MBX supports 2x2 MIMO, which means each modem uses two antenna elements simultaneously to double the theoretical throughput compared to a single-element connection. For MIMO to work properly, the two antenna elements need to be spatially separated or cross-polarised. Most commercial MIMO panel antennas handle this internally with cross-polarised elements, so you do not need to worry about physical spacing.
What you do need to worry about is cable quality and length. Cellular frequencies, particularly in the higher 5G bands, are sensitive to cable loss. A 5-metre run of low-quality RG58 coaxial cable can lose 3 to 5 dB at 3.5 GHz, which is enough to turn a usable 5G connection into a marginal one. Use LMR-400 or equivalent low-loss cable for any run over 2 metres. For roof-mounted antennas on an OB vehicle, keep the cable runs as short as possible and mount the HD4 MBX close to the antenna penetration point.
5G: what it changes and what it does not
5G is relevant to OB cellular bonding, but perhaps not in the way the marketing materials suggest.
Sub-6 GHz 5G (n78, n77, n1, n3) is the variant that matters for OB work in the UK today. It offers genuine throughput improvements over 4G LTE, typically 100 to 300 Mbps per connection in areas with good n78 coverage. Bonding four sub-6 5G connections gives you an aggregate that comfortably exceeds what most OB productions need, even for high-bitrate 1080p50 or 4K contribution feeds.
Millimetre-wave 5G (mmWave, n257, n258) is largely irrelevant for OB today. It requires line of sight to the cell antenna, has a range measured in hundreds of metres rather than kilometres, and is only deployed in a handful of dense urban locations in the UK. You cannot plan an OB around mmWave availability.
The practical difference 5G makes for OB work is headroom. With four bonded 4G connections, you can reliably sustain a 10 to 15 Mbps contribution feed in most locations. With four bonded 5G connections, you can sustain 25 to 40 Mbps, which opens the door to higher-bitrate codecs, 4K contribution, or multiple simultaneous feeds from a single bonding router. It does not change the fundamental architecture. It just gives you more bandwidth to work with when it is available.
The HD4 MBX supports 5G across all four modems. If 5G is available at your OB location, the modems will use it automatically. If not, they fall back to 4G LTE. The SpeedFusion tunnel does not care which generation of cellular technology the underlying modems are using.
Latency and bandwidth: real-world numbers
Spec sheets and theoretical maximums are useful for product comparisons but useless for OB planning. Here is what we actually see in the field with an HD4 MBX bonding four UK cellular connections through a FusionHub instance hosted in a London data centre.
Latency (glass-to-glass, including encoding and decoding): 500 ms to 1.5 seconds depending on encoder settings, SRT buffer, and codec. With a Haivision Makito X4 encoding H.265 at 10 Mbps with an SRT latency buffer of 250 ms, we consistently measure 800 ms to 1.2 seconds glass-to-glass. That is competitive with satellite (which typically runs 600 ms to 2 seconds depending on the transponder and codec) and dramatically better than the 4 to 8 seconds you see with some consumer-grade streaming solutions.
Throughput (sustained, with WAN Smoothing on High): 15 to 40 Mbps aggregate uplink depending on location and cellular conditions. In central London, we routinely see 25 to 35 Mbps with four 4G connections. In suburban areas, 15 to 25 Mbps. In rural locations with weaker coverage, 8 to 15 Mbps. These figures account for the overhead of WAN Smoothing, which consumes roughly 30 to 50 per cent of raw throughput for packet duplication on the "High" setting.
Stability: With WAN Smoothing on High, we have sustained 10 Mbps CBR streams for 8+ hours without a single visible artefact on the decoded output. Individual cellular connections dropped and reconnected multiple times during those sessions. The SpeedFusion tunnel absorbed every disruption transparently.
These numbers are not guaranteed. Cellular is a shared medium, and performance varies with location, time of day, network congestion, and how many other people at a 40,000-seat stadium are trying to upload videos simultaneously. The point is that bonded cellular with SpeedFusion WAN Smoothing delivers broadcast-usable performance in the vast majority of UK locations, most of the time.
When satellite still makes sense
Cellular bonding is not a universal replacement for satellite. There are scenarios where satellite remains the only viable option, and it is important to be honest about those.
Open ocean: No cellular coverage. If you are covering a yacht race, an offshore platform, or anything more than 20 km from shore, satellite (VSAT or LEO constellations like Starlink Maritime) is your only option. Bonded cellular is strictly a terrestrial technology.
Polar and extreme-latitude regions: Geostationary satellite coverage degrades above roughly 70 degrees latitude, but LEO constellations cover the poles reasonably well. Cellular coverage at extreme latitudes is sparse to non-existent. For Arctic or Antarctic OB work, satellite wins by default.
Genuinely remote terrestrial locations: Deep rural Scotland, central Wales, parts of the Highlands. If you are more than 10 km from the nearest cell tower in any direction, cellular bonding will not deliver the throughput you need for a reliable contribution feed. In these locations, satellite (or a hybrid satellite-plus-cellular approach, using whatever cellular signal is available to supplement a satellite uplink) is the right answer.
Guaranteed bandwidth requirements: If your broadcaster or client requires a contractually guaranteed bitrate for the contribution feed, satellite can provide that in a way that cellular cannot. Cellular throughput is best-effort by nature. You can mitigate that with bonding, smoothing, and over-provisioning, but you cannot guarantee a specific bitrate over cellular in the way you can book a specific allocation on a satellite transponder.
For everything else, which in practice covers 90 to 95 per cent of UK and European OB work, bonded cellular is the more practical and more economical choice.
Deployment examples
Regional news gathering, UK. A regional news operation replaced two satellite trucks with four HD4 MBX-based bonding kits, each in a Peli case with a battery, an omni MIMO antenna on a magnetic mount, and a compact H.265 encoder. Reporters deploy from estate cars rather than satellite trucks. Setup time dropped from 30 minutes to under 5 minutes. The annual saving on satellite truck hire exceeded £80,000. The bonding kits have been in daily use for over 18 months with a success rate (defined as "went live on time without visible quality issues") above 98 per cent.
Live music festival, outdoor multi-stage. A production company deployed two HD4 MBX units with directional antennas on 6-metre pneumatic masts to cover a three-day music festival. One unit fed the main stage programme to a London MCR. The other provided a backup path and carried production comms. Peak aggregate throughput across the weekend was 38 Mbps. The directional antennas were critical here: during headlining acts, the omni antennas dropped to single-digit throughput as 30,000 attendees hammered the local cell infrastructure, but the directional antennas maintained 15+ Mbps by targeting a tower on the far side of the site that the crowd's phones could not reach.
Motorsport, European circuit. An OB team covering a multi-day motorsport event used an HD4 MBX with four 5G SIMs (two on local operators, two on roaming UK SIMs) to deliver 1080p50 contribution feeds from the paddock. Latency was consistently under 900 ms glass-to-glass. The rig ran for four consecutive days, 10 hours per day, without interruption. Total data consumed across all four SIMs over the four days was approximately 480 GB.
Hardware summary
For broadcast OB, the two key Peplink components are:
Peplink HD4 MBX: Four embedded cellular modems (Cat-20 LTE or 5G depending on variant), dual-band Wi-Fi, Ethernet WAN and LAN ports, GPS, and a ruggedised chassis designed for vehicle and mobile deployment. This is the field unit that sits in the OB vehicle or flight case. It runs SpeedFusion natively.
FusionHub Solo: A virtual appliance that runs on any hypervisor or cloud platform. It acts as the SpeedFusion tunnel endpoint and provides the stable, public-IP decode point where your MCR or playout system receives the contribution feed. Licensed for a single SpeedFusion peer, which is all you need for a one-to-one OB contribution link.
Supporting hardware includes MIMO panel antennas (roof or mast mounted), low-loss coaxial cables, SIM cards from multiple operators, and whatever encoder your production workflow requires.
Planning your first cellular OB
If you are evaluating bonded cellular for OB contribution, here is a practical checklist.
Test before you commit. Borrow or hire an HD4 MBX and test it at your most challenging OB location. Not your easiest venue. The difficult one. The one in the valley, or the one where 50,000 people show up and destroy the cellular network. If it works there, it will work everywhere else.
Use four different operators. Do not put all four SIMs on the same network. The entire point of bonding is diversity. If one operator's tower is congested, the other three should still be performing. In the UK, use EE, Three, Vodafone, and O2 (or their MVNOs).
Set WAN Smoothing to High for live transmissions. Yes, it uses more data. Yes, the effective throughput drops. The reliability gain is worth it. You can drop to Low or Off for non-critical traffic like file transfers or IP comms.
Budget for antennas properly. A £5,000 router connected to the stubby antennas that came in the box will underperform a £5,000 router connected to properly specified, properly mounted external MIMO antennas. Antenna investment is not optional for broadcast-grade reliability.
Have a satellite or Starlink backup plan for genuinely remote locations. Bonded cellular is not magic. If there is no cellular coverage, there is no cellular bonding. Know your venue's cellular situation before you leave the office.
Engage with your FusionHub hosting early. Spin up the cloud instance, test the tunnel, confirm the throughput, and verify that your decoder can pull the SRT or RIST stream from FusionHub before the day of the broadcast. Do not leave cloud configuration to the morning of the event.
The bottom line
Bonded cellular does not replace satellite for every OB scenario. But for the majority of live broadcast contribution work in the UK and across Europe, it delivers equivalent or better performance at a fraction of the cost, with dramatically faster setup times and no dependency on satellite truck availability or orbital slot bookings.
The technology is mature. SpeedFusion bonding with WAN Smoothing is not experimental; it has been in production broadcast environments for years. The protocols (SRT, RIST) are broadcast-standard. The hardware (HD4 MBX, FusionHub Solo) is proven and field-tested.
The remaining challenge is confidence. Broadcast engineers who have relied on satellite for twenty years are understandably cautious about trusting cellular for a live contribution feed. The answer is not to argue theory. The answer is to test it, measure it, and deploy it alongside your existing satellite workflow until you trust the numbers. Most teams that do this end up cancelling the satellite truck within a few months.
The satellite truck is not dead. But for the vast majority of OB work, it is no longer necessary. And the economics are only going in one direction.