Mission library · Media & events

Field newsgathering when everyone is streaming.

The bigger the story, the worse the uplink: the crowd that makes a scene worth covering is the same crowd saturating every network around it. This mission is about a backpack kit that treats the carriers as one pool — harvesting whatever capacity exists, holding a floor under the live shot, and moving with the crew instead of resetting on air.

Who runs it

News crews, freelance journalists and agency field desks covering protests, disasters, elections and celebrations.

What breaks

The crowd saturates every network at exactly the scene worth covering — and the uplink, the direction the live shot needs, dies first.

What Atlas contributes

Carriers that saturate differently, bonded into one encrypted tunnel; a reserved floor under the return feed; sessions that roam when the crew moves.

Runs on

A backpack-sized Linux computer between the camera chain and the modems the crew already carries. One binary, one configuration file.

The mission

A march that turned into something bigger, a flooded town, an election night in the main square. The crew arrives with what a crew carries: a camera, an encoder, a backpack of modems, and a slot on the evening programme that will happen whether the uplink cooperates or not. The desk wants a live shot on the hour, raw files between hits, and a crew it can actually reach. Everything about the assignment is routine — except that the story has drawn ten thousand people, and every one of them is streaming from the same cells the crew needs.

What breaks

Crowds do not take the network down; they take its capacity — and they take the uplink first, because the crowd is streaming outward too, in exactly the direction the live shot needs. A single carrier that carried a clean rehearsal at noon delivers stutter and multi-second latency at the head of the march, and it fails worst at the peak of the story, because news value and phone use spike together. Displacement compounds it: crews move — down a street, onto a rooftop, ahead of a police line — and a conventional single-link chain treats every move as a disconnection, with the stream re-established while the programme waits.

The classical answers each fall short somewhere. Satellite trucks cannot park inside the crowd and cannot follow it. Bonded-cellular hardware modules help with saturation but weld the newsroom to one vendor’s appliance and its per-unit price. And carrier priority schemes, where they exist, belong to public safety — not to the press standing in the same square.

The architecture on this mission

Saturated-scene bonding geometry A backpack field kit in a crowded square bonds two carriers that still have capacity while a third, saturated by the crowd, is left behind; the feed reaches the broadcast centre over both good paths at once. the square — ten thousand phones, all streaming upward CARRIER A CARRIER B CARRIER C · SATURATED CARRIERS THAT SATURATE DIFFERENTLY Field kit backpack, beside the camera Broadcast centre master control
The scene: the kit bonds the carriers that still have capacity while the one the crowd has flattened is left behind (dashed). Different networks have different cells, different backhaul and different subscriber mixes — they do not fail identically at the same moment, and the bond harvests the difference.

The kit is a small Linux computer between the camera chain and the modems, running one node that bonds two or three carriers into a single encrypted tunnel to the broadcast centre. The load-bearing observation is the same one that saves the stadium mission: different networks do not saturate identically or simultaneously. Their cells, backhaul and subscriber mixes differ, so their worst minutes interleave — and a bond across them, probed several times a second and reweighted in well under a second, delivers capacity no single carrier at the scene can. The encoder is unmodified; to it, the tunnel is simply a network that behaves.

On top of the bond, the class system holds the line the desk cares about. The return feed rides the video class with a reserved floor; file transfers, the raw-footage drain and everything else elastic yield to it the moment capacity tightens. Degradation is ordered: files pause, the live picture sheds bitrate, and the talkback to the gallery — a few kilobits — does not flinch. When measured loss climbs across every path at the peak of the story, adaptive redundancy duplicates the critical classes across carriers, spending scarce bandwidth on certainty exactly when certainty is scarce.

And because crews move, the tunnel moves with them. Sessions ride the bond’s stable addresses, not any one modem’s: when the crew displaces two streets over mid-broadcast, carriers drop and reappear beneath the tunnel while the shot stays up. What could not be sent live queues on the kit and drains automatically, priority first, the moment any path improves — the raw files arrive late instead of never.

How the story unfolds

  1. Arrival. The crew walks in with the backpack; the kit bonds the modems it carries into one tunnel to the broadcast centre. The first hit goes out clean; the desk’s dashboard shows every carrier’s measured state at the scene, not bars on dongles.
  2. The crowd builds. The cells begin to load, unevenly. The bond shifts weight toward whichever carrier currently has uplink left — continuously, with no switchover and nothing for the operator to press.
  3. The peak. The story breaks properly and ten thousand phones stream at once. Ordered degradation engages: file transfers pause, the live picture sheds bitrate, and the reserved floor under the return feed and the talkback holds.
  4. The displacement. The crew moves down a side street to stay with the story — during the live shot. Carriers vanish and reattach under the tunnel as they go; the session roams with the kit and the shot never resets on air.
  5. Between hits. From the new position the queued raw files drain in priority order over whatever capacity the carriers recover. By the next hit the desk has the footage it thought the day had lost.
  6. The debrief. The journal holds a per-carrier, per-minute record of the whole assignment — which network held at that square and which folded. The next crew sent to the same scene goes in knowing, not guessing.

What each mechanism contributes

  • Multi-carrier bonding — carriers that saturate differently combined into one tunnel; capacity harvested wherever it currently exists. Shipped.
  • Class floors and ordered degradation — the return feed and talkback reserved; files elastic; classification reads the traffic itself, the encoder unmodified. Shipped, opt-in.
  • Endpoint roaming — the crew displaces mid-broadcast and sessions move with the kit instead of resetting on air. Shipped.
  • Adaptive redundancy and store-and-forward — critical traffic duplicated across carriers as measured conditions worsen; what cannot be sent live queues and delivers on recovery. Shipped.
  • Continuous measurement and the journal — a per-carrier record of what every network did at the scene, for the debrief and the next assignment. Shipped.

The honest boundary: Atlas cannot conjure capacity the cells do not have — at the absolute peak of saturation the guarantee is ordered degradation and the best use of what exists, not a fixed bitrate. The cameras, the encoder, the editorial judgement and the crew’s safety in the crowd are the newsroom’s; Atlas carries the bits and proves afterwards what each carrier actually did.

What a pilot should prove

  • Side-by-side link records from a real crowded event: each single carrier versus the bond, from the built-in measurement.
  • Live-shot continuity through the worst measured interval of the event — floor held, files yielding, bitrate shed before continuity.
  • A displacement mid-stream: the crew moves position during a live hit and the session roams without an on-air reset.
  • The debrief artefact: a per-carrier journal of the scene, usable by the desk to plan the next deployment to the same ground.

One kit, one crew, one big story. The evaluation format covers the structure.

Request a briefing

Equipping crews for saturated scenes?

Bring one kit, one crew and the carriers they already carry — we will define what a pilot deployment should prove.