Aligned Output (Cross-Node Alignment)

Two edge nodes forwarding the same contribution feed do not put it on the wire at the same moment. Each emits when its own network path delivers, so they differ by the difference in path latency. Cut between them on a downstream switcher and you get a timing jump.

An alignment group removes that difference. The manager works out one shared timeline, hands it to every member node, and each edge holds its packets until that timeline says to release. Every member then runs the same arithmetic on the same numbers, so they line up analytically rather than by chasing each other.

The result is a clean cut between nodes. It is not a seamless SMPTE 2022-7 merge — the two streams remain separate streams, they simply arrive in step.

Find it at Alignment in the sidebar. The entry appears only when at least one of your nodes supports it.

What can join a group

This is deliberately narrow. Each member is one UDP or RTP output, on a flow that:

• has exactly one input;
• is not an assembled (PID-bus) flow;
• has its input set to passthrough clock, or uses a bonded input;
• has the output set to PCR egress pacing, with no transcoding on it.

A group needs at least two members and at most eight. A node can only appear once — two outputs on the same box already share a clock, so aligning them against each other measures nothing.

Alignment and clock regeneration cannot both be on

Alignment works by reading the timing that is already in the stream. That only works if the node forwards it untouched, which is what passthrough clock means.

Turning alignment on therefore means giving up PCR/PTS regeneration on that input. If you rely on the node re-stamping timing for a downstream receiver, alignment is not available for that flow.

The manager checks all of this for you as you add each member, and tells you exactly which condition a candidate fails.

Creating a group

1. Go to Alignment and click + New group. Give it a name.
2. Set the offset — see Choosing the offset below.
3. Add members using the node → flow → output picker. Each row runs a live readiness check.
4. Save.

The group starts at Pending while it collects a reading from every member, then moves to Armed.

The same offset must also be set on each output itself, under Configure → Outputs on that node. The manager verifies the two agree and refuses the group if they do not.

Why the manager verifies the offset instead of pushing it

Changing the hold time on a live output steps its output timing. That is a disruptive act, so it belongs in a config change you make deliberately — not something that happens as a side effect of editing a group.

A mismatch is worth catching precisely because it is invisible otherwise: the group would be misaligned by exactly the difference, and every node would still report healthy.

Choosing the offset

The offset is the headroom each node has to hold packets before releasing them. It is bounded 150–800 ms.

It is headroom for the difference between your nodes, not for total latency. If your two paths are 200 ms and 260 ms, the spread is 60 ms — not 260 ms. Sizing the offset against total end-to-end latency is the common mistake, and it pushes the output past what the edge will hold, at which point it sheds most of the stream.

Start at the default and adjust using the two states the page shows:

State	What it means	What to do
Deficit	A member released late — its path needs more headroom than the offset allows.	Raise the offset.
Clamped	A member released early — the offset asks it to hold longer than it will.	Lower the offset.

These call for opposite remedies, which is why they are shown as separate states rather than one “misaligned” warning.

Reading the group state

State	Meaning
Pending	Collecting readings from members. Not yet aligned.
Armed	Shared timeline published, and every member is holding it.
Degraded	Still armed and still on air, but at least one member has fallen off the timeline.
Spread too wide	The measured difference between nodes is larger than the offset can absorb, so the group deliberately did not arm.

Spread too wide is a refusal, not a failure

Arming a group whose members are further apart than the offset can bridge would align nothing while showing green. The manager would rather tell you. Either raise the offset, or reduce the difference between the paths.

Degraded exists because “armed” would otherwise mean two different things at once — that the timeline was published, and that the group is actually aligned. A source restart takes every member off the timeline while leaving the published timeline perfectly intact. Degraded members stay on air; the group recovers in place once the cause clears.

Re-minting

If a group has drifted, press Re-mint. This genuinely re-derives the timeline from current traffic: it withdraws the old one from every member first, waits for fresh readings, then re-arms all of them on the same packet.

Expect a brief step in output timing while this happens. That step is the correction.

Alarms

Alarms land on the member node, under category alignment, so they appear next to that node’s other events.

Event	Meaning
alignment_member_deficit	Releasing late — raise the offset.
alignment_member_clamped	Releasing early — lower the offset.
alignment_member_disengaged	Fell back to local pacing. Running, but not aligned.
alignment_member_not_engaged	Holding the timeline but no stream is reaching it — the feed to that node has stopped.
alignment_member_unreachable	Stopped reporting.
alignment_offset_mismatch	That output’s configured offset disagrees with the group’s.
alignment_generation_skew	Stuck on an older timeline. Usually heals by itself within a few seconds.
alignment_spread_exceeded	The nodes are further apart than the offset can absorb.

Requirements and limits

• Every member node must support alignment. Older nodes ignore the setting silently — which looks exactly like success — so the manager hides the controls for them rather than letting you configure something that will not happen. Upgrade the node if it does not appear in the picker.
• Compressed UDP/RTP forwarding only. Two independent encoders never produce the same bitstream, so transcoded outputs can never be aligned this way.
• This gives a clean cut, not a hitless merge. For hitless, use SMPTE 2022-7 redundancy on a single node.

See also

• Master Clock & A/V Sync — how the edge derives output timing, and why a node’s own clock cannot be shared
• Wire-Time Precision — the release mechanism alignment builds on
• Configuration — the per-output epoch_lock fields