Hold time to a
razor's edge.
The O-DU and O-RU must agree on time to ±1.5 µs — the width of light travelling 450 m — or the cell falls apart. Here we derive that budget, distribute time with PTP (G.8275.1 full timing) and SyncE, place the four sync topologies LLS-C1 → C4, follow the Annex-H error waterfall, and diagnose a clock in HOLDOVER → FREERUN through the M-plane. Verbatim from O-RAN WG4 CUS.0 v21 §11 + Annex H.
The S-plane — ±1.5 µs, PTP, and the four topologies
The player screen is a live animation stage — clocks tick, the ±1.5 µs budget appears, the four LLS-C topologies redraw their timing path, the error waterfall accumulates, and a clock slips from LOCKED to FREERUN, exactly as the narration reaches each idea. Karaoke subtitles EN + Persian, fullscreen.
The four sync topologies
The configurations differ in one thing: the path timing takes to reach the O-RU. Click one to see where the grandmaster sits and how time flows.
LLS-C1 → C4 — who owns time
All four hold the same ±1.5 µs relative budget; they differ in where the grandmaster sits and how many boundary clocks time must cross.
| Config | Grandmaster | Path to O-RU | When you use it | Clause |
|---|---|---|---|---|
| LLS-C1 | the O-DU | direct cable, no switches | simplest · most timing margin | §11.2.2.2 |
| LLS-C2 | the O-DU | bridged fronthaul (boundary-clock switches) | one O-DU → many O-RUs | §11.2.2.2 |
| LLS-C3 | a T-GM in the transport network | PRTC → T-GM → T-BCs → both O-DU & O-RU | timing offered as a network service | §11.2.2.3 |
| LLS-C4 | a local GNSS at the O-RU | none — no transport involvement | fronthaul timing absent / untrusted | §11.2.2.4 |
For LLS-C1/C2, "the relative time error of the S-plane measurement signals between the O-DU and O-RU is within the limit of 3 µs (±1.5 µs)." That single number, driven by the air interface, governs how many boundary clocks your fronthaul may contain — the Annex H waterfall must still fit inside it at the O-RU.
How time is distributed
| Mechanism | Carries | Profile | Note |
|---|---|---|---|
| PTP (IEEE 1588) | phase + time | ITU-T G.8275.1 — full timing support | every switch is a boundary clock |
| PTP — partial | phase + time | ITU-T G.8275.2 — partial timing support | for networks that cannot upgrade every hop |
| SyncE / PLFS | frequency only | physical-layer frequency signal | keeps frequency rock-steady beneath PTP |
PTP carries phase and time; SyncE carries frequency. Full timing support (G.8275.1) is the O-RAN gold standard: each switch recovers time and re-launches it cleanly, so error does not pile up uncontrolled. Partial timing support exists for brownfield transport, but you design for full timing wherever you can.
Locked, holdover, freerun
| State | Meaning | Trustworthy? |
|---|---|---|
| LOCKED | tracking its reference — all is well | yes |
| HOLDOVER | lost reference, coasting on last known rate | temporarily — error grows |
| FREERUN | never locked, or holdover expired | no — stop transmitting |
Time error accumulates hop by hop: PRTC (GPS) → T-GM → T-BC → T-BC → … → O-RU. Each hop is allocated a slice of the ±1.5 µs. Add one boundary clock too many and the budget no longer closes — the hop count is a hard design limit, not a detail. Allowed clock classes per §11.3.1.
🛠 Engineer detail block — what you'll actually meet in the field
- Read six values before you touch the radio. Off the O-RU (MP.0 §13): lock-state, PTP status, GNSS status, offset-from-master, holdover remaining, and the sync source. A "coverage hole" is a sync fault nine times out of ten.
- Hop count is a hard limit. Each boundary clock spends time-error budget. If a site needs one switch too many, LLS-C3 or a local GNSS (C4) may be the only way to close ±1.5 µs.
- FREERUN means stop, not degrade. A well-behaved O-RU that loses sync ceases transmitting rather than radiating on a wrong clock — so a sync fault takes the cell down cleanly, which is a feature.
- SyncE without PTP is not enough. Frequency alone gives you no phase/time. You need PTP (G.8275.1) for the ±1.5 µs phase alignment; SyncE only stabilises the frequency underneath it.
- Holdover is a countdown, not a fix. The longer an O-RU coasts, the further it drifts. Treat every holdover alarm as a ticking clock and chase the upstream reference immediately.