ECSS Communications — Time Synchronization (space-systems/ecss/e50-time-synchronization)
Use when two spacecraft on an inter-spacecraft network have to keep their clocks together, per ECSS-E-ST-50C clause 5.7.4.5 — how well one exchange pins the offset, and how long that answer survives before the next one.
Domain quick reference
- Synchronization between a pair is a different problem from distributing a reference down a tree. Here there is no master to walk away from: the two clocks are measured against each other, and the quantity under control is the offset between them.
- The four-timestamp exchange cancels whatever is reciprocal about the path. Request and reply share the same range and the same media, so the common part drops out of the offset and only the difference survives — which is why the whole budget is a list of the ways the two directions are not the same.
- Relative motion breaks reciprocity even on a perfect link. While the responder holds the reply the range changes, so the reply travels a different distance from the request; the error scales with range rate and with hold time, which makes the hold time a design parameter rather than an implementation detail.
- An offset of any size is the measurement, not an error. A responder clock reading minutes away from the initiator is exactly what the exchange exists to find, so ordering checks apply within each clock's own frame and never across the two.
- Between exchanges the pair walks apart at the relative drift rate, not at either clock's absolute rate. Two oscillators of the same build can hold each other far better than either holds an absolute scale, and budgeting with the absolute figure over-synchronizes.
- The interval is the knob, and it inverts cleanly. The window less the exchange error, divided by the relative drift, is the longest interval that still holds — and when the exchange alone spends the window, no interval does.
Workflow
- Take the four timestamps of one exchange and check ordering inside each clock's own frame only. Reject a reply that precedes its request, or a responder hold longer than the round trip.
- Recover the round trip with the responder hold removed, the one-way delay, and the clock offset.
- Budget the exchange error: half the static path asymmetry, half the motion asymmetry from range rate over the hold time, half the timestamp resolution, and the jitter carried separately.
- Compute how far the pair walks apart over the synchronization interval at the relative drift rate.
- Compare the sum against the coordination window with a relative tolerance, so a budget landing exactly on the window passes on every build host.
- Invert for the remedy: the longest interval that still fits. Report zero where the exchange alone spends the window, and unconstrained where the pair has no declared relative drift.
- Say which term dominates. Where motion beats static asymmetry, the cheaper fix is a shorter responder hold, not a better cable.
- Where the network carries more nodes than this pair, the exchange should be repeated between each node and the one holding the common reference, so that what the service reports is alignment to that reference rather than agreement between two clocks that have both walked away from it.
Obligations
| Item | Step |
|---|---|
| ECSS-E-ST-50C Rev.2 5.7.4.5a | 8 |
Pitfalls
- Treating a large measured offset as a bad exchange. The offset is the output; validating timestamps across the two clocks rejects exactly the case the exchange was built to measure.
- Leaving the responder hold time out of the round trip. Including it inflates the one-way delay and biases the offset by half the hold, which is usually larger than every other term combined.
- Assuming reciprocity because the link is symmetric. Range rate breaks it without any hardware asymmetry at all, and a formation closing at kilometres per second breaks it by more than the cable ever could.
- Budgeting the interval with an absolute oscillator stability figure. The pair is graded on relative drift, and using the absolute number buys synchronization traffic the design did not need.
- Deciding the window with a bare inequality. Two arithmetically identical budgets can straddle the window on different machines, and the verdict then depends on the build host.
- Answering a missed window with a shorter interval without checking the exchange error first. When the exchange alone spends the window, synchronizing more often changes nothing and the honest answer says which term to attack instead.
Behavior contract (gate 3)
Timestamp validation inside each clock frame, offset and round-trip recovery, the half-weighted asymmetry and resolution terms, the range rate and hold-time motion term, the relative-drift walk, the exact window boundary, the longest-interval inverse with its zero and unconstrained cases and the dominant-term finding are exercised by the gate 3 contract test: scripts/test_e50_time_synchronization.py against scripts/e50_time_synchronization_logic.py (stdlib unittest, offline). Run: python3 scripts/test_e50_time_synchronization.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
