Time and coordinate frames
A position is meaningless without saying when it was, and in what frame. Most disagreements between two orbital tools come down to one of these two being different rather than to any physics being wrong.
Time
User-facing scenario times are UTC, written as RFC 3339:
2026-01-01T00:00:00ZInternally a run uses a deterministic epoch plus elapsed seconds, so a simulation is reproducible: the same inputs give the same outputs, independent of when it is executed.
| Scale | What it is | Where it appears |
|---|---|---|
| UTC | Civil time, with leap seconds inserted to track Earth rotation | All user input and output |
| UT1 | Time tied to actual Earth rotation angle | Frame conversion, via Earth orientation parameters |
| TAI | Continuous atomic time, no leap seconds | Intermediate, internal |
| TT | Terrestrial time, used for solar system ephemerides | Third-body positions |
Leap seconds are the trap. UTC is not continuous: it occasionally has a 61-second minute. Any arithmetic that assumes a fixed 86,400 seconds per day will drift against Earth rotation. This is why the high-fidelity frame path accounts for them explicitly and the fast path does not.
Epoch age
The single largest error source in propagating a real object is usually not the force model but the age of the state you started from. A two-line element set is a fit to observations around its epoch; propagating it far from that epoch extrapolates a fit rather than integrating physics.
Fresh elements with a simple model routinely beat stale elements with a sophisticated one.
Frames
| Frame | Rotates with Earth | Use |
|---|---|---|
| ECI, Earth-centered inertial | No | Propagation. Newton’s laws need a non-rotating frame |
| ECEF, Earth-centered Earth-fixed | Yes | Anything ground-relative: stations, coverage, footprints |
| LLA, latitude longitude altitude | Yes | Human-readable positions on the Earth model |
Propagation happens in ECI. Ground analysis happens in ECEF. Every coverage result therefore depends on a conversion between them, and that conversion depends on knowing precisely how far the Earth has rotated.
The branch in the middle is the whole story: everything ground-relative inherits whatever error the rotation model carries.
The two fidelity settings
| Setting | What it does | Cost |
|---|---|---|
fast (default) | Greenwich mean sidereal time only | Cheapest |
full | IAU-76/FK5 with leap seconds; also rotates SGP4 output from TEME into GCRF | Higher |
fast ignores precession, nutation, polar motion, and the difference between UT1
and UTC. For visualization and early trade studies that is fine. For a result
that will be compared against another tool, the residual shows up as a fixed
angular offset in longitude, which reads as a systematic ground-track shift.
SGP4 produces its output in TEME, which is not the same frame as the GCRF used
elsewhere. The full path performs that rotation. Under the fast path the
difference is folded into the same approximation as everything else. If you are
propagating TLEs and comparing ground tracks against another tool, this is the first
thing to check.
Earth orientation parameters come from an IERS finals2000A.all file. When one
is not supplied the system degrades gracefully to zero EOP corrections rather
than failing, which is convenient and worth knowing about, because the degraded
result looks exactly like the corrected one.
What a run records
Every simulation records enough metadata to be audited later:
- Scenario start time, step size, duration
- Propagation model, and force model where applicable
- Earth model and coordinate frame
- Software version
That list is what makes a result reproducible months later. A number without its frame, epoch, and model is not a result; it is a rumor.
Next steps
- Units and conventions for sign conventions and angle units.
- Propagation fidelity for the force models these frames carry.
main (pre-release)