Time & reference frames
A position only means something with a time and a frame attached. The Orbit Visualizer works in UTC, renders in the J2000 (EME2000) inertial frame, converts SGP4's native TEME states into it, and rotates to an Earth-fixed frame through GMST for ground tracks and access. This page states each convention — and its documented approximations — honestly.
Time: UTC and epochs
All times in the app are UTC. An epoch is the instant at which a set of elements or a state vector is valid; propagation carries that state forward or back from its epoch. Internally the app uses a Julian Date defined as jd = unixMillis / 86400000 + 2440587.5. One deliberate simplification: the frame math uses UTC in place of Terrestrial Time (TT). In 2026 that difference is about 69 seconds, which enters the rotation angles at roughly 10−7 radians — a documented, standard approximation for TLE work and far below SGP4's physical accuracy.
What an inertial frame is
An inertial frame is one that does not rotate or accelerate — its axes stay fixed relative to the distant stars. Newton's laws (and therefore the clean form of the orbital equations) hold in such a frame, so orbits are naturally described and integrated there. A frame fixed to the spinning Earth is not inertial; the Earth turns underneath the orbit, which is exactly why a ground track drifts westward.
J2000 / EME2000
J2000 (also written EME2000, Earth Mean Equator and Equinox of J2000.0) is the inertial frame the Orbit Visualizer uses for its 3-D scene and for exported ephemeris. Its axes are pinned to the Earth's mean equator and the vernal equinox at the reference epoch 2000 January 1, 12:00 TT, so they no longer move with the slowly precessing real equator. When you set the camera reference to ECI, you are looking at this frame.
TEME and the conversion to J2000
SGP4/SDP4 does not output J2000. Its native frame is TEME (True Equator, Mean Equinox), a frame peculiar to the SGP4 theory. The app therefore rotates every SGP4 state into J2000 through an explicit, documented chain, following Vallado's TEME reference:
- TEME → TOD by the geometric equation of the equinoxes (the small kinematic terms are omitted, per Vallado's TEME convention).
- TOD → MOD by the IAU-1980 nutation series, truncated to its 6 largest terms of the full 106-term series.
- MOD → J2000 by IAU-76 precession (the ζ, θ, z polynomials).
That nutation truncation is the single largest deliberate approximation: it leaves a frame error of |Δψ| < ~0.1″ (arcsecond level) — a few metres at LEO radius, well below SGP4's own kilometre-level physical accuracy. Polar motion and the full IAU-2006/2000A reductions are out of scope for this build.
Earth-fixed frames and GMST
Ground tracks, ground sites, and access passes need an Earth-fixed (ECEF) frame that rotates with the planet. The app obtains it by rotating about the Z axis through the Greenwich Mean Sidereal Time (GMST) angle (evaluated from the IAU-82 GMST polynomial), with no polar motion — the standard TEME + GMST → ECEF approximation. J2000 objects use the same GMST rotation. Geodetic latitude, longitude, and altitude then come from the WGS-84 ellipsoid (via Bowring's method). Switching the camera reference to ECEF shows this rotating frame, in which a geostationary satellite appears to hover.
The RIC frame (radial / in-track / cross-track)
Relative motion — in conjunction screening and in maneuver planning — is expressed in a local RIC frame built from a chief object's state (r, v):
- Radial (R̂) = r / |r| — straight up, away from Earth.
- Cross-track (Ĉ) = (r × v) / |r × v| — along the orbit normal.
- In-track (Î) = Ĉ × R̂ — completing the right-handed triad, roughly along the velocity.
See conjunction screening and maneuvers, which use this frame.
▶ Load the ISS and toggle the ECI/ECEF camera reference
Related
Mini-FAQ
Why not use the full IAU-2006/2000A reductions?
Because the residual from the truncated nutation and omitted polar motion is a few metres at LEO — hundreds to thousands of times smaller than SGP4's own kilometre-level accuracy. The full reduction would add complexity without changing any answer the app can meaningfully give.
Does using UTC instead of TT matter?
Negligibly. The ~69-second offset in 2026 enters the frame rotation at about 10−7 radians, which is a documented, standard approximation for TLE-based work and far below the model's physical error.