sidereon

GNSS and astrodynamics for Python, with answers you can trust.

sidereon is the Python interface to the sidereon engine: a single GNSS and astrodynamics core, written in Rust, exposed here as an idiomatic Python package. You get orbit propagation, precise positioning, frames and time, ephemeris handling, and format parsing through plain Python objects and numpy arrays. The engine is reference-validated (SGP4 is bit-exact to Vallado's implementation; frames and time check against Skyfield and IERS; the positioning stack checks against IGS products), so the numbers match the sources the field already trusts.

The Rust core is compiled into the wheel and linked statically, so the package is fast and the only runtime dependency is numpy. There is no separate native install.

Install

pip install sidereon

import sidereon
print(sidereon.__version__)

Example: where is the ISS in the sky right now?

No data files and no setup: give it a two-line element set and a ground station, and ask for the look angles. Everything that takes time takes unix microseconds (int64) and arrays propagate in one call.

import numpy as np
import sidereon

tle = sidereon.Tle(
    "1 25544U 98067A   24001.50000000  .00016717  00000-0  10270-3 0  9009",
    "2 25544  51.6400 208.8657 0002644 250.3037 109.7782 15.49560812999990",
)
station = sidereon.GroundStation(latitude_deg=51.5, longitude_deg=-0.1, altitude_m=10.0)

# Epochs are unix microseconds (int64); arrays propagate in one call.
epochs_us = np.array([1_704_110_400_000_000], dtype=np.int64)
look = tle.look_angles(station, epochs_us)
print(look.azimuth_deg, look.elevation_deg, look.range_km)

Tle also gives you propagate() (TEME state arcs as numpy arrays) and find_passes() (rise/set/peak over a window). The positioning side has the same pattern: a typed config in, a result object with numpy positions and scalar attributes out.

import sidereon

sp3 = sidereon.load_sp3(open("igs_product.sp3", "rb").read())

config = sidereon.SppConfig(
    observations=[
        sidereon.SppObservation("G08", 23_825_519.8),  # PRN, pseudorange (m)
        sidereon.SppObservation("G10", 22_717_690.1),
        # ...more satellites
    ],
    t_rx_j2000_s=646_272_000.0,
    t_rx_second_of_day_s=43_200.0,
    day_of_year=176.5,
    initial_guess=[4_500_000, 500_000, 4_500_000, 0.0],
    corrections=sidereon.SppCorrections(ionosphere=True, troposphere=True),
    with_geodetic=True,
)

solution = sidereon.solve_spp(sp3, config)
print(solution.position)     # numpy [x, y, z] ECEF metres
print(solution.rx_clock_s)   # receiver clock bias, seconds

PROJ-compatible EGM96 vertical-grid interpolation is available when bit-level agreement with a particular PROJ build matters. Load the public egm96_15.gtx grid and select the reference build's floating-point arithmetic explicitly:

import sidereon

grid = sidereon.GeoidGrid.from_proj_egm96_gtx(open("egm96_15.gtx", "rb").read())
undulation_m = grid.undulation_proj_rad(
    latitude_rad,
    longitude_rad,
    sidereon.ProjVgridshiftArithmetic.FUSED_MULTIPLY_ADD,
)

Use SEPARATE_MULTIPLY_ADD for a PROJ build that does not contract the multiply-add operations. Invalid coordinates raise a typed ProjVgridshiftError subclass instead of panicking or extrapolating.

Capabilities

The Python package mirrors the full breadth of the engine.

• Orbit propagation: SGP4/SDP4 from TLE/OMM, numerical propagation with a composable force model (spherical-harmonic geopotential to selectable degree and order, Sun/Moon third-body, solar radiation pressure, relativistic correction, atmospheric drag) and orbital-decay estimation with a post-decay validity latch, batch and constellation arcs, pass prediction, look angles, and coverage analysis.
• Orbital mechanics: classical, equinoctial, and modified equinoctial elements, anomaly conversions and Kepler propagation, Lambert transfers, initial orbit determination (IOD), batch least-squares orbit fitting against precise ephemerides (including terrestrial-frame SP3 products through the Earth-orientation chain) with a per-satellite residual ledger, and relative motion in RIC/RTN/LVLH frames with Clohessy-Wiltshire propagation.
• GNSS positioning: single-point positioning (SPP), public solve_static multi-epoch static positioning with covariance, RINEX observation assembly and solve helpers (spp_inputs_from_rinex_obs and solve_spp_from_rinex_obs), leave-one-out redundancy diagnostics, and robust weighting, RTK (float and fixed), PPP (float and fixed), static PPP temporal-correlation covariance with calibrated day-length bounds, optional elevation cutoff, optional tropospheric-gradient estimation, DGNSS, a Huber-reweighted solve driver, and DOP.
• Integrity and error bounds: RAIM fault detection and exclusion, multi-constellation ARAIM protection levels, SBAS protection levels (DO-229), per-observation reliability (minimal detectable bias, internal/external), typed RAIM inputs and RAIM over existing SPP solutions, broadcast-ephemeris FDE, observability classification of every solution (rank, redundancy, conditioning), and covariance-derived error metrics (CEP, R95, SEP, error ellipse) that report wide or flagged bounds for weak geometry rather than fabricated confidence.
• GNSS corrections and products: SBAS and RTCM SSR corrections applied to broadcast ephemeris, RTCM 3 broadcast ephemeris decode for GPS (1019), GLONASS (1020), Galileo (1045/1046), BeiDou (1042), and QZSS (1044), each real-data validated, Bias-SINEX code and phase biases (DCB/OSB), Klobuchar and NeQuick-G ionosphere, IONEX maps, troposphere models, top-level SBAS/SSR decode and SSR store helpers, and NTRIP client stream handling.
• Ephemeris and time: broadcast ephemeris and precise SP3 products, JPL SPK (DAF/.bsp) kernels, uniform satellite-state sampling across broadcast and precise sources with batched multi-satellite interpolation, scale-aware time (UTC/TAI/TT/TDB/UT1 and the GNSS system times) with leap-second handling, and Earth orientation parameters (EOP).
• Timing and clocks: Allan-family stability analysis (ADEV/MDEV/HDEV/TDEV) and power-law clock-noise identification with a five-coefficient fit (IEEE 1139).
• Estimation and detection: scalar Kalman and alpha-beta trackers, innovation gating, MAD statistics, CFAR detection thresholds, and source localization (ToA/TDOA) from arrival times at known sensors.
• Geodesy and monitoring: geodesic direct and inverse problems (Karney), an epoch-aware terrestrial reference frame catalog with published ITRF and ETRF Helmert parameter sets, station velocity (MIDAS), trajectory fitting with seasonal terms and offsets, step detection, network motion fields with common-mode removal, and repeating-geometry (sidereal) filtering.
• Geometry and events: reference frames, geodetic and ECEF conversions, look angles, eclipse, conjunction screening with collision probability, and angular geometry (separation, position angle, phase angle, beta angle).
• Observation and almanac: apparent places for the Sun, Moon, and any SPK body (astrometric and apparent RA/Dec plus az/el, with refraction and polar motion), sub-solar and sub-observer points, the terminator, parallactic angle, satellite visual magnitude, moonrise/moonset, seasons, moon phases, planetary events, meridian transits, and lunar and solar eclipses.
• Observation quality and integrity: RINEX observation QC (completeness, multipath, cycle slips), post-solve RAIM fault detection, ARAIM protection levels, carrier-phase combinations, and Hatch smoothing.
• Terrain: DTED elevation lookup with batch probes, a memory-mappable terrain store with tile-list builders and store byte/tile metadata, and geoid height conversion from EGM96 and EGM2008 grids, including PROJ 9.3.0 EGM96 GTX interpolation with explicit fused or separately rounded arithmetic.
• RF: link budget (FSPL, EIRP, C/N0, antenna gain).
• GNSS/INS fusion: strapdown mechanization with an error-state EKF (UKF option), loose and tight coupling, IGG-III loose updates with an outlier guard, an RTS fixed-interval smoother, checkpointed time synchronization, a serializable filter state, and field mode: zero-velocity and zero-angular-rate updates with a stationarity detector, non-holonomic vehicle constraints, per-fix-status measurement weighting, and an IMU-to-body mounting matrix, all off by default.
• Reference-station static solve: rover and reference observations across epochs in, one station coordinate with covariance and typed per-mode errors out, selecting fixed carrier, then code-DGNSS, then float.
• Scenario simulation: deterministic synthetic GNSS observables from a versioned scenario with a per-term error budget, plus a ground-truth ledger attributing solver error to each term; same scenario and seed give identical bytes.
• Signal analysis: closed-form BPSK/BOC spectra, spectral separation coefficients, DLL thermal-noise jitter, and multipath error envelopes, validated against published constants.
• Formats: parse and serialize TLE/OMM, CCSDS OEM/OPM/CDM/TDM, RINEX, CRINEX, SP3, IONEX, ANTEX, Bias-SINEX, SBAS logs, RTCM, and NMEA.
• Data acquisition: the sidereon.data module downloads and caches GNSS products (SP3 and IONEX from IGS/MGEX analysis centers, including merged multi-center SP3) and DTED terrain tiles. Exact-product requests can select direct archives, NASA CDDIS/Earthdata, local files, or caller-provided bytes without changing product identity; see Public GNSS distribution sources.

RAIM residual tests must be weighted with per-satellite residual variances. The Python API takes inverse-variance weights, so build them from your range noise model instead of using unit weights on metre-scale residuals:

import math
import sidereon

used_sats = ["G01", "G02", "G03", "G04", "G05", "G06"]
residuals_m = [0.4, -0.6, 0.3, 0.1, -0.2, 0.5]
elevation_deg = {"G01": 72.0, "G02": 42.0, "G03": 35.0, "G04": 64.0, "G05": 50.0, "G06": 28.0}
base_sigma_m = 0.8
variances_m2 = {
    sat: (base_sigma_m / max(math.sin(math.radians(el)), 0.2)) ** 2
    for sat, el in elevation_deg.items()
}
weights = {sat: 1.0 / variances_m2[sat] for sat in used_sats}
result = sidereon.raim(used_sats, residuals_m, weights=weights)

The binding adds no modeling of its own: every result is exactly what the engine computes, returned as numpy arrays, typed objects, and real Python exceptions (sidereon.SidereonError and friends). Full signatures live in the bundled type stubs (sidereon/__init__.pyi).

One engine, every language

sidereon is one validated core with first-class interfaces, so the numbers are the same everywhere:

• sidereon: the Rust core and engine
• sidereon-c: C interface
• sidereon-ex: Elixir interface
• sidereon-wasm: WebAssembly interface

See the live demo and docs at sidereon.dev.

Building from source

For contributors: pip install maturin, then maturin develop from the repo. Run the tests with pytest.

License

MIT.