geoveil-mp

GNSS code multipath analysis — Rust core, Python API

Anubis/TEQC-style code multipath estimation for every pseudorange observable in a RINEX file — per-code MP linear combinations, arc-wise ambiguity debiasing, interval-aware cycle-slip detection, and SNR series export. Built for continuous station monitoring on long time series of 1 s or 30 s data, following the methodology of Hunegnaw & Teferle (Sensors 2022).

Part of the GeoVeil suite together with geoveil-cn0 (CN0 signal quality and threat detection).

Live demo: batch.geoveil-rinex.eu — the GeoVeil batch dashboard runs this library in production: per-code MP RMS, cycle-slip counts, SNR-residual wavelet spectra, Fresnel zone maps, and long-term multipath trend monitoring on daily 30 s station data.

Website: miluta7.github.io/geoveil-mp

What's new in 1.0

• SNR wavelet analysis in Rust. Morlet continuous wavelet transform of detrended SNR per satellite, 95 % red-noise significance, and band power in period bands matched to the sampling interval.
• First Fresnel zone mapping. Zone semi-axes and specular-point distance for every track cell, for any antenna height and carrier (Hunegnaw & Teferle 2022, eq. 6).
• Same numbers, about 20x faster than the numpy reference that runs on the GeoVeil platform. Satellites run in parallel and the GIL is released.
• License: PolyForm Noncommercial 1.0.0 with attribution. Versions up to 0.2.1 remain MIT. See License.

Installation

pip install geoveil-mp

Pre-built wheels for Linux, Windows, and macOS — no Rust toolchain required.

# Rust
[dependencies]
geoveil_mp = "0.2"

Quick Start

import geoveil_mp as gm

obs = gm.read_rinex_obs("BOR100POL_R_20240010000_01D_30S_MO.rnx")
print(obs.num_epochs, obs.num_satellites, obs.interval)   # 2880 104 30.0

analyzer = gm.MultipathAnalyzer(obs, elevation_cutoff=10.0, systems=["G", "R", "E", "C"])
results = analyzer.analyze()

# One statistics row per signal code: GPSM1C, GPSM2W, GALM8X, GLOM1P, BDSM2I, ...
for s in sorted(results.statistics, key=lambda s: s.signal):
    print(f"{s.signal}  rms={s.rms:.3f} m  n={s.count}  slips={s.cycle_slips}")

# Cycle slips, counted per carrier-phase signal
print(results.cycle_slip_counts)          # {'GPSL1C': 210, 'GPSL2W': 168, ...}

# Attach precise elevations from SP3 (recomputes elevation-weighted RMS)
sp3 = gm.read_sp3("COD0MGXFIN_20240010000_01D_05M_ORB.SP3")
computed, failed = results.compute_elevations(sp3, obs.approx_position)

# SNR time series per satellite and S-code (for SNR-residual multipath analysis)
for series in obs.get_snr_series("G07"):
    print(series.code, len(series), series.values[:3])

What it computes

Per-code multipath combinations

For every pseudorange code observable P_k (C1C, C1X, C2W, C2X, C5X, ...), the ionosphere-free, geometry-free multipath combination is formed with two carrier phases:

MP_k = P_k − (1 + 2/(α−1))·Φ_i + (2/(α−1))·Φ_j        α = (f_i / f_j)²

• Φ_i — phase on the code's own band; Φ_j — phase on a partner band chosen from a deterministic per-system priority list
• GLONASS FDMA frequencies use the per-satellite channel numbers from the RINEX header
• The phase-ambiguity bias is removed per continuous arc: whole-arc mean for short arcs, centered moving average (default 1500 s) for long ones
• Arcs reset at data gaps, cycle slips, and MP jumps — slips never smear into the RMS

Signals are named in the convention used by GNSS monitoring literature: GPSM1C, GPSM2W, GLOM1P, GALM8X, BDSM2I.

Cycle-slip detection

Three detectors, thresholds that scale with the sampling interval (|Δ| > base + rate·dt):

At 30 s sampling this detects single-cycle L1 slips (GF signature ≈ 0.29 m vs. threshold 0.19 m) — rate-based thresholds tuned for 1 s data cannot. Slips are attributed and counted per signal code (GPSL2W, GLOL1C, ...).

SNR series export

RinexObsData.get_snr_series(satellite, code=None) returns per-S-code time series (unix-second timestamps, dB-Hz values) — the input for SNR-residual multipath analysis (polynomial detrending, wavelet spectra) without re-parsing the file.

SNR wavelets and Fresnel zones (1.0)

w = gm.analyze_snr_wavelets(obs)          # all satellites, in parallel
w["bands"]                                # [[60, 120], [120, 300], [300, 900], [900, 1800]] at 30 s
w["band_power"]["R02"]["power"]           # binned scale-averaged power, one series per band

r = w["residuals"]["R02"]                 # detrended linear-amplitude SNR
s = gm.scalogram(r["t"], r["v"], w["interval"])   # power, periods, significance, coi

a, b, dist = gm.fresnel_zone(15.0, 2.0)   # 2 m antenna, 15 deg, GPS L1 -> (4.896, 1.267, 7.464) m
zones = gm.fresnel_map([("G05", azimuths, elevations)], antenna_height=2.0)

• SNR is converted from dB-Hz to linear amplitude, split into arcs at gaps, and detrended per arc with the polynomial (order 2 to 9) that leaves the smallest residual RMS
• Morlet CWT follows the Torrence & Compo (1998) FFT formulation; significance is tested against a lag-1 red-noise background
• Period bands adapt to the interval because the article's 1 s bands are unreachable at 30 s sampling
• Fresnel footprints are sampled once per 10° × 10° azimuth/elevation cell above the elevation cutoff

Results on real data

24 h of BOR1 (EPN, Poland) 30 s multi-GNSS RINEX 3, default settings:

The ranking (Galileo AltBOC best, GLONASS C/A worst) reproduces published station-monitoring results. Parsing the file takes ~0.4 s; the full 15-signal analysis with slip detection ~1.4 s (306 k estimates).

Analyzer options

gm.MultipathAnalyzer(
    obs,
    elevation_cutoff=10.0,        # degrees; applied when elevations are known
    systems=["G", "R", "E", "C"], # G R E C J S I
    bias_window_seconds=1500.0,   # moving-average window; None = whole-arc mean
    min_arc_seconds=300.0,        # drop shorter arcs
    arc_gap_factor=5.0,           # arc break at gap > factor × interval
    include_codes=["C1C", "GC5X"],# restrict codes ("C1C" any system, "GC5X" GPS only)
    exclude_codes=[],
    max_epochs=None,              # uniform decimation guard for huge files
    detect_cycle_slips=True,
    ion_delta_base=0.10, ion_delta_rate=0.003,   # GF slip threshold (m, m/s)
    cp_delta_base=5.0,   cp_delta_rate=0.10,     # code-phase slip threshold
)

API surface

Supported input: RINEX v2 / v3 / v4 observation files, SP3-c/d orbits, broadcast ephemerides (Keplerian + GLONASS RK4).

Pipeline

flowchart LR
    A[RINEX v2/v3/v4] --> B[Parse<br/>code + phase + SNR]
    B --> C[Per-code combos<br/>deterministic phase pairing]
    B --> D[Cycle-slip detection<br/>GF · code-phase · LLI]
    D --> E[Arc splitting]
    C --> E
    E --> F[Ambiguity debiasing<br/>mean / moving average]
    F --> G[Per-signal statistics<br/>RMS · weighted RMS]
    H[SP3 orbits] -.-> I[Elevations / azimuths]
    I -.-> G
    B --> J[SNR detrend<br/>per arc]
    J --> K[Morlet CWT<br/>band power · significance]
    I -.-> L[First Fresnel zones]

CLI

geoveil-mp analyze --obs station.rnx --sp3 orbits.sp3 --elevation 10 --output results/
geoveil-mp info --obs station.rnx

References

• Hunegnaw, A.; Teferle, F.N. Evaluation of the Multipath Environment Using Electromagnetic-Absorbing Materials at Continuous GNSS Stations. Sensors 2022, 22, 3384.
• Estey, L.H.; Meertens, C.M. TEQC: The Multi-Purpose Toolkit for GPS/GLONASS Data. GPS Solutions 1999, 3, 42–49.
• Václavovic, P.; Douša, J. G-Nut/Anubis: Open-Source Tool for Multi-GNSS Data Monitoring. IAG Symposia 2016, 143.

License

PolyForm Noncommercial 1.0.0 with an attribution term — see LICENSE.

Free for research, education, personal and other non-commercial use, provided you credit the author and cite the library (see Citation below). Commercial use requires a separate license — contact miluta.flueras@cartografie.ro.

Author

Miluta Dulea-Flueras — miluta.flueras@cartografie.ro

Citation

@software{geoveil_mp,
  author  = {Dulea-Flueras, Miluta},
  title   = {geoveil-mp: GNSS Code Multipath Analysis Library},
  year    = {2026},
  version = {1.0.1},
  url     = {https://github.com/miluta7/geoveil-mp},
  license = {PolyForm-Noncommercial-1.0.0}
}