Satellite-to-Ground Channel Characterization for LEO Constellations

LEO-SG channel characterization models RF propagation dynamics b/w LEO sats (~300–2000 km alt) & terrestrial GSs. Key drivers: rapid Doppler shift (±10–50 kHz @ S/C-band, ↑ w/ freq & ↓ orb height), TEC-induced ionospheric delay (~1–10 m range error, freq⁻² dep), Faraday rotation (impacts poln purity, ∝ TEC & B-field), tropo scintillation (Ku/Ka-band fade, ↑ @ low elev, wet env), and shadowing/blockage (urban/canyon effects). Time-scales: coherence time ~10–100 ms (dictates CP length), Doppler spread 1–10 kHz. Path loss: FSPL ≈ 150–180 dB (↓ w/ elev ↑), modeled via ITU-R P.530/618. Atmospheric atten: ~0.2 dB (clear-sky S-band), ↑ to 5–20 dB @ Ka-band rain events (P.618). Link margin design must account for pointing loss (sat antenna beamwidth ~0.1–2°), tracking error (≈1–3 dB), and freq agility for iono mitigation. Channel models: 3GPP NTN TR 38.811 defines geometry-based stochastic model (GBSM) w/ LEO-specific parameters: satellite speed ~7 km/s, vis time ~5–15 min/sat, handover rate ~1–3/min. Model includes LoS/NLoS states, elevation-dep attenuation, and time-varying shadowing (EGC/SCM variants). Empirical data from Starlink, OneWeb, Iridium NEXT confirms Rician K-factor ~10–20 dB (LoS-dominant), fading dist ≈ Rician → Rayleigh in urban. Rain fade diversity: spatial sep of GSs >10 km reduces outage (cross-correl <0.5). Polarization: dual-LHCP/RHCP common; XPD degraded by multipath/iono (↓ 3–10 dB). MIMO feasibility limited by sat size; SIMO (massive GS arr) preferred for beamforming & diversity. Beam-hopping & intelligent reflecting surfaces (IRS) under study for coverage opt. Real-time channel est (CE) challenged by high Doppler; pilots must be dense (e.g., DMRS every 1–2 OFDM symb). CSI feedback latency critical—LMS-based prediction used. Hybrid ARQ less effective due to RTT ~10–50 ms (vs. GEO ~500 ms). Standardization: 3GPP R17/18 NTN integrates NR over LEO, enabling direct-to-device (D2D) links. Co-primary spectrum sharing w/ terrestrial nets requires geo-sep & power control to avoid interference. Machine learning (ML) applied for CE (LSTM/Transformer nets), rain fade forecasting (RF/SVR), and handover opt (RL agents). Key pitfalls: overestimating LoS availability in urban; underestimating Doppler-induced ICI; ignoring cross-layer imp (PHY-MAC sync). Testbeds: ESA’s SAGA, NASA’s SCaN, and commercial DTs (e.g., SpaceX RF farms). Future: AI-driven dynamic resource alloc, Q/V-band exploration (↑ BW, ↑ atten), and LEO-IoT convergence. Open challenges: accurate NLoS modeling, standardization of adaptive coding & mod (ACM) schemes, and quantum key distribution (QKD) over LEO-SG links.