Link Adaptation Algorithms for Dual Polarization Mobile Satellite Systems

The use of dual polarization in mobile satellite systems is very promising as a means for increasing the transmission capacity. In this paper we study a system which uses simultaneously two orthogonal polarizations in order to communicate with the users. The application of MIMO signal processing techniques along with Adaptive Coding and Modulation in the forward link can provide remarkable throughput gains up to 100 % when compared with the single polarization system. The gateway is allowed to vary the MIMO and Modulation and Coding Schemes for each frame. The selection is done by means of a link adaptation algorithm which uses a tunable margin to achieve a prede ned target Frame Error Rate.

💡 Research Summary

The paper investigates the use of dual orthogonal polarizations in low‑frequency mobile satellite systems (L‑ and S‑band) as a means to create a 2×2 MIMO link that can significantly increase spectral efficiency. Traditionally, these bands employ a single circular polarization (RHCP or LHCP) to mitigate Faraday rotation, but the authors propose transmitting simultaneously on both polarizations, thereby enabling four transmission modes: SISO, Orthogonal Polarization‑Time Block Code (OPTBC), Polarization Modulation (PMoD), and V‑BLAST.

A system model is defined where the received signal vector yₙ = √P·Hₙ·xₙ + wₙ, with Hₙ ∈ ℂ^{2×2} representing the dual‑polarization channel, P the transmit power, and wₙ additive white Gaussian noise. Frames consist of 2560 QPSK symbols (80 ms duration, 33.6 kSym/s). To avoid full baseband processing, the authors adopt a physical‑layer abstraction based on Mutual‑Information Effective SNR. For each symbol the instantaneous SNR γₙ is computed according to the active MIMO mode, and the frame‑level effective SNR (SNR_eff) is obtained via the Φ⁻¹ averaging method.

The four MIMO modes have distinct SNR expressions: SISO uses the magnitude of a single channel coefficient; OPTBC applies Alamouti coding across the two polarizations, yielding a gain of 2 in the effective SNR term; PMoD combines the power of both polarizations and adds one bit of polarization selection; V‑BLAST employs two independent streams with an MMSE receiver, leading to a more complex per‑stream SNR formula. None of the modes require channel state information at the transmitter, which is advantageous given the long round‑trip times in GEO satellite links.

Link adaptation is split into two tasks. The mobile terminal (MT) evaluates the effective SNR for all four modes and selects the mode that maximizes instantaneous throughput. This decision is fed back together with the ACK/NACK flag and the effective SNR of the chosen mode, thereby limiting feedback overhead. The gateway (GW) then determines the Modulation and Coding Scheme (MCS) using a lookup table (LUT) that maps SNR intervals to MCS indices. Crucially, an adaptive margin c is added to the reported SNR before the LUT lookup. The margin is updated recursively:

c_{i+1} = c_i – μθ² + SNR_i² – 2d (Γ_{i–d} – p₀)θ

where μ = 1, θ = 10, d is the feedback delay measured in frames (7 frames ≈ 560 ms for GEO), Γ is the ACK/NACK indicator, and p₀ is the target Frame Error Rate (FER). This formulation stems from minimizing the squared deviation between the observed FER and the target FER, ensuring that the system converges to the desired reliability while adapting to channel variations.

The channel model follows the three‑component Rice formulation used in SatNEx‑IV: a line‑of‑sight (LoS) component, a specular reflected component, and a Rayleigh‑distributed diffuse component. The matrices β and ξ describe cross‑polarization coupling, while D contains complex Gaussian entries with a prescribed covariance. Time correlation and Doppler spread are generated using Clarke’s model, with the coherence time τ_c = 3λ/(4v√π) and Doppler spread D_s = v/λ.

Simulation settings emulate a maritime scenario with a vessel moving at 50 km/h, carrier frequency 1.6 GHz, QPSK modulation, and the BGAN F80T1Q‑1B bearer. For each average SNR (from –5 dB to 25 dB in 2.5 dB steps) 60 000 frames are transmitted. Performance metrics include average spectral efficiency (bits/s/Hz) and cumulative FER.

Results show that dual‑polarization MIMO provides substantial gains over the single‑polarization baseline. At low SNR (–5 dB) OPTBC enables communication where SISO would fail, delivering 0.68 bps/Hz. In the mid‑SNR range (–2.5 dB to 7.5 dB) PMoD becomes optimal, achieving 1.02–2.32 bps/Hz (≈50–55 % improvement). Around 10 dB both PMoD and V‑BLAST are used, yielding an average efficiency of 2.47 bps/Hz (≈47 % gain). At high SNR (12.5 dB–25 dB) V‑BLAST dominates, doubling capacity with efficiencies up to 3.48 bps/Hz (60–100 % gain).

Comparing a fixed margin (–1 dB) with the adaptive‑margin scheme for target FERs of 10 % and 1 % shows that spectral efficiency curves are similar, but the adaptive margin tightly controls the actual FER to the desired level. Lower target FER (1 %) yields a modest increase in efficiency, confirming prior observations that an optimal FER exists for maximizing throughput.

In summary, the paper demonstrates that employing dual orthogonal polarizations together with a two‑stage link adaptation (mode selection at the terminal, MCS selection with adaptive margin at the gateway) can double the achievable throughput of mobile satellite links while maintaining a prescribed reliability, even under realistic GEO feedback delays. The authors suggest future work on multi‑user extensions, higher frequency bands (Ku/Ka), and experimental validation.