Efficient Multi-User Detection and Turbulent Compensation in Optical Links for Small Satellite Communications

Future non-terrestrial networks (NTNs) will stream gigabit-class data through hundreds of CubeSats whose tight size-weight-and-power (SWaP) budgets exclude multiple receive telescopes. Severe atmospheric scintillation and wave-front mixing make simultaneous multi-user reception on a single aperture difficult, reducing both channel capacity and link margin. We propose Data-Interpolated Blind-Source Separation (DI-BSS) as a lightweight digital-signal-processing solution that: (i) separates two co-channel 1 Gb s⁻¹ optical carriers on one compact telescope; (ii) restores several decibels of link margin by cancelling co-channel interference without pilot-tone overhead; and (iii) fits inside sub-watt FPGA power envelopes. While simulations demonstrate the algorithm’s efficacy for QPSK and higher-order QAM, we experimentally validate the system using two statistically independent 1 GBd BPSK streams combined through a 2×2 fading matrix emulating strong atmospheric turbulence σ²_I ≈ 0.99. We introduce a Sign Reversal injection strategy that preserves signal geometry, allowing the demixing matrix to be reestimated only once every N = 50–100 symbol blocks. Hardware-in-the-loop results confirm that this approach maintains a Signal-to-Interference Ratio (SIR) penalty of < 1 dB relative to full-rate Independent Component Analysis (ICA) while reducing the computational iteration count by a factor of 3.5 compared to standard methods. Validated on a bench-top coherent receiver, DI-BSS offers a practical path to high-capacity, low-SWaP optical links across the next generation of satellite constellations.