Adaptation and optimization of C libraries for 5G NTN orbit propagation on embedded satellite payloads

Accurate satellite position and velocity prediction is essential for 5G NTN to maintain Doppler frequency synchronization within the 3GPP TS 38.101-5 limit of 200 Hz at 2 GHz. For operators relying on publicly available TLE sets to initialize orbital propagators, a key dilemma emerges: analytical propagators like SGP4 are structurally designed for TLE propagation, while numerical methods offer superior force models but are not built to use mean orbital elements. This thesis compares SGP4, Cowell's method, Encke's method, and MATLAB Cowell orbital propagators using the Sentinel-6A satellite and precise reference ephemerides. Three experiments assess: 24-hour propagation accuracy and Doppler shift error after this mid-term propagation; the statistical 30-minute pass performance of each propagator across orbital phases; and the operational trade-off decision of TLE refresh timing. Results show all propagators satisfy 3GPP compliance in short-term passes, yet SGP4 is superior: it maintains position errors below 1 km over 24 hours while requiring two orders of magnitude less computation time. After 23 hours of propagation, SGP4 remains the only candidate with robust compliance margin, whereas numerical integrators can exceed the 200 Hz limit due to accumulated transversal position errors. TLE initial condition uncertainty dominates the error budget: numerical propagators show their advantage only with high-precision states, which are operationally not feasible. Critically, SGP4 can safely propagate aging TLEs with state discontinuities below 100 m, while numerical methods diverge by kilometers within days. The thesis provides quantified TLE-driven propagation limits, operational guidance for propagator selection, and demonstrates that the structural compatibility with the input data source is more important than theoretical sophistication.