Starlink receiver prototyping for opportunistic positioning

Traditional GNSS systems for positioning (PNT), such as GPS and Galileo, use Medium Earth Orbits (MEO). Recently, the possibility to use Low Earth Orbit (LEO) orbits for PNT has been investigated, which offer several advantages over the traditional MEO, e.g., higher power and wider band. Among the available signals of opportunity (SOOPs), this thesis project investigates the feasibility of utilizing Starlink signals, primarily designed for global Internet coverage, for positioning purposes. The Starlink signal structure is not publicly available, but the literature suggests the presence of nine equidistant spectral peaks within a band of approximately 1 MHz in the signal spectrum of each satellite, centered at frequency 11,325 GHz. The method proposed in this thesis involves the acquisition and tracking of these peaks, on a signal sampled at a lower frequency than the estimated bandwidth for the entire Starlink channel of 240 MHz, in order to reduce receiver complexity. For the acquisition phase, once the IQ components have been extracted from the signal, the optimal acquisition window length is selected as the trade-off between noise and Fast Fourier Transform (FFT) computational performance. The peak detection threshold is chosen based on the Gaussian distribution of noise and a predefined false alarm probability. This enables the selection of peaks above the noise floor in each acquisition instance, facilitating the detection of potential satellites. Then, similar to standard GNSS receivers, a tracking loop (a third-order PLL assisted by a second-order FLL) is implemented to estimate the Doppler frequency shift of the peaks over the entire captured window. However, as opposed to standard GNSS signals, Starlink does not use a PRN code to identify the individual satellites. To resolve the ambiguity in satellite identification, a method is proposed to compare the Doppler frequency shifts estimated from peak tracking with the Doppler frequency shifts predicted by a visibility prediction tool, which provides the ability to associate each identified peak with a specific Starlink satellite. The tool uses Two-Line Element Sets (TLEs) data and a simplified perturbation model (SGP4) to propagate the satellite orbits. The method is applied to a signal captured using a basic configuration with a Ku-band Low Noise Block (LNB) converter, and the data acquired consist of raw In-phase and Quadrature-phase (IQ) samples with a bandwidth of 4,096 MHz around 11,325 GHz. The results show that the method allows to acquire several satellites in the captured signal, and to track the corresponding peaks for positioning purposes.