Spectrum analysis using a trellis algorithm

Carrier frequency offset is a significant problem in achieving signal acquisition in communications equipment. In a receiver the frequency offset must be corrected before symbol timing and phase recovery operations can be completed. Several methods have been applied to frequency estimation, for example, the fast Fourier transform (FFT) method and the differential phase method. However, the FFT method suffers from threshold effects and the differential phase method, although simple, suffers an SNR penalty though its differential detection and is sensitive to large frequency offsets. This paper describes a trellis based scheme for frequency and phase estimation. A recently reported technique, per-survivor processing (Raheli, 1995) has been adapted to make use of the trellis structure and is described. The technique, referred to herein as frequency estimation by PSP (FEPSP), gives near optimum performance under certain conditions. Firstly Cramer-Rao lower bounds (CRLB) are derived for the frequency estimator and the phase estimator. The PSP estimation algorithm is then described. Lastly, the performance of the FEPSP is discussed and compared with that of the CRLB and FFT methods.