Timing Recovery For Blind Equalization

Timing recovery for blind equalization requires a non-equalizer-based, non-decision-directed approach, due to slow convergence of the adaptive equalizer taps and the resulting initial unrellability of the data decisions. The leading possibilities for timing recovery are either band-edge, or conventional envelope-derived using the common nonlinearities of magnitude, square, or fourth-power. A novel implementation is proposed for band-edge timing recovery. With this method, the vector representing the band spectral line is derived by digital correlation techniques, and then this vector is filtered in such a way that the timing recovery loop can attain fast convergence while tracking worst case frequency offset between the transmitter and receiver master oscillators. It is shown experimentally that for signals chosen from the CCITT V.33 trellis-coded 128-QAM constellation having 12.5% excess bandwidth and square-root Nyquist shaping at the transmitter, band-edge timing recovery is preferred over conventional envelope-derived, as it is the only one which can recover the baud spectral line in a reasonably jitter-free manner over both moderately and severely impaired channels.