This paper investigates the performance of a flexible decision-aided maximum likelihood (DA-ML) phase estimator in coherent optical phase-shift-keying (PSK) system recently proposed by us. We take into account the time-varying phase noise, instead of the assumption of constant phase noise over the observation interval in conventional DA-ML phase estimator.
The weighted coefficients are introduced to recover carrier phase more accurately than conventional DA-ML scheme. The phase error variance of our proposed scheme is derived and the effect of different block length is also analysed. Simulation results show that there is a bit error rate (BER) floor at different block length and this can eliminate the block length effect of the conventional DA-ML scheme.