A group of dynamic phasor measurement algorithms, including dynamic synchrophasor estimation algorithm (DPEA), dynamic phasor estimator based on frequency-domain model (DPEFM) and time-frequency information-based dynamic phasor estimator (TFDPE), are revealed to be linear time-invariant (LTI) systems, and the expressions of their equivalent filters are derived. In addition, these algorithms are extended to more generalized forms in the sense of filter equivalence. A group of filter specifications are defined based on the frequency response of the equivalent filter, and the quantitative relations are built between these filter specifications and the measurement errors caused by non-nominal fundamental tone, harmonics, decaying DC, and wideband noise. In terms of these relations, on the one hand, the measurement performance of these algorithms are conveniently evaluated and compared based on their frequency response; on the other hand, filter optimization is formulated to design optimum filters so that the algorithms achieve the best overall measurement performance. Optimization results of P-class algorithm are presented in the cases of 2-cycle and 3-cycle data length. Filter specifications and simulation results verify that, the overall measurement performance of the optimized algorithms are significantly improved as compared to their original versions, where excellent performance of rejecting decaying DC disturbance is achieved in particular.