“…In contrast to offline calibration algorithms, online filtering methods usually need less memory space and have better harmonic elimination performance against the changeable and complex electromagnetic environment in practice. Online filtering techniques mainly include Kalman filter (KF) [10], Fourier analysis [11,12], artificial neural networks (ANN) [13], probabilistic learning [14], type-III angle tracking observer (ATO) [15,16], and various phase-locked loop-based (PLL) methods, such as decoupled double synchronous reference frame (DSRF)-PLL [17,18], adaptive notch filter (ANF)-PLL [19], two-phase-type PLL [20], non-linear and linearized PLL [21], and other types [22,23]. For example, an improved PLL was presented in [24] and the maximum position error was nearly 15 • when the machine speed changes.…”