Hardware Realization and Resource Analysis of Continuous-Time Sliding Mode State Estimators on FPGA

Rotor-flux estimation on FPGA requires attention to both estimation accuracy and hardware resources. This study implements sliding-mode mean-square (SMMS) and sliding-mode mean-module (SMMM) filters for an induction motor and compares them with a Kalman–Bucy (KB) implementation. The continuous-time filter equations are realized as fixed-point updates using recorded electrical signals and measured rotor speed. Software Monte Carlo analysis is followed by separate FPGA experiments in which a stochastic motor reference model and one estimator operate concurrently. Software evaluation uses model-generated stator currents, whereas the FPGA experiments feed the estimators with replayed measured currents from the same recorded operating interval. The comparison shows that the relative accuracy of the estimators varies with the noise level: the sliding mode filters achieve lower errors across several tested conditions while requiring fewer logic resources than KB. FPGA outputs were recorded under the tested laboratory conditions, although post-fit timing analysis did not establish closure at the specified base-clock constraint.