Permanent magnet synchronous motors (PMSMs) are widely used in industrial automation due to their high power density and efficiency. However, achieving precise speed regulation in practical applications is challenging because of inherent system nonlinearities, external time-varying load disturbances, and high-frequency measurement noise from position sensors. Traditional extended state observers face a fundamental limitation: increasing the observer bandwidth to accurately estimate fast disturbances inevitably amplifies high-frequency measurement noise. To overcome this, a novel composite control strategy that resolves the conflict between noise suppression and disturbance-observation bandwidth is proposed.

The proposed method combines a generalized nonlinear extended state observer (GNESO) and continuous terminal sliding mode control (CTSMC). System uncertainties and external load variations are treated as a lumped disturbance. The GNESO features an integral output mechanism that inherently filters high-frequency measurement noise, eliminating the need for cascaded low-pass filters and avoiding the associated phase lag. It also incorporates a multi-layer differential disturbance-observation structure and a nonlinear fixed-time convergence function to accelerate initial error convergence and improve estimation accuracy. Based on these precise estimates, a CTSMC speed controller is designed to ensure finite-time convergence of tracking errors while significantly mitigating the chattering commonly associated with traditional sliding mode control (SMC). To validate the proposed strategy, a real-time hardware-in-the-loop experimental platform using a TMS320F28335 architecture with two coaxially connected PMSMs is constructed.

Comprehensive experiments are conducted to compare the proposed method against Continuous Sliding Mode Control (CSMC) and Continuous Terminal Sliding Mode Control (CTSMC)+traditional extended state observer (TESO). Under no-load startup at 1000 r·min^–1, the proposed method achieves a minimal speed fluctuation rate of 0.286%, significantly outperforming CTSMC+TESO (0.326%) and CSMC (0.574%), demonstrating exceptional noise filtering. During sudden load disturbance tests (0.25 N·m at 600 r·min^–1), the proposed strategy exhibits superior robustness, limiting the maximum speed drop to only 18 r·min^–1 and restoring steady state in just 0.07 s. The Integral of Time-weighted Absolute Error and Integral of Time-weighted Squared Error indices are reduced by 49% and 44%, respectively, compared to the TESO-based method. In addition, under continuous time-varying load disturbances, the control framework maintains superior dynamic tracking with a maximum speed deviation of only 4.1 r·min^–1.

The proposed control strategy successfully balances measurement noise suppression with rapid disturbance rejection for PMSM systems. The GNESO effectively overcomes the bandwidth limitations of traditional observers, and the CTSMC ensures higher steady-state precision and remarkably lower chattering. Rigorous experimental results confirm that this robust framework offers exceptional dynamic performance and show strong potential for application in high-precision motor servo drives.