An Improved Low-Noise Square-Wave Voltage Injection Method for IPMSM Sensorless Control

Injecting high-frequency (HF) square-wave voltage signals into the estimated d-axis serves as an effective strategy for achieving zero and low-speed position sensorless control in interior permanent magnet synchronous motor (IPMSM) drives. However, the application of this method in diverse fields is circumscribed. The root cause lies in the fact that the injection of high - frequency voltage engenders noise that is not only perceptible but also causes discomfort, thereby impeding the extensive adoption of this method in practical applications. To reduce unnecessary audible noise, this paper proposes an improved low-noise random voltage injection method adapting Markov chain and frequency adaptive complex coefficient filter (FACCF). For the estimated d - axis, two square - wave voltage signals with randomly varying frequencies and amplitudes are injected. Subsequently, the position information is extracted using a corresponding decoupled signal. The power spectrum density (PSD) of the induced HF currents is contrasted and anatomized with respect to the fixed - frequency injection approach and the optimized methodology. Comparing with fixed frequency injection method, the simulation and experiment demonstrates that the Markov chain based random signal switching can further reduce the audible noise, and FACCF based position information extraction method can solve the phase delay problem caused by using low-pass filter (LPF).