PMSM sensorless angle optimisation

Hello,

I am using the Motor control Blockset for simulating and programming a motor control board.

I have identified my motor's parameters with a LCR meter and double checked with a classic first order identification. (delta winding but Y assumption)

I use the block Flux Observer (https://fr.mathworks.com/help/mcb/ref/fluxobserver.html) for angle and speed estimation.

As a robustness test, I have changed the Ld and Rs parameters on the fly and find out I could reduce the Iq value for the same load by increasing the Ld parameter (up to x2). It works only during test and not simulation.

Does it means my angle estimation is necessary better with a x2 inductor value? What could cause this ?

Best regards,

Jérémy

Answer 1

[Satyam]

Hi Jeremy,

Increasing Ld inside the Flux Observer can appear to improve torque‑per‑ampere (lower Iq​ for the same load) because it changes the internal decoupling and back‑EMF, shifting the estimated electrical angle closer to the true rotor position when your nominal parameters are biased. That effect shows up on hardware (with inverter non‑idealities, sensor offsets, temperature drift, magnetic saturation, delays, etc.) but not in an ideal simulation, so it’s more a parameter‑mismatch compensation than the true motor Ld​ being 2×. The right fix is to calibrate and/or adapt parameters and tune the observer/PLL for the real plant.