Motor suddenly turns in the wrong direction

[RReic.2]

[PN]: Customer Board (Very similiar IHM08M1 and STM32G431RB)
[VERSION]: 6.3.1
[TOOL]: Motor Pilot
[DETAILS]: at heavy load step the motor turns suddenly in wrong direction with full speed (see plot below)
[EXPECTED BEHAVIOR]: stable speed or motor stops with speed feedback error or over current error
[HOW TO REPRODUCE]: Use PMSM Motor with FOC (STO+PLL as primary Sensor, 3shunt),  STM32CubeMX 6.12.1, Target Toolchain ST STM32CubeIDE, Firmware Package Version STM32 FW V1.6.0, Drive Type: LL

[GMA]

Hello @RReic.2,

Could you plot the IQ_MEAS and IQ_REF during the error use case?
You can decrease the STO-PLL variance threshold to make the system more sensitive to incorrect speed estimation.

[RReic.2]

Hello  @GMA

Thank you for supporting me.

Attached you will find two plots showing IQ_MEAS and IQ_REF at the moment when the motor starts turning in the wrong direction.

I also tried reducing the variance threshold from 25% to 10%, but unfortunately, this did not resolve the issue.

However, I discovered that lowering the PLL Kp in the Observer from 250 to 120 makes the problem less severe or almost eliminates it. That said, this adjustment results in poor control at low speeds.

[GMA]

Hello @RReic.2,

Thank you for your feedback. In the meantime, have you evaluated the "Observer + Cordic (Sensorless)" Speed Sensor Mode?

[RReic.2]

Hello @GMA 

No, we currently use PLL. What are the advantages of Cordic? When is Cordic recommended?

[GMA]

Hello @RReic.2,

It is to determine whether the Cordic, which estimates the angle differently from the PLL, exhibits such behavior. If so, whether it can generate an error when reverse rotation is initiated.

[RReic.2]

Hi @GMA 

It seems to be better with Cordic, it hasn't happened again.
Any tips on how to optimize C1 and C2?

[GMA]

Hello @RReic.2,

If you are referring to G1 and G2 observer gains, review the 2. Back-EMF state observer (STO) chapter in the Speed and position feedback sensorless algorithms section of the user manual. This documentation is available through the Workbench tool by navigating to About > Documentations > Documentation.

Gain1 is related to the current estimator, and Gain2 is related to the back-EMF/speed estimator.

[RReic.2]

Hallo @GMA 

cordic does not seem to be a solution either, as it exhibits the same behavior.

At the moment, I see the only option to check the speed manually with MC_GetMecSpeedReferenceMotor1_F() and MC_GetAverageMecSpeedMotor1_F(). If there is a (hugh) deviation, the motor is stopped and restarted.
This is annoying for our customer because the product is about to go into series production and we are not getting any better support from ST.

[GMA]

Hello @RReic.2,

According to this post, you already faced the same issue. Did Cordic resolve the issue in this case, as you mentioned?