HSO recommended hardware setup

[RTirapegui]

Hi,

While reading the "HSO Startup Guide" included in the release notes of Motor Workbench 6.3.2, I came across a recommendation regarding the filtering of voltage:

This raised the following question: considering that the only development kit offered by ST to test the HSO algorithm consists of the B-G473E-ZEST1S + STEVAL-LVLP01 + B-MOTOR-PMSMA1, where the B-MOTOR-PMSMA1 is based on the R57BLB50L2 motor with 2 pole pairs and supports a maximum speed of 5000 RPM (i.e., 167 Hz of electrical frequency), the suggested phase voltage filter cutoff frequency of 500 Hz seems appropriate.

However, if I were to test the algorithm using the B-G473E-ZEST1S in combination with a custom-designed power board to control a PMSM with 2 pole pairs capable of reaching 90 kRPM (i.e., 3 kHz of electrical frequency), should I maintain the 500 Hz filtering for the phase voltages, or would it need to be adjusted accordingly?

In other words, does the HSO algorithm use phase voltages to calculate the angle independently of the motor speed, or does it rely on phase currents after a certain frequency range?

Additionally, regarding the phase currents, could I use more restrictive filters? For instance, could a cutoff frequency of 16 kHz be used instead of the recommended 300 kHz?

Thanks!

[Gael A]

Hello,

Concerning the currents sensing, the 300 kHz filtering is there to get rid of really high dynamics spikes happening while switching and that does not depend on your motor's behaviour, so keeping that out of your sampling is its main objective. I guess you could use a lower filter frequency, but 16 kHz is too low in my opinion and will not improve the behaviour. Avoid getting under 100 kHz.

Concerning the phase voltages sensing, you would like a high filter cross-over frequency to allow a good emf estimation even during high dynamic changes in torque / speed. However, you would also like to avoid aliasing as much as possible, which can be achieved with a lower cross-over frequency. We generally want a cross-over frequency staying roughly below 10% of your sampling frequency, which happens 4 times per PWM period thanks to oversampling. All in all : Frc < 0.1 * 4 * Fpwm