Other Parts Discussed in Thread: DRV8312
Hi, I am using a DRV8312 Kit and am in the project PM_Sensorless.
Right now i am tuning the PID controllers for Id and Iq currents. Using the guide in the pid_grando.pdf within the control suite folder I provide a step response of Iqref= 0.15 and monitor the response Iq_feedback[k] (pid1_iq.term.Fbk) with the DATALOG module.
I currently cant seem to satisfy my rise time and overshoot requirements with just a PI controller, have tried for a few days now. I tried to even model the Plant P[k] of the system along with my controller parameters C[k] in matlab to get the correct coefficients. Does not seem to be working.
Thus I rather experimentally tune the current controllers and introduce a Derivative term. I see your linear approximation for a digital lowpass filter coefficients c1 = a/(1+a*T) and c2 = 1/(1+a*T) where a = cutoff frequency in radians/s and T = ISR period.
You describe the derivative term of the grando module as follows: ud[k] = Kd(c2*ud[k-1] + c1*e[k] - c1*e[k-1]) where e[k] = Km*r[k-1] - y[k-1]
I follow this difference equation and am happy with the reference weighting options. However, I am a bit confused as to what value for the cutoff frequency i should be using. I am using very small Kd terms, and set my cutoff frequency a(rad/s) to around 5 Hz thus a = 2*pi*(5) = 32.
I have seen some instability with values of Kd around 0.10 Q24 when paired with this 5Hz cutoff frequency. What kind of noise should i be expecting in the derivative portion of the loop?
