Hi,

I am trying to tune our system using frequency domain analysis, i.e. bode plots of the system response.

I am trying to push the limits of the bandwidth while minimising the steady state error. I can do the first without any problems when Ki is zero.

However, as the motor position is an integrating process, specifying any non-zero value for KI causes the system to go unstable due to the introduction of an additional phase lag of 90 degrees, and a gain > 1 at a phase of -180 somewhere in the low frequency range.

My only thought at this stage is to put a zero at the -180 degree phase crossover frequency, but do not have a clear understanding of the structure of the notch transfer function from the manual.

Can you please explicitly tell me the notch transfer function?

From NF corresponding to the frequency in Hz, and NB/NZ corresponding to the real parts of the pole and zero respectively, my best guess is:

N(z) = [z + NZ*(1+i*atan(NF*2*pi*TM/10^6)][z + NZ*(1-i*atan(NF*2*pi*TM/10^6)]/[z + NB*(1+i*atan(NF*2*pi*TM/10^6)][z + NB*(1-i*atan(NF*2*pi*TM/10^6)]

Also, as a suggestion, could a phase compensator be added to the control loop, or perhaps even the control be based on velocity rather than position, both which would cause, at least at low frequencies, the system output to be in phase with the system input, thus avoiding the aforementioned problem?

Steve