> > No not interested in any of the above more than possible what you write > > about integration. > > > > That's fine, it takes a lot of study to move past experimentation with > PID. And the literature is full of dead ends. The question I have about > putting the integrator at the model error is how that would drive a steady > state positioning error to zero. Seems like your measurement and model > could be forced to agreement without driving the positioning error to > zero. Of course, I don't understand your system at all, so I could be > wrong. In general, putting integrators into a system adds instability > because it's a marginally stable dynamic element. Even if it does just > remain an integrator in the closed loop system, there is always the issue > of windup. > Eric
The model go to zero without an integrator. Integrating error between what signal is supposed to be with output will make error between model and reality goto zero. I am not totally sure this already happen in some documents I read. For a "dead beat" controller it will not make a difference but then gain is lowered as is probably always the case it will make a difference. I have not yet been able to put together the theory and are not totally sure it have already been written down or if idea is bad. _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
