Kirk Wallace wrote: > I now have an encoder mounted to my tool carousel. I linked the encoder > to a PID and found that each pocket is 720 counts apart. I can travel in > either way to get to a pocket, but there is no motion control, so the > start and stop is pretty abrupt. I am starting to consider implementing > some sort of motion planning to control the acceleration on each side of > the move, but now that I actually need to create the solution it becomes > more complicated. Finding the closest path to the target pocket should > be easy. The next step is to feed position commands to the PID to > control the trajectory. A couple of issues are that the carousel will > not be an axis and the motion component is not available during a tool > change, also position overshoot (but eventually corrected) is > acceptable. As is, the straight PID works okay, but if I increase the > maximum carousel speed (very likely), the start/stop jerk may become a > problem. Any pointers on how to attack this problem would be > appreciated. > You cold build a stupid trajectory planner as a hal component. It would take a couple parameters such as max speed and acceleration, and a pin for final position. It would accelerate to max speed and then compute the point at which it needs to start decelerating. I did one of these in an FPGA driving a stepper where we wanted rapid movement to the final point without the risk of stalling the stepper. It was all done with just adds and subtracts, although all constants were hard-coded.
I wouldn't think this would take more than a dozen lines of C code. You can still use the PID and whatever servo driver and drive would be used, just like it was an axis. Jon ------------------------------------------------------------------------------ All the data continuously generated in your IT infrastructure contains a definitive record of customers, application performance, security threats, fraudulent activity and more. Splunk takes this data and makes sense of it. Business sense. IT sense. Common sense. http://p.sf.net/sfu/splunk-d2d-oct _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
