On Mon, 2009-11-30 at 11:58 -0600, Jon Elson wrote: > Pat Lyons wrote: > > Yea, I'd say it falls into the <15Hz category. > > > > I was thinking about finding a way to dampen the Table's mechanical response > > by using the gibbs (I'm not sure thats the right name for em, but im > > reffering to the screws that are used to apply pressure to the dovetails on > > the ways) > > > (Technically, the gibs are the bars that ride against the inside of the > dovetail, not the > screws that apply the pressure to them.) > Well, adding lots of friction is not a good approach. OK, with analog > velocity > servos, you have a bunch of control loops within the servo amp, and then > another > control loop outside it (EMC). So, the interactions get complicated. > If the torque loop > or the velocity loop in the servo amp are unstable, then EMC certainly > can't fix it. > > > Funny you mention FF2 as 0.0002, because i think the highest I have mine > > (Xaxis) is like 0.0005. My FF values in all instances are below 1. I > > managed to get it to settle a little with the error, but the biggest trouble > > seems to be the initial error, which spikes- I would slowly increase D and > > would end up with much resonance as a result... I was thinking, this could > > also be due to an improper backlash measurement I've made (i suspect this > > because the worst spikes occur when changing direction... > > > I have never liked the way D works on the EMC PID. It seems like more D > should just > make the response sluggish, but it causes oscillation instead. I think > this is because > D only looks one sample interval behind. But, certainly too much D does > cause instability. > With a velocity servo amp doing all the work (meaning loop gain) you > don't need a lot > of P gain. > > If you have backlash compensation turned on, definitely try again with > it turned off. > I don't use backlash comp on my systems, as it just makes movement messy. > > Jon
Jon, Any progress on filtering D? Some commercial PID's apparently allow one to adjust rolloff on D. I've never found D or I as useful as I hope they are. If I can't fix it with P, FF1 and FF2 then I probably won't be able to fix it. I suspect the accel and decel spikes cannot really be fixed on systems with backlash. I've never had a system tight enough to prove that. ;-) Dave > > ------------------------------------------------------------------------------ > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day > trial. Simplify your report design, integration and deployment - and focus on > what you do best, core application coding. Discover what's new with > Crystal Reports now. http://p.sf.net/sfu/bobj-july > _______________________________________________ > Emc-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-users ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
