On Tue, 8 May 2012 21:22:29 +0100 andy pugh <bodge...@gmail.com> wrote:
> On 8 May 2012 21:06, Dave <e...@dc9.tzo.com> wrote: > > > You have no Pgain, but Igain and Dgain. That's a problem. An all > > I and D loop is almost always unstable. > > That depends…. > > For a spindle P gain is not much help, as you need a non-zero PID > output even when at the set speed (and when the error is zero). That > is why the Wiki example is all-I. I assume that FF0 wasn't available > when that page was written. > > I spent most of this afternoon at work tuning something that turned > out to work best with 99% of the control coming from the D-term and 1% > from I. But that was a rather unusual situation. > Machine: Cinci Contourmaster: Servo Dynamics 1525BR amps, SEM 30 series servo motors, 2500 ppr automation direct/Koyo encoders direct coupled to the ballscrew. Mesa 5120/7i33/7i37. I used to think I knew something about servo tuning but I've since encountered a couple of nasty cases. First of all I assume you have a servo motor with a tach for velocity feedback. If you are using torque then all bets are off. Step 1: set amp gain to achieve max velocity plus a bit of headroom. Step 2: adjust tach feedback so you get control, stiff spindle when the input to the amp is zero. Step 3: Set P, I, D, FF0, FF1, FF2 to zero. Step 4: Increase P until you get you just get sustained oscillation. Step 5: Multiply that value by .6 and use that as a starting P. Step 6: Set FF1 to 1.0 and test (halscope) Step 7: adjust value of FF1 to balance following error around zero. Step 8: add small values of FF2 to clean up accel/decel. With a tight axis this works well; if you have significant backlash you usually don't gain much. Step 9: add D and I only if they yield better following error. I've seen P values as low as 15 and as high as 600 & FF1 values from 18 (ugh) down to -0.01. In torque mode all bets are off. Just look at the values in the configs for the Mazak. (torque mode). ;-) Those are probably extremes but.... Best of luck. A utility to get a bode plot and then some guidance on how to use it would take a lot of the guess work out of this. I keep waiting. Dave ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
