On Tue, 25 Jan 2011, jros wrote: > Date: Tue, 25 Jan 2011 14:44:05 +0100 > From: jros <j...@unavarra.es> > To: joachim.fra...@pibf.de, > "Enhanced Machine Controller (EMC)" <emc-users@lists.sourceforge.net> > Subject: Re: [Emc-users] DIY BLDC Motors - FCM8201,2 and IRAMS > > >> I get some used PMSM from auctions. Replacing the >> bearings gives them a second life. Keep the angle position >> of the resolver on the rotor! So you can test it with >> an old (with analog velocity input) 3 phase power supply. >> Usually this motors run very smooth if the stator pack >> is twisted giving nearly no torque ripple (assuming >> rotating magnets). >> >> Motors with torque ripple maybe used as a drive >> if the control can do a predistortion to the >> current to compensate for the ripple. But I >> have not seen this until now with my eyes. > > I can see you've be seriously dealing with the internals of PMSM, and I > would like to discuss a little bit about the subject,as I'm facing some > problems related to torque rippling issue. > > I agree with your comments, but it looks like complicated, as some sort > of > identification/calibration should be needed. Plus a resolver, I mean you > need to track the absolute position of the shaft during a single > revolution (may be, there are suitable encoders for that work). > > Plus, I think that the logic should be implemented directly in the > electronics driving directly the motor, as the compensation strategy > should be fairly quick. Driving a variable intensity command from EMC > will certainly limit the strategy to low rpm.
I dont think so (unless you are considering high speed spindle motors). For servo motors its not really a problem. For example with a 4 pole servo motor (so 120 Hz for a reasonable 3600 RPM max speed) a 4 KHz servo loop will keep the maximum electrical angle error < 10.8 degrees at 3600 RPM which will keep torque ripple contributed by angular error issues less than 1.8%, and much better at slower speeds, decreasing to ~0.2 percent at 1200 RPM My guess is that magnetic irregularities and lack of good sine wave drive are more likely culprits for your cogging issues. > > I've seen some guys trying to do such compensation dealing with this > when using linear motors, that I presume are very similar to PMSM. > > I suppose that this problem important to them probably because the > aforementioned lack of twisting, but also because it is a Direct-Drive > application. That is, no reduction, and then low velocity. > > What do you think about this?: > > I've seen referred this torque ripple as "cogging", I can feel it > in my setup (www.imac.unavarra.es/~cnc) rotating the shaft by hand my > motors switched off. I'm pretty sure that this is the main responsable > of the vibrations I feel in the robot at low velocities. I'm using an > almost direct driving. That is strange, all the (larger) PMSMs we have used have very near 0 cogging when no power is applied. > > Do you think that this is an indication that my stator is not twisted?. > > So I'm thinking about using pulleys to introduce some sort of reduction, > but I'm not sure if this is going to lower the problem. I mean, for a > frictionless reduction strategy, I would get a increase in rotor > velocity, but also a lower level of torque. I see a benefit as the > rippling frequency will be increased. But wouldn't worsen the ripple to > torque ratio, and then introduce more problems?. > > I've been thinking about increasing the bandwidth of the PID (I'm > controlling intensity), hoping that the feedback can be able to > compensate the cogging, but may be then other sources of noise can > become more important. If your system has a high enough bandwidth, running the servo thread faster will help. I've noticed that with our test bed dynamomter and 3 KW PMSM, a 1 KHz servo loop rate is too slow for good performance, 2 or 4 KHz being noticebly better, allowing much larger P and D terms and coorespondingly lower following errors. > > Then may be an identification + feedforward compensation can be the best > solution. I think it can be implemented in EMC, up to some bandwidth. > Identification can be the hard part, but if a good homing strategy is > available, is looks like possible. > > I'm thinking about making such a implementation. Do you think that it > can improve things?. > > Also what is your position about the best motor/driver type for > direct-drive position control applications. > > > I hope this discussion helps some body else to enlighten thing. It > certainly helps me. > > Cheers, > > Javier Peter Wallace Mesa Electronics ------------------------------------------------------------------------------ Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)! Finally, a world-class log management solution at an even better price-free! Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY! http://p.sf.net/sfu/arcsight-sfd2d _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users