On Dec 6, 2007, at 6:56 PM, Jon Elson wrote: > Kirk Wallace wrote: >> I got the first pass on my Bridgeport X axis working and I am having >> trouble with tuning. I am using a brushed DC motor with a 3:1 belt >> ratio >> between the motor and ballscrew. The motor is driven by a Pico PWM >> amp >> and controlled by a Pico UPC and EMC 2.1.7. Because a glass slide was >> already mounted, I used it for my encoder. My primary concern was >> to do >> a sanity check on the motor because it is a cheap treadmill motor and >> not a true servo motor. The motor seems to have no trouble driving >> the >> axis at my theoretical rapid of 390"/minute or 6.5"/second. My next >> consideration is the low speed performance, which seems much more >> dependent on the system tuning than just raw system power. Since I >> have >> a similar system on my lathe, I started with its tuning parameters. I >> need to post pictures and Halscope plots but some general issues >> came to >> mind. >> >> Glass slides do not seem to be recommended for CNC applications. >> Could >> someone remind me why? I think the resolution of glass scales adequate i.e. 10, 5 and 1 um. However, they require really tight systems to work well. That is, no backlash in the ball screws and almost perfect alignment with the table. Something like 0.005" over the length of the scale. If your system has backlash, going to extremes on resolution doesn't do much good; 5-10 times the accuracy you require. I'm always reminded of something Ray Henry once said and it goes like this, "machine manufacturers are more concerned about really good performance than costs; and they use high resolution encoders on the end of the ball screw". If your system has 0.002 backlash then it will tune MUCH better with an encoder on the ball screw, and of course a tach on the servo motor. Been there, done that.
Having said that it would be most interesting to have a system that used the information from both a linear scale and an encoder on the ball screw. I'll let the really bright guys dope out how to make that work. ;-) > Low resolution and possible backlash between motor and encoder. > I would think for positioning accuracy you would >> want as little as possible between the tool and the mechanism >> doing the >> position measuring. On the other hand for motor control, you would >> want >> as little as possible between the motor shaft position and the >> controller. Is there a way to cater to both? >> > Yes, have the backlash reduced to the absolute maximum. Make > sure the slider on the glass scale moves very freely (may need > internal cleaning and work on the shield that keeps junk out of > the encoder. Make sure the link between the encoder read head > and the machine is straight. A kink in the linkage can allow it > to bend under load, and so the encoder itself can develop > backlash. This is especially true of the ones that use a piece > of music wire to push/pull the head. >> I seem to have a consistent (non-self energizing) oscillation >> across a >> wide P range (30 to 280). I am wondering, is this is due to the .002" >> screw backlash, or this backlash being on the motor side of the >> encoder, >> the relatively low encoder (slide) resolution or all of the above? >> Which >> has the greater influence? >> > You're sunk! 0.002" backlash is HUGE! You need to have the > ballscrews repaired, or the end bearings on the screws, wherever > that backlash is, you can't expect proper operation with that > much backlash. The only way to supress this right now is to set > the deadband to .003" or so. >> The end position always seems to come to within an encoder count >> (0.0005") but the rate seems to vary widely during the traverse. > Yes, the servo is trying to get to where it is commanded, but it > is hard to do with the backlash. You probably need more gain, > but the system must be stable, first, or the vibration will > cause damage. Try some deadband, then maybe you can turn the P up. Dave > > Jon > > ---------------------------------------------------------------------- > --- > SF.Net email is sponsored by: > Check out the new SourceForge.net Marketplace. > It's the best place to buy or sell services for > just about anything Open Source. > http://sourceforge.net/services/buy/index.php > _______________________________________________ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users ------------------------------------------------------------------------- SF.Net email is sponsored by: Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://sourceforge.net/services/buy/index.php _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
