I was told by a Boeing engineer to make my (linear) scales 5 times better than the position accuracy I needed. On a well used machine 0.2 thou is way better than I need. My experience is that one should use the linear scale for position and add a incremental encoder directly coupled to the end of the ball screw for PD. On one of my machines I actually use a small idler gear to run the encoder and position this idler to run off the timing belt between the servo motor and the ballscrew. 100K+ counts/inch certainly makes for easy tuning; but this is also helped by the analog tach on the servo motor going back to the servo amp.
To summarize: Glass scales gave me excellent position but were very difficult to tune. Encoders on the ballscrews were easy to tune but were worse in positional accuracy (because of backlash). Since you already have the expensive part (linear) adding encoders on the ballscrew for P and D and using the scale for I would give you the best of both worlds. Having said that; best wishes on your endeavour. Dave On 05/30/2015 12:29 PM, Mark Johnsen wrote: > Chris, > > Part of your reasoning is why I thought the higher resolution scales would > be an answer because I would never in a million years worry about dither at > .00004"... But, I do think I need to see how well I can get it going > before that because it seems as though I should be able to get the mill > working well enough for my needs. > > In John Thornton's tuning tutorial, he talks about having 0 output to the > drive and not having any creep. I do have some slight creep, but then it > stops, so I could look at that (and then add it back). But, I think the > point is that I need to work w/ the system, look at the lash and see where > I'm at. > > Also, PCW mentioned adding a smidgen (0.010) of pid-bias, which is very > similar to what you're suggesting. I'll play around w/ both and how the > system responds. > > Thanks for the suggestions, bottom line is I need to see what I have and > will get out to the garage in a bit... > > Thanks, > Mark > > > On Fri, May 29, 2015 at 11:38:54PM -0400, John Kasunich wrote: >> Have you tried deadband equal to one or two counts of the scale? > I think adding deadband of one count will make it dither across two > counts instead of one. The problem is when pid's output is zero, > the position creeps. After a while there's an error of one count, > pid corrects it, and then pid output goes back to zero and it starts > again. Adding deadband just makes it not correct until you get more > error. > > Dithering across one count is very very normal. > > I think you can slow or stop the dither simply by turning the offset > knob on the amp. > > For velocity mode I recommend I=0 in your tuning. > > You are unlikely to permanently stop the dither because analog > adjustments aren't very permanent, but you can make it slow enough > it won't bother you. > ------------------------------------------------------------------------------ > _______________________________________________ > Emc-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-users ------------------------------------------------------------------------------ _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
