Tom wrote: > > Since I have not had the opportunity to witness other EMC2 servo conversions, > I > am guessing / interpreting what you refer to as "stable".
Stable means that when the commanded position changes and then remains stationary, the axis moves to the new position and settles there without hunting around. A single overshoot (if small) and return to the proper position might be considered OK, but if it goes back and forth several times, the tuning isn't right. > I assume you mean > that the mid frequency moaning/buzzing goes away. My mill will quiet down > after > sitting at rest for about 30 seconds. I can see this on halscope in the form > of > a sawtooth trace that hovers close to zero. Perhaps my deadband and D term > can > be adjusted so that it quiets down in a few seconds. But then, from > experience > I find that following error goes up. > > Question: > If I tune the PID for the lowest following error while making the buzzing > sound > (sawtooth trace) go away, (and set MIN_FERROR & FERROR accordingly), but then > if I am not happy with the accuracy of my cuts at normal feedrates (20-70 > ipm), > will I then need to change to higher count encoders? (Currently Renco 500 > line > / 2000 ppr) No servo system can control to better than one encoder count. Unless your drives, motors, tuning, and everything else are perfect, you will have multiple counts (maybe even several tens of counts) of error while moving, even more during accel and decel, and a count or two when you stop. I was googling for some halscope PID waveforms, and came across a couple of nice blog posts by Anders Wallin: http://www.anderswallin.net/tag/pid/ , complete with scope screenshots. It is a good read - he initially had 0.02m (0.0008") peak error (about 32 counts), but eventually got it down to about 0.01mm (16 counts, 0.0004"). The peak errors are during accel and decel - during the cruise portion of the move it was 5-10 times better. Ideally, an encoder count is several times smaller than your accuracy target. If I was building a machine that wants to hold 0.0001" tolerances, I'd aim for roughly 100,000 counts per inch. Of course, if I'm also aiming for 60 ipm = 1 ips, then I need to be able to count 100,000 counts per second - that is where hardware counters start to shine. Anders' machine has encoder counts of 0.000625mm - that is 40,640 counts per inch. If you have lower speed or lower accuracy requirements, then you can get away with software counting. For example, if I want to work to 0.0005, then I'd aim for about 20,000 counts per inch. Software encoder counting can work to 20-30,000 counts per second (depending on your base period of course). That means the machine could rapid at 1.0 to 1.5 inches per secone, or 60-90 ipm. Everything is a trade-off, and machine designers must do the math to make the right decisions. Regards, John Kasunich ------------------------------------------------------------------------------ This SF.net email is sponsored by: High Quality Requirements in a Collaborative Environment. Download a free trial of Rational Requirements Composer Now! http://p.sf.net/sfu/www-ibm-com _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
