Servo Tune Newbie

This is my first attempt at tuning a velocity servo loop. I feel it could be better but unsure what direction to take on settings. I mostly followed this tutorial: gnipsel.com/linuxcnc/tuning/servo.html . Are there any other ones to look at? Anyways here are the details...

Drive is an old Fanuc with DC (perm magnet) Gettys motor Model 20
Screw pitch is 10mm

P 40
I 0.065
D 0.001
Bias 0
FF0 0.001
FF1 0.625
FF2 0.0015
Deadband 0.00004

Is this a metric or an Imperial machine?

Your worst-case following-error looks to be about 1 milli-unit. So that is either 1 micron (really pretty good) or 1 thousandth (less good, but not awful during what would be (in that case) a 200 in/min move.

It is a imperial machine. Any tips for getting it better?

I calculated my encoder resolution at 0.0000393701in for a 10mm pitch screw on a 2500 quadrature encoder. Maybe you can check to see if my math is correct? Should I be using this number as the deadband or up it a little as I did?

DaOne wrote: It is a imperial machine. Any tips for getting it better?

0.001" f-error during a rapid isn't horrible. What is the f-error during a cutting-speed move?

I will test and report back. Any specific speed you want me to try? Say 20ipm good? Or slower?

DaOne wrote: I calculated my encoder resolution at 0.0000393701in for a 10mm pitch screw on a 2500 quadrature encoder. Maybe you can check to see if my math is correct?

I get the same number, assuming it is 2500-slots for 10,000 counts.

DaOne wrote: I will test and report back. Any specific speed you want me to try? Say 20ipm good? Or slower?

I don't know what is typical. I am still re-calibrating away from my mini-mill where 20mm/min was about as much as it could do. I still get nervous when I program a 100mm/min cut (4 in/min)

Maybe a more experienced imperial machinist can suggest a sensible number.

andypugh wrote:

DaOne wrote: I calculated my encoder resolution at 0.0000393701in for a 10mm pitch screw on a 2500 quadrature encoder. Maybe you can check to see if my math is correct?

I get the same number, assuming it is 2500-slots for 10,000 counts.

Yes that's correct 10,000 counts. Thank you for checking my math. Should I use that as my deadband number? Also I noticed that before I adjusted the "I" value all of my moves were exactly 1 thousandth off in the positive direction. putting in some "I" helped with that problem.

andypugh wrote:

DaOne wrote: I will test and report back. Any specific speed you want me to try? Say 20ipm good? Or slower?

I don't know what is typical. I am still re-calibrating away from my mini-mill where 20mm/min was about as much as it could do. I still get nervous when I program a 100mm/min cut (4 in/min)

Maybe a more experienced imperial machinist can suggest a sensible number.

I will post the scope screens with a few different cut speeds then for you to look at. Heading over to the shop now.