pid tuning in torque mode

I'm trying to tune a fairly large knee mill. I cant seem to get my x or y axis below a 0.0025 ferror with only a 120 inch per minute movement. I also cant seem to get it to over come the friction for say a 0.0001 movement while trying to find zero on a g54 or g55, if you know what I mean. does anyone have any suggestions on the proper order of adjustment with regards to p first I second and d last, or am I wrong. my I is aproxametly 10x my p and d is small

on my y axis
p is 1500
I is 17000
d is 3.5

with that setting the ferror is as low and as clean as I could get it.

That seems like a lot of I. Your I is usually the smallest number. If memory serves me.right, ff1 ,in small amounts helps in getting current mode amps tighter. There are quite a few tuning tutorials on the net. I swear it is a black art......

any ff1 I add makes it worse ill try from scratch with p and ff1. all of the tutorials I see are for velocity loops. it seems like the more I I add the better it got and after 2 days this is where I ended up it seemed high to me too.

Sometimes you can get an improvement by increasing the servo thread rate (if you have not already done this)
increasing the servo thread rate will increase the PID loop bandwidth, allowing higher stable D and P gains

Also for torque mode a little FF2 can be used to compensate for inertia

Increasing the thread rate helped me a lot on one of my lathes. I was using amc amps in current mode .

Im using amc amps too. Ill try this as soon as I get back to the shop tomorrow. Would you have any suggestions as to what I should increase the servo period too. or where to start.

What is your encoder resolution (steps/inch)?

The torque mode servos system I tuned was very sensitive to ff1 and ff2, using only very small amounts. But the little bit that I did use seemed to make a big difference. If I remember right I also had pretty high I term numbers on at least one of the axis between 8-200 times the size of the D term. The ff1 was between 0.2 and 2.5 and ff2 was around 0.1. These seemed to help a lot to combat stiction. The velocity drives I've tuned (if the drive is tuned right first) were a lot simpler to tune. The feed forward terms seemed to behave very differently on the torque mode drives than on the velocity drives.

Here is a link to the thread where I was going through some of my troubles tuning those torque mode servos.
linuxcnc.org/index.php/english/forum/10-...ning-advice?start=30

Hopefully the thread isn't to long or confusing, it may have also uncovered a couple of bugs (that are probably fixed or at least been addressed)

Encoders are 40640 steps per inch. Thank you for your reply. Im going to check your post out.