Problems tuning servos DMC-2162 Galil

[Aeroshade]

Hello all!!! This is my very first post on CNCZone!!

OK, I am having a problem tuning my servos on a Republic Lagun 4824VMC CNC machine. This is a vertical 4axis machine (X, Y, Z and A is a rotary). It had a Vickers A2100 controller originally, and we have purchased the Camsoft Pro software along with the Galil DMC-2162 motion card with ICM 2900's.

The servo amps were made by Glentek, which we changed to AMC (Advanced Motion Control) cards (Part# 30A20AC).

The servo motors for X,Y and Z are made by SEM (Part# MT40W4-68) all had tach's on the drives. X and Y axis have Heidenhain glass scales (linear scales) and Z axis has a rotary encoder.

We have had this machine since 1998 and it has served us very well, but a few months before the retrofit it started having problems with the controller.

Since we started the retrofit we have ran into many problems. The main problem was that the original servo amp cards were set for velocity mode, which Galil wants the amp cards set for current mode. This is the reason we purchased the AMC cards.

We have purchased the Galil tools program and have ran X and Y axis through this software multi. times. The main problem is with X axis. When tuning the X axis I set the step command to 500counts. Then I start moving the KD and KP in 1/10 increments until I see a good reading on the scope (25-35% overshot). I then have a program I run in the Galil tools that runs the X axis +.5in then back -.5in. It's in this test that I see that when the X axis is moving in the - direction that there is oscillation. I can't hear it, unless I turn the KD and KP up, but I can see it in my command volts. I have attached a few SS (screen shots) of my plots so you can see what I mean.

We have tried everything to get this noise out of the system. We have cleaned the glass scales, had the ball screws reground and we have installed new bearings for the ball screws. I have swapped X and Y motors. I have even hooked up the rotary encoder from Z axis to eliminate that the glass scales might be the problem.




Here is a SS with low PID's. And a 500count step command. Notice the encoder, pretty sloppy.



This is the .5in+ and - move I make, notice the blue command volt line, nice in both directions.



A little higher PID's, good over shot but it's a little slow (I think).



OK starting to see some abnormal spikes in the command volts line on the return (negative direction). This gets a lot worse the higher I move the PID's.



I talked to a tech at Galil and he suggested me to change the FV (Velocity Feedfoward) to 45, but I just noticed that I didn't put FVA=45... I will try that when I get done here.



In this SS I adjusted the accel and decel, it didn't help but you can see the command volts a little better here.



Lowered the accel and decel even more.



Here is a step command with the PID's D=157, P=16 and I=0. Better looking step, but look at the next SS for the .5in -/+ longer move.



The PID's were D=157, P=16 and I=0. Notice the command volts line now... WOW!! And this makes the machine really vibrate in the negative direction.




If anyone might have an idea why this is happening I would appreciate the help!! This machine has been down a good 7 months, and we would like to see it running again, sooner rather than later :P.

If you need any more INFO please let me know.


THANKS!!!
Aero

[Al_The_Man]

Did you watch the tuning tutorials on the Galil site?
They include one for dual feedback system tuning the PID for Glass scale and encoder.
I usually run the auto tune and then adjust as necessary to get optimum results, they very rarely run off the auto tune without some manual adjustment.
What is it like with just the motor encoder and no scale?
Al.

[Aeroshade]

Did you watch the tuning tutorials on the Galil site?
They include one for dual feedback system tuning the PID for Glass scale and encoder.
I usually run the auto tune and then adjust as necessary to get optimum results, they very rarely run off the auto tune without some manual adjustment.
What is it like with just the motor encoder and no scale?
Al.

I have talked to the techs about the dual feedback system at Galil and that is something they suggested until he saw the plot and that they are not even in both directions. It is still on the table to try, but I was hoping that since it work before it might be something silly that might be easy to fix.

When I used the rotary encoder on the X axis I got the exact same results with the scope.

Everything is pointing to the dual feedback system, but the only thing that is strange is that Y axis doesn't seem to have the same problem. I am having issues tuning it, but not like the X axis.

I have watched most of the videos on Galil's site for tuning, but I have also talked to them about what I have been doing and have sent them all these plots. Everybody seems to be at a loss. I have talked to Camsoft, Galil and AMC many times trying to figure all this out, everybody is stumped and can't come up with a true fix.


Thanks for the response!!
Aero

[Aeroshade]

I just ran Autotune a couple times, and every time I run it it comes up with different numbers for the PID. Does Autotune run pretty fast? it seemed to only do one or two step moves then it's done.

[Al_The_Man]

I have an older version of the tuning program, you have a few more features, but I usually find the auto tune usually takes a while, while it plugs in different frequency values.
Al.

[Aeroshade]

When I ran the Autotune it took about 3sec then stopped, but every time I used it it came up with new PID's.

[Farmers Machine]

I used a ratio for P I D of 10 / 1 / 100 starting at 50. My gantry was a little coarse and jerky. kept changing until it leveled out at 35 / 3.5/ 350 . I never hooked it up to the galil servo tuning software, it was more like blind hog finding an ear of corn. The ratio was suggested by the cam soft tech dept.
Good Luck
The Farmer