Servo Tuning detailed How To.

More
16 Nov 2020 19:01 - 16 Nov 2020 19:12 #189530 by viesturs.lacis

tommylight wrote: Vhat VFD? Any schematics ?


I meant that Commander VFD in Biesse machines. Nevermind, I managed to find the manual: www.acontrol.com.pl/uploads/pdf/instrukc...er-CD-user-guide.pdf

I have a question about your tutorial. And particularly about the moment, when trying to jog a motor leads to a big jump. My problem is that the first thing I tried was changing the sign of "output scale", but that does not help. Is there any other potential cause for such behavior?
One thing that makes me think that there might be some bug, is that in INI file I manually entered "max output = 6, output scale = 6". Saved the file, started LinuxCNC, opened "calibration" and there I see "output scale = 1, max output = 6". If I close LinuxCNC and open INI file in text editor, I see "max output = 6, output scale = 6".
What I did was connected pwmgen-value to a slider widget in pyvcp panel and manually set the pwmgen-value to a very low number, like 0,003. Motor movement is really slow and smooth. That means that it is all about my config. Positive pwmgen-value brings moves the motor in negative direction. And vice versa. That makes me think that "output scale" should be negative to reverse this. Encoder works correctly. Is that correct? But the motor still makes a jump, if I try to jog it.
There is only one config - there is only one folder in "configs" and there is only one INI file in that folder, so that excludes possibility that I run LinuxCNC with one config and try to edit another config. Any ideas, how to check, what am I doing wrong?
Last edit: 16 Nov 2020 19:12 by viesturs.lacis.

Please Log in or Create an account to join the conversation.

More
16 Nov 2020 19:38 #189534 by tommylight
Check the hal file and make sure the right encoder is using the right PWM.
When it jumps, does it trigger a disable ?

Please Log in or Create an account to join the conversation.

More
16 Nov 2020 20:03 - 16 Nov 2020 20:08 #189537 by bevins

viesturs.lacis wrote:

tommylight wrote: Vhat VFD? Any schematics ?


I meant that Commander VFD in Biesse machines. Nevermind, I managed to find the manual: www.acontrol.com.pl/uploads/pdf/instrukc...er-CD-user-guide.pdf


There is a schematic wiring diagram on the cover of the green commander. There is also a pot interface. Use that, real easy for spindle output.

I cant help with the 7i48. Never even seen one. With a 7i77 I tuned my CACR drives/motors in an hour 3 years ago and never been back in the control. I would suspect your pwm stuff before the analog out to be the issue. I don't know, never used that card.
Last edit: 16 Nov 2020 20:08 by bevins.

Please Log in or Create an account to join the conversation.

More
16 Nov 2020 20:13 #189539 by viesturs.lacis

tommylight wrote: Check the hal file and make sure the right encoder is using the right PWM.
When it jumps, does it trigger a disable ?


Yes, it disables the drives. Well, X and Z axes are fine, so there is only one pwmgen left and encoder works fine. Ok, I will check for 14th time tomorrow.

Please Log in or Create an account to join the conversation.

More
16 Nov 2020 20:16 #189540 by viesturs.lacis

bevins wrote: I would suspect your pwm stuff before the analog out to be the issue. I don't know, never used that card.


I have used them previously, they are great. I will return there tomorrow and probably will paste my config files somewhere, because I completely agree that it is something about config before pwmgen. Because it works great, if I manually set pwmgen-value.

Please Log in or Create an account to join the conversation.

More
18 Nov 2020 10:32 #189654 by viesturs.lacis
So I have managed to get the axes almost tuned.
I am happy with X axis (see attached picture with HalScope screenshot). Basically I have get max spike of following error 0,02 mm at jog speed 45 m/min. Just like I am happy about Z axis - following error of 0,01 mm at 6,4 m/min. And that is only with P and FF1 parameter, like Tomy suggested.

What I do not understand, is WTF is wrong with Y axis. My main issue are the spikes at the beginning and end of movements. I tried to tune this the same way I did all the rest - set the P parameter and then worked on FF1 gradually increasing the jog speed. Up to 36 m/min.
I made it to the point in which it seems to me that FF1 is good enough that movement does not lag behind nor rushes forward command - see screenshot of y_axis_36m-min. All that is left are those spikes. I have tried to remove them with FF2, but this is the best I can do - if I increase this parameter, the graph changes so that there is an increase of 0,2 mm for whole accel or decel phase.
And one more thing that puzzles me is that I dropped the velocity to 9 m/min and now the graph looks like FF1 is too much, because it runs ahead of commanded position.
I tried addding I term. Only thing it does is makes the whole graph go up or down, it does not help with the spikes. I tried also D term. It started vibrating at 0.006, so I tested it with 0.004, but the spikes are exactly the same.
I would appreciate any ideas, what is wrong with my approach.
Attachments:

Please Log in or Create an account to join the conversation.

More
18 Nov 2020 13:46 #189667 by Todd Zuercher
Can you describe the mechanics of the Y-axis? Is the feedback encoder mounted directly to the Y motor or is it mounted to a ball-screw or is it a linear feedback device?

Tuning FF2 can be very touchy. Your plot shows you were using an FF2=0.000002. What happens with FF2=0.000003 or 0.0000025

Please Log in or Create an account to join the conversation.

More
18 Nov 2020 13:55 #189669 by viesturs.lacis

Todd Zuercher wrote: Can you describe the mechanics of the Y-axis? Is the feedback encoder mounted directly to the Y motor or is it mounted to a ball-screw or is it a linear feedback device?


Encoder on motor shaft, motor coupled to ballscrew with belt

Todd Zuercher wrote: Tuning FF2 can be very touchy. Your plot shows you were using an FF2=0.000002. What happens with FF2=0.000003 or 0.0000025


So you mean that my approach was correct and that the FF2 value is not good enough?

Please Log in or Create an account to join the conversation.

More
18 Nov 2020 14:05 #189673 by Todd Zuercher
That is my suspicion. The plots of your other axis also look like FF2 could use some more attention. (And maybe a little less P for the X.)

A plot of the PID output might also be a little telling, to try to see if the f-error spikes are due to delay in response of the servo or simply a lack of FF2.

Please Log in or Create an account to join the conversation.

More
18 Nov 2020 14:34 #189675 by viesturs.lacis

Todd Zuercher wrote: That is my suspicion. The plots of your other axis also look like FF2 could use some more attention. (And maybe a little less P for the X.)

A plot of the PID output might also be a little telling, to try to see if the f-error spikes are due to delay in response of the servo or simply a lack of FF2.


Ok, thank you for a suggestion! I will try that the next time I visit that place!

Please Log in or Create an account to join the conversation.

Time to create page: 0.173 seconds
Powered by Kunena Forum