Welcome, Guest. Please login or register.
Did you miss your activation email?
November 17, 2017, 11:38:52 PM

Login with username, password and session length
Search:     Advanced search
* Home Help Search Calendar Links Login Register
+  Machsupport Forum
|-+  Third party software and hardware support forums.
| |-+  HiCON Motion Controller (Moderator: Vital System Support)
| | |-+  Backlash copensation issues
Pages: 1 2 »   Go Down
Print
Author Topic: Backlash copensation issues  (Read 2176 times)
0 Members and 1 Guest are viewing this topic.
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« on: May 05, 2017, 10:26:17 AM »

Hello Vital System Team

I have noticed a little issue with Mach4, Hicon7766 and backlash compensation. It is not adding backlash comp to .0001inch moves while jogging, Not sure if it is a Mach4, but assume it is a hicon issue because that is where the backlash is entered in the system. .001 and .01inch moves backlash works on most of the time every now and again it does not apply backlash


Mach4 version 2.0.3233 and the latest test version both have been used
Hicon 7766, FW:3.46 FPGA:7E7C
5mm Per revolution ballscrews
10000PPR Servo motors using encoder feedback to the Hicon Closed Loop from motor encoder direct drive to the screw
50800 steps per inch in the motor setup units
velocity set at 175
acceleration 50
backlash entered was .0009 Hicon calculated tp .00088583

This was tested with MDI, MPG pendant plugged into Hicon MPG port, a Modbus MPG pendant that is being worked on, and with incremental jog controls From the Mach4 screen.

First Thing that was noted is backlash comp will not be added to .0001 moves using jog controls or MDI.  The Machine is capable of the move and it does when going one direction. as soon as you reverse direction at .0001 increments the DRO on the mach screen will show the axis moving the backlash amount before it shows movement on the indicator. If the increments are set to .001 it works fine most of the time.

Second thing that was noted is  testing backlash on a machine using a part program and not a jog move. The program  moves to point where the indicator was zeroed, off the block past zero by a inch then back to the Zero point. if the indicator shows that it is loaded then comp is added by the loaded amount, If the indicator shows that it is light then the comp is too much and need to be reduced.

here is the program that is used for Y axis.
G01 X0.0000 Y-1.0000 F10
G01 X0.0000 Y0.0000 F10     (Part Zero)
M01
G01 X0.5000 Y0.0000 F10
G01 X0.5000 Y1.0000 F40
G01 X0.5000 Y 0.0000 F40
G01 X0.0000 Y0.0000 F10     (return to Part Zero)
M01
M30

The program is one that is run several times before it is determined the the backlash comp Value good. The reason I do this is if you over comp the machine by .00005 it will cause the Zero to walk. When testing our machines at work we do it at least 3 times knowing that the indicator we use is only accurate to .0001 we allow .0001 of error. but If we see the indicator start moving a little when we go back to zero each time we will run it 10 time in a row to look for over comp.

I was doing this and noticed on the 9th run of the program that compensation was not applies at all. Opened up the diagnostics screen and started watch the encoder counts for first touch of zero count and second touch of zero count.
when the backlash is being compensated for the first touch of Zero was 238386, Return to Zero was 238343
when backlash comp was skipped first touch of zero was 238386, return to zero was 238386. This was also reflected on the indicator  showing about .0009 loaded
ran the program 21 times and 5 time it did not apply the backlash compensation
commanded position for each run for first touch of zero and return to zero was 238364 if backlash move was applied or if it was skipped.

Ran the motion planner from the diagnostics page it look like it just shows the encoder position or step generator of what ever axis you select but does not show commanded to compare it to.

Is there any thing I can provide to help with this issue or help diagnose the issue I seem to be having.
I would swear the not comping backlash for a .0001 jog move worked several plugins ago,  But i can not be for sure  will the early 2016 plugins, FW and FPGA still work with the latest revision of Mach4?  I would go back and try them out.

Thank you
Robert


Logged
Vital System Support
Active Member

Offline Offline

Posts: 279


View Profile
« Reply #1 on: May 05, 2017, 03:58:54 PM »

We tested using 0.01, 0.001 and 0.0001 backlash compensation and it is being applied properly. We used jog, MDI and g-code. The test was performed using plugin version 2.01.54, Mach4 build 3365 and Integra firmware 3.51.

- Ricardo
Logged
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« Reply #2 on: May 05, 2017, 08:20:02 PM »

That is great when will the new plugin and firmware be released? It is not on the Vital system website.

Thank you
Logged
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« Reply #3 on: May 18, 2017, 02:19:51 PM »

Mauri
somebody was up late you and Rufi both.
Rufi
I do some embedded work my self if any one understands, fix one thing break another not even related,  I am with you. Then to top it all off add in you are working with a ever changing user interface that could be driven by widows7,8,or 10.
That is hard.  The backlash issue is not a show stopper for me. let us know i am willing to test any thing you need, if there is a better way we can collect data for you.

Rufi, don't let this consume your week end.

Thank You
Logged
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« Reply #4 on: May 19, 2017, 09:13:36 PM »

I have seen it all ready. tried it with 3390 and 3233.  I am collecting data to send to them.  Are you sending what you find via email? does Rufi have a direct email.

still a little backlash issue left.  When running the Gcode test one after another it will miss the backlash comp all together 4 out of 20 times. but it does comp with a .0001 move now.

down side that was observed is, after referencing the machine it will get locked in a sync error between mach4 and the Hicon, telling me to check that all the motors mapped should not be set to undefined.
it will continue in this loop till the disabled button is hit, then I can resume. start the gcode program when it stops at the M01. it does the same thing again. After a disable toggle it will run the rest of the program until the end  
then show the error again. after about thirty minutes it will sync up and run fine. shut down Mach and it will start all over again with errors.

Can you or Rufi get me a link to 3383 so we are all testing the same version of Mach.

THank you
Robert

Logged
Mauri
Active Member

Offline Offline

Posts: 211


View Profile
« Reply #5 on: May 19, 2017, 09:37:37 PM »

Sparky3,
Rufi, found my problem must have had the wrong profile Too many tests.
I can now do a proper test with V3383
I do not have any issue with Backlash I use it down to .01mm.

The link for V3383 has been taken out they only have V3390 on the Mach4 Website at present.
Send me a personal message with your email address and I can send it to you.
Best Regards,
Mauri.
« Last Edit: May 19, 2017, 11:39:43 PM by Vital System Support » Logged
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« Reply #6 on: May 20, 2017, 12:01:06 AM »

Mauri, Rufi
Changed the polling back up to 100 that took care of the communication issue, just had to power cycle the Hicon to get it to stick.
testing new plug in now.
Logged
Mauri
Active Member

Offline Offline

Posts: 211


View Profile
« Reply #7 on: May 20, 2017, 01:42:42 AM »

Sparky3,
The latest Plugin works for me with V3383 no more issues.
Air test went well with no stoppages.

Wait for Rufi's new V3390 Plugin before testing V3390.
Did you get my File?

Thanks Rufi for your hard work.
Regards,
Mauri.

Logged
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« Reply #8 on: May 20, 2017, 09:07:45 PM »

Rufi.

I will try the plugin later tonight.
Backlash is working well now. just the set still thing going on.
I am using closed loop.
tuning may be a little loose. but the backlash is still the same when it is open loop. I will turn the feed back off and test again to confirm.

I don't have a SDK for 3390, this is the link on the FTP server http://ftp://anonymous@ftp.machsupport.com/Mach4/DevlopmentVersions/. it is just a EXE file.

I think it would be great if it was built against the release 3233.

Thank you
Sparky3
Logged
sparky3
Active Member

Offline Offline

Posts: 31


View Profile
« Reply #9 on: May 21, 2017, 09:48:21 PM »

Mauri
I did down load the Mach4 3383 and dropped it on top of the existing installation. it works

Rufi
the 63 plugin did in fact get rid of the Set still delayed issue I was having. I would like to know what is different between my system and Mauri's system, why 62 work so well for him but not for me.

My system
64bit windwow10 pro
Hicon
Novakon mill with Maxsine Servos  and  drives, step and direction input

After watching the servo moves with the Hicon plug in last night each move showed set the up 150IPM and velocity of 40 and they would show about 3200 following error. the motor was lagging the comanded move during the move but would end up at the right destination just a little behind.
so I did as you suggested and disabled the encoder feed back.
the results are
encoders off- incremental jog and MPG will compensate for the backlash for each move.
encoders on- it is the same as above
encoders off- gcode program testing for backlash still misses the backlash comp 4 or 5 out of 20 back to back runs.
encoder on-  same as above.  on the gcode test you can watch the commanded and feedback counts are the same the the step generator is different for  Zero and return to Zero position, when they all match that is when it missed the compensation.

so now it is time to dig into the servo tuning, and well what I found is the gains are set at factory defaults for the drives only thing that was changed were inputs and output settings for the drive set up.
Played with P021 and P009 and applied more gain to the position loop , and some feed forward, it changed the following error from 3200 to around 190 - 200 counts for the same 1 inch move at.  then I notice that the other gain is set to the Factory default. Before any more backlash testing is done I will play with Servosoft see if the tuning can be done from the software, or may just have to wing the drive tuning and using  Hicon plugin and my ear to determine the tuning.  

Thanks for the 3233 plugin I will most likely go back to using the release version for now.

Thank you for the great service
Sparky3

Logged
Pages: 1 2 »   Go Up
Print
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.20 | SMF © 2013, Simple Machines Valid XHTML 1.0! Valid CSS!