Dynomotion

Group: DynoMotion Message: 13115 From: juhakoljonen Date: 4/8/2016
Subject: KFlop open loop axis jerk - large time slice

Hey,


I have been having some axis jerk problems using KFlop + Granite Devices Argon servo drive. The position loop is closed at Argon, so KFlop is sending step+dir in open loop mode. Everything usually works ok, but sometimes, maybe with 1 hour intervals, both my X and Y servo axes jump about 0,5-2 mm. I have verified that Argon still knows where the axis is, so then I started to suspect that the problem is at KFlop's side. I tried both step+dir and quadrature outputs with various speeds, those made no difference.


I then ran into a topic at cnc forums: https://en.industryarena.com/forum/intermittent-axis-position-issue--251655.html and there I got a max time slice computing code, which I basically copy pasted to my init.c thread 1. Then I waited for the axis jump, and there was really a huge step in the time slice recorded! Screenshot included. I have also included my other .c files, although only the first thread runs while running my gcode. It does not seem to matter which gcode I am running, it is 4 axes, as I am using KFlop to control a large 3D printer.


The version command replies: Version
KFLOP 4.33q Build 17:47:12 Aug 29 2015
KFLOP 4.33q


Any advice on what to try next would be really helpful!



Thank you,

Juha Koljonen

 

  @@attachment@@
Group: DynoMotion Message: 13139 From: Tom Kerekes Date: 4/12/2016
Subject: Re: KFlop open loop axis jerk - large time slice [2 Attachments]
Hi Juha,

Did you ever resolve this?  Is there any more data that you can provide?

We are not able to replicate the problem.  We have tested the latest code running for days without detecting any Time Slice over 182us.

We have tested with your exact Version 4.33q + KFLOP + Kanalog + your C file as well.

Do you have a noisy environment?  Anything special about your system?  Could you test with all your I/O Disconnected and/or with Kanalog disconnected to see if the problem still occurs?

Regards
TK



On 4/8/2016 3:18 AM, juhakoljonen@... [DynoMotion] wrote:
 

Hey,


I have been having some axis jerk problems using KFlop + Granite Devices Argon servo drive. The position loop is closed at Argon, so KFlop is sending step+dir in open loop mode. Everything usually works ok, but sometimes, maybe with 1 hour intervals, both my X and Y servo axes jump about 0,5-2 mm. I have verified that Argon still knows where the axis is, so then I started to suspect that the problem is at KFlop's side. I tried both step+dir and quadrature outputs with various speeds, those made no difference.


I then ran into a topic at cnc forums: https://en.industryarena.com/forum/intermittent-axis-position-issue--251655.html and there I got a max time slice computing code, which I basically copy pasted to my init.c thread 1. Then I waited for the axis jump, and there was really a huge step in the time slice recorded! Screenshot included. I have also included my other .c files, although only the first thread runs while running my gcode. It does not seem to matter which gcode I am running, it is 4 axes, as I am using KFlop to control a large 3D printer.


The version command replies: Version
KFLOP 4.33q Build 17:47:12 Aug 29 2015
KFLOP 4.33q


Any advice on what to try next would be really helpful!



Thank you,

Juha Koljonen

 


Group: DynoMotion Message: 13142 From: juhakoljonen Date: 4/13/2016
Subject: Re: KFlop open loop axis jerk - large time slice
Hi Tom,

thank you for your reply. I have found a parameter change that seems to have fixed the issue for now, I had not reported it here yet because I wanted to be sure before making such statements..

After getting the time slice data flowing, I noticed that the issue appeared often at a certain point of my G-code. Then I played around with many parameters and found that the "Break Angle" in KMotionCNC trajectory planner causes axis jump in my test case when it is high, like 35 degrees. When lowered to for example 18 degrees, the machine runs great, no axis jumps even after about 10 hours of testing.

My tool setup->trajectory planner settings and test G-code are attached. The test G-code should produce jump in about 10 seconds from beginning when using 35 degree break angle and otherwise same settings as in the screen capture. It seems to be repeatable, which is nice.

There is nothing really special or strange in my setup, and as you know my KFlop is even running in open loop so there should be very little trouble at KFlop side. All axis drivers have filters before them and my logic power supplies are made by Murr, so they should be pretty high quality supplies.

I hope that you can now reproduce the issue with your hardware.


Juha


---In DynoMotion@yahoogroups.com, <tk@...> wrote :

Hi Juha,

Did you ever resolve this?  Is there any more data that you can provide?

We are not able to replicate the problem.  We have tested the latest code running for days without detecting any Time Slice over 182us.

We have tested with your exact Version 4.33q + KFLOP + Kanalog + your C file as well.

Do you have a noisy environment?  Anything special about your system?  Could you test with all your I/O Disconnected and/or with Kanalog disconnected to see if the problem still occurs?

Regards
TK



On 4/8/2016 3:18 AM, juhakoljonen@... [DynoMotion] wrote:
 

Hey,


I have been having some axis jerk problems using KFlop + Granite Devices Argon servo drive. The position loop is closed at Argon, so KFlop is sending step+dir in open loop mode. Everything usually works ok, but sometimes, maybe with 1 hour intervals, both my X and Y servo axes jump about 0,5-2 mm. I have verified that Argon still knows where the axis is, so then I started to suspect that the problem is at KFlop's side. I tried both step+dir and quadrature outputs with various speeds, those made no difference.


I then ran into a topic at cnc forums: https://en.industryarena.com/forum/intermittent-axis-position-issue--251655.html and there I got a max time slice computing code, which I basically copy pasted to my init.c thread 1. Then I waited for the axis jump, and there was really a huge step in the time slice recorded! Screenshot included. I have also included my other .c files, although only the first thread runs while running my gcode. It does not seem to matter which gcode I am running, it is 4 axes, as I am using KFlop to control a large 3D printer.


The version command replies: Version
KFLOP 4.33q Build 17:47:12 Aug 29 2015
KFLOP 4.33q


Any advice on what to try next would be really helpful!



Thank you,

Juha Koljonen

 


  @@attachment@@
Group: DynoMotion Message: 13166 From: Tom Kerekes Date: 4/19/2016
Subject: Re: KFlop open loop axis jerk - large time slice [2 Attachments]
Hi Juha,

I think the main issue is the "facet angle" setting on 0.1 degrees.  I believe this is excessively small.   That makes a single 35 degree corner generate 350 linear segments to round the corner.  But we certainly should handle that issue more gracefully.   Please try a setting of 1 degrees to see how it works for you.

Regards
TK

On 4/13/2016 1:54 AM, juhakoljonen@... [DynoMotion] wrote:
 

Hi Tom,

thank you for your reply. I have found a parameter change that seems to have fixed the issue for now, I had not reported it here yet because I wanted to be sure before making such statements..

After getting the time slice data flowing, I noticed that the issue appeared often at a certain point of my G-code. Then I played around with many parameters and found that the "Break Angle" in KMotionCNC trajectory planner causes axis jump in my test case when it is high, like 35 degrees. When lowered to for example 18 degrees, the machine runs great, no axis jumps even after about 10 hours of testing.

My tool setup->trajectory planner settings and test G-code are attached. The test G-code should produce jump in about 10 seconds from beginning when using 35 degree break angle and otherwise same settings as in the screen capture. It seems to be repeatable, which is nice.

There is nothing really special or strange in my setup, and as you know my KFlop is even running in open loop so there should be very little trouble at KFlop side. All axis drivers have filters before them and my logic power supplies are made by Murr, so they should be pretty high quality supplies.

I hope that you can now reproduce the issue with your hardware.


Juha



---In DynoMotion@yahoogroups.com, <tk@...> wrote :

Hi Juha,

Did you ever resolve this?  Is there any more data that you can provide?

We are not able to replicate the problem.  We have tested the latest code running for days without detecting any Time Slice over 182us.

We have tested with your exact Version 4.33q + KFLOP + Kanalog + your C file as well.

Do you have a noisy environment?  Anything special about your system?  Could you test with all your I/O Disconnected and/or with Kanalog disconnected to see if the problem still occurs?

Regards
TK



On 4/8/2016 3:18 AM, juhakoljonen@... [DynoMotion] wrote:
 

Hey,


I have been having some axis jerk problems using KFlop + Granite Devices Argon servo drive. The position loop is closed at Argon, so KFlop is sending step+dir in open loop mode. Everything usually works ok, but sometimes, maybe with 1 hour intervals, both my X and Y servo axes jump about 0,5-2 mm. I have verified that Argon still knows where the axis is, so then I started to suspect that the problem is at KFlop's side. I tried both step+dir and quadrature outputs with various speeds, those made no difference.


I then ran into a topic at cnc forums: https://en.industryarena.com/forum/intermittent-axis-position-issue--251655.html and there I got a max time slice computing code, which I basically copy pasted to my init.c thread 1. Then I waited for the axis jump, and there was really a huge step in the time slice recorded! Screenshot included. I have also included my other .c files, although only the first thread runs while running my gcode. It does not seem to matter which gcode I am running, it is 4 axes, as I am using KFlop to control a large 3D printer.


The version command replies: Version
KFLOP 4.33q Build 17:47:12 Aug 29 2015
KFLOP 4.33q


Any advice on what to try next would be really helpful!



Thank you,

Juha Koljonen

 



Group: DynoMotion Message: 13167 From: juhakoljonen Date: 4/19/2016
Subject: Re: KFlop open loop axis jerk - large time slice
Hi Tom,

alright, thank you, I will definitely try that! I will test for a while and report back my findings.


Juha