Dynomotion

Group: DynoMotion Message: 9197 From: Hugh Sontag Date: 2/12/2014
Subject: KMotionCNC movements better, but still a "jiggle"
Hi Tom,

I tried running the same file we've been working with again today.

I tried turning off "Arcs to segments", and the movements of the CNC became much smoother, and faster, too.

I ran the CNC with the G-code for a while, actually not cutting metal because the bit was moving in areas that had been cut before.

In the larger holes, there is still a "jiggle" in which the machine moves jerkily. I am concerned that this may still break a bit.

I collected the same data for X, X1, Y, and Z axes for one of the jiggles. The data, and a screen shot of the current Trajectory Planner settings for KMotionCNC, are in a file at the URL


I estimate that I pressed the switch about 0.1 seconds after the last jiggle.

Do you have any suggestions on how to make the movement in the larger hole more smooth?

Thanks,
Hugh
Group: DynoMotion Message: 9199 From: Tom Kerekes Date: 2/13/2014
Subject: Re: KMotionCNC movements better, but still a "jiggle"
Attachments :
    Hi Hugh,

    Thanks for all the great data.  I think there are a lot of interesting things going on.  Such as:

    - GCode issues
    - Trajectory Planner settings
    - Encoder Noise

    I generated several reports with the data.  See attached.

    BTW There is also a V4.31s patch to fix the GViewer Crash when viewing large numbers of Arcs.  It was missing a mutex.  It is located here (copy to KMotion\Release):
    http://dynomotion.com/Software/Patch/FixGViewerCrashLotsOfArcs4.31s/KMotionCNC.exe

    I think the primary issue is TP setting of 5 degree break angle is too small.  The original GCode has angles of up to 10 degrees in spots.  Such a small break angle should not be necessary.  If there is a "mistake" and what should be an exact stop is actually "Rounded" with a 0.5 mil radius it shouldn't be a big problem.

    Also the Collinear Tolerance of 0.001 is actually quite large.  On small circles this can result in curves being changed to linear segments with large angles (ie 20 degrees).

    I didn't have much time to look at the Mach3 problems.  The Trajectory data does show a discontinuity.  One thing to try is to turn off the Display Plot.  This is a known issue with Mach3.  There should be a switch on the Diagnostics screen to turn it off.

    What type of encoders are you using.  Capacitive?

    Regards
    TK





    Group: DynoMotion Message: 9200 From: Hugh Sontag Date: 2/13/2014
    Subject: Re: KMotionCNC movements better, but still a "jiggle"
    Thanks, Tom, for the great analysis.

    I'll try your recommendations tomorrow.

    I have difficulty zooming the G Viewer up enough to see the detail you show in your pictures. When I zoom past showing the whole circle in the window, the entire path disappears and I can't find it by panning around. Is there a trick to zooming about a point, while keeping that point in the window?

    We're using the CUI AMT-102 capacitive encoders. I tried running the G-code open loop, and I still got some jiggles. And when cutting open-loop, the bit broke within about three circuits around the inside of a 0.3" hole. That's quite a lot worse than we did when running closed-loop.

    I'll work on the noise. How did you make the X-Y plots? How is "X theoretical" and "Y theoretical" calculated? If it's from a tool that you wrote, would it be possible for me to use it in pursuit of the noise on the Y encoder?

    Is the KLP coordinated motion lowpass filter still in KFLOP? I think I might try that.

    Thanks,
    Hugh



    On Thu, Feb 13, 2014 at 11:46 AM, Tom Kerekes <tk@...> wrote:
     

    Hi Hugh,

    Thanks for all the great data.  I think there are a lot of interesting things going on.  Such as:

    - GCode issues
    - Trajectory Planner settings
    - Encoder Noise

    I generated several reports with the data.  See attached.

    BTW There is also a V4.31s patch to fix the GViewer Crash when viewing large numbers of Arcs.  It was missing a mutex.  It is located here (copy to KMotion\Release):

    I think the primary issue is TP setting of 5 degree break angle is too small.  The original GCode has angles of up to 10 degrees in spots.  Such a small break angle should not be necessary.  If there is a "mistake" and what should be an exact stop is actually "Rounded" with a 0.5 mil radius it shouldn't be a big problem.

    Also the Collinear Tolerance of 0.001 is actually quite large.  On small circles this can result in curves being changed to linear segments with large angles (ie 20 degrees).

    I didn't have much time to look at the Mach3 problems.  The Trajectory data does show a discontinuity.  One thing to try is to turn off the Display Plot.  This is a known issue with Mach3.  There should be a switch on the Diagnostics screen to turn it off.

    What type of encoders are you using.  Capacitive?

    Regards
    TK





    Group: DynoMotion Message: 9203 From: Hugh Sontag Date: 2/15/2014
    Subject: Re: KMotionCNC movements better, but still a "jiggle"
    Thank you, Tom, for your help in getting our CNC machine running smoothly. With your analysis and suggestions, we were able to run the same G-code in KMotionCNC without any jiggles or bumps!

    Hugh


    On Thu, Feb 13, 2014 at 11:46 AM, Tom Kerekes <tk@...> wrote:
     

    Hi Hugh,

    Thanks for all the great data.  I think there are a lot of interesting things going on.  Such as:

    - GCode issues
    - Trajectory Planner settings
    - Encoder Noise

    I generated several reports with the data.  See attached.

    BTW There is also a V4.31s patch to fix the GViewer Crash when viewing large numbers of Arcs.  It was missing a mutex.  It is located here (copy to KMotion\Release):

    I think the primary issue is TP setting of 5 degree break angle is too small.  The original GCode has angles of up to 10 degrees in spots.  Such a small break angle should not be necessary.  If there is a "mistake" and what should be an exact stop is actually "Rounded" with a 0.5 mil radius it shouldn't be a big problem.

    Also the Collinear Tolerance of 0.001 is actually quite large.  On small circles this can result in curves being changed to linear segments with large angles (ie 20 degrees).

    I didn't have much time to look at the Mach3 problems.  The Trajectory data does show a discontinuity.  One thing to try is to turn off the Display Plot.  This is a known issue with Mach3.  There should be a switch on the Diagnostics screen to turn it off.

    What type of encoders are you using.  Capacitive?

    Regards
    TK