LinuxCNC 2.7.3 Trajectory Planner, circle problem

In sim is the same problem...

It seems to me there was some discussion during the testing of the new trajectory planner about the last or next to last move and acceleration because of the need to come to a complete halt at the end of the program.


What happens if you try this...
or this...
Rick G

I can't test this program on machine right now, but on simulation is the same problem. You can see on graph that velocity slow down on last circle. I think if that slow down is on simulation then machine will behave the same.

Hey Guys,

I would like to bump this topic because It is exactly the problem I am facing with the new trajectory planner and I could not find something newer. I have had this problem for a long time already and since I did a lot of circular cutouts last weekend I finally decided to investigate further.

So, I am on 2.8.0-pre1-3902 and whenever I go down in circles I get a stutter after the first segment because linuxcnc comes to a complete stop, then reaccelerates and continues. I sat down today at home and used the simulator. It shows the exact same behaviour. I did play around with G64 but that had no impact at all. Also what I did to the ARC_BLEND settings didn't have any impact.

When I disable the new planner and switch to 1 segment look ahead it behaves as expected. No slow down after the first segment!

Is this a known problem? Is there a solution (apart from using the old planner)? Since I could not find anything else but this thread, are there any other threads that deal with this problem and came up with a solution?

For my tests I used the "gantry_mm" sim config and changed the accelerations to 750mm/s^2, the g-code is below.

Cheers,
Christian.

Hi

What you are describing sound like the bugs reported on GIT, 447 and or 426. They cause me no end of grief when plasma cutting.

John

Personally I believe that it is a bug. But there are some others who may argue that it doesn't really exist.
If you play with the sequence of moves before or after the glitch. Sometimes you can make it go away. For example if you take your example code and change the g1 move before the 1st helical move to a g0 move, the pause goes away.

Thanks for pointing me to these issues. This is exactly what I am encountering. From what I read on git there should be fixes for both these issues but they don't seem to be merged because ARC_BLEND_FALLBACK_ENABLE doesn't seem to do anything in my case.

Strange things are happening when you play around with the code: If you set all the Z heights to 0 it will not slow down. If you set the Z height in the G2 line to 0 it will start to slow down a line further down in the code… And also not as much?? If something goes away by playing with what you do before or after I think it is safe to consider it a bug.

Someone commented on GIT an hour ago and asked if the fix for #426 is ready to be merged, so maybe something is going to happen... If you guys need anything to be tested feel free to contact me. I have quite a lot more circular cutouts to do this weekend.

Cheers,
Christian.