scurve trajectory planner

@Julian,

The file seems to run now under -> ./start_axis_9
I removed some segment length checks like : < 1e-12
This is a very tiny value, but it was enough to halt the planner so now and then.
Code is updated.
 

Well done!
I will update/patch and try some more codes on my testmachines this evening (living in Germany btw.). I will report back then.

Joco wrote:

And that behaviour continues under scurve planner?  That was what I was referring to. Not what standard planner does.

@Joco - my apologies, I completely misunderstood your original warning.  I thought you were concerned that the current TP maybe didn't lock out FH and motion-inhibit as advertised, and perhaps that situation was carried in to Grotius' s-curve TP.

I have not tested the s-curve TP's behavior.

Sorry to all for the unnecessary tangent.

Hi Grotius,

had some time to test. First of all. No freezes, overflows what so ever. So well done!

I found two things and I made two videos to show it:
1.) When stopping or pausing the code the DRO dont stop. this is with my 4 axis testmachine. With the 9axis sim machine this dont happen. Both machines show strange DTG values for the A axis.

Video 1

2.) Of course in the first video the max jerk limits the feed override. So I tested a insane 2000% feed override with insane max jerk and something happened. Up to 500% everything looks fine. Over 500% it stops the code. Then I have to back up to ~350% and everthing starts again. So, yeah, not a real usecase, but I found it while going crazy at testing - sorry.

Video 2

Found another one.
The slider for max velocity is working incorrect.

Video

As one can see, I test stupid high velocities, accelerations, jerk values and so on. With such experiments I am able to crash the tp. But no freezes of linuxcnc and I am able to stop the program and restart it without a problem.

@Hi Julian,

The display value for "dtg" abc, uvw is fixed. This was indeed a bug.

When moving in a blend segment, the "dtg" abc, uvw values are calculated slightly different.
This was not calculated correctly for the display value.

Then i tried max velocity overide : 20.0 and a max jerk of 20.000.
However i was not able to get the planner to run crazy.

Can you post video's with higher resolution or full screen. I cannot read the hal text in the video.
Your last video, maybe provide the file, and run values, so i can try to get same behaviour.

How do you update the code when github is updated?

living in Germany btw
Living in Holland, Next door.
 
Code is updated.
 

Hi Grotius.
I copy cmake and run ./patch and the ./install.

I will do some more testing later and will try to capture the screen. Are you sure you maximized the resolutuon? I can read everthing in my videos.

And of course I will provide the code. I think it will be best to further test with your provided 9axis sim machine so that we will have the same circumstances. But I will test also with my test 4 axis machine with connected 7i92 mesa card just to be sure I dont miss something.

Grotius,

I tested it with "20.0 and a max jerk of 20.000". With this settings the tp dont stop. Try stupidly high values (jerk 20.000.000 for example) and you will see, that anywhere in the override range of 500% to 1500% the tp stops. I think something overflows...

To be clear. I dont WANT to find a bug. I found it because I wanted to test different codes the fast way

Ok. Your new code works a lot better.

With the given 9axis sim machine everything works as it should I think.

But, if you use a my 4axis testmachine (very high kinematic values) than you are able to stop the tp and also get trouble with the axis screen. But nevertheless. Under normal circumstances the 4axis testmachine works now also without axis screen problems.

Hi Grotius,

I made a small test nc file for better testing. This is a video with my 4 axis test machine capable of:

- 100m/min velocity
- 1 g acceleration
- 6.000 rpm A axis
- wild acceleration of the A axis

My current project will be capable of 30m/min, 0,5 g and the A axis will be a turning axis (of curse with way less acceleration). So in my opinion the test machine is not sooooo way of.

Video

I try different feed override values and as you can see I set the max jerk nearly to maximum (as if there is no scurve).
The behaviour of the A axis is still strange. And as you can see I am able to stop the tp. but it seems that the planer works on and only the axis gui si stoped. Nevertheless strange.

Please feel free to mention if I am testing to wild! I can live with boundaries so to say.

Last thing I just realised. 2.9.3 need a few percentage of my cpu (i5 4570T), maybe 3 to 10%. 2.10 with your scure tp needs over 40% while in idle and up to 50% while running the program.