LinuxCNC S-Curve Accelerations

Hi,

Just finished a template, where so far as i can see all scenario's are implemented for a traditional lineair curve.
This is a output that happens within 1 gcode line.

This includes :

• User changes max velocity during program run.
• Target velocity can not be reached by input displacement. 
• Curve up, curve down, curve steady.
• Curve respects max velocity.

Now i can try to apply the the scurve algo to this working demo.

code
video

Hi,

I just tested the scurve on top of the traditional linair curve impementation.
It works.

For using the gain value, time periods are different. I have to code this in the next "final" step.
Using the gain value can speed up the machine acc, dcc stages.

So far so good !

code
video

The "final" step is coded, and it worked!
The 4 pictures containing different gain value's, are showing the difference in machine acc, dcc and traject time outputs

code
picture


 

Hi,

I tested the algo in hal-core. It runs ok.

video

write on wrong place ... sorry

I thought I might shime in with some experience with how atleast some comercial cnc's handles this acceleration profile (Mazak Integrex and Cincinnati Arrow).

My Cincinnati Arrow has Siemens 2100e controller with die & mold package, which means S curve accelerations among other things.
This is a regular 3 axis vmc so no super advanced planning required. When doing moves that requires spindle sync, ie. rigid tapping or feed-per-tooth, it reverts back to trapezoidal acceleration profile. It will also use trapezoidal profile for jogging and handwheel.

The Integrex does not have true S curve profile but not trapezoidal either, rather something in between. This is a 5 axis millturn btw.
But this machine will also revert back to trapezoidal acceleration when doing rigid tapping and also thread turning or hobbing.
It will not use trapezoidal for feed-per-revolution feedrate, and thank god for that as a lot of turning involves 90 degree direction changes.

The point is that for lathes and mills an S curve sync'ed spindle is not really used in any situation. Trying to do rigid tapping when moving on the last bit of that S profile would probably stall the spindle during such a move.

Now of course linuxCNC probably have uses for S curve synced spindle far beyond what I can see, but when it comes to milling or turning it seems that is not required. In fact both above machines have a modal G code to activate S profile (and some advanced lookahead) separately from standard trapezoidal.

Other than that, thank you for pushing this forward! It's a really crucial feature.

hi arvidb
I haven't been on the forum for a long time.
I wanted to know if it was possible to solve the problem that I understood on forum.linuxcnc.org/38-general-linuxcnc-q...elerations?start=210

If you are referring to installing ethercat, yes its solved by using a repository containing IgH Etherlabmaster maintained by them
I wrote a Sticky on it in the Ethercat section on how to do this.
forum.linuxcnc.org/ethercat/45336-etherc...-how-to-step-by-step
Everything else uses an obsolate version of the IgH source code on Mercurial when their repositories have moved to gitlab
Next on the agenda is a custom homing module for cia402 ethercat drives. 

I would suggest building linuxcnc on Debian Bullseye from source and follow that sticky using these instructions
www.qtpyvcp.com/install/bullseye.html
 

TheRoslyak wrote:

hi arvidb
I haven't been on the forum for a long time.
I wanted to know if it was possible to solve the problem that I understood on forum.linuxcnc.org/38-general-linuxcnc-q...elerations?start=210

Hi TheRoslyak,

The integration of my jerk limited jog planner is not done yet, no. I'm still working on it now and then, when I can find enough time and energy. I did get some significant preparatory patches into master recently. Lots of work still left to do though. Untangling the innards of motmod enough to enable modular jog planning is a large undertaking, and that's pretty much what I'm doing.

Grotius wrote:

Hi,

Now i can try to apply the the scurve algo to this working demo.

code
video
[/list]

Colleague, all the Githab links you cited above are dead, I can't download anything. Please share information about the current state of things.
I would like to implement the s-curve feature in my LinuxCNC2.8 example and would appreciate any help, guidance, instructions, links, etc.