LinuxCNC S-Curve Accelerations

Grandixximo just wanted to ask have you done any work to fix the lookahead on multiaxis movements.  on the old trajectory planner the lookahead reverts to slower movement (almost no lookahead) when using a 4th axis.  or any extra axis after 3 axis  it's a real pain when machining

i have been playing with scurve sim but can't work out if it fixes the issue yet. so thought i would ask a question about it here

cheers

New TP2 shall have 9 axis look ahead, we are working on a different base source, i can't share a work in progress, I will be porting later, still working out feed override spikes at the moment, and thinking about remaking the TP base completely.

For multiaxis, I started also on porting and improving G12.1 (switchkins from rushabhGH), and I'm juggling with an idea to have tool length be something adjacent to a virtual axis, and if those are in place, have the proper axis limits be derived from joints + current kins + current tool

Time will tell...

Wow, ambitious, looking forward to seeing your results.

Thankyou for your contributions.

looking forward to your TP2. I would consider a few jerks in feed override an acceptable loss for using override as long as the max degredation of velocity is trapezoidal. Even a slight violation of max acceleration of an axis during the feed override is an acceptable degradation in my opinion.

If you are looking for a discussion on switchkins, I can help. When I first wrote switchkins, I could not think of an easy method to force a interpreter sync and yield type of queue buster which stops readahead. So looking at Fanuc and Hass, I added G12.1 and carried it through the complete chain of interpreter and motion. However dgarret came up with a much easier way to affect the queuebuster using M66 and remap which avoided touching so many files in the source tree for the switchkins feature.

For using axis limits during planning, the tool length which will become a parameter for the kinematics (including kinematics on the user side), could be carried through HAL or as a status variable from the #variables (#5401 to #5409).
I am looking forward to your release.
-automata

took a log of servo drives commands on a deckel dmc 70v, MELDAS 500 circa 1992, 25MHz 68040 CPU. runs at 1.77(7) ms servo cycle. rapids use trapezoidal velocity profile, cutting feed is s-curve. interestingly, the later overshoots target position up to 20 um and backs off...

 

Hello,
Now that I have EtherCAT working,
I noticed something regarding the `polygon.ngc` program: with 8 polygon points and a feed rate of 5000 mm/s,
the corner-rounding speed is 467 mm/s.
with a feed rate of 1000 mm/s, the corner-rounding speed is 172 mm/s.
As I understand it, with a feed rate of 1000 mm/s, the corner-rounding speed should also be 467 mm/s.
Is my assumption correct?
With 4 polygon points, the corner-rounding speed is the same.

Gruß
Rüdiger

Were you changing your speed from 5000 to 1000 using the feed override or by changing the feed F command in your G-code?