SCARA hand region issues

Hi all ,
I just put together the control for an epson scara arm .
After tuning the servos and home procedure in MDI I can't move past a point .
The arm is not physically limited in any way , there's plenty of travel , but it wont go past that point .
I raised the max and min travel to very large numbers and the problem persists .
Axis returns following error on all the joints involved in the move . That is joint 0, 1 and 3 .

Looking forward to any ideas .

Cheers

Nikolai

Which version of LinuxCNC?

The axis limits are rectangular, but the joint limits are not. Perhaps this is the problem?

I'll check the version tomorrow . Should be 2.7.8
So max limit and min limit correspond to X Y Z limits and not joint limits ?
The joints are set as angular so I expected them to be degrees of freedom .

Niko wrote: I'll check the version tomorrow . Should be 2.7.8
So max limit and min limit correspond to X Y Z limits and not joint limits ?

I am not sure with 2.7.8. With 2.8 there are separate limits for both joints and axes (but the axis limits are still a cuboid)

The joints are set as angular so I expected them to be degrees of freedom .

That's probably right. Did you check for any limits applied in the kinematics?

It's 2.7.0 .
Where can I get 2.8 from ?
Max and min limit in INI are set to more than a meter (twice the length of the arm).
Where can I check the limits in the kinematics ?
Other than hal and ini I'm pretty clueless .

Niko wrote: It's 2.7.0

You should update to 2.7.8, though it won't help here.

Where can I get 2.8 from ?

Currently from buildbot.linuxcnc.org, though it might not help with your problem it will give you a system better-suited to robots.

Where can I check the limits in the kinematics ?
Other than hal and ini I'm pretty clueless .

That would be in the scarkins code:
github.com/LinuxCNC/linuxcnc/blob/master...nematics/scarakins.c
Possibly the acos in the inverse kins always returns an angle < 90?

Upgraded to 2.8 .
It works but in a rather weird fashion . As a joint Z works fine but when I switch to real world position it switches direction .

Is there any logic to debugging nonlinearity in straight moves ?
I've already checked the lengths and encoder count rates and still get a slight curve .

P.S. in MDI linear moves on x or y result in "linear move on line 0 would exceed joint z positive limit".

And the point which limits it is at 90 degrees between arms .

I am a bit surprised that you have a "joint" called "Z". Do you mean Joint 3?

Joints and axes are fine . Right now I'm wondering how to recompile stuff .

For a simple kinematics module you might well be able to use halcompile. It's certainly worth a try.