Kinematics for XYZAB mill

Thanks Rod,
1) I noticed the wrong rotation sense as well. We did invert the sines on the rotation matrices as described in Rudy's document but since the machine in question has the rotation sense defined as being opposite the right hand rule it seem that we don't need to do that.

2) I think you might be using the Identity kinematics. With the TCP on you should not see a toolpath in the vismach model when rotating B as the tool will be following the rotating coordinate point. Try to click on the GUI button 'TCP: XYZAB' and test again. What I see is that is seems to work for the A axis but B is not correct.

Need to put my thinking cap on, again.

No thats TCP mode for sure. You can see XYZAB is selected in the black window. Jogging the rotary axes will prove that.
When I was working with ChapEng, I got him to change  the B axis rotation but he felt it wasn't required on the A so it wasn't done.
If you compare your work with R(x0) in (2), you'll see you are using the non inverted sines.

Anyway, just grab my revised component and have  a play!

What was confusing in my video was I moved 2 axes at the one time. If you move just the B axis or the A axis, the tool will track the same point. I should have done that.

Actually, you are right, its not TCP. I just had a bit more of a play

Actually, the A axis and the spindle do  coordinated TCP moves but the movement is not coordinated on the B axis (it just uses identity)
Eg, on A xis rotations, the behaivour is correct and thats the one I modified.
Maybe something is not connected correctly. I'll have  a look at the pins later..

You jumped the gun a bit there Rod, presenting that.
Finally found the mistake that was staring me in the eye. 
Here is a presentable solution that actually works for both A and B axes. So you might want to halcompile that and shoot that video again

 

Anyway, the other think I had was that Dy is not needed because the axes of our A and B DO actually intersect.
 

Sorry to keep posting but the problem with the B axis was there before I made my changes.
If you go back to the screen dumps of vismach on this post, the tool paths are displayed there on a purely  b axis rotation.
forum.linuxcnc.org/10-advanced-configura...mill?start=40#258349
It seems really odd to have made a mistake on just one rotating axis..
Sorry I just don't know enough on this branch of maths to help..

We don't want the public to laugh me off the streets, do we?

Just kidding Rod, I forgot the smiley when I wrote that you jumped the gun

Doesn't take much to screw up a kinematic, a 'CB' instead of a 'CA' will do.

Anyway I'll redo the pdf with the math and I'll need to have a look at the kinematics comp because I'm having issues setting the rotation point in the ini file.

[Edit]
Just to be clear, the error was in the comp I posted earlier. Nothing to do with your changes.

[edit2]
Also tool length offset is not implemented yet.

Aciera wrote:

You jumped the gun a bit there Rod, presenting that.
Finally found the mistake that was staring me in the eye. 
Here is a presentable solution that actually works for both A and B axes. So you might want to halcompile that and shoot that video again

 

File Attachment:

File Name: xyzab_trt.comp
File Size:9 KB

Anyway, the other think I had was that Dy is not needed because the axes of our A and B DO actually intersect.

 

Well the translations work but in TCP mode, I'm not sure if the axes are right. Eg to lift the spindle by 50mm while turning A & B I need to do
g0 b360 a -90 y 50
IS that expected behaviour? We might have the simple axes connected to the wrong joints.
I'll look later and redo that video.

Maybe compare with the xyzac-trt and xyzbc-trt simulation configs.