Please Help - How to Configure a 5 Axis Head-Head

Sorry for the delay.

Tommy
I wasn't suggesting this would be best for my machine but I was proposing an idea for machines in general and specifically for a company trying to build and sell machines.  My machine does not have two y motors.  It uses an "anti-racking" cable that the company millright CNC came up with (the base of my machine is Millright CNC).  I have attached a photo to demonstrate how it works, the round things are pulleys and there are two kevlar cables (one solid one dotted) mounted to the machine going through the pulleys. Btw, there is are separate pulleys for each cable. The idea is that any twisting of the gantry would induce greater tension in one of the cables.  Millright does not use this any more and now uses two y motors.  

Its an interesting idea that maybe this cable system is causing my out of square issues and that this could potentially be fixed by adding another y motor.  But the fact that the machine is not consistently out of square (square in front right and very not square in back left) makes me think that perhaps something else is going on here.  I would consider investing the energy and money into a second y motor if I was convinced this would address the issue, but I am not sure.  The cable system has never given me the warm and fuzzies though..


Aciera
Interesting, I am not sure why 'Method of least squares' is the best system here, are you using this method to take an average of the different joint positions?

Forgive my lack of computer prowess, but how do I run this script?  I tried running it through VS code on my mac and got a 'zsh' error.  I can try running it on the LCNC machine in a moment.  Will it prompt me for the data and then I feed it in? 

I will be happy to try it, thank you. 

EDIT: by tommy I mean rodw

Tommy
See edit  

Aciera
Okay that makes sense to get the highest accuracy over a range.  I will try that script after lunch. 

My machine was getting aluminum swarf in the pulley reduction for the 5 axis head, so I have had to cook up some 3D printed dust covers, since there was no way this machine would be running without your help I thought some adornment was in order.  

Aciera
Okay had to use python3 to run the script but seemed to work.  So I need to modify the attached text file with the data (I am assuming more data points is more better?) with the form Aangle, Bangle, Zheight? 

So, with z height, should I have a programmed 'z pivot' when I am taking my measurements, aka my best guess? If not the z delta is going to be very large. 

 

Aciera
Sorry about the delay, another project needed my attention (and honestly my brain needed a break from the Hackbot). 

I tried out your software and I got some, results...

At first things seemed very good as the y pivot was within 4 tenths of what I measured initially. For the range that I measured (+/-20 degrees) the results seemed pretty good.  For the most part staying at 0 change in height, but in a few places moving by as much as 5 thou (which is less good but not insanely bad).  Btw, does the sigma value give the codes confidence level?

As I pushed outside of that range things got weird, in some cases moving in z by over 20 thou (very not good).  The further away the worse things got. This part could well be contributed to a pivot length error.

I also observed a very strange behavior that I am trying to determine what is causing it. When I am tilted way over (like 40ish degrees) and then I change direction to head back to 0, the z height jumps WAY up.  Like probably 60-100 thou.  When the head is kicked over like that more torque is needed to move the machine.  I do not think that the rotary axis is missing steps because z height seems to return to 0 when I return the rotary axis to 0.  But something strange is happening.  I think maybe the backlash is increasing under heavier load. This would cause the rotary axis to remain stationary as the z starts to climb (more z and less x or y in TCP when the head is well tipped over) to deal with where the machine thinks the rotary axis is.  Back lash is the only thing I can think of that would cause a direction change error like this, but if anyone can think of something else, please let me know. Is there a way that I could modify backlash control based on rotary position? It seems like something like this may be possible in INI. 


My plan:
My z axis needs to be replaced, I have bad bearings in the ball screw of z.  I am going to re-deck the surface of my fixture plate (the machine shifted when I moved recently) then I am going to machine up the parts for the replacement of the Z.  Then I will replace the z axis.  This will eliminate the z axis error which will help. This may take a couple weeks. 

Then I am going to spend some time dialing in backlash as best I can and making sure the machine is perfectly trammed. 

Then I will go to work with indicators as well as your python script Aciera and see what I can get to work..

Again let me know if you guys have any suggestions.