Puma 6 axis Robotic Arm with Mesa 7i80/7i40

I have been using LinuxCNC for a long time for my Mill and Lathe setup and parallel controlled stepper motors but this new project is making me a little crazy (it was straight forward on my other machines). 

I was given (in an incomplete state) a Kawasaki Robotics/PUMA 200 arm which has a Mesa 7i80 controller with 7i40 daughter cards running the servo motors.  

CORRECTION:   I believe this is more appropriately identified as a PUMA 260 arm - if that makes a difference

I have things "moving" generally with INI and HAL files from the previous owner and of the axes respond generally to control (I don't know enough about servos with LinuxCNC/mesa  but I am having all kinds of issues with Joint #1 the "shoulder" joint (the rotation, waist pillar joint in the center is Joint 0).  With a higher P value (200-300ish) the motors will hold it in position and move it up (positive direction) but when I reverse direction it stutters, pops up a follow error and estops.  I can very slowly/carefully/intermittently move the joint down but it eventually errors out.   I have a feeling that there needs to be some kind of asymmetric control of the positive/negative direction for this joint due to gravity/weight of the mechanism. 

I'm not even sure where to start here (I've been combing through some of the documentation leveraging my previous knowledge of LinuxCNC - which is a little rusty to begin with) or what questions to ask.   

Does anyone have a similar setup who would be willing to share their configuration, or maybe help me formulate some intelligent questions and find the answers. 

Thank you, 

Michael Jones

so I'm not familiar with PUMA 200, but most robots have brakes, i have had all kinds of trouble with them. I would look at them and any counterbalance system. if it has a counterbalance and the brake is on often you can get it to move up but not back down.

 
 

The system does have brakes which I have disabled (the joint free drops when it gets the follow error)

it doesn’t have a physical counterbalance that I can tell, maybe the gearbox has some sort of clutch?  

When idle, the motors are able to hold the joint in place even if I push on it.

I have a feeling that there’s something off  between the motors and encoders.   When it moves, it seems to be miscalculating the actual vs expected travel - but that’s just gut feeling.   It sometimes errors moving up, but not always.. it always errors moving down.  

Here's the files.   I'll work on verifying the encoders shortly.  

One thing you might try is setting the PID MAX_OUTPUT to say 9.5 (95% of full scale)
As the 7I40 cannot have 100% duty cycle (the high side MOSFET bootstrap
requires a continuous PWM signal)

Good point.. I've set all of the axes to 9.5 but it didn't resolve my issue.

 

I attached a digital angle finder to the axes, moved it a specific amount (45 degrees), calculated how many encoder steps were read, divided by the number of degrees and entered that number into the joint config in the ini under encoder scale.   The numbers I entered were only a couple encoder ticks off from what was already there.

Is there some other spot to enter the encoder resolution/scale/ppr?

I am unable to find any documentation whatsoever on the specs for the encoders: Tamagawa TS1902N117 or TS3083

I am also considering swapping 2 of the joint connections to see if the issue follows the daughtercard.
 

Do you get overcurrents on the 7I40 (red LEDs blinking)?
Also are the 7I40 current limits set to the high range?

Good thought.  I'll check tonight - I sort of made an assumption that the previous owner checked that, but that's probably not a safe assumption.

I checked the jumpers, they were all set to Low (default)  I set the jumper to high and it didn’t make any difference.  

There are no red lights, just green and amber.  

I was told by the previous owner that half of two of the 7i40s don’t work.  This joint is on the working half of one if those cards.  I’m wondering if there might be more wrong with that card than I originally thought.   

The next test I guess is to swap two joints.