LinuxCNC Pick and Place

Hi all,

My last project was a refurbished CNC controlled gantry. It was my introduction to using LinuxCNC and Mesa7i76e controller. Now fresh out of college I got a job at an electronics fab company. I am not appeased by our current prototype line, or proprietary machinery in general. I have been doing a literature dive into what it would take to make an open source line. Using LinuxCNC and MESA controllers paired with OpenPNP, but the support just isn't there from what I am seeing, so I came here to find out if someone is hiding away some secrets on how they have used LinuxCNC for pick and place.

pich and place systems most work on optical placmentation
this is not what regular CNC do
so Best woudt be to get the OpenPNP system to the mashining

and do Regular cnc stuff like PCB prototyping with the 7i76e

i agree with Andy
as Open PAP is the one that is made for the Job and always updated
here you may need to get alot of work your own without the Special help of people that are in this Business

been working on a tiny DIY PnP machine for a while. planned to use OpenPNP, but haven't managed to do it so far. so the machine is currently under LinuxCNC control with a bunch of g-code subroutines and a python script for vision. wanted to have complete control over motion to exploit LinuxCNC's trajectory blending - and that's seems to be not that easy with generic OpenPNP machine driver model.

and yes, i'm biased toward LinuxCNC because i'm familiar with it and a few other reasons which have nothing to do with this particular application. so probably i gonna stay with the LinuxCNC...

andypugh wrote:

LinuxCNC has pluggable "tasks" but only "milltask" currently exists.

I wonder if pickandplacetask would be an idea worth pursuing? I am already thinking that a plasma task might be sensible.

I'm still half interested in a plasma task but Phill has done such a good job with plasmac, I suspect there are only a couple of plasma "tasks" that could benefit. I think it would also need a pluggable plasma io module as well. One day I might be able to retire and play with stuff like this....

i have no idea what part of the functionality is handled by the milltask. task.hh looks pretty generic...

a few thoughts...

i'd like to use TnM6 for nozzle change but having two "spindles" and two selected "tools" doesn't seem to be possible. not sure if there're examples of such configuration and how it can be handled in the g-code. so instead i'm using the tool table for selecting head offsets: T0 for the top camera, T1 for the left head and T2 for the right.

would be nice to have coordinate systems defined by an affine transform. could use that for reversing z-axis direction when selecting heads and correcting machine geometry.

not using rotary axes for nozzle rotation as it breaks trajectory blending - can't live with that. so i opted for limit3 comp instead.

currently running LinuxCNC under debian Buster as ancient OpenCV dist in Stretch constantly crashed when handling USB cams. still had to patch it to disable v4l buffering which gave unacceptable latency in video. there're OpenCV 4.0 packages in debian testing but python libs packages seem to be missing yet.. not sure if i can port them myself, so took the easy way for now.

one thing which may qualify for an "advanced configuration": played a bit with the "torque limit" function on z servo drive. tied the "in torque limit" output of the drive to the motion.probe-input and used that for z-probing. worked surprisingly well, considering a timing belt drive of the head...