stuff going on

[gfish]

I finally got a wire-feed (flux/MIG) welder last week, and a sharp looking (if kind of cheap) rolling cart to keep it all on. It's... just amazing. Normally when you buy a new tool, it gives you a new capability, and that's cool. But sometimes you get the magical experience of a new tool changing how you do everything. This is one of those times. Being able to weld cheaper/faster/better means I'm already turning to welding as a solution for all kinds of things I never would have bothered with when using my old oxy/accetylene set.

Other than cleaning up the workroom and shop last night, my focus continues to be on incrementally working towards the grand hexapod project. I have a working hydraulic testbed now, with a tiny little electric pump driving a small cylinder through a real directional control valve. I just finished mounting the cylinder to a pivoting arm, to give me a complete mockup of a single (1 of 18!) hexapod joint. I have the electronics working to control the valve from an Arduino, a frustrating but very educational process. Next: attaching a position encoder on the arm and implementing precision control. I might make a series of videos showing the relative performance of hysterisis, proportional, derivative and integral control loops, that would be neat. Assuming I can get that working, it's time to start working on a full-size leg. Which means a larger electric pump, amongst other things. That will be the final sanity check before getting absolutely stupid about this project. The point of no return. Whee!

Comments

[nojay.livejournal.com]

Sadly what you're describing isn't as easy as it looks. I was involved in research to do this sort of thing about 25 years ago and the results weren't pretty. We were using pneumatic actuators and valves coupled with microprocessor controllers but the problems with hydraulics are the same, pretty much.

Positioning using a hydraulic cylinder and a proportional valve gets messy when you consider inertia, mass flow through the valve and into the cylinder, the differential characteristics of the two cylinder piston faces (the rod face has less active area than the non-rod face hence the same pressure on both sides of the piston causes differential force and, absent a braking system will cause creep when the system is supposedly static) and a whole load of other bugger factors.

We got it to work, sort of, eventually developing a cheap industrial robotic arm but precision was not its strong point and hunting and overshoot were a fact of life. It was not a commercial success.

If you are seeking rotary motion at the joints rather than linear extension/retraction you might want to consider using vane-type rotary actuators rather than cylinders. They do have the advantage that the vanes have the same pressure/force on each side as the rotating shaft comes out the side of the vane chamber, not through one end of a cylinder. They also simplify the actuation at the joint. Rotary encoders are also cheaper and more robust than linear ones; we got decent results from using wirewound potentiometers to measure angular movement, building a resistance/angle lookup table in the MPU controller's memory for each joint but we ended up with 8-bit Gray encoders in the production version for engineering reasons.

[randomdreams.livejournal.com]

I felt much the same way when I got my welder. Suddenly you just want to stick *everything* together. I used to spend a lot of time casting and fabricating. Now I weld everything. It's glorious.