In case you’re curious what it looks like after you drain the liquid out:

Fun, very fun.

Here’s the Sony controller. 6 boards (4 computing, 2 amplifiers), about 12″ square each.

And here’s the Galil 2143 that replaces 2 of those boards along with the terminal blocks that break out all the feedback from the arm. Smaller.

The four isolated power supplies, 2 for each arm. These I reused. ~105V for θ1, ~70V for the other 3 axes.

The other board I reused:

One inductor for each motor, interface for the Z axis motor brake (found that out the hard way), relay and power resistors to slow the arm after an estop. Upon an estop, the relay clicks and the motor windings for θ1 and θ2 are separately shorted through the resistors. I removed these components. This thing isn’t that dangerous.

It’s getting there. Just need someone smart to look into these encoder noise issues.

I had a busy spell that lasted most of the winter, and spring, and some of the summer. This blog turned out to be the lowest priority. Maybe I can be more consistent from now on.

Lots to mention, I’ll try to fill things in.

I sold my beautiful German Alzmetall drill press last month, sort of sad.

Despite it’s power feeding, foot pedal reversing, 1400lb geared head badassery, I really didn’t have a use for it. I never did manage to find a cheap 480V transformer to run it.

Due to the generosity of a friend I got this:

Yes, Its a Sony SRX-340 two arm SCARA robot. Complete, with controller, though due to it’s 15 year age it was essentially impossible to get the controller to talk to a computer. Brushed motors on all axes, harmonic drives on θ1 and θ2, ballscrew and ball spline on Z. Mechanically in very good shape. I trashed the Sony hardware and threw a controller together for one arm using the original power supply, a 4 axis Galil motion controller and four servo amps. It was actually pretty straightforward to identify all the (single ended) encoder lines, the limit switches and the motor wires and get things mostly running. The servo amps I’m using might be a bit too noisy though, I’m getting enough ~18kHz nise to screw up my encoder counts on the axes with the longest wires.

Still some possible solutions to explore, it will be running correctly before long.

Anybody know of a fun use for a SCARA robot?

This actuator can output 2.5Nm of continuous torque with an angular resolution of .0036 degrees. Accuracy is around .03 degrees. All of the control electronics are integrated into the unit, only 4 wires are needed for power and communication. When more than one unit is used they can be daisy chained together and motions can be coordinated.

Each unit can be configured to rotate continuously or through a fixed range. In the fixed configuration the axis homes itself upon powerup and runs an absolute position controller.

If I cared more about marketing I’d try to find some other uses for this. It’s pretty specialized and the high accuracy and large load capacity of the unit make it expensive.

Anybody need a compact, precise 1 or 2 axis pointing system?

Garrolite ball from McMaster, some allthread, some aluminum and one screw. The plates are surface milled to match the profile of the ball so it’s easy to adjust for the proper amount of friction. A nice, small effort with a good result.

I did this back in June. Thanks to an auction find by my friend Tom I have a -120C deep freezer in my shop (along with a 13′x21′ rigid wall cleanroom). I plugged it in when I got it to make sure it worked and didn’t touch it for 16 months afterward.

Fast forward to June. While trying to come up with a gimmick to improve the karaoke parties my roommate and I throw periodically I thought of freezing liquor into cubes. Imagine: a gin and tonic which starts as a glass of gin cubes filled in with tonic! Your drink gets stronger rather than weaker when the ice melts!

Well, it sort of works. The cubes never actually get completely solid, they’re more like a firm slush. Interestingly, different 80 proof liquors freeze completely differently. Some crystal structure issue maybe? Gin was the worst, the cubes would melt in about 2 minutes.

My plan of freezing cubes in the shop and hauling them home for a party was scrapped after this. They’d start melting immediately after I pulled them from the freezer. My brilliant idea of white russian popsicles was also somewhat of a failure. They’d freeze solid alright but your tongue would stick to it if you tried to lick one.

This experiment will happen again, there’s too much potential here.

I made up a messy little board to sum the signals coming from the sensors and to interface with the motion controller. This is the second incarnation. I thought I could do it with one chip until I realized that I couldn’t do it with one chip. Oops. There’s something satisfying about soldering like this so I didn’t mind doing another one.

This board works perfectly. The complete robot has been tested for many hours now with no failures. Things look good!

It went for much less than I’d expected, I’m pretty pleased. Came with 2 faceplates, 3 chucks and a decent set of collets.

These lathes are some of the nicest ever made. The company still produces a few of this model a year at a cost of over $60K each.

I also bought 3000lbs worth of Vidmar cabinets, I couldn’t resist. I got a lot of 7 for the cost of 1 new one. I’ll keep 3 or 4 and eBay the rest.

I’ll head down with a 20′ truck in early November to grab everything.

The water bottles simulate the weight of the piece the robot will be carrying. I ran the robot in circles to get a sense of how it behaves and also how long the batteries are likely to last. I did about 4 hours total which wore a nice groove into the plywood. I had 2 friends sitting on it at once while it was circling, no issues. This test proves that the motors and amps are capable of the speed and torque needed for the robot. The gallery has polished concrete floors which will be much easier to drive over then the crappy plywood over uneven floor that I’ve been running on.

I’ve decided to ditch the inexpensive analog sensors I’d chosen to detect obstacles. They’re much too noisy in the range I need.  I’m going with a similar sensor but with a digital output that switches at a fixed range. This change has the benefit of making the programming a bit easier.

This pic was taken before they were polished, they came out great. The form was surfaced with a .5″ ball end mill taking very close passes to minimize the scallop. The program was about 10,000 lines long, lots and lots of tiny moves. Only took 5 minutes to run at 100IPM. It only took a few minutes of sanding with crocus cloth and a few more on the buffing wheel to get a near-mirror finish.

These will be used to make the silicone mold in which Ben will cast the actual lipsticks for the prototype out of red urethane. I’ll do some secondary machining on the urethanes after they’re cast to add the precise features needed for them to interface with the rest of the installation.

The form is taken from a NARS lipstick, actual size.