TL;WR: http://www.github.com/marginallyclever/arm3

I know the total steps and the maximum velocity I want to reach.  I can find how many steps I should accelerate and how many I should decelerate.  So I think I spent 2h writing the code to do exactly that and 2h tweaking the acceleration and top speed to minimize overshoot.  How good is it?  The overshoot is almost completely gone.  I'm picking up a bit of vibration from the motors and I wonder now if gear backlash is becoming the dominant issue.

The GcodeSender app has been improved to create "your message here" gcode for the arm.  I'm finding that lines are very straight, but the pen holder is now the weakest link - the grip on the pen is loose and the pen wobbles in the holder.  I tried a rubber band to grip more tightly and that created a new set of problems.

Philisophical question: how do you know the difference between a problem and a symptom?  Can't a problem be a symptom of a deeper, more subtle problem?  I need a loose grip on the pen because the arm's distance from the table isn't perfect.  I also notice that long horizontal lines on paper are bowed.  Perhaps I'm getting the machine home position wrong.    If it were perfect the arm would move in straight lines with constant Z (if desired).  that would mean I could put a tighter grip and reduce the play in the system.

Since I'm at maker faire I think I'll wander over to the TechShop booth.  They have a laser cutter.  It shouldn't be any trouble for them to cut a better fitting pen holder as an experiment.  Someone around here has got to have some glue.

You'd be surprised how hard it is to get makers to make something at a maker faire.  Wierd, huh?