Third Field Test, Ongoing (Success)

The feeder's robustness is markedly improved.  I made several changes before putting it out for its latest field test, and it's run for several days without trouble.

Trophy for a Successful Field Test

I added a piece of thick cardboard inside the bucket to direct the feed towards the hopper.  This was a fun reminder that process often trumps planning.
 I started by tracing the top of the bucket onto the cardboard, then constructed the center of the circle, traced an ellipse with minor radius equal to the circle's radius using pins to mark the two focii, then cut out that ellipse.  While the geometric constructions are fun, this produced an ellipse that didn't even fit in the bucket, because I hadn't accounted for the taper of the bucket as it reaches the bottom, and my cut through the thick cardboard with a wide blade was imprecise.  I was able to quickly refine the shape (and could have refined even a simple rectangle just as quickly) by testing for fit and then cutting off the high points.  This produced a snugly fitting shape that matched the actual bucket, rather than my idealized model of the bucket.  I also enlarged the slot that allow feed from the bucket to enter the hopper.

Internal baffle, cut to fit.

To help the feed flow smoothly into the hopper I propped up the back of the bucket.  This tiled the solar cell away from the sun, so I move it to the lid.  I measured 25ma of charge current into the battery under direct sunlight at 6.12 volts in this new configuration.

New solar cell placement on the lid.

Wire routing across hinge

Finally, I cut a new door out of the lid of a trashed scanner/printer, which will have better dimensional stability than the old wooden lid.  I retained the section of the old door that connected to the crank arm, since it had proven to be a reliable mechanism.

Along with some general tidying of wires and permanent attachment of the battery to the hopper, I think we're in pretty good shape right now.  The only remaining concern is the limit switch, which needed to have some grit cleaned out to allow the rocker arm its full range of motion.  If that happens again, I may need to replace it with a non-mechanical sensor, like a reed switch or hall effect sensor.