As I was mulling the problems over in my head (mechanical systems are complex, have parts that can break, etc), I noticed one of the plastic buckets lying in the box of supplies. It occurred to me that if I cut a few holes in it and dump the water into the bucket, it will push some of the sediment out too. One problem with the other AutoSed designs is they required dry sediment. This combines problems related to the mess of loose, dry sediment getting all over the place (into gears, etc) with problems of drying the sediment for re-use. You can just scoop the sediment from the drain catch and plop it right back into the AutoSed bukkit.
Imagine my surprise when not only did it supply sediment, but it didn't let all the sediment out at once and all it flushed all of the sediment from inside the bucket - it didn't leave big blobs stuck to the sides. The amount of sediment supplied was proportional to the discharge from the pump, so fewer issues when you quickly drop base level and the pump starts pushing out more water as the hydraulic head increases. Filling the one quart bucket requires less attention than dumping individual scoops by hand. Cheap, no moving parts, requires minor modifications, and is easy to use. Perfect. Exactly the kind of thing I'm looking for with my Emriver hacks.
Notice the lovely concave up profile
After dropping base level - notice the terraces.
AutoSed 3.0 from Matt Kuchta on Vimeo.
And then there are the time lapse videos. They look lovely. Under longwave UV, it looks surreal. There are a few fluorescent bits in the color-coded sediment, so you can track particle movement. I added some fluorescent yellow dye to the water so it shows up. Allogenic controls on stream response never looked better.
Experimental Fluvial Geomorphology from Matt Kuchta on Vimeo.
A little experimenting with the Em2 stream table. Now with a simpler and more effective automated sediment delivery system.