That's why I want to model the plasma. The whole system is way too complicated. It's possible to think about one section at a time, but the whole system interacts with every part of itself continuously and it's not so easy to figure out what the effects of each subsection of the problem will be. It will be interesting to see how the models behave depending on the assumptions made.
I finally found a clue about what Bussard was talking about in one of his earliest papers (PDF). On page 6 he shows what he means by a truncated cube, and that indeed would provide the geometry he specified.
Excellent find! The remarks on the polyhedral currents come from a French reference from 1966 (ref 14). Completely different thought process than where I was coming from, so this helps a lot.
In the "Should Google Go Nuclear" pdf at askmar.com there's a picture of some single-turn devices emc2 used to test the polyhedron concept, look on page 14.
I think it was also shown in the video.
They are also mentioned in the IAC paper on page 10, but there isn't a picture.