The catch with any "modern" preference voting systems is that none of them are "strategyproof", meaning two things. (1) There will always be cases where a voter has an incentive to state something other than their true preferences. (2) There will always be obscure corner cases where everybody can look at the numbers and say that, obviously, Candidate A beat B and C, but in fact, C won. This is the crux of Arrow's Impossibility Theorem, for which he got a Nobel prize in economics.

electionmethods.org is a reasonable site, although it's clearly biased in favor of Condorcet elections. The wikipedia has a pretty good page on Condorcet election methods and voting systems in general. I actually have a soft spot for approval voting. Voters give up the precision of being able to rank candidates, but they get something much easier to understand. You can cast one or zero votes for each candidate. All of those votes are added up, and whoever gets the most, wins. One of the interesting side benefits of approval voting is that it works right away with traditional voting technologies. You just redefine overvotes as legal and now your old punchcards or optical scan system works just fine.

One of the most intriguing possibilities that I heard of recently is called "open primaries", as in California's Proposition 62, which apparently failed in the California election. The general idea is that the two candidates for November 2 will be decided in one, wild swinging open primary election with multiple candidates. In effect, the primary becomes an open election and Election Day becomes a runoff election. The only way you could make it work is with some kind of preference system (as above), which wasn't part of California's prop 62, as far as I know. Still, the idea of a multiround preference voting system might be a way to work around some of the undesirable features of single-round preference voting, but still end up with something that voters can wrap their brains around without causing too much pain.