Just to add a few points, the reason that Japanese students can't hear the difference is because to them there is no difference. All the folks in Japan probably know this very well by now, but I just wanted to offer the actual linguistic terminology for it. For them, r and l are allophones, meaning that no matter which sound you choose to use, they would always be perceived as the same sound. People have conducted really fascinating tests with these sounds, and I wish I could direct you to those results.
If anyone's still in search of a diagram, the thing to look up is the vocal tract or vocal tract diagram. There's actually an interactive one here:
http://www.chass.utoronto.ca/~danhal...ics/sammy.html
l's and r's are hard to describe because they're consonants that are very nearly vowels. If you think it would help, the way we describe [l] is as a lateral glide. That means, if you take a breath with your tongue in the [l] position, you should feel air all along both sides of it. [r] is a retroflex, meaning that the tongue is slightly curled backwards or around the sides, and if you breathe in, you'll feel air over the tip of the tongue as well (because it's not exactly touching the roof of your mouth).
I'm thinking the most perfect Japanese r/l sound is somewhere between the two that we know, so taking the sounds to extremes and exaggerating (onomotapoeia is a good call, with the "growling" idea for [r]) might be a good place to start. Or just "eeeeeeeeeeeerrrr" vs. "eeeellllllll", "are" "all" and other single syllable minimal pairs will also get them started with hearing the difference. Someone posted about the minimal pairs elsewhere, but it's really important to just write the sounds where they can see it, make them, and them have them select the correct ones as you go.
Ok, that may or may not have been helpful, but I just wanted to throw it out there!