No schematic right now, but if you were really interested I could get one, or at least walk you through it. It's a very flexible chip, and the reference designs in the datasheet (plus the application note they publish) are pretty good.
I'm using it in boost topology, since that gets a few more % efficiency and works better with some of the other components I'm using. The one problem with boost is that the "full off" state from the PWM signal translates to the base voltage of the power supply going through the LEDs, so the dimming range is chopped off at the bottom end - you can't turn them down any farther than about 10% brightness. I have it driving 8 HB LEDs from a 24V DC supply, so it's better than some of the commercial solutions. Lotsa reef people are using buckpucks, which will drive 6 LEDs each from a 24v supply (in buck mode, of course) and cost $15 - $20. This driver is costing me about $6 in parts and drives more LEDs.
I'm right with you, the lighting is one of the main reasons I'm doing a controller. I have the arduino fading three separate banks of LEDs up and down to simulate sunrise and sunset (plus the sun moving across the sky, to an extent, since the banks are oriented as east, center, and west.) Right now, I'm working on coding a "storm simulator" such that every few days, the lights dim for a random period of time, to imitate cloud cover. Next up is a moonlight that will simulate a 20-day lunar cycle.
All that said, I'm trying to figure out if paralleling output pins from a TLC5940, and using them to drive an external transistor, would allow me to drive HB LEDs directly from the TLC. I was going to use one as a PWM port expander to let me control more channels of LEDs, but if it can drive the LEDs without the NCP3066 circuits, that's even better.