Commitment
It's official. I have now invested actual money on this business. I've ordered a batch of parts that I can use to assemble my prototypes.
I spent most of the day optimising and testing my design. Switchmode power supply design is hard, especially if you have space and cost restrictions. Moreso, because the 'one part in the world that can do what I want' is pretty marginal for that task. There's been some trimming of requirements, but I don't think they're unreasonable. The only loser is 3-4V operation (single lithium cell), where you might not quite get the full 1.5A output. The efficiency also blows. Running a 6W LED off a single lithium cell isn't exactly the sanest thing to do if you're expecting power efficiency.
The Luxeon K2's don't actually get much brighter once you jump from 1A to 1.5A current. 1A operation is perfect across the entire voltage range. In general, the K2 range is specced to be brighter at the same current than the Luxeon III's, so even running at 1A should be impressively bright. (It also means that you can mod your old LED lights to use Luxeon K2's and gain some brightness).
I'm tempted to open up the Buckpuck that I have here and see how they solved the problems I encountered today. My suspicion is that generally people don't, because the parts to do exactly what I wan't simply don't exist yet. All is well in theory-land - building an actual device is far more difficult!
I was also surprised at how tedious the whole purchasing process was. I might be looking at my schematic or BOM thinking "ok, I need a plain old 10k resistor here", but there are hundreds of parts that fit that description. Thick film, thin film, power ratings, case styles, RoHS messiness - it's all so much effort for something that I don't normally think about!
On one hand, I have new-found respect for people who do this for a living, but more because of their tolerance for pain. I can imagine a piece of software that will take your criteria ('resistor', '10k', 'surface mount') and pick the best (cheapest) option based on that. Problem is, every vendor has a different format for shuffling this data around. At the moment, I'm getting my stuff from Farnell, but they're renowned for being expensive in Australia. Still, my first two prototypes won't cost me that much. One part takes half of the cost, and it costs 25% as much if I order from overseas. There are definitely cost reductions to be had once I'm building larger quantities.
RoHS (Reduction of Hazardous Substances - or the 'lead-free' initiative) is a hassle. At the moment, I'm not going to deal with it apart from checking if I can export to Europe anyway given the small quantities I'll be moving. If I can't do it, no biggie. My geographic targets are the United States and Australia. Most (if not all) of the parts that I've specced for my prototype are in fact RoHS compliant; I just won't be assembling them using lead-free techniques. That'll come down the track when (if) I'm outsourcing the production to someone who knows what they're doing. They can deal with lead-free soldering fun.
PCB design is next. Since I'm a cheapass and am using free PCB layout software at the moment, I have to define all of my own parts libraries. Sigh.
After spending all of yesterday dealing with food poisoning, and all of today staring at LTSpice and spreadsheets, I need to get out of here. I'm thinking Mexican food might be nice...