mouldy.org

I've been able to use LTSpice to verify that one of my two chip choices cannot do what I want. So good work there for Linear Technology, I suppose. I know not to buy one of their parts!

It's slowly becoming apparent why people don't just bang these designs together overnight. At the worst-case design parameters that I have in mind - three LEDs, 1.5A current, 3.5V li-ion cell - the current that the circuit tries to draw is absolutely massive. This isn't surprising; three LEDs is about 12V, giving 18W output power. From a 3.5V lithium cell, that's over 5 amps (even at 100% efficiency).

For more sensible circuits, my design is doing just fine, however. I can crank 1.5A through a single LED from a single lithium cell without any real trouble. The obvious application for that is tiny high-powered flashlights, and it makes a great demo. It remains to be seen whether the circuit actually works in reality (pfft!) but I can be reasonably confident that it will.

I still find it funny to watch the imaginary battery in my circuit try to crank out 8 amps of virtual current. "Give me more!" the regulator says, and the battery says "Nooo! I can't! I'll dieeeeee!".

Maybe I should get out more.  

From a couple of hour's fiddling, I now know that the chip I've chosen can do the job in any realistic scenario. Which is good, because there's nothing better right now.

Tomorrow, optimisation work begins. I need to keep costs low, output power quality high, space small, efficiency high, and do it all with parts that exist in real life. 

And all of this before I even see any silicon!