The file (posted in ABSE) was posted here recently, I don't know who posted it. I built it, but changed the Q2 to a BC338, and I added a 47 uF across the supply lines. The problem I have is that it's balky, sometimes it doesn't want to start up, and sometimes it will. It usually won't start if I turn on the PS, but if the PS is on and I hook it up, then it'll start. I think it needs to have a faster risetime for the supply voltage, if it's slow, then it won't start. I thought that if someone had this modeled in their spice prog, then they could try slowly ramping up the supply voltage and see what causes it to be so snotty.
I compared it to Stepan's boost circuit, and I find that it's a lot simpler and going by the column on the right, it's a very efficient circuit. But somehow I doubt that it could be that efficient, 75 to 80%. It regulates the LED current by robbing the output transistor of its base bias. That seems to me to be wasteful. I like Stepan's circuit a lot better, the one on the right of URL
But I think both circuits waste power across the current sensing resistor in series with the LED. And Stepan's circuit wastes more power across the schottky rectifier; it may be better to just run the LED directly off the choke, and let it do the rectifying - after all, it _is_ a diode.
I remember a circuit that was better at sensing the current, and used a much lower voltage and a lower resistor, 10 or 6.8 ohms. I'll have to come up with that circuit.