Time to fess up....
I guess it's time to let ya in on a bit of a secret, I've taught this good stuff (part time) for more years than I care to confess to.....
Al's circuit is here on CPF, somewhere, and while I'm not at all good at such stuff, I'll try to find it. Perhaps a search on "LM 334"?. Anyway it's taken directly off the spec sheet for the device (always a good place to go, the maker wants engineers to use his parts, he makes it easy......). You can see the sheet on the Jameco web site (highly recommended for mail order parts, BTW) at: <
http://www.jameco.com/Jameco/Products/ProdDS/23739.pdf> . Go to page 8, the "typical applications" section (always interesting IMO) and check out the 'higher output current' example. 2.2 ohms for R(set), 1,000 for R1 and .1 micro farad for C1 have worked super for 30 mA drive on 'Supor White 5 mm LEDs', The LED goes instead of the line connecting C1 and R(set) (in series with the transistor collector). I've been changing the transistor to a slightly higher capacity part (2N4918, 19 or 20 also from Jameco for under half a buck) and a hand wound R(set) made of about 2 feet of 30 AWG wire wrap wire for Luxeon LEDs. You can start long and 'cut and try' it until you get the current you want. For margin I tend to set it at 320 or so mA,
The LM 334 comes as a 'TO-92' case part, a small 3 lead transistor package, the same as the 2N3906, making the entire circuit easily 'sugar cube size' or smaller. Total parts cost is under $2, exclusive of LED and battery.
Paying attention to the possible need to add a heatsink to the transistor, cutting the R(set) wire in half and raising the cell pack to 6 (nominal 7.2 Volts) or perhaps even 7 depending and you're 'good to go' with 5 Watt Stars if they ever get here (come on Wayne!).
The neat thing about this circuit is the transistor makes up sort of a 'smart resistor' that changes it's effective resistance as needed (according to the IC regulator, LM 334) to maintain the proper current in the LED no 'matter what the battery voltage'. In reality, too high a voltage is a waste and only heats the transistor, and the circuit 'runs out of gas' when the battery gets very close to the voltage the LED needs (properly done this is less than .1 Volt). NiMH cells are a great fit here. Since at reasonable current draw then can easily stay above 1.15 volts until the bitter end, this means we can support a 3.35 Vf LED at almost 100% efficiency. At this point, a higher Vf diode will still get almost full current, probably not possible to detect the difference by eye.
Works a treat as the Brits would say, give it a lash.....
Cheers.
Doug Owen