Do be aware of the efficiency issue. At 12v source, with or without the 5v reg, your efficiency is 30%. This is of course a significant amount of heat as well. Any linear solution will have this problem. The HV9910 is around 95% efficient provided you choose a decent inductor.

If you're buying devices for area light rather than a flashlight, I would urge you to look up Lamina ceramics devices. Their BL2000 have a far, far more efficient heat strategy. It's a paper-thin wafer you attach with thermal epoxy to a flat surface of any thick chunk of aluminum. They're around $13 for colors, $18 for white. Laminas are much more powerful than Luxeons. Their light comes from a number of emitter chips on the wafer and is thus more difficult to focus, so it generally does not make a great flashlight.
http://www.laminaceramics.com/docs/BL_2_White_43.pdf

Bitchin cool- it hardly stands out from the heatsink at all, but makes LOTS of light. Luxeon emitters can't really be mounted straight on heatsinks because they didn't electrically isolate the thermal backing. The Luxeon Star doesn't solve that problem either.

That brings up a good point. Do you have the star's board completely isolated from power supply's ground? If not you have a major problem.

Hiya Oznog,
Those Lamina led's look awesum and I did notice there isn't a dealer in Australia which means I'll have to get them from the US. So postage could be the big killer but i'll e-mail mouser first as I don't want to get stung with $110 US in postage again ( that happened with digikey for 10-16 pin chips :shock: ). As most of my friends live off the grid I'm sure they would love a few for themselves so when I can afford it I'll get a few over and set them up. Oh and the LED is totally isolated from the box, I made sure it was before I glued it on the heatsink.
Thanks for the link too mate

Cheers Bryan

P.S. Eh WilliB, I Lathe tested the genny I worked out around 250 watts output but in real life it will be different.

Originally Posted by williB

what current does it draw..?

Hee, hee. :lol: I think Bryan is going to use the motor as a windmill's generator, not as a wind-making machine. :lol:
Can you imagine trying to make a wind in the opposite direction to try and cancel the real wind?
President Bush should get NASA to try and cancel those darn hurricanes.
I hope they don't begin using nukes to get rid of hurricanes. Then bye, bye Cuba and Jamaica (and maybe also Florida?)

Originally Posted by Optikon

where can I get a Mickey Mouse diode?

Disneyland. :lol:
Mickey Mouse diodes at the output of a voltage regulator ruin its superb regulation.
Mickey Mouse diodes are also used with a pullup or pull-down resistor as OR gates for Cmos inverters.

How do I insert this image?

Hiya Guy's,
Well I got my sample HV9910 chips from supertex today and they ended up coming from a local supplier Anyway below is a schematic for a high brightness LED, the output is 4.5 volts @ 900mA. Now I've just spent a couple of hours studying the datasheet and trying to workout how to get the output down to 3.6 volts @850mA. I'm going to use a 12 volt 7 amp/hour battery for the power supply. But I'm a bit stumped on how to work it out. If anyone can help me out on this I'll apprecieate it.

Here is a link to the datasheet http://www.supertex.com/pdf/datasheets/HV9910.pdf

Cheers Bryan

What part confuses you?
You need a 0.277 ohm resistor for the feedback. A huge variety of SMD low ohm resistors are available and are easy to solder onto the backside of prototyping board. Leaded ones have far less selection available.
You can use a lower ohm resistor (higher is not possible) if you go with the external voltage reference solution.
SOIC-8 transistors are a really good idea for this. Little heat and they have very low rds for efficient switching.

Diode should be Schottky type. Rating it for at least the 900mA output current is good. There are very common 3 amp ones and I'd recommend those.

That schematic you showed is an example which rarely works smoothly in the real world. You need a RC filter between CS and R10 with an RC time constant of 300nS. The surge (which can be large) of current turning on the gate also travels down the source and makes a momentarily higher voltage on R10, which may result in resetting the output latch. The problem becomes more likely when you use larger inductances which make a nice, clean, low ripple output. Low ripple means that the inductor may only be at 875mA when it switches the transistor on- gate takes 100mA from the HV9910, R10 reads like it's 975mA, way over the theshold, and shuts off the drive immediately.

BTW- since I've already been through this- these are great inductors for this job (and cheap!):
http://www.bourns.com/pdfs/sdr1806.pdf
101K is probably the best choice for 900mA. Maybe 151K.

Hiya Oznog,
Eh mate as you know thats a schematic I pulled off the datasheet. When you say use an R/C time constrant to 300uS could you possibbly show me how it's done. I'm not really sure how to implement that and even the general schematic shown on the data sheet doesn't show it either. I'm a mechanical bloke and electronic's are a great learning experience for me with real life things I can do. So any help I can get is both a help and a learning experince too.

Cheers Bryan

Pin 5 is an ENABLE/PWM DIMMING pin. If you don't want to dim it with PWM you tie it to Vdd and LD as per pg 3 of the spec sheet. "Vdd" is a poor choice of name for that pin, it is not a power input but rather the output of the internal 7.5v reg. The power in is connected to Vin.

The 300nS RC filter is:
1. Replace the wire between R10 and CS (pin 2) with a resistor.
2. Add a capacitor from pin 2 to ground.

Resistance times capacitance must be greater than 300 x 10^-9. For example, 91pF and 3.3K. It can be a bit greater than 300x10^-9 but don't make it wildly greater. The resistor you choose should not be really high, stay under 50k, maybe even 30k. A cheap ceramic disk cap will do fine.

I got that recommendation on the RC from Supertex directly. It's an obvious problem with the principle of the circuit and fixed all my probs. I love their part but find their negligence in documenting this fundamental requirement rather unfortunate!