Rebel 6x Cree Star?

[Oznog]

I was trying to think more on how to mount a number of Rebels to get the most power in a small place. I need to work with "red-orange", I did see the 3-emitter boards but 3 barely meets my specs.

Well, I did have one genius idea that might top that:
http://www.dealextreme.com/details.dx/sku.4512

The Cree Star looks like it can definitely mount 4x if I just slice the contact strip in 2 places on each side to make 1 strip into 3 sections, for the leaving each side in series. I keep going over the spec sheet and it looks just right!

In fact I am trying to figure out if that Cree Star could fit 3x Rebels abreast for a total of 6. I don't have a board yet so I can't get the exact measure.

Of course there are real thermal limits, 6 Rebels would be over 15W at 700mA. I will be mounting that Star on a good heatsink but I doubt that the die temp could be kept reasonable at 15W. But for sure you can make more light with 6 devices than 3 and we'll try to see what power level generates the most light without exceeding the maximum die temp and staying reasonable efficient (knowing how much reds lose output when the die temp rises). And the red-orange are cheap anyways, $2.21 @ futurelec, so this is looking like a pretty wise choice.

 

[LukeA]

I don't see how you could solder a significant number of Rebels to that star.

Edit: I see you're going in series-parallel, not all parallel, it might be doable, but heat will be the 800lb gorilla.

3 wide won't fit, that needs 9mm, and a Cree XR-E is only 7mm wide. Hmm, maybe if you scrape a little of the black off of the strip, but then the heat transfer won't be very good and the solder might not stick unless you do a very clean job of removing it.

 

[Oznog]

Well, the Cree is 7mm, but the Cree Star's pads ARE wider. Since I don't have one, I'm not 100% sure how much. Now the Rebel is 3mm wide, though actually 0.1mm is a margin on the side. Also the solder on the Rebel's pad may provide some lateral conduction on an overhanging area to the pad.

What it looks like is even if the Rebels on the ends had its thermal conductivity reduced to only 80% because 20% of the thermal pad overhung and could not be soldered to the star, we're still talking about 50% more devices here. If we reduce the power per device to 80% to maintain die temp, there's still 20% more output than with 4x devices by my calc. Since the red-orange devices are cheap this could be justifiable. What I didn't calc though is the devices are going to be less efficient unless I can keep the heatsink's temp down under this high wattage stuff.

Also worthy to note is that unlike most LEDs, the die can handle up to 135C without serious degradation, it just reduces the output. So they're in series and share the same current, the one in the center will be cooler and brighter so even if the ones on the ends are losing efficiency they'll still be contributing light to it and maintain their lifespan.

I wish I could make up my own boards, but the aluminum-core PCB is surely not a cheap option.

Ah well this'll be a cheap experiment. Come on, we all wanna see how much light you could actually get when you put 600 lumens of device rating in a space under 1/2". Elevated die temp lowering the efficiency thus the lumen output... yeah really I do know that's gonna be a real significant factor. We'll see.

 

[Gryloc]

Oznog,

I had the same idea several months ago and I tried soldering on 6 of the Rebel 0090 emitters, but my soldering iron was not up to the task (the Al star board dissipated heat too quickly to allow the solder to melt). I had no access to any reflow soldering ovens (a stove or toaster oven). Give it a try, but you will find that if you choose to run over 400mA through each, things will get rather toasty! I ended up soldering 4 of those emitters in the same configuration on a piece of copper (the -/+ pads were hanging over the copper sheet), and it now resides in an old Streamlight 3AA TaskLED light modded to fit a Fatman driver. Even when the copper piece with the 4 LEDs are attached to a decent heat sink with a thin layer of paste, the copper gets hot when I tried to run 400mA through everything. The power density is so high that you have to have a good medium to transfer the heat. Osram seemed to get everything right with the Ostar.

Well, have fun with this. Even 6 of them re-flow soldered neatly onto the Cree star powered at 350mA each will provide you with some great amounts of light! You could slowly ramp up the current to see when things get too warm. It is easy to forget how quickly the power ramps up when you are slowly increasing the current.

As for a small source of light for mating with a big reflector, do not get too excited. When I put my 4 symmetrically positioned Rebel 0090s behind a smooth Mag reflector, I got an ugly donut when correctly focused, just as I anticipated. When de-focused some, you can obtain a small spot, but you will have an equally ugly flower pattern in the beam. If you had a much larger reflector, or one that was stippled, you may obtain a nicer, cleaner beam. I put a small diffuser on the above mentioned Streamlight flashlight to smooth out the beam, but I ended up with a pretty wide flood.

I have a good method to test beam shapes when you have multiple light sources (like a tiny cluster of Rebel emitters), but I have not had enough time latley to post in enough detail with pictures to demonstrate. I tested multiple emitter configurations way before I can even get the emitters soldered up. It is really simple to make your own mini light sources with tissue paper and electrical tape. I knew from the start that people would be dissatisfied with the tri-Rebel stars.

Anyway, I would like to see you do this. Please keep up updated throughout the process! Thank you.

-Tony

 

[Oznog]

Me, I don't need or want a collimated beam. It's a position light.

IIRC there IS a funky 3-emitter optic MADE for a 3-Rebel Star. It's got 3 separate collimators. At least I'm pretty sure I remember it being for a 3-Star not 3 adjacent Stars.

That was intended for an RGB setup but there's no law saying the emitters can't be the same color. Since the optics are smaller I expect a somewhat lower capture ratio, which could potentially lower the effective efficiency by tens of percents unfortunately.

 

[Oznog]

Well I just found the 3x Rebel "Endor" Star datasheet and it provides hints on thermal resistance of Stars with close-mounted emitters:

http://www.luxeonstar.com/7007-endor-datasheet.pdf

OK, stay with me here. The thermal resistance, junction-to-pad, of the white Rebel is 10C/W. The Endor spec says with only ONE emitter installed it's 13C/W from junction to Star backing. So the Endor Star's adding 3C/W.

Now with 3x devices installed, looking at is as one big emitter, the Rebels' junction-to-pad simply totals 3.3C/W.

So, with the heat originating from more points, does that reduce the Star's overall thermal resistance from the pad(s) to the backing? Or was the heat spreading from the pad not an issue in the first place so the thermal resistance is unchanged when we have more pads?

The Endor datasheet seems to say the ladder case- the thermal resistance of the Star's mouting is not improved with the number of devices. I say that because it says the total resistance is 6.3C/W for 3x installed. We already know the Rebel's total junction-to-pad of 3.3C/W so the Star's adding the same 3C/W to the device(s). So that says that if we have 1 device running at 1W or 3x devices at 333mW, the device(s) pad(s) will be at the same temp.