Rebel 4X, no, 6X! Feel the power!

[Oznog]

Well, here it is in all its glory. 4x Rebel Red-Orange 50's on a Cree mount!

They float together when you reflow. You can't space them even if you wanted to. 200lumens @350mA, and putting it side-by-side with a red Lamnia BL-2000, this is slightly brighter at a fraction of the power.
4W, not really asking too much of the mounting especially considering the device-to-Star interface is solder.

Reflow with a heat gun to the bottom.

I had one of the 4x in series blow up... I was having trouble with the test lead though, I was having trouble adjusting the power supply and I think I had it set too high and it gave a pulse from the supply's output cap before the current limit kicked in. Or maybe the reflow didn't work under the pad, because that first time I was using the toaster oven and maybe there was a problem, I'm not sure. The remaining devices seem ok.

I rearranged it so there are two 2x series strings in parallel and turning up to 700mA. They don't turn on at the same time, and at low power one side shines much brighter. Still, once I turn the power up they look pretty equal. In any case, the devices are each rated up to 700mA so even they don't share well it's still within the limits.

Actually when I still had them in series, I tried it at 350mA and 700mA with the BL2000 on a fixed current beside it for reference. Honestly, not a lot of intensity difference when doubling the current, in this configuration apparently the increased die temp is hurting nearly as much as the extra current is helping.

 

[Oznog]

Re: Rebel 4X! Feel the power!

Behold, the Rebel 6x Spidereye!

I had another look. The heat spreader, a copper fill area that the thermal pad is made of, has wings off to either side. They're just covered by soldermask. By removing about 1mm from either side, there's actually enough space here to mount *6* Rebels, as shown above.

Well, 6 watts is kind of a lot of power, but it fits right in with my ideal requirements. My ideal voltage match would have series strings of 3x devices. I don't have enough room for a second Star, not unless I cut it up, and it could only carry 2 additional devices on it and still meet my voltage limitation.

Oh yeah it's bright. Like unbelievably bright here. It tasered my retinas.

 

[TigerhawkT3]

Cool stuff!

You might want to current-balance the two strings to avoid thermal runaway.

 

[Oznog]

Well, I think I may not have thermal runaway issues. The thermal pads are joined by the copper foil heat spreader on top of the Star, due to package geometry the thermal pads are nearly adjacent. And I'm providing a regulated 700mA, trying to get 350mA per string to optimize the light output but 700mA is electrically and thermally permissible.

It does look like it's sharing properly once the power's kicked up.

Thing is, I was trying to keep this a 2-wire device and there's really not room for resistors, esp ones which generate a lot of heat. Well, I'm looking at the issue. Dropping 1V @ 350mA would be plenty of ballast but that's 350mW of heat and the way I'd do this is heat shrink it inline before it joins up and that's not tolerant of high temps.

I gotta say, it's pretty awesome. The BL2000 was 13.6W at full rated power, but even with a pretty low thermal resistance that brought the die temp up and it degraded the light output pretty heavy. This is giving out way more light in less than half the power, it's not as flat but the diameter's not too different.

It really sucks that Future raised its low-quantity Rebel prices, that's gonna hurt big time. I need a bunch but still not 1000x.

 

[frenzee]

Great work Oznog. I had the same imbalance issues at lower currents even when everything was in series. I did four white rebel 100s a couple months ago and the the light bounce off the wall 20 ft away with no optics at all was like that of a welding arc. In fact at 1A, if you held a sheet of Kleenex in front of it, poof!

 

[Oznog]

In fact at 1A, if you held a sheet of Kleenex in front of it, poof!

"Poof"? What, the kleenex caught fire or the LEDs smoked?

What type of board is that one you've pictured there?

 

[jeffosborne]

I like using a few re-orange Rebels with several neutral whites to warm up the light for photography. I drive them at 850ma. They do have a very respectable output. So Oznog, what are you lighting up with so many red-orange lights? Cheers, Jeff

 

[Oznog]

I like using a few re-orange Rebels with several neutral whites to warm up the light for photography. I drive them at 850ma. They do have a very respectable output. So Oznog, what are you lighting up with so many red-orange lights? Cheers, Jeff

Rebel was only rated for 700mA.
Have you checked the graphs? The problem is that Rebel red-orange generated about 3.3x the heat at 850mA, and the thermal resistance of the package is on the high side. The efficiency of the die goes down quite a bit, you get only a 65% light gain when doubling the current even if the heatsink maintains 25C. That penalty's basically instantaneous when turning the device on, the heatsink will be hotter unless it's enormous and that will produce additional losses over a few minutes as it heats up- which is kinda unfortunate for photography because your color balance will change over that time.

I crunched the numbers and 6 devices at 350mA was much better than 3 devices at 700mA. The additional forward voltage at 700mA resulted in an additional 24% of heat, AND the 3 devices would only go up by 65% even if I could maintain the same sink temp whereas doubling the number of devices to 6x at at 350mA is double the output.

So it really does come down to increasing the number of devices is more beneficial than increasing the current past the suggested 350mA. Thus the reason for uber-packing the Star instead of jacking up the current. The technical justification is sound.

 

[frenzee]

"Poof"? What, the kleenex caught fire or the LEDs smoked?

What type of board is that one you've pictured there?

LOL, the Kleenex, of course. Therefore, I suggest you don't use your fingertip to see if how hot these emitters get

The board is a 21mm dia. Bergquist copper core disk. Here's the trace layer if you're interested, and few other crazy stuff, in case you decide to put up to 18 of these little guys on a single star:

 

[Jarl]

How are they connected to each other?

 

[rizky_p]

How are they connected to each other?

they are basicaly 2 series and 3 paralel right in 6x setup? (imagining Cree board)

 

[Oznog]

The Cree board has a long trace on either side for the power, each one has a solder pad on each side. Each trace was cut in 3 places and the Rebel's power terminals straddle the cut. So on one side 3 devices are in series without doing any wacky soldered micro-wiring hidden in there. Which is genius, if I do say so myself.

So we've got 4 wires available, we can make the 2 strings into series (18V @ 350mA) or 2 in parallel (9V @ 700mA). Like I say, the current sharing isn't all that bad with the thermal pads connected to each other by copper, and the packages are rated to 700mA anyways so there's no real risk of destructive runaway.

It is essential to use solder paste and reflow it. Hand work is impossible because there's no exposed pad area to apply a soldering iron to!

 

[Oznog]

LOL, the Kleenex, of course. Therefore, I suggest you don't use your fingertip to see if how hot these emitters get

The board is a 21mm dia. Bergquist copper core disk. Here's the trace layer if you're interested, and few other crazy stuff, in case you decide to put up to 18 of these little guys on a single star:

frenzee, do these things exist or are they hypothetical designs? It is obtainable?

On the Cree Star, with 6 devices running @ 1W each, I used a tiny thermocouple to measure on a tiny blob of solder off to the side of the thermal pad. I get about 15C above ambient after about 30 sec, and frankly the Star's mounted like crap, I only have 1 screw I can tighten down onto the sink so a some of that rise is simply Star-to-sink thermal resistance.

 

[frenzee]

frenzee, do these things exist or are they hypothetical designs? It is obtainable?...

I've actually made One and Two and they exist and I've posted pictures before. Three and Four would be very straightforward to build. Five is, well, unobtainium at the moment. Although the pad topology would allow 18 Rebels to coexist, designing the trace layer (for a series connection) would be quite a challenge, not to mention the driver.

 

[Gomer]

 

[mrsinbad]

Gomer, I like that... tell me more!

 

[Gomer]

http://www.asiansignals.com/RebelBoards.htm

I have ZERO affiliation with them nor can I vouch for their products in any way shape or form.

 

[Oznog]

http://www.asiansignals.com/RebelBoards.htm

I have ZERO affiliation with them nor can I vouch for their products in any way shape or form.

I have a sample of that specific 6x board from that company.

It is NOT an aluminum-core thermal board. It's a fiberglass core double-sided PCB. It's got plated throughholes and all to try to conduct the heat to the bottom, but the thermal impedance looks very poor to me.

 

[Gomer]

I have a sample of that specific 6x board from that company.

It is NOT an aluminum-core thermal board. It's a fiberglass core double-sided PCB. It's got plated throughholes and all to try to conduct the heat to the bottom, but the thermal impedance looks very poor to me.

Any chance you have measurements on the thermal conductance?

 

[Oznog]

No, but the board's 0.042" total, foil and glass. The foil appears quite thin, the difference in depth between the white paint and the gold plated foil areas is 0.002". Since I don't know the paint thickness I can't say for sure what the foil thickness really is. There is a tiny rim where the foil on back ends before the pcb edge, no paint... it's real hard to read, and I may be reading edge imperfections from the cutting process, but I get a foil thickness of ~3.7 mils. Probably 2 oz copper before plating then.

Aside from the 3 large mounting screw PTHs, each of the three "pairs" has one Rebel mounting area surrounded by 4 small PTHs and one mounting area with 11 small PTHs. In addition, there are 2 *extremely* small PTHs directly under the Rebel's thermal pad. There are also 2 holes between the mounts making up a pair, but they removed the copper under them on the backside, thus these have no thermal conduction value.

I know Luxeon Application Brief AB32 shows a number of designs relying on the PTH copper plating to bring the heat to the back, achieving a thermal resistance of 7C/W or even 3C/W to the back (note: not to ambient). This board does not have a similar density of PTHs. So, assuming the conduction is all about PTH (which AB32 says a lot about) and the 7C/W board with 31 large PTHs can be compared directly with the 5 PTH of this board... I get an estimate of over 40C/W on that 6x circle board. If that estimate's correct this is a failed, ineffective thermal design and inappropriate for a powerful Rebel application.

I don't know what they were thinking. Maybe you are supposed to use aluminum screws and they'll do the lion's share of the conduction to a sink? However, I did some calculations awhile back, the thermal resistance of copper foil when used to conduct heat laterally is limited, which is probably why Luxeon did not recommend adding more thermal vias outside of about 1mm-2mm from the perimeter; the thermal resistance from pad to the outer vias would be so great that it would yield little additional benefit. So even the screw would not get a great thermal resistance because it's only on one of the 4 sides of the pad.