Luxeon Rebel Questions

[hamilton]

Hello all,

I'm currently working on some plans to build a stage lighting instrument using Luxeon Rebels, and just want to confirm a few things before I go out and waste a bunch of money.

1. Do I need a heatsink if I am using the Rebels at 350 ma? I haven't been able to find much information on this, but what I have seems to say probably not. I will also be throwing one of these into the case (http://www.tigerdirect.ca/applications/SearchTools/item-details.asp?EdpNo=311342&CatId=494) to help with cooling, and I'm hoping that it'll be enough since I'm not running them at 700 ma.

2. As far as holders go, does anybody have any experience with these? (http://www.asiansignals.com/Products/LED-Circuit-Boards/Rebel-Single-22mm-Square.aspx) Does anybody have any suggestions as to other good holders for the Rebels?

3. Related to the above question - How exactly do lenses and optics attach to holders? Is this a universal system? I can tell that different holders are made for different size lenses, but if the two sizes match (i.e, 20mm) should any lens work with any holder?

4. Can anybody foresee any problem running 6 rebels in series?



Thanks!

James

 

[lolzertank]

Hello all,

I'm currently working on some plans to build a stage lighting instrument using Luxeon Rebels, and just want to confirm a few things before I go out and waste a bunch of money.

1. Do I need a heatsink if I am using the Rebels at 350 ma? I haven't been able to find much information on this, but what I have seems to say probably not. I will also be throwing one of these into the case (http://www.tigerdirect.ca/applications/SearchTools/item-details.asp?EdpNo=311342&CatId=494) to help with cooling, and I'm hoping that it'll be enough since I'm not running them at 700 ma.

2. As far as holders go, does anybody have any experience with these? (http://www.asiansignals.com/Products/LED-Circuit-Boards/Rebel-Single-22mm-Square.aspx) Does anybody have any suggestions as to other good holders for the Rebels?

3. Related to the above question - How exactly do lenses and optics attach to holders? Is this a universal system? I can tell that different holders are made for different size lenses, but if the two sizes match (i.e, 20mm) should any lens work with any holder?

4. Can anybody foresee any problem running 6 rebels in series?



Thanks!

James

1. Definitely. No heatsink = dead Rebel. The only time you might be able to get away without one is if you're driving under 100ma.

2. That board is not very good. From the product picture it doesn't have enough vias to be a good thermal conductor.

3. The system is not universal.

4. No problems.

 

[Oznog]

Well, here's the thing. Actually that Asian Signals board should have "fair" thermal performance, thermal vias can easily beat out MCPCB for performance. I wish they had 2x-3x more but it's already enough that for non-high-performance apps (not trying for 1A or even 700mA), the board's thermal resistance will probably not be an issue.

BUT, neither the AS nor any MCPCB "star" is a great sink. It is a good heat SPREADER. LEDs have small thermal pads. Due to it being impossible to solder to an aluminum heatsink, the difficulty of attaching leads, the somewhat poor performance of using thermal epoxy on small thermal pads carrying a lot of heat, and the difficulty and lack of versatility in epoxying to sinks, we often use one of these boards as a spreader to attach to a heatsink with thermal compound.

A 1" MCPCB with both sides exposed to air, or the 1" AS boards, will survive 1W (350mA) of heat but it's "not good". The die heats up significantly which reduces light output and reduces LED lifespan prematurely.

The case fan won't do diddley squat. If you try to sink the board by soldering the thermal pad to the top copper of a single-sided board, the temp difference between the air and the LED die is very high. Exchanging the air inside the case so the air doesn't warm up won't change the fact that the die's much hotter than the air.

 

[hamilton]

Thanks for the replies everybody!

What I am thinking of doing is spreading these holders along a piece of copper (roughly 2" X 1' for 6 dies) and then attaching heat sinks to this copper. Does this seem like a plausible idea? I am debating between using a couple of larger heat sinks, or one small one on top of the copper on top of each die (like this, http://www.futureelectronics.com/en...-accessories/Pages/3411758-507302J00000G.aspx)

On that, is copper the best material for heat transmission? That's the impression that I have gotten.

So, is the consensus that the case fan is pointless?

Thanks again everybody!

 

[lolzertank]

You don't need to get copper. For LED purposes, aluminum is probably better since you can get more of it for less $$$ and it is sufficient for most applications. Also, if you're not using a fan, you'll need something with high emissivity to radiate the heat, something that black anodized aluminum handles very well. The decision between large or small heatsinks is mainly price and labor. Large heatsinks with multiple dies would be much easier to use since you wouldn't have to fiddle around with lots of little heatsinks but small heatsinks might be more common and cheaper.

If your dies are spaced fairly far apart, a case fan is probably pointless since passive cooling should be enough.

 

[yellow]

forget heatsinks and mount the led - or their aluminium mounting plate - directly to the metal housing.
That way this one acts as sink.
General agreement "way" of building led lights

 

[LED Boatguy]

Kinda late to this thread, but I used 18 Asian Signals 3-up square Rebel boards to make 54-LED floodlights. Made 3 of them so far and no complaints. Moving heat from the boards to the heatsink was no issue at all--even during testing at 1+ amp--and have never roasted a Rebel. Those AS boards transfer heat just fine--just make sure you corrrectly mount the Rebels to the boards. I use a reflow oven.

If you are going to Arctic Silver the boards to a heat sink, I would lap them flat first--after soldering.

$.02

 

[KeithInAsia]

"BUT, neither the AS nor any MCPCB "star" is a great sink. It is a good heat SPREADER"

(chin scratch)... hum..... can someone explain to me the difference between a heat sink and a heat spreader?

Wiki doesn't cover it.

By the way people just to let you know -- I am the guy behind "AS".

 

[snarfer]

I have always used the term "heat spreader" to mean a thermal interface device. On the other hand, by "heatsink" I usually mean a device used to dissipate heat, typically into the atmosphere or a fluid.

 

[yellow]

with "heat sink" I think of those junks of metal featuring fins to widen the surface.
Problem: much weight and where to place?
with the aluminium housing of a light, one has a wide surface to get heat away and this one is light (lighter than an additional sink), is in moving air and already encapsules all the parts.
no need for useless sink

spreader? :thinking:
probably what I call mounting plate. The aluminium "sheet" where the led (or the star) is mounted on and that is press fit into the housing to get the heat from the led to the outside.

 

[KeithInAsia]

Copper is far better then aluminum -- but it does cost more. You can also get a marginal performance boost with silver -- but the percentage of thermal improvement won't justify the cost. And you won't find silver in sheets and bars so you can work with it.

Want a better thermal solution? Diamond.

Copper is the best choice if you need higher thermal performace. Don't need that much performance? - use aluminum.

Match the material to your need.


I recently built 80 LED lights for PAR 38 cans. We used the Rebel Spot 36 board and the Synjet PAR 38 product. A link for the board is here.

http://www.led-professional.com/content/view/1302/56/

(for all you guys who are hung up on vias -- there are 1,100 of them in this board)

I personally don't care for the Synjet product -- I don't believe whole-heartedly in the design approach, but that product fits right inside a PAR 38 can. So, it's an easy fit and most importantly - the customer liked it.

This solution is powerful but tends to be on the expensive side.

The question is -- how much performace do you want to buy?

If you only need 8 or 10 Rebels, then you may use the Rebel Single board and apply a LEDil Lens to it. Or the Fraen lens (22mm square lenses) -- both will fit reasonibly well. However, keep in mind that NO lens manufacturer on the planet plans very well for an LED board. They have this weak minded notion that a lens/holder should be glued or double face taped to an LEDs system.

So, the 2 holes in my board will need to be transferred over to the lens holders. You'll need to mod that lens holder to accept the screw heads.

On the Heat Sink / Heat Spreader comments ...... it's all heat sinking. (the key word here is "sink" -- place where water drains...) Meaning that heat is abosorbed. It's not any more complicated than that. Anything -- including your finger is a heat sink -- if applied to a hot area.


I think the emissivity idea is highly over rated. If you're talking about headers on a small block Chevy 350, then black anodizing is a big deal. On LEDs, I believe air flow is more significant and emissivity doesn't do a whole lot for you. Get your metal surfaces in contact with air flow. That should take care of you.

Black coverings not only emit heat engergy but they also abosorb it too. If your LED housing/heat sink is exposed to IR radiation (sunlight or a roaring fire in a fire place, you LED will be toast)


You should have no problems running 6 Rebels in series if you set up the driver correctly.


Please note -- the Rebel Square 22mm board is not a full heat sink. It's just a small amount of copper. It's a great interface device. You will need to screw it down to a full size heat sink (yes, people the screw holes are a required components in that board). The heat sink must match up to your performace requirements.

For the last piece of advice please note. When you see comments like this "That board is not very good. From the product picture it doesn't have enough vias to be a good thermal conductor."

Please understand that not everone agrees or understands what they see. That board work and works very well. It might work better with a few more vias in some situations .... but it is certainly effetive.

I have a 3D television company running a single Rebel on that board at 1Amp (on a heat sink of course). They have not burned out the LED yet. I think it's foolish to push the part that hard. Better to buy more LEDs and play it safe. I don't expect they are going to be successful with that part at 1Amp (otherwise Phillips would change the spec on it and brag to the world).

Anyhow -- well. There is go. Oznog -- you're up next for the rebuttal.

 

[snarfer]

I recently built 80 LED lights for PAR 38 cans. We used the Rebel Spot 36 board and the Synjet PAR 38 product. A link for the board is here.

http://www.led-professional.com/content/view/1302/56/

(for all you guys who are hung up on vias -- there are 1,100 of them in this board)

I personally don't care for the Synjet product -- I don't believe whole-heartedly in the design approach, but that product fits right inside a PAR 38 can. So, it's an easy fit and most importantly - the customer liked it.

I'm curious about your experience with the Synjet. I need something like that - silent, but much larger. If they made a 200 watt version or larger it would be nice.

Anyway, you said you didn't really like it, so did you have another product you liked more? I am not into the CPU cooler approach at all. They are just too small and noisy.

 

[KeithInAsia]

The Synjet people preach the virtues of their IP to no end. So much so that they simply ignore the pitfalls of their design approach.

I had one call me - and we kind of go into an argument. They get really defense when you ask them to defend their position. Kind of like a preacher who can only says "take it on faith".

The Synjet is not silent. It hums. It's basically a speaker type diaphragm that is vibrated in a sawtooth wave pattern. In contracts fast and relaxes slowly. This pushes jets of air forward. The low pressure of that action pulls other air into a passage way. The same amount of air that is jetted into that area is also sucked back in to the diaphragm area, but the net result is a very mild air flow through the passage way.

The air flow is so small, that you almost can't feel it by hand. You have to put your face up to it to feel the air moving.

They claim that the focus of that air against the metal heat sink walls is enough to do the job -- but their heat sink design is so primitive (cheap to manufacture) that even with a fan, it probably doesn't do a great job. (the heat sink was cast and it was a soft aluminum alloy)

The bottom line is this -- fans work better than a Synjet. When they called and I said to them -- Job #1 is for cooling -- why don't you cool better than a fan? - they got all flustered and started shooting down the fan --

They said the fan will fail sooner and could get noisy in the process ... I said I use ball bearing fans that are rated for 80k hours. They responded by saying the Synjet is rated to 100k to 200k hours. I said, yea but the LED won't last that long. It will be dead long before because it's running to hot!

I said, -- hey -- what about the surround of the speaker component? That can fail... what about that noise?

They simply could not defend their design.

It works to some degree, but not nearly as nicely as a higher grade heat sink and a fan. Fans that run at RPM of 4k or less are just about silent.

Fan control chips can detect fan failures and report that back to the driving system....

So, I don't see a good reason to spend more on a technology that does less. (and at full price, the cost is a great deal more)

If you have a very tight space - or the need for a lighter unit or a severe power constraint, then Synjet might be an advantage.


200 Watts is a fairly large load of LEDs.

My instinct would be to build a copper board that is the mounting solution as well as the heat sink. Then build a cooling solution into the board.

 

[Oznog]

The higher thermal conductivity of copper won't benefit you that much, thick aluminum sinks are usually limited by fin area. Copper allows heat to transfer down very tall, thin fins. Probably won't matter.

What DOES make copper a great Rebel sink is that you can solder directly to it, which is a near 0 thermal resistance for that interface. Now you could use Arctic Silver epoxy, but AS epoxy is not "as good" as AS thermal compound, and the contact area is VERY small, like a square mm. A bigger slug like 10 sq mm would have 1/10th the resistance for the same thickness of AS epoxy.

And MCPCBs are "ok", but they're actually pretty high thermal resistances. I get the impression that there are huge differences in resistance among the MCPCBs but almost nobody specifies it. MCPCBs put a thin electrical insulator between the top copper and aluminum. Rebels do not require electrical insulation, and this is also somewhat of a thermal insulator.

The prob with Rebels and copper is the bottom pads. If we try to solder- or even AS epoxy- them to the copper, it'll short out the bottom pads and there's no way to connect them. If you use an MCPCB or PTH thermal board, then the copper is a complete waste of time and money, aluminum will work almost the same.

 

[snarfer]

Thanks for the info about the Synjet. That is exactly the kind of thing I needed to know. It really does look like a cheap piece of junk. For 60 watts you could just use a passive heatsink anyway.

Actually the light I make right now is 180 watts passively cooled with no copper at all (except in the wiring, duh). The thermal design works just fine. But I want to make something more like 1000 watts, so a new approach is needed. Maybe a really big fan running slowly, or piezo fans.

 

[KeithInAsia]

"The higher thermal conductivity of copper won't benefit you that much".

Copper will always benefit you. I particularly like the heat sinks that are made with the process called skiving. This where a machine cuts small thin fins into a coppper block and pushes them up into a vertical position. This is repeated until a full heat sink with fins is created. You can read about this more here: http://www.cooljag.com/

Copper is 25% better than aluminum and always will be. It's more expensive too but probably neccesary for you guys who love to "over-clock".

If you're going to put a rebel down on a copper heat sink - use a copper clad mounting board.

MCPCBs are not ok. They are a poor concept -- however, some LEDs require an insulated layer under their thermal pads. In my opinion, these LEDs are not worth the trouble. (ever read the full instructions on how to use the special adhesive on these LEDs Yikes!)

Luxeon I, III, and V BAD! -- Luxeon Rebel Good! Cree LEDs Good!

"The prob with Rebels and copper is the bottom pads. If we try to solder- or even AS epoxy- them to the copper, it'll short out the bottom pads and there's no way to connect them. If you use an MCPCB or PTH thermal board, then the copper is a complete waste of time and money, aluminum will work almost the same. "

The problem with this "problem" is that you need to use the right product for the job -- then you won't have a problem. I never have any problems with heat sinking and heat dissipation. I can't imagine how you guys manage beat this topic like a dead horse. Holy smokes.

 

[BillyNoMates]

I have used Rebels in my builds (mainly bike lights) and I have always used Plated Through Holes PCBs.

For those of you who are interested I calculated the Thermal Resistance per via of a standard 1.6mm thick PCB with 35um plating. For a 0.6mm drill, the thermal resistance of a single via is about 105 K/W. For a layout that uses 16 vias you end up with 6.5 K/W. Thats not great, but one needs to bear in mind the thermal resistance of the Rebel at 10 K/W and make a decision on where to draw the line of deminishing returns.....

My choice is always to keep the LED modestly driven (ie below the datasheet max value) but move to multiple emitters to get more light output - but that's just me.

For most heatsink applications that rely on natural convection in air, the surface temp of the heatsink needs to be HOT - ie significantly hotter than ambient. If you look closely at heatsink data sheets one will often find that the rating of xx K/W is valid for a 50K or sometimes a 70K temperature delta.....

 

[georges80]

We aren't dealing with black magic here - vias have been used for thermal paths for a long time - most any QFN/DFN high power IC package relies on vias for transferring heat from the IC thermal pad through the board (or to inner layers).

There are some pretty simple formulae that can be used to determine the thermal resistance of plated vias (diameter, copper weight, PCB thickness) and how many are required to do the job.

FR4 boards with plated vias do a great job with the Rebel.

Instead of wondering what does or doesn't work - read Lumiled's app note....

http://www.philipslumileds.com/pdfs/AB33.pdf

cheers,
george.

 

[Oznog]

Yep, but the success of thermal vias depends on a lot of factors.
The lateral thermal resistance of the top copper is significant. You can only place so many vias around the pad itself, forcing you to place additional vias further out. Not only does this add lateral resistance, but the copper foil between the pad and outer via is perforated by the inner holes and has less conduction that a whole layer. Thicker top copper does help.

The thermal resistance is directly related to PCB thickness. A board of half the FR4 thickness means all thermal vias have half the resistance. This doesn't exactly mean the whole board has half the thermal resistance because the lateral top resistance TO the outer vias is not reduced, but it's still most of the way to half.

It takes a LOT of vias to compete with good MCPCBs, as well as thick top copper and thin FR4. Which isn't really a problem, most mfgs don't charge by the via. The PCB machine's fast. The extra top copper will cost you, though, as will a thinner PCB.

 

[Oznog]

"The higher thermal conductivity of copper won't benefit you that much".

Copper will always benefit you. I particularly like the heat sinks that are made with the process called skiving. This where a machine cuts small thin fins into a coppper block and pushes them up into a vertical position. This is repeated until a full heat sink with fins is created. You can read about this more here: http://www.cooljag.com/

Copper is 25% better than aluminum and always will be. It's more expensive too but probably neccesary for you guys who love to "over-clock".

The thermal resistance to air is primarily determined by the number of fins. The thermal resistance of the aluminum itself is 235W/m*K. That means that for a 1 cm * 1 cm bar, the resistance is 0.426K/W per cm. But a heatsink typically places the device in the middle and the heat radiates out. This makes the thermal path much wider. As such, the thermal resistance of the aluminum is usually fairly inconsequential and copper does not reduce it.

There are 2 exceptions. With exceptionally small devices which have a small thermal contact area (not on a big Star), there can be a significant increase in K/W in the material immediately around it. The second is with very long or thin fins. Thinner fins have improved airflow and higher density but the thermal resistance from the base of a long, thin fin to the far end may be so high that the end of the fin becomes ineffective. Higher thermal conductivity remedies this.

MCPCBs are not ok. They are a poor concept -- however, some LEDs require an insulated layer under their thermal pads. In my opinion, these LEDs are not worth the trouble. (ever read the full instructions on how to use the special adhesive on these LEDs Yikes!)

Luxeon I, III, and V BAD! -- Luxeon Rebel Good! Cree LEDs Good!

The MCPCB accommodates isolated and nonisolated thermal pads equally well. The thermal via solution REQUIRES isolated thermal pads.
The sucky case is non-solderable pads, because the thermal pad on the LED is aluminum. That requires thermal epoxy, which is higher resistance than soldering. Some people use specific techniques to reduce that resistance but it's tough. You can't easily clamp down the LED because the top surface won't take the pressure.

How are MCPCBs "not ok"? IIRC Asian Signals-type thermal vias are only equivalent to the older and/or more poorly designed types of MCPCB, like 10-12C/W, and they have thermal resistance issues on board-to-sink because of board deformation when pressure is applied. Modern MCPCB seems to be at most 2C/W, the performance stuff is 0.5C/W or less. Some figures went to 0.2C/W (not sure I believe that one, maybe a typo). AFAIK no thermal via construction can compete with that.