Cree XR-E Thermal Concerns

cmacclel

Flashlight Enthusiast
Joined
Jul 15, 2003
Messages
5,018
Location
Sweden
Many have you have been asking about the thermal condictivity of the Cree round and star bases so here are your pics.

As you can see there is a thin layer of PCB between the thermal pad of the Cree and the actual aluminum heatsink base. Not a very effecient solution in my opinion. Lumileds emitters are directly thermal epoxied directly to the aluminum for you would think a much better thermal path.


IMG_3023-vi.jpg



IMG_3024-vi.jpg



IMG_3025-vi.jpg



IMG_3026-vi.jpg
 

LumenHound

Flashlight Enthusiast
Joined
Apr 16, 2005
Messages
1,797
Location
Toronto
Nice photos!
Looks like an inexpensive phase change adhesive "thermal pad" between an emitter and bare metal heatsink would give the same so-so results of those round and star shaped bases.
 

frenzee

Enlightened
Joined
Sep 4, 2006
Messages
216
I think that's how Lux I's are attached too. Not an efficient way IMO either.
 

Anglepoise

Flashlight Enthusiast
Joined
Nov 4, 2004
Messages
1,554
Location
Pacific Northwest
Thanks for taking the time to chop up and photograph.
I will let others, more knowledgeable than I, tell us what this will mean in real world terms.
 

IsaacHayes

Flashlight Enthusiast
Joined
Jan 30, 2003
Messages
5,876
Location
Missouri
Not good for 1amp drive level that's for sure. Luxeons are durable but I've killed one with the stock star at 1amp drive. The thermal resistance of the pcb was too high.

700ma the cree will probably survive, but won't run well. 350ma should be ok. That's why luxeon1s at 350ma use the pcb, and lux3 at 700ma dont.

I personally like a max of TWO thermal interfaces in my lights, and that's with thermal epoxy. Emitter to heatsink, and heatsink to flashlight. One would be awesome.

A copper star would be excellent, with a pcb cut out for the center to attach to the copper directly, and pads above for the connections. Sure the solder would be a bit thicker but I think it would conduct heat better than even a thin layer of thermal epoxy...
 

NewBie

*Retired*
Joined
Feb 18, 2004
Messages
4,944
Location
Oregon- United States of America
FYI, those stars and rounds are *NOT* produced by CREE, but by other companies, and some distributors. The ETGTech version is a very common one. Unfortunately, it is lacking.

All MCPCBs are most definitely not created equal, and I most definitely applaud your taking one apart to show us all the short commings of the first round of MCPCBs produced by various third parties for the CREE LED. You get what you pay for.

They buy the CREE LEDs and mount them on the boards themselves.

The FR-4 you see there is a great insulator (epoxy and fiberglass), and it has to be made extremely thin just to get modest thermal transfer.

An example of the problems this causes is shown here:
etgmcpcb.png



Directly soldered to copper, for comparision:
xrecopp.jpg
 
Last edited:

cmacclel

Flashlight Enthusiast
Joined
Jul 15, 2003
Messages
5,018
Location
Sweden
So Newbie what boards be it round or star provides the best thermal transfer?

Mac
 

chimo

Flashlight Enthusiast
Joined
Sep 16, 2004
Messages
1,905
Location
Ottawa, Canada
Yes, LumiLEDs changed their star heatsinks as well. Here is an old LuxIII vs a new LuxI heatsink. Note that they made the copper under the slug as large as possible for heat spreading.




Paul
 

davidefromitaly

Enlightened
Joined
May 16, 2004
Messages
635
Location
43.66 N - 13.13 E
IsaacHayes said:
Not good for 1amp drive level that's for sure. Luxeons are durable but I've killed one with the stock star at 1amp drive. The thermal resistance of the pcb was too high.

700ma the cree will probably survive, but won't run well. 350ma should be ok. That's why luxeon1s at 350ma use the pcb, and lux3 at 700ma dont.

I personally like a max of TWO thermal interfaces in my lights, and that's with thermal epoxy. Emitter to heatsink, and heatsink to flashlight. One would be awesome.

A copper star would be excellent, with a pcb cut out for the center to attach to the copper directly, and pads above for the connections. Sure the solder would be a bit thicker but I think it would conduct heat better than even a thin layer of thermal epoxy...

yeah a copper star!!! some modder can start to di it ;) solder it with tin mean to have 10 times more heat tranfer vs. the best arctic product.
 

NewBie

*Retired*
Joined
Feb 18, 2004
Messages
4,944
Location
Oregon- United States of America
cmacclel said:
So Newbie what boards be it round or star provides the best thermal transfer?

Mac


Well, thats a tough one to call. They do function, but why would these distributors purposefully throw the cat in with the wash?

Here we have a really nice XR-E product that CREE worked so hard at minimizing thermal resistance, so you could pull the heat out of the die, in order to maximize the light output (lumen output drops as the LED heats up), then some distributor comes up with a cheap hack that works the opposite way, and launches them off to yet another board house for mounting. I don't know what nutcase designed these things.

They could have very easily put 150 copper vias under the thermal pad area, which would have given direct thermal transfer thru the board directly to the aluminum (common in RF designs), and also flooded the copper from the thermal pad across the surface of the board, to also act as a heat spreader. This would have resulted in a thermal resistance well under 0.5 C/W. Instead of using what looks to be 0.000707" thick copper (1/2 oz.), they could have easily used 4 oz copper to help maximize the spreading across the entire surface of the board. Sure, so it would have actually cost nothing for the vias as well as nothing for the copper flood. If they were wanting to maximize performance, they could have spent another 0.04 dollars for the thick copper, but it sure would have helped further. I don't know, but this just reeks of cheap.

If I'd spend some of my own hard earned cash on these myself, you bet I'd be on the phone and someone's ears would be on fire...this isn't rocket science, just Thermal Management 098.

It would be cool if CREE were to take the mess these distributors made out of these and make something proper. Though, I don't think there are any plans for this- but I've never asked CREE about it.

Which works better? Harumpf, neither?

In reality, if you don't care much for getting great performance, and don't mind loosing some of your lumens due to die temperature rise, as well as color shift due to higher temps, I guess they work adequately.
 

Erasmus

Flashlight Enthusiast
Joined
Dec 29, 2004
Messages
1,077
Location
Belgium
Interesting post and good to know for those who buy mounted Cree's. But maybe the threadstarter can change the title. It's no thermal concern about the Cree LED but about the third party PCB's. I hope there will be better PCB's soon!
 

NewBie

*Retired*
Joined
Feb 18, 2004
Messages
4,944
Location
Oregon- United States of America
chesterqw said:
maybe there is some super secret formula for a thermal conductive pcb?

Nah, the techniques have been pretty common for the past decade, nothing new to see here. Usually, where the FR-4 is used in the ones ETGTech supplies, you use instead, a thermally conductitive dielectric material, instead of an insulator material.

I'm looking at one of these and the FR-4 layer they use looks to be about 12 mils thick, but I do not have one that is unmounted for measurement. If this is in fact correct, the thermal resistance would be about 25 C/W.

Typical MCPCB's from way back when used to use 3 mil thick FR-4, which results in a thermal resistance of about 7 C/W.

Something that came out a few years ago, was a special thermal dielectric material that enhanced the thermal conductivity of the dielectric layer (typically the board material was black in color inside), which has a 3 C/W thermal resistance.

Just simply using 150 vias (which is 100% free in volume manufacturing) under the part can lower the thermal resistance to way below 0.5 C/W. This can be lowered even further by utilizing top and bottom copper layers, flooding from the thermal pad (the copper spot you solder to) across the board surface.

Using vias as thermal transfer points can be quite effective. You could make the board 39 mils thick (1mm), and do nothing more than put only 9 vias of 0.5mm in diameter, and still end up with a thermal resistance of 9 C/W. If you reduce the board thickness to one of the standard pre-preg board materials, like the 3 mil thick material (0.0762 mm), you end up with a thermal resistance of only 0.69 C/W.

Keep in mind, with the CREE, the ceramic material used is not the best for thermal spreading. As such, one would most definitely want to put the vias right under the die area.


So, how does it work in the overall scheme? I took a A19 head with a 2x123 cell body, and with the surface area, very roughly estimate the flashlight has a 30C/W thermal resistance to the ambient air. Using the 12 mil thick FR-4 MCPCB would end up nearly doubling the thermal resistance of the overall solution from the LED die to the ambient air.

However, if one were to carry the flashlight in their hand, as most people do, you always sweat a little, causing a very good thermal path from the hand, to the blood steam, which then pumps the heat away, much like a coolant system does for the engine in a car. Unfortunately, I do not know the exact C/W of this interface, but in this case, the use of the ETGTech MCPCB would be by and far be the highest thermal resistance point, and have a highly significant impact on the overall thermals.


On another note, if you want to go high tech instead:

There have been some developments in the past few years where they use graphite as a thermal spreader (done right, it has a lower spreading thermal resistance than even copper, about 20% better), then use the area advantage to lower thermal transfer thru the board.

An example of one of these newer technologies:
http://www.graftechaet.com/Technical-Documents/Technical-Bullitin/EG-ZS-PP-098.pdf
 
Last edited:

chimo

Flashlight Enthusiast
Joined
Sep 16, 2004
Messages
1,905
Location
Ottawa, Canada
I just sacrificed a blank ETG board (not the same as above). The FR4/copper layer on that one is around 4-5 mil. I may have to hang on to these.

I checked the FR4 layer on new LumiLED star and it measured 5 mil.

Paul
 

cmacclel

Flashlight Enthusiast
Joined
Jul 15, 2003
Messages
5,018
Location
Sweden
Last night I did a quick test with the MCPCB's I ran the mounted cree for about 1 minute while measuring the LED die temp then the back of the MCPCB the difference in temp was ony 15 degrees so they seem like they are working ok.

Mac
 

SemiMan

Banned
Joined
Jan 13, 2005
Messages
3,899
Best MCPCB Material

As NEWBIE has pointed out, there are vast differences between MCPCB materials. The best in my experience is LAIRD THERMAGON 4mil thermal pre-preg with 2ounce copper. I have only used it on an Aluminum core, but I hear it is even better on a copper core. It is not as good as mounting directly to copper, but it is probably as close as you are going to come with current products where you need isolation. It appears to be as almost if not as good as some of the esoteric methods such as Anotherm where they place traces right on the heat sink. The material is 0.053C in2/W so for a 5mm*5mm conduction path, you can get about 1.3C/W. With some extra copper for spreading you can get this down to about 1/2 of that.

As Newbie has pointed out though, using lots of vias, especially if they are filled, will be even better. However, if you need isolation, the Thermagon is one of the best options.

Semiman
 

Anglepoise

Flashlight Enthusiast
Joined
Nov 4, 2004
Messages
1,554
Location
Pacific Northwest
Things are getting a bit technical and above my head.

Where can we buy these 'good' MCPB's?

Are they generally available for sale or as I rather suspect, have to be custom made for each application.

If the later, how does one go about getting 1000 or more made up just for us.
 

NewBie

*Retired*
Joined
Feb 18, 2004
Messages
4,944
Location
Oregon- United States of America
cmacclel said:
Last night I did a quick test with the MCPCB's I ran the mounted cree for about 1 minute while measuring the LED die temp then the back of the MCPCB the difference in temp was ony 15 degrees so they seem like they are working ok.

Mac


Good deal.

How did you go about measuring the die temperature on yours?

I hope it wasn't an IR thermometer, they are not that accurate in this situation.

I'm seeing 10-20C rise just from the aluminum of the MCPCB to the body of the CREE, which means the die is much hotter than that.
 
Top