Cree XR-E Thermal Concerns

[ViReN]

One Alternate solution some one had recently posted (was it milkey?) is to put Cree on New Generation Luxeon's (they have much better copper below the surface) solder able? not sure, but it can be epoxied (like the luxeon slug)... legal? may be not, but certainly more efficient than the ETG and Third Party MCPCB

It would be best if Cree themselves come out with Star's and Rounds, they would hopefully be doing much better job, wonder what's stopping them? is it some thing with Luxeon Patents???

 

[Calina]

Nevertheless, is there something better available?

 

[45/70]

Well Don, I hope you are correct. As I pointed out earlier, you guys are way ahead of me! As NewBie has pointed out, these boards just don't seem right, and the fact that Lumileds mounts their higher power LED's directly onto the aluminum heatsink, I just have to wonder?

I was planning on trying one (mine will be mounted on stars) and seeing what happens. I would like to be able to run them, at least intermittently, at 1000mA+ and be able to continuosly run them at 750mA in a brass (not as heat friendly as aluminum) light. What I'm seeing here just doesn't look too good. :sigh:

Again, I hope your right. :thumbsup:

 

[Calina]

Viren, this solution has not been proven yet to be any better than ETG's MCPCB.

It look fine in theory but before it is tested we don't know!

 

[NewBie]

According to ETG and a spec sheet they provided me, the isolation layer used on the MCPCB's is rated at less than .6 C/W resistance.

Their numbers do not add up.

3 mil thick FR-4, has a thermal resistance of 7.4 C/W, as measured by Flomerics and OSRAM.

To work out, ETGTech's FR-4 layer would have to be 0.00025" thick.

It is not.

For a sanity check take another look at the photos, and note the temperature scale on the side for various points, like the LED near the metal ring, towards the ends, and then look a surface of MCPCB area.

Next, take a look at this one, note the scale for temperatures on the side, and compare the LED metal ring, the various areas of the flat spot on the CREE body, and then also the copper surface that the XR-E is mounted to:

I'm looking at the blown up version of Chimo's photo of the ETGTech's MCPCB again, and I'm wondering if the solder was even reflowed at all in the center of the CREE thermal pad. It could be just as likely that Chimo melted the solder under the thermal pad, and it never even got reflowed there in the first place:

 

[McGizmo]

Newbie,
Both images present a good case for questioning. :shrug:

ViRen,
The direct mount of the Luxeon to the Al core of the star is fine if your LED leads are in a plane above the sink plane. (it's a shallow hole there) If you mount a XR-E to this board, you have a gap to fill and with what?!?!

I have mounted someof these directly to Al and had the lead pad areas relieved to avoid electrical contact. This is fine but I am also avoiding thermal passage from the package to the sink here.

In terms of Luxeon's being mounted to the bare core of the boards, consider the fact that the K2 is the first Luxeon that even has the reflow option. Any K2's mounted on stars by Lumileds?

There are theoretical ideal methods where cost and effort are not factored in and then there are commercialy viable methods which compromise based on cost and time.

We have seen some examples here where it looks like QC may be more of a problem than materials or design even.

None of our lights are going to perform as well as a LED mounted to a big hunk of copper unless the design includes that big hunk of copper. There will likely be some loss in efficiencies but the important thing to consider is the magnitude of these losses and whether they are acceptable when it is all said and done.

Clamp one of these to a 10kt emerald cut diamond and you are on your way!

 

[Calina]

You might be glad to know that the price of diamonds is way down since they started producing artificial gems.

Hey don't rush to get some; I'm just teasing.

 

[IsaacHayes]

Don, if you snipe the corners with wire cutters you sever the electrical connection of the bottom of the pads and they become neutral. Solder to the top then for connection. Attach with AA epoxy, or solder to copper. No need then to make reliefs/depressions in the heatsink to prevent a short. It's really easy and effective.

IIRC Newbies copper was a very thin copper sheet. Not the most perfect setup either, but good enough for showing what a direct solder to copper can do.

Right now I'm torn between just AA'ing it to aluminum, or soldering it to a copper disc the thickness of a penny, and AA'ing the "penny" to the aluminum light. The former would only be 1 thermal interface, but the later would help spread the heat better so it had a larger area to go through the less efficient thermal interface of AA epoxy... :shrug:

 

[ViReN]

ViRen,
The direct mount of the Luxeon to the Al core of the star is fine if your LED leads are in a plane above the sink plane. (it's a shallow hole there) If you mount a XR-E to this board, you have a gap to fill and with what?!?!

Dear McGizmo,

I was referring to the chimo's Luxeon Pictures mentioned in the following post.
https://www.candlepowerforums.com/posts/1688325&postcount=9

of course, direct mounting on solid aluminum / soldering to copper would be much preferred

perhaps Cree could change LED design so as to easily mount on heat sink.

 

[NewBie]

Even a thick, I think it was 39.3701 mils (1mm) FR-4 board in their example, glued to aluminum, with only 9 vias in it has a thermal resistance of 9.7C/W. If one reduces the thickness of the board to a standard pre-peg FR-4 thickness of 3 mils, AND keeps the same nine thermal vias, the thermal resistance drops to 0.746 C/W. There is a tremendous reduction of thermal resistance, just by adding vias to nothing more than standard FR-4. A person can easily decrease the thermal resistance further, by utilizing smaller vias, and more of them, to increase the copper area that goes thru the board. This doesn't count solder that partially fills the vias, which lowers if further. I'd seriously consider putting the vias under the LED die area and filling the whole thermal pad area with them.
http://www.flomerics.de/Produkte/Osram/Golden_Dragon_LED.pdf

 

[Christexan]

Speaking of copper sheets etc, a common issue I've seen (but haven't seen this solution to) is soldering wires to the top contacts of the Cree. The common issues seem to be:
#1 - Difficult to solder cleanlyi
#2 - Wires interfere with reflector/optics placement to a larger or smaller degree...

Solution to both of these, get some very thin copper sheet / heavy copper foil... cut the sheet into a thin strip (roughly the size of the Cree contacts, with a little extended part either off the short or long edge depending on desired wire routing...)
tin the Cree contacts, and the copper sheet, then lay the copper sheet over the contact as desired (with a dab of flux to improve the process, as desired). Apply soldering iron briefly to top of sheet (should heat up very quickly considering thickness)... and remove... probably would only take a second or two to make the joint. Now you have a "solder tab" off the edge of the package, you can solder to directly (flat), or roll/fold into a tube shape, and insert wire then solder for a stouter joint.
Tweak sheet thickness, overhang, etc, as desired for your needs, but now the entire LED package "top" is still flat (a few mils higher, but still flat and level for the most part).
Hope someone finds the tip handy, if it works (can't imagine why it wouldn't).

 

[45/70]

In terms of Luxeon's being mounted to the bare core of the boards, consider the fact that the K2 is the first Luxeon that even has the reflow option. Any K2's mounted on stars by Lumileds?

Well, no but, there are no XR-E's mounted on stars by Cree either. Hummm. :thinking:

Dave

 

[45/70]

Newbie,
Clamp one of these to a 10kt emerald cut diamond and you are on your way!

Group Buy?

Dave

 

[45/70]

Even a thick, I think it was 39.3701 mils (1mm) FR-4 board in their example, glued to aluminum, with only 9 vias in it has a thermal resistance of 9.7C/W. If one reduces the thickness of the board to a standard pre-peg FR-4 thickness of 3 mils, AND keeps the same nine thermal vias, the thermal resistance drops to 0.746 C/W. There is a tremendous reduction of thermal resistance, just by adding vias to nothing more than standard FR-4. A person can easily decrease the thermal resistance further, by utilizing smaller vias, and more of them, to increase the copper area that goes thru the board. This doesn't count solder that partially fills the vias, which lowers if further. I'd seriously consider putting the vias under the LED die area and filling the whole thermal pad area with them.
http://www.flomerics.de/Produkte/Osram/Golden_Dragon_LED.pdf

Sounds good to me, NewBie. When can you start? :naughty:

Dave

 

[McGizmo]

Soldering to the top is a PITA and the fact that the lens retaining ring is tinned and more than happy to receive solder doesn't help!! :green:

Instead of nipping the corners off to remove the vias feeding the bottom side, has anyone tried just taking a burr to the vias themselves to break continuity?

ViRen,

The current XR package mounts easily to a heat sink using reflow technology. The issue here is more in regards to quality control, quality and effective thermal relief of various MCPCB's and what have you.

Now for custom or modded lights where production and assembly costs are not of primary concern, there are other alternatives that are likely more efficient interms of thermal resistance. This is one link in the chain of the complete light package. Efficiencies are compromised at many levels. The key is to consider the total package and determin if the summation of losses are at a permissible level or not, or so I would think. The more we understand each link in the chain, the better we can control and dictate the eventual strength and nature of the full chain.

Newbie's IR image here is contrary to what one would expect from the MCPCB. A similar image may be contrary to what one would expect from another solution that looked good in print. :shrug:

This image is of a hand assembled part that does not have a good and complete thermal path from MCPCB to can (view the delta from LED to the perimeter which I believe is the can lip which is above the contact plane). The LED is being driven beyond spec. This is a sample of one. It looks like a 10C difference and this is with a drive current of 825 mA or about 3 watts? Does that mean we are looking at a delta of about 3C/Watt from source across the MCPCB itself (including not only the boundaries from LED pad through the isolator film to Al core but spread through the core and then back up through the boundary layer and into the white paint of film that is the surface finish on the MCPCB and actually being photographed)? Some of these thermal epoxies have a great thermal conductivity when used in very thin films but those numbers are also based on surface area. What do we get with a real IR photo like this when the part is mounted on a sink?

I won't pretend that this thermal image is an image I can understand or properly interpret. Had Newbie scrapped a spot on the MCPCB to show the raw Al core of the MCPCB, would we see this as a hot spot in the image? What is with the streak from right to left that shows a much cooler section across the whole image? I assume I am to disregard this portion of the image but why disregard this but accept the balance and pretend to understand what I am seeing? Just what am I seeing here? How am I to read this? GlowBug has done a number of IR images of complete flashlights and I recall one he did of a chrome Aleph that was just bunk. It seems that chrome is not a reasonable surface to get an IR image from. The chrome seems to hide the IR energy from the camera. :shrug:

Newbie has stated that he coated the part with some material that will provide the same emmisivity (sp) to the camera. This makes me think that the camera is imaging the surface and not the core beneath. Isn't the core beneath insulated by both the isolation layer as well as some top film? If yes, perhaps this insulation is not significant.

I have learned from personal experience that an IR thermometer is not to be trusted for absolute temperature readings. However, an IR image is? :thinking: I can imagine that an IR image is viable for relative comparison but for absolute measure I need some coaxing.

 

[IsaacHayes]

I've had no problem soldering on the top. I use very fine teflon wire prebent where I want it to go. :shrug: This is even after the corners have been snipped off so there is even less to solder to. Christexan's idea of making LED leads is pretty good too, but might be hard when you solder the wire, it might unsolder from the LED at the same time. But would make it look more like a luxeon lead package.

Don, yes at first people were grinding the pads off the bottom, but clipping seems so much faster.

Yes it's hard to tell exactly whats going even with the IR picture, but it's easy to see that the cree runs about 2x more hot with the pcb. Weither or not you have much actual lumen loss is another thing to ponder as if you don't then all this worry is moot point. Unless it shortens the life of the led considerably.

A test with a cree on MCPCB, to measure lumens with a ceiling bounce or light box, then take the same cree off, re-flow it to a copper disc/sheet metal, and re-test would be needed to see how much of an improvement we get I think is the next logical step.

 

[Anglepoise]

Instead of nipping the corners off to remove the vias feeding the bottom side, has anyone tried just taking a burr to the vias themselves to break continuity?

Yes Don. I have done this with a diamond grinder.
See this post. 3rd one in. Click

 

[Christexan]

I thought about my "solder tab" idea unsoldering itself when connecting the wires to the "tab" part of it shortly after I posted it, but 2 thoughts on that...

#1 - Once soldered directly to the package, the package itself will act as a heatsink for any heat applied to the "tab" part (open to ambient), so the temperature gradient from hanging tab part to package soldered part should be dramatic (meaning, the solder to the package itself is unlikely to heat up across the width enough with brief "end tab" soldering to loosen completely from the package (the heat being "sunk" into the unit itself).
#2 Once the "tab" is fixed to the Cree package, a simple alligator clip or other "solder heat sink" method (putty, etc) can be used to isolate the free-hanging end from the rest of the tab/package contact joint. Again, with copper this thin, any soldering should be momentary, at most 1-2 seconds contact probably to make a good connection, so the heat isn't likely to have time to heatup the package/contact top and break the connection before the iron is removed from the wire connection being made. Anyhow, those were my post-posting thoughts.

 

[McGizmo]

David,
I meant going into the via itself with like a ball burr and just removing enough copper (think counter sink or divet) to break the connection. This would leave full pad and surface contact of LED to sink. That ceramic is a great thermal conductor!

Isaac,
I believe much of the 2X temp is due to the small LE trying to compete with thermal relief against the larger and more massive copper plate. I think it reasonable to look at the two images in terms of thermal differentials across the components but not to draw conclusions on the absolute temp of the LED in both cases. A full and clear picture may be worth 1000 words but what is the value of a partial or misintrepreted picture? :thinking:

I leave it to the qualified to draw the proper conclusions.

 

[Anglepoise]

I thought about my "solder tab" idea unsoldering itself when connecting the wires to the "tab" part of it shortly after I posted it, but 2 thoughts on that...

#1 - Once soldered directly to the package, the package itself will act as a heatsink for any heat applied to the "tab" part (open to ambient), so the temperature gradient from hanging tab part to package soldered part should be dramatic (meaning, the solder to the package itself is unlikely to heat up across the width enough with brief "end tab" soldering to loosen completely from the package (the heat being "sunk" into the unit itself).
#2 Once the "tab" is fixed to the Cree package, a simple alligator clip or other "solder heat sink" method (putty, etc) can be used to isolate the free-hanging end from the rest of the tab/package contact joint. Again, with copper this thin, any soldering should be momentary, at most 1-2 seconds contact probably to make a good connection, so the heat isn't likely to have time to heatup the package/contact top and break the connection before the iron is removed from the wire connection being made. Anyhow, those were my post-posting thoughts.

I tried your idea and it worked but did not really help at all. As far as I can see, any soldering done to the top is potentially going to interfere with the reflector.
However I have a photo somewhere of Don's new McR-17EX and it looks as if he has machined a shoulder inside the back of the reflector so when it sits on the tinned Cree reflector ring, it is proud of the connection plane.

From the photo it looks like there might be enough room to solder some 24 gauge wire to the top pads.

My next test is going to mount the LED on a machined pedestal with AA in the traditional way , and solder the + and - underneath.

I have been running a P2 , XR-E Cree epoxied directly to a 3/4" X 3" aluminum rod with a thin layer of AA. Steady state temp stabilized at 37° C ( Fluke Thermocouple ) with 350 ma going to the LED. Think that is a bit hot?