Regarding the thinness of the 20mm copper star, the star's mass is less important than how fast a conduit it is to the heatsink beneath it, right? My expectation is that eliminating the thermally insulating laminate layer between the LED's center pad and the star, by soldering (the center pad only) directly to the star should make for less thermal resistance between the LED and the heatsink than any other 20mm star where laminate sits between it and all of the emitter.
If I were designing (and capable of manufacturing) everything from scratch, then obviously, there are more effective designs than the star, however, that requires far, far more effort than a simple de-solder, scrape and re-solder operation. I know my limitations. Tweaking the existing design is my forte, custom manufacturing something entirely new is not. I've tried my hand at creating completely custom parts with what tools I have at my disposal, and the results are seldom more rewarding than the results with the easy tweaks.
Laird Theramgon Tlam SS 1KA04 or DS 1KA06
That said I have seen the data sheet for the Opulent thermal prepreg (many years ago) and I remember it being very good.
Before I started using any volume, I did my first ones the old fashioned way. I asked nicely for a bare board sample, and like any good hobbyist I rolled my own using my trusty printer and etched my own.
"What works with one LED will work with another." .... not true. I regularly use FR4 with vias for Luxeon Rebels and I used to use it for Cree XRE. I never use it for Cree XPE/XPG. Rebels and the XRE have a large thermal path that I can take advantage on on an FR4 board. On the XPE/XPG, there is almost no way I can spread the heat out on the surface, so I need to have a low thermal resistance downwards first and foremost.
"I am using what has started to become an industry standard for LED construction." ... and what would that be? Since last time I checked, LED on FR4, metal core, and direct thermal connection to the heat sink are all regularly practiced though you run into more patent issues with direct thermal contact unfortunately.
How do I know what those meters do? I asked! No no yesterday, but when I was looking at meters. They do not adjust for spectrum or color temperature. They simply adjust the gain (in s/w) based on a calibration done with standard sources .... which is great when you are measuring standard sources but no so great when you are doing other things.
You are contradicting yourself. You claim to be trying to achieve "the best possible". So why even discuss circuit boards? Best is direct solder and/or ultra thin high thermal conductivity glue direct to copper or even better a graphite/copper hybrid.
Yes seriously! Showing experimental results without understanding what they truly mean is useless. I did work with someone who had ruled out FR4 as being no good. What was really no good was his implementation.
Will there be an improvement between ablated ceramic and non ablated ceramic? ... of course there will be! Yes you can shave off the ceramic, hopefully you do not damage the part, hopefully you create the exact same thermal interface conditions with the non shaved and shaved version, etc. OR ... you could find out the material (or assume it is some aluminum/beryllium oxide derivative), find out the characteristics for that material, estimate the width of the thermal path, and then calculate fairly accurately how much improvement you will get thermally based on how much material is removed. That may sound complicated, but it really is not and it would not be susceptible to measurement error such as not having consistent mounting to the heat sink. The calculations will tell you it is worth doing (or not) and then you test it out. If it is not better in the test, then it is likely the test implementation, not the concept. However, if you just did tests, you could assume it was no better.
My glass is niether half full nor half empty.
I have worked with many good "feel" engineers and they are great when things are simple, easily understood and when you were not truly pushing the boundaries. However, their lack of proper methodology ususually was their (and ultimately the companies) downfall when things got complex and/or difficult.
Trust your experience ... I do too. But I am not arrogant enough to think that I do not make mistakes and/or I can't find better ways of doing things .... just like my GPS tells me often of routes I had not considered .... yet my experience sometimes knows that traffic will make certain of those routes unviable.
What was that assinine comment about the K2? In fact, older gen LEDs would benefit more from proper heat sinking that many new ones. Older LEDs generally had lower max Tjunction and they had worse output response w.r.t. temperature. The real change with new LEDs is that they are on small packages not much bigger than the die itself and hence you no longer have what is essentially a copper heat spreader. So yes now, more than ever it is important to get the LED connected to something that can spread the heat out as quickly as possible since your contact surface area is so small.
I agree, the thickness of the copper is pretty meaningless for what you are doing. It's all about reducing the thermal resistance between the LED and primary heatsink and the air. \
That said, I think that SAA's complaint was about the copper circuit layer. Thicker surface copper will spread the heat more sideways so there is less resistance downwards into the metal core. For some LEDS, that can be highly advantageous, for some it does not make as much of a difference.
YES ... let's do some calculations!
I don't know the material, but I will assume it is relatively good, 0.5 C/cm2/watt.
Assume Cree XPG/XPE which has a thermal pad 4.4mm2. I am going to assume some reasonable heat spreading and say the effective heat path is 8mm2. If I knew the copper thickness I could make a better calculation.
So the thermal resistance is going to be 0.5 * 100/8 = 6C / watt. I would say that is best case and it could be as bad as 0.8 * 100 / 6 = 13C / watt. Most likely it is in the middle. If you are pumping 5 watts into LED, a 30-50C decrease in LED temp may be possible.
You are correct that mass has nothing to do with it. A heatpipe has very little mass but can transfer heat exceedingly efficiently. I say go right ahead and try your idea. If you can do it right and not gouge the underlying copper you will definitely see improved performance. I just don't think it will be very easy.
It's a pity that a thread, where a discussion on different aluminium PC boards took place, is lost. There I wrote about what level of thermal conductivity one can achieve if he bothers to make PCB design of his own. Here in Moscow local manufacturers offer different materials for PCB design, and the best combination available is like this: Al1100 alloy with 222 W/(m*K) thermal coductivity, 0,05 mm thin insulation layer with 1,8 W/(m*K) conductivity and copper layer 0,07 mm thin. Thickness of copper layer matters when you are trying to enlarge effective area for heat dissipation through insulation layer which has the least thermal conductivity among named materials. This picture illustrates difference between poor (left) and competent (right) design.
I'm too lazy to make calculations again, but I remember that total conductivity estimation was near 2 W/K for XM-L on 2 mm thick PCB 22 mm in diameter. It could be much less if I ordered PCBs with 0,140 mm or even 0,35 mm copper layer, but there were some technological difficulties.
So, looking at Lux-rc's attempts of optimising thermal qualities of PCB, I thought that they are not worth charges.
Last edited by vaska; 04-02-2011 at 03:03 AM.
" That being - decreasing the thermal resistance increases the performance of the LED. "
But then you can't argue with that now can you.
Second, "the best possible" is different with each build. Different lights impose different design restraints and they may be forced to use something other than a heatpipe. If it is the case that they will be beholden to the use of mcpcbs then they need to know what is the "best possible" for that situation.
I would agree with you if we were designing flashlights for a virtual world but we aren't. You say calculating first would be superior because you remove the possible error or deviation that can occur in real life application but ignore that the world we live in and the products we use are based on actual implementation where there can and are variances in our results. Therefore the best way to know real world results is to make real world tests.
I get what your saying about the fact that calculations can show potentially bad implementation but I really think you are presuming that I don't know proper implementation. You'd be wrong. I also think you are severely over thinking our hobby. Again, on an industrial level dealing with very complex components sure. Then I would agree with everything you are saying. Not here however.
Vaska, good thermal materials for boards are available up to 7-8 W (m*K). Your board suppliers should be able to order in for you.
Flashlight designers? .. no .... LED Luminaire circuit board designers ... YES.
I never promised to post a link to a star ... but I did show where YOU could get board material and I assume from your posts you are capable of making your own boards. That said, if there was enough interest, having some truly high performance stars made up would be a worthy pursuit, though if I did that, I would likely not use board material at all but direct go direct trace on copper with Anotherm. Not sure how much people would be willing to pay for high performance bare stars though.
Just to report back that I received the DX copper star XML a short while ago, just got some time to play with it.
The led was visibly lifted from the board, there's a visible gap on one end, bad news if one want to use it "As is", may perform worse than a well done alu. star due to the "gap". Cutter leds are usually re-flowed on to the star with better quality job.
So I proceeded to a re-re-flow, added a little bid of solder paste to the gap side,
used a heat gun & managed to "close the gap".
Tested it on a big copper heat sink running between 5-5.5 Amp, so far so good.
Last edited by ma_sha1; 04-04-2011 at 02:13 PM.
BTW, this wasn't my first time, my 1st time using this technique was trying to mount
triple SST-90 onto copper, I ended up killing 2 out of 3 SST-90s.
Last edited by ma_sha1; 04-04-2011 at 04:03 PM.
The best way for me is a oven like this
, using the top side , i don't know how do we call this in english.
On the lowest setting, it is getting around 150-200 degrees, depending on the point you are measuring.
I used an IR thermometer to measure this.
Is is the best way to reflow a led. You have complete control and stable temperature, using two nice and long needles it is very easy.
To add, finding the right way to reflow, i have managed to destroy 2 mce,1 sst90,2 sst50, one xm-l and a few xr-e, fortunately no xp-g xp-e!
I too use the heatgun and needlenose method at times and I have no problem getting it down and pushing the LED down before it sets up. For most production work I use the lab oven as is specified in the data sheets.
Last edited by saabluster; 04-05-2011 at 02:07 AM. Reason: spelling..grr
There is NO surface tension force that will naturally squeeze out any excess solder from between the two surfaces.
Since the best solder has a Thermal Conductivity far worse than Copper, it's essential that the two Copper faces are in as close a contact as possible, with solder only filling gaps cause by surface roughness. This will ONLY happen if press the two parts together.
If you follow EBAY, you can find surplus temp controlled hot plates that are somewhat accurate and work quite well. Now If I could just find a really good way to put down thin paste reliably for prototyping, then I would not need to press out the solder which seems almost standard practice when using paste manually. I sometimes find that just tinning the pad and using flux gives me better control.
Do someone perheps know how the metall heat contact plate (it's grind away in saablusters thinning process) are attached to the ceramic base of these kind of emitters? Are they glued on whith something? In that case that glue must have some thermal resistans that you get rid of too.
Last edited by Neondiod; 04-07-2011 at 03:39 AM.
Now, saabluster, you just need to sand an XR-E down to the reverse-polarity diode and test again!
@vaska (and lux-rc/serge), wow, I hope your implementation of that metal heat spreader makes it to production!
@TJ, now with copper metal stars, this idea seems plausible enough to try! I say go for it!
i can no longer do prac but am still engrossed by these threads.
its a shame that cree dont supply the led on a larger piece of ceramic.
anyway when you mount the ceramic (what you guys call reflow ), which is just under 1 mm thick? , would it help to build a bead of heat conducting goo around the edge the ceramic tile, effectively conducting some heat thu the sides of the tile ? being pretty close to epicenter and all. if that tile is 4mm by 4mm by 1mm thats 16 mm sqared on the base and 16 mm squared on the vertical border surface.
gee i hope this isnt already in this thread somewhere .
Just want to provide some feedback on the DX copper star XML, it didn't work,
When I put a light together, it wouldn't lit up, some times emit a blink of light.
I ordered & received another one, same thing, initial test with a battery fine, after soldering
leads onto the star, it no longer work, again some rimes emit a blink of light.
I am down two for two on this Copper start XML, so I am out.
Speaking of Cutter vs other MCPCBs for the XM-L
from a comparison I did.Did a comparison of the Cutter XM-L boards and KD XM-L boards. Mounted a U2 2S XM-L to one of each PCB (reflowing onto pre-tinned pads with RMA flux)
Screwed the pcb to a heatsink I pulled from an old mobo for the mosfets, using thermal compound(one at a time of course). Ran each LED at 3.73A for 10 minutes to reach a somewhat steady equilibrium. I recorded the Vf at the power supply and light output on a lux-meter 6.5" away, neither the clamp for the heatsink nor the lux-meter moved between tests.
Even soldering the wire leads to the PCBs with them mounted to the heatsink, I noticed the KD board was harder to solder. I also noticed the heatsink became hotter much faster with the KD compared to the cutter PCB (hotter faster is good!)
Initial lux 11.65k 9.17k
T+10 Lux 10.31k 7.47k
Lux drop 11.5% 18.5%
Initial Vf* 3.78v 3.73v
T+10 Vf* 3.70v 3.63
Vf drop .08v .1v
*The Vf was at the power supply, not at the LED, so there is ~6ft of wire that is probably a touch small for 4A between the LED and the voltage measurement.
I think the results pretty much speak for themselves, the KD boards are the better of the two
I'm not sure how much difference the boards made on the initial lux readings, and I'd say the % drop would be more meaningful then the initial lux difference, but FWIW the LEDs were right next to each other in the cut-tape from the same bin, so I couldnt see them varying that much.
for those looking to do more then a little bit of reflow work,
http://www.circuitspecialists.com/prod.itml/icOid/8010 may be of interest, very nicely priced at $89, powerful enough for doing a star on its own, and regulated temperature so you dont burn your PCBs or LEDs by leaving them on for just second too long. I use one for SMD work at home also...
http://www.amazon.com/Steinel-34890-.../dp/B002NKM1MK if youre feeling more spendy, and what I use at work for bigger stuff, and would work fine for LEDs and smaller stuff alike, other then it being a bit large. Ive reflowed LEDs onto MCPCBs with this, while they were attached to the heatsink.
Last edited by CKOD; 04-27-2011 at 11:21 AM.