November 26th, 2006
CREE's new XR-E LED that started off a whole new phase in high power LEDs (and also enabled the low cost Seoul P4- uses the CREE die inside of it)).
(I have edited things however I see fit, to remove junk, clarify items, and such)
First off, CREE made a vast improvement in the Vf of their LEDs, and it holds quite well with increasing current. In fact, I know of no other power LED company producing LEDs with Vf's this low. Why does it matter? Power consumed is Current * Voltage. So, here is the first chart, compared with a Philips LumiLEDs Luxeon K2, their latest device:
Next, I measured the lumen output (I figure accuracy is +/-10%), vs. current:
This is also a vast improvement vs. any other LED I've tested to date, and pretty much, producing roughly 2x more light than other similar LEDs at the same current. Or looking at it another way, at the same light output, this part consumes about half the current, so your batteries will last 2x as long, and the LED will produce half the heat.
Due to the extra low Vf, and the outstanding amount of lumens coming out of the device at a given current, when you multiply these two together, the total power is often less than half, compared to most devices.
The extra high lumen output of this device, vs. power consumed, means it is more efficient, and more of you power leaves the device as light, instead of being converted to heat. Heat is the bane of LEDs and reduces both their expected lifetime and lumen output.
Yes, you can find the rare unobtainium bin of some whizbang Philips Luxeon, but they are rare, and still won't come close to matching the lumens vs. current of the CREE XR-E. CREE also makes some premium bins, this one tested here is just a run of the mill part.
For testing, the device was directly soldered to a 2" by 3" plate of 0.162" copper slab, and set on a table with no cooling air. Your results may be better, if you are more aggressive in your heatsinking.
[CPF-chimo}
Great data Newbie!
Wow, this is more than an incremental improvement - more like a leap. Looks like Cree has a real winner here. The conformal phosphor coating is icing on the cake (no pun intended).
It's VI curve looks like it would resemble a UxxH LuxIII (but with a greater light output). Luxeon did not do themselves much of a favour in releasing the K2 with such high Vfs. It was a more of one step forward and two steps back.
Paul
Correction: Oops - I was looking at the wrong point (500mA instead of 700mA) binning for the Cree. Looks like a good VxxH instead of a good UxxH bin!!! Outstanding.
[CPF-Pinter]
I also got higher values for Vf.
Cree XR-E, bin WC-P4, right out-of box, epoxied to heatsink. Measured on the emitter. Current source: nFlex.
20 mA - 2,71 V
58 mA - 2,84 V
100 mA - 2,94 V
175 mA - 3,05 V
350 mA - 3,24 V
500 mA - 3,34 V
700 mA - 3,49 V
1000 mA - 3,59 V
Last one is a little suspicious, I will check it later, maybe I was blind and missed a segment on my multimeter.
Vf's will typically vary from part to part, but your Vfs are much closer to what I measured.
The reason why they are closer, is that it may be because you measured at the emitter for the voltage, typically, when you have current going down a test lead, it causes a voltage drop, which means voltages measured at the power supply will be higher than normal.
But remember, Vf's will vary from part to part, just like the Luxeons do.
Luxeon III binning current is 700mA
Your sample looks like it would be a Philips Luxeon III bin J.
My sample looks like it would be a rare Philips Luxeon III bin H.
@ 701.5 mA and half the voltage of a Lux V (3.2716
Volts), I'm getting ~134.075 lumens and 2.2950274 Watts.
This would make it Luxeon III a VXOH bin, and a very upper end of the flux bin at that.
I don't have much time tonight.
Though, I took some photos that I thought you might enjoy:
[CPF-x2x3x2]
since u were already testing the input current vs output, didnt u take any beamshots?
Like this?
I stopped the picture way, way down, so you could see the CREE better:
Okay, I poked around a bit and found that I did find a loose reflector handy, it is 45mm in diameter.
I know that this isn't a fair comparison, but what I happened to have that was portable is a circuit that delivered 450mA to the CREE XR-E. The exact wattage drawn by this CREE at this current is 1.423 Watts.
I do not know how many mA the Fenix P1 delivers to the emitter, but I had one with a fresh battery I put in earlier today, so I used that as a reference. I just checked the current draw from the cell, and it is measuring 625mA, with the cell voltage dropping to 3.05V. This is 1.91 Watts. I suspect the converter is ~80% efficient, so the power delivered to the Luxeon III in the P1 would be roughly 1.525 Watts- but this is not fact.
I held the reflector in my hand, over the CREE XR-E and thus it isn't centered, and the photo is at an angle, as the truck was in the way. So, we have the larger reflector I used (so please keep this in mind!), and this is what I saw:
Anyhow, I need to run off to bed, maybe I can put together something tomorrow that is a much more fair comparison.
[CPF-jtr1962]
Amazing results and thanks for testing these NewBie!
To show how much less heat these make for a given lumen output let's say the design goal is 80 lumens. You can get that with a typical Luxeon III running at 700 mA. Typical Vf is roughly 3.7 volts so input power would be 2.59 watts. Efficiency is about 10% so you would have to deal with about 2.33 watts of heat.
Now let's do the same thing with a Cree XR-E. You only need about 350 mA. Vf would be around 3.1 volts so power input would be only 1.085 watts, a reduction of 58% compared to using the Lux III. The heat reduction is even more impressive. Since this part is about 25% efficient you would only be dealing with about 0.81 watts of heat, a reduction of 65% compared to the Lux III. In other words, one-third the heat for the same lumen output.
The good times are just beginning. A few years ago I was waiting patiently until LEDs finally reached or bettered fluorescent tube efficiency. That time has finally arrived. I can hardly wait to see what Cree has in store for us next year!
The flood brightness in the CREE is tremendous. Instead of a really "contrasty" hotspot with a weak flood, it has easily as strong of a hotspot, and a very strong flood. It makes for a *very* useful light, that can throw and has a flood beam that is much closer to the hotspot, imagine flood that can be throw. I found it very useful for searching for things.
Anyhow, I got my A19 XR-E (heheh- 230.00 dollars later), which has a bit of an issue. Within two hours of receiving the package, dat2zip was all over things to get the problem rectified. So, this isn't a good comparision either, as it is, the XR-E is "crippled". The issue was that the emitter was *way* off center and not flat
The XR-E was not mounted properly centered by ETGTech on the ETGTech MCPCB, and the MCPCB was not mounted centered in the Ecan:(yes, that is thermal epoxy on the emitter lens below)
And, the MCPCB was not sitting flush in the Ecan, but lifted at a serious angle:
But I bracketed the photos (different exposure levels), and the reflectors are so close in size, that it takes out the reflector factor-just remember the XR-E is now crippled in the McGizmo/dat2zip A19 flashlight:
[CPF-Chronos]
The cree looks so promising. I too thank you Newbie for the information! I bet this would be a great lightsource in a Gladius, my Minimag LED, and I'd love to see one retrofitted to my KL5 head. Hell, I bet this will be the LED to have in '07.
LOL!
It is the LED to have for 2006.
Next year will bring further improvements, trust me.
I've been goofing around a bit with the XR-E, if anyone is interested in some beam shots:
http://candlepowerforums.com/vb/sho...042#post1664042
http://candlepowerforums.com/vb/sho...044#post1664044
Many folks have talked about how utterly ice cold the CREE looks, or how it is blue.
This is usually due to a monitor that is not calibrated, or a camera that was not white balanced "correctly".
To give folks an idea, here is my A19 CREE XR-E measurements. If you look at the block labeled T, you will see that it measured 6402, which works out to 6402K:
Dave,
The one used in my A19 is clearly a different bin than the raw CREE LEDs I have received. Yes, it is an X1 bin, but barely in that bin, and doesn't appear green to my eyes. (actually it was CREE WF bin) I however, have plenty of Luxeons that are X1 bins, and are very clearly green in comparison. I have another A19 light engine that is cooler in color temperature. I hope to measure it soon.
Yes, driving an LED at other currents than what it was binned at, will cause a tint shift from what it was binned at.
Also keep in mind, that reflectors also change the tint of an LED.
Plot of lm/W vs. Current on one of my samples, directly soldered to copper 2" by 3" by 0.165", no MCPCB:
I have a second A19 XR-E light engine to check out, and obtained the following information:
Note the color temperature on this one, 7207K.
A few folks have asked me about the drawbacks of MCPCB.(this one is the ETGTech MCPCB, and keep in mind that all MCPCB are definitely not created equal!)
There are high end and low end MCPCBs on the market, and just like the difference between a 5mm LED and a CREE XR-E, they are not created equal, but both are LEDs.
A common low cost MCPCB that is used by some of the folks selling the CREE XR-E is shown in the thermal photo below.
When doing the same test with a copper plate, there is a whole world of difference, maybe I'll show you if I have a chance.
I took a known Binned WC CREE XR-E emitter and ran a color test on it.
Measurement Guaranteed Chromaticity Accuracy: x: ± 0.0015, y: ± 0.001
Repeatability: Chromaticity: ± 0.0002
[CPF-LEDninja]
What bins and what amps were the XR-Es run at in the spectrographs on posts 77 & 84? Not at all familiar with the binning - colour temperature relationship, but my guess is 7207 degrees would be outside the 'XO' bin box.
I am guessing daylight=~5500 degree=~WO bin.
No, 5500K is V0 bin towards the center of it, on the left side, which is not very popular. The X0 bin runs up to 7000K
Typically, XO (6300K-7000K), X1 (7000-5650K on the green side, above the BBL radiator "white" line), and WO (6300K-5650K) are the most popular "killer" bins (the ones that everyone on CPF seems to pay premium prices for). A few of the outdoor types like the X1 bin, which is a green tint of white, as they feel it offers better color rendering outdoors.(for myself, the X1 brings out green leaves and grass, but I am *not* usually hunting rhododendrons or fir trees).
A great way to understand K temperatures, is to simply take your monitor and push the settings button, and you will typically find settings for 9300K white, 6500K white, and 5600K "white".
The producers of the A19 will not specify which bin is used in their lights. On one of them the current at the LED itself was 800mA.and another was 940mA
Keep in mind, if things heat up more, there will be additional shift in the color.
So, if you mount it on the MCPCBs that are common for the CREE at the moment, you will see additional tint shift. Also, if you hold the temperature more constant than my copper plate, such as having more surface area to get rid of the heat, better thermal transfer, etc, the shift will be less.
Heating is also cause by converters, which produce their own heat. Some converters will produce one third of the heat in the flashlight, causing extra tint shift, where as higher efficiency converters will cause under 2% of the overall heat. Resistors also produce heat, which heats everything up.
You also have heat produced by the power source, such as batteries, which contributes to flashlight heat and tint shift.
[CPF-IsaacHayes]
yeah convertors (especially boost) will produce a lot of heat.
Most especially Buck-Boost, which produce the most heat of all, as they are in fact, the least efficient.
However, with very careful design, you can achieve 90% with a buck-boost, but not likely in the form factor of a dime, especially at 1 Amp of output current.
A boost or a buck, has one-half of the switches of a buck-boost, and this is easier to get efficient. They can hit 90% in this form factor, and with a bit more room, 96-98% efficiency is not at all impossible.
With an efficient converter, you will produce less heat from the converter, you will pull less power from the cells which creates less heat from the cells, each of which will help keep the LED cooler, and raise the LED efficiency. If the LED is cooler, and running more efficiently, it will produce less heat, and in turn will load the converter less and make it more efficient making it also run cooler, and the converter will pull even less power from the cell which reduces the heat produced by the cell, helping the LED to be more efficient.
On top of all this, your cells end up lasting longer, due to the reduced load, and you can get more power out of the cells, with less going up as heat.
Eventually, you reach equilibrium.
[CPF-Christexan]
Parallel isn't so bad, so long as it's CURRENT regulated (not voltage), and the Vf and current specs of each LED at the operating current is well within max limits (not overdriving any of them outside of ratings or thermal envelope).
You might however see visible output differences in that scenario. Adding low ohm resistors would help even it out some, yes it's wasteful, but high series voltage may not be an option, and small resistors (say 1 ohm or less) won't waste much power, but will definitely stabilize things.
If your LEDs individually at 500mA range from Vf of 3.3 to 3.6, and the LOWEST one (say the 3.3 in this case) runs at 700mA at 3.6Vf, then you are still within spec so long as you can keep the heat from escalating. Brightness will DEFINITELY vary however in this scenario.
In other words, choose the current you wish to operate at, and you have a few choices..
Option 1 (riskiest but reasonable): Find the LEDs voltage range at the desired current (say 500mA)... ranked from high to low (use 3.3 low to 3.5V high as an example). Take the highest Vf LED rating (at test current).. for example, 3.5V... now run the lowest Vf LED at 3.5V (ramp up to avoid burning it out) and see what current it pulls. If it's still within safe margins (say 600mA), then you can run them in parallel, so long as you are current regulated. (Pushing lower LEDs to match highest while keeping in rated specifications, brightness COULD vary significantly)
Option 2 (very safe): Find the range, and test the highest Vf (3.5Vat 500mA) LED, but test it at the voltage of the lowest rated LED's Vf at 500mA (3.3V)... Set the current to whatever current makes the highest rated Vf LED equal to the Vf of the lowest rated, and almost no chance of problems (Limiting the circuit to the lowest end denominator, dimmer performance likely in this case, and varying brightness of LEDs)
Option 3 (best safe performance) : Test LEDs Vf range, put resistors on all LEDs ("resistor match" them), adjusting resistance to match current levels at a given voltage (I'd choose a tiny bit more voltage output than the the top Vf, and work back from there). So put say a 0.2 ohm resistor on the top LED (just to be safe, all the LEDs should have some resistance, or none, otherwise you might have problems (the non-resisted ones would have a nonlinear resistance curve to the rest of the set)). This gives a regulator output of 3.6V to drive the top LED at 3.5Vf. The 3.3Vf LED should have a resistor of 3.6-3.3 = 0.3/.5A = 0.6 ohms
Losses to resistance (in this 2-LED case for example) would be small if you start with small enough resistance. In this example, if it were just these 2, you'od have .1*.5=.05 watts (tiny) plus .3*.5 = .15 watts. Total loss of 0.2 watts, plus maybe a little more loss from driving the regulator 0.1V higher. Anyhow, you can use smaller resistors to reduce this further assuming you can find them smaller (or make your own).
Actually that is one of the worst ideas I have heard since AWR started suggesting it for someone that was doing a headlight.
One major factor you have neglected to think about.
Temperature.
One of the LEDs will start to pull more power than the other. The result is the LED starts to heat up. This causes that LED's Vf to drop, which causes it to pull more power.
As the first LED starts to pull more current, the one next to it will begin to pull less current, and it will cool down, and it's Vf will rise. This causes the current in that LED to drop, which then diverts more current to the other LED, which makes that one get hotter.
Pretty soon, you end up with a radical difference in current between the LEDs.
A person *might* possibly be able to get by with this half baked idea, if the temperature was held constant, and the LEDs were somewhat matched, and also had resistors on them.
The situation becomes worse with these high power LEDs, as compared to the lower power LEDs.
Here are the basics on what you are attempting to do, but done with a low power LED:
http://www.lumileds.com/pdfs/AB20-3.pdf
An addendum if you have more interest in the subject:
http://www.lumileds.com/pdfs/AB20-3A.PDF
Of course, I've seen plenty of examples where designers have used your technique, and sections of road signal lamps as well as automotive rear lamp arrays, have sections burned out. No surprise here.
[CPF-Archangel]
Is that "worse" than you'd get if you grabbed ten random luxeon that were binned only by output?
[CPF-milkyspit]
Good question, and I don't know the answer... but I do know that with ten Luxeon I SWOH stars, for example, I'll typically see a low Vf around 3.12V and a high Vf around 3.23V... much tighter variance.
When you grab 10 random Luxeon SWOH stars, the H indicates the voltage bin, which can range from 3.03-3.27, with a die temperature of 25C. You are grabbing 10 random Luxeon parts from the same Vf bin. This variation will change as the die temperature heats up, since the die doesn't sit at 25C, once power is applied.
I have received Luxeon III with Vf bins of G (as binned low as 2.79V), and Luxeon III Vf bins of L, (which is binned as high as 3.99V). That is a pretty darn wide Vf variation for a product!
Not only does the Vf shift with temperature, with some LEDs it also shifts further with time.
And the amount of Vf shift is different between two otherwise apparently similar parts. I've seen this Vf shift as much as 0.3V.
I have not had a chance to test the new XR-E that much yet.
But, the previous 3XL7090 series, and the one before it, had very little Vf shift. Here are my test results for the 3XL7090:
Here is an example of the new Luxeon K2:
[CPF-Chimo]
I did a quick test on the three XR-Es I have on hand. They were cut consecutively from the same strip.
Here's a quick list of their Vfs at 350 and 700mA:
XR-E #1
2.91V @350mA
3.03V @700mA
XR-E #2
3.03V @350mA
3.18V @700mA
XR-E #3
3.18V @350mA
3.53V @700mA
Paul
From some of my thermal testing:
Here is the setup with no emissivity coating, the temperatures are *HIGHLY INACCURATE*(the low emissivities of the various surfaces make things read lower than normal
Here is the same setup, with a high thermal emissivity coating on all parts, so we can see the actual temperatures accurately (NOTE, TEMPERATURE SCALE ON SIDE HAS CHANGED!):
The parts that were mounted with Arctic Alumina were bonded under weight for over 12 hours.
The copper sheet metal that the parts were mounted to was placed upon another plate to help get rid of the ~9W produced.
As you can see, the Arctic Alumina Thermal Epoxy to the copper plate nearly works as well as soldering directly to the plate. The ETGTech MCPCB doesn't work nearly as well as the other two options.
The AA Thermal epoxy XR-E has a dome temperature of 42 degrees C.
The direct solder XR-E has a dome temperature of 39 degrees C.
The ETGTech MCPCB has a dome temperature of 54 degrees C.
Notice the wire going from the MCPCB to the direct copper soldered one, and how the wire is conducting heat from the MCPCB mounted XR-E, and how it fades along the wire length.
Keep in mind, these are package temperatures, on the outside, and the die temperatures are even higher.
Ran at a different current, but used for illustration, without emissivity coating:
I just noticed Fraen has optics available for the XR-E, and the older Seoul P3 an P5 parts:
http://www.fraensrl.com/prodinfo.html
Other companies offering optical solutions for the XR-E are G&L, Fraen, LEDIL, Carclo, Khatod, Polymer Optics:
http://www.cree.com/products/xlamp_part.asp
NIST (National Institute of Standards and Technology), tested the CREE XR-E for color and Luminous Flux (lumens), and lm/W, back in October last year:
http://www.cree.com/products/pdf/NI...%20Document.pdf
Triad Flashlights has chosen the CREE XR-E for their product:
http://www.cree.com/press/press_det...i=1168609850571
Surefire has also chosen the CREE XR-E for use in several of their flashlights:
http://www.cree.com/press/press_det...i=1167920349046
One of the largest distributors in North America is now carrying the CREE LED line, Arrow Electronics:
http://www.cree.com/press/press_det...i=1162560893953
The parts, and the pricing for the CREE LEDs is shown at the Arrow North America website, just type in XR7090 in the search block on the left, there are three pages of parts already(they also have a *very* low cost five pack of XR-E parts coming out very shortly!!!):
http://www.arrownac.com/
There are plenty of other distributors all over the world, you can find them listed here:
http://www.cree.com/products/xlamp_dist.htm
CREE was chosen as one of the of 300 companies chosen for the new Ocean Tomo 300 Patent Index, the first equity index based on the value of corporate intellectual property:
http://www.cree.com/investor/patent_index.htm
In response to questions on the die:
Well, I'd refer you to CREE's datasheet for CREE's EZ1000 die for starters:
http://www.cree.com/products/pdf/CPR3CR.-.pdf
If you look at the die, you will notice it is one of the thin InGaN active layer types, with a metalization layer right underneath the active area. On the top surface, you will find a rough patterned surface to also help extract light from the thin active layer.
In the XR-E, you will find the very robust and highly thermally conductive SiC ESD diode underneath the die, and then this is also soldered to the package substrate.
Then if I was you, for further information, I'd be looking around on the various patent search engines.
[CPF-FloggedSynapse]
Hopefully they're all of higher caliber than the common 5 mm LEDs out there. Apparently the XR-E uses the latest (EZBright1000) core, while the XLamp uses the older (XB9000?) core.
Asked before, and I'll ask again now: does anyone have any additional information on these cores? Considering the XRE is new, has set an efficiency record, and considering the competition it's not surprising there's a paucity of information on the geometry and functional layout of the die. Still curious though.
Since you asked, I thought I'd also enclose some additional photos for you. This should help several things to become more clear.
CREE and LumiLEDs use metallurgical and solder bonds at several points in the construction of their LEDs which helps the resulting part to be robust from the die and thermal standpoint. Others do also, but fyi, the Seoul P4 does not. (in the Seoul P4 they are epoxied on, like you find in low end 5mm LEDs) LumiLEDs and CREE both utilize a robust ESD diode, with CREE utilizing an extremely robust and highly thermally conductive Silicon Carbide ESD diode.(in the Seoul P4 they have a small cheap ESD diode off to the side, if you look carefully)
Here is the top side of the Luxeon III:
Here is the underside of the Luxeon III die that I unsoldered from LumiLED's ESD diode, the round and oblong bumps are the solder points and thermal transfer points:
This is the ESD Diode that the Luxeon III die is soldered to:
Cree takes a bit of a different approach, which results in better thermal transfer than the Luxeon III and better than even the new Luxeon K2. Cree uses metallurgical bonding as well as solder bonding. Try as I might, to the point of melting the gold bond wires, I have not managed to get CREE's EZ1000 die off of their SiC ESD diode. Later in the series of pictures, you'll see where the part is soldered at, as well as a demonstration.
Unsoldering the highly thermally conductive Silicon Carbide (SiC) ESD diode from the substrate:
Unsoldering series above end.
Third in series of photos, light extraction coating on Thin active layer of InGaN:
End-Coating series
[jtice]
VERY cool !!!!!
I have always loved macros, really another world once you get a good look at tiny things like that.
Lot of good info there, gives a better understanding of what goes into these,
and how they are actually built.
~John
Thanks Jtice!
I just noticed that around midnight last night, before bed, I was in a hurry to get these finished up, and I got a few labels on some of the photos wrong.
I'll fix them tonight.
For now, this one is correct:
[PEU]
waaay cool photos, THANKS!
Pablo
[CPF-Doug S]
I'll second that!! Great stuff, Jar.
I note that you have identified the surface texturing as a coating. Does this in fact appear to you to be a coating as opposed to the result of an etching operation?
Also, the die is not centered on the SiC diode so as to accommodate the bond wires. Some of your photos make it appear that there is also an angular rotation of the die relative to the SiC. Is this for real or is it just an artifact of how the photo was taken? If for real, it would seem to indicate a bit of slop in the die placement step of assembly.
Yes, it is a coating that is both over the die and over the negative current spreader.
Yes, the off center placement of the LED die on the SiC ESD diode is in fact to accommodate the positive bond wire connections.
Part of the apparent magnitude of rotation is due to how the shot was taken, but there is some actual rotation. The rotation doesn't hurt anything, and it does not actually extend beyond the Silicon Carbide ESD diode edge (such as the die hanging over the edge of the SiC ESD diode- I have never seen that actually happen). I've seen an occasional very slight sliver of the underlayment area where the metallurgical bond is, caused angular rotation as you see here. Remember, we are at pretty extreme magnifications here, in some cases exceeding 1000x
[CPF-Gryloc]
Newbie,
That is some very nice work there! Good job with getting up close.
Thank you.
[CPF-Gryloc]
How difficult was it to desolder the carrier? Do you think that you could remove the ESD carrier and die from the substrate without mutilating the phosphor and permanently damage the die (decrease its output and such)? If I would heat it quickly then take away the heat, will there still be some damage? The XR-E was made to be re-flow soldered.
If you are used to working with stuff this small it can be done. Re-attaching the gold bond wires (assuming you sever from the substrate) is tricky, as gold dissolves into solder. You can solder to the gold bond wire, but you have to be hella fast about it. Just take your time, and maybe buy some lower cost red ones just to practice technique. Don't apply force to anything, the die will chip and usually be destroyed. Also, if you get any solder up the side of the SiC or the LED die (say like you used the CREE EZ1000 die from a Seoul- its silicone is hard and makes doing the phosphor intact thing much harder) then you will often end up with a short. I'd utilize non-metallic handling tools, lowers chance of chipping the die. I know in fact it can be done, been there, done that. Also, the gel will need to be re-protected. With the Seoul P4, you risk a pretty good chance of severing bond wires due to the hard silicone they utilize, fyi.
[CPF-Gryloc]
What did you mean by "metallurgical bonding". Where was this applied, versus the solder bonding?
Essentially it is a metal to metal bond, that can be accomplished a number of ways. No solder, no epoxy, and in this case, *extremely* robust and rugged.
[CPF-Gryloc]
I was thinking that I can carefully cut the bond wires at the board, then de-solder the whole die and carrier. I would have to use an exacto knife to lightly press on the carrier to push it aside, of course.
See my comment on non-metallic above. Once you are in there, you'll soon realize how small things actually are. The LED die itself is just under 1mm by 1mm, pretty close to the size of the Luxeon and OSRAM die.(in fact, a number of the OSRAM Golden Dragon devices have used die made by CREE) The lens magnifies the apparent size dramatically.
[CPF-Gryloc]
Did the phosphor coating melt in your first pictures? I understand if you cannot remove the die from the carrier. Thats quite alright.
No melt. That is the remaining gel over the phosphor.
[CPF-Gryloc]
I was wondering this because maybe I can sacrifice 4 P2 pr P3 binned XR-Es and fulfill my desires with making a mini cluster (as Newbie understands). Then I can attempt to re-solder them. Then a tiny dab of solder can attach the bond wires, too. It would be tricky to do this, but that would be awesome! This thread has made my day and helped get a better understanding of the Cree innards. Keep up the crazy work...
-Tony
I'm very glad you liked the photos and found them useful. It makes the effort needed to format them for posting on cpf worth it.
It can be done, but how well things will hold up, that is more up to your soldering skills.
Oh, and chill out before you do all this, and make sure you have plenty of time so that nothing is rushed. You might want to skip your morning coffee/tea/coke before attempting the "surgery"
Good Luck!
[CPF-srvctec]
Simply amazing photos, Newbie! Macro is my favorite kind of photography.
What in the world did you use to capture all of those macro shots? Almost looks like it was done with a microscope!
Some are just with a Canon S3 IS camera, some are done holding the S3 IS looking into the microscope eyepiece. (unlike many cameras, if you have enough light, you can actually set something down upon the lens and take a photo of it!
Updated the naming errors I noticed.
Thanks, my pleasure.
If you'd like to see the Seoul P4 that uses the CREE EZ1000 die, here is the direct link to the photo posts:
Seoul P4 page with more die photos
[CPF-FloggedSynapse]
NewBie, cxcellent! Thanks for taking the time to throw those images out. I don't think most people realize the scale of the die.. I mean it's less than a mm square (right?). The LED looks, and operates, like a solar (PV) cell in reverse.
It's good the see CREE is serious about trying to keep the actual p/n junction cool. It's very difficult to measure the junction temperature directly. So I imagine a cheaper implementation might appear to run cooler, but that would only be because the design does not let as much heat flow out of the working junction. So the CREE parts may appear to run hotter, but that's simply because they have better thermal management, while the cheaper die actually has a warmer working junction. Very very important for the reliability of these devices - they must stay cool as possible for long life and reliability.
Does the SiC ESD serve two purposes - static protection and heatsink/thermal path? I'm curious what CREE did to improve the efficiency of the XRE - was the actual quantum efficiency of the junction improved, or is it more a matter of changing the die construction so more light can escape? Don't imagine anyone can shed some light on this (bad pun intended)?
You might find this interview of interest. Shuji Nakamura was one of the pioneers in the development of high powered blue LEDs and violet laser diodes:
http://www.sciencewatch.com/jan-feb...b2000_page3.htm
Yes, the LED die is 0.98mm by 0.98mm and the high thermal conductivity and robust Silicon Carbide ESD diode is slightly larger.
[CPF-Gryloc]
Newbie,
I have a few questions for you. I hope that you understand what I am talking about...
I didnt think of this before, but how is the ESD carrier wired? I know the top of the die has two negative contacts, and the bottom side of the die, that is metallurgically bonded to the carrier, is the positive side. What about the carrier? There is a bond wire on the top of the carrier that goes to the positive copper pad, so is that entire side electrically connected to the bottom (+) side of the die? Then what about the bottom side that is soldered to the same positive copper pad? Is this soldered joint just for thermal conductivity? Is the carrier of the XR-E similar to Lumiled's carrier? I mean, is both the positive and negative contacts of the ESD diode on the very top side? Like this:
Originally Posted by NewBie:
The reason I ask is because if the very bottom side of the ESD carrier is electrically neutral or attached to the positive side, I now know how to design the solder pads.
If they are neutral, I can have one larger pad for all four dies to be soldered to, with the bond wires going off to the side to their own pads.
If the bottom of the ESD carrier is the exact same as the positive (like after the actual ESD diode part), then I could really do away with the single little positive bond wire. With this, I can create four separate pads that attach to the negative side of the next series connected LED, and only have two bond wires to have to attach instead of three. This is how I designed the first LED module. With that drawing, I didn't have to worry about the ESD carrier and I just had to solder the positive side of the die to the pad for electrical contact.
If the bottom of the ESD diode is not neutral, but it also is not the main positive contact for the LED and ESD diode, then things will be more difficult.
Finally, do you know how the SSC P4 is wired? I see the negative wires disappearing to the side to connect to the negative lead. The bottom side of the die is attached to the aluminum slug, which is attached to the bottom side of the little diode cube. Then the top of that diode has a bond wire disappears and is attached to the positive lead. So does that mean if you attach the positive power wire (from the circuit or battery) to the aluminum slug, it bypasses the ESD diode and the whole LED is susceptible to damage?
This is interesting stuff. I always liked thinking and working on the micro scale of things. Thanks...
-Tony
The bottom side of the ESD diode in the XR-E is minus. Topside of the Silicon Carbide ESD diode is positive as is the bottom of the die.
Wire runs from common junction to + substrate pad.
Two minus wires run from top side of die to minus substrate pad.
Seoul P4 has die epoxied to slug. A tiny ESD diode that sits off to the side is attached to the same slug. This is why the slug is the + connection.
Three wires, two from the Seoul P4 CREE EZ1000 die minus, as well as one from the ESD diode run to the minus lead.
Then if you look carefully, there is yet another wire, the fourth one, which runs from the positive lead to the slug. If you happen to blow this wire, you can still power it by attaching your + wire to the slug.
No bypassing of the ESD diode, unless you cut into the dome and cut the ESD diode wire, but these die are rather static sensitive, so I'd not recommend doing that.
Be careful with the Seoul P4, don't push or load the dome, you can sever the bond wires on accident by the sheer forces, since a hard silicone instead of a gel was used inside you don't get much relief from the forces.
[CPF-Gryloc]
Ahh, thanks. That makes perfect sense. I forgot that the ESD diode was in parallel with the die. It never looked like they were parallel just by looking at the internals of the package. Interesting setup...
Great. Now I see how to wire this thing up. This will be tough because I have to deal with three bond wires and the designs of the module will be far more complex.
I know the ESD diode made of Silicon Carbide has great thermal conductivity, but I wonder why Cree doesn't solder the dies straight to the pad. It appears to have plenty of real-estate on the board under the dome, compared to Lumileds parts that have very limited space. I was just thinking.
I see that fourth bond wire in one of your closeup pictures. Interesting! What is the power handling of a single, average length bond wire? It is nice knowing that if the fourth bond wire fails, I have a way to power it back up.
My understanding...
Different materials have different co-efficients of thermal expansion (CTE). So, when to items of differing material change temperature (lets say raise), they actually expand. But materials made from different "matter", expand at different rates. So, when you bond two items with different expansion rates or different CTE's, you get a lot of stress between them. Usually the bond fails, but sometimes you can get failure of the materials near the interface, or you cause enough stress to fracture it in other areas.
The SiC matches the CTE of the InGaN fairly well, and falls between the solder/substrate, and the InGaN, so it reduces the stresses at each interface.
There are also some additional advantages for crystal structure, or lattice, and growing the InGaN on SiC substrate, instead of sapphire.
In reality, the bulk material is Silicon, but is also Silicon Carbide. An example is the diagram at the bottom of page 1:
http://www.cree.com/products/pdf/CPR3CC.00c.pdf
Another example:
http://www.creelighting.com/products/pdf/CPR3DC.000.pdf
"Based on Cree's EZBright 1000 LED chip, the XLamp 7090 XR-E is produced on a silicon-carbide (SiC) substrate that has an indium-gallium-nitride (InGaN) epitaxial layer grown on it (Fig. 1). "
http://www.elecdesign.com/Articles/...3982/13982.html
Take a look at the patent here:
http://www.freepatentsonline.com/6630690.html
They can directly grow the InGaN right on the SiC.
I don't think the high thermal conductivity of the rather thin SiC ESD diode that CREE uses in the XR-E is much of the equation as far as overall thermal resistance. See my comments later, about comparision to metal.
One would want to look in other areas like the ceramic substrate, which can be made from a variety of fillers.
Notice the low thermal conductivity is silica (silicon) and Alumina.
http://en.wikipedia.org/wiki/Alumina
Notice how the chemical formula is Al2O3. Turns out, that sapphire (mineral glass) is also aluminum oxide. In the crystal form, thermal conductivity of sapphire is only 33% of Silicon Carbide. Silicon Carbide also has 2x the thermal conductivity of pure Silicon.
"At room temperature, Silicon Carbide has higher thermal conductivity than any metal."
http://www.cree.com/products/sic_sub_prop.asp
CREE is a leading producer of Silicon Carbide, and Silicon Carbide components, so it makes sense that they'd make their own ESD diodes in house. It is a very tough, and thermally conductive material, and is one of the hardest substances out there, and also performs quite well at elevated temperatures. SiC also has a very high electric field breakdown strength, and a very high maximum current density. It also has a low CTE.
There is a grade of SiC that CREE makes, that is colorless, and is used as a diamond replacement, even in jewelry. I forget the trade name for it off-hand...ah, here it is, Moissanite. It is produced by the C3 division of CREE, the only source of gem grade Moissanite. Even the thermal probe test, used to detect cubic Zirconia, is often fooled by Moissanite. One of the neat things about this material, that differs from diamond, is that it creates double refraction, and demonstrates birefringence. This causes a doubling of the facets cut in it then you look into it. It also has a dispersive power that is 2.5x greater than diamond and creates extra live "fire" than many feel is more beautiful. They are still expensive though, at 569.00 per carot or so.
With diamond at a hardness of 10, SiC at 9.5 Mohs is pretty darn close. The Mohs scale is a non-linear scale.
Main site
Mirror:
Main site mirror