Die luminance, emittance and advanced die heatsinking

[saabluster]

Re: Die luminance (surface brightness, important for throw)

Uhhh, SiC ??? :duh2:

SiC has a thermal conductivity of 120 W/mK.. Copper has 401 !

http://www.accuratus.com/silicar.html


Regards,

Ra.

SiC is an excellent thermal conductor. Heat will flow more readily through SiC than other semiconductor materials. In fact, at room temperature, SiC has a higher thermal conductivity than any metal. This property enables SiC devices to operate at extremely high power levels and still dissipate the large amounts of excess heat generated.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

SiC is an excellent thermal conductor. Heat will flow more readily through SiC than other semiconductor materials. In fact, at room temperature, SiC has a higher thermal conductivity than any metal. This property enables SiC devices to operate at extremely high power levels and still dissipate the large amounts of excess heat generated.

That's obviously a typo !!! At the Cree site, SiC is compared to other semi-conductor materials, not metal.

Their specs:

3.0-3.8 W/cmK @ Room temperature equals 300-380 W/mK So just below copper...

Edit: According to wiki, the highest possible value is for SiC is 460 W/mK, that would indeed be higher than any metal !! Still 1540 W/mK to go to meet CVD diamond...


All the best,

Ra.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

Just curious...so is the primary limitation regarding driving an emitter conducting the heat away or is there something inherent in the die design?

Meaning...I understand when driving within an emitter's spec that conducting heat away is crucial to maintaining that output but when pushing above spec is the main limitation still just conducting heat away or is there an internal limitation to the emitter itself? I assume the answer is yes but I'd appreciate a little clarification.

The heat from the die needs to travel through the base of the led before you can take it away with a heatsink. Heat only flows to colder regions, with a heatsink, you create that colder region. The colder that region is, the quicker the heat flows away from the die.
The limits are in the thermal conductivity of the material (s) between the die and the heatsink (because you cannot change them..). But that's not a hard boundary: Better (colder) heatsinks always result in lower temperatures of the die, but at the edge of possibillities, the differences will be small.


Regards,

Ra.

 

[Walterk]

Re: Die luminance (surface brightness, important for throw)

@Saabluster; I think you found your wonderflonium. From Wiki:
While rare on Earth, silicon carbide is remarkably common in space. It is a common form of stardust found around carbon-rich stars, and examples of this stardust have been found in pristine condition in primitive (unaltered) meteorites.
The low thermal expansion coefficient, high hardness, rigidity and thermal conductivity make silicon carbide a desirable mirror material for astronomical telescopes.
Moissanite has become popular as a diamond substitute, and may be misidentified as diamond, since its thermal conductivity is much closer to that of diamond than any other diamond substitute.

Back on topic, die surface and brightness: using my left eye, my petzl and caliper;

XRE R2 - 1,85 x 1,85mm ( authentic, bincode will follow)
XRE R2 - 1,8 x 1,8mm ( bincode unknown, from 5$ headlamp advertised as R2)
SST90 - 5,0 x 4,85mm (bin unknown, marked 004022)

The dome of the SST90 made the die look having a belly on the sides.
Counting pixels might be more suitable, but I like caliper better.

 

[Mettee]

Re: Die luminance (surface brightness, important for throw)

very cool topic and interesting reading, I will try to keep up:thinking:

 

[Dr.Jones]

Re: Die luminance (surface brightness, important for throw)

Although you may have inferred it you did not directly state. Are you saying to use a lens to project an image of the die or not?

No (at least no other than the camera/microscope lens for apparent die size measurement). That 'apparent' die size is really just the size of the die as you see it (enlarged) through the dome, and the intensity is bare emitter forward intensity without any additional lenses, just LED and lux-meter.

Consider a bare emitter (no dome, but think of adding a dummy glass pane to get the same losses), and you measure illuminance (lux) in 1m distance (no lens in between) to get it's forward intensity (cd, same as lux in that distance). To get the luminance (surface brightness), you divide the intensity by the die area. That's what we want.

Now you have the dome, which is a small collimator. It creates an enlarged virtual image of the die, which can be seen as a new, 'virtual' light source. It gives more beam intensity (due to the bigger virtual emitter area, or equivalently, because it's a bit collimated), thus the lux-meter will show a higher value. At the same time the viewing angle is reduced, but the luminance (surface brightness) of that virtual light source is the same.
Luminance is the higher intensity divided by the bigger virtual (apparent) die area resulting in the same luminance as above.

Since you usually don't want to shear the dome off and add a loss-equivalent dummy, just use the second method with the dome, yielding the same result.

 

[saabluster]

Re: Die luminance (surface brightness, important for throw)

Edit: According to wiki, the highest possible value is 460 W/mK, that would indeed be higher than any metal !! Still 1540 W/mK to go to meet CVD diamond...

Yes it doesn't match diamond but I was just looking at all the alternatives and trying to find something better than silver. Every little bit helps. The thing is even CVD diamond can't compete with with a cheap heatpipe that you can get off ebay for a couple dollars. Of course it is not as exotic as the diamond and diamond would allow one to directly mount the die without having to have the heatsink be electrically active.

How do you plan to mount the LED to the diamond? Epoxy I assume? What kind? I have a suggestion in that regard.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

Yes it doesn't match diamond but I was just looking at all the alternatives and trying to find something better than silver. Every little bit helps. The thing is even CVD diamond can't compete with with a cheap heatpipe that you can get off ebay for a couple dollars. Of course it is not as exotic as the diamond and diamond would allow one to directly mount the die without having to have the heatsink be electrically active.

How do you plan to mount the LED to the diamond? Epoxy I assume? What kind? I have a suggestion in that regard.

Ok, only because this post will still de interesting for the members who want to stay on topic, I'll answer this..( because it will have suggestions for reaching the highest surface brightness )

I got my diamond for free !! So much cheaper than any heatpipe.. You think commercially, you want to sell lights. That makes the use of diamond a no go. I don't think commercially, I just want to create another 'one of a kind' superlight..


Back to topic:

@Dr.Jones: (cc walterk...)

I don't know if you already did the following during die-measurement: You didn't mention it in your first post..

Yesterday, I effectively measured the apparent die-size of a Seoul SSC-P4 led: 1.77 mm2

This is how: Any type of Li-Ion cell with a 10k-Ohm resistor makes it light up faintly, now, with a microscope or eyepiece, you can clearly see, and measure the (apparent) active region of the die. That should be close enough for our needs I think..

BTW: Using a diamond disk, yesterday, I safely pushed the Seoul P4 to 130 lux@1m !

When I do the math whith a 30mm TIR-optic, It would give a mere 49,250 cp !!
With a 75mm TIR optic: 325,000 cp !!!

I have no reason to doubt these numbers: All my former TIR-based ledlights almost exactly behave according to these theoretical calculations..

But there is more: (answering part of Saablusters question)...

I found that indeed it's very important how to mount the led: I nearly blasted a few led's to 'the next world' by not paying (much) attention to this..

Thermal paste has a thermal conductivity of around 8 W/mK.. Quite low for our needs. But this also means that we want an amount, as small as possible between led and heatsink. With the smaller led-footprints of today, this will be hard to do, one little dustpartikle between led and heatsink can (and will!) ruin things.

So it is very understandable that most choose the already star-mounted types: Soldered to the star, proper heatexchange is granted. Solder-tin has a heat conductivity of approx 40 W/mK, way, way better than any paste.

Not all of you know this, but I have direct access to optical multilayer deposition machines (sorry, don't recal the english term for it) So I can provide the diamond, and the led's with a silver layer, that alowes me to solder them together! So that's what I'm going to do... UNLESS>>

I also have the possibillity of polishing things, with extreme accuracy.. Grinding the led's base up to just 0.2mm away from 'disasterpoint', I then can polish the base. And now it comes: Polishing the ceramic of a Cree-led can create the possibillity of 'optical bonding' !!

Now I hear you shout: Optical bonding ?? What the heck is that !!??

Let me tell you:

In optics, when you have two, extremely flat surfaces (or curved, exactly fitting together), then you have the oppertunity of bonding them without the use of any glue, or paste! When the two surface are clean enough, they can come extermely close to eachother, within the region of monecular attraction !! For optics, this is the best glue there is: There is no glue (there is no spoon.. remember?), so no refractive index, no thermal resistance, no aging of the bond ! But very difficult to do.

So my next step: Trying to give a led the super flat surface needed for optical bonding, my diamond disk already has a super flat surface.. Hope it's going to work..

Ill keep you posted on this..


Any questions??

Regards,

Ra.

 

[Th232]

Re: Die luminance (surface brightness, important for throw)

One small question. With the optical bonding, what kind of surface roughness do you require? I'm curious because my undergrad thesis involves loose abrasive polishing, reported surface roughness with this particular technique is ~3.25 nm, but that was when polishing silica glass.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

One small question. With the optical bonding, what kind of surface roughness do you require? I'm curious because my undergrad thesis involves loose abrasive polishing, reported surface roughness with this particular technique is ~3.25 nm, but that was when polishing silica glass.

Surface roughness approx 2nm, overall surface flattness Lambda/20..

But you don't need to go that far: Whith a roughness of 3.25nm, you can come very close. The beauty of the diamond disc is that it's transparent! Trough the other side, I can see how close I am by looking at the interference pattern. Colors mean distance, but close to optical bonding, the colors dissapear and the pattern goes white. I played with that using two diamond disks: White is good enough for exellent thermal conduction.. I can do this when I can't reach the optical bonding requirements. But then I have to close the side of the led airtight by putting some epoxy to the sides, closing the outer rim. Then it will stay put and cannot come off, because of the vacuum it would create inside.. Ofcource, this is much harder (nearly impossible) to do this on a copper heatsink..


Regards,

Ra.

 

[saabluster]

Re: Die luminance (surface brightness, important for throw)

Ok, only because this post will still de interesting for the members who want to stay on topic, I'll answer this..( because it will have suggestions for reaching the highest surface brightness )

I got my diamond for free !! So much cheaper than any heatpipe.. You think commercially, you want to sell lights. That makes the use of diamond a no go. I don't think commercially, I just want to create another 'one of a kind' superlight..

No I don't think commercially. That's why I don't make money. I like to figure out what the absolute best way to make something is and rarely is that ever something the marketplace could bare. And I do it anyway but don't charge enough.

When I do the math whith a 30mm TIR-optic, It would give a mere 49,250 cp !!
With a 75mm TIR optic: 325,000 cp !!!

I have no reason to doubt these numbers: All my former TIR-based ledlights almost exactly behave according to these theoretical calculations..

:thumbsup: Be very interesting to see how well the end product matches those numbers.

But there is more: (answering part of Saablusters question)...

I found that indeed it's very important how to mount the led: I nearly blasted a few led's to 'the next world' by not paying (much) attention to this..

Thermal paste has a thermal conductivity of around 8 W/mK.. Quite low for our needs. But this also means that we want an amount, as small as possible between led and heatsink. With the smaller led-footprints of today, this will be hard to do, one little dustpartikle between led and heatsink can (and will!) ruin things.

So it is very understandable that most choose the already star-mounted types: Soldered to the star, proper heatexchange is granted. Solder-tin has a heat conductivity of approx 40 W/mK, way, way better than any paste.

You are of course referring to the pathetic stuff most people buy. Here is what I buy. It's the one on the very right and at the top.

Not all of you know this, but I have direct access to optical multilayer deposition machines (sorry, don't recal the english term for it) So I can provide the diamond, and the led's with a silver layer, that alowes me to solder them together! So that's what I'm going to do... UNLESS>>

That is a huge ace up your sleeve. It was the first thought of mine too but I have no access.:mecry: Oh the things I could accomplish with more tools. They need a DIY kit.

I also have the possibillity of polishing things, with extreme accuracy.. Grinding the led's base up to just 0.2mm away from 'disasterpoint', I then can polish the base. And now it comes: Polishing the ceramic of a Cree-led can create the possibillity of 'optical bonding' !!

Now I hear you shout: Optical bonding ?? What the heck is that !!??

Let me tell you:

In optics, when you have two, extremely flat surfaces (or curved, exactly fitting together), then you have the oppertunity of bonding them without the use of any glue, or paste! When the two surface are clean enough, they can come extermely close to eachother, within the region of monecular attraction !! For optics, this is the best glue there is: There is no glue (there is no spoon.. remember?), so no refractive index, no thermal resistance, no aging of the bond ! But very difficult to do.

So my next step: Trying to give a led the super flat surface needed for optical bonding, my diamond disk already has a super flat surface.. Hope it's going to work..

Ill keep you posted on this..


Any questions??

Regards,

Ra.

I got to admit that is a new one on me. So are you saying that if you get it just right when you push down it will not come back up? Would it need some glue around the edges to insure it didn't move?

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

I got to admit that is a new one on me. So are you saying that if you get it just right when you push down it will not come back up? Would it need some glue around the edges to insure it didn't move?

Yep... But like all types of glue, optical bonding has it's limits in strength, mostly caused by very small imperfections at the two surfaces, like the surface roughness we mentioned earlier (which is very important for this)
Given that fact, and the fact that the led's are getting smaller and smaller, then yes, it would be wise to apply glue around the edges to fix it in place..

EDIT: @Saabluster... The Diemat link you posted: I never thought that thermal epoxy's could ever go that far !!! You made my day !! Thanks...


Regards,

Ra.

 

[HarryN]

Re: Die luminance (surface brightness, important for throw)

Hi Ra, just for fun, double check your diamond for conductivity by direction. Usually very high conductivity materials like that have quite large differences in conductivity with the crystal directions.

CVD diamond is pretty amazing stuff. I used to sell equipment that did that in a funny sort of egg shaped plasma cvd machine.

Coming back to the original topic - surface brightness measurements, I was curious how we should approach a package like a Rebel? Some of them have the phos over quite a large area of the dome, not just the die, at least it looks like that.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

Hi Ra, just for fun, double check your diamond for conductivity by direction. Usually very high conductivity materials like that have quite large differences in conductivity with the crystal directions.

CVD diamond is pretty amazing stuff. I used to sell equipment that did that in a funny sort of egg shaped plasma cvd machine.

Coming back to the original topic - surface brightness measurements, I was curious how we should approach a package like a Rebel? Some of them have the phos over quite a large area of the dome, not just the die, at least it looks like that.

I'll see into it.. But playing with them didn't give any hint of directionabillityness..

About the Rebel: I don't have any (yet..) to take measurements from.. However, the pictures on the web always show a very clear Rebel-dome, so I think the method I mentioned in post #24 should work, right?


Regards,

Ra.

 

[Dr.Jones]

Re: Die luminance (surface brightness, important for throw)

Awesome!
Can the ceramics be polished that microscopically flat?

Ra wrote:
Yesterday, I effectively measured the apparent die-size of a Seoul SSC-P4 led: 1.77 mm2
This is how: Any type of Li-Ion cell with a 10k-Ohm resistor makes it light up faintly, now, with a microscope or eyepiece, you can clearly see, and measure the (apparent) active region of the die. That should be close enough for our needs I think.

Good idea. With the LEDs I measured I didn't need that, but for the SSCs it's the way to go. Did you subtract the contact pads?
Can you provide binning and luminance at various currents, preferably including 350mA, as that is the standard current for the specs?
And preferably at 25°C, using only short current pulses, similar to what the manufacturers do.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

Awesome!
Can the ceramics be polished that microscopically flat?

Yes.. The aluminium oxide base of the Cree led's has almost the same density as saphire.. Hard to polish, but small surfaces are doable..

Good idea. With the LEDs I measured I didn't need that, but for the SSCs it's the way to go. Did you subtract the contact pads?
Can you provide binning and luminance at various currents, preferably including 350mA, as that is the standard current for the specs?
And preferably at 25°C, using only short current pulses, similar to what the manufacturers do.

Don't start asking difficult questions now !! I don't have the electrical equipment (yet...) to do this with reliable results.. And it takes time too, I don't have that with all the things I want to test and build.. So for the moment, I concentrate on the things I want to test before I'm going to create the allmighty 'LedBlaster' ....wait for it...
And.... I did an estimate on the contact pads of the SSC...

EDIT: With my new diamond setup, I immidiately noticed that the output from the led's stayes very stable over time: It's one way to determine the temperature increase of the die: Most manufacturers post nice graphs in which you can see the relationship between flux and die-temperature.. When you make a test-setup with a lux meter at fixed distance, you can clearly notice the flux-drop caused by the increase of the die's temperature. Lower flux-drop means higher heatsink efficiency.. (note that you need stable power to led.. no batteries, because increasing internal risistance also causes flux-drop..)



Regards,

Ra.

 

[Paul Baldwin]

Re: Die luminance (surface brightness, important for throw)

Hi,
good thread I thought I may be able to contribute slightly? http://en.wikipedia.org/wiki/Gauge_block

I used to use these when inspecting cnc work I used to do. Theres a bit of text on that page explaining the phenomena. It is suprising just how well they do stick to each other with no glue involved!

Paul.

 

[Al Combs]

Re: Die luminance (surface brightness, important for throw)

You are of course referring to the pathetic stuff most people buy. Here is what I buy. It's the one on the very right and at the top.

So I had to check out a silver epoxy with thermal conductivity better than solder. The data sheet for DM6030Hk states a minimum cure temp of 175°C or 347°F for a dwell time interval of 45 minutes. Remarkable stuff, but is that high a temperature practical for use with LED's? The data sheet for the DM4030LD says it requires baking at 150°C for 30-60 minutes. That's a little better, but it only has 1/4 the thermal conductivity of the DM6030. SSC's pdf on the P4 and P7 both list a max storage temp of 100°C. Perhaps some reflection on the plastic case material they use? The doc file for the Cree XR-E doesn't list a maximum storage temp. Just a temp as it refers to possible damage from soldering caused by moisture absorption from relative humidity.

 

[Ra]

Re: Die luminance (surface brightness, important for throw)

So I had to check out a silver epoxy with thermal conductivity better than solder. The data sheet for DM6030Hk states a minimum cure temp of 175°C or 347°F for a dwell time interval of 45 minutes. Remarkable stuff, but is that high a temperature practical for use with LED's? The data sheet for the DM4030LD says it requires baking at 150°C for 30-60 minutes. That's a little better, but it only has 1/4 the thermal conductivity of the DM6030. SSC's pdf on the P4 and P7 both list a max storage temp of 100°C. Perhaps some reflection on the plastic case material they use? The doc file for the Cree XR-E doesn't list a maximum storage temp. Just a temp as it refers to possible damage from soldering caused by moisture absorption from relative humidity.

When you read the reflow diagram of the Cree led's, temperatures of 260°C are mentioned, for a short time of cource. But the ramping temp already is 150°C..
I think they will survive a 45 mintute 175°C curing time... They are quite rugged: I got one smoking during testing (bad heatsink contact, or it just likes to smoke..) and it still works fine.. only the dome is a frosted dome now...(XP-G r4) I don't know how this is for the SSC's...


Regards,

Ra.

 

[saabluster]

Re: Die luminance (surface brightness, important for throw)

Grinding the led's base up to just 0.2mm away from 'disasterpoint',

BTW Have you been reading my threads?

So I had to check out a silver epoxy with thermal conductivity better than solder. The data sheet for DM6030Hk states a minimum cure temp of 175°C or 347°F for a dwell time interval of 45 minutes. Remarkable stuff, but is that high a temperature practical for use with LED's? The data sheet for the DM4030LD says it requires baking at 150°C for 30-60 minutes. That's a little better, but it only has 1/4 the thermal conductivity of the DM6030. SSC's pdf on the P4 and P7 both list a max storage temp of 100°C. Perhaps some reflection on the plastic case material they use? The doc file for the Cree XR-E doesn't list a maximum storage temp. Just a temp as it refers to possible damage from soldering caused by moisture absorption from relative humidity.

I have used this with many XR-Es and they can handle the cure schedule needed for the epoxy just fine. As far as SSC's stuff I don't know. I have played with their stuff a bunch but not any "serious" fun as they just can't handle the high current densities like the XR-E can.

Keep in mind the storage temp is not the one you want to be looking at. In the data sheets you will find solder schedules/profile for the LEDs and the maximum allowable temp will be listed. For the XR-E I think it is 260C.