Future development of the maximum luminance of LEDs

303 posts / 0 new
Last post
Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

Enderman wrote:
Agro wrote:
There are 2 problems with Conductonaut though:
  • it is rated to 140 °C, so you can’t use it together with soldering (although I don’t know what happens once you exceed that temp).
  • it is fluid which means there’s a risk of run-out. Or pump-out.

The LED should never exceed 140C.
If you use liquid metal there is no need to solder anything, that’s the entire point of using it instead of soldering.
If you mean soldering the LED to the MCPCB, you do that before you put the MCPCB on the heatsink and apply thermal compound.

Also if you apply it correctly there is nothing that comes out.
If you overapply and some comes out around the MCPCB you can just wipe it off.
Nothing else from below the MCPCB will come out so it is completely safe.
This is assuming the MCPCB is securely screwed down of course.


I meant using it on the thermal pad of a led while joining electric pads otherwise. If the other joining method is soldering – LM seems off the table. If it’s glueing heat is not a problem.

There was quite a number of people reporting how Conductonaut leaked and caused shorts inside their laptops. That’s why people recommend CLU there – it’s thicker.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

Enderman wrote:
The_Driver wrote:
German Modder Vinz uses self-made diamond based thermal glue. It has 85% synthetic diamond content. The diamond has a thermal conductivity of 3300 W/(m·K). It theoretically has a better thermal conductivity than copper. He doesn’t use it for the central thermal pad though. He uses normal DTP copper pcbs and a solder with a high silver content, Mundorf MSolder Silver Gold. This solder is very expensive unfortunately (50€ for a 100g roll). MSolder Silver Gold Supreme has an even higher Silver content (9.6g instead of 3.8g), but it’s twice as expensive.

I would prefer we talk about more fundamental changes. Using better solder will not increase the luminance by visible amounts. How do we get from 300cd/mm2 to 500 or even 1000?


I know this post is a bit old but I wanted to mention that diamond is 2200W/mk not 3300, and that thermal compounds made of diamond do not actually work very well.
ICdiamond only has ~10w/mk.
The way to get the best thermal connection is to use liquid metal such as coolaboratory liquid ultra/pro or thermal grizzly conductonaut.
The pro and conductonaut have ~76W/mk.
Using this is a better choice than solder because being a liquid it can contact all the small cracks and crevices.
If you tried to solder a copper MCPCB to a copper heatsink it would be almost impossible to get a perfect joint due to the large area and how fast the heat moves away.
High silver content in solder also gives a higher thermal conductivity, about 70-80 IIRC, but it makes pretty much no noticeable difference.
I use 3% silver/tin solder for all of the stuff I do.

Applying and removing the liquid metal is actually really easy to do.
Having an easily removable MCPCB is great, I don’t recommend people solder their MCPCBs to a heatsink.

I don’t know where he got the number, different places state different things. A range makes more sense to me than a single value. The highest I have read is 2500.

I do find it interesting though that you stated that it doesn’t work that well because this one company can’t or won’t do it that good.

In the end polishing the surfaces will reduce the effect of the paste anyhow. Vinz preferred this stuff especially for potting his lights.

I think it could be used for replacing a DTP pcb. That would be interesting.

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Agro wrote:

I meant using it on the thermal pad of a led while joining electric pads otherwise. If the other joining method is soldering – LM seems off the table. If it’s glueing heat is not a problem.

There was quite a number of people reporting how Conductonaut leaked and caused shorts inside their laptops. That’s why people recommend CLU there – it’s thicker.


Oh I see.
Yeah liquid metal wouldn’t be ideal to use so close to the LED pads.
For stuff like the synios LEDs though the thermal pad is also the cathode so it needs to be soldered anyway.
Using conductonaut between the MCPCB and heatsink is fine, it does not seep out the sides.

The_Driver wrote:

I don’t know where he got the number, different places state different things. A range makes more sense to me than a single value. The highest I have read is 2500.

I do find it interesting though that you stated that it doesn’t work that well because this one company can’t or won’t do it that good.

In the end polishing the surfaces will reduce the effect of the paste anyhow. Vinz preferred this stuff especially for potting his lights.

I think it could be used for replacing a DTP pcb. That would be interesting.

I just double checked, the 3300 is for single crystal artificially made synthetic diamonds. Apparently moissanite also has a high thermal conductivity close to diamond.

Still, diamond isn’t fluid so the best that can be done is grinding it into powder and putting it inside another thermal compound.
Due to the surface tension of liquid metal, I doubt it the diamond dust would mix well with that while still allowing good contact between the fluid and solid grains.

This is most likely why the only diamond thermal compounds are regular viscous pastes and not liquid metals.
Instead of having a uniform mixture it has to go mediocre thermal compound -> diamond grain -> mediocre thermal compound -> diamond grain over and over again, which likely leads to the bad performance.

Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

Enderman wrote:
Agro wrote:

I meant using it on the thermal pad of a led while joining electric pads otherwise. If the other joining method is soldering – LM seems off the table. If it’s glueing heat is not a problem.

There was quite a number of people reporting how Conductonaut leaked and caused shorts inside their laptops. That’s why people recommend CLU there – it’s thicker.


Oh I see.
Yeah liquid metal wouldn’t be ideal to use so close to the LED pads.
For stuff like the synios LEDs though the thermal pad is also the cathode so it needs to be soldered anyway.
Using conductonaut between the MCPCB and heatsink is fine, it does not seep out the sides.

Did you perform drop tests?
Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Agro wrote:

Did you perform drop tests?

Uh, no, if I dropped a $2000 11lb light I would have much bigger problems than a tiny drop of liquid metal.

But just from looking at it, the gap between the mcpcb and heatsink is so small that the liquid metal can’t run out.
The only runout happens if you apply too much initially, it can seep out when you tighten the MCPCB down, but as I said you can just wipe that off.

Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

I meant that this tiny drop can damage your $400 reflector.
And also that I simply don’t like general statements like “Using conductonaut between the MCPCB and heatsink is fine, it does not seep out the sides.” as they may not be always applicable.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

You could put expoxy or similar around the sides of the pcb to seal it off.

Or you could grind and polish both surfaces to perfect flatness. Are you gonna do that?

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Agro wrote:
I meant that this tiny drop can damage your $400 reflector.
And also that I simply don’t like general statements like “Using conductonaut between the MCPCB and heatsink is fine, it does not seep out the sides.” as they may not be always applicable.

The liquid metal simply rolls off of a surface unless it is purposefully rubbed in using the swab it comes with.
Kinda like mercury where it just forms small balls that don’t stick.

The only concern using liquid metal is that it is conductive and can short stuff which is why you need to be very careful using it in a PC.
Here, worst case scenario is that a $5 LED driver gets fried.

But no, I can say that there is certainly not going to be any liquid metal coming out from between the MCPCB and cooling block, there is literally a few microns of space and it is impossible for a fluid with such high surface tension to seep out.
This is like being concerned that regular thermal paste will leak out from between a CPU and heatsink due to them not being perfectly flat together. It simply never happens. The only stuff that can possibly come out is during the initial application.

The_Driver wrote:
You could put expoxy or similar around the sides of the pcb to seal it off.

Or you could grind and polish both surfaces to perfect flatness. Are you gonna do that?


The block is faced on a lathe and the MCPCB is perfectly flat as far as I can tell.
Any polishing I try to do would make things worse.
The liquid metal already gives a large improvement over thermal paste, and that only results in slightly higher current handling.
The only thing which would possibly be of any further improvement is directly lapping both surfaces together, but that would take tens or hundreds of hours for basically 0 improvement (in temperatures).
Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

Enderman wrote:
Agro wrote:
I meant that this tiny drop can damage your $400 reflector.
And also that I simply don’t like general statements like “Using conductonaut between the MCPCB and heatsink is fine, it does not seep out the sides.” as they may not be always applicable.

The liquid metal simply rolls off of a surface unless it is purposefully rubbed in using the swab it comes with.
Kinda like mercury where it just forms small balls that don’t stick.

The only concern using liquid metal is that it is conductive and can short stuff which is why you need to be very careful using it in a PC.
Here, worst case scenario is that a $5 LED driver gets fried.


It’s extremely corrosive to aluminium.
Pöbel
Pöbel's picture
Offline
Last seen: 1 month 2 weeks ago
Joined: 02/14/2013 - 08:27
Posts: 259
Location: Germany

Yes, exactly. Get liquid metal on any aluminium part and that part will be gone soon. The aluminium will simply fall into pieces Big Smile

Also Copper + Copper might be a more eternal joint than you would hope, although there is no danger of the copper being corroded away by the liquid metal. If the copper is gold plated like on most boards, this won’t be a problem though.

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Agro wrote:

It’s extremely corrosive to aluminium.

Yes I have always known that liquid metal reacts with aluminum.
The MCPCB and block are both copper.
As I said before, there is no risk of any liquid metal coming out anyway, there is a fraction of a gram compressed between the two pieces of metal and not enough space for it to exit.
And again, as I said before, even if there is a drop, it forms a ball just like mercury that rolls off of a surface, it does not stick.
It would not cause any damage to the aluminum on the reflector unless it was rubbed onto the surface such that it broke the surface tension and made contact. This would also require a lot more liquid metal than is even applied in the light.

It’s honestly not as scary as people make it out to be.
You even get a ton of applications with the 1g syringe, like more than 10.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

I have updated the table because more practical tests have shown that the Synios LEDs are not quite as good as we thought. The luminance of the CFT-90 is also not certain yet. Vinhnguyen54 got 1.58Mcd in a BLF GT and 900kcd in a TN-42 at sub-maximal current. I don’t know of any other values.

Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

He said he got just 20 amps in GT. Frankly, it beats me why so low.
Maybe he posted it as a conservative number that any modded light will reach with much used batteries, but anyway he has a lot of headroom.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

Maybe there are too many losses in the main springs.
It’s also a DD light. It’s difficult to base anything off of those kinds of measurements. Measuring high currents obviously needs to be done correctly, otherwise the measurement will reduce the actual current.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

I have updated the values in the table to reflect koef3’s newest findings.

I didn’t anticipate having to change them so often… :D.

EDIT:
Since the luminance values of some LEDs are lower than anticipated, we need to also question those of the Luminus CFT-90.

vinhnguyen54 measured 3800otf Lumens and 900kcd with this TN42vn90.The reflector openings have 89mm and 28mm diameter (so 5605mm^2). Lets say that the lens transmission rate is 98% and the reflectivity of the reflector is 85%. Lets also say that the reflector collects 75% of the emitted light.

I thus get 4390 led lumens and 193cd/mm^2. sma measured around 185cd/mm^2 at 4400 lumens, so it can be concluded that the values match.

djozz
djozz's picture
Offline
Last seen: 10 hours 16 min ago
Joined: 09/07/2012 - 17:04
Posts: 16669
Location: Amsterdam

Some leds of which I would be curious where they end up in the table of the OP:

-Osram Oslon Compact PL
-newest Lumileds Luxeon Z ES (they just announced a brightness upgrade)
-new Lumileds Luxeon CZ (=dome-less Luxeon C)

These all are not available yet and all have non-standard footprints that have to be overcome.

Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

I would be also interested in OSRAM OSTAR Projection Compact KW CSLNM1.TG. In many ways similar to Black HWQP, but newer.
djozz, when I search for Luxeon CX I see some COB

djozz
djozz's picture
Offline
Last seen: 10 hours 16 min ago
Joined: 09/07/2012 - 17:04
Posts: 16669
Location: Amsterdam

Sorry: CZ, corrected

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Agro wrote:
I would be also interested in OSRAM OSTAR Projection Compact KW CSLNM1.TG. In many ways similar to Black HWQP, but newer.
djozz, when I search for Luxeon CX I see some COB

You’re right, these seem to be even better than the black flat, assuming they can be overdriven just as much.
LUW HWQP: White 3.05 W 325 lm 108 cd 1000 mW 334 mW/sr 120°
KW CSLNM1.TG: White 3.00 W 325 lm 108 cd 1000 mW 334 mW/sr 120°
KW CELNM1.TG: White 3.00 W 374 lm 124 cd 1150 mW 384 mW/sr 120°

All three have 1.03mm square die, but the CELNM1 is both brighter and more efficient than the other two.
Also, no notch at all for these new LEDs Smile

—————————-

There is also this larger version, closer to the Q8WP koef3 tested recently that got almost as high cd/mm^2 as the black flat.
LE UW Q8WP: White 4.20 W 533 lm 153 cd 1650 mW 473 mW/sr 130°
(die size 1.91mm^2)
KW CSLPM1.TG: White 4.20 W 515 lm 171 cd 1590 mW 529 mW/sr 120°
(die size: 1.99mm^2)

The main difference is that the KW has no side emission like the Q8WP did so possibly more of that light is going forward and could potentially give higher cd/mm^2 despite larger area and lower stock output.

Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

Enderman wrote:
Agro wrote:
I would be also interested in OSRAM OSTAR Projection Compact KW CSLNM1.TG. In many ways similar to Black HWQP, but newer.
djozz, when I search for Luxeon CX I see some COB

You’re right, these seem to be even better than the black flat, assuming they can be overdriven just as much.
LUW HWQP: White 3.05 W 325 lm 108 cd 1000 mW 334 mW/sr 120°
KW CSLNM1.TG: White 3.00 W 325 lm 108 cd 1000 mW 334 mW/sr 120°
KW CELNM1.TG: White 3.00 W 374 lm 124 cd 1150 mW 384 mW/sr 120°

All three have 1.03mm square die, but the CELNM1 is both brighter and more efficient than the other two.
Also, no notch at all for these new LEDs Smile


Yeah, but CELNM1.TG has a tiny thermal pad. And it doesn’t make up for this with smaller thermal resistance inside the package – actually it’s a bit worse here too. So likely it won’t be possible to overdrive it so much.

Enderman wrote:
There is also this larger version, closer to the Q8WP koef3 tested recently that got almost as high cd/mm^2 as the black flat.
LE UW Q8WP: White 4.20 W 533 lm 153 cd 1650 mW 473 mW/sr 130°
(die size 1.91mm^2)
KW CSLPM1.TG: White 4.20 W 515 lm 171 cd 1590 mW 529 mW/sr 120°
(die size: 1.99mm^2)

The main difference is that the KW has no side emission like the Q8WP did so possibly more of that light is going forward and could potentially give higher cd/mm^2 despite larger area and lower stock output.


This also has a much smaller thermal pad than the older LED.
Due to better efficiency I find it interesting, but I don’t expect it to beat Q8WP.

It’s interesting that now we’re discovering several years old emitters that are great for our application.
And sad how the later generations are often just inferior for what we do.
And what’s worse – we’re not a big enough market to even keep an old product alive, let have one developed for us.

Agro
Agro's picture
Online
Last seen: 25 sec ago
Joined: 05/14/2017 - 11:16
Posts: 5742
Location: Ślōnsk

BTW, I don’t see it mentioned:
There’s some chance that Luxeon F turns out good.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

I think the most realistic way to get higher luminance is with a Q8WP using higher bins.

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Agro wrote:

Yeah, but CELNM1.TG has a tiny thermal pad. And it doesn’t make up for this with smaller thermal resistance inside the package – actually it’s a bit worse here too. So likely it won’t be possible to overdrive it so much.

Oh you’re right, I wonder why the CELNM1 has a higher stock output then, if it’s basically the same die.
The CELNM does have higher thermalresistance due to the smaller pad, but what’s cool is that the CSLNM actually has less thermal resistance than the black flat.
.
For the CSLPM1 even though it has a smaller thermal pad it is still the same size as the CSLNM1 and has the same thermal resistance as the Q8WP, so it might not be that difficult to cool well.
.
So in summary:
LUW HWQP: White 3.05 W 325 lm 108 cd 1000 mW 334 mW/sr 120° = reference
KW CSLNM1.TG: White 3.00 W 325 lm 108 cd 1000 mW 334 mW/sr 120° = same contact pad size, lower thermal resistance
KW CELNM1.TG: White 3.00 W 374 lm 124 cd 1150 mW 384 mW/sr 120° = small contact pad, higher thermal resistance, for some reason better performance at stock? will not work on any normal MCPCB

LE UW Q8WP: White 4.20 W 533 lm 153 cd 1650 mW 473 mW/sr 130° = large contact pad, lower thermal resistance than black flat
(die size 1.91mm^2)
KW CSLPM1.TG: White 4.20 W 515 lm 171 cd 1590 mW 529 mW/sr 120° = same contact pad size as black flat, same/slightly lower thermal resistance as Q8WP (much lower than black flat)
(die size: 1.99mm^2)

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

I just noticed something interesting about the Black Flat:
It only has one bond wire. In his test Köf3 noted that one of his Black Flats died at 5.4A because the bond wire melted. So we can conclude that it can tolerate up to ~5A. This also means that it gets warm at currents close to that. This might be one reason why the Vf of the Black Flat is so high at high currents, which of course makes it run hotter than it would otherwise.

The Cree XP-G2, XM-L2, XP-L (HI/HD) and Luminus SST-40 all have two bond wires. The latter goes up to 9.5A.
Now of course we don’t know if they use bond wires with different diameters 8Cree might have reduced the diameter of the XM-L2 bond wires at sme point because they now die at lower currents than in the past).

I do think it’s interesting though that the Osram Q8WP has not two, but four, very thick looking bond wires. They should not be heating up nearly as much as the single one in the Black Flat.

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

I run all my black flats at 6-6.5A though…

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany
Enderman wrote:
I run all my black flats at 6-6.5A though…

That’s a good point, I had forgotten about your use of high currents.

Xandre
Xandre's picture
Offline
Last seen: 2 hours 12 min ago
Joined: 03/06/2011 - 07:02
Posts: 434
Location: South Germany

That is still a mystery for me ?

Cause of the tests the Black Flat is running best at 4,5A.

Do you make other tests and the Led get brighter with more A ?

Regards Xandre

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany

Here are the results Enderman posted back then. His voltage measurements seem incorrect though (way too high), there was probably some additional resistance involved.

 

The problem here is that he can't really prove why his LEDs performed better in that setup. You can only test it for yourself by emulating parts of the process (his specific thermal paste, active cooling, his way of soldering etc.).

Enderman
Enderman's picture
Offline
Last seen: 8 hours 30 min ago
Joined: 11/03/2016 - 22:42
Posts: 4027
Location: Vancouver, Canada

Xandre wrote:
That is still a mystery for me ?

Cause of the tests the Black Flat is running best at 4,5A.

Do you make other tests and the Led get brighter with more A ?

Regards Xandre


I tested my own black flats with a fan cooled, copper plate, 5 heatpipe CPU cooler and MX-4 thermal paste.
I was getting lux increases up to 5.2 amps which was the limit on the power supply, and the curve showed it kept increasing a bit past that.

And yes as the driver said the voltage measurements are off by a lot, I didn’t measure the voltage at the LED properly, that was the voltage at the power supply.

The buck drivers I use are 6A, but they can spike up to 6.5A according to mtn electronics. I think last time I measured the current I get about 5.7-5.9A stable out of them
I have killed one or two black flats by turning them on max too fast for the first time.
If I go to low, then medium, then high after a few minutes there are no problems. After some use it seems like I can quickly turn it to max with no issues.

I want to do a new test using liquid metal, but first I need to buy a power supply that can do more than 5.2A.
I’m planning to get the DPS5020 soon.

The_Driver
The_Driver's picture
Offline
Last seen: 14 hours 43 min ago
Joined: 10/20/2016 - 05:51
Posts: 1441
Location: Germany
Enderman wrote:
ICdiamond only has ~10w/mk.

I just found an interesting, old test of diamond thermal grease here .

Pages