Future development of the maximum luminance of LEDs

303 posts / 0 new
Last post
Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

Yesterday it occurred to me that Oslon Black HWQP has a 2.7 mm² thermal pad and 4 times as much exposed ceramic.
Could we harness it to improve heat extraction efficiency?

BTW, the table in the fist post shows 3,06 mm². In the datasheet I see 2.7×1 mm minus rounded corners.

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Agro wrote:
Yesterday it occurred to me that Oslon Black HWQP has a 2.7 mm² thermal pad and 4 times as much exposed ceramic. Could we harness it to improve heat extraction efficiency? BTW, the table in the fist post shows 3,06 mm². In the datasheet I see 2.7x1 mm minus rounded corners.

 

Thanks, I changed it to 2.67mm2 (using 0.1mm corner radius). I wonder why I had that value there. I think I might have accidentally used an older version of the H9QP datasheet.

 

First we need to understand the internal layout of these LEDs. The Black Flats use the UX:3 Technology. There is a pdf on this somewhere.

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Here is an interesting interview of some Osram OS employees from 12-2016 (in German).

They talk about Osram Stage LEDs. The white variants of these LEDs have new kind of ceramic-phosphor on top. It does not contain silicone and instead is sintered, compressed and baked at 1000°C. It is supposed to tolerate higher temperatures better and does not show signs of degradation.

They also consider 2mm2 die size at high-currents to be an innovation.

The UX:3 technology is a different way of getting the current to the die. Instead of putting a wire mesh on top (which reduces light output and adds electrical resistance), they input current from the bottom side and spread it inside the semiconductor. This also results in a more even current distribution.

 

EDIT: I found it, the paper regarding the development of the UX:3 technology: Development of high-efficiency and high-power vertical light emitting diodes (2014).

Generally this technology seems to have been a very big step for Osram. It caused big jumps in performance and the current Black Flat and Q8WP are definitely already highly optimized LEDs for high luminance and reliability.

 

The lower resistance of the current spreader caused a reduction in Vf of the LEDs compared to older models.

 

Interesting statement: "The maximum power of the devices is only limited by the resistivity of the current distribution layer and the current density of the n-contacts, which can both be engineered according to the application requirements."

Another one: "Owing to the low-Rth(<1.4 K/W for a 1mm2 chip) design, formed by a silicon carrier with solder die attached, the maximum current can be maintained up to 150 °C junction temperature."

That thermal resistance is a lot lower than the value in the datasheet of the Black Flat.

 

The LE UW Q8WP seems to be the 2mm2 SHP (super high power) chip they talk about there, the largest variant of the technology at that point. It looks exactly like their diagrams (except for only being 1.8mm2).

 

 

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

Thanks, interesting.
Though I don’t understand it nearly well enough to be able to recognise the thermal paths, let alone estimate their resistance to the ceramic…

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

You could try sanding down the bottom of the LED, but it will be difficult. Saabluster managed to do it with the Cree XR-E because that LED actually also has contacts on top. This enabled him to just sand down the entire LED.

These Osram LEDs don’t have those, so you would need to sand down only the center solder pad.

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

Speaking of Osram Stage…

LE RTDUW S2WP has 1.8 mm² white die (and 3 others that we can ignore).
Specs look largely like a quad Q8WP, quarter thermal resistance, four times thermal pad size. Actually marginally worse than that.
Now…if a single die can’t widen its thermal path to use that of neighbouring dies, it should be slightly worse than Q8WP.
However if it can move some meaningful heat through the other parts of the package, it could be a winner.

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Agro wrote:
Speaking of Osram Stage…

LE RTDUW S2WP has 1.8 mm² white die (and 3 others that we can ignore).
Specs look largely like a quad Q8WP, quarter thermal resistance, four times thermal pad size. Actually marginally worse than that.
Now…if a single die can’t widen its thermal path to use that of neighbouring dies, it should be slightly worse than Q8WP.
However if it can move some meaningful heat through the other parts of the package, it could be a winner.

That’s an interesting idea, but very impractical! Centering the white die of that LED will be “fun”. It seems to have the same die and lumen range of the Q8WP.

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

The_Driver wrote:
Agro wrote:
Speaking of Osram Stage…

LE RTDUW S2WP has 1.8 mm² white die (and 3 others that we can ignore).
Specs look largely like a quad Q8WP, quarter thermal resistance, four times thermal pad size. Actually marginally worse than that.
Now…if a single die can’t widen its thermal path to use that of neighbouring dies, it should be slightly worse than Q8WP.
However if it can move some meaningful heat through the other parts of the package, it could be a winner.

That’s an interesting idea, but very impractical! Centering the white die of that LED will be “fun”. It seems to have the same die and lumen range of the Q8WP.


You surely know this stuff much better than I do…could you please elaborate why is it going to be significantly harder to centre than Q8WP?
I would think that with both lights having their dies offset would cause the same kind of troubles. OK here the die is offset in 2 dimensions.
The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Agro wrote:
The_Driver wrote:
Agro wrote:
Speaking of Osram Stage…

LE RTDUW S2WP has 1.8 mm² white die (and 3 others that we can ignore).
Specs look largely like a quad Q8WP, quarter thermal resistance, four times thermal pad size. Actually marginally worse than that.
Now…if a single die can’t widen its thermal path to use that of neighbouring dies, it should be slightly worse than Q8WP.
However if it can move some meaningful heat through the other parts of the package, it could be a winner.

That’s an interesting idea, but very impractical! Centering the white die of that LED will be “fun”. It seems to have the same die and lumen range of the Q8WP.


You surely know this stuff much better than I do…could you please elaborate why is it going to be significantly harder to centre than Q8WP?
I would think that with both lights having their dies offset would cause the same kind of troubles. OK here the die is offset in 2 dimensions.

Well it has a non-standard solder footprint. It wont center itself correctly when you put it on a pcb for 5050 (XM-L etc.) LEDs. The next problem you will have (which has nothing to do with focussing though) is that the solder connections underneath for the indvidual dies have a really impractical layout for this scenario. Anode and kathode of a single die are on the same side of the LED. So soldering it onto a 5050 pcb will not work.

Many people wont drill new holes for correct centering when modding a flashlight. You might need to even drill new holes in the light, not just in the pcb. With Q8WP you just need to widen the holes a bit. But I’m not sure, maybe it’s not so bad after all.

It also depends on what kind of light you want to build or mod. The bigger the reflector, the more headaches little things like this will give you.

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

Thanks for the explanation.

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

I’m still waiting for some people to actually try the Q8WP. I will definitely use it in the light in my sig (the light was designed with this LED in mind).

A380
A380's picture
Offline
Last seen: 4 days 10 hours ago
Joined: 01/02/2012 - 16:48
Posts: 103
Location: Sabadell, Catalonia.

The_Driver,

Any link to purchase them?

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

A380 wrote:
The_Driver,

Any link to purchase them?

Rs-online, taobao

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

The_Driver wrote:
Well it has a non-standard solder footprint. It wont center itself correctly when you put it on a pcb for 5050 (XM-L etc.) LEDs. The next problem you will have (which has nothing to do with focussing though) is that the solder connections underneath for the indvidual dies have a really impractical layout for this scenario. Anode and kathode of a single die are on the same side of the LED. So soldering it onto a 5050 pcb will not work.

Some time ago Hoop offered to make custom MCPCBs for a reasonable price.
He hasn’t been active for quite some time, but maybe it would be possible to get some other machinist to do such job?
The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany
Agro wrote:
The_Driver wrote:
Well it has a non-standard solder footprint. It wont center itself correctly when you put it on a pcb for 5050 (XM-L etc.) LEDs. The next problem you will have (which has nothing to do with focussing though) is that the solder connections underneath for the indvidual dies have a really impractical layout for this scenario. Anode and kathode of a single die are on the same side of the LED. So soldering it onto a 5050 pcb will not work.
Some time ago Hoop offered to make custom MCPCBs for a reasonable price. He hasn’t been active for quite some time, but maybe it would be possible to get some other machinist to do such job?

Why a machinist? We just need a standard pcb company to do it. We could ask someone like led4power to make pcbs for the Q8WP.

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

I was talking about S2WP. For now (and likely forever) too niche for even small volume order.

But I agree, it should be possible to gather enough interest for Q8WP.

Enderman
Enderman's picture
Offline
Last seen: 19 min 3 sec ago
Joined: 11/03/2016 - 22:42
Posts: 3947
Location: Vancouver, Canada

Ok so I found out two things:

1) The KW CSLNM1.TG (image A) has a smaller lm and cd value than the KW CELNM1.TG (Image B) simply because the datasheet has a larger bin range for the second one. The website then lists the average lumens and candela as the (min+max) / 2 making it seem like the CELNM is better.

image A, CSLNM)

image B, CELNM)

Even though the CELNM may have a higher bin, we likely will never have a chance to purchase those anyway, so effectively it is the same as the CSLNM.

2) Also, as expected, you can see in the datasheet that the smaller thermal pad does hinder the overdriving ability, as the curve only shows 1.7x output at 2 amps, where the CSLNM has 1.75x output.
For this reason, the CSLNM is a much better choice because not only does it perform the same, assuming you get the same bin, but it has the same thermal pad as the black flat allowing it to be used on XP-size boards and gets better overcurrent ability than the CELNM.


As for the performance compared to the black flat, lower thermal resistance only seems to make it maybe 0.05x higher output at 1.5A when compared to 1A according to the flux-current graph.
The main benefit will be not having the notch in the corner.

———————————————————————————————————————

Despite smaller contact pad size, the CSLPM1.TG (seen below left) has almost the same thermal resistance as the Q8WP and gives it the same overcurrent capability according to the datasheet graphs.

The benefit will be that it can now fit on an XP-size board easily just like the black flat.
It is unclear how having the new only-surface-emitting design will affect the intensity, it has no weird polar distribution from side LES like the Q8WP does, The die area is 0.08mm^2 larger, and the flux for the minimum bin is 400 now instead of 355.
With similar or more lumens, and no side LES, but larger die area, I would expect similar cd/mm^2 than the Q8WP, which is already extremely high 245cd/mm^2.

The specs from the website (in this post) show the Q8WP as higher flux because the highest bin goes up to 710lm while the new LED only goes to 630.
But since we seem to never get the highest bin anyway when buying our LEDs this likely does not matter as much as the minimum bin does.

—————————————————————————————————————-

Based on this, there are two LEDs which are important to test as soon as they are available to buy:
OSRAM OSTAR® Projection Compact, KW CSLNM1.TG
OSRAM OSTAR® Projection Compact, KW CSLPM1.TG

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Very interesting, thanks for the heads-up! These looks like a very nice, practical upgrade! The bigger die of the KW CSLPM1.TG could allow for higher max lumens. The central solder pad is smaller though compared to the Q8WP, but the thermal resistance is actually the same. 

 

 

Enderman
Enderman's picture
Offline
Last seen: 19 min 3 sec ago
Joined: 11/03/2016 - 22:42
Posts: 3947
Location: Vancouver, Canada

Yup!
I would guess the CSLPM will get similar intensity to the “new black flat” CSLNM and just more lumens.
The insulated center pad and no cut corner makes these LEDs really appealing, even if they don’t perform much better than the current ~250cd/mm^2 LEDs.
But of course I do hope that they perform a bit better.
I can’t wait for them to be available Smile

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

It would be interesting to test Stanley HCNW115AJTE

  • 0.96 mm²(unclear, up to 1 mm²) (Blackie: 1.122)
  • 265-410 lm at 1A*3.25V (Blackie: 250-400 at 1A*3.05V)
  • 3 K/W (Blackie: 4.3)
  • 2.096 mm² thermal pad (Blackie: 2.67)

If top bins are orderable, that could be a win.

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Agro wrote:
It would be interesting to test Stanley HCNW115AJTE
  • 0.96 mm²(unclear, up to 1 mm²) (Blackie: 1.122)
  • 265-410 lm at 1A*3.25V (Blackie: 250-400 at 1A*3.05V)
  • 3 K/W (Blackie: 4.3)
  • 2.096 mm² thermal pad (Blackie: 2.67)
  • dome with small diameter

If top bins are orderable, that could be a win.

Interesting, the package does look a bit weird though!
One difference: the Stanley has two bond wires.

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

Luminus now has a new LED on the site, the XP-G2 sized brother of the SST-40, the SST-20 (datasheet).

Features:

  • Thermal resistance: 3.5°C/W (XP-G2: 4°C/W, CSLPM1.TG: 2.6°C/W)
  • two different solder profiles:
  • => A120: XP solder profile with neutral center solder pad
  • => B120: smaller Osram (Black Flat, Square etc.) solder profile with neutral center solder pad
  • rated up to 3A
  • square die (not retangular)
  • brightness at 1.5A: up to 10.7% higher compared to XP-G2 S4, average might be closer to S3

Maybe 220-230cd/mm^2 are possible? Benefits here are that the new Luminus LEDs seem to be chaeap and easy to obtain. They also seem to still have a cool-white tint after de-doming.

KawiBoy1428
KawiBoy1428's picture
Offline
Last seen: 3 hours 26 min ago
Joined: 04/11/2014 - 18:05
Posts: 3710
Location: The Motor City

Love those bond wires… Facepalm

KB1428 “Live Life WOT

The_Driver
The_Driver's picture
Offline
Last seen: 9 hours 35 min ago
Joined: 10/20/2016 - 05:51
Posts: 1384
Location: Germany

I wouldn’t worry about those right now. The SST-40 goes up to 9A and also had just two. It’s possible that they use the same ones in this LED.

djozz
djozz's picture
Offline
Last seen: 6 hours 54 min ago
Joined: 09/07/2012 - 17:04
Posts: 15456
Location: Amsterdam

And, though not the scope of this thread, the SST-20 comes with some warm 95 CRI versions.

KawiBoy1428
KawiBoy1428's picture
Offline
Last seen: 3 hours 26 min ago
Joined: 04/11/2014 - 18:05
Posts: 3710
Location: The Motor City

The_Driver wrote:
I wouldn’t worry about those right now. The SST-40 goes up to 9A and also had just two. It’s possible that they use the same ones in this LED.

Not too concerned with current capabilities, those long bond wires make it very hard for a clean de-dome. They tend to break when the dome releases during the chemical de-doming process. I went thru over 30 SST-40’s. Vinh gave up on them too, he just shaves them now.

KB1428 “Live Life WOT

gchart
gchart's picture
Offline
Last seen: 2 hours 24 min ago
Joined: 03/19/2016 - 11:57
Posts: 1634
Location: Central IL

I wish a purely chemical dedome would be easier with these, but I’m not holding my breath. I’ve had good luck with a combination of shaving / slicing / chemical dedome. Considerably more work, but still good results.


click for larger image

g_damian
g_damian's picture
Offline
Last seen: 9 hours 4 min ago
Joined: 08/20/2017 - 06:52
Posts: 428

SST-20 L2 6500K is available on kaidomain for $3.19.

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

The_Driver wrote:
Agro wrote:
Seenov doesn’t make custom LEDs but only custom PCBs. I asked prophotonix, though looking at their site more closely I suspect they do the same.

Thats too bad, but I guess it was to be expected. Thanks for keeping us up to date. Wink

prophotonix never answered.

I found 2 more companies which may be willing to make a batch of custom LEDs:
https://www.marktechopto.com
https://www.chips4light.com

At least the latter has the necessary expertise:

I haven’t mailed them, if one wants to do it – just do it.

Agro
Agro's picture
Online
Last seen: 8 min 10 sec ago
Joined: 05/14/2017 - 11:16
Posts: 4400
Location: Ślōnsk

What is the die size of old XP-G2?
Here I see 2.16 mm².
In the CRX thread 1.66 mm² for old and 2.18 mm² for new…

200 cd/mm² seems semi-consistent with the smaller number.

Pages