During the last few years the maximum luminance which can be reached with an LED in a flashlight, without additional aiding measures, increased by about 25%. It's specified in Candela per square millimeter of die area, the actual glowing part part the LED (cd/mm2). Together with the front surface area of the reflector (or optic) it determines the throw of a light (just multiply the values). The current leader of the white LEDs manages at least 300cd/mm^2. Phosphor-converted green LEDs can reach an even higher luminance, but with a large reduction in light quality and color contrast. In comparison to years past these are of course very high values, but they're still much lower compared to even low powered short-arc HID bulbs. The 75W Xenon short-arc bulb in the Maxabeam does around 784cd/mm2 and even more at 85W (~1300cd/mm2). The 120-350W UHP bulbs in projectors easily manage 2000cd/mm2.
In this thread I want to discuss why some LEDs manage a higher luminance than others and what can be done to increase these values even further.
Generally luminance values can be found in this thread by sma (which lists values for a lot of other light sources as well giving a great overview!) and here in the tests of Köf3 (he has recently switched over to BLF). Most datasheets unfortunately don't offer this data. Those which do are usually missing values for the maximum possible current.
So how does one get a high luminance?
One needs a cool-white LED with low CRI (high proportion of yellow/green light in the spectrum) and without a dome on it, which tolerates a high power density (Watts per square millimeter of die area).
Temperature of the LED and everything that influences it
Temperature of the phosphor
Not fully understood/researched current limit of some LEDs (see the Luxeon V for example)
By far the most important criteria is heat. Most LEDs are rated for a maximum temperature of 150°C.
Here I have created a table where I compare the currently best LEDs in this regard:
max. Power [W]
Power density in regard to the die-area[W/mm2]
Current density in regard to the die-area [A/mm2]
Area of the heat conducting solder pad [mm2]
Power density in regard to the heat conducting solder pad[W/mm^2]
real thermal resistance, calculated values assume 20% efficiency and 1.293 °C/Wmm² thermal resistance of solder[°C/W]
min. LED-temperature at max. luminance and 25°C heatsink temp[°C]
I got these values from the datasheets, from tests in the forums and some I have calculated. Please tell me if you find any mistakes.
What I noticed:
Most of the LEDs reach their maximum luminance at around 90-108°C
Only the LEDs from Lumileds reach their maximum at temp less than that
Only a few reach their maximum above that
The Luminus CFT-90 seems to be a step ahead of the other LEDs from a technological standpoint. Its luminance is almost unbelievably high for such a large die. It does have the largest thermal pad though (I calculated using the thermal pad size of the SST-90 and also using the size of the integrated copper pcb)
The Osram "White Flat" KW CSLNM1.TG has the highest power density (of the whiter LEDs) regarding the die size, but not regarding the thermal pad size
What I want is a LED with a small die (2mm2), no dome, directly mounted on a copper pcb like the CFT-90. This pcb would optimally have the 16mm standard footprint we often use here.
What do you guys think?
Calculation of real thermal rsistance which includes solder: real_thermal_resistance = typ_datasheet_therm_resis + (1,293°C/Wmm2 / area_of_heat_conducting_solder_pad)
Calculation of min temp in table: min_temp = (max_power * 0,8 * real_therm_resis) + 25°C
Nice compilation, but I suspect your luminance value quoted for the XPG3 (and possibly others) is too high. Often the luminance is calculated by measuring the intensity above the LED (in candela) then dividing by the die area. But if light is being emitted by area other than the actual die than the calculated luminance will be inflated. I showed this with the XPG3. In these cases measuring the resulting beam center intensity from a flashlight and dividing by the reflector area is a better way of determining the luminance since, if properly focused, this method measures the light coming from the die itself. This phenomenon happens with other LEDs, too. If you look at the lit LED and see light coming from parts other than the die than the above luminance calculation will be inflated. This stray light is contributing to the total output from the LED, but contributes to the beam corona rather than the beam center intensity.
Electrical thermal resistance must be calculated at certain conditions to be valuable. I don’t think that whatever the manufacturer specifies here is very relevant to our leds over-driven near the limits.
No, I meant the measured value of the XP-G3, which I found, only pertains to the use with aspherics.
I actually state the real thermal resistance, which accounts for the efficiency of the LED. Here is an application note from Osram explaining the differences regarding the real and electical thermal resistance.
The way I understand it…
Let’s say for a while that 1 W of white light is 310 lm.
XP-G3 S5 at binning conditions does about 191 lm/W, 62% efficiency.
Let’s assume also that that their stated thermal resistance is electrical (I don’t know if that’s the case).
For XP-G3 3 °C/W electrical is 7.81 °C/W thermal.
Overdrive it so hard that efficacy goes down to 60 lm/W and electrical thermal resistance rises to 6.3 °C/W.
CFT-90 at binning conditions does 61 lm/W. By overdriving it you won’t make its electrical thermal resistance rise much.
Sorry, I think I edited my post while you wrote your reply. I state the real thermal resistance unless noted otherwise.
The Osram Black Flat HWQP has a stated typical real thermal resistance of 4.3°C/W. They also note a maximum of 5.5°C/W. Do you think this pertains to individual variation of these LEDs or do these values also change when the power goes up?
Since what you state is real thermal resistance - that’s good. Though it would be interesting to know LED efficiency (or efficacy) at peak luminosity to get a more accurate picture of temperature rise.
As to Oslon Black HWQP…
I did a quick search and I see that thermal conductivity of materials changes with temperature. If wikipedia is correct, for metals it improves proportionally to temp rise (in K) while for non-metal solids it stays largely constant. So within LED operating conditions the change can’t be nearly as large. Individual differences? Intuition tells me the difference is too large to be explained this way. Maybe there are different variants of the LED?
Really I have no idea.
I don’t know what you mean with “different variants of this LED”. The datasheet is only for this one type, the HWQP.
So, getting back to the actual topic of this thread, what do you guys think needs to be done in order to improve the maximum luminance? How high can LEDs go considering the properties of the used materials?
Are laser-phosphor modules the only way forward? They offer two distinct advantages:
A blue laser diode coupled with a focussing lens will put a much higher concentration of blue light onto the phosphor compared to LEDs at a much lower power consumption if one ignores the size of this area
This design also allows for thermally separating the phosphor from the blue light source and mounting it in a thermally beneficial way (i.e. between two glass plates)
Cree changed production processes of several emitters during their lifetimes. Maybe Osram does the same and what we see is accounting for that?
As to lasers:
There’s always the scare of the laser moving against the enclosure and its beam hitting the reflector, hurting somebody. Personally I’d rather not go this route.
However, there’s a safer way: use Yuji phosphors and UV laser together with UV filter.
Hmm… laser excited phosphor sounds rather expensive and inefficient.
The dielectric layer between chip and LED board could probably be improved further?
Or do they already use nano carbon stuff for that (which is (can be) better than copper and not electrically conductive)?
Cooling on the phosphor side is maybe something to develop.
Diamond top layer is probably good for that, but expensive…
I should have a better look at how an LED is built up…
but as a layered cake:
blue pump LED
“nano carbon stuff” base
The diamond layers held in place by thermally conductive stuff with a thermal path to the “nano carbon stuff”, or even integral with the nano carbon stuff base.
Hmm… With a blue pump you use the blue light as a part of the spectrum.
With “Yuji phosphor” all (or at least much more) of the radiation from the UV pump is used to excite the phosphor, and it’s the phosphor that radiates the full spectrum.
I think it’s safe to assume that it will always be less efficient than a blue pump LED.
As for risks of lasers, this depends on how you build it.
I think it’s not hard to mechanically prevent the laser from missing the target.
But the laser could accidentally burn through the phosphor, which sounds like a problem…
You mean someone has measured the beam intensity and divided by the lens area and got 220cd/mm^2?
I measured a sliced XPG3 in a UF1504 and got similar, lower than expected, numbers as in a reflector light.
As far as your question about the future of high luminance LEDs: my understanding is that it’s a matter of internal photon production efficiency and the photon extraction efficiency. Using higher quality InGaN active material and designing better thermal paths can increase the internal efficiency.
Ensuring all photons exit through the die will increase the extraction efficiency and luminance. For example if the LED design is such that a lot of light escapes out the side, like the XPG3, this will lower the luminance.
This happens to some extent in other LEDs. I was working with dedomed SST40s recently and noticed some light escaping out the bottom where the chip meets the package. Not nearly as much light escapes as the XPG3, but I think I have observed the effect on the luminance. Comparing the dedomed SST40 with the dedomed XPL V6 in an eagle eye X6, I measured about 20% more output with the SST40 (because of the higher current), but only 9% higher beam intensity.
Did you take into account that the SST-40 has a larger die (3.994mm2)? It needs more lumens for the same luminance compared to the XP-L (3.55mm2). You can find both values in my table above (the XP-L HD should have the same die size as the XP-L HI). ;)
I don’t think the electrical power factors in to my reasoning. If the die sizes are the same and all light is emitted from the die area, 20% more output should also mean 20% higher luminance and beam intensity. The fact that they are different could be explained by the fact that some of the light from the SST40 is coming from the side of the die rather then from the die itself. The result of this is a slightly brighter corona.