Highest efficiency LED?

  • Hero Member

    So far, I've found one that's notionally equivalent to 314 lumens per watt: https://www.digikey.com/product-detail/en/61-238%2FKK2C-S30306F4GB2%2FET/1080-1217-1-ND/2675951/?itemSeq=318239104

    Just to be clear, though, it doesn't actually achieve 314 lumens, because it doesn't use that much wattage. What it does is achieve 20.42 lumens at 20ma at 3.25v, which is 65 mWatt, which is a far more useful range for us. So, I'm not sure what the right unit would be for expressing its efficiency. Lumens per milliwatt?

    Regardless, that seems quite efficient compared to most LED's. I imagine it takes only a very short blink to see it.

    Anyone know of an LED that has even more lumens/mWatt?

  • Hardware Contributor

    @NeverDie said in Highest efficiency LED?:

    What it does is achieve 20.42 lumens at 20ma at 3.25v

    No it doesn't if I believe page 5.

    Range of Luminous Intensity*1
    *1 When three LED dies are operated simultaneously.

    *2 For each die.

    So my understanding is it's achieving a maximum of 20.xx lm at 20*3 = 60mA and it's just 10x lm/W so nothing extraordinary here.

    If you only just want a light signal visible by user, then go for a green led. It's the color the eye is the most sensible to, so it's the one you will be able to see the most for a specific input power.

  • Plugin Developer

    If you look at the datasheet is says, depending on the bin, between 17 and 18 lm max output. At max power of 110 mW that is 17.5/0.110 or 159 lm/w. I haven't looked for comparisons but I suspect that is typical of most modern LEDs. Keep in mind that, for commercial LED bulbs, there is power converter efficiency that has to be factored into the efficacy and they still achieve > 100 lm/W with pretty average LEDs.

  • Hero Member

    Thanks guys for pointing this out. I had based my numbers not on the datasheet but only just Digikey's posted information in their product description:
    Millicandela Rating 6900mcd
    Voltage - Forward (Vf) (Typ) 3.25V
    Current - Test 20mA
    Viewing Angle 120°

    I didn't have much luck finding high-efficiency green LED's on digikey with >100 lumen/watt output, but I did find some on mouser. I agree that for notification purposes, it would make more sense to focus on an LED whose light output is clustered around 555 nm, since that appears to be what human eyes are most sensitive to. Curiously, some LED manufacturer's describe that region as "lime" instead of "green".

    Is there a minimum length of time an LED needs to be lit for the human eye to see it? If so, I was toying with the idea of using a 365 nm ultraviolet LED to only very briefly flash into some phosphorescent paint, which would then glow for quite a while afterward, certainly long enough for the eye to see it. There is a newer breed of phosphorescent whose glowing is strong enough that it's even visible in daylight (even in direct sunlight). I know this because I used some in a tile groute, and we had to dilute it down because we only wanted the glowing to be visible at night, not during the day.

    Anyhow, since there are conversion losses from energy input into a phosphorescent coating vs light emitted out afterward, I'm a little skeptical it would be better than lighting a green/lime LED for long enough for the human eye to see it. Not sure though. It doesn't seem to take long for the human eye to see the flash from a camera, but then again I've never measured how long those flashes actually take, since it needs to be long enough to fully excite the pixels in the camera.

    Nick gammon's advice was 5-10 milliseconds for a low power LED. But that was advise from the year 2012. Perhaps it would be less for a current generation, high efficiency, high power LED flashed in a higher current range but for much less time?
    "Go easy on "power" LEDs. LEDs can use quite a bit of current (say, 10 mA). If possible just flash one briefly rather than have it on all the time, or use a low-power LED. Experimentation shows that flashing an LED for 5 mS to 10 mS is quite noticeable. So, one 10 mS flash per second is only an overhead of one percent." (https://www.gammon.com.au/forum/?id=11497)

  • Hardware Contributor

    @NeverDie said in Highest efficiency LED?:

    flashing an LED for 5 mS to 10 mS is quite noticeable.

    I checked in an old sketch for a board that was using a 0603 green led (cheap one so not even bright light) and I had a 20ms on time. But this was producing quite a strong light flash, easily noticeable in daytime through 2mm of white PLA (battery was CR1632 and resistor I think around 2K).
    I think it's a combination on "on time" and the current the led receives, so you will need to experiment with the leds you buy, the "problem" is not really about a duration but of a quantity of light you manage to get from the LED compared to ambient light.

    I don't know if you saw the answers to this question, but the experiment is quite interesting.

  • Hero Member

    @Nca78 From what I've read, the eye averages the amount of energy it sees over some period of time. That's how the whole persistance of vision thing works with moving pictures. So, it may turn out that it doesn't matter so much whether I use high power for incredibly brief period or lower power for longer. Maybe it's all equal. I'm sure this has been well studied, but so far I haven't been able to google up the answer. I guess I'll just order a few different types and try some experiments. In the meantime, if somebody reading this knows the answer as to which approach is a more efficient use of energy, please do post. I'm wagering whichever green/lime LED has the highest lumens/watt will be the winner, and that everything else is secondary.

    The guy in the post that @Nca78 linked to above said he could see a flash from a very ordinary LED that lasted just 1 microsecond, but with rather special conditions like cupping his hands around it to make a dark contrast and expecting the flash and staring right at it. So, maybe a much brighter LED (say 150 lumen) pulsed for one microsecond wouldn't require those special conditions. That would be pretty amazing, because then the energy consumption would be tiny. Somehow I'm both hopeful and skeptical at the same time.

  • Hero Member

    It turns out there are two different types of peak eye sensitivity: scotopic at 507nm and photopic at 555nm:
    So, I need to get that sorted as to which one I need/want, so that I can buy some LED's centered on the appropriate wavelength. Or maybe I should split the difference and get something that overlaps both? Or perhaps actually get both and only flash the one that matches the circumstances. I think I'll pick photopic, because I probably care more about daylight vision than dark adapted vision.
    alt text

  • @NeverDie Not so scientific, but am acutely aware even with peripheral vision when my nodes flash blue up to 4m away indoors even in artificial light, duration 5ms.

  • Hero Member

    @zboblamont Do you happen to know the make and model number?

  • @NeverDie Sorry no information as not the constructor, it is on a pre-built board, now 3 years old.
    The LED is a 1mm wide SMD driven on 3.3v via a 220R, but exceptionally bright compared to a second yellow LED on an identical circuit, probably from the same manufacturer.
    My intended point was that blue may an option worth examining alongside the suggested green.

  • Hero Member

    @zboblamont I've read that bright blue LED's can be harmful to the eyes, and that you can pretty quickly damage your eyes unintentionally. I think the explanation offered was that since the eye isn't particularly sensitive to blue, your eyes can suffer damage before the pupils close. Or, maybe because blue is so close to UV, and so you might get extra invisible dosing that way? Not sure.

    Anyhow, I haven't read any similar warnings regarding green, so maybe they're safer.

  • @NeverDie I can't think why any particular tone in the visible light spectrum should harm the eye, else LEDs might be issued with a Health Warning 😉
    Hopefully your experiments will identify what you're after...

  • Hero Member

    @zboblamont Thanks. FWIW, I first became aware of the issue from reading CREE datasheets. At some point CREE put out an advisory: https://www.cree.com/led-components/media/documents/XLamp_EyeSafety.pdf

    Cree classifies its own XR-E royal blue LED's as RG-3 High Risk, which it defines as "Hazardous even for momentary exposure".

    "To date, the testing shows that Cree’s blue and royal blue LED components (450-485 nm dominant wavelengths) pose a higher potential eye safety hazard than its white LED components. Other colors of LED components, such as green and red LED components, do not pose as significant of an eye safety risk. Regardless of LED color, Cree advises users to not look directly at any operating LED component. "

    In the context of this thread, it's that last sentence that's tricky. I suppose for an indicator common sense suggests it's just plain safer to stick with a lower power LED and run it for a longer duration. So, the challenge will be to find high lumens/watt but at not too high absolute lumens.

  • Hero Member

    Closing the loop, I finally found the information I was looking for: "Also note from Figure 17 that — if the flashes are separated by at least 20mS — the brain ‘sees’ the individual flashes at full brilliance if they have a duration of 10mS or greater, but sees them at diminishing brilliance at durations below 10mS (a 2mS flash appears at roughly 1/5th of true brilliance; the perceived brilliance falls off rapidly at durations below 1mS). The perceived duration of a 20mS flash (30mS) is only 50% greater than that of a 10mS flash (20mS)." https://www.nutsvolts.com/magazine/article/practical-led-indicator-and-flasher-circuits


  • Hero Member

    Reporting back:

    Testing some cheap aliexpress 3mm red or yellow through-hole LED's, I found they required around 10ua to have decent visibility. Other LEDs that I got from amazon required several hundred microamps to be equally visible. However, running this 25 cent CREE green LED at 1ua at around 2.5v was sufficient to be dimly visible: https://www.mouser.com/ProductDetail/941-C503BGASCB0F0792
    Even 500na was enough, though at less than that it becomes dodgy. Part of the credit goes to narrowly focused optics.

    Also, 10ms was enough. So, 1ua at 1Hz with a 1% duty cycle means just 10na on average. Not bad! The pulsing actually makes it easier to see than it just being constantly turned on, I presume because of better contrast.

    I haven't yet tried it at night, but so far this is already better than what I was expecting a priori. Putting it at the bottom of a black tube might also be interesting to try: you'd get improved contrast, and with the narrow optics maybe little light would be sacrificed to the blackness of the tube

    Anyone here found an LED that's more efficient than this one? Or is this already about as good as it's likely to get?

Log in to reply

Suggested Topics

  • 1
  • 8
  • 3
  • 3
  • 19
  • 933
  • 1
  • 15