LED overpowering limits?

[drummerdimitri]

I have the following led package:

https://www.aliexpress.com/item/Cree-XTE-100W-Led-White-Warm-White-High-Power-LED-the-thermoelectric-separation-Cooper-PCB-For/32366427694.html

I have slapped it on a 1 pound or so of aluminium finned heatsink with a fan at the back to be used as my projector to illuminate my workshop.

My question is how much can I overdrive these things? I am using a lux meter to see how much power I can dump into them before i start losing power as heat and stop there.

The heatsink temperature will not  rise above 40C at the rated 100W power consumption. How far is far enough do you think?

[Siwastaja]

The Ebay/Aliexpress LEDs vary a lot (this is not bashing, but from experience. I have used them a lot because the price is appealing, especially for hobby projects).

I don't recommend "overpowering" any LED, no matter how well you cool them (unless you go to extremes, think about liquid nitrogen or regrigeration based things). Why do you think their marketing would undersell the products? The title ratings, even in good brand products bought through proper channels, tend to be close to the actual maximum because the bigger the numbers, the better they sell. Everybody wants to operate near borderline "scamming".

For Ali/Ebay specials, I would actually recommend derating by 20-30% at least.

Other than that, no one can say for your particular LED. I have had DOAs, some with order of magnitude lower efficiency than expected (i.e, not much light output per power input, something resembling good old incandescent bulb), some that require derating by 50% to work reliably, and some have been absolutely great without any derating, or even slightly over the limits.

[DaJMasta]

+1 to no overdrive.

Generally speaking, rated current yields a shorter lifetime than I want, from a reputable source it should be the same lifetime they're rated for, but I like the idea of not replacing LEDs in my designs, so I always underdrive.

I've had some high power lights that overdrove well characterized CREEs by something like 20% with a substantial heatsink and designed for operation in open air without any enclosure.... and they died right around 3 years of use at about 8 hours a day.


The cause of death is basically always heat, and I don't think that style LED package is optimal for dissipating it well, so even with a beefy heatsink the die to case thermal resistance could be high enough as to cause significant problems even when the heatsink is fairly cool.  You could try to find a place overdriven where they operate well, but expect the lifetime to be low enough that they will need to be replaced in the nearer term.  I believe with most LEDs the peak efficiency is actually below rated output - by the time you're running full rated current into it the percentage of power lost to heat is already well over where it operates most efficiently.


I wonder if something like a thermal camera would be good for checking out die temperature more closely.  You could also try mashing a thermocouple in there and just sacrificing one by gradually increasing current and mapping light and heat output.  By plotting them you should get an idea of where the crest of performance is in the module, even if it costs you one.

[Zero999]

Generally I agree with the above: don't overpower.

However, I've had some luck driving LEDs at well over ten times their rated power for very short pulses, so it can be done in strobe/flash applications., but not continuously.

[wraper]

I would be surprised if at 100W LEDs are not already overdriven. You should think about their "power rating" with skepticism. Not to say than most of such cheap modules consist of faulty LEDs.

https://youtu.be/hR8iu5X17kE

[drummerdimitri]

+1 to no overdrive.

Generally speaking, rated current yields a shorter lifetime than I want, from a reputable source it should be the same lifetime they're rated for, but I like the idea of not replacing LEDs in my designs, so I always underdrive.

I've had some high power lights that overdrove well characterized CREEs by something like 20% with a substantial heatsink and designed for operation in open air without any enclosure.... and they died right around 3 years of use at about 8 hours a day.


The cause of death is basically always heat, and I don't think that style LED package is optimal for dissipating it well, so even with a beefy heatsink the die to case thermal resistance could be high enough as to cause significant problems even when the heatsink is fairly cool.  You could try to find a place overdriven where they operate well, but expect the lifetime to be low enough that they will need to be replaced in the nearer term.  I believe with most LEDs the peak efficiency is actually below rated output - by the time you're running full rated current into it the percentage of power lost to heat is already well over where it operates most efficiently.


I wonder if something like a thermal camera would be good for checking out die temperature more closely.  You could also try mashing a thermocouple in there and just sacrificing one by gradually increasing current and mapping light and heat output.  By plotting them you should get an idea of where the crest of performance is in the module, even if it costs you one.

I have taken the below picture of the LED package. The strip on the right is electrical tape to get an idea of the heatsink temp which was about 47C. Are these good thermals? I am running a 12V fan at 5V for silent operation.

[Siwastaja]

Doesn't look too bad IMO.

[Nusa]

I suggest looking at the datasheet for starters, with the assumption they are what they say they are.* After that, you're officially on your own. You'll know you've gone too far when it fails.

https://www.cree.com/led-components/media/documents/XLampXTE.pdf

*If they aren't no answer we can give will be right anyway.

[DaJMasta]

Doesn't look too bad IMO.

Agreed, I don't know enough about thermography to know if there's emissivity differences with the LED plastic housing or whatnot, but mid 60s die temp would certainly suggest headroom for pushing it further.  How much work does a failure due to overcurrent cost you?  If not too much, could be a worthwhile experiment.

[drummerdimitri]

Doesn't look too bad IMO.

Agreed, I don't know enough about thermography to know if there's emissivity differences with the LED plastic housing or whatnot, but mid 60s die temp would certainly suggest headroom for pushing it further.  How much work does a failure due to overcurrent cost you?  If not too much, could be a worthwhile experiment.

I was able to push 400W into the damn thing before my lux meter started reading lower light output beyond that point.

Will be running it at 100W for now and see if the light output is enough for my needs otherwise will consider pushing it to 150W and increase the fan speed a little bit.

[sokoloff]

Would running two units at 75W each be a better plan for a one-off?

[Zero999]

Doesn't look too bad IMO.

Agreed, I don't know enough about thermography to know if there's emissivity differences with the LED plastic housing or whatnot, but mid 60s die temp would certainly suggest headroom for pushing it further.  How much work does a failure due to overcurrent cost you?  If not too much, could be a worthwhile experiment.

I was able to push 400W into the damn thing before my lux meter started reading lower light output beyond that point.

Will be running it at 100W for now and see if the light output is enough for my needs otherwise will consider pushing it to 150W and increase the fan speed a little bit.

Yes this is known as the efficiency droop. Powering LEDs at higher currents is seldom a good idea, since efficiency drops at higher currents and junction temperatures. I did get away with it, when using the LED as a strobe, but that's because the duty cycle was so low, heating was negligible and efficiency wasn't the prime design goal.

[Siwastaja]

It's a tradeoff:

Having more LEDs for the same light output results in better efficiency and lower energy cost, but the LEDs cost more.

Driving fewer LEDs more strongly results in less investment in the LEDs, but energy cost is higher due to lower efficiency. If taken too far, the LED lifetime shortens so much that it's not wise investment-wise either.

For short duty cycles, optimize for higher power; for 24/7 operation, derate more.

Sweet spot for money spent on LEDs vs. money spent on electricity tends to be roughly around 75% of the nominal power rating for general illumination purposes. OTOH, for Ebay/Aliexpress parts, the "nominal" power rating can be anything.

I'd say typically, driving at 200% power results in around 130-140% the light output. It stops making sense really quick. For pulsing, that may still be desirable, I mean, short pulses at 1000% rated current for 250% the light output does sometimes happen. I'm working with a product where the LED investment cost is meaningful. Efficiency is important, too, but high product cost would be a showstopper; hence the need to drive the LEDs quite hard.

[schmitt trigger]

Many moons ago, I had access to a lab where they had the equipment to measure lumens, including a large integrating sphere.

I was there as a lab assistant helping characterize the lumens vs watt output.

If I remember correctly, the lumen/watt ratio started to decrease at over 80% of their rated power.
Not only that, but as the LED warmed up, the efficiency would also drop.

Thus, I will add to the chorus: No, don't overdrive.