Hope you’re all doing well. I’m still new to modding, and I have a question about LED data sheets. How strict are the limits they list for an LED?
For example, on the Luminus SST-20 sheet they say the max drive current is 2 amps. Does that mean you will definitely fry it if you go to 2.1, or are they playing it safe, and you could technically run it at 2.2 or 2.5 for a short time? And if there is some wiggle room, how much is it?
Same question for the forward voltage. Some LED’s list a range, like 2.9-3.1, and some are more specific at 3.03V. Does the driver need to convert the voltage from the battery to be in that range, or are most LED’s (the ones we typically use for flashlights, anyways) just fine at 4.2V and they just need to regulate the current?
I believe they’re mostly the limit of where the manufacturer guarantees its claims.
If the manufacturer guarantees 98cri and a 50,000 hours lifespan, for example, that guarantee is only good up until the the maximum forward current. Doesn’t necessarily mean the thing will explode as soon as you go a little over the absolute maximum in the datasheet, but you might not get that 50,000 hours lifespan if you do
Very much playing it safe. LEDs for our use do not stay within the recommended settings, for example the sst20 dies at around 7A compared to the 2A rated maximum. 5A is a good current for an sst20, more than double the rated current
Manufacturers play it safe because thats the current where they can be sure that the LED will perform as specified for whatever the guaranteed lifetime is, but those runtimes for LEDs are intended for usually constant-on uses inside lamps and other devices. Luminus and Nichia are very conservative with their current ratings, usually 2x rating is safe for flashlight use, but CRI might go down and the lifetime will be shorter (not that you will likely notice, the LED probably will outlive the flashlight it is in)
As your example said, the sst-20 specifies 2A but the output maxes out at 5A, and the LED dies at around 7A
Usually I recommend looking here on BLF for a test from djozz or maukka, it will easily tell you what the max-output current is, and what the current that the LED is destroyed at (usually much higher).
Most LEDs burn off the extra voltage just fine, as long as the current is right, but some can be more fragile, like the sst20 deep red because of the significantly lower Vf (around 2.2v) and work better with a driver that burns off some extra voltage (linear regulated or a buck driver).
It’s interesting that some manufacturers do play it so safe though. You’d think they’d be in an LED output arms race with their competition and want to push the specs as high as they can comfortably go. If the spec is 2A but we know they can go to say 7A, that gap seems pretty wide to me.
The arms race isn’t as much in output, and more about lm/W and lifetime because lighting fixtures and lamps are where most LEDs end up going, turbo-lumens matter mostly to flashlight manufacturers and us enthusiasts.
Nowadays lots of lighting fixtures use large COBs which are cheap and efficient though instead of SMD LEDs, so I wonder how that will change. I guess it might explain the recent R9050 versions of the third gen XHP50 and XHP70 after years of low-R9 high-CRI Cree leds.
Ah, that makes a lot of sense. So in many cases the max listed current is probably where efficiency drops out of spec. I know for some LEDs the listed max current actually is pretty close to the realistic max. That makes the testing that some members here do all the more valuable.
My thinking is, some major company buys 2,000,000 LEDs for idk what, to replace all the street lights in LA or something. And some lights start failing at 49,000 hours instead of 50,000 hours because Nichia thought they could get another 40 lumens out of them and it’d be alright. But they guaranteed 50,000 hours in the datasheet and the city followed every guideline so now Nichia is being sued by California for breach of contract or something and they gotta warranty all 2,000,000 of these things or whatever it is they do.
And then there is something like statistics. My daddy used to say: there are lies, there are big lies, and there are statistics. Just kidding.
A lifespan of 50,000h does not mean that every one of those led’s will live exactly (or at least) 50,000h. Remember the famous Gauss-curve, als known as the Bell-curve? The total surface covered by the curve is 1, or 100%. It shows us the “normal distribution” of a certain object, event, performance, or anything. The big problem is, there is always a fraction to the far left side of the curve that performs below average.
Well, led’s are bought by the manufacturers of fixtures. Fixtures are bought by firms that deal in lighting as a commodity. And those firms are hired by cities to lighten up so-and-so streets with so-and-so lumens. As pointed out earlier flashlights are not the main goal for manufacturing led’s, but the firms that provide lighting as a product. And when things go wrong, every link in the chain will try to blame the previous link. The above mentioned Gauss-curve is linked to the spec’s of the led. and the led manufacturing companies are covering their little hineys by certainly NOT pushing the envelope.
Which we (flashaholics) are doing all the time. Will your light be used for 50,000h or for 5,000h, or less. Look around what fellow members have found when they tested the led’s beyond manufacturer spec’s. Caveat: there is one thing you always have to take very serious. Make sure the light can handle the extra heat.
No, of course they all won’t. But if none of them do thatll be an issue.
And Im not sure normal distribution applies here in the exact same way we were taught in school. I mean, that’s what quality control is for right. The ones that deviate too far from the mean are thrown away.
At the very least, they aren’t using 50,000 hours as the mean. That would mean half of everything they sell they’d potentially have to warranty
Say they are using normal distribution, then 50,000 hours I would think would already be a deviation or two away
Manufacturers have to specify for the worst case - a 100% duty cycle at high current with thermal design that removes just enough heat to keep the junction and die temperatures below the failure point. Flashlights of course run low duty cycles with brief spurts of intense current, thus can be considerably overdriven … within thermal limits (i.e. I recall some experiments with early Luxeons determining that the bond wires were the ultimate weak point, failing at 5A no matter how much cooling was thrown at the LED thermal slug).
And you see these factors at play looking at the differences between fixed lighting and flashlights. The former is usually designed to operate below the LEDs’ maximum current with significant attention paid to maximizing thermal performance due to expected high duty cycle. The latter is designed to be overdriven with the use case (or thermal controls) preventing premature failure of the LEDs and thermal performance designed around average demand and to briefly extend peak performance.
I recall seeing some reliability data from Micro Switch years ago on one of their snap-action switch lines rated to 10M clicks. Sub-10M failures were on the order of 1-in-100K. MTBF was just shy of 100M. These margins likely allowed for actual field reliability - with thermal cycling, moisture, dust, vibration, mechanical stress, some exceeding of operating limits - closer to 10M than 100M.
Engineering to this degree of reliability makes sense given that the cost of failing to meet warranty performance is more than just the cost of RMA’ing a ~$1 switch - it’s the cost of servicing or replacing units in the field.
For anybody still interested, I was looking at some LED light bulbs we got at work today, and on the back of the packaging they actually had an explanation of the 50,000 hour rating. It was the amount of time that it took 50% of the LED’s in a test batch to go down to 70% output.