Why 100 Watt eBay LEDs Are Not Your Friends

LEDs are amazing things. As time marches on, they are being used in more and more lighting applications – everything from household bulbs to automotive headlights. But the push for smaller, cheaper, and brighter LEDs seems to have hit a snag for some of the less reputable manufactures.

Case in point, [bigclivedotcom] has been testing of some 100 Watt LEDs from eBay. When these LEDs work correctly, they put out a face-melting beam of light that you wouldn’t dare looking into (picture the scene from Raiders of the Lost Ark). They also have some unusual specs for an LED, like running on 30 Volts – and that’s a lot compared to the forward voltage of most LEDs at around 2.5 volts@20 mA.

So what gives? Well many of these high-wattage LEDs use a string of several LEDs in series. And as [bigclivedotcom] points out, this can be a real problem when a few of the LEDs begin to fail and act more like a low value resistor than a typical LED. In the videos after the break you can see [bigclivedotcom] test the LEDs to get a better look and what’s happening and why.

93 thoughts on “Why 100 Watt eBay LEDs Are Not Your Friends

    1. same here, I put one in a lamp housing with an old pentium 4 heatsink sometime in 2013. I use it as a worklamp when doing automotive work. runs like a champ, puts out a decent amount of light for the power and i’ve had no issues.

    2. Did you have fun showing that the led’s from China are crap. If you have this time and so smart as to what current should run through, perhaps use your time more wisely and make your own led’s. Bet you end up going to china for a chip supplier. Hope they don’t sue you for defamation. No comment section so I replied to this one to stat this, about his comments on shitty 100 watt leds. WHAT A DAM JOKE, CRACKS ME UP WHAT PEOPLE DO WITH THERE TIME TODAY. WOW

      1. It isn’t defamation if it is provably true, and the phenomena has be well demonstrated in the videos. If there are any legal repercussions it would be consumer rights laws used against the “bottom feeder” suppliers for knowingly reselling “factory back-door” reject modules on ebay when they were supposed to be destroyed.

        Do you honestly think that people haven’t worked out what is really going on by now?

      2. +1 actually went to bigdaddy comment.
        I cannot understand the “fury” against some Chinese products because, after all, the almighty and venerated Apple iPhone is ALSO made in China!!
        Makes you wander …

        1. Are you retarded? When did research and fact checking become bad? Some of us actually try to build real stuff that actually works and need to meet some efficiency standards. This is not a rant against Chinese components (I’ve sourced leds from china now, thousands of single chips I can bin myself, and _they_ hold up) – it’s a rant against shit products that don’t do what’s claimed! I was saved a lot of time by these tests – and a lot of money.
          What the fuck do you guys do? Post bull shit in response to peoples well invested efforts just because you bought crap products and are in denial? You and bigdaddy should just shut the fuck up if you’ve got nothing good to contribute, instead of bashing solid people that I rely on that do actual research.
          Fucking low life retards!

        2. So where can I buy good 100w led that light evenly? All I am seeing is that the cheap led plates are probably cap and I don’t see much difference between the sellers of the $5 and the $20 led plates. I don’t want to buy a crappy $2 one for $20

          1. If you can’t figure this out, you aren’t competent to design a product using 100W LEDs anyway. They are of little use unless a very high intensity light has to be put in a very small area with well engineered active cooling, like in a projector. Otherwise take a hint from the major lighting manufacturers to use smaller wattage LEDs and spread them out so that the thermal density is reduced.

            Realize that the generics have resistive shorts in them so what they call a 100W LED start out producing about 60W worth of light, then as each die in them with lowest forward voltage fails one by one, they end up producing very little light at all, just become a space heater until they completely burn out.

            Realize that what you are buying on eBay for dirt cheap are factory rejects. No major manufacturer would accept these LEDs and put their name on a product using them. There are some that look identical that aren’t defective but you won’t find them dirt cheap shipped around the world.

            More to the point, if you want a quality LED then buy from a known reputable manufacturer. Philips, Osram, Cree, etc. If you want 100W output for general lighting, consider using 10 pieces of 10W LEDs.

      3. If you were sane you would realize that he is pointing out issues of interest about specific LEDs bought from a specific website, NOT some blanket statement that ALL LEDS coming from China are crap. It is a VERY GOOD use of time for one person to thoroughly test something then present the findings to the masses for distribution of info, peer review. and so people know what they are getting up front because specs provided for these LEDs are laughable, you literally have to guess what you’re going to get, if they even fill the order correctly. Fortunately there are a small # of different designs for the 10W, 20W, and 100W with one design each being most likely.

        As for lawsuits you have it pretty backwards. The vast majority of generic LED products coming out of China have fraudulent specs like claiming a 1 xAA flashlight is producing 700 lumens from a counterfeit/fake pseudo-Cree 3W LED driven at only 2W. It truly is ludicrous like they’re just making things up as they go along. At least it IS a flashlight or an LED whichever the case may be, instead of a duck or bowling shoes. lol. Clive called it appropriately with the phrase “Clown Science”.

        Know what is ironic? You’re writing about what someone else does with their time, whining about it by spending your own time doing absolutely nothing useful. If you would like to test LEDs and demonstrate to us the virtues you find, we will watch your video and appreciate it. Hint: People stop watching videos they don’t appreciate. Nobody has a gun to their head and is being forced to watch them.

        1. I bought a 4 cob “800 watt true” from this pos company called anjeet. You can actually test the capabilities with a few tools less than 25-35 buck. They know this. Buttttttttttt. I tell em theyre selling shotty product, and can prove it, and all i want is my $ back and for them to pay to ship. They counter with ” well thats our mistake, you are completely right, and how bout you keep the light and full refund? But only by removing negative feedback and never ordering or giving feedback on that particular site ( ebay amazon) under that account. The fact is this. These chinese investors, nay, any type of investor, not to sound racist..they exploit the cheap materials, labor, overhead, etc and get huge profits by distributing the false specs to people like us. Sure it looks too good to be true. 120 buck for a 800 watt cob 4x2oow. The company i bought from will have too many people leaving shit feedback (even if they offer the money back AND a free light. which i will be using the componets and 4fans, driver, etc for a diy cree build) THERE WILL ALWAYS BE AN UNSUSPECTING CONSUMER THAT WILL BE TAKEN ADVANTAGE of. Thousands most likely. The next month or two there will be anothercompany under a different name offering this product better when in reality..its the same monster. Not being held accountable. Just a different company name. Think. You will get what you pay for. If the light specs look tooooo good to be true and the company is an up and comer from china withmaybe no reviews or some serious street cred. Dont do it. The free supposed 800 watt led ful spec they gave me for free to remove truth prob cost the pennies. And most likely were made by little kids. Just sayin

      4. The entire business culture in China is rotten.

        I like Chinese people and cannot wait to visit China – but lets be honest, in USA where we print the worlds money the businessmen have run amok with corruption. In China where they make the worlds products, the same is true.

        This is what happens when vicious, evil, war mongering people run countries.

        1. This stuff is actually the fault of the west for accepting trade agreements designed to cause our lifestyles and business regulations to match those of developing countries.

      5. He’s NOT saying
        “Don’t buy LEDs from china it’s all crap”
        What he IS saying is
        “If you buy cheap LEDs from China, you get what you pay for”

        Clearly demonstrated by the tests in the video, and the science / numbers add up. If you get CHEAP LED cobs from china, you can expect them to use more current / be less efficient than properly sourced ones.

        Also note, it’s not that you can’t get a lot of light out of a cheap “100W” LED – you still can get plenty of light. The thing is because of the inefficientcies within the cob you’ll waste a LOT of current. A cheap (defective) 100W LED and a properly-sourced 25W LED can put out the SAME amount of light, but the 25W will use a lot less power to do so

  1. LEDs in series are not a problem when used with a constant current driver. If an LED fails (which they really shouldn’t, if being operated within spec) it either fails open (and the entire series chain just stops working) or fails short (and the constant current driver just reduces its output voltage, keeping the same amount of current going through each of the individual LEDs as there was at the beginning). Series strings of LEDs are used all the time (most LCD backlights these days are a string of white LEDs being driven at anywhere from 18 to 50 volts).

    The issue with these cheap high-power LEDs is that they put multiple series strings in parallel. LEDs in parallel, without something to individually limit current along each path, is almost always a recipe for disaster. If you put two LEDs (or series chains) in parallel, you are depending on their resistances being exactly equal to split the current equally between them. But that’s never the case—one is always going to have a little bit lower resistance, so it takes more of the current. If this extra current is more than the LEDs can handle, they start to burn out.

    The next problem happens as a result of the burn-out. Because you’re pumping in current assuming that it’s going to split equally along all of the series paths, and now one of those paths is suddenly an open circuit because an LED failed open, the remaining paths are getting more current than intended. You can see where this is going—the remaining paths are now even more heavily stressed, another one fails, and now what’s left gets even more current. The arrangement is inherently unstable—rather than failures reducing the stress on what remains, things just get rapidly worse.

    These cheap high-power LEDs *can* be usable, but only if operated well below their “rated” power, so that even the lower resistance paths are still within a range of current that the LEDs can handle.

          1. “There is always someone who can make something a bit cheaper and a bit worse. Those who consider price only are that person’s lawful prey.”

            Also, although it’s not perfect, you can get some control over the runaway situation if resistance increases with temperature. IIRC I think this happens with LEDs, but no guarantees. I think I’d be fine with using parallel LEDs for some noncritical application.

          2. -“Those who consider price only are that person’s lawful prey”

            In the real world, knowledge costs money. You have to buy one to know that it doesn’t work, because you can’t trust what the manufacturers say about themselves or of each other, or what the people in general say about products, which means there’s always a market for crappy products.

            The only personal “remedy” to it is brand loyalty, which has the downside that the brands are then able to push a certain portion of crap as well because of this self-imposed inertia. As long as it costs more for you to find, evaluate and select a better supplier, you”ll stay regardless of problems and keep buying crap products.

            This is why e.g. on average 1/5 of all laptops sold break within three years. For something that may cost thousands of dollars, you would expect the failure rate to be significantly less, but it simply isn’t. Not even with Apple or Sony etc. They’re all bad manufacturers, and deliberately so, because they’ve calculated that the probability of any individual customer getting a lemon twice in a row is smaller than the critical mass required to hurt your brand’s established reputation.

            It takes a certain number of people to complain before anyone notices anything from under the general noise. Brands even deliberately add noise by marketing and hype, paying for reviews etc. and doing tricks like making a new model every year, so people couldn’t make direct comparisons, so everyone would have only the vaguest idea of how reliable or good brand X products really are.

            Caveat emptor doesn’t apply here. It’s a market failure in an imperfect world – not a fault of the buyer.

        1. Actually ‘cheapskate westerners buying Quality Chinese Engineering Seconds,Thirds and Fouths’. You get what you pay for. The expensive ones are at spec or better; these are the ones that still sort of work but fail one or more QC tests.

          I have a bunch of ceramic bowls I bought at a Dollar store. They had some forgettable name on a decal-applied label on the bottom. After a few washings, the decals started to come off, revealing ‘Ikea’ labels set in the glaze. They have odd pits and irregularity in the glazes here and there. Perfectly useful for bowls, but not up to Ikea’s standards. Rather than throwing them away, they’re relabelled and sold at a hefty discount.

          1. CRT manufacturers used to sell their duff-tubes through subsidiaries that sounded deliberately like cheap knockoffs so people would think they had nothing to do with the parent company. Things like Samsung -> Samtron.

            Nowadays with LCDs being made by about three different companies anyhow, the duds are simply put into the mix. The quality checks at the end manufacturers are done by spot checks instead of testing every single unit, so they can pass a certain percentage of bad product as long as it’s dispersed well enough in a larger mass of working units. Then they end up with the consumers, where a few dead pixels or off-colored backlights usually go completely unnoticed.

        2. Ehhhh, if you watch the videos you’ll see that Clive comes to the conclusion that these are likely name brand chips that somebody dumpster dived out of the Q.C. Failed pile. One way or another these chips end up on the market at all sorts of price points, paying more isn’t a guarantee of a product I think we all learned that from the FTDI Bricking scandal and even buying from reputable sources is no guarantee anymore, See Mouser carrying Adafruit’s marked up Chinese crud (they are using the same product pictures as the AliExpress sellers even).

          1. The fundamental problem is the logical design, not QC/rejects. They make the dies and place and wire them, THEN because there are multiple dies in series AND multiple series in parallel, almost NONE of the completed LED packages (people refer to as 1 x 100W LED) are going to have perfect matching current across all LED dies within. Not even their BEST specimens would in a quantity that could be depended upon to design something with trust in that. Any more than one parallel LED without laser trimmed current limiting resistors in series with each, custom to each individual package, is a design failure.

    1. When a product fails, as in a led string gives up the ghost, its nice that the entire module isn’t a write off immediatly, instead over driving the remaining strings is a very good value for money as it is still soldiering on, maybe shorter than ideal, but you’re not left in the dark. (The glass is half full, not half empty on this one)

      1. It’s not a good value for the money to have the failure result in greatly reduced lifespan of the rest of the product. Being left in the dark is definitely something to avoid, but there are other ways to design with this in mind like at least two parallel PSU separately powering parallel LED packages so if the power supply fails (really, who is going to use a top end HQ expensive PSU for a dirt cheap generic Chinese LED?!) instead of the LED, OR the LED instead of the power supply, THEN you still have the other 50% producing light still.

        The key thing here is that you use twice as many 100W LEDs and run each at 50% or lower below rated power. This will also reduce heatsink cost at the expense of a slightly larger encloser and yet, a simpler longer heatsink extrusion for the greater mounting surface area of double the LEDs, also tends to be cheaper than one that has to be machined with fins and with a thicker base for higher thermal density dissipation.

    2. It does sound however that the system is inherently stable to some degree. If one chain gets too much current because the other chain has 1 led burnt out, then the weakest LED from the overdriven chain will burn out; the whole system will be back in balance.

      Your points are still overall valid and I didn’t really consider the parralel LED issue.

      1. The problem here is that when the next weakest LED fails and the “over driven chain” fails, the rest of the LED strings are now given even more amperage, and driven even further above rated current.

        The constant current supply is only going to regulate the current to the whole module, not to each string of LED’s in the module, since they are all wired in parallel. It doesn’t know if it’s driving one monolithic LED at 10W, or 1000 smaller LED’s in some mix of parallel or series strings. It only supplies the rated power, and if the module suicides on that power, it doesn’t know anything is wrong until the whole thing goes complete short circuit, or open circuit.

      2. Except that’s not what actually happens. The LEDs continue to burn out one after another and the more there are burnt out, the lower the chances that the average forward voltage will be equal between the two. Mathematically speaking a smaller # has higher variance from an average Vf than a larger # does.
        The second problem is that something suspicious is going on. LEDs blow OPEN, so the whole series should have failed when one does. How do they avoid that? Seems like they have a resistor in series with every LED die so while it consumes 10W, it wastes a far higher % as heat, heat which also contributes to the failure. It’s like they build them to self destruct. Ironically even the cheapest of generic drivers tend to claim a certain wattage then provide less than that when used exactly as specified.

    3. You’re absolutely wrong about the constant current driver! Say, we have 5×2.5Vx20mA LEDs in series, connected to a 100mA CC driver. So its output voltage would be about 12.5V@100mA. If an LED fails, there would be 4 LEDs and still 100mA, giving each of them 25mA, above their specifications. While it’s true that the voltage would be decreased from 12.5V to maintain the constant 100mA current, but it wouldn’t be the optimal 10V for 4 LEDs either. This probably won’t cause any serious damage, but imagine if 4 LEDs die and only one remains, we’d have an LED with 100mA instead of 20mA. The voltage would be about 5V instead of 2.5V. Your claim is only correct only if we have separate current source for each LED.

      1. When you have 5 2.5V 20mA LEDs connected in series, you don’t connect it to 100mA CC driver – all LEDs would get 100mA and fry. You connect it to a 20mA driver. Output voltage is 12.5V@20mA, if one LED fails 4 LEDs are still getting their rated 20mA, at 10V total, 2.5V per LED.
        When 4 LEDs die, you are left with 20mA@2.5V, exactly what that last remaining LED is rated for. For driving a single string of series connected LEDs, constant current driver is indeed the right choice.

        Problems you are describing are correct for a set of 5 parallel strings of 5 LEDs in series, but problems with potential cascade failure mode of parallel LED strings were mentioned already.

        1. People keep writing “if one LED fails” as if it is expected to fail shorted. It isn’t expected to fail shorted. That is usually a manufacturing defect while the later resultant cascade failures will eventually cause them to fail open, not shorted.

      2. Please review Kirchoff’s circuit laws before calling me absolutely wrong. Current is not divided between circuit elements in series; as there is only one path through the series for current to flow, whatever current is flowing through one element must be flowing through all of the elements in the series. Connecting a 100 mA CC driver to a series string of 20 mA LEDs is going to push 100 mA through all of them, regardless of how many there are. They can also be completely different LEDs with different internal resistances and forward voltage drops—they are all going to see the same 100 mA current (provided that the total forward voltage drop of the string is within the limit of what the driver can provide). The *voltage* is split across elements in series proportional to each element’s resistance, but the *current* is always the same.

        What you say would be true if the LEDs were in parallel: the 100 mA would be split between them, and it would only take just over 2.5V to push that much current through the parallel set. The amount of current that goes through each LED in parallel depends on its internal resistance. If they are all truly identical, the current will be equal, but in practice, small variations in internal resistance are going to mean that some will get slightly more than 20mA and some will get less. For elements in parallel, the *current* is split between them proportional to each element’s resistance, and the *voltage* will be the same across each element.

    4. All COB (Chip On Board) LEDs are actually many series strings wired in parallel, not just the cheap ones; check out a datasheet from Cree, Sharp, or Bridgelux. The issue is that the cheap LEDs have poor Vf consistency so the parallel strings don’t share current equally because of the Vf mismatch.

  2. can that burn out the driver?

    i have a ge led light bulb i got at walmart and after a year’s use it went dead and after removing the led module i wired it to a series of 3 9 volt batteries and 1 12 volt ups like battery getting about 40 volts not all of the leds was lit so i thought one burned out and made the driver fail by being overloaded.

    here it looks like ge may be using the same “duff” leds or they was driven too hard.

    i also noticed the metal part was hot.

    1. It is not easy to speculate about this without seeing the driver circuit design but generally no. Generally they are a constant current driver and burn out of a single LED or even the resultant open circuit for a while series would just make it easier for the driver to produce same current at lower voltage. If it were too low it could destabilize, but at that point the bulb has already failed in a practical sense and isn’t cost effective to repair back to same state since you would expect it to fail again with same design in same environment. All you can really take from it is the same as with any product, that a single thing can fail and leave the rest of the components viable, though in the case of AC/mains connected light bulbs, power surges are a common cause of damage. They rate them for 10’s of thousands of hours but a single surge on day one can still fry it.

  3. run them at 80% and you are fine, just like everything else that is really cheap and from china. never ever expect any cheap ebay products are what they claim they are. China sellers are kings at lying.

    1. I love looking at the Chinese sellers claiming 2000 Lumens from a Cree XM-L LED, even though the Cree datashet specs it at about 1000 lm (that’s not accounting for a true FL-1 standard reading either). I think they are probably rated at 16 MP VGA video resolution too (you’ll understand if you but a lot of Chinese stuff)

      I should buy one to see how it stacks up against my FL-1 320lm rated flashlight…

      1. Are you still assuming it’s really a Cree?

        If I buy it direct from China, I don’t assume the brand is actually what they say it is. To that end, the data sheet is really only useful for p8nouts and the oike and even then marginally so. Makes it awkward when it’s Chinese only crap though…

      2. I saw a nice stereo Bluetooth earpiece from an Ebay supplier. A single earpiece, in one ear, “stereo”.

        I asked them what they thought “stereo” actually meant, the guy who could write the best English in the company informed me “stereo means can play music”.

        No idea about the quality of the earpiece. And it’s still better English than half the English speakers on Facebook. But seems like the future is some horrible pidgin of English and Cantonese that nobody can spell.

        1. I was looking at this yesterday as I want a single earpiece for listening to music while exercising.

          The ‘stereo’ unit does phone calls and music. The ‘mono’ unit only does phone calls. I assume it doesn’t use the A2DP protocol.)

        2. Keep in mind that there are plenty of Chinese with a good education and good English, but they tend to be able to get better employment than having to sell items for pennies profit at a time on eBay. In some ways it is admirable that these impoverished sellers can speak a 2nd language at all, or perhaps some are using online translators that can’t get it right either.

    2. Ebay has a 100% money back guarantee, you should always expect what they are spouting.
      I have been burned by a few buyers buying something used from me and then getting their money back, if thats the case you can certainly get your money back from a Chinese seller.

      When you get your money back, you get to keep the product anyway as the Chinese seller does not want to pay shipping back to china. Obviously this can be abused, but being honest if you really didn’t get what you paid for open a claim and get your cash back.

      1. Problem is in the return, and many Chinese sellers know this. Return with tracking is required for eBay’s money back guarrantee, and shipping to China is quite pricey, sometimes more than what the product is worth. No return = no refund.

        eBay of course does require seller to provide return label on SNAD claims but it is not enforced unless the buyer calls eBay because Chinese seller can’t buy USA shipping label that the buyer can print and use. eBay being weird without considering all angles like international seller footing the cost of return postage.

  4. If you buy some cheaper muti-LED flashlight from eBay (there are 3LED, 7LED, more LED variants) those LEDs are always connected in parallel and driven at about 30-50% of the current they can handle. That however doesn’t stop seller to mutiply number of LEDs with maximum output of single LED and write BS like “9000 lumen” for lamp that outputs merely 2000 lumens.
    Driving them properly would required much more electronics, one choice is to have separate driver for every LED, some 3LED flashlights had it in the beginning, but they realize it’s cheaper to connect them in parallel and drive with one driver. Second choice is to wire them in series and use boost converter to provide enough voltage and current to drive them safely. Brand name flashlights like Led Lenser do it correctly but cost much much more than Chinese ones.
    Cost effective option is to buy eBay 100W LED, drive it as it was 50W LED and in the end you have much light for not so much money.

    1. If I remember correctly, he explains in the video that he was intending to use them at a lower wattage but gave up because they were just so much worse than the actual LEDs rated for that voltage.

    2. Ironically, LEDs should only be run at 30-50% of their rated power anyways, because the rated figures are for 25 C junction temperature, which remains true for about 1 milliseconds after switching on. After the junction heats up, the lumen output drops, the efficiency drops, the lifetime, the color temprature, etc. all changes.

      The whole LED industry is standardized on that little 25 C fib. With the more honest manufacturers, there’s simply a derating formula in the datasheets that tells you what the LED will actually do when you actually use it, which is why no 100 W LED even from a reputable manufacturer is -actually- a 100 Watt LED in practice.

      1. Errr no. Every datasheet specifies Ta = 25degC. That’s AMBIENT temperature. Good datasheets will then go on and give derating factors for above ambient temperatures and the thermal resistance values required to keep Tj at the correct operating conditions. I.e. you bolt a heatsink on with the correct thermal characteristics as per the datasheet and then drive the LED at full power if your room is at 25degC. Most of them only need to start to derate if you have an ambient temperature above 40 degrees or you intentionally balls up your heatsink design.

        You most absolutely can drive any reputable LED at its full rated wattage at room temperature provided you have the appropriately rated heatsink on the back.

        And I don’t think a single semiconductor device I’ve ever seen has a rated junction temperature of 25degC. Most of them including LEDs are rated to around 120degC. Problem with LEDs is that thermal runaway will cause you to hit that in the blink of an eye.

        1. No they often rate for die temperature. Ambient is meaningless as it depends on the design heatsink and most datasheets don’t even bother trying to specify a heatsink, assuming that if you’re this dug in, you don’t need to be told how to calculate a heatsink given the specs for the LED and knowing your target Ta.

          Rated junction temp of 25C does not mean you have to try to keep it at 25C. It merely means that is the temperature at which the other specs hold true like lumens. It is mandatory to spec die temperature as a constant for the other values even if they do provide graphs for deviation from that, just like they have to spec a specific current because both have statistically significant effect.

      2. All the modern Cree LEDs are actually tested and characterized at a junction temperature of 85 C. They also provide estimated outputs at 25 for some of them, but all their production testing and binning is done at 85.

    3. Or just parallel 9 5mm white LEDs and directly drive them at 4.5 volts from three AAA cells. I want to find a source for those AAA cell holders without having to buy a flashlight. They’d be ideal for lightsaber power but then what do I do with the rest of the now useless flashlight?

      1. and watch those LEDs burn out one after another like they have done quite often in the 3xAAA flashlights designed the same way. It’s quite easy to find those on eBay for a buck or two, but you might consider just using a Li-Ion cell instead.

  5. Is it really as bad as the video indicates? In the first video the LEDs are used at a current of max 20mA. That’s less than 1% of the rated current (which is approximately 3A for 100W at 30-35V). It is very well possible that the LEDs light up a lot more equally at 10-50% of the rated current. By the way some LED manufacturers explicitly recommend not to dim the LEDs below a certain current (and use PWM instead if you need even less brightness). So it would actually be nice to see photos of the LEDs at higher current. Taking the photo will probably require a strong ND filter and some kind of heatsink (but you need that anyways if you actually want to use the LED).

    1. Only if the LEDs have leakage do they light up at different times in a series, and if they do then they won’t actually light up more evenly because each string gets a different amount of current, as explained in the video.

      The uneven current distribution is a dealbreaking fault, because it will cause a cascade failure sooner or later. Even if you de-rate the lamp, the leaking string will overheat and fail pretty soon, and then the next, and the next…

      1. Again wrong. Lightness of LEDs and the forward current at varying forward voltages are NOT linear. Effects of discrepancies between brightness vs current and by extension imbalances between strings due to current vs forward voltage will DECREASE as you approach the normal operating point of the LED.

        The problem is most definitely exaggerated when underdriving the LED matrix. I will give you one point for a change for noting correctly that uneven current distribution can cause a cascading fault but you can keep that in control by slightly underdriving the LED, but certainly not 50% underdriving like you claimed in another post.

        1. Wrong. As the current through the LED increases, so does the forward voltage, and so more current shunts through the parallel fault resistance. Heat also increases the forward voltage of the diode for the same current, but less so for the resistance, again sending more current through the fault instead of through the diode junction. These effects are largely compensating.

          The faulty diode will always remain less bright than the rest. It’s just difficult to see under the glare of the other dies.

          And the 50% comment has nothing to do with this case in particular. LEDs in general are specced at 25 C junction tempearture and that’s a fact. It’s a vestige from times when LEDs had outputs in the single milliwatts at best, so they could not physically heat themselves much above ambient temperature.

          For real world operation they have to be de-rated to actual operating conditions, and if you run them directly at their rated specs you’re not going to get the specified output. In practice, there’s no way to make the LEDs actually live up to the claims for more than a few seconds. For maximum efficiency (lm/W) and lifetime, they typically have to be run at 1/3 the rated current, up to 1/2 the rated for more light at the expense of lifetime.

          At 100% current, the thermal resistance between the junction and the diode casing is too high to remove heat even when you’ve properly heatsinked the chip, so the junction temperature goes up anyways, the luminous output goes down, permanently so after a couple minutes, and the diode dims/burns out in a couple thousand hours.

          1. See a typical example:

            The temperature of the diode chip running at a significant power is usually some 15 degrees higher than the heatsink, and the junction temperature inside the chip some 30 degrees higher than that, so when your heatsink is at 55 C, your junction is already somewhere around 90 C, and the diode suffers efficiency loss and reduced light output from the heat.

            So yes – in a sense – when you increase the current through the string with a faulty diode, the brightness does tend to even out near the full rated current because you’re overdriving the other LEDs

          2. While you’re right about the relationship between a diode’s forward voltage and current, the temperature coefficient is actually negative. So increasing the junction temperature reduces the forward voltage.

        2. Nope Garbz, I have built a few things with these that under drove the LEDs by more than 50% and they still failed. Ironically a couple still had one die lit from the leakage current but were only consuming about 7mA and 15mA. I think I got as little as 200 hours out of many before enough dies had failed to make it dim enough to notice.

          I’ve given up on the 100W generics and won’t even use the 10W generics at more than 2W. That’s how badly mismatched the leakage current makes them.

  6. I’m sorry that these wonderful LEDs don’t fit into your expectation for how the world works. Under load, once you overcome the internal resistance, it become negligible in practice. The LEDs work exactly as prescribed* when under load. I understand you (and many engineers like you) have certain expectations of how the world “ought” to work, but the world has moved on. We’ve decided that allowing for some resistance across our LED is A-okay.

    TL;DR Junk is the wrong word. Financially efficient is probably a better one.

    * = Apply the 25% Chinese data sheet rule (De-rate all extremes by 25%) I.e. never operate a 100W LED past 75W.

  7. OK moving on from the econo-political issues others have focused on, where can I find a good tutorial on how to get the smaller single units for the lowest price, test them, and then build a driver circuit so I can gang up as many as I need in a way that manages the current flow better?

  8. Love the Coco (Koko ?) the clown science explanation. Thank you.
    I’m sure this didnt just start happening but has been going on for a while and involves the calculus of seperating as much money from people as you can for the least value given you can manage.
    The parallel string architecture is designed to fail at the weakest link.

  9. @ Mike Lu
    I bought two types of 100W LEDs (10×10 Chips) at one shop, one type was warm white and one cold white.
    The price for one LED-Module was: 4,32 Euro.
    The warm white 100W LEDs were totally garbage, they had unselected chips and because of this a very different current flow over the LED-Chips!

    The cold white 100W LEDs did have selected Chips and they are lit up all equal bright and to the same time, but this chips are smaller then normal chips.
    There are chips with a edge length of 30mil and 45mil. The chip-area of the 45mil chips is nearly double so big as the 30mil chips.

    Here is a video from other not selected 100W LEDs which I have bought earlier.

  10. The encapsulation is flexibe!, the bond wires will break if the encapsulant gets even so much as touched.
    He handles them poorly. nothing must touch the yellow. Fine gold wire bonded strings in hard jelly!
    too much whiskey for me!

    1. Remember the scene toward the end when the bad guys ceremoniously open the arc with our heros staked out as intended victoms? Remember the head bad guys faces melting?
      That’s what he’s talking about.

  11. If they’re parrallel connected wouldn’t you simply be better driving them with a constant voltage, 1 failed led would result in a string failing in short order. odds are one will blow open. If not some sort of cheap overcurrent protection (fuse) should cover it all and you loose the lot, instead of a string. Sounds perfectly fine to me.

  12. Never ever, never ever ever ever have series parallel LED strings. Not at 2.3 VDC, not at an VDC. No LED company worthwhile makes such a part because this is EXACTLY what can happen.

  13. *Reads Article
    *Watches Video
    *Reads Comments and Replies
    *Nods head in understanding and approval…

    So, Can anyone give proper recommendation and maybe extrapolate their experience on what are good affordable suppliers of 100 Watt LEDs?

  14. Interested in a 100w COB chip horror saga?

    I bought qty 220 100w Bridgelux COB LED high bay lamps with Meanwell drivers from a factory in China. Within a year, one chip had failed. I convinced the factory to allow me to conduct the warranty (5-year) repair here instead of us both incurring the expense of shipping the entire lamp units to the factory for repair (my cost to China, their cost back). They pre-soldered the leads onto the chips and shipped 4 chips to me. The 4 chips they sent were the wrong CCT. OK fine, at least the repaired lamps were working for now; I’ll just have to replace them again with the correct CCT chips later. Then 2 more chips failed, so I had the factory send 10 more chips just in case, but this time, getting them to send the chips was like pulling teeth.

    Understandably, the vendor wants to know why the problem is occurring. I made video of the lamps in their location, checked ambient temperature, and kept the old chips. They also want me to ship them one of the defective lamps for diagnosis, but I’m asking them to foot the bill for shipping, considering I’m saving them TONS by doing the repairs myself. I also offered to include one lamp that I have repaired so that they can inspect my repair job.

    The bottom line in this warranty resolution process is that the aggravation of me repairing the lamps was less painful than the expense of shipping. Problem solved? Not quite.

    I got a call from my customer today that 13 more lamps failed within the last 2 weeks. This is a total of 17 failed lamps in 18 months. My customer is s#itting bricks. He trusts me that I will take care of him one way or another, but is afraid of his business being crippled by a mass chip failure that could happen any day.

    With the environmental factors controlled, drivers tested OK, the explanation for failure is one or both of the following: bad chips, improperly-applied thermal paste, or both. I’m now asking the factory to replace all 220 chips because 17/220 is way too large a failure rate to just wait for them to die individually. If I cannot get support from the factory – if they refuse to proactively replace all 220 chips – I will have to source them myself.

    Here is a picture of the original, bad, supposedly “Bridgelux” 100w chip. It has a maker’s mark of some sort. I do not know if this is a series mark, a mfr mark or what. Can anyone reading this identify it? https://gyazo.com/42e9165fd44f0ed32427ed5383c7c310

    If anyone is interested in seeing how this comes out, let me know, and I’ll update this post.

    Here’s the lesson: Get documentation of what parts are being used. What chip manufacturer, series and model #, who makes the driver, etc.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s