Extreme heat from a 100W LED COB. I melted my thermometer!

[lmester]

I bought some cheap 100W LED COB's and have been testing them. They have 10 parallel sets of 10 series LED's. I mounted one on an old Pentium CPU heat sink. I ran it up to 100W power and used my infrared thermometer to check temperatures. Heat sink temp was good. Only about 90 Deg F. I then put the thermometer in front of the LED and was moving it around to find the hottest spot. I smelled hot burning plastic! The temperature reading was 400 Deg! The front of the thermometer was melting!

Is this normal? The only data sheet I can find for these generic Chinese LED's shows a max temp of  80 Deg C (176 F). I'm getting temps far higher than  that.

Possibly the LED's are really efficient and pump out most of their power as light.

Maybe I should be asking what the efficiency of an LED is. If this was an incandescent light, 100W input would mean nearly 100W of power dissipated in the lamp. I thought that LED's were about 10% efficient, incandescent only 1 or 2%.

Blazing heat from the front of this LED would look like much better efficiency.

[lmester]

If the LED runs super hot, then it is likely your heat sinking is not adequate. A CPU fan is designed to take away ~100W of power, but keep in mind that packaging thermal resistance of a CPU is much lower than a cheap LED module, so it tends to run hotter. To bring temperature down, you need a beefier heat sink or a more powerful fan.

Since I knew that I'd be removing this LED from the heat sink, I used my cheap white heat sink compound. I should try this again with some of my good CPU heat sink compound. Heat sink temp is low. I have no way to determine the thermal resistance between the LED and heat sink.

Also, I toasted my thermometer some more and did more tests. Turning off the power to the LED and immediately taking the temperature shows a large temp drop. After a few more seconds it's down to the heat sink temp. Even if this high temp is just due to the light, I can think of some reliability problems.  The bonding wires are in the light. Huge temp changes from power on to ambient temp. Not good for reliability...

[janoc]

Also, I toasted my thermometer some more and did more tests. Turning off the power to the LED and immediately taking the temperature shows a large temp drop. After a few more seconds it's down to the heat sink temp. Even if this high temp is just due to the light, I can think of some reliability problems.  The bonding wires are in the light. Huge temp changes from power on to ambient temp. Not good for reliability...

The bonding wires are welded, not soldered and the LED is unlikely to produce sufficient heat to cause them to detach or to break due to thermal expansion. Also metals reflect most of the incident light (even if not polished), which is one reason why they are so difficult to cut by laser. So I wouldn't be worried about those too much.

On the other hand, if you are worried about reliability, a soft power on/off driver will help to alleviate the thermal shocks.

[Zero999]

Are you sure the 100W specification isn't a lie? I'm always weary of the specifications of cheap LEDs.

[amyk]

It probably is blindingly bright, if you look at the second picture. Presumably the room is still lit at an ordinary intensity, and you can see lit LED strips under the shelf, but from the picture it looks like the rest of the room is pitch black. The camera will automatically adjust down the exposure so as not to cause the whole picture to become pure white.

[Zero999]

Are you sure the 100W specification isn't a lie? I'm always weary of the specifications of cheap LEDs.

According to his last picture it is consuming 100W.

That doesn't mean it's able to dissipate 100W reliably and safely. It's possible it's a scam. Perhaps it's a 20W LED sold as being able to dissipate 100W?

[stj]

how bright are these - compared to a car headlamp for example??
i'm getting idea's for a killer flashlight - with very short battery life!!! 

[Audioguru]

Of course the LED is a scam. It is cheap. The heatsink compound was also cheap.
Instead of ebay Chinese fakes then buy Philips Luxeon Lumileds instead.

[sleemanj]

That doesn't mean it's able to dissipate 100W reliably and safely. It's possible it's a scam. Perhaps it's a 20W LED sold as being able to dissipate 100W?

The images showsa typical 10s10p arrangement, and the die look relatively large (compared to other typical chinese power leds of this style).

It's undoubtedly 100W (rated as the power consumption, not to the light it throws out).

Yes, they get hot, it's unavoidable, LED technology is more efficient than traditional lighting, but it's still not very efficient, like saying my Lada is faster than walking, that ain't saying much.

Personally without any specs I prefer to see 80C or under measured by themocouple or thermistor attached at the led's tab, in the steady state, taking into account potentially higher ambient temperature than it was measured at of course, if each die has a 20C/watt resistance to the case tab (I think/hope that's conservative and most would be less, although this is where die size comes into play of course), then we are at 100C die/junction temperature which is about as high as you want to go preferably, but I've seen 120C as maximum operating junction temperature before for common power leds so there is probably still a reasonable wiggle room.

http://www.newark.com/pdfs/techarticles/lumex/ImportanceLEDThermalManagement.pdf
https://en.wikipedia.org/wiki/Thermal_management_of_high-power_LEDs

[lmester]

I guess I'm showing my age. It just amazed me that I could melt plastic with LED's. For decades LED's were a wimpy light source that was only useful as an indicator light. Worthless for outdoor use. Really worthless for lighting. My first calculator had an LED display. I tried using it outdoors. I couldn't read the display. Blueskull said about 20% efficient for a quality LED. Really amazing!

Please don't go crazy on the "Cheap Chinese junk" topic.

Are these Chinese junk? Yep! I bought these just to have some fun. They were the absolute cheapest ones that I could find on EBay. I bought four of them for about $10. I was hoping to get at least one that was relatively functional.

I was surprised. All 100 LED's light on all four of them. They also appear to be fairly well balanced. LED's start to light up with only 30 uA of current. All LED's appear equally bright at 150uA. Much better than I've seen in some Youtube video's. These guys were getting ones that had dead strings of LED's at normal operating voltage. The dim picture that I posted before was with one running at 30uA.

How bright are they?

The simple answer is painfully bright! I made a stupid mistake when I first tested one. I set the power supply to 2A constant current. About 2/3 of LED rated capacity. Hooked up the LED and turned on the power. Stupid me! It was face up on the bench. I was looking right at it from about two feet away. Eyeball OUCH! Had a square afterimage in my vision for several minutes.

I decided to find a way to do a little more accurate and less painful brightness test than the eyeball OUCH test.

About a year ago I installed an LED floodlight above my garage door. It's a quality fixture with CREE LED's. Cost me about $150. When I have to get up on a ladder to install something, I don't want a piece of junk that's going to fail in a year or two. I want to be dead before it needs replaced!

I think that the fixture was listed as 65W input power. Can't remember the lumens. I'll try to find the specs for it.

I don't have a light meter. I installed a light meter app on my tablet. I measured the garage light from about 5 feet away. The app showed about 5000lm. At about the same distance, the Chinese module gave about 4000lm. It looks like a quality LED fixture puts out more light. The lumen values from this light meter app would only be useful as a relative brightness comparison. No way to calibrate the light meter app. I can only hope that it's readings are at least fairly linear. I doubt that my tablet's light sensor was designed to be used as a precision light meter.

These are functional but not very efficient. Puts out less light than a quality 65W LED fixture. Still, not bad when you consider that I'm comparing a $2.50 LED with a $150 fixture. No way to know how accurate this comparison really is. I need to borrow a light meter.


Someone had asked if these were really 100W LED's. I gave one a stress test last night. I ran it at 100W for about 3 hours. All LED's are still lit. It's probably over-rated. After all it's cheap and from China. Probably not grossly over-rated. 3 hours at 100W and it didn't go poof! Also I ended up giving it more of a temperature stress test than I'd intended. I'm using cheap white heat sink compound. Not going to waste the good stuff just for testing cheap parts. Even worse, I noticed that I'd placed the heat sink on the bench with the fins pointing vertical. I was blocking half of the airflow out of the heat sink. When I test the next one I'll measure the temp with the heat sink the wrong way.

Attached is a microscope picture showing the LED chips. You can see the bonding wires jumping from one to the next.

[stj]

your 150$ light may have used pwm to reduce the duty cycle, it's how i would drive an led.

[lmester]

your 150$ light may have used pwm to reduce the duty cycle, it's how i would drive an led.

I don't see why you would do this. I'm thinking that if you reduce the duty cycle or average power, my $150 light would look less bright. It's not a cheapo light but the manufacturer would still want to get the most brightness out of the LED for the least cost.  They might derate the LED to extend It's life. Not an extreme amount of derating.  Maybe use a better quality power supply. From what I've read, the power supply is the weak link in an LED fixture.  It's common to have LED's with a 50,000 hour life. Cheap electrolytic caps in a switching power supply will never last that long. I just hope that I didn't get a Chinese cheapie rebranded fixture for $150. Even for equipment bought from a U.S vendor you still need to worry about the quality.  After all, they're competing with China:) And, probably bought it from a Chinese vendor.

I'll be really pissed off if I'm up on the ladder in a few years to replace it!

Also, I'm now testing the next LED at maximum power. Actually slightly above max (101W power). With the heat sink placed correctly it's temp is 88 Deg. Wrong way up (half of the output airflow blocked) it's 96 Deg. Ambient temp is 75 Deg. Temp right in front of the LED is 400 Deg. The same as the first one. I'll let it cook for a few hours and see if all LED's are still lighting up.

[Psi]

The cheap 100W leds from china are genuine 100W leds but they're all the factor seconds.
So your mileage may vary, sometimes they last, sometimes they don't. Best to order a few

I got 10, about half had bright or dead spots so i put those aside for low power use, since they would obviously fail quickly if run at full power.
Of the rest about 3 of them looked perfect. I've been running one of those at 100W as my room lighting for 2 years and its still good. Although it may be a little dimmer than when new.  I've been meaning to swap it around and compare brightness.

[lmester]

412 deg F per 80W just for junction-baseplate isn't normal. That translates to 2.86K/W. Normal value should be less than 0.5K/W.
I suggest you to double check your thermal paste. Make sure to apply only a very thin layer of them, also make sure to press the LED firmly to heat sink from one corner to avoid air bubbles from being trapped in.

I've been installing power transistors onto heat sinks for about 30 years (Mostly TO-3's). A thin coat of compound on the bottom of the transistor. Ease it down onto the heat sink from one end and slowly apply hand pressure. Install the screw onto one end and tighten slightly. Install the other screw. Then tighten them both down. Check the screw tightness later. They tend to loosen up. I'm guessing that the compound is slowly squeezing out of the gap.

This is an LED with four screws. I used the same procedure. Just tighten two screws first and then the other two screws. Wait & then tighten again.

I'm using the standard white compound. Using my expensive arctic silver compound should decrease the thermal resistance. It's $8 for a small syringe. I'll save that for when I decide what to do with these cheap LED's. Quality heat sink compound costs more than I paid for these LED's!

I'm thinking of building an insanely bright flashlight. It'll take several pounds of NiMH batteries to run it for a useful amount of time. Not practical but loads of fun!

[Fungus]

I was surprised. All 100 LED's light on all four of them. They also appear to be fairly well balanced. LED's start to light up with only 30 uA of current. All LED's appear equally bright at 150uA. Much better than I've seen in some Youtube video's. These guys were getting ones that had dead strings of LED's at normal operating voltage. The dim picture that I posted before was with one running at 30uA.

Yes, you can get good ones on eBay but it's a lottery. You just never know what you'll receive.

How bright are they?

100 LEDs crammed into 1cm²? Very bright!

If they're white then they're about 3.6V. If there's 100 LEDs at 20mA each then that gives approx 72W.

I think your "100W" figure is a stretch.

Me? I'd run them at 50W or less if I was interested in reliability. They'll still be very bright.

Remember: Brightness is a curve. Above the LED's power rating you generate much more heat than light (as you found out). Running them at a bit less power might not reduce the light very much and they'll have a chance at surviving.

The front of the thermometer was melting! Is this normal?

 

Yes, they radiate quite a lot. I once dropped a piece of paper on one and it instantly caught fire. I have to be really careful because I own a cat who instantly sits on any bright lights in the room.

PS: Have you ever been on stage with one of those super troupers. It's like a heat ray when they hit you with it.

[sleemanj]

If they're white then they're about 3.6V. If there's 100 LEDs at 20mA each then that gives approx 72W.

20mA?  Either you missed a zero, or you really are in the ... dark ages :-)

These are power leds, up to 300mA through each die usually, but of course less = cooler and cooler = longer life.

[Fungus]

If they're white then they're about 3.6V. If there's 100 LEDs at 20mA each then that gives approx 72W.

20mA?  Either you missed a zero, or you really are in the ... dark ages :-)

These are power leds, up to 300mA through each die usually, but of course less = cooler and cooler = longer life.

Ummmm, the LEDs are arranged in parallel on the die.

If the LEDs are 20mA/3.6V then you can arrange them in different ways, eg.:

10 strings of 10 LEDs would need 200mA through the die at 36V.

20 strings of 5 would need 400mA /18V.

etc.

[Zero999]

Someone had asked if these were really 100W LED's. I gave one a stress test last night. I ran it at 100W for about 3 hours. All LED's are still lit. It's probably over-rated. After all it's cheap and from China. Probably not grossly over-rated. 3 hours at 100W and it didn't go poof! Also I ended up giving it more of a temperature stress test than I'd intended. I'm using cheap white heat sink compound. Not going to waste the good stuff just for testing cheap parts. Even worse, I noticed that I'd placed the heat sink on the bench with the fins pointing vertical. I was blocking half of the airflow out of the heat sink. When I test the next one I'll measure the temp with the heat sink the wrong way.

Attached is a microscope picture showing the LED chips. You can see the bonding wires jumping from one to the next.

How big are the LEDs, compared to the one used in your high quality Cree fixture?

The size should give you a rough estimation of how much power they should be able to dissipate, but bear in mind that they might not have such good heat transfer as a high quality LED.

[janoc]

Yes, they radiate quite a lot. I once dropped a piece of paper on one and it instantly caught fire. I have to be really careful because I own a cat who instantly sits on any bright lights in the room.

Well, I am pretty sure that the cat would attempt to sit on a LED only once  Cats are not dumb, they remember where they got their whiskers burned last time.

[sleemanj]

Ummmm, the LEDs are arranged in parallel on the die.

I suspect it is nomenclature that is tripping you up.  When dealing with power leds it is common to refer to the parts making up the LED on a macroscopic level as below.



(Note that I've rotated the actual direction of the strings by 90 degrees if you look closely you can actually see the bond wires in the LED form the series strings vertically)

As I hope you can see on this type of LED there are 10 strings of 10 DIE in series on this common type of power led. As I do not have to explain, devices in series pass the same current through them all, and in parallel the current divides, therefore each die sees 10% of the total current through the LED and 100% of the current through it's series string.

Each die has a forward voltage of around 3.3v, and has a typical forward current of around 300mA (not 20mA like you have said). 

Each die therefore has a consumption of 3.3v * 0.3A = 0.99W
Each string therefore has a consumption of 33v * 0.3A = 9.9W
And the led therefore has a consumption of 33v * 3A = 99W

[lmester]

Me? I'd run them at 50W or less if I was interested in reliability. They'll still be very bright.

Remember: Brightness is a curve. Above the LED's power rating you generate much more heat than light (as you found out). Running them at a bit less power might not reduce the light very much and they'll have a chance at surviving.

Sounds like a good idea. Especially if I'm running it with battery power.

I found a data sheet for a 100W LED. Probably not for this unit but it should be similar. You can see on the graph that 3A gives only about double the light produced at 1A. I'll have to see what batteries I have in my junk box.  Getting a useful life from available batteries will probably be what sets the operating current.

[CJay]

Yes, they radiate quite a lot. I once dropped a piece of paper on one and it instantly caught fire. I have to be really careful because I own a cat who instantly sits on any bright lights in the room.

The cat would have to be rather dumb to sit on that one twice. Admittedly having a flaming cat running around the house might not be ideal but...

[lmester]

How big are the LEDs, compared to the one used in your high quality Cree fixture?

The size should give you a rough estimation of how much power they should be able to dissipate, but bear in mind that they might not have such good heat transfer as a high quality LED.

The LED fixture has two LED modules. Each is mounted in it's own reflector. Each looks like one big yellow square. Either one big LED or I can't see the individual LED's through the coating. Probably just can't see them.