Interested in getting the real information on Heat and LEDs? Check out this informative blog post [originally posted on esthersantos.com] from one of our engineers, regarding managing LEDs’ temperature in a light fixture and why what we do at Noribachi is one of the best solutions in the industry.
A lot of times when we have visitors to our factory floor, I get the question, “We know that LEDs have a heat issue, how do you handle the heat issue”?
(I find it funny that an HID bulb running at 750º F does not have a heat issue.)
High-powered LEDs do not have a heat issue, they have a HEAT TRANSFER issue. Power goes in and light and heat come out. The small size of the LEDs requires that the heat generated at the LED “junction” (where it attaches to the circuit) be removed (dissipated) continuously and efficiently. Otherwise, the LEDs will degrade and not achieve their expected long life.
As everyone knows, heat transfer is accomplished one of three ways: radiation, convection or conduction. You can use any one of these to dissipate the heat. The best method, meaning the most efficient method of heat transfer in the case of high-powered LEDs, is conduction.¹
Conduction heat transfer is the basic process of movement of energy of some particles to their neighboring particles. It is all about proximity and it is therefore very efficient in a solid-to-solid interface. All high-powered LEDs have three solder pads on the back of them, one for anode, one for cathode and one for thermal transfer. The thermal pad normally is the largest of the three and sits right underneath the actual diode. The task of the thermal engineer in designing a good LED light is to create a high-energy/heat-transfer flow between that thermal pad and the ambient atmosphere. By far, the most important part of this task is what to do right underneath the thermal pad of the LED. So, a large solid-to-solid interface will conduct the heat the best.
The right thing to do to keep the LEDs thermally happy is to use a thick, highly conductive metal clad PC board with a heavy copper interface. Then, the heat from the LED will conduct through the thermal pad, through the solder, to the copper interface to the metal clad on the PC board. As you can imagine, this is expensive; more expensive than a normal FR4 board. It gets even more expensive since you will need a much bigger oven to do your SMT work. Many places do not even offer the service.
But the better thing you can do is to reduce drive current. We been experimenting continuously for over 4 years with our light boards running at different drive currents. Here is what we found: if you drive the LEDs under 525mA, your need for complicated heat sinks and complicated heat management is minimal.
Let’s walk through an example that will make things more clear. Let’s use two examples where in both of them, the LED manufacturer and the power supply used are exactly the same. In scenario one, we have a fixture with 49 LEDs in strings of 7 on a 4 inch diameter FR4 board running at 1,300mA. The FR4 board is attached to a massive heat pipe design heat sink (8″ X 8″ X 10″). This light uses 240W of power and will nominally generate 18,865 lumens of light. In the second scenario, we have a fixture using 140 LEDs in strings of 4 on a 11″ X 13″ aluminum backed PC board with 3 oz of copper sitting on a 19 1/2″ diameter sheet of 1/8″ aluminum running at 500mA. This light also uses 240W of power and will nominally generate 26,460 lumens of light.
So, what is happening here? The difference in nominal lumens comes from the fact that at 1,300mA, the LEDs use 4 times as much power – with its attendant heat issues – from a baseline of 350mA and generate about 2.6 times more lumens. At 525mA, the LEDs use 1.5 times as much power from baseline and generate 1.45 times more lumens. In other words, much more efficient. Going after lumens by cranking up the drive current is the major cause of heat transfer problems in LED lights.
The above scenarios are real (come visit and we will gladly show you). If you assume that LEDs cost about $1 each, the second scenario nominally costs $91 more to produce. The first scenario is more likely to suffer a heat-related failure than the second. In the first scenario, the power for a failed string will – after failure – get transferred to the other strings where the current will go above 1,500mA, which will result in the all the LEDs failing. The light is dead. In the second scenario, if one of the LEDs fail, the current to the other 19 LED strings increases from 500mA to 526mA. Not catastrophic.
At Noribachi, we have decided not to drive our LEDs beyond 525mA for any of our applications. Over the last 4 years with millions of LEDs used in the tens of thousands of lights that we have sold, we have had 6 of our LEDs fail.
We do a 24 hour burn-in for all lights. If it’s a new design, we do thermal testing of the light in that 24 hour period as well. We do not ship any product that has a junction temperature of over 85º C.²
I leave it up to the reader to decide which of these scenarios is better. At Noribachi we design, build and manufacture based on the second scenario. I find it also funny that in the first scenario, the manufacturer spec sheet of the light above has 3 pages about the heat sink and only 1 line of 1 page mentions the 1,300mA and no mention of what specific LED they use.
¹ I call high powered LEDs CREE XTE, LG 3535, Osram Square, and similar.
² This is a full 40 deg C below the allowable junction temperature for most high-powered LEDs.