Rules of thumb are wonderful. They tend to get passed down from people who have actually done the math and generally let everyone else do their jobs better and faster. Still, when you have used a particular rule of thumb so regularly that it becomes habit, you start to wonder why exactly this is regarded as the best solution. It becomes time to do the math yourself.
Or, in my case, it becomes time to find the fire extinguisher and go make a lot of graphs.
The Rule of Thumb
Most of my work involves relatively high currents on small boards. LED power supplies tend to need to provide a couple amps for an effective design and the LEDs themselves must be able to tolerate that current as well. Sizing the copper traces and such are certainly tactics to keep your system from suffering, but the most important part of my designs is always thermal management. Whether it’s a switching MOSFET in the driver or a lighting module capable of running at an amp or more, I have to be able to get the generated heat out of the component package and safely away from the source and any neighboring components. The rules of thumb I use are to make the copper pour under the heat-generating component as large as possible and to add a handful of thermal vias under the largest pad on the package. Most LED projects would benefit from a proper metal heat sink, but budget and space considerations often make that difficult in practice.
The Test
I created PCBs that represent the four ways I’ve seen thermals managed for a lighting LED. An LED was easier than other components because it could operate off a current limiting bench top power supply without any additional circuitry.
Ähnliches Produkt:
The boards are quite small, about 1cm by 2cm, and each have identical power/ground pads with an XLAMP lighting LED from Cree centered between them.
I also attached a thermal test point pad to the center pad of each LED for a uniform spot to attach a thermocouple. The boards are named No Relief (no thermal relief on the LED at all with 6mil traces), Copper Relief (wide traces and a full-board copper pour connected to the thermal pad), Copper Relief +10 (same as copper relief but with 10 thermal vias under the pad), and Extreme Thermal Vias (copper relief with as many thermal vias as I could fit onto the board).
The test used two thermocouples for each board: one attached to the thermal test point (T2) and one under the board behind the LED (T1). The board was given 0.5A from a current limited supply at t=0. The current was increased to 1A at 10 minutes (these LEDs can handle 1.5A) and power was turned off at 20 minutes. Ten minutes is not technically long enough to let the temperature of the system stabilize at a particular current, but it is long enough to compare the thermal strategies across the boards.
The Results (Boards Only)
The maximum junction temperature allowed by this LED’s datasheet is 150oC. The junction is inside the package, so the maximum test point temperature should be lower by at least a few degrees. As you can see, the board without any relief fails this test pretty hard. In fact, heating up to 318oC caused the silicone dome to burn fairly badly and the coating on the PCB to start smoking. It was dramatic and a little frightening, but most importantly it would drastically reduce the usability lifetime of the LEDs if this were ever actually done in practice. Adding even a small copper pour brings the maximum temperature down significantly, but thermal vias seem to be the most helpful in controlling the temperature. I was honestly surprised that adding more than 10 vias did not significantly improve thermal performance and came in at the same peak temperature as only a few vias under the pad.
The Results (Boards with Heat Sink)
After testing all the boards while they sat on the table, I re-tested each on a 35mm square heat sink with a built-in adhesive.
This represents the optimal passive thermal solution. Blowing air over a system helps tremendously, but spinning a fan requires power and is considered active rather than passive cooling.
Ah, the beauty of heat sinks. Simply adding any sort of relief to the board without any copper relief brings the max temperature down over 100oC. It still dramatically exceeds the maximum junction temperature specified for the LED and is a poor design, but proves that even the worst board designs can be mitigated if necessary. The other boards performed better as well; even the board without any thermal vias maxed out well under 100oC and are considered safe for the chosen LED. I still wouldn’t hold my hand on it, but we will not be starting any fires at these temperatures.
While the maximum junction temperature is 150oC, the datasheet for these LEDs shows that luminosity begins to decrease below 100% when the temperature exceeds 75oC. This may not be a concern for every design, but most LED projects are centered on efficiency. The only designs that cleanly fit below this line are the designs with thermal vias on heat sinks.
Conclusion
I think we can all agree that running an LED or any power device without thermal relief is going to go poorly. All the well-designed boards performed better on their own than the relief-less board with a heat sink, which was not unexpected. With or without a heat sink, the design with vias directly under the largest pad on the package performs almost equally as well as the design with all the vias that we could fit on the board. Creating vias is not free, so a few selectively placed vias will be more cost effective than the monstrosity I created as the final board. We have verified beyond doubt that the rule of thumb holds, and that to maximize efficiency in both design and heat, a handful of thermal vias through a board onto a heat sink will always be our best bet.