Thermal vias are merely one tool, albeit sometimes an important tool, in the thermal engineer’s arsenal of techniques for getting heat away from potentially overheating electronics components. The problem is, vias don’t actually “cool” a hot device in and of themselves. As their name implies (“vias”), what they really are is nothing more than a short pathway for unwanted heat to travel in order to get to where it can actually be dissipated.
In other words, a thermal via doesn’t do diddly-squat unless there’s something at the receiving end of the via that can actually do something with the heat. For instance, if you’ve got a great thermal spreading plane in the interior of a PCB, or a big heatsink mounted on the back side of the PCB opposite the hot component you’re trying to cool, then a thermal via gets the heat to that dissipater with a little less temperature delta than if the via wasn’t there. However if you’ve got that great thermal dissipater there in the first place, it’s already going to have made a far more significant contribution to lowering your device’s theta-JA than without it. Lastly, to reiterate the original point, if the heatsink wasn’t there, the thermal vias are just thermal dead-ends. Ten years ago my rule of thumb was that if you had a good heat spreader available, then adding thermal vias might drop your overall theta-JA by about 10 percent.
What’s changed since ten years ago? Actually, not a whole lot, except that packages continue to shrink, and customers continue to try to pack even more heat per square inch of PCB. But as a package shrinks, its size relative to the thickness of the PCB gets smaller, which means that the relative through-plane PCB contribution to the overall theta-JA of a device gets to be a larger fraction of the total. A few well-designed thermal vias can therefore have a much larger impact on that PCB contribution to the overall thermal resistance. The flip-side of the package-shrink issue, however, is that board traces become relatively more important, thermally, especially as Rds-on goes down and the nominal current limits for tiny packages go up. (In particular, if the traces are appropriately sized for the higher current, the traces themselves may provide as much thermal path as a couple of thermal vias.) As for my rule of thumb, today I might generously bump that 10 percent via-contribution figure all the way up to 20 percent. Hopefully you understand what I’m trying to say here: You still have to provide a thermal sink to connect to the output end of the via. If you don’t, the via still doesn’t do anything. I’d say the biggest error in thermal design continues to be underestimating how much that “external” heatsink is needed. Vias are not a substitute.
Vias also come at a price. Unless they’re filled, they can actually cause problems with solder-wicking, air and moisture entrapment, and associated lower long-term reliability. Further, if you’re really after the maximum thermal effect, then obviously the vias must be filled with a thermally conductive material (read “copper,” and that’s not a cheap process), because even after plating, the area of the remaining hole is likely to be most of the original drilled hole area. Ironically, I’ve even seen examples of where a board designer used “spider” traces around his thermal vias that punched serious holes in the thermal path, effectively isolating the thermal vias from the spreader plane, making them next to useless. The moral here is, thermal vias can do more harm than good if not used wisely.
So, be sure to connect your thermal vias to a heatsink, and stay cool!
– Roger Stout
