PCB design: A close look at facts and myths about thermal vias

Article By : Douglas Brooks and Johannes Adam

Thermal vias are useful in controlling the temperature of a hot component, but it’s important to recognize they must terminate somewhere.

It is fairly common to have a hotter than desired component on a PCB. Often, the way to control the heat from such a component is to (a) create a copper pad underneath it as solid as possible, and then (b) place vias between the pad to a heat conducting surface somewhere below the pad. Such vias are called “thermal vias.” The idea is that the thermal vias will conduct heat away from the pad, thereby helping to control the temperature of the hot component.

Various sources suggest—without much theoretical or experimental verification—that the optimum size for such a via is 0.3 mm in diameter and that it should be copper-filled. And since each via offers a marginally smaller improvement in temperature, the practical limit on the number of vias is around 50 to 100.

Taking a closer look

In most articles related to thermal vias, the authors fail to recognize a very important point. Thermal vias have to go from the pad to “somewhere.” And that “somewhere” is typically a copper plane located in the stack-up below the heated pad. We have shown in a previous article that the presence of an underlying plane significantly lowers the temperature of a trace. Similarly, an underlying plane by itself will lower the temperature of a heated pad. So, what’s important to recognize is which factors more strongly affect the temperature of the pad: the thermal via(s) or the underlying plane.

We examine these factors using a thermal simulation tool called TRM. We describe such thermal modeling extensively in our recent book, PCB Design Guide to Via and Trace Currents and Temperatures. We start with a typical, 1600-μm thick FR4 board sized 100 x 100 mm2. We simulate a heated component with a 25 x 25 mm2 pad. A unique capability of a TRM model is that we can apply a certain number of watts to a copper pad to heat it instead of applying current across the pad. This avoids having to calculate various current flows throughout the pads, vias, and planes. In our case, we will apply 2.5 watts to the pad, heating the bare pad to 95.7oC—75.7oC above the ambient temperature. Figure 1 illustrates the thermal distribution across the top layer of the board under these conditions.

Figure 1 Thermal distribution is shown on the heated pad without any underlying planes.

Note that the pad temperature is highest at the center. It is also higher along the edges. This is because the corners cool more effectively than do the sides of the pad, and the sides cool more effectively than does the center. We cover the reason in Chapter 14 of our book.

Unfortunately, there are an almost unlimited number of ways we can bring thermal vias into a design. Designs vary in dimensions, materials, number and sizes of thermal vias and heat generation, to name just a few. So, there is no “typical” design that we can simulate. Therefore, we offer the following discussion from which we draw a few conclusions.

But first, we want to emphasize two points:

  1. Thermal via designs almost always (almost by definition) have the via terminating on a copper “plane” of some size.
  2. That plane offers more cooling potential than does a number of thermal vias.

We will look at two different plane configurations. One will be a “plane” (“small”) that is the same size as the pad. The other will be a plane (“large”, think in terms of a power supply plane) that covers the entire area of the board at some layer in the board. The planes will be placed at two depths in the board. One will be 300 μm—“near”, approximately 12 mils—below the pad. The other will be on the “far” side of the board, almost 1.6 mm—approximately 63 mils—below the pad.

These four simulations will cool “naturally,” meaning heat will flow through them to the board material and to the ambient air. In one additional pair of simulations, these planes will become “heat sinks.” That is, their temperatures shall be held constant at 20oC.

Each thermal via will be 0.3 mm—approximately 12 mils—in diameter. We will assume the thermal vias are filled with plated copper, which is virtually pure copper. This assumes we will have the best possible thermal conductivity through the via. If the via walls were only plated to, say, a 1.5 mil thickness, their thermal conductivity would be dramatically less.

It’s interesting to compare the heat conducting qualities of the thermal via to the board material. The formula for thermal conduction—ignoring convection, radiation, and heat spreading—is:

Q/t = KA (ΔT)/d           (1)


Q/t = Rate of heat transfer (watts, or joule/sec)

K = Thermal conductivity coefficient (W/mK)

About 0.6 for our FR4 model

About 385 for copper

ΔT = Change in temperature (oC = oK)

A = Overlapping area

About 625 mm2 for the pad

πr2 = (3.14) * (0.152) = 0.0707 mm2 for each thermal via

d = Distance between pad and plane

300 μm for the “near” plane

1.6 mm for the “far” plane

The heat transfer rates for the pad and for the thermal vias are different. We can compare their magnitudes by forming the ratio (the ΔT and the d’s cancel):

(Q/t)p/(Q/t)tv = (kA)p/(kA)tv = (0.6)(625)/(385)(0.0707) = 13.8         (2)

That is, the thermal conductivity through the board material in this specific design is almost 14 times greater than that through the thermal vias. But the situation is much more one-sided than that. It is extremely important to note that the mere presence of an underlying plane lowers the temperature of the pad. Therefore, the subsequent thermal conductivity of the thermal via is further reduced because the ΔT term has been reduced by the presence of the plane.

Simulation results

Figure 2 graphically shows the simulation results. The 2.5-watt source heats the bare pad by itself to 95.7oC. That is a 75.7oC difference above the 20oC ambient. The graph plots the maximum temperature on the pad for each combination of plane and number of thermal vias. Different combinations of planes and thermal vias appear to have some effect on the plate’s maximum temperature. But some factors are much more important than others.

Figure 2 Maximum temperature on the pad vs. the number of thermal vias is shown for different sub-plane configurations.

Impact of planes dominates

The presence of an underlying plane significantly lowers the temperature of the pad. Intuitively, it’s not too difficult to understand. But it also raises the temperature of the underlying plane. For example, Figure 3 illustrates the thermal distribution on the bottom layer of the case of a “small” plane on the far side of the board. The temperature on the plane is in the 80-degree range, but drops off quickly toward the ambient temperature as you move off the plane.

Figure 3 Thermal distribution of bottom layer is shown for the case of a “small plane, far.”

The stable temperature of the pad/plane combination depends on the relative size of the plane. Since the heat source is at the pad, the temperature of the pad raises the temperature of a small plane. A larger plane, with more cooling capability, tends to lower the temperature of the pad. In either case, the temperature difference between the pad and the plane is relatively small, something less than 10oC in our models.

The stabilized (absolute) temperature will be somewhere between the bare pad temperature and the ambient temperature—in our case fairly near the middle range of temperature. However, in the extreme case of the heatsink, the pad temperature is lowered almost all the way down to the heatsink temperature.

As a result, the difference in temperature between the pad and any plane (ΔT in Equation 1) in every simulation is significantly reduced. Table 1 illustrates what is happening in our simulations. Since the presence of the planes causes the ΔT to fall so sharply, the thermal conductivity through the thermal via (Equation 1) is reduced to the point that subsequent thermal vias can have little or no effect.

The presence of a plane lowers the pad temperature but raises the plane temperature

Table 1 The presence of a plane lowers the pad temperature but raises the plane temperature.

The above statements apply to all of our simulations. But it’s important to note that a small plane added to a board to help cool a heated pad is counterproductive. Rather than lowering the temperature of the pad, the effect is to raise the temperature of the plane, thus negating the primary objective in the first place.

Thermal vias offer only “point” solutions

What little additional benefit thermal vias may offer is limited to a very narrow area around the via itself. Figure 4 shows the thermal distribution for the pad in Figure 1 with the addition of a “large plane, far.” The case of no thermal vias is shown on the left and 25 thermal vias shown on the right. Note the expanded thermal scale for these images. The characteristic cooling at the edges and corners of the pad is nearly identical to that for the pad without a plane. The maximum temperature for this pad under these conditions without any thermal vias (from Table 1) is around 58oC. The vias provide only a couple of degrees difference in temperature from the case without vias, and that difference is localized very near to the vias themselves.

Figure 4 Thermal vias only make a minor difference in thermal distribution on the pad. This simulation is for “large plane, far.”

Thermal vias, almost by definition, require a copper surface to terminate on. The presence of this copper surface impacts the thermal distribution through the board. The net impact is that the difference in temperature (ΔT) between the heated pad and the copper surface reduces dramatically. This reduces the thermal conductivity through any thermal via to the point that thermal vias add very little additional benefit.

Furthermore, if the additional copper surface area is small, then the stabilized temperature of the pad may only be marginally lower than without the pad. The best improvement comes from a relatively large copper area—think power/reference plane—placed relatively near the far side of the board.

Finally, any thermal vias that are added to the board tend to only have a “point” impact. That is, they tend to only impact the pad at the point where they are placed. That is the main reason why so many previous authors think that a large number of thermal vias is typically required.

Douglas Brooks has written two books and numerous technical articles on PCB design. He gives seminars on PCB designs around the world.

Johannes Adam has worked on numerical simulations of electronics cooling at companies like Cisi Ingenierie, Flomerics and Mentor Graphics. He currently works as a technical consultant.

Other articles in this series:

Related Content


Leave a comment