Minimizing EMI from Heat Sinks

N.J. Ryan, D.A. Stone, and B. Chambers

Research on heat sinks points to several design rules that can help engineers reduce the EMI resonated by these structures.

Heat sinks play a vital role in enhancing circuit design reliability by keeping sensitive components, such as power-switching semiconductors, within a particular operating-temperature range. However, the heat sinks can also add to overall circuit EMI by resonating the EMI that the temperature-sensitive components radiate. Power switching actually generates much of the noise in a circuit, and this noise may be subsequently coupled directly to the heat sink. Devices that generate high-order harmonics of the switching currents are often mounted onto large metallic objects with potential multimode resonant characteristics. Measurements of one circuit have shown an increase of 10 dB in the level of radiated EMI solely because of the addition of a heat sink to the circuit.

Because heat sinks have such potential to add to overall EMI, engineers need to be sure to take into account the possible EMI resonance of heat sinks in their designs. It would be inefficient, however, to take radiated EMI measurements for every possible circuit configuration that uses a heat sink. Instead, the authors of this article conducted in-depth studies to devise general rules that engineers can use to minimize EMI when working with heat sinks in their designs. Using these rules, engineers can reduce the radiated EMI levels from circuits and produce products with higher integrity and a greater margin of compliance.

Modeling the Problem

To study heat sink EMI, the authors used electromagnetic modeling techniques to predict the effect of different heat sink configurations on radiated emissions without having to physically construct and measure circuits. The modeling technique used was the finite difference time domain (FDTD) method.

One of several standard methods of electromagnetic modeling, FDTD splits up the computational space into a three-dimensional grid. Field values at each position are calculated in an iterative fashion. A new field value is calculated from the previous value and the values of the surrounding field components. The material parameters of each element in the mesh can be individually defined to simulate the structure to be modeled. Once the heat sink is defined within the mesh, the model can then be used to calculate the radiation enhancement from that structure within the mesh.

The model is initially developed without the heat sink in the mesh; and then created a second time with the heat sink structure added to the mesh. The two sets of results are then compared to yield the increase in the radiated EMI caused by the heat sink.

The heat sinks in the model were excited and the results measured. In a conventional power circuit, excitation of the heat sink may occur in a number of ways. In the most common scenario, the scenario used for these studies, the device and the heat sink form the two plates of a parallel-plate capacitor. The insulating washer acts as the dielectric of the capacitor (see Figure 1). The resulting E-field generated in the insulating washer is perpendicular to the base of the heat sink and is represented by a simple source placed within the FDTD mesh.

Figure 1. Generation of an E-field excitation.

What the Models Reveal

The authors have carried out an extensive study, resulting in the derivation of a number of design rules that can significantly reduce the radiated emissions from power electronics systems. A number of simulations are presented here to illustrate these design rules.

In one model, a heat sink that has a square base with sides of 300 mm each is used (see Figure 2). The sources are placed at the points marked A to D, and the radiation enhancement at the various points is measured.

Figure 2. Model and EMI results of a system using a square-base heat sink.

Examination of the radiation enhancement illustrates several phenomena. First, the level of enhancement is dependent upon frequency, an expected result because the heat sink is acting as an antenna. The frequency dependence of an antenna is linked to the physical dimensions of that antenna; therefore, the first peak, seen at approximately 320 MHz, can be linked to the 300-mm side length. Like an antenna, the heat sink has a radiation pattern that is directional and linked to the source position. When the source is placed at A, the structure is excited in a direction along an axis through A and B, while a source placed at D excites the structure in a direction along an axis passing through D and B.

Second, the results of the model show that EMI enhancement is least when the source is placed at point B, away from the edges.

Third, comparison of source points A and D in the model shows that fin direction is important. With the source at D, the fins are running in the same direction as the excitation, or longitudinally. A source at A produces an excitation transverse to fin direction. Comparison of the results demonstrates that at the peak of enhancement, a transverse fin direction will produce a greater enhancement than a longitudinal fin direction. If the fin size is increased, then this effect is also increased.

Finally, the model shows that when the source is mounted at the corner, C, excitation occurs in both directions, producing the greatest excitation.

Another model in the author's study is a smaller heat sink with a rectangular base (see Figure 3). This model demonstrates the effect of the length-to-width ratio. If a structure has one dimension that is significantly longer than the others, that dimension will be more susceptible to excitation. When the source is positioned at A, the enhancement is less than when the source is placed at D, even though fin orientation would suggest that the source at A would produce greater enhancement. Also, the rectangular model reinforces the finding that a source placed at the corner produces the greatest enhancement.

Figure 3. Model and EMI results of a system using a rectangular-base heat sink.

The authors also model a heat sink with central channels in which to mount components, a configuration that often offers the best thermal performance (see Figure 4). This design also improves the EMI performance of the heat sink. When the source is mounted at the end of the channel, the peak enhancement is reduced to 11 dB. When the source is mounted in the center of the heat sink, there is no enhancement; the structure acts as a partial shield and reduces the level of radiated emissions. At higher frequencies, this channel could form a resonant cavity; however, these frequencies are usually above the frequencies covered by current standards.

Figure 4. Model and EMI results of a system using a heat sink that has central channels in which to mount components.

Design Rules

The modeling of heat sinks provided several general design rules for minimizing EMI.