Numerical models predict EMC performance
Preparing for electromagnetic compatibility (EMC) emissions and susceptibility tests can raise your blood pressure. Will your product pass? Following good design practices and relying on intuition for printed-circuit-board (PCB) routing, grounding, and shielding should ease some of the stress, but you can still face uncertainty as you head to the EMC lab.
To alleviate your anxiety, you can take advantage of software that models integrated circuits (ICs), PCB routing, component placement, module design, and system design. These modeling tools provide insight into how a design will perform before you commit to building a prototype.
Their simulation features let you evaluate design tradeoffs and how they can affect EMC performance. Just as EMC design is often a tradeoff between compliance, cost, and time, modeling is a tradeoff in accuracy versus complexity and simulation time.
Many circuit designers, system engineers, and EMC engineers use modeling software to predict EMC performance. In fact, EMC engineers often model a system based on assumptions or design specifications before hardware engineers develop their circuits and before PCB designers lay out their boards. In some companies, mechanical engineers use EMC modeling tools to understand how, for example, a design that optimizes cooling can adversely affect EMC.
Figure 1. PCB modeling software can display predicted emissions using simulated oscilloscopes and spectrum analyzers. Courtesy of Mentor Graphics.
The engineers may also work in parallel, with EMC engineers modeling a system while designers are developing circuits and PCB layouts. A system designer or EMC engineer, knowing the approximate locations of sources of electromagnetic interference (EMI) and knowing the enclosure sizes and openings, can begin to develop a system model while the PCB designer routes traces. EMC engineers often start with incomplete information and add details to their models—such as clock frequencies and rise and fall times—based on input from circuit designers as designs progress.
Engineers can also reduce EMC problems by using PCB design tools that analyze circuit designs for emissions and susceptibility. HyperLynx from Mentor Graphics, for example, produces simulated oscilloscope and spectrum analyzer displays of currents in a PCB (Figure 1). The software analyzes currents in PCB interconnects and can provide emissions estimates.
Check design rules
Rule-checking software can apply design rules to a board and is one of the most useful tools available for EMC design, said Bruce Archambault, distinguished engineer at IBM. For example, a rule checker can tell you if clock signals are placed too close to the edge of a PCB or if an IC needs a decoupling capacitor.
There are several numerical methods in use today that solve Maxwell’s equations and predict EMC performance. They include finite-element modeling (FEM) analysis, transmission-line method (TLM), finite-difference time-domain (FDTD) method, and others.
Commercial software programs generally take advantage of just one of the modeling methods. For example, IceWave software from Ansys uses FDTD, Comsol’s Multiphysics software uses FEM, and FLO/EMC from Flomerics uses TLM. A popular paper by Dr. Todd Hubing explains these methods and the tradeoffs among them (Ref. 1).
|See links to more resources on numerical EMC modeling.|
Numerical-modeling software can model the E- and H-fields surrounding a PCB in 2-D or 3-D space by solving Maxwell’s equations. To solve Maxwell’s equations, you must define the geometry through which fields propagate. The software applies a time-domain Gaussian pulse to the model and plots the response. A pulse introduces a wide range of frequencies to your model. You can convert the response to the frequency domain and limit your analysis to frequencies of interest.
Figure 2. System models use color to show emissions "hot spots." Courtesy of Flomerics.
At the system level, EMC engineers use modeling tools in which the models are based on assumptions and preliminary specifications. Initial models may contain the locations of displays, cables, and emissions sources such as power supplies and PCBs, but not much else. As designs progress, engineers will add IC locations, heat sinks, shields, and more accurate geometries from CAD software packages. EMC engineers may also get data on the materials used in a design that modeling software can use in its calculations.
Figure 2 shows a simulation of radiated emission in a computer. The image uses color to indicate field strength. Typically, red indicates the highest field-strength level. For each plot, you must specify the frequency range of interest if you suspect fields at certain frequencies such as clock frequencies or their harmonics.
EMC modeling isn’t just for boards, modules, and systems, though; you can also use it at the device level. Figure 3 shows an EMC model of bond wires inside an IC. Intentional or unintentional current in one wire can couple into adjacent wires through mutual inductance, which can cause a device to malfunction.
Figure 3. Software models of IC bond wires require information about wire geometry and materials. Courtesy of Comsol.
Even this relatively simple model highlights the complexities in accurately modeling a device or system. A model must describe the voltage between the ends of a wire (ΔV). If ΔV is greater than 0 V, then current will flow in the bond wire.
An admittance model of four wires requires that you know the properties of the materials involved. The model needs each wire’s conductivity (δ), permeability (µ), and permittivity (ε). You then need to assign voltages to the ends of the two adjacent wires (V1 through V4) and calculate the current that one wire induces in the other using the matrix shown below, in which the admittance values for y11, y22, y33, and y44 represent the effects that each wire has on itself. All y values are based on the values of δ, µ, and ε that device manufacturers provide. (Modeling software often contains a library of materials properties such as wires, enclosures, and insulators.)
Given that a simple model can contain considerable information, you can see that modeling in such detail is impractical at the system level. Thus, EMC models often lump areas together under the assumption that the characteristics for a given area or volume are constant. These areas form a “mesh.” Figure 4 shows a mesh model for a PCB where the software creates an equivalent circuit model for each segment.
Figure 4. EMC models of a PCB divide the board into segments that form a mesh where each segment is modeled with an equivalent circuit. Courtesy of Ansys.
Clearly, the finer the mesh, the more accurate the system model. But that resolution comes at a price: simulation time. “A complex model may take several days to run,” said Chetan Desai, IceWave product manager at Ansys. “The tradeoff is detail versus time.”
Making the tradeoff decision requires experience and intuition. “You can’t simulate the entire world,” added Bjorn Sjodin, VP of applications at Comsol. “You must define the boundaries of what you can simulate.”
When developing a model, concentrate on the areas that are most likely to cause EMC problems. At the board level, that’s IC placement and trace routing. At the system level, concentrate on the location of emissions sources (ICs and oscillators) and the location of heat sinks. You also should focus on openings in an enclosure such as those for displays, controls, cables, and seams. EMC engineers can often make modeling recommendations based on experience.
Tradeoffs also occur between EMC and other parameters, particularly heat. Case openings let heat escape, which makes for cooler-running components, but they also let emissions out and outside interference in. Thus, engineers often use software that simulates a system’s thermal properties in conjunction with EMC simulations. “You’re often faced with competing design issues,” warned David Johns, VP of EM engineering at Flomerics.
Numerical EMC modeling can reduce the stress you face when testing a physical product, but it won’t guarantee success in the test lab. “There’s no replacement for the insight of an experienced EMC engineer,” said Archambault.
|1. Hubing, Todd H., “Survey of Numerical Electromagnetic Modeling Techniques,” University of Missouri-Rolla, 1991. www.emclab.umr.edu/pdf/TR91-1-001.pdf.|