Electromagnetics Best Practices

Best practices for setting the values for the following electromagnetics physics include:
  • Minimum size for Global fidelity should be the size of the metallic features that you would like to resolve throughout the simulation region. For example, for traces that are 1mm in width or a helical antenna made from a wire with a diameter of 1mm, set this field to 1mm at the largest, and most safely to about 0.8mm (80% width or diameter). The global feature size should be used if all the metal geometry in the simulation has a similar feature size.
  • Local Fidelity can be defined for metal geometry with small features requiring resolution in certain regions. You can use the same rules you set for global Minimum size but you only need to apply them to the required geometry. Using local fidelity can make the simulation run faster and consume less memory than setting a smaller global Minimum size. For example, if you have a patch antenna with traces that are 1mm in width, but one small region has 0.25mm wide traces, then it is best practice to use 1mm Minimum size and 0.25mm Local Fidelity.
  • Thin, axis-aligned traces ideally should be modeled as 2D surface bodies. However, if you choose to use 3D geometry, you do not need to reduce the minimum feature size (global or local) to the physical thickness of the trace, but only consider the width of the traces (and the proximity of different traces) when setting the minimum feature size.
  • Thin, non-planar or non-axis aligned structures are currently implemented as axis-aligned 2D structures. Therefore, if the structure is not axis-aligned, you must treat it as 3D geometry and must set a local fidelity equal to 80-100% of the thickness of the structure. It is assumed that the thickness is the smallest feature of the structure.
  • Circuit ports do not need local fidelities unless those ports have a width less than the global Minimum size. In such cases, higher accuracy is obtained by setting local fidelity to be the same as the width of the port.
  • Scaling of simulation time with the number of ports is optimal for antenna structures with a small number of ports, since the total time to solve all ports scales with the number of ports.
    • For devices with a high degree of symmetry, where the performance can be characterized only by considering one column (or a few columns) of the S-matrix, only one port (or a subset of ports) should be excited during initial exploration of the design, which can be achieved by making some of the ports passive. After initial exploration, the full S-matrix can be calculated.
  • Maximum simulation duration should be the largest of 50/fmin and 2Q/fres where:
    • fmin is the minimum frequency specified in the properties of your electromagnetic region,
    • Q is the quality factor, and
    • fres is the resonant frequency of any resonant mode of your device.

    Since Q and fres may not be known prior to simulation, envision Q as the maximum value you want resolved in the results.

  • Target numerical convergence should rarely be changed from the default of 1e-5. The only exception is for very high-quality factor resonances in a broadband simulation, where it may be as low as 1e-8.