Author: 孤筝(lvbowen040427@163.com)

Design Note: Aimed at university students with strong learning ability, this tutorial outline deliberately de-emphasizes mechanical “step-by-step screenshots” and instead focuses on the underlying electromagnetics and numerical computation logic of why things are set up this way. Mastering the physical and mathematical core is the key to applying knowledge flexibly.

1. Software Overview and Computational Electromagnetics Fundamentals

This chapter aims to build a foundational understanding of Ansys HFSS (High Frequency Structure Simulator). For beginners, the biggest pitfall is treating the simulation software as a “black box” and staying at the level of clicking through the interface. Understanding HFSS’s applicable boundaries and computational core (the Finite Element Method) is a prerequisite for evaluating simulation result credibility and diagnosing non-convergence issues.

1.1 Ansys HFSS Introduction and Applicable Boundaries

• Core Positioning: HFSS is a commercial software based on full-wave 3D electromagnetic field solving (Full-wave solver), with a computational foundation built on the rigorous solution of Maxwell’s Equations.
• Strengths: It excels at handling high-frequency electromagnetic field problems for arbitrary complex 3D structures, such as antenna design, microwave passive components (filters, couplers), RF connectors, package structures, and Radar Cross Section (RCS) analysis.
• Applicable Boundaries (When Not to Use HFSS):
  • Ultra-Low Frequency / Pure Circuit Problems: For lumped-parameter circuits or low-frequency motor designs, HFSS’s computational cost is excessively high and it is prone to low-frequency breakdown. In such cases, you should turn to SPICE or Maxwell (based on time-domain or low-frequency domain solvers).
  • Electrically Very Large Targets: When the physical dimensions of the simulation target (e.g., an entire aircraft or a large ship) are far greater than the operating wavelength, the finite element mesh scale typically grows rapidly with the electrical size (often approaching cubic growth in practice), and HFSS will consume massive amounts of memory. These problems are better suited to the Method of Moments (MoM) or Physical Optics (PO/SBR) solvers (such as Ansys Savant or HFSS’s SBR+ solution domain).
  • Pure 2D Layered Structures: For standard microstrip lines or multilayer PCB boards, using a 2.5D planar solver (such as HFSS 3D Layout or Siwave) is far more efficient than a full 3D solver.

Source: Ansys official training document Ansys HFSS User’s Guide, Chapter 1: HFSS Overview and Solver Technology.

1.2 Finite Element Method (FEM) Core Logic Overview

The Finite Element Method (FEM) is the underlying mathematical engine of HFSS. Computers cannot directly parse and process arbitrarily complex geometries in continuous space. The core idea of FEM is to “divide the whole into parts, and then reassemble the parts into the whole.”

• Spatial Discretization (Meshing): HFSS cuts the user-drawn 3D continuous space (solution domain) into hundreds of thousands or even millions of non-overlapping tiny “tetrahedra.” This process is called mesh generation.
• Basis Functions and Local Solving: Inside each tetrahedron, the electromagnetic field distribution is assumed to follow a simple mathematical polynomial (basis function). HFSS approximates the true electromagnetic field distribution by solving for the field strength at each tetrahedron’s nodes or edges.
• Assembly and Global Solving (Matrix Equation): All tetrahedron equations are assembled according to boundary conditions to form a massive sparse matrix equation (Ax = b). The HFSS solving process is essentially performing matrix inversion or iterative solving on this enormous matrix.
• Engineering Implications (Why FEM Matters):
  • Mesh Density Determines Accuracy: In regions where the electromagnetic field changes most dramatically (e.g., metal edges, gaps), a denser tetrahedral mesh is required.
  • Adaptive Meshing: What sets HFSS apart from other software is its automation. It first performs a trial calculation with a sparse mesh, identifies regions with high error, then automatically refines the mesh in those regions and recalculates—until the S-parameter variation (Delta S) between two consecutive calculations falls below a set threshold.

Source: Jian-Ming Jin, The Finite Element Method in Electromagnetics, with emphasis on the mathematical derivations of 3D vector finite elements and tetrahedral meshing.

1.3 Standard Simulation Workflow Analysis

Every HFSS project follows a rigorous linear logic loop. Skipping or reversing steps often leads to physical model contradictions or solver failures. The standard workflow can be broken down into the following six core steps:

• Step 1: Geometry Modeling
  • Draw the physical shape of objects in the 3D modeling interface, or import CAD models from external software such as SolidWorks or AutoCAD.
  • Key Point: Simplify mechanical features that have no impact on electromagnetic characteristics—such as chamfers and screw holes—to reduce unnecessary mesh generation.
• Step 2: Materials Assignment
  • Assign physical properties (dielectric constant, permeability, conductivity, loss tangent, etc.) to all geometric bodies.
  • Key Point: By default, HFSS treats the unassigned background space as absolute vacuum. If geometric bodies overlap, you must explicitly define material override priority (Material Override rules).
• Step 3: Boundaries
  • Define the boundary characteristics of the computational space. Since a computer cannot simulate an infinitely large universe, a boundary layer (such as an absorbing Radiation boundary or a Perfectly Matched Layer—PML) must be wrapped around the finite model exterior to simulate electromagnetic waves radiating toward infinity.
  • Key Point: Boundaries are also used to idealize object surfaces (e.g., setting them to Perfect E), eliminating the need to compute fields inside the metal and drastically reducing computational cost.
• Step 4: Excitations (Ports)
  • Define the “gateway” through which electromagnetic waves enter or leave the system. Common types include Wave Ports (for external feeding, such as waveguides or coaxial lines) and Lumped Ports (for internal feeding, such as microstrip lines).
  • Key Point: Excitations not only input energy but also compute the characteristic impedance (Z₀) at the port.
• Step 5: Analysis Setup (Solver Settings)
  • Define the solution frequency (typically set to the highest operating frequency or resonant frequency, to guide the adaptive meshing size), convergence criteria (Delta S), and sweep frequency range.
• Step 6: Post-Processing
  • After solving, extract the engineering-required data representations. This includes network parameters (S/Y/Z parameter matrices), spatial field distributions (3D color maps or vector plots of electric/magnetic fields), and antenna radiation characteristics (radiation patterns, gain, axial ratio).

Source: David M. Pozar, Microwave Engineering, on network analysis and the theoretical definition of S-parameters for microwave networks, mapped to the practical application of HFSS post-processing parameter extraction.

2. Project Management and Interface Navigation

This chapter focuses on analyzing the top-level architecture and interaction logic of the Ansys software. For beginners, clarifying the data hierarchy of project files and selecting the correct “Design Type” according to project goals is the cornerstone of avoiding later rework.

2.1 Ansys Electronics Desktop (AEDT) Architecture

Earlier versions of HFSS were standalone software, but modern versions have all been integrated into the unified Ansys Electronics Desktop (AEDT) desktop environment. Understanding AEDT’s hierarchical architecture is crucial for managing complex multi-physics co-simulations.

• AEDT’s “Shell” and “Engine”: AEDT itself is merely a Graphical User Interface (GUI) and project management platform. Internally, it encapsulates multiple solver engines (e.g., HFSS for high-frequency, Maxwell for low-frequency motors, Q3D for parasitic parameters, Icepak for thermodynamics).
• Data Hierarchy: In the AEDT Project Manager tree view, the following nesting logic is strictly followed:
  • Project: The highest level, corresponding to a single .aedt file on disk. A project can contain multiple different designs.
  • Design (e.g., HFSSDesign1): This is the specific workspace containing a particular geometry model, boundary conditions, and solution setup. You can create an antenna design (HFSS) and an RF circuit design (Circuit) within the same project and co-link them.
  • Setup: Mounted under a specific design, defining at what precision and what frequency the model will be computed by the engine.
• Underlying Logic: All geometry modifications, material assignments, and boundary conditions are reflected in real time in the Project Manager tree nodes. When debugging, checking the tree nodes top-to-bottom (looking for red X marks or yellow warnings) is the most efficient method.

Source: Ansys official documentation Ansys Electronics Desktop Help, System Architecture and Project Management chapter.

2.2 Core Design Type Analysis: 3D vs 3D Layout

When creating a new HFSS project in AEDT, beginners are often confused by Insert HFSS Design and Insert HFSS 3D Layout Design. Both invoke the same underlying FEM full-wave solver engine, but the target physical forms and modeling logic are entirely different.

• HFSS (3D / MCAD Mode):
  • Applicable Scenarios: Complex structures in arbitrary 3D space, such as horn antennas, waveguide cavities, coaxial connectors, radar reflectors, etc.
  • Modeling Logic: Based on mechanical CAD (MCAD) Boolean operation thinking. Space is free—you can extrude or rotate solids along any X, Y, or Z axis.
• HFSS 3D Layout (ECAD Mode):
  • Applicable Scenarios: Layered structures, printed circuit boards (PCB), chip packages, microstrip line networks.
  • Modeling Logic: Based on electronic CAD (ECAD) 2.5D thinking. It pre-defines the concept of “layers (stackup)” (e.g., Top layer, Dielectric layer, Bottom layer). You only need to draw traces and vias on a 2D plane, and the software automatically assigns Z-axis thickness based on the stackup.
• Engineering Implication: Never manually draw a 10-layer complex PCB in the HFSS 3D environment—this will result in extremely low modeling efficiency and massive mesh fragmentation. Similarly, do not attempt to build a spherical lens antenna in 3D Layout.

Source: Eric Bogatin, Signal and Power Integrity - Simplified, combined with Ansys SI/PI training materials on ECAD model extraction engineering practices.

2.3 View Controls, Shortcuts, and Underlying Options

Familiarity with shortcuts is not just about efficiency—it is key to building spatial awareness. HFSS’s 3D interaction logic differs slightly from traditional mechanical drafting software.

• Frequently Used Navigation Shortcuts:
  • Rotate: Hold Alt key + left mouse button drag.
  • Pan: Hold Shift key + left mouse button drag.
  • Zoom: Scroll the mouse wheel, or hold Alt + Shift + left mouse button drag up/down.
  • Fit All: Shortcut Ctrl + D—recovers the view in one click when the model flies off screen.
• Smart Selection Filters:
  • When applying boundary conditions or meshing, accurately selecting the target is critical.
  • Shortcut O (Object): Default state; selects the entire 3D solid.
  • Shortcut F (Face): Extremely commonly used; selects only the solid’s surface (for specifying surface current or surface boundary conditions).
  • Shortcut E (Edge) / V (Vertex): Selects edges or vertices (usually for measurements or establishing local coordinate systems).
• Essential Underlying Options (Tools > Options):
  • Auto-Save Mechanism: Since high-frequency simulations can easily exhaust memory and crash the software, always enable Auto Save in General Options > Project Options (recommended interval: 15–30 minutes).
  • Default Material Override: In HFSS Options, understand the Material Override mechanism. By default, HFSS allows the material of an inner solid to override that of an outer solid (e.g., drawing a copper block inside a vacuum cavity—the overlapping region is automatically treated as copper). If you are unaware of this mechanism, it is very easy to end up with a physical model that doesn’t match your intentions.

Source: Ansys software built-in Help Keyboard Shortcuts manual and engineering application best practices summary.

3. Geometry Modeling and Parametrics

This chapter explores how to build a digital twin of the physical world in HFSS. For advanced users, modeling is not just about “drawing a shape”—it is about establishing a mathematically topologically connected model. Poor modeling habits can cause severe mesh degradation or make automated optimization impossible later.

3.1 3D Solid Modeling Based on Boolean Operations

HFSS’s built-in 3D modeler adopts a typical Constructive Solid Geometry (CSG) logic. Unlike artistic polygon modeling, the core idea here is to “assemble” and “carve” complex structures from primitive solids through logical operations.

• Reduction to Primitives: Every complex structure can be decomposed into combinations of boxes, cylinders, spheres, or 2D sheets.
• The Three Core Boolean Operations:
  • Unite (Union): Merges multiple intersecting solids of the same material into one, eliminating interior redundant faces. This effectively reduces the mesh generator’s computational burden at interfaces.
  • Subtract (Difference): Uses one solid (tool) to “hollow out” another solid (blank). Commonly used for constructing resonant cavities, coaxial cable insulation, or waveguide slots.
  • Intersect: Retains only the overlapping spatial volume of two solids.
• Underlying Modeling Philosophy: “Do not add entities unless necessary.” In electromagnetic simulation, you must resolutely remove mechanical details that have minimal impact on electromagnetic wave propagation (such as screw threads, tiny chamfers, and non-critical support structures). These minute features (sliver faces) will force the FEM engine to generate extremely dense, deformed tetrahedral meshes in those regions, directly causing out-of-memory errors.

Source: Ansys official training document Ansys HFSS 3D Modeler User’s Guide, chapters on Boolean operations and model simplification.

3.2 Variables and Parametric Structure Creation

Parametrics is the dividing line between a “novice” and an “engineer.” In HFSS, never hard-code specific values in dimension input fields (e.g., directly entering 5mm). Instead, always input variable names (e.g., patch_length). Because simulation is fundamentally a process of continuous trial-and-error and optimization.

• Local Variables vs Project Variables (Global Variables):
  • Local Variables: Defined directly (e.g., radius = 2mm), with scope limited to the current Design (e.g., effective only in this particular antenna design).
  • Global Variables: Prefixed with $ (e.g., $substrate_h = 1.6mm), with scope spanning the entire Project file. When performing multi-physics co-simulation (e.g., HFSS for electromagnetic, Icepak for thermal) and unified dimensions are required, global variables must be used.
• Mathematical Expressions and Topological Constraints: Variables are not just static values—they can be trigonometric functions or interdependent equations. For example, when defining a quarter-wavelength impedance transformer, the length can be written as lambda/4 (where lambda is derived from the speed of light and the center frequency).
• Underlying Logic of Parametrics: Establish constraint relationships. For example, when placing a microstrip patch on a dielectric substrate, the patch’s Z-axis coordinate should not be hard-coded—it should be set to the substrate thickness variable h. This way, when you later sweep the substrate thickness during optimization, the patch will automatically “float” on the surface rather than sinking into the substrate or hanging in mid-air.

Source: Stephen H. Hall, Advanced Signal Integrity for High-Speed Digital Designs, borrowing its engineering specification for building fully parametric topological structures in channel modeling.

3.3 Material Library Management

The propagation speed and attenuation characteristics of electromagnetic waves in different media are entirely governed by the material’s constitutive parameters: relative permittivity ($\varepsilon_r$), relative permeability ($\mu_r$), conductivity ($\sigma$), and dielectric loss tangent ($\tan\delta$).

• Material Assignment Mechanism and Override Rules: By default, the HFSS background is vacuum. In practice, solid overlaps should be avoided as much as possible. If overlaps exist for modeling convenience, you can enable Enable material override in HFSS > Design Settings and explicitly define which category of geometry overrides the other (e.g., a via overrides the corresponding volume in the dielectric substrate).
• Anisotropic Materials: The vast majority of basic simulations assume materials are isotropic (i.e., the permittivity is the same in all directions). However, in advanced designs (e.g., using sapphire substrates or special liquid crystal polymer—LCP), the $\varepsilon_r$ along the X, Y, and Z axes differs. HFSS allows inputting these anisotropic parameters via tensor notation, which is critical for high-frequency millimeter-wave circuits.
• Dispersive Materials: For ultra-wideband (UWB) antennas or optical-frequency simulations, $\varepsilon_r$ and loss vary dramatically with frequency. In such cases, you cannot input a constant—you must introduce a dispersion model (such as Debye or Drude models) or import frequency-dependent measured datasets.
• Underlying Logic: Pay close attention to $\tan\delta$. Many beginners find that their simulated insertion loss (S₂₁) doesn’t match measurements—this is often because they overlooked conductor surface roughness and the nonlinear variation of the dielectric loss tangent at high frequencies.

Source: Constantine A. Balanis, Advanced Engineering Electromagnetics, on constitutive relations and dielectric polarization theory.

4. Boundary Conditions: Spatial Truncation and Physical Idealization

This chapter explores the extremely core concept of “boundary conditions” in HFSS. Electromagnetic fields extend infinitely in the real universe, but computer memory is finite. The essence of boundary conditions is to use the laws of electromagnetics and mathematical boundary value problems (BVP) to trick the computer within a finite computational domain, making its solution equivalent to that of an infinite space or an extremely complex physical structure.

4.1 Physical Meaning of Boundary Conditions

• Spatial Truncation: The Finite Element Method (FEM) cannot handle an infinitely large mesh matrix. Boundary conditions act as a “wall” built around the computational domain, telling the solver: “At this boundary, the electromagnetic wave behavior follows a specific mathematical rule—there’s no need to compute further outward.”
• Physical Idealization and Dimensionality Reduction: In the microwave frequency range, the skin depth of metals is extremely shallow (typically on the order of micrometers). If 3D meshing were used to partition the interior of a thick copper plate, the mesh count would explode. By applying boundary conditions on the metal surface, we can reduce a 3D solid problem to a 2D surface current problem, saving at least 90% of computational resources.
• Underlying Logic: If no open boundary (such as Radiation or PML) is explicitly set, the outer boundary of the finite solution domain will behave as a closed boundary. The model effectively becomes enclosed in a cavity, which can introduce non-physical reflections.

Source: Constantine A. Balanis, Advanced Engineering Electromagnetics, on the mathematical uniqueness theorem for electromagnetic field boundary value problems (BVP).

4.2 Common Material Surface Boundaries

• Perfect E (Perfect Electric Conductor):
  • Physical Characteristics: Represents an ideal conductor with infinite conductivity. The electric field vector must be perpendicular to this surface; the tangential electric field is zero ($E_{tan} = 0$).
  • Engineering Application: Used to simulate ideal metal ground planes, waveguide inner walls, or antenna radiating patches.
• Perfect H (Perfect Magnetic Conductor):
  • Physical Characteristics: A true magnetic conductor does not exist in nature. It forces the magnetic field vector to be perpendicular to the surface; the tangential magnetic field is zero ($H_{tan} = 0$).
  • Engineering Application (Core Technique): Often used in conjunction with Perfect E to exploit electromagnetic symmetry (Symmetry) to bisect the model. For example, cutting a symmetric horn antenna into quarters along its symmetry planes and assigning Perfect E and Perfect H boundaries yields identical radiation characteristics, while memory consumption and solution time are each reduced to 1/4 of the original.
• Finite Conductivity (Finite Conductivity Boundary):
  • Physical Characteristics: Accounts for ohmic losses in real metals (e.g., copper, aluminum, gold). It allows a minuscule electromagnetic wave transmission (skin depth) and computes the active power loss on the surface.
  • Engineering Application: When calculating the quality factor of a high-Q resonant cavity or precisely evaluating the insertion loss of a microstrip line at millimeter-wave frequencies, this boundary must be used. Surface roughness models (e.g., Groisse or Huray models) can be further introduced.

4.3 Open-Space Truncation Strategies

For problems involving outward-radiating energy such as antennas or Radar Cross Section (RCS), the outermost layer of the model must be wrapped with an “absorbing material” to ensure zero reflection when electromagnetic waves pass through the boundary.

• Radiation (Absorbing Boundary Condition — ABC):
  • Principle: An approximate mathematical boundary based on the Sommerfeld radiation condition. It assumes that electromagnetic waves arriving at this boundary are normally incident plane waves, thereby absorbing them completely.
  • Sizing Specification (λ/4 Rule): Since it only absorbs normally incident waves effectively, it must be placed at least λ/4 (quarter-wavelength at the operating frequency) away from the radiating body. This is because within the reactive near-field region (less than λ/4), electromagnetic energy exists in the form of stored energy rather than radiation. Forcing a truncation here causes severe non-physical reflections that directly compromise the accuracy of S-parameters.
• PML (Perfectly Matched Layer):
  • Principle: PML is not a simple boundary but an artificially constructed, extremely special anisotropic non-physical “lossy material.” Theoretically, it can perfectly absorb electromagnetic waves from any angle of incidence at any frequency.
  • Engineering Comparison and Application: Compared to an ordinary Radiation boundary, PML’s absorption performance is far superior. Therefore, it can be placed much closer to the radiating body (e.g., λ/10), reducing the total computational volume. The trade-off is that PML itself requires a denser mesh and more complex matrix solving. Generally, PML is mandatory for extremely high-precision antenna gain calculations or grazing-incidence scattering problems.

Source: Ansys official documentation HFSS Boundaries and Excitations; theoretical derivation references Jean-Pierre Bérenger’s original paper on PML invention and Antenna Theory: Analysis and Design.

5. Excitations: Electromagnetic Wave Input and Extraction

This chapter analyzes how electromagnetic energy enters and exits the simulation model. The port is not only a physical interface for feeding but also the absolute reference plane for computing S-parameters (scattering parameters) and extracting characteristic impedance ($Z_0$). Incorrect port configuration is the most common cause of S-parameter divergence and impedance mismatch artifacts.

5.1 Core Concepts of Excitations: Dimensionality Reduction and Expansion from 2D to 3D

• Two-Step Solving Mechanism: When HFSS begins computing, it first performs a 2D Eigenmode Solution on the port cross-section you defined. The purpose of this step is to determine which electromagnetic wave modes (Modes)—such as TE, TM, or TEM—the cross-section (e.g., a microstrip line or waveguide cross-section) can support, and to compute the characteristic impedance of that cross-section.
• Transfer to the Volumetric Solver: HFSS then takes the field distribution obtained from the 2D solution and injects it as a boundary condition into the 3D Full-Wave Solution domain.
• Underlying Logic: Understanding this mechanism is critical. If your port cross-section is drawn too small, the 2D solution will fail to capture the complete fringing fields; if drawn too large, the port itself may become a waveguide and excite higher-order modes that don’t exist in reality.

Source: Ansys official documentation HFSS Excitations and Ports, on the underlying mathematical principles of the 2D port solver.

5.2 Wave Port Principles and Specifications

The Wave Port is the most precise and most commonly used external excitation method in HFSS. It assumes the port is connected to an infinitely long semi-rigid transmission line.

• Applicable Scenarios: Waveguide cavity end faces, coaxial line cross-sections, microstrip lines/striplines located at the edge of an absorbing boundary.
• External vs Internal Wave Ports:
  • External (Default): The Wave Port must be placed on the outermost boundary (background) of the entire 3D model. Waves can only radiate inward in a single direction.
  • Internal Wave Port: If forced to use a Wave Port inside the model, you must draw a Perfect E (ideal conductor) cap on the back of the port, equal in size to the port, to force the electromagnetic wave to propagate in only one direction. Otherwise, the wave radiates in both directions, causing completely incorrect S-parameter calculations.
• Sizing Empirical Formulas (Microstrip Line Example):
  • A Wave Port cannot be placed only on the metal trace—it must include the dielectric and air surrounding the trace to fully encapsulate the electromagnetic field.
  • Height: Typically set to 6 to 10 times the dielectric substrate thickness $h$ ($6h \sim 10h$).
  • Width: Typically set to about 10 times the trace width $w$, or dynamically adjusted based on the ratio of $w$ to $h$ (e.g., when $w \ge h$, the width is set to $10w$).
  • Pitfall to Avoid: Never let the edge of the Wave Port intersect with or be too close to the model’s Radiation (absorbing) boundary—this will cause severe boundary conflict errors.

5.3 Lumped Port Details

A Lumped Port is analogous to an ideal voltage/current source in a circuit—it directly imposes an electric field between two metal surfaces.

• Applicable Scenarios: Internal feeding (e.g., coaxial feed probes inside microstrip antennas, equivalent excitation for SMT pads of patch components), low-frequency bands or extremely electrically small slots.
• Internal Excitation Characteristics: Lumped Ports must be located inside the model and must connect two conducting surfaces (e.g., connecting a signal line to a ground plane). It does not perform 2D eigenmode pre-computation; instead, it forces the assumption that the electric field on the port face is uniform (TEM mode).
• Normalized Impedance Setting: Since it does not perform a 2D computation, a Lumped Port cannot automatically determine the characteristic impedance. The user must manually specify a reference impedance (typically $50\Omega$). All computed S-parameters will be normalized based on this given impedance.
• Usage Constraints: The physical length and width of a Lumped Port must be much smaller than one-tenth of the wavelength at the highest operating frequency ($< \lambda/10$); otherwise, the uniform field assumption breaks down, producing severe parasitic inductance and capacitance.

Source: David M. Pozar, Microwave Engineering, on the theory of equivalent differences between lumped elements and distributed transmission lines at microwave frequencies.

5.4 Modes and Integration Lines

• The Bridge to Impedance Calculation: In microwave engineering, “voltage” and “current” at high frequencies are ambiguous concepts (because the field is distributed in space). HFSS calculates voltage by defining an Integration Line on the port face: $V = \int E \cdot dl$. The characteristic impedance of the mode is then computed via $Z = V^2 / (2 \cdot Power)$.
• Phase Reference: The integration line specifies the positive direction of the electric field vector (typically from ground toward the signal line). In multi-port networks (such as power dividers and couplers), if the integration line directions are inconsistent across ports, the extracted S-parameter phases (e.g., the phase of $S_{21}$) will have a 180-degree inversion error.
• Multi-Mode Transmission: For structures like waveguides that may simultaneously transmit the dominant mode (e.g., $TE_{10}$) and higher-order modes, you must increase the “Number of Modes” in the port settings and draw separate integration lines for each potential mode to extract their respective S-parameters. This is critical when designing mode converters or filters.

6. Solver Configuration and Adaptive Meshing Mechanism

This chapter explains the configuration logic of HFSS’s core computational engine. Properly setting solution parameters is a critical step in ensuring that Finite Element Analysis (FEM) results are physically meaningful and that computational resource consumption remains within a controllable range.

6.1 Solution Type Selection Criteria

HFSS offers three mainstream full-wave 3D solution modes, with fundamentally different underlying mathematical derivations and boundary condition settings. Before designing, you must accurately select based on the physical characteristics of the Device Under Test (DUT).

• Driven Modal:
  • Computational Basis: Based on microwave network theory. By solving for the eigenmode distribution at the port (e.g., TE, TM, TEM modes), it calculates the ratio of incident to reflected power for different modes, yielding a generalized S-parameter matrix based on modes.
  • Applicable Scenarios: The vast majority of passive microwave devices, such as antennas, waveguide cavities, microstrip filters, power dividers, and Radar Cross Section (RCS) calculations.
• Driven Terminal:
  • Computational Basis: Based on multi-conductor transmission line theory. It models the system in terms of “node voltages” and “branch currents,” directly extracting the terminal S-parameter matrix and corresponding node impedances for multi-port networks.
  • Applicable Scenarios: Signal Integrity (SI) and Power Integrity (PI) analysis, multi-pin IC package modeling, differential pair routing. In this mode, extracting differential-mode/common-mode impedance is more straightforward.
• Eigenmode:
  • Computational Basis: A pure mathematical eigenvalue solving problem (passive, no external excitation). By solving Maxwell’s equations for resonant solutions, it obtains the natural resonant frequencies of the structure and their corresponding 3D field distributions.
  • Applicable Scenarios: Designing high quality factor (Q) resonant cavities and dielectric resonators, and extracting the unloaded Q ($Q_u$) of specific cavities.

Source: Ansys Electronics Desktop official Help documentation: HFSS Solution Types and Excitation Fundamentals.

6.2 Solution Frequency Specification

The Solution Frequency determines the target wavelength ($\lambda = c/f$) for HFSS’s adaptive meshing. Since FEM requires a sufficient number of tetrahedral elements per wavelength (typically 6–10), the choice of solution frequency directly determines the mesh density and computational accuracy.

• Narrowband and Resonant Structures (e.g., microstrip antennas, narrowband filters):
  • Specification: Set to the expected center resonant frequency. Near the resonant frequency, the standing-wave distribution of the electromagnetic field is most intense and the gradient variation is greatest. Meshing at this frequency captures the critical electric field characteristics with the greatest precision.
• Broadband Structures (e.g., broadband antennas, high-speed connectors):
  • Specification: Set to the highest operating frequency in the sweep range. The highest frequency corresponds to the shortest spatial wavelength; the mesh generated based on this baseline is the finest and can accommodate lower frequency accuracy requirements by downward coverage. If set to the center frequency or a lower frequency, the mesh at high frequencies will be too coarse, causing non-physical divergence in high-frequency S-parameters.
• Multi-Band and Filter Passbands:
  • Specification: For multi-band antennas, enable the Multi-Frequencies option in Setup and input multiple frequency points to guide meshing separately. For wide stopband filters, it is recommended to set the solution frequency at the highest edge frequency of the passband.

Source: David M. Pozar, Microwave Engineering, combined with wavelength and mesh discretization error theory from finite element analysis.

6.3 Frequency Sweep Algorithm Comparison

After obtaining the mesh at the center frequency, an appropriate algorithm must be selected to compute the response curve across the entire frequency band. A trade-off between computation time and data accuracy is required.

• Discrete Sweep:
  • Mechanism: At each user-specified frequency point, a complete matrix equation solving is performed from scratch.
  • Characteristics: Highest data accuracy (considered the ground truth), but extremely long computation time and large storage requirements.
  • Scope: Strict verification of final results, or a mandatory choice when the DUT has a very high Q value with multiple extremely sharp resonant peaks (interpolation algorithms may miss peaks).
• Fast Sweep:
  • Mechanism: Based on the ALWE (Asymptotic Linear Waveform Evaluation) algorithm, it extracts derivative information of the electromagnetic field at the solution frequency and uses Taylor series or Padé approximation to extrapolate the response at adjacent frequencies.
  • Characteristics: Fastest computation speed, but error increases exponentially in frequency bands far from the solution frequency.
  • Scope: Extremely narrowband structures (e.g., high-frequency narrowband bandpass filters); the applicable bandwidth typically does not exceed 10%–15% of the center frequency.
• Interpolating Sweep:
  • Mechanism: Automatically selects key frequency points within the band for Discrete solving, then uses rational function fitting to smoothly interpolate these discrete points. If the fitting error between adjacent points exceeds the set threshold, the algorithm automatically adds solution frequency points until the entire S-parameter curve meets the convergence criterion.
  • Characteristics: Achieves the optimal balance between computational efficiency and broadband accuracy.
  • Scope: Broadband antennas, high-speed digital signal links, and most other modern microwave engineering scenarios—it is currently the default and preferred method.

6.4 Adaptive Meshing Mechanism Analysis

HFSS’s core advantage lies in its automated error control and mesh refinement process. Understanding this mechanism helps diagnose “non-convergence” engineering errors.

• Iteration Process (Passes):
  1. Generate an initial coarse mesh (Initial Mesh) covering the entire physical space.
  2. Perform a full-wave calculation at the set solution frequency.
  3. Evaluate the electric field gradient and error throughout the space, identifying regions with the highest error (e.g., metal edges, gaps, near ports).
  4. Perform mesh refinement in high-error regions, increasing the number of tetrahedral elements (by default, 30% per iteration).
  5. Repeat step 2 and compare the S-parameter matrix differences between consecutive iterations.
• Core Convergence Criterion: Delta S ($|\Delta S|$):
  • Mathematical Definition: The maximum absolute value of the variation of all elements in the entire S-parameter matrix between two consecutive mesh iterations. The formula can be simplified as $\max_{i,j} |S_{ij}^{N} - S_{ij}^{N-1}|$.
  • Typical Values:
    • Basic antennas and microwave passive devices: Typically set to 0.02 (i.e., allowing a maximum 2% error).
    • High-precision multiplexers, strongly coupled arrays, or high-speed differential lines: Must be increased to 0.01 or lower.
• Maximum Number of Passes:
  • Function: A forced termination mechanism. Because certain geometric models contain non-physical singularities (e.g., the electric field strength at the edge of an infinitely thin ideal metal is theoretically infinite), these regions may never converge, causing the mesh to enter an infinite refinement loop.
  • Recommendation: Typically set between 15 and 20. If Delta S still hasn’t dropped below the threshold after reaching the maximum, the model likely contains small unreasonable gaps or self-intersecting boundaries. You should return to the 3D Modeler to check the geometry topology rather than blindly increasing the iteration count.

Source: Ansys official training material HFSS Solvers and Meshing Strategies, on error calculation and matrix convergence criteria.

7. Simulation Verification and Multi-Dimensional Post-Processing

This chapter details data extraction and physical characterization analysis after the simulation computation is complete. The HFSS solver essentially generates a massive electromagnetic field database. The core capability of post-processing lies in extracting S-parameter network characteristics, spatial field distributions, and far-field radiation performance from this database according to engineering requirements, thereby providing an objective evaluation of the design.

7.1 Validation Check and HPC Settings

Before clicking “Analyze All” to launch the solver, strict pre-project checks and computational resource configuration must be performed to avoid computational resource waste caused by trivial errors.

• Validation Check (Automatic Verification Mechanism):
  • Execution Logic: The software sequentially checks the project’s 3D geometry topology (inspecting for unprocessed solid intersections), material properties, boundary condition assignment conflicts, and whether excitation port settings conform to the selected solution mode (e.g., whether a reference ground is missing in Terminal mode).
  • Engineering Standard: You must ensure all items in the Validation Check window display a green checkmark. Any yellow warnings must be investigated, and red errors prevent the simulation from running.
• HPC (High-Performance Computing) Hardware Resource Scheduling:
  • Multi-Core Parallel Allocation: The FEM matrix solving in HFSS is highly dependent on CPU performance and memory throughput. In Simulation Setup > HPC and Analysis Options, you need to allocate Tasks and Cores per Task based on the workstation’s actual physical core count.
  • Sweep Parallelization: For projects using Discrete sweeps with many frequency points, frequency-parallel computation can be enabled through HPC settings, distributing computation tasks for different frequency points across multiple cores or even multiple cluster nodes simultaneously, reducing computation time in a roughly linear fashion.

Source: Ansys official documentation Ansys Electronics Desktop HPC Admin Guide.

7.2 Network Parameter Extraction

Network parameters are the most intuitive frequency-domain metrics for evaluating microwave passive devices and antenna feed networks, located at Results > Create Modal Solution Data Report.

• S-Parameter Matrix (Scattering Parameters):
  • $S_{11}$ (Return Loss): The core metric characterizing the impedance matching quality of a port. Typically, $S_{11} < -10 \text{ dB}$ (i.e., reflected power less than 10%) is required across the operating band.
  • $S_{21}$ (Insertion Loss): Characterizes the efficiency of energy transfer from Port 1 to Port 2. An ideal filter’s passband $S_{21}$ should be close to $0 \text{ dB}$, and the stopband should be well below $-30 \text{ dB}$.
  • Plotting Standard: In engineering plots, the S-parameter vertical axis must use a logarithmic decibel (dB) scale, and the horizontal axis uses a linear frequency scale.
• Smith Chart:
  • Analytical Value: A polar plot that maps the complex reflection coefficient ($\Gamma$) to the normalized complex impedance ($Z/Z_0$) plane. By observing the trajectory of the frequency band curve on the Smith Chart, you can intuitively determine whether the device is capacitive or inductive.
  • Engineering Application: Combined with broadband matching theory, design matching networks (L-type, $\pi$-type circuits) to converge the core portion of the curve to the $50\Omega$ matching point at the center of the chart.
• Y Parameters and Z Parameters: Typically used to export the HFSS 3D model as a broadband equivalent circuit model (SPICE netlist) for subsequent system-level circuit simulation.

7.3 Spatial Field Distribution Visualization

Field Overlays can intuitively reveal the propagation path, resonant modes, and energy dissipation regions of electromagnetic waves within the structure, forming the core basis for optimizing structural design.

• Prerequisites for Plotting: You must select a specific solid face or spatial cross-section plane in the 3D Modeler interface before you can add a Field Plot via right-click.
• Mag E / Mag H (Scalar Field Intensity Distribution):
  • Displays color maps of the absolute magnitude of the electric or magnetic field.
  • Debugging Application: Used to locate electric field breakdown risk points in high-power microwave devices (regions of concentrated field intensity) or magnetic field energy accumulation zones within filter cavities.
• Vector E / Vector H (Vector Field Distribution):
  • Displays both the direction and magnitude of the field in arrow form.
  • Physical Analysis: Allows clear observation of TE/TM mode distribution in waveguides or the coupling path of fringing fields along microstrip line edges.
• Surface Current ($J_{surf}$):
  • Core for Antenna Design: Current is the source of radiation. Surface current plots precisely show the resonant path on the antenna and inactive regions (where current is minimal—structure can typically be trimmed here without affecting performance).
  • Dynamic Animation: By performing a phase ($\omega t$) sweep animation, you can observe the current flow process on the conductor surface and the wave propagation pattern.

7.4 Far-Field Radiation Characteristics

For antenna and Radar Cross Section (RCS) analysis, all core performance metrics are extracted based on the far-field environment. You must pre-define an infinite radiation sphere in Radiation > Insert Far Field Setup, typically covering the full $\theta$ ($0^\circ \sim 180^\circ$) and $\phi$ ($0^\circ \sim 360^\circ$) angular sweeps.

• Radiation Pattern:
  • 2D Polar Pattern: Select specific cut planes (e.g., E-plane, H-plane) and plot radiation intensity versus angle. Used to evaluate Half-Power Beamwidth (HPBW) and Side Lobe Level (SLL).
  • 3D Pattern: Intuitively displays the antenna’s beam coverage shape in 3D space.
• Gain and Directivity:
  • Directivity: Depends only on antenna geometry and pattern focusing, assuming no internal losses and perfect matching.
  • Gain: Accounts for dielectric losses and metallic ohmic losses of the antenna materials ($Gain = e_{rad} \times Directivity$).
  • Realized Gain: The most rigorous metric in engineering. Building on gain, it further deducts the reflection loss due to port impedance mismatch ($Realized Gain = Gain \times (1 - |S_{11}|^2)$).
• Axial Ratio (AR):
  • The decisive parameter for characterizing circularly polarized antenna performance.
  • Evaluation Criterion: An ideal circular polarization has an AR of $0 \text{ dB}$. In engineering, the frequency band where $AR < 3 \text{ dB}$ is typically defined as the effective operating bandwidth of a circularly polarized antenna.

Source: Constantine A. Balanis, Antenna Theory: Analysis and Design, chapters on far-field radiation region parameter definitions.

8. Advanced Capabilities: Parametric Sweeps and Engineering Optimization (Optimetrics)

This chapter introduces HFSS’s automated optimization module (Optimetrics). After completing a basic single-dimension simulation, the core challenge of engineering design lies in multi-variable coupled tuning. Using Optimetrics, you can transform manual “trial-and-error” into an automated global optimization mechanism based on mathematical algorithms.

8.1 Parametric Sweep Quick Start

Parametric sweeps are the most fundamental method for establishing the mapping between multiple variables and electromagnetic performance. Their essence is to have the solver automatically execute exhaustive calculations based on pre-set variable step sizes.

• Prerequisite: The 3D model must be fully parameterized (i.e., all key dimensions are driven by local or global variables rather than fixed numerical values).
• Sweep Space Definition:
  • Linear Step: Specify the variable’s start value, end value, and step size (e.g., width from 1mm to 5mm, step 0.5mm).
  • Linear Count: Specify the range and total number of calculation points; the software automatically divides the interval evenly.
  • Multi-Variable Nested Computation: When scanning multiple variables simultaneously (e.g., Variable A scans 5 points, Variable B scans 5 points), the computational cost grows exponentially ($5 \times 5 = 25$ complete full-wave solutions). When computational resources are limited, combinations must be chosen carefully.
• Result Representation and Stacked Comparison:
  • In the Report interface, setting the variable as a Family attribute allows you to plot all scanned S-parameter curves or antenna radiation patterns on a single 2D chart. This is used to intuitively identify the optimal parameter range.
• Computational Resource Optimization: Exercise extreme caution when checking Save Fields and Mesh in Optimetrics settings. Saving 3D field data for every sweep point will rapidly exhaust hard drive storage. Usually, saving only network parameters (S/Y/Z) is sufficient for evaluation purposes.

Source: Ansys Electronics Desktop Help - Optimetrics: Parametric Analysis.

8.2 Optimization Algorithms

When design variables exceed 3 and are mutually coupled, the computational cost of exhaustive scanning becomes unacceptable. The optimization module introduces mathematical algorithms to automatically search the multi-dimensional solution space for the optimal parameter combination that satisfies engineering objectives.

• Cost Function Setup:
  • The core of optimization lies in converting physical requirements into mathematical criteria.
  • You can set multiple objectives in Optimizer Setup > Goals. For example: require $S_{11} < -15 \text{ dB}$ at a target frequency point while setting Gain $> 5 \text{ dBi}$.
  • Weight: When multiple objectives have physical conflicts (e.g., simultaneously requiring very wide bandwidth and very high gain), different weight values (e.g., setting return loss weight to 1 and gain to 2) must be assigned to guide the algorithm’s emphasis.
• Core Optimization Algorithm Comparison and Application:
  • SNLP (Sequential Non-Linear Programming) / Gradient Descent:
    • Mechanism: Computes the partial derivatives of the cost function with respect to each variable and follows the direction of steepest gradient descent to find a minimum.
    • Characteristics: Extremely fast convergence.
    • Applicable Boundary: A local optimizer. Effective only when the initial design parameters are already close to the optimal solution. If the solution space is complex, it easily gets trapped in local minima.
  • Genetic Algorithm (GA) / Particle Swarm Optimization (PSO):
    • Mechanism: Simulates crossover, mutation, and selection in biological evolution, or the information-sharing mechanism of bird flocking foraging.
    • Characteristics: Global optimizers. Capable of large-scale search across the full parameter space with the ability to escape local minima.
    • Applicable Boundary: Must be used when no good initial reference dimensions are available and complex structures need to be explored from scratch. The downside is the need to compute massive sample sets, which is extremely time-consuming.

Source: Kalyanmoy Deb, Multi-Objective Optimization using Evolutionary Algorithms, on convergence theory of engineering optimization algorithms.

8.3 Sensitivity and Tuning Analysis

After achieving optimal electromagnetic performance, the design must be transitioned to manufacturing. Since machining inevitably involves tolerances, you must evaluate the robustness of the design to manufacturing errors.

• Sensitivity Analysis:
  • Physical Meaning: Quantifies the sensitivity of model output results to small variations in a single input variable. Its mathematical essence is extracting the first-order derivative (linear sensitivity) of the cost function at the current design point via Taylor series.
  • Engineering Application: Analysis identifies the “high-sensitivity variables” with the greatest impact on performance (e.g., the distance between the microstrip line edge and the resonant cavity wall). In manufacturing drawings, stricter tolerance classes must be specified for these critical dimensions, while requirements for low-sensitivity variables can be relaxed to reduce machining costs.
• Tuning Analysis:
  • Execution Logic: HFSS does not recompute the full-wave matrix every time the user drags a slider. Instead, the user pre-defines the variable tuning range, and the software automatically pre-computes the solution space matrix for these variable combinations in the background.
  • Interactive Representation: Generates a control panel with real-time sliders. As you drag the sliders, the S-parameter curves on the chart update smoothly in real time, providing excellent physical intuition and a “what you see is what you get” parameter tuning experience. This is commonly used for final fine-tuning of impedance matching points.

Source: Ansys official training document Optimetrics Sensitivity and Tuning Analysis.

9. Troubleshooting Guide and Authoritative Data Source Index

This chapter focuses on the exception diagnosis logic in engineering practice and the expansion framework of the knowledge system. The ability to independently resolve errors and the knowledge of how to access underlying documentation are core markers of graduating from the “beginner phase.”

9.1 High-Frequency Beginner Error Quick Reference and Troubleshooting Chains

In HFSS simulation, errors generally fall into three categories: geometry topology errors, physical definition conflicts, and computational resource exhaustion. Mastering structured troubleshooting chains can dramatically reduce trial-and-error time.

• Wave Port Location Error (Port Setup Error):
  • Error Signature: The message indicates that the Wave Port is not in contact with the Background, or that the port area is insufficient to support the specified number of modes.
  • Troubleshooting Chain: Check whether the Wave Port is completely flush with the outermost radiation boundary. If a Wave Port is forced to be used inside the model, check whether a Perfect E (ideal electric conductor) material has been manually drawn and assigned on the back of the port as a wave reflection truncation surface (Cap).
• Mesh Non-Convergence:
  • Error Signature: Adaptive meshing has reached the maximum number of passes, but Delta S ($|\Delta S|$) is still far above the convergence threshold.
  • Troubleshooting Chain:
    1. Check the physical model for singularities: such as infinitely sharp metal thin-plate edges, or extremely small (nanometer-scale) unintended gaps between solids. These regions cause the electric field strength to theoretically approach infinity, and the mesh can never converge.
    2. Return to the 3D Modeler for geometry cleanup (defeaturing), removing tiny chamfers, threads, or extremely fragmented solid surfaces (sliver faces) that have no impact on electromagnetic propagation.
    3. For high-Q resonant cavities, convergence is inherently very slow. Confirm whether this is caused by physical phenomena and consider increasing the minimum number of passes as appropriate.
• Out of Memory (OOM):
  • Error Signature: The solving process is interrupted, and the system reports insufficient memory to parse the current finite element matrix.
  • Troubleshooting Chain:
    1. Check whether the Solution Frequency is set too high. If the solution frequency is mistakenly set to an extremely high band, a massive number of tiny tetrahedral meshes will be generated.
    2. Assess the electrical size. If the DUT’s physical dimensions exceed tens of wavelengths (e.g., large reflector antennas, full-vehicle RCS), switch to an integral-equation-based MoM solver or SBR+ (ray-tracing) solver.
    3. Exploit electromagnetic symmetry: bisect the model along a symmetry plane to one-half or one-quarter, apply Perfect E / Perfect H boundaries, and directly reduce memory consumption to the corresponding fraction.

9.2 References and Theoretical Foundations

The HFSS interface is merely the surface; its core lies in computational electromagnetics and microwave network theory. The following literature forms the knowledge anchors for understanding the software’s underlying logic:

• Software Mechanism and Underlying Algorithm References:
  • “Ansys HFSS User’s Guide” and “Ansys Electronics Desktop Help”: Ansys official documentation. Provides the most rigorous mathematical explanations for every boundary condition, port definition, and solving engine.
  • “The Finite Element Method in Electromagnetics” (by Jian-Ming Jin): For in-depth understanding of 3D vector finite element methods, mathematical convergence of adaptive meshing, and the underlying matrix inversion logic of high-frequency computational electromagnetics.
• Microwave Network and Electromagnetic Theory Anchors:
  • “Microwave Engineering” (by David M. Pozar): Corresponds to HFSS’s multi-mode Wave Port theory, S/Y/Z parameter matrix derivations, characteristic impedance definitions, and Smith Chart matching network design—the theoretical cornerstone.
  • “Antenna Theory: Analysis and Design” (by Constantine A. Balanis): Corresponds to HFSS’s mathematical definitions for the far-field radiation region, the physical mechanisms of absorbing boundary conditions (ABC) and PML, and theoretical derivations of antenna gain and directivity.
• Practical Engineering References:
  • “Ansys HFSS Electromagnetic Simulation Design Application Detailed” (by Li Mingyang et al.): A widely used practical reference in China. Provides numerous localized engineering examples and standard operating procedures based on classic microwave devices (microstrip antennas, waveguide filters, power dividers).

9.3 Continuous Learning Path Recommendations

• Deep Utilization of Official Help Documentation (F1):
  • HFSS has an extremely detailed built-in documentation system. Pressing F1 in any settings window will jump directly to the underlying mathematical and physical mechanism explanation for that window. For unknown error codes, the first step should be to enter the code in the Help search engine for investigation.
• Reverse Engineering of Built-in Examples:
  • Via File > Open Examples, you can access a large number of standardized project files provided by the official source. Reverse-engineering these officially verified models (observing their material settings, boundary condition override priorities, and adaptive meshing settings) is the fastest path to mastering advanced engineering standards.
• Verification Simulation:
  • Before applying any complex or unfamiliar new software feature, always first build an extremely simple model with a known analytical solution (such as a standard rectangular waveguide or a 50-ohm coaxial line). By comparing the software simulation results with analytical formula calculations, you can verify whether your understanding of the software settings is correct.

Published on 2026-03-28 at 孤筝の温暖小家, last modified on 2026-03-28

All articles on this blog are licensed under the BY-NC-SA license agreement unless otherwise stated. Please indicate the source when reprinting!