PCB Antenna and Feed Resources

Antenna Simulation Workflow: HFSS & CST Setup to Export

A step-by-step antenna simulation workflow covering geometry import, port and boundary setup, meshing strategy, convergence criteria, result interpretation, and export to PCB fabrication in both HFSS and CST Studio Suite.

Microstrip patch antenna on high-frequency laminate with gold feed trace

What Does an Antenna Simulation Workflow Look Like?

An antenna simulation workflow is the repeatable sequence of steps that takes a radiating structure from a 3-D model in a full-wave solver—typically Ansys HFSS or Dassault CST Studio Suite—through excitation, meshing, convergence, post-processing, and finally export of geometry and stackup data to PCB fabrication. Getting the workflow right matters because a single mis-assigned port or an under-resolved mesh can shift a patch antenna’s resonant frequency by 100 MHz or more at 5.8 GHz, turning a first-spin success into a costly respin.

The sections below walk through each stage in order, flag the places where HFSS and CST diverge, and end with the export steps that hand a validated design to manufacturing.

Step 1: Geometry and Material Assignment

Exploded cross-section view of a multilayer antenna PCB stackup

Importing or Building the Model

Most projects start by importing a STEP or SAT solid model from a mechanical CAD tool, or by drawing the antenna directly inside the solver. For PCB-based antennas—patches, IFAs, slot antennas—drawing in the solver is usually faster because you can parameterize trace width, slot length, and feed offset from the start. If you are working with a full board model, the HFSS PCB import and solve workflow covers the EDB-to-HFSS path in detail.

Material Properties That Matter

Assign the substrate its frequency-dependent Dk and Df from the laminate datasheet, not a generic “FR4” library entry. For a 5.8 GHz patch on RO4350B, use Dk = 3.48 ± 0.05 and Df = 0.0037 at 10 GHz (the closest published test frequency). Copper layers should be modeled as finite-conductivity sheets (σ = 5.8 × 10⁷ S/m) rather than perfect-E boundaries once you care about efficiency within ±0.5 dB. When choosing a laminate grade, the RF PCB material selection guide explains how Dk tolerance and Df affect antenna performance.

Step 2: Ports and Boundary Conditions

Excitation Types

The port you assign determines the impedance reference and the fields the solver seeds at the feed point. The two most common choices for antenna simulation are:

  • Wave port (HFSS) / Waveguide port (CST) — a 2-D eigenmode solution on a cross-section. Best for microstrip-fed or coplanar-waveguide-fed antennas where you want a well-defined characteristic impedance. The port rectangle should be at least 6–10 times the substrate height wide and 6–10 times high to capture fringing fields.
  • Lumped port (HFSS) / Discrete port (CST) — a voltage or current gap between two conductors. Use it for probe-fed patches or differential feeds. Lumped ports are simpler to set up but do not de-embed transmission-line length, so place them right at the antenna reference plane.

For more on port sizing, de-embedding, and common mistakes, see the EM simulation setup guide on ports, boundaries, and meshing.

Radiation Boundary / Open Space

In HFSS, a radiation (absorbing) boundary must sit at least λ/4 from the nearest radiating edge—at 2.45 GHz that is roughly 30 mm. In CST’s time-domain solver, the default open-space boundary with perfectly matched layers (PML) handles this automatically, which is one reason CST’s transient solver is popular for broadband antenna sweeps. The CST Studio PCB workflow covers solver-type trade-offs in more detail.

Step 3: Meshing Strategy

Meshing is where simulation accuracy and run time are won or lost. Both solvers use adaptive meshing, but the starting point matters.

Parameter HFSS (FEM) CST Time-Domain (FIT)
Element type Tetrahedral Hexahedral
Initial mesh target λ/6 at solution frequency λ/10 default, refine near edges
Adaptive passes (typical) 8–15 for a patch antenna Not applicable; broadband in one run
Key manual refinement Seed mesh on feed gap and slot edges Local mesh density on thin substrates (< 0.5 mm)
Memory for a single-patch model 2–6 GB 1–4 GB

A practical rule: if doubling the mesh density changes S₁₁ by less than 0.5 dB and shifts resonance by less than 0.5 %, the mesh is adequate. Run this check at least once before trusting any gain or efficiency number.

Step 4: Convergence and Frequency Sweep

In HFSS, set a convergence criterion of ΔS ≤ 0.02 (the maximum change in S-parameter magnitude between adaptive passes). Most single-element antennas converge in 10–12 passes. If the solver stalls at ΔS ≈ 0.03, the radiation boundary is too close or a port is under-sized.

CST’s time-domain solver runs a single broadband excitation (Gaussian pulse) and extracts S-parameters via Fourier transform. Convergence here means the time-domain energy has decayed to –30 dB or better. For high-Q antennas—narrow-band patches, structures near the fundamental Q limit—the simulation time can be long; increase the solver’s maximum simulation time to at least 20 pulse widths.

Step 5: Post-Processing and Result Validation

3-D radiation pattern visualization over an antenna PCB ground plane

Key Outputs to Check

  1. S₁₁ (return loss): Confirm the –10 dB bandwidth covers your specification. A 2.4 GHz ISM patch on 1.524 mm RO4350B typically shows 80–100 MHz of –10 dB bandwidth.
  2. Input impedance (Smith chart): The loop should pass through or near 50 Ω at center frequency. A persistent reactive offset means the feed point needs adjustment. Understanding why measured 50 Ω is never exactly 50 Ω saves time in tuning.
  3. 3-D radiation pattern and gain: Look at both co-pol and cross-pol. For a linearly polarized patch, cross-pol should be at least 15–20 dB below co-pol in the main beam.
  4. Efficiency: Total efficiency accounts for mismatch and ohmic loss. Values below 80 % on a low-loss laminate warrant investigation—usually a meshing artifact or an incorrect loss tangent.

Exporting Your Results as PDF

Once the simulation is validated, archive the key plots—S-parameters, Smith chart, gain pattern, and efficiency—into a single PDF for design review or customer hand-off. Enter your project details below to generate a formatted results summary.

Loading Estimator Results PDF Export…

A typical export should include the frequency sweep range, convergence status (ΔS value or energy decay level), peak realized gain, and the –10 dB bandwidth. Attach the PDF to your version-controlled project folder alongside the solver project file so that any future respin starts from a known-good baseline.

Step 6: Stackup Definition and Fabrication Export

Translating the Simulation Stackup to a Fab Stackup

The stackup you simulated—substrate material, thickness, copper weight, prepreg layers—must be communicated to the fabricator exactly. Discrepancies here are the single largest source of sim-to-measurement mismatch. A 0.05 mm error in dielectric thickness shifts a 5.8 GHz patch resonance by roughly 50 MHz.

Document every layer: material grade (e.g., RO4350B core 0.508 mm, RO4450F prepreg 0.1 mm), copper weight (typically ½ oz for antenna layers, 1 oz for ground), and finished plating (ENIG or immersion silver). Use the tool below to generate a stackup PDF that your fabricator can quote against directly.

Loading Stackup PDF Export…

After exporting, compare the PDF layer-by-layer against the solver’s material assignments. Confirm that the Dk value listed matches the datasheet value at your operating frequency, not the “process” Dk the laminate vendor sometimes quotes for a different test method. This single cross-check prevents more first-article failures than any other step in the workflow.

Geometry Export: ODB++ or Gerber

Export the antenna metallization as Gerber X2 (preferred) or ODB++. If the antenna geometry was drawn inside the solver rather than imported from a PCB tool, export a DXF from HFSS or CST and import it into your layout tool (Altium, KiCad, Cadence Allegro) to generate manufacturing files. Maintain a minimum of 0.05 mm (2 mil) positional accuracy on the antenna outline and feed structures. The antenna prototype-to-production guide covers tolerance budgets and yield considerations for the fabrication stage.

Common Pitfalls and How to Avoid Them

  • Using a generic Dk: “FR4, Dk = 4.4” is meaningless above 1 GHz. Always use the laminate vendor’s frequency-dependent data. Rogers’ RO4350B datasheet is a good model for how to read and apply these numbers.
  • Radiation boundary too close: Less than λ/4 clearance introduces artificial reflections that distort the pattern and shift S₁₁.
  • Ignoring surface roughness: At 24 GHz and above, copper roughness (typically 1–3 µm RMS on standard HTE foil) can reduce antenna efficiency by 1–2 dB. Enable the Huray or Hammerstad roughness model in your solver.
  • Skipping the mesh convergence check: A pretty radiation pattern from a coarse mesh is not a valid result. Always verify with a mesh-density sweep.
  • Not de-embedding the port: If the wave port is placed 5 mm back from the antenna edge, de-embed that length or the phase of S₁₁ will be wrong and impedance matching in the layout will drift.

HFSS vs CST: Which to Choose for Antenna Simulation

Criterion HFSS (Frequency Domain / FEM) CST (Time Domain / FIT)
Best for Narrowband, resonant antennas; high-Q structures Broadband antennas; UWB; multi-band sweeps
Speed on a single-frequency patch Fast (one frequency, adaptive) Comparable (broadband in one run)
Array handling Domain decomposition + finite array DDM Time-domain + unit cell for infinite arrays
Learning curve Steeper port/boundary setup Easier defaults for open-space problems
Fabrication export ecosystem Tight integration with Ansys EDB / SIwave Export via DXF/STEP; less direct PCB-tool link

Neither solver is universally better. Many teams run both: CST for initial broadband exploration, HFSS for final narrowband optimization and correlation to measurement. The RF simulation workflow overview explains how schematic-level and EM co-simulation fit together across tools.

FAQ

How long does a typical antenna simulation take?

A single-element patch antenna at 5.8 GHz on a modern workstation (32 GB RAM, 8-core CPU) converges in 5–15 minutes in HFSS or CST. A 16-element array at 28 GHz can take 2–8 hours depending on symmetry exploitation and solver settings.

Can I simulate an antenna on FR-4?

Yes, but use measured Dk/Df data at your operating frequency. Standard FR-4 has Dk ≈ 4.2–4.5 and Df ≈ 0.02 at 2.4 GHz, which limits efficiency to roughly 60–75 % for a patch antenna. For prototyping below 3 GHz it is acceptable; above that, move to a controlled-Dk laminate.

What is the minimum mesh density for accurate gain results?

Start at λ/6 (HFSS) or λ/10 (CST) and refine adaptively. Gain results typically stabilize within ±0.3 dBi once the mesh passes a convergence check where doubling density changes gain by less than 0.2 dBi.

How do I validate simulation against measurement?

Compare S₁₁ center frequency (expect ±1–3 % shift due to fabrication tolerance), –10 dB bandwidth, and peak gain. If the frequency shift exceeds 3 %, re-examine the Dk value and substrate thickness used in simulation versus the fabricated stackup.

Do I need a separate license for antenna pattern export?

Both HFSS and CST include far-field pattern export (gain, directivity, axial ratio) in their base electromagnetic solver licenses. Exporting embedded element patterns for array beamforming may require an additional post-processing module depending on the version. Check the CST Studio Suite product page or Ansys licensing documentation for current module breakdowns.

Ready for a quotation?

Send your requirements for a material, stackup and manufacturing review. Price and schedule are confirmed for your project.

Upload Gerber