Choose EMI simulation software by the decision you need to make. A circuit model can help explain which source drives a disturbance; a board field solver can show where current flows or couples; a system model can test whether a cable or enclosure changes radiation. None of these replaces an emissions test on the actual product.
Use this guide when a PCB design is still editable and you need to choose what to model before committing to another prototype. The running example is a hypothetical controller with a fast clock and an external coax connection. The question is whether to revise the board return path or investigate current on the outside of the cable shield first.
EMI simulation software
Start with the simplest model that includes the relevant source, coupling path and observation point. The useful output may be a port current, a relative field map or a predicted field at a specified location. Choose that output before picking a tool: an attractive field image is not a substitute for the quantity your decision requires.
| Model type | Decision it supports | What it may miss |
| Circuit model | Compare source spectra, termination and damping. | Distributed geometry or radiation unless represented explicitly. |
| Board field solver | Locate return-current detours and compare layout changes. | Cables or chassis features omitted from the model. |
| 3D assembly model | Compare coupling through connectors, cable and enclosure. | Operating modes or test arrangements not included in the model. |
For the controller, first determine which clock harmonics need investigation. Ansys documents both board-level network extraction and near-field analysis in SIwave. These are distinct outputs, not interchangeable labels for an emissions prediction.
PCB EMI simulation
Define the source, return and measurable quantity in the same setup. Import the actual board revision and stack-up, then verify reference planes, plane cuts, vias and connector bonds. A model with a missing plane opening can be numerically stable and still represent a different board.
Set each excitation between the relevant conductors rather than connecting everything to a single ideal ground. Use a measured or documented source spectrum when amplitude matters. If a unit current is used to compare paths, label the result as a transfer comparison. It cannot be interpreted as the absolute noise generated by the operating controller.
PCB field solver
Use a board-focused solver for a predominantly planar path; use a 3D model when the return path runs through connector contacts, shielding or enclosure features not represented by the board model. Start with the suspected region and expand it when the result changes with the boundary or cutout.
For the controller, preserve the copper between the clock source and the connector, including the connector ground contacts and local return path. Compare the as-built board with a version where that return remains continuous. If the two cases give nearly the same connector current, don’t assume the detour is irrelevant; first check whether the excitation and the connector model are capable of revealing the difference.
Mesh convergence study
Refine the mesh until the quantity used for the decision stops changing materially between passes. A stable overall scattering-parameter result is helpful, but a local current at the connector can still be sensitive to how a narrow gap, via or port is modeled. Track the chosen observable as well as the solver’s convergence flag.
Suppose the project uses a 1 dB refinement tolerance for connector-current magnitude at a suspected harmonic. If successive passes change the magnitude by 0.4 dB but move the peak to a different part of the band, the model is not yet adequate for choosing a frequency-specific fix. The 1 dB is a teaching assumption, not a compliance limit. Recheck the sweep, refinement frequencies and port before proceeding.
Near field simulation
Use near-field simulation to compare where energy is concentrated close to the board, not to infer a chamber reading. Record the field component, sampling plane, distance from the board and frequency. Compare maps using the same color scale; automatically normalized pictures can make two very different amplitudes look identical.
On the first prototype, use repeatable probe height, orientation and fixture position to look for the predicted hotspot. A magnetic loop responds to field components according to its orientation. A relative hotspot check is useful even when the instrument or model is not calibrated for absolute field strength. The near-field probe guide covers the bench technique in more detail.
Common mode current in coaxial cable
Current on the exterior of a coax shield can take a different return path from the intended center-conductor and inner-shield current. Include the connector bond and the external path when the predicted board hotspot does not explain why emissions change as the cable is moved.
Compare two controlled configurations: same board and operating mode, but different documented cable placement or shield bond. Measure the outside-current proxy with a suitable RF current probe around the entire cable. If the cable-current peak shifts while the nearby board hotspot stays fixed, investigate cable and chassis coupling before widening traces or replacing the clock.
Radiated emissions simulation
A radiated-emissions model needs a defined assembly, source level, observation distance and orientation. Keep track of what geometry and modes it excludes. A board-only model may be adequate to rank layout options; an absolute prediction for the final assembly needs a model of the relevant return and radiating structures.
| Comparison | Keep fixed | What to do with disagreement |
| Modeled versus measured peak frequency | Clock mode, workload and frequency axis | Check source spectrum and resonance geometry. |
| Modeled versus measured hotspot | Scan height, location and field component | Check probe orientation, local mesh and return path. |
| Modeled versus measured cable trend | Cable termination and repeatable routing | Review shield bond and omitted chassis paths. |
EMC pre compliance testing
Use a physical pre-compliance scan to test the model’s predictions. Save baseline and changed configurations, record operating modes and fixture geometry, and compare the same frequencies and observables. Set aside any model whose claimed improvement does not survive a repeated measurement.
For the controller, make the smallest change that separates the two hypotheses: improve the board return with the cable unchanged, then repeat with only the cable arrangement changed. If the result follows one change, update the model accordingly. Close the design review when the predicted direction is reproduced and the remaining margins are understood; formal conformity still depends on the applicable test procedure and actual assembly.
FAQ
Can a schematic-only simulator find the radiating cable?
Not unless the cable and its coupling are represented in the model. A schematic can establish the likely noise source and spectrum, but two circuits with the same connectivity can radiate differently if the geometry, return path or shield termination changes.
Do I need the complete PCB database to start?
A representative region can be enough to evaluate a local hypothesis. Document the model cutout and terminations, then enlarge it if predicted currents depend on a boundary or if the return path leaves the cropped region.
Can I use one harmonic to validate the whole model?
No. Agreement at one frequency can be coincidental. Compare additional bands and changes in geometry or excitation; a model that reproduces those trends is more useful for deciding which design change to make.
Is a stronger near-field hotspot always worse?
No. Local concentration is not the same as efficient coupling into a radiator. Check whether the hotspot couples into a cable, enclosure feature or another structure that can carry or radiate the energy.
What should I export from the PCB layout?
Use a format supported by the chosen solver, then check the imported stack-up, copper geometry, via spans, plane boundaries and net mapping. Save the source revision and the model’s simplifications so the comparison can be reproduced.
How should I model an uncertain connector bond?
Compare plausible cases rather than assigning a precise value without evidence. If your conclusion reverses across those cases, measure or characterize the actual connection before treating the simulation as a design decision.
Is the best solver the one with the most detailed 3D model?
Not necessarily. A higher-detail model is helpful if those details control the observable. A simpler model with verified ports and geometry can be more useful than a complete-looking assembly with unverified excitation and material inputs.
What is the minimum evidence for a recommended layout change?
Retain the baseline, one controlled layout change, the convergence check and a matching prototype observation. State which frequencies and configurations the result covers. Do not generalize the finding to another board or enclosure without checking the relevant coupling path.
References
- Ansys SIwave analysis types
- Ansys HFSS adaptive meshing
- Keysight conducted and radiated emissions measurements
Discuss the PCB build
For a Best Technology PCB fabrication review, send the layout revision, intended stack-up and critical connector or return-path geometry to sales@bestpcbs.com.
You may also like
Tags: board level emi shielding, EMC, emc test, pcb design