Ham Radio Antenna Model Verification: NEC2 to Real World
A NEC2 antenna model is only as useful as the confidence you have in its accuracy. Model verification — systematically comparing predicted performance against real measurements — closes the loop between simulation and reality. This guide shows you exactly how to compare NEC2 output against NanoVNA measurements, identify the sources of discrepancy, update your model to match reality, and know when the model is trustworthy enough to guide design decisions.
Antenna modelling software is not a truth machine — it is a sophisticated mathematical approximation of reality that makes simplifying assumptions about conductor geometry, ground properties, nearby structures, and material constants. When a NEC2 model says your dipole resonates at 14.200 MHz with a feed point impedance of 67 + j2 Ω and you measure 14.050 MHz and 58 − j18 Ω on the NanoVNA, something in the model does not match reality. Finding and correcting that discrepancy transforms the model from an approximation into a validated tool that you can trust for design decisions.
Verification also builds fundamental understanding of what the modelling software is actually computing versus what the real antenna is doing. Every discrepancy you track down and explain teaches you something about the physics of your antenna and the limitations of the modelling approach — knowledge that transfers to every future design project.
What verification confirms
Resonant frequency prediction accuracy, feed point impedance accuracy, SWR bandwidth accuracy, and whether the model's structural representation correctly captures the real antenna geometry. A verified model earns the right to be trusted for gain and pattern predictions.
What verification cannot confirm
Absolute gain in dBi (requires calibrated reference antenna), radiation pattern shape (requires a calibrated outdoor range or far-field measurement facility), or performance advantages smaller than the measurement uncertainty of the tools available.
The verification hierarchy
Resonant frequency is easiest to verify (NanoVNA, ±1%). Feed point impedance is next (NanoVNA, ±5%). SWR bandwidth is straightforward (±10%). Gain requires comparative testing (RBN, WSPR, or calibrated field strength meter). Pattern shape is hardest and rarely verified by amateurs.
Before comparing model to measurement, verify that the model accurately represents the physical antenna. This step is often skipped but is the source of the majority of model-measurement discrepancies — not flaws in the NEC2 code, but differences between the antenna that was planned and the antenna that was actually built.
Do not assume that the antenna was cut to the planned length. Measure each arm of a dipole from the feed point insulator to the tip, each element of a Yagi from the boom centreline to the tip (× 2 for total length). Record measurements to the nearest 5 mm. Include any wire folded back at insulators in the electrical length. If the wire is insulated, note the insulation type — it affects the velocity factor.
Ground proximity significantly affects feed point impedance and resonant frequency — 10 cm of height error on a 40 m dipole at 10 m height causes measurable impedance change. Measure actual height at the feed point and at the element tips (which may be lower for an inverted-V). Use a tape measure along the support rope if direct measurement is impractical, correcting for rope angle with basic trigonometry.
Ground conductivity and permittivity directly affect the antenna's performance over real ground. Identify your ground type from the standard tables (agricultural soil, suburban clay, sandy soil, etc.) and use the corresponding σ and εr values in your NEC2 model. If you are unsure, try several values and compare the resulting resonant frequency and impedance against measurement — the best-matching ground parameters are likely close to your actual site.
Buildings, fences, downpipes, other antennas, and large trees all interact with the antenna's near field and alter the measured impedance and resonant frequency. Structures within approximately 0.2λ of the antenna have the most significant effect. Catalogue everything within this radius — even if you cannot include all structures in the NEC2 model, knowing they are there explains discrepancies that might otherwise be mysterious.
The most direct and reliable verification measurement is a VNA sweep at the antenna feed point — connected directly to the antenna terminals before any feedline, balun, or matching network. This gives the true feed point impedance as a function of frequency, which can be compared directly against the NEC2 model output.
Perform the SOLT (Short-Open-Load-Through) calibration at the end of a short reference cable that you will connect at the antenna feed point. The calibration plane must be at the feed point terminals — not at the shack end of the feedline. This eliminates the effect of the reference cable and any adapters from the measurements. A 300–500 mm length of RG-58 with a connector at each end is ideal as the reference cable for this purpose.
Connect the calibrated NanoVNA directly to the antenna feed terminals — removing any existing feedline, balun, or connector that would normally be between the antenna and the measurement point. Sweep across a frequency range spanning approximately ±20% either side of the expected resonant frequency. This provides enough data to see the full impedance trajectory on the Smith chart and the complete SWR curve through resonance.
From the NanoVNA display, record: the resonant frequency (where reactance X = 0 or nearest to zero), the resistance R at resonance, the SWR at resonance, and the SWR 2:1 bandwidth (frequencies where SWR first exceeds 2.0:1 above and below resonance). Export the Smith chart data as a screenshot or CSV if the firmware supports it — this provides a comprehensive dataset for model comparison.
After the feed point measurement, reconnect the normal feedline and measure again at the shack end. This gives you a second dataset showing how the feedline transforms the feed point impedance. The difference between these two measurements characterises the feedline's electrical length and loss. A validated model of the feedline (correct length and velocity factor entered) should correctly predict the shack-end impedance from the feed-point measurement.
NEC2 vs. NanoVNA Comparison Tool
Enter predicted (model) and measured values to quantify the discrepancy and identify likely causes.
Resonant frequency lower than predicted
The most common discrepancy. The measured antenna resonates at a lower frequency than the NEC2 model predicts — meaning the antenna is electrically longer than the model assumes. The leading causes are:
- Insulated wire not accounted for: Insulated wire has a velocity factor of 0.94–0.97 rather than 1.0 for bare wire in air. If the model uses bare wire and the antenna uses insulated wire, the model overestimates the resonant frequency by 3–6%. Fix: use the correct K-factor in element length calculations
- Actual height lower than modelled: Lower height above ground increases the effect of ground proximity, which shifts resonance downward. Measure actual height and update the model
- Extra wire at end insulators: Wire folded back at end insulators adds electrically significant length. A 50 mm fold-back adds approximately 100 mm to the effective element length
- Nearby objects: Metal guttering, fencing, or building framework within 0.2λ electrically loads the antenna, lowering its resonant frequency
Feed point resistance significantly different from predicted
The measured R at resonance differs from the model prediction by more than 10 Ω. Causes include:
- Height error: Feed point resistance is sensitive to height above ground. A dipole at 8 m presents different R than a dipole at 10 m — update the model height to the actual measured value
- Ground conductivity mismatch: The model's ground parameters (σ and εr) may not match the actual site. Try several ground types and find the one that best matches both the resonant frequency and the R measurement simultaneously
- Unmodelled loss: Real antenna conductors, connectors, and baluns add small resistances not included in perfect-conductor NEC2 models. Add an LD (distributed load) card with the conductor resistivity or add a small series resistance at the source segment
SWR bandwidth narrower than predicted
The measured 2:1 SWR bandwidth is significantly narrower than the model predicts. This indicates the antenna's effective Q is higher than the model assumes — usually because the element diameter in the model is larger than the actual element, or because the model does not include resistive losses that would broaden the bandwidth by reducing Q. Check that the wire or tubing diameter in the model exactly matches the physical element.
SWR bandwidth wider than predicted
Wider-than-predicted bandwidth usually indicates unmodelled losses — the antenna is less efficient than the model assumes, and the efficiency losses are broadening the apparent bandwidth. This is the bandwidth-efficiency trade-off: a lossy antenna has lower Q and wider SWR bandwidth. Measure the actual feed point resistance — if it is higher than predicted, the excess resistance is loss resistance, not radiation resistance, and the antenna is less efficient than the model predicts.
Updating the Model to Match MeasurementsOnce you have identified the sources of discrepancy, update the model systematically — one change at a time — and re-run the simulation after each change to verify that it moves the model in the right direction. Making multiple simultaneous changes makes it impossible to know which change caused which improvement.
| Discrepancy | Model parameter to adjust | Direction of adjustment | Expected effect |
|---|---|---|---|
| Resonance too low | Element length | Shorten elements by frequency-ratio scaling | Resonance shifts up |
| Resonance too high | Element length or wire K-factor | Lengthen elements or reduce K from 0.98 to 0.95 | Resonance shifts down |
| R too low | Antenna height above ground | Reduce modelled height to match actual | R decreases, resonance shifts |
| R too high | Ground conductivity (σ) | Increase σ (better soil) or add ground radials | Ground losses reduce, R drops |
| X too capacitive | Element length | Increase element length in model | Resonance shifts down, X less negative |
| X too inductive | Element length | Decrease element length in model | Resonance shifts up, X less positive |
| BW too narrow | Wire diameter | Decrease element diameter in model | Q increases, BW narrows further — reconsider |
| BW too wide | Loss resistance (LD card) | Add series resistance to source segment | Efficiency and Q reduce, BW widens |
Gain verification is fundamentally harder than impedance verification because you cannot measure absolute gain with a NanoVNA. Absolute gain measurement requires a calibrated reference antenna, a controlled measurement environment, and a way to separate path loss from antenna gain. Most amateurs do not have access to these resources. The practical alternatives are:
Comparative RBN testing
Use the Reverse Beacon Network to compare your antenna against a known reference antenna (typically a half-wave dipole at a known height) on the same band. If the model predicts 3 dBd gain for your Yagi over a dipole at equal heights, and RBN data consistently shows your Yagi giving 3 dB stronger spots than a neighbour's dipole on the same paths, the gain prediction is confirmed. This is not absolute gain measurement but it validates the relative gain prediction, which is what matters for most design decisions.
Pattern verification by RBN azimuth comparison
Point a rotatable beam in different directions and compare RBN spot strengths from spotters in each direction. A 3-element Yagi model predicting 7 dBd forward gain and 20 dB F/B should show approximately 7 dB better spots from forward-direction spotters and 20 dB worse spots from directly behind the beam compared to broadside spots from a reference dipole. Systematic azimuth-versus-RBN-spot plots can reveal whether the model's pattern prediction is approximately correct.
WSPR signal strength comparison
WSPR reports include SNR from each receiving station. Running WSPR for 24 hours on both the antenna under test and a reference antenna (in separate sessions) and comparing the median SNR received by common stations in each direction gives quantitative gain comparison data. The advantage over RBN is that WSPR produces many more data points (one every 2 minutes continuously) with less operator involvement.
Validation Thresholds — When Is the Model Good Enough?| Parameter | Excellent agreement | Good agreement | Investigate further | Model unreliable |
|---|---|---|---|---|
| Resonant frequency | <0.5% | 0.5–2% | 2–5% | >5% |
| Feed point R | <5 Ω | 5–10 Ω | 10–20 Ω | >20 Ω |
| Feed point X at f₀ | <5 Ω | 5–15 Ω | 15–30 Ω | >30 Ω |
| SWR 2:1 bandwidth | <5% | 5–15% | 15–30% | >30% |
| Gain (RBN comparative) | <0.5 dB | 0.5–1.5 dB | 1.5–3 dB | >3 dB |
| F/B ratio | <2 dB | 2–5 dB | 5–10 dB | >10 dB |
Every antenna you verify creates a validated reference model that you can reuse as the foundation for future designs. A library of verified models for your common antenna types — a dipole at your site, an inverted-V with your specific support geometry, your vertical with its actual radial system — provides a set of validated starting points that dramatically accelerates the next design iteration.
Document each verified model with: the physical dimensions as built, the NEC2 model file, the VNA measurement data, the ground parameters used, the date of verification, and any notes about nearby objects or environmental factors that affect the model's applicability. Store these alongside screenshots of the Smith chart and SWR sweep for the verified model. This documentation transforms a one-time verification exercise into a permanent asset for your station development.
Frequently Asked QuestionsMy model matches well for frequency but not for impedance — which should I fix first?
Fix frequency first. The impedance at resonance depends on the resonant frequency being correct — if the model resonates at the wrong frequency, the impedance comparison is being made at the wrong point on the impedance curve. Update element lengths until the model resonant frequency matches the measured frequency, then investigate remaining impedance discrepancies.
How do I model a wire with a velocity factor in NEC2?
NEC2 does not directly model insulated wire velocity factor — it assumes all wires are in free space (VF = 1.0 for bare wire in air). The practical approach is to shorten the modelled wire length by the appropriate factor: for insulated wire with VF = 0.96, use 0.96 × the physical wire length as the model element length. This gives the correct electrical length even though the insulation is not explicitly represented.
Should I verify the model before or after optimising it?
Build a baseline model first, then verify it against the built antenna. If the model matches well, you have a valid foundation for further optimisation. If it does not match, update the model to match reality before optimising — optimising an unvalidated model optimises mathematical abstractions rather than the real antenna. The correct sequence is: build → model → verify → update model → optimise model → modify antenna → re-verify.
My model is perfect on the workbench but wrong when installed at height — why?
Ground proximity is the primary cause of height-dependent discrepancies. A dipole on a bench 1 m above the floor is in a very different electromagnetic environment from the same dipole at 10 m above ground outdoors. Always model with the actual installation height and the correct ground parameters. The bench model is useful for checking element geometry but should not be compared to in-situ measurements without updating the height and ground parameters.
How accurate is NEC2 for antenna modelling in general?
For thin wire antennas (diameter much smaller than wavelength) above realistic ground, NEC2 predicts resonant frequency within 1–3%, feed point impedance within 5–15%, and gain within 0.5–2 dB of measured values for antennas in uncluttered environments. Accuracy degrades significantly for thick elements near junctions, very low antenna heights, buried conductors, and antennas near complex structures. Within its valid domain, NEC2 is a highly reliable tool for practical antenna design.
Is a verified model still valid if I change the antenna later?
A verified model applies to the specific physical configuration that was measured. If you change element lengths, add or remove radials, change the antenna height, or modify nearby structures, the model must be updated to reflect those changes and re-verified. Fortunately, a well-validated model can be updated incrementally — if you raise the antenna by 2 m, updating the height in the model and computing the new predicted impedance is likely to be accurate without a full re-verification, because the basic model structure is already validated.