Understanding Near-Field Measurements in Antenna Testing

Near-field antenna measurements capture electromagnetic field data in the immediate vicinity of the antenna under test (AUT), typically within a distance of a few wavelengths. Unlike far-field testing, which requires large anechoic chambers or outdoor ranges to simulate plane wave conditions, near-field techniques allow compact and controlled indoor setups. The measured complex field (amplitude and phase) is recorded over a known geometric surface—planar, cylindrical, or spherical—using a precision scanning system and a probe antenna. This raw data does not directly represent the antenna’s far-field radiation performance or its network parameters. Through rigorous mathematical transformations and probe correction algorithms, engineers can derive both the radiation pattern and the antenna’s scattering parameters (S parameters) with high accuracy.

Near-field measurement systems are indispensable when testing large phased arrays, reflector antennas, or modern 5G and satellite communication devices that are physically too large or too directional for far-field ranges. By moving the probe across a sampled grid, the system records the magnitude and phase of the tangential electric or magnetic fields. The underlying theory, grounded in electromagnetic equivalence principles, allows reconstruction of the field anywhere outside the measurement surface—most importantly, in the far-field region. Modern systems achieve dynamic ranges exceeding 60 dB, making them suitable for characterizing low-sidelobe and high-gain antennas where small errors in S parameters can significantly impact system performance.

Near-field scanning also provides the advantage of measuring the antenna in a controlled environment, reducing weather dependencies and interference from external sources. This repeatable setting is critical for production-level testing where consistency across units must be maintained with minimal deviation. The ability to extract network parameters directly from field data reduces the reliance on separate vector network analyzer (VNA) measurements, streamlining the test flow and lowering instrumentation costs.

Why S Parameters Matter for Antenna Characterization

S parameters describe the linear behavior of a device as seen at its ports. For antenna testing, S11 represents the input reflection coefficient, which directly relates to impedance matching and return loss. S21 (or S12) characterizes the transmission between two antennas or between an antenna and a probe, providing gain and coupling information. Extracting these parameters from near-field data eliminates the need for separate network analyzer measurements in some setups, streamlining the test process and reducing instrumentation overhead. This integration is especially valuable in production environments where every minute of test time counts.

When an antenna operates in a system, poor S11 means power is reflected back into the transmitter, reducing efficiency and risking damage. Accurate S parameter extraction during the prototype phase helps engineers optimize matching networks, verify bandwidth, and confirm that mutual coupling with adjacent elements (in arrays) stays within design limits. In near-field testing, S parameters are not measured directly at the antenna terminals with a VNA; they are derived from the transformed far-field data or from the modal coefficients obtained during the near-field to far-field transformation. This indirect approach demands careful calibration and probe characterization, but it yields results comparable to direct far-field or network analyzer measurements when proper procedures are followed.

The S-parameter matrix for an antenna is intimately connected to its radiation characteristics. For instance, the total radiated power computed from the far-field pattern must equal the net power entering the antenna port, minus ohmic and dielectric losses. By combining near-field derived far-field data with calibrated input power measurements, engineers can extract S11 magnitude and phase with confidence. Similarly, multi-port S parameters are derived by sequentially exciting each port and measuring the response at all others using the near-field probe as a calibrated receiver.

Preparing for Near-Field Data Acquisition

Equipment Calibration and Setup

The foundation of reliable S parameter extraction is a meticulously calibrated measurement system. A vector network analyzer (VNA) connects to both the AUT and the near-field probe, capturing magnitude and phase of the transmitted and reflected signals. Full two-port calibration, such as SOLT (Short-Open-Load-Through) or TRL (Through-Reflect-Line), must be performed at the probe tip and the AUT connector reference planes. Any drift, cable movement, or temperature variation during the scan can introduce phase errors that ripple into the final S parameter values. It is standard practice to perform a calibration check at the beginning and end of each measurement session to ensure stability.

Alignment of the scanning robot and the AUT is equally critical. Mechanical misalignments cause the probe to deviate from the intended sampling surface, distorting the measured field distribution. Laser trackers or optical alignment systems are used to map the actual scan geometry. The measurement chamber itself should be free of excessive reflections; absorber placement and time-domain gating techniques can further suppress residual echoes. For high-accuracy S parameter work, the chamber’s quiet zone reflection level should be at least 40 dB below the direct signal. The choice of VNA intermediate frequency bandwidth (IFBW) also affects measurement noise; a smaller IFBW improves dynamic range at the cost of longer sweep times. Balancing these parameters is essential for efficient data acquisition.

Probe Selection and Compensation

The near-field probe is itself a small antenna with a directional pattern. Its finite size and non-ideal response distort the measured field. Probe correction is essential to recover the actual field of the AUT. The probe’s receiving characteristics are characterized beforehand, typically using a known reference antenna or by performing multiple polarization measurements. Modern transformation algorithms incorporate probe pattern data to deconvolve the probe effect from the raw data, yielding the true tangential field components on the measurement surface. For S parameter extraction, the probe must also be characterized for its own input impedance and gain to compute the transmission coefficients correctly.

Probe selection depends on the frequency range, polarization, and the type of scan. Open-ended waveguide probes offer well-defined patterns and are common for planar scanning. For spherical near-field measurements, dual-polarized probes speed up data collection. High-gain probes improve signal-to-noise ratio but require more careful correction due to their larger aperture. In all cases, the probe must remain stable and linear across the entire dynamic range of the measurement. The probe radiation pattern is typically measured on a different range or obtained from electromagnetic simulation. Additionally, the probe’s polarization purity must be known to avoid cross-polarization errors that can degrade S21 accuracy in dual-polarized applications.

Defining the Measurement Grid

The spatial sampling interval on the scan surface is dictated by the Nyquist criterion: the maximum wavenumber of the radiated field determines the required sample spacing. For a planar scan of an antenna with a maximum far-field direction cosine extent, the recommended spacing is typically half-wavelength or less at the highest frequency of interest. Oversampling provides redundancy that aids in error detection and filtering. The scan area must be large enough to capture the significant field energy; truncation of the measurement plane leads to aliasing and errors in the far-field pattern and S parameters. A common rule of thumb is to extend the scan plane until the field amplitude drops at least 30 dB below the peak.

For cylindrical and spherical scans, angular sample spacing and elevation increments follow analogous rules. The total number of measurement points can reach hundreds of thousands, necessitating automated data acquisition and robust signal processing. Each additional measurement point increases acquisition time, so a trade-off must be made between spatial resolution and overall test duration. For S parameter extraction, the grid density directly impacts the accuracy of the radiated power integration, which feeds into the S11 computation. Finite scan area also truncates the evanescent fields close to the antenna surface; these fields carry stored energy that contributes to the input impedance. If the scan plane is too close to the AUT, the stored energy may be inadequately captured, leading to S11 errors. A distance of 3–5 wavelengths from the AUT is typical for planar scanning, though this depends on antenna size and frequency.

Near-Field to Far-Field Transformation Fundamentals

Planar, Cylindrical, and Spherical Scanning

The transformation method depends on the scan geometry. Planar near-field measurements use the plane wave spectrum (PWS) representation. By applying a two-dimensional Fourier transform to the measured tangential field components, the angular spectrum of plane waves is obtained. The far-field pattern is then directly computed from the spectrum for visible (propagating) directions. This technique is widely adopted for high-gain antennas because it naturally captures the forward beam with minimal computational effort. For S11 computation, the plane wave spectrum is integrated over all real angles to compute total radiated power.

Cylindrical near-field scanning is suitable for antennas with omnidirectional patterns in one plane, such as base station antennas. The field decomposes into cylindrical wave functions, and the far-field is recovered via mode expansion and integration. Spherical near-field measurements represent the most general case, offering full 3-D pattern and S parameter extraction for any antenna. The measured data is expanded in spherical vector wave functions, and the complex expansion coefficients become the basis for computing far-field and S parameters. International standards, such as the IEEE Recommended Practice for Near-Field Antenna Measurements, provide detailed guidance on these transformations (IEEE 149-2021).

Each scanning geometry has implications for the S-parameter extraction error budget. Planar scans are fast and computationally efficient but require a large scan area to capture wide-angle radiation, which affects low-elevation S21 accuracy. Cylindrical scans are good for antennas with a fixed beam in elevation, while spherical scans provide the most complete data but require longer acquisition times and more complex positioners. The selection of geometry should be based on the AUT’s radiation pattern and the required accuracy of the S parameters.

Mathematical Transformations: Fourier and Mode Expansion Methods

At the heart of planar transformation lies the Fast Fourier Transform (FFT). The measured tangential electric field components Ex and Ey are Fourier transformed to compute the plane wave spectrum A(kx,ky). For each spectral component, the far-field pattern is proportional to the spectrum evaluated at the wavenumbers corresponding to the visible region (kx2+ky2 < k02). Probe correction is applied either in the spectral domain as a deconvolution or by transforming the probe response into a weighting function. The post-processing must account for the finite scan area; windowing functions like Tukey may be applied to reduce truncation effects on S parameter accuracy.

Cylindrical and spherical transformations use Fourier series expansion in angular coordinates and Hankel or spherical Hankel functions in radial directions. The mode coefficients obtained are then used to compute transmitted and reflected power at the port, leading directly to S parameters. The mathematical relationship between the mode coefficients and the antenna’s scattering matrix is established by enforcing the boundary conditions at the measurement surface and the antenna terminals. In practice, these computations are performed by commercial software packages that handle the numerical details, but an understanding of the underlying math is essential for interpreting errors. For spherical near-field, the truncation of the mode index n is an important parameter; a rule of thumb is nmax = k0a + 10, where a is the radius of the minimum sphere enclosing the AUT. An insufficient number of modes leads to pattern aliasing and S11 errors at high angles.

Probe Correction Techniques

Probe correction can be performed in the spectral domain using the probe’s receiving coefficients. For planar scanning, the measured data is divided by the probe’s plane wave receiving spectrum, provided the probe pattern is known. This step is sensitive to noise, so regularisation filters like the Wiener filter may be applied. Alternatively, the probe can be modeled as a summation of elementary dipoles, and the unknown AUT coefficients are solved using an iterative least-squares approach. Accurate probe correction prevents systematic errors in S parameter magnitude and phase, especially for wide-angle radiation and elliptical polarization antennas. For S21 measurements, the probe correction must also account for the impedance mismatch between the probe and the receiver.

Modern software often includes built-in probe correction routines that accept a probe pattern file in standard formats. It is critical that the probe’s pattern data be measured at the same frequencies and with the same polarization definitions used in the AUT scan. For multi-port antennas, the probe correction must be applied consistently to the field data from each port excitation, ensuring that any cross-polarization coupling in the probe is de-embedded uniformly.

Step-by-Step Extraction of S Parameters

Computing Reflection Coefficients (S11)

The reflection coefficient S11 quantifies the impedance matching at the antenna’s input port. After the near-field to far-field transformation, the total radiated power is computed by integrating the far-field pattern over the full 4π steradian sphere. The input power is known from the incident wave amplitude at the reference plane. S11 magnitude is derived from the power balance: Pradiated = (1 - |S11|2) Pincident, assuming antenna losses are negligible or separately characterized. Phase of S11 is obtained from the complex input reflection data measured during calibration. Many modern software suites directly output S11 as part of the modal decomposition, giving both magnitude and phase. When measuring on a standard-gain horn, a direct VNA measurement can confirm the transformation-derived S11 within 0.1 dB.

For lossy antennas, the radiated power is less than (1 - |S11|2) Pincident due to ohmic losses in the conductor and dielectric. These losses can be estimated from the antenna’s efficiency, which itself can be derived from the near-field gain comparison or from separate calorimetric measurements. If the loss is significant (greater than 0.5 dB), the power balance equation must be corrected to avoid S11 errors. Alternatively, one can use the complex reflection coefficient obtained directly from the VNA during data acquisition, referencing the AUT port. This direct measurement is often more accurate for S11 than the derived value, but the near-field derived S11 provides a consistency check on the entire measurement chain.

Determining Transmission Coefficients (S21, S12)

Transmission S parameters between two ports—for instance, between two antenna elements in an array, or between a reference antenna and the AUT—can be extracted by comparing the far-field gain with the known gain of a standard horn. In a near-field system, the probe itself can serve as the second port if calibrated. By placing the probe at a known distance and measuring the transmission coefficient as a function of position, the far-field transmission factor is de-embedded. The final S21 is calculated by relating the received wave amplitude at the probe’s output port to the incident wave at the AUT’s input, incorporating all path losses and probe gains. Full S-parameter matrices for multi-port antennas are built by repeating the measurement for each port excitation using a switching matrix and correcting for cable phase differences.

In multi-port extraction, the isolated element pattern approach assumes that the presence of the probe does not disturb the excitation of the AUT ports. This assumption holds when the mutual coupling between the probe and the AUT is low, which is typically the case for open-ended waveguide probes with low aperture. For higher gain probes, multiple reflections may introduce coupling errors; time-domain gating can help isolate the direct transmission path. The extracted S21 for an array element can be validated by comparing the cumulative coupling from all elements to the far-field pattern of the full array, a technique known as active impedance characterization.

Post-Processing and Validation

After transformation, the extracted S parameters must be validated. One common check is to compare S11 measured directly with a VNA on the same setup (by breaking the scan connection) with the value obtained from the near‑field method. Agreement within 0.2 dB and 2 degrees of phase indicates a solid calibration. The far-field pattern can also be compared against compact range or anechoic chamber results. Data smoothing or windowing may be applied to reduce truncation ripples without distorting the true S parameter response. For multi-port measurements, the symmetry of the S-parameter matrix (Sij = Sji for reciprocal passive antennas) provides an additional validation check.

Validation must also include an assessment of measurement repeatability. Running the same scan multiple times and analyzing the standard deviation of the extracted S11 across the frequency band gives confidence intervals. A typical acceptable repeatability standard deviation is 0.05 dB in magnitude and 0.5 degrees in phase for frequencies below 40 GHz. For higher millimeter-wave frequencies, tighter mechanical tolerances and thermal stability become more critical, and repeatability may degrade. Documentation of all calibration factors, probe files, and transformation settings is necessary for traceability and for reproducing results months later.

Software Tools and Resources

Several commercial and open-source packages incorporate robust near‑field to far‑field algorithms with built-in probe correction and S parameter post-processing:

  • CST Studio Suite (Dassault Systèmes) – Offers a comprehensive near‑field import module and coupled EM‑circuit co‑simulation for S parameter extraction.
  • ANSYS HFSS – Widely used for antenna design, it includes near‑field data processing and S‑parameter computation with HFSS‑IE and hybrid solver capabilities (ANSYS HFSS).
  • Altair FEKO – Excels in large‑scale antenna placement and near‑field analysis; its POSTFEKO module handles transformation and modal decomposition.
  • NSI-MI Technologies – Provides dedicated near‑field measurement software like NSI2000, which drives scanners and performs real‑time transformation with integrated S parameter extraction.
  • MATLAB Antenna Toolbox – Custom scripts can be written for planar, cylindrical, and spherical transformations using built‑in Fourier and special function libraries (MATLAB Field Transformations).
  • Scikit-rf – An open-source Python library for RF/microwave engineering; useful for manipulating and comparing S-parameter files once data has been exported.
  • EMCoS Antenna VNA – A specialized software tool for near-field measurement data processing and full S-parameter matrix generation (EMCoS Antenna VNA).

When selecting a tool, consider the scan geometry, probe model complexity, and the level of automation required. Many test ranges develop proprietary software that integrates directly with their VNA and positioner controllers, providing a seamless workflow from raw data to final S parameters. Open-source options like Scikit-rf are valuable for custom post-processing but require more programming effort for the transformation core.

Best Practices for Accurate S Parameter Extraction

Implementing the following guidelines consistently reduces measurement uncertainty and improves repeatability:

  • Calibrate the VNA at the probe and AUT connectors. Use a mechanical calibration kit with known standards and verify with a verification device. Temperature changes during long scans call for periodic recalibration. Electronic calibration units can automate this process but may introduce drift over time. Consider performing a calibration every hour for scans lasting longer than 30 minutes.
  • Characterize the probe pattern thoroughly. A single probe file may not suffice for wide bandwidths. Use interpolation or multiple calibration files for different frequency bands. The probe’s port impedance must also be known for accurate S21 extraction. Additionally, measure the probe’s polarization axial ratio and orientation angle with respect to the scan coordinate system.
  • Verify scan area truncation limits. Measure the AUT’s near-field amplitude taper and ensure the scan plane extends at least 15 dB below the peak. Apply tapered windows only if necessary and account for their effect on S parameter values. For high-gain antennas, a larger scan area may be required to capture reactively stored energy. Use a conservative scan margin of at least 3 half-wavelengths beyond the -30 dB contour.
  • Maintain precise mechanical alignment. Misposition errors as small as 0.01λ can create significant phase errors. Routinely check the robot’s repeatability with a laser interferometer. For spherical scans, the center of rotation must coincide with the AUT’s phase center to within 0.02λ for accurate S parameter phase.
  • Acquire data with high dynamic range. Average multiple samples and use a low‑noise amplifier near the probe to improve signal-to-noise ratio, which is vital for weak field regions that affect far-field sidelobes and therefore S11 accuracy. A dynamic range of 60 dB or more is recommended for most applications. For arrays with deep nulls in the pattern, 70 dB may be necessary.
  • Apply probe correction symmetrically. For planar measurements, using symmetrical scan grids (e.g., odd number of points) avoids subtle spectrum shifting. The same principle applies to angular sampling in cylindrical and spherical scans. Also ensure that the probe correction algorithm accounts for the orientation of the probe during the scan (e.g., y-axis rotation for spherical scans).
  • Compare with a reference antenna. Validate the whole chain against a standard‑gain horn or dipole whose S parameters and gain are known from a calibrated lab. This step catches systematic errors in calibration, probe correction, or transformation. Perform this validation at multiple frequencies across the band of interest.
  • Document all measurement parameters. Record the scan geometry, step sizes, probe type, calibration dates, environmental conditions (temperature and humidity), and software version. This documentation supports root-cause analysis if deviations appear later in system-level simulations.

Common Challenges and How to Overcome Them

Probe-AUT Multiple Reflections. In some setups, signals bounce between the probe and the AUT, causing ripples in the near-field data. Time-domain gating in the VNA or use of absorbing materials around the probe can mitigate this. In post-processing, filtering in the spatial frequency domain can reduce the effect, though care must be taken not to eliminate legitimate fine structure. The use of a slightly offset probe trajectory can help identify and remove these artifacts. For arrays, these reflections may also couple into adjacent ports, corrupting S21 measurements.

Truncation and Aliasing Errors. Insufficient scan size truncates the plane wave spectrum, introducing ringing in the far-field pattern and S11 errors. The solution is to model the AUT beforehand to estimate its reactive field extent and increase the scan area accordingly. If physical limits exist, extrapolation algorithms can extend the data, but these must be validated against known results. Oversampling in the spatial domain also helps mitigate aliasing. For electrically large antennas, a rule of thumb is to make the scan plane at least 2.5 times the antenna’s diagonal dimension.

Probe Positioning Errors. Thermal drift, backlash, and gravity sag degrade the scan geometry. Regular laser alignment and closed‑loop position correction can reduce such errors to below λ/100. Post‑processing correction using known fiducial points on the AUT helps if the positioning system cannot meet the required tolerances. In spherical scans, a misaligned rotation axis introduces phase offsets that directly affect S parameter phase values. Perform periodic positioner characterization using a calibrated reference antenna with known pattern symmetry.

Multi‑Port and Array Measurements. For arrays with many ports, switching between ports introduces repeatability challenges. Automated switching matrices and full calibration at each port path are necessary. Computation time for multi‑port transformation can be high, so efficient algorithms and parallel processing are recommended. Mutual coupling between array elements can affect the measured S parameters; the scan grid and probe correction must account for the element patterns individually. In phased arrays, the active S-parameter (S11 under all ports excited) differs from the passive S11 of each element due to mutual coupling. Near-field extraction can provide active S parameters if all ports are excited simultaneously using a multi-beam network, but this requires more complex setup and calibration.

Frequency and Bandwidth Constraints. Near-field measurements are typically performed at single frequencies or across discrete steps. For wideband antennas, the scan parameters and probe correction must be valid across the entire bandwidth. Performing separate scans at multiple frequency points and interpolating S parameters may be inefficient. Alternatively, time-domain techniques can extract wideband S parameters from a single pulsed measurement, but this approach is less mature for near-field applications. Currently, most systems rely on stepped-frequency sweeps with the VNA, which can be time-consuming for hundreds of frequency points. Using a parallel multi-channel receiver can speed up the process.

Applications and Real-World Examples

The ability to extract S parameters from a single near‑field scan is particularly valuable during the development of 5G massive MIMO arrays. Each of the 64 or 128 antenna elements must be characterized for both radiation pattern and mutual coupling (Sxy). Near‑field scanning of the fully assembled array under controlled conditions accelerates design verification and reduces costly far‑field tests. In satellite communications, where reflectors can be tens of meters in diameter, near‑field techniques inside a moderate‑sized chamber allow full S‑parameter matrix extraction that feeds into link budget analysis. Automotive radar antennas operating at 77 GHz benefit from this approach because on‑chip integration makes direct probing difficult; near‑field scanning combined with over‑the‑air characterization yields accurate impedance and coupling data. Even in biomedical antenna design, where antennas are small and operated close to the human body, near‑field measurements in tissue‑simulating liquids allow S11 extraction that accounts for the lossy environment.

In each case, the extracted S parameters are exported as Touchstone files and used in system‑level simulations, directly linking physical measurements with circuit and array modeling tools. For example, a 5G base station array’s S-parameter matrix can be imported into a system simulator to predict beamforming performance and power amplifier loading. Aerospace and defense applications also use near-field S parameter extraction for phased-array calibration and to validate the mutual coupling coefficients that are critical for digital beamforming algorithms. The trend toward integrated active antenna systems (AAS) with embedded amplifiers and filters further motivates the need for combined pattern and S-parameter characterization in a single test session.

Conclusion

Extracting S parameters from near‑field antenna measurements transforms a traditional radiation-pattern test into a complete RF performance assessment. By combining precise positioning, meticulous VNA calibration, probe correction, and robust transformation algorithms, engineers can obtain S11, S21, and full multi‑port S‑parameter matrices without a separate network‑analyzer measurement. This integrated methodology reduces test time, cuts facility costs, and provides a deeper understanding of antenna behavior. Adhering to best practices in scan geometry, alignment, and validation ensures the results match or exceed the accuracy of direct far‑field measurements. As antenna systems grow more complex and integrated, the role of near‑field S parameter extraction will only continue to expand, driving innovation in wireless communications, radar, and satellite technology. The move toward over-the-air (OTA) testing of integrated devices reinforces the importance of this technique, enabling accurate characterization of active antennas in the same chamber used for passive performance evaluation. With ongoing advances in computational efficiency and probe technology, near-field S parameter extraction is becoming a standard tool in antenna metrology, offering a practical path to comprehensive RF characterization in a single, cost-effective measurement session.