Definition and purpose

A network analyzer is an instrument used to characterize the behavior of primarily linear electrical networks, with extensions for nonlinear analysis, in the frequency domain across radio frequency (RF) and microwave ranges, typically spanning from kilohertz (kHz) to terahertz (THz) frequencies.[5][6][7]

The primary purpose of a network analyzer is to measure key parameters such as scattering parameters (S-parameters), impedance, admittance, insertion loss, and gain, enabling engineers to evaluate device performance, ensure quality, and optimize impedance matching in applications like antennas, filters, amplifiers, and transmission lines.[5][7][8]

In basic operation, the instrument generates a swept-frequency signal that is applied to the device under test, then analyzes the reflected and transmitted responses to determine both magnitude and phase characteristics as a function of frequency.[9][10]

Vector network analyzers remain the most prevalent type in modern use.[5]

Types of network analyzers

Network analyzers are categorized primarily by their measurement capabilities, such as whether they capture magnitude only or both magnitude and phase, as well as by their handling of signal power levels and specialized applications.[9] The main types include scalar, vector, and large-signal variants, each suited to different testing needs in RF and microwave engineering.[11][8]

Scalar network analyzers (SNAs) measure only the magnitude, or amplitude, of network parameters, providing a cost-effective option for basic assessments without phase information.[12] They are ideal for transmission and reflection tests, such as evaluating insertion loss, return loss, gain, and standing wave ratio (SWR) in linear devices like filters and cables.[9] Swept-tuned SNAs emerged in the 1960s and 1970s as commercial instruments for these magnitude-based measurements.[13]

Vector network analyzers (VNAs) measure both magnitude and phase to determine complex scattering parameters (S-parameters), enabling full characterization of linear network behavior.[11] VNAs, which became the dominant type following their commercial introduction in the mid-1960s, support multiport configurations from 2 to 48 ports for testing complex systems like antennas and multi-antenna arrays.[14][7]

Large-signal network analyzers (LSNAs) extend VNA capabilities to nonlinear regimes under high-power conditions, measuring X-parameters that describe amplitude-dependent behaviors and harmonics. They are essential for analyzing devices like power amplifiers and mixers where traditional S-parameters fail due to nonlinearity.[15]

Other specialized types include time-domain network analyzers, which use time-domain reflectometry (TDR) to assess signal integrity in high-speed interconnects by transforming frequency-domain data.[16] Pulsed RF analyzers support radar and pulsed system testing by capturing transient responses, while millimeter-wave analyzers extend operations up to 1.1 THz for 5G, 6G, and terahertz applications.[17][7]

Early instruments

The origins of network analyzers in electrical engineering emerged in the mid-20th century with the development of slotted line instruments designed for precise impedance measurements in radio-frequency transmission lines. In 1951, the General Radio Company offered the Type 874-L8 Slotted Line, a coaxial device that sampled electric field patterns along a slotted transmission line to detect standing waves, enabling calculations of voltage standing wave ratio (VSWR) and impedance using tools like Smith charts. This instrument operated in the RF range, typically 300 to 900 MHz, and represented a foundational advancement over prior manual methods for characterizing network reflections and mismatches.[20]

Throughout the 1950s and 1960s, scalar network analyzers became standard tools, relying on bridge circuits—such as substitution or reflection bridges—and crystal detectors to isolate and measure signal magnitudes in waveguide systems at microwave frequencies. These setups, often manual and frequency-specific, focused on amplitude responses like insertion loss, return loss, and gain, without capturing phase data, making them suitable for evaluating passive components in high-frequency environments. For instance, Hewlett-Packard's early slotted line sections, like the 805A introduced around 1950, complemented bridge-based measurements by providing direct VSWR readings in waveguides from 500 MHz to 4 GHz, facilitating quick assessments in laboratory and production settings. Limitations included labor-intensive manual probe adjustments, susceptibility to environmental factors, and restricted bandwidth, often requiring multiple setups for broad frequency coverage.[21][22]

A pivotal milestone in this era was Hewlett-Packard's 8405A Vector Voltmeter, launched in 1966, which introduced simultaneous amplitude and phase measurements using dual-channel detection and step-recovery diode sampling, extending capabilities to 1 GHz with sensitivities down to 100 μV. Despite these improvements, the 8405A retained analog limitations like manual frequency tuning and narrow instantaneous bandwidth, preventing swept-frequency automation. These early scalar and nascent vector instruments were instrumental in microwave engineering applications, particularly for antenna pattern evaluation and filter response characterization during Cold War radar system developments, where precise component testing supported defense initiatives like early-warning and tracking technologies.[23][24]

This foundation of analog, magnitude-focused tools paved the way for automated vector analyzers in later decades.

Evolution to vector analyzers

The transition from scalar to vector network analyzers marked a pivotal advancement in microwave measurement technology during the late 20th century, enabling the capture of both magnitude and phase information for more accurate characterization of electrical networks. A key breakthrough occurred in the late 1960s with the introduction of the Hewlett-Packard (HP) 8410A vector network analyzer (VNA) in 1967, which integrated a swept microwave source, directional bridge, and phase-sensitive receiver to automate the measurement of complex scattering parameters across frequencies from 10 MHz to 12 GHz.[25] This system revolutionized component testing by providing direct vector measurements, replacing manual phase adjustments and graphical constructions common in earlier scalar approaches, and laid the foundation for automated swept magnitude and phase analysis.[26]

In the 1980s and 1990s, VNAs evolved to support multiport configurations and extended frequency ranges, addressing the growing complexity of integrated circuits and communication systems. The HP 8510 series, introduced in 1985, exemplified this progress with its modular architecture that facilitated multiport measurements through switch matrices and test sets, allowing simultaneous characterization of multiple ports with full two-port error correction.[27] Frequency extension modules, developed during this period, pushed VNA capabilities to 110 GHz by the early 1980s using harmonic multipliers and waveguide mixers, enabling precise vector analysis in millimeter-wave applications such as radar and satellite components.[28] Concurrently, software advancements in error correction, including 12-term models implemented in instruments like the HP 8510, mathematically compensated for systematic imperfections in the measurement path, enhancing accuracy without hardware modifications.[29] Calibration improvements during this era further enabled reliable vector measurements by standardizing error models across systems.[30]

Entering the 2000s, VNAs incorporated advanced signal processing and nonlinear capabilities, driven by the demands of high-speed digital and RF integrated circuits. Time-domain transforms became seamlessly integrated into VNA firmware, allowing inverse Fourier analysis of frequency-domain S-parameters to visualize reflections and discontinuities as step or impulse responses, a feature refined in models like the Agilent (now Keysight) PNA series for fault isolation in transmission lines.[31] Nonlinear vector network analysis (NVNA) emerged as a significant innovation, extending VNA functionality to large-signal regimes by measuring amplitude and phase of harmonics and intermodulation products, with early commercial implementations in the mid-2000s providing vector-corrected data for active device modeling.[32] The Keysight PNA-X series, launched around 2008, represented a high-frequency milestone with base capabilities up to 67 GHz and extension modules reaching 500 GHz, supporting comprehensive tests for amplifiers and converters in 5G and beyond applications.[33][28] The exponential growth in computational power, aligned with Moore's Law, accelerated VNA processing speeds for complex operations like real-time error correction and time-domain simulations, reducing measurement times from minutes to seconds in modern systems.[34]

In the 2010s and 2020s, VNAs continued to advance with extensions into terahertz frequencies, reaching up to 1.1 THz as of 2025 through specialized frequency extenders and improved dynamic ranges exceeding 140 dB. These developments, led by manufacturers like Keysight, Rohde & Schwarz, and Anritsu, have been crucial for testing components in 5G millimeter-wave systems, 6G research, and emerging THz applications such as high-speed communications and sensing. Innovations include modular designs for multiport (up to 48 ports) and nonlinear measurements, alongside software for automated calibration and integration with simulation tools, ensuring VNAs remain essential for bridging design and validation in advanced RF and wireless technologies.[3][35][36]

Leading companies such as Keysight Technologies (formerly HP/Agilent), Rohde & Schwarz, and Anritsu have played instrumental roles in VNA standardization by developing interoperable hardware, contributing to IEEE measurement standards, and promoting consistent calibration protocols that ensure reproducibility across global R&D and manufacturing environments.[37][36]

Scalar and vector measurement techniques

Network analyzers employ scalar measurement techniques to determine the magnitude of electrical signals without capturing phase information, making them suitable for basic assessments of device performance. These methods typically utilize power detectors, such as diode bridges or thermocouples, which convert high-frequency microwave signals into detectable low-frequency AC or DC voltages proportional to the signal power. In operation, a swept-frequency source generates an incident signal that is directed toward the device under test (DUT), and the reflected or transmitted responses are isolated and measured relative to the incident wave. The reflection coefficient magnitude is calculated as ∣Γ∣=VrVi|\Gamma| = \frac{V_r}{V_i}∣Γ∣=Vi​Vr​​, where VrV_rVr​ is the reflected voltage and ViV_iVi​ is the incident voltage, providing metrics like return loss or insertion loss for quick pass/fail evaluations.[13][14]

Vector measurement techniques extend scalar methods by incorporating phase information, enabling a complete characterization of both magnitude and phase shifts introduced by the DUT. These rely on coherent detection schemes, such as homodyne or heterodyne receivers, where the test signal is compared against a stable reference to preserve phase coherence across frequencies. The vector ratio is expressed as VoutVin=∣A∣ejϕ\frac{V_{out}}{V_{in}} = |A| e^{j\phi}Vin​Vout​​=∣A∣ejϕ, quantifying the complex response that accounts for both amplitude ∣A∣|A|∣A∣ and phase ϕ\phiϕ. Signal flow begins with the generation of an incident wave from the source, which is separated from the reflected or transmitted waves using directional couplers or bridges to ensure accurate isolation of forward and reverse signals before detection.[14][13]

The shared signal flow in both techniques involves launching an incident wave onto the DUT via a transmission line, separating it from responses with directional elements, and detecting the resultant signals to compute ratios. However, scalar approaches prioritize simplicity and broadband operation for applications like gain verification in amplifiers or filters, offering lower cost and faster measurements but limited accuracy due to the absence of phase data, which can obscure impedance details or multi-port interactions. In contrast, vector techniques provide superior precision for full network modeling, such as determining complex scattering parameters, though they require more sophisticated synchronization and calibration to mitigate phase errors.[14][13]

Frequency and time-domain analysis

Network analyzers primarily operate in the frequency domain by generating a swept sine wave signal across a broad range, typically from as low as 10 Hz to over 1 THz in advanced systems, to measure the steady-state responses of linear time-invariant networks. This approach leverages the Fourier basis, where the device under test (DUT) is excited at discrete frequencies, allowing the analyzer to capture amplitude and phase information for each point independently.[38][39][40]

In the time domain, vector network analyzers derive responses by applying the inverse Fourier transform (IFT) to the frequency-domain S-parameter data, yielding step or impulse responses that reveal transient behaviors. This transformation enables time-domain reflectometry (TDR) for reflection measurements and time-domain transmission (TDT) for transmission, facilitating the precise location of discontinuities such as impedance mismatches or faults in the DUT.[41][31][42]

Two main techniques are employed for time-domain analysis: bandpass mode, which simulates a chirp-like swept excitation over a frequency band starting above DC and produces impulse responses suitable for narrowband or harmonic grids, and broadband (low-pass) mode, which includes DC and emulates impulse or step excitations for full-spectrum analysis. The temporal resolution in these transforms is fundamentally limited by the frequency span Δf\Delta fΔf, approximated as Δt≈1Δf\Delta t \approx \frac{1}{\Delta f}Δt≈Δf1​, where wider spans yield finer resolution, such as approximately 2 mm for a 70 GHz span in air.[43][44][43]

Frequency-domain analysis is particularly applied to characterize filters by evaluating passband ripple, insertion loss, and group delay across the spectrum, while time-domain methods excel in cable fault detection, identifying break locations and reflection coefficients through TDR waveforms.[31][45]

Signal generation and test set

The signal generation subsystem in a vector network analyzer (VNA) employs a synthesized RF source to produce stable continuous wave (CW) or frequency-swept signals that stimulate the device under test (DUT). These sources typically utilize phase-locked loop (PLL) synthesizers for precise frequency control and low phase noise, supporting operational ranges from 9 kHz to 67 GHz or higher in advanced models.[36] Output power levels span -85 dBm to +13 dBm to accommodate diverse measurement conditions, with phase noise performance below -100 dBc/Hz at a 10 kHz offset to reduce noise contributions in sensitive applications.[46]

The test set serves as the interface between the signal source and the DUT, incorporating directional bridges or couplers to separate incident, reflected, and transmitted signals with high isolation. These components achieve directivity of 20 to 50 dB, enabling accurate discrimination between forward and reverse power flows essential for reflection measurements. In multiport VNAs, integrated switch matrices facilitate automated routing, supporting configurations from 2-port to 4-port or beyond by electronically selecting signal paths without physical reconnection.[47]

Protective and conditioning elements within the test set include bias tees, which allow DC biasing of active DUTs such as amplifiers while maintaining RF continuity, and variable or fixed attenuators to safeguard the instrument from excessive input power or to fine-tune signal levels. Impedance matching to the 50 Ω reference standard is inherent across all ports to minimize reflections and ensure measurement fidelity.

Overall integration of the signal generator and test set yields precise control, with source power accuracy of ±0.1 dB and frequency settling times under 1 ms, supporting rapid sweeps and repeatable results in high-throughput testing environments.[36]

Receiver and processing

The receiver in a network analyzer captures the reflected and transmitted signals from the device under test (DUT) through multi-channel intermediate frequency (IF) detection systems. These receivers downconvert the high-frequency RF signals to a lower IF using superheterodyne architectures with local oscillators to preserve signal integrity for subsequent digitization. The IF bandwidth for filtering is typically 1 Hz to 100 kHz. This process enables simultaneous extraction of magnitude and phase information across multiple receiver channels (e.g., reference, reflection, and transmission paths), supporting vector measurements essential for accurate characterization.[2][38]

Receiver performance is characterized by a dynamic range of 100-140 dB, which defines the span between the maximum receivable signal power (often limited by compression points around +10 to +20 dBm) and the minimum detectable signal, allowing precise quantification of both strong reflections and weak transmissions. The noise floor, typically around -120 dBm at narrow IF bandwidths (e.g., 10 Hz), sets the lower limit and is influenced by thermal noise, receiver gain, and IF filtering; wider bandwidths increase the noise floor but speed up measurements. These specifications ensure reliable detection in applications ranging from low-loss filters to high-attenuation devices.[48][2]

The processing unit serves as the computational backbone, utilizing digital signal processors (DSPs) to refine raw receiver data into usable measurements. DSP algorithms perform time-domain averaging (e.g., 1 to 1000 sweeps) to suppress random noise by a factor of √N, where N is the number of averages, improving signal-to-noise ratio without sacrificing sweep speed. Interpolation techniques generate intermediate data points for smoother traces, while built-in error correction applies vector-calibrated models to mitigate directivity, source match, and other systematic errors. Firmware handles advanced trace mathematics, including transformations for Smith chart plotting, which maps complex impedance data into polar coordinates for intuitive visualization of reflection coefficients.[48][38][2]

The display interface integrates these processed results into real-time graphical formats, plotting logarithmic magnitude (dB scale for gain/loss), unwrapped phase (degrees for delay analysis), and derived metrics like group delay (time derivative of phase). These visualizations update dynamically during sweeps, aiding immediate DUT assessment, with customizable overlays for multi-trace comparisons. Processed data supports export in industry-standard formats, such as Touchstone (.s2p files for two-port S-parameters), enabling seamless integration with circuit simulators like Keysight ADS or Ansys HFSS.[49]

Contemporary network analyzers incorporate high-performance embedded controllers with clock speeds reaching 1 GHz, leveraging multi-core architectures and optimized firmware to achieve sweep rates exceeding 1000 points per second. This enables full 100,000-point frequency sweeps across multi-GHz spans in mere seconds, even with averaging and error correction applied, drastically reducing test times in production environments while maintaining resolution for fine details like resonances.[50][51]

S-parameter characterization

Scattering parameters, commonly known as S-parameters, describe the linear behavior of electrical networks at high frequencies by relating incident and reflected traveling waves at the network ports. Introduced by Kaneyuki Kurokawa, these parameters are defined in terms of power waves, where the incident wave aia_iai​ and reflected wave bib_ibi​ at port iii are normalized such that their magnitudes squared represent normalized power. For a two-port network, the input reflection coefficient is S11=b1/a1S_{11} = b_1 / a_1S11​=b1​/a1​ when port 2 is terminated in a matched load, quantifying the fraction of power reflected back to port 1. The forward transmission coefficient is S21=b2/a1S_{21} = b_2 / a_1S21​=b2​/a1​, representing the transmitted power from port 1 to port 2 under the same conditions.[52][52][52]

In matrix form, S-parameters for an NNN-port network are expressed as the scattering matrix [S][S][S], where the vector of outgoing waves relatestothevectorofincomingwavesrelates to the vector of incoming wavesrelatestothevectorofincomingwaves by =[S]= [S]=[S], or equivalently [S]=−1[S] = ^{-1}[S]=−1. This formulation assumes the network is linear and operates under small-signal conditions, making it ideal for characterizing passive and active RF/microwave devices.[52][53]

To measure the full two-port S-parameter matrix with a vector network analyzer (VNA), forward and reverse sweeps are performed across the frequency range of interest. In the forward configuration, the VNA source excites port 1 while port 2 is matched, yielding S11S_{11}S11​ and S21S_{21}S21​; the reverse sweep excites port 2 to measure S22S_{22}S22​ and S12S_{12}S12​, ensuring reciprocity (S12=S21S_{12} = S_{21}S12​=S21​) for passive networks. When fixtures such as probes or test adapters are involved, de-embedding algorithms mathematically remove their parasitic effects from the raw measurements to isolate the device under test (DUT) S-parameters, often using models derived from open, short, and thru standards.[53][53][54]

Interpretation of S-parameters provides key insights into network performance. The magnitude ∣S11∣|S_{11}|∣S11​∣ determines the voltage standing wave ratio (VSWR) via VSWR=1+∣S11∣1−∣S11∣\text{VSWR} = \frac{1 + |S_{11}|}{1 - |S_{11}|}VSWR=1−∣S11​∣1+∣S11​∣​, where values close to 1 indicate good impedance matching and minimal reflections. Insertion loss, a measure of signal attenuation, is derived from the transmission parameter as −20log⁡10∣S21∣-20 \log_{10} |S_{21}|−20log10​∣S21​∣ in decibels, with lower values signifying efficient power transfer. Measurement uncertainties, particularly from trace noise in the VNA's receivers, degrade accuracy for low-magnitude parameters, introducing phase and amplitude errors that can exceed 1 dB in noisy conditions.[55][55][56]

For differential circuits, such as balanced transmission lines in high-speed interconnects, balanced or mixed-mode S-parameters extend the conventional formulation by decomposing signals into differential and common-mode components. These parameters, defined using even- and odd-mode impedances, include terms like differential reflection Sdd11S_{dd11}Sdd11​ and mode conversion Sdc21S_{dc21}Sdc21​, facilitating analysis of crosstalk and imbalance in symmetric structures.

Noise figure and gain measurements

Noise figure quantifies the degradation in the signal-to-noise ratio (SNR) caused by a device under test (DUT), defined as the noise factor F=SNRinSNRoutF = \frac{\mathrm{SNR_{in}}}{\mathrm{SNR_{out}}}F=SNRout​SNRin​​, where noise figure NF=10log⁡10F\mathrm{NF} = 10 \log_{10} FNF=10log10​F is expressed in decibels.[57] This metric is essential for evaluating the noise performance of amplifiers and receivers in RF and microwave systems.[58]

The Y-factor method is a widely adopted technique for measuring noise figure, involving the use of hot and cold noise sources to determine the noise parameters of the DUT.[57] In this approach, the Y-factor is calculated as the ratio of the output noise power with the hot noise source (NhN_hNh​) to that with the cold noise source (NcN_cNc​), where Y=NhNcY = \frac{N_h}{N_c}Y=Nc​Nh​​, and the noise factor is derived as F=ENRY−1F = \frac{\mathrm{ENR}}{Y - 1}F=Y−1ENR​ with ENR being the excess noise ratio of the source.[59] Measurements are typically performed at spot frequencies across a swept range, contrasting with broadband methods that average over wider bandwidths for overall system assessment.[57]

Gain measurements in the context of noise figure characterization often involve the associated gain GaG_aGa​, derived from the magnitude squared of the transmission parameter ∣S21∣2|S_{21}|^2∣S21​∣2, with corrections applied for input and output mismatches to ensure accuracy.[60]

Vector network analyzers (VNAs) equipped with a noise figure option utilize the receiver as a noise receiver, enabling direct noise power measurements without additional hardware in some configurations.[61] These systems achieve high accuracy, with typical uncertainties of ±0.1 dB for noise figure measurements up to 50 GHz when using low-noise receivers and proper calibration.[62]

The effective noise temperature TeT_eTe​ provides an equivalent representation of the noise added by the DUT, given by Te=T0(F−1)T_e = T_0 (F - 1)Te​=T0​(F−1), where T0=290T_0 = 290T0​=290 K is the standard reference temperature.[63] This parameter is particularly useful for cryogenic or low-noise applications, linking noise figure to thermal noise equivalents.[57]

Error models and correction methods

In vector network analyzer (VNA) measurements, systematic errors arise from imperfections in the instrument's signal path, including reflections, losses, and crosstalk, which can significantly distort the characterization of devices under test. The 12-term error model addresses these for two-port measurements by representing the forward and reverse signal paths separately, capturing twelve independent error coefficients that account for the analyzer's non-ideal behavior. These terms include directivity (leakage from incident to reflected signal paths, denoted as e00e_{00}e00​ and e33e_{33}e33​), source match (reflection at the source port, e11e_{11}e11​ and e44e_{44}e44​), load match (reflection at the load port, e22e_{22}e22​ and e55e_{55}e55​), reflection tracking (phase and magnitude errors in reflection measurements, e10e01e_{10}e_{01}e10​e01​ and e43e34e_{43}e_{34}e43​e34​), transmission tracking (errors in forward and reverse transmission, e10e32e_{10}e_{32}e10​e32​ and e23e01e_{23}e_{01}e23​e01​), and isolation or crosstalk (leakage terms, e30e_{30}e30​ and e03e_{03}e03​). The model conceptualizes the signal flow through directional couplers, switches, and mixers, where incident waves (a1,a2a_1, a_2a1​,a2​) are separated from reflected/transmitted waves (b1,b2b_1, b_2b1​,b2​) before down-conversion and digitization, with error adapters inserted mathematically between the VNA ports and an ideal device to model deviations.[64]

For one-port measurements, the Short-Open-Load (SOL) calibration simplifies the model to three terms—directivity (D=e00D = e_{00}D=e00​), source match (M=e11M = e_{11}M=e11​), and reflection tracking—using three known standards to solve for the coefficients. The corrected reflection coefficient Γ\GammaΓ is renormalized to ideal conditions via the equation:

where Γm\Gamma_mΓm​ is the measured reflection coefficient, derived by solving the system of equations from the standards' known responses to isolate the device under test (DUT) from systematic errors. This approach yields high accuracy for reflection-only characterizations, such as antenna or filter inputs. For two-port extensions, the full 12-term model incorporates these one-port corrections into a bilinear transformation framework, enabling comprehensive S-parameter renormalization across both directions.[64][65]

The Through-Reflect-Line (TRL) calibration method enhances two-port accuracy, particularly for broadband applications, by employing transmission line standards instead of precise loads, reducing dependency on connector-specific artifacts. Developed for non-coaxial environments like on-wafer or in-fixture testing, TRL uses a zero-length thru, a high-reflection standard (e.g., short or open), and a precisely known-length transmission line to determine the 7- or 8-term error model (a subset of the 12-term for reciprocal networks), solving for propagation constants and reference impedances over wide frequency bands. This technique achieves superior performance up to millimeter-wave frequencies (e.g., beyond 110 GHz with appropriate lines), as the line standards' characteristic impedance and electrical length provide robust renormalization without needing low-loss loads, minimizing uncertainties in high-frequency scattering parameters. As of 2025, specialized calibration kits, such as Keysight's 85065A for up to 250 GHz, and photonic frequency extenders enabling measurements to 520 GHz, further extend these capabilities for THz applications.[65][66]

After calibration, residual uncertainties propagate through the error terms, typically quantified using the root-sum-square (RSS) method to combine contributions from directivity, match, and tracking residuals, along with noise and drift. For well-calibrated systems, magnitude uncertainties are often on the order of 0.05 dB, while phase uncertainties reach about 1° across microwave bands, though these degrade at higher frequencies or lower signal levels due to covariance in the error model coefficients. Automated software implementations facilitate these corrections by iteratively solving the model equations during measurement sweeps.[67][68]

Automated fixtures and procedures

Automated fixtures and procedures enhance the efficiency and repeatability of calibration in vector network analyzers (VNAs) by integrating hardware standards and software-driven workflows that minimize manual intervention. These tools address systematic errors through predefined sequences, enabling precise measurements in research, development, and production environments.[69]

Mechanical calibration kits, such as Short-Open-Load-Through (SOLT) types, form the foundation of traditional automated setups. These kits typically include a short circuit standard for zero reflection, an open circuit for infinite reflection, a matched load (usually 50 Ω) for absorption, and a through connection for transmission characterization; precision airlines—coaxial sections with known lengths—are often incorporated to support verification or TRL extensions.[69][70] Such kits connect sequentially via automated prompts, ensuring consistent torque and alignment to reduce variability. Electronic calibration (ECal) kits offer faster switching by replacing mechanical standards with solid-state switches that electronically simulate shorts, opens, loads, and through paths, enabling one-connection calibration from DC to 67 GHz with reduced setup time—often under a minute per port—compared to multi-step mechanical processes.[71][72]

Software-guided procedures automate the execution of these kits, particularly for 12-term full two-port error correction models that account for forward and reverse directivity, source/load match, reflection and transmission tracking, and isolation. Tools like the Keysight Calibration Wizard provide step-by-step interfaces that detect device-under-test (DUT) connectors, recommend compatible kits, and sequence measurements for SOLT or ECal methods, supporting flexible port selection and thru configurations.[73][74] For multiport VNAs, these wizards extend to N-port calibrations by automating pairwise thru connections and error term computations, streamlining setups for devices with 4 or more ports.[73]

Test fixture simulators (TFS) automate de-embedding of PCB launch effects, where measured S-parameters include unwanted fixture parasitics like connectors and traces. TFS software models the fixture using S-parameter files or electromagnetic simulations, then mathematically removes its contribution to isolate DUT responses, improving accuracy for embedded components up to millimeter-wave frequencies.[75] In high-volume production, automated robotic systems handle calibration and probing, such as six-axis manipulators for on-wafer RF testing that align probes and execute VNA sweeps with sub-micron repeatability, reducing cycle times for batch characterization.[76]

Traceable standards, like 7-mm coaxial kits compliant with IEEE 287, ensure metrological integrity through artifacts verified against national references. These kits include precision shorts, opens, loads, and optional sliding loads for broadband use up to 18 GHz. Manufacturers recommend annual verification cycles to maintain performance, involving measurements against check standards to confirm return loss and insertion loss within specifications.[77][78]

Verification kits and protocols

Verification kits for vector network analyzers (VNAs) consist of precisely characterized standards designed to independently validate the accuracy and performance of the instrument after calibration. These kits typically include known-good components such as offset loads, flush shorts, and broadband terminations, which allow users to measure residual errors in parameters like directivity, source match, and load match by comparing obtained results against provided certificate data. For instance, offset shorts with varying electrical lengths enable the evaluation of phase tracking linearity across the frequency range, while flush shorts help assess reflection coefficient accuracy at specific points.[79]

A representative example is the Keysight 85058V verification kit, which supports frequencies up to 67 GHz and incorporates airline standards and adapters for coaxial and waveguide interfaces to confirm system stability and error correction efficacy.[80] These kits are essential for detecting subtle imperfections, such as imperfect connections or drift in the VNA's internal references, ensuring measurements remain reliable for device characterization.

Verification protocols involve systematic testing procedures to quantify uncertainty and repeatability, often employing software tools like the METAS VNA Tools II for computing S-parameter uncertainties based on propagation of error models from the kit measurements. Key protocols include sweep-to-sweep repeatability assessments, where multiple consecutive sweeps on a stable standard, such as a matched load, must exhibit magnitude variations below 0.01 dB and phase variations below 0.2 degrees to confirm short-term stability. Uncertainty evaluation follows established frameworks, integrating contributions from random effects like noise and systematic effects like residual directivity, with tools providing propagation calculations compliant with metrology guidelines.[81]

Critical metrics in these protocols include effective directivity exceeding 40 dB, which verifies minimal signal leakage and high isolation in the test setup, and reflection tracking linearity, ensuring the VNA accurately reproduces known reflection responses without distortion. If verification indicates deviations, such as reduced directivity below specifications, post-verification adjustments like connector re-torquing or partial recalibration are applied to restore performance. At mm-wave frequencies above 100 GHz, verification faces unique challenges due to increased waveguide losses and potential higher-order mode excitation; overmoded waveguides are employed in these protocols to introduce and quantify multi-mode propagation effects, allowing assessment of mode purity and coupling in high-frequency standards.[82][83]

Traceability to international standards

Traceability of network analyzer measurements to international standards ensures the reliability and comparability of results across global laboratories by establishing an unbroken chain from primary references to end-user calibrations. This chain originates with primary standards at national metrology institutes, such as the National Institute of Standards and Technology (NIST) in the United States and the Physikalisch-Technische Bundesanstalt (PTB) in Germany, which provide the foundational accuracy for vector network analyzer (VNA) S-parameter determinations. For example, NIST utilizes multiline thru-reflect-line (TRL) calibrations with standards of traceable dimensions—such as line lengths from 14.999 mm to 29.981 mm with uncertainties of ±0.009 mm—to characterize electronic calibration units, propagating uncertainties via the NIST Microwave Uncertainty Framework for 95% confidence bounds. PTB has developed a similar traceability path for on-wafer measurements up to 110 GHz, employing multiline TRL on custom alumina wafers to characterize commercial impedance standard substrates like the GGB CS-5, ensuring single-mode propagation and alignment with SI units through documented material properties and low-loss approximations. Cryogenic attenuators serve as key primary standards in these chains, enabling low-noise power measurements at millikelvin temperatures to support high-frequency accuracy while minimizing thermal loading.[84][85][86]

Commercial calibration kits and verification standards link to these primary references through laboratories accredited under ISO/IEC 17025, which mandates an uninterrupted calibration hierarchy with quantified uncertainties for each step, including statistical correlations and sensitivity analyses. This accreditation facilitates the transfer of metrological reliability from national institutes to industry, where VNAs are calibrated using traceable kits to achieve measurement uncertainties typically below 0.1 dB in magnitude and 1° in phase up to 50 GHz. Developments in vector calibration standards since the 1990s have strengthened this chain, incorporating advanced error models and electronic units to replace mechanical kits for more robust, repeatable characterizations.[87][88]

International bodies further standardize these practices to promote global consistency. The International Electrotechnical Commission (IEC) standard 62153-4-3 outlines the triaxial method for assessing screening attenuation via surface transfer impedance of metallic cable shields, employing VNAs with S-parameter test sets or power splitters for measurements up to 100 MHz, with calibration at connector reference planes to ensure traceability. The IEEE Standard 370-2020 addresses de-embedding of fixture effects in printed circuit board interconnects up to 50 GHz, recommending test structures like lines and Beatty standards alongside quality metrics for S-parameter validation, directly supporting vector calibration in high-speed RF applications. These standards, integrated into accredited workflows, enable comparable results for electromagnetic compatibility and signal integrity testing worldwide.[89][90]

RF and microwave device testing

Network analyzers play a crucial role in characterizing RF and microwave components, enabling precise evaluation of their linear performance parameters such as scattering parameters (S-parameters), gain, and impedance. For amplifiers, vector network analyzers (VNAs) measure key metrics including gain flatness across frequency bands and input/output return loss to ensure minimal signal reflections and stable amplification. These measurements help identify deviations from ideal behavior, such as ripple in gain response, which can degrade system performance in high-frequency applications.[94]

Filters, essential for signal selectivity in RF systems, are tested using VNAs to assess bandpass response, insertion loss, and out-of-band rejection. By sweeping frequencies and analyzing magnitude and phase of transmitted signals, engineers verify filter alignment with design specifications, ensuring effective suppression of unwanted frequencies while passing desired bands with minimal attenuation.[94] Antennas, particularly phased arrays for modern systems, undergo VNA-based testing in compact antenna test ranges (CATR) to evaluate radiation patterns, beam steering accuracy, and S-parameters under far-field conditions. This setup simulates real-world propagation, allowing measurement of gain, directivity, and phase alignment without anechoic chamber limitations.[95]

In the telecommunications industry, VNAs support testing of 5G base stations across sub-6 GHz (FR1) and mm-wave (FR2) bands, where they characterize components like duplexers and transceivers for low return loss and high isolation to meet stringent throughput requirements. For satellite communications, Ka-band (26.5-40 GHz) S-parameter measurements using extended-range VNAs ensure reliable link performance in transceivers and feeds, verifying low insertion loss and phase stability critical for high-data-rate uplinks and downlinks.[96][97]

The typical workflow for RF device development integrates VNA measurements with simulation tools like Keysight's Advanced Design System (ADS), where measured S-parameter data is imported to validate and refine circuit models before hardware iteration. In production, automated test equipment (ATE) incorporating VNAs performs high-volume screening with limit lines for pass/fail criteria on parameters like gain and return loss, facilitating yield analysis by tracking defect rates and optimizing manufacturing processes.[98][98]

A representative case involves testing power amplifier (PA) modules in smartphone front-end modules, where VNAs assess linear gain and return loss alongside brief evaluations of intermodulation distortion (IMD) and efficiency under multi-tone stimuli to ensure compliance with mobile network standards.[99][100]

Nonlinear and large-signal analysis

Network analyzers equipped with nonlinear vector network analyzer (NVNA) capabilities enable the characterization of device nonlinearities through harmonic balance simulations, which solve for steady-state responses under large-signal conditions by balancing harmonic components in the frequency domain.[101] This approach is particularly useful for measuring third-order intermodulation distortion (IMD3), where two-tone signals are applied to assess distortion products that arise from nonlinear mixing, providing insights into amplifier linearity under multi-tone excitation. For instance, NVNA systems can capture vector-corrected measurements up to 50 GHz, allowing accurate extraction of IMD3 levels that traditional scalar methods overlook.[102]

X-parameters extend the linear S-parameter framework to nonlinear regimes, offering a measurement-based black-box model that describes incident and reflected waves under large-signal drive, including harmonic generation and compression effects.[103] Developed as a device-independent representation, X-parameters facilitate behavioral modeling of active components like transistors without requiring internal circuit details, enabling faster simulation integration in design tools.[104] They are extracted using NVNA setups, where multi-tone excitations reveal scattering behavior across power levels, supporting applications in predicting nonlinear interactions in cascaded systems.[105]

In large-signal analysis, hot S-parameters measure scattering characteristics under biased and powered conditions, capturing real operating impedances and gain compression that differ from cold (low-power) measurements.[101] This technique uses vector-corrected receivers to quantify changes in S21 and output power as input drive increases, essential for evaluating amplifier performance near saturation.[16] Load-pull measurements complement this by systematically varying source and load impedances to map contours of optimum performance metrics, such as maximum output power or power-added efficiency (PAE).[106] For example, tuner-based load-pull systems generate constant-gain or constant-PA circles on the Smith chart, guiding matching network design for devices like GaN HEMTs.[107]

Modern network analyzers like the Keysight PNA-X incorporate NVNA software options to perform these nonlinear assessments directly, integrating source power calibration and harmonic receivers for comprehensive large-signal characterization up to millimeter-wave frequencies.[17] The nonlinear noise option extends this to evaluate noise contributions under compressed operation, combining vector signal analysis with low-noise receivers for metrics like noise figure degradation at high power.[101] For thermal-sensitive devices such as GaN HEMTs, pulsed IV and RF measurements mitigate self-heating effects by applying short pulses (e.g., 100 ns to 1 µs duty cycles), enabling extraction of intrinsic device parameters and thermal resistance without thermal runaway. This pulsed approach reveals trap-related dispersions in I-V curves, informing thermal management strategies like substrate selection for high-power operation.

Definition and purpose

A network analyzer is an instrument used to characterize the behavior of primarily linear electrical networks, with extensions for nonlinear analysis, in the frequency domain across radio frequency (RF) and microwave ranges, typically spanning from kilohertz (kHz) to terahertz (THz) frequencies.[5][6][7]

The primary purpose of a network analyzer is to measure key parameters such as scattering parameters (S-parameters), impedance, admittance, insertion loss, and gain, enabling engineers to evaluate device performance, ensure quality, and optimize impedance matching in applications like antennas, filters, amplifiers, and transmission lines.[5][7][8]

In basic operation, the instrument generates a swept-frequency signal that is applied to the device under test, then analyzes the reflected and transmitted responses to determine both magnitude and phase characteristics as a function of frequency.[9][10]

Vector network analyzers remain the most prevalent type in modern use.[5]

Types of network analyzers

Network analyzers are categorized primarily by their measurement capabilities, such as whether they capture magnitude only or both magnitude and phase, as well as by their handling of signal power levels and specialized applications.[9] The main types include scalar, vector, and large-signal variants, each suited to different testing needs in RF and microwave engineering.[11][8]

Scalar network analyzers (SNAs) measure only the magnitude, or amplitude, of network parameters, providing a cost-effective option for basic assessments without phase information.[12] They are ideal for transmission and reflection tests, such as evaluating insertion loss, return loss, gain, and standing wave ratio (SWR) in linear devices like filters and cables.[9] Swept-tuned SNAs emerged in the 1960s and 1970s as commercial instruments for these magnitude-based measurements.[13]

Vector network analyzers (VNAs) measure both magnitude and phase to determine complex scattering parameters (S-parameters), enabling full characterization of linear network behavior.[11] VNAs, which became the dominant type following their commercial introduction in the mid-1960s, support multiport configurations from 2 to 48 ports for testing complex systems like antennas and multi-antenna arrays.[14][7]

Large-signal network analyzers (LSNAs) extend VNA capabilities to nonlinear regimes under high-power conditions, measuring X-parameters that describe amplitude-dependent behaviors and harmonics. They are essential for analyzing devices like power amplifiers and mixers where traditional S-parameters fail due to nonlinearity.[15]

Other specialized types include time-domain network analyzers, which use time-domain reflectometry (TDR) to assess signal integrity in high-speed interconnects by transforming frequency-domain data.[16] Pulsed RF analyzers support radar and pulsed system testing by capturing transient responses, while millimeter-wave analyzers extend operations up to 1.1 THz for 5G, 6G, and terahertz applications.[17][7]

Early instruments

The origins of network analyzers in electrical engineering emerged in the mid-20th century with the development of slotted line instruments designed for precise impedance measurements in radio-frequency transmission lines. In 1951, the General Radio Company offered the Type 874-L8 Slotted Line, a coaxial device that sampled electric field patterns along a slotted transmission line to detect standing waves, enabling calculations of voltage standing wave ratio (VSWR) and impedance using tools like Smith charts. This instrument operated in the RF range, typically 300 to 900 MHz, and represented a foundational advancement over prior manual methods for characterizing network reflections and mismatches.[20]

Throughout the 1950s and 1960s, scalar network analyzers became standard tools, relying on bridge circuits—such as substitution or reflection bridges—and crystal detectors to isolate and measure signal magnitudes in waveguide systems at microwave frequencies. These setups, often manual and frequency-specific, focused on amplitude responses like insertion loss, return loss, and gain, without capturing phase data, making them suitable for evaluating passive components in high-frequency environments. For instance, Hewlett-Packard's early slotted line sections, like the 805A introduced around 1950, complemented bridge-based measurements by providing direct VSWR readings in waveguides from 500 MHz to 4 GHz, facilitating quick assessments in laboratory and production settings. Limitations included labor-intensive manual probe adjustments, susceptibility to environmental factors, and restricted bandwidth, often requiring multiple setups for broad frequency coverage.[21][22]

A pivotal milestone in this era was Hewlett-Packard's 8405A Vector Voltmeter, launched in 1966, which introduced simultaneous amplitude and phase measurements using dual-channel detection and step-recovery diode sampling, extending capabilities to 1 GHz with sensitivities down to 100 μV. Despite these improvements, the 8405A retained analog limitations like manual frequency tuning and narrow instantaneous bandwidth, preventing swept-frequency automation. These early scalar and nascent vector instruments were instrumental in microwave engineering applications, particularly for antenna pattern evaluation and filter response characterization during Cold War radar system developments, where precise component testing supported defense initiatives like early-warning and tracking technologies.[23][24]

This foundation of analog, magnitude-focused tools paved the way for automated vector analyzers in later decades.

Evolution to vector analyzers

The transition from scalar to vector network analyzers marked a pivotal advancement in microwave measurement technology during the late 20th century, enabling the capture of both magnitude and phase information for more accurate characterization of electrical networks. A key breakthrough occurred in the late 1960s with the introduction of the Hewlett-Packard (HP) 8410A vector network analyzer (VNA) in 1967, which integrated a swept microwave source, directional bridge, and phase-sensitive receiver to automate the measurement of complex scattering parameters across frequencies from 10 MHz to 12 GHz.[25] This system revolutionized component testing by providing direct vector measurements, replacing manual phase adjustments and graphical constructions common in earlier scalar approaches, and laid the foundation for automated swept magnitude and phase analysis.[26]

In the 1980s and 1990s, VNAs evolved to support multiport configurations and extended frequency ranges, addressing the growing complexity of integrated circuits and communication systems. The HP 8510 series, introduced in 1985, exemplified this progress with its modular architecture that facilitated multiport measurements through switch matrices and test sets, allowing simultaneous characterization of multiple ports with full two-port error correction.[27] Frequency extension modules, developed during this period, pushed VNA capabilities to 110 GHz by the early 1980s using harmonic multipliers and waveguide mixers, enabling precise vector analysis in millimeter-wave applications such as radar and satellite components.[28] Concurrently, software advancements in error correction, including 12-term models implemented in instruments like the HP 8510, mathematically compensated for systematic imperfections in the measurement path, enhancing accuracy without hardware modifications.[29] Calibration improvements during this era further enabled reliable vector measurements by standardizing error models across systems.[30]

Entering the 2000s, VNAs incorporated advanced signal processing and nonlinear capabilities, driven by the demands of high-speed digital and RF integrated circuits. Time-domain transforms became seamlessly integrated into VNA firmware, allowing inverse Fourier analysis of frequency-domain S-parameters to visualize reflections and discontinuities as step or impulse responses, a feature refined in models like the Agilent (now Keysight) PNA series for fault isolation in transmission lines.[31] Nonlinear vector network analysis (NVNA) emerged as a significant innovation, extending VNA functionality to large-signal regimes by measuring amplitude and phase of harmonics and intermodulation products, with early commercial implementations in the mid-2000s providing vector-corrected data for active device modeling.[32] The Keysight PNA-X series, launched around 2008, represented a high-frequency milestone with base capabilities up to 67 GHz and extension modules reaching 500 GHz, supporting comprehensive tests for amplifiers and converters in 5G and beyond applications.[33][28] The exponential growth in computational power, aligned with Moore's Law, accelerated VNA processing speeds for complex operations like real-time error correction and time-domain simulations, reducing measurement times from minutes to seconds in modern systems.[34]

In the 2010s and 2020s, VNAs continued to advance with extensions into terahertz frequencies, reaching up to 1.1 THz as of 2025 through specialized frequency extenders and improved dynamic ranges exceeding 140 dB. These developments, led by manufacturers like Keysight, Rohde & Schwarz, and Anritsu, have been crucial for testing components in 5G millimeter-wave systems, 6G research, and emerging THz applications such as high-speed communications and sensing. Innovations include modular designs for multiport (up to 48 ports) and nonlinear measurements, alongside software for automated calibration and integration with simulation tools, ensuring VNAs remain essential for bridging design and validation in advanced RF and wireless technologies.[3][35][36]

Leading companies such as Keysight Technologies (formerly HP/Agilent), Rohde & Schwarz, and Anritsu have played instrumental roles in VNA standardization by developing interoperable hardware, contributing to IEEE measurement standards, and promoting consistent calibration protocols that ensure reproducibility across global R&D and manufacturing environments.[37][36]

Scalar and vector measurement techniques

Network analyzers employ scalar measurement techniques to determine the magnitude of electrical signals without capturing phase information, making them suitable for basic assessments of device performance. These methods typically utilize power detectors, such as diode bridges or thermocouples, which convert high-frequency microwave signals into detectable low-frequency AC or DC voltages proportional to the signal power. In operation, a swept-frequency source generates an incident signal that is directed toward the device under test (DUT), and the reflected or transmitted responses are isolated and measured relative to the incident wave. The reflection coefficient magnitude is calculated as ∣Γ∣=VrVi|\Gamma| = \frac{V_r}{V_i}∣Γ∣=Vi​Vr​​, where VrV_rVr​ is the reflected voltage and ViV_iVi​ is the incident voltage, providing metrics like return loss or insertion loss for quick pass/fail evaluations.[13][14]

Vector measurement techniques extend scalar methods by incorporating phase information, enabling a complete characterization of both magnitude and phase shifts introduced by the DUT. These rely on coherent detection schemes, such as homodyne or heterodyne receivers, where the test signal is compared against a stable reference to preserve phase coherence across frequencies. The vector ratio is expressed as VoutVin=∣A∣ejϕ\frac{V_{out}}{V_{in}} = |A| e^{j\phi}Vin​Vout​​=∣A∣ejϕ, quantifying the complex response that accounts for both amplitude ∣A∣|A|∣A∣ and phase ϕ\phiϕ. Signal flow begins with the generation of an incident wave from the source, which is separated from the reflected or transmitted waves using directional couplers or bridges to ensure accurate isolation of forward and reverse signals before detection.[14][13]

The shared signal flow in both techniques involves launching an incident wave onto the DUT via a transmission line, separating it from responses with directional elements, and detecting the resultant signals to compute ratios. However, scalar approaches prioritize simplicity and broadband operation for applications like gain verification in amplifiers or filters, offering lower cost and faster measurements but limited accuracy due to the absence of phase data, which can obscure impedance details or multi-port interactions. In contrast, vector techniques provide superior precision for full network modeling, such as determining complex scattering parameters, though they require more sophisticated synchronization and calibration to mitigate phase errors.[14][13]

Frequency and time-domain analysis

Network analyzers primarily operate in the frequency domain by generating a swept sine wave signal across a broad range, typically from as low as 10 Hz to over 1 THz in advanced systems, to measure the steady-state responses of linear time-invariant networks. This approach leverages the Fourier basis, where the device under test (DUT) is excited at discrete frequencies, allowing the analyzer to capture amplitude and phase information for each point independently.[38][39][40]

In the time domain, vector network analyzers derive responses by applying the inverse Fourier transform (IFT) to the frequency-domain S-parameter data, yielding step or impulse responses that reveal transient behaviors. This transformation enables time-domain reflectometry (TDR) for reflection measurements and time-domain transmission (TDT) for transmission, facilitating the precise location of discontinuities such as impedance mismatches or faults in the DUT.[41][31][42]

Two main techniques are employed for time-domain analysis: bandpass mode, which simulates a chirp-like swept excitation over a frequency band starting above DC and produces impulse responses suitable for narrowband or harmonic grids, and broadband (low-pass) mode, which includes DC and emulates impulse or step excitations for full-spectrum analysis. The temporal resolution in these transforms is fundamentally limited by the frequency span Δf\Delta fΔf, approximated as Δt≈1Δf\Delta t \approx \frac{1}{\Delta f}Δt≈Δf1​, where wider spans yield finer resolution, such as approximately 2 mm for a 70 GHz span in air.[43][44][43]

Frequency-domain analysis is particularly applied to characterize filters by evaluating passband ripple, insertion loss, and group delay across the spectrum, while time-domain methods excel in cable fault detection, identifying break locations and reflection coefficients through TDR waveforms.[31][45]

Signal generation and test set

The signal generation subsystem in a vector network analyzer (VNA) employs a synthesized RF source to produce stable continuous wave (CW) or frequency-swept signals that stimulate the device under test (DUT). These sources typically utilize phase-locked loop (PLL) synthesizers for precise frequency control and low phase noise, supporting operational ranges from 9 kHz to 67 GHz or higher in advanced models.[36] Output power levels span -85 dBm to +13 dBm to accommodate diverse measurement conditions, with phase noise performance below -100 dBc/Hz at a 10 kHz offset to reduce noise contributions in sensitive applications.[46]

The test set serves as the interface between the signal source and the DUT, incorporating directional bridges or couplers to separate incident, reflected, and transmitted signals with high isolation. These components achieve directivity of 20 to 50 dB, enabling accurate discrimination between forward and reverse power flows essential for reflection measurements. In multiport VNAs, integrated switch matrices facilitate automated routing, supporting configurations from 2-port to 4-port or beyond by electronically selecting signal paths without physical reconnection.[47]

Protective and conditioning elements within the test set include bias tees, which allow DC biasing of active DUTs such as amplifiers while maintaining RF continuity, and variable or fixed attenuators to safeguard the instrument from excessive input power or to fine-tune signal levels. Impedance matching to the 50 Ω reference standard is inherent across all ports to minimize reflections and ensure measurement fidelity.

Overall integration of the signal generator and test set yields precise control, with source power accuracy of ±0.1 dB and frequency settling times under 1 ms, supporting rapid sweeps and repeatable results in high-throughput testing environments.[36]

Receiver and processing

The receiver in a network analyzer captures the reflected and transmitted signals from the device under test (DUT) through multi-channel intermediate frequency (IF) detection systems. These receivers downconvert the high-frequency RF signals to a lower IF using superheterodyne architectures with local oscillators to preserve signal integrity for subsequent digitization. The IF bandwidth for filtering is typically 1 Hz to 100 kHz. This process enables simultaneous extraction of magnitude and phase information across multiple receiver channels (e.g., reference, reflection, and transmission paths), supporting vector measurements essential for accurate characterization.[2][38]

Receiver performance is characterized by a dynamic range of 100-140 dB, which defines the span between the maximum receivable signal power (often limited by compression points around +10 to +20 dBm) and the minimum detectable signal, allowing precise quantification of both strong reflections and weak transmissions. The noise floor, typically around -120 dBm at narrow IF bandwidths (e.g., 10 Hz), sets the lower limit and is influenced by thermal noise, receiver gain, and IF filtering; wider bandwidths increase the noise floor but speed up measurements. These specifications ensure reliable detection in applications ranging from low-loss filters to high-attenuation devices.[48][2]

The processing unit serves as the computational backbone, utilizing digital signal processors (DSPs) to refine raw receiver data into usable measurements. DSP algorithms perform time-domain averaging (e.g., 1 to 1000 sweeps) to suppress random noise by a factor of √N, where N is the number of averages, improving signal-to-noise ratio without sacrificing sweep speed. Interpolation techniques generate intermediate data points for smoother traces, while built-in error correction applies vector-calibrated models to mitigate directivity, source match, and other systematic errors. Firmware handles advanced trace mathematics, including transformations for Smith chart plotting, which maps complex impedance data into polar coordinates for intuitive visualization of reflection coefficients.[48][38][2]

The display interface integrates these processed results into real-time graphical formats, plotting logarithmic magnitude (dB scale for gain/loss), unwrapped phase (degrees for delay analysis), and derived metrics like group delay (time derivative of phase). These visualizations update dynamically during sweeps, aiding immediate DUT assessment, with customizable overlays for multi-trace comparisons. Processed data supports export in industry-standard formats, such as Touchstone (.s2p files for two-port S-parameters), enabling seamless integration with circuit simulators like Keysight ADS or Ansys HFSS.[49]

Contemporary network analyzers incorporate high-performance embedded controllers with clock speeds reaching 1 GHz, leveraging multi-core architectures and optimized firmware to achieve sweep rates exceeding 1000 points per second. This enables full 100,000-point frequency sweeps across multi-GHz spans in mere seconds, even with averaging and error correction applied, drastically reducing test times in production environments while maintaining resolution for fine details like resonances.[50][51]

S-parameter characterization

Scattering parameters, commonly known as S-parameters, describe the linear behavior of electrical networks at high frequencies by relating incident and reflected traveling waves at the network ports. Introduced by Kaneyuki Kurokawa, these parameters are defined in terms of power waves, where the incident wave aia_iai​ and reflected wave bib_ibi​ at port iii are normalized such that their magnitudes squared represent normalized power. For a two-port network, the input reflection coefficient is S11=b1/a1S_{11} = b_1 / a_1S11​=b1​/a1​ when port 2 is terminated in a matched load, quantifying the fraction of power reflected back to port 1. The forward transmission coefficient is S21=b2/a1S_{21} = b_2 / a_1S21​=b2​/a1​, representing the transmitted power from port 1 to port 2 under the same conditions.[52][52][52]

In matrix form, S-parameters for an NNN-port network are expressed as the scattering matrix [S][S][S], where the vector of outgoing waves relatestothevectorofincomingwavesrelates to the vector of incoming wavesrelatestothevectorofincomingwaves by =[S]= [S]=[S], or equivalently [S]=−1[S] = ^{-1}[S]=−1. This formulation assumes the network is linear and operates under small-signal conditions, making it ideal for characterizing passive and active RF/microwave devices.[52][53]

To measure the full two-port S-parameter matrix with a vector network analyzer (VNA), forward and reverse sweeps are performed across the frequency range of interest. In the forward configuration, the VNA source excites port 1 while port 2 is matched, yielding S11S_{11}S11​ and S21S_{21}S21​; the reverse sweep excites port 2 to measure S22S_{22}S22​ and S12S_{12}S12​, ensuring reciprocity (S12=S21S_{12} = S_{21}S12​=S21​) for passive networks. When fixtures such as probes or test adapters are involved, de-embedding algorithms mathematically remove their parasitic effects from the raw measurements to isolate the device under test (DUT) S-parameters, often using models derived from open, short, and thru standards.[53][53][54]

Interpretation of S-parameters provides key insights into network performance. The magnitude ∣S11∣|S_{11}|∣S11​∣ determines the voltage standing wave ratio (VSWR) via VSWR=1+∣S11∣1−∣S11∣\text{VSWR} = \frac{1 + |S_{11}|}{1 - |S_{11}|}VSWR=1−∣S11​∣1+∣S11​∣​, where values close to 1 indicate good impedance matching and minimal reflections. Insertion loss, a measure of signal attenuation, is derived from the transmission parameter as −20log⁡10∣S21∣-20 \log_{10} |S_{21}|−20log10​∣S21​∣ in decibels, with lower values signifying efficient power transfer. Measurement uncertainties, particularly from trace noise in the VNA's receivers, degrade accuracy for low-magnitude parameters, introducing phase and amplitude errors that can exceed 1 dB in noisy conditions.[55][55][56]

For differential circuits, such as balanced transmission lines in high-speed interconnects, balanced or mixed-mode S-parameters extend the conventional formulation by decomposing signals into differential and common-mode components. These parameters, defined using even- and odd-mode impedances, include terms like differential reflection Sdd11S_{dd11}Sdd11​ and mode conversion Sdc21S_{dc21}Sdc21​, facilitating analysis of crosstalk and imbalance in symmetric structures.

Noise figure and gain measurements

Noise figure quantifies the degradation in the signal-to-noise ratio (SNR) caused by a device under test (DUT), defined as the noise factor F=SNRinSNRoutF = \frac{\mathrm{SNR_{in}}}{\mathrm{SNR_{out}}}F=SNRout​SNRin​​, where noise figure NF=10log⁡10F\mathrm{NF} = 10 \log_{10} FNF=10log10​F is expressed in decibels.[57] This metric is essential for evaluating the noise performance of amplifiers and receivers in RF and microwave systems.[58]

The Y-factor method is a widely adopted technique for measuring noise figure, involving the use of hot and cold noise sources to determine the noise parameters of the DUT.[57] In this approach, the Y-factor is calculated as the ratio of the output noise power with the hot noise source (NhN_hNh​) to that with the cold noise source (NcN_cNc​), where Y=NhNcY = \frac{N_h}{N_c}Y=Nc​Nh​​, and the noise factor is derived as F=ENRY−1F = \frac{\mathrm{ENR}}{Y - 1}F=Y−1ENR​ with ENR being the excess noise ratio of the source.[59] Measurements are typically performed at spot frequencies across a swept range, contrasting with broadband methods that average over wider bandwidths for overall system assessment.[57]

Gain measurements in the context of noise figure characterization often involve the associated gain GaG_aGa​, derived from the magnitude squared of the transmission parameter ∣S21∣2|S_{21}|^2∣S21​∣2, with corrections applied for input and output mismatches to ensure accuracy.[60]

Vector network analyzers (VNAs) equipped with a noise figure option utilize the receiver as a noise receiver, enabling direct noise power measurements without additional hardware in some configurations.[61] These systems achieve high accuracy, with typical uncertainties of ±0.1 dB for noise figure measurements up to 50 GHz when using low-noise receivers and proper calibration.[62]

The effective noise temperature TeT_eTe​ provides an equivalent representation of the noise added by the DUT, given by Te=T0(F−1)T_e = T_0 (F - 1)Te​=T0​(F−1), where T0=290T_0 = 290T0​=290 K is the standard reference temperature.[63] This parameter is particularly useful for cryogenic or low-noise applications, linking noise figure to thermal noise equivalents.[57]

Error models and correction methods

In vector network analyzer (VNA) measurements, systematic errors arise from imperfections in the instrument's signal path, including reflections, losses, and crosstalk, which can significantly distort the characterization of devices under test. The 12-term error model addresses these for two-port measurements by representing the forward and reverse signal paths separately, capturing twelve independent error coefficients that account for the analyzer's non-ideal behavior. These terms include directivity (leakage from incident to reflected signal paths, denoted as e00e_{00}e00​ and e33e_{33}e33​), source match (reflection at the source port, e11e_{11}e11​ and e44e_{44}e44​), load match (reflection at the load port, e22e_{22}e22​ and e55e_{55}e55​), reflection tracking (phase and magnitude errors in reflection measurements, e10e01e_{10}e_{01}e10​e01​ and e43e34e_{43}e_{34}e43​e34​), transmission tracking (errors in forward and reverse transmission, e10e32e_{10}e_{32}e10​e32​ and e23e01e_{23}e_{01}e23​e01​), and isolation or crosstalk (leakage terms, e30e_{30}e30​ and e03e_{03}e03​). The model conceptualizes the signal flow through directional couplers, switches, and mixers, where incident waves (a1,a2a_1, a_2a1​,a2​) are separated from reflected/transmitted waves (b1,b2b_1, b_2b1​,b2​) before down-conversion and digitization, with error adapters inserted mathematically between the VNA ports and an ideal device to model deviations.[64]

For one-port measurements, the Short-Open-Load (SOL) calibration simplifies the model to three terms—directivity (D=e00D = e_{00}D=e00​), source match (M=e11M = e_{11}M=e11​), and reflection tracking—using three known standards to solve for the coefficients. The corrected reflection coefficient Γ\GammaΓ is renormalized to ideal conditions via the equation:

where Γm\Gamma_mΓm​ is the measured reflection coefficient, derived by solving the system of equations from the standards' known responses to isolate the device under test (DUT) from systematic errors. This approach yields high accuracy for reflection-only characterizations, such as antenna or filter inputs. For two-port extensions, the full 12-term model incorporates these one-port corrections into a bilinear transformation framework, enabling comprehensive S-parameter renormalization across both directions.[64][65]

The Through-Reflect-Line (TRL) calibration method enhances two-port accuracy, particularly for broadband applications, by employing transmission line standards instead of precise loads, reducing dependency on connector-specific artifacts. Developed for non-coaxial environments like on-wafer or in-fixture testing, TRL uses a zero-length thru, a high-reflection standard (e.g., short or open), and a precisely known-length transmission line to determine the 7- or 8-term error model (a subset of the 12-term for reciprocal networks), solving for propagation constants and reference impedances over wide frequency bands. This technique achieves superior performance up to millimeter-wave frequencies (e.g., beyond 110 GHz with appropriate lines), as the line standards' characteristic impedance and electrical length provide robust renormalization without needing low-loss loads, minimizing uncertainties in high-frequency scattering parameters. As of 2025, specialized calibration kits, such as Keysight's 85065A for up to 250 GHz, and photonic frequency extenders enabling measurements to 520 GHz, further extend these capabilities for THz applications.[65][66]

After calibration, residual uncertainties propagate through the error terms, typically quantified using the root-sum-square (RSS) method to combine contributions from directivity, match, and tracking residuals, along with noise and drift. For well-calibrated systems, magnitude uncertainties are often on the order of 0.05 dB, while phase uncertainties reach about 1° across microwave bands, though these degrade at higher frequencies or lower signal levels due to covariance in the error model coefficients. Automated software implementations facilitate these corrections by iteratively solving the model equations during measurement sweeps.[67][68]

Automated fixtures and procedures

Automated fixtures and procedures enhance the efficiency and repeatability of calibration in vector network analyzers (VNAs) by integrating hardware standards and software-driven workflows that minimize manual intervention. These tools address systematic errors through predefined sequences, enabling precise measurements in research, development, and production environments.[69]

Mechanical calibration kits, such as Short-Open-Load-Through (SOLT) types, form the foundation of traditional automated setups. These kits typically include a short circuit standard for zero reflection, an open circuit for infinite reflection, a matched load (usually 50 Ω) for absorption, and a through connection for transmission characterization; precision airlines—coaxial sections with known lengths—are often incorporated to support verification or TRL extensions.[69][70] Such kits connect sequentially via automated prompts, ensuring consistent torque and alignment to reduce variability. Electronic calibration (ECal) kits offer faster switching by replacing mechanical standards with solid-state switches that electronically simulate shorts, opens, loads, and through paths, enabling one-connection calibration from DC to 67 GHz with reduced setup time—often under a minute per port—compared to multi-step mechanical processes.[71][72]

Software-guided procedures automate the execution of these kits, particularly for 12-term full two-port error correction models that account for forward and reverse directivity, source/load match, reflection and transmission tracking, and isolation. Tools like the Keysight Calibration Wizard provide step-by-step interfaces that detect device-under-test (DUT) connectors, recommend compatible kits, and sequence measurements for SOLT or ECal methods, supporting flexible port selection and thru configurations.[73][74] For multiport VNAs, these wizards extend to N-port calibrations by automating pairwise thru connections and error term computations, streamlining setups for devices with 4 or more ports.[73]

Test fixture simulators (TFS) automate de-embedding of PCB launch effects, where measured S-parameters include unwanted fixture parasitics like connectors and traces. TFS software models the fixture using S-parameter files or electromagnetic simulations, then mathematically removes its contribution to isolate DUT responses, improving accuracy for embedded components up to millimeter-wave frequencies.[75] In high-volume production, automated robotic systems handle calibration and probing, such as six-axis manipulators for on-wafer RF testing that align probes and execute VNA sweeps with sub-micron repeatability, reducing cycle times for batch characterization.[76]

Traceable standards, like 7-mm coaxial kits compliant with IEEE 287, ensure metrological integrity through artifacts verified against national references. These kits include precision shorts, opens, loads, and optional sliding loads for broadband use up to 18 GHz. Manufacturers recommend annual verification cycles to maintain performance, involving measurements against check standards to confirm return loss and insertion loss within specifications.[77][78]

Verification kits and protocols

Verification kits for vector network analyzers (VNAs) consist of precisely characterized standards designed to independently validate the accuracy and performance of the instrument after calibration. These kits typically include known-good components such as offset loads, flush shorts, and broadband terminations, which allow users to measure residual errors in parameters like directivity, source match, and load match by comparing obtained results against provided certificate data. For instance, offset shorts with varying electrical lengths enable the evaluation of phase tracking linearity across the frequency range, while flush shorts help assess reflection coefficient accuracy at specific points.[79]

A representative example is the Keysight 85058V verification kit, which supports frequencies up to 67 GHz and incorporates airline standards and adapters for coaxial and waveguide interfaces to confirm system stability and error correction efficacy.[80] These kits are essential for detecting subtle imperfections, such as imperfect connections or drift in the VNA's internal references, ensuring measurements remain reliable for device characterization.

Verification protocols involve systematic testing procedures to quantify uncertainty and repeatability, often employing software tools like the METAS VNA Tools II for computing S-parameter uncertainties based on propagation of error models from the kit measurements. Key protocols include sweep-to-sweep repeatability assessments, where multiple consecutive sweeps on a stable standard, such as a matched load, must exhibit magnitude variations below 0.01 dB and phase variations below 0.2 degrees to confirm short-term stability. Uncertainty evaluation follows established frameworks, integrating contributions from random effects like noise and systematic effects like residual directivity, with tools providing propagation calculations compliant with metrology guidelines.[81]

Critical metrics in these protocols include effective directivity exceeding 40 dB, which verifies minimal signal leakage and high isolation in the test setup, and reflection tracking linearity, ensuring the VNA accurately reproduces known reflection responses without distortion. If verification indicates deviations, such as reduced directivity below specifications, post-verification adjustments like connector re-torquing or partial recalibration are applied to restore performance. At mm-wave frequencies above 100 GHz, verification faces unique challenges due to increased waveguide losses and potential higher-order mode excitation; overmoded waveguides are employed in these protocols to introduce and quantify multi-mode propagation effects, allowing assessment of mode purity and coupling in high-frequency standards.[82][83]

Traceability to international standards

Traceability of network analyzer measurements to international standards ensures the reliability and comparability of results across global laboratories by establishing an unbroken chain from primary references to end-user calibrations. This chain originates with primary standards at national metrology institutes, such as the National Institute of Standards and Technology (NIST) in the United States and the Physikalisch-Technische Bundesanstalt (PTB) in Germany, which provide the foundational accuracy for vector network analyzer (VNA) S-parameter determinations. For example, NIST utilizes multiline thru-reflect-line (TRL) calibrations with standards of traceable dimensions—such as line lengths from 14.999 mm to 29.981 mm with uncertainties of ±0.009 mm—to characterize electronic calibration units, propagating uncertainties via the NIST Microwave Uncertainty Framework for 95% confidence bounds. PTB has developed a similar traceability path for on-wafer measurements up to 110 GHz, employing multiline TRL on custom alumina wafers to characterize commercial impedance standard substrates like the GGB CS-5, ensuring single-mode propagation and alignment with SI units through documented material properties and low-loss approximations. Cryogenic attenuators serve as key primary standards in these chains, enabling low-noise power measurements at millikelvin temperatures to support high-frequency accuracy while minimizing thermal loading.[84][85][86]

Commercial calibration kits and verification standards link to these primary references through laboratories accredited under ISO/IEC 17025, which mandates an uninterrupted calibration hierarchy with quantified uncertainties for each step, including statistical correlations and sensitivity analyses. This accreditation facilitates the transfer of metrological reliability from national institutes to industry, where VNAs are calibrated using traceable kits to achieve measurement uncertainties typically below 0.1 dB in magnitude and 1° in phase up to 50 GHz. Developments in vector calibration standards since the 1990s have strengthened this chain, incorporating advanced error models and electronic units to replace mechanical kits for more robust, repeatable characterizations.[87][88]

International bodies further standardize these practices to promote global consistency. The International Electrotechnical Commission (IEC) standard 62153-4-3 outlines the triaxial method for assessing screening attenuation via surface transfer impedance of metallic cable shields, employing VNAs with S-parameter test sets or power splitters for measurements up to 100 MHz, with calibration at connector reference planes to ensure traceability. The IEEE Standard 370-2020 addresses de-embedding of fixture effects in printed circuit board interconnects up to 50 GHz, recommending test structures like lines and Beatty standards alongside quality metrics for S-parameter validation, directly supporting vector calibration in high-speed RF applications. These standards, integrated into accredited workflows, enable comparable results for electromagnetic compatibility and signal integrity testing worldwide.[89][90]

RF and microwave device testing

Network analyzers play a crucial role in characterizing RF and microwave components, enabling precise evaluation of their linear performance parameters such as scattering parameters (S-parameters), gain, and impedance. For amplifiers, vector network analyzers (VNAs) measure key metrics including gain flatness across frequency bands and input/output return loss to ensure minimal signal reflections and stable amplification. These measurements help identify deviations from ideal behavior, such as ripple in gain response, which can degrade system performance in high-frequency applications.[94]

Filters, essential for signal selectivity in RF systems, are tested using VNAs to assess bandpass response, insertion loss, and out-of-band rejection. By sweeping frequencies and analyzing magnitude and phase of transmitted signals, engineers verify filter alignment with design specifications, ensuring effective suppression of unwanted frequencies while passing desired bands with minimal attenuation.[94] Antennas, particularly phased arrays for modern systems, undergo VNA-based testing in compact antenna test ranges (CATR) to evaluate radiation patterns, beam steering accuracy, and S-parameters under far-field conditions. This setup simulates real-world propagation, allowing measurement of gain, directivity, and phase alignment without anechoic chamber limitations.[95]

In the telecommunications industry, VNAs support testing of 5G base stations across sub-6 GHz (FR1) and mm-wave (FR2) bands, where they characterize components like duplexers and transceivers for low return loss and high isolation to meet stringent throughput requirements. For satellite communications, Ka-band (26.5-40 GHz) S-parameter measurements using extended-range VNAs ensure reliable link performance in transceivers and feeds, verifying low insertion loss and phase stability critical for high-data-rate uplinks and downlinks.[96][97]

The typical workflow for RF device development integrates VNA measurements with simulation tools like Keysight's Advanced Design System (ADS), where measured S-parameter data is imported to validate and refine circuit models before hardware iteration. In production, automated test equipment (ATE) incorporating VNAs performs high-volume screening with limit lines for pass/fail criteria on parameters like gain and return loss, facilitating yield analysis by tracking defect rates and optimizing manufacturing processes.[98][98]

A representative case involves testing power amplifier (PA) modules in smartphone front-end modules, where VNAs assess linear gain and return loss alongside brief evaluations of intermodulation distortion (IMD) and efficiency under multi-tone stimuli to ensure compliance with mobile network standards.[99][100]

Nonlinear and large-signal analysis

Network analyzers equipped with nonlinear vector network analyzer (NVNA) capabilities enable the characterization of device nonlinearities through harmonic balance simulations, which solve for steady-state responses under large-signal conditions by balancing harmonic components in the frequency domain.[101] This approach is particularly useful for measuring third-order intermodulation distortion (IMD3), where two-tone signals are applied to assess distortion products that arise from nonlinear mixing, providing insights into amplifier linearity under multi-tone excitation. For instance, NVNA systems can capture vector-corrected measurements up to 50 GHz, allowing accurate extraction of IMD3 levels that traditional scalar methods overlook.[102]

X-parameters extend the linear S-parameter framework to nonlinear regimes, offering a measurement-based black-box model that describes incident and reflected waves under large-signal drive, including harmonic generation and compression effects.[103] Developed as a device-independent representation, X-parameters facilitate behavioral modeling of active components like transistors without requiring internal circuit details, enabling faster simulation integration in design tools.[104] They are extracted using NVNA setups, where multi-tone excitations reveal scattering behavior across power levels, supporting applications in predicting nonlinear interactions in cascaded systems.[105]

In large-signal analysis, hot S-parameters measure scattering characteristics under biased and powered conditions, capturing real operating impedances and gain compression that differ from cold (low-power) measurements.[101] This technique uses vector-corrected receivers to quantify changes in S21 and output power as input drive increases, essential for evaluating amplifier performance near saturation.[16] Load-pull measurements complement this by systematically varying source and load impedances to map contours of optimum performance metrics, such as maximum output power or power-added efficiency (PAE).[106] For example, tuner-based load-pull systems generate constant-gain or constant-PA circles on the Smith chart, guiding matching network design for devices like GaN HEMTs.[107]

Modern network analyzers like the Keysight PNA-X incorporate NVNA software options to perform these nonlinear assessments directly, integrating source power calibration and harmonic receivers for comprehensive large-signal characterization up to millimeter-wave frequencies.[17] The nonlinear noise option extends this to evaluate noise contributions under compressed operation, combining vector signal analysis with low-noise receivers for metrics like noise figure degradation at high power.[101] For thermal-sensitive devices such as GaN HEMTs, pulsed IV and RF measurements mitigate self-heating effects by applying short pulses (e.g., 100 ns to 1 µs duty cycles), enabling extraction of intrinsic device parameters and thermal resistance without thermal runaway. This pulsed approach reveals trap-related dispersions in I-V curves, informing thermal management strategies like substrate selection for high-power operation.