How to choose a Vector Network Analyzer? A Complete Guide to Selection

1. Introduction

The vector network analyzer (VNA) is arguably the single most important instrument in RF, microwave, and signal integrity engineering. Whether you’re characterizing a filter, verifying an antenna’s match, qualifying a high-speed PCB interconnect, or profiling a pulsed radar transmitter, the VNA sits at the center of the measurement chain. Unlike a spectrum analyzer, which tells you what frequencies are present, or a power meter, which tells you how much energy is there, a VNA tells you how a device transforms a signal — in both amplitude and phase — across frequency. That combination of amplitude and phase is what makes it possible to model, de-embed, and predict the behavior of components and systems with real engineering precision.

This article is written for RF and microwave engineers, signal integrity (SI) engineers, radar engineers, R&D teams, production test engineers, and university labs. It is meant to be technical but readable: we favor clear explanations over exhaustive derivations, and we point to more rigorous treatments where they matter — from textbooks such as Pozar’s Microwave Engineering to application notes from the major test equipment vendors.

Here’s the map of what follows: we start from first principles (N-port network theory and the matrices used to characterize them), move through the distinction between scalar, vector, and nonlinear network analyzers, explain how a modern VNA actually works internally, then walk through S-parameter measurements on passive networks, de-embedding and signal integrity topics, dynamic range, and finally the advanced capabilities — pulsed radar measurements and mixer characterization — that separate an entry-level VNA from a fully equipped microwave test bench.

💡 Note: At the end of this article you’ll find our interactive online selection tool, which automatically recommends the right VNA configuration and options based on your measurement requirements.

---

2. Circuit Theory: N-Port Networks

2.1 Definition of an N-Port

Before we can talk meaningfully about S-parameters, we need to agree on what a „port“ actually is. A port is a pair of terminals at which a well-defined voltage and current can be measured, with the current entering one terminal equal to the current leaving the other (the port condition). An N-port network is simply the generalization of a two-terminal (one-port) or four-terminal (two-port) network to a device with N such ports — think of a splitter with one input and three outputs, or a multi-channel PCB backplane connector.

This entire framework rests on a critical assumption: the network must be linear and time-invariant (LTI), and must operate in a quasi-stationary regime at the frequencies of interest (i.e., the physical dimensions of the network are such that voltage and current are well-defined at each port, without significant radiation or higher-order mode effects at the port boundary). This LTI assumption is not a minor technicality — it is the foundation that every characterization matrix described below depends on. Violate it (for example, by driving an amplifier into compression) and the entire matrix formalism breaks down, requiring the nonlinear extensions discussed in Section 3.3.

For a rigorous treatment of N-port theory, see David Pozar’s Microwave Engineering (University of Massachusetts Amherst), the lecture notes of the Chair of High-Frequency Engineering at the Technical University of Munich, and the microwave engineering course materials at ETH Zürich.

2.2 Characterization Matrices

Several matrix formalisms exist to characterize a linear N-port network, each suited to different measurement conditions and circuit topologies:

Matrix	Symbol	Stimulus	Response	Measurement Condition	Typical Application
Impedance	[Z]	Currents Ij	Voltages Vi	Unexcited ports open-circuited	Low-frequency circuits
Admittance	[Y]	Voltages Vj	Currents Ii	Unexcited ports short-circuited	Networks in parallel
Hybrid H	[H]	I1, V2	V1, I2	Mixed open/short	BJT transistors
Hybrid G	[G]	V1, I2	I1, V2	Mixed open/short	FET transistors
ABCD (chain)	[T]	V2, I2	V1, I1	Cascade of two-ports	Cascaded networks
Scattering	[S]	Incident waves aj	Reflected waves bi	Matched loads Z0	RF/microwave, VNA

The classic low-frequency matrices relate voltages and currents directly: V = Z·I and I = Y·V. At microwave frequencies, however, these definitions become impractical: creating a true open or short circuit at a port — required to measure Z or Y parameters — is extremely difficult once wavelengths become comparable to circuit dimensions, and stray reactance dominates the result.

This is precisely why the scattering matrix [S] is preferred at high frequencies. Instead of open or short terminations, S-parameters are defined in terms of power waves — incident waves a_i and reflected waves b_i — measured with every port terminated in the reference impedance Z₀ (typically 50 Ω), which is a condition that’s easy to realize accurately at any frequency.

Two properties of the S-matrix are worth internalizing early:

• Reciprocity: for a reciprocal network, S^T = S (the matrix is symmetric — S_ij = S_ji).
• Losslessness (energy conservation): for a lossless network, S^†·S = I (the matrix is unitary — no net power is absorbed).

For the full derivation, see Pozar’s Microwave Engineering, Chapter 4; Collin’s Foundations for Microwave Engineering; and the lecture notes of Prof. Peter Russer at the Technical University of Munich.

2.3 Conditions of Validity

Every matrix described above depends on the LTI assumption holding true. Three practical consequences follow:

• Linearity and time-invariance are not optional — they are the precondition for every matrix (Z, Y, H, G, ABCD, S) to be meaningful.
• Narrowband stationarity: S-parameters are, strictly speaking, valid in a narrow band around the frequency at which they were measured; a swept measurement stitches together many such narrowband snapshots.
• Nonlinearity breaks the model: harmonics and intermodulation products generated by a nonlinear device under test violate the linear model outright, which is exactly why X-parameters (and related Volterra-series approaches) were developed as a nonlinear generalization — see Section 3.3.

Further reading: the work of Dr. Jan Verspecht on large-signal network analysis (available via arXiv and IEEE Microwave Magazine), NIST technical notes on nonlinear measurement, and research on advanced calibration methods from institutions such as the Max Planck Institute for the Physics of Complex Systems.

---

3. Scalar, Vector, and Nonlinear VNAs

3.1 Scalar Network Analyzer (SNA)

A scalar network analyzer measures only the magnitude of voltage or power ratios — for example |S₂₁| or |S₁₁| — with no coherent phase reference. Historically, SNAs were built around log/diode detectors and were the workhorse for broadband insertion-loss and return-loss measurements before coherent receivers became affordable.

The limitations are significant: no phase information means no group delay, no time-domain transforms, and no sophisticated de-embedding. For anything beyond a basic pass/fail loss check, a scalar instrument simply doesn’t have the information needed.

Further reading: introductory VNA primers from Tektronix and Anritsu, and NI’s „Overview of Vector Network Analyzers.“

3.2 Vector Network Analyzer (VNA)

A true vector network analyzer measures both the amplitude and phase of incident, reflected, and transmitted waves. This requires a coherent architecture: receivers phase-locked to the source, capable of measuring complex ratios rather than just magnitudes.

The payoff is substantial. With complex S-parameters and a stable phase reference, a VNA enables:

• Sophisticated impedance matching design
• Advanced de-embedding of test fixtures and connectors
• Mixed-mode S-parameters for differential (balanced) channels
• Time-domain analysis (TDR-like) via inverse Fourier transform of the frequency-domain data

Further reading: „Fundamentals of Vector Network Analysis“ primers and NI’s VNA documentation.

3.3 Nonlinear Vector Network Analyzer (NVNA)

The NVNA extends vector network analysis into large-signal, nonlinear operating conditions. It measures harmonics, intermodulation products, and fully calibrated waveforms (amplitude and phase, referenced to absolute quantities) under real operating conditions — something a standard small-signal VNA simply cannot do.

This is where X-parameters come in: a nonlinear generalization of S-parameters that reduces to ordinary S-parameters in the small-signal limit, but remains valid under large-signal drive. Typical NVNA applications include:

• Power amplifier characterization under real load conditions
• Mixer characterization at realistic drive levels
• Behavioral modeling for circuit simulators such as Keysight’s Advanced Design System (ADS)

Further reading: Keysight’s NVNA product literature, the Wikipedia entry on X-parameters, and the ECE 451 course materials at the University of Illinois.

---

4. Passive vs. Active Networks

4.1 Passive Networks

A passive network cannot generate net power: formally, Re[S^†S] ≤ I — there is no power gain anywhere in the matrix. Filters, transformers, transmission lines, antennas, connectors, and PCB traces all fall into this category.

Passive networks are typically reciprocal (S_ij = S_ji), with the notable exception of non-reciprocal materials and structures such as isolators and circulators, which rely on ferrite materials biased by a magnetic field to break reciprocity.

Further reading: Pozar, Chapter 4; Collin’s Foundations for Microwave Engineering.

4.2 Active Networks

Active networks can amplify — power gain greater than 1 — and are not necessarily reciprocal. Amplifiers, active mixers, VCOs, oscillators, and LNAs are all active networks.

Measuring active devices introduces several practical complications:

• Oscillation risk: an active device presented with certain load impedances can become unstable and oscillate during the measurement.
• Controlled biasing: DC bias must be applied and controlled carefully, often synchronized with the measurement sweep.
• Linearity breakdown at high drive: push an active device hard enough and the small-signal S-parameter model simply stops applying.

Additional figures of merit become relevant for active devices: Noise Figure (NF), gain compression (P1dB), IIP3, and — for genuinely nonlinear characterization — X-parameters.

Further reading: Agilent/Keysight Application Note AN 1287-4; the work of Prof. Frank Ellinger at TU Dresden; Pozar, Chapter 11.

---

5. How a Modern VNA Works: Measurement Principles

5.1 Basic Architecture

At a block-diagram level, a modern VNA consists of a synthesized source, a reference splitter, directional couplers (to separate incident from reflected waves), heterodyne receivers, and analog-to-digital converters. Crucially, the instrument is ratiometric: it measures ratios of signals, not absolute values, which cancels out variations in the source itself.

Because real hardware has imperfections — finite directivity in the couplers, mismatch, cable loss, frequency response errors — a VNA must be calibrated before a trustworthy measurement can be made. Standard calibration techniques include SOLT (Short-Open-Load-Thru), TRL (Thru-Reflect-Line), and SOLR (Short-Open-Load-Reciprocal-Thru), all of which correct for systematic errors using a well-established 12-term error model for two-port measurements.

Further reading: Keysight Application Note AN 1287-3; Rohde & Schwarz Application Note 1EZ29; IEEE Transactions on Instrumentation and Measurement.

5.2 Swept-CW Method (Pure Sinusoidal Frequency)

The dominant measurement method in modern VNAs is swept continuous-wave (CW): a synthesized source steps through frequencies with a defined step size (Δf), and amplitude/phase are measured at each point.

A critical, often underappreciated detail: the phase noise of the source directly impacts the receiver’s noise floor, since phase noise sidebands leak energy away from the carrier and raise the effective noise level near it. There is also a direct trade-off between IF bandwidth (IFBW), sweep time, and dynamic range: narrowing the IFBW improves the noise floor but slows down the sweep.

Top-tier instruments (e.g., Rohde & Schwarz ZVA and Keysight PNA-X families) typically specify SSB phase noise better than −120 dBc/Hz at a 10 kHz offset.

Further reading: Rohde & Schwarz ZVA datasheet; Keysight PNA-X datasheet; published research on advances in phase noise measurement techniques.

5.3 Pulsed Method (TDR / Broadband Pulse)

The alternative philosophy is impulsive measurement, based on a simple duality: the Fourier transform of an ideal impulse δ(t) is a constant — meaning an ideal impulse contains all frequencies simultaneously. This is the principle behind classic time-domain reflectometry (TDR).

The advantages are speed and intuition: a single-shot measurement gives a direct time-domain view and lets you localize discontinuities immediately. The trade-offs: dynamic range is generally lower than swept-CW, and frequency resolution is limited by the length of the time record.

TDR and VNA-based time-domain analysis (via IFFT of swept S-parameter data) represent two related but distinct philosophies for looking at the same physical reality — one measures directly in time, the other synthesizes a time-domain view from frequency-domain data.

Further reading: Keysight Application Note „Time Domain Analysis Using a Network Analyzer“; IEEE MTT-S publications.

---

6. Measurements on Passive Networks: S-Parameters and Use Cases

6.1 S-Parameters — Definitions and Interpretation

• S₁₁: port 1 reflection coefficient → return loss, VSWR
• S₂₁: forward transmission → insertion loss (or gain)
• S₁₂: reverse transmission → isolation
• S₂₂: port 2 reflection coefficient

For N-port devices — multi-way connectors, splitters, combiners — the same logic extends to a full N × N matrix, with every off-diagonal term representing transmission or coupling between a specific pair of ports.

Further reading: „S-Parameters and Two-Port Measurements,“ part of the Keysight University Lab Series.

6.2 Use Cases for Passive Networks

Device	Key Measurement	Relevant S-Parameters	Typical Instrument
RF/microwave filters	Passband, out-of-band rejection, ripple	S21, S11	Standard VNA
Connectors/cables	Return loss, insertion loss	S11, S21	Handheld VNA
Antennas	S11 = VSWR, efficiency	S11	VNA + anechoic chamber
Splitters/combiners	Isolation, balance	All Sij	4-port VNA
PCB/backplane	Impedance, crosstalk, signal integrity	Full N-port matrix	WavePulser, PNA-X
Dielectric materials	Permittivity εr, permeability μr	S11, S21	VNA + material fixture

---

7. De-embedding, Crosstalk, and Eye Diagrams

7.1 De-embedding

In practice, a device under test (DUT) is never connected directly to the VNA’s reference plane — there’s always a test fixture, connectors, or PCB traces in between, and their characteristics inevitably „contaminate“ the raw measurement. De-embedding is the mathematical process of removing those unwanted effects to recover the true response of the DUT alone.

Common techniques include SOLT- and TRL-based calibration performed as close as possible to the DUT, mathematical port-shifting, and the 2×Thru method formalized in the IEEE 370 standard. De-embedding is essential for reliable measurements on PCBs and non-coaxial fixtures, where you simply cannot calibrate directly at the DUT’s terminals.

Further reading: Keysight Application Note „The ABCs of De-Embedding“; the IEEE P370 standard; „Test Fixture De-embedding 101.“

7.2 Crosstalk

Crosstalk is unwanted coupling between adjacent transmission lines, commonly categorized as near-end crosstalk (NEXT) and far-end crosstalk (FEXT). On a VNA, crosstalk is measured directly as the off-diagonal S-parameters of an N-port matrix.

Crosstalk limits are formally specified by standards bodies such as IEEE 802.3 (Ethernet), JEDEC, and PCI-SIG.

Further reading: Signal Integrity Journal; IPC-TM-650 test methods.

7.3 Eye Diagrams and the Limits of the Classic VNA

An eye diagram is, by nature, a time-domain measurement on a digital signal — built by overlaying many bit periods of a pseudo-random bit sequence (PRBS) on top of each other. A standard VNA, however, measures in CW and cannot directly generate or apply a PRBS pattern.

The workaround: hybrid instruments — such as Teledyne LeCroy’s WavePulser 40iX — that synthesize an eye diagram from S-parameter data using an inverse FFT, effectively predicting what a real digital eye would look like based purely on the linear frequency response of the channel.

Further reading: Teledyne LeCroy WavePulser product literature; Keysight PLTS software documentation; DesignCon conference papers on S-parameter-based eye synthesis.

---

8. VNA Dynamic Range: What It Is, Why It Matters

8.1 Definition of Dynamic Range

Dynamic range is defined as the maximum source power minus the receiver’s noise floor, expressed in dB. Professional VNAs typically achieve 100–160 dB of dynamic range depending on frequency.

Dynamic range and IF bandwidth (IFBW) are directly linked: reducing IFBW by a factor of 10 improves the noise floor by roughly 10 dB (following a √10 relationship), at the cost of a proportionally slower sweep.

Further reading: „How to Estimate and Enhance VNA Dynamic Range“; NI VNA specifications documentation.

8.2 Problems That Dynamic Range Solves

• Measuring notch filters with high rejection (> 80 dB)
• Characterizing highly matched components (return loss > 40 dB)
• Measuring long, high-attenuation cables
• Measuring antennas in the presence of parasitic reflections

Further reading: AllAboutCircuits tutorials; Rohde & Schwarz application notes on dynamic range.

---

9. Basic vs. Advanced VNA Functions

9.1 What Almost Every VNA Does

• Linear two-port S-parameters (|S₁₁|, |S₂₁|, phase)
• Time-domain transforms via IFFT (with limited fidelity)
• Basic SOLT calibration
• Standard passive-device measurements (filters, cables, antennas, matching networks)

Further reading: „Fundamentals of Vector Network Analysis“ primers.

9.2 What Only Advanced VNAs Do (With Extra Hardware/Software)

• Mixer measurements: frequency-offset mode, a dedicated mixer test set, and application-specific software
• Pulsed S-parameters / radar measurements: integrated pulse modulators, IF gating, and synchronization with device bias
• Noise figure via the cold-source method, two-tone IMD, and X-parameters (NVNA)
• Advanced signal integrity: 4-port mixed-mode measurements, IEEE 370 de-embedding, and synthesized eye diagrams from S-parameter data

Note: mixer measurements and pulsed measurements are not available on an entry-level VNA. They require dedicated architecture (a second source, pulse modulation hardware, a mixer test set), specific software, and additional operator expertise.

Further reading: Keysight’s NVNA product literature; Anritsu white papers on radar measurement; Rohde & Schwarz application notes on mixer testing.

---

10. Pulsed Measurements with a VNA for Radar Applications

Radar devices typically operate at a low duty cycle, meaning the device under test cannot simply be driven continuously (CW) without risking damage or triggering thermal effects that don’t represent real operating conditions. This is the core motivation for pulsed S-parameter measurement: precise synchronization between the VNA’s pulse generator and the DUT’s bias conditions.

Two complementary techniques are used to acquire pulsed data:

• Narrowband IF gating — higher sensitivity, at the cost of measurement speed
• Wideband pulsed acquisition — higher speed, at some cost to sensitivity

Beyond basic pulsed S-parameters, VNA-based radar test setups also support Radar Cross Section (RCS) measurement (VNA plus antenna in an anechoic chamber), pulse profiling, point-to-point (P2P) analysis, and pulse-in-pulse (PIP) measurement for complex radar waveforms.

Representative products with dedicated pulsed-measurement capability include the Anritsu MS4640B with its pulsed measurement option, and the Keysight PNA-X with integrated pulse modulators.

Further reading: Anritsu’s white paper „VNA’s High-Speed Architecture Advances Radar Pulse Measurements“; Electronic Design magazine coverage of pulsed VNA techniques.

---

11. Mixer Measurements

11.1 Key Specifications of a Mixer

Mixers occupy a special place in VNA measurement: they are three-port, frequency-translating devices (RF, LO, and IF ports operate at different frequencies), which means standard two-port S-parameter measurement — implicitly assuming the input and output frequencies are identical — simply doesn’t apply without a frequency-offset measurement mode. The key specifications engineers characterize on a mixer include:

• Conversion loss/gain: CL = 10 log(P_IF / P_RF) — how much power is lost (or gained, for an active mixer) in the frequency translation process.
• Port match (RF, LO, IF): the reflection coefficients S₁₁ and S₂₂ at each port, measured at the appropriate frequency for that port.
• Isolation: how much LO power leaks into the RF port, how much LO power leaks into the IF port, and how much RF power leaks into the IF port (LO→RF, LO→IF, RF→IF). Poor isolation is a common source of spurious signals and self-interference in a receiver chain.
• Intermodulation distortion: characterized via the second- and third-order input intercept points, IIP2 and IIP3, which quantify how quickly unwanted intermodulation products grow relative to the desired signal as input power increases.

Measuring these parameters correctly requires a VNA configured in frequency-offset mode — where the source and receivers are set to different frequencies for the RF and IF ports — together with an LO source (either the VNA’s own internal source or an external synthesizer phase-locked to the VNA’s reference). Full characterization, including conversion compression and two-tone IMD, typically also requires the mixer test set hardware and software options discussed in Section 9.2.

Further reading: Rohde & Schwarz and Keysight application notes on mixer/converter measurements with a VNA; general mixer theory in Pozar’s Microwave Engineering, Chapter 6.

---

12. Conclusion

A vector network analyzer earns its place as the central instrument of RF and microwave test precisely because it captures the one thing scalar and power-based instruments cannot: the complex, frequency-dependent relationship between what goes into a device and what comes out — in both magnitude and phase. That single capability underlies everything from basic filter characterization to advanced signal integrity de-embedding, pulsed radar profiling, and full nonlinear X-parameter modeling of power amplifiers.

Choosing the right VNA — and the right options on top of it — comes down to matching the instrument’s dynamic range, frequency coverage, port count, and measurement modes (CW, pulsed, mixer, nonlinear) to your actual application. An entry-level two-port VNA is more than capable of characterizing passive components and simple matching networks; pulsed radar work, mixer testing, and nonlinear amplifier characterization all demand dedicated hardware, software, and calibration expertise well beyond the base instrument.