Microwave Network Theory and S-Parameters

Introduction and context

Microwave circuits differ fundamentally from low-frequency lumped circuits. At microwave frequencies, the physical dimensions of interconnections are comparable with the wavelength, so voltage and current are functions of position as well as time. Parasitic capacitance, inductance, radiation, discontinuities, and propagation delay are therefore part of the circuit itself. A conventional two-terminal voltage–current description is often inconvenient or even physically misleading.

Microwave network theory provides a systematic way to represent sources, transmission lines, discontinuities, passive components, amplifiers, filters, antennas, and complete subsystems through port variables. Among the available network representations, scattering parameters, or S-parameters, are the most widely used at microwave frequencies because they relate incident and reflected travelling waves. These quantities can be measured directly with directional couplers and vector network analysers without requiring an ideal open circuit or short circuit at the measurement plane.

S-parameters are central to the analysis and design of amplifiers, filters, oscillators, antenna feeds, radar front ends, wireless transceivers, satellite payloads, and millimetre-wave integrated circuits.

Learning objectives

After completing this chapter, the reader should be able to:

1. Explain why travelling-wave variables are preferred at microwave frequencies.
2. Define ports, incident waves, reflected waves, and reference impedances.
3. Derive and interpret the S-parameter matrix of an N-port network.
4. Relate S-parameters to impedance, admittance, and transmission matrices.
5. Analyse reciprocity, passivity, losslessness, symmetry, and matching using S-parameters.
6. Calculate input and output reflection coefficients under arbitrary terminations.
7. Understand cascading, renormalisation, de-embedding, and measurement uncertainty.
8. Interpret measured S-parameter data for practical microwave design.

1. Microwave networks and ports

A microwave network is a physical structure or subsystem connected to the external world through one or more ports. A port is a pair of terminals through which power enters or leaves the network. In a waveguide, a port is usually associated with a propagating mode and a reference plane. In a coaxial or planar circuit, it is associated with a transmission-line cross-section.

A valid port description assumes that the port supports a known set of modes. In the simplest single-mode treatment, each port has one incident wave and one outgoing wave. Multimode structures require mode-indexed waves and may be represented by a larger matrix.

The reference plane is important. Moving it along a lossless transmission line changes the phase of the measured S-parameters, although it does not change the intrinsic network. In a lossy line, both phase and magnitude may change.

2. Travelling-wave description

For a uniform transmission line with characteristic impedance Z₀, the total voltage and current can be written as the sum of forward and reverse waves:

Here, V^+ and V^- are complex voltage-wave amplitudes, γ = α + jβ is the propagation constant in m⁻¹, α is attenuation, and β is phase constant. The signs follow the convention that the forward wave travels in the +z direction.

For a lossless line, α = 0 and γ = jβ. The time-average power carried by a wave is proportional to the squared magnitude of its amplitude. A convenient power-wave definition, for a real reference impedance Z₀ > 0, is

where a is the incident-wave amplitude and b is the outgoing-wave amplitude. Both have units of √W when voltage and current are expressed in SI units. The inverse relations are

For real Z₀, the net power entering the port is

This equation applies to power waves with a real reference impedance. For complex reference impedances, pseudo-wave or power-wave definitions must be stated explicitly because the simple power interpretation changes.

3. Definition of S-parameters

For an N-port network, let a_i denote the incident wave at port i and b_i the outgoing wave. The scattering relation is

or, component by component,

The coefficient S_ij is defined as the outgoing wave at port i caused by an incident wave at port j, with all other ports terminated in their reference impedances. Thus,

The condition a_k = 0 means that port k is matched to its reference impedance; it does not mean that the port is physically disconnected.

For a two-port network,

The four coefficients have standard interpretations:

The magnitude is commonly expressed in decibels:

The corresponding power ratio is 10 log₁₀(|S_ij|²), which is identical to the expression above. The phase is the argument of the complex coefficient, normally reported in degrees or radians.

4. Reflection coefficient and matching

For a load impedance Z_L connected to a line of characteristic impedance Z₀, the voltage reflection coefficient is

The formula assumes a single-mode line and consistent voltage-wave definitions. A matched load has Z_L = Z₀, giving Γ_L = 0. An open circuit has Γ_L = +1, while an ideal short circuit has Γ_L = −1. For a passive load, |Γ_L| ≤ 1.

The voltage standing-wave ratio is

for |Γ| < 1. The return loss is

A small reflection coefficient corresponds to a large return loss and a VSWR close to unity.

For a two-port network terminated by a load reflection coefficient Γ_L, the input reflection coefficient is

Similarly, if the input is terminated by Γ_S, the output reflection coefficient is

These expressions assume consistent reference impedances and a stable linear two-port operating at a specified frequency. They show that the measured input match is generally not just S₁₁ when the output is not matched.

5. Power gain and available power

Let the source have reflection coefficient Γ_S and the load have reflection coefficient Γ_L. For a two-port, the transducer gain is

This is the ratio of power delivered to the load to the maximum available power from the source, under the stated wave convention. It includes source mismatch, network transmission, output mismatch, and multiple reflections.

If reverse transmission is negligible, S₁₂ ≈ 0, the denominator simplifies to

The simplified unilateral approximation is useful for initial design but must not be used when reverse coupling is significant or when feedback affects stability.

6. Physical properties represented by the S-matrix

Reciprocity

A reciprocal network has equal transmission in opposite directions when the port reference conditions are consistently defined:

For a reciprocal two-port, S₁₂ = S₂₁. Passive transmission lines, attenuators, and many passive filters are reciprocal. Ferrite circulators and isolators are deliberately nonreciprocal.

Symmetry

A symmetric two-port has identical behaviour when ports 1 and 2 are interchanged. With equal reference impedances, symmetry implies

Reciprocity and symmetry are different properties: a network can be reciprocal without being symmetric.

Passivity

A passive network cannot deliver more net power than it receives. For a lossless network, conservation of power requires

where [S]ᴴ is the conjugate transpose and [I] is the identity matrix. For a one-port lossless network, this reduces to |S₁₁| = 1. For a passive network, the scattering matrix is contractive, meaning that its singular values do not exceed unity. Individual S-parameters may sometimes appear larger than unity in an active network, but this does not by itself prove instability.

Matched networks

A network is matched at port 1 under the specified termination if the input reflection coefficient is zero. For a two-port measured with port 2 matched, this is S₁₁ = 0. Simultaneous matching at both ports requires S₁₁ = S₂₂ = 0 under matched measurement conditions. It does not imply zero insertion loss or reciprocity.

7. Conversion between network representations

S-parameters are ideal for measurement and high-frequency interpretation, whereas impedance and admittance matrices may be more convenient for circuit synthesis. For equal real reference impedance Z₀, define diagonal matrices [Z₀]. The general conversion from S to impedance is

For equal reference impedance, this becomes

The inverse conversion is

These formulas require the indicated matrix inverses to exist. A singular matrix may occur at resonances or under special ideal conditions and must be handled using limiting or alternative formulations.

For a two-port, transmission or ABCD parameters relate input and output variables, commonly as

The sign convention for I₂ must be declared because alternative conventions exist. ABCD matrices are particularly useful for cascading two-ports: the total ABCD matrix is the matrix product of the individual ABCD matrices in physical order. Direct multiplication of S-matrices is generally invalid except for special interconnections.

8. Cascading and interconnection of microwave networks

When two two-port networks are cascaded, waves emerging from the first network become incident on the second, and internal multiple reflections occur. Conversion to ABCD parameters followed by matrix multiplication is a common approach. A more general method uses signal-flow graphs or the Redheffer star product, which connects subnetwork ports while accounting rigorously for internal wave interactions.

For practical design, the following rules are important:

• S-parameters may be cascaded directly only when the mathematical operation correctly eliminates internal ports.
• The reference impedance at every connected port must be consistent.
• Discontinuity effects at connectors and transitions must be included or de-embedded.
• A two-port model is valid only over the frequency range and mode content for which it was obtained.

9. Measurement with a vector network analyser

A vector network analyser, or VNA, measures magnitude and phase of travelling waves. A typical measurement includes a source, test ports, directional couplers, receivers, calibration standards, and a display or data-processing system.

The most common one-port calibration is open–short–load, while a two-port calibration may use short, open, load, and through standards. Modern calibration methods include SOLT, TRL, LRM, and multiline TRL. TRL is especially valuable at high frequencies because it relies less on accurately known lumped standards and is well suited to planar fixtures and waveguide.

Calibration establishes a measurement reference plane and corrects systematic errors such as directivity, source/load mismatch, reflection tracking, and transmission tracking. Random noise, connector repeatability, cable movement, thermal drift, and imperfect standards remain sources of uncertainty.

De-embedding removes the electrical effects of fixtures, probes, launchers, or transmission-line sections. Embedding performs the reverse operation: it predicts the response after specified networks are added. These operations require reliable fixture models and careful control of reference-plane orientation.

Measurement best practices

1. Select the correct frequency range, intermediate-frequency bandwidth, power level, and number of points.
2. Use a calibration kit appropriate to the connector and frequency band.
3. Inspect and clean connectors; apply the specified torque.
4. Avoid cable movement after calibration.
5. Verify calibration using an independent check standard.
6. Record reference impedance, calibration plane, temperature, power, and bias conditions.
7. Check for compression and nonlinear behaviour when measuring active devices.
8. Preserve complex data, not only plots of magnitude.

10. Worked example: a matched attenuator

Consider an ideal reciprocal 6 dB attenuator with equal 50 Ω reference impedances and negligible reflection. Its transmission magnitude is

Thus,

The input and output are matched, so S₁₁ = S₂₂ = 0. If 1 mW is incident at port 1 and port 2 is matched, the output power is

The power ratio is approximately 0.25, corresponding to a 6 dB loss. Since the network is reciprocal, the same result occurs in the reverse direction. The example also illustrates why voltage-wave magnitude is not itself a power ratio: power depends on the squared magnitude.

11. Worked example: input match with a terminated two-port

Suppose a two-port has

S₁₁=0.20∠−30°, S₂₁=2.0∠80°, S₁₂=0.05∠20°, and S₂₂=0.30∠10°. Let the load reflection coefficient be Γ_L=0.40∠45°. The input reflection coefficient is found from

The second term represents a wave transmitted to the load, reflected there, and returned through the reverse transmission path. A numerical calculation should be performed using complex rectangular coordinates rather than independently adding magnitudes and phases. This distinction is essential: S-parameters are complex quantities, and phase determines constructive or destructive interference.

The example demonstrates a general design principle: S₁₁ is the input reflection only when the output is matched. Once the load differs from the reference impedance, the complete two-port relation must be used.

12. Applied case: characterising a microwave low-noise amplifier

Background

Consider an illustrative 10 GHz low-noise amplifier intended for a satellite receiver. The design team must verify gain, input and output matching, reverse isolation, stability, and compliance with the permitted operating power.

Problem

The transistor is characterised by measured two-port S-parameters at several bias points. The amplifier must provide adequate gain while remaining stable with the intended source and load. A poor input match can increase noise figure, while an unstable bias condition can produce oscillation outside the measurement band.

Analysis

The team first calibrates a VNA to the probe reference planes using multiline TRL. The measured data are exported as complex Touchstone files. S₂₁ is examined for gain, S₁₁ and S₂₂ for matching, and S₁₂ for reverse isolation. Stability is assessed using a suitable criterion, such as Rollett’s factor:

where

For unconditional stability under the usual two-port assumptions, a commonly used sufficient condition is K > 1 and |∆| < 1. These conditions should be evaluated across frequency, bias, temperature, and expected source/load reflection regions; they are not a substitute for a complete stability analysis.

Approach and solution

The engineer adds a small resistive or feedback stabilisation element, then repeats the measurement. The modification reduces gain slightly but improves input/output stability margins. A matching network is synthesised using the measured complex data rather than nominal transistor data. Fixture effects are de-embedded, and the final amplifier is tested at the intended bias and power levels.

Outcome

The resulting design meets its gain and matching targets over the required band and remains stable for the specified source and load conditions. The measured data also reveal a trade-off: improving input match may degrade noise performance or gain. Therefore, S-parameter optimisation must be integrated with noise, linearity, thermal, and reliability requirements.

Learning points

• S-parameters are operating-point dependent for active devices.
• A matched VNA termination does not represent every system termination.
• Stability must be evaluated over frequency and reflection-coefficient space.
• Calibration and de-embedding can affect design conclusions as strongly as the circuit model.
• Measurement data require engineering judgement, not merely visual inspection of plots.

13. Limitations and contemporary developments

S-parameters are small-signal linear quantities. They do not directly describe gain compression, intermodulation, harmonic generation, memory effects, or large-signal switching. For nonlinear devices, X-parameters, large-signal S-parameters, load-pull measurements, and harmonic-balance models may be more appropriate.

Conventional S-parameters also assume a defined mode at each port. In waveguide transitions, differential connectors, coupled lines, and radiating structures, multimode S-parameters may be required. Mixed-mode S-parameters describe differential-mode and common-mode behaviour and are important in high-speed digital and RF interconnects.

Contemporary VNAs support broadband frequency sweeps, time-domain transformation, nonlinear measurements, on-wafer probing, noise-parameter extraction, and automated de-embedding. Machine-learning methods can assist with calibration-error modelling, surrogate modelling, anomaly detection, and design-space exploration, but they do not remove the need for physical validation, traceable calibration, uncertainty analysis, or protection of proprietary measurement data.

14. Engineering and governance considerations

Accurate S-parameter data can influence safety-critical communication, radar, medical, and aerospace decisions. Engineers should document calibration traceability, instrument status, environmental conditions, data provenance, and uncertainty. Data files should include port ordering, reference impedance, frequency units, format, and calibration information. Extrapolating a model beyond its measured frequency, power, temperature, or bias range can produce unsafe designs.

Standards and laboratory procedures should be followed for connector handling, electromagnetic compatibility, equipment calibration, and export-controlled or security-sensitive RF data. Reproducibility requires that another qualified engineer can reconstruct the measurement configuration and processing steps.

Chapter Summary

Microwave network theory represents high-frequency structures through ports and travelling waves. S-parameters relate incident and outgoing waves and are especially suitable because they can be measured directly with matched instruments. S₁₁ and S₂₂ describe reflection under matched opposite-port conditions, while S₂₁ and S₁₂ describe forward and reverse transmission.

Reflection coefficient, return loss, VSWR, gain, reciprocity, symmetry, passivity, and losslessness provide physical interpretations of the S-matrix. Conversion between S-, Z-, Y-, and ABCD-parameters enables different forms of analysis. Cascades require proper network interconnection methods, not casual multiplication of S-matrices. VNA calibration, de-embedding, reference-plane control, and uncertainty management are essential for trustworthy results. Finally, S-parameters are linear, operating-condition-dependent models and must be supplemented by nonlinear, noise, thermal, and stability analysis when required.

Key Terms

Port: A defined interface through which microwave power enters or leaves a network.

Reference plane: The physical plane at which port waves and network parameters are defined.

Incident wave: A travelling wave directed toward a network port.

Outgoing wave: A travelling wave directed away from a network port.

Reference impedance: The impedance used to define incident and outgoing waves.

S-parameter: A complex ratio relating an outgoing wave to an incident wave under matched conditions at all other ports.

Reflection coefficient: The complex ratio of reflected to incident voltage-wave amplitude.

Return loss: The negative decibel value of reflection-coefficient magnitude.

VSWR: Voltage standing-wave ratio produced by a mismatch.

Reciprocity: Property in which transmission between two ports is the same in both directions.

Symmetry: Equality of network behaviour after interchanging ports.

Passivity: The inability of a network to deliver more net power than it receives.

Losslessness: Ideal conservation of net power in a network.

ABCD parameters: Transmission parameters useful for cascading two-port networks.

VNA: Instrument that measures complex reflection and transmission coefficients.

Calibration: Process of correcting systematic measurement errors and establishing a reference plane.

De-embedding: Removal of fixture or interconnect effects from measured data.

Mixed-mode S-parameters: Parameters expressed in differential- and common-mode variables.

Touchstone file: A standard text-based format for frequency-dependent network data.

Review Questions

Short Answer Questions

1. Why are travelling-wave variables preferred to terminal voltage and current at microwave frequencies?
2. What does S₂₁ represent in a two-port network?
3. State the conditions under which a two-port is reciprocal and lossless.
4. Define reference plane and explain why it matters in measurement.
5. Distinguish return loss from VSWR.

Long Answer Questions

1. Derive the incident and outgoing wave definitions from voltage and current on a transmission line, and explain their power interpretation.
2. Explain the complete physical meaning of the four S-parameters of a two-port.
3. Discuss reciprocity, symmetry, passivity, and losslessness using matrix conditions.
4. Compare S-, Z-, Y-, and ABCD-parameter representations and identify suitable applications for each.
5. Describe VNA calibration, de-embedding, uncertainty sources, and best measurement practice.

Analytical / Application Questions

1. A two-port is specified by complex S-parameters and connected to a non-matched load. Derive the input reflection coefficient and explain why S₁₁ alone is insufficient.
2. Design a measurement and validation plan for a 10 GHz microwave amplifier, including calibration, stability, matching, bias, and power checks.
3. Explain how you would determine whether a measured two-port can be safely used in a cascade simulation over a wider frequency and temperature range than the original measurement.

Multiple Choice Questions with Answers

Question 1

For a two-port network, what does S₂₁ represent when port 2 is matched?

A. Input reflection coefficient

B. Forward transmission coefficient

C. Reverse transmission coefficient

D. Output reflection coefficient

Correct Answer: B. Forward transmission coefficient

Question 2

A load is perfectly matched to a line when its reflection coefficient is:

A. −1

B. +1

C. 0

D. j

Correct Answer: C. 0

Question 3

Which condition characterises reciprocity in a consistently referenced multiport network?

A. S_ij=0

B. S_ij=S_ji

C. S_ii=1

D. [S]ᴴ[S]=[0]

Correct Answer: B. S_ij=S_ji

Question 4

An ideal 6 dB attenuator has a transmission magnitude closest to:

A. 0.063

B. 0.25

C. 0.501

D. 2.00

Correct Answer: C. 0.501

Question 5

For a lossless network, conservation of power is expressed by:

A. [S]ᴴ[S]=[I]

B. [S]ᴴ[S]=[0]

C. [S]=[Z]

D. S₁₁=S₂₂=1

Correct Answer: A. [S]ᴴ[S]=[I]

Question 6

When measuring S₂₁, the other ports are normally:

A. Open-circuited

B. Short-circuited

C. Terminated in their reference impedances

D. Connected in parallel

Correct Answer: C. Terminated in their reference impedances

Question 7

Which parameter representation is especially convenient for cascading two-port networks by matrix multiplication?

A. ABCD parameters

B. Only S₁₁

C. Scalar VSWR

D. Return loss

Correct Answer: A. ABCD parameters

Question 8

If |Γ|=0.2, the return loss is approximately:

A. −13.98 dB

B. −3.98 dB

C. 13.98 dB

D. 27.96 dB

Correct Answer: C. 13.98 dB

Question 9

Which measurement process removes the electrical effects of a fixture from measured device data?

A. Embedding

B. De-embedding

C. Modulation

D. Down-conversion

Correct Answer: B. De-embedding

Question 10

The simple power interpretation P=|a|²−|b|² assumes, in the stated formulation, that the reference impedance is:

A. Zero

B. Purely imaginary

C. Real and positive

D. Frequency independent and negative

Correct Answer: C. Real and positive