Microwave Measurements and Metrology

Introduction and Context

Microwave engineering is an experimental discipline as much as a theoretical one. At microwave frequencies, voltage and current are not conveniently measured as lumped-circuit quantities because conductor dimensions become comparable with wavelength, parasitic effects become significant, and electromagnetic fields propagate as guided waves. Consequently, microwave measurements are usually expressed in terms of travelling-wave quantities, reflection coefficients, scattering parameters, impedance, power, noise, frequency, and phase.

Microwave metrology is the science of establishing measurement standards, traceability, uncertainty, and calibration procedures for these quantities. It links a laboratory measurement to a recognized reference through an unbroken chain of calibrations. Reliable metrology is essential in radar, satellite communication, wireless systems, electronic warfare, semiconductor characterization, antenna testing, and production quality control.

A measurement result is not simply a numerical reading. It should include the measurand, measurement method, calibration status, environmental conditions, uncertainty, and—in regulated work—traceability to national or international standards.

Learning Objectives

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

1. Explain why microwave measurements differ from low-frequency measurements.
2. Describe power, frequency, impedance, reflection, transmission, phase, and noise measurements.
3. Derive and apply relationships among reflection coefficient, VSWR, return loss, and impedance.
4. Explain calibration and error-correction methods for vector network analyzers.
5. Evaluate measurement uncertainty and identify dominant error sources.
6. Select suitable instruments, standards, and fixtures for practical microwave tests.
7. Interpret a microwave measurement case from problem definition through validation.

1. The Measurement Model at Microwave Frequencies

For a sinusoidal wave on a transmission line, the voltage and current may be represented as the sum of incident and reflected waves. At a reference plane, the complex reflection coefficient is

where Γ is the complex reflection coefficient, a and b are incident and reflected wave amplitudes, Z_L is the load impedance, and Z_0 is the characteristic impedance of the line. For a real reference impedance, usually 50 Ω, this relation applies to a single-mode transmission line under steady-state sinusoidal excitation.

The inverse relation is

The magnitude |Γ| indicates the fraction of voltage-wave amplitude reflected at the load, while the phase of Γ gives the position of the impedance relative to the reference plane. Power reflection is |Γ|² when the incident and reflected waves are normalized consistently.

The power delivered to a load, neglecting conductor and dielectric loss between the measurement plane and the load, is

where P_inc is incident power. This expression assumes a single propagating mode and no additional losses in the defined measurement plane.

1.1 Scattering Parameters

At microwave frequencies, scattering parameters are preferred to open-circuit voltage or short-circuit current parameters because matched terminations are easier to realize accurately than ideal opens or shorts. For an n-port network,

where a_j and b_i are normalized incident and outgoing wave amplitudes at ports j and i, respectively. S_ij is measured with all ports other than j terminated in their reference impedances.

For a two-port network,

S_11 and S_22 are input and output reflection coefficients under matched termination; S_21 is forward transmission and S_12 is reverse transmission. In decibels, a power-related transmission magnitude is commonly reported as 20 log₁₀|S_21| when S-parameters are wave-amplitude ratios. For a matched two-port, available power gain associated with transmission is |S_21|², and its decibel value is 10 log₁₀|S_21|² = 20 log₁₀|S_21|.

2. Fundamental Microwave Measurands

2.1 Frequency and Frequency Stability

Frequency is measured using frequency counters, spectrum analyzers, signal analyzers, phase-locked receivers, and frequency references. A counter measures the number of cycles during a gate interval. If N cycles are counted during time T_g,

The result is affected by time-base accuracy, trigger uncertainty, reciprocal-counting error, phase noise, and signal purity. A rubidium, cesium, or disciplined quartz reference may be used where high long-term accuracy is required. Frequency accuracy and frequency stability are distinct: accuracy describes closeness to the nominal value, whereas stability describes variation with time.

2.2 Power

Microwave power may be measured using thermistor, thermocouple, diode, bolometric, calorimetric, and broadband sensor technologies.

Thermal sensors measure the heating effect of absorbed power and are generally accurate over a defined frequency range, but may be relatively slow. Diode sensors provide high sensitivity and speed but require linearization and careful calibration, particularly for modulated or high-crest-factor signals. Calorimeters determine power from a temperature rise and are useful as primary or transfer standards, although they are physically larger and slower.

Average power, peak power, pulse power, and power spectral density are different measurands. A sensor calibrated for continuous-wave average power cannot automatically be assumed accurate for a pulsed waveform. For a pulse with energy E_p and pulse duration τ_p, the average power during the pulse is

If pulses repeat with period T_r, the duty cycle is D = τ_p/T_r and, for an approximately rectangular pulse train,

The last relation assumes negligible rise and fall-time effects and constant pulse amplitude.

2.3 VSWR, Return Loss, and Mismatch Loss

The voltage standing-wave ratio is related to the reflection coefficient by

for 0 ≤ |Γ| < 1. Return loss is

and mismatch loss is

Return loss is zero for total reflection and increases as the match improves. Mismatch loss is zero for a perfect match and increases as reflected power increases. These quantities should not be treated as interchangeable.

Worked example. A device has |Γ| = 0.20 at its input. Then

Thus, the device reflects 4% of the incident power, has a VSWR of approximately 1.5:1, and introduces approximately 0.18 dB mismatch loss under the stated assumptions.

2.4 Impedance and Admittance

A vector network analyzer (VNA) measures complex reflection coefficient and converts it to impedance using the inverse relation given earlier. The conversion is valid only when the reference impedance and reference plane are correctly defined. A phase error caused by an unknown cable length can produce a large apparent impedance error, particularly for highly reactive loads.

On a Smith chart, normalized impedance is plotted as z = Z_L/Z_0. Constant-resistance and constant-reactance circles provide a graphical means of interpreting matching networks, transmission-line transformations, and stability regions.

2.5 Phase and Group Delay

The phase of a transmission coefficient is measured relative to a reference. For a two-port network, group delay is

where φ is the unwrapped transmission phase in radians and ω is angular frequency in radians per second. For approximately linear phase over a narrow frequency interval, τ_g may be estimated by a finite difference:

Phase must be unwrapped before differentiation; otherwise discontinuities of ±2π can create false delay spikes. Group delay is important in broadband communication channels, filters, radar pulse compression, and phased-array systems.

2.6 Noise Measurements

Noise temperature and noise figure characterize the degradation of signal-to-noise ratio by a component. The noise factor is

and noise figure is

For a matched source at the standard reference temperature T₀ = 290 K, the equivalent input noise temperature is

For cascaded stages, Friis’ formula is

where F_i are linear noise factors and G_i are preceding available power gains, not decibel values. This equation assumes stages are connected in the intended reference conditions and that the noise contributions can be represented by equivalent input noise factors.

3. Principal Microwave Measurement Instruments

3.1 Scalar and Vector Network Analyzers

A scalar network analyzer measures magnitude only, whereas a VNA measures both magnitude and phase of reflected and transmitted waves. A typical VNA contains a stimulus source, directional couplers or bridges, receivers, a reference clock, data-processing hardware, and calibration software.

The VNA does not measure an ideal S-parameter directly. Coupler leakage, receiver imbalance, imperfect source match, switch repeatability, drift, cable loss, connector mismatch, and fixture effects contaminate the raw data. Calibration estimates these systematic errors and mathematically removes them from subsequent measurements.

3.2 Spectrum and Signal Analyzers

A spectrum analyzer displays signal amplitude versus frequency, usually through a swept-tuned or Fourier-transform architecture. Important settings include frequency span, resolution bandwidth, video bandwidth, detector type, attenuation, reference level, sweep time, and averaging.

Resolution bandwidth determines the ability to separate nearby spectral components and affects displayed noise level. Video bandwidth smooths the detected trace but does not improve the fundamental frequency resolution. Excessive input attenuation protects the mixer but reduces sensitivity; insufficient attenuation can cause compression or internal distortion.

3.3 Power Meters and Sensors

A power meter should be selected according to frequency range, dynamic range, connector type, maximum power, modulation bandwidth, and required uncertainty. Sensor calibration factors are frequency dependent. The indicated value generally requires application of the calibration factor supplied for the sensor and frequency.

3.4 Time-Domain and Modulation Measurements

A VNA can transform frequency-domain data into a time-domain response using an inverse Fourier transform. Time-domain gating can isolate a connector discontinuity or fixture reflection, but it cannot recover information that was excluded by insufficient frequency span or inappropriate windowing. Modulation analyzers measure error-vector magnitude, occupied bandwidth, phase error, and other digital-communication quantities; these are distinct from conventional small-signal S-parameters.

4. Calibration, Reference Planes, and Error Correction

4.1 Calibration versus Verification

Calibration determines correction coefficients using standards with known or characterized properties. Verification checks the calibrated system using an independent device or check standard. Verification is evidence that the measurement system remains within an acceptance criterion; it is not a substitute for calibration.

A calibration is valid only for its frequency range, connector system, power level, temperature range, port configuration, and reference plane. Moving a cable, changing a test port adapter, or altering power level may invalidate the correction.

4.2 One-Port Calibration

A one-port calibration corrects reflection measurements. The classical SOL calibration uses short, open, and load standards. The standards provide known reflection coefficients across frequency, enabling estimation of directivity, source match, and reflection tracking.

A simplified one-port measurement model is

where Γ_m is measured reflection coefficient, Γ is the true reflection coefficient, E_D is directivity error, E_R is reflection tracking, and E_S is source-match error. Practical instruments solve for the error terms using a defined calibration algorithm.

4.3 Two-Port Calibration

Two-port calibration must address reflection and transmission errors at both ports. Common methods include SOLT, TRL, LRM, and electronic calibration.

• SOLT uses short, open, load, and through standards. It is convenient when accurate coaxial standards are available.
• TRL uses through, reflect, and line standards. It is especially valuable in planar, waveguide, and non-coaxial environments because it reduces dependence on accurately known open and short standards.
• LRM uses line, reflect, and match standards and is useful in some fixture geometries.
• Electronic calibration uses an integrated module containing characterized impedance states, improving repeatability and reducing manual connection errors.

The calibration reference plane is the plane at which corrected S-parameters are defined. Port extension mathematically shifts the reference plane by adding or removing a modeled transmission-line delay and loss. It should not be used to conceal an unmodeled discontinuity.

4.4 Connector and Fixture Practice

Connector cleanliness, alignment, torque, wear, and gender compatibility are major determinants of repeatability. Precision connectors should be inspected with a microscope when appropriate and tightened using the specified torque wrench. A damaged connector can create a calibration that appears valid but produces erroneous device data.

For on-wafer or fixture measurements, de-embedding removes the electrical effect of probes, pads, launches, and interconnects. De-embedding is model dependent; an incorrect fixture model may introduce more error than it removes. Fixture design should therefore support electromagnetic characterization, repeatable contact, and a well-defined calibration plane.

5. Measurement Uncertainty and Metrological Traceability

5.1 Sources of Uncertainty

Uncertainty may arise from calibration-standard characterization, instrument noise, drift, connector repeatability, mismatch, resolution, temperature, cable movement, fixture modeling, and data-processing assumptions. Random effects are often estimated from repeated observations; systematic effects are evaluated using calibration data, specifications, models, or prior measurements.

The standard uncertainty of a quantity x is denoted u(x). For independent input quantities x_i and a measurand y = f(x₁, x₂, …), the law of propagation of uncertainty is

when covariance terms are negligible. More generally,

where u(x_i,x_j) is the covariance. An expanded uncertainty is commonly reported as

where k is a coverage factor, often close to 2 for approximately normal distributions and approximately 95% coverage, provided the effective degrees of freedom justify that interpretation.

5.2 Example of Uncertainty Combination

Suppose a power measurement is affected by three independent standard uncertainties: sensor calibration 0.20 dB, meter repeatability 0.10 dB, and mismatch contribution 0.15 dB. Treating the quantities as expressed in the same additive dB representation,

For k = 2, the expanded uncertainty is approximately 0.54 dB. This result is valid only under the stated independence and distribution assumptions. Correlation or a nonlinear measurement model requires a revised analysis.

5.3 Traceability and Reporting

Traceability requires a documented, unbroken chain of calibrations, each contributing to measurement uncertainty. A sound report identifies the instrument and serial number, calibration date, standards, reference plane, frequency and power conditions, environmental conditions, measured value, uncertainty, coverage factor, and acceptance criterion.

Significant digits should reflect uncertainty. Reporting 12 digits from an instrument whose uncertainty is ±0.2 dB creates false precision. Measurement data should also preserve raw files, calibration coefficients, software version, and test configuration so that results are reproducible.

6. Practical Measurement Procedures

A disciplined measurement begins with a measurand definition. The engineer should specify whether the objective is input return loss, insertion loss, gain, noise figure, harmonic power, spurious emission, group delay, or another quantity.

A robust procedure is:

1. Define frequency range, dynamic range, power level, ports, reference impedance, and acceptance limits.
2. Select instruments and standards whose specifications support the required uncertainty.
3. Allow instruments and standards to reach thermal equilibrium.
4. Inspect and clean connectors; control cable routing and mechanical stress.
5. Perform the appropriate calibration at the intended reference plane.
6. Verify calibration using an independent check device.
7. Measure the device with suitable averaging, IF bandwidth, power, and receiver settings.
8. Repeat selected connections to estimate repeatability.
9. Check for compression, drift, leakage, resonances, and implausible discontinuities.
10. Document the result with uncertainty and configuration information.

A useful diagnostic principle is to change one condition at a time. If moving a cable changes the result, the system may be sensitive to phase stability or connector movement. If reducing source power changes gain, compression or nonlinear behavior may be present. If a result changes with IF bandwidth, noise and averaging settings may be influencing the observation.

7. Illustrative Case Study: Characterizing a 5 GHz Low-Noise Amplifier

Background. An engineering laboratory must validate a 5 GHz low-noise amplifier (LNA) intended for a receiver front end. Required measurements are gain, input and output return loss, noise figure, and stability over 4.8–5.2 GHz.

Problem. Initial measurements show gain variation of nearly 1 dB between operators. The apparent noise figure also changes when the test cables are repositioned.

Analysis. The LNA is a high-gain, low-noise device. Cable movement changes phase and mismatch at the input, while inadequate isolation may allow output-to-input feedback. The laboratory’s first calibration was performed with a short cable arrangement, but the device was subsequently measured through an adapter and a mechanically unsupported cable. No independent verification standard was used.

Approach. The team mounts both cables securely and defines the calibration plane at the LNA connectors. A two-port VNA calibration is performed using suitable coaxial standards, followed by verification with a characterized airline and attenuator. Gain is measured at a power level sufficiently below compression. Input and output reflection coefficients are recorded simultaneously. Noise figure is measured using a calibrated noise source and receiver, with the noise-source excess-noise ratio entered at each frequency. The team repeats the connection three times and includes connector repeatability in the uncertainty budget.

For a measured forward coefficient |S₂₁| = 10, the small-signal transmission magnitude is

If the input reflection magnitude is 0.316, the input return loss is approximately

The team also calculates available gain and checks whether the measured S-parameters satisfy the intended stability margin. If unconditional stability is required, the Rollett factor may be used:

where Δ = S₁₁S₂₂ − S₁₂S₂₁. For the conventional criterion, K > 1 and |Δ| < 1 indicate unconditional stability under the assumptions of the model and reference impedances.

Outcome. After mechanical stabilization, correct reference-plane calibration, and verification, gain variation falls within the laboratory repeatability target. The noise-figure variation is reduced substantially. The remaining uncertainty is dominated by noise-source calibration and input mismatch rather than cable motion.

Learning points. Calibration is a system operation, not a button pressed without regard to the test setup. Mechanical repeatability, reference-plane definition, power level, independent verification, and an uncertainty budget are all necessary for defensible microwave results.

8. Limitations, Regulation, and Contemporary Developments

No measurement method is universally superior. A VNA provides phase and coherent information but can be sensitive to calibration quality and dynamic range. A power meter is accurate for defined power conditions but generally lacks spectral information. A spectrum analyzer reveals frequency content but may suffer from overload, leakage, and detector-dependent readings. Time-domain transformations are useful for localization but are limited by bandwidth and windowing.

Regulatory measurements such as radiated emissions, occupied bandwidth, and spurious radiation require compliance with applicable national and international standards. The test site, antenna factor, cable loss, detector function, bandwidth, and uncertainty must conform to the relevant procedure. It is poor practice to report a pass/fail decision without stating the measurement uncertainty and decision rule where regulations require it.

Modern systems increasingly use automated calibration, multiport VNAs, on-wafer probing, broadband millimeter-wave extenders, and real-time signal analysis. Machine learning can assist in anomaly detection, connector-failure prediction, drift monitoring, and calibration-quality assessment. It does not replace traceability: an algorithmic correction must be validated against standards, its training domain must be documented, and out-of-domain behavior must be recognized. Automated systems should retain raw data, calibration states, software versions, and audit logs.

Best practice is therefore a combination of electromagnetic understanding, careful instrumentation, controlled mechanics, metrological discipline, and transparent reporting.

Chapter Summary

Microwave measurements use travelling-wave quantities because lumped voltage and current measurements become inadequate at high frequency. Reflection coefficient, impedance, VSWR, return loss, mismatch loss, S-parameters, phase, group delay, power, noise figure, and noise temperature are central measurands. Vector network analyzers measure complex reflection and transmission but require calibration to remove systematic errors. SOLT, TRL, LRM, and electronic calibration methods establish corrected reference planes under defined conditions.

Reliable results require connector control, fixture characterization, verification, uncertainty evaluation, traceability, and appropriate reporting. Power, frequency, noise, and spectral measurements require different instruments and calibration concepts. In professional practice, a measurement is defensible only when its measurand, method, conditions, uncertainty, and traceability are clear.

Key Terms

Reflection coefficient: Complex ratio of reflected to incident wave amplitude.

Scattering parameter: Ratio relating outgoing and incident waves at a port under specified terminations.

Reference plane: Electrical plane at which a microwave quantity is defined.

VSWR: Ratio of maximum to minimum standing-wave voltage.

Return loss: Negative decibel value of reflection-coefficient magnitude.

Mismatch loss: Power loss caused by reflection at an imperfect match.

Vector network analyzer: Instrument measuring magnitude and phase of microwave reflection and transmission.

Scalar network analyzer: Instrument measuring magnitude without phase information.

SOLT calibration: Calibration using short, open, load, and through standards.

TRL calibration: Calibration using through, reflect, and line standards.

De-embedding: Mathematical removal of fixture or interconnect effects.

Noise factor: Ratio of input to output signal-to-noise ratio.

Noise figure: Noise factor expressed in decibels.

Group delay: Negative derivative of transmission phase with respect to angular frequency.

Traceability: Documented chain connecting a result to a recognized standard.

Standard uncertainty: Uncertainty expressed as an estimated standard deviation.

Expanded uncertainty: Combined standard uncertainty multiplied by a coverage factor.

Calibration verification: Independent check that a calibrated system performs acceptably.

Reference impedance: Impedance used to normalize travelling-wave quantities, commonly 50 Ω.

De-embedding: Process of mathematically shifting the measurement plane through a known network.

Review Questions

Short Answer Questions

1. Why are scattering parameters preferred to open-circuit impedance parameters at microwave frequencies?
2. Define return loss and mismatch loss, and distinguish their physical meanings.
3. What is the purpose of a VNA calibration?
4. State two differences between frequency accuracy and frequency stability.
5. Why must phase be unwrapped before calculating group delay?

Long Answer Questions

1. Explain the architecture and operation of a two-port vector network analyzer.
2. Derive the relationships among reflection coefficient, impedance, VSWR, and return loss.
3. Compare SOLT, TRL, LRM, and electronic calibration methods.
4. Explain the major contributors to uncertainty in a microwave power measurement.
5. Discuss the practical requirements for accurate on-wafer or fixture-based microwave measurements.

Analytical / Application Questions

1. A load has |Γ| = 0.1. Calculate its VSWR, return loss, reflected-power fraction, and mismatch loss.
2. Design a measurement plan for a 10 GHz filter requiring insertion-loss uncertainty below 0.2 dB. Specify calibration, verification, instrument settings, and uncertainty contributors.
3. An LNA has three cascaded stages with linear gains G₁ = 10, G₂ = 5, G₃ = 4 and noise factors F₁ = 2, F₂ = 3, F₃ = 4. Use Friis’ formula to calculate total noise factor and noise figure, and explain which stage dominates the result.

Multiple Choice Questions with Answers

Question 1

For a load with |Γ| = 0.20, what is the approximate VSWR?

A. 1.05:1

B. 1.25:1

C. 1.50:1

D. 2.00:1

Correct Answer: C. 1.50:1

Question 2

Which calibration uses short, open, load, and through standards?

A. TRL

B. SOLT

C. LRM

D. Thru-reflect-line only

Correct Answer: B. SOLT

Question 3

Return loss is calculated from reflection coefficient magnitude using which expression?

A. 20 log₁₀|Γ|

B. −20 log₁₀|Γ|

C. 10 log₁₀(1−|Γ|)

D. −10 log₁₀|Γ|² only for power meters

Correct Answer: B. −20 log₁₀|Γ|

Question 4

What does a vector network analyzer measure that a scalar network analyzer does not?

A. Frequency

B. Average power

C. Phase

D. Temperature

Correct Answer: C. Phase

Question 5

If a pulse has duration 2 μs and repetition period 10 μs, what is its duty cycle?

A. 2%

B. 5%

C. 20%

D. 50%

Correct Answer: C. 20%

Question 6

Which quantity is defined as −dφ/dω?

A. Return loss

B. Group delay

C. Noise factor

D. Mismatch loss

Correct Answer: B. Group delay

Question 7

For independent uncertainty components of 0.20 dB, 0.10 dB, and 0.15 dB, the combined standard uncertainty is approximately:

A. 0.15 dB

B. 0.27 dB

C. 0.45 dB

D. 0.90 dB

Correct Answer: B. 0.27 dB

Question 8

In a two-port S-parameter measurement, S₂₁ primarily represents:

A. Reverse reflection

B. Forward transmission

C. Input reflection

D. Output reflection

Correct Answer: B. Forward transmission

Question 9

Which method is especially useful when accurately characterized open and short standards are difficult to realize?

A. TRL

B. Power-meter calibration

C. Frequency counting

D. Detector linearization

Correct Answer: A. TRL

Question 10

For a matched two-port, the decibel value of |S₂₁| = 10 is:

A. 10 dB

B. 20 dB

C. 30 dB

D. 100 dB

Correct Answer: B. 20 dB