Microwave Amplifier Design

Introduction

A microwave amplifier is an active two-port network that increases the power, voltage, or signal-to-noise performance of a microwave signal while operating over a frequency range in which parasitic capacitances, transmission-line effects, impedance transformation, and device instability are significant. Unlike low-frequency amplifier design, microwave design is normally performed in terms of travelling waves and scattering parameters rather than simple open-circuit voltage and short-circuit current parameters.

Microwave amplifiers are used in receiver front ends, radar transmitters, satellite payloads, wireless infrastructure, instrumentation, electronic warfare systems, and test equipment. The design objective depends on the application. A low-noise amplifier (LNA) prioritizes noise figure and stable operation; a power amplifier prioritizes output power, efficiency, and linearity; a wideband amplifier prioritizes bandwidth and gain flatness; and a driver amplifier must often balance all these properties.

The central design problem is to select and bias a transistor, determine suitable source and load impedances, implement those impedances using microwave matching networks, and verify gain, stability, noise, power, and thermal performance.

Learning Objectives

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

1. Explain the role of scattering parameters in microwave amplifier design.
2. Distinguish transducer gain, available gain, operating gain, and unilateral gain.
3. Analyze unconditional and conditional stability using Rollett's stability factor and the μ criterion.
4. Design input and output matching networks using reflection coefficients and Smith-chart concepts.
5. Apply noise-parameter theory to low-noise amplifier design.
6. Evaluate power gain, compression, intermodulation, and efficiency.
7. Develop a practical design workflow and assess its limitations through an applied case.

1. Microwave Two-Port Model

Consider a transistor represented as a linear two-port network with reference impedance Z₀, commonly 50 Ω. The incident and reflected travelling-wave amplitudes at ports 1 and 2 are a₁, a₂ and b₁, b₂. The scattering relation is

Here Sᵢⱼ is the output wave at port i caused by an incident wave at port j when all other ports are terminated in Z₀. Thus S₁₁ and S₂₂ are input and output reflection coefficients under matched conditions, S₂₁ is forward transmission, and S₁₂ is reverse transmission. S-parameters are frequency-dependent complex quantities and are normally obtained from measurement or a small-signal device model.

If the source and load reflection coefficients are ΓS and ΓL, respectively, the input and output reflection coefficients of the loaded amplifier are

These equations assume linear operation, consistent reference impedances, and stable interconnection of the source, transistor, and load.

The source and load matching networks are described by ΓS and ΓL. For a passive network referenced to Z₀, |Γ| ≤ 1. The corresponding impedance is

where Z and Z₀ are complex impedances in ohms. Conversely,

A matched termination has Γ = 0 and Z = Z₀.

2. Gain Definitions and Derivation

Several gain definitions are necessary because a microwave amplifier may not be matched at either port.

2.1 Transducer gain

Transducer gain is the ratio of power delivered to the load to the power available from the source:

For a two-port described by S-parameters,

This is the most useful gain expression for a practical matched-network design because it includes source mismatch, load mismatch, reverse transmission, and feedback through S₁₂.

2.2 Available gain and operating gain

Available gain is the ratio of the maximum power available from the amplifier to the power available from the source. Operating gain is the ratio of power delivered to the load to the power accepted by the amplifier input. They are useful when analyzing source and load matching separately.

If the device is unilateral, meaning S₁₂ ≈ 0, the transducer gain simplifies to

where

Gains are often expressed in decibels:

For a voltage or wave-amplitude ratio, 20 log₁₀ is used only when the corresponding impedance conditions justify that relationship.

2.3 Maximum available gain

For a potentially unstable bilateral device, maximum available gain (MAG) is defined under simultaneous conjugate matching and is meaningful only when the device is unconditionally stable. In one common form,

where K is Rollett's stability factor and K > 1. This expression assumes S₁₂ ≠ 0 and linear small-signal operation. If S₁₂ is extremely small, the unilateral approximation is usually more practical.

3. Stability Analysis

An amplifier is unconditionally stable at a frequency if every passive source and load termination produces a stable network. Conditional stability means that only a restricted region of source and load terminations is stable.

Rollett's stability factor is

where

For S₁₂S₂₁ ≠ 0, a commonly used unconditional-stability condition is

A numerically convenient alternative is the μ factor:

where the asterisk denotes complex conjugation. The device is unconditionally stable when μ > 1; the corresponding reverse factor μ′ must also exceed unity when both source and load terminations are considered. Stability must be checked across the complete operating band, bias range, temperature range, and expected layout variation.

When a transistor is conditionally stable, stabilization may be achieved by adding resistive loss, feedback, series or shunt resistors, lossy matching sections, or a neutralization network. These methods generally reduce gain, increase noise, reduce efficiency, or narrow the bandwidth. A design should therefore use the least harmful stabilization that provides an adequate margin.

4. Matching-Network Design

The purpose of an input matching network is to transform the source impedance into the desired ΓS at the transistor reference plane. The output network transforms the transistor's desired load reflection coefficient ΓL into the physical load impedance.

For maximum power transfer in a passive, lossless matching network, the termination presented to a device is the complex conjugate of the device impedance. In a bilateral amplifier, however, Γin depends on ΓL and Γout depends on ΓS. Therefore, exact simultaneous conjugate matching may be impossible or undesirable. In a unilateral design, the input and output matches can be designed independently.

Common matching structures include:

For a lossless transmission line of characteristic impedance Z₀ and electrical length θ terminated in ZL,

At θ = π/2, this becomes the quarter-wave transformation

These expressions assume an ideal, lossless, TEM or quasi-TEM line and must be corrected in practice for conductor loss, dielectric loss, dispersion, bends, vias, pads, and discontinuities.

5. Gain Circles and Design Trade-offs

When ΓS or ΓL is varied over the passive Smith chart, constant-gain contours can be plotted. Input and output constant-gain circles help the designer choose a compromise among gain, stability, bandwidth, and noise.

For a unilateral device, the source and load decisions are decoupled. The input network controls GS and the output network controls GL. For a bilateral device, gain circles are coupled because reverse transmission produces feedback. In such a case, simultaneous conjugate matching may maximize gain but may lead to poor stability or unacceptable noise performance.

A practical design rarely maximizes gain alone. Excessive gain can increase oscillation risk, sensitivity to component tolerances, and nonlinear distortion. A controlled gain target with adequate stability and noise margin is usually preferable.

6. Noise in Microwave Amplifiers

For an amplifier with source reflection coefficient ΓS, the noise figure can be represented by four noise parameters:

Here F is the noise factor (dimensionless), Fmin is the minimum noise factor, Rn is the equivalent noise resistance in ohms, and Γopt is the optimum source reflection coefficient. These parameters are specified at a particular frequency and bias condition.

Noise figure in decibels is

Minimum noise is obtained when ΓS = Γopt, provided the matching network is lossless and the model is valid. This source match is generally different from the source match for maximum gain. Consequently, LNA design is a constrained optimization problem.

For cascaded stages, the overall noise factor is given by Friis' formula:

Here Fi are linear noise factors and Gi are preceding-stage available power gains, not decibel values. The first stage dominates system noise performance, which explains the importance of a low-loss input network and appropriate biasing.

A passive loss L at the physical input, expressed as a linear power-loss ratio L ≥ 1, has noise factor F = L at the reference temperature. Thus a 1 dB input loss contributes approximately 1 dB of noise figure before any active gain is obtained.

7. Power, Linearity, and Efficiency

Small-signal S-parameters do not predict large-signal amplifier behavior. As input power increases, transistor current and voltage limits, gain compression, thermal effects, and waveform clipping become important.

The 1 dB compression point is the input or output power at which the measured gain is 1 dB below the extrapolated small-signal gain. It is commonly denoted P1dB,in or P1dB,out. The saturated output power, Psat, is the approximate maximum output power under a specified drive and bias condition.

For a two-tone test, third-order intermodulation products occur at 2f₁ − f₂ and 2f₂ − f₁. The third-order intercept point is an extrapolated measure of linearity, not a physically reached operating point. If the fundamental and third-order product powers are measured in dBm at a common reference, the output intercept point may be estimated by

where Pfund is the fundamental output power and PIM3 is the third-order product output power. The formula assumes the fundamental increases approximately 1 dB per dB of input increase while the third-order product increases approximately 3 dB per dB in the low-distortion region.

Power-added efficiency is

where Pout is RF output power, Pin is RF input power, and PDC is DC power consumed. Drain efficiency is Pout/PDC for a transistor stage when defined with the same power conventions. PAE is especially useful for comparing driver and power amplifier performance.

8. Biasing and Device Selection

The device must be selected with regard to frequency, gain, noise, breakdown voltage, current capability, thermal resistance, linearity, reliability, and packaging. Gallium nitride devices are attractive for high power and high voltage; GaAs and pHEMT devices are widely used for low-noise and microwave applications; silicon-germanium and RF CMOS technologies support integration and cost reduction.

Bias networks should provide:

• a stable operating point over temperature and supply variation;
• high impedance at RF where appropriate;
• low impedance at the intended bias modulation frequencies;
• adequate decoupling and suppression of bias-line feedback;
• safe current limiting and transient protection.

The transistor's S-parameters and noise parameters must be used at the actual bias point. Data measured at a different bias or reference plane can produce a misleading design. Bias tees, bond wires, package leads, vias, and decoupling capacitors are part of the RF network, not merely external accessories.

9. Practical Design Workflow

A robust design sequence is as follows:

1. Specify frequency range, gain, noise figure, output power, linearity, supply, bandwidth, and environmental limits.
2. Select a device and bias point using manufacturer data or measured characterization.
3. Import or measure S-parameters, noise parameters, and large-signal data at the selected bias.
4. Check K, μ, and stability circles throughout the frequency and bias ranges.
5. Decide whether the design is gain-driven, noise-driven, power-driven, or a constrained compromise.
6. Select ΓS and ΓL using noise circles, gain circles, stability regions, and load-pull data.
7. Synthesize input and output matching networks.
8. Include transmission-line losses, discontinuities, package models, and bias networks in electromagnetic simulation.
9. Simulate small-signal gain, return loss, noise figure, stability, compression, harmonics, and thermal behavior.
10. Fabricate a calibrated test board and measure at the correct reference planes.
11. Correlate simulation and measurement, identify parasitic effects, and retune the layout.

The reference plane is critical. S-parameters measured at the connector plane cannot be directly combined with a transistor model defined at the package pin unless the intervening network is de-embedded or explicitly included.

10. Worked Example: Unilateral Gain Estimate

Consider an illustrative transistor at 10 GHz with

S₁₁ = 0.45∠−60°, S₂₁ = 3.0∠80°, S₂₂ = 0.35∠−25°, and S₁₂ ≈ 0.

Assume the source and load networks are designed for ΓS = S₁₁ and ΓL = S₂₂. For a unilateral device, these are conjugate matches. The source and load factors become

The intrinsic forward power gain is

Therefore,

and

This result is an ideal small-signal estimate. It excludes matching-network loss, bias-network loss, layout parasitics, noise constraints, and large-signal compression. If the input and output networks each introduce 0.5 dB loss, the realized gain will be approximately 10.1 dB rather than 11.1 dB, assuming the losses occur in the signal path and other effects are negligible.

11. Applied Case: 5.8 GHz Receiver LNA

Background

The following is an illustrative case involving a 5.8 GHz receiver front end requiring approximately 15 dB gain, a noise figure below 2 dB, unconditional stability across 5.5–6.0 GHz, and a minimum input return loss of 10 dB.

Problem

A device offers its minimum noise figure when ΓS is near Γopt, but the corresponding source match gives only moderate input return loss. The transistor is also conditionally stable near the lower edge of the band. A direct maximum-gain conjugate match would violate the stability requirement and degrade noise performance.

Analysis and approach

The designer first selects a bias point with adequate gain and low noise current. Stability factors are calculated at frequency increments across the band, including package and bias networks. A small series resistor in the gate path and a lossy bias-feed arrangement increase the stability margin. Their loss is included in the noise calculation.

The input network is designed to transform 50 Ω close to Γopt while maintaining the required return loss. The output network is selected for a moderate gain match rather than maximum available gain. Electromagnetic simulation includes microstrip bends, via inductance, transistor pads, and the decoupling network. The design is then checked with a two-tone test for linearity and with a temperature sweep for bias stability.

Outcome

A suitable design would be accepted only if measured gain, noise figure, input return loss, stability margin, and current consumption all satisfy the specification over frequency and temperature. If gain is excessive, a small attenuating or feedback element may improve stability and reduce the risk of oscillation. If noise figure is too high, the first remedy should be reduction of input loss and reconsideration of the input matching network rather than simply increasing bias current.

Learning points

The case demonstrates that microwave amplifier design is a multi-objective optimization problem. Maximum gain is not necessarily the correct design target. Stability, noise, matching, manufacturability, and measurement uncertainty must be considered simultaneously.

12. Limitations, Reliability, and Contemporary Practice

Models are valid only within specified frequency, bias, temperature, and power domains. Small-signal models cannot predict compression; linear S-parameters cannot establish thermal reliability; and a simulated stable circuit can oscillate after packaging or enclosure integration. Electromagnetic co-simulation and nonlinear device models reduce these risks but do not eliminate the need for measurement.

Modern design increasingly uses automated optimization, surrogate models, load-pull databases, and machine-learning-assisted tuning. Such tools can efficiently search a multidimensional design space, but they do not replace physical constraints, model validation, or engineering judgment. Training data may not represent process corners or unusual bias conditions, and an optimizer may exploit an artefact in an inaccurate model.

Professional practice requires traceable data, calibrated instruments, controlled reference planes, repeatable fixtures, appropriate electromagnetic compatibility procedures, and compliance with applicable emissions, safety, and spectrum regulations. High-power amplifiers also require protection against excessive voltage, current, temperature, and reflected power.

Best-Practice Checklist

• Use measured or validated device data at the actual bias point.
• Check stability over frequency, bias, temperature, and manufacturing variation.
• Keep noise parameters and S-parameters in linear form during calculations; convert to decibels only for reporting.
• Include matching-network and input-loss contributions in the noise budget.
• Distinguish small-signal gain from compressed and large-signal gain.
• De-embed measurement planes consistently.
• Simulate and measure both differential and common-mode bias behaviour where relevant.
• Provide adequate thermal paths and protection circuits.
• Optimize for the system specification rather than a single isolated metric.

Chapter Summary

Microwave amplifier design uses S-parameters and travelling-wave concepts to determine gain, matching, stability, noise, and power performance. The transducer-gain equation accounts for source and load mismatch as well as bilateral feedback. Stability is evaluated using K, Δ, μ, and related stability regions; unconditional stability requires suitable margins across all relevant operating conditions. Matching networks transform practical source and load impedances into desired reflection coefficients, but their loss and parasitics must be included.

Low-noise design uses noise parameters and often requires a compromise between Γopt and maximum-gain matching. Power-amplifier design additionally requires compression, intermodulation, efficiency, thermal, and reliability analysis. A complete engineering workflow proceeds from specifications and device selection through biasing, stability analysis, matching synthesis, electromagnetic and nonlinear simulation, fabrication, calibration, and measurement. The essential professional principle is that gain, noise, stability, linearity, efficiency, bandwidth, cost, and reliability must be optimized together.

Key Terms

S-parameter: Complex travelling-wave ratio used to characterize a microwave network.

Reflection coefficient: Ratio of reflected to incident wave at a port.

Transducer gain: Power delivered to the load divided by power available from the source.

Available gain: Maximum power available from an amplifier divided by power available from the source.

Operating gain: Power delivered to the load divided by power accepted at the amplifier input.

Unilateral device: Device for which reverse transmission S₁₂ is negligible.

Bilateral device: Device with significant reverse transmission.

Rollett stability factor: K-based measure used to assess unconditional stability.

μ factor: Alternative stability measure; μ > 1 indicates unconditional stability under the stated criterion.

Stability circle: Locus separating stable and potentially unstable source or load terminations.

Conjugate match: Termination whose impedance is the complex conjugate of the device impedance.

Noise figure: Degradation of signal-to-noise ratio caused by a network, expressed in decibels.

Noise factor: Linear ratio corresponding to noise figure.

Γopt: Source reflection coefficient producing minimum modeled noise figure.

1 dB compression point: Power level at which gain falls 1 dB below its small-signal extrapolation.

Third-order intercept point: Extrapolated metric of third-order nonlinearity.

Power-added efficiency: Added RF output power divided by DC power consumed.

Load-pull: Measurement or simulation method for determining performance versus load impedance.

Reference plane: Physical or modeled plane at which network quantities are defined.

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

Review Questions

Short Answer Questions

1. What physical quantities do S₁₁, S₂₁, S₁₂, and S₂₂ represent?
2. State the unconditional-stability conditions based on K and Δ.
3. Why is ΓS = Γopt important in low-noise amplifier design?
4. Distinguish noise factor from noise figure.
5. Why must matching-network loss be included in an LNA noise budget?

Long Answer Questions

1. Derive and explain the transducer-gain expression for a bilateral microwave two-port.
2. Explain the design and limitations of quarter-wave and stub matching networks.
3. Discuss the trade-off among gain, noise figure, stability, bandwidth, and linearity.
4. Explain how K, Δ, μ, and stability circles are used in a practical design.
5. Describe a complete workflow for designing and validating a microwave power amplifier.

Analytical / Application Questions

1. A transistor has |S₂₁| = 4, |S₁₁| = 0.4, and |S₂₂| = 0.3 and is assumed unilateral and conjugately matched. Calculate the approximate transducer gain in dB.
2. An amplifier has a fundamental output of 10 dBm and a third-order product of −20 dBm. Estimate OIP3 using the low-distortion two-tone approximation.
3. A two-stage receiver has F₁ = 2, G₁ = 10, and F₂ = 5 in linear units. Calculate the cascaded noise factor and explain why the first stage is important.

Multiple Choice Questions with Answers

Question 1

Which parameter represents forward transmission from port 1 to port 2?

A. S₁₁

B. S₁₂

C. S₂₁

D. S₂₂

Correct Answer: C. S₂₁

Question 2

For unconditional stability using Rollett's criterion, which condition is required?

A. K < 1 and |Δ| > 1

B. K > 1 and |Δ| < 1

C. K = 0 and |Δ| = 1

D. K > 0 and |Δ| > 1

Correct Answer: B. K > 1 and |Δ| < 1

Question 3

What is the principal purpose of an input matching network?

A. To transform the source impedance to the desired ΓS

B. To increase DC power consumption

C. To eliminate all transistor noise

D. To force S₁₂ to zero

Correct Answer: A. To transform the source impedance to the desired ΓS

Question 4

For a unilateral device, the approximate transducer gain is expressed as:

A. G_T ≈ G_S + G₀ + G_L

B. G_T ≈ G_S G₀ G_L

C. G_T ≈ G_S/G₀/G_L

D. G_T ≈ G₀ − G_S − G_L

Correct Answer: B. G_T ≈ G_S G₀ G_L

Question 5

Which source reflection coefficient gives minimum noise according to the four-noise-parameter model?

A. ΓS = 0 always

B. ΓS = ΓL

C. ΓS = Γopt

D. ΓS = S₂₂

Correct Answer: C. ΓS = Γopt

Question 6

A 1 dB passive loss placed before the first active stage approximately contributes what noise figure?

A. 0 dB

B. 0.1 dB

C. 1 dB

D. 10 dB

Correct Answer: C. 1 dB

Question 7

If |S₂₁| = 3 for a unilateral amplifier, what is the intrinsic forward power gain?

A. 3

B. 6 dB

C. 9

D. 27

Correct Answer: C. 9

Question 8

A transistor has output fundamental power 10 dBm and third-order product power −20 dBm. What is the estimated OIP3?

A. 15 dBm

B. 20 dBm

C. 25 dBm

D. 30 dBm

Correct Answer: C. 25 dBm

Question 9

Which metric includes both RF output power and RF input power in relation to DC power?

A. Noise figure

B. Power-added efficiency

C. Return loss

D. Rollett factor

Correct Answer: B. Power-added efficiency

Question 10

Why is the first stage of a receiver especially important in Friis' formula?

A. Its noise contribution is divided by preceding gain

B. It has no effect on total noise

C. Later-stage noise is divided by preceding gain

D. It always has the highest DC power

Correct Answer: C. Later-stage noise is divided by preceding gain