Reconfigurable Power Amplifiers: How Switched Matching Networks Extend PA Efficiency Across Bands
A conventional RF power amplifier is optimized at one operating point: one band, one load impedance, one power level. A phone, base station or jammer that has to work across several bands and power levels therefore carries several PAs, or runs one PA backed off far from its efficient peak. A reconfigurable power amplifier changes that trade by retuning its matching network, bias or output structure on command — and most practical architectures do it with RF switches and passives rather than with exotic transistor technology.
This article explains how switched matching makes a PA reconfigurable, what the MEMS switch contributes, and how the approach compares with fixed and fully tunable alternatives.
Why a fixed PA cannot be efficient everywhere
A PA delivers peak efficiency only when the transistor sees the optimum load impedance at the fundamental (and properly terminated harmonics for switching classes). Move frequency, power or antenna VSWR, and that load match is lost. Efficiency in backed-off class-AB operation collapses, and a narrowband PA delivered into a mismatched multi-band antenna can lose several dB of output power or fail reliability tests.
Two classic responses are:
- use multiple PAs plus filters and switches (the carrier-aggregation phone model);
- accept broadband but lower-efficiency operation with distributed or balanced PA structures.
Reconfigurable PAs take a third path: keep one (or fewer) PA cores and move its optimum impedance around electrically.
The anatomy of a reconfigurable PA
Beyond the transistor itself, a reconfigurable PA typically contains three tunable blocks.
1. Switched input/interstage matching. Capacitor or inductor banks selectable by RF switches change the input match and interstage impedance so that each gain band stays flat and stable.
2. Reconfigurable output matching network (OMN). This is where most of the efficiency gain lives. The output match is built as a ladder or pi network in which shunt capacitors, series inductors or transmission-line stubs are switched in and out. An ohmic RF MEMS switch closes a path to a matched component with under 0.2 dB insertion loss and negligible quiescent current — far better than a PIN diode in both loss and DC budget, and more linear than an SOI FET at high power.
3. Bias and harmonic control. Switched or tunable terminations retune class-of-operation (for example selecting continuous-class modes) and harmonic traps for different bands.
Switch technology inside the PA output network
| Switch option | Insertion loss | Linearity | DC power | Practical limit |
|---|---|---|---|---|
| PIN diode | Low | Moderate (junction) | High bias current | Hot switching fine, but inefficient in handsets |
| SOI CMOS FET | Moderate, scaling with R_on | Limited at high voltage swing | Low | Voltage breakdown caps PA power |
| RF MEMS ohmic | Very low | Excellent (no junction) | Near zero (electrostatic) | Hot-switch life and packaging |
| MEMS varactor / BST | Low at mmWave | Good | Near zero | Tuning range and linearity limits |
At the PA output, where even 0.5 dB of switch loss directly cuts radiated power and heats the handset, the near-zero on-resistance and absence of junction curvature make MEMS contacts the preferred element for switched-LC and switched-line topologies. Capacitive MEMS structures appear in the input match and at mmWave, where no DC path is needed.
Common reconfiguration architectures
- Discrete band switching. The PA stores N pre-designed matching states, one per band. Simple, robust, and the closest analog to multiple PAs with shared transistor area. Filter banks in the front-end are switched in parallel using reconfigurable filters.
- Continuous (load-modulated) matching. Varactors and switched banks interpolate across a region of the Smith chart, tracking antenna VSWR. More flexible but harder to guarantee stability and linearity over all states.
- Doherty / outphasing with switched elements. Switched quarter-wave lines and compensating stubs reconfigure the load-modulation behavior itself, extending the efficient back-off region band-to-band.
- Power-level reconfiguration. Switching in/out PA cells or matching sections lets a small efficient PA handle low power and a stacked array handle peak power — useful for waveforms with high PAPR.
Design pitfalls
- Switch hot voltage. The PA output swings far above the supply voltage. Switch off-state voltage rating and the hot-switching power rating must be checked against peak envelope voltage, not average power.
- State-dependent stability. Every matching state is effectively a different amplifier; Rollett K-factor and loop stability must be verified per state, including bias transients during switching.
- Control and settling time. MEMS switching in microseconds is fast for band changes, but envelope tracking needs nanoseconds — switched matching targets band and mode changes, not symbol-level tuning.
- Loss accounting. The "reconfigurable PA is better" claim is only true if the switched network's loss plus its control overhead is below the loss of the alternative (extra PAs, isolators, diplexers). Do the system-level power budget.
- Tolerance and repeatability. Switched states using precision thin-film passives on a stable carrier (glass IPD or alumina) are far more repeatable state-to-state than assemblies of discrete 0402 components.
Where reconfigurable PAs ship today
Multi-mode, multi-band cellular transmit chains; wideband military and EW amplifiers covering several octaves; software-defined radios; satellite communications terminals that uplink across multiple operators; and small-cell infrastructure all use switched matching in production. The common pattern: a modest number of well-chosen discrete states built with low-loss MEMS switches and tight-tolerance passives beats either an enormous bank of fixed PAs or an aggressive continuously tunable design.
For component selection, start from the output network: define bands, power per band, VSWR conditions and allowable switching time, then choose switches and passives whose loss leaves the PA efficiency budget intact.