Gallium Nitride (GaN)

GaN high-electron-mobility transistors (HEMTs) achieve power densities exceeding 5–12 W/mm of gate periphery, compared to ~1 W/mm for GaAs. This is a direct consequence of the 3.4 eV bandgap and high critical electric field (~3.3 MV/cm), which allows for much higher drain bias voltages (28–48V) without avalanche breakdown.

• Impact on mmWave: Reduces the total semiconductor area required for a given output power, minimizing parasitic capacitances.
• Result: Enables compact power amplifiers (PAs) with higher output impedance, simplifying broadband matching network design at 28 GHz and 39 GHz.

GaN-on-SiC (Silicon Carbide) substrates exhibit a thermal conductivity of ~200–400 W/m·K, which is 3–5x higher than GaAs. This efficient heat extraction is critical because DC-to-RF efficiency directly impacts junction temperature.

• Thermal Memory Effects: High thermal conductivity mitigates slow self-heating dynamics that cause long-term gain and phase drift.
• DPD Simplification: By reducing the magnitude of thermal memory effects, GaN-on-SiC reduces the complexity required in the behavioral model's memory depth, easing the burden on real-time digital predistortion (DPD) coefficient estimation.

The high power density of GaN results in physically smaller transistors with higher output impedance for a given RF power target. This is a distinct advantage for mmWave phased arrays where antenna element spacing is constrained to half-wavelength (~5 mm at 28 GHz).

• Impedance Transformation Ratio: A higher device impedance requires a lower transformation ratio to match to the 50-ohm system, inherently yielding wider bandwidth.
• Doherty Architectures: This facilitates the design of broadband mmWave Doherty PAs, where the impedance inverter and offset lines can be realized with lower loss and higher fractional bandwidth.

A critical trade-off in GaN HEMTs is the presence of charge trapping at the surface, buffer, or barrier layers. These traps capture and emit electrons on a microsecond-to-millisecond timescale, causing gate-lag and drain-lag.

• AM-AM/AM-PM Dispersions: Trapping introduces a dynamic nonlinearity that is a function of the signal's envelope history, not just the instantaneous power.
• DPD Challenge: This necessitates DPD models with long temporal memory (e.g., Generalized Memory Polynomials or LSTM-based neural networks) to compensate for the history-dependent charge state of the traps, especially under wideband modulated signals.

Operating at drain voltages of 28V to 48V, GaN PAs provide a large voltage swing headroom. This directly translates to improved linearity for a given output power back-off (OBO).

• AM-AM/AM-PM Flatness: The intrinsic transconductance (gm) and input capacitance (Cgs) nonlinearities are diluted by the high supply voltage, resulting in a flatter gain profile before compression.
• EVM Floor: The high linear-gain region allows GaN PAs to meet stringent Error Vector Magnitude (EVM) requirements (e.g., 3.5% for 256-QAM) with less aggressive DPD correction, reducing the crest factor expansion risk.

While GaN excels in saturated efficiency, its true advantage for mmWave 5G lies in Power-Added Efficiency (PAE) at deep output back-off (OBO). The high gain allows the driver stage to be driven into compression while the final stage remains linear.

• Envelope Tracking Compatibility: GaN's high breakdown voltage supports supply modulation techniques like Envelope Tracking (ET) to boost average efficiency for high-PAPR signals.
• Thermal Budget: Superior PAE directly reduces the DC power dissipated as heat, preserving the Mean Time To Failure (MTTF) of the phased-array antenna module in dense urban deployments.