ET-DPD for Doherty PAs

A specialized modeling framework that captures the distinct nonlinear behaviors of the carrier and peaking amplifier paths under dynamic supply modulation. Unlike single-input models, this approach treats the Doherty PA as a two-branch system with:

• Carrier amplifier: Operates in Class-AB with supply-dependent gain compression
• Peaking amplifier: Operates in Class-C with turn-on threshold nonlinearity
• Impedance modulation: Load-pulling effects between branches vary with envelope tracking voltage

The model accepts both the RF input signal and the dynamic supply voltage as independent variables, accurately predicting the composite output distortion including AM/AM and AM/PM components.

Doherty PAs rely on active load modulation between the carrier and peaking amplifiers to achieve high efficiency at back-off power levels. Under envelope tracking, the dynamic supply voltage alters the impedance presented to each amplifier, creating supply-dependent load trajectories that must be characterized and inverted.

Key considerations include:

• Carrier impedance shift: Varies with both input drive and instantaneous drain voltage
• Peaking turn-on point: Shifts with supply voltage, altering the Doherty transition region
• Phase discontinuity: Abrupt phase changes occur at the peaking amplifier activation threshold

The DPD must compensate for these modulation-dependent impedance variations to maintain linearity across the full ET operating range.

Doherty PAs exhibit asymmetric memory effects due to the different thermal and trapping characteristics of the carrier and peaking amplifier devices. Under envelope tracking, these effects become more pronounced:

• Carrier thermal memory: Continuous conduction creates long-term thermal time constants
• Peaking trapping effects: GaN HEMT devices exhibit gate-lag and drain-lag under pulsed Class-C operation
• ET-induced memory: Supply modulator dynamics introduce additional frequency-dependent memory

The ET-DPD model must incorporate multi-branch memory polynomial structures with separate memory depth allocations for each amplifier path, capturing both short-term (nanosecond) and long-term (microsecond) memory effects.

A three-dimensional predistortion LUT indexed by instantaneous input power, supply voltage, and the Doherty operating region (back-off vs. saturation). This structure compensates for the static nonlinearities unique to ET-Doherty systems:

• Back-off region: Carrier amplifier dominates; LUT corrects supply-dependent gain compression
• Transition region: Both amplifiers active; LUT compensates for impedance modulation nonlinearity
• Saturation region: Peaking amplifier fully on; LUT addresses combined compression characteristics

The 3D LUT provides memoryless correction with fast lookup times suitable for real-time implementation, while augmented Volterra terms handle residual memory effects.

The Class-C peaking amplifier introduces a hard nonlinearity at its turn-on threshold that creates severe distortion in the Doherty transition region. Under envelope tracking, this threshold becomes supply-dependent, complicating linearization:

• Gain expansion: Peaking amplifier gain increases rapidly after turn-on, creating AM/AM kinks
• Phase jump: Abrupt phase shift of 30-60 degrees occurs at peaking activation
• ET interaction: Dynamic supply voltage shifts the turn-on point, requiring adaptive compensation

The ET-DPD must apply pre-distortion gain expansion in the transition region to cancel the peaking amplifier's nonlinear turn-on characteristic while maintaining stability.

An extension of the Volterra series that incorporates dynamic supply voltage terms and Doherty-specific cross-terms to capture the complex nonlinear interactions between the carrier and peaking paths under envelope tracking:

• Supply-dependent kernels: Model gain variation as a function of instantaneous drain voltage
• Cross-branch terms: Capture intermodulation products from carrier-peaking interaction
• ET-Doherty cross-terms: Account for the combined effect of supply modulation and load modulation

The augmented Volterra model provides high-fidelity linearization for wideband signals where memory effects dominate, though at the cost of increased computational complexity compared to memory polynomial approaches.

A load modulation architecture combining a carrier amplifier (biased Class AB) and a peaking amplifier (biased Class C) connected via an impedance inverter. At low power, only the carrier operates at high efficiency. At peak power, the peaking amplifier turns on, modulating the load impedance seen by the carrier to maintain efficiency over a 6-9 dB back-off range. This active load-pull mechanism creates a unique, power-dependent nonlinearity profile that ET-DPD must characterize and invert.

The dynamic interaction between the carrier and peaking amplifier stages creates a dual-path nonlinearity that single-branch models fail to capture. Key effects include:

• Impedance modulation: The peaking amplifier's turn-on characteristic dynamically changes the carrier's load line
• Phase discontinuity: The peaking path introduces a phase shift at the combining node that varies with drive level
• AM/PM asymmetry: The carrier and peaking amplifiers exhibit different phase distortion profiles under ET

A dual-input behavioral model is required to independently characterize each path before combining.

A memoryless predistortion structure indexed by instantaneous input power and instantaneous supply voltage to apply a complex gain correction. For Doherty PAs, the 3D LUT must capture the discontinuous gain expansion that occurs when the peaking amplifier transitions from cutoff to active operation. The LUT is populated using iso-gain contours measured across the full supply voltage range, with denser interpolation points around the peaking turn-on threshold to prevent spectral regrowth at the efficiency sweet spot.

An extension of the Volterra series that incorporates dynamic supply voltage terms and cross-path memory effects to model the Doherty PA under envelope tracking. The model includes:

• Carrier path kernels: Standard Volterra terms with supply-dependent coefficients
• Peaking path kernels: Terms activated only above the peaking threshold
• Cross-path kernels: Interaction terms capturing the impedance modulation effect of the peaking amplifier on the carrier's memory response

This structure enables a single predistorter to compensate for the entire Doherty ET transmitter chain.

A critical distortion mechanism in ET-Doherty systems where the supply voltage modulation alters the peaking amplifier's turn-on threshold. As the drain voltage drops during envelope tracking, the peaking amplifier's Class C bias point shifts, causing it to turn on earlier or later than designed. This creates a supply-dependent gain expansion that changes shape dynamically with the envelope. The DPD model must include supply-dependent threshold terms to track and pre-compensate for this time-varying nonlinearity.

An adaptive architecture using a feedback observation receiver to continuously monitor the Doherty PA output and update predistortion coefficients in real-time. The closed-loop system must:

• Track thermal drift: Doherty PAs exhibit asymmetric heating between carrier and peaking stages
• Compensate for aging: GaN device trapping and degradation shift the peaking threshold over time
• Adapt to load mismatch: Antenna VSWR changes alter the impedance inverter's behavior

The coefficient update rate must be fast enough to track envelope-rate supply modulation effects while maintaining stability.