Power Amplifier Non-Linearity

The amplitude-to-amplitude conversion curve defines how the PA's output signal magnitude deviates from the ideal linear gain as input power increases. This manifests as gain compression at high drive levels, where the amplifier saturates and can no longer produce proportional output power. The 1 dB compression point (P1dB) marks the input power at which gain drops by 1 dB from the small-signal value.

• Small-signal region: Linear operation with constant gain
• Compression region: Gradual gain reduction as input power increases
• Saturation region: Output power plateaus regardless of input increase
• Modeled using Rapp, Saleh, or polynomial memoryless models

The amplitude-to-phase conversion curve captures the unwanted phase shift imposed on the output signal as a function of the instantaneous input envelope power. Unlike AM-AM distortion, AM-PM is a memory effect that causes constellation rotation and spectral asymmetry. This phase deviation is highly sensitive to the PA's semiconductor physics and biasing network.

• Caused by varactor-like behavior in transistor junction capacitances
• Results in unequal upper and lower sidebands in spectral regrowth
• Critical for modulation schemes with phase information (QPSK, QAM)
• Measured in degrees per dB of input power change

Memory effects occur when the PA's output depends not only on the instantaneous input but also on previous signal states. These are categorized as electrical memory (bias network impedance, trapping effects) and thermal memory (junction temperature fluctuations). Memory effects cause hysteresis in the AM-AM and AM-PM curves, making the distortion signal-history dependent.

• Short-term memory: Envelope frequency-dependent impedance variations
• Long-term memory: Thermal time constants and bias circuit charging
• Creates spread in the AM-AM/AM-PM curves when plotted dynamically
• Modeled using Volterra series or memory polynomial structures

Spectral regrowth is the appearance of signal power in adjacent frequency channels caused by PA non-linearity. When a modulated signal with a non-constant envelope passes through a non-linear amplifier, intermodulation products spread energy beyond the intended bandwidth. This regrowth pattern is a unique, device-specific fingerprint.

• Quantified by Adjacent Channel Leakage Ratio (ACLR)
• The shape and asymmetry of regrowth reveals AM-PM characteristics
• Higher PAPR signals (OFDM) produce more severe regrowth
• Regrowth cannot be removed by linear filtering at the receiver

The Volterra series is the most comprehensive mathematical framework for modeling PA non-linearity with memory. It represents the output as a sum of multi-dimensional convolution integrals, capturing both static non-linearity and dynamic memory effects in a single unified structure.

• 1st-order kernel: Linear impulse response
• 3rd-order kernel: Dominant non-linear term causing spectral regrowth
• 5th-order and higher: Capture severe compression behavior
• Complexity grows exponentially with order; pruned Volterra models reduce coefficients
• Forms the theoretical basis for digital pre-distortion (DPD) algorithms

The Saleh model is a widely-used two-parameter memoryless behavioral model originally developed for traveling-wave tube amplifiers (TWTAs) but adapted for solid-state PAs. It provides closed-form expressions for both AM-AM and AM-PM conversion using only four scalar parameters.

• AM-AM: A(r) = αₐr / (1 + βₐr²)
• AM-PM: Φ(r) = αᵩr² / (1 + βᵩr²)
• Parameters (αₐ, βₐ, αᵩ, βᵩ) are extracted from measured data
• Computationally efficient for real-time simulation and synthetic data generation
• Does not capture memory effects; used for narrowband signals

The simulated non-linear relationship between input signal amplitude and output signal amplitude in a power amplifier. This impairment causes gain compression at high power levels, where the amplifier saturates and can no longer produce a proportional output. In synthetic generation, AM-AM curves are parameterized by metrics like the 1 dB compression point (P1dB) and third-order intercept point (IP3). The distortion creates in-band signal warping and generates spectral regrowth into adjacent channels, making it a primary source of unique, device-specific fingerprint features.

A critical memory effect where the input signal amplitude causes an unwanted phase shift in the output signal. Unlike AM-AM distortion, which affects magnitude, AM-PM introduces phase modulation that rotates the constellation diagram dynamically. This impairment is particularly significant in spectrally efficient modulations like QAM and OFDM. Synthetic models replicate AM-PM conversion using a phase shift vs. input power curve, often expressed in degrees per dB. The resulting constellation warping provides a rich, device-specific signature for deep learning fingerprinting models.

A mathematical framework for modeling non-linear dynamic systems with memory, such as power amplifiers. The Volterra series represents the amplifier output as a sum of multi-dimensional convolution integrals, capturing both instantaneous non-linearity and time-dependent memory effects. Key components include:

• Linear kernel: Standard filtering behavior
• Second-order kernel: Quadratic non-linearity and memory
• Third-order kernel: Cubic distortion, dominant for intermodulation This model is computationally intensive but provides the highest fidelity for generating realistic synthetic impairments in digital twins.

A critical metric defining the ratio of power leaked into adjacent frequency channels versus the power in the assigned channel, caused by transmitter non-linearity. ACLR is measured in dBc (decibels relative to carrier) and is a primary regulatory compliance parameter for wireless standards like 3GPP LTE and 5G NR. In synthetic impairment generation, ACLR serves as a validation metric to ensure the simulated spectral regrowth matches real hardware behavior. Typical values range from -30 dBc to -45 dBc depending on the amplifier class and linearization quality.

A signal characteristic representing the ratio of instantaneous peak power to average power, typically expressed in dB. High PAPR signals, common in OFDM and multi-carrier waveforms, drive power amplifiers into non-linear saturation during peaks. This causes clipping distortion and spectral regrowth. In synthetic generation, PAPR is a critical simulation parameter:

• Low PAPR (0-3 dB): Constant envelope modulations like GMSK
• Medium PAPR (3-6 dB): Single-carrier QPSK
• High PAPR (8-13 dB): OFDM with many subcarriers Crest factor reduction techniques are often modeled alongside PA non-linearity to create realistic device signatures.

Residual, unique signal distortions that remain after a linearization algorithm compensates for power amplifier non-linearity. DPD applies an inverse model of the PA's transfer function to the input signal, but imperfect estimation and limited polynomial order leave behind subtle artifacts. These residuals form a highly identifiable hardware signature because they are specific to the interaction between the DPD algorithm and the physical amplifier. Synthetic generation must model both the PA non-linearity and the DPD correction loop to produce realistic post-compensation fingerprints for secure device authentication.