AM-AM Distortion

As the input drive level increases toward the amplifier's saturation point, the output amplitude stops increasing linearly. The gain compresses, typically defined by the 1 dB compression point (P1dB) where actual gain drops 1 dB below the ideal linear gain. This is the most common AM-AM characteristic in Class AB and Class B amplifiers.

The static nonlinearity is characterized by plotting instantaneous output amplitude vs. instantaneous input amplitude. Key features include:

• Linear region: Small-signal operation with constant gain
• Compression region: Gradual gain reduction as saturation approaches
• Saturation region: Output amplitude plateaus regardless of input increase This curve is the foundation for memoryless behavioral models and Look-Up Table (LUT) predistorters.

AM-AM nonlinearity generates odd-order intermodulation products when a multi-tone or modulated signal passes through the amplifier. These products fall in-band and in adjacent channels, causing:

• Spectral regrowth measured by Adjacent Channel Power Ratio (ACPR)
• In-band distortion degrading Error Vector Magnitude (EVM) Third-order products (IM3) are typically the most problematic due to their proximity to the carrier.

The Rapp model provides a mathematically tractable AM-AM transfer function specifically for solid-state power amplifiers (SSPAs). It characterizes the smooth transition from linear operation to hard saturation using a smoothness factor (p). A higher p-value produces a sharper saturation knee, typical of modern GaN and LDMOS amplifier technologies.

The Saleh model was originally developed for Traveling Wave Tube Amplifiers (TWTAs) used in satellite communications. It uses a two-parameter rational function to capture both AM-AM and AM-PM distortion simultaneously. The AM-AM component exhibits a characteristic peak-then-compress behavior distinct from the monotonic compression of solid-state devices.

AM-AM distortion causes constellation warping in digitally modulated signals:

• Outer constellation points compress inward due to gain reduction at high amplitudes
• QAM and APSK schemes are particularly vulnerable
• Results in increased Symbol Error Rate (SER) and degraded Bit Error Rate (BER) This is distinct from AM-PM distortion, which causes rotational constellation errors.

The dependence of a power amplifier's current output on past input values due to:

• Thermal memory: Die temperature changes from recent signal peaks
• Electrical memory: Bias circuit impedance variations at envelope frequencies
• Trapping effects: Charge capture/release in semiconductor defects

Memory effects make AM-AM distortion frequency-dependent, requiring models beyond simple memoryless nonlinearity.

A simplified Volterra series that retains only diagonal terms to efficiently represent nonlinear memory effects. The model captures AM-AM and AM-PM distortion across multiple time delays while maintaining computational tractability for FPGA implementation. Its generalized form adds cross-terms between different delays and nonlinear orders for improved accuracy with strong memory effects.

The primary regulatory metric quantifying spectral regrowth caused by AM-AM distortion. Measured as the ratio of power leaked into adjacent frequency channels versus the main channel power. Typical requirements:

• -45 dBc for 3GPP LTE/5G NR base stations
• -60 dBc for stringent military applications

DPD systems target ACPR improvement of 15-25 dB.

A lowpass equivalent signal representation that captures both in-phase (I) and quadrature (Q) components while omitting the high-frequency carrier. This representation is essential for behavioral modeling because AM-AM distortion manifests as a nonlinear mapping of the complex envelope magnitude to the output envelope, simplifying simulation and DPD coefficient extraction.

A memory-mapping approach that stores predistortion gain values in a table indexed by instantaneous input amplitude. Simple LUTs correct memoryless AM-AM distortion efficiently, while 2D LUTs incorporating past envelope values can address memory effects. Widely used in real-time FPGA implementations due to deterministic latency and low computational overhead compared to polynomial evaluation.