AM-AM Distortion

As input drive level increases toward the amplifier's saturation point, the incremental gain decreases progressively. This gain compression occurs because the active device's transconductance reduces at large signal swings. The 1 dB compression point (P1dB) marks where gain drops by exactly 1 dB from the small-signal value, serving as a critical boundary between quasi-linear and nonlinear operation. Beyond P1dB, each additional input power increment produces diminishing output power returns until the amplifier reaches full saturation where output power plateaus completely.

The AM-AM characteristic curve plots normalized output amplitude against normalized input amplitude. An ideal linear amplifier produces a straight 45-degree line through the origin. Real amplifiers deviate from this line in predictable patterns:

• Class A amplifiers: Exhibit soft, gradual compression with a rounded knee
• Class AB amplifiers: Show moderate compression with sharper transition near saturation
• Doherty amplifiers: Display complex multi-stage compression due to carrier and peaking amplifier interaction
• GaN devices: Often demonstrate sharper saturation knees compared to LDMOS

AM-AM behavior divides into two distinct operating regimes separated by the compression threshold. In the small-signal regime, output amplitude scales linearly with input, and the amplifier behaves as a linear gain block. In the large-signal regime, nonlinear mechanisms dominate: clipping from voltage rail limitations, current starvation in the active device, and load-line modulation all contribute to amplitude-dependent gain variation. The transition between these regimes is not abrupt but follows a gradual roll-off determined by the amplifier's linearity figure of merit.

AM-AM distortion directly generates odd-order harmonics and intermodulation products. A single-tone input experiencing amplitude nonlinearity produces harmonics at integer multiples of the fundamental frequency. For bandpass communication signals, the third-order intermodulation products (IM3) fall within or adjacent to the operating band, causing spectral regrowth. The amplitude of these distortion products follows a predictable slope: third-order products grow at 3 dB per 1 dB of fundamental power increase, making them the dominant source of adjacent channel interference.

Pure AM-AM distortion is considered memoryless when the instantaneous output amplitude depends only on the instantaneous input amplitude, independent of prior signal history. However, practical amplifiers exhibit quasi-memoryless behavior where AM-AM characteristics shift subtly with envelope frequency due to bias network impedance variations and device parasitics. This frequency-dependent compression means the AM-AM curve measured with a slow ramp differs from that measured with a wideband modulated signal, complicating predistorter design for signals with high peak-to-average ratios.

AM-AM distortion degrades Error Vector Magnitude (EVM) by compressing constellation points at higher amplitudes while leaving lower-amplitude points relatively unaffected. For 256-QAM and higher-order modulations used in 5G NR, even modest compression creates asymmetric constellation warping that closes the decision boundaries between adjacent symbols. The outer constellation points experience the greatest displacement, creating a characteristic 'pinched' appearance in constellation diagrams. This amplitude-dependent degradation is distinct from the rotational errors caused by AM-PM conversion.

The companion distortion mechanism to AM-AM, where the phase shift introduced by a power amplifier varies as a function of the instantaneous input signal amplitude. While AM-AM distorts the magnitude, AM-PM distorts the phase, causing spectral regrowth and constellation rotation. In wideband mmWave systems, AM-PM is often the dominant source of nonlinearity and must be compensated jointly with AM-AM for effective linearization.

A DPD training method that identifies the predistorter by placing it after the power amplifier model in the estimation loop. The ILA avoids the need to compute an explicit inverse of the AM-AM characteristic:

• The postdistorter coefficients are estimated to linearize the PA output
• These coefficients are copied directly to the predistorter
• Assumes the postdistorter and predistorter are interchangeable This architecture is widely used because it simplifies the coefficient extraction problem to a standard system identification task.

The amount by which a power amplifier's average output power is reduced below its saturation point to operate in a more linear region of the AM-AM characteristic. Key tradeoffs:

• Higher OBO: Better linearity, lower efficiency (PAE drops significantly)
• Lower OBO: Higher efficiency, severe AM-AM compression
• Typical values: 6-12 dB for OFDM signals without DPD Digital predistortion enables operation at lower OBO by compensating for the resulting AM-AM distortion, directly improving power-added efficiency.

Slowly varying changes in power amplifier gain caused by self-heating and substrate temperature fluctuations dependent on signal history. These effects modulate the AM-AM characteristic over time:

• Short-term: Die-level heating from instantaneous power dissipation
• Long-term: Package-level thermal time constants (milliseconds to seconds)
• Impact: Causes the AM-AM curve to shift dynamically, requiring adaptive DPD GaN-on-SiC amplifiers exhibit particularly strong thermal memory due to high power density and the nonlinear thermal conductivity of the substrate.

The variation in load impedance seen by each power amplifier in a phased array due to beam-steering, causing channel-specific AM-AM distortion. As the beam is electronically steered:

• Mutual coupling between antenna elements changes
• Each PA sees a different complex load impedance
• The AM-AM characteristic becomes angle-dependent This is a critical challenge for mmWave DPD, requiring either per-element linearization or over-the-air techniques that capture the combined array nonlinearity.