AM-PM Distortion

The fundamental mechanism of AM-PM distortion is the variation of the amplifier's transmission phase (∠S21) with input drive level. As the instantaneous envelope power increases, the phase shift through the device changes due to nonlinear input capacitance (varactor effects) and bias-dependent transconductance in the active device.

• At low input power, the amplifier operates in small-signal mode with a baseline phase shift
• As the envelope approaches compression, the phase typically advances or lags by several degrees
• In GaN HEMT devices, phase shifts of 5-15 degrees are common near the 1dB compression point
• The phase variation is often modeled as a polynomial function of instantaneous envelope power: Δφ = Σ kₙ|v(t)|²ⁿ

AM-PM distortion is the primary cause of spectral asymmetry in amplified signals. While AM-AM distortion produces symmetric spectral regrowth, the combination of amplitude and phase nonlinearities creates upper and lower sidebands with unequal power levels.

• The asymmetry results from the interaction between the nonlinear phase response and the signal's complex modulation
• Upper and lower adjacent channel power can differ by 2-5 dB due to AM-PM effects alone
• This asymmetry is particularly problematic for regulatory compliance where both sidebands must meet ACLR limits
• Memory effects in the AM-PM response create frequency-dependent asymmetry that varies across the modulation bandwidth

AM-PM distortion causes signal-dependent phase rotation that directly degrades modulation quality. Higher-amplitude constellation points experience greater phase shifts than lower-amplitude points, producing a characteristic 'spiral' distortion pattern in the constellation diagram.

• Outer constellation points rotate more than inner points, creating a non-uniform phase error distribution
• For 256-QAM signals, AM-PM distortion can increase EVM by 2-4 percentage points
• The phase error is correlated with instantaneous envelope amplitude, making it predictable but requiring nonlinear correction
• Unlike thermal noise, AM-PM-induced phase errors are deterministic and can be compensated by digital predistortion with phase correction capability

The AM-PM characteristic is strongly dependent on the amplifier's DC bias conditions. The operating class and quiescent point determine both the magnitude and direction of phase conversion.

• Class A amplifiers exhibit relatively low AM-PM conversion due to constant current draw, but suffer from poor efficiency
• Class AB amplifiers show moderate AM-PM with phase typically advancing as the device transitions toward Class B operation during envelope peaks
• Class C amplifiers demonstrate severe AM-PM distortion as the conduction angle varies dramatically with drive level
• Doherty amplifiers have complex AM-PM characteristics due to the interaction between carrier and peaking amplifiers, often requiring dedicated phase alignment in the DPD

AM-PM distortion exhibits significant frequency-dependent memory effects caused by thermal dynamics, trapping phenomena, and bias circuit impedance variations. These memory effects make the instantaneous phase shift dependent on both current and past envelope values.

• Thermal memory: Device junction temperature changes with average power, altering the AM-PM characteristic over millisecond timescales
• Electrical memory: Bias network impedance at the envelope frequency modulates the instantaneous bias point, creating dynamic AM-PM variations
• Trapping effects: In GaN and GaAs devices, surface and buffer traps introduce long time-constant (microseconds to milliseconds) variations in phase response
• Memory effects require Volterra series or memory polynomial models rather than static AM-PM lookup tables for accurate compensation

The amplitude-to-amplitude nonlinearity that compresses gain at high input power levels. While AM-PM distorts phase, AM-AM distorts magnitude—both contribute to spectral regrowth. AM-AM is characterized by the 1dB compression point (P1dB) and is typically modeled alongside AM-PM in Saleh or Rapp behavioral models. Unlike AM-PM, AM-AM distortion is often symmetric and easier to compensate with gain expansion predistortion.

A phenomenon where a power amplifier's current output depends on past input states, not just the instantaneous envelope. Memory effects cause frequency-dependent AM-AM and AM-PM distortion, making the phase shift a function of both signal amplitude and modulation bandwidth. Sources include:

• Thermal memory: Die temperature changes with signal envelope history
• Electrical memory: Bias circuit impedance variations at envelope frequencies
• Trapping effects: Slow charge capture/release in GaN HEMT devices Memory effects create asymmetric spectral regrowth that static AM-PM models cannot capture.

The primary regulatory compliance metric that AM-PM distortion directly degrades. ACLR measures the ratio of in-channel power to power leaking into adjacent channels, typically specified at -45 dBc for 3GPP LTE/5G NR. AM-PM distortion creates asymmetric ACLR—the upper and lower adjacent channels exhibit different leakage levels due to the phase nonlinearity interacting with the complex modulation envelope. This asymmetry is a telltale signature of AM-PM impairment.

A modulation quality metric that quantifies the vector difference between ideal constellation points and actual transmitted symbols. AM-PM distortion rotates constellation points by an amplitude-dependent phase angle, directly increasing EVM. For 256-QAM and 1024-QAM in 5G NR, EVM requirements are stringent (≤3.5%), making AM-PM compensation critical. EVM captures in-band distortion while ACLR captures out-of-band effects—both must be optimized simultaneously.

A figure of merit for amplifier linearity derived from two-tone testing. While IP3 primarily characterizes AM-AM nonlinearity through IMD3 products, it correlates with AM-PM performance because both originate from the same device physics. Amplifiers with higher OIP3 (Output IP3) typically exhibit lower AM-PM conversion coefficients. IP3 is extrapolated from low-power measurements and provides a single-number linearity benchmark useful for comparing PA technologies.

The primary linearization technique used to compensate for both AM-AM and AM-PM distortion. A DPD system applies an inverse nonlinearity to the baseband signal before the PA, such that the cascade of DPD + PA produces a linear output. Modern DPD architectures use memory polynomial models or neural networks to jointly correct amplitude and phase distortion across frequency. Effective AM-PM compensation requires the DPD model to capture the amplitude-dependent phase rotation with high precision.