AM-PM Distortion

AM-PM distortion arises primarily from the voltage-dependent capacitance of transistor junctions within the power amplifier. As the input signal amplitude swings, the depletion region width in the semiconductor varies, changing the junction capacitance. This variable capacitance alters the phase of the signal passing through the transistor, creating an amplitude-dependent phase shift. Key contributing factors include:

• Varactor-like behavior of the base-collector or gate-drain junction
• Non-linear input capacitance (Cgs in FETs, Cbe in BJTs) that varies with signal swing
• Self-heating effects that dynamically alter electron mobility and transit time
• Knee voltage modulation causing phase lag at high instantaneous power levels

The exact capacitance-voltage (C-V) curve is a direct consequence of the specific doping profile and physical geometry of the transistor, which varies microscopically even between devices on the same wafer.

Power amplifier non-linearity is fully characterized by two distinct but coupled conversion mechanisms:

• AM-AM Distortion: The non-linear relationship between input amplitude and output amplitude, causing gain compression at high power levels. This is the magnitude-domain distortion.
• AM-PM Distortion: The non-linear relationship between input amplitude and output phase, causing phase shift that varies with signal envelope power. This is the phase-domain distortion.

While AM-AM is often corrected through gain expansion techniques, AM-PM is more subtle and harder to compensate, making it a more persistent and unique fingerprint. The two effects are not independent—they both stem from the same underlying device physics—but their specific trajectories differ per device, creating a two-dimensional non-linearity signature that is exceptionally difficult to clone.

For modern spectrally efficient modulation schemes with high peak-to-average power ratios (PAPR), AM-PM distortion is particularly damaging and revealing:

• Constellation Warping: In QAM and PSK modulations, AM-PM causes constellation points to rotate by different amounts depending on their distance from the origin. Outer points experience greater phase rotation than inner points, creating a spiral-like distortion pattern unique to each amplifier.
• Spectral Regrowth: The phase modulation induced by amplitude variations creates intermodulation products that broaden the transmitted spectrum into adjacent channels.
• EVM Degradation: The phase error component of Error Vector Magnitude increases, with the distribution of phase errors across constellation points forming a device-specific histogram.
• OFDM Vulnerability: Orthogonal Frequency Division Multiplexing signals, with their high PAPR, are especially susceptible, as each subcarrier experiences a different instantaneous amplitude and thus a different phase shift.

AM-PM distortion is not purely instantaneous—it exhibits memory effects where the current phase shift depends on previous signal amplitudes. This history-dependent behavior creates a hysteresis-like phase trajectory that is highly individual to each amplifier:

• Thermal Memory: Signal amplitude changes cause instantaneous power dissipation variations, heating or cooling the transistor junction with time constants in the microsecond to millisecond range. Temperature changes alter electron mobility, which in turn shifts the AM-PM characteristic.
• Electrical Memory: Bias circuit impedances and decoupling capacitors create low-frequency poles that cause the transistor's operating point to shift dynamically with the signal envelope. This dynamic bias modulation changes the junction capacitance and thus the AM-PM curve.
• Trapping Effects: In GaAs and GaN devices, charge trapping in surface states and buffer layers creates time constants from nanoseconds to seconds, introducing a complex, multi-rate memory signature.

These memory effects transform a simple static AM-PM curve into a multi-dimensional dynamic signature that is extraordinarily difficult to replicate.

AM-PM distortion is a first-order fingerprinting feature because it directly reflects the analog physics of the individual transistor die. Extraction methods include:

• Constellation Analysis: Measuring the phase deviation of each received symbol relative to its ideal position and correlating it with the symbol's instantaneous amplitude. The resulting scatter plot reveals the AM-PM transfer function.
• Two-Tone Testing: Injecting two closely spaced tones and measuring the phase of the resulting intermodulation products. The asymmetry in upper and lower sideband phases indicates AM-PM conversion.
• Complex Gain Modeling: Fitting a polynomial or Volterra series model where the complex gain (magnitude and phase) is a function of instantaneous input power. The polynomial coefficients form a compact fingerprint vector.
• Deep Learning Feature Extraction: Training convolutional neural networks on raw I/Q samples to autonomously learn the AM-PM distortion manifold without explicit parametric modeling.

The phase-only nature of this impairment makes it robust against channel amplitude fading, as frequency-flat fading affects magnitude but preserves relative phase relationships.

In traditional communications, AM-PM distortion is an impairment to be corrected through digital pre-distortion (DPD). In fingerprinting, it is a feature to be preserved and analyzed:

• DPD Approach: A lookup table or polynomial is trained to apply the inverse phase rotation before the signal reaches the amplifier, linearizing the output. Modern DPD systems can suppress AM-PM by 20-30 dB.
• Fingerprinting Approach: The residual AM-PM after DPD—or the raw AM-PM in systems without DPD—is treated as a biometric. Even with DPD, imperfect compensation leaves a residual signature because DPD cannot perfectly track memory effects and thermal drift.
• Adversarial Consideration: A sophisticated adversary could attempt to measure and replicate a target's AM-PM profile, but the memory effects and thermal dynamics make real-time cloning extremely difficult without possessing the exact same physical hardware.
• Dual-Use Systems: Advanced systems can simultaneously use DPD for spectral compliance while extracting the residual AM-PM error signal for continuous device authentication.

AM-PM distortion creates a unique phase-modulation fingerprint that cannot be cloned or extracted from a device's digital memory. Unlike MAC addresses or digital certificates, this analog impairment is physically unclonable because it arises from sub-micron manufacturing variances in the power amplifier's semiconductor lattice.

• Enables zero-trust authentication at the waveform level before demodulation
• Immune to replay attacks since the signature is an inherent property of the transmission chain
• Operates transparently without requiring cryptographic handshakes or key exchanges

Genuine integrated circuits exhibit statistically consistent AM-PM distortion profiles determined by the foundry's process control. Counterfeit or remarked components display measurably different phase-distortion curves due to variations in doping profiles and gate oxide thickness.

• Detects remarked or recycled ICs that pass functional testing but fail analog fingerprint verification
• Compares measured AM-PM transfer functions against a golden device database from trusted fabrication batches
• Critical for defense and aerospace procurement where supply chain integrity is paramount

Malicious actors deploy IMSI catchers and fake cell towers to intercept mobile communications. These rogue base stations can mimic protocol-level identifiers but cannot replicate the AM-PM distortion signature of legitimate carrier-grade equipment.

• User equipment can passively fingerprint the serving cell's downlink signal
• Detects man-in-the-middle attacks by comparing real-time distortion measurements against a whitelist of authorized base station signatures
• Provides physical-layer situational awareness for secure communications in contested environments

AM-PM distortion enables zero-touch provisioning where a network controller authenticates a new IoT device solely from its initial transmission burst. This eliminates the need for pre-shared keys or manual certificate installation.

• Few-shot enrollment: A single transmission capture can extract sufficient AM-PM features for initial classification
• Tracks device aging and drift by monitoring gradual changes in the phase-distortion curve over months of operation
• Automatically flags devices exhibiting anomalous signature shifts indicative of hardware tampering or replacement

Regulatory agencies use AM-PM distortion analysis to identify specific transmitter units causing harmful interference, even when the interfering device transmits sporadically or uses spoofed identifiers.

• Correlates interference events to a single physical transmitter by matching phase-distortion fingerprints across multiple monitoring stations
• Distinguishes between multiple identical-model devices operating in the same band
• Provides forensic evidence for enforcement actions by demonstrating a unique hardware link between the interfering signal and the seized equipment

While digital identifiers can be trivially spoofed, replicating a specific device's AM-PM distortion curve requires physically modifying or replacing the power amplifier with an identically defective unit—a practically insurmountable barrier.

• The non-linear memory effect component of AM-PM distortion is particularly resistant to emulation because it depends on thermal time constants and parasitic capacitances deep within the semiconductor
• Multi-dimensional fingerprinting combining AM-PM with AM-AM and phase noise creates a signature space too complex for real-time adversarial synthesis
• Provides defense-in-depth when layered with conventional cryptographic authentication