Phase Error

The probability density function (PDF) of phase error over thousands of symbols forms a hardware-specific signature. While ideal transmitters exhibit Gaussian distributions, real devices show skewness, kurtosis, and heavy tails caused by amplifier memory effects and power supply ripple. Two identical-model radios can be distinguished by comparing their phase error histograms using Kullback-Leibler divergence or Wasserstein distance metrics.

Phase error is not uniform across the constellation. Each symbol transition produces a unique dynamic phase overshoot and settling pattern determined by the transmitter's PLL bandwidth and loop filter components. Key characteristics include:

• QPSK corner symbols often exhibit larger peak phase errors than inner symbols
• Consecutive identical symbols show reduced error due to reduced slew rate demand
• Diagonal transitions stress both I and Q paths simultaneously, revealing I/Q timing skew

Phase error at the symbol level is the integrated phase noise over the PLL loop bandwidth. Each local oscillator has a unique phase noise mask with distinct:

• 1/f³ corner frequency where flicker noise dominates
• PLL peaking at the loop bandwidth edge
• Reference spur amplitudes at specific offset frequencies These characteristics directly imprint onto the phase error variance measured at the symbol decision points.

The same transmitter exhibits different phase error signatures across modulation schemes due to varying peak-to-average power ratios and symbol rates:

• QPSK reveals amplifier AM-PM conversion at constant envelope
• 16-QAM stresses the linearity range, exposing compression-induced phase rotation
• OFDM symbols with high PAPR trigger transient thermal memory effects Cross-modulation analysis provides a richer fingerprint than single-scheme measurements.

Phase error is not static. Thermal drift causes slow variation as the transmitter warms up, following a device-specific time constant determined by the thermal mass of the oscillator enclosure and PCB layout. Aging drift over months reflects crystal resonator degradation. Robust fingerprinting systems model this drift as a slowly-varying baseline and extract the stable residual pattern that persists despite temperature and aging effects.

Phase and amplitude errors are not independent. The AM-PM distortion of the power amplifier creates a deterministic coupling where amplitude excursions produce proportional phase shifts. The slope and shape of this AM-PM transfer function is unique to each amplifier's semiconductor physics. Analyzing the joint distribution of I/Q error vectors reveals this coupling as a characteristic rotation pattern in the error vector field.