DC Offset

DC offset originates primarily from local oscillator (LO) leakage and finite port-to-port isolation in the mixer stage. The LO signal couples into the RF or IF path, self-mixes, and produces a DC component at baseband. This leakage is a function of the physical layout, shielding, and semiconductor doping variances unique to each transmitter.

• Self-mixing mechanism: LO signal leaks to the RF port, reflects, and mixes with itself
• Isolation variance: Typical mixer isolation ranges from 25-40 dB, varying per device
• Temperature dependence: DC offset drifts predictably with thermal changes, adding a secondary identifying dimension

In the frequency domain, DC offset appears as an unmodulated tone at the carrier frequency—a spectral spike that should not exist in an ideal suppressed-carrier modulation scheme. The amplitude and phase of this residual carrier are stable over time and unique to each transmitter.

• Amplitude signature: The power of the carrier leak relative to the modulated signal is a device-specific metric
• Phase relationship: The fixed phase offset of the carrier leak relative to the modulated signal provides a second dimension for discrimination
• Measurement: Typically expressed in dBc (decibels relative to carrier), with values ranging from -25 dBc to -45 dBc in commercial transmitters

DC offset causes a rigid translation of the entire constellation diagram away from the origin. Unlike I/Q imbalance, which warps the constellation shape, DC offset shifts every symbol point by the same vector. The magnitude and direction of this shift form a stable, two-dimensional fingerprint.

• Vector representation: The offset is characterized by a fixed (I_offset, Q_offset) coordinate pair
• Modulation independence: The shift vector remains constant regardless of the modulation scheme or data being transmitted
• Visual indicator: The centroid of the symbol clusters is displaced from the origin by a measurable distance

DC offset must be carefully isolated from other hardware impairments that also affect the constellation. Unlike I/Q imbalance, which creates an elliptical distortion, DC offset is a pure translation. Unlike phase noise, which causes rotational smearing, DC offset is a static displacement.

• vs. I/Q Imbalance: DC offset shifts the origin; I/Q imbalance warps the grid
• vs. Carrier Frequency Offset: CFO causes constellation rotation; DC offset causes translation
• vs. Phase Noise: Phase noise is stochastic; DC offset is deterministic and stable
• Joint estimation: Advanced algorithms separate these impairments using statistical decomposition techniques

DC offset is extracted by computing the mean of the I and Q sample streams over a sufficient observation window. The DC component is the time-average of the baseband signal, and modern receivers can estimate it with high precision using digital signal processing.

• Time-domain averaging: Compute the arithmetic mean of I and Q samples independently
• FFT bin analysis: Measure the magnitude of the zero-frequency bin in the baseband spectrum
• Compensation-aware extraction: Some transmitters apply DC offset correction; fingerprinting systems must characterize the residual after compensation
• Required SNR: Reliable extraction typically requires a signal-to-noise ratio above 15 dB

DC offset exhibits excellent long-term stability, making it a reliable fingerprint for persistent device identification. However, it does drift slowly due to component aging and thermal cycling. Fingerprinting systems must model this drift to maintain accuracy over months or years of operation.

• Short-term stability: Variance is typically less than 0.1% of the offset magnitude over hours
• Thermal drift coefficient: Predictable shift of 0.05-0.2% per degree Celsius
• Aging rate: Gradual change of 1-3% per year due to semiconductor junction degradation
• Drift compensation: Kalman filters or exponential moving averages track the slow evolution of the DC offset vector

A metric measuring the deviation of actual transmitted symbols from their ideal constellation points. The statistical distribution of this error vector—its mean, variance, and higher-order moments—serves as a device fingerprint. A consistent DC Offset contributes a fixed bias component to the error vector, shifting the centroid of each symbol cluster in a specific direction.

• EVM = |Error Vector| / |Reference Vector| × 100%
• DC Offset produces a non-zero mean error in one direction
• The error vector's phase and magnitude distribution is device-specific

The path traced by the signal's instantaneous phase over time. Subtle, device-specific variations in the transition between symbols reveal a unique hardware signature. While DC Offset primarily affects the steady-state symbol positions, it also influences the dynamic trajectory as the signal passes near the origin during symbol transitions, creating a distinctive curvature or asymmetry in the phase path.

• Captures both transient and steady-state behavior
• DC Offset shifts the center of rotation for phase transitions
• Useful for identifying devices with similar constellation maps but different dynamic signatures

The distortion introduced by a power amplifier operating near its saturation point, characterized by AM/AM (amplitude-to-amplitude) and AM/PM (amplitude-to-phase) conversion curves unique to each physical device. While distinct from DC Offset, non-linearity often interacts with carrier leakage—the DC Offset can shift the operating point of the amplifier, altering the non-linear distortion pattern in a device-specific manner.

• AM/AM distortion: Output amplitude compresses relative to input
• AM/PM distortion: Phase shift varies with input amplitude
• The combined effect with DC Offset creates a multi-dimensional fingerprint