I/Q Imbalance

Gain imbalance occurs when the amplification applied to the I (in-phase) and Q (quadrature) signal paths differs. This asymmetry causes the ideal square constellation to stretch into a rectangular shape.

• Typical range: 0.1 dB to 2 dB in commercial transmitters
• Effect: Symbol points are displaced along one axis more than the other
• Measurement: Ratio of I-channel gain to Q-channel gain (α = G_I / G_Q)
• Stability: Remains remarkably consistent over device lifetime due to fixed resistor tolerances in the amplifier feedback network

Phase imbalance (quadrature error) arises when the I and Q local oscillator signals are not precisely 90 degrees apart. This non-orthogonality causes the constellation to skew or shear diagonally.

• Typical range: 1 to 10 degrees in low-cost transceivers
• Effect: I/Q cross-talk where energy from one channel leaks into the other
• Mathematical model: The received baseband signal becomes r(t) = I(t) + j·Q(t)·sin(φ) + j·Q(t)·cos(φ) where φ is the phase error
• Origin: Mismatched trace lengths in PCB layout or LO path component tolerances

The Image Rejection Ratio (IRR) quantifies the combined effect of gain and phase imbalance. It measures how well the modulator suppresses the unwanted sideband image signal.

• Formula: IRR (dB) = 10·log₁₀[(1 + 2·α·cos(φ) + α²) / (1 - 2·α·cos(φ) + α²)]
• High-quality transmitters: 35–45 dB IRR
• Low-cost IoT devices: 20–30 dB IRR
• Fingerprinting value: IRR varies by 5–15 dB between individual units of the same model, making it a strong discriminator

While often modeled as frequency-independent, real I/Q imbalance varies across the signal bandwidth. Frequency-selective imbalance is caused by mismatched I and Q low-pass filters.

• Source: Component tolerances in analog baseband filters create different frequency responses in each branch
• Wideband signals: OFDM and spread-spectrum waveforms reveal this frequency-dependent behavior
• Fingerprinting advantage: Provides a richer, multi-dimensional signature compared to flat imbalance
• Compensation complexity: Requires adaptive FIR filters rather than simple scalar correction

The combined gain and phase imbalance produces a characteristic warping pattern in the I/Q constellation diagram that is unique to each transmitter.

• Gain-only imbalance: Produces a rectangular stretch along one axis
• Phase-only imbalance: Produces a rhomboid skew
• Combined imbalance: Creates an elliptical distortion where the ideal circular decision boundaries become oval
• Device-specific: The exact warping parameters (α, φ) form a continuous 2D fingerprint space
• Modulation impact: Higher-order QAM (64-QAM, 256-QAM) is more sensitive, making imbalance more detectable

I/Q imbalance exhibits strong temporal stability, making it a reliable long-term fingerprint. Unlike phase noise, which fluctuates, imbalance is primarily determined by fixed component values.

• Temperature coefficient: Typically < 0.01 dB/°C for gain imbalance
• Aging drift: Negligible over 5–10 year device lifetimes due to stable resistor ratios
• Calibration persistence: Factory calibration values remain valid for years
• Contrast with oscillator drift: Carrier frequency offset drifts with temperature, but I/Q imbalance ratios remain constant because both branches are affected equally by environmental changes

A constant voltage bias added to the baseband signal caused by local oscillator leakage or mixer port isolation. This results in a carrier leak—a distinct spectral tone at the center frequency—that manifests as a unique, measurable signature in the constellation diagram. Unlike I/Q imbalance which creates elliptical warping, DC offset shifts the entire constellation away from the origin.

• Caused by LO-to-RF port isolation failures
• Appears as a fixed displacement in the I/Q plane
• Highly stable over time and temperature

The visual and quantitative examination of the scatter plot of in-phase versus quadrature signal samples. Hardware impairments manifest as distinct geometric distortions: gain imbalance stretches the constellation into an ellipse, phase imbalance shears it into a parallelogram, and DC offset displaces the entire structure. Statistical clustering metrics derived from these diagrams serve as robust device fingerprints.

• Gain imbalance: elliptical stretching along one axis
• Phase imbalance: non-orthogonal skewing of symbol clusters
• Combined effects create unique, device-specific warping patterns

A metric measuring the deviation of actual transmitted symbols from their ideal constellation points. The statistical distribution of this error vector—not just its magnitude—serves as a device fingerprint. I/Q imbalance produces a characteristic error pattern where deviations correlate with symbol position, creating a signature error cloud unique to each transmitter's analog front-end.

• EVM = |S_actual - S_ideal| / |S_ideal|
• I/Q imbalance creates position-dependent error vectors
• Distribution shape is more identifying than RMS EVM alone

The extraction of device-specific features directly from the demodulated symbol sequence, focusing on errors in the ideal symbol constellation caused by hardware impairments. This approach operates after carrier and timing recovery, isolating the residual distortion that I/Q imbalance, phase noise, and amplifier non-linearity imprint on each symbol decision point.

• Works on fully demodulated baseband symbols
• Captures the combined effect of all analog impairments
• Enables fingerprinting without raw waveform storage

The application of neural networks to correct non-linear signal distortion caused by power amplifiers. While primarily targeting AM/AM and AM/PM conversion, modern DPD systems also compensate for I/Q imbalance by learning an inverse model of the transmitter's complex baseband impairment. The residual uncorrected distortion after DPD becomes a subtle, device-specific fingerprint.

• Neural DPD learns joint impairment compensation
• Residual error after correction is device-unique
• Adaptive algorithms track thermal drift in real-time

A technique that uses the known, repetitive structure of a packet preamble to isolate and analyze subtle hardware-induced distortions. By correlating the received preamble against an ideal reference, the residual error signal reveals I/Q imbalance, DC offset, and other impairments with high precision, as the deterministic preamble eliminates modulation uncertainty.

• Uses 802.11 STS/LTF or similar training sequences
• Correlation isolates impairment from channel effects
• Enables per-packet fingerprint extraction