Local Oscillator Leakage

LO Leakage causes the entire I/Q constellation to shift away from the origin by a fixed vector. This DC offset is the direct result of the unmodulated carrier bleeding through the mixer stage. Unlike thermal noise, which is stochastic, this offset is a deterministic impairment unique to each transmitter's mixer isolation characteristics.

• The offset vector has a specific magnitude and phase angle
• Appears as a non-zero mean in the baseband I and Q sample streams
• Remains constant across modulation schemes for a given device

The physical origin of LO Leakage is finite LO-to-RF isolation in the mixer. Ideally, the mixer suppresses the local oscillator at its output, but parasitic capacitive and substrate coupling paths allow a fraction of the LO power to radiate directly. This leakage level is a function of semiconductor process variations and die layout asymmetries.

• Typical isolation values range from -30 dBm to -50 dBm relative to LO power
• Varies unit-to-unit even within the same chip batch
• Sensitive to temperature and voltage, creating a slow drift signature

In the frequency domain, LO Leakage appears as a distinct narrowband spur at the exact carrier frequency, superimposed on the modulated signal. This spur is unmodulated and coherent with the LO. Its amplitude relative to the modulated signal power defines the carrier suppression ratio, a key fingerprinting metric.

• Spurs are easily identifiable in a power spectral density plot
• Amplitude is independent of the transmitted data payload
• Can be measured even during idle transmission periods

The magnitude and phase of LO Leakage exhibit a slow, predictable drift as the transmitter's die temperature changes. This thermal dependence is governed by the temperature coefficients of the mixer's active and passive components. Unlike random impairments, this drift follows a repeatable trajectory for each device.

• Drift rate is typically on the order of microvolts per degree Celsius
• Creates a hysteresis loop during thermal cycling
• Can be modeled and compensated for in long-term fingerprinting systems

LO Leakage is often confused with I/Q Imbalance, but they are orthogonal impairments. LO Leakage adds a fixed vector to the entire constellation, while I/Q Imbalance causes a gain and phase mismatch between the I and Q rails, resulting in an elliptical distortion. A device's fingerprint is the composite vector of both effects.

• LO Leakage: additive offset, independent of signal amplitude
• I/Q Imbalance: multiplicative distortion, scales with signal amplitude
• Both are simultaneously present in real hardware and must be jointly estimated

In Synthetic RF Impairment Generation, LO Leakage is injected by adding a complex DC term to the clean baseband I/Q samples before upconversion. The impairment is parameterized by a leakage magnitude and phase angle. Domain randomization varies these parameters to train robust Deep Learning Signal Identification models.

• Modeled as: y(t) = x(t) + K * exp(j * phi)
• K represents the leakage magnitude relative to signal RMS
• phi is the fixed phase offset of the leaked carrier

The deliberate reduction of a signal's modulation accuracy by injecting synthetic impairments, serving as a holistic aggregate metric to quantify the severity of combined hardware distortions. LO leakage contributes directly to EVM by shifting the entire constellation away from the origin. Key relationships:

• A -30 dBc LO leakage typically degrades EVM by 2-3%
• EVM is calculated as the root-mean-square error between ideal and impaired symbol positions
• System-level simulators use EVM targets to back-calculate permissible impairment budgets

The synthetic modeling of the irreducible rounding error introduced when a digital baseband waveform is converted to an analog voltage. While LO leakage is an analog mixer artifact, DAC quantization error is a digital-domain impairment that creates spurious tones and an elevated noise floor. In a complete digital twin, these impairments cascade: quantization noise from the DAC feeds into the mixer, where it can intermodulate with the LO leakage tone, producing complex spectral artifacts that become part of the device's unique signature.

The emulation of AM-AM and AM-PM distortion in a transmitter's final stage, characterized by compression curves and memory effects. LO leakage interacts with PA non-linearity in a critical way: the unmodulated carrier component from LO leakage is amplified along with the desired signal, consuming headroom in the PA's linear region. This pushes the composite signal deeper into saturation, exacerbating spectral regrowth and creating a coupled impairment signature that is highly device-specific and difficult to clone.

The process of adding synthesized short-term frequency instability to a clean carrier signal to emulate oscillator imperfections. LO leakage and phase noise originate from the same physical component—the local oscillator—but represent different failure modes:

• LO leakage: A deterministic, constant-amplitude spur at the carrier frequency
• Phase noise: A stochastic process that spreads the carrier energy into skirt-shaped sidebands In high-fidelity simulators, these are modeled as correlated impairments derived from a shared oscillator phase noise mask.

The algorithmic creation of modulated RF signals with precisely controlled, labeled hardware impairments to serve as training data for deep learning fingerprinting models. LO leakage is injected as a complex DC offset in the digital baseband before upconversion. The generation pipeline typically follows this sequence:

• Generate ideal constellation points
• Apply I/Q imbalance and LO leakage as a matrix transformation
• Pass through a digital PA non-linearity model
• Convolve with a channel impulse response
• Add AWGN at the target SNR