Multipath Fading

The superposition of multiple delayed signal copies at the receiver antenna causes the composite signal amplitude to vary dramatically. When signal components arrive in-phase, they combine constructively, boosting received power. When they arrive out-of-phase, they cancel destructively, creating deep fades that can drop the signal below the noise floor. This interference pattern is highly sensitive to the transmitter's position, carrier frequency, and the geometry of the physical environment.

Multipath components travel different distances, arriving at the receiver with varying propagation delays. The delay spread—the time difference between the first and last significant arriving path—causes inter-symbol interference (ISI) in digital communications. When the delay spread exceeds the symbol period, energy from one symbol spills into adjacent symbols, distorting the received constellation and complicating the extraction of stable hardware impairment fingerprints.

When the signal bandwidth exceeds the coherence bandwidth of the channel (inversely proportional to delay spread), different frequency components experience uncorrelated fading. This creates a non-uniform frequency response across the signal spectrum, with some subcarriers experiencing deep nulls while others remain strong. For RF fingerprinting, this means the channel distorts different spectral regions of the device signature independently, requiring robust feature extraction across the entire band.

Relative motion between transmitter, receiver, or scatterers introduces Doppler shift, causing spectral broadening of the received signal. The coherence time—the duration over which the channel impulse response remains approximately constant—is inversely proportional to the maximum Doppler spread. Fast-fading channels change within a single transmission burst, forcing fingerprinting models to extract device signatures from transient, non-stationary observations rather than relying on long-term averaging.

Multipath fading is categorized by the relationship between signal bandwidth and channel coherence bandwidth. Flat fading occurs when the signal bandwidth is narrower than the coherence bandwidth, causing all frequency components to fade together—preserving the spectral shape but scaling amplitude uniformly. Frequency-selective fading occurs when the signal bandwidth exceeds the coherence bandwidth, creating a non-uniform spectral response. Domain adversarial training must handle both regimes to produce channel-robust device fingerprints.

Multipath fading introduces a multiplicative distortion that convolves with the transmitter's intrinsic hardware impairment signature. The received signal is the convolution of the transmitted signal with the channel impulse response, making it difficult to disentangle device-specific features from propagation artifacts. Techniques like contrastive learning and feature disentanglement explicitly aim to separate these channel-induced variations from the stable, device-specific embedding that constitutes the RF fingerprint.