Multipath Fading Emulation

The foundational mathematical kernel of multipath emulation. A Channel Impulse Response (CIR) is a time-domain filter that captures the amplitude, delay, and phase of every resolvable propagation path between a transmitter and receiver. In emulation, the synthetic signal is convolved with a time-varying CIR to replicate dynamic environments where the transmitter, receiver, or scatterers are in motion. The CIR is typically implemented using a Tapped Delay Line (TDL) structure, where each tap represents a discrete multipath component with its own complex coefficient and Doppler spectrum.

The envelope of the received signal is modeled using well-defined statistical distributions that characterize the propagation environment. Rayleigh fading models dense multipath with no dominant line-of-sight (LoS) component, where the received envelope follows a Rayleigh distribution. Rician fading adds a dominant LoS component to the scattered paths, parameterized by a K-factor (the ratio of LoS power to scattered power). A K-factor of 0 reduces to Rayleigh; a high K-factor approaches an AWGN-only channel. These distributions are critical for training fingerprinting models to be robust across diverse deployment scenarios.

Relative motion between transmitter, receiver, and scatterers causes Doppler shift, a frequency-domain spreading that makes the channel time-selective. Emulators apply a Doppler spectrum—commonly the Jakes model for isotropic scattering—to each tap of the TDL. The maximum Doppler shift (f_d) is proportional to velocity and carrier frequency. This temporal selectivity introduces coherence time, the interval over which the channel remains approximately constant. Accurate Doppler emulation is essential for testing fingerprinting models against real-world mobility patterns.

A Power Delay Profile (PDP) defines the relative power and excess delay of each multipath component in a channel model. Standardized PDPs—such as those from ITU (Pedestrian A, Vehicular A) or 3GPP (Tapped Delay Line models)—provide parameter sets for emulating specific environments. Key PDP metrics include:

• Mean excess delay: The first moment of the PDP
• RMS delay spread: The square root of the second central moment, quantifying time dispersion
• Maximum excess delay: The delay at which the PDP falls below a threshold These parameters directly determine the coherence bandwidth, the frequency range over which the channel is flat.

When the signal bandwidth exceeds the coherence bandwidth (approximately the inverse of the RMS delay spread), the channel becomes frequency-selective. Different frequency components of the signal experience uncorrelated fading, causing inter-symbol interference (ISI). Emulators must accurately reproduce this selectivity by ensuring the TDL tap spacing is sufficiently fine to resolve the signal bandwidth. Frequency selectivity is a critical stress test for fingerprinting models, as it distorts the very hardware impairments the model relies upon for identification.

Emulators operate in two fundamental regimes based on the relationship between signal bandwidth and coherence bandwidth:

• Flat fading: Signal bandwidth << coherence bandwidth. All frequency components fade simultaneously. Simpler to implement but less realistic for wideband signals.
• Frequency-selective fading: Signal bandwidth > coherence bandwidth. Requires a multi-tap TDL to model the channel's frequency-dependent behavior. This is the standard mode for emulating modern wideband waveforms (LTE, 5G, Wi-Fi) and is essential for generating realistic training data for deep learning fingerprinting models.