WSSUS Assumption

The temporal statistics of the channel impulse response remain invariant over short observation intervals. Specifically, the autocorrelation function depends only on the time difference Δt, not on absolute time t.

• Fading statistics (mean, variance) are constant within a stationarity region
• Doppler spectrum remains fixed during the coherence time
• Enables estimation of channel state information (CSI) before it becomes outdated
• Violated when a mobile moves between drastically different scattering environments

Multipath components arriving at different delay bins are statistically uncorrelated. Scatterers at delay τ₁ and τ₂ contribute independently to the received signal.

• The scattering function S(τ, f_D) is separable into independent delay and Doppler profiles
• Delay power spectrum fully characterizes the power distribution across taps
• Simplifies channel estimation by allowing per-tap processing
• Holds when scatterers are physically separated by more than a wavelength

Under WSSUS, the complete second-order channel statistic—the scattering function—factorizes into independent marginal distributions.

• S(τ, f_D) = P(τ) · D(f_D) where P(τ) is the power delay profile and D(f_D) is the Doppler power spectrum
• Delay spread and Doppler spread become independent design parameters
• Enables separate frequency-domain and time-domain equalizer design
• Validated extensively in macrocellular environments below 6 GHz

The WSSUS assumption breaks down in several practically significant scenarios, requiring more complex non-WSSUS or geometry-based stochastic models.

• High mobility: Rapid environmental changes violate stationarity over typical packet durations
• mmWave and sub-THz: Sparse scattering and beamforming create correlated delay taps
• Indoor hotspots: Clustered scatterers introduce delay-Doppler coupling
• Vehicular channels: Non-stationary birth-death processes of scatterers require time-varying statistics

Bello's 1963 framework formalized WSSUS by defining four equivalent system functions interconnected by Fourier transforms, each revealing different channel characteristics.

• Time-variant impulse response h(t, τ): Direct multipath structure
• Time-variant transfer function H(t, f): Frequency selectivity evolution
• Delay-Doppler spread function s(τ, f_D): Physical scatterer distribution
• Output Doppler-spread function H(f, f_D): Joint frequency-Doppler coupling
• All four representations are equivalent under the WSSUS assumption

The duration over which the channel impulse response remains highly correlated—effectively constant. Under the wide-sense stationary (WSS) component of the assumption, the channel's statistical properties are invariant only within this window. For a carrier frequency f and maximum Doppler shift f_d, coherence time is approximated as T_c ≈ 0.423 / f_d. Transmissions with symbol durations shorter than T_c experience slow fading, while longer transmissions require adaptive equalization to track channel variation.

The frequency range over which the channel response is considered flat or highly correlated. It is inversely proportional to the delay spread (σ_τ): B_c ≈ 1 / (5σ_τ). The uncorrelated scattering (US) assumption ensures that multipath components arriving at different delays fade independently, making coherence bandwidth a statistically meaningful metric. Signals with bandwidths smaller than B_c experience flat fading; wider signals encounter frequency-selective fading requiring OFDM or equalization.

A two-dimensional power spectral density S(τ, f_d) that maps the channel's average power output as a joint function of delay τ and Doppler shift f_d. The WSSUS assumption is the necessary condition for the scattering function to be a complete statistical description of the channel. It reveals:

• Delay power profile: Integrating over Doppler yields average power per delay bin
• Doppler power spectrum: Integrating over delay yields spectral broadening per frequency shift This function is the primary input for designing pilot patterns and equalizers.

The finite time window—typically on the order of milliseconds to tens of milliseconds in vehicular channels—during which the WSS assumption holds. Beyond this interval, large-scale effects like shadowing and path loss alter the channel's statistics, violating stationarity. In RF digital twin environments, the stationarity interval defines the maximum simulation step size before the channel model must be updated with new geometric parameters. Exceeding this interval invalidates ergodic capacity calculations and adaptive modulation decisions.

A discrete-time implementation of the WSSUS channel where each tap represents a resolvable multipath cluster with a specific delay and a time-varying complex gain. Under the US assumption, the fading processes on different taps are statistically independent. Each tap's gain is generated by filtering white Gaussian noise through a Doppler spectrum—typically a Jakes spectrum for isotropic scattering. This model is the workhorse for link-level simulation in standards like 3GPP and IEEE 802.11.

The maximum mutual information averaged over all channel realizations, valid only when the channel is ergodic within the coding block length. The WSSUS assumption provides the statistical stationarity required for ergodicity to hold. In practice, ergodic capacity is achievable when the codeword spans many coherence intervals, allowing the receiver to average over both time and frequency diversity. This metric drives adaptive coding and modulation (ACM) decisions in LTE and 5G NR systems.