Ch. 3 — Statistical Multipath Channel Models

Chapter 3 of Wireless Communications (2nd ed. Draft) — Andrea Goldsmith

Abstract

This chapter establishes the statistical framework required to characterize small-scale multipath fading in wireless communication channels where deterministic path geometry is unknown. It introduces the time-varying channel impulse response $c(\tau ,t)$ to model the constructive and destructive addition of multipath components over time. The text derives the narrowband fading model under the Central Limit Theorem, leading to Rayleigh and Rician distributions for the signal envelope, and analyzes temporal statistics including Level Crossing Rate and Average Fade Duration. These models are critical for predicting link reliability and designing diversity schemes in systems operating over mobile channels.

Key Concepts

• Time-Varying Channel Impulse Response: Defined as $c(\tau ,t)$, this function represents the channel output at time $t$ in response to an impulse launched at time $t-\tau$, capturing the dynamic nature of multipath delays ${\tau }_i(t)$ and amplitudes ${\alpha }_i(t)$. It generalizes the time-invariant impulse response to account for Doppler shifts caused by transmitter, receiver, or scatterer movement, forming the basis for all subsequent statistical models.
• Narrowband Fading Model: This model assumes the channel delay spread $T_m$ is small relative to the inverse signal bandwidth ($T_m\ll 1/B_u$), effectively treating the channel as non-dispersive. This simplification allows the received signal to be represented as a scaled version of the transmitted signal, enabling the separation of multipath effects into in-phase and quadrature components for statistical analysis.
• Resolvable vs. Non-Resolvable Multipath: Multipath components are considered resolvable if their delay difference significantly exceeds the inverse signal bandwidth. In narrowband channels, components often fall below this threshold, combining into a single resolvable component where phase interactions cause rapid amplitude fluctuations known as fading.
• Jakes/Clarke Uniform Scattering Model: A canonical theoretical model assuming a dense scattering environment where multipath components arrive uniformly from all angles with equal average power. This assumption allows the derivation of a closed-form autocorrelation function using the zero-order Bessel function, predicting how signal correlation decays with distance.
• Rayleigh Fading Distribution: Derived via the Central Limit Theorem for scenarios without a dominant Line-of-Sight (LOS) component, where in-phase and quadrature components are zero-mean Gaussian variables. The signal envelope follows this distribution, implying severe fluctuations where the received power can drop significantly below the average value.
• Rician Fading Distribution: Applicable when a dominant LOS component exists alongside scattered non-LOS paths. The fading parameter $K$ quantifies the ratio of LOS power to scattered power, modifying the envelope distribution to account for the persistent signal component that mitigates deep fades.
• Nakagami-$m$ Fading Distribution: A generalized fading model parameterized by $m$ and average power $P_r$, capable of fitting a wider range of empirical measurements than Rayleigh or Rician models. It reduces to Rayleigh fading when $m=1$ and represents no fading when $m=\infty$, providing flexibility for environments with varying scattering conditions.
• Doppler Power Spectral Density (PSD): Describes the distribution of signal power as a function of Doppler frequency shift, typically exhibiting a “U-shaped” spectrum (Jakes spectrum). The PSD peaks at the maximum Doppler shift $f_D$, indicating that multipath components arriving head-on or tail-on contribute most to the frequency spread.
• Level Crossing Rate (LCR): Represents the expected rate at which the signal envelope crosses a specified threshold level in the downward direction. This statistical metric quantifies the frequency of fades, which is directly related to the Doppler spread of the channel and the velocity of the mobile unit.
• Average Fade Duration (AFD): Measures the mean time the signal envelope remains continuously below a target threshold level once a fade begins. It is inversely proportional to the Doppler frequency, implying that faster-moving channels experience shorter but more frequent fades compared to slower channels.
• Block-Fading Channel: A simplification where the channel gain is assumed constant over a specific block of time (the coherence time) and changes independently between blocks. This model facilitates performance analysis by discretizing the continuous fading process into manageable states for coding and modulation design.
• Finite-State Markov Channel (FSMC): A specialized block-fading model where channel states transition according to Markov probabilities, restricting transitions to adjacent states. This approach captures the temporal correlation of fading, providing a more accurate representation of channel dynamics for system design than independent block models.

Key Equations and Algorithms

• Received Signal Model: $r(t)={\sum }_{i=0}^{N(t)-1}{\alpha }_i(t)\cos (2\pi f_c(t-{\tau }_i(t))+{\varphi }_{Di}(t))$. This summation expresses the received signal as a superposition of $N(t)$ multipath components, where ${\alpha }_i(t)$ represents path loss and shadowing, and ${\varphi }_{Di}(t)$ accounts for Doppler phase shifts due to motion.
• Time-Varying Impulse Response: $c(\tau ,t)={\sum }_{i=0}^{N(t)-1}{\alpha }_i(t)e^{-j{\varphi }_i(t)}\delta (\tau -{\tau }_i(t))$. This expression defines the equivalent lowpass impulse response, mapping the input $u(t)$ to output $r_{LP}(t)$ via convolution, and explicitly shows the dependence on delay $\tau$ and observation time $t$.
• In-Phase and Quadrature Autocorrelation: $A_{rI}(\tau )=\frac{P_r}{2}{J}_0(2\pi f_D\tau )$. Derived under uniform scattering, this equation relates the correlation of the signal components to the zero-order Bessel function, determining how quickly the channel decorrelates over time $\tau$ given maximum Doppler $f_D$.
• Power Spectral Density (Jakes Spectrum): $S_{rI}(f)=\frac{P_r}{2\pi f_D\sqrt{1-(f/f_D{)}^2}}$ for $|f|\le f_D$. This PSD describes the power density distribution over Doppler frequencies, characterizing the bandwidth of the fading process and confirming infinite density at the maximum Doppler frequency limits.
• Rayleigh Distribution: $p(z)=\frac{z}{{\sigma }^2}\exp \left(-\frac{z^2}{2{\sigma }^2}\right)$ for $z\ge 0$. This probability density function characterizes the envelope $z(t)$ of a narrowband signal in the absence of a LOS component, with mean power $P_r=2{\sigma }^2$.
• Rician Distribution: $p(z)=\frac{z}{{\sigma }^2}\exp \left(-\frac{z^2+s^2}{2{\sigma }^2}\right)I_0\left(\frac{zs}{{\sigma }^2}\right)$. This distribution modifies the Rayleigh model by including the LOS amplitude $s$, where $I_0$ is the modified Bessel function of zeroth order, and fading severity is controlled by the $K$-factor ratio $s^2/2{\sigma }^2$.
• Level Crossing Rate (Rayleigh): $L_Z=\sqrt{2\pi }{f}_D\rho e^{-{\rho }^2}$. This formula quantifies the frequency of fades below a normalized threshold $\rho$, showing that crossing rates increase with Doppler frequency $f_D$ and decrease exponentially with the threshold level.
• Average Fade Duration (Rayleigh): ${\bar{\tau }}_Z=\frac{e^{{\rho }^2}-1}{f_D\rho }$. This metric calculates the mean duration of a fade event, demonstrating that higher Doppler frequencies reduce the time spent in outage for a given signal threshold.
• FSMC State Transition Condition: Channel state $R_j$ transitions only to $R_{j-1}$, $R_j$, or $R_{j+1}$ within a block time $T$. This constraint models the temporal continuity of fading, ensuring that SNR does not change abruptly between states, which is physically consistent with finite Doppler spread.

Key Claims and Findings

• Signal decorrelation occurs over a distance of approximately $0.4\lambda$, meaning antenna spacing of this magnitude is required to obtain independent fading paths for diversity combining.
• Narrowband fading is characterized by rapid variations (on the order of a wavelength) superimposed on large-scale path loss and shadowing, necessitating separate statistical treatment for each phenomenon.
• In a uniform scattering environment, the received signal power is exponentially distributed, leading to deep fades that occur with significant probability even when average received power is high.
• The maximum Doppler frequency $f_D$ scales linearly with the velocity of the mobile unit, directly dictating the coherence time and the rate at which the channel changes.
• Average Fade Duration decreases as the Doppler frequency increases, implying that while fast-moving users experience fades more frequently, they spend less absolute time in any single fade state.
• The Nakagami-$m$ distribution provides a versatile approximation for empirical fading data, capable of modeling both lighter-than-Rayleigh and heavier-than-Rayleigh fading conditions through its shape parameter.
• Finite-State Markov Channels offer a computationally tractable approximation for continuous fading processes, valid when the transition probability between adjacent SNR regions is well-defined by the Doppler spread.

Terminology

• Delay Spread ($T_m$): The time difference between the arrival of the first and last significant multipath components of a signal, characterizing the time-dispersion capability of the channel.
• Coherence Time: The time duration over which the channel impulse response is considered invariant, typically inversely proportional to the maximum Doppler shift.
• Doppler Frequency Shift ($f_{Di}$): The change in frequency of a signal component due to relative motion, calculated as $v\cos ({\theta }_i)/\lambda$ where $v$ is velocity and ${\theta }_i$ is the angle of arrival.
• Level Crossing Rate ($L_Z$): The expected number of times per second that the signal envelope crosses a specific threshold level in a specified direction (typically downward).
• Average Fade Duration (${\bar{\tau }}_Z$): The average length of time the signal envelope remains continuously below a specified threshold level once a fade event is initiated.
• In-Phase Component ($r_I$): The quadrature component of the received signal that is in phase with the local carrier reference, treated as a zero-mean Gaussian random process in narrowband analysis.
• Quadrature Component ($r_Q$): The component of the received signal shifted by 90 degrees relative to the in-phase component, uncorrelated with $r_I$ in uniform scattering environments.
• Power Angle Spectrum (PAS): A function characterizing the received signal power as a function of the angle of arrival of multipath components, central to defining scattering models like the Jakes model.
• Block Time ($T$): The time interval over which the channel state in a block-fading model is assumed to remain constant, often set to match the channel coherence time.
• Fading Parameter ($K$ or $m$): Shape parameters defining specific fading distributions; $K$ describes the LOS-to-scatterer power ratio in Rician fading, while $m$ generalizes the envelope shape in Nakagami fading.