Ch. 15 — Cellular Systems and Infrastructure-Based

Chapter 15 of Wireless Communications (2nd ed. Draft) — Andrea Goldsmith

Abstract

This chapter establishes the fundamental information-theoretic rate limits of infrastructure-based cellular systems, addressing the complexities of intercell interference and frequency reuse. It analyzes the Shannon capacity under two distinct regimes: full base station cooperation and treating interference as Gaussian noise. The central contribution is the formulation of Area Spectral Efficiency (ASE) as a metric to optimize cellular system structure relative to fundamental capacity, specifically evaluating the trade-off between reuse distance and interference power. This analysis provides a theoretical benchmark for evaluating practical resource allocation strategies like power control and dynamic channel assignment.

Key Concepts

• Wyner Model: A simplified cellular channel model where base stations are arranged in one or two-dimensional arrays. It assumes a channel gain of unity within a cell and an intercell attenuation factor $\alpha$ ($0\le \alpha \le 1$) between cells. This model allows for the analytical derivation of per-user capacity $C(\alpha )$ under base station cooperation, characterizing how intercell coupling affects system throughput.
• Full Base Station Cooperation: An assumption where base stations jointly encode and decode all signals, effectively eliminating the notion of separate cells. The network is treated as a single multiple-input multiple-output (MIMO) system, where the uplink becomes a multiple-access channel and the downlink becomes a broadcast channel. This represents the upper bound on cellular capacity.
• Interference as Gaussian Noise: A practical analysis approach where receivers in each cell treat signals from other cells as noise rather than decoding them. This mirrors real-world systems where full cooperation is unavailable. While suboptimal compared to joint decoding, it simplifies capacity analysis to single-cell expressions dependent on the signal-to-interference ratio (SIR).
• Area Spectral Efficiency (ASE): A capacity measure defined as the throughput per hertz per unit area supported by a cell’s resources. It quantifies the trade-off between efficient resource reuse (small reuse distance) and the resulting intercell interference. ASE is critical for optimizing the cellular structure, as it normalizes capacity by the physical area covered.
• Reuse Distance ($D$): The minimum distance between the centers of any two cells using the same channel frequency. The area covered by a single channel is roughly determined by $A=\pi (0.5D{)}^2$. Reducing $D$ increases resource reuse efficiency but increases intercell interference, thereby reducing the per-cell capacity region.
• Intercell Interference Power: The aggregate power received from transmissions in neighboring cells using the same frequency. In the Wyner model, this is characterized by the attenuation factor $\alpha$. In physical models, it depends on the path loss exponent $\gamma$ and the geometric arrangement of interfering cells (typically the first tier of neighbors).
• SINR Optimization: The process of maximizing the signal-to-interference-plus-noise ratio for users. Under interference-as-noise assumptions, maximizing SINR often correlates with maximizing capacity. Power control algorithms are frequently employed to maintain target SIR levels, though feasibility depends on the existence of a power vector $\mathbf{P}$ such that $\mathbf{P}>0$.
• Orthogonal Multiple Access: Techniques like TDMA or FDMA that assign disjoint resources to users within a cell. The text indicates that for the uplink with interference treated as noise, orthogonal channelization is capacity-achieving within the cell, even though non-orthogonal schemes may offer other benefits.

Key Equations and Algorithms

• Wyner Uplink Per-User Capacity: $C(\alpha )=\frac{B}{K}{\log }_2\left(1+\frac{P}{N_{0}B/K+\alpha P}\right)$. This equation defines the maximum rate per user in the Wyner model with full cooperation. $B$ is total bandwidth, $K$ is users per cell, $N_0$ is noise spectral density, $P$ is transmit power, and $\alpha$ is the intercell interference attenuation.
• System Throughput (Sum-Rate): $C_{SR}={\sum }_{k=1}^K{R}_k$. This represents the total capacity region sum rate for a cell with $K$ users, where $R_k$ is the rate of user $k$. This metric is used to derive the ASE by normalizing throughput against bandwidth and area.
• Area Spectral Efficiency Definition: $A_e=\frac{C_{SR}}{B\times \pi (D/2{)}^2}$. This formula calculates the throughput per Hz per square meter. It explicitly links the system throughput $C_{SR}$ to the reuse distance $D$, highlighting that increasing $D$ reduces the denominator (area per channel) but increases the numerator (capacity per cell) due to reduced interference.
• Path Loss Model for Interference: $P_r=P_t{d}^{-\gamma }$. This simplified model characterizes signal attenuation. Within a cell, the path loss exponent is typically $\gamma \le 4$, while between cells, it may vary. The received signal power $P_k$ is $P{R}^{-2}$ (where $R$ is cell radius), and interference power $I_k$ is derived from interfering transmitters at distance $D-R$.
• User Rate with Constant SINR: $R_k={\tau }_{k}B{\log }_2(1+{\gamma }_k)$. This expression gives the achievable rate for user $k$, where ${\tau }_k$ is the assigned time fraction and ${\gamma }_k$ is the received signal-to-interference power. This is used to compute $C_{SR}$ when ${\gamma }_k$ is assumed constant.
• Distributed Power Control Algorithm: $P_k(i+1)=P_k(i)\frac{{\gamma }_k^{\ast }}{{\text{SINR}}_k(i)}$. This iterative update rule allows transmitters to adjust power to meet target SINR ${\gamma }_k^{\ast }$. If ${\text{SINR}}_k(i)$ is below target, power increases. Convergence to a Pareto optimal solution ${\mathbf{P}}^{\ast }$ is guaranteed if the Perron-Frobenius eigenvalue of the channel gain matrix is less than unity.

Key Claims and Findings

• Full Cooperation Capacity Limits: Under full base station cooperation, the capacity region of cellular uplinks is known for specific models, with the Wyner model showing that per-user capacity is determined primarily by the number of users $K$ and the interference attenuation $\alpha$.
• Behavior of Capacity with $\alpha$: At high SNR, per-user capacity $C(\alpha )$ generally increases with $\alpha$ because strong interference aids decoding and subsequent subtraction. Conversely, at low SNR, $C(\alpha )$ initially decreases with $\alpha$ because weak interference cannot be reliably decoded, acting purely as noise.
• Optimality of Orthogonal Access: When intercell interference is treated as Gaussian noise, orthogonal channelization methods (e.g., TDMA) within a cell are capacity-achieving. This holds true even when channel-inversion power control is utilized within the cell.
• Area Spectral Efficiency Trade-off: There exists an optimal reuse distance $D$ that maximizes Area Spectral Efficiency. Increasing $D$ reduces interference (increasing sum-rate) but increases the area associated with each channel (decreasing efficiency per unit area).
• Dependence on Path Loss Exponent: The optimal reuse distance is influenced by the intercell path loss exponent $\gamma$. If intercell interference falls off more slowly (smaller $\gamma$), the Area Spectral Efficiency is decreased, and surprisingly, the optimal reuse distance may also be decreased.
• Downlink Capacity Gap: While capacity results exist for the uplink under various assumptions (cooperation or noise assumption), no such closed-form results are available for the downlink under broad assumptions, creating a significant open problem in information theory.

Terminology

• Area Spectral Efficiency (ASE): A performance metric representing the throughput per hertz per unit area of the cellular system. It integrates both spectral efficiency and spatial reuse efficiency into a single value.
• Wyner Model: A theoretical cellular model assuming an infinite grid of cells with uniform attenuation $\alpha$ to nearest neighbors. It simplifies spatial geometry to facilitate closed-form capacity analysis.
• Intercell Interference Factor ($\alpha$): A parameter in the Wyner model representing the channel gain between a mobile in one cell and the base station in a neighboring cell, normalized to the intracell gain.
• Reuse Distance ($D$): The shortest distance between the centers of two cells assigned the same frequency channel. It determines the frequency reuse pattern and the magnitude of co-channel interference.
• Sum-Rate Capacity: The maximum total data rate supported by the system, denoted as $C_{SR}=\sum R_k$. In cellular contexts, this is often maximized subject to power and interference constraints.
• Pareto Optimal Solution: A power allocation vector ${\mathbf{P}}^{\ast }$ where improving the SINR of one link requires reducing the SINR of another. In power control, this represents the minimum power state satisfying all SINR targets.
• Channel Gain Matrix ($\mathbf{F}$): A matrix where elements $F_{kj}$ represent the normalized interference gain from transmitter $j$ to receiver $k$. The spectral radius of this matrix determines the feasibility of power control.
• Cooperative Diversity: A technique derived from cooperation principles where nodes relay signals to improve reliability. While not fully resolved for downlink in this chapter, it relates to the potential of joint decoding strategies.